object-storage: Don't catch generic exceptions around xattr calls

We should really avoid catching generic Exception objects around any of our system calls. This simple fix brings this code in line with the other code wrapping xattr calls. Change-Id: I75492a61081fbfbec34c8d91a47214858c67e405 Signed-off-by: Peter Portante <peter.portante@redhat.com> Reviewed-on: http://review.gluster.org/4099 Reviewed-by: Jeff Darcy <jdarcy@redhat.com> Reviewed-by: Anand Avati <avati@redhat.com> Tested-by: Anand Avati <avati@redhat.com>
author: Peter Portante <peter.portante@redhat.com> 2012-10-17 15:25:34 -0400
committer: Anand Avati <avati@redhat.com> 2012-10-17 15:45:56 -0700
commit: 2b8043986e6f35a91bf1919c657d73f67b905ef7 (patch)
tree: 7a2198fb94fd9cb415ea88b83f3aa8e67b86b55a /contrib/stdlib
parent: dbe793bcde5d0ba936eceb19bcb8a6f376a0dbc4 (diff)
0 files changed, 0 insertions, 0 deletions
diff --git a/xlators/Makefile.am b/xlators/Makefile.am
index b1643d26c..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 system
+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 60bf3c9e5..337c983ec 100644
--- a/xlators/bindings/python/src/gluster.py
+++ b/xlators/bindings/python/src/gluster.py
@@ -1,12 +1,11 @@
-"""
-  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.
-"""
+
+#  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.
 
 from ctypes import *
 from glustertypes import *
diff --git a/xlators/bindings/python/src/glusterstack.py b/xlators/bindings/python/src/glusterstack.py
index 19da219fd..0c071ae98 100644
--- a/xlators/bindings/python/src/glusterstack.py
+++ b/xlators/bindings/python/src/glusterstack.py
@@ -1,12 +1,11 @@
-"""
-  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.
-"""
+#  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.
 
 from ctypes import *
 from glustertypes import *
diff --git a/xlators/bindings/python/src/glustertypes.py b/xlators/bindings/python/src/glustertypes.py
index a75b7d090..98437d22e 100644
--- a/xlators/bindings/python/src/glustertypes.py
+++ b/xlators/bindings/python/src/glustertypes.py
@@ -1,12 +1,11 @@
-"""
-  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.
-"""
+
+#  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.
 
 from ctypes import *
 import collections
diff --git a/xlators/bindings/python/src/python.c b/xlators/bindings/python/src/python.c
index 8ba008c23..9b96790de 100644
--- a/xlators/bindings/python/src/python.c
+++ b/xlators/bindings/python/src/python.c
@@ -1,13 +1,12 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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
@@ -36,7 +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;
@@ -139,7 +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;
   }
@@ -157,7 +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);
@@ -179,7 +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())
   {
@@ -208,7 +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 7f6d9f5d8..59a991dca 100644
--- a/xlators/bindings/python/src/testxlator.py
+++ b/xlators/bindings/python/src/testxlator.py
@@ -1,5 +1,5 @@
 """
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  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
diff --git a/xlators/cluster/afr/src/Makefile.am b/xlators/cluster/afr/src/Makefile.am
index 95db5dd96..35d18a6c0 100644
--- a/xlators/cluster/afr/src/Makefile.am
+++ b/xlators/cluster/afr/src/Makefile.am
@@ -7,11 +7,11 @@ afr_common_source = afr-dir-read.c afr-dir-write.c afr-inode-read.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
 
@@ -21,9 +21,11 @@ noinst_HEADERS = afr.h afr-transaction.h afr-inode-write.h afr-inode-read.h \
 	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) \
+AM_CPPFLAGS = $(GF_CPPFLAGS) \
 	-I$(top_srcdir)/libglusterfs/src -I$(top_srcdir)/xlators/lib/src \
-	-I$(top_srcdir)/rpc/rpc-lib/src -shared -nostartfiles $(GF_CFLAGS)
+	-I$(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 ddcd0bca3..af01f2ef2 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -49,10 +49,9 @@
 #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);
@@ -203,59 +202,86 @@ out:
         return ret;
 }
 
-afr_inode_ctx_t*
-afr_inode_ctx_get_from_addr (uint64_t addr, int32_t child_count)
+void
+afr_inode_ctx_destroy (afr_inode_ctx_t *ctx)
 {
-        int             ret  = -1;
-        afr_inode_ctx_t *ctx = NULL;
-        size_t          size = 0;
+        if (!ctx)
+                return;
+        GF_FREE (ctx->fresh_children);
+        GF_FREE (ctx);
+}
 
-        GF_ASSERT (child_count > 0);
+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;
 
-        if (!addr) {
-                ctx = GF_CALLOC (1, sizeof (*ctx),
-                                 gf_afr_mt_inode_ctx_t);
-                if (!ctx)
-                        goto out;
-                size = sizeof (*ctx->fresh_children);
-                ctx->fresh_children = GF_CALLOC (child_count, size,
-                                                 gf_afr_mt_int32_t);
-                if (!ctx->fresh_children)
-                        goto out;
-        } else {
-                ctx = (afr_inode_ctx_t*) (long) addr;
+        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;
         }
-        ret = 0;
+        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;
+        }
+
 out:
-        if (ret && ctx) {
-                GF_FREE (ctx->fresh_children);
-                GF_FREE (ctx);
-                ctx = NULL;
+        return ctx;
+
+fail:
+        afr_inode_ctx_destroy (ctx);
+        return NULL;
+}
+
+afr_inode_ctx_t*
+afr_inode_ctx_get (inode_t *inode, xlator_t *this)
+{
+        afr_inode_ctx_t *ctx = NULL;
+
+        LOCK (&inode->lock);
+        {
+                ctx = __afr_inode_ctx_get (inode, this);
         }
+        UNLOCK (&inode->lock);
         return ctx;
 }
 
 void
-afr_inode_get_ctx (xlator_t *this, inode_t *inode, afr_inode_params_t *params)
+afr_inode_get_ctx_params (xlator_t *this, inode_t *inode,
+                          afr_inode_params_t *params)
 {
         GF_ASSERT (inode);
         GF_ASSERT (params);
 
-        int             ret = 0;
         afr_inode_ctx_t *ctx = NULL;
         afr_private_t   *priv = NULL;
         int             i = 0;
-        uint64_t        ctx_addr   = 0;
         int32_t         read_child = -1;
         int32_t         *fresh_children = NULL;
 
         priv = this->private;
         LOCK (&inode->lock);
         {
-                ret = __inode_ctx_get (inode, this, &ctx_addr);
-                if (ret < 0)
-                        goto unlock;
-                ctx = afr_inode_ctx_get_from_addr (ctx_addr, priv->child_count);
+                ctx = __afr_inode_ctx_get (inode, this);
                 if (!ctx)
                         goto unlock;
                 switch (params->op) {
@@ -274,12 +300,6 @@ afr_inode_get_ctx (xlator_t *this, inode_t *inode, afr_inode_params_t *params)
                         if (ctx->masks & AFR_ICTX_OPENDIR_DONE_MASK)
                                 params->u.value = _gf_true;
                         break;
-                case AFR_INODE_GET_SPLIT_BRAIN:
-                        params->u.value = _gf_false;
-                        if (ctx->masks & AFR_ICTX_SPLIT_BRAIN_MASK)
-                                params->u.value = _gf_true;
-                        ;
-                        break;
                 default:
                         GF_ASSERT (0);
                         break;
@@ -292,11 +312,16 @@ unlock:
 gf_boolean_t
 afr_is_split_brain (xlator_t *this, inode_t *inode)
 {
-        afr_inode_params_t params = {0};
+        afr_inode_ctx_t *ctx = NULL;
+        gf_boolean_t    spb  = _gf_false;
 
-        params.op = AFR_INODE_GET_SPLIT_BRAIN;
-        afr_inode_get_ctx (this, inode, &params);
-        return params.u.value;
+        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 spb;
 }
 
 gf_boolean_t
@@ -305,11 +330,10 @@ afr_is_opendir_done (xlator_t *this, inode_t *inode)
         afr_inode_params_t params = {0};
 
         params.op = AFR_INODE_GET_OPENDIR_DONE;
-        afr_inode_get_ctx (this, inode, &params);
+        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)
 {
@@ -317,7 +341,7 @@ afr_inode_get_read_ctx (xlator_t *this, inode_t *inode, int32_t *fresh_children)
 
         params.op = AFR_INODE_GET_READ_CTX;
         params.u.read_ctx.children = fresh_children;
-        afr_inode_get_ctx (this, inode, &params);
+        afr_inode_get_ctx_params (this, inode, &params);
         return params.u.read_ctx.read_child;
 }
 
@@ -379,31 +403,14 @@ afr_inode_ctx_set_opendir_done (afr_inode_ctx_t *ctx)
 }
 
 void
-afr_inode_ctx_set_splitbrain (afr_inode_ctx_t *ctx, gf_boolean_t set)
-{
-        uint64_t        remaining_mask = 0;
-        uint64_t        mask = 0;
-
-        if (set) {
-                remaining_mask = (~AFR_ICTX_SPLIT_BRAIN_MASK & ctx->masks);
-                mask = (0xFFFFFFFFFFFFFFFFULL & AFR_ICTX_SPLIT_BRAIN_MASK);
-                ctx->masks = remaining_mask | mask;
-        } else {
-                ctx->masks = (~AFR_ICTX_SPLIT_BRAIN_MASK & ctx->masks);
-        }
-}
-
-void
-afr_inode_set_ctx (xlator_t *this, inode_t *inode, afr_inode_params_t *params)
+afr_inode_set_ctx_params (xlator_t *this, inode_t *inode,
+                          afr_inode_params_t *params)
 {
         GF_ASSERT (inode);
         GF_ASSERT (params);
 
-        int             ret = 0;
         afr_inode_ctx_t *ctx            = NULL;
         afr_private_t   *priv           = NULL;
-        uint64_t        ctx_addr        = 0;
-        gf_boolean_t    set             = _gf_false;
         int32_t         read_child      = -1;
         int32_t         *fresh_children = NULL;
         int32_t         *stale_children = NULL;
@@ -411,10 +418,7 @@ afr_inode_set_ctx (xlator_t *this, inode_t *inode, afr_inode_params_t *params)
         priv = this->private;
         LOCK (&inode->lock);
         {
-                ret = __inode_ctx_get (inode, this, &ctx_addr);
-                if (ret < 0)
-                        ctx_addr = 0;
-                ctx = afr_inode_ctx_get_from_addr (ctx_addr, priv->child_count);
+                ctx = __afr_inode_ctx_get (inode, this);
                 if (!ctx)
                         goto unlock;
                 switch (params->op) {
@@ -434,33 +438,26 @@ afr_inode_set_ctx (xlator_t *this, inode_t *inode, afr_inode_params_t *params)
                 case AFR_INODE_SET_OPENDIR_DONE:
                         afr_inode_ctx_set_opendir_done (ctx);
                         break;
-                case AFR_INODE_SET_SPLIT_BRAIN:
-                        set = params->u.value;
-                        afr_inode_ctx_set_splitbrain (ctx, set);
-                        break;
                 default:
                         GF_ASSERT (0);
                         break;
                 }
-                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));
-                }
         }
 unlock:
         UNLOCK (&inode->lock);
 }
 
 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)
 {
-        afr_inode_params_t      params = {0};
+        afr_inode_ctx_t *ctx = NULL;
 
-        params.op = AFR_INODE_SET_SPLIT_BRAIN;
-        params.u.value          = set;
-        afr_inode_set_ctx (this, inode, &params);
+        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
@@ -469,7 +466,7 @@ afr_set_opendir_done (xlator_t *this, inode_t *inode)
         afr_inode_params_t params = {0};
 
         params.op = AFR_INODE_SET_OPENDIR_DONE;
-        afr_inode_set_ctx (this, inode, &params);
+        afr_inode_set_ctx_params (this, inode, &params);
 }
 
 void
@@ -488,7 +485,7 @@ afr_inode_set_read_ctx (xlator_t *this, inode_t *inode, int32_t 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 (this, inode, &params);
+        afr_inode_set_ctx_params (this, inode, &params);
 }
 
 void
@@ -501,7 +498,7 @@ afr_inode_rm_stale_children (xlator_t *this, inode_t *inode,
 
         params.op = AFR_INODE_RM_STALE_CHILDREN;
         params.u.read_ctx.children = stale_children;
-        afr_inode_set_ctx (this, inode, &params);
+        afr_inode_set_ctx_params (this, inode, &params);
 }
 
 gf_boolean_t
@@ -570,6 +567,7 @@ 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;
 
         if (!hmode) {
                 return -1;
@@ -588,7 +586,8 @@ afr_hash_child (int32_t *success_children, int32_t child_count,
                  * perfection here.  All we need is a low probability that
                  * multiple clients won't converge on the same subvolume.
                  */
-                *((pid_t *)gfid_copy) = getpid();
+                pid = getpid();
+                memcpy (gfid_copy, &pid, sizeof(pid));
         }
 
         return SuperFastHash((char *)gfid_copy,
@@ -765,6 +764,13 @@ out:
 }
 
 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;
@@ -773,6 +779,12 @@ afr_local_sh_cleanup (afr_local_t *local, xlator_t *this)
         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->metadata_sh_info && strcmp (sh->metadata_sh_info, ""))
+                GF_FREE (sh->metadata_sh_info);
+
         GF_FREE (sh->buf);
 
         GF_FREE (sh->parentbufs);
@@ -780,10 +792,7 @@ afr_local_sh_cleanup (afr_local_t *local, xlator_t *this)
         if (sh->inode)
                 inode_unref (sh->inode);
 
-        if (sh->xattr) {
-                afr_reset_xattr (sh->xattr, priv->child_count);
-                GF_FREE (sh->xattr);
-        }
+        afr_xattr_array_destroy (sh->xattr, priv->child_count);
 
         GF_FREE (sh->child_errno);
 
@@ -823,7 +832,8 @@ afr_local_sh_cleanup (afr_local_t *local, xlator_t *this)
 void
 afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this)
 {
-        afr_private_t * priv = NULL;
+        afr_private_t *priv    = NULL;
+        int           i        = 0;
 
         priv = this->private;
 
@@ -833,12 +843,13 @@ afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this)
 
         GF_FREE (local->internal_lock.locked_nodes);
 
-        GF_FREE (local->internal_lock.inode_locked_nodes);
-
-        GF_FREE (local->internal_lock.entry_locked_nodes);
+        for (i = 0; local->internal_lock.inodelk[i].domain; i++) {
+                GF_FREE (local->internal_lock.inodelk[i].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);
@@ -848,6 +859,8 @@ afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this)
 
         loc_wipe (&local->transaction.parent_loc);
         loc_wipe (&local->transaction.new_parent_loc);
+
+        GF_FREE (local->transaction.postop_piggybacked);
 }
 
 
@@ -877,14 +890,14 @@ afr_local_cleanup (afr_local_t *local, xlator_t *this)
         if (local->dict)
                 dict_unref (local->dict);
 
+	GF_FREE(local->replies);
+
         GF_FREE (local->child_up);
 
         GF_FREE (local->child_errno);
 
         GF_FREE (local->fresh_children);
 
-        GF_FREE (local->fd_open_on);
-
         { /* lookup */
                 if (local->cont.lookup.xattrs) {
                         afr_reset_xattr (local->cont.lookup.xattrs,
@@ -1067,6 +1080,88 @@ afr_update_loc_gfids (loc_t *loc, struct iatt *buf, struct iatt *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'.
+ * */
+
+static void
+afr_handle_quota_size (afr_local_t *local, xlator_t *this,
+                       dict_t *rsp_dict)
+{
+        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;
+        }
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (sources[i]) {
+                        source_present = _gf_true;
+                        break;
+                }
+        }
+
+        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);
+                }
+        }
+
+        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);
+                }
+
+                data_unref (max_data);
+        }
+}
+
 int
 afr_lookup_build_response_params (afr_local_t *local, xlator_t *this)
 {
@@ -1111,13 +1206,18 @@ afr_lookup_build_response_params (afr_local_t *local, xlator_t *this)
                 ret = -1;
                 goto out;
         }
+
         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;
@@ -1155,31 +1255,57 @@ afr_lookup_update_lk_counts (afr_local_t *local, xlator_t *this,
                 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_set_self_heal_params_by_xattr (afr_local_t *local, xlator_t *this,
                                           dict_t *xattr)
 {
+        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);
 
-        if (afr_sh_has_metadata_pending (xattr, this)) {
-                local->self_heal.do_metadata_self_heal = _gf_true;
-                gf_log(this->name, GF_LOG_DEBUG,
-                       "metadata self-heal is pending for %s.",
-                       local->loc.path);
-        }
+        priv = this->private;
 
-        if (afr_sh_has_entry_pending (xattr, this)) {
-                local->self_heal.do_entry_self_heal = _gf_true;
-                gf_log(this->name, GF_LOG_DEBUG,
-                       "entry self-heal is pending for %s.", local->loc.path);
-        }
+        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 (afr_sh_has_data_pending (xattr, this)) {
-                local->self_heal.do_data_self_heal = _gf_true;
-                gf_log(this->name, GF_LOG_DEBUG,
-                       "data self-heal is pending for %s.", local->loc.path);
+                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;
+                }
+
+                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;
+                }
+
+                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;
+                }
         }
 }
 
@@ -1240,7 +1366,7 @@ afr_detect_self_heal_by_iatt (afr_local_t *local, xlator_t *this,
 
         if (uuid_compare (buf->ia_gfid, lookup_buf->ia_gfid)) {
                 /* mismatching gfid */
-                gf_log (this->name, GF_LOG_WARNING,
+                gf_log (this->name, GF_LOG_DEBUG,
                         "%s: gfid different on subvolume", local->loc.path);
         }
 }
@@ -1464,7 +1590,7 @@ afr_conflicting_iattrs (struct iatt *bufs, int32_t *success_children,
 
                 child2 = &bufs[success_children[i-1]];
                 if (FILETYPE_DIFFERS (child1, child2)) {
-                        gf_log (xlator_name, GF_LOG_WARNING, "%s: filetype "
+                        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;
@@ -1473,7 +1599,7 @@ afr_conflicting_iattrs (struct iatt *bufs, int32_t *success_children,
                 if (!gfid || uuid_is_null (child1->ia_gfid))
                         continue;
                 if (uuid_compare (*gfid, child1->ia_gfid)) {
-                       gf_log (xlator_name, GF_LOG_WARNING, "%s: gfid differs"
+                       gf_log (xlator_name, GF_LOG_DEBUG, "%s: gfid differs"
                                " on subvolume %d", path, success_children[i]);
                        conflicting = _gf_true;
                        goto out;
@@ -1607,8 +1733,8 @@ afr_self_heal_lookup_unwind (call_frame_t *frame, xlator_t *this,
 
         if (op_ret == -1) {
                 local->op_ret = -1;
-                if (afr_error_more_important (local->op_errno, op_errno))
-                        local->op_errno = op_errno;
+		local->op_errno = afr_most_important_error(local->op_errno,
+							   op_errno, _gf_true);
 
                 goto out;
         } else {
@@ -1622,6 +1748,16 @@ afr_self_heal_lookup_unwind (call_frame_t *frame, xlator_t *this,
                 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,
@@ -1696,7 +1832,8 @@ afr_lookup_perform_self_heal (call_frame_t *frame, xlator_t *this,
 
         afr_lookup_set_self_heal_params (local, this);
         if (afr_can_self_heal_proceed (&local->self_heal, priv)) {
-                if  (afr_is_transaction_running (local))
+                if  (afr_is_transaction_running (local) &&
+                     (!local->allow_sh_for_running_transaction))
                         goto out;
 
                 reason = "lookup detected pending operations";
@@ -1875,6 +2012,57 @@ out:
         return;
 }
 
+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;
+}
+
+
 static void
 afr_lookup_done (call_frame_t *frame, xlator_t *this)
 {
@@ -1891,6 +2079,12 @@ afr_lookup_done (call_frame_t *frame, xlator_t *this)
         priv  = this->private;
         local = frame->local;
 
+	if (afr_is_entry_possibly_under_creation (local, this)) {
+		local->op_ret = -1;
+		local->op_errno = ENOENT;
+		goto unwind;
+	}
+
         if (local->op_ret < 0)
                 goto unwind;
 
@@ -1948,25 +2142,20 @@ afr_lookup_done (call_frame_t *frame, xlator_t *this)
  * others in that they must be given higher priority while
  * returning to the user.
  *
- * The hierarchy is ESTALE > ENOENT > others
- *
+ * The hierarchy is ESTALE > EIO > ENOENT > others
  */
-
-gf_boolean_t
-afr_error_more_important (int32_t old_errno, int32_t new_errno)
+int32_t
+afr_most_important_error(int32_t old_errno, int32_t new_errno,
+			 gf_boolean_t eio)
 {
-        gf_boolean_t ret = _gf_true;
-
-        /* Nothing should ever overwrite ESTALE */
-        if (old_errno == ESTALE)
-                ret = _gf_false;
+	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;
 
-        /* Nothing should overwrite ENOENT, except ESTALE/EIO*/
-        else if ((old_errno == ENOENT) && (new_errno != ESTALE)
-                 && (new_errno != EIO))
-                ret = _gf_false;
-
-        return ret;
+	return new_errno;
 }
 
 int32_t
@@ -1985,8 +2174,9 @@ afr_resultant_errno_get (int32_t *children,
                 } else {
                         child = i;
                 }
-                if (afr_error_more_important (op_errno, child_errno[child]))
-                                op_errno = child_errno[child];
+		op_errno = afr_most_important_error(op_errno,
+						    child_errno[child],
+						    _gf_false);
         }
         return op_errno;
 }
@@ -1998,8 +2188,8 @@ afr_lookup_handle_error (afr_local_t *local, int32_t op_ret,  int32_t op_errno)
         if (op_errno == ENOENT)
                 local->enoent_count++;
 
-        if (afr_error_more_important (local->op_errno, op_errno))
-                local->op_errno = op_errno;
+	local->op_errno = afr_most_important_error(local->op_errno, op_errno,
+						   _gf_false);
 
         if (local->op_errno == ESTALE) {
                 local->op_ret = -1;
@@ -2238,7 +2428,7 @@ afr_lookup (call_frame_t *frame, xlator_t *this,
         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);
@@ -2255,6 +2445,13 @@ afr_lookup (call_frame_t *frame, xlator_t *this,
                 goto out;
         }
 
+        if (local->loc.path &&
+            (strcmp (local->loc.path, "/" GF_REPLICATE_TRASH_DIR) == 0)) {
+                op_errno = EPERM;
+                ret = -1;
+                goto out;
+        }
+
         ret = inode_ctx_get (local->loc.inode, this, &ctx);
         if (ret == 0) {
                 /* lookup is a revalidate */
@@ -2303,6 +2500,11 @@ afr_lookup (call_frame_t *frame, xlator_t *this,
         /* By default assume ENOTCONN. On success it will be set to 0. */
         local->op_errno = ENOTCONN;
 
+        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;
+
         ret = afr_lookup_xattr_req_prepare (local, this, xattr_req, &local->loc,
                                             &gfid_req);
         if (ret) {
@@ -2419,8 +2621,10 @@ __afr_fd_ctx_set (xlator_t *this, fd_t *fd)
         }
 
 	pthread_mutex_init (&fd_ctx->delay_lock, NULL);
-        INIT_LIST_HEAD (&fd_ctx->paused_calls);
         INIT_LIST_HEAD (&fd_ctx->entries);
+        fd_ctx->call_child = -1;
+
+        INIT_LIST_HEAD (&fd_ctx->eager_locked);
 
         ret = __fd_ctx_set (fd, this, (uint64_t)(long) fd_ctx);
         if (ret)
@@ -2448,135 +2652,70 @@ afr_fd_ctx_set (xlator_t *this, fd_t *fd)
 /* {{{ 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,
-                                  NULL);
-        }
-
-        return 0;
-}
-
-
-int
-afr_flush_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+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;
-        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);
 
-        call_count = afr_frame_return (frame);
-
-        if (call_count == 0) {
-                local->transaction.resume (frame, this);
-        }
+	if (call_count == 0)
+		AFR_STACK_UNWIND(flush, frame, local->op_ret,
+				 local->op_errno, NULL);
 
         return 0;
 }
 
-
-int
-afr_flush_wind (call_frame_t *frame, xlator_t *this)
+static int
+afr_flush_wrapper (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
 {
-        afr_local_t *local = NULL;
-        afr_private_t *priv = NULL;
-        int i = 0;
-        int call_count = -1;
+        int           i      = 0;
+        afr_local_t   *local = NULL;
+        afr_private_t *priv  = NULL;
+        int call_count       = -1;
 
-        local = frame->local;
         priv = this->private;
-
-        call_count = afr_up_children_count (local->child_up, priv->child_count);
-
-        if (call_count == 0) {
-                local->transaction.resume (frame, this);
-                return 0;
-        }
-
-        local->call_count = call_count;
+        local = frame->local;
+        call_count = local->call_count;
 
         for (i = 0; i < priv->child_count; i++) {
                 if (local->child_up[i]) {
-                        STACK_WIND_COOKIE (frame, afr_flush_wind_cbk,
+                        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;
+
                 }
         }
 
         return 0;
 }
 
-
-int
-afr_flush_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;
-}
-
-
 int
 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;
+        call_stub_t   *stub = NULL;
         int            ret        = -1;
         int            op_errno   = 0;
 
@@ -2586,47 +2725,27 @@ afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
 
         priv = this->private;
 
-        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->op = GF_FOP_FLUSH;
+	AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
+	local = frame->local;
 
-        local->transaction.fop    = afr_flush_wind;
-        local->transaction.done   = afr_flush_done;
-        local->transaction.unwind = afr_flush_unwind;
+	ret = afr_local_init(local, priv, &op_errno);
+	if (ret < 0)
+		goto out;
 
-        local->fd                 = fd_ref (fd);
-
-        local->transaction.main_frame = frame;
-        local->transaction.start  = 0;
-        local->transaction.len    = 0;
-
-        ret = afr_open_fd_fix (transaction_frame, this, _gf_false);
-        if (ret) {
-                op_errno = -ret;
+	local->fd = fd_ref(fd);
+        stub = fop_flush_stub (frame, afr_flush_wrapper, fd, xdata);
+        if (!stub) {
+                ret = -1;
+                op_errno = ENOMEM;
                 goto out;
         }
-        afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
 
+        afr_delayed_changelog_wake_resume (this, fd, stub);
+	ret = 0;
 
-        ret = 0;
 out:
-        if (ret < 0) {
-                if (transaction_frame)
-                        AFR_STACK_DESTROY (transaction_frame);
-
-                AFR_STACK_UNWIND (flush, frame, -1, op_errno, NULL);
-        }
+	if (ret < 0)
+		AFR_STACK_UNWIND(flush, frame, -1, op_errno, NULL);
 
         return 0;
 }
@@ -2640,8 +2759,6 @@ afr_cleanup_fd_ctx (xlator_t *this, fd_t *fd)
         uint64_t        ctx = 0;
         afr_fd_ctx_t    *fd_ctx = NULL;
         int             ret = 0;
-        afr_fd_paused_call_t *paused_call = NULL;
-        afr_fd_paused_call_t *tmp = NULL;
 
         ret = fd_ctx_get (fd, this, &ctx);
         if (ret < 0)
@@ -2657,12 +2774,6 @@ afr_cleanup_fd_ctx (xlator_t *this, fd_t *fd)
                 GF_FREE (fd_ctx->locked_on);
 
                 GF_FREE (fd_ctx->pre_op_piggyback);
-                list_for_each_entry_safe (paused_call, tmp, &fd_ctx->paused_calls,
-                                          call_list) {
-                        list_del_init (&paused_call->call_list);
-                        GF_FREE (paused_call);
-                }
-
                 GF_FREE (fd_ctx->lock_piggyback);
 
                 GF_FREE (fd_ctx->lock_acquired);
@@ -2705,6 +2816,16 @@ 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, dict_t *xdata)
@@ -2713,6 +2834,7 @@ afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         int call_count = -1;
         int child_index = (long) cookie;
         int read_child  = 0;
+	call_stub_t *stub = NULL;
 
         local = frame->local;
 
@@ -2728,13 +2850,13 @@ afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->op_ret = 0;
 
                         if (local->success_count == 0) {
-                                local->cont.fsync.prebuf  = *prebuf;
-                                local->cont.fsync.postbuf = *postbuf;
+                                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++;
@@ -2747,10 +2869,32 @@ afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         call_count = afr_frame_return (frame);
 
         if (call_count == 0) {
-                AFR_STACK_UNWIND (fsync, frame, local->op_ret, local->op_errno,
-                                  &local->cont.fsync.prebuf,
-                                  &local->cont.fsync.postbuf,
-                                  NULL);
+		/* 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;
+		}
+
+		/* 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;
@@ -2785,6 +2929,10 @@ afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
 
         local->fd             = fd_ref (fd);
 
+	if (afr_fd_has_witnessed_unstable_write (this, fd)) {
+		/* don't care. we only wanted to CLEAR the bit */
+	}
+
         for (i = 0; i < priv->child_count; i++) {
                 if (local->child_up[i]) {
                         STACK_WIND_COOKIE (frame, afr_fsync_cbk,
@@ -3942,6 +4090,17 @@ afr_local_init (afr_local_t *local, afr_private_t *priv, int32_t *op_errno)
                 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;
@@ -3953,16 +4112,6 @@ afr_internal_lock_init (afr_internal_lock_t *lk, size_t child_count,
 {
         int             ret = -ENOMEM;
 
-        lk->inode_locked_nodes = GF_CALLOC (sizeof (*lk->inode_locked_nodes),
-                                            child_count, gf_afr_mt_char);
-        if (NULL == lk->inode_locked_nodes)
-                goto out;
-
-        lk->entry_locked_nodes = GF_CALLOC (sizeof (*lk->entry_locked_nodes),
-                                            child_count, gf_afr_mt_char);
-        if (NULL == lk->entry_locked_nodes)
-                goto out;
-
         lk->locked_nodes = GF_CALLOC (sizeof (*lk->locked_nodes),
                                       child_count, gf_afr_mt_char);
         if (NULL == lk->locked_nodes)
@@ -4021,6 +4170,21 @@ out:
 }
 
 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;
@@ -4033,6 +4197,14 @@ afr_transaction_local_init (afr_local_t *local, xlator_t *this)
         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);
@@ -4054,14 +4226,6 @@ afr_transaction_local_init (afr_local_t *local, xlator_t *this)
         if (!local->fresh_children)
                 goto out;
 
-        if (local->fd) {
-                local->fd_open_on = GF_CALLOC (sizeof (*local->fd_open_on),
-                                               priv->child_count,
-                                               gf_afr_mt_char);
-                if (!local->fd_open_on)
-                        goto out;
-        }
-
         local->transaction.pre_op = GF_CALLOC (sizeof (*local->transaction.pre_op),
                                                priv->child_count,
                                                gf_afr_mt_char);
@@ -4077,6 +4241,9 @@ afr_transaction_local_init (afr_local_t *local, xlator_t *this)
                                                            AFR_NUM_CHANGE_LOGS);
         if (!local->transaction.txn_changelog)
                 goto out;
+
+	INIT_LIST_HEAD (&local->transaction.eager_locked);
+
         ret = 0;
 out:
         return ret;
@@ -4264,6 +4431,16 @@ afr_priv_destroy (afr_private_t *priv)
         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++)
@@ -4290,3 +4467,125 @@ xlator_subvolume_count (xlator_t *this)
                 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 ce91ffba7..689dd84e6 100644
--- a/xlators/cluster/afr/src/afr-dir-read.c
+++ b/xlators/cluster/afr/src/afr-dir-read.c
@@ -253,7 +253,7 @@ unlock:
                         goto out;
 
                 if (!afr_is_opendir_done (this, local->fd->inode) &&
-                    up_children_count > 1) {
+                    up_children_count > 1 && priv->entry_self_heal) {
 
                         /*
                          * This is the first opendir on this inode. We need
@@ -358,85 +358,6 @@ struct entry_name {
         struct list_head list;
 };
 
-
-static gf_boolean_t
-remembered_name (const char *name, struct list_head *entries)
-{
-        struct entry_name *e   = NULL;
-        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    = 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 (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)
-{
-        gf_dirent_t  *entry  = NULL;
-        gf_dirent_t  *tmp    = NULL;
-        int           ret    = 0;
-        uint64_t      ctx    = 0;
-        afr_fd_ctx_t *fd_ctx = NULL;
-        off_t         offset = 0;
-
-        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 -1;
-        }
-
-        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;
-}
-
-
 static void
 afr_forget_entries (fd_t *fd)
 {
@@ -462,14 +383,36 @@ afr_forget_entries (fd_t *fd)
         }
 }
 
+static void
+afr_readdir_filter_trash_dir (gf_dirent_t *entries, fd_t *fd)
+{
+        gf_dirent_t *   entry       = NULL;
+        gf_dirent_t *   tmp         = NULL;
+
+        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, dict_t *xdata)
 {
-        AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, entries, NULL);
+        afr_local_t     *local = NULL;
+
+        if (op_ret == -1)
+                goto out;
+
+        local = frame->local;
+        afr_readdir_filter_trash_dir (entries, local->fd);
 
+out:
+        AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, entries, NULL);
         return 0;
 }
 
@@ -479,113 +422,16 @@ afr_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)
 {
-        afr_private_t *  priv            = NULL;
-        afr_local_t *    local           = NULL;
-        xlator_t **      children        = NULL;
-        int32_t          next_call_child = -1;
-        int              ret             = 0;
-        int32_t          *last_index     = NULL;
-        int32_t          read_child      = -1;
-        int32_t         *fresh_children   = NULL;
-        uint64_t         ctx             = 0;
-        afr_fd_ctx_t    *fd_ctx          = NULL;
-        off_t            offset          = 0;
-        int32_t         call_child       = -1;
+        afr_local_t     *local = NULL;
 
-        priv     = this->private;
-        children = priv->children;
+        if (op_ret == -1)
+                goto out;
 
         local = frame->local;
-
-        read_child = (long) cookie;
-        last_index = &local->cont.readdir.last_index;
-        fresh_children = local->fresh_children;
-
-        /* the value of the last_index changes if afr_next_call_child is
-         * called. So to find the call_child of this callback use last_index
-         * before the next_call_child call.
-         */
-        if (*last_index == -1)
-                call_child = read_child;
-        else
-                call_child = fresh_children[*last_index];
-
-        if (priv->strict_readdir) {
-                ret = fd_ctx_get (local->fd, this, &ctx);
-                if (ret < 0) {
-                        gf_log (this->name, GF_LOG_INFO,
-                                "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 (op_ret == -1) {
-                        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;
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "starting readdir afresh on child %d, offset %"PRId64,
-                                next_call_child, (uint64_t) 0);
-
-                        fd_ctx->failed_over = _gf_true;
-
-                        STACK_WIND_COOKIE (frame, afr_readdirp_cbk,
-                                           (void *) (long) read_child,
-                                           children[next_call_child],
-                                           children[next_call_child]->fops->readdirp,
-                                           local->fd,
-                                           local->cont.readdir.size, 0,
-                                           local->cont.readdir.dict);
-                        return 0;
-                }
-        }
-
-        if (priv->strict_readdir) {
-                if (fd_ctx->failed_over) {
-                        if (list_empty (&entries->list)) {
-                                gf_log (this->name, GF_LOG_DEBUG,
-                                        "no entries found");
-                                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 will 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[call_child]->name);
-
-                                STACK_WIND_COOKIE (frame, afr_readdirp_cbk,
-                                                   (void *) (long) read_child,
-                                                   children[call_child],
-                                                   children[call_child]->fops->readdirp,
-                                                   local->fd, local->cont.readdir.size, offset,
-                                                   local->cont.readdir.dict);
-                                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, NULL);
-
         return 0;
 }
 
@@ -593,15 +439,14 @@ int32_t
 afr_do_readdir (call_frame_t *frame, xlator_t *this,
                 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;
-        uint64_t         ctx        = 0;
-        afr_fd_ctx_t    *fd_ctx     = NULL;
-        int              ret        = -1;
-        int32_t          op_errno   = 0;
-        uint64_t         read_child = 0;
+        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);
@@ -626,42 +471,32 @@ afr_do_readdir (call_frame_t *frame, xlator_t *this,
         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);
+                                  local->fresh_children,
+                                  &call_child,
+                                  &local->cont.readdir.last_index);
         if (ret < 0) {
                 op_errno = -ret;
                 goto out;
         }
 
+        fd_ctx  = afr_fd_ctx_get (fd, this);
+        if (!fd_ctx) {
+                op_errno = EBADF;
+                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;
         local->cont.readdir.dict   = (dict)? dict_ref (dict) : NULL;
 
-        if (priv->strict_readdir) {
-                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);
-                        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;
-        }
-
         if (whichop == GF_FOP_READDIR)
                 STACK_WIND_COOKIE (frame, afr_readdir_cbk,
                                    (void *) (long) call_child,
@@ -675,10 +510,9 @@ afr_do_readdir (call_frame_t *frame, xlator_t *this,
                                    children[call_child]->fops->readdirp, fd,
                                    size, offset, dict);
 
-        ret = 0;
+        return 0;
 out:
-        if (ret < 0)
-                AFR_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
+        AFR_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
diff --git a/xlators/cluster/afr/src/afr-dir-write.c b/xlators/cluster/afr/src/afr-dir-write.c
index 304103541..1943b719b 100644
--- a/xlators/cluster/afr/src/afr-dir-write.c
+++ b/xlators/cluster/afr/src/afr-dir-write.c
@@ -106,11 +106,134 @@ __dir_entry_fop_common_cbk (call_frame_t *frame, int child_index,
 
                 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;
         }
 
         local->op_errno = op_errno;
 }
 
+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)
 {
@@ -129,6 +252,7 @@ afr_dir_fop_done (call_frame_t *frame, xlator_t *this)
                                       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);
 }
 
@@ -295,11 +419,12 @@ afr_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)
 {
-        afr_private_t  *priv  = NULL;
-        afr_local_t    *local = NULL;
-        call_frame_t   *transaction_frame = NULL;
-        int             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);
@@ -331,6 +456,7 @@ afr_create (call_frame_t *frame, xlator_t *this,
         }
         UNLOCK (&priv->read_child_lock);
 
+        local->op                = GF_FOP_CREATE;
         local->cont.create.flags = flags;
         local->cont.create.mode  = mode;
         local->cont.create.fd    = fd_ref (fd);
@@ -349,8 +475,22 @@ afr_create (call_frame_t *frame, xlator_t *this,
 
         local->transaction.main_frame = frame;
         local->transaction.basename = AFR_BASENAME (loc->path);
+        int_lock = &local->internal_lock;
 
-        afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        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:
@@ -484,11 +624,12 @@ int
 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;
-        int 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);
@@ -520,6 +661,7 @@ afr_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
         }
         UNLOCK (&priv->read_child_lock);
 
+        local->op               = GF_FOP_MKNOD;
         local->cont.mknod.mode  = mode;
         local->cont.mknod.dev   = dev;
         local->umask = umask;
@@ -537,8 +679,22 @@ afr_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
 
         local->transaction.main_frame = frame;
         local->transaction.basename = AFR_BASENAME (loc->path);
+        int_lock = &local->internal_lock;
 
-        afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        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:
@@ -669,16 +825,16 @@ afr_mkdir_done (call_frame_t *frame, xlator_t *this)
         return 0;
 }
 
-
 int
 afr_mkdir (call_frame_t *frame, xlator_t *this,
            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_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);
@@ -715,6 +871,7 @@ afr_mkdir (call_frame_t *frame, xlator_t *this,
         if (params)
                 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;
@@ -726,8 +883,22 @@ afr_mkdir (call_frame_t *frame, xlator_t *this,
 
         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;
 
-        afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        int_lock->lockee_count++;
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
 
         ret = 0;
 out:
@@ -829,7 +1000,8 @@ afr_link_wind (call_frame_t *frame, xlator_t *this)
 
         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,
+                        STACK_WIND_COOKIE (frame, afr_link_wind_cbk,
+                                           (void *) (long) i,
                                            priv->children[i],
                                            priv->children[i]->fops->link,
                                            &local->loc,
@@ -861,11 +1033,12 @@ int
 afr_link (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;
-        call_frame_t  * transaction_frame = NULL;
-        int 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);
@@ -900,6 +1073,7 @@ afr_link (call_frame_t *frame, xlator_t *this,
         }
         UNLOCK (&priv->read_child_lock);
 
+        local->op = GF_FOP_LINK;
         local->transaction.fop    = afr_link_wind;
         local->transaction.done   = afr_link_done;
         local->transaction.unwind = afr_link_unwind;
@@ -911,9 +1085,22 @@ afr_link (call_frame_t *frame, xlator_t *this,
 
         local->transaction.main_frame   = frame;
         local->transaction.basename     = AFR_BASENAME (newloc->path);
+        int_lock = &local->internal_lock;
 
-        afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        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 (ret < 0) {
@@ -1049,11 +1236,12 @@ int
 afr_symlink (call_frame_t *frame, xlator_t *this,
              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;
-        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);
@@ -1090,6 +1278,7 @@ afr_symlink (call_frame_t *frame, xlator_t *this,
         if (params)
                 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;
@@ -1101,8 +1290,22 @@ afr_symlink (call_frame_t *frame, xlator_t *this,
 
         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;
 
-        afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        int_lock->lockee_count++;
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
 
         ret = 0;
 out:
@@ -1172,6 +1375,7 @@ afr_rename_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 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,
@@ -1246,11 +1450,13 @@ 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;
-        call_frame_t  * transaction_frame = NULL;
-        int 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;
+        int                     nlockee                 = 0;
 
         VALIDATE_OR_GOTO (frame, out);
         VALIDATE_OR_GOTO (this, out);
@@ -1278,6 +1484,7 @@ afr_rename (call_frame_t *frame, xlator_t *this,
 
         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;
@@ -1294,8 +1501,44 @@ afr_rename (call_frame_t *frame, xlator_t *this,
         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;
 
-        afr_transaction (transaction_frame, this, AFR_ENTRY_RENAME_TRANSACTION);
+        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;
+        }
 
         ret = 0;
 out:
@@ -1429,11 +1672,12 @@ int32_t
 afr_unlink (call_frame_t *frame, xlator_t *this,
             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;
-        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);
@@ -1461,6 +1705,7 @@ afr_unlink (call_frame_t *frame, xlator_t *this,
         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;
@@ -1472,8 +1717,22 @@ afr_unlink (call_frame_t *frame, xlator_t *this,
 
         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;
 
-        afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        int_lock->lockee_count++;
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
 
         ret = 0;
 out:
@@ -1542,6 +1801,7 @@ afr_rmdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 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,
@@ -1615,11 +1875,13 @@ int
 afr_rmdir (call_frame_t *frame, xlator_t *this,
            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_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;
+        int                     nlockee                 = 0;
 
