protocol/server: reflect lru limit in inode table also

Upon reconfigure, when lru limit of the inode table is changed, the new value was just saved in the private structure of the protocol/server xlator and the inode table used to have the older values still. A brick start was required for the changes to get reflected. To handle it, traverse through the xlator tree and check whether a xlator is a bound_xl or not (if it is a bound_xl it would have its itable pointer set). If a xlator is a bound_xl, then get the inode table of that bound_xl and set its lru limit to new value given via cli. Also prune the inode table so that extra inodes are purged from the inode table. Change-Id: I6909be028c116adaa1d1a5108470015b5fc6f09d BUG: 1103756 Signed-off-by: Raghavendra Bhat <raghavendra@redhat.com> Reviewed-on: http://review.gluster.org/7957 Tested-by: Gluster Build System <jenkins@build.gluster.com> Reviewed-by: Raghavendra G <rgowdapp@redhat.com> Tested-by: Raghavendra G <rgowdapp@redhat.com>
author: Raghavendra Bhat <raghavendra@redhat.com> 2014-06-03 00:28:08 +0530
committer: Raghavendra G <rgowdapp@redhat.com> 2014-06-13 01:49:10 -0700
commit: 6ba178fd9ebf9fc98415c30bcd338a68ee5eb601 (patch)
tree: b050a02506263651ba24603f25672bfd13d1f0b3 /cli
parent: 4d656f9008747172db52f6d532b610c487528bfb (diff)
0 files changed, 0 insertions, 0 deletions
diff --git a/xlators/cluster/afr/src/Makefile.am b/xlators/cluster/afr/src/Makefile.am
index 4e4b4c752..ea5a90abb 100644
--- a/xlators/cluster/afr/src/Makefile.am
+++ b/xlators/cluster/afr/src/Makefile.am
@@ -1,26 +1,38 @@
 xlator_LTLIBRARIES = afr.la pump.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/cluster
 
-afr_common_source = afr-dir-read.c afr-dir-write.c afr-inode-read.c afr-inode-write.c afr-open.c afr-transaction.c afr-self-heal-data.c afr-self-heal-common.c afr-self-heal-metadata.c afr-self-heal-entry.c afr-self-heal-algorithm.c afr-lk-common.c
+afr_common_source = afr-dir-read.c afr-dir-write.c afr-inode-read.c \
+	afr-inode-write.c afr-open.c afr-transaction.c afr-lk-common.c \
+	afr-read-txn.c \
+	$(top_builddir)/xlators/lib/src/libxlator.c
 
-afr_la_LDFLAGS = -module -avoidversion
-afr_la_SOURCES = $(afr_common_source) afr.c
+AFR_SELFHEAL_SOURCES = afr-self-heal-common.c afr-self-heal-data.c \
+	afr-self-heal-entry.c afr-self-heal-metadata.c afr-self-heald.c \
+	afr-self-heal-name.c
+
+afr_la_LDFLAGS = -module -avoid-version
+afr_la_SOURCES = $(afr_common_source) $(AFR_SELFHEAL_SOURCES) afr.c
 afr_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-pump_la_LDFLAGS = -module -avoidversion
-pump_la_SOURCES =  $(afr_common_source) pump.c
+pump_la_LDFLAGS = -module -avoid-version
+pump_la_SOURCES =  $(afr_common_source) $(AFR_SELFHEAL_SOURCES) pump.c
 pump_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-noinst_HEADERS = afr.h afr-transaction.h afr-inode-write.h afr-inode-read.h afr-dir-read.h afr-dir-write.h afr-self-heal.h afr-self-heal-common.h afr-self-heal-algorithm.h pump.h afr-mem-types.h afr-common.c
+noinst_HEADERS = afr.h afr-transaction.h afr-inode-write.h afr-inode-read.h \
+	afr-dir-read.h afr-dir-write.h afr-self-heal.h afr-mem-types.h \
+	afr-common.c afr-self-heald.h pump.h \
+	$(top_builddir)/xlators/lib/src/libxlator.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) \
+	-I$(top_srcdir)/libglusterfs/src -I$(top_srcdir)/xlators/lib/src \
+	-I$(top_srcdir)/rpc/rpc-lib/src
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	    -I$(top_srcdir)/libglusterfs/src -I$(top_srcdir)/contrib/md5 -shared -nostartfiles $(GF_CFLAGS)
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES =
 
 uninstall-local:
 	rm -f $(DESTDIR)$(xlatordir)/replicate.so
-	rm -f $(DESTDIR)$(xlatordir)/pump.so
 
 install-data-hook:
 	ln -sf afr.so $(DESTDIR)$(xlatordir)/replicate.so
diff --git a/xlators/cluster/afr/src/afr-common.c b/xlators/cluster/afr/src/afr-common.c
index 4ef11cb1b..6bd231600 100644
--- a/xlators/cluster/afr/src/afr-common.c
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include <libgen.h>
@@ -44,6 +35,7 @@
 #include "compat.h"
 #include "byte-order.h"
 #include "statedump.h"
+#include "inode.h"
 
 #include "fd.h"
 
@@ -53,308 +45,873 @@
 #include "afr-dir-write.h"
 #include "afr-transaction.h"
 #include "afr-self-heal.h"
-#include "afr-self-heal-common.h"
-#include "pump.h"
+#include "afr-self-heald.h"
 
-#define AFR_ICTX_OPENDIR_DONE_MASK     0x0000000200000000ULL
-#define AFR_ICTX_SPLIT_BRAIN_MASK      0x0000000100000000ULL
-#define AFR_ICTX_READ_CHILD_MASK       0x00000000FFFFFFFFULL
 
-int32_t
-afr_set_dict_gfid (dict_t *dict, uuid_t gfid)
+call_frame_t *
+afr_copy_frame (call_frame_t *base)
 {
-        int ret = 0;
+	afr_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	int op_errno = 0;
+
+	frame = copy_frame (base);
+	if (!frame)
+		return NULL;
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local) {
+		AFR_STACK_DESTROY (frame);
+		return NULL;
+	}
+
+	return frame;
+}
 
-        GF_ASSERT (gfid);
+/*
+ * INODE CTX 64-bit VALUE FORMAT FOR SMALL (<= 16) SUBVOL COUNTS:
+ *
+ * |<----------   64bit   ------------>|
+ *  63           32 31    16 15       0
+ * |   EVENT_GEN   |  DATA  | METADATA |
+ *
+ *
+ *  METADATA (bit-0 .. bit-15): bitmap representing subvolumes from which
+ *                              metadata can be attempted to be read.
+ *
+ *                              bit-0 => priv->subvolumes[0]
+ *                              bit-1 => priv->subvolumes[1]
+ *                              ... etc. till bit-15
+ *
+ *  DATA (bit-16 .. bit-31): bitmap representing subvolumes from which data
+ *                           can be attempted to be read.
+ *
+ *                           bit-16 => priv->subvolumes[0]
+ *                           bit-17 => priv->subvolumes[1]
+ *                           ... etc. till bit-31
+ *
+ *  EVENT_GEN (bit-32 .. bit-63): event generation (i.e priv->event_generation)
+ *                                when DATA and METADATA was last updated.
+ *
+ *                                If EVENT_GEN is < priv->event_generation,
+ *                                or is 0, it means afr_inode_refresh() needs
+ *                                to be called to recalculate the bitmaps.
+ */
 
-        ret = dict_set_static_bin (dict, "gfid-req", gfid, 16);
-        if (ret)
-                gf_log (THIS->name, GF_LOG_DEBUG, "gfid set failed");
+int
+__afr_inode_read_subvol_get_small (inode_t *inode, xlator_t *this,
+				   unsigned char *data, unsigned char *metadata,
+				   int *event_p)
+{
+	afr_private_t *priv = NULL;
+	int ret = -1;
+	uint16_t datamap = 0;
+	uint16_t metadatamap = 0;
+	uint32_t event = 0;
+	uint64_t val = 0;
+	int i = 0;
+
+	priv = this->private;
+
+	ret = __inode_ctx_get (inode, this, &val);
+	if (ret < 0)
+		return ret;
+
+	metadatamap = (val & 0x000000000000ffff);
+	datamap =     (val & 0x00000000ffff0000) >> 16;
+	event =       (val & 0xffffffff00000000) >> 32;
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (metadata)
+			metadata[i] = (metadatamap >> i) & 1;
+		if (data)
+			data[i] = (datamap >> i) & 1;
+	}
+
+	if (event_p)
+		*event_p = event;
+	return ret;
+}
 
-        return ret;
+
+int
+__afr_inode_read_subvol_set_small (inode_t *inode, xlator_t *this,
+				   unsigned char *data, unsigned char *metadata,
+				   int event)
+{
+	afr_private_t *priv = NULL;
+	uint16_t datamap = 0;
+	uint16_t metadatamap = 0;
+	uint64_t val = 0;
+	int i = 0;
+
+	priv = this->private;
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (data[i])
+			datamap |= (1 << i);
+		if (metadata[i])
+			metadatamap |= (1 << i);
+	}
+
+	val = ((uint64_t) metadatamap) |
+		(((uint64_t) datamap) << 16) |
+		(((uint64_t) event) << 32);
+
+	return __inode_ctx_set (inode, this, &val);
 }
 
-uint64_t
-afr_is_split_brain (xlator_t *this, inode_t *inode)
+
+int
+__afr_inode_read_subvol_reset_small (inode_t *inode, xlator_t *this)
 {
-        int ret = 0;
+	int ret = -1;
+	uint16_t datamap = 0;
+	uint16_t metadatamap = 0;
+	uint32_t event = 0;
+	uint64_t val = 0;
 
-        uint64_t ctx         = 0;
-        uint64_t split_brain = 0;
+	ret = __inode_ctx_get (inode, this, &val);
+	(void) ret;
 
-        VALIDATE_OR_GOTO (inode, out);
+	metadatamap = (val & 0x000000000000ffff) >> 0;
+	datamap =     (val & 0x00000000ffff0000) >> 16;
+	event = 0;
 
-        LOCK (&inode->lock);
-        {
-                ret = __inode_ctx_get (inode, this, &ctx);
+	val = ((uint64_t) metadatamap) |
+		(((uint64_t) datamap) << 16) |
+		(((uint64_t) event) << 32);
 
-                if (ret < 0)
-                        goto unlock;
+	return __inode_ctx_set (inode, this, &val);
+}
 
-                split_brain = ctx & AFR_ICTX_SPLIT_BRAIN_MASK;
-        }
-unlock:
-        UNLOCK (&inode->lock);
 
-out:
-        return split_brain;
+int
+__afr_inode_read_subvol_get (inode_t *inode, xlator_t *this,
+			     unsigned char *data, unsigned char *metadata,
+			     int *event_p)
+{
+	afr_private_t *priv = NULL;
+	int ret = -1;
+
+	priv = this->private;
+
+	if (priv->child_count <= 16)
+		ret = __afr_inode_read_subvol_get_small (inode, this, data,
+							 metadata, event_p);
+	else
+		/* TBD: allocate structure with array and read from it */
+		ret = -1;
+
+	return ret;
 }
 
 
-void
-afr_set_split_brain (xlator_t *this, inode_t *inode, gf_boolean_t set)
+int
+__afr_inode_read_subvol_set (inode_t *inode, xlator_t *this, unsigned char *data,
+			     unsigned char *metadata, int event)
 {
-        uint64_t ctx = 0;
-        int      ret = 0;
+	afr_private_t *priv = NULL;
+	int ret = -1;
 
-        VALIDATE_OR_GOTO (inode, out);
+	priv = this->private;
 