         VALIDATE_OR_GOTO (frame, out);
         VALIDATE_OR_GOTO (this, out);
@@ -1645,6 +1907,7 @@ afr_rmdir (call_frame_t *frame, xlator_t *this,
         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;
@@ -1656,8 +1919,34 @@ afr_rmdir (call_frame_t *frame, xlator_t *this,
 
         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;
 
-        afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        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;
+        }
 
         ret = 0;
 out:
diff --git a/xlators/cluster/afr/src/afr-inode-read.c b/xlators/cluster/afr/src/afr-inode-read.c
index 4679e4610..e06e3b2f2 100644
--- a/xlators/cluster/afr/src/afr-inode-read.c
+++ b/xlators/cluster/afr/src/afr-inode-read.c
@@ -117,6 +117,8 @@ afr_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
 
         children = priv->children;
 
+        AFR_SBRAIN_CHECK_LOC (loc, out);
+
         AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
         local = frame->local;
 
@@ -232,6 +234,8 @@ afr_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 
         children = priv->children;
 
+        AFR_SBRAIN_CHECK_LOC (loc, out);
+
         AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
         local = frame->local;
 
@@ -348,6 +352,8 @@ afr_fstat (call_frame_t *frame, xlator_t *this,
 
         VALIDATE_OR_GOTO (fd->inode, out);
 
+        AFR_SBRAIN_CHECK_FD (fd, out);
+
         AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
         local = frame->local;
 
@@ -377,11 +383,8 @@ afr_fstat (call_frame_t *frame, xlator_t *this,
 
         local->fd = fd_ref (fd);
 
-        ret = afr_open_fd_fix (frame, this, _gf_false);
-        if (ret) {
-                op_errno = -ret;
-                goto out;
-        }
+        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,
@@ -470,6 +473,8 @@ afr_readlink (call_frame_t *frame, xlator_t *this,
 
         children = priv->children;
 
+        AFR_SBRAIN_CHECK_LOC (loc, out);
+
         AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
         local = frame->local;
 
@@ -629,6 +634,96 @@ afr_getxattr_unwind (call_frame_t *frame, int op_ret, int op_errno,
 }
 
 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_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)
@@ -762,12 +857,364 @@ afr_getxattr_node_uuid_cbk (call_frame_t *frame, void *cookie,
 
  unwind:
         if (unwind)
-                AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, NULL);
+                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;
+                }
+
+                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 (fgetxattr, frame, op_ret,
+                                  op_errno, newdict,
+                                  local->xdata_rsp);
+        }
+
+        dict_unref (lockinfo);
+
+        return 0;
+}
+
+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;
+        }
+
+        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 (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)
@@ -879,8 +1326,65 @@ afr_getxattr_pathinfo_cbk (call_frame_t *frame, void *cookie,
         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);
+        }
+
+out:
+        return 0;
+}
+
+
 static gf_boolean_t
-afr_is_special_xattr (const char *name, fop_getxattr_cbk_t *cbk)
+afr_is_special_xattr (const char *name, fop_getxattr_cbk_t *cbk,
+                      gf_boolean_t is_fgetxattr)
 {
         gf_boolean_t    is_spl = _gf_true;
 
@@ -890,14 +1394,31 @@ afr_is_special_xattr (const char *name, fop_getxattr_cbk_t *cbk)
                 goto out;
         }
 
-        if (!strcmp (name, GF_XATTR_PATHINFO_KEY))
-                *cbk = afr_getxattr_pathinfo_cbk;
-
-        else if (!strncmp (name, GF_XATTR_CLRLK_CMD,
-                           strlen (GF_XATTR_CLRLK_CMD)))
-                *cbk = afr_getxattr_clrlk_cbk;
-        else
+        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 {
                 is_spl = _gf_false;
+        }
 
 out:
         return is_spl;
@@ -943,7 +1464,7 @@ afr_getxattr (call_frame_t *frame, xlator_t *this,
         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);
@@ -954,6 +1475,8 @@ afr_getxattr (call_frame_t *frame, xlator_t *this,
 
         children = priv->children;
 
+        AFR_SBRAIN_CHECK_LOC (loc, out);
+
         AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
         local = frame->local;
 
@@ -976,7 +1499,7 @@ afr_getxattr (call_frame_t *frame, xlator_t *this,
                 goto out;
         }
         if ((strcmp (GF_XATTR_MARKER_KEY, name) == 0)
-            && (-1 == frame->root->pid)) {
+            && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
 
                 local->marker.call_count = priv->child_count;
 
@@ -992,6 +1515,7 @@ afr_getxattr (call_frame_t *frame, xlator_t *this,
                                            sub_volumes,
                                            priv->child_count,
                                            MARKER_UUID_TYPE,
+                                           marker_uuid_default_gauge,
                                            priv->vol_uuid)) {
 
                         gf_log (this->name, GF_LOG_INFO,
@@ -1008,7 +1532,7 @@ afr_getxattr (call_frame_t *frame, xlator_t *this,
          * if we are doing getxattr with pathinfo as the key then we
          * collect information from all childs
          */
-        if (afr_is_special_xattr (name, &cbk)) {
+        if (afr_is_special_xattr (name, &cbk, 0)) {
                 afr_getxattr_frm_all_children (this, frame, name,
                                                loc, cbk);
                 return 0;
@@ -1026,10 +1550,11 @@ afr_getxattr (call_frame_t *frame, xlator_t *this,
 
         if (*priv->vol_uuid) {
                 if ((match_uuid_local (name, priv->vol_uuid) == 0)
-                    && (-1 == frame->root->pid)) {
+                    && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
                         local->marker.call_count = priv->child_count;
 
-                        sub_volumes = alloca ( priv->child_count * sizeof (xlator_t *));
+                        sub_volumes = alloca ( priv->child_count
+                                               * sizeof (xlator_t *));
                         for (i = 0, trav = this->children; trav ;
                              trav = trav->next, i++) {
 
@@ -1037,12 +1562,20 @@ afr_getxattr (call_frame_t *frame, xlator_t *this,
 
                         }
 
+                        /* 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)",
@@ -1062,7 +1595,8 @@ no_name:
                 goto out;
         }
 
-        read_child = afr_inode_get_read_ctx (this, loc->inode, local->fresh_children);
+        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,
@@ -1150,18 +1684,44 @@ afr_fgetxattr_unwind (call_frame_t *frame,
         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;
-
+        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);
@@ -1172,6 +1732,8 @@ afr_fgetxattr (call_frame_t *frame, xlator_t *this,
 
         children = priv->children;
 
+        AFR_SBRAIN_CHECK_FD (fd, out);
+
         AFR_LOCAL_ALLOC_OR_GOTO (local, out);
         frame->local = local;
 
@@ -1185,7 +1747,17 @@ afr_fgetxattr (call_frame_t *frame, xlator_t *this,
         if (name)
                 local->cont.getxattr.name = gf_strdup (name);
 
-        /* pathinfo gets handled only in getxattr() */
+        /* 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) {
@@ -1193,7 +1765,8 @@ afr_fgetxattr (call_frame_t *frame, xlator_t *this,
                 goto out;
         }
 
-        read_child = afr_inode_get_read_ctx (this, fd->inode, local->fresh_children);
+        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,
@@ -1213,7 +1786,8 @@ afr_fgetxattr (call_frame_t *frame, xlator_t *this,
         op_ret = 0;
 out:
         if (op_ret == -1) {
-                AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, NULL, NULL);
+                AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, NULL,
+                                  NULL);
         }
         return 0;
 }
@@ -1315,6 +1889,8 @@ afr_readv (call_frame_t *frame, xlator_t *this,
         priv     = this->private;
         children = priv->children;
 
+        AFR_SBRAIN_CHECK_FD (fd, out);
+
         AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
         local = frame->local;
 
@@ -1344,11 +1920,8 @@ afr_readv (call_frame_t *frame, xlator_t *this,
         local->cont.readv.offset     = offset;
         local->cont.readv.flags      = flags;
 
-        ret = afr_open_fd_fix (frame, this, _gf_false);
-        if (ret) {
-                op_errno = -ret;
-                goto out;
-        }
+        afr_open_fd_fix (fd, this);
+
         STACK_WIND_COOKIE (frame, afr_readv_cbk,
                            (void *) (long) call_child,
                            children[call_child],
diff --git a/xlators/cluster/afr/src/afr-inode-write.c b/xlators/cluster/afr/src/afr-inode-write.c
index 148a939ca..c1ec69a55 100644
--- a/xlators/cluster/afr/src/afr-inode-write.c
+++ b/xlators/cluster/afr/src/afr-inode-write.c
@@ -39,34 +39,133 @@
 #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
+void
+afr_writev_copy_outvars (call_frame_t *src_frame, call_frame_t *dst_frame)
+{
+        afr_local_t *src_local = NULL;
+        afr_local_t *dst_local = NULL;
+
+        src_local = src_frame->local;
+        dst_local = dst_frame->local;
+
+        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;
+}
+
+void
 afr_writev_unwind (call_frame_t *frame, xlator_t *this)
 {
         afr_local_t *   local = NULL;
-        call_frame_t   *main_frame = NULL;
+        local = frame->local;
+
+        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);
         {
-                if (local->transaction.main_frame)
-                        main_frame = local->transaction.main_frame;
+                fop_frame = local->transaction.main_frame;
                 local->transaction.main_frame = NULL;
         }
         UNLOCK (&frame->lock);
 
-        if (main_frame) {
-                AFR_STACK_UNWIND (writev, main_frame,
-                                  local->op_ret, local->op_errno,
-                                  &local->cont.writev.prebuf,
-                                  &local->cont.writev.postbuf,
-                                  NULL);
+        return fop_frame;
+}
+
+int
+afr_transaction_writev_unwind (call_frame_t *frame, xlator_t *this)
+{
+        call_frame_t *fop_frame = NULL;
+
+        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,
@@ -74,11 +173,17 @@ afr_writev_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                      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;
+        priv  = this->private;
 
         read_child = afr_inode_get_read_ctx (this, local->fd->inode, NULL);
 
@@ -88,32 +193,81 @@ afr_writev_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         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;
-                        }
-
-                        if (child_index == read_child) {
-                                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);
+
+		local->replies[child_index].valid = 1;
+		local->replies[child_index].op_ret = op_ret;
+		local->replies[child_index].op_errno = op_errno;
+
+
+		/* 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 (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->op_errno = op_errno;
+		}
         }
         UNLOCK (&frame->lock);
 
         call_count = afr_frame_return (frame);
 
         if (call_count == 0) {
-                local->transaction.unwind (frame, this);
 
-                local->transaction.resume (frame, this);
+                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;
 }
@@ -125,6 +279,8 @@ afr_writev_wind (call_frame_t *frame, xlator_t *this)
         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;
@@ -138,6 +294,28 @@ afr_writev_wind (call_frame_t *frame, xlator_t *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;
+	}
+
+        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]) {
@@ -151,13 +329,16 @@ afr_writev_wind (call_frame_t *frame, xlator_t *this)
                                            local->cont.writev.offset,
                                            local->cont.writev.flags,
                                            local->cont.writev.iobref,
-                                           NULL);
+                                           xdata);
 
                         if (!--call_count)
                                 break;
                 }
         }
 
+        if (xdata)
+                dict_unref (xdata);
+
         return 0;
 }
 
@@ -197,7 +378,7 @@ afr_do_writev (call_frame_t *frame, xlator_t *this)
         }
 
         transaction_frame->local = local;
-        frame->local = NULL;
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
 
         local->op = GF_FOP_WRITE;
 
@@ -205,10 +386,17 @@ afr_do_writev (call_frame_t *frame, xlator_t *this)
 
         local->transaction.fop    = afr_writev_wind;
         local->transaction.done   = afr_writev_done;
-        local->transaction.unwind = afr_writev_unwind;
+        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 {
@@ -217,11 +405,15 @@ afr_do_writev (call_frame_t *frame, xlator_t *this)
                                                          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 (op_ret < 0) {
                 if (transaction_frame)
                         AFR_STACK_DESTROY (transaction_frame);
                 AFR_STACK_UNWIND (writev, frame, op_ret, op_errno, NULL, NULL, NULL);
@@ -230,153 +422,74 @@ out:
         return 0;
 }
 
-static int
-afr_prepare_loc (call_frame_t *frame, fd_t *fd)
+static void
+afr_trigger_open_fd_self_heal (fd_t *fd, xlator_t *this)
 {
-        afr_local_t    *local = NULL;
-        char           *name = NULL;
-        char           *path = NULL;
-        int             ret = 0;
-
-        if ((!fd) || (!fd->inode))
-                return -1;
-
-        local = frame->local;
-        ret = inode_path (fd->inode, NULL, (char **)&path);
-        if (ret <= 0) {
-                gf_log (frame->this->name, GF_LOG_DEBUG,
-                        "Unable to get path for gfid: %s",
-                        uuid_utoa (fd->inode->gfid));
-                return -1;
-        }
-
-        if (local->loc.path) {
-                if (strcmp (path, local->loc.path))
-                        gf_log (frame->this->name, GF_LOG_DEBUG,
-                                "overwriting old loc->path %s with %s",
-                                local->loc.path, path);
-                GF_FREE ((char *)local->loc.path);
-        }
-        local->loc.path = path;
-
-        name = strrchr (local->loc.path, '/');
-        if (name)
-                name++;
-        local->loc.name = name;
-
-        if (local->loc.inode) {
-                inode_unref (local->loc.inode);
-        }
-        local->loc.inode = inode_ref (fd->inode);
-
-        if (local->loc.parent) {
-                inode_unref (local->loc.parent);
+        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;
         }
 
-        local->loc.parent = inode_parent (local->loc.inode, 0, NULL);
-
-        return 0;
-}
-
-afr_fd_paused_call_t*
-afr_paused_call_create (call_frame_t *frame)
-{
-        afr_local_t             *local = NULL;
-        afr_fd_paused_call_t   *paused_call = NULL;
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame)
+                goto out;
 
+        AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
         local = frame->local;
-        GF_ASSERT (local->fop_call_continue);
-
-        paused_call = GF_CALLOC (1, sizeof (*paused_call),
-                                  gf_afr_fd_paused_call_t);
-        if (paused_call) {
-                INIT_LIST_HEAD (&paused_call->call_list);
-                paused_call->frame = frame;
-        }
-
-        return paused_call;
-}
-
-static int
-afr_pause_fd_fop (call_frame_t *frame, xlator_t *this, afr_fd_ctx_t *fd_ctx)
-{
-        afr_fd_paused_call_t *paused_call = NULL;
-        int                    ret = 0;
-
-        paused_call = afr_paused_call_create (frame);
-        if (paused_call)
-                list_add (&paused_call->call_list, &fd_ctx->paused_calls);
-        else
-                ret = -ENOMEM;
-
-        return ret;
-}
+        ret = afr_local_init (local, this->private, &op_errno);
+        if (ret < 0)
+                goto out;
 
-static void
-afr_trigger_open_fd_self_heal (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t             *local = NULL;
-        afr_self_heal_t         *sh = NULL;
-        inode_t                 *inode = NULL;
-        char                    *reason = NULL;
+        local->loc.inode = inode_ref (fd->inode);
+        ret = loc_path (&local->loc, NULL);
+        if (ret < 0)
+                goto out;
 
-        local = frame->local;
         sh    = &local->self_heal;
-        inode = local->fd->inode;
-
-        sh->do_missing_entry_self_heal = _gf_true;
-        sh->do_gfid_self_heal = _gf_true;
-        sh->do_data_self_heal = _gf_true;
+        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, inode, _gf_true, inode->ia_type,
-                              reason, NULL, NULL);
+        afr_launch_self_heal (frame, this, fd->inode, _gf_true,
+                              fd->inode->ia_type, reason, NULL, NULL);
+        return;
+out:
+        AFR_STACK_DESTROY (frame);
 }
 
-int
-afr_open_fd_fix (call_frame_t *frame, xlator_t *this, gf_boolean_t pause_fop)
-{
-        int                     ret = 0;
-        int                     i   = 0;
-        afr_fd_ctx_t            *fd_ctx = NULL;
-        inode_t                 *inode = NULL;
-        gf_boolean_t            need_self_heal = _gf_false;
-        int                     *need_open = NULL;
-        int                     need_open_count = 0;
-        afr_local_t             *local = NULL;
-        afr_private_t           *priv = NULL;
-        gf_boolean_t            fop_continue = _gf_true;
+void
+afr_open_fd_fix (fd_t *fd, xlator_t *this)
+{
+        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;
 
-        local = frame->local;
         priv  = this->private;
 
-        GF_ASSERT (local->fd);
-
-        inode = local->fd->inode;
-        //gfid is not set in rebalance, that case needs to be handled.
-        if (fd_is_anonymous (local->fd) ||
-            !inode || uuid_is_null (inode->gfid)) {
-                fop_continue = _gf_true;
-                goto out;
-        }
-
-        if (pause_fop)
-                GF_ASSERT (local->fop_call_continue);
-
-        ret = afr_prepare_loc (frame, local->fd);
-        if (ret < 0) {
-                //File does not exist we cant open it.
-                ret = 0;
+        if (!afr_is_fd_fixable (fd))
                 goto out;
-        }
 
-        fd_ctx = afr_fd_ctx_get (local->fd, this);
-        if (!fd_ctx) {
-                ret = -EINVAL;
+        fd_ctx = afr_fd_ctx_get (fd, this);
+        if (!fd_ctx)
                 goto out;
-        }
 
-        LOCK (&local->fd->lock);
+        LOCK (&fd->lock);
         {
                 if (fd_ctx->up_count < priv->up_count) {
                         need_self_heal = _gf_true;
@@ -384,55 +497,34 @@ afr_open_fd_fix (call_frame_t *frame, xlator_t *this, gf_boolean_t pause_fop)
                         fd_ctx->down_count = priv->down_count;
                 }
 
+                need_open = alloca (priv->child_count * sizeof (*need_open));
                 for (i = 0; i < priv->child_count; i++) {
-                        if ((fd_ctx->opened_on[i] == AFR_FD_NOT_OPENED) &&
-                            local->child_up[i]) {
-                                fd_ctx->opened_on[i] = AFR_FD_OPENING;
-                                if (!need_open)
-                                        need_open = GF_CALLOC (priv->child_count,
-                                                               sizeof (*need_open),
-                                                               gf_afr_mt_int32_t);
-                                need_open[i] = 1;
-                                need_open_count++;
-                        } else if (pause_fop && local->child_up[i] &&
-                                   (fd_ctx->opened_on[i] == AFR_FD_OPENING)) {
-                                local->fop_paused = _gf_true;
-                        }
-                }
+                        need_open[i] = 0;
+                        if (fd_ctx->opened_on[i] != AFR_FD_NOT_OPENED)
+                                continue;
+
+                        if (!priv->child_up[i])
+                                continue;
 
-                if (local->fop_paused) {
-                        GF_ASSERT (pause_fop);
-                        gf_log (this->name, GF_LOG_INFO, "Pause fd %p",
-                                local->fd);
-                        ret = afr_pause_fd_fop (frame, this, fd_ctx);
-                        if (ret)
-                                goto unlock;
-                        fop_continue = _gf_false;
+                        fd_ctx->opened_on[i] = AFR_FD_OPENING;
+
+                        need_open[i] = 1;
+                        need_open_count++;
                 }
         }
-unlock:
-        UNLOCK (&local->fd->lock);
-        if (ret) {
-                gf_log (this->name, GF_LOG_ERROR, "Failed to fix fd for %s",
-                        local->loc.path);
-                fop_continue = _gf_false;
+        UNLOCK (&fd->lock);
+        if (ret)
                 goto out;
-        }
 
         if (need_self_heal)
-                afr_trigger_open_fd_self_heal (frame, this);
+                afr_trigger_open_fd_self_heal (fd, this);
 
         if (!need_open_count)
                 goto out;
 
-        gf_log (this->name, GF_LOG_INFO, "Opening fd %p", local->fd);
-        afr_fix_open (frame, this, fd_ctx, need_open_count, need_open);
-        fop_continue = _gf_false;
+        afr_fix_open (this, fd, need_open_count, need_open);
 out:
-        GF_FREE (need_open);
-        if (fop_continue && local->fop_call_continue)
-                local->fop_call_continue (frame, this);
-        return ret;
+        return;
 }
 
 int
@@ -451,6 +543,11 @@ afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
 
         priv = this->private;
 
+        if (afr_is_split_brain (this, fd->inode)) {
+                op_errno = EIO;
+                goto out;
+        }
+
         QUORUM_CHECK(writev,out);
 
         AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
@@ -467,13 +564,15 @@ afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
         local->cont.writev.iobref     = iobref_ref (iobref);
 
         local->fd                = fd_ref (fd);
-        local->fop_call_continue = afr_do_writev;
 
-        ret = afr_open_fd_fix (frame, this, _gf_true);
-        if (ret) {
-                op_errno = -ret;
-                goto out;
-        }
+	/* 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);
+
+        afr_do_writev (frame, this);
 
         ret = 0;
 out:
@@ -507,8 +606,8 @@ afr_truncate_unwind (call_frame_t *frame, xlator_t *this)
         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);
         }
 
@@ -522,14 +621,11 @@ afr_truncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                        struct iatt *postbuf, dict_t *xdata)
 {
         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;
 
         local = frame->local;
-        priv  = this->private;
 
         read_child = afr_inode_get_read_ctx (this, local->loc.inode, NULL);
 
@@ -539,38 +635,22 @@ afr_truncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         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;
-                        }
+			if (prebuf->ia_size != postbuf->ia_size)
+				local->stable_write = _gf_false;
                 }
-                local->op_errno = op_errno;
+                __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);
 
         if (call_count == 0) {
+		if (local->stable_write && afr_txn_nothing_failed (frame, this))
+			local->transaction.unwind (frame, this);
+
                 local->transaction.resume (frame, this);
         }
 
@@ -598,6 +678,7 @@ afr_truncate_wind (call_frame_t *frame, xlator_t *this)
         }
 
         local->call_count = call_count;
+	local->stable_write = _gf_true;
 
         for (i = 0; i < priv->child_count; i++) {
                 if (local->transaction.pre_op[i]) {
@@ -676,7 +757,11 @@ afr_truncate (call_frame_t *frame, xlator_t *this,
         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;
+        }
 
         ret = 0;
 out:
@@ -714,8 +799,8 @@ afr_ftruncate_unwind (call_frame_t *frame, xlator_t *this)
         if (main_frame) {
                 AFR_STACK_UNWIND (ftruncate, main_frame, local->op_ret,
                                   local->op_errno,
-                                  &local->cont.ftruncate.prebuf,
-                                  &local->cont.ftruncate.postbuf,
+                                  &local->cont.inode_wfop.prebuf,
+                                  &local->cont.inode_wfop.postbuf,
                                   NULL);
         }
         return 0;
@@ -728,14 +813,11 @@ afr_ftruncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         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);
 
@@ -745,38 +827,22 @@ afr_ftruncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         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 ((local->success_count >= priv->wait_count)
-                            && local->read_child_returned) {
-                                need_unwind = 1;
-                        }
+			if (prebuf->ia_size != postbuf->ia_size)
+				local->stable_write = _gf_false;
                 }
-                local->op_errno = op_errno;
+                __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);
 
         if (call_count == 0) {
+		if (local->stable_write && afr_txn_nothing_failed (frame, this))
+			local->transaction.unwind (frame, this);
+
                 local->transaction.resume (frame, this);
         }
 
@@ -804,6 +870,7 @@ afr_ftruncate_wind (call_frame_t *frame, xlator_t *this)
         }
 
         local->call_count = call_count;
+	local->stable_write = _gf_true;
 
         for (i = 0; i < priv->child_count; i++) {
                 if (local->transaction.pre_op[i]) {
@@ -868,11 +935,15 @@ afr_do_ftruncate (call_frame_t *frame, xlator_t *this)
         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 (op_ret < 0) {
                 if (transaction_frame)
                         AFR_STACK_DESTROY (transaction_frame);
                 AFR_STACK_UNWIND (ftruncate, frame, op_ret, op_errno, NULL,
@@ -899,6 +970,10 @@ afr_ftruncate (call_frame_t *frame, xlator_t *this,
 
         priv = this->private;
 
+        if (afr_is_split_brain (this, fd->inode)) {
+                op_errno = EIO;
+                goto out;
+        }
         QUORUM_CHECK(ftruncate,out);
 
         AFR_LOCAL_ALLOC_OR_GOTO (frame->local, out);
@@ -911,13 +986,10 @@ afr_ftruncate (call_frame_t *frame, xlator_t *this,
         local->cont.ftruncate.offset  = offset;
 
         local->fd = fd_ref (fd);
-        local->fop_call_continue = afr_do_ftruncate;
 
-        ret = afr_open_fd_fix (frame, this, _gf_true);
-        if (ret) {
-                op_errno = -ret;
-                goto out;
-        }
+        afr_open_fd_fix (fd, this);
+
+        afr_do_ftruncate (frame, this);
 
         ret = 0;
 out:
@@ -953,8 +1025,8 @@ afr_setattr_unwind (call_frame_t *frame, xlator_t *this)
         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);
         }
 
@@ -985,29 +1057,14 @@ afr_setattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         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;
-                        }
-
-                        local->success_count++;
+                __inode_write_fop_cbk (frame, child_index, read_child, this,
+                                       &op_ret, &op_errno, preop, postop,
+                                       xdata);
 
-                        if ((local->success_count >= priv->wait_count)
-                            && local->read_child_returned) {
-                                need_unwind = 1;
-                        }
+                if ((local->success_count >= priv->wait_count)
+                    && local->read_child_returned) {
+                        need_unwind = 1;
                 }
-                local->op_errno = op_errno;
         }
         UNLOCK (&frame->lock);
 
@@ -1124,7 +1181,11 @@ afr_setattr (call_frame_t *frame, xlator_t *this,
         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;
+        }
 
         ret = 0;
 out:
@@ -1158,8 +1219,8 @@ afr_fsetattr_unwind (call_frame_t *frame, xlator_t *this)
         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);
         }
 
@@ -1190,29 +1251,14 @@ afr_fsetattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         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.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;
-                        }
+                __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;
-                        }
+                if ((local->success_count >= priv->wait_count)
+                    && local->read_child_returned) {
+                        need_unwind = 1;
                 }
-                local->op_errno = op_errno;
         }
         UNLOCK (&frame->lock);
 
@@ -1300,6 +1346,11 @@ afr_fsetattr (call_frame_t *frame, xlator_t *this,
 
         priv = this->private;
 
+        if (afr_is_split_brain (this, fd->inode)) {
+                op_errno = EIO;
+                goto out;
+        }
+
         QUORUM_CHECK(fsetattr,out);
 
         transaction_frame = copy_frame (frame);
@@ -1324,17 +1375,17 @@ afr_fsetattr (call_frame_t *frame, xlator_t *this,
 
         local->fd                 = fd_ref (fd);
 
-        ret = afr_open_fd_fix (transaction_frame, this, _gf_false);
-        if (ret) {
-                op_errno = -ret;
-                goto out;
-        }
+        afr_open_fd_fix (fd, this);
 
         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;
+        }
 
         ret = 0;
 out:
@@ -1380,28 +1431,23 @@ int
 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;
-        int               call_count    = -1;
-        int               need_unwind   = 0;
+        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);
         {
-                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;
-                        }
+                __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;
                 }
-
-                local->op_errno = op_errno;
         }
         UNLOCK (&frame->lock);
 
@@ -1521,7 +1567,11 @@ afr_setxattr (call_frame_t *frame, xlator_t *this,
         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;
+        }
 
         ret = 0;
 out:
@@ -1566,28 +1616,24 @@ int
 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_local_t     *local       = NULL;
-        afr_private_t   *priv        = NULL;
-        int              call_count  = -1;
-        int              need_unwind = 0;
+        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);
         {
-                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;
-                        }
+                __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;
                 }
-
-                local->op_errno = op_errno;
         }
         UNLOCK (&frame->lock);
 
@@ -1677,11 +1723,13 @@ afr_fsetxattr (call_frame_t *frame, xlator_t *this,
         GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.afr.*", dict,
                                    op_errno, out);
 
-        if (ret)
-                goto 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);
@@ -1712,7 +1760,11 @@ afr_fsetxattr (call_frame_t *frame, xlator_t *this,
         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;
+        }
 
         ret = 0;
 out:
@@ -1760,28 +1812,23 @@ int
 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;
-        int call_count  = -1;
-        int need_unwind = 0;
+        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);
         {
-                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;
-                        }
+                __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;
                 }
-
-                local->op_errno = op_errno;
         }
         UNLOCK (&frame->lock);
 
@@ -1902,7 +1949,11 @@ afr_removexattr (call_frame_t *frame, xlator_t *this,
         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;
+        }
 
         ret = 0;
 out:
@@ -1945,28 +1996,24 @@ 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;
+        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);
         {
-                if (op_ret != -1) {
-                        if (local->success_count == 0) {
-                                local->op_ret = op_ret;
-                        }
-                        local->success_count++;
+                __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;
-                        }
+                if (local->success_count == priv->wait_count) {
+                        need_unwind = 1;
                 }
-
-                local->op_errno = op_errno;
         }
         UNLOCK (&frame->lock);
 
@@ -2059,6 +2106,10 @@ afr_fremovexattr (call_frame_t *frame, xlator_t *this,
         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);
 
@@ -2091,11 +2142,15 @@ afr_fremovexattr (call_frame_t *frame, xlator_t *this,
         local->transaction.start   = LLONG_MAX - 1;
         local->transaction.len     = 0;
 
-        afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        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 == -1) {
+        if (op_ret < 0) {
                 if (transaction_frame)
                         AFR_STACK_DESTROY (transaction_frame);
                 AFR_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, NULL);
@@ -2103,3 +2158,704 @@ out:
 
         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;
+
+        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 ed11079fd..8e93ca44a 100644
--- a/xlators/cluster/afr/src/afr-inode-write.h
+++ b/xlators/cluster/afr/src/afr-inode-write.h
@@ -68,4 +68,15 @@ 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 5e61be4d4..060d78f35 100644
--- a/xlators/cluster/afr/src/afr-lk-common.c
+++ b/xlators/cluster/afr/src/afr-lk-common.c
@@ -55,7 +55,33 @@
         } 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);
@@ -125,16 +151,9 @@ internal_lock_count (call_frame_t *frame, xlator_t *this)
         local = frame->local;
         priv  = this->private;
 
-        if (local->fd) {
-                for (i = 0; i < priv->child_count; i++) {
-                        if (local->child_up[i] && local->fd_open_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;
@@ -332,10 +351,13 @@ static void
 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)
 {
         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];
@@ -343,28 +365,40 @@ afr_trace_entrylk_in (call_frame_t *frame, xlator_t *this,
 
         local    = frame->local;
         int_lock = &local->internal_lock;
+        priv     = this->private;
+
+        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_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_INFO,
-                "[%s %s] Lock={%s} Lockee={%s} Number={%llu}",
+                "[%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, 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 child_index)
+                       int op_ret, int op_errno, int32_t cookie)
 {
         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];
@@ -373,20 +407,30 @@ afr_trace_entrylk_out (call_frame_t *frame, xlator_t *this,
 
         local    = frame->local;
         int_lock = &local->internal_lock;
+        priv     = this->private;
 
-        afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
+        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, &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_INFO,
-                "[%s %s] [%s] Lock={%s} Lockee={%s} Number={%llu}",
+                "[%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);
 
 }
 
@@ -439,6 +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)
 {
@@ -456,8 +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;
@@ -469,19 +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;
 }
@@ -503,6 +597,18 @@ lower_path (loc_t *l1, const char *b1, loc_t *l2, const char *b2)
 }
 
 int
+afr_lockee_locked_nodes_count (afr_internal_lock_t *int_lock)
+{
+        int call_count  = 0;
+        int i           = 0;
+
+        for (i = 0; i < int_lock->lockee_count; i++)
+                call_count += int_lock->lockee[i].locked_count;
+
+        return call_count;
+}
+
+int
 afr_locked_nodes_count (unsigned char *locked_nodes, int child_count)
 
 {
@@ -550,7 +656,9 @@ afr_unlock_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 {
         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;
@@ -559,14 +667,18 @@ afr_unlock_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                AFR_UNLOCK_OP, NULL, op_ret,
                                op_errno, child_index);
 
+        priv = this->private;
+
         if (op_ret < 0 && op_errno != ENOTCONN && op_errno != EBADFD) {
-                gf_log (this->name, GF_LOG_INFO, "%s: unlock failed on %d "
-                        "unlock by %s", local->loc.path, child_index,
+                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));
         }
 
 
-        int_lock->inode_locked_nodes[child_index] &= LOCKED_NO;
+        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;
 
@@ -580,6 +692,7 @@ 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 gf_flock flock = {0,};
@@ -595,12 +708,14 @@ afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
         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;
 
         full_flock.l_type = F_UNLCK;
-        call_count = afr_locked_nodes_count (int_lock->inode_locked_nodes,
+        call_count = afr_locked_nodes_count (inodelk->locked_nodes,
                                              priv->child_count);
 
         int_lock->lk_call_count = call_count;
@@ -616,8 +731,7 @@ afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
                 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)
-                    != LOCKED_YES)
+                if ((inodelk->locked_nodes[i] & LOCKED_YES) != LOCKED_YES)
                         continue;
 
                 if (local->fd) {
@@ -658,7 +772,7 @@ afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
                                            (void *) (long)i,
                                            priv->children[i],
                                            priv->children[i]->fops->finodelk,
-                                           this->name, local->fd,
+                                           int_lock->domain, local->fd,
                                            F_SETLK, flock_use, NULL);
 
                         if (!--call_count)
@@ -673,7 +787,7 @@ afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
                                            (void *) (long)i,
                                            priv->children[i],
                                            priv->children[i]->fops->inodelk,
-                                           this->name, &local->loc,
+                                           int_lock->domain, &local->loc,
                                            F_SETLK, &flock, NULL);
 
                         if (!--call_count)
@@ -689,13 +803,22 @@ afr_unlock_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;
-        int32_t             child_index = (long)cookie;
+        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_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
-                               AFR_UNLOCK_OP, NULL, op_ret,
-                               op_errno, child_index);
+                               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,
@@ -703,6 +826,7 @@ afr_unlock_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         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;
@@ -711,24 +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){
@@ -738,18 +860,22 @@ 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, this,
-                                              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,
+                                           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)
@@ -766,13 +892,20 @@ static int32_t
 afr_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
               int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = 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) {
@@ -788,6 +921,8 @@ afr_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->op_errno              = op_errno;
                         int_lock->lock_op_errno      = op_errno;
                 }
+
+		int_lock->lk_attempted_count++;
         }
         UNLOCK (&frame->lock);
 
@@ -796,10 +931,17 @@ afr_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 afr_unlock (frame, this);
         } else {
                 if (op_ret == 0) {
-                        int_lock->locked_nodes[child_index] |= LOCKED_YES;
-                        int_lock->lock_count++;
+                        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, child_index + 1);
+                afr_lock_blocking (frame, this, cky + 1);
         }
 
         return 0;
@@ -819,79 +961,6 @@ afr_blocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 }
 
 static int32_t
-afr_lock_lower_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
-{
-        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          *higher_name = NULL;
-        int child_index = (long) cookie;
-
-        priv     = this->private;
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-
-        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/locks xlator on server");
-
-                                local->op_ret   = op_ret;
-                        }
-
-                        local->op_errno = op_errno;
-                }
-        }
-        UNLOCK (&frame->lock);
-
-        if (op_ret != 0) {
-                afr_copy_locked_nodes (frame, this);
-                afr_unlock (frame, this);
-                goto out;
-        } else {
-                int_lock->lower_locked_nodes[child_index] |= LOCKED_LOWER;
-                int_lock->lock_count++;
-        }
-
-        /* 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);
-
-        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, this, AFR_ENTRYLK_TRANSACTION,
-                              AFR_LOCK_OP, higher_name, child_index);
-
-
-        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, NULL);
-
-out:
-        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, dict_t *xdata)
 {
@@ -907,6 +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;
 
@@ -917,18 +987,16 @@ afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this)
         switch (local->transaction.type) {
         case AFR_DATA_TRANSACTION:
         case AFR_METADATA_TRANSACTION:
-                memcpy (int_lock->inode_locked_nodes,
-                        int_lock->locked_nodes,
-                        priv->child_count);
-                int_lock->inodelk_lock_count = int_lock->lock_count;
+                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;
         }
 
@@ -936,25 +1004,67 @@ 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;
-        loc_t               *lower       = NULL;
-        const char          *lower_name  = 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);
@@ -973,42 +1083,26 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
 
                         return 0;
                 }
-
-                /* skip over children that or down
-                   or don't have the fd open */
-
-                while ((child_index < priv->child_count)
-                       && (!local->child_up[child_index] ||
-                          !local->fd_open_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_INFO,
-                        "unable to lock on even one child");
-
-                local->op_ret           = -1;
-                int_lock->lock_op_ret   = -1;
+        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");
 
-                afr_copy_locked_nodes (frame, this);
+                        local->op_ret           = -1;
+                        int_lock->lock_op_ret   = -1;
 
-                afr_unlock(frame, this);
+                        afr_copy_locked_nodes (frame, this);
 
-                return 0;
+                        afr_unlock(frame, this);
 
+                        return 0;
+                }
         }
 
-        if ((child_index == priv->child_count)
-            || (int_lock->lock_count == int_lock->lk_expected_count)) {
-
+        if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
                 /* we're done locking */
 
                 gf_log (this->name, GF_LOG_DEBUG,
@@ -1021,6 +1115,11 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
                 return 0;
         }
 
+        if (!_is_lock_wind_needed (local, child_index)) {
+                afr_lock_blocking (frame, this, cookie + 1);
+                return 0;
+        }
+
         switch (local->transaction.type) {
         case AFR_DATA_TRANSACTION:
         case AFR_METADATA_TRANSACTION:
@@ -1035,7 +1134,7 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
                                            (void *) (long) child_index,
                                            priv->children[child_index],
                                            priv->children[child_index]->fops->finodelk,
-                                           this->name, local->fd,
+                                           int_lock->domain, local->fd,
                                            F_SETLKW, &flock, NULL);
 
                 } else {
@@ -1048,50 +1147,29 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
                                            (void *) (long) child_index,
                                            priv->children[child_index],
                                            priv->children[child_index]->fops->inodelk,
-                                           this->name, &local->loc,
+                                           int_lock->domain, &local->loc,
                                            F_SETLKW, &flock, NULL);
                 }
 
                 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);
-
-                AFR_TRACE_ENTRYLK_IN (frame, this, 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, NULL);
-
-                break;
-        }
-
         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, local->transaction.basename,
-                                              child_index);
+                        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) child_index,
+                                           (void *) (long) cookie,
                                            priv->children[child_index],
                                            priv->children[child_index]->fops->fentrylk,
-                                           this->name, local->fd,
-                                           local->transaction.basename,
+                                           int_lock->domain, local->fd,
+                                           int_lock->lockee[lockee_no].basename,
                                            ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
                 } else {
                         AFR_TRACE_ENTRYLK_IN (frame, this,
@@ -1100,12 +1178,12 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
                                               child_index);
 
                         STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
-                                           (void *) (long) child_index,
+                                           (void *) (long) cookie,
                                            priv->children[child_index],
                                            priv->children[child_index]->fops->entrylk,
-                                           this->name,
-                                           &local->transaction.parent_loc,
-                                           local->transaction.basename,
+                                           int_lock->domain,
+                                           &int_lock->lockee[lockee_no].loc,
+                                           int_lock->lockee[lockee_no].basename,
                                            ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
                 }
 
@@ -1134,11 +1212,12 @@ afr_blocking_lock (call_frame_t *frame, xlator_t *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_expected_count = 2 * up_count;
-                //fallthrough
-        case AFR_ENTRY_TRANSACTION:
+                int_lock->lk_call_count = int_lock->lk_expected_count
+                                        = (int_lock->lockee_count *
+                                           up_count);
                 initialize_entrylk_variables (frame, this);
                 break;
         }
@@ -1154,35 +1233,48 @@ afr_nonblocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 {
         afr_internal_lock_t *int_lock = NULL;
         afr_local_t         *local    = NULL;
-        int                 call_count  = 0;
-        int                 child_index = (long) cookie;
+        int call_count          = 0;
+        int child_index         = (long) cookie;
+        int copies              = 0;
+        int index               = 0;
+        int lockee_no           = 0;
+        afr_private_t       *priv = NULL;
+
+        priv = this->private;
+
+        copies = priv->child_count;
+        index = child_index % copies;
+        lockee_no = child_index / copies;
 
         local    = frame->local;
         int_lock = &local->internal_lock;
 
         AFR_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
-                               AFR_LOCK_OP, NULL, op_ret,
+                               AFR_LOCK_OP,
+                               int_lock->lockee[lockee_no].basename, op_ret,
                                op_errno, (long) cookie);
 
-        if (op_ret < 0 ) {
-                if (op_errno == ENOSYS) {
+	LOCK (&frame->lock);
+	{
+		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/locks xlator on server");
-                        local->op_ret         = op_ret;
-                        int_lock->lock_op_ret = op_ret;
-
-                        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++;
-        }
+				gf_log (this->name, GF_LOG_ERROR,
+					"subvolume does not support locking. "
+					"please load features/locks xlator on server");
+				local->op_ret         = op_ret;
+				int_lock->lock_op_ret = op_ret;
+
+				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++;
+		}
 
-        LOCK (&frame->lock);
-        {
                 call_count = --int_lock->lk_call_count;
         }
         UNLOCK (&frame->lock);
@@ -1212,42 +1304,26 @@ afr_nonblocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         return 0;
 }
 
-void
-afr_mark_fd_open_on (afr_local_t *local, afr_fd_ctx_t *fd_ctx,
-                       size_t child_count)
-{
-        int             i = 0;
-
-        GF_ASSERT (local->fd_open_on);
-
-        memset (local->fd_open_on, 0, sizeof (*local->fd_open_on)*child_count);
-        for (i = 0; i < child_count; i++)
-                if (fd_ctx->opened_on[i] == AFR_FD_OPENED)
-                        local->fd_open_on[i] = 1;
-}
-
 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;
 
         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) {
                 fd_ctx = afr_fd_ctx_get (local->fd, this);
                 if (!fd_ctx) {
@@ -1260,11 +1336,11 @@ 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;
                 }
 
-                afr_mark_fd_open_on (local, fd_ctx, priv->child_count);
-                call_count = internal_lock_count (frame, this);
+                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;
 
@@ -1277,46 +1353,52 @@ afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this)
 
                 /* 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] && local->fd_open_on[i]) {
-                                AFR_TRACE_ENTRYLK_IN (frame, this,
-                                                      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,
+                                                   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);
+                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, this,
-                                                      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,
+                                                   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;
-
                         }
                 }
         }
@@ -1329,6 +1411,7 @@ afr_nonblocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                              int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
         afr_internal_lock_t *int_lock = NULL;
+        afr_inodelk_t       *inodelk  = NULL;
         afr_local_t         *local    = NULL;
         int call_count  = 0;
         int child_index = (long) cookie;
@@ -1337,58 +1420,57 @@ afr_nonblocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         local    = frame->local;
         int_lock = &local->internal_lock;
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
 
         AFR_TRACE_INODELK_OUT (frame, this, AFR_INODELK_NB_TRANSACTION,
                                AFR_LOCK_OP, NULL, op_ret,
                                op_errno, (long) cookie);
 
-        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/locks xlator on server");
-                        local->op_ret                = op_ret;
-                        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 {
-                int_lock->inode_locked_nodes[child_index]
-                        |= LOCKED_YES;
-                int_lock->inodelk_lock_count++;
-
-                if (local->transaction.eager_lock &&
-                    local->transaction.eager_lock[child_index] && local->fd) {
-                        fd_ctx = afr_fd_ctx_get (local->fd, this);
-                        /* piggybacked */
-
-                        if (op_ret == 1) {
-                                /* piggybacked */
-                        } else if (op_ret == 0) {
-                                /* lock acquired from server */
-                                LOCK (&local->fd->lock);
-                                {
-                                        fd_ctx->lock_acquired[child_index]++;
-                                }
-                                UNLOCK (&local->fd->lock);
-                        }
-                }
-        }
+	if (local->fd)
+		fd_ctx = afr_fd_ctx_get (local->fd, this);
 
         LOCK (&frame->lock);
         {
+		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/locks xlator on "
+					"server");
+				local->op_ret                = op_ret;
+				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 successful. Proceed to call FOP */
-                if (int_lock->inodelk_lock_count ==
-                                int_lock->lk_expected_count) {
+                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;
@@ -1412,6 +1494,7 @@ 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;
@@ -1427,11 +1510,13 @@ afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
         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 = int_lock->lk_flock.l_type;
+        full_flock.l_type = inodelk->flock.l_type;
 
         initialize_inodelk_variables (frame, this);
 
@@ -1447,11 +1532,11 @@ 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;
                 }
 
-                afr_mark_fd_open_on (local, fd_ctx, priv->child_count);
                 call_count = internal_lock_count (frame, this);
                 int_lock->lk_call_count = call_count;
                 int_lock->lk_expected_count = call_count;
@@ -1466,11 +1551,11 @@ afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
                 /* 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] || !local->fd_open_on[i])
+                        if (!local->child_up[i])
                                 continue;
 
                         flock_use = &flock;
-                        if (!priv->eager_lock) {
+                        if (!local->transaction.eager_lock_on) {
                                 goto wind;
                         }
 
@@ -1506,7 +1591,7 @@ afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
                                            (void *) (long) i,
                                            priv->children[i],
                                            priv->children[i]->fops->finodelk,
-                                           this->name, local->fd,
+                                           int_lock->domain, local->fd,
                                            F_SETLK, flock_use, NULL);
 
                         if (!--call_count)
@@ -1528,7 +1613,7 @@ afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
                                            (void *) (long) i,
                                            priv->children[i],
                                            priv->children[i]->fops->inodelk,
-                                           this->name, &local->loc,
+                                           int_lock->domain, &local->loc,
                                            F_SETLK, &flock, NULL);
 
                         if (!--call_count)
@@ -1539,201 +1624,6 @@ out:
         return ret;
 }
 
-static int
-__is_lower_locked (call_frame_t *frame, xlator_t *this)
-{
-        afr_internal_lock_t *int_lock = NULL;
-        afr_private_t       *priv     = NULL;
-        afr_local_t         *local    = NULL;
-        int count = 0;
-        int i     = 0;
-
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-        priv     = this->private;
-
-        for (i = 0; i < priv->child_count; i++) {
-                if (int_lock->lower_locked_nodes[i] & LOCKED_LOWER)
-                        count++;
-        }
-
-        return count;
-
-}
-
-static int
-__is_higher_locked (call_frame_t *frame, xlator_t *this)
-{
-        afr_internal_lock_t *int_lock = NULL;
-        afr_private_t       *priv     = NULL;
-        afr_local_t         *local    = NULL;
-        int count = 0;
-        int i = 0;
-
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-        priv     = this->private;
-
-        for (i = 0; i < priv->child_count; i++) {
-                if (int_lock->locked_nodes[i] & LOCKED_YES)
-                        count++;
-        }
-
-        return count;
-
-}
-
-static int
-afr_unlock_lower_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;
-
-        local = frame->local;
-        int_lock = &local->internal_lock;
-        priv  = this->private;
-
-        basename = int_lock->lk_basename;
-        if (int_lock->lk_loc)
-                loc = int_lock->lk_loc;
-
-        call_count = __is_lower_locked (frame, this);
-        int_lock->lk_call_count = call_count;
-
-        if (!call_count){
-                gf_log (this->name, GF_LOG_TRACE,
-                        "No internal locks unlocked");
-                int_lock->lock_cbk (frame, this);
-                goto out;
-        }
-
-        for (i = 0; i < priv->child_count; i++) {
-                if (int_lock->lower_locked_nodes[i] & LOCKED_LOWER) {
-                        AFR_TRACE_ENTRYLK_IN (frame, this,
-                                              AFR_ENTRYLK_NB_TRANSACTION,
-                                              AFR_UNLOCK_OP, 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, NULL);
-
-                        if (!--call_count)
-                                break;
-
-                }
-        }
-
-out:
-        return 0;
-
-}
-
-
-static int
-afr_post_unlock_higher_cbk (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t *local = NULL;
-
-        local = frame->local;
-
-        local->transaction.done (frame, this);
-        return 0;
-}
-
-static int
-afr_post_unlock_lower_cbk (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          *higher_name = NULL;
-
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-
-        lower = lower_path (&local->transaction.parent_loc,
-                            local->transaction.basename,
-                            &local->transaction.new_parent_loc,
-                            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);
-
-        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;
-
-                afr_unlock_entrylk (frame, this);
-        } else
-                local->transaction.done (frame, this);
-
-        return 0;
-}
-
-static int
-afr_rename_unlock (call_frame_t *frame, xlator_t *this)
-{
-        afr_internal_lock_t *int_lock    = NULL;
-        afr_local_t         *local       = NULL;
-        loc_t               *lower       = NULL;
-        const char          *lower_name  = NULL;
-
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-
-        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);
-
-        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;
-
-                afr_unlock_lower_entrylk (frame, this);
-        } else
-                afr_post_unlock_lower_cbk (frame, this);
-
-        return 0;
-}
-
-static int
-afr_rename_transaction (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t *local = NULL;
-
-        local = frame->local;
-
-        return (local->transaction.type ==
-                AFR_ENTRY_RENAME_TRANSACTION);
-
-}
-
 int32_t
 afr_unlock (call_frame_t *frame, xlator_t *this)
 {
@@ -1745,10 +1635,8 @@ afr_unlock (call_frame_t *frame, xlator_t *this)
                 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);
+                        afr_unlock_entrylk (frame, this);
+
         } else {
                 if (is_afr_lock_selfheal (local))
                         afr_unlock_inodelk (frame, this);
@@ -2249,29 +2137,38 @@ out:
         return ret;
 }
 
-void
-afr_lk_transfer_datalock (call_frame_t *dst, call_frame_t *src,
+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_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;
 
-        dst_local = dst->local;
-        dst_lock  = &dst_local->internal_lock;
         src_local = src->local;
         src_lock  = &src_local->internal_lock;
-        if (src_lock->inode_locked_nodes) {
-                memcpy (dst_lock->inode_locked_nodes,
-                        src_lock->inode_locked_nodes,
-                        sizeof (*dst_lock->inode_locked_nodes) * child_count);
-                memset (src_lock->inode_locked_nodes, 0,
-                        sizeof (*src_lock->inode_locked_nodes) * child_count);
+        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_lock->inodelk_lock_count = src_lock->inodelk_lock_count;
-        src_lock->inodelk_lock_count = 0;
+        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 7e684801f..73594f265 100644
--- a/xlators/cluster/afr/src/afr-mem-types.h
+++ b/xlators/cluster/afr/src/afr-mem-types.h
@@ -41,6 +41,10 @@ enum gf_afr_mem_types_ {
         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 c0be197f2..643a5d692 100644
--- a/xlators/cluster/afr/src/afr-open.c
+++ b/xlators/cluster/afr/src/afr-open.c
@@ -249,189 +249,126 @@ out:
         return 0;
 }
 
-//NOTE: this function should be called with holding the lock on
-//fd to which fd_ctx belongs
-void
-afr_get_resumable_calls (xlator_t *this, afr_fd_ctx_t *fd_ctx,
-                         struct list_head *list)
-{
-        afr_fd_paused_call_t *paused_call = NULL;
-        afr_fd_paused_call_t *tmp = NULL;
-        afr_local_t           *call_local  = NULL;
-        afr_private_t         *priv        = NULL;
-        int                    i = 0;
-        gf_boolean_t           call = _gf_false;
-
-        priv = this->private;
-        list_for_each_entry_safe (paused_call, tmp, &fd_ctx->paused_calls,
-                                  call_list) {
-                call = _gf_true;
-                call_local = paused_call->frame->local;
-                for (i = 0; i < priv->child_count; i++) {
-                        if (call_local->child_up[i] &&
-                            (fd_ctx->opened_on[i] == AFR_FD_OPENING))
-                                call = _gf_false;
-                }
-
-                if (call) {
-                        list_del_init (&paused_call->call_list);
-                        list_add (&paused_call->call_list, list);
-                }
-        }
-}
-
-void
-afr_resume_calls (xlator_t *this, struct list_head *list)
-{
-        afr_fd_paused_call_t *paused_call = NULL;
-        afr_fd_paused_call_t *tmp = NULL;
-        afr_local_t           *call_local  = NULL;
-
-        list_for_each_entry_safe (paused_call, tmp, list, call_list) {
-                list_del_init (&paused_call->call_list);
-                call_local = paused_call->frame->local;
-                call_local->fop_call_continue (paused_call->frame, this);
-                GF_FREE (paused_call);
-        }
-}
-
 int
 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)
+                         int32_t op_ret, int32_t op_errno, fd_t *fd,
+                         dict_t *xdata)
 {
-        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;
-        struct list_head       paused_calls = {0};
-        gf_boolean_t           fop_paused = _gf_false;
-        fd_t                   *local_fd  = 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;
 
         priv     = this->private;
         local    = frame->local;
-        fop_paused = local->fop_paused;
 
         if (op_ret >= 0) {
-                gf_log (this->name, GF_LOG_INFO, "fd for %s opened "
+                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);
         }
 
-        local_fd = fd_ref (local->fd);
-        call_count = afr_frame_return (frame);
-        //Note: Do not access any thing using the frame outside call_count 0
-
-        //Note: No frame locking needed for this block of code
-        fd_ctx = afr_fd_ctx_get (local_fd, this);
+        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);
+                        "failed to get fd context, %p", local->fd);
                 goto out;
         }
 
-        LOCK (&local_fd->lock);
+        LOCK (&local->fd->lock);
         {
                 if (op_ret >= 0) {
                         fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
                 } else {
                         fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
                 }
-                if (call_count == 0) {
-                        INIT_LIST_HEAD (&paused_calls);
-                        afr_get_resumable_calls (this, fd_ctx, &paused_calls);
-                }
         }
-        UNLOCK (&local_fd->lock);
+        UNLOCK (&local->fd->lock);
 out:
-        if (call_count == 0) {
-                afr_resume_calls (this, &paused_calls);
-                //If the fop is paused then resume_calls will continue the fop
-                if (fop_paused)
-                        goto done;
-
-                if (local->fop_call_continue)
-                        local->fop_call_continue (frame, this);
-                else
-                        AFR_STACK_DESTROY (frame);
-        }
+        call_count = afr_frame_return (frame);
+        if (call_count == 0)
+                AFR_STACK_DESTROY (frame);
 
-done:
-        fd_unref (local_fd);
         return 0;
 }
 
-int
-afr_fix_open (call_frame_t *frame, xlator_t *this, afr_fd_ctx_t *fd_ctx,
-              int need_open_count, int *need_open)
+void
+afr_fix_open (xlator_t *this, fd_t *fd, size_t need_open_count, int *need_open)
 {
-        afr_local_t       *local = NULL;
-        afr_private_t     *priv  = NULL;
-        int               i      = 0;
-        call_frame_t      *open_frame = NULL;
-        afr_local_t      *open_local = NULL;
-        int               ret    = -1;
-        ia_type_t         ia_type = IA_INVAL;
-        int32_t           op_errno = 0;
-
-        GF_ASSERT (fd_ctx);
-        GF_ASSERT (need_open_count > 0);
-        GF_ASSERT (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;
 
-        local = frame->local;
         priv  = this->private;
-        if (!local->fop_call_continue) {
-                open_frame = copy_frame (frame);
-                if (!open_frame) {
-                        ret = -ENOMEM;
-                        goto out;
-                }
-                AFR_LOCAL_ALLOC_OR_GOTO (open_frame->local, out);
-                open_local = open_frame->local;
-                ret = afr_local_init (open_local, priv, &op_errno);
-                if (ret < 0)
-                        goto out;
-                loc_copy (&open_local->loc, &local->loc);
-                open_local->fd = fd_ref (local->fd);
-        } else {
-                ret = 0;
-                open_frame = frame;
-                open_local = local;
+
+        if (!afr_is_fd_fixable (fd) || !need_open || !need_open_count)
+                goto out;
+
+        fd_ctx = afr_fd_ctx_get (fd, this);
+        if (!fd_ctx) {
+                ret = -1;
+                goto out;
         }
 
-        open_local->call_count = need_open_count;
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame) {
+                ret = -1;
+                goto out;
+        }
 
-        gf_log (this->name, GF_LOG_DEBUG, "need open count: %d",
+        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->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;
+
+        gf_log (this->name, GF_LOG_DEBUG, "need open count: %zd",
                 need_open_count);
 
-        ia_type = open_local->fd->inode->ia_type;
-        GF_ASSERT (ia_type != IA_INVAL);
         for (i = 0; i < priv->child_count; i++) {
                 if (!need_open[i])
                         continue;
-                if (IA_IFDIR == ia_type) {
+
+                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);
 
-                        STACK_WIND_COOKIE (open_frame, afr_openfd_fix_open_cbk,
+                        STACK_WIND_COOKIE (frame, afr_openfd_fix_open_cbk,
                                            (void*) (long) i,
                                            priv->children[i],
                                            priv->children[i]->fops->opendir,
-                                           &open_local->loc, open_local->fd,
+                                           &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);
 
-                        STACK_WIND_COOKIE (open_frame, afr_openfd_fix_open_cbk,
+                        STACK_WIND_COOKIE (frame, afr_openfd_fix_open_cbk,
                                            (void *)(long) i,
                                            priv->children[i],
                                            priv->children[i]->fops->open,
-                                           &open_local->loc,
+                                           &local->loc,
                                            fd_ctx->flags & (~O_TRUNC),
-                                           open_local->fd, NULL);
+                                           local->fd, NULL);
                 }
 
         }
@@ -439,8 +376,7 @@ afr_fix_open (call_frame_t *frame, xlator_t *this, afr_fd_ctx_t *fd_ctx,
         ret = 0;
 out:
         if (op_errno)
-                ret = -op_errno;
-        if (ret && open_frame)
-                AFR_STACK_DESTROY (open_frame);
-        return ret;
+                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 9ad405920..83846f152 100644
--- a/xlators/cluster/afr/src/afr-self-heal-algorithm.c
+++ b/xlators/cluster/afr/src/afr-self-heal-algorithm.c
@@ -100,7 +100,7 @@ sh_loop_driver_done (call_frame_t *sh_frame, xlator_t *this,
         }
 
         sh_private_cleanup (sh_frame, this);
-        if (sh->op_failed) {
+        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,
@@ -143,7 +143,7 @@ sh_loop_finish (call_frame_t *loop_frame, xlator_t *this)
         }
 
         if (loop_sh && loop_sh->data_lock_held) {
-                afr_sh_data_unlock (loop_frame, this,
+                afr_sh_data_unlock (loop_frame, this, this->name,
                                     sh_destroy_frame);
         } else {
                 sh_destroy_frame (loop_frame, this);
@@ -214,7 +214,7 @@ sh_loop_frame_create (call_frame_t *sh_frame, xlator_t *this,
                 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_local_copy (local, this);
+        new_loop_local = afr_self_heal_local_init (local, this);
         if (!new_loop_local)
                 goto out;
         new_loop_frame->local = new_loop_local;
@@ -273,10 +273,10 @@ sh_loop_start (call_frame_t *sh_frame, xlator_t *this, off_t offset,
         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,
-                          sh_loop_lock_success, sh_loop_lock_failure);
+                          _gf_true, this->name, sh_loop_lock_success, sh_loop_lock_failure);
         return 0;
 out:
-        sh->op_failed = 1;
+        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);
@@ -307,8 +307,9 @@ sh_loop_driver (call_frame_t *sh_frame, xlator_t *this,
                         sh_priv->loops_running--;
                 offset = sh_priv->offset;
                 block_size = sh->block_size;
-                while ((!sh->eof_reached) && (0 == sh->op_failed) &&
-                       (sh_priv->loops_running < priv->data_self_heal_window_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++;
@@ -327,7 +328,8 @@ sh_loop_driver (call_frame_t *sh_frame, xlator_t *this,
         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 && !sh->op_failed)
+                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);
@@ -338,7 +340,7 @@ sh_loop_driver (call_frame_t *sh_frame, xlator_t *this,
         //If we have more loops to form we should finish previous loop after
         //the next loop lock
         while (loop--) {
-                if (sh->op_failed) {
+                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);
@@ -384,7 +386,7 @@ sh_loop_return (call_frame_t *sh_frame, xlator_t *this, call_frame_t *loop_frame
         }
 
         if (op_ret == -1) {
-                sh->op_failed = 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);
@@ -432,13 +434,22 @@ sh_loop_write_cbk (call_frame_t *loop_frame, void *cookie, xlator_t *this,
                         priv->children[child_index]->name,
                         strerror (op_errno));
 
-                sh->op_failed = 1;
+                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);
         }
 
         call_count = afr_frame_return (loop_frame);
 
         if (call_count == 0) {
+		iobref_unref(loop_local->cont.writev.iobref);
+
                 sh_loop_return (sh_frame, this, loop_frame,
                                 loop_sh->op_ret, loop_sh->op_errno);
         }
@@ -505,7 +516,7 @@ sh_loop_read_cbk (call_frame_t *loop_frame, void *cookie,
 
         if (op_ret <= 0) {
                 if (op_ret < 0) {
-                        sh->op_failed = 1;
+                        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));
@@ -527,8 +538,17 @@ sh_loop_read_cbk (call_frame_t *loop_frame, void *cookie,
                 sh_loop_return (sh_frame, this, loop_frame, 0, 0);
                 goto out;
         }
+
         loop_local->call_count = call_count;
 
+	/*
+	 * 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);
+
         for (i = 0; i < priv->child_count; i++) {
                 if (!loop_sh->write_needed[i])
                         continue;
@@ -606,7 +626,7 @@ sh_diff_checksum_cbk (call_frame_t *loop_frame, void *cookie, xlator_t *this,
                         "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_sh->checksum + child_index * MD5_DIGEST_LENGTH,
                         strong_checksum, MD5_DIGEST_LENGTH);
@@ -644,7 +664,8 @@ sh_diff_checksum_cbk (call_frame_t *loop_frame, void *cookie, xlator_t *this,
                 }
                 UNLOCK (&sh_priv->lock);
 
-                if (write_needed && !sh->op_failed) {
+                if (write_needed &&
+                    !is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
                         sh_loop_read (loop_frame, this);
                 } else {
                         sh_loop_return (sh_frame, this, loop_frame,
@@ -733,14 +754,15 @@ out:
         return sh_priv;
 }
 
-void
-afr_sh_transfer_lock (call_frame_t *dst, call_frame_t *src,
+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;
@@ -748,9 +770,12 @@ afr_sh_transfer_lock (call_frame_t *dst, call_frame_t *src,
         src_sh = &src_local->self_heal;
         GF_ASSERT (src_sh->data_lock_held);
         GF_ASSERT (!dst_sh->data_lock_held);
-        afr_lk_transfer_datalock (dst, src, child_count);
+        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;
 }
 
 int
@@ -772,7 +797,10 @@ afr_sh_start_loops (call_frame_t *sh_frame, xlator_t *this,
         ret = sh_loop_frame_create (sh_frame, this, NULL, &first_loop_frame);
         if (ret)
                 goto out;
-        afr_sh_transfer_lock (first_loop_frame, sh_frame, priv->child_count);
+        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;
@@ -782,7 +810,7 @@ afr_sh_start_loops (call_frame_t *sh_frame, xlator_t *this,
         ret = 0;
 out:
         if (ret) {
-                sh->op_failed = 1;
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
                 sh_loop_driver_done (sh_frame, this, NULL);
         }
         return 0;
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.c b/xlators/cluster/afr/src/afr-self-heal-common.c
index b6593a7ec..ef92b4205 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.c
+++ b/xlators/cluster/afr/src/afr-self-heal-common.c
@@ -18,6 +18,28 @@
 #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)
 {
@@ -112,8 +134,8 @@ void
 afr_sh_set_error (afr_self_heal_t *sh, int32_t op_errno)
 {
         sh->op_ret = -1;
-        if (afr_error_more_important (sh->op_errno, op_errno))
-                sh->op_errno = op_errno;
+	sh->op_errno = afr_most_important_error(sh->op_errno, op_errno,
+						_gf_false);
 }
 
 void
@@ -141,6 +163,79 @@ afr_sh_print_pending_matrix (int32_t *pending_matrix[], xlator_t *this)
         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_split_brain_log (int32_t *pending_matrix[], xlator_t *this,
+                              const char *loc)
+{
+        char *buf      = NULL;
+        char *free_ptr = NULL;
+
+        buf = afr_get_pending_matrix_str (pending_matrix, this);
+        if (buf)
+                free_ptr = buf;
+        else
+                buf = "";
+
+
+        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;
+}
+
+
 void
 afr_init_pending_matrix (int32_t **pending_matrix, size_t child_count)
 {
@@ -406,6 +501,8 @@ afr_find_biggest_witness_among_fools (int32_t *witnesses,
 {
         int i               = 0;
         int biggest_witness = -1;
+        int biggest_witness_idx = -1;
+        int biggest_witness_cnt = -1;
 
         GF_ASSERT (witnesses);
         GF_ASSERT (characters);
@@ -415,10 +512,21 @@ afr_find_biggest_witness_among_fools (int32_t *witnesses,
                 if (characters[i].type != AFR_NODE_FOOL)
                         continue;
 
-                if (biggest_witness < witnesses[i])
+                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 biggest_witness;
+
+	if (biggest_witness_cnt != 1)
+		return -1;
+
+        return biggest_witness_idx;
 }
 
 int
@@ -446,10 +554,84 @@ afr_mark_fool_as_source_by_witness (int32_t *sources, int32_t *witnesses,
         return nsources;
 }
 
+
+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_find_largest_file_size (struct iatt *bufs, int32_t *success_children,
+			    int child_count)
+{
+	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,
-                                     int child_count)
+				     int32_t *success_children,
+                                     int child_count, struct iatt *bufs)
 {
         int32_t       biggest_witness = 0;
         int           nsources        = 0;
@@ -457,6 +639,11 @@ afr_mark_biggest_of_fools_as_source (int32_t *sources, int32_t **pending_matrix,
 
         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) {
@@ -469,9 +656,15 @@ afr_mark_biggest_of_fools_as_source (int32_t *sources, int32_t **pending_matrix,
         biggest_witness = afr_find_biggest_witness_among_fools (witnesses,
                                                                 characters,
                                                                 child_count);
-        nsources = afr_mark_fool_as_source_by_witness (sources, witnesses,
-                                                       characters, child_count,
-                                                       biggest_witness);
+	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;
@@ -815,7 +1008,8 @@ afr_mark_sources (xlator_t *this, int32_t *sources, int32_t **pending_matrix,
                 nsources = afr_mark_biggest_of_fools_as_source (sources,
                                                                 pending_matrix,
                                                                 characters,
-                                                                child_count);
+								success_children,
+                                                                child_count, bufs);
         }
 
 out:
@@ -833,14 +1027,46 @@ afr_sh_pending_to_delta (afr_private_t *priv, dict_t **xattr,
                          int32_t *delta_matrix[], unsigned char success[],
                          int child_count, afr_transaction_type type)
 {
-        int i = 0;
-        int j = 0;
+        int     tgt     = 0;
+        int     src     = 0;
+        int     value   = 0;
 
         afr_build_pending_matrix (priv->pending_key, delta_matrix, NULL,
                                   xattr, type, priv->child_count);
-        for (i = 0; i < priv->child_count; i++)
-                for (j = 0; j < priv->child_count; j++)
-                        delta_matrix[i][j] = -delta_matrix[i][j];
+
+        /*
+         * 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;
+                        }
+                }
+        }
 }
 
 
@@ -867,8 +1093,14 @@ afr_sh_delta_to_xattr (xlator_t *this,
                         pending = GF_CALLOC (sizeof (int32_t), 3,
                                              gf_afr_mt_int32_t);
 
-                        if (!pending)
+                        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);
@@ -904,98 +1136,6 @@ afr_sh_delta_to_xattr (xlator_t *this,
 
 
 int
-afr_sh_has_metadata_pending (dict_t *xattr, xlator_t *this)
-{
-        /* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
-        int32_t        pending[3]  = {0,};
-        void          *pending_raw = NULL;
-        afr_private_t *priv        = NULL;
-        int            ret         = -1;
-        int            i           = 0;
-        int            j           = 0;
-
-        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)
-                        return 0;
-
-                memcpy (pending, pending_raw, sizeof(pending));
-                j = afr_index_for_transaction_type (AFR_METADATA_TRANSACTION);
-
-                if (pending[j])
-                        return 1;
-        }
-
-        return 0;
-}
-
-
-int
-afr_sh_has_data_pending (dict_t *xattr, xlator_t *this)
-{
-        /* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
-        int32_t        pending[3]  = {0,};
-        void          *pending_raw = NULL;
-        afr_private_t *priv        = NULL;
-        int            ret         = -1;
-        int            i           = 0;
-        int            j           = 0;
-
-        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)
-                        return 0;
-
-                memcpy (pending, pending_raw, sizeof(pending));
-                j = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
-
-                if (pending[j])
-                        return 1;
-        }
-
-        return 0;
-}
-
-
-int
-afr_sh_has_entry_pending (dict_t *xattr, xlator_t *this)
-{
-        /* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
-        int32_t        pending[3]  = {0,};
-        void          *pending_raw = NULL;
-        afr_private_t *priv        = NULL;
-        int            ret         = -1;
-        int            i           = 0;
-        int            j           = 0;
-
-        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)
-                        return 0;
-
-                memcpy (pending, pending_raw, sizeof(pending));
-                j = afr_index_for_transaction_type (AFR_ENTRY_TRANSACTION);
-
-                if (pending[j])
-                        return 1;
-        }
-
-        return 0;
-}
-
-int
 afr_sh_missing_entries_done (call_frame_t *frame, xlator_t *this)
 {
         afr_local_t     *local = NULL;
@@ -1006,14 +1146,13 @@ afr_sh_missing_entries_done (call_frame_t *frame, xlator_t *this)
 
         afr_sh_reset (frame, this);
 
-        if (local->govinda_gOvinda) {
+        if (local->unhealable) {
                 gf_log (this->name, GF_LOG_DEBUG,
                         "split brain found, aborting selfheal of %s",
                         local->loc.path);
-                sh->op_failed = 1;
         }
 
-        if (sh->op_failed) {
+        if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
                 sh->completion_cbk (frame, this);
         } else {
                 gf_log (this->name, GF_LOG_TRACE,
@@ -1245,7 +1384,7 @@ out:
         if (ret) {
                 gf_log (this->name, GF_LOG_ERROR, "impunge of %s failed, "
                         "reason: %s", local->loc.path, strerror (-ret));
-                sh->op_failed = 1;
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
         }
         afr_sh_missing_entries_finish (frame, this);
 }
@@ -1260,7 +1399,7 @@ afr_sh_create_entry_cbk (call_frame_t *frame, xlator_t *this,
         local = frame->local;
         sh = &local->self_heal;
         if (op_ret < 0)
-                sh->op_failed = 1;
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
         afr_sh_missing_entries_finish (frame, this);
         return 0;
 }
@@ -1284,7 +1423,7 @@ sh_missing_entries_create (call_frame_t *frame, xlator_t *this)
         if (!afr_valid_ia_type (type)) {
                 gf_log (this->name, GF_LOG_ERROR,
                         "%s: unknown file type: 0%o", local->loc.path, type);
-                local->govinda_gOvinda = 1;
+                afr_set_local_for_unhealable (local);
                 afr_sh_missing_entries_finish (frame, this);
                 goto out;
         }
@@ -1317,8 +1456,9 @@ afr_sh_missing_entries_lookup_done (call_frame_t *frame, xlator_t *this,
         loc = &local->loc;
 
         if (op_ret < 0) {
-                if (op_errno == EIO)
-                        local->govinda_gOvinda = 1;
+                if (op_errno == EIO) {
+                        afr_set_local_for_unhealable (local);
+                }
                 // EIO can happen if finding the fresh parent dir failed
                 goto out;
         }
@@ -1380,7 +1520,7 @@ afr_sh_missing_entries_lookup_done (call_frame_t *frame, xlator_t *this,
         }
         return;
 out:
-        sh->op_failed = 1;
+        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;
@@ -1464,7 +1604,7 @@ afr_sh_remove_entry_cbk (call_frame_t *frame, xlator_t *this, int child,
                 LOCK (&frame->lock);
                 {
                         afr_sh_set_error (sh, EIO);
-                        sh->op_failed = 1;
+                        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
                 }
                 UNLOCK (&frame->lock);
         }
@@ -1546,7 +1686,7 @@ afr_sh_purge_stale_entries_done (call_frame_t *frame, xlator_t *this)
         sh       = &local->self_heal;
         priv     = this->private;
 
-        if (sh->op_failed) {
+        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,
@@ -1639,7 +1779,7 @@ afr_sh_purge_entry_common (call_frame_t *frame, xlator_t *this,
                 if (!purge_condition (local, priv, i))
                         continue;
                 gf_log (this->name, GF_LOG_INFO, "purging the stale entry %s "
-                        "on %d", local->loc.path, i);
+                        "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],
@@ -1759,10 +1899,8 @@ afr_sh_children_lookup_done (call_frame_t *frame, xlator_t *this,
                                                sh->child_errno,
                                                priv->child_count, ENOENT);
         if (fresh_child_enoents == fresh_parent_count) {
-                gf_log (this->name, GF_LOG_INFO, "Deleting stale file %s",
-                        local->loc.path);
                 afr_sh_set_error (sh, ENOENT);
-                sh->op_failed = 1;
+                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,
@@ -1776,14 +1914,14 @@ afr_sh_children_lookup_done (call_frame_t *frame, xlator_t *this,
                 afr_sh_purge_stale_entry (frame, this);
         } else {
                 op_errno = EIO;
-                local->govinda_gOvinda = 1;
+                afr_set_local_for_unhealable (local);
                 goto fail;
         }
 
         return;
 
 fail:
-        sh->op_failed = 1;
+        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;
@@ -1854,8 +1992,8 @@ afr_sh_find_fresh_parents (call_frame_t *frame, xlator_t *this,
 
 out:
         afr_sh_set_error (sh, op_errno);
-        sh->op_failed = 1;
-        afr_sh_missing_entries_finish (frame, this);
+        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+	afr_sh_missing_entries_finish (frame, this);
         return;
 }
 
@@ -1944,7 +2082,8 @@ afr_sh_common_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
 
 
 int
-afr_sh_post_nb_entrylk_conflicting_sh_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;
@@ -1957,7 +2096,7 @@ afr_sh_post_nb_entrylk_conflicting_sh_cbk (call_frame_t *frame, xlator_t *this)
         if (int_lock->lock_op_ret < 0) {
                 gf_log (this->name, GF_LOG_INFO,
                         "Non blocking entrylks failed.");
-                sh->op_failed = -1;
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
                 afr_sh_missing_entries_done (frame, this);
         } else {
 
@@ -1973,40 +2112,14 @@ afr_sh_post_nb_entrylk_conflicting_sh_cbk (call_frame_t *frame, xlator_t *this)
 }
 
 int
-afr_sh_post_nb_entrylk_gfid_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;
-        sh       = &local->self_heal;
-        int_lock = &local->internal_lock;
-
-        if (int_lock->lock_op_ret < 0) {
-                gf_log (this->name, GF_LOG_INFO,
-                        "Non blocking entrylks failed.");
-                afr_sh_missing_entries_done (frame, this);
-        } else {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Non blocking entrylks done. Proceeding to FOP");
-                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);
-        }
-
-        return 0;
-}
-
-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_private_t       *priv     = NULL;
 
+        priv     = this->private;
         local    = frame->local;
         int_lock = &local->internal_lock;
 
@@ -2018,7 +2131,12 @@ afr_sh_entrylk (call_frame_t *frame, xlator_t *this, loc_t *loc,
         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;
@@ -2056,27 +2174,31 @@ out:
 }
 
 static int
-afr_self_heal_conflicting_entries (call_frame_t *frame, xlator_t *this)
+afr_self_heal_missing_entries (call_frame_t *frame, xlator_t *this)
 {
-        afr_self_heal_parent_entrylk (frame, this,
-                                      afr_sh_post_nb_entrylk_conflicting_sh_cbk);
-        return 0;
-}
+        afr_local_t     *local = NULL;
+        afr_self_heal_t *sh = 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);
 
-static int
-afr_self_heal_gfids (call_frame_t *frame, xlator_t *this)
-{
         afr_self_heal_parent_entrylk (frame, this,
-                                      afr_sh_post_nb_entrylk_gfid_sh_cbk);
+                                      afr_sh_post_nb_entrylk_missing_entry_sh_cbk);
         return 0;
 }
 
-afr_local_t *afr_local_copy (afr_local_t *l, xlator_t *this)
+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;
+        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;
 
@@ -2099,13 +2221,23 @@ afr_local_t *afr_local_copy (afr_local_t *l, xlator_t *this)
         shc->forced_merge = sh->forced_merge;
         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)
-                loc_copy (&lc->loc, &l->loc);
+        if (l->loc.path) {
+                ret = loc_copy (&lc->loc, &l->loc);
+                if (ret < 0)
+                        goto out;
+        }
 
         lc->child_up  = memdup (l->child_up,
                                 sizeof (*lc->child_up) * priv->child_count);
+        if (!lc->child_up) {
+                ret = -1;
+                goto out;
+        }
+
         if (l->xattr_req)
                 lc->xattr_req = dict_ref (l->xattr_req);
 
@@ -2113,40 +2245,25 @@ 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,
-                                sizeof (*lc->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,
-                                sizeof (*lc->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,
-                                sizeof (*lc->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;
 }
 
@@ -2156,41 +2273,39 @@ afr_self_heal_completion_cbk (call_frame_t *bgsh_frame, xlator_t *this)
         afr_private_t *   priv  = 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_boolean_t      split_brain = _gf_false;
+        gf_loglevel_t     loglevel = 0;
 
         priv  = this->private;
         local = bgsh_frame->local;
         sh    = &local->self_heal;
 
-        if (local->govinda_gOvinda || sh->mdata_spb || sh->data_spb) {
-                split_brain = _gf_true;
-                sh->op_failed = 1;
+        if (local->unhealable) {
+                afr_set_split_brain (this, sh->inode, SPB, SPB);
         }
 
-        afr_set_split_brain (this, sh->inode, split_brain);
-
         afr_self_heal_type_str_get (sh, sh_type_str,
                                     sizeof(sh_type_str));
-        if (sh->op_failed) {
-                gf_loglevel_t     loglevel = GF_LOG_ERROR;
-                if (priv->shd.iamshd)
-                        loglevel = GF_LOG_DEBUG;
-
-                gf_log (this->name, loglevel, "background %s self-heal "
-                        "failed on %s", sh_type_str, local->loc.path);
-
+        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 {
-                gf_log (this->name, GF_LOG_DEBUG, "background %s self-heal "
-                        "completed on %s", sh_type_str, local->loc.path);
-
+                loglevel = GF_LOG_DEBUG;
         }
 
+        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,
-                            sh->op_failed);
+                            is_self_heal_failed (sh, AFR_CHECK_ALL));
         }
 
         if (sh->background) {
@@ -2239,7 +2354,7 @@ afr_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode)
         afr_set_lk_owner (sh_frame, this, sh_frame->root);
         afr_set_low_priority (sh_frame);
 
-        sh_local        = afr_local_copy (local, this);
+        sh_local        = afr_self_heal_local_init (local, this);
         if (!sh_local)
                 goto out;
         sh_frame->local = sh_local;
@@ -2304,12 +2419,11 @@ afr_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode)
                 sh->do_gfid_self_heal = _gf_false;
         }
 
+        sh->sh_type_in_action = AFR_SELF_HEAL_INVALID;
+
         FRAME_SU_DO (sh_frame, afr_local_t);
-        if (sh->do_missing_entry_self_heal) {
-                afr_self_heal_conflicting_entries (sh_frame, this);
-        } else if (sh->do_gfid_self_heal) {
-                GF_ASSERT (!uuid_is_null (sh->sh_gfid_req));
-                afr_self_heal_gfids (sh_frame, this);
+        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)) {
@@ -2516,9 +2630,183 @@ out:
         GF_FREE (erase_xattr);
 
         if (ret < 0) {
-                sh->op_failed = _gf_true;
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
                 finish (frame, this);
         }
 
         return 0;
 }
+
+void
+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)
+{
+        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;
+
+        }
+
+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;
+        }
+
+        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 6eabd1766..473264776 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.h
+++ b/xlators/cluster/afr/src/afr-self-heal-common.h
@@ -15,13 +15,6 @@
 #define AFR_SH_MIN_PARTICIPANTS 2
 
 typedef enum {
-        AFR_SELF_HEAL_ENTRY,
-        AFR_SELF_HEAL_METADATA,
-        AFR_SELF_HEAL_DATA,
-        AFR_SELF_HEAL_INVALID = -1,
-} afr_self_heal_type;
-
-typedef enum {
         AFR_LOOKUP_FAIL_CONFLICTS = 1,
         AFR_LOOKUP_FAIL_MISSING_GFIDS = 2,
 } afr_lookup_flags_t;
@@ -35,6 +28,10 @@ afr_sh_source_count (int sources[], int child_count);
 void
 afr_sh_print_pending_matrix (int32_t *pending_matrix[], xlator_t *this);
 
+void
+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,
@@ -94,13 +91,13 @@ 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,
+afr_sh_data_unlock (call_frame_t *frame, xlator_t *this, char *dom,
                     afr_lock_cbk_t lock_cbk);
 afr_local_t *
-afr_local_copy (afr_local_t *l, xlator_t *this);
+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,
+                  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
@@ -129,4 +126,19 @@ 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 6501e596c..9de26ee56 100644
--- a/xlators/cluster/afr/src/afr-self-heal-data.c
+++ b/xlators/cluster/afr/src/afr-self-heal-data.c
@@ -156,6 +156,25 @@ afr_sh_data_close (call_frame_t *frame, xlator_t *this)
 }
 
 int
+afr_sh_dom_unlock (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;
+
+        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;
+}
+
+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)
@@ -190,29 +209,20 @@ afr_sh_data_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 }
 
 int
-afr_sh_data_setattr (call_frame_t *frame, xlator_t *this)
+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;
-        int              source     = 0;
         int32_t          valid      = 0;
-        struct iatt      stbuf      = {0,};
 
         local = frame->local;
         sh    = &local->self_heal;
         priv  = this->private;
 
-        source = sh->source;
-
-        valid |= (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME);
-
-        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;
+        valid = (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME);
 
         call_count        = afr_set_elem_count_get (sh->success,
                                                     priv->child_count);
@@ -232,7 +242,7 @@ afr_sh_data_setattr (call_frame_t *frame, xlator_t *this)
                                    (void *) (long) i,
                                    priv->children[i],
                                    priv->children[i]->fops->setattr,
-                                   &local->loc, &stbuf, valid, NULL);
+                                   &local->loc, stbuf, valid, NULL);
 
                 if (!--call_count)
                         break;
@@ -256,7 +266,7 @@ afr_sh_data_setattr_fstat_cbk (call_frame_t *frame, void *cookie,
         GF_ASSERT (sh->source == child_index);
         if (op_ret != -1) {
                 sh->buf[child_index] = *buf;
-                afr_sh_data_setattr (frame, this);
+                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);
@@ -292,31 +302,45 @@ afr_sh_set_timestamps (call_frame_t *frame, xlator_t *this)
 
 //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_private_t       *priv     = NULL;
+        int                 ret       = 0;
 
         local    = frame->local;
         int_lock = &local->internal_lock;
         sh       = &local->self_heal;
+        priv     = this->private;
 
-        GF_ASSERT (sh->data_lock_held);
-
-        sh->data_lock_held = _gf_false;
+        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);
 
+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;
@@ -325,9 +349,9 @@ afr_sh_data_finish (call_frame_t *frame, xlator_t *this)
                 "finishing data selfheal of %s", local->loc.path);
 
         if (sh->data_lock_held)
-                afr_sh_data_unlock (frame, this, afr_sh_data_close);
+                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;
 }
@@ -344,11 +368,8 @@ afr_sh_data_fail (call_frame_t *frame, xlator_t *this)
         gf_log (this->name, GF_LOG_DEBUG,
                 "finishing failed data selfheal of %s", local->loc.path);
 
-        sh->op_failed = 1;
-        if (sh->data_lock_held)
-                afr_sh_data_unlock (frame, this, afr_sh_data_close);
-        else
-                afr_sh_data_close (frame, this);
+        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+        afr_sh_data_finish (frame, this);
         return 0;
 }
 
@@ -371,13 +392,13 @@ afr_sh_data_erase_pending_cbk (call_frame_t *frame, void *cookie,
                         "log failed on %s for subvol %s, reason: %s",
                         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);
         }
 
         call_count = afr_frame_return (frame);
 
         if (call_count == 0) {
-                if (sh->op_failed) {
+                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;
@@ -389,7 +410,7 @@ afr_sh_data_erase_pending_cbk (call_frame_t *frame, void *cookie,
                         goto out;
                 }
                 GF_ASSERT (sh->old_loop_frame);
-                afr_sh_data_lock (frame, this, 0, 0,
+                afr_sh_data_lock (frame, this, 0, 0, _gf_true, this->name,
                                   afr_post_sh_big_lock_success,
                                   afr_post_sh_big_lock_failure);
         }
@@ -406,12 +427,89 @@ afr_sh_data_erase_pending (call_frame_t *frame, xlator_t *this)
         return 0;
 }
 