-        LOCK (&inode->lock);
-        {
-                ret = __inode_ctx_get (inode, this, &ctx);
+	if (priv->child_count <= 16)
+		ret = __afr_inode_read_subvol_set_small (inode, this, data,
+							 metadata, event);
+	else
+		ret = -1;
 
-                if (ret < 0) {
-                        ctx = 0;
-                }
+	return ret;
+}
 
-                if (set) {
-                        ctx = (~AFR_ICTX_SPLIT_BRAIN_MASK & ctx)
-                                | (0xFFFFFFFFFFFFFFFFULL & AFR_ICTX_SPLIT_BRAIN_MASK);
-                } else {
-                        ctx = (~AFR_ICTX_SPLIT_BRAIN_MASK & ctx);
-                }
-                __inode_ctx_put (inode, this, ctx);
-        }
-        UNLOCK (&inode->lock);
-out:
-        return;
+
+int
+__afr_inode_read_subvol_reset (inode_t *inode, xlator_t *this)
+{
+	afr_private_t *priv = NULL;
+	int ret = -1;
+
+	priv = this->private;
+
+	if (priv->child_count <= 16)
+		ret = __afr_inode_read_subvol_reset_small (inode, this);
+	else
+		ret = -1;
+
+	return ret;
 }
 
 
-uint64_t
-afr_is_opendir_done (xlator_t *this, inode_t *inode)
+int
+afr_inode_read_subvol_get (inode_t *inode, xlator_t *this, unsigned char *data,
+			   unsigned char *metadata, int *event_p)
 {
-        int ret = 0;
+	int ret = -1;
 
-        uint64_t ctx          = 0;
-        uint64_t opendir_done = 0;
+	LOCK(&inode->lock);
+	{
+		ret = __afr_inode_read_subvol_get (inode, this, data,
+						   metadata, event_p);
+	}
+	UNLOCK(&inode->lock);
 
-        VALIDATE_OR_GOTO (inode, out);
+	return ret;
+}
 
-        LOCK (&inode->lock);
-        {
-                ret = __inode_ctx_get (inode, this, &ctx);
 
-                if (ret < 0)
-                        goto unlock;
+int
+afr_inode_read_subvol_set (inode_t *inode, xlator_t *this, unsigned char *data,
+			   unsigned char *metadata, int event)
+{
+	int ret = -1;
 
-                opendir_done = ctx & AFR_ICTX_OPENDIR_DONE_MASK;
-        }
-unlock:
-        UNLOCK (&inode->lock);
+	LOCK(&inode->lock);
+	{
+		ret = __afr_inode_read_subvol_set (inode, this, data, metadata,
+						   event);
+	}
+	UNLOCK(&inode->lock);
 
-out:
-        return opendir_done;
+	return ret;
 }
 
 
-void
-afr_set_opendir_done (xlator_t *this, inode_t *inode)
+int
+afr_inode_read_subvol_reset (inode_t *inode, xlator_t *this)
 {
-        uint64_t ctx = 0;
-        int      ret = 0;
+	int ret = -1;
 
-        VALIDATE_OR_GOTO (inode, out);
+	LOCK(&inode->lock);
+	{
+		ret = __afr_inode_read_subvol_reset (inode, this);
+	}
+	UNLOCK(&inode->lock);
 
-        LOCK (&inode->lock);
-        {
-                ret = __inode_ctx_get (inode, this, &ctx);
+	return ret;
+}
 
-                if (ret < 0) {
-                        ctx = 0;
-                }
 
-                ctx = (~AFR_ICTX_OPENDIR_DONE_MASK & ctx)
-                        | (0xFFFFFFFFFFFFFFFFULL & AFR_ICTX_OPENDIR_DONE_MASK);
+int
+afr_accused_fill (xlator_t *this, dict_t *xdata, unsigned char *accused,
+		  afr_transaction_type type)
+{
+	afr_private_t *priv = NULL;
+	int i = 0;
+	int idx = afr_index_for_transaction_type (type);
+	void *pending_raw = NULL;
+	int pending[3];
+	int ret = 0;
+
+	priv = this->private;
+
+	for (i = 0; i < priv->child_count; i++) {
+		ret = dict_get_ptr (xdata, priv->pending_key[i],
+				    &pending_raw);
+		if (ret) /* no pending flags */
+			continue;
+		memcpy (pending, pending_raw, sizeof(pending));
+
+		if (ntoh32 (pending[idx]))
+			accused[i] = 1;
+	}
+
+	return 0;
+}
 
-                __inode_ctx_put (inode, this, ctx);
-        }
-        UNLOCK (&inode->lock);
-out:
-        return;
+
+int
+afr_accuse_smallfiles (xlator_t *this, struct afr_reply *replies,
+		       unsigned char *data_accused)
+{
+	int i = 0;
+	afr_private_t *priv = NULL;
+	uint64_t maxsize = 0;
+
+	priv = this->private;
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (data_accused[i])
+			continue;
+		if (replies[i].poststat.ia_size > maxsize)
+			maxsize = replies[i].poststat.ia_size;
+	}
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (data_accused[i])
+			continue;
+		if (replies[i].poststat.ia_size < maxsize)
+			data_accused[i] = 1;
+	}
+
+	return 0;
 }
 
 
-uint64_t
-afr_read_child (xlator_t *this, inode_t *inode)
+int
+afr_replies_interpret (call_frame_t *frame, xlator_t *this, inode_t *inode)
 {
-        int ret = 0;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	struct afr_reply *replies = NULL;
+	int event_generation = 0;
+	int i = 0;
+	unsigned char *data_accused = NULL;
+	unsigned char *metadata_accused = NULL;
+	unsigned char *data_readable = NULL;
+	unsigned char *metadata_readable = NULL;
+	int ret = 0;
+
+	local = frame->local;
+	priv = this->private;
+	replies = local->replies;
+	event_generation = local->event_generation;
+
+	data_accused = alloca0 (priv->child_count);
+	data_readable = alloca0 (priv->child_count);
+	metadata_accused = alloca0 (priv->child_count);
+	metadata_readable = alloca0 (priv->child_count);
+
+	for (i = 0; i < priv->child_count; i++) {
+		data_readable[i] = 1;
+		metadata_readable[i] = 1;
+	}
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].valid) {
+			data_readable[i] = 0;
+			metadata_readable[i] = 0;
+			continue;
+		}
+
+		if (replies[i].op_ret == -1) {
+			data_readable[i] = 0;
+			metadata_readable[i] = 0;
+			continue;
+		}
+
+		afr_accused_fill (this, replies[i].xdata, data_accused,
+				  (inode->ia_type == IA_IFDIR) ?
+				   AFR_ENTRY_TRANSACTION : AFR_DATA_TRANSACTION);
+
+		afr_accused_fill (this, replies[i].xdata,
+				  metadata_accused, AFR_METADATA_TRANSACTION);
+
+	}
+
+	if (inode->ia_type != IA_IFDIR)
+		afr_accuse_smallfiles (this, replies, data_accused);
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (data_accused[i]) {
+			data_readable[i] = 0;
+			ret = 1;
+		}
+		if (metadata_accused[i]) {
+			metadata_readable[i] = 0;
+			ret = 1;
+		}
+	}
+
+	afr_inode_read_subvol_set (inode, this, data_readable,
+				   metadata_readable, event_generation);
+	return ret;
+}
 
-        uint64_t ctx         = 0;
-        uint64_t read_child  = 0;
 
-        VALIDATE_OR_GOTO (inode, out);
 
-        LOCK (&inode->lock);
-        {
-                ret = __inode_ctx_get (inode, this, &ctx);
+int
+afr_refresh_selfheal_done (int ret, call_frame_t *heal, void *opaque)
+{
+	if (heal)
+		STACK_DESTROY (heal->root);
+	return 0;
+}
 
-                if (ret < 0)
-                        goto unlock;
+int
+afr_inode_refresh_err (call_frame_t *frame, xlator_t *this)
+{
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	int i = 0;
+	int err = 0;
+
+	local = frame->local;
+	priv = this->private;
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (local->replies[i].valid && !local->replies[i].op_ret) {
+			err = 0;
+			goto ret;
+		}
+	}
+
+	err = afr_final_errno (local, priv);
+ret:
+	return -err;
+}
 
-                read_child = ctx & AFR_ICTX_READ_CHILD_MASK;
-        }
-unlock:
-        UNLOCK (&inode->lock);
 
-out:
-        return read_child;
+int
+afr_refresh_selfheal_wrap (void *opaque)
+{
+	call_frame_t *frame = opaque;
+	afr_local_t *local = NULL;
+	xlator_t *this = NULL;
+	int err = 0;
+
+	local = frame->local;
+	this = frame->this;
+
+	afr_selfheal (frame->this, local->refreshinode->gfid);
+
+	afr_selfheal_unlocked_discover (frame, local->refreshinode,
+					local->refreshinode->gfid,
+					local->replies);
+
+	afr_replies_interpret (frame, this, local->refreshinode);
+
+	err = afr_inode_refresh_err (frame, this);
+
+	afr_replies_wipe (local, this->private);
+
+	local->refreshfn (frame, this, err);
+
+	return 0;
 }
 
 
-void
-afr_set_read_child (xlator_t *this, inode_t *inode, int32_t read_child)
+gf_boolean_t
+afr_selfheal_enabled (xlator_t *this)
 {
-        uint64_t ctx = 0;
-        int      ret = 0;
+	afr_private_t *priv = NULL;
+	gf_boolean_t data = _gf_false;
 
-        VALIDATE_OR_GOTO (inode, out);
+	priv = this->private;
 
-        LOCK (&inode->lock);
-        {
-                ret = __inode_ctx_get (inode, this, &ctx);
+	gf_string2boolean (priv->data_self_heal, &data);
 
-                if (ret < 0) {
-                        ctx = 0;
-                }
+	return data || priv->metadata_self_heal || priv->entry_self_heal;
+}
 
-                ctx = (~AFR_ICTX_READ_CHILD_MASK & ctx)
-                        | (AFR_ICTX_READ_CHILD_MASK & read_child);
 
-                __inode_ctx_put (inode, this, ctx);
-        }
-        UNLOCK (&inode->lock);
 
-out:
-        return;
+int
+afr_inode_refresh_done (call_frame_t *frame, xlator_t *this)
+{
+	call_frame_t *heal = NULL;
+	afr_local_t *local = NULL;
+	int ret = 0;
+	int err = 0;
+
+	local = frame->local;
+
+	ret = afr_replies_interpret (frame, this, local->refreshinode);
+
+	err = afr_inode_refresh_err (frame, this);
+
+	afr_replies_wipe (local, this->private);
+
+	if (ret && afr_selfheal_enabled (this)) {
+		heal = copy_frame (frame);
+		if (heal)
+			heal->root->pid = -1;
+		ret = synctask_new (this->ctx->env, afr_refresh_selfheal_wrap,
+				    afr_refresh_selfheal_done, heal, frame);
+		if (ret)
+			goto refresh_done;
+	} else {
+	refresh_done:
+		local->refreshfn (frame, this, err);
+	}
+
+	return 0;
 }
 
 
-/**
- * afr_local_cleanup - cleanup everything in frame->local
- */
+int
+afr_inode_refresh_subvol_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+			      int op_ret, int op_errno, inode_t *inode,
+			      struct iatt *buf, dict_t *xdata, struct iatt *par)
+{
+	afr_local_t *local = NULL;
+	int call_child = (long) cookie;
+	int call_count = 0;
 