+int
+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_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;
+        }
+
+        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;
+}
+
+/*
+ * 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 = frame->local;
+        priv = this->private;
+        sh   = &local->self_heal;
+
+        call_count        = sh->active_sinks;
+        if (call_count == 0) {
+                afr_sh_data_erase_pending (frame, this);
+                return 0;
+        }
+
+        local->call_count = call_count;
+        for (i = 0; i < priv->child_count; i++) {
+                if (!sh->success[i] || sh->sources[i])
+                        continue;
+
+                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;
+}
 
 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];
@@ -420,6 +518,7 @@ sh_algo_from_name (xlator_t *this, char *name)
                 i++;
         }
 
+out:
         return NULL;
 }
 
@@ -499,7 +598,7 @@ afr_sh_data_sync_prepare (call_frame_t *frame, xlator_t *this)
         local = frame->local;
         sh = &local->self_heal;
 
-        sh->algo_completion_cbk = afr_sh_data_erase_pending;
+        sh->algo_completion_cbk = afr_sh_data_fsync;
         sh->algo_abort_cbk      = afr_sh_data_fail;
 
         sh_algo = afr_sh_data_pick_algo (frame, this);
@@ -535,7 +634,7 @@ afr_sh_data_trim_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                 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 {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "ftruncate of %s on subvolume %s completed",
@@ -548,7 +647,7 @@ afr_sh_data_trim_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         call_count = afr_frame_return (frame);
 
         if (call_count == 0) {
-                if (sh->op_failed)
+                if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC))
                         afr_sh_data_fail (frame, this);
                 else
                         afr_sh_data_sync_prepare (frame, this);
@@ -628,6 +727,199 @@ out:
         return ret;
 }
 
+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++;
+                }
+
+        }
+
+        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);
+        }
+
+        len++;//for '\0'
+
+        sizes_str = GF_CALLOC (len, sizeof (char), gf_common_mt_char);
+
+        if (!sizes_str)
+                return NULL;
+
+        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);
+                }
+        }
+
+        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);
+                }
+        }
+
+        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);
+                }
+        }
+
+        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);
+                }
+        }
+
+        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);
+
+                }
+        }
+
+        return sizes_str;
+}
+
+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);
+                }
+        }
+
+        len ++;
+
+        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);
+                }
+        }
+
+        return sinks_str;
+
+}
+
+
+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)
 {
@@ -649,7 +941,7 @@ afr_sh_data_fix (call_frame_t *frame, xlator_t *this)
 
         if (sh->background && sh->unwind && !sh->unwound) {
                 sh->unwind (sh->orig_frame, this, sh->op_ret, sh->op_errno,
-                            sh->op_failed);
+                            is_self_heal_failed (sh, AFR_CHECK_SPECIFIC));
                 sh->unwound = _gf_true;
         }
 
@@ -666,6 +958,9 @@ afr_sh_data_fix (call_frame_t *frame, xlator_t *this)
                 "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);
 }
 
@@ -677,6 +972,9 @@ afr_sh_data_fxattrop_fstat_done (call_frame_t *frame, xlator_t *this)
         afr_private_t   *priv = NULL;
         int              nsources = 0;
         int              ret = 0;
+        int             *old_sources = NULL;
+        int             tstamp_source = 0;
+        int             i = 0;
 
         local = frame->local;
         sh = &local->self_heal;
@@ -684,6 +982,13 @@ afr_sh_data_fxattrop_fstat_done (call_frame_t *frame, xlator_t *this)
 
         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));
+        }
 
         nsources = afr_build_sources (this, sh->xattr, sh->buf, sh->pending_matrix,
                                       sh->sources, sh->success_children,
@@ -705,17 +1010,16 @@ afr_sh_data_fxattrop_fstat_done (call_frame_t *frame, xlator_t *this)
         }
 
         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);
+                afr_sh_print_split_brain_log (sh->pending_matrix, this,
+                                              local->loc.path);
+                afr_set_split_brain (this, sh->inode, DONT_KNOW, SPB);
 
-                sh->data_spb = _gf_true;
                 afr_sh_data_fail (frame, this);
                 return 0;
         }
 
-        sh->data_spb = _gf_false;
+        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,
@@ -726,8 +1030,18 @@ afr_sh_data_fxattrop_fstat_done (call_frame_t *frame, xlator_t *this)
         }
 
         if (sh->sync_done) {
-                afr_sh_data_setattr (frame, this);
+                /* 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",
@@ -985,11 +1299,12 @@ 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;
+        dict_t          **xattr_req;
         int32_t         *zero_pending = NULL;
         int call_count = 0;
         int i = 0;
         int ret = 0;
+	int j;
 
         priv  = this->private;
         local = frame->local;
@@ -1000,30 +1315,39 @@ afr_sh_data_fxattrop (call_frame_t *frame, xlator_t *this)
 
         local->call_count = call_count;
 
-        xattr_req = dict_new();
-        if (!xattr_req) {
-                ret = -1;
-                goto out;
-        }
-
-        for (i = 0; i < priv->child_count; i++) {
-                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, priv->pending_key[i],
-                                       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;
-                }
-        }
+	xattr_req = GF_CALLOC(priv->child_count, sizeof(struct dict_t *),
+			      gf_afr_mt_dict_t);
+	if (!xattr_req)
+		goto out;
+
+	for (i = 0; i < priv->child_count; i++) {
+		xattr_req[i] = dict_new();
+		if (!xattr_req[i]) {
+			ret = -1;
+			goto out;
+		}
+	}
+
+	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);
@@ -1037,7 +1361,7 @@ afr_sh_data_fxattrop (call_frame_t *frame, xlator_t *this)
                                            priv->children[i],
                                            priv->children[i]->fops->fxattrop,
                                            sh->healing_fd, GF_XATTROP_ADD_ARRAY,
-                                           xattr_req, NULL);
+                                           xattr_req[i], NULL);
 
                         if (!--call_count)
                                 break;
@@ -1045,8 +1369,12 @@ afr_sh_data_fxattrop (call_frame_t *frame, xlator_t *this)
         }
 
 out:
-        if (xattr_req)
-                dict_unref (xattr_req);
+	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 (ret) {
                 GF_FREE (zero_pending);
@@ -1071,6 +1399,22 @@ afr_sh_data_big_lock_success (call_frame_t *frame, xlator_t *this)
 }
 
 int
+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_blocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
 {
         afr_internal_lock_t *int_lock = NULL;
@@ -1115,8 +1459,13 @@ afr_sh_data_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
                         "failed for %s. by %s",
                         local->loc.path, lkowner_utoa (&frame->root->lk_owner));
 
-                int_lock->lock_cbk = afr_sh_data_post_blocking_inodelk_cbk;
-                afr_blocking_lock (frame, this);
+		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 "
@@ -1129,9 +1478,11 @@ afr_sh_data_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
 }
 
 int
-afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this, off_t start, off_t len)
+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;
@@ -1142,11 +1493,14 @@ afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this, off_t start, off_t le
 
         afr_set_lock_number (frame, this);
 
-        int_lock->lk_flock.l_start = start;
-        int_lock->lk_flock.l_len   = len;
-        int_lock->lk_flock.l_type  = F_WRLCK;
         int_lock->lock_cbk         = afr_sh_data_post_nonblocking_inodelk_cbk;
 
+        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;
@@ -1189,8 +1543,8 @@ afr_post_sh_big_lock_failure (call_frame_t *frame, xlator_t *this)
 
 int
 afr_sh_data_lock (call_frame_t *frame, xlator_t *this,
-                  off_t start, off_t len,
-                  afr_lock_cbk_t success_handler,
+                  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;
@@ -1201,7 +1555,8 @@ afr_sh_data_lock (call_frame_t *frame, xlator_t *this,
 
         sh->data_lock_success_handler = success_handler;
         sh->data_lock_failure_handler = failure_handler;
-        return afr_sh_data_lock_rec (frame, this, start, len);
+	sh->data_lock_block = block;
+        return afr_sh_data_lock_rec (frame, this, dom, start, len);
 }
 
 int
@@ -1232,7 +1587,7 @@ afr_sh_data_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                 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 {
                         gf_log (this->name, GF_LOG_TRACE,
                                 "open of %s succeeded on child %s",
@@ -1245,7 +1600,7 @@ afr_sh_data_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         call_count = afr_frame_return (frame);
 
         if (call_count == 0) {
-                if (sh->op_failed) {
+                if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
                         afr_sh_data_fail (frame, this);
                         return 0;
                 }
@@ -1254,9 +1609,8 @@ afr_sh_data_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         "fd for %s opened, commencing sync",
                         local->loc.path);
 
-                afr_sh_data_lock (frame, this, 0, 0,
-                                  afr_sh_data_big_lock_success,
-                                  afr_sh_data_fail);
+                afr_sh_data_lock (frame, this, 0, 0, _gf_true, priv->sh_domain,
+                                  afr_sh_dom_lock_success, afr_sh_data_fail);
         }
 
         return 0;
@@ -1361,25 +1715,31 @@ afr_can_start_data_self_heal (afr_self_heal_t *sh, afr_private_t *priv)
 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;
 
-        /* Self-heal completion cbk changes inode split-brain status based on
-         * govinda_gOvinda, mdata_spb, data_spb value.
-         * Initialize data_spb with current split-brain status.
-         * If for some reason self-heal fails(locking phase etc), it makes sure
-         * we retain the split-brain status before this self-heal started.
-         */
-        sh->data_spb = afr_is_split_brain (this, sh->inode);
+        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 (IA_ISREG (sh->type)) {
                         afr_sh_data_open (frame, this);
                 } else {
-                        afr_sh_data_lock (frame, this, 0, 0,
+                        afr_sh_data_lock (frame, this, 0, 0, _gf_true,
+                                          this->name,
                                           afr_sh_non_reg_lock_success,
                                           afr_sh_data_fail);
                 }
diff --git a/xlators/cluster/afr/src/afr-self-heal-entry.c b/xlators/cluster/afr/src/afr-self-heal-entry.c
index e7f74c752..53491a1d7 100644
--- a/xlators/cluster/afr/src/afr-self-heal-entry.c
+++ b/xlators/cluster/afr/src/afr-self-heal-entry.c
@@ -162,7 +162,7 @@ afr_sh_entry_erase_pending (call_frame_t *frame, xlator_t *this)
         sh = &local->self_heal;
 
         if (sh->entries_skipped) {
-                sh->op_failed = _gf_true;
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
                 goto out;
         }
         afr_sh_erase_pending (frame, this, AFR_ENTRY_TRANSACTION,
@@ -580,7 +580,8 @@ afr_sh_entry_expunge_entry_cbk (call_frame_t *expunge_frame, void *cookie,
 
         if (need_expunge) {
                 gf_log (this->name, GF_LOG_INFO,
-                        "missing entry %s on %s",
+                        "Entry %s is missing on %s and deleting from "
+                        "replica's other bricks",
                         expunge_local->loc.path,
                         priv->children[source]->name);
 
@@ -612,6 +613,19 @@ out:
         return 0;
 }
 
+static gf_boolean_t
+can_skip_entry_self_heal (char *name, loc_t *parent_loc)
+{
+        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_entry (call_frame_t *frame, xlator_t *this,
@@ -639,13 +653,7 @@ afr_sh_entry_expunge_entry (call_frame_t *frame, xlator_t *this,
         sh->expunge_done = afr_sh_entry_expunge_entry_done;
 
         name = entry->d_name;
-
-        if ((strcmp (name, ".") == 0)
-            || (strcmp (name, "..") == 0)) {
-
-                gf_log (this->name, GF_LOG_TRACE,
-                        "skipping inspection of %s under %s",
-                        name, local->loc.path);
+        if (can_skip_entry_self_heal (name, &local->loc)) {
                 op_ret = 0;
                 goto out;
         }
@@ -798,7 +806,7 @@ afr_sh_entry_expunge_all (call_frame_t *frame, xlator_t *this)
         active_src = next_active_sink (frame, this, sh->active_source);
         sh->active_source = active_src;
 
-        if (sh->op_failed) {
+        if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
                 goto out;
         }
 
@@ -1017,33 +1025,6 @@ out:
         return 0;
 }
 
-void
-afr_sh_prepare_new_entry_pending_matrix (int32_t **pending,
-                                         int *child_errno,
-                                         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 (child_errno[i])
-                        continue;
-                pending[i][midx] = hton32 (1);
-                if (idx == -1)
-                        continue;
-                pending[i][idx] = hton32 (1);
-        }
-}
-
 int
 afr_sh_entry_impunge_perform_xattrop (call_frame_t *impunge_frame,
                                       xlator_t *this)
@@ -1060,10 +1041,11 @@ afr_sh_entry_impunge_perform_xattrop (call_frame_t *impunge_frame,
         impunge_sh = &impunge_local->self_heal;
         active_src = impunge_sh->active_source;
 
-        afr_sh_prepare_new_entry_pending_matrix (impunge_local->pending,
-                                                 impunge_sh->child_errno,
-                                                 &impunge_sh->entrybuf,
-                                                 priv->child_count);
+        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;
@@ -1281,6 +1263,35 @@ afr_sh_entry_impunge_mknod (call_frame_t *impunge_frame, xlator_t *this,
                 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],
@@ -1897,12 +1908,7 @@ afr_sh_entry_impunge_entry (call_frame_t *frame, xlator_t *this,
         active_src = sh->active_source;
         sh->impunge_done = afr_sh_entry_impunge_entry_done;
 
-        if ((strcmp (entry->d_name, ".") == 0)
-            || (strcmp (entry->d_name, "..") == 0)) {
-
-                gf_log (this->name, GF_LOG_TRACE,
-                        "skipping inspection of %s under %s",
-                        entry->d_name, local->loc.path);
+        if (can_skip_entry_self_heal (entry->d_name, &local->loc)) {
                 op_ret = 0;
                 goto out;
         }
@@ -1971,7 +1977,7 @@ afr_sh_entry_impunge_readdir_cbk (call_frame_t *frame, void *cookie,
                                 local->loc.path,
                                 priv->children[active_src]->name,
                                 strerror (op_errno));
-                        sh->op_failed = 1;
+                        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",
@@ -2044,7 +2050,7 @@ afr_sh_entry_impunge_all (call_frame_t *frame, xlator_t *this)
         active_src = next_active_source (frame, this, sh->active_source);
         sh->active_source = active_src;
 
-        if (sh->op_failed) {
+        if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
                 afr_sh_entry_finish (frame, this);
                 return 0;
         }
@@ -2093,7 +2099,7 @@ afr_sh_entry_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                 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);
                 }
         }
         UNLOCK (&frame->lock);
@@ -2101,7 +2107,7 @@ afr_sh_entry_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         call_count = afr_frame_return (frame);
 
         if (call_count == 0) {
-                if (sh->op_failed) {
+                if (is_self_heal_failed (sh, AFR_CHECK_SPECIFIC)) {
                         afr_sh_entry_finish (frame, this);
                         return 0;
                 }
@@ -2139,7 +2145,7 @@ afr_sh_entry_open (call_frame_t *frame, xlator_t *this)
         source  = local->self_heal.source;
         sources = local->self_heal.sources;
 
-        sh->block_size = 65536; //131072
+        sh->block_size = priv->sh_readdir_size;
         sh->offset = 0;
 
         call_count = sh->active_sinks;
@@ -2233,6 +2239,8 @@ afr_sh_entry_sync_prepare (call_frame_t *frame, xlator_t *this)
                         "merging all entries as a conservative decision",
                         local->loc.path);
 
+        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;
@@ -2255,7 +2263,7 @@ afr_sh_entry_fix (call_frame_t *frame, xlator_t *this,
         priv = this->private;
 
         if (op_ret < 0) {
-                sh->op_failed = 1;
+                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;
@@ -2318,7 +2326,7 @@ afr_sh_post_nonblocking_entry_cbk (call_frame_t *frame, xlator_t *this)
         if (int_lock->lock_op_ret < 0) {
                 gf_log (this->name, GF_LOG_ERROR, "Non Blocking entrylks "
                         "failed for %s.", local->loc.path);
-                sh->op_failed = 1;
+                afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
                 afr_sh_entry_done (frame, this);
         } else {
 
@@ -2337,14 +2345,18 @@ afr_sh_post_nonblocking_entry_cbk (call_frame_t *frame, xlator_t *this)
 int
 afr_self_heal_entry (call_frame_t *frame, xlator_t *this)
 {
-        afr_local_t   *local = 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;
+
+        sh->sh_type_in_action = AFR_SELF_HEAL_ENTRY;
 
         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 {
diff --git a/xlators/cluster/afr/src/afr-self-heal-metadata.c b/xlators/cluster/afr/src/afr-self-heal-metadata.c
index 6e98f2a63..fd5da6cfd 100644
--- a/xlators/cluster/afr/src/afr-self-heal-metadata.c
+++ b/xlators/cluster/afr/src/afr-self-heal-metadata.c
@@ -88,6 +88,19 @@ afr_sh_metadata_finish (call_frame_t *frame, xlator_t *this)
         return 0;
 }
 
+int
+afr_sh_metadata_fail (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;
+
+        afr_set_self_heal_status (sh, AFR_SELF_HEAL_FAILED);
+        afr_sh_metadata_finish (frame, this);
+        return 0;
+}
 
 int
 afr_sh_metadata_erase_pending_cbk (call_frame_t *frame, void *cookie,
@@ -167,8 +180,13 @@ afr_sh_metadata_sync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         call_count = afr_frame_return (frame);
 
-        if (call_count == 0)
+        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;
 }
@@ -194,6 +212,86 @@ afr_sh_metadata_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         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)
@@ -219,9 +317,10 @@ afr_sh_metadata_sync (call_frame_t *frame, xlator_t *this, dict_t *xattr)
         /*
          * 2 calls per sink - setattr, setxattr
          */
-        if (xattr)
+        if (xattr) {
                 call_count = active_sinks * 2;
-        else
+                local->xattr_req = dict_ref (xattr);
+        } else
                 call_count = active_sinks;
 
         local->call_count = call_count;
@@ -264,11 +363,11 @@ afr_sh_metadata_sync (call_frame_t *frame, xlator_t *this, dict_t *xattr)
                 if (!xattr)
                         continue;
 
-                STACK_WIND_COOKIE (frame, afr_sh_metadata_xattr_cbk,
+                STACK_WIND_COOKIE (frame, afr_sh_getxattr_cbk,
                                    (void *) (long) i,
                                    priv->children[i],
-                                   priv->children[i]->fops->setxattr,
-                                   &local->loc, xattr, 0, NULL);
+                                   priv->children[i]->fops->getxattr,
+                                   &local->loc, NULL, NULL);
                 call_count--;
         }
 
@@ -286,8 +385,6 @@ afr_sh_metadata_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         afr_private_t   *priv = NULL;
         int              source = 0;
 
-        int i;
-
         local = frame->local;
         sh = &local->self_heal;
         priv = this->private;
@@ -302,16 +399,147 @@ afr_sh_metadata_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
                 afr_sh_metadata_sync (frame, this, NULL);
         } else {
-                for (i = 0; i < priv->child_count; i++) {
-                        dict_del (xattr, priv->pending_key[i]);
-                }
-
+                afr_prune_pending_keys (xattr, priv);
                 afr_sh_metadata_sync (frame, this, xattr);
         }
 
         return 0;
 }
 
+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++;
+                }
+        }
+
+        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);
+                }
+        }
+
+        len ++;
+
+        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);
+        }
+
+        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);
+                }
+        }
+
+        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);
+        }
+
+        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);
+                }
+        }
+
+        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);
+        }
+
+        gf_asprintf (&sh->metadata_sh_info, "%s metadata %s,", string,
+                     pending_matrix_str);
+
+        if (pending_matrix_str && strcmp (pending_matrix_str, ""))
+                GF_FREE (pending_matrix_str);
+
+        if (string && strcmp (string, ""))
+                GF_FREE (string);
+}
 
 int
 afr_sh_metadata_sync_prepare (call_frame_t *frame, xlator_t *this)
@@ -341,6 +569,9 @@ afr_sh_metadata_sync_prepare (call_frame_t *frame, xlator_t *this)
                 local->loc.path, priv->children[source]->name,
                 sh->active_sinks);
 
+        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,
@@ -366,7 +597,7 @@ afr_sh_metadata_fix (call_frame_t *frame, xlator_t *this,
         priv = this->private;
 
         if (op_ret < 0) {
-                sh->op_failed = 1;
+                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;
@@ -391,18 +622,14 @@ afr_sh_metadata_fix (call_frame_t *frame, xlator_t *this,
         }
 
         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);
-
-                sh->mdata_spb = _gf_true;
-
-                afr_sh_metadata_finish (frame, this);
+                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;
         }
 
-        sh->mdata_spb = _gf_false;
+        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",
@@ -488,19 +715,22 @@ 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 = LLONG_MAX - 1;
-        int_lock->lk_flock.l_len   = 0;
-        int_lock->lk_flock.l_type  = F_WRLCK;
+        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);
@@ -527,16 +757,10 @@ afr_self_heal_metadata (call_frame_t *frame, xlator_t *this)
 
         local = frame->local;
         sh = &local->self_heal;
-
-        /* Self-heal completion cbk changes inode split-brain status based on
-         * govinda_gOvinda, mdata_spb, data_spb value.
-         * Initialize mdata_spb with current split-brain status.
-         * If for some reason self-heal fails(locking phase etc), it makes sure
-         * we retain the split-brain status before this self-heal started.
-         */
-        sh->mdata_spb = afr_is_split_brain (this, sh->inode);
+        sh->sh_type_in_action = AFR_SELF_HEAL_METADATA;
 
         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);
diff --git a/xlators/cluster/afr/src/afr-self-heal.h b/xlators/cluster/afr/src/afr-self-heal.h
index 2efc1116d..7c9bc8111 100644
--- a/xlators/cluster/afr/src/afr-self-heal.h
+++ b/xlators/cluster/afr/src/afr-self-heal.h
@@ -21,13 +21,6 @@
 #define SIZE_GREATER(buf1,buf2) ((buf1)->ia_size > (buf2)->ia_size)
 
 int
-afr_sh_has_metadata_pending (dict_t *xattr, xlator_t *this);
-int
-afr_sh_has_entry_pending (dict_t *xattr, xlator_t *this);
-int
-afr_sh_has_data_pending (dict_t *xattr, xlator_t *this);
-
-int
 afr_self_heal_entry (call_frame_t *frame, xlator_t *this);
 
 int
diff --git a/xlators/cluster/afr/src/afr-self-heald.c b/xlators/cluster/afr/src/afr-self-heald.c
index b5653dee7..1b48a1bca 100644
--- a/xlators/cluster/afr/src/afr-self-heald.c
+++ b/xlators/cluster/afr/src/afr-self-heald.c
@@ -25,7 +25,8 @@ typedef enum {
 
 typedef enum {
         HEAL = 1,
-        INFO
+        INFO,
+        STATISTICS_TO_BE_HEALED,
 } shd_crawl_op;
 
 typedef struct shd_dump {
@@ -57,6 +58,11 @@ afr_start_crawl (xlator_t *this, int idx, afr_crawl_type_t crawl,
 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);
 
@@ -79,6 +85,33 @@ _loc_assign_gfid_path (loc_t *loc)
 }
 
 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;
@@ -123,6 +156,123 @@ _build_index_loc (xlator_t *this, loc_t *loc, char *name, loc_t *parent)
 }
 
 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)
 {
@@ -177,16 +327,18 @@ out:
 
 int
 _get_path_from_gfid_loc (xlator_t *this, xlator_t *readdir_xl, loc_t *child,
-                         char **fpath)
+                         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)
+        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 "
@@ -226,6 +378,20 @@ out:
 }
 
 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};
@@ -237,32 +403,24 @@ _add_eh_to_dict (xlator_t *this, eh_t *eh, dict_t *dict, int child)
         return 0;
 }
 
+
 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)
+_add_statistics_to_dict (xlator_t *this, dict_t *dict, int child)
 {
-        dict_t          *output = NULL;
-        xlator_t        *readdir_xl = NULL;
-        int             ret = -1;
-        char            *path = NULL;
-
-        if (uuid_is_null (childloc->gfid))
-                goto out;
+        shd_dump_t              dump_data = {0};
+        afr_private_t           *priv  = NULL;
+        afr_self_heald_t        *shd = NULL;
 
-        output = crawl_data->op_data;
-        readdir_xl = crawl_data->readdir_xl;
+        priv = this->private;
+        shd = &priv->shd;
 
-        ret = _get_path_from_gfid_loc (this, readdir_xl, childloc, &path);
-        if (ret)
-                goto out;
+        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;
 
-        ret = _add_path_to_dict (this, output, crawl_data->child, path, NULL,
-                                 _gf_true);
-out:
-        if (ret && path)
-                GF_FREE (path);
-        return ret;
 }
 
 void
@@ -289,30 +447,108 @@ 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;
-        shd_event_t      *event = NULL;
-        int32_t          sh_failed = 0;
-        gf_boolean_t     split_brain = 0;
+        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 (child->gfid));
+                                             parent, uuid_utoa_r (child->gfid,
+                                                                  gfid_str));
                         goto out;
                 }
                 ret = _get_path_from_gfid_loc (this, crawl_data->readdir_xl,
-                                               child, &path);
+                                               child, &path, NULL);
                 if (ret)
                         goto out;
         } else {
@@ -323,26 +559,45 @@ _crawl_post_sh_action (xlator_t *this, loc_t *parent, loc_t *child,
                 }
         }
 
-        if (xattr_rsp)
+        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)
+        if ((op_ret < 0 && op_errno == EIO) || split_brain) {
                 eh = shd->split_brain;
-        else if ((op_ret < 0) || sh_failed)
+                shd_crawl_event[crawl_data->child]->split_brain_count += 1;
+        } else if ((op_ret < 0) || sh_failed) {
                 eh = shd->heal_failed;
-        else
+                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;
-        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;
+
+        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:
@@ -352,21 +607,56 @@ out:
 }
 
 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, NULL,
+        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;
 }
 
@@ -398,11 +688,11 @@ _crawl_proceed (xlator_t *this, int child, int crawl_flags, char **reason)
         shd  = &priv->shd;
         if (!shd->enabled) {
                 msg = "Self-heal daemon is not enabled";
-                gf_log (this->name, GF_LOG_ERROR, "%s", msg);
+                gf_log (this->name, GF_LOG_DEBUG, "%s", msg);
                 goto out;
         }
         if (!priv->child_up[child]) {
-                gf_log (this->name, GF_LOG_ERROR, "Stopping crawl for %s , "
+                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;
@@ -411,7 +701,7 @@ _crawl_proceed (xlator_t *this, int child, int crawl_flags, char **reason)
         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_ERROR, "Stopping crawl as "
+                        gf_log (this->name, GF_LOG_DEBUG, "Stopping crawl as "
                                 "< 2 children are up");
                         msg = "< 2 bricks in replica are running";
                         goto out;
@@ -443,20 +733,28 @@ _do_crawl_op_on_local_subvols (xlator_t *this, afr_crawl_type_t crawl,
         if (op == HEAL)
                 crawl_flags |= STOP_CRAWL_ON_SINGLE_SUBVOL;
 
-        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;
+        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;
-                        ret = synctask_new (this->ctx->env,
-                                            afr_syncop_find_child_position,
-                                            NULL, NULL, &pos_data);
+                        /*
+                         * 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) {
@@ -467,22 +765,35 @@ _do_crawl_op_on_local_subvols (xlator_t *this, afr_crawl_type_t crawl,
                                         status = "Started self-heal";
                                         _do_self_heal_on_subvol (this, i,
                                                                  crawl);
-                                } else {
+                                } 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);
                                 }
                         }
-                        snprintf (key, sizeof (key), "%d-%d-%s", xl_id,
-                                  i, subkey);
-                        ret = dict_set_str (output, key, status);
+                        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;
                 }
-                snprintf (key, sizeof (key), "%d-%d-%s", xl_id, i, subkey);
-                ret = dict_set_str (output, key, status);
+                if (output) {
+                        snprintf (key, sizeof (key), "%d-%d-%s", xl_id, i,
+                                  subkey);
+                        ret = dict_set_str (output, key, status);
+                }
         }
 out:
         return op_ret;
@@ -501,8 +812,103 @@ _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_all_subvols_eh_to_dict (xlator_t *this, eh_t *eh, dict_t *dict)
+_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;
@@ -552,16 +958,25 @@ afr_xl_op (xlator_t *this, dict_t *input, dict_t *output)
                 ret = 0;
                 break;
         case GF_AFR_OP_HEALED_FILES:
-                ret = _add_all_subvols_eh_to_dict (this, shd->healed, output);
+                ret = _add_local_subvols_eh_to_dict (this, shd->healed, output);
                 break;
         case GF_AFR_OP_HEAL_FAILED_FILES:
-                ret = _add_all_subvols_eh_to_dict (this, shd->heal_failed,
+                ret = _add_local_subvols_eh_to_dict (this, shd->heal_failed,
                                                    output);
                 break;
         case GF_AFR_OP_SPLIT_BRAIN_FILES:
-                ret = _add_all_subvols_eh_to_dict (this, shd->split_brain,
+                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;
@@ -576,7 +991,7 @@ afr_poll_self_heal (void *data)
 {
         afr_private_t    *priv = NULL;
         afr_self_heald_t *shd = NULL;
-        struct timeval   timeout = {0};
+        struct timespec  timeout = {0};
         xlator_t         *this = NULL;
         long             child = (long)data;
         gf_timer_t       *old_timer = NULL;
@@ -600,7 +1015,7 @@ afr_poll_self_heal (void *data)
         if (shd->enabled && (shd->pos[child] == AFR_POS_LOCAL))
                 _do_self_heal_on_subvol (this, child, INDEX);
         timeout.tv_sec = shd->timeout;
-        timeout.tv_usec = 0;
+        timeout.tv_nsec = 0;
         //notify and previous timer should be synchronized.
         LOCK (&priv->lock);
         {
@@ -626,7 +1041,7 @@ unlock:
 }
 
 static int
-afr_local_child_poll_self_heal  (int ret, call_frame_t *sync_frame, void *data)
+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;
@@ -640,6 +1055,7 @@ afr_local_child_poll_self_heal  (int ret, call_frame_t *sync_frame, void *data)
         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;
@@ -663,7 +1079,7 @@ afr_proactive_self_heal (void *data)
         pos_data->this = this;
         pos_data->child = child;
         ret = synctask_new (this->ctx->env, afr_syncop_find_child_position,
-                            afr_local_child_poll_self_heal, NULL, pos_data);
+                            afr_handle_child_up, NULL, pos_data);
         if (ret)
                 goto out;
 out:
@@ -698,25 +1114,6 @@ out:
 }
 
 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
 afr_local_pathinfo (char *pathinfo, gf_boolean_t *local)
 {
         int             ret   = 0;
@@ -752,6 +1149,7 @@ afr_crawl_build_start_loc (xlator_t *this, afr_crawl_data_t *crawl_data,
         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};
@@ -761,7 +1159,7 @@ afr_crawl_build_start_loc (xlator_t *this, afr_crawl_data_t *crawl_data,
         priv = this->private;
         if (crawl_data->crawl == FULL) {
                 afr_build_root_loc (this, dirloc);
-        } else {
+        } else if (crawl_data->crawl == INDEX) {
                 afr_build_root_loc (this, &rootloc);
                 ret = syncop_getxattr (readdir_xl, &rootloc, &xattr,
                                        GF_XATTROP_INDEX_GFID);
@@ -795,6 +1193,47 @@ afr_crawl_build_start_loc (xlator_t *this, afr_crawl_data_t *crawl_data,
                 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:
@@ -859,6 +1298,16 @@ afr_crawl_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent,
         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)
@@ -908,13 +1357,14 @@ _process_entries (xlator_t *this, loc_t *parentloc, gf_dirent_t *entries,
                 ret = crawl_data->process_entry (this, crawl_data, entry,
                                                  &entry_loc, parentloc, &iattr);
 
-                if (ret)
+                if (crawl_data->crawl == INDEX_TO_BE_HEALED && ret) {
+                       goto out;
+                } else if (ret) {
                         continue;
+                }
 
-                ret = _link_inode_update_loc (this, &entry_loc, &iattr);
-                if (ret)
-                        goto out;
-                if (crawl_data->crawl == INDEX)
+                if ((crawl_data->crawl == INDEX) ||
+                    (crawl_data->crawl == INDEX_TO_BE_HEALED))
                         continue;
 
                 if (!IA_ISDIR (iattr.ia_type))
@@ -929,6 +1379,10 @@ _process_entries (xlator_t *this, loc_t *parentloc, gf_dirent_t *entries,
         }
         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;
 }
@@ -976,6 +1430,9 @@ _crawl_directory (fd_t *fd, loc_t *loc, afr_crawl_data_t *crawl_data)
 
                 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;
         }
@@ -1017,7 +1474,7 @@ afr_find_child_position (xlator_t *this, int child, afr_child_pos_t *pos)
 
         ret = syncop_getxattr (priv->children[child], &loc, &xattr_rsp,
                                GF_XATTR_NODE_UUID_KEY);
-        if (ret) {
+        if (ret < 0) {
                 gf_log (this->name, GF_LOG_ERROR, "getxattr failed on %s - "
                         "(%s)", priv->children[child]->name, strerror (errno));
                 goto out;
@@ -1081,8 +1538,13 @@ afr_dir_crawl (void *data)
                 goto out;
 
         ret = afr_crawl_opendir (this, crawl_data, &fd, &dirloc);
-        if (ret)
+        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)
@@ -1096,12 +1558,102 @@ afr_dir_crawl (void *data)
 out:
         if (fd)
                 fd_unref (fd);
-        if (crawl_data->crawl == INDEX)
+        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)
 {
@@ -1137,7 +1689,15 @@ afr_dir_exclusive_crawl (void *data)
         }
 
         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) {
@@ -1194,8 +1754,8 @@ afr_start_crawl (xlator_t *this, int idx, afr_crawl_type_t crawl,
         ret = synctask_new (this->ctx->env, crawler,
                             crawl_done, frame, crawl_data);
         if (ret)
-                gf_log (this->name, GF_LOG_ERROR, "Could not create the "
-                        "task for %d ret %d", idx, ret);
+                gf_log (this->name, GF_LOG_ERROR, "afr crawl failed for child"
+                        " %d with ret %d", idx, ret);
 out:
         return;
 }
@@ -1225,4 +1785,3 @@ afr_set_root_gfid (dict_t *dict)
 
         return ret;
 }
-
diff --git a/xlators/cluster/afr/src/afr-self-heald.h b/xlators/cluster/afr/src/afr-self-heald.h
index 32a8aaca5..e0c083754 100644
--- a/xlators/cluster/afr/src/afr-self-heald.h
+++ b/xlators/cluster/afr/src/afr-self-heald.h
@@ -29,6 +29,19 @@ typedef struct afr_crawl_data_ {
                               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);
diff --git a/xlators/cluster/afr/src/afr-transaction.c b/xlators/cluster/afr/src/afr-transaction.c
index ddf0af317..20306e469 100644
--- a/xlators/cluster/afr/src/afr-transaction.c
+++ b/xlators/cluster/afr/src/afr-transaction.c
@@ -79,7 +79,7 @@ afr_save_lk_owner (call_frame_t *frame)
 
         local = frame->local;
 
-	local->saved_lk_owner = frame->root->lk_owner;
+        local->saved_lk_owner = frame->root->lk_owner;
 }
 
 
@@ -90,10 +90,9 @@ afr_restore_lk_owner (call_frame_t *frame)
 
         local = frame->local;
 
-	frame->root->lk_owner = local->saved_lk_owner;
+        frame->root->lk_owner = local->saved_lk_owner;
 }
 
-
 static void
 __mark_all_pending (int32_t *pending[], int child_count,
                     afr_transaction_type type)
@@ -146,34 +145,6 @@ out:
         return;
 }
 
-
-static void
-__mark_pre_op_undone_on_fd (call_frame_t *frame, xlator_t *this, int child_index)
-{
-        afr_local_t   *local = NULL;
-        afr_fd_ctx_t  *fd_ctx = NULL;
-
-        local = frame->local;
-
-        if (!local->fd)
-                return;
-
-        fd_ctx = afr_fd_ctx_get (local->fd, this);
-
-        if (!fd_ctx)
-                goto out;
-
-        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_non_participant_children (int32_t *pending[], int child_count,
                                  unsigned char *participants,
@@ -190,7 +161,7 @@ __mark_non_participant_children (int32_t *pending[], int child_count,
 }
 
 
-static void
+void
 __mark_all_success (int32_t *pending[], int child_count,
                     afr_transaction_type type)
 {
@@ -203,6 +174,54 @@ __mark_all_success (int32_t *pending[], int child_count,
         }
 }
 
+void
+_set_all_child_errno (int *child_errno, unsigned int child_count)
+{
+        int     i = 0;
+
+        for (i = 0; i < child_count; i++)
+                if (child_errno[i] == 0)
+                        child_errno[i] = ENOTCONN;
+}
+
+void
+afr_transaction_perform_fop (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t     *local = NULL;
+        afr_private_t   *priv = NULL;
+        fd_t            *fd   = NULL;
+
+        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
 __changelog_enabled (afr_private_t *priv, afr_transaction_type type)
@@ -348,6 +367,11 @@ afr_changelog_post_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         UNLOCK (&frame->lock);
 
         if (call_count == 0) {
+                if (local->transaction.resume_stub) {
+			call_resume (local->transaction.resume_stub);
+                        local->transaction.resume_stub = NULL;
+                }
+
                 if (afr_lock_server_count (priv, local->transaction.type) == 0) {
                         local->transaction.done (frame, this);
                 } else {
@@ -418,19 +442,37 @@ out:
         return;
 }
 
+afr_inodelk_t*
+afr_get_inodelk (afr_internal_lock_t *int_lock, char *dom)
+{
+        afr_inodelk_t *inodelk = NULL;
+        int           i = 0;
+
+        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)
 {
         unsigned char *locked_nodes = NULL;
+        afr_inodelk_t *inodelk = NULL;
         switch (type) {
         case AFR_DATA_TRANSACTION:
         case AFR_METADATA_TRANSACTION:
-                locked_nodes = int_lock->inode_locked_nodes;
+                inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+                locked_nodes = inodelk->locked_nodes;
         break;
 
         case AFR_ENTRY_TRANSACTION:
         case AFR_ENTRY_RENAME_TRANSACTION:
-                locked_nodes = int_lock->entry_locked_nodes;
+                /*Because same set of subvols participate in all lockee
+                 * entities*/
+                locked_nodes = int_lock->lockee[0].locked_nodes;
         break;
         }
         return locked_nodes;
@@ -522,6 +564,82 @@ afr_set_postop_dict (afr_local_t *local, xlator_t *this, dict_t *xattr,
                         "failed to set pending entry");
 }
 
+
+gf_boolean_t
+afr_txn_nothing_failed (call_frame_t *frame, xlator_t *this)
+{
+        afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        int index = -1;
+        int i = 0;
+
+        local = frame->local;
+        priv = this->private;
+
+        index = afr_index_for_transaction_type (local->transaction.type);
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->pending[i][index] == 0)
+                        return _gf_false;
+        }
+
+        return _gf_true;
+}
+
+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;
+
+        this = frame->this;
+        local = frame->local;
+        priv = this->private;
+
+        if ((local->transaction.type != AFR_ENTRY_TRANSACTION) &&
+            (local->transaction.type != AFR_ENTRY_RENAME_TRANSACTION))
+                return;
+
+        if (local->op_ret >= 0)
+                goto out;
+
+        __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;
+
+        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)
 {
@@ -534,7 +652,6 @@ afr_changelog_post_op_now (call_frame_t *frame, xlator_t *this)
         afr_fd_ctx_t  *fdctx = NULL;
         dict_t        **xattr = NULL;
         int            piggyback = 0;
-        int            index = 0;
         int            nothing_failed = 1;
 
         local    = frame->local;
@@ -544,6 +661,9 @@ afr_changelog_post_op_now (call_frame_t *frame, xlator_t *this)
                                          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,
@@ -570,15 +690,7 @@ afr_changelog_post_op_now (call_frame_t *frame, xlator_t *this)
                 goto out;
         }
 
-        /* check if something has failed, to handle piggybacking */
-        nothing_failed = 1;
-        index = afr_index_for_transaction_type (local->transaction.type);
-        for (i = 0; i < priv->child_count; i++) {
-                if (local->pending[i][index] == 0) {
-                        nothing_failed = 0;
-                        break;
-                }
-        }
+        nothing_failed = afr_txn_nothing_failed (frame, this);
 
         afr_compute_txn_changelog (local , priv);
 
@@ -604,15 +716,14 @@ afr_changelog_post_op_now (call_frame_t *frame, xlator_t *this)
                                 break;
                         }
 
-                        LOCK (&local->fd->lock);
-                        {
-                                piggyback = 0;
-                                if (fdctx->pre_op_piggyback[i]) {
-                                        fdctx->pre_op_piggyback[i]--;
-                                        piggyback = 1;
-                                }
-                        }
-                        UNLOCK (&local->fd->lock);
+                        /* 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);
@@ -621,7 +732,6 @@ afr_changelog_post_op_now (call_frame_t *frame, xlator_t *this)
                                 afr_changelog_post_op_cbk (frame, (void *)(long)i,
                                                            this, 1, 0, xattr[i], NULL);
                         } else {
-                                __mark_pre_op_undone_on_fd (frame, this, i);
                                 STACK_WIND_COOKIE (frame,
                                                    afr_changelog_post_op_cbk,
                                                    (void *) (long) i,
@@ -778,18 +888,7 @@ afr_changelog_pre_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     (local->op_errno == ENOTSUP)) {
                         local->transaction.resume (frame, this);
                 } else {
-                        __mark_all_success (local->pending, priv->child_count,
-                                            local->transaction.type);
-
-			/*  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;
-
-                        local->transaction.fop (frame, this);
+                        afr_transaction_perform_fop (frame, this);
                 }
         }
 
@@ -878,7 +977,7 @@ afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
                         }
                         UNLOCK (&local->fd->lock);
 
-			afr_set_delayed_post_op (frame, this);
+                        afr_set_delayed_post_op (frame, this);
 
                         if (piggyback)
                                 afr_changelog_pre_op_cbk (frame, (void *)(long)i,
@@ -1151,12 +1250,14 @@ 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;
 }
@@ -1171,6 +1272,7 @@ 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:
@@ -1184,8 +1286,8 @@ afr_lock_rec (call_frame_t *frame, xlator_t *this)
 
         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:
@@ -1218,28 +1320,10 @@ 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;
-
-        priv  = this->private;
-        local = frame->local;
-
         if (__fop_changelog_needed (frame, this)) {
                 afr_changelog_pre_op (frame, this);
         } else {
-                __mark_all_success (local->pending, priv->child_count,
-                                    local->transaction.type);
-
-
-		/*  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;
-
-                local->transaction.fop (frame, this);
+                afr_transaction_perform_fop (frame, this);
         }
 
         return 0;
@@ -1249,87 +1333,384 @@ afr_internal_lock_finish (call_frame_t *frame, xlator_t *this)
 void
 afr_set_delayed_post_op (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;
 
-	/* call this function from any of the related optimizations
-	   which benefit from delaying post op are enabled, namely:
+        /* call this function from any of the related optimizations
+           which benefit from delaying post op are enabled, namely:
 
-	   - changelog piggybacking
-	   - eager locking
-	*/
+           - changelog piggybacking
+           - eager locking
+        */
 
-	priv = this->private;
-	if (!priv)
-		return;
+        priv = this->private;
+        if (!priv)
+                return;
 
-	if (!priv->post_op_delay_secs)
-		return;
+        if (!priv->post_op_delay_secs)
+                return;
 
-	local = frame->local;
-	if (!local)
-		return;
+        local = frame->local;
+        if (!local->transaction.eager_lock_on)
+                return;
+
+        if (!local)
+                return;
+
+        if (!local->fd)
+                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 (local->op == GF_FOP_WRITE)
-		local->delayed_post_op = _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_local_t      *local = NULL;
+        gf_boolean_t      res = _gf_false;
+        afr_private_t    *priv  = NULL;
 
-	local = frame->local;
-	if (!local)
-		goto out;
+        priv  = this->private;
+
+        local = frame->local;
+        if (!local)
+                goto out;
 
-	if (!local->delayed_post_op)
-		goto out;
+        if (!local->delayed_post_op)
+                goto out;
 
-	res = _gf_true;
+        //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;
+        return res;
 }
 
 
 void
-afr_delayed_changelog_post_op (xlator_t *this, call_frame_t *frame, fd_t *fd);
-
-void
-afr_delayed_changelog_wake_up (xlator_t *this, fd_t *fd);
+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_t           *fd = NULL;
 
-	fd = data;
+        fd = data;
 
-	afr_delayed_changelog_wake_up (THIS, fd);
+        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;
+
+        fdctx = afr_fd_ctx_get (fd, this);
+
+        LOCK(&fd->lock);
+        {
+                fdctx->witnessed_unstable_write = _gf_true;
+        }
+        UNLOCK(&fd->lock);
+
+        return 0;
+}
+
+/* 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)
+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 timeval     delta = {0, };
+	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)
-		return;
+                goto out;
 
 	delta.tv_sec = priv->post_op_delay_secs;
-	delta.tv_usec = 0;
+	delta.tv_nsec = 0;
 
 	pthread_mutex_lock (&fd_ctx->delay_lock);
 	{
@@ -1348,8 +1729,13 @@ afr_delayed_changelog_post_op (xlator_t *this, call_frame_t *frame, fd_t *fd)
 unlock:
 	pthread_mutex_unlock (&fd_ctx->delay_lock);
 
+out:
 	if (prev_frame) {
-		afr_changelog_post_op_now (prev_frame, this);
+		local = prev_frame->local;
+		local->transaction.resume_stub = stub;
+		afr_changelog_post_op_safe (prev_frame, this);
+	} else if (stub) {
+		call_resume (stub);
 	}
 }
 
@@ -1357,49 +1743,62 @@ unlock:
 void
 afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t  *local = NULL;
+        afr_local_t  *local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	if (is_afr_delayed_changelog_post_op_needed (frame, this))
-		afr_delayed_changelog_post_op (this, frame, local->fd);
-	else
-		afr_changelog_post_op_now (frame, this);
+        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);
+        afr_delayed_changelog_post_op (this, NULL, fd, NULL);
 }
 
 
-int
+        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;
-	fd_t                *fd = NULL;
 
         local    = frame->local;
         int_lock = &local->internal_lock;
         priv     = this->private;
-	fd       = local->fd;
 
-	if (fd)
-		/* 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.
-		*/
-		afr_delayed_changelog_wake_up (this, fd);
+        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);
+        afr_restore_lk_owner (frame);
 
         if (__fop_changelog_needed (frame, this)) {
                 afr_changelog_post_op (frame, this);
@@ -1421,7 +1820,8 @@ 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;
@@ -1430,34 +1830,134 @@ afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this, int child_index
         priv  = this->private;
 
         __mark_child_dead (local->pending, priv->child_count,
-                           child_index, local->transaction.type);
+                        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);
+}
+
+
 int
 afr_transaction (call_frame_t *frame, xlator_t *this, afr_transaction_type type)
 {
         afr_local_t *   local = NULL;
         afr_private_t * priv  = NULL;
+        fd_t            *fd   = NULL;
+        int             ret   = -1;
 
         local = frame->local;
         priv  = this->private;
 
-	if (local->fd && priv->eager_lock &&
-	    local->transaction.type == AFR_DATA_TRANSACTION)
-		afr_set_lk_owner (frame, this, local->fd);
-	else
-		afr_set_lk_owner (frame, this, frame->root);
-
-        afr_transaction_local_init (local, this);
-
         local->transaction.resume = afr_transaction_resume;
         local->transaction.type   = type;
 
+        ret = afr_transaction_local_init (local, this);
+        if (ret < 0)
+            goto out;
+
+        afr_transaction_eager_lock_init (local, this);
+
+        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);
+
+        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);
+                }
+        }
+
         if (afr_lock_server_count (priv, local->transaction.type) == 0) {
                 afr_internal_lock_finish (frame, this);
         } else {
                 afr_lock (frame, this);
         }
-
-        return 0;
+        ret = 0;
+out:
+        return ret;
 }
diff --git a/xlators/cluster/afr/src/afr-transaction.h b/xlators/cluster/afr/src/afr-transaction.h
index 102599633..fa626fd0d 100644
--- a/xlators/cluster/afr/src/afr-transaction.h
+++ b/xlators/cluster/afr/src/afr-transaction.h
@@ -23,6 +23,9 @@ 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);
 
@@ -33,4 +36,16 @@ 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 77f00ef72..c724eb2ae 100644
--- a/xlators/cluster/afr/src/afr.c
+++ b/xlators/cluster/afr/src/afr.c
@@ -186,7 +186,13 @@ reconfigure (xlator_t *this, dict_t *options)
 	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;
@@ -325,9 +331,14 @@ init (xlator_t *this)
         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;
 
@@ -369,8 +380,6 @@ init (xlator_t *this)
                                    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;
                 }
@@ -379,6 +388,13 @@ init (xlator_t *this)
                 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) {
@@ -423,15 +439,18 @@ init (xlator_t *this)
         if (!priv->shd.timer)
                 goto out;
 
-        priv->shd.healed = eh_new (AFR_EH_HEALED_LIMIT, _gf_false);
+        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);
+        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);
+        priv->shd.split_brain = eh_new (AFR_EH_SPLIT_BRAIN_LIMIT, _gf_false,
+                                        _destroy_shd_event_data);
         if (!priv->shd.split_brain)
                 goto out;
 
@@ -441,7 +460,9 @@ init (xlator_t *this)
         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;
@@ -476,6 +497,9 @@ struct xlator_fops fops = {
         .finodelk    = afr_finodelk,
         .entrylk     = afr_entrylk,
         .fentrylk    = afr_fentrylk,
+	.fallocate   = afr_fallocate,
+	.discard     = afr_discard,
+        .zerofill    = afr_zerofill,
 
         /* inode read */
         .access      = afr_access,
@@ -528,35 +552,53 @@ struct xlator_cbks cbks = {
 
 struct volume_options options[] = {
         { .key  = {"read-subvolume" },
-          .type = GF_OPTION_TYPE_XLATOR
+          .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 = "0 = first responder, "
-                         "1 = hash by GFID (all clients use same subvolume), "
-                         "2 = hash by GFID and client PID",
+          .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 to read from if "
+          .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
+          .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,
           .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,
@@ -564,18 +606,25 @@ struct volume_options options[] = {
                     "0", "off", "no", "false", "disable",
                     "open"},
           .default_value = "on",
-          .description   = "\"open\" means data self-heal action will"
-                           "only be triggered by file open operations."
+          .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-algorithm"},
           .type = GF_OPTION_TYPE_STR,
-          .default_value = "",
           .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.",
-          .value = { "diff", "full", "" }
+                           "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,
@@ -588,26 +637,43 @@ struct volume_options options[] = {
         { .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,
@@ -616,26 +682,55 @@ struct volume_options options[] = {
         { .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", "" },
+          .value = { "none", "auto", "fixed"},
           .default_value = "none",
           .description = "If value is \"fixed\" only allow writes if "
                          "quorum-count bricks are present.  If value is "
@@ -654,14 +749,17 @@ struct volume_options options[] = {
         },
         { .key  = {"node-uuid"},
           .type = GF_OPTION_TYPE_STR,
-          .description = "Local glusterd uuid string",
+          .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 = "Poll timeout for checking the need to self-heal"
+          .description = "time interval for checking the need to self-heal "
+                         "in self-heal-daemon"
         },
         { .key  = {"post-op-delay-secs"},
           .type = GF_OPTION_TYPE_INT,
@@ -672,5 +770,24 @@ struct volume_options options[] = {
 	                 "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  = {NULL} },
 };
diff --git a/xlators/cluster/afr/src/afr.h b/xlators/cluster/afr/src/afr.h
index 9ae154a3d..21064db58 100644
--- a/xlators/cluster/afr/src/afr.h
+++ b/xlators/cluster/afr/src/afr.h
@@ -27,6 +27,11 @@
 
 #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;
 
@@ -57,10 +62,8 @@ typedef enum {
         AFR_INODE_SET_READ_CTX = 1,
         AFR_INODE_RM_STALE_CHILDREN,
         AFR_INODE_SET_OPENDIR_DONE,
-        AFR_INODE_SET_SPLIT_BRAIN,
         AFR_INODE_GET_READ_CTX,
         AFR_INODE_GET_OPENDIR_DONE,
-        AFR_INODE_GET_SPLIT_BRAIN,
 } afr_inode_op_t;
 
 typedef struct afr_inode_params_ {
@@ -74,29 +77,41 @@ typedef struct afr_inode_params_ {
         } 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;
-        char             *node_uuid;
-        int              timeout;
+        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 {
@@ -159,9 +174,40 @@ typedef struct _afr_private {
         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 */
 
@@ -238,10 +284,10 @@ typedef struct {
         const char *linkname;
         gf_boolean_t entries_skipped;
 
-        int   op_failed;
-
+        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;
@@ -252,27 +298,32 @@ typedef struct {
         uint8_t *checksum;
         afr_post_remove_call_t post_remove_call;
 
-        loc_t parent_loc;
+        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;
 
         afr_sh_algo_private_t *private;
+        afr_sh_status_for_all_type  afr_all_sh_status;
+        afr_self_heal_type       sh_type_in_action;
 
         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);
-        gf_boolean_t    mdata_spb;
-        gf_boolean_t    data_spb;
 
         call_frame_t *sh_frame;
-} afr_self_heal_t;
+};
+
+typedef struct afr_self_heal_ afr_self_heal_t;
 
 typedef enum {
         AFR_DATA_TRANSACTION,          /* truncate, write, ... */
@@ -334,11 +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;
+
+} 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 gf_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;
@@ -347,23 +418,22 @@ 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;
 
 typedef struct _afr_locked_fd {
@@ -371,15 +441,23 @@ typedef struct _afr_locked_fd {
         struct list_head list;
 } afr_locked_fd_t;
 
+struct afr_reply {
+	int	valid;
+	int32_t	op_ret;
+	int32_t	op_errno;
+};
+
 typedef struct _afr_local {
         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 char read_child_returned;
@@ -396,7 +474,6 @@ typedef struct _afr_local {
         loc_t newloc;
 
         fd_t *fd;
-        unsigned char *fd_open_on;
 
         glusterfs_fop_t fop;
 
@@ -420,12 +497,23 @@ typedef struct _afr_local {
         int      optimistic_change_log;
 	gf_boolean_t      delayed_post_op;
 
-        gf_boolean_t    fop_paused;
-        int (*fop_call_continue) (call_frame_t *frame, xlator_t *this);
 
-        /*
-          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;
+
+        /* 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;
+
+        int allow_sh_for_running_transaction;
+
+
+        /* This struct contains the arguments for the "continuation"
+           (scheme-like) of fops
         */
 
         int   op;
@@ -521,7 +609,9 @@ typedef struct _afr_local {
                 struct {
                         struct iatt prebuf;
                         struct iatt postbuf;
+                } inode_wfop; //common structure for all inode-write-fops
 
+                struct {
                         int32_t op_ret;
 
                         struct iovec *vector;
@@ -532,34 +622,21 @@ typedef struct _afr_local {
                 } writev;
 
                 struct {
-                        struct iatt prebuf;
-                        struct iatt postbuf;
-                } fsync;
-
-                struct {
                         off_t offset;
-                        struct iatt prebuf;
-                        struct iatt postbuf;
                 } truncate;
 
                 struct {
                         off_t offset;
-                        struct iatt prebuf;
-                        struct iatt postbuf;
                 } ftruncate;
 
                 struct {
                         struct iatt in_buf;
                         int32_t valid;
-                        struct iatt preop_buf;
-                        struct iatt postop_buf;
                 } setattr;
 
                 struct {
                         struct iatt in_buf;
                         int32_t valid;
-                        struct iatt preop_buf;
-                        struct iatt postop_buf;
                 } fsetattr;
 
                 struct {
@@ -621,11 +698,32 @@ typedef struct _afr_local {
                         dict_t *params;
                         char *linkpath;
                 } symlink;
+
+		struct {
+			int32_t mode;
+			off_t offset;
+			size_t len;
+		} fallocate;
+
+		struct {
+			off_t offset;
+			size_t len;
+		} discard;
+
+                struct {
+                        off_t offset;
+                        size_t len;
+                        struct iatt prebuf;
+                        struct iatt postbuf;
+                } zerofill;
+
+
         } cont;
 
         struct {
                 off_t start, len;
 
+                gf_boolean_t    eager_lock_on;
                 int *eager_lock;
 
                 char *basename;
@@ -636,6 +734,17 @@ typedef struct _afr_local {
 
                 afr_transaction_type type;
 
+		/* pre-compute the post piggyback status before
+		   entering POST-OP phase
+		*/
+		int              *postop_piggybacked;
+
+		/* stub to resume on destruction
+		   of the transaction frame */
+		call_stub_t      *resume_stub;
+
+		struct list_head  eager_locked;
+
                 int32_t         **txn_changelog;//changelog after pre+post ops
                 unsigned char   *pre_op;
 
@@ -663,6 +772,7 @@ typedef struct _afr_local {
         mode_t          umask;
         int             xflag;
         gf_boolean_t    do_discovery;
+	struct afr_reply *replies;
 } afr_local_t;
 
 typedef enum {
@@ -672,11 +782,6 @@ typedef enum {
 } afr_fd_open_status_t;
 
 typedef struct {
-        struct list_head call_list;
-        call_frame_t    *frame;
-} afr_fd_paused_call_t;
-
-typedef struct {
         unsigned int *pre_op_done;
         afr_fd_open_status_t *opened_on; /* which subvolumes the fd is open on */
         unsigned int *pre_op_piggyback;
@@ -695,12 +800,20 @@ typedef struct {
         struct list_head entries; /* needed for readdir failover */
 
         unsigned char *locked_on; /* which subvolumes locks have been successful */
-	struct list_head  paused_calls; /* queued calls while fix_open happens  */
 
 	/* 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;
 
 
@@ -736,6 +849,13 @@ int32_t
 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_entry_lockee_cleanup (afr_internal_lock_t *int_lock);
+
+int
 afr_attempt_lock_recovery (xlator_t *this, int32_t child_index);
 
 int
@@ -770,8 +890,8 @@ afr_blocking_lock (call_frame_t *frame, xlator_t *this);
 int
 afr_internal_lock_finish (call_frame_t *frame, xlator_t *this);
 
-void
-afr_lk_transfer_datalock (call_frame_t *dst, call_frame_t *src,
+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);
@@ -821,7 +941,8 @@ 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,
@@ -936,8 +1057,9 @@ 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);
-gf_boolean_t
-afr_error_more_important (int32_t old_errno, int32_t new_errno);
+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);
@@ -980,11 +1102,11 @@ afr_launch_self_heal (call_frame_t *frame, xlator_t *this, inode_t *inode,
                       int (*unwind) (call_frame_t *frame, xlator_t *this,
                                      int32_t op_ret, int32_t op_errno,
                                      int32_t sh_failed));
-int
-afr_fix_open (call_frame_t *frame, xlator_t *this, afr_fd_ctx_t *fd_ctx,
-              int need_open_count, int *need_open);
-int
-afr_open_fd_fix (call_frame_t *frame, xlator_t *this, gf_boolean_t pause_fop);
+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);
 
@@ -1011,6 +1133,23 @@ afr_matrix_cleanup (int32_t **pending, unsigned int m);
 
 int32_t**
 afr_matrix_create (unsigned int m, unsigned int n);
+
+gf_boolean_t
+afr_is_errno_set (int *child_errno, int child);
+
+gf_boolean_t
+afr_is_errno_unset (int *child_errno, int child);
+
+gf_boolean_t
+afr_is_fd_fixable (fd_t *fd);
+
+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).
@@ -1030,4 +1169,44 @@ afr_matrix_create (unsigned int m, unsigned int n);
         }                                                                \
 } 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 e42c955d0..a7f72fb30 100644
--- a/xlators/cluster/afr/src/pump.c
+++ b/xlators/cluster/afr/src/pump.c
@@ -942,6 +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;
@@ -970,12 +971,15 @@ 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 ();
 
@@ -2433,14 +2437,13 @@ init (xlator_t *this)
 	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 */
 
@@ -2498,6 +2501,12 @@ 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;
 
@@ -2529,7 +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");
@@ -2570,9 +2579,6 @@ fini (xlator_t *this)
         if (!pump_priv)
                 goto afr_priv;
 
-        if (pump_priv->env)
-                syncenv_destroy (pump_priv->env);
-
         GF_FREE (pump_priv->resume_path);
         LOCK_DESTROY (&pump_priv->resume_path_lock);
         LOCK_DESTROY (&pump_priv->pump_state_lock);
diff --git a/xlators/cluster/dht/src/Makefile.am b/xlators/cluster/dht/src/Makefile.am
index e35058d65..174bea841 100644
--- a/xlators/cluster/dht/src/Makefile.am
+++ b/xlators/cluster/dht/src/Makefile.am
@@ -4,7 +4,7 @@ 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-common.c dht-inode-write.c dht-inode-read.c dht-shared.c \
 	$(top_builddir)/xlators/lib/src/libxlator.c
 
 dht_la_SOURCES = $(dht_common_source) dht.c
@@ -12,22 +12,23 @@ 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-mem-types.h \
 	$(top_builddir)/xlators/lib/src/libxlator.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/lib/src
 
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
 CLEANFILES =
 
 uninstall-local:
diff --git a/xlators/cluster/dht/src/dht-common.c b/xlators/cluster/dht/src/dht-common.c
index 24f28c1fe..8f61339e6 100644
--- a/xlators/cluster/dht/src/dht-common.c
+++ b/xlators/cluster/dht/src/dht-common.c
@@ -22,6 +22,7 @@
 #include "dht-common.h"
 #include "defaults.h"
 #include "byte-order.h"
+#include "glusterfs-acl.h"
 
 #include <sys/time.h>
 #include <libgen.h>
@@ -62,6 +63,11 @@ dht_aggregate (dict_t *this, char *key, data_t *value, void *data)
                 }
 
                 *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."))) {
@@ -125,7 +131,10 @@ dht_lookup_selfheal_cbk (call_frame_t *frame, void *cookie,
                 ret = dht_layout_set (this, local->inode, layout);
         }
 
-        WIPE (&local->postparent);
+        if (local->loc.parent) {
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           &local->postparent, 1);
+        }
 
         DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
 
@@ -145,9 +154,11 @@ dht_discover_complete (xlator_t *this, call_frame_t *discover_frame)
         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);
         {
@@ -190,11 +201,14 @@ dht_discover_complete (xlator_t *this, call_frame_t *discover_frame)
                                 "(overlaps/holes present: %s, "
                                 "ENOENT errors: %d)", local->loc.path,
                                 (ret < 0) ? "yes" : "no", (ret > 0) ? ret : 0);
-                        op_errno = EINVAL;
-                        goto out;
+                        if ((ret > 0) && (ret == conf->subvolume_cnt)) {
+                                op_errno = ESTALE;
+                                goto out;
+                        }
                 }
 
-                dht_layout_set (this, local->inode, layout);
+                if (local->inode)
+                        dht_layout_set (this, local->inode, layout);
         }
 
         DHT_STACK_UNWIND (lookup, main_frame, local->op_ret, local->op_errno,
@@ -223,6 +237,7 @@ dht_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         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);
@@ -232,6 +247,7 @@ dht_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         local = frame->local;
         prev  = cookie;
+        conf = this->private;
 
         layout = local->layout;
 
@@ -266,7 +282,8 @@ dht_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         goto unlock;
                 }
 
-                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) {
@@ -325,23 +342,20 @@ dht_discover (call_frame_t *frame, xlator_t *this, loc_t *loc)
         int          i = 0;
         call_frame_t *discover_frame = NULL;
 
-
         conf = this->private;
         local = frame->local;
 
-        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)
                 gf_log (this->name, GF_LOG_WARNING,
-                        "%s: failed to set 'trusted.glusterfs.dht' key",
-                        loc->path);
+                        "%s: failed to set '%s' key",
+                        loc->path, conf->xattr_name);
 
-        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)
                 gf_log (this->name, GF_LOG_WARNING,
-                        "%s: failed to set 'trusted.glusterfs.dht.linkto' key",
-                        loc->path);
+                        "%s: failed to set '%s' key",
+                        loc->path, conf->link_xattr_name);
 
         call_cnt        = conf->subvolume_cnt;
         local->call_cnt = call_cnt;
@@ -427,7 +441,7 @@ dht_lookup_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                         op_ret, op_errno, xattr);
 
                 if (op_ret == -1) {
-                        local->op_errno = ENOENT;
+                        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,
@@ -486,6 +500,11 @@ unlock:
                         dht_layout_set (this, local->inode, layout);
                 }
 
+                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,
@@ -517,6 +536,8 @@ dht_revalidate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         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);
@@ -575,7 +596,8 @@ dht_revalidate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 layout = local->layout;
 
                 is_dir = check_is_dir (inode, stbuf, xattr);
-                is_linkfile = check_is_linkfile (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,
@@ -587,6 +609,23 @@ dht_revalidate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 }
 
                 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);
@@ -625,7 +664,28 @@ out:
                     && (conf && conf->unhashed_sticky_bit)) {
                         local->stbuf.ia_prot.sticky = 1;
                 }
-                if (local->layout_mismatch) {
+                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) {
                         /* Found layout mismatch in the directory, need to
                            fix this in the inode context */
                         dht_layout_unref (this, local->layout);
@@ -651,7 +711,10 @@ out:
                         local->op_errno = ESTALE;
                 }
 
-                WIPE (&local->postparent);
+                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,
@@ -703,8 +766,14 @@ dht_lookup_linkfile_create_cbk (call_frame_t *frame, void *cookie,
                 local->stbuf.ia_prot.sticky = 1;
         }
 
+        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_STRIP_PHASE1_FLAGS (&local->stbuf);
         DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
@@ -777,7 +846,10 @@ dht_lookup_everywhere_done (call_frame_t *frame, xlator_t *this)
                                                   "<nil>"));
                 }
 
-                WIPE (&local->postparent);
+                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,
@@ -807,7 +879,10 @@ dht_lookup_everywhere_done (call_frame_t *frame, xlator_t *this)
                         local->op_errno = EINVAL;
                 }
 
-                WIPE (&local->postparent);
+                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,
@@ -823,7 +898,7 @@ dht_lookup_everywhere_done (call_frame_t *frame, xlator_t *this)
                 hashed_subvol->name);
 
         ret = dht_linkfile_create (frame,
-                                   dht_lookup_linkfile_create_cbk,
+                                   dht_lookup_linkfile_create_cbk, this,
                                    cached_subvol, hashed_subvol, &local->loc);
 
         return ret;
@@ -861,8 +936,9 @@ dht_lookup_everywhere_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         xlator_t     *subvol        = NULL;
         loc_t        *loc           = NULL;
         xlator_t     *link_subvol   = NULL;
-        int           ret = -1;
-        int32_t       fd_count = 0;
+        int           ret           = -1;
+        int32_t       fd_count      = 0;
+        dht_conf_t   *conf          = NULL;
 
         GF_VALIDATE_OR_GOTO ("dht", frame, out);
         GF_VALIDATE_OR_GOTO ("dht", this, out);
@@ -872,6 +948,7 @@ dht_lookup_everywhere_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         local  = frame->local;
         loc    = &local->loc;
+        conf   = this->private;
 
         prev   = cookie;
         subvol = prev->this;
@@ -893,7 +970,8 @@ dht_lookup_everywhere_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                 loc->path, prev->this->name);
                 }
 
-                is_linkfile = check_is_linkfile (inode, buf, xattr);
+                is_linkfile = check_is_linkfile (inode, buf, xattr,
+                                                 conf->link_xattr_name);
                 is_dir = check_is_dir (inode, buf, xattr);
 
                 if (is_linkfile) {
@@ -1035,7 +1113,16 @@ dht_lookup_linkfile_cbk (call_frame_t *frame, void *cookie,
                 gf_log (this->name, GF_LOG_INFO,
                         "lookup of %s on %s (following linkfile) failed (%s)",
                         local->loc.path, subvol->name, strerror (op_errno));
-                goto err;
+
+                /* 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)) {
@@ -1045,7 +1132,7 @@ dht_lookup_linkfile_cbk (call_frame_t *frame, void *cookie,
                 goto err;
         }
 
-        if (check_is_linkfile (inode, stbuf, xattr)) {
+        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);
@@ -1073,9 +1160,12 @@ dht_lookup_linkfile_cbk (call_frame_t *frame, void *cookie,
                 op_errno = EINVAL;
         }
 
-unwind:
-        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);
@@ -1157,7 +1247,6 @@ 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);
@@ -1188,8 +1277,10 @@ dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 }
                 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);
@@ -1218,7 +1309,8 @@ dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 goto out;
         }
 
-        is_linkfile = check_is_linkfile (inode, stbuf, xattr);
+        is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+                                         conf->link_xattr_name);
 
         if (!is_linkfile) {
                 /* non-directory and not a linkfile */
@@ -1258,7 +1350,10 @@ out:
          * 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,
@@ -1320,7 +1415,6 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
         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;
         if (!conf)
@@ -1364,16 +1458,8 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
                 return 0;
         }
 
-        if (!hashed_subvol) {
+        if (!hashed_subvol)
                 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;
-                }
-        }
         local->hashed_subvol = hashed_subvol;
 
         if (is_revalidate (loc)) {
@@ -1403,7 +1489,7 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
                  *       revalidates directly go to the cached-subvolume.
                  */
                 ret = dict_set_uint32 (local->xattr_req,
-                                       "trusted.glusterfs.dht", 4 * 4);
+                                       conf->xattr_name, 4 * 4);
 
                 if (IA_ISDIR (local->inode->ia_type)) {
                         local->call_cnt = call_cnt = conf->subvolume_cnt;
@@ -1438,10 +1524,10 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
         do_fresh_lookup:
                 /* TODO: remove the hard-coding */
                 ret = dict_set_uint32 (local->xattr_req,
-                                       "trusted.glusterfs.dht", 4 * 4);
+                                       conf->xattr_name, 4 * 4);
 
                 ret = dict_set_uint32 (local->xattr_req,
-                                       DHT_LINKFILE_KEY, 256);
+                                       conf->link_xattr_name, 256);
 
                 /* need it for self-healing linkfiles which is
                    'in-migration' state */
@@ -1517,8 +1603,12 @@ dht_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 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);
@@ -1545,7 +1635,8 @@ dht_unlink_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         LOCK (&frame->lock);
         {
-                if (op_ret == -1) {
+                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)",
@@ -1558,7 +1649,7 @@ dht_unlink_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 unlock:
         UNLOCK (&frame->lock);
 
-        if (op_ret == -1)
+        if (local->op_ret == -1)
                 goto err;
 
         cached_subvol = dht_subvol_get_cached (this, local->loc.inode);
@@ -1582,41 +1673,6 @@ err:
         return 0;
 }
 
-static int
-dht_ufo_xattr_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;
-
-        local = frame->local;
-        prev = cookie;
-
-        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;
-                }
-        }
-unlock:
-        UNLOCK (&frame->lock);
-
-        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);
-        }
-
-        return 0;
-}
-
-
 int
 dht_err_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
              int op_ret, int op_errno, dict_t *xdata)
@@ -1689,115 +1745,222 @@ dht_fill_pathinfo_xattr (xlator_t *this, dht_local_t *local,
 }
 
 int
-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_vgetxattr_alloc_and_fill (dht_local_t *local, dict_t *xattr, xlator_t *this,
+                              int op_errno)
 {
-        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     = NULL;
-        dict_t      *dict          = NULL;
-        int32_t      alloc_len     = 0;
-        int32_t      plen          = 0;
-        call_frame_t *prev         = NULL;
+        int      ret       = -1;
+        char    *value     = NULL;
+        int32_t  plen      = 0;
 
-        local = frame->local;
-        prev = cookie;
+        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;
+        }
 
-        if (op_ret >= 0) {
-                ret = dict_get_str (xattr, local->xsel, &value_got);
-                if (!ret) {
-                        alloc_len = strlen (value_got);
+        local->alloc_len += strlen(value);
 
-                        /**
-                         * allocate the buffer:- we allocate 10 bytes extra in
-                         * case we need to append ' Link: ' in the buffer for
-                         * another STACK_WIND
-                         */
+        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;
+                }
+        }
+
+        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) {
-                                alloc_len += (strlen (DHT_PATHINFO_HEADER) + 10);
-                                local->xattr_val =
-                                        GF_CALLOC (alloc_len,
-                                                   sizeof (char),
-                                                   gf_common_mt_char);
+                                ret = -1;
+                                goto out;
                         }
+                }
 
-                        if (local->xattr_val) {
-                                plen = strlen (local->xattr_val);
-                                if (plen) {
-                                        void *p;
-                                        /* extra byte(s) for \0 to be safe */
-                                        alloc_len += (plen + 2);
-                                        p = GF_REALLOC (local->xattr_val,
-                                                        alloc_len);
-                                        if (!p)
-                                                goto out;
-                                        local->xattr_val = p;
-                                }
+                (void) strcat (local->xattr_val, value);
+                (void) strcat (local->xattr_val, " ");
+                local->op_ret = 0;
+        }
 
-                                strcat (local->xattr_val, value_got);
-                        }
-                        local->op_ret = 0;
-                }
+        ret = 0;
+
+ out:
+        return ret;
+}
+
+int
+dht_vgetxattr_fill_and_set (dht_local_t *local, dict_t **dict, xlator_t *this,
+                            gf_boolean_t flag)
+{
+        int   ret             = -1;
+        char *xattr_buf       = NULL;
+        char layout_buf[8192] = {0,};
+
+        if (flag)
+                fill_layout_info (local->layout, layout_buf);
+
+        *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 {
-                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));
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Unknown local->xsel (%s)", local->xsel);
+                goto out;
         }
 
+        ret = dict_set_dynstr (*dict, local->xsel, xattr_buf);
+        GF_FREE (local->xattr_val);
+
  out:
-        this_call_cnt = dht_frame_return (frame);
-        if (is_last_call (this_call_cnt)) {
+        return ret;
+}
 
-                if (local->op_ret == -1) {
-                        goto unwind;
-                }
+int
+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)
+{
+        int          ret           = 0;
+        dht_local_t *local         = NULL;
+        int          this_call_cnt = 0;
+        dict_t      *dict          = NULL;
 
-                if (local->layout->cnt > 1) {
-                        /* Set it for directory */
-                        fill_layout_info (local->layout, layout_buf);
-                        flag = 1;
-                }
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (frame->local, out);
 
-                dict = dict_new ();
-
-                /* we would need max this many bytes to create xattr string */
-                alloc_len += (2 * strlen (this->name))
-                          + strlen (layout_buf)
-                          + 40;
-                xattr_buf = GF_CALLOC (alloc_len,
-                                       sizeof (char), gf_common_mt_char);
-
-                if (XATTR_IS_PATHINFO (local->xsel)) {
-                        (void) dht_fill_pathinfo_xattr (this, local, xattr_buf,
-                                                        alloc_len, flag,
-                                                        layout_buf);
-                } else if (XATTR_IS_NODE_UUID (local->xsel)) {
-                        (void) snprintf (xattr_buf, alloc_len, "%s",
-                                         local->xattr_val);
-                } else
-                        gf_log (this->name, GF_LOG_WARNING,
-                                "Unknown local->xsel (%s)", local->xsel);
+        local = frame->local;
 
-                ret = dict_set_dynstr (dict, local->xsel, xattr_buf);
+        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;
+                        }
 
-                GF_FREE (local->xattr_val);
+                        goto unlock;
+                }
 
-                DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict,
-                                  xdata);
+                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);
 
-                if (dict)
-                        dict_unref (dict);
+        if (!is_last_call (this_call_cnt))
+                goto out;
 
-                return 0;
+        /* -- last call: do patch ups -- */
+
+        if (local->op_ret == -1) {
+                goto unwind;
         }
 
-unwind:
+        ret = dht_vgetxattr_fill_and_set (local, &dict, this, _gf_true);
+        if (ret)
+                goto unwind;
+
+        DHT_STACK_UNWIND (getxattr, frame, 0, 0, dict, xdata);
+        goto cleanup;
+
+ unwind:
         DHT_STACK_UNWIND (getxattr, frame, -1, local->op_errno, NULL, NULL);
+ cleanup:
+        if (dict)
+                dict_unref (dict);
+ out:
+        return 0;
+}
+
+int
+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;
+        dict_t       *dict          = NULL;
+        call_frame_t *prev          = NULL;
+        gf_boolean_t  flag          = _gf_true;
+
+        local = frame->local;
+        prev = cookie;
+
+        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;
+        }
+
+        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;
+        }
+
+        flag = (local->layout->cnt > 1) ? _gf_true : _gf_false;
+
+        ret = dht_vgetxattr_fill_and_set (local, &dict, this, flag);
+        if (ret)
+                goto unwind;
+
+        DHT_STACK_UNWIND (getxattr, frame, 0, 0, dict, xdata);
+        goto cleanup;
+
+ unwind:
+        DHT_STACK_UNWIND (getxattr, frame, -1, local->op_errno,
+                          NULL, NULL);
+ cleanup:
+        if (dict)
+                dict_unref (dict);
+
         return 0;
 }
 
@@ -1830,10 +1993,13 @@ dht_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 {
         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);
@@ -1841,8 +2007,8 @@ dht_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         if (!xattr || (op_ret == -1))
                 goto out;
 
-        if (dict_get (xattr, "trusted.glusterfs.dht")) {
-                dict_del (xattr, "trusted.glusterfs.dht");
+        if (dict_get (xattr, conf->xattr_name)) {
+                dict_del (xattr, conf->xattr_name);
         }
         local->op_ret = 0;
 
@@ -1875,9 +2041,72 @@ dht_getxattr_unwind (call_frame_t *frame,
 
 
 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, 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;
@@ -1893,7 +2122,6 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
         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;
@@ -1921,8 +2149,40 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
                 }
         }
 
-        if (key && ((strcmp (key, GF_XATTR_PATHINFO_KEY) == 0)
-                    || strcmp (key, GF_XATTR_NODE_UUID_KEY) == 0)) {
+	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;
+        }
+
+        /* 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);
 
@@ -1932,6 +2192,7 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
 
                 return 0;
         }
+
         if (key && (strcmp (key, GF_XATTR_LINKINFO_KEY) == 0)) {
                 hashed_subvol = dht_subvol_get_hashed (this, loc);
                 if (!hashed_subvol) {
@@ -1964,13 +2225,13 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
         }
 
         if (key && (!strcmp (GF_XATTR_MARKER_KEY, key))
-            && (-1 == frame->root->pid)) {
-
-                if (loc->inode-> ia_type == IA_IFDIR) {
+            && (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;
@@ -1978,7 +2239,8 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
                 if (cluster_getmarkerattr (frame, this, loc, key,
                                            local, dht_getxattr_unwind,
                                            sub_volumes, cnt,
-                                           MARKER_UUID_TYPE, conf->vol_uuid)) {
+                                           MARKER_UUID_TYPE, marker_uuid_default_gauge,
+                                           conf->vol_uuid)) {
                         op_errno = EINVAL;
                         goto err;
                 }
@@ -1988,8 +2250,8 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
 
         if (key && *conf->vol_uuid) {
                 if ((match_uuid_local (key, conf->vol_uuid) == 0) &&
-                    (-1 == frame->root->pid)) {
-                        if (loc->inode-> ia_type == IA_IFDIR) {
+                    (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+                        if (DHT_IS_DIR(layout)) {
                                 cnt = layout->cnt;
                         } else {
                                 cnt = 1;
@@ -2002,6 +2264,7 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
                                                    local, dht_getxattr_unwind,
                                                    sub_volumes, cnt,
                                                    MARKER_XTIME_TYPE,
+                                                   marker_xtime_default_gauge,
                                                    conf->vol_uuid)) {
                                 op_errno = EINVAL;
                                 goto err;
@@ -2011,7 +2274,7 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
                 }
         }
 
-        if (loc->inode-> ia_type == IA_IFDIR) {
+        if (DHT_IS_DIR(layout)) {
                 cnt = local->call_cnt = layout->cnt;
         } else {
                 cnt = local->call_cnt  = 1;
@@ -2031,6 +2294,7 @@ err:
 
         return 0;
 }
+#undef DHT_IS_DIR
 
 int
 dht_fgetxattr (call_frame_t *frame, xlator_t *this,
@@ -2072,7 +2336,9 @@ dht_fgetxattr (call_frame_t *frame, xlator_t *this,
                 }
         }
 
-        if (fd->inode->ia_type == IA_IFDIR) {
+        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;
@@ -2100,13 +2366,17 @@ dht_fsetxattr (call_frame_t *frame, xlator_t *this,
         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 (fd, err);
         VALIDATE_OR_GOTO (fd->inode, err);
+        VALIDATE_OR_GOTO (this->private, err);
+
+        conf = this->private;
 
-        GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.dht*", xattr,
+        GF_IF_INTERNAL_XATTR_GOTO (conf->wild_xattr_name, xattr,
                                    op_errno, err);
 
         local = dht_local_init (frame, NULL, fd, GF_FOP_FSETXATTR);
@@ -2210,12 +2480,12 @@ dht_setxattr (call_frame_t *frame, xlator_t *this,
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (loc, err);
         VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
 
-        GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.dht*", xattr,
+        conf   = this->private;
+
+        GF_IF_INTERNAL_XATTR_GOTO (conf->wild_xattr_name, xattr,
                                    op_errno, err);
 
-        conf   = this->private;
         local = dht_local_init (frame, loc, NULL, GF_FOP_SETXATTR);
         if (!local) {
                 op_errno = ENOMEM;
@@ -2240,25 +2510,6 @@ dht_setxattr (call_frame_t *frame, xlator_t *this,
 
         local->call_cnt = call_cnt = layout->cnt;
 
-        /*  This key is sent by Unified File and Object storage
-         *  to test xattr support in backend.
-         */
-        tmp = dict_get (xattr, "user.ufo-test");
-        if (tmp) {
-                if (IA_ISREG (loc->inode->ia_type)) {
-                        op_errno = ENOTSUP;
-                        goto err;
-                }
-                local->op_ret = 0;
-                for (i = 0; i < call_cnt; i++) {
-                        STACK_WIND (frame, dht_ufo_xattr_cbk,
-                                    layout->list[i].xlator,
-                                    layout->list[i].xlator->fops->setxattr,
-                                    loc, xattr, flags, NULL);
-                }
-                return 0;
-        }
-
         tmp = dict_get (xattr, "distribute.migrate-data");
         if (tmp) {
                 if (IA_ISDIR (loc->inode->ia_type)) {
@@ -2333,9 +2584,13 @@ dht_setxattr (call_frame_t *frame, xlator_t *this,
                 gf_log (this->name, GF_LOG_INFO,
                         "fixing the layout of %s", loc->path);
 
-                dht_fix_directory_layout (frame, dht_common_setxattr_cbk,
-                                          layout);
-                return 0;
+                ret = dht_fix_directory_layout (frame, dht_common_setxattr_cbk,
+                                                layout);
+                if (ret) {
+                        op_errno = ENOTCONN;
+                        goto err;
+                }
+                return ret;
         }
 
         tmp = dict_get (xattr, "distribute.directory-spread-count");
@@ -2347,10 +2602,14 @@ dht_setxattr (call_frame_t *frame, xlator_t *this,
                              (dir_spread > 0))) {
                         layout->spread_cnt = dir_spread;
 
-                        dht_fix_directory_layout (frame,
-                                                  dht_common_setxattr_cbk,
-                                                  layout);
-                        return 0;
+                        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);
@@ -2420,18 +2679,20 @@ dht_removexattr (call_frame_t *frame, xlator_t *this,
         dht_local_t  *local = NULL;
         dht_layout_t *layout = NULL;
         int           call_cnt = 0;
+        dht_conf_t   *conf = NULL;
 
         int i;
 
         VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-        GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.dht*",
-                                 key, op_errno, 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);
-        VALIDATE_OR_GOTO (loc->path, err);
 
         local = dht_local_init (frame, loc, NULL, GF_FOP_REMOVEXATTR);
         if (!local) {
@@ -2483,13 +2744,16 @@ dht_fremovexattr (call_frame_t *frame, xlator_t *this,
         dht_local_t  *local = NULL;
         dht_layout_t *layout = NULL;
         int           call_cnt = 0;
+        dht_conf_t   *conf = 0;
 
         int i;
 
         VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (this->private, err);
+
+        conf = this->private;
 
-        GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.dht*",
-                                 key, op_errno, err);
+        GF_IF_NATIVE_XATTR_GOTO (conf->wild_xattr_name, key, op_errno, err);
 
         VALIDATE_OR_GOTO (frame, err);
 
@@ -2660,7 +2924,6 @@ dht_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
         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;
@@ -2780,10 +3043,13 @@ dht_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
 
         list_for_each_entry (orig_entry, (&orig_entries->list), list) {
                 next_offset = orig_entry->d_off;
-                if ((check_is_dir (NULL, (&orig_entry->d_stat), NULL) &&
-                     (prev->this != dht_first_up_subvol (this))) ||
-                    check_is_linkfile (NULL, (&orig_entry->d_stat),
-                                       orig_entry->dict)) {
+                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;
                 }
 
@@ -2825,6 +3091,9 @@ dht_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_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);
@@ -2857,13 +3126,16 @@ done:
                 }
 
 		if (conf->readdir_optimize == _gf_true) {
-                        if (next_subvol != dht_first_up_subvol (this)) {
+                        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,
@@ -2996,6 +3268,7 @@ dht_do_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
         conf = this->private;
 
@@ -3008,6 +3281,7 @@ dht_do_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t 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);
 