-void
-afr_local_sh_cleanup (afr_local_t *local, xlator_t *this)
+	local = frame->local;
+
+	local->replies[call_child].valid = 1;
+	local->replies[call_child].op_ret = op_ret;
+	local->replies[call_child].op_errno = op_errno;
+	if (op_ret != -1) {
+		local->replies[call_child].poststat = *buf;
+		local->replies[call_child].postparent = *par;
+		local->replies[call_child].xdata = dict_ref (xdata);
+	}
+
+	call_count = afr_frame_return (frame);
+
+	if (call_count == 0)
+		afr_inode_refresh_done (frame, this);
+
+	return 0;
+}
+
+
+int
+afr_inode_refresh_subvol (call_frame_t *frame, xlator_t *this, int i,
+			  inode_t *inode, dict_t *xdata)
 {
-        afr_self_heal_t *sh = NULL;
-        afr_private_t   *priv = NULL;
-        int              i = 0;
+	loc_t loc = {0, };
+	afr_private_t *priv = NULL;
 
+	priv = this->private;
 
-        sh = &local->self_heal;
-        priv = this->private;
+	loc.inode = inode;
+	uuid_copy (loc.gfid, inode->gfid);
 
-        if (sh->buf)
-                GF_FREE (sh->buf);
+	STACK_WIND_COOKIE (frame, afr_inode_refresh_subvol_cbk,
+			   (void *) (long) i, priv->children[i],
+			   priv->children[i]->fops->lookup, &loc, xdata);
+	return 0;
+}
 
-        if (sh->xattr) {
-                for (i = 0; i < priv->child_count; i++) {
-                        if (sh->xattr[i]) {
-                                dict_unref (sh->xattr[i]);
-                                sh->xattr[i] = NULL;
-                        }
-                }
-                GF_FREE (sh->xattr);
+
+int
+afr_inode_refresh_do (call_frame_t *frame, xlator_t *this)
+{
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	int call_count = 0;
+	int i = 0;
+	dict_t *xdata = NULL;
+
+	priv = this->private;
+	local = frame->local;
+
+	afr_replies_wipe (local, priv);
+
+	xdata = dict_new ();
+	if (!xdata) {
+		afr_inode_refresh_done (frame, this);
+		return 0;
+	}
+
+	if (afr_xattr_req_prepare (this, xdata) != 0) {
+		dict_unref (xdata);
+		afr_inode_refresh_done (frame, this);
+		return 0;
+	}
+
+	local->call_count = AFR_COUNT (local->child_up, priv->child_count);
+
+	call_count = local->call_count;
+	for (i = 0; i < priv->child_count; i++) {
+		if (!local->child_up[i])
+			continue;
+
+		afr_inode_refresh_subvol (frame, this, i, local->refreshinode,
+					  xdata);
+
+		if (!--call_count)
+			break;
+	}
+
+	dict_unref (xdata);
+
+	return 0;
+}
+
+
+int
+afr_inode_refresh (call_frame_t *frame, xlator_t *this, inode_t *inode,
+		   afr_inode_refresh_cbk_t refreshfn)
+{
+	afr_local_t *local = NULL;
+
+	local = frame->local;
+
+	local->refreshfn = refreshfn;
+
+	if (local->refreshinode) {
+		inode_unref (local->refreshinode);
+		local->refreshinode = NULL;
+	}
+
+	local->refreshinode = inode_ref (inode);
+
+	afr_inode_refresh_do (frame, this);
+
+	return 0;
+}
+
+
+int
+afr_xattr_req_prepare (xlator_t *this, dict_t *xattr_req)
+{
+        int             i           = 0;
+        afr_private_t   *priv       = NULL;
+        int             ret         = 0;
+
+        priv   = this->private;
+
+        for (i = 0; i < priv->child_count; i++) {
+                ret = dict_set_uint64 (xattr_req, priv->pending_key[i],
+                                       AFR_NUM_CHANGE_LOGS * sizeof(int));
+                if (ret < 0)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "Unable to set dict value for %s",
+                                priv->pending_key[i]);
+                /* 3 = data+metadata+entry */
+        }
+        ret = dict_set_uint64 (xattr_req, AFR_DIRTY,
+			       AFR_NUM_CHANGE_LOGS * sizeof(int));
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "failed to set dirty "
+                        "query flag");
         }
 
-        if (sh->child_errno)
-                GF_FREE (sh->child_errno);
+	return ret;
+}
 
-        if (sh->pending_matrix) {
-                for (i = 0; i < priv->child_count; i++) {
-                        GF_FREE (sh->pending_matrix[i]);
-                }
-                GF_FREE (sh->pending_matrix);
+int
+afr_lookup_xattr_req_prepare (afr_local_t *local, xlator_t *this,
+                              dict_t *xattr_req, loc_t *loc)
+{
+        int     ret = -ENOMEM;
+
+        local->xattr_req = dict_new ();
+        if (!local->xattr_req)
+                goto out;
+        if (xattr_req)
+                dict_copy (xattr_req, local->xattr_req);
+
+        ret = afr_xattr_req_prepare (this, local->xattr_req);
+	if (ret < 0) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"%s: Unable to prepare xattr_req", loc->path);
+	}
+
+        ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_INODELK_COUNT, 0);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: Unable to set dict value for %s",
+                        loc->path, GLUSTERFS_INODELK_COUNT);
+        }
+        ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_ENTRYLK_COUNT, 0);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: Unable to set dict value for %s",
+                        loc->path, GLUSTERFS_ENTRYLK_COUNT);
         }
 
-        if (sh->delta_matrix) {
-                for (i = 0; i < priv->child_count; i++) {
-                        GF_FREE (sh->delta_matrix[i]);
-                }
-                GF_FREE (sh->delta_matrix);
+        ret = dict_set_uint32 (local->xattr_req, GLUSTERFS_PARENT_ENTRYLK, 0);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: Unable to set dict value for %s",
+                        loc->path, GLUSTERFS_PARENT_ENTRYLK);
         }
 
-        if (sh->sources)
-                GF_FREE (sh->sources);
+        ret = 0;
+out:
+        return ret;
+}
 
-        if (sh->success)
-                GF_FREE (sh->success);
 
-        if (sh->locked_nodes)
-                GF_FREE (sh->locked_nodes);
+int
+afr_hash_child (inode_t *inode, int32_t child_count, int hashmode)
+{
+        uuid_t gfid_copy = {0,};
+        pid_t pid;
 
-        if (sh->healing_fd && !sh->healing_fd_opened) {
-                fd_unref (sh->healing_fd);
-                sh->healing_fd = NULL;
+        if (!hashmode) {
+                return -1;
         }
 
-        if (sh->linkname)
-                GF_FREE ((char *)sh->linkname);
+        if (inode) {
+               uuid_copy (gfid_copy, inode->gfid);
+        }
 
-        loc_wipe (&sh->parent_loc);
+        if (hashmode > 1) {
+                /*
+                 * Why getpid?  Because it's one of the cheapest calls
+                 * available - faster than gethostname etc. - and returns a
+                 * constant-length value that's sure to be shorter than a UUID.
+                 * It's still very unlikely to be the same across clients, so
+                 * it still provides good mixing.  We're not trying for
+                 * perfection here.  All we need is a low probability that
+                 * multiple clients won't converge on the same subvolume.
+                 */
+                pid = getpid();
+                memcpy (gfid_copy, &pid, sizeof(pid));
+        }
+
+        return SuperFastHash((char *)gfid_copy,
+                             sizeof(gfid_copy)) % child_count;
+}
+
+
+int
+afr_read_subvol_select_by_policy (inode_t *inode, xlator_t *this,
+				  unsigned char *readable)
+{
+	afr_private_t *priv = NULL;
+	int read_subvol = -1;
+	int i = 0;
+
+	priv = this->private;
+
+	/* first preference - explicitly specified or local subvolume */
+	if (priv->read_child >= 0 && readable[priv->read_child])
+		return priv->read_child;
+
+	/* second preference - use hashed mode */
+	read_subvol = afr_hash_child (inode, priv->child_count,
+				      priv->hash_mode);
+	if (read_subvol >= 0 && readable[read_subvol])
+		return read_subvol;
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (readable[i])
+			return i;
+	}
+
+	/* no readable subvolumes, either split brain or all subvols down */
+
+	return -1;
+}
+
+
+int
+afr_inode_read_subvol_type_get (inode_t *inode, xlator_t *this,
+				unsigned char *readable, int *event_p,
+				int type)
+{
+	int ret = -1;
+
+	if (type == AFR_METADATA_TRANSACTION)
+		ret = afr_inode_read_subvol_get (inode, this, 0, readable,
+						 event_p);
+	else
+		ret = afr_inode_read_subvol_get (inode, this, readable, 0,
+						 event_p);
+	return ret;
+}
+
+
+int
+afr_read_subvol_get (inode_t *inode, xlator_t *this, int *subvol_p,
+		     int *event_p, afr_transaction_type type)
+{
+	afr_private_t *priv = NULL;
+	unsigned char *data_readable = NULL;
+	unsigned char *metadata_readable = NULL;
+	unsigned char *readable = NULL;
+	unsigned char *intersection = NULL;
+	int subvol = -1;
+	int event = 0;
+
+	priv = this->private;
+
+	readable = alloca0 (priv->child_count);
+	data_readable = alloca0 (priv->child_count);
+	metadata_readable = alloca0 (priv->child_count);
+	intersection = alloca0 (priv->child_count);
+
+	afr_inode_read_subvol_type_get (inode, this, readable, &event, type);
+
+	afr_inode_read_subvol_get (inode, this, data_readable, metadata_readable,
+				   &event);
+
+	AFR_INTERSECT (intersection, data_readable, metadata_readable,
+		       priv->child_count);
+
+	if (AFR_COUNT (intersection, priv->child_count) > 0)
+		subvol = afr_read_subvol_select_by_policy (inode, this,
+							   intersection);
+	else
+		subvol = afr_read_subvol_select_by_policy (inode, this,
+							   readable);
+	if (subvol_p)
+		*subvol_p = subvol;
+	if (event_p)
+		*event_p = event;
+	return subvol;
 }
 
 
 void
 afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this)
 {
-        int             i = 0;
-        afr_private_t * priv = NULL;
+        afr_private_t *priv    = NULL;
+        int           i        = 0;
 
         priv = this->private;
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (local->pending && local->pending[i])
-                        GF_FREE (local->pending[i]);
-        }
+        afr_matrix_cleanup (local->pending, priv->child_count);
 
-        GF_FREE (local->pending);
+        GF_FREE (local->internal_lock.locked_nodes);
 
-        if (local->internal_lock.locked_nodes)
-                GF_FREE (local->internal_lock.locked_nodes);
-
-        if (local->internal_lock.inode_locked_nodes)
-                GF_FREE (local->internal_lock.inode_locked_nodes);
-
-        if (local->internal_lock.entry_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);
+        }
 
-        if (local->internal_lock.lower_locked_nodes)
-                GF_FREE (local->internal_lock.lower_locked_nodes);
+        GF_FREE (local->internal_lock.lower_locked_nodes);
 
+        afr_entry_lockee_cleanup (&local->internal_lock);
 
-        GF_FREE (local->transaction.child_errno);
-        GF_FREE (local->child_errno);
+        GF_FREE (local->transaction.pre_op);
+        GF_FREE (local->transaction.eager_lock);
 
         GF_FREE (local->transaction.basename);
         GF_FREE (local->transaction.new_basename);
 
         loc_wipe (&local->transaction.parent_loc);
         loc_wipe (&local->transaction.new_parent_loc);
+
 }
 
 
 void
+afr_replies_wipe (afr_local_t *local, afr_private_t *priv)
+{
+	int i;
+
+	if (!local->replies)
+		return;
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (local->replies[i].xdata) {
+			dict_unref (local->replies[i].xdata);
+			local->replies[i].xdata = NULL;
+		}
+	}
+
+	memset (local->replies, 0, sizeof(*local->replies) * priv->child_count);
+}
+
+void
+afr_remove_eager_lock_stub (afr_local_t *local)
+{
+        LOCK (&local->fd->lock);
+        {
+                list_del_init (&local->transaction.eager_locked);
+        }
+        UNLOCK (&local->fd->lock);
+}
+
+void
 afr_local_cleanup (afr_local_t *local, xlator_t *this)
 {
-        int i;
         afr_private_t * priv = NULL;
 
         if (!local)
                 return;
 