@@ -3020,20 +3294,22 @@ dht_do_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
 
                 if (local->xattr) {
                         ret = dict_set_uint32 (local->xattr,
-                                               "trusted.glusterfs.dht.linkto",
-                                               256);
+                                               conf->link_xattr_name, 256);
                         if (ret)
                                 gf_log (this->name, GF_LOG_WARNING,
-                                        "failed to set 'glusterfs.dht.linkto'"
-                                        " key");
+                                        "failed to set '%s' key",
+                                        conf->link_xattr_name);
 			if (conf->readdir_optimize == _gf_true) {
-                                if (xvol != dht_first_up_subvol (this)) {
+                                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);
                                 }
 			}
                 }
@@ -3169,7 +3445,7 @@ dht_newfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                  inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
                  struct iatt *postparent, dict_t *xdata)
 {
-        call_frame_t *prev = NULL;
+        xlator_t     *prev = NULL;
         int           ret = -1;
         dht_local_t  *local = NULL;
 
@@ -3187,19 +3463,24 @@ dht_newfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         prev = cookie;
 
         if (local->loc.parent) {
-                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);
+        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->this->name);
+                        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
@@ -3208,7 +3489,6 @@ out:
          * corresponding values from each of the subvolume.
          * See dht_iatt_merge for reference.
          */
-
         DHT_STRIP_PHASE1_FLAGS (stbuf);
         DHT_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode, stbuf,
                           preparent, postparent, xdata);
@@ -3230,12 +3510,17 @@ dht_mknod_linkfile_create_cbk (call_frame_t *frame, void *cookie,
                 goto err;
 
         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->umask,
-                    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:
@@ -3278,11 +3563,13 @@ dht_mknod (call_frame_t *frame, xlator_t *this,
                 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, umask, 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
                            files there */
@@ -3294,14 +3581,15 @@ dht_mknod (call_frame_t *frame, xlator_t *this,
                         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, umask, params);
+                        STACK_WIND_COOKIE (frame, dht_newfile_cbk,
+                                           (void *)subvol, subvol,
+                                           subvol->fops->mknod, loc, mode,
+                                           rdev, umask, params);
                 }
         }
 
@@ -3346,9 +3634,9 @@ dht_symlink (call_frame_t *frame, xlator_t *this,
         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, umask, params);
+        STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol, subvol,
+                           subvol->fops->symlink, linkname, loc, umask,
+                           params);
 
         return 0;
 
@@ -3437,9 +3725,12 @@ dht_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 {
         call_frame_t *prev = NULL;
         dht_layout_t *layout = NULL;
+        dht_local_t  *local = NULL;
 
         prev = cookie;
 
+        local = frame->local;
+
         if (op_ret == -1)
                 goto out;
 
@@ -3453,9 +3744,16 @@ dht_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 goto out;
         }
 
-        WIPE (preparent);
-        WIPE (postparent);
-
+        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,
@@ -3542,7 +3840,7 @@ dht_link (call_frame_t *frame, xlator_t *this,
 
         if (hashed_subvol != cached_subvol) {
                 uuid_copy (local->gfid, oldloc->inode->gfid);
-                dht_linkfile_create (frame, dht_link_linkfile_cbk,
+                dht_linkfile_create (frame, dht_link_linkfile_cbk, this,
                                      cached_subvol, hashed_subvol, newloc);
         } else {
                 STACK_WIND (frame, dht_link_cbk,
@@ -3583,8 +3881,11 @@ dht_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         prev = cookie;
 
         if (local->loc.parent) {
-                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);
@@ -3596,7 +3897,10 @@ dht_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 op_errno = EINVAL;
                 goto out;
         }
-
+        if (local->linked == _gf_true) {
+                local->stbuf = *stbuf;
+                dht_linkfile_attr_heal (frame, this);
+        }
 out:
         DHT_STRIP_PHASE1_FLAGS (stbuf);
         DHT_STACK_UNWIND (create, frame, op_ret, op_errno, fd, inode, stbuf, preparent,
@@ -3686,7 +3990,7 @@ dht_create (call_frame_t *frame, xlator_t *this,
         }
         /* Choose the minimum filled volume, and create the
            files there */
-        avail_subvol = dht_free_disk_available_subvol (this, subvol);
+        avail_subvol = dht_free_disk_available_subvol (this, subvol, local);
         if (avail_subvol != subvol) {
                 local->params = dict_ref (params);
                 local->flags = flags;
@@ -3697,9 +4001,8 @@ dht_create (call_frame_t *frame, xlator_t *this,
                 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,
@@ -3733,8 +4036,11 @@ dht_mkdir_selfheal_cbk (call_frame_t *frame, void *cookie,
         if (op_ret == 0) {
                 dht_layout_set (this, local->inode, layout);
                 if (local->loc.parent) {
-                        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);
                 }
         }
 
@@ -3769,6 +4075,15 @@ dht_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         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);
                 }
@@ -4001,8 +4316,15 @@ unlock:
                } else {
 
                         if (local->loc.parent) {
-                                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,
@@ -4088,8 +4410,17 @@ unlock:
                         /* All subvol's have responded, proceed */
 
                         if (local->loc.parent) {
-                                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,
@@ -4210,6 +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;
@@ -4221,7 +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;
 
@@ -4256,6 +4588,7 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
         dht_local_t        *lookup_local = NULL;
         dht_local_t        *local = NULL;
         dict_t             *xattrs = NULL;
+        dht_conf_t         *conf = this->private;
 
         local = frame->local;
 
@@ -4264,7 +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), trav->dict)) {
+                if (check_is_linkfile (NULL, (&trav->d_stat), trav->dict,
+                                              conf->link_xattr_name)) {
                         ret++;
                         continue;
                 }
@@ -4282,7 +4616,7 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
                 return -1;
         }
 
-        ret = dict_set_uint32 (xattrs, DHT_LINKFILE_KEY, 256);
+        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");
@@ -4405,6 +4739,7 @@ dht_rmdir_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         call_frame_t *prev = NULL;
         dict_t       *dict = NULL;
         int           ret = 0;
+        dht_conf_t   *conf = this->private;
 
         local = frame->local;
         prev  = cookie;
@@ -4428,12 +4763,11 @@ dht_rmdir_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 goto err;
         }
 
-        ret = dict_set_uint32 (dict,
-                               "trusted.glusterfs.dht.linkto", 256);
+        ret = dict_set_uint32 (dict, conf->link_xattr_name, 256);
         if (ret)
                 gf_log (this->name, GF_LOG_WARNING,
-                        "%s: failed to set 'trusted.glusterfs.dht.linkto' key",
-                        local->loc.path);
+                        "%s: failed to set '%s' key",
+                        local->loc.path, conf->link_xattr_name);
 
         STACK_WIND (frame, dht_rmdir_readdirp_cbk,
                     prev->this, prev->this->fops->readdirp,
@@ -4532,7 +4866,6 @@ dht_entrylk (call_frame_t *frame, xlator_t *this,
         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_ENTRYLK);
         if (!local) {
@@ -4611,16 +4944,21 @@ err:
 int
 dht_forget (xlator_t *this, inode_t *inode)
 {
-        uint64_t      tmp_layout = 0;
+        uint64_t        ctx_int = 0;
+        dht_inode_ctx_t *ctx    = NULL;
         dht_layout_t *layout = NULL;
 
-        inode_ctx_del (inode, this, &tmp_layout);
+        inode_ctx_del (inode, this, &ctx_int);
 
-        if (!tmp_layout)
+        if (!ctx_int)
                 return 0;
 
-        layout = (dht_layout_t *)(long)tmp_layout;
+        ctx = (dht_inode_ctx_t *) (long) ctx_int;
+
+        layout = ctx->layout;
+        ctx->layout = NULL;
         dht_layout_unref (this, layout);
+        GF_FREE (ctx);
 
         return 0;
 }
@@ -4808,15 +5146,7 @@ unlock:
            or wait for anything else. Just propagate blindly */
         if (have_heard_from_all) {
                 propagate = 1;
-                if (conf->defrag) {
-                        ret = pthread_create (&conf->defrag->th, NULL,
-                                              gf_defrag_start, this);
-                        if (ret) {
-                                conf->defrag = NULL;
-                                GF_FREE (conf->defrag);
-                                kill (getpid(), SIGTERM);
-                        }
-                }
+
         }
 
 
@@ -4838,6 +5168,19 @@ unlock:
                                 /* continue to check other events for CHILD_UP */
                         }
                 }
+
+                /* 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);
+                        }
+                }
         }
 
         ret = 0;
@@ -4846,3 +5189,24 @@ unlock:
 
         return ret;
 }
+
+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;
+
+        ret = dht_inode_ctx_get (inode, this, &ctx);
+
+        if (!ret && ctx) {
+                if (ctx->layout) {
+                        if (layout)
+                                *layout = ctx->layout;
+                        ret = 0;
+                } else {
+                        ret = -1;
+                }
+        }
+
+        return ret;
+}
diff --git a/xlators/cluster/dht/src/dht-common.h b/xlators/cluster/dht/src/dht-common.h
index c02c0b83d..5ccd66799 100644
--- a/xlators/cluster/dht/src/dht-common.h
+++ b/xlators/cluster/dht/src/dht-common.h
@@ -13,6 +13,8 @@
 #include "config.h"
 #endif
 
+#include <regex.h>
+
 #include "dht-mem-types.h"
 #include "libxlator.h"
 #include "syncop.h"
@@ -53,10 +55,28 @@ struct dht_layout {
                 uint32_t   start;
                 uint32_t   stop;
                 xlator_t  *xlator;
-        } list[0];
+        } 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_inode_ctx dht_inode_ctx_t;
+
 
 typedef enum {
         DHT_HASH_TYPE_DM,
@@ -122,7 +142,6 @@ struct dht_local {
         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;
@@ -143,6 +162,7 @@ struct dht_local {
 
         /* which xattr request? */
         char xsel[256];
+        int32_t alloc_len;
 
         char   *newpath;
 
@@ -159,7 +179,11 @@ struct dht_local {
 
         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;
@@ -188,15 +212,27 @@ enum gf_defrag_status_t {
         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;
@@ -209,7 +245,7 @@ struct gf_defrag_info_ {
         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;
@@ -225,8 +261,8 @@ struct dht_conf {
         gf_boolean_t   search_unhashed;
         int            gen;
         dht_du_t      *du_stats;
-        uint64_t       min_free_disk;
-	uint32_t       min_free_inodes;
+        double         min_free_disk;
+	double         min_free_inodes;
         char           disk_unit;
         int32_t        refresh_interval;
         gf_boolean_t   unhashed_sticky_bit;
@@ -246,6 +282,7 @@ struct dht_conf {
         /* 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;
@@ -253,6 +290,17 @@ struct dht_conf {
         /* 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;
 
@@ -276,20 +324,19 @@ typedef enum {
 
 #define ENTRY_MISSING(op_ret, op_errno) (op_ret == -1 && op_errno == ENOENT)
 
-#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_MIGRATION_IN_PROGRESS 1
 #define DHT_MIGRATION_COMPLETED   2
 
-#define DHT_LINKFILE_KEY         "trusted.glusterfs.dht.linkto"
 #define DHT_LINKFILE_MODE        (S_ISVTX)
 
-#define check_is_linkfile(i,s,x) (                                      \
+#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, DHT_LINKFILE_KEY))
+                 == DHT_LINKFILE_MODE) &&                                 \
+                dict_get (x, n))
 
 #define IS_DHT_MIGRATION_PHASE2(buf)  (                                 \
                 IA_ISREG ((buf)->ia_type) &&                            \
@@ -334,6 +381,25 @@ typedef enum {
                 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);
@@ -343,7 +409,7 @@ int                                      dht_layout_normalize (xlator_t *this, l
 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      *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);
 
@@ -377,12 +443,14 @@ int dht_iatt_merge (xlator_t                 *this, struct iatt *to, struct iatt
 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);
+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);
+                         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
@@ -400,7 +468,8 @@ dht_layout_sort_volname (dht_layout_t *layout);
 int dht_get_du_info (call_frame_t *frame, xlator_t *this, loc_t *loc);
 
 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);
+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);
@@ -622,7 +691,16 @@ 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 */
@@ -667,4 +745,43 @@ gf_defrag_handle_hardlink (xlator_t *this, loc_t *loc, dict_t  *xattrs,
 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 */
diff --git a/xlators/cluster/dht/src/dht-diskusage.c b/xlators/cluster/dht/src/dht-diskusage.c
index 52ea3a32a..fe3955ecb 100644
--- a/xlators/cluster/dht/src/dht-diskusage.c
+++ b/xlators/cluster/dht/src/dht-diskusage.c
@@ -248,50 +248,167 @@ dht_is_subvol_filled (xlator_t *this, xlator_t *subvol)
 	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_free_disk_available_subvol (xlator_t *this, xlator_t *subvol,
+                                dht_local_t *local)
 {
-	int         i = 0;
-	double      max = 0;
-	double      max_inodes = 0;
 	xlator_t   *avail_subvol = NULL;
 	dht_conf_t *conf = NULL;
+        dht_layout_t *layout = NULL;
+        loc_t      *loc = NULL;
 
 	conf = this->private;
-
-	LOCK (&conf->subvolume_lock);
+        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);
+        }
+
+        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)
-				    && (conf->du_stats[i].avail_inodes > max_inodes)) {
-					max = conf->du_stats[i].avail_percent;
-					max_inodes = conf->du_stats[i].avail_inodes;
-					avail_subvol = conf->subvolumes[i];
-				}
-			} else {
-				if ((conf->du_stats[i].avail_space > max)
-				    && (conf->du_stats[i].avail_inodes > max_inodes)) {
-					max = conf->du_stats[i].avail_space;
-					max_inodes = conf->du_stats[i].avail_inodes;
-					avail_subvol = conf->subvolumes[i];
-				}
+                avail_subvol = dht_subvol_with_free_space_inodes(this, subvol,
+                                                                 layout);
+                if(!avail_subvol)
+                {
+                        avail_subvol = dht_subvol_maxspace_nonzeroinode(this,
+                                                                        subvol,
+                                                                        layout);
+                }
 
-			}
-		}
 	}
 	UNLOCK (&conf->subvolume_lock);
-
+out:
 	if (!avail_subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume has enough free space and inodes to create");
+		gf_log (this->name,
+                        GF_LOG_DEBUG,
+			"no subvolume has enough free space and/or inodes\
+                         to create");
+                avail_subvol = subvol;
 	}
 
-	if ((max < conf->min_free_disk) && (max_inodes < conf->min_free_inodes))
-		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_subvol_with_free_space_inodes(xlator_t *this, xlator_t *subvol,
+                                  dht_layout_t *layout)
+{
+        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;
+
+        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];
+                        }
+                }
+
+                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];
+                        }
+                }
+        }
+
+        return avail_subvol;
+}
 
-	if (!avail_subvol)
-		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;
+
+        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 97eb1f05f..656cf23a0 100644
--- a/xlators/cluster/dht/src/dht-hashfn.c
+++ b/xlators/cluster/dht/src/dht-hashfn.c
@@ -44,30 +44,68 @@ dht_hash_compute_internal (int type, const char *name, uint32_t *hash_p)
 }
 
 
-#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;
+        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.
+         */
 
-        MAKE_RSYNC_FRIENDLY_NAME (rsync_friendly_name, name);
+        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 f0e74a154..311a48112 100644
--- a/xlators/cluster/dht/src/dht-helper.c
+++ b/xlators/cluster/dht/src/dht-helper.c
@@ -18,6 +18,28 @@
 #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)
@@ -40,6 +62,43 @@ dht_frame_return (call_frame_t *frame)
 }
 
 
+static uint64_t
+dht_bits_for (uint64_t num)
+{
+	uint64_t bits = 0, ctrl = 1;
+
+	while (ctrl < num) {
+		ctrl *= 2;
+		bits ++;
+	}
+
+	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)
 {
@@ -47,6 +106,9 @@ dht_itransform (xlator_t *this, xlator_t *subvol, uint64_t x, uint64_t *y_p)
         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;
@@ -60,7 +122,23 @@ dht_itransform (xlator_t *this, xlator_t *subvol, uint64_t x, uint64_t *y_p)
         max = conf->subvolume_cnt;
         cnt = dht_subvol_cnt (this, subvol);
 
-        y = ((x * max) + cnt);
+	if (max == 1) {
+		y = x;
+		goto out;
+	}
+
+        max_bits = dht_bits_for (max);
+
+        hi_mask = ~(PRESENT_MASK >> (max_bits + 1));
+
+        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)
@@ -135,16 +213,38 @@ dht_deitransform (xlator_t *this, uint64_t y, xlator_t **subvol_p,
         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)
-                goto out;
+                return -1;
 
         conf = this->private;
         max = conf->subvolume_cnt;
 
-        cnt = y % max;
-        x   = y / max;
+	if (max == 1) {
+		x = y;
+		cnt = 0;
+		goto out;
+	}
+
+        if (y & TOP_BIT) {
+                /* HUGE d_off */
+                max_bits = dht_bits_for (max);
+                off_mask = (MASK << max_bits);
+                host_mask = ~(off_mask);
+
+                x = ((y & ~TOP_BIT) & off_mask) << SHIFT_BITS;
+
+                cnt = y & host_mask;
+	} else {
+                /* small d_off */
+                cnt = y % max;
+                x = y / max;
+        }
 
+out:
         subvol = conf->subvolumes[cnt];
 
         if (subvol_p)
@@ -153,7 +253,6 @@ dht_deitransform (xlator_t *this, uint64_t y, xlator_t **subvol_p,
         if (x_p)
                 *x_p = x;
 
-out:
         return 0;
 }
 
@@ -263,20 +362,6 @@ out:
         return local;
 }
 
-
-char *
-basestr (const char *str)
-{
-        char *basestr = NULL;
-
-        basestr = strrchr (str, '/');
-        if (basestr)
-                basestr ++;
-
-        return basestr;
-}
-
-
 xlator_t *
 dht_first_up_subvol (xlator_t *this)
 {
@@ -428,7 +513,36 @@ 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)
 {
@@ -620,20 +734,36 @@ dht_migration_complete_check_task (void *data)
         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;
 
-        /* getxattr on cached_subvol for 'linkto' value */
-        if (!local->loc.inode)
+        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,
-                                        DHT_LINKFILE_KEY);
-        else
+                                        conf->link_xattr_name);
+        } else {
+                SYNCTASK_SETID (0, 0);
                 ret = syncop_getxattr (src_node, &local->loc, &dict,
-                                       DHT_LINKFILE_KEY);
+                                       conf->link_xattr_name);
+                SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+        }
 
         if (!ret)
                 dst_node = dht_linkfile_subvol (this, NULL, NULL, dict);
@@ -684,10 +814,7 @@ dht_migration_complete_check_task (void *data)
         /* update inode ctx (the layout) */
         dht_layout_unref (this, local->layout);
 
-        if (!local->loc.inode)
-                ret = dht_layout_preset (this, dst_node, local->fd->inode);
-        else
-                ret = dht_layout_preset (this, dst_node, local->loc.inode);
+        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",
@@ -705,10 +832,7 @@ dht_migration_complete_check_task (void *data)
                 goto out;
         }
 
-        if (!local->loc.inode)
-                ret = dht_layout_set (this, local->fd->inode, layout);
-        else
-                ret = dht_layout_set (this, local->loc.inode, layout);
+        ret = dht_layout_set (this, inode, layout);
         if (ret) {
                 gf_log (this->name, GF_LOG_ERROR,
                         "%s: failed to set the new layout",
@@ -719,42 +843,46 @@ dht_migration_complete_check_task (void *data)
         local->cached_subvol = dst_node;
         ret = 0;
 
-        if (!local->fd)
+        /* 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;
 
-        /* once we detect the migration complete, the fd-ctx is no more
-           required.. delete the ctx, and do one extra 'fd_unref' for open fd */
-        ret = fd_ctx_del (local->fd, this, NULL);
-        if (!ret) {
-                fd_unref (local->fd);
-                ret = 0;
+        if (list_empty (&inode->fd_list))
                 goto out;
-        }
 
-        /* if 'local->fd' (ie, fd based operation), send a 'open()' on
-           destination if not already done */
-        if (local->loc.inode) {
-                ret = syncop_open (dst_node, &local->loc,
-                                   local->fd->flags, local->fd);
-        } else {
-                tmp_loc.inode = local->fd->inode;
-                inode_path (local->fd->inode, NULL, &path);
-                if (path)
-                        tmp_loc.path = path;
-                ret = syncop_open (dst_node, &tmp_loc,
-                                   local->fd->flags, local->fd);
-                GF_FREE (path);
+        /* 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;
+                }
         }
-        if (ret == -1) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "%s: failed to send open() on target file at %s",
-                        local->loc.path, dst_node->name);
+        GF_FREE (path);
+
+        SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+        if (open_failed) {
+                ret = -1;
                 goto out;
         }
-
-        /* need this unref for the fd on src_node */
-        fd_unref (local->fd);
         ret = 0;
 out:
 
@@ -798,20 +926,34 @@ dht_rebalance_inprogress_task (void *data)
         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;
 
-        /* getxattr on cached_subvol for 'linkto' value */
-        if (local->loc.inode)
+        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,
-                                       DHT_LINKFILE_KEY);
-        else
+                                       conf->link_xattr_name);
+                SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+        } else {
                 ret = syncop_fgetxattr (src_node, local->fd, &dict,
-                                        DHT_LINKFILE_KEY);
+                                        conf->link_xattr_name);
+        }
 
         if (ret) {
                 gf_log (this->name, GF_LOG_ERROR,
@@ -853,33 +995,46 @@ dht_rebalance_inprogress_task (void *data)
 
         ret = 0;
 
-        if (!local->fd)
-                goto out;
+        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;
 
-        if (local->loc.inode) {
-                ret = syncop_open (dst_node, &local->loc,
-                                   local->fd->flags, local->fd);
-        } else {
-                tmp_loc.inode = local->fd->inode;
-                inode_path (local->fd->inode, NULL, &path);
-                if (path)
-                        tmp_loc.path = path;
                 ret = syncop_open (dst_node, &tmp_loc,
-                                   local->fd->flags, local->fd);
-                GF_FREE (path);
+                                   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);
 
-        if (ret == -1) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "%s: failed to send open() on target file at %s",
-                        local->loc.path, dst_node->name);
+        SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+
+        if (open_failed) {
+                ret = -1;
                 goto out;
         }
 
-        ret = fd_ctx_set (local->fd, this, (uint64_t)(long)dst_node);
+done:
+        ret = dht_inode_ctx_set1 (this, inode, dst_node);
         if (ret) {
                 gf_log (this->name, GF_LOG_ERROR,
-                        "%s: failed to set fd-ctx target file at %s",
+                        "%s: failed to set inode-ctx target file at %s",
                         local->loc.path, dst_node->name);
                 goto out;
         }
@@ -900,3 +1055,93 @@ dht_rebalance_in_progress_check (xlator_t *this, call_frame_t *frame)
                             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
index 7e29a7a8c..ece84151a 100644
--- a/xlators/cluster/dht/src/dht-inode-read.c
+++ b/xlators/cluster/dht/src/dht-inode-read.c
@@ -130,10 +130,11 @@ 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)
 {
-        uint64_t      tmp_subvol = 0;
+        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);
@@ -154,21 +155,23 @@ dht_file_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         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)) {
-                if (local->fd)
-                        ret = fd_ctx_get (local->fd, this, &tmp_subvol);
-                if (ret) {
+                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);
-                }
-                if (!ret)
                         return 0;
+                }
         }
 
 out:
@@ -381,6 +384,8 @@ dht_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 {
         dht_local_t *local      = NULL;
         int          ret        = 0;
+        inode_t     *inode      = NULL;
+        xlator_t    *subvol = 0;
 
         local = frame->local;
         if (!local) {
@@ -396,19 +401,21 @@ dht_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         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 = fd_ctx_get (local->fd, this, NULL);
-                if (ret) {
+                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);
-                }
-                if (!ret)
                         return 0;
+                }
         }
 
 out:
@@ -498,14 +505,35 @@ dht_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 {
         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)
@@ -593,8 +621,9 @@ 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;
-        int           ret = -1;
+        dht_local_t  *local  = NULL;
+        inode_t      *inode  = NULL;
+        xlator_t     *subvol = 0;
 
         local = frame->local;
 
@@ -604,8 +633,8 @@ dht_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 goto out;
 
         /* If context is set, then send flush() it to the destination */
-        ret = fd_ctx_get (local->fd, this, NULL);
-        if (!ret) {
+        dht_inode_ctx_get1 (this, inode, &subvol);
+        if (subvol) {
                 dht_flush2 (this, frame, 0);
                 return 0;
         }
@@ -621,14 +650,10 @@ dht_flush2 (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;
 
-        ret = fd_ctx_get (local->fd, this, &tmp_subvol);
-        if (!ret)
-                subvol = (xlator_t *)(long)tmp_subvol;
+        dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
 
         if (!subvol)
                 subvol = local->cached_subvol;
@@ -690,12 +715,14 @@ dht_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
         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) {
+        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));
@@ -710,8 +737,9 @@ dht_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
                 goto out;
         }
 
-        ret = fd_ctx_get (local->fd, this, NULL);
-        if (ret) {
+        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 */
@@ -726,11 +754,12 @@ dht_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
                 if (IS_DHT_MIGRATION_PHASE2 (postbuf)) {
                         ret = dht_rebalance_complete_check (this, frame);
                 }
+                if (!ret)
+                        return 0;
         } else {
                 dht_fsync2 (this, frame, 0);
-        }
-        if (!ret)
                 return 0;
+        }
 
 out:
         DHT_STRIP_PHASE1_FLAGS (postbuf);
@@ -746,15 +775,10 @@ dht_fsync2 (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;
 
-        ret = fd_ctx_get (local->fd, this, &tmp_subvol);
-        if (!ret)
-                subvol = (xlator_t *)(long)tmp_subvol;
-
+        dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
         if (!subvol)
                 subvol = local->cached_subvol;
 
diff --git a/xlators/cluster/dht/src/dht-inode-write.c b/xlators/cluster/dht/src/dht-inode-write.c
index d4a3ecc39..4b3f3a049 100644
--- a/xlators/cluster/dht/src/dht-inode-write.c
+++ b/xlators/cluster/dht/src/dht-inode-write.c
@@ -19,6 +19,9 @@
 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,
@@ -27,8 +30,9 @@ dht_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 {
         dht_local_t *local = NULL;
         int          ret   = -1;
+        xlator_t    *subvol = NULL;
 
-        if (op_ret == -1) {
+        if (op_ret == -1 && (op_errno != ENOENT)) {
                 goto out;
         }
 
@@ -50,6 +54,7 @@ dht_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         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);
@@ -62,8 +67,8 @@ dht_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 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) {
+                ret = dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+                if (subvol) {
                         dht_writev2 (this, frame, 0);
                         return 0;
                 }
@@ -87,14 +92,10 @@ dht_writev2 (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;
 
-        ret = fd_ctx_get (local->fd, this, &tmp_subvol);
-        if (!ret)
-                subvol = (xlator_t *)(long)tmp_subvol;
+        dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
 
         if (!subvol)
                 subvol = local->cached_subvol;
@@ -169,6 +170,8 @@ dht_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         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);
@@ -198,6 +201,7 @@ dht_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         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);
@@ -209,8 +213,9 @@ dht_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         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) {
+                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;
                 }
@@ -234,16 +239,13 @@ dht_truncate2 (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;
+        inode_t      *inode = NULL;
 
         local = frame->local;
 
-        if (local->fd)
-                ret = fd_ctx_get (local->fd, this, &tmp_subvol);
-        if (!ret)
-                subvol = (xlator_t *)(long)tmp_subvol;
+        inode = local->fd ? local->fd->inode : local->loc.inode;
 
+        dht_inode_ctx_get1 (this, inode, &subvol);
         if (!subvol)
                 subvol = local->cached_subvol;
 
@@ -346,6 +348,407 @@ err:
         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,
@@ -397,15 +800,13 @@ dht_setattr2 (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;
+        inode_t      *inode = NULL;
 
         local = frame->local;
 
-        if (local->fd)
-                ret = fd_ctx_get (local->fd, this, &tmp_subvol);
-        if (!ret)
-                subvol = (xlator_t *)(long)tmp_subvol;
+        inode = (local->fd) ? local->fd->inode : local->loc.inode;
+
+        dht_inode_ctx_get1 (this, inode, &subvol);
 
         if (!subvol)
                 subvol = local->cached_subvol;
diff --git a/xlators/cluster/dht/src/dht-layout.c b/xlators/cluster/dht/src/dht-layout.c
index 97bbe987f..38e9970a7 100644
--- a/xlators/cluster/dht/src/dht-layout.c
+++ b/xlators/cluster/dht/src/dht-layout.c
@@ -59,9 +59,7 @@ 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)
@@ -69,9 +67,8 @@ dht_layout_get (xlator_t *this, inode_t *inode)
 
         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++;
                 }
         }
@@ -89,7 +86,6 @@ 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)
@@ -97,16 +93,13 @@ dht_layout_set (xlator_t *this, inode_t *inode, dht_layout_t *layout)
 
         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);
         }
 
@@ -121,7 +114,7 @@ dht_layout_unref (xlator_t *this, dht_layout_t *layout)
         dht_conf_t  *conf = NULL;
         int          ref = 0;
 
-        if (layout->preset || !this->private)
+        if (!layout || layout->preset || !this->private)
                 return;
 
         conf = this->private;
@@ -165,7 +158,7 @@ dht_layout_search (xlator_t *this, dht_layout_t *layout, const char *name)
         int        ret = 0;
 
 
-        ret = dht_hash_compute (layout->type, name, &hash);
+        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",
@@ -342,11 +335,12 @@ int
 dht_layout_merge (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
                   int op_ret, int op_errno, dict_t *xattr)
 {
-        int      i     = 0;
-        int      ret   = -1;
-        int      err   = -1;
-        void    *disk_layout_raw = NULL;
-        int      disk_layout_len = 0;
+        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;
@@ -367,12 +361,12 @@ dht_layout_merge (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
 
         if (xattr) {
                 /* during lookup and not mkdir */
-                ret = dict_get_ptr_and_len (xattr, "trusted.glusterfs.dht",
+                ret = dict_get_ptr_and_len (xattr, conf->xattr_name,
 					    &disk_layout_raw, &disk_layout_len);
         }
 
         if (ret != 0) {
-                layout->list[i].err = -1;
+                layout->list[i].err = 0;
                 gf_log (this->name, GF_LOG_TRACE,
                         "missing disk layout on %s. err = %d",
                         subvol->name, err);
@@ -419,6 +413,22 @@ dht_layout_entry_swap (dht_layout_t *layout, int i, int j)
         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)
 {
@@ -426,17 +436,37 @@ dht_layout_entry_cmp_volname (dht_layout_t *layout, int i, int j)
                         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;
 
+        /* 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;
 
+out:
         return diff;
 }
 
@@ -485,7 +515,8 @@ dht_layout_sort_volname (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 *missing_p, uint32_t *down_p, uint32_t *misc_p,
+                      uint32_t *no_space_p)
 {
         uint32_t    overlaps = 0;
         uint32_t    missing  = 0;
@@ -498,6 +529,7 @@ dht_layout_anomalies (xlator_t *this, loc_t *loc, dht_layout_t *layout,
         uint32_t    prev_stop = 0;
         uint32_t    last_stop = 0;
         char        is_virgin = 1;
+        uint32_t    no_space  = 0;
 
         /* TODO: explain what is happening */
 
@@ -505,23 +537,30 @@ dht_layout_anomalies (xlator_t *this, loc_t *loc, dht_layout_t *layout,
         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++;
+                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;
 
@@ -554,6 +593,9 @@ dht_layout_anomalies (xlator_t *this, loc_t *loc, dht_layout_t *layout,
         if (misc_p)
                 *misc_p = misc;
 
+        if (no_space_p)
+                *no_space_p = no_space;
+
         return ret;
 }
 
@@ -578,7 +620,7 @@ dht_layout_normalize (xlator_t *this, loc_t *loc, dht_layout_t *layout)
 
         ret = dht_layout_anomalies (this, loc, layout,
                                     &holes, &overlaps,
-                                    &missing, &down, &misc);
+                                    &missing, &down, &misc, NULL);
         if (ret == -1) {
                 gf_log (this->name, GF_LOG_WARNING,
                         "error while finding anomalies in %s -- not good news",
@@ -622,21 +664,30 @@ out:
         return ret;
 }
 
+int
+dht_dir_has_layout (dict_t *xattr, char *name)
+{
+
+        void     *disk_layout_raw = NULL;
+
+        return dict_get_ptr (xattr, name, &disk_layout_raw);
+}
 
 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;
+        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++) {
@@ -666,7 +717,7 @@ dht_layout_dir_mismatch (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
                 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) {
@@ -731,7 +782,7 @@ dht_layout_preset (xlator_t *this, xlator_t *subvol, inode_t *inode)
 
         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 803f344af..dbc9d0b3c 100644
--- a/xlators/cluster/dht/src/dht-linkfile.c
+++ b/xlators/cluster/dht/src/dht-linkfile.c
@@ -19,8 +19,37 @@
 #include "compat.h"
 #include "dht-common.h"
 
+int
+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)
+{
+        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;
+
+        if (op_ret)
+                goto out;
 
+        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,
@@ -28,29 +57,68 @@ dht_linkfile_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                          struct iatt *postparent, dict_t *xdata)
 {
         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,
                                       xdata);
+        if (xattrs)
+                dict_unref (xattrs);
         return 0;
 }
 
 
 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)
 {
         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 ();
@@ -71,8 +139,7 @@ dht_linkfile_create (call_frame_t *frame, fop_mknod_cbk_t linkfile_cbk,
                 gf_log ("dht-linkfile", GF_LOG_INFO,
                         "%s: internal-fop set failed", loc->path);
 
-        ret = dict_set_str (dict, "trusted.glusterfs.dht.linkto",
-                                   tovol->name);
+        ret = dict_set_str (dict, conf->link_xattr_name, tovol->name);
 
         if (ret < 0) {
                 gf_log (frame->this->name, GF_LOG_INFO,
@@ -81,6 +148,10 @@ dht_linkfile_create (call_frame_t *frame, fop_mknod_cbk_t linkfile_cbk,
                 goto out;
         }
 
+        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);
@@ -173,7 +244,7 @@ dht_linkfile_subvol (xlator_t *this, inode_t *inode, struct iatt *stbuf,
         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;
@@ -188,3 +259,70 @@ dht_linkfile_subvol (xlator_t *this, inode_t *inode, struct iatt *stbuf,
 out:
         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 4f3a90e91..e893eb48f 100644
--- a/xlators/cluster/dht/src/dht-mem-types.h
+++ b/xlators/cluster/dht/src/dht-mem-types.h
@@ -28,6 +28,8 @@ enum gf_dht_mem_types_ {
         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
index 2989d6a43..bcb19f23e 100644
--- a/xlators/cluster/dht/src/dht-rebalance.c
+++ b/xlators/cluster/dht/src/dht-rebalance.c
@@ -16,6 +16,7 @@
 
 #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 */
@@ -101,12 +102,16 @@ gf_defrag_handle_hardlink (xlator_t *this, loc_t *loc, dict_t  *xattrs,
         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 "
@@ -137,10 +142,10 @@ gf_defrag_handle_hardlink (xlator_t *this, loc_t *loc, dict_t  *xattrs,
         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, DHT_LINKFILE_KEY);
+        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, DHT_LINKFILE_KEY,
+                ret = dict_set_str (xattrs, conf->link_xattr_name,
                                     hashed_subvol->name);
                 if (ret) {
                         gf_log (this->name, GF_LOG_ERROR, "Failed to set "
@@ -238,14 +243,16 @@ out:
 
 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 *dict, fd_t **dst_fd, dict_t *xattr)
 {
-        xlator_t *this = NULL;
-        int       ret  = -1;
-        fd_t     *fd   = NULL;
-        struct iatt new_stbuf = {0,};
+        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) {
@@ -254,7 +261,7 @@ __dht_rebalance_create_dst_file (xlator_t *to, xlator_t *from, loc_t *loc, struc
                 goto out;
         }
 
-        ret = dict_set_str (dict, DHT_LINKFILE_KEY, from->name);
+        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);
@@ -292,7 +299,7 @@ __dht_rebalance_create_dst_file (xlator_t *to, xlator_t *from, loc_t *loc, struc
         /* 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);
+                             dict, &new_stbuf);
         if (ret < 0) {
                 gf_log (this->name, GF_LOG_ERROR,
                         "failed to create %s on %s (%s)",
@@ -300,6 +307,18 @@ __dht_rebalance_create_dst_file (xlator_t *to, xlator_t *from, loc_t *loc, struc
                 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);
@@ -327,6 +346,9 @@ __dht_check_free_space (xlator_t *to, xlator_t *from, loc_t *loc,
         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);
@@ -350,22 +372,34 @@ __dht_check_free_space (xlator_t *to, xlator_t *from, loc_t *loc,
         if (flag != GF_DHT_MIGRATE_DATA)
                 goto check_avail_space;
 
-        if (((dst_statfs.f_bavail *
-              dst_statfs.f_bsize) / GF_DISK_SECTOR_SIZE) <
-            (((src_statfs.f_bavail * src_statfs.f_bsize) /
-              GF_DISK_SECTOR_SIZE) - 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:
+           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) {
@@ -442,8 +476,10 @@ __dht_rebalance_open_src_file (xlator_t *from, xlator_t *to, loc_t *loc,
         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) {
@@ -466,7 +502,7 @@ __dht_rebalance_open_src_file (xlator_t *from, xlator_t *to, loc_t *loc,
         if (!dict)
                 goto out;
 
-        ret = dict_set_str (dict, DHT_LINKFILE_KEY, to->name);
+        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)",
@@ -519,12 +555,13 @@ migrate_special_files (xlator_t *this, xlator_t *from, xlator_t *to, loc_t *loc,
         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, DHT_LINKFILE_KEY, 256);
+        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);
@@ -540,12 +577,13 @@ migrate_special_files (xlator_t *this, xlator_t *from, xlator_t *to, loc_t *loc,
         }
 
         /* we no more require this key */
-        dict_del (dict, DHT_LINKFILE_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)) {
+                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;
@@ -581,7 +619,7 @@ migrate_special_files (xlator_t *this, xlator_t *from, xlator_t *to, loc_t *loc,
                         goto out;
                 }
 
-                ret = syncop_symlink (to, loc, link, dict);
+                ret = syncop_symlink (to, loc, link, dict, 0);
                 if (ret) {
                         gf_log (this->name, GF_LOG_WARNING,
                                 "%s: creating symlink failed (%s)",
@@ -595,7 +633,7 @@ migrate_special_files (xlator_t *this, xlator_t *from, xlator_t *to, loc_t *loc,
         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);
+                                     ia_minor (buf->ia_rdev)), dict, 0);
         if (ret) {
                 gf_log (this->name, GF_LOG_WARNING, "%s: mknod failed (%s)",
                         loc->path, strerror (errno));
@@ -603,6 +641,15 @@ migrate_special_files (xlator_t *this, xlator_t *from, xlator_t *to, loc_t *loc,
         }
 
 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)",
@@ -640,6 +687,7 @@ dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
         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);
@@ -648,7 +696,7 @@ dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
         if (!dict)
                 goto out;
 
-        ret = dict_set_int32 (dict, DHT_LINKFILE_KEY, 256);
+        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);
@@ -664,7 +712,7 @@ dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
         }
 
         /* we no more require this key */
-        dict_del (dict, DHT_LINKFILE_KEY);
+        dict_del (dict, conf->link_xattr_name);
 
         /* preserve source mode, so set the same to the destination */
         src_ia_prot = stbuf.ia_prot;
@@ -681,9 +729,16 @@ dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
                 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);
+                                               dict, &dst_fd, xattr);
         if (ret)
                 goto out;
 
@@ -700,6 +755,7 @@ dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
                 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)",
@@ -729,19 +785,6 @@ dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
                 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));
-
-        ret = syncop_setxattr (to, loc, 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));
-
         /* TODO: Sync the locks */
 
         ret = syncop_fsync (to, dst_fd, 0);
@@ -805,6 +848,23 @@ dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
                 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) {
@@ -825,23 +885,6 @@ dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
                 }
         }
 
-        /* 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, DHT_LINKFILE_KEY);
-        if (ret) {
-                gf_log (this->name, GF_LOG_WARNING,
-                        "%s: failed to perform removexattr on %s (%s)",
-                        loc->path, to->name, strerror (errno));
-        }
-
         ret = syncop_lookup (this, loc, NULL, NULL, NULL, NULL);
         if (ret) {
                 gf_log (this->name, GF_LOG_DEBUG,
@@ -1031,6 +1074,31 @@ gf_defrag_handle_migrate_error (int32_t op_errno, gf_defrag_info_t *defrag)
         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
@@ -1058,6 +1126,7 @@ gf_defrag_migrate_data (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
         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);
@@ -1080,17 +1149,24 @@ gf_defrag_migrate_data (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
 
         while ((ret = syncop_readdirp (this, fd, 131072, offset, NULL,
                                        &entries)) != 0) {
-                if (ret < 0)
-                        break;
+
+                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;
 
-                free_entries = _gf_true;
-
                 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;
@@ -1110,6 +1186,12 @@ gf_defrag_migrate_data (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
                         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);
@@ -1202,9 +1284,21 @@ gf_defrag_migrate_data (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
                         ret = syncop_setxattr (this, &entry_loc, migrate_data,
                                                0);
                         if (ret) {
-                                gf_log (this->name, GF_LOG_ERROR, "migrate-data"
-                                        " failed for %s", entry_loc.path);
-                                defrag->total_failures +=1;
+                                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) {
@@ -1284,6 +1378,7 @@ gf_defrag_fix_layout (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
         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) {
@@ -1319,12 +1414,22 @@ gf_defrag_fix_layout (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
         while ((ret = syncop_readdirp (this, fd, 131072, offset, NULL,
                 &entries)) != 0)
         {
-                if ((ret < 0) || (ret && (errno == ENOENT)))
-                        break;
-                free_entries = _gf_true;
+
+                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;
@@ -1351,7 +1456,7 @@ gf_defrag_fix_layout (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
 
                         if (uuid_is_null (entry->d_stat.ia_gfid)) {
                                 gf_log (this->name, GF_LOG_ERROR, "%s/%s"
-                                        "gfid not present", loc->path,
+                                        " gfid not present", loc->path,
                                          entry->d_name);
                                 continue;
                         }
@@ -1361,7 +1466,7 @@ gf_defrag_fix_layout (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
                         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,
+                                        " gfid not present", loc->path,
                                          entry->d_name);
                                 continue;
                         }
@@ -1383,6 +1488,7 @@ gf_defrag_fix_layout (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
                                         "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,
@@ -1391,6 +1497,7 @@ gf_defrag_fix_layout (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
                         if (ret) {
                                 gf_log (this->name, GF_LOG_ERROR, "Fix layout "
                                         "failed for %s", entry_loc.path);
+                                defrag->total_failures++;
                                 goto out;
                         }
 