-        afr_local_sh_cleanup (local, this);
+	syncbarrier_destroy (&local->barrier);
+
+        if (local->transaction.eager_lock_on &&
+            !list_empty (&local->transaction.eager_locked))
+                afr_remove_eager_lock_stub (local);
 
         afr_local_transaction_cleanup (local, this);
 
@@ -369,37 +926,36 @@ afr_local_cleanup (afr_local_t *local, xlator_t *this)
         if (local->xattr_req)
                 dict_unref (local->xattr_req);
 
+        if (local->dict)
+                dict_unref (local->dict);
+
+	afr_replies_wipe (local, priv);
+	GF_FREE(local->replies);
+
         GF_FREE (local->child_up);
 
-        { /* lookup */
-                if (local->cont.lookup.xattrs) {
-                        for (i = 0; i < priv->child_count; i++) {
-                                if (local->cont.lookup.xattrs[i]) {
-                                        dict_unref (local->cont.lookup.xattrs[i]);
-                                        local->cont.lookup.xattrs[i] = NULL;
-                                }
-                        }
-                        GF_FREE (local->cont.lookup.xattrs);
-                        local->cont.lookup.xattrs = NULL;
-                }
+        GF_FREE (local->read_attempted);
 
-                if (local->cont.lookup.xattr) {
-                        dict_unref (local->cont.lookup.xattr);
-                }
+        GF_FREE (local->readable);
 
-                if (local->cont.lookup.inode) {
-                        inode_unref (local->cont.lookup.inode);
-                }
-        }
+	if (local->inode)
+		inode_unref (local->inode);
+
+	if (local->parent)
+		inode_unref (local->parent);
+
+	if (local->parent2)
+		inode_unref (local->parent2);
+
+	if (local->refreshinode)
+		inode_unref (local->refreshinode);
 
         { /* getxattr */
-                if (local->cont.getxattr.name)
-                        GF_FREE (local->cont.getxattr.name);
+                GF_FREE (local->cont.getxattr.name);
         }
 
         { /* lk */
-                if (local->cont.lk.locked_nodes)
-                        GF_FREE (local->cont.lk.locked_nodes);
+                GF_FREE (local->cont.lk.locked_nodes);
         }
 
         { /* create */
@@ -433,18 +989,40 @@ afr_local_cleanup (afr_local_t *local, xlator_t *this)
                         dict_unref (local->cont.setxattr.dict);
         }
 
+        { /* fsetxattr */
+                if (local->cont.fsetxattr.dict)
+                        dict_unref (local->cont.fsetxattr.dict);
+        }
+
         { /* removexattr */
                 GF_FREE (local->cont.removexattr.name);
         }
-
+        { /* xattrop */
+                if (local->cont.xattrop.xattr)
+                        dict_unref (local->cont.xattrop.xattr);
+        }
+        { /* fxattrop */
+                if (local->cont.fxattrop.xattr)
+                        dict_unref (local->cont.fxattrop.xattr);
+        }
         { /* symlink */
                 GF_FREE (local->cont.symlink.linkpath);
         }
 
         { /* opendir */
-                if (local->cont.opendir.checksum)
-                        GF_FREE (local->cont.opendir.checksum);
+                GF_FREE (local->cont.opendir.checksum);
+        }
+
+        { /* readdirp */
+                if (local->cont.readdir.dict)
+                        dict_unref (local->cont.readdir.dict);
         }
+
+        if (local->xdata_req)
+                dict_unref (local->xdata_req);
+
+        if (local->xdata_rsp)
+                dict_unref (local->xdata_rsp);
 }
 
 
@@ -466,949 +1044,1060 @@ afr_frame_return (call_frame_t *frame)
 }
 
 
-/**
- * up_children_count - return the number of children that are up
- */
-
-int
-afr_up_children_count (int child_count, unsigned char *child_up)
+gf_boolean_t
+afr_is_entry_possibly_under_txn (afr_local_t *local, xlator_t *this)
 {
-        int i   = 0;
-        int ret = 0;
+	int i = 0;
+	int tmp = 0;
+	afr_private_t *priv = NULL;
 
-        for (i = 0; i < child_count; i++)
-                if (child_up[i])
-                        ret++;
-        return ret;
-}
+	priv = this->private;
 
+	for (i = 0; i < priv->child_count; i++) {
+		if (!local->replies[i].xdata)
+			continue;
+		if (dict_get_int32 (local->replies[i].xdata,
+				    GLUSTERFS_PARENT_ENTRYLK,
+				    &tmp) == 0)
+			if (tmp)
+				return _gf_true;
+	}
 
-ino64_t
-afr_itransform (ino64_t ino, int child_count, int child_index)
-{
-        ino64_t scaled_ino = -1;
+	return _gf_false;
+}
 
-        if (ino == ((uint64_t) -1)) {
-                scaled_ino = ((uint64_t) -1);
-                goto out;
-        }
 
-        scaled_ino = (ino * child_count) + child_index;
+/*
+ * 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'.
+ * */
 
-out:
-        return scaled_ino;
+static void
+afr_handle_quota_size (call_frame_t *frame, xlator_t *this)
+{
+	unsigned char *readable = NULL;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	struct afr_reply *replies = NULL;
+	int i = 0;
+	uint64_t size = 0;
+	uint64_t max_size = 0;
+	int readable_cnt = 0;
+
+	local = frame->local;
+	priv = this->private;
+	replies = local->replies;
+
+	readable = alloca0 (priv->child_count);
+
+	afr_inode_read_subvol_get (local->inode, this, readable, 0, 0);
+
+	readable_cnt = AFR_COUNT (readable, priv->child_count);
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].valid || replies[i].op_ret == -1)
+			continue;
+		if (readable_cnt && !readable[i])
+			continue;
+		if (!replies[i].xdata)
+			continue;
+		if (dict_get_uint64 (replies[i].xdata, QUOTA_SIZE_KEY, &size))
+			continue;
+		if (size > max_size)
+			max_size = size;
+	}
+
+	if (!max_size)
+		return;
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].valid || replies[i].op_ret == -1)
+			continue;
+		if (readable_cnt && !readable[i])
+			continue;
+		if (!replies[i].xdata)
+			continue;
+		if (dict_set_uint64 (replies[i].xdata, QUOTA_SIZE_KEY, max_size))
+			continue;
+	}
 }
 
 
-int
-afr_deitransform_orig (ino64_t ino, int child_count)
+static void
+afr_lookup_done (call_frame_t *frame, xlator_t *this)
 {
-        int index = -1;
+        afr_private_t       *priv  = NULL;
+        afr_local_t         *local = NULL;
+	int                 i = -1;
+	int                 op_errno = 0;
+	int                 read_subvol = 0;
+	unsigned char      *readable = NULL;
+	int                 event = 0;
+	struct afr_reply   *replies = NULL;
+	uuid_t              read_gfid = {0, };
+	gf_boolean_t        locked_entry = _gf_false;
+	gf_boolean_t        can_interpret = _gf_true;
+
+        priv  = this->private;
+        local = frame->local;
+	replies = local->replies;
 
-        index = ino % child_count;
+	locked_entry = afr_is_entry_possibly_under_txn (local, this);
 
-        return index;
+	readable = alloca0 (priv->child_count);
+
+	afr_inode_read_subvol_get (local->loc.parent, this, readable,
+				   NULL, &event);
+
+	/* First, check if we have an ESTALE from somewhere,
+	   If so, propagate that so that a revalidate can be
+	   issued
+	*/
+	op_errno = afr_final_errno (frame->local, this->private);
+	local->op_errno = op_errno;
+        if (op_errno == ESTALE) {
+		local->op_errno = op_errno;
+		local->op_ret = -1;
+                goto unwind;
+	}
+
+	read_subvol = -1;
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].valid)
+			continue;
+
+		if (locked_entry && replies[i].op_ret == -1 &&
+		    replies[i].op_errno == ENOENT) {
+			/* Second, check entry is still
+			   "underway" in creation */
+			local->op_ret = -1;
+			local->op_errno = ENOENT;
+			read_subvol = i;
+			goto unwind;
+		}
+
+		if (replies[i].op_ret == -1)
+			continue;
+
+		if (read_subvol == -1 || !readable[read_subvol]) {
+			read_subvol = i;
+			uuid_copy (read_gfid, replies[i].poststat.ia_gfid);
+			local->op_ret = 0;
+		}
+	}
+
+	if (read_subvol == -1)
+		goto unwind;
+	/* We now have a read_subvol, which is readable[] (if there
+	   were any). Next we look for GFID mismatches. We don't
+	   consider a GFID mismatch as an error if read_subvol is
+	   readable[] but the mismatching GFID subvol is not.
+	*/
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].valid || replies[i].op_ret == -1) {
+			if (priv->child_up[i])
+				can_interpret = _gf_false;
+			continue;
+		}
+
+		if (!uuid_compare (replies[i].poststat.ia_gfid,
+				   read_gfid))
+			continue;
+
+		can_interpret = _gf_false;
+
+		if (locked_entry)
+			continue;
+
+		/* Now GFIDs mismatch. It's OK as long as this subvol
+		   is not readable[] but read_subvol is */
+		if (readable[read_subvol] && !readable[i])
+			continue;
+
+		/* LOG ERROR */
+		local->op_ret = -1;
+		local->op_errno = EIO;
+		goto unwind;
+	}
+
+	/* Forth, for the finalized GFID, pick the best subvolume
+	   to return stats from.
+	*/
+	if (can_interpret) {
+		/* It is safe to call afr_replies_interpret() because we have
+		   a response from all the UP subvolumes and all of them resolved
+		   to the same GFID
+		*/
+		if (afr_replies_interpret (frame, this, local->inode)) {
+			read_subvol = afr_data_subvol_get (local->inode, this,
+							   0, 0);
+			afr_inode_read_subvol_reset (local->inode, this);
+			goto cant_interpret;
+		} else {
+			read_subvol = afr_data_subvol_get (local->inode, this,
+							   0, 0);
+		}
+	} else {
+	cant_interpret:
+		if (read_subvol == -1)
+			dict_del (replies[0].xdata, GF_CONTENT_KEY);
+		else
+			dict_del (replies[read_subvol].xdata, GF_CONTENT_KEY);
+	}
+
+	afr_handle_quota_size (frame, this);
+
+unwind:
+	if (read_subvol == -1)
+		read_subvol = 0;
+
+	AFR_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
+			  local->inode, &local->replies[read_subvol].poststat,
+			  local->replies[read_subvol].xdata,
+			  &local->replies[read_subvol].postparent);
 }
 
+/*
+ * During a lookup, some errors are more "important" than
+ * others in that they must be given higher priority while
+ * returning to the user.
+ *
+ * The hierarchy is ESTALE > ENOENT > others
+ */
 
 int
-afr_deitransform (ino64_t ino, int child_count)
+afr_higher_errno (int32_t old_errno, int32_t new_errno)
 {
-        return 0;
+	if (old_errno == ENODATA || new_errno == ENODATA)
+		return ENODATA;
+	if (old_errno == ESTALE || new_errno == ESTALE)
+		return ESTALE;
+	if (old_errno == ENOENT || new_errno == ENOENT)
+		return ENOENT;
+
+	return new_errno;
 }
 
 
 int
-afr_self_heal_lookup_unwind (call_frame_t *frame, xlator_t *this)
+afr_final_errno (afr_local_t *local, afr_private_t *priv)
 {
-        afr_local_t *local = NULL;
+	int i = 0;
+	int op_errno = 0;
+	int tmp_errno = 0;
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (!local->replies[i].valid)
+			continue;
+		if (local->replies[i].op_ret == 0)
+			continue;
+		tmp_errno = local->replies[i].op_errno;
+		op_errno = afr_higher_errno (op_errno, tmp_errno);
+	}
+
+	return op_errno;
+}
 
-        local = frame->local;
+static int
+get_pathinfo_host (char *pathinfo, char *hostname, size_t size)
+{
+        char    *start = NULL;
+        char    *end = NULL;
+        int     ret  = -1;
+        int     i    = 0;
 
-        if (local->govinda_gOvinda) {
-                afr_set_split_brain (this, local->cont.lookup.inode, _gf_true);
-        }
+        if (!pathinfo)
+                goto out;
 
-        AFR_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
-                          local->cont.lookup.inode,
-                          &local->cont.lookup.buf,
-                          local->cont.lookup.xattr,
-                          &local->cont.lookup.postparent);
+        start = strchr (pathinfo, ':');
+        if (!start)
+                goto out;
+        end = strrchr (pathinfo, ':');
+        if (start == end)
+                goto out;
 
-        return 0;
+        memset (hostname, 0, size);
+        i = 0;
+        while (++start != end)
+                hostname[i++] = *start;
+        ret = 0;
+out:
+        return ret;
 }
 
-
-static void
-afr_lookup_collect_xattr (afr_local_t *local, xlator_t *this,
-                          int child_index, dict_t *xattr)
+int
+afr_local_pathinfo (char *pathinfo, gf_boolean_t *local)
 {
-        uint32_t        inodelk_count = 0;
-        uint32_t        entrylk_count = 0;
-
-        int ret = 0;
-
-        if (afr_sh_has_metadata_pending (xattr, child_index, this)) {
-                local->self_heal.need_metadata_self_heal = _gf_true;
-                gf_log(this->name, GF_LOG_DEBUG,
-                       "metadata self-heal is pending for %s.",
-                       local->loc.path);
-        }
-
-        if (afr_sh_has_entry_pending (xattr, child_index, this)) {
-                local->self_heal.need_entry_self_heal = _gf_true;
-                gf_log(this->name, GF_LOG_DEBUG,
-                       "entry self-heal is pending for %s.", local->loc.path);
+        int             ret   = 0;
+        char            pathinfohost[1024] = {0};
+        char            localhost[1024] = {0};
+        xlator_t        *this = THIS;
+
+        *local = _gf_false;
+        ret = get_pathinfo_host (pathinfo, pathinfohost, sizeof (pathinfohost));
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Invalid pathinfo: %s",
+                        pathinfo);
+                goto out;
         }
 
-        if (afr_sh_has_data_pending (xattr, child_index, this)) {
-                local->self_heal.need_data_self_heal = _gf_true;
-                gf_log(this->name, GF_LOG_DEBUG,
-                       "data self-heal is pending for %s.", local->loc.path);
+        ret = gethostname (localhost, sizeof (localhost));
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "gethostname() failed, "
+                        "reason: %s", strerror (errno));
+                goto out;
         }
 
-        ret = dict_get_uint32 (xattr, GLUSTERFS_INODELK_COUNT,
-                               &inodelk_count);
-        if (ret == 0)
-                local->inodelk_count += inodelk_count;
-
-        ret = dict_get_uint32 (xattr, GLUSTERFS_ENTRYLK_COUNT,
-                               &entrylk_count);
-        if (ret == 0)
-                local->entrylk_count += entrylk_count;
+        if (!strcmp (localhost, pathinfohost))
+                *local = _gf_true;
+out:
+        return ret;
 }
 
-
-static void
-afr_lookup_self_heal_check (xlator_t *this, afr_local_t *local,
-                            struct iatt *buf, struct iatt *lookup_buf)
+static int32_t
+afr_local_discovery_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+			 int32_t op_ret, int32_t op_errno, dict_t *dict,
+			 dict_t *xdata)
 {
-        if (FILETYPE_DIFFERS (buf, lookup_buf)) {
-                /* mismatching filetypes with same name
-                */
-
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "filetype differs for %s ", local->loc.path);
+        int              ret            = 0;
+        char            *pathinfo       = NULL;
+        gf_boolean_t     is_local        = _gf_false;
+        afr_private_t   *priv           = NULL;
+        int32_t          child_index    = -1;
 
-                local->govinda_gOvinda = 1;
+        if (op_ret != 0) {
+                goto out;
         }
 
-        if (PERMISSION_DIFFERS (buf, lookup_buf)) {
-                /* mismatching permissions */
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "permissions differ for %s ", local->loc.path);
-                local->self_heal.need_metadata_self_heal = _gf_true;
-        }
+	priv = this->private;
+	child_index = (int32_t)(long)cookie;
 
-        if (OWNERSHIP_DIFFERS (buf, lookup_buf)) {
-                /* mismatching permissions */
-                local->self_heal.need_metadata_self_heal = _gf_true;
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "ownership differs for %s ", local->loc.path);
+        ret = dict_get_str (dict, GF_XATTR_PATHINFO_KEY, &pathinfo);
+        if (ret != 0) {
+                goto out;
         }
 
-        if (SIZE_DIFFERS (buf, lookup_buf)
-            && IA_ISREG (buf->ia_type)) {
-                gf_log (this->name, GF_LOG_NORMAL,
-                        "size differs for %s ", local->loc.path);
-                local->self_heal.need_data_self_heal = _gf_true;
+        ret = afr_local_pathinfo (pathinfo, &is_local);
+        if (ret) {
+                goto out;
         }
 
+        /*
+         * Note that one local subvolume will override another here.  The only
+         * way to avoid that would be to retain extra information about whether
+         * the previous read_child is local, and it's just not worth it.  Even
+         * the slowest local subvolume is far preferable to a remote one.
+         */
+        if (is_local) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "selecting local read_child %s",
+                        priv->children[child_index]->name);
+                priv->read_child = child_index;
+        }
+out:
+        STACK_DESTROY(frame->root);
+        return 0;
 }
 
-
 static void
-afr_lookup_done (call_frame_t *frame, xlator_t *this, struct iatt *lookup_buf)
+afr_attempt_local_discovery (xlator_t *this, int32_t child_index)
 {
-        int  unwind = 1;
-        int  source = -1;
-        int  up_count = 0;
-        char sh_type_str[256] = {0,};
+        call_frame_t    *newframe = NULL;
+        loc_t            tmploc = {0,};
+        afr_private_t   *priv = this->private;
 
-        afr_private_t *priv  = NULL;
-        afr_local_t   *local = NULL;
-
-        priv  = this->private;
-        local = frame->local;
-
-        local->cont.lookup.postparent.ia_ino  = local->cont.lookup.parent_ino;
-
-        if (local->cont.lookup.ino) {
-                local->cont.lookup.buf.ia_ino = local->cont.lookup.ino;
-        }
-
-        up_count = afr_up_children_count (priv->child_count, priv->child_up);
-        if (up_count == 1) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Only 1 child up - do not attempt to detect self heal");
-                goto unwind;
-        }
-
-        if (local->op_ret == 0) {
-                /* KLUDGE: assuming DHT will not itransform in
-                   revalidate */
-                if (local->cont.lookup.inode->ino) {
-                        local->cont.lookup.buf.ia_ino =
-                                local->cont.lookup.inode->ino;
-                }
-        }
-
-        if (local->success_count && local->enoent_count) {
-                local->self_heal.need_metadata_self_heal = _gf_true;
-                local->self_heal.need_data_self_heal     = _gf_true;
-                local->self_heal.need_entry_self_heal    = _gf_true;
-                gf_log(this->name, GF_LOG_NORMAL,
-                       "entries are missing in lookup of %s.",
-                       local->loc.path);
-        }
-
-        if (local->success_count) {
-                /* check for split-brain case in previous lookup */
-                if (afr_is_split_brain (this,
-                                        local->cont.lookup.inode)) {
-                        local->self_heal.need_data_self_heal = _gf_true;
-                        gf_log(this->name, GF_LOG_NORMAL,
-                               "split brain detected during lookup of "
-                               "%s.", local->loc.path);
-                }
+        newframe = create_frame(this,this->ctx->pool);
+        if (!newframe) {
+                return;
         }
 
-        if ((local->self_heal.need_metadata_self_heal
-             || local->self_heal.need_data_self_heal
-             || local->self_heal.need_entry_self_heal)
-            && ((!local->cont.lookup.is_revalidate)
-                || (local->op_ret != -1))) {
-
-                if (local->inodelk_count || local->entrylk_count) {
-
-                        /* Someone else is doing self-heal on this file.
-                           So just make a best effort to set the read-subvolume
-                           and return */
+        tmploc.gfid[sizeof(tmploc.gfid)-1] = 1;
+        STACK_WIND_COOKIE (newframe, afr_local_discovery_cbk,
+                           (void *)(long)child_index,
+                           priv->children[child_index],
+                           priv->children[child_index]->fops->getxattr,
+                           &tmploc, GF_XATTR_PATHINFO_KEY, NULL);
+}
 
-                        if (IA_ISREG (local->cont.lookup.inode->ia_type)) {
-                                source = afr_self_heal_get_source (this, local, local->cont.lookup.xattrs);
 
-                                if (source >= 0) {
-                                        afr_set_read_child (this,
-                                                            local->cont.lookup.inode,
-                                                            source);
-                                }
-                        }
-                } else {
-                        if (!local->cont.lookup.inode->ia_type) {
-                                /* fix for RT #602 */
-                                local->cont.lookup.inode->ia_type =
-                                        lookup_buf->ia_type;
-                        }
-
-                        local->self_heal.background = _gf_true;
-                        local->self_heal.type       = local->cont.lookup.buf.ia_type;
-                        local->self_heal.unwind     = afr_self_heal_lookup_unwind;
+int
+afr_lookup_selfheal_wrap (void *opaque)
+{
+	call_frame_t *frame = opaque;
+	afr_local_t *local = NULL;
+	xlator_t *this = NULL;
+	inode_t *inode = NULL;
 
-                        unwind = 0;
+	local = frame->local;
+	this = frame->this;
 
-                        afr_self_heal_type_str_get(&local->self_heal,
-                                                   sh_type_str,
-                                                   sizeof(sh_type_str));
+	afr_selfheal_name (frame->this, local->loc.pargfid, local->loc.name);
 
-                        gf_log (this->name, GF_LOG_NORMAL, "background %s "
-                                "self-heal triggered. path: %s",
-                                sh_type_str, local->loc.path);
+	afr_replies_wipe (local, this->private);
 
-                        afr_self_heal (frame, this);
-                }
-        }
+	inode = afr_selfheal_unlocked_lookup_on (frame, local->loc.parent,
+						 local->loc.name, local->replies,
+						 local->child_up);
+	if (inode)
+		inode_unref (inode);
+	afr_lookup_done (frame, this);
 
-unwind:
-        if (unwind) {
-                AFR_STACK_UNWIND (lookup, frame, local->op_ret,
-                                  local->op_errno,
-                                  local->cont.lookup.inode,
-                                  &local->cont.lookup.buf,
-                                  local->cont.lookup.xattr,
-                                  &local->cont.lookup.postparent);
-        }
+	return 0;
 }
 