@@ -1398,6 +1505,8 @@ gf_defrag_fix_layout (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
                 gf_dirent_free (&entries);
                 free_entries = _gf_false;
                 INIT_LIST_HEAD (&entries.list);
+                if (readdirp_errno == ENOENT)
+                        break;
         }
 
         ret = 0;
@@ -1481,6 +1590,7 @@ gf_defrag_start_crawl (void *data)
         if (ret) {
                 gf_log (this->name, GF_LOG_ERROR, "fix layout on %s failed",
                         loc.path);
+                defrag->total_failures++;
                 goto out;
         }
 
@@ -1587,6 +1697,7 @@ gf_defrag_status_get (gf_defrag_info_t *defrag, dict_t *dict)
         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,};
@@ -1603,6 +1714,7 @@ gf_defrag_status_get (gf_defrag_info_t *defrag, dict_t *dict)
         size   = defrag->total_data;
         lookup = defrag->num_files_lookedup;
         failures = defrag->total_failures;
+        skipped = defrag->skipped;
 
         gettimeofday (&end, NULL);
 
@@ -1626,6 +1738,7 @@ gf_defrag_status_get (gf_defrag_info_t *defrag, dict_t *dict)
                 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,
@@ -1638,6 +1751,14 @@ gf_defrag_status_get (gf_defrag_info_t *defrag, dict_t *dict)
         }
 
         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:
@@ -1655,13 +1776,15 @@ log:
         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, files, size,
-                lookup, failures);
+                PRIu64", lookups: %"PRIu64", failures: %"PRIu64", skipped: "
+                "%"PRIu64, files, size, lookup, failures, skipped);
 
 
 out:
@@ -1680,7 +1803,7 @@ gf_defrag_stop (gf_defrag_info_t *defrag, dict_t *output)
                 goto out;
         }
 
-        gf_log ("", GF_LOG_INFO, "Recieved stop command on rebalance");
+        gf_log ("", GF_LOG_INFO, "Received stop command on rebalance");
         defrag->defrag_status = GF_DEFRAG_STATUS_STOPPED;
 
         if (output)
diff --git a/xlators/cluster/dht/src/dht-rename.c b/xlators/cluster/dht/src/dht-rename.c
index f1a4364df..5d6f4f232 100644
--- a/xlators/cluster/dht/src/dht-rename.c
+++ b/xlators/cluster/dht/src/dht-rename.c
@@ -306,7 +306,38 @@ err:
                           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,
@@ -340,11 +371,7 @@ dht_rename_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         WIPE (&local->postparent);
 
         if (is_last_call (this_call_cnt)) {
-                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);
+                dht_rename_done (frame, this);
         }
 
 out:
@@ -362,7 +389,7 @@ dht_rename_cleanup (call_frame_t *frame)
         xlator_t    *dst_hashed = NULL;
         xlator_t    *dst_cached = NULL;
         int          call_cnt = 0;
-
+        dict_t      *xattr      = NULL;
 
         local = frame->local;
         this  = frame->this;
@@ -386,13 +413,15 @@ dht_rename_cleanup (call_frame_t *frame)
         if (!call_cnt)
                 goto nolinks;
 
+        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, 0, NULL);
+                            &local->loc, 0, xattr);
         }
 
         if (src_cached != dst_hashed) {
@@ -401,9 +430,12 @@ dht_rename_cleanup (call_frame_t *frame)
                         local->loc2.path, src_cached->name);
                 STACK_WIND (frame, dht_rename_unlink_cbk,
                             src_cached, src_cached->fops->unlink,
-                            &local->loc2, 0, NULL);
+                            &local->loc2, 0, xattr);
         }
 
+        if (xattr)
+                dict_unref (xattr);
+
         return 0;
 
 nolinks:
@@ -441,6 +473,10 @@ dht_rename_links_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         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;
@@ -463,6 +499,7 @@ dht_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         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;
@@ -472,6 +509,8 @@ dht_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         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,
@@ -501,15 +540,25 @@ dht_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 uuid_copy (link_local->gfid, local->loc.inode->gfid);
 
                 dht_linkfile_create (link_frame, dht_rename_links_create_cbk,
-                                     src_cached, dst_hashed, &link_local->loc);
+                                     this, src_cached, dst_hashed,
+                                     &link_local->loc);
         }
 
 err:
-        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);
+        /* 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);
+        }
+
 
         /* NOTE: rename_subvol is the same subvolume from which dht_rename_cbk
          *       is called. since rename has already happened on rename_subvol,
@@ -534,6 +583,8 @@ err:
         if (local->call_cnt == 0)
                 goto unwind;
 
+        DHT_MARK_FOP_INTERNAL (xattr);
+
         if (src_cached != dst_hashed && src_cached != dst_cached) {
                 gf_log (this->name, GF_LOG_TRACE,
                         "deleting old src datafile %s @ %s",
@@ -541,7 +592,7 @@ err:
 
                 STACK_WIND (frame, dht_rename_unlink_cbk,
                             src_cached, src_cached->fops->unlink,
-                            &local->loc, 0, NULL);
+                            &local->loc, 0, xattr);
         }
 
         if (src_hashed != rename_subvol && src_hashed != src_cached) {
@@ -551,7 +602,7 @@ err:
 
                 STACK_WIND (frame, dht_rename_unlink_cbk,
                             src_hashed, src_hashed->fops->unlink,
-                            &local->loc, 0, NULL);
+                            &local->loc, 0, xattr);
         }
 
         if (dst_cached
@@ -563,8 +614,10 @@ err:
 
                 STACK_WIND (frame, dht_rename_unlink_cbk,
                             dst_cached, dst_cached->fops->unlink,
-                            &local->loc2, 0, NULL);
+                            &local->loc2, 0, xattr);
         }
+        if (xattr)
+                dict_unref (xattr);
         return 0;
 
 unwind:
@@ -572,16 +625,16 @@ unwind:
         WIPE (&local->postoldparent);
         WIPE (&local->preparent);
         WIPE (&local->postparent);
+        if (xattr)
+                dict_unref (xattr);
 
-        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);
+        dht_rename_done (frame, this);
 
         return 0;
 
 cleanup:
+        if (xattr)
+                dict_unref (xattr);
         dht_rename_cleanup (frame);
 
         return 0;
@@ -615,6 +668,8 @@ dht_do_rename (call_frame_t *frame)
                 "renaming %s => %s (%s)",
                 local->loc.path, local->loc2.path, rename_subvol->name);
 
+        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);
@@ -645,6 +700,9 @@ dht_rename_links_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 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);
@@ -710,6 +768,7 @@ dht_rename_create_links (call_frame_t *frame)
         xlator_t    *dst_hashed = NULL;
         xlator_t    *dst_cached = NULL;
         int          call_cnt = 0;
+        dict_t      *xattr = NULL;
 
 
         local = frame->local;
@@ -720,6 +779,7 @@ dht_rename_create_links (call_frame_t *frame)
         dst_hashed = local->dst_hashed;
         dst_cached = local->dst_cached;
 
+        DHT_MARK_FOP_INTERNAL (xattr);
 
         if (src_cached == dst_cached) {
                 if (dst_hashed == dst_cached)
@@ -731,7 +791,7 @@ dht_rename_create_links (call_frame_t *frame)
 
 		STACK_WIND (frame, dht_rename_unlink_links_cbk,
 			    dst_hashed, dst_hashed->fops->unlink,
-			    &local->loc2, 0, NULL);
+			    &local->loc2, 0, xattr);
                 return 0;
         }
 
@@ -748,7 +808,7 @@ dht_rename_create_links (call_frame_t *frame)
                         "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,
+		dht_linkfile_create (frame, dht_rename_links_cbk, this,
 				     src_cached, dst_hashed, &local->loc);
 	}
 
@@ -758,7 +818,7 @@ 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, NULL);
+			    &local->loc, &local->loc2, xattr);
 	}
 
 nolinks:
@@ -766,6 +826,8 @@ nolinks:
                 /* skip to next step */
                 dht_do_rename (frame);
         }
+        if (xattr)
+                dict_unref (xattr);
 
         return 0;
 }
diff --git a/xlators/cluster/dht/src/dht-selfheal.c b/xlators/cluster/dht/src/dht-selfheal.c
index 8b7d15ce5..3fe96b1c7 100644
--- a/xlators/cluster/dht/src/dht-selfheal.c
+++ b/xlators/cluster/dht/src/dht-selfheal.c
@@ -17,6 +17,7 @@
 #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;                            \
@@ -28,42 +29,40 @@
                         layout->list[i].xlator->name, path);            \
         } while (0)
 
-static inline uint32_t
-dht_find_overlap (int idx, int cnk_idx, uint32_t start, uint32_t stop,
-                  uint32_t chunk_size)
-{
-        uint32_t overlap = 0;
-        uint32_t chunk_begin = 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)
 
-        chunk_begin = cnk_idx * chunk_size;
+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;
 
-        /* There is no chance of overlap */
-        if ((chunk_begin > stop) ||
-            ((chunk_begin + chunk_size) < start))
-                goto out;
+	if (old->list[o].err > 0 || new->list[n].err > 0)
+		return 0;
 
-        if ((chunk_begin <= start) &&
-            ((chunk_begin + chunk_size) <= stop)) {
-                overlap = ((chunk_begin + chunk_size) - start);
-                goto out;
+        if (old->list[o].start == old->list[o].stop) {
+                return 0;
         }
 
-        if ((chunk_begin <= start) &&
-            ((chunk_begin + chunk_size) >= stop)) {
-                overlap = (stop - start);
-                goto out;
+        if (new->list[n].start == new->list[n].stop) {
+                return 0;
         }
 
-        if ((chunk_begin < stop) &&
-            ((chunk_begin + chunk_size) >= stop)) {
-                overlap = (stop - chunk_begin);
-                goto out;
-        }
+	if ((old->list[o].start > new->list[n].stop) ||
+	    (old->list[o].stop < new->list[n].start))
+		return 0;
 
-out:
-        return overlap;
+	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)
 {
@@ -118,7 +117,8 @@ dht_selfheal_dir_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 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;
@@ -126,16 +126,22 @@ dht_selfheal_dir_xattr_persubvol (call_frame_t *frame, loc_t *loc,
         xlator_t          *this = NULL;
         int32_t           *disk_layout = NULL;
         dht_local_t       *local = NULL;
-
+        dht_conf_t        *conf = NULL;
 
         local = frame->local;
-        subvol = layout->list[i].xlator;
+        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) {
@@ -150,8 +156,7 @@ dht_selfheal_dir_xattr_persubvol (call_frame_t *frame, loc_t *loc,
                 goto err;
         }
 
-        ret = dict_set_bin (xattr, "trusted.glusterfs.dht",
-                            disk_layout, 4 * 4);
+        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",
@@ -196,21 +201,42 @@ dht_fix_dir_xattr (call_frame_t *frame, loc_t *loc, dht_layout_t *layout)
         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 = layout->cnt;
+        local->call_cnt = count = conf->subvolume_cnt;
 
         for (i = 0; i < layout->cnt; i++) {
-                dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i);
+                dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i, NULL);
 
                 if (--count == 0)
-                        break;
+                        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;
 }
 
@@ -221,9 +247,12 @@ dht_selfheal_dir_xattr (call_frame_t *frame, loc_t *loc, dht_layout_t *layout)
         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) {
@@ -252,11 +281,23 @@ dht_selfheal_dir_xattr (call_frame_t *frame, loc_t *loc, dht_layout_t *layout)
                 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;
         }
+        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;
 }
 
@@ -493,7 +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);
         }
@@ -516,7 +557,7 @@ dht_get_layout_count (xlator_t *this, dht_layout_t *layout, int new_layout)
                 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;
+                                layout->list[i].err = EINVAL;
                                 break;
                         }
                 }
@@ -524,9 +565,33 @@ dht_get_layout_count (xlator_t *this, dht_layout_t *layout, int new_layout)
 
         for (i = 0; i < layout->cnt; i++) {
                 err = layout->list[i].err;
-                if (err == -1 || err == 0) {
-                        layout->list[i].err = -1;
+                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;
                 }
         }
 
@@ -541,35 +606,118 @@ dht_get_layout_count (xlator_t *this, dht_layout_t *layout, int new_layout)
                 }
         }
 
-        count = ((layout->spread_cnt) ? layout->spread_cnt :
-                 ((count) ? count : 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));
 
         return count;
 }
 
 
+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;
+                                }
+                        }
+                }
+
+		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)
 {
-        uint32_t      chunk        = 0;
-        uint32_t      start        = 0;
-        uint32_t      stop         = 0;
-        uint32_t      overlap      = 0;
-        uint32_t      max_overlap  = 0;
-        uint32_t      chunk_begin  = 0;
-        int           count        = 0;
-        int           cnt          = 0;
         int           i            = 0;
-        int           j            = 0;
-        int           k            = 0;
-        int           loop_cnt     = 0;
-        int           start_subvol = 0;
-        int          *fix_array    = NULL;
         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;
@@ -581,114 +729,37 @@ dht_fix_layout_of_directory (call_frame_t *frame, loc_t *loc,
                 goto done;
         }
 
-        count = cnt = dht_get_layout_count (this, layout, 0);
-
-        chunk = ((unsigned long) 0xffffffff) / ((cnt) ? cnt : 1);
-
-        start_subvol = dht_selfheal_layout_alloc_start (this, loc, layout);
-
-        fix_array = GF_CALLOC (sizeof (int), layout->cnt, gf_common_mt_char);
-        if (!fix_array) {
-                /* No fix, use the existing layout itself */
-                goto done;
-        }
-
         new_layout = dht_layout_new (this, priv->subvolume_cnt);
         if (!new_layout)
                 goto done;
 
-        for (i = 0; i < new_layout->cnt; i++) {
-                /* TODO: fix this in layout_alloc() itself */
-                new_layout->list[i].err = -ENOENT;
-                if (i < layout->cnt)
-                        new_layout->list[i].xlator = layout->list[i].xlator;
-        }
-
-        /* Check if there are any overlap in layout, and give the proper fix */
-        for (i = 0; i < layout->cnt; i++) {
-                /* No need to fix if 'err' is not '-1' */
-                if (layout->list[i].err != -1)
-                        continue;
-
-                /* If already existing layout is having no range, skip it */
-                start = layout->list[i].start;
-                stop  = layout->list[i].stop;
-                if ((stop - start) == 0)
-                        continue;
-
-                max_overlap = 0;
-
-                /* 'j' is used as starting point of each chunk */
-                for (j = 1; j <= count; j++) {
-                        /* if chunk is already used, don't use it again */
-                        for (k = 0; k < i; k++)
-                                if (j == fix_array[k])
-                                        break;
-                        if (k < i)
-                                continue;
-
-                        overlap = dht_find_overlap (i, (j-1), start, stop, chunk);
-                        if (max_overlap < overlap) {
-                                max_overlap = overlap;
-                                fix_array[i] = j;
-                        }
-                }
-
-                /* If we have any overlap, then use that itself as new
-                   layout for the subvolume */
-                if (fix_array[i]) {
-                        chunk_begin = chunk * (fix_array[i] - 1);
-                        new_layout->list[i].err = -1;
-                        DHT_SET_LAYOUT_RANGE (new_layout, i, chunk_begin,
-                                              chunk, cnt, loc->path);
-                        /* make sure to give (max - 1) as 'stop' range,
-                           if it is last chunk */
-                        if (fix_array[i] == count)
-                                new_layout->list[i].stop = 0xffffffff;
-                        if (--cnt == 0)
-                                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;
         }
 
-        /* Now, look for layouts which are not having any overlaps
-           and give it a fix */
-        for (loop_cnt = 0, i = start_subvol; loop_cnt < new_layout->cnt;
-             i++, loop_cnt++) {
-                if (i == new_layout->cnt)
-                        i = 0;
-
-                /* If 'fix_array[i]' is set, the layout is already fixed. */
-                if (fix_array[i])
-                        continue;
+        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;
 
-                if (layout->list[i].err != -1) {
-                        new_layout->list[i].err = layout->list[i].err;
-                        continue;
-                }
+		new_layout->list[i].xlator = layout->list[i].xlator;
+        }
 
-                for (k = 1; k <= count; k++) {
-                        for (j = 0; j < new_layout->cnt; j++) {
-                                if (k == fix_array[j])
-                                        break;
-                        }
-                        /* Didn't find any of the list begining with 'k' */
-                        if (j == new_layout->cnt)
-                                break;
-                }
+	/* 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);
 
-                fix_array[i] = k;
-                chunk_begin = (k - 1) * chunk;
-                new_layout->list[i].err = -1;
-                DHT_SET_LAYOUT_RANGE (new_layout, i, chunk_begin, chunk, cnt,
-                                      loc->path);
-                /* make sure to give (max - 1) as 'stop' range,
-                   if it is last chunk */
-                if (k == count)
-                        new_layout->list[i].stop = 0xffffffff;
-                if (--cnt == 0)
-                        goto done;
-        }
+	/* Now selectively re-assign ranges only when it helps */
+	dht_selfheal_layout_maximize_overlap (frame, loc, new_layout, layout);
 
 done:
         if (new_layout) {
@@ -702,8 +773,6 @@ done:
                 local->layout = new_layout;
         }
 
-        GF_FREE (fix_array);
-
         return local->layout;
 }
 
@@ -728,9 +797,11 @@ dht_selfheal_layout_new_directory (call_frame_t *frame, loc_t *loc,
 
         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) {
+                if (err == -1 || err == ENOENT) {
                         DHT_SET_LAYOUT_RANGE(layout, i, start, chunk,
                                              cnt, loc->path);
                         if (--cnt == 0) {
@@ -743,7 +814,7 @@ dht_selfheal_layout_new_directory (call_frame_t *frame, loc_t *loc,
 
         for (i = 0; i < start_subvol; i++) {
                 err = layout->list[i].err;
-                if (err == -1) {
+                if (err == -1 || err == ENOENT) {
                         DHT_SET_LAYOUT_RANGE(layout, i, start, chunk,
                                              cnt, loc->path);
                         if (--cnt == 0) {
@@ -762,35 +833,17 @@ int
 dht_selfheal_dir_getafix (call_frame_t *frame, loc_t *loc,
                           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;
+        uint32_t     holes = 0;
         int          ret = -1;
         int          i = -1;
-        int          overlaps = -1;
+        uint32_t     overlaps = 0;
 
-        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;
-        }
-
-        if (holes <= down) {
-                /* the down subvol might fill up the holes */
-                ret = 0;
-        }
-
         if (holes || overlaps) {
                 dht_selfheal_layout_new_directory (frame, loc, layout);
                 ret = 0;
@@ -842,6 +895,9 @@ dht_fix_directory_layout (call_frame_t *frame,
 
         /* 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);
 
         return 0;
@@ -864,9 +920,8 @@ dht_selfheal_directory (call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
         dht_layout_anomalies (this, loc, layout,
                               &local->selfheal.hole_cnt,
                               &local->selfheal.overlaps_cnt,
-                              &local->selfheal.missing,
-                              &local->selfheal.down,
-                              &local->selfheal.misc);
+                              NULL, &local->selfheal.down,
+                              &local->selfheal.misc, NULL);
 
         down     = local->selfheal.down;
         misc     = local->selfheal.misc;
@@ -925,3 +980,50 @@ dht_selfheal_restore (call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
 
         return ret;
 }
+
+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;
+}
+
+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 d8aa33233..fc0ca2f77 100644
--- a/xlators/cluster/dht/src/dht.c
+++ b/xlators/cluster/dht/src/dht.c
@@ -14,487 +14,15 @@
 #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;
-
-        GF_VALIDATE_OR_GOTO ("dht", layout, out);
-        GF_VALIDATE_OR_GOTO ("dht", prefix, 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);
-        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);
-                }
-        }
-
-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;
-
-        GF_VALIDATE_OR_GOTO ("dht", this, out);
-
-        conf = this->private;
-
-        if (!conf)
-                return -1;
-
-        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", "%lu", conf->min_free_disk);
-	gf_proc_dump_write("min_free_inodes", "%lu", 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;
-        uint64_t        tmp_layout = 0;
-
-        GF_VALIDATE_OR_GOTO ("dht", this, out);
-        GF_VALIDATE_OR_GOTO ("dht", inode, out);
-
-        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_add_section("xlator.cluster.dht.%s.inode", this->name);
-        dht_layout_dump(layout, "layout");
-
-out:
-        return ret;
-}
-
-int
-notify (xlator_t *this, int event, void *data, ...)
-{
-        int              ret = -1;
-        va_list          ap;
-        dict_t          *output = NULL;
-
-        GF_VALIDATE_OR_GOTO ("dht", this, out);
-
-
-        if (!data)
-                goto out;
-
-        va_start (ap, data);
-        output = va_arg (ap, dict_t*);
-
-        ret = dht_notify (this, event, data, output);
-
-out:
-        return ret;
-}
-
-void
-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];
-                                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
-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);
-	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;
-        }
-
-        ret = 0;
-out:
-        return ret;
-}
-
-
-int
-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);
-        }
-
-        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;
-        }
-
-        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;
-        }
-
-        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);
-        }
-
-        return -1;
-}
-
+class_methods_t class_methods = {
+        .init           = dht_init,
+        .fini           = dht_fini,
+        .reconfigure    = dht_reconfigure,
+        .notify         = dht_notify
+};
 
 struct xlator_fops fops = {
         .lookup      = dht_lookup,
@@ -542,6 +70,9 @@ struct xlator_fops fops = {
         .fxattrop    = dht_fxattrop,
         .setattr     = dht_setattr,
         .fsetattr    = dht_fsetattr,
+	.fallocate   = dht_fallocate,
+	.discard     = dht_discard,
+        .zerofill    = dht_zerofill,
 };
 
 struct xlator_dumpops dumpops = {
@@ -555,59 +86,4 @@ struct xlator_cbks cbks = {
 //      .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,
-          .default_value = "on",
-        },
-        { .key  = {"min-free-disk"},
-          .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
-          .default_value = "10%",
-          .description = "Percentage/Size of disk space that must be "
-                         "kept free."
-        },
-	{ .key  = {"min-free-inodes"},
-          .type = GF_OPTION_TYPE_PERCENT,
-          .default_value = "5%",
-          .description = "Percentage inodes that must be "
-                         "kept free."
-        },
-        { .key = {"unhashed-sticky-bit"},
-          .type = GF_OPTION_TYPE_BOOL,
-          .default_value = "off",
-        },
-        { .key = {"use-readdirp"},
-          .type = GF_OPTION_TYPE_BOOL,
-          .default_value = "on",
-        },
-        { .key = {"assert-no-child-down"},
-          .type = GF_OPTION_TYPE_BOOL,
-          .default_value = "off",
-        },
-        { .key  = {"directory-layout-spread"},
-          .type = GF_OPTION_TYPE_INT,
-        },
-        { .key  = {"decommissioned-bricks"},
-          .type = GF_OPTION_TYPE_ANY,
-        },
-        { .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",
-        },
-        { .key = {"readdir-optimize"},
-          .type = GF_OPTION_TYPE_BOOL,
-          .default_value = "off",
-        },
-
-        { .key  = {NULL} },
-};
+;
diff --git a/xlators/cluster/dht/src/nufa.c b/xlators/cluster/dht/src/nufa.c
index 53c66aa45..e934acdf0 100644
--- a/xlators/cluster/dht/src/nufa.c
+++ b/xlators/cluster/dht/src/nufa.c
@@ -18,6 +18,8 @@
 
 /* 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,
@@ -52,7 +54,8 @@ nufa_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         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) {
@@ -201,7 +204,7 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
                  *       revalidates directly go to the cached-subvolume.
                  */
                 ret = dict_set_uint32 (local->xattr_req,
-                                       "trusted.glusterfs.dht", 4 * 4);
+                                       conf->xattr_name, 4 * 4);
                 if (ret < 0) {
                         gf_log (this->name, GF_LOG_ERROR,
                                 "Failed to set dict value.");
@@ -222,7 +225,7 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
         } else {
         do_fresh_lookup:
                 ret = dict_set_uint32 (local->xattr_req,
-                                       "trusted.glusterfs.dht", 4 * 4);
+                                       conf->xattr_name, 4 * 4);
                 if (ret < 0) {
                         gf_log (this->name, GF_LOG_ERROR,
                                 "Failed to set dict value.");
@@ -231,7 +234,7 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
                 }
 
                 ret = dict_set_uint32 (local->xattr_req,
-                                       "trusted.glusterfs.dht.linkto", 256);
+                                       conf->link_xattr_name, 256);
                 if (ret < 0) {
                         gf_log (this->name, GF_LOG_ERROR,
                                 "Failed to set dict value.");
@@ -320,7 +323,8 @@ nufa_create (call_frame_t *frame, xlator_t *this,
         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) {
@@ -330,9 +334,8 @@ nufa_create (call_frame_t *frame, xlator_t *this,
                 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;
         }
 
@@ -362,17 +365,22 @@ nufa_mknod_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         dht_local_t  *local = NULL;
 
         local = frame->local;
+        if (!local || !local->cached_subvol) {
+                op_errno = EINVAL;
+                op_ret = -1;
+                goto err;
+        }
 
         if (op_ret >= 0) {
-                STACK_WIND (frame, dht_newfile_cbk,
-                            local->cached_subvol,
+                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);
 
@@ -420,7 +428,8 @@ nufa_mknod (call_frame_t *frame, xlator_t *this,
         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) {
@@ -432,7 +441,7 @@ nufa_mknod (call_frame_t *frame, xlator_t *this,
                 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;
         }
@@ -440,9 +449,9 @@ nufa_mknod (call_frame_t *frame, xlator_t *this,
         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, umask, params);
+        STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol, subvol,
+                           subvol->fops->mknod, loc, mode, rdev, umask,
+                           params);
 
         return 0;
 
@@ -455,189 +464,170 @@ err:
 }
 
 
-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)
-{
-        int         i = 0;
-        dht_conf_t *conf = NULL;
+typedef struct nufa_args {
+        xlator_t    *this;
+        char        *volname;
+        gf_boolean_t addr_match;
+} nufa_args_t;
 
-        conf = this->private;
+static void
+nufa_find_local_brick (xlator_t *xl, void *data)
+{
+        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;
+        }
 
-        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 (!addr_match)
+                return;
 
-                GF_FREE (conf->subvolumes);
+        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;
+        }
 
-                GF_FREE (conf->subvolume_status);
+}
 
-                GF_FREE (conf);
-        }
+static void
+nufa_to_dht (xlator_t *this)
+{
+        GF_ASSERT (this);
+        GF_ASSERT (this->fops);
 
-        return;
+        this->fops->lookup = dht_lookup;
+        this->fops->create = dht_create;
+        this->fops->mknod  = dht_mknod;
 }
 
 int
-init (xlator_t *this)
+nufa_find_local_subvol (xlator_t *this,
+                        void (*fn) (xlator_t *each, void* data), void *data)
 {
-        dht_conf_t    *conf = NULL;
-        xlator_list_t *trav = NULL;
-        data_t        *data = NULL;
-        char          *local_volname = NULL;
-        char          *temp_str = 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");
+        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;
         }
 
-        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;
-        }
-
-        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;
-        }
+        candidate = conf->private;
+        trav = candidate->parents;
+        while (trav) {
 
-        ret = dht_init_subvolumes (this, conf);
-        if (ret == -1) {
-                goto err;
-        }
+                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;
+                }
 
-        ret = dht_layouts_init (this, conf);
-        if (ret == -1) {
-                goto err;
+                candidate = parent;
+                trav = parent->parents;
         }
 
-        LOCK_INIT (&conf->subvolume_lock);
-        LOCK_INIT (&conf->layout_lock);
+        return ret;
+}
 
-        conf->gen = 1;
+int
+nufa_init (xlator_t *this)
+{
+        data_t        *data = NULL;
+        char          *local_volname = NULL;
+        int            ret = -1;
+        char           my_hostname[256];
+        gf_boolean_t   addr_match = _gf_false;
+        nufa_args_t    args = {0, };
 
-        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));
+        ret = dht_init(this);
+        if (ret) {
+                return ret;
         }
 
-        if (ret == 0)
-                local_volname = my_hostname;
-
-        data = dict_get (this->options, "local-volume-name");
-        if (data) {
+        if ((data = dict_get (this->options, "local-volume-name"))) {
                 local_volname = data->data;
-        }
 
-        trav = this->children;
-        while (trav) {
-                if (strcmp (trav->xlator->name, local_volname) == 0)
-                        break;
-                trav = trav->next;
-        }
+        } else {
+                addr_match = _gf_true;
+                local_volname = "localhost";
+                ret = gethostname (my_hostname, 256);
+                if (ret == 0)
+                        local_volname = my_hostname;
 
-        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';
-                }
-        }
+                else
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "could not find hostname (%s)",
+                                strerror (errno));
 
-        conf->du_stats = GF_CALLOC (conf->subvolume_cnt, sizeof (dht_du_t),
-                                    gf_dht_mt_dht_du_t);
-        if (!conf->du_stats) {
-                goto err;
         }
 
-        this->local_pool = mem_pool_new (dht_local_t, 128);
-        if (!this->local_pool) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "failed to create local_t's memory pool");
-                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);
-                }
-
-                GF_FREE (conf->subvolumes);
-
-                GF_FREE (conf->subvolume_status);
-
-                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 = {
@@ -684,19 +674,3 @@ struct xlator_fops fops = {
 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} },
-};
diff --git a/xlators/cluster/dht/src/switch.c b/xlators/cluster/dht/src/switch.c
index 0542d7f9a..d3ea90ba8 100644
--- a/xlators/cluster/dht/src/switch.c
+++ b/xlators/cluster/dht/src/switch.c
@@ -22,6 +22,8 @@
 #include <fnmatch.h>
 #include <string.h>
 
+extern struct volume_options options[];
+
 struct switch_sched_array {
         xlator_t *xl;
         int32_t   eligible;
@@ -135,7 +137,8 @@ switch_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         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) {
@@ -289,11 +292,11 @@ switch_lookup (call_frame_t *frame, xlator_t *this,
                  * attribute, revalidates directly go to the cached-subvolume.
                  */
                 ret = dict_set_uint32 (local->xattr_req,
-                                       "trusted.glusterfs.dht", 4 * 4);
+                                       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;
@@ -308,18 +311,18 @@ switch_lookup (call_frame_t *frame, xlator_t *this,
         } else {
         do_fresh_lookup:
                 ret = dict_set_uint32 (local->xattr_req,
-                                       "trusted.glusterfs.dht", 4 * 4);
+                                       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);
 
                 ret = dict_set_uint32 (local->xattr_req,
-                                       "trusted.glusterfs.dht.linkto", 256);
+                                       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");
+                                "failed to set dict value for %s",
+                                conf->link_xattr_name);
 
                 if (!hashed_subvol) {
                         gf_log (this->name, GF_LOG_DEBUG,
@@ -434,7 +437,8 @@ switch_create (call_frame_t *frame, xlator_t *this,
         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) {
@@ -443,9 +447,8 @@ switch_create (call_frame_t *frame, xlator_t *this,
                 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;
         }
 
@@ -475,17 +478,22 @@ switch_mknod_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         dht_local_t  *local = NULL;
 
         local = frame->local;
+        if (!local || !local->cached_subvol) {
+                op_errno = EINVAL;
+                op_ret = -1;
+                goto err;
+        }
 
         if (op_ret >= 0) {
-                STACK_WIND (frame, dht_newfile_cbk,
-                            local->cached_subvol,
+                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:
         DHT_STACK_UNWIND (link, frame, op_ret, op_errno,
                           inode, stbuf, preparent, postparent, xdata);
         return 0;
@@ -529,7 +537,8 @@ switch_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
         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) {
@@ -542,16 +551,16 @@ switch_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
                 local->cached_subvol = avail_subvol;
 
                 dht_linkfile_create (frame, switch_mknod_linkfile_cbk,
-                                     avail_subvol, subvol, loc);
+                                     this, avail_subvol, subvol, loc);
                 return 0;
         }
 
         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, umask, params);
+        STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol, subvol,
+                           subvol->fops->mknod, loc, mode, rdev, umask,
+                           params);
 
         return 0;
 
@@ -564,20 +573,9 @@ err:
 }
 
 
-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;
@@ -593,22 +591,9 @@ fini (xlator_t *this)
                         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);
-                }
-
-                GF_FREE (conf->subvolumes);
-
-                GF_FREE (conf->subvolume_status);
-
-                GF_FREE (conf);
         }
 
-        return;
+        dht_fini(this);
 }
 
 int
@@ -834,68 +819,18 @@ 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;
         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) {
-                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) {
@@ -906,60 +841,23 @@ init (xlator_t *this)
                 }
         }
 
-        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) {
-                goto err;
-        }
-
-        this->local_pool = mem_pool_new (dht_local_t, 128);
-        if (!this->local_pool) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "failed to create local_t's memory pool");
-                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);
-        }
-
+        dht_fini(this);
         return -1;
 }
 
 
+class_methods_t class_methods = {
+        .init           = switch_init,
+        .fini           = switch_fini,
+        .reconfigure    = dht_reconfigure,
+        .notify         = dht_notify
+};
+
+
 struct xlator_fops fops = {
         .lookup      = switch_lookup,
         .create      = switch_create,
@@ -1004,19 +902,3 @@ struct xlator_fops fops = {
 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  = {"pattern.switch.case"},
-          .type = GF_OPTION_TYPE_ANY
-        },
-        { .key  = {"min-free-disk"},
-          .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
-        },
-        { .key  = {NULL} },
-};
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 a68f7ddfd..19be1ed27 100644
--- a/xlators/cluster/ha/src/ha-helpers.c
+++ b/xlators/cluster/ha/src/ha-helpers.c
@@ -1,13 +1,12 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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 931bd33a4..e5e97d237 100644
--- a/xlators/cluster/ha/src/ha-mem-types.h
+++ b/xlators/cluster/ha/src/ha-mem-types.h
@@ -1,14 +1,13 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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 9a3e60889..3eccb516b 100644
--- a/xlators/cluster/ha/src/ha.c
+++ b/xlators/cluster/ha/src/ha.c
@@ -1,13 +1,12 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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
  */
diff --git a/xlators/cluster/ha/src/ha.h b/xlators/cluster/ha/src/ha.h
index e732cfd08..e2ed7eaa6 100644
--- a/xlators/cluster/ha/src/ha.h
+++ b/xlators/cluster/ha/src/ha.h
@@ -1,13 +1,12 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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 2c360dc8b..851397b68 100644
--- a/xlators/cluster/map/src/map-helper.c
+++ b/xlators/cluster/map/src/map-helper.c
@@ -1,13 +1,12 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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 7abb40b28..3e89f4736 100644
--- a/xlators/cluster/map/src/map-mem-types.h
+++ b/xlators/cluster/map/src/map-mem-types.h
@@ -1,14 +1,13 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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 78b356554..6150a33ce 100644
--- a/xlators/cluster/map/src/map.c
+++ b/xlators/cluster/map/src/map.c
@@ -1,13 +1,12 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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.h b/xlators/cluster/map/src/map.h
index b87112eb7..7703a543e 100644
--- a/xlators/cluster/map/src/map.h
+++ b/xlators/cluster/map/src/map.h
@@ -1,13 +1,12 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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 8c48d3410..2d151422a 100644
--- a/xlators/cluster/stripe/src/Makefile.am
+++ b/xlators/cluster/stripe/src/Makefile.am
@@ -2,7 +2,7 @@
 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
@@ -11,9 +11,10 @@ stripe_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
 noinst_HEADERS = stripe.h stripe-mem-types.h $(top_builddir)/xlators/lib/src/libxlator.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/lib/src
 
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
 CLEANFILES = 
 
diff --git a/xlators/cluster/stripe/src/stripe-helpers.c b/xlators/cluster/stripe/src/stripe-helpers.c
index e85661c14..a83abdc72 100644
--- a/xlators/cluster/stripe/src/stripe-helpers.c
+++ b/xlators/cluster/stripe/src/stripe-helpers.c
@@ -12,6 +12,7 @@
 
 #include "stripe.h"
 #include "byte-order.h"
+#include "mem-types.h"
 
 void
 stripe_local_wipe (stripe_local_t *local)
@@ -158,6 +159,92 @@ stripe_free_xattr_str (stripe_local_t *local)
 
 
 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)
 {
diff --git a/xlators/cluster/stripe/src/stripe-mem-types.h b/xlators/cluster/stripe/src/stripe-mem-types.h
index e05ba0c29..e9ac9cf46 100644
--- a/xlators/cluster/stripe/src/stripe-mem-types.h
+++ b/xlators/cluster/stripe/src/stripe-mem-types.h
@@ -20,6 +20,7 @@ enum gf_stripe_mem_types_ {
         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,
diff --git a/xlators/cluster/stripe/src/stripe.c b/xlators/cluster/stripe/src/stripe.c
index 76d912a09..69b510e23 100644
--- a/xlators/cluster/stripe/src/stripe.c
+++ b/xlators/cluster/stripe/src/stripe.c
@@ -3774,6 +3774,524 @@ err:
 
 
 int32_t
+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);
+
+        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;
+
+        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;
+        }
+        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;
@@ -3809,6 +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;
@@ -3820,30 +4339,34 @@ notify (xlator_t *this, int32_t event, void *data, ...)
         switch (event)
         {
         case GF_EVENT_CHILD_UP:
-        case GF_EVENT_CHILD_CONNECTING:
         {
                 /* get an index number to set */
                 for (i = 0; i < priv->child_count; i++) {
                         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 */
@@ -3851,20 +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);
         }
@@ -3874,10 +4396,30 @@ 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;
 }
 
@@ -3923,6 +4465,37 @@ stripe_setxattr_cbk (call_frame_t *frame, void *cookie,
         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)
@@ -3932,6 +4505,7 @@ stripe_setxattr (call_frame_t *frame, xlator_t *this,
         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);
@@ -3954,11 +4528,15 @@ stripe_setxattr (call_frame_t *frame, xlator_t *this,
         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.
+         * this power is only given to a special client. Bd xlator
+         * also needs xattrs for regular files (ie LVs)
          */
-        if ((frame->root->pid == -1) && IA_ISDIR (loc->inode->ia_type)) {
+        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,
@@ -3987,12 +4565,111 @@ stripe_fsetxattr_cbk (call_frame_t *frame, void *cookie,
         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;
+                }
+        }
+        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;
+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;
-        int32_t         op_errno = EINVAL;
+        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);
@@ -4001,12 +4678,25 @@ stripe_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
         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:
+err:
         STRIPE_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, NULL);
         return 0;
 }
@@ -4486,9 +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;
@@ -4589,6 +5279,7 @@ fini (xlator_t *this)
                         trav = trav->next;
                         GF_FREE (prev);
                 }
+                GF_FREE (priv->last_event);
                 LOCK_DESTROY (&priv->lock);
                 GF_FREE (priv);
         }
@@ -4687,8 +5378,8 @@ stripe_vgetxattr_cbk (call_frame_t *frame, void *cookie,
         int32_t              callcnt       = 0;
         int32_t              ret           = -1;
         long                 cky           = 0;
-        char                *xattr_val     = NULL;
-        char                *xattr_serz    = NULL;
+        void                *xattr_val     = NULL;
+        void                *xattr_serz    = NULL;
         stripe_xattr_sort_t *xattr         = NULL;
         dict_t              *stripe_xattr  = NULL;
 
@@ -4719,18 +5410,20 @@ stripe_vgetxattr_cbk (call_frame_t *frame, void *cookie,
                                            gf_stripe_mt_xattr_sort_t);
 
                 if (local->xattr_list) {
-                        ret = dict_get_str (dict, local->xsel, &xattr_val);
-                        if (ret)
-                                goto out;
-
                         xattr = local->xattr_list + (int32_t) cky;
 
-                        xattr_val = gf_strdup (xattr_val);
+                        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 = xattr_val;
-                        xattr->xattr_len = strlen (xattr_val);
+                        xattr->xattr_value = gf_memdup (xattr_val,
+                                                        xattr->xattr_len);
 
-                        local->xattr_total_len += xattr->xattr_len + 1;
+                        if (xattr->xattr_value != NULL)
+                                local->xattr_total_len += xattr->xattr_len + 1;
                 }
         }
  out:
@@ -4747,19 +5440,24 @@ stripe_vgetxattr_cbk (call_frame_t *frame, void *cookie,
                 /* 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 {
+                } else {
                         gf_log (this->name, GF_LOG_WARNING,
                                 "Unknown xattr in xattr request");
                         goto unwind;
                 }
 
                 if (!ret) {
-                        ret = dict_set_dynstr (stripe_xattr, local->xsel,
-                                               xattr_serz);
+                        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);
+                                        "Can't set %s key in dict",
+                                        local->xsel);
                 }
 
         unwind:
@@ -4810,7 +5508,7 @@ stripe_getxattr (call_frame_t *frame, xlator_t *this,
 
 
         if (name && (strcmp (GF_XATTR_MARKER_KEY, name) == 0)
-            && (-1 == frame->root->pid)) {
+            && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
                 local->marker.call_count = priv->child_count;
 
                 sub_volumes = alloca ( priv->child_count *
@@ -4825,7 +5523,8 @@ stripe_getxattr (call_frame_t *frame, xlator_t *this,
                 if (cluster_getmarkerattr (frame, this, loc, name,
                                            local, stripe_getxattr_unwind,
                                            sub_volumes, priv->child_count,
-                                           MARKER_UUID_TYPE, priv->vol_uuid)) {
+                                           MARKER_UUID_TYPE, marker_uuid_default_gauge,
+                                           priv->vol_uuid)) {
                         op_errno = EINVAL;
                         goto err;
                 }
@@ -4881,7 +5580,7 @@ stripe_getxattr (call_frame_t *frame, xlator_t *this,
 
         if (name &&(*priv->vol_uuid)) {
                 if ((match_uuid_local (name, priv->vol_uuid) == 0)
-                    && (-1 == frame->root->pid)) {
+                    && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
 
                         if (!IA_FILE_OR_DIR (loc->inode->ia_type))
                                 local->marker.call_count = 1;
@@ -4904,6 +5603,7 @@ stripe_getxattr (call_frame_t *frame, xlator_t *this,
                                                    sub_volumes,
                                                    local->marker.call_count,
                                                    MARKER_XTIME_TYPE,
+                                                   marker_xtime_default_gauge,
                                                    priv->vol_uuid)) {
                                 op_errno = EINVAL;
                                 goto err;
@@ -4924,6 +5624,79 @@ err:
         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
@@ -5000,9 +5773,13 @@ struct xlator_fops fops = {
         .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 = {
@@ -5028,10 +5805,10 @@ struct volume_options options[] = {
         },
 	{ .key = {"coalesce"},
 	  .type = GF_OPTION_TYPE_BOOL,
-	  .default_value = "false",
-	  .description = "Enable coalesce mode to flatten striped files as "
-			 "stored on the server (i.e., eliminate holes caused "
-			 "by the traditional format)."
+	  .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 a440f87ba..5673d18f3 100644
--- a/xlators/cluster/stripe/src/stripe.h
+++ b/xlators/cluster/stripe/src/stripe.h
@@ -73,9 +73,9 @@
        } while (0)
 
 typedef struct stripe_xattr_sort {
-        int32_t  pos;
-        int32_t  xattr_len;
-        char    *xattr_value;
+        int   pos;
+        int   xattr_len;
+        char *xattr_value;
 } stripe_xattr_sort_t;
 