 
-/*
- * During a lookup, some errors are more "important" than
- * others in that they must be given higher priority while
- * returning to the user.
- *
- * The hierarchy is ESTALE > ENOENT > others
- *
- */
-
-static gf_boolean_t
-__error_more_important (int32_t old_errno, int32_t new_errno)
+int
+afr_lookup_entry_heal (call_frame_t *frame, xlator_t *this)
 {
-        gf_boolean_t ret = _gf_true;
-
-        /* Nothing should ever overwrite ESTALE */
-        if (old_errno == ESTALE)
-                ret = _gf_false;
-
-        /* Nothing should overwrite ENOENT, except ESTALE */
-        else if ((old_errno == ENOENT) && (new_errno != ESTALE))
-                ret = _gf_false;
-
-        return ret;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	call_frame_t *heal = NULL;
+	int i = 0, first = -1;
+	gf_boolean_t need_heal = _gf_false;
+	struct afr_reply *replies = NULL;
+	int ret = 0;
+
+	local = frame->local;
+	replies = local->replies;
+	priv = this->private;
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].valid)
+			continue;
+
+		if (first == -1) {
+			first = i;
+			continue;
+		}
+
+		if (replies[i].op_ret != replies[first].op_ret) {
+			need_heal = _gf_true;
+			break;
+		}
+
+		if (uuid_compare (replies[i].poststat.ia_gfid,
+				  replies[first].poststat.ia_gfid)) {
+			need_heal = _gf_true;
+			break;
+		}
+	}
+
+	if (need_heal) {
+		heal = copy_frame (frame);
+		if (heal)
+			heal->root->pid = -1;
+		ret = synctask_new (this->ctx->env, afr_lookup_selfheal_wrap,
+				    afr_refresh_selfheal_done, heal, frame);
+		if (ret)
+			goto lookup_done;
+	} else {
+	lookup_done:
+		afr_lookup_done (frame, this);
+	}
+
+	return ret;
 }
 
 
 int
-afr_fresh_lookup_cbk (call_frame_t *frame, void *cookie,
-                      xlator_t *this,  int32_t op_ret,  int32_t op_errno,
-                      inode_t *inode,   struct iatt *buf, dict_t *xattr,
-                      struct iatt *postparent)
+afr_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		int op_ret, int op_errno, inode_t *inode, struct iatt *buf,
+		dict_t *xdata, struct iatt *postparent)
 {
         afr_local_t *   local = NULL;
-        afr_private_t * priv  = NULL;
-        struct iatt *   lookup_buf = NULL;
-
         int             call_count      = -1;
         int             child_index     = -1;
-        int             first_up_child  = -1;
 
-        child_index = (long) cookie;
-        priv = this->private;
+	child_index = (long) cookie;
 
-        LOCK (&frame->lock);
-        {
-                local = frame->local;
+	local = frame->local;
 
-                lookup_buf = &local->cont.lookup.buf;
+	local->replies[child_index].valid = 1;
+	local->replies[child_index].op_ret = op_ret;
+	local->replies[child_index].op_errno = op_errno;
+	if (op_ret != -1) {
+		local->replies[child_index].poststat = *buf;
+		local->replies[child_index].postparent = *postparent;
+		if (xdata)
+			local->replies[child_index].xdata = dict_ref (xdata);
+	}
 
-                if (op_ret == -1) {
-                        if (op_errno == ENOENT)
-                                local->enoent_count++;
+        call_count = afr_frame_return (frame);
+        if (call_count == 0) {
+		afr_lookup_entry_heal (frame, this);
+        }
 
-                        if (__error_more_important (local->op_errno, op_errno))
-                                local->op_errno = op_errno;
+	return 0;
+}
 
-                        if (local->op_errno == ESTALE) {
-                                local->op_ret = -1;
-                        }
 
-                        goto unlock;
-                }
 
-                afr_lookup_collect_xattr (local, this, child_index, xattr);
+static void
+afr_discover_done (call_frame_t *frame, xlator_t *this)
+{
+        afr_private_t       *priv  = NULL;
+        afr_local_t         *local = NULL;
+	int                 i = -1;
+	int                 op_errno = 0;
+	int                 read_subvol = 0;
 
-                first_up_child = afr_first_up_child (priv);
+        priv  = this->private;
+        local = frame->local;
 
-                if (child_index == first_up_child) {
-                        local->cont.lookup.ino =
-                                afr_itransform (buf->ia_ino,
-                                                priv->child_count,
-                                                first_up_child);
-                }
+	for (i = 0; i < priv->child_count; i++) {
+		if (!local->replies[i].valid)
+			continue;
+		if (local->replies[i].op_ret == 0)
+			local->op_ret = 0;
+	}
 
-                if (local->success_count == 0) {
-                        if (local->op_errno != ESTALE)
-                                local->op_ret = op_ret;
+	op_errno = afr_final_errno (frame->local, this->private);
 
-                        local->cont.lookup.inode               = inode_ref (inode);
-                        local->cont.lookup.xattr               = dict_ref (xattr);
-                        local->cont.lookup.xattrs[child_index] = dict_ref (xattr);
-                        local->cont.lookup.postparent          = *postparent;
+        if (local->op_ret < 0) {
+		local->op_errno = op_errno;
+		local->op_ret = -1;
+                goto unwind;
+	}
 
-                        if (priv->first_lookup && inode->ino == 1) {
-                                gf_log (this->name, GF_LOG_NORMAL,
-                                        "added root inode");
-                                priv->root_inode = inode_ref (inode);
-                                priv->first_lookup = 0;
-                        }
+	afr_replies_interpret (frame, this, local->inode);
 
-                        *lookup_buf = *buf;
+	read_subvol = afr_data_subvol_get (local->inode, this, 0, 0);
+	if (read_subvol == -1) {
+		gf_log (this->name, GF_LOG_WARNING, "no read subvols for %s",
+			local->loc.path);
 
-                        lookup_buf->ia_ino = afr_itransform (buf->ia_ino,
-                                                             priv->child_count,
-                                                             child_index);
-                        if (priv->read_child >= 0) {
-                                afr_set_read_child (this,
-                                                    local->cont.lookup.inode,
-                                                    priv->read_child);
-                        } else {
-                                afr_set_read_child (this,
-                                                    local->cont.lookup.inode,
-                                                    child_index);
-                        }
+		for (i = 0; i < priv->child_count; i++) {
+			if (!local->replies[i].valid ||
+			    local->replies[i].op_ret == -1)
+				continue;
+			read_subvol = i;
+			break;
+		}
+	}
 
-                } else {
-                        afr_lookup_self_heal_check (this, local, buf, lookup_buf);
-
-                        if (child_index == local->read_child_index) {
-                                /*
-                                   lookup has succeeded on the read child.
-                                   So use its inode number
-                                */
-                                if (local->cont.lookup.xattr)
-                                        dict_unref (local->cont.lookup.xattr);
-
-                                local->cont.lookup.xattr = dict_ref (xattr);
-                                local->cont.lookup.xattrs[child_index] = dict_ref (xattr);
-                                local->cont.lookup.postparent          = *postparent;
-
-                                *lookup_buf = *buf;
-
-                                if (priv->read_child >= 0) {
-                                        afr_set_read_child (this,
-                                                            local->cont.lookup.inode,
-                                                            priv->read_child);
-                                } else {
-                                        afr_set_read_child (this,
-                                                            local->cont.lookup.inode,
-                                                            local->read_child_index);
-                                }
-                        }
+unwind:
+	if (read_subvol == -1)
+		read_subvol = 0;
 
-                }
+	AFR_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
+			  local->inode, &local->replies[read_subvol].poststat,
+			  local->replies[read_subvol].xdata,
+			  &local->replies[read_subvol].postparent);
+}
 
-                local->success_count++;
-        }
-unlock:
-        UNLOCK (&frame->lock);
 
-        call_count = afr_frame_return (frame);
+int
+afr_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		  int op_ret, int op_errno, inode_t *inode, struct iatt *buf,
+		  dict_t *xdata, struct iatt *postparent)
+{
+        afr_local_t *   local = NULL;
+        int             call_count      = -1;
+        int             child_index     = -1;
+
+	child_index = (long) cookie;
+
+	local = frame->local;
 
+	local->replies[child_index].valid = 1;
+	local->replies[child_index].op_ret = op_ret;
+	local->replies[child_index].op_errno = op_errno;
+	if (op_ret != -1) {
+		local->replies[child_index].poststat = *buf;
+		local->replies[child_index].postparent = *postparent;
+		if (xdata)
+			local->replies[child_index].xdata = dict_ref (xdata);
+	}
+
+        if (local->do_discovery && (op_ret == 0))
+                afr_attempt_local_discovery (this, child_index);
+
+        call_count = afr_frame_return (frame);
         if (call_count == 0) {
-                afr_lookup_done (frame, this, lookup_buf);
+               afr_discover_done (frame, this);
         }
 
-        return 0;
+	return 0;
 }
 
 
 int
-afr_revalidate_lookup_cbk (call_frame_t *frame, void *cookie,
-                           xlator_t *this, int32_t op_ret, int32_t op_errno,
-                           inode_t *inode, struct iatt *buf, dict_t *xattr,
-                           struct iatt *postparent)
+afr_discover_do (call_frame_t *frame, xlator_t *this, int err)
 {
-        afr_local_t *   local = NULL;
-        afr_private_t * priv  = NULL;
-        struct iatt *   lookup_buf = NULL;
+	int ret = 0;
+	int i = 0;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	int call_count = 0;
 
-        int             call_count      = -1;
-        int             child_index     = -1;
-        int             first_up_child  = -1;
+	local = frame->local;
+	priv = this->private;
 
-        child_index = (long) cookie;
-        priv = this->private;
+	if (err) {
+		local->op_errno = -err;
+		ret = -1;
+		goto out;
+	}
 
-        LOCK (&frame->lock);
-        {
-                local = frame->local;
+	call_count = local->call_count = AFR_COUNT (local->child_up,
+						    priv->child_count);
 
-                lookup_buf = &local->cont.lookup.buf;
+        ret = afr_lookup_xattr_req_prepare (local, this, local->xattr_req,
+					    &local->loc);
+        if (ret) {
+                local->op_errno = -ret;
+		ret = -1;
+                goto out;
+        }
 
-                if (op_ret == -1) {
-                        if (op_errno == ENOENT)
-                                local->enoent_count++;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND_COOKIE (frame, afr_discover_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->lookup,
+                                           &local->loc, local->xattr_req);
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-                        if (__error_more_important (local->op_errno, op_errno))
-                            local->op_errno = op_errno;
+	return 0;
+out:
+	AFR_STACK_UNWIND (lookup, frame, -1, local->op_errno, 0, 0, 0, 0);
+	return 0;
+}
 
-                            if (local->op_errno == ESTALE) {
-                                    local->op_ret = -1;
-                            }
 
-                            goto unlock;
-                }
+int
+afr_discover (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr_req)
+{
+	int op_errno = ENOMEM;
+	afr_private_t *priv = NULL;
+	afr_local_t *local = NULL;
+	int event = 0;
 
-                afr_lookup_collect_xattr (local, this, child_index, xattr);
+	priv = this->private;
 
-                first_up_child = afr_first_up_child (priv);
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-                if (child_index == first_up_child) {
-                        local->cont.lookup.ino =
-                                afr_itransform (buf->ia_ino,
-                                                priv->child_count,
-                                                first_up_child);
-                }
+        if (!local->call_count) {
+                op_errno = ENOTCONN;
+                goto out;
+        }
 
-                /* in case of revalidate, we need to send stat of the
-                 * child whose stat was sent during the first lookup.
-                 * (so that time stamp does not vary with revalidate.
-                 * in case it is down, stat of the fist success will
-                 * be replied */
+	if (__is_root_gfid (loc->inode->gfid)) {
+		if (!this->itable)
+			this->itable = loc->inode->table;
+		if (!priv->root_inode)
+			priv->root_inode = inode_ref (loc->inode);
 
-                /* inode number should be preserved across revalidates */
+		if (priv->choose_local && !priv->did_discovery) {
+			/* Logic to detect which subvolumes of AFR are
+			   local, in order to prefer them for reads
+			*/
+			local->do_discovery = _gf_true;
+                        priv->did_discovery = _gf_true;
+                }
+	}
 
-                if (local->success_count == 0) {
-                        if (local->op_errno != ESTALE)
-                                local->op_ret = op_ret;
+        local->op = GF_FOP_LOOKUP;
 
-                        local->cont.lookup.inode               = inode_ref (inode);
-                        local->cont.lookup.xattr               = dict_ref (xattr);
-                        local->cont.lookup.xattrs[child_index] = dict_ref (xattr);
-                        local->cont.lookup.postparent          = *postparent;
+        loc_copy (&local->loc, loc);
 
-                        *lookup_buf = *buf;
+	local->inode = inode_ref (loc->inode);
 
-                        lookup_buf->ia_ino = afr_itransform (buf->ia_ino,
-                                                             priv->child_count,
-                                                             child_index);
+	if (xattr_req)
+		/* If xattr_req was null, afr_lookup_xattr_req_prepare() will
+		   allocate one for us */
+		local->xattr_req = dict_ref (xattr_req);
 
-                        if (priv->read_child >= 0) {
-                                afr_set_read_child (this,
-                                                    local->cont.lookup.inode,
-                                                    priv->read_child);
-                        } else {
-                                afr_set_read_child (this,
-                                                    local->cont.lookup.inode,
-                                                    child_index);
-                        }
+	if (uuid_is_null (loc->inode->gfid)) {
+		afr_discover_do (frame, this, 0);
+		return 0;
+	}
 
-                } else {
-                        afr_lookup_self_heal_check (this, local, buf, lookup_buf);
+	afr_read_subvol_get (loc->inode, this, NULL, &event,
+			     AFR_DATA_TRANSACTION);
 
-                        if (child_index == local->read_child_index) {
+	if (event != local->event_generation)
+		afr_inode_refresh (frame, this, loc->inode, afr_discover_do);
+	else
+		afr_discover_do (frame, this, 0);
 
-                                /*
-                                   lookup has succeeded on the read child.
-                                   So use its inode number
-                                */
+	return 0;
+out:
+	AFR_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+	return 0;
+}
 
-                                if (local->cont.lookup.xattr)
-                                        dict_unref (local->cont.lookup.xattr);
 
-                                local->cont.lookup.xattr               = dict_ref (xattr);
-                                local->cont.lookup.xattrs[child_index] = dict_ref (xattr);
-                                local->cont.lookup.postparent          = *postparent;
+int
+afr_lookup_do (call_frame_t *frame, xlator_t *this, int err)
+{
+	int ret = 0;
+	int i = 0;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	int call_count = 0;
 
-                                *lookup_buf = *buf;
+	local = frame->local;
+	priv = this->private;
 
-                                if (priv->read_child >= 0) {
-                                        afr_set_read_child (this,
-                                                            local->cont.lookup.inode,
-                                                            priv->read_child);
-                                } else {
-                                        afr_set_read_child (this,
-                                                            local->cont.lookup.inode,
-                                                            local->read_child_index);
-                                }
-                        }
+	if (err < 0) {
+		local->op_errno = -err;
+		ret = -1;
+		goto out;
+	}
 
-                }
+	call_count = local->call_count = AFR_COUNT (local->child_up,
+						    priv->child_count);
 
-                local->success_count++;
+        ret = afr_lookup_xattr_req_prepare (local, this, local->xattr_req,
+					    &local->loc);
+        if (ret) {
+                local->op_errno = -ret;
+		ret = -1;
+                goto out;
         }
-unlock:
-        UNLOCK (&frame->lock);
-
-        call_count = afr_frame_return (frame);
 
-        if (call_count == 0) {
-                afr_lookup_done (frame, this, lookup_buf);
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND_COOKIE (frame, afr_lookup_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->lookup,
+                                           &local->loc, local->xattr_req);
+                        if (!--call_count)
+                                break;
+                }
         }
-
-        return 0;
+	return 0;
+out:
+	AFR_STACK_UNWIND (lookup, frame, -1, local->op_errno, 0, 0, 0, 0);
+	return 0;
 }
 
+/*
+ * afr_lookup()
+ *
+ * The goal here is to figure out what the element getting looked up is.
+ * i.e what is the GFID, inode type and a conservative estimate of the
+ * inode attributes are.
+ *
+ * As we lookup, operations may be underway on the entry name and the
+ * inode. In lookup() we are primarily concerned only with the entry
+ * operations. If the entry is getting unlinked or renamed, we detect
+ * what operation is underway by querying for on-going transactions and
+ * pending self-healing on the entry through xdata.
+ *
+ * If the entry is a file/dir, it may need self-heal and/or in a
+ * split-brain condition. Lookup is not the place to worry about these
+ * conditions. Outcast marking will naturally handle them in the read
+ * paths.
+ *
+ * Here is a brief goal of what we are trying to achieve:
+ *
+ * - LOOKUP on all subvolumes concurrently, querying on-going transaction
+ *   and pending self-heal info from the servers.
+ *
+ * - If all servers reply the same inode type and GFID, the overall call
+ *   MUST be a success.
+ *
+ * - If inode types or GFIDs mismatch, and there IS either an on-going
+ *   transaction or pending self-heal, inspect what the nature of the
+ *   transaction or pending heal is, and select the appropriate subvolume's
+ *   reply as the winner.
+ *
+ * - If inode types or GFIDs mismatch, and there are no on-going transactions
+ *   or pending self-heal on the entry name on any of the servers, fail the
+ *   lookup with EIO. Something has gone wrong beyond reasonable action.
+ */
 
 int
-afr_lookup (call_frame_t *frame, xlator_t *this,
-            loc_t *loc, dict_t *xattr_req)
+afr_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr_req)
 {
-        afr_private_t *priv = NULL;
         afr_local_t   *local = NULL;
-        int            ret = -1;
-        int            i = 0;
-
-        fop_lookup_cbk_t callback;
-
-        int call_count = 0;
-
-        uint64_t       ctx;
-
         int32_t        op_errno = 0;
+	int            event = 0;
 
-        priv = this->private;
-
-        ALLOC_OR_GOTO (local, afr_local_t, out);
+	if (!loc->parent) {
+		afr_discover (frame, this, loc, xattr_req);
+		return 0;
+	}
 
-        local->op_ret = -1;
+	if (__is_root_gfid (loc->parent->gfid)) {
+		if (!strcmp (loc->name, GF_REPLICATE_TRASH_DIR)) {
+			op_errno = EPERM;
+			goto out;
+		}
+	}
 
-        frame->local = local;
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-        if (!strcmp (loc->path, "/" GF_REPLICATE_TRASH_DIR)) {
-                op_errno = ENOENT;
+        if (!local->call_count) {
+                op_errno = ENOTCONN;
                 goto out;
         }
 
+        local->op = GF_FOP_LOOKUP;
+
         loc_copy (&local->loc, loc);
 
-        ret = inode_ctx_get (loc->inode, this, &ctx);
-        if (ret == 0) {
-                /* lookup is a revalidate */
+	local->inode = inode_ref (loc->inode);
 
-                callback = afr_revalidate_lookup_cbk;
+	if (xattr_req)
+		/* If xattr_req was null, afr_lookup_xattr_req_prepare() will
+		   allocate one for us */
+		local->xattr_req = dict_ref (xattr_req);
 
-                local->cont.lookup.is_revalidate = _gf_true;
-                local->read_child_index          = afr_read_child (this,
-                                                                   loc->inode);
-        } else {
-                callback = afr_fresh_lookup_cbk;
+	afr_read_subvol_get (loc->parent, this, NULL, &event,
+			     AFR_DATA_TRANSACTION);
 
-                LOCK (&priv->read_child_lock);
-                {
-                        local->read_child_index = (++priv->read_child_rr)
-                                % (priv->child_count);
-                }
-                UNLOCK (&priv->read_child_lock);
-        }
+	if (event != local->event_generation)
+		afr_inode_refresh (frame, this, loc->parent, afr_lookup_do);
+	else
+		afr_lookup_do (frame, this, 0);
 
-        if (loc->parent)
-                local->cont.lookup.parent_ino = loc->parent->ino;
+	return 0;
+out:
+	AFR_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
 
-        local->child_up = memdup (priv->child_up, priv->child_count);
+        return 0;
+}
 
-        local->cont.lookup.xattrs = GF_CALLOC (priv->child_count,
-                                    sizeof (*local->cont.lookup.xattr),
-                                    gf_afr_mt_dict_t);
 
-        local->call_count = afr_up_children_count (priv->child_count,
-                                                   local->child_up);
-        call_count = local->call_count;
+/* {{{ open */
 
-        if (local->call_count == 0) {
-                ret      = -1;
-                op_errno = ENOTCONN;
-                goto out;
-        }
+afr_fd_ctx_t *
+__afr_fd_ctx_get (fd_t *fd, xlator_t *this)
+{
+        uint64_t       ctx = 0;
+        int            ret = 0;
+        afr_fd_ctx_t  *fd_ctx = NULL;
 
-        /* By default assume ENOTCONN. On success it will be set to 0. */
-        local->op_errno = ENOTCONN;
+        ret = __fd_ctx_get (fd, this, &ctx);
 
-        if (xattr_req == NULL)
-                local->xattr_req = dict_new ();
-        else
-                local->xattr_req = dict_ref (xattr_req);
+        if (ret < 0) {
+                ret = __afr_fd_ctx_set (this, fd);
+                if (ret < 0)
+                        goto out;
 
-        for (i = 0; i < priv->child_count; i++) {
-                ret = dict_set_uint64 (local->xattr_req, priv->pending_key[i],
-                                       3 * sizeof(int32_t));
+                ret = __fd_ctx_get (fd, this, &ctx);
                 if (ret < 0)
-                        gf_log (this->name, GF_LOG_WARNING,
-                                "Unable to set dict value.");
-                /* 3 = data+metadata+entry */
+                        goto out;
         }
 
-        ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_INODELK_COUNT, 0);
-        if (ret < 0) {
-                gf_log (this->name, GF_LOG_WARNING,
-                        "Unable to set dict value.");
-        }
+        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+out:
+        return fd_ctx;
+}
 
-        ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_ENTRYLK_COUNT, 0);
-        if (ret < 0) {
-                gf_log (this->name, GF_LOG_WARNING,
-                        "Unable to set dict value.");
-        }
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (local->child_up[i]) {
-                        STACK_WIND_COOKIE (frame, callback, (void *) (long) i,
-                                           priv->children[i],
-                                           priv->children[i]->fops->lookup,
-                                           loc, local->xattr_req);
-                        if (!--call_count)
-                                break;
-                }
-        }
+afr_fd_ctx_t *
+afr_fd_ctx_get (fd_t *fd, xlator_t *this)
+{
+        afr_fd_ctx_t  *fd_ctx = NULL;
 
-        ret = 0;
-out:
-        if (ret == -1)
-                AFR_STACK_UNWIND (lookup, frame, -1, op_errno,
-                                  NULL, NULL, NULL, NULL);
+        LOCK(&fd->lock);
+        {
+                fd_ctx = __afr_fd_ctx_get (fd, this);
+        }
+        UNLOCK(&fd->lock);
 
-        return 0;
+        return fd_ctx;
 }
 
 
-/* {{{ open */
-
 int
-afr_fd_ctx_set (xlator_t *this, fd_t *fd)
+__afr_fd_ctx_set (xlator_t *this, fd_t *fd)
 {
-        afr_private_t * priv = NULL;
-
-        int ret    = 0;
-
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
+        afr_private_t * priv   = NULL;
+        int             ret    = -1;
+        uint64_t        ctx    = 0;
+        afr_fd_ctx_t *  fd_ctx = NULL;
+	int             i = 0;
 
         VALIDATE_OR_GOTO (this->private, out);
         VALIDATE_OR_GOTO (fd, out);
 
         priv = this->private;
 
-        LOCK (&fd->lock);
-        {
-                ret = __fd_ctx_get (fd, this, &ctx);
-
-                if (ret == 0)
-                        goto unlock;
+        ret = __fd_ctx_get (fd, this, &ctx);
 
-                fd_ctx = GF_CALLOC (1, sizeof (afr_fd_ctx_t),
-                                    gf_afr_mt_afr_fd_ctx_t);
-                if (!fd_ctx) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
+        if (ret == 0)
+                goto out;
 
-                        ret = -ENOMEM;
-                        goto unlock;
-                }
+        fd_ctx = GF_CALLOC (1, sizeof (afr_fd_ctx_t),
+                            gf_afr_mt_afr_fd_ctx_t);
+        if (!fd_ctx) {
+                ret = -ENOMEM;
+                goto out;
+        }
 
-                fd_ctx->pre_op_done = GF_CALLOC (sizeof (*fd_ctx->pre_op_done),
-                                                 priv->child_count,
-                                                 gf_afr_mt_char);
-                if (!fd_ctx->pre_op_done) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        ret = -ENOMEM;
-                        goto unlock;
-                }
+	for (i = 0; i < AFR_NUM_CHANGE_LOGS; i++) {
+		fd_ctx->pre_op_done[i] = GF_CALLOC (sizeof (*fd_ctx->pre_op_done[i]),
+						    priv->child_count,
+						    gf_afr_mt_int32_t);
+		if (!fd_ctx->pre_op_done[i]) {
+			ret = -ENOMEM;
+			goto out;
+		}
+	}
+
+        fd_ctx->opened_on = GF_CALLOC (sizeof (*fd_ctx->opened_on),
+                                       priv->child_count,
+                                       gf_afr_mt_int32_t);
+        if (!fd_ctx->opened_on) {
+                ret = -ENOMEM;
+                goto out;
+        }
 
-                fd_ctx->pre_op_piggyback = GF_CALLOC (sizeof (*fd_ctx->pre_op_piggyback),
-                                                      priv->child_count,
-                                                      gf_afr_mt_char);
-                if (!fd_ctx->pre_op_piggyback) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        ret = -ENOMEM;
-                        goto unlock;
-                }
+	for (i = 0; i < priv->child_count; i++) {
+		if (fd_is_anonymous (fd))
+			fd_ctx->opened_on[i] = AFR_FD_OPENED;
+		else
+			fd_ctx->opened_on[i] = AFR_FD_NOT_OPENED;
+	}
 
-                fd_ctx->opened_on = GF_CALLOC (sizeof (*fd_ctx->opened_on),
-                                               priv->child_count,
-                                               gf_afr_mt_char);
-                if (!fd_ctx->opened_on) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        ret = -ENOMEM;
-                        goto unlock;
-                }
+        fd_ctx->lock_piggyback = GF_CALLOC (sizeof (*fd_ctx->lock_piggyback),
+                                            priv->child_count,
+                                            gf_afr_mt_char);
+        if (!fd_ctx->lock_piggyback) {
+                ret = -ENOMEM;
+                goto out;
+        }
 
-                fd_ctx->up_count   = priv->up_count;
-                fd_ctx->down_count = priv->down_count;
+        fd_ctx->lock_acquired = GF_CALLOC (sizeof (*fd_ctx->lock_acquired),
+                                           priv->child_count,
+                                           gf_afr_mt_char);
+        if (!fd_ctx->lock_acquired) {
+                ret = -ENOMEM;
+                goto out;
+        }
 
-                fd_ctx->locked_on = GF_CALLOC (sizeof (*fd_ctx->locked_on),
-                                               priv->child_count,
-                                               gf_afr_mt_char);
-                if (!fd_ctx->locked_on) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        ret = -ENOMEM;
-                        goto unlock;
-                }
+	pthread_mutex_init (&fd_ctx->delay_lock, NULL);
 
-                ret = __fd_ctx_set (fd, this, (uint64_t)(long) fd_ctx);
+        INIT_LIST_HEAD (&fd_ctx->eager_locked);
 
-                INIT_LIST_HEAD (&fd_ctx->entries);
-        }
-unlock:
-        UNLOCK (&fd->lock);
+        ret = __fd_ctx_set (fd, this, (uint64_t)(long) fd_ctx);
+        if (ret)
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "failed to set fd ctx (%p)", fd);
 out:
         return ret;
 }
 
-/* {{{ flush */
 
 int
-afr_flush_unwind (call_frame_t *frame, xlator_t *this)
+afr_fd_ctx_set (xlator_t *this, fd_t *fd)
 {
-        afr_local_t *   local = NULL;
-        call_frame_t   *main_frame = NULL;
-
-        local = frame->local;
+        int ret = -1;
 
-        LOCK (&frame->lock);
+        LOCK (&fd->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);
+                ret = __afr_fd_ctx_set (this, fd);
         }
+        UNLOCK (&fd->lock);
 
-        return 0;
+        return ret;
 }
 
+/* {{{ flush */
 
 int
-afr_flush_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno)
+afr_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        afr_local_t *   local = NULL;
-        afr_private_t * priv  = NULL;
-
-        int call_count  = -1;
-        int child_index = (long) cookie;
-        int need_unwind = 0;
+        afr_local_t *local = NULL;
+        int call_count = -1;
 
         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;
+			local->op_ret = op_ret;
+			if (!local->xdata_rsp && xdata)
+				local->xdata_rsp = dict_ref (xdata);
+		} else {
+			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, local->xdata_rsp);
 
         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 (priv->child_count, local->child_up);
-
-        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);
-
+                                           local->fd, xdata);
                         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)
+afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
 {
-        afr_private_t * priv  = NULL;
-        afr_local_t   * local = NULL;
-
-        call_frame_t  * transaction_frame = NULL;
-
-        int ret        = -1;
-
-        int op_ret   = -1;
-        int op_errno = 0;
-
-        int call_count = 0;
-
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
+        afr_local_t   *local = NULL;
+        call_stub_t   *stub = NULL;
+        int            op_errno   = ENOMEM;
 
-        priv = this->private;
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-        ALLOC_OR_GOTO (local, afr_local_t, out);
+	if (!local->call_count) {
+		op_errno = ENOTCONN;
+		goto out;
+	}
 
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
-                goto out;
-        }
+	local->fd = fd_ref(fd);
 
-        call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-        transaction_frame = copy_frame (frame);
-        if (!transaction_frame) {
-                op_errno = ENOMEM;
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
+        stub = fop_flush_stub (frame, afr_flush_wrapper, fd, xdata);
+        if (!stub)
                 goto out;
-        }
-
-        transaction_frame->local = local;
 
-        local->op = GF_FOP_FLUSH;
+        afr_delayed_changelog_wake_resume (this, fd, stub);
 
-        local->transaction.fop    = afr_flush_wind;
-        local->transaction.done   = afr_flush_done;
-        local->transaction.unwind = afr_flush_unwind;
-
-        local->fd                 = fd_ref (fd);
-
-        local->transaction.main_frame = frame;
-        local->transaction.start  = 0;
-        local->transaction.len    = 0;
-
-        afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
-
-
-        op_ret = 0;
+	return 0;
 out:
-        if (op_ret == -1) {
-                if (transaction_frame)
-                        AFR_STACK_DESTROY (transaction_frame);
-
-                AFR_STACK_UNWIND (flush, frame, op_ret, op_errno);
-        }
-
+	AFR_STACK_UNWIND (flush, frame, -1, op_errno, NULL);
         return 0;
 }
 
@@ -1421,23 +2110,31 @@ afr_cleanup_fd_ctx (xlator_t *this, fd_t *fd)
         uint64_t        ctx = 0;
         afr_fd_ctx_t    *fd_ctx = NULL;
         int             ret = 0;
+	int             i = 0;
 
         ret = fd_ctx_get (fd, this, &ctx);
-
         if (ret < 0)
                 goto out;
 
         fd_ctx = (afr_fd_ctx_t *)(long) ctx;
 
         if (fd_ctx) {
-                if (fd_ctx->pre_op_done)
-                        GF_FREE (fd_ctx->pre_op_done);
+                //no need to take any locks
+                if (!list_empty (&fd_ctx->eager_locked))
+                        gf_log (this->name, GF_LOG_WARNING, "%s: Stale "
+                                "Eager-lock stubs found",
+                                uuid_utoa (fd->inode->gfid));
 
-                if (fd_ctx->opened_on)
-                        GF_FREE (fd_ctx->opened_on);
+		for (i = 0; i < AFR_NUM_CHANGE_LOGS; i++)
+			GF_FREE (fd_ctx->pre_op_done[i]);
 
-                if (fd_ctx->locked_on)
-                        GF_FREE (fd_ctx->locked_on);
+                GF_FREE (fd_ctx->opened_on);
+
+                GF_FREE (fd_ctx->lock_piggyback);
+
+                GF_FREE (fd_ctx->lock_acquired);
+
+		pthread_mutex_destroy (&fd_ctx->delay_lock);
 
                 GF_FREE (fd_ctx);
         }
@@ -1450,24 +2147,8 @@ out:
 int
 afr_release (xlator_t *this, fd_t *fd)
 {
-        afr_locked_fd_t *locked_fd = NULL;
-        afr_locked_fd_t *tmp       = NULL;
-        afr_private_t   *priv      = NULL;
-
-        priv = this->private;
-
         afr_cleanup_fd_ctx (this, fd);
 
-        list_for_each_entry_safe (locked_fd, tmp, &priv->saved_fds,
-                                  list) {
-
-                if (locked_fd->fd == fd) {
-                        list_del_init (&locked_fd->list);
-                        GF_FREE (locked_fd);
-                }
-
-        }
-
         return 0;
 }
 
@@ -1475,56 +2156,87 @@ afr_release (xlator_t *this, fd_t *fd)
 /* {{{ fsync */
 
 int
+afr_fsync_unwind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                      struct iatt *postbuf, dict_t *xdata)
+{
+        AFR_STACK_UNWIND (fsync, frame, op_ret, op_errno, prebuf, postbuf,
+                          xdata);
+        return 0;
+}
+
+int
 afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-               struct iatt *postbuf)
+               struct iatt *postbuf, dict_t *xdata)
 {
         afr_local_t *local = NULL;
-
         int call_count = -1;
-
         int child_index = (long) cookie;
-        int read_child  = 0;
+	int read_subvol = 0;
+	call_stub_t *stub = NULL;
 
         local = frame->local;
 
-        read_child = afr_read_child (this, local->fd->inode);
+	read_subvol = afr_data_subvol_get (local->inode, this, 0, 0);
 
         LOCK (&frame->lock);
         {
-                if (child_index == read_child) {
-                        local->read_child_returned = _gf_true;
-                }
-
                 if (op_ret == 0) {
-                        local->op_ret = 0;
+                        if (local->op_ret == -1) {
+				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;
+				if (xdata)
+					local->xdata_rsp = dict_ref (xdata);
                         }
 
-                        local->success_count++;
-                }
-
-                local->op_errno = op_errno;
+                        if (child_index == read_subvol) {