 /**
@@ -98,8 +98,8 @@ 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];
@@ -262,6 +262,9 @@ int32_t stripe_xattr_aggregate (char *buffer, stripe_local_t *local,
 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.
diff --git a/xlators/debug/error-gen/src/Makefile.am b/xlators/debug/error-gen/src/Makefile.am
index df9080358..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 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
index 3926b707e..f02280535 100644
--- a/xlators/debug/error-gen/src/error-gen-mem-types.h
+++ b/xlators/debug/error-gen/src/error-gen-mem-types.h
@@ -1,14 +1,13 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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__
 
diff --git a/xlators/debug/error-gen/src/error-gen.c b/xlators/debug/error-gen/src/error-gen.c
index 4a1890ec2..ec0874b35 100644
--- a/xlators/debug/error-gen/src/error-gen.c
+++ b/xlators/debug/error-gen/src/error-gen.c
@@ -1,13 +1,12 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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"
@@ -15,6 +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,
@@ -82,9 +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,
@@ -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;
 }
@@ -1105,10 +1108,26 @@ 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) {
+	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;
+        	return 0;
 	}
 
 	STACK_WIND (frame, error_gen_writev_cbk,
@@ -1935,6 +1954,85 @@ error_gen_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
 	return 0;
 }
 
+static void
+error_gen_set_failure (eg_t *pvt, int percent)
+{
+        GF_ASSERT (pvt);
+
+        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)
+{
+        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;
+
+        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)
 {
@@ -1955,18 +2053,43 @@ mem_acct_init (xlator_t *this)
 }
 
 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;
-        gf_boolean_t    random_failure = _gf_false;
         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) {
@@ -1981,79 +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 = 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);
 
-        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;
-                        }
-                }
-        }
+        GF_OPTION_INIT ("enable", error_enable_fops, str, out);
 
-        random_failure = dict_get_str_boolean (this->options, "random-failure",
-                                               _gf_false);
-        pvt->random_failure = random_failure;
+        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);
 
         this->private = pvt;
 
         /* Give some seed value here */
         srand (time(NULL));
 out:
+        if (ret)
+                GF_FREE (pvt);
         return ret;
 }
 
@@ -2075,9 +2153,12 @@ fini (xlator_t *this)
         return;
 }
 
-struct xlator_fops cbks = {
+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,
@@ -2125,17 +2206,27 @@ struct xlator_fops fops = {
 
 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},
+          .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 81baef58e..d92c23062 100644
--- a/xlators/debug/error-gen/src/error-gen.h
+++ b/xlators/debug/error-gen/src/error-gen.h
@@ -1,13 +1,12 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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
 
@@ -20,6 +19,18 @@
 
 #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;
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 80c6b5604..c30dfb17e 100644
--- a/xlators/debug/io-stats/src/io-stats-mem-types.h
+++ b/xlators/debug/io-stats/src/io-stats-mem-types.h
@@ -1,14 +1,13 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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__
 
diff --git a/xlators/debug/io-stats/src/io-stats.c b/xlators/debug/io-stats/src/io-stats.c
index 82159fcac..7fb697ae4 100644
--- a/xlators/debug/io-stats/src/io-stats.c
+++ b/xlators/debug/io-stats/src/io-stats.c
@@ -1,13 +1,12 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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"
@@ -1088,9 +1087,10 @@ io_stats_dump_stats_to_dict (xlator_t *this, dict_t *resp,
                                 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);
+                                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)
@@ -1724,6 +1724,40 @@ io_stats_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 
 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 gf_flock *lock, dict_t *xdata)
 {
@@ -2166,7 +2200,6 @@ conditional_dump (dict_t *dict, char *key, data_t *value, void *data)
                         return -1;
                 }
                 logfp = fopen (filename, "w+");
-                GF_ASSERT (logfp);
                 if (!logfp) {
                         gf_log (this->name, GF_LOG_ERROR, "failed to open %s "
                                 "for writing", filename);
@@ -2393,6 +2426,45 @@ io_stats_fstat (call_frame_t *frame, xlator_t *this,
 
 
 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 gf_flock *lock, dict_t *xdata)
 {
@@ -2818,6 +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 = {
@@ -2861,7 +2936,7 @@ struct volume_options options[] = {
           .type = GF_OPTION_TYPE_STR,
           .default_value = "CRITICAL",
           .description = "Gluster's syslog log-level",
-          .value = { "WARNING", "ERROR", "CRITICAL"}
+          .value = { "WARNING", "ERROR", "INFO", "CRITICAL"}
         },
         { .key = {"brick-log-level"},
           .type = GF_OPTION_TYPE_STR,
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 5c0ad1a4c..c9d839356 100644
--- a/xlators/debug/trace/src/trace.c
+++ b/xlators/debug/trace/src/trace.c
@@ -1,5 +1,5 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  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
@@ -8,10 +8,8 @@
   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 :
@@ -20,858 +18,977 @@
  *    Very helpful translator for debugging.
  */
 
-#include <time.h>
-#include <errno.h>
-#include "glusterfs.h"
-#include "xlator.h"
-#include "common-utils.h"
-
-
-struct {
-        char *name;
-        int enabled;
-} trace_fop_names[GF_FOP_MAXVALUE];
-
-int trace_log_level = GF_LOG_INFO;
-
-static char *
-trace_stat_to_str (struct iatt *buf)
+int
+dump_history_trace (circular_buffer_t *cb, void *data)
 {
-        char     *statstr           = NULL;
-        char      atime_buf[64]     = {0,};
-        char      mtime_buf[64]     = {0,};
-        char      ctime_buf[64]     = {0,};
-        int       asprint_ret_value = 0;
+        char    *string      = NULL;
+        struct  tm      *tm  = NULL;
+        char    timestr[256] = {0,};
 
-        if (!buf) {
-                statstr = NULL;
-                goto out;
-        }
+        string = (char *)cb->data;
+        tm    = localtime (&cb->tv.tv_sec);
 
-        gf_time_fmt (atime_buf, sizeof atime_buf, buf->ia_atime, gf_timefmt_bdT);
-        gf_time_fmt (mtime_buf, sizeof mtime_buf, buf->ia_mtime, gf_timefmt_bdT);
-        gf_time_fmt (ctime_buf, sizeof ctime_buf, buf->ia_ctime, gf_timefmt_bdT);
-        asprint_ret_value = gf_asprintf (&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);
+        /* 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);
 
-out:
-        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, dict_t *xdata)
+                  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_INFO,
-                                "%"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);
-
-                        GF_FREE (statstr);
-                        GF_FREE (preparentstr);
-                        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_INFO,
-                                "%"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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
-                             preparent, postparent, xdata);
+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, 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_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d, op_errno=%d, *fd=%p",
-                        frame->root->unique, uuid_utoa (frame->local), 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);
         }
 
+out:
         /* for 'release' log */
         if (op_ret >= 0)
                 fd_ctx_set (fd, this, 0);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
+        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, dict_t *xdata)
+                int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                dict_t *xdata)
 {
-        char *statstr = NULL;
-        if (trace_fop_names[GF_FOP_STAT].enabled) {
-                if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        gf_log (this->name, GF_LOG_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d buf=%s",
-                                frame->root->unique, uuid_utoa (frame->local),
-                                op_ret, statstr);
+        char          statstr[4096] = {0, };
+        trace_conf_t  *conf         = NULL;
+
+        conf = this->private;
 
-                        GF_FREE (statstr);
+        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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, op_errno=%d)",
-                                frame->root->unique, uuid_utoa (frame->local), 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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (stat, frame, op_ret, op_errno, buf, xdata);
+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, dict_t *xdata)
+                 int32_t count, struct iatt *buf, struct iobref *iobref,
+                 dict_t *xdata)
 {
-        char  *statstr = NULL;
+        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) {
-                        statstr = trace_stat_to_str (buf);
-                        gf_log (this->name, GF_LOG_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d buf=%s",
-                                frame->root->unique, uuid_utoa (frame->local),
-                                op_ret, statstr);
-
-                        GF_FREE (statstr);
+                        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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, op_errno=%d)",
-                                frame->root->unique, uuid_utoa (frame->local),
-                                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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector, count,
-                             buf, iobref, xdata);
+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, 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_INFO,
-                                "%"PRId64": (op_ret=%d, *prebuf = {%s}, "
-                                "*postbuf = {%s})",
-                                frame->root->unique, op_ret,
-                                preopstr, postopstr);
-
-                        GF_FREE (preopstr);
-
-                        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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                                frame->root->unique, uuid_utoa (frame->local),
-                                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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+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, dict_t *xdata)
+                   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_INFO,
-                        "%"PRId64" : gfid=%s op_ret=%d, op_errno=%d",
-                        frame->root->unique, uuid_utoa (frame->local), 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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, buf, xdata);
+                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, dict_t *xdata)
+                    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_INFO,
-                        "%"PRId64" : gfid=%s op_ret=%d, op_errno=%d",
-                        frame->root->unique, uuid_utoa (frame->local), 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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, buf, xdata);
+                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, 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);
-
-                        gf_log (this->name, GF_LOG_INFO,
-                                "%"PRId64": (op_ret=%d, *prebuf = {%s}, "
-                                "*postbuf = {%s}",
-                                frame->root->unique, op_ret,
-                                preopstr, postopstr);
+        char          preopstr[4096]  = {0, };
+        char          postopstr[4096] = {0, };
+        trace_conf_t  *conf           = NULL;
 
-                        GF_FREE (preopstr);
+        conf = this->private;
 
-                        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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                                frame->root->unique, uuid_utoa (frame->local),
-                                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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+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, dict_t *xdata)
 {
-        char  *preopstr = NULL;
-        char  *postopstr = NULL;
-
-        if (trace_fop_names[GF_FOP_SETATTR].enabled) {
-                if (op_ret >= 0) {
-                        preopstr = trace_stat_to_str (statpre);
-                        postopstr = trace_stat_to_str (statpost);
-
-                        gf_log (this->name, GF_LOG_INFO,
-                                "%"PRId64": (op_ret=%d, *prebuf = {%s}, "
-                                "*postbuf = {%s})",
-                                frame->root->unique, op_ret,
-                                preopstr, postopstr);
+        char          preopstr[4096]  = {0, };
+        char          postopstr[4096] = {0, };
+        trace_conf_t  *conf           = NULL;
 
-                        GF_FREE (preopstr);
+        conf = this->private;
 
-                        GF_FREE (postopstr);
+        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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, op_errno=%d)",
-                                frame->root->unique, uuid_utoa (frame->local), 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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (setattr, frame, op_ret, op_errno, statpre, statpost, xdata);
+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, dict_t *xdata)
 {
-        char  *preopstr = NULL;
-        char  *postopstr = NULL;
-
-        if (trace_fop_names[GF_FOP_FSETATTR].enabled) {
-                if (op_ret >= 0) {
-                        preopstr = trace_stat_to_str (statpre);
-                        postopstr = trace_stat_to_str (statpost);
-
-                        gf_log (this->name, GF_LOG_INFO,
-                                "%"PRId64": (op_ret=%d, *prebuf = {%s}, "
-                                "*postbuf = {%s})",
-                                frame->root->unique, op_ret,
-                                preopstr, postopstr);
+        char          preopstr[4096]  = {0, };
+        char          postopstr[4096] = {0, };
+        trace_conf_t  *conf           = NULL;
 
-                        GF_FREE (preopstr);
+        conf = this->private;
 
-                        GF_FREE (postopstr);
+        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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, op_errno=%d)",
-                                frame->root->unique, uuid_utoa (frame->local), 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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (fsetattr, frame, op_ret, op_errno,
-                             statpre, statpost, xdata);
+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, dict_t *xdata)
+                  struct iatt *preparent, struct iatt *postparent,
+                  dict_t *xdata)
 {
-        char *preparentstr = NULL;
-        char *postparentstr = NULL;
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = 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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, *preparent = {%s}, "
-                                "*postparent = {%s})",
-                                frame->root->unique, uuid_utoa (frame->local), op_ret, preparentstr,
-                                postparentstr);
+        conf = this->private;
 
-                        GF_FREE (preparentstr);
-
-                        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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, op_errno=%d)",
-                                frame->root->unique, uuid_utoa (frame->local), 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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno,
-                             preparent, postparent, xdata);
+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, dict_t *xdata)
+                  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_INFO,
-                                "%"PRId64": (op_ret=%d, *stbuf = {%s}, "
-                                "*preoldparent = {%s}, *postoldparent = {%s}"
-                                " *prenewparent = {%s}, *postnewparent = {%s})",
-                                frame->root->unique, op_ret, statstr,
-                                preoldparentstr, postoldparentstr,
-                                prenewparentstr, postnewparentstr);
-
-                        GF_FREE (statstr);
+        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;
 
-                        GF_FREE (preoldparentstr);
+        conf = this->private;
 
-                        GF_FREE (postoldparentstr);
-
-                        GF_FREE (prenewparentstr);
-
-                        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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                                frame->root->unique, uuid_utoa (frame->local),
-                                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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (rename, frame, op_ret, op_errno, buf,
-                             preoldparent, postoldparent,
-                             prenewparent, postnewparent, xdata);
+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, 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_INFO,
-                                "%"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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                                frame->root->unique, uuid_utoa (frame->local),
-                                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);
+                }
 
-                GF_FREE (statstr);
+                LOG_ELEMENT (conf, string);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, buf, stbuf, xdata);
+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 *xdata, struct iatt *postparent)
 {
-        char  *statstr = NULL;
-        char  *postparentstr = NULL;
+        char          statstr[4096]       = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
 
-        if (trace_fop_names[GF_FOP_LOOKUP].enabled) {
-                if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        postparentstr = trace_stat_to_str (postparent);
+        conf = this->private;
 
-                        gf_log (this->name, GF_LOG_INFO,
-                                "%"PRId64": gfid=%s (op_ret=%d "
-                                "*buf {%s}, *postparent {%s}",
-                                frame->root->unique, uuid_utoa (inode->gfid),
-                                op_ret, statstr, postparentstr);
-
-                        GF_FREE (statstr);
-                        GF_FREE (postparentstr);
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_LOOKUP].enabled) {
+                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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, op_errno=%d)",
-                                frame->root->unique, uuid_utoa (frame->local),
-                                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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
-                             xdata, 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, dict_t *xdata)
+                   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);
-
-                        gf_log (this->name, GF_LOG_INFO,
-                                "%"PRId64": gfid=%s (op_ret=%d "
-                                "*stbuf = {%s}, *preparent = {%s}, "
-                                "*postparent = {%s})",
-                                frame->root->unique, uuid_utoa (inode->gfid),
-                                op_ret, statstr, preparentstr, postparentstr);
-
-                        GF_FREE (statstr);
-
-                        GF_FREE (preparentstr);
+        char          statstr[4096]       = {0, };
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
 
-                        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_INFO,
-                                "%"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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent, xdata);
+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, 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);
-
-                        gf_log (this->name, GF_LOG_INFO,
-                                "%"PRId64": gfid=%s (op_ret=%d "
-                                "*stbuf = {%s}, *preparent = {%s}, "
-                                "*postparent = {%s})",
-                                frame->root->unique, uuid_utoa (inode->gfid),
-                                op_ret, statstr, preparentstr, postparentstr);
-
-                        GF_FREE (statstr);
+        char          statstr[4096]       = {0, };
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
 
-                        GF_FREE (preparentstr);
+        conf = this->private;
 
-                        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_INFO,
-                                "%"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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (mknod, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent, xdata);
+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, 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_INFO,
-                                "%"PRId64": gfid=%s (op_ret=%d "
-                                ", *stbuf = {%s}, *prebuf = {%s}, "
-                                "*postbuf = {%s} )",
-                                frame->root->unique, uuid_utoa (inode->gfid),
-                                op_ret, statstr, preparentstr, postparentstr);
-
-                        GF_FREE (statstr);
+        char          statstr[4096]       = {0, };
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
 
-                        GF_FREE (preparentstr);
+        conf = this->private;
 
-                        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_INFO,
-                                "%"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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent, xdata);
+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, 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_INFO,
-                                "%"PRId64": (op_ret=%d, *stbuf = {%s}, "
-                                " *prebuf = {%s}, *postbuf = {%s})",
-                                frame->root->unique, op_ret,
-                                statstr, preparentstr, postparentstr);
+        char          statstr[4096]       = {0, };
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
 
-                        GF_FREE (statstr);
+        conf = this->private;
 
-                        GF_FREE (preparentstr);
-
-                        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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                                frame->root->unique, uuid_utoa (frame->local),
-                                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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent, xdata);
+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, 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_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                        frame->root->unique, uuid_utoa (frame->local),
-                        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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno, xdata);
+                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, 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_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d, op_errno=%d, fd=%p",
-                        frame->root->unique, uuid_utoa (frame->local),
-                        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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd, xdata);
+        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, dict_t *xdata)
 {
-        char  *preparentstr = NULL;
-        char  *postparentstr = NULL;
+        char           preparentstr[4096]  = {0, };
+        char           postparentstr[4096] = {0, };
+        trace_conf_t   *conf               = 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);
+        conf = this->private;
 
-                        gf_log (this->name, GF_LOG_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, *prebuf = {%s}, "
-                                "*postbuf = {%s}",
-                                frame->root->unique, uuid_utoa (frame->local),
-                                op_ret, preparentstr,  postparentstr);
-
-                        GF_FREE (preparentstr);
-
-                        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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                                frame->root->unique, uuid_utoa (frame->local),
-                                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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno,
-                             preparent, postparent, xdata);
+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, dict_t *xdata)
 {
-        char  *preopstr = NULL;
-        char  *postopstr = 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);
+        char           preopstr[4096]  = {0, };
+        char           postopstr[4096] = {0, };
+        trace_conf_t   *conf           = NULL;
 
-                        gf_log (this->name, GF_LOG_INFO,
-                                "%"PRId64": (op_ret=%d, *prebuf = {%s}, "
-                                "*postbuf = {%s} )",
-                                frame->root->unique, op_ret, preopstr,
-                                postopstr);
+        conf = this->private;
 
-                        GF_FREE (preopstr);
-
-                        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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                                frame->root->unique, uuid_utoa (frame->local),
-                                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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+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, dict_t *xdata)
+                  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_INFO,
-                                "%"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_INFO,
-                                "%"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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, buf, xdata);
+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, 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_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                        frame->root->unique, uuid_utoa (frame->local),
-                        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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
+                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, dict_t *xdata)
+                    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_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d, op_errno=%d, dict=%p",
-                        frame->root->unique, uuid_utoa (frame->local), 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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
 
         return 0;
 }
@@ -880,32 +997,49 @@ 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) {
-                gf_log (this->name, GF_LOG_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                        frame->root->unique, uuid_utoa (frame->local), 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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
+                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)
+                     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) {
-                gf_log (this->name, GF_LOG_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d, op_errno=%d, dict=%p",
-                        frame->root->unique, uuid_utoa (frame->local), 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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict, xdata);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, dict, xdata);
 
         return 0;
 }
@@ -914,160 +1048,208 @@ int
 trace_removexattr_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_REMOVEXATTR].enabled) {
-                gf_log (this->name, GF_LOG_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                        frame->root->unique, uuid_utoa (frame->local),
-                        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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
+                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, 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_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                        frame->root->unique, uuid_utoa (frame->local),
-                        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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (fsyncdir, frame, op_ret, op_errno, xdata);
+                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, 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_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d, op_errno=%d)",
-                        frame->root->unique, uuid_utoa (frame->local),
-                        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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (access, frame, op_ret, op_errno, xdata);
+                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, 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_INFO,
-                                "%"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;
 
-                        GF_FREE (prebufstr);
-
-                        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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                                frame->root->unique, uuid_utoa (frame->local),
-                                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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+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, dict_t *xdata)
 {
-        char *statstr = NULL;
+        char          statstr[4096] = {0, };
+        trace_conf_t  *conf         = NULL;
 
-        if (trace_fop_names[GF_FOP_FSTAT].enabled) {
-                if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        gf_log (this->name, GF_LOG_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d buf=%s",
-                                frame->root->unique, uuid_utoa (frame->local),
-                                op_ret, statstr);
+        conf = this->private;
 
-                        GF_FREE (statstr);
+        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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                                frame->root->unique, uuid_utoa (frame->local),
-                                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);
         }
-
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, buf, xdata);
+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 gf_flock *lock, dict_t *xdata)
+              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_INFO,
-                                "%"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);
+                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_INFO,
-                                "%"PRId64": gfid=%s op_ret=%d, op_errno=%d)",
-                                frame->root->unique, uuid_utoa (frame->local),
-                                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);
                 }
-        }
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, lock, xdata);
+                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, 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_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                        frame->root->unique, uuid_utoa (frame->local),
-                        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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno, xdata);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (entrylk, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
@@ -1075,66 +1257,97 @@ 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) {
-                gf_log (this->name, GF_LOG_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                        frame->root->unique, uuid_utoa (frame->local),
-                        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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (fentrylk, frame, op_ret, op_errno, xdata);
+                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, dict_t *xdata)
+                   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_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                        frame->root->unique, uuid_utoa (frame->local),
-                        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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (xattrop, frame, op_ret, op_errno, dict, xdata);
+                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, dict_t *xdata)
+                    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_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                        frame->root->unique, uuid_utoa (frame->local),
-                        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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (fxattrop, frame, op_ret, op_errno, dict, xdata);
+                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, 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_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                        frame->root->unique, uuid_utoa (frame->local),
-                        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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno, xdata);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (inodelk, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
@@ -1142,32 +1355,51 @@ 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) {
-                gf_log (this->name, GF_LOG_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
-                        frame->root->unique, uuid_utoa (frame->local), 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);
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (finodelk, frame, op_ret, op_errno, xdata);
+                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)
+                     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) {
-                gf_log (this->name, GF_LOG_INFO,
-                        "%"PRId64": gfid=%s op_ret=%d op_errno=%d",
-                        frame->root->unique, uuid_utoa (frame->local), 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);
         }
 
-        frame->local = NULL;
-        STACK_UNWIND_STRICT (rchecksum, frame, op_ret, op_errno, weak_checksum,
-                             strong_checksum, xdata);
+out:
+        TRACE_STACK_UNWIND (rchecksum, frame, op_ret, op_errno, weak_checksum,
+                            strong_checksum, xdata);
 
         return 0;
 }
@@ -1179,17 +1411,31 @@ trace_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)
 {
+        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_INFO,
-                        "%"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"));
+                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,
@@ -1197,15 +1443,20 @@ trace_entrylk (call_frame_t *frame, xlator_t *this,
         return 0;
 }
 
-
 int
 trace_inodelk (call_frame_t *frame, xlator_t *this, const char *volume,
                loc_t *loc, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
 {
-        char *cmd_str = NULL;
-        char *type_str = NULL;
+        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) {
+                char string[4096]  = {0,};
                 switch (cmd) {
 #if F_GETLK != F_GETLK64
                 case F_GETLK64:
@@ -1248,17 +1499,22 @@ trace_inodelk (call_frame_t *frame, xlator_t *this, const char *volume,
                         break;
                 }
 
-                gf_log (this->name, GF_LOG_INFO,
-                        "%"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);
+                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,
@@ -1266,14 +1522,20 @@ trace_inodelk (call_frame_t *frame, xlator_t *this, const char *volume,
         return 0;
 }
 
-
 int
 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, *type_str = NULL;
+        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) {
+                char  string[4096] = {0,};
                 switch (cmd) {
 #if F_GETLK != F_GETLK64
                 case F_GETLK64:
@@ -1316,16 +1578,21 @@ trace_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
                         break;
                 }
 
-                gf_log (this->name, GF_LOG_INFO,
-                        "%"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);
+                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,
@@ -1333,19 +1600,30 @@ trace_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
         return 0;
 }
 
-
 int
 trace_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
                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_INFO,
-                        "%"PRId64": gfid=%s (path=%s flags=%d)",
-                        frame->root->unique, uuid_utoa (loc->inode->gfid),
-                        loc->path, 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,
@@ -1354,19 +1632,29 @@ trace_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
         return 0;
 }
 
-
 int
 trace_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
                 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_INFO,
-                        "%"PRId64": gfid=%s fd=%p, flags=%d",
-                        frame->root->unique, uuid_utoa (fd->inode->gfid),
-                        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,
@@ -1375,20 +1663,30 @@ trace_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
         return 0;
 }
 
-
 int
 trace_lookup (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_LOOKUP].enabled) {
+                char string[4096] = {0,};
                 /* TODO: print all the keys mentioned in xattr_req */
-                gf_log (this->name, GF_LOG_INFO,
-                        "%"PRId64": gfid=%s path=%s",
-                        frame->root->unique, uuid_utoa (loc->inode->gfid),
-                        loc->path);
+                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,
@@ -1397,18 +1695,28 @@ trace_lookup (call_frame_t *frame, xlator_t *this,
         return 0;
 }
 
-
 int
 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_INFO,
-                        "%"PRId64": gfid=%s path=%s",
-                        frame->root->unique, uuid_utoa (loc->inode->gfid),
-                        loc->path);
+                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,
@@ -1417,18 +1725,30 @@ trace_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
         return 0;
 }
 
-
 int
-trace_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size, dict_t *xdata)
+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_INFO,
-                        "%"PRId64": gfid=%s path=%s, size=%"GF_PRI_SIZET")",
-                        frame->root->unique, uuid_utoa (loc->inode->gfid),
-                        loc->path, 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,
@@ -1437,19 +1757,29 @@ trace_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size, di
         return 0;
 }
 
-
 int
 trace_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
              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_INFO,
-                        "%"PRId64": gfid=%s path=%s mode=0%o umask=0%o "
-                        "dev=%"GF_PRI_DEV")",
-                        frame->root->unique, uuid_utoa (loc->inode->gfid),
-                        loc->path, mode, umask, 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,
@@ -1458,18 +1788,28 @@ trace_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
         return 0;
 }
 
-
 int
 trace_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
              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_INFO,
-                        "%"PRId64": gfid=%s path=%s mode=0%o umask=0%o",
-                        frame->root->unique, uuid_utoa (loc->inode->gfid),
-                        loc->path, mode, umask);
+                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,
@@ -1477,19 +1817,29 @@ trace_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
         return 0;
 }
 
-
 int
 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_INFO,
-                        "%"PRId64": gfid=%s path=%s flag=%d",
-                        frame->root->unique, uuid_utoa (loc->inode->gfid),
-                        loc->path, xflag);
+                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,
@@ -1497,19 +1847,30 @@ trace_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
         return 0;
 }
 
-
 int
 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_INFO,
-                        "%"PRId64": gfid=%s path=%s flags=%d",
-                        frame->root->unique, uuid_utoa (loc->inode->gfid),
-                        loc->path, 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_rmdir_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->rmdir,
@@ -1518,18 +1879,28 @@ trace_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
         return 0;
 }
 
-
 int
 trace_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
                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_INFO,
-                        "%"PRId64": gfid=%s linkpath=%s, path=%s umask=0%o",
-                        frame->root->unique, uuid_utoa (loc->inode->gfid),
-                        linkpath, loc->path, umask);
+                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,
@@ -1538,15 +1909,20 @@ trace_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
         return 0;
 }
 
-
 int
 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,};
+        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) {
+                char string[4096] = {0,};
                 if (newloc->inode)
                         uuid_utoa_r (newloc->inode->gfid, newgfid);
                 else
@@ -1554,13 +1930,18 @@ trace_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
 
                 uuid_utoa_r (oldloc->inode->gfid, oldgfid);
 
-                gf_log (this->name, GF_LOG_INFO,
-                        "%"PRId64": oldgfid=%s oldpath=%s --> newgfid=%s newpath=%s",
-                        frame->root->unique, oldgfid, oldloc->path, newgfid, newloc->path);
+                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,
@@ -1569,15 +1950,20 @@ trace_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
         return 0;
 }
 
-
 int
 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,};
+        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) {
+                char string[4096]  = {0,};
                 if (newloc->inode)
                         uuid_utoa_r (newloc->inode->gfid, newgfid);
                 else
@@ -1585,13 +1971,18 @@ trace_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
 
                 uuid_utoa_r (oldloc->inode->gfid, oldgfid);
 
-                gf_log (this->name, GF_LOG_INFO,
-                        "%"PRId64": oldgfid=%s oldpath=%s --> newgfid=%s newpath=%s",
-                        frame->root->unique, oldgfid, oldloc->path,
-                        newgfid, newloc->path);
+                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,
@@ -1599,43 +1990,69 @@ trace_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
         return 0;
 }
 
-
 int
 trace_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
                struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
-        char      actime_str[64]   = {0,};
-        char      modtime_str[64]  = {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_INFO,
-                                "%"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));
+                        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_INFO,
-                                "%"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);
+                        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)) {
-                        gf_time_fmt (actime_str, sizeof actime_str,
-                                     stbuf->ia_atime, gf_timefmt_bdT);
-                        gf_time_fmt (modtime_str, sizeof modtime_str,
-                                     stbuf->ia_mtime, gf_timefmt_bdT);
-
-                        gf_log (this->name, GF_LOG_INFO,
-                                "%"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);
+                        ia_time = stbuf->ia_atime;
+                        strftime (actime_str, 256, "[%b %d %H:%M:%S]",
+                                  localtime ((time_t *)&ia_time));
+
+                        ia_time = stbuf->ia_mtime;
+                        strftime (modtime_str, 256, "[%b %d %H:%M:%S]",
+                                  localtime ((time_t *)&ia_time));
+
+                        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,
@@ -1644,43 +2061,67 @@ trace_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
         return 0;
 }
 
-
 int
 trace_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
                 struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
-        char      actime_str[64]  = {0,};
-        char      modtime_str[64] = {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_INFO,
-                                "%"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));
+                        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_INFO,
-                                "%"PRId64": gfid=%s fd=%p, uid=%o, gid=%o",
-                                frame->root->unique, uuid_utoa (fd->inode->gfid),
-                                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)) {
-                        gf_time_fmt (actime_str, sizeof actime_str,
-                                     stbuf->ia_atime, gf_timefmt_bdT);
-                        gf_time_fmt (modtime_str, sizeof modtime_str,
-                                     stbuf->ia_mtime, gf_timefmt_bdT);
-
-                        gf_log (this->name, GF_LOG_INFO,
-                                "%"PRId64": gfid=%s fd=%p ia_atime=%s, ia_mtime=%s",
-                                frame->root->unique, uuid_utoa (fd->inode->gfid),
-                                fd, actime_str, modtime_str);
+                        ia_time = stbuf->ia_atime;
+                        strftime (actime_str, 256, "[%b %d %H:%M:%S]",
+                                  localtime ((time_t *)&ia_time));
+
+                        ia_time = stbuf->ia_mtime;
+                        strftime (modtime_str, 256, "[%b %d %H:%M:%S]",
+                                  localtime ((time_t *)&ia_time));
+
+                        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,
@@ -1689,19 +2130,30 @@ trace_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
         return 0;
 }
 
-
 int
 trace_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc,
                 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_INFO,
-                        "%"PRId64": gfid=%s path=%s, offset=%"PRId64"",
-                        frame->root->unique, uuid_utoa (loc->inode->gfid),
-                        loc->path, 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,
@@ -1710,19 +2162,30 @@ trace_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc,
         return 0;
 }
 
-
 int
 trace_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
             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_INFO,
-                        "%"PRId64": gfid=%s path=%s flags=%d fd=%p",
-                        frame->root->unique, uuid_utoa (loc->inode->gfid),
-                        loc->path, flags, fd);
+                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,
@@ -1730,42 +2193,62 @@ trace_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
         return 0;
 }
 
-
 int
 trace_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)
 {
+        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_INFO,
-                        "%"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);
+                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, umask, fd, xdata);
-
         return 0;
 }
 
-
 int
 trace_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
              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_INFO,
-                        "%"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);
+                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,
@@ -1773,21 +2256,32 @@ trace_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
         return 0;
 }
 
-
 int
 trace_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)
 {
+        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_INFO,
-                        "%"PRId64": gfid=%s fd=%p, count=%d, offset=%"PRId64
-                        " flag=0%x)",
-                        frame->root->unique, uuid_utoa (fd->inode->gfid),
-                        fd, count, offset, flags);
+                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,
@@ -1795,17 +2289,26 @@ trace_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
         return 0;
 }
 
-
 int
 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_INFO,
-                        "%"PRId64": gfid=%s path=%s",
-                        frame->root->unique, (loc->inode)?
-                        uuid_utoa (loc->inode->gfid):"0", loc->path);
+                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,
@@ -1813,17 +2316,28 @@ trace_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
         return 0;
 }
 
-
 int
 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_INFO,
-                        "%"PRId64": gfid=%s fd=%p",
-                        frame->root->unique, uuid_utoa (fd->inode->gfid), 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,
@@ -1831,17 +2345,29 @@ trace_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
         return 0;
 }
 
-
 int
-trace_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags, dict_t *xdata)
+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_INFO,
-                        "%"PRId64": gfid=%s flags=%d fd=%p",
-                        frame->root->unique, uuid_utoa (fd->inode->gfid), 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,
@@ -1849,19 +2375,30 @@ trace_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags, dict_
         return 0;
 }
 
-
 int
 trace_setxattr (call_frame_t *frame, xlator_t *this,
                 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_INFO,
-                        "%"PRId64": gfid=%s path=%s flags=%d",
-                        frame->root->unique, uuid_utoa (loc->inode->gfid),
-                        loc->path, 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,
@@ -1869,19 +2406,30 @@ trace_setxattr (call_frame_t *frame, xlator_t *this,
         return 0;
 }
 
-
 int
 trace_getxattr (call_frame_t *frame, xlator_t *this,
                 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_INFO,
-                        "%"PRId64": gfid=%s path=%s name=%s",
-                        frame->root->unique, uuid_utoa (loc->inode->gfid),
-                        loc->path, 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,
@@ -1889,19 +2437,30 @@ trace_getxattr (call_frame_t *frame, xlator_t *this,
         return 0;
 }
 
-
 int
 trace_removexattr (call_frame_t *frame, xlator_t *this,
                    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_INFO,
-                        "%"PRId64": gfid=%s path=%s name=%s",
-                        frame->root->unique, uuid_utoa (loc->inode->gfid),
-                        loc->path, 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,
@@ -1910,18 +2469,29 @@ trace_removexattr (call_frame_t *frame, xlator_t *this,
         return 0;
 }
 
-
 int
-trace_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd, dict_t *xdata)
+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_INFO,
-                        "%"PRId64": gfid=%s path=%s fd=%p",
-                        frame->root->unique, uuid_utoa (loc->inode->gfid),
-                        loc->path, 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,
@@ -1933,15 +2503,27 @@ int
 trace_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
                 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_INFO,
-                        "%"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);
+                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,
@@ -1950,19 +2532,31 @@ trace_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
         return 0;
 }
 
-
 int
 trace_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
                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_INFO,
-                        "%"PRId64": gfid=%s fd=%p, size=%"GF_PRI_SIZET", offset=%"PRId64,
-                        frame->root->unique, uuid_utoa (fd->inode->gfid),
-                        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,
@@ -1971,19 +2565,29 @@ trace_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
         return 0;
 }
 
-
 int
 trace_fsyncdir (call_frame_t *frame, xlator_t *this,
                 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_INFO,
-                        "%"PRId64": gfid=%s datasync=%d fd=%p",
-                        frame->root->unique, uuid_utoa (fd->inode->gfid),
-                        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,
@@ -1991,18 +2595,30 @@ trace_fsyncdir (call_frame_t *frame, xlator_t *this,
         return 0;
 }
 
-
 int
-trace_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask, dict_t *xdata)
+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_INFO,
-                        "%"PRId64": gfid=%s path=%s mask=0%o",
-                        frame->root->unique, uuid_utoa (loc->inode->gfid),
-                        loc->path, 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,
@@ -2010,19 +2626,30 @@ trace_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask, dic
         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) {
-                gf_log (this->name, GF_LOG_INFO,
-                        "%"PRId64": gfid=%s offset=%"PRId64" len=%u fd=%p",
-                        frame->root->unique, uuid_utoa (fd->inode->gfid),
-                        offset, len, fd);
+                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,
@@ -2037,17 +2664,31 @@ 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) {
-                gf_log (this->name, GF_LOG_INFO,
-                        "%"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"));
+                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,
@@ -2060,14 +2701,25 @@ 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) {
-                gf_log (this->name, GF_LOG_INFO,
-                        "%"PRId64": gfid=%s fd=%p name=%s",
-                        frame->root->unique, uuid_utoa (fd->inode->gfid),
-                        fd, name);
+                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,
@@ -2079,14 +2731,25 @@ 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) {
-                gf_log (this->name, GF_LOG_INFO,
-                        "%"PRId64": gfid=%s fd=%p flags=%d",
-                        frame->root->unique, uuid_utoa (fd->inode->gfid),
-                        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_fsetxattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fsetxattr,
@@ -2098,14 +2761,25 @@ int
 trace_ftruncate (call_frame_t *frame, xlator_t *this,
                  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_INFO,
-                        "%"PRId64": gfid=%s offset=%"PRId64" fd=%p",
-                        frame->root->unique, uuid_utoa (fd->inode->gfid),
-                        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,
@@ -2114,17 +2788,28 @@ trace_ftruncate (call_frame_t *frame, xlator_t *this,
         return 0;
 }
 
-
 int
 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_INFO,
-                        "%"PRId64": gfid=%s fd=%p",
-                        frame->root->unique, uuid_utoa (fd->inode->gfid), 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,
@@ -2132,21 +2817,34 @@ trace_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
         return 0;
 }
 
-
 int
 trace_lk (call_frame_t *frame, xlator_t *this, fd_t *fd,
           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_INFO,
-                        "%"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);
+                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,
@@ -2157,40 +2855,71 @@ trace_lk (call_frame_t *frame, xlator_t *this, fd_t *fd,
 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) {
-                gf_log (this->name, GF_LOG_INFO,
-                        "gfid=%s", uuid_utoa (inode->gfid));
+                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) {
-                gf_log (this->name, GF_LOG_INFO,
-                        "gfid=%s fd=%p", uuid_utoa (fd->inode->gfid), fd);
+                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) {
-                gf_log (this->name, GF_LOG_INFO,
-                        "gfid=%s fd=%p", uuid_utoa (fd->inode->gfid), fd);
+                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
 enable_all_calls (int enabled)
 {
@@ -2200,7 +2929,6 @@ enable_all_calls (int enabled)
                 trace_fop_names[i].enabled = enabled;
 }
 
-
 void
 enable_call (const char *name, int enabled)
 {
@@ -2228,6 +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)
@@ -2235,6 +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;
@@ -2249,6 +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"));
@@ -2257,9 +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] ?
-                                                   (char *)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;
                 }
         }
@@ -2267,14 +3108,43 @@ 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"));
@@ -2282,25 +3152,34 @@ init (xlator_t *this)
                         goto setloglevel;
 
                 if (strcmp (forced_loglevel, "INFO") == 0)
-                        trace_log_level = GF_LOG_INFO;
+                        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
@@ -2309,6 +3188,9 @@ fini (xlator_t *this)
         if (!this)
                 return;
 
+        if (this->history)
+                eh_destroy (this->history);
+
         gf_log (this->name, GF_LOG_INFO,
                 "trace translator unloaded");
         return;
@@ -2357,7 +3239,6 @@ struct xlator_fops fops = {
         .fsetattr    = trace_fsetattr,
 };
 
-
 struct xlator_cbks cbks = {
         .release     = trace_release,
         .releasedir  = trace_releasedir,
@@ -2373,5 +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/encryption/crypt/Makefile.am b/xlators/encryption/crypt/Makefile.am
new file mode 100644
index 000000000..d471a3f92
--- /dev/null
+++ b/xlators/encryption/crypt/Makefile.am
@@ -0,0 +1,3 @@
+SUBDIRS = src
+
+CLEANFILES = 
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 5243af791..b9ac29a72 100644
--- a/xlators/encryption/rot-13/src/rot-13.c
+++ b/xlators/encryption/rot-13/src/rot-13.c
@@ -1,13 +1,12 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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>
 
@@ -23,13 +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;
@@ -62,7 +61,7 @@ rot13_readv_cbk (call_frame_t *frame,
                  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);
 
@@ -129,7 +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;
 	}
@@ -187,15 +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 56b10068c..3e9fc19c7 100644
--- a/xlators/encryption/rot-13/src/rot-13.h
+++ b/xlators/encryption/rot-13/src/rot-13.h
@@ -1,13 +1,12 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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 be48f93fa..d2f5ef192 100644
--- a/xlators/features/Makefile.am
+++ b/xlators/features/Makefile.am
@@ -1,3 +1,4 @@
-SUBDIRS = locks quota read-only mac-compat quiesce marker index # trash 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 =
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/changelog/lib/Makefile.am b/xlators/features/changelog/lib/Makefile.am
new file mode 100644
index 000000000..a985f42a8
--- /dev/null
+++ b/xlators/features/changelog/lib/Makefile.am
@@ -0,0 +1,3 @@
+SUBDIRS = src
+
+CLEANFILES =
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/features/compress/Makefile.am b/xlators/features/compress/Makefile.am
new file mode 100644
index 000000000..a985f42a8
--- /dev/null
+++ b/xlators/features/compress/Makefile.am
@@ -0,0 +1,3 @@
+SUBDIRS = src
+
+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 2620ce4c7..47a17249b 100644
--- a/xlators/features/filter/src/filter-mem-types.h
+++ b/xlators/features/filter/src/filter-mem-types.h
@@ -1,13 +1,12 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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 aa68b0318..1d4887b71 100644
--- a/xlators/features/filter/src/filter.c
+++ b/xlators/features/filter/src/filter.c
@@ -1,13 +1,12 @@
 /*
-  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
-  This file is part of GlusterFS.
+   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.
+   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/gfid-access/Makefile.am b/xlators/features/gfid-access/Makefile.am
new file mode 100644
index 000000000..af437a64d
--- /dev/null
+++ b/xlators/features/gfid-access/Makefile.am
@@ -0,0 +1 @@
+SUBDIRS = src
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/features/glupy/Makefile.am b/xlators/features/glupy/Makefile.am
new file mode 100644
index 000000000..a985f42a8
--- /dev/null
+++ b/xlators/features/glupy/Makefile.am
@@ -0,0 +1,3 @@
+SUBDIRS = src
+
+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);
+        }
+
author	Peter Portante <peter.portante@redhat.com>	2012-10-17 15:25:34 -0400
committer	Anand Avati <avati@redhat.com>	2012-10-17 15:45:56 -0700
commit	2b8043986e6f35a91bf1919c657d73f67b905ef7 (patch)
tree	7a2198fb94fd9cb415ea88b83f3aa8e67b86b55a /contrib/stdlib
parent	dbe793bcde5d0ba936eceb19bcb8a6f376a0dbc4 (diff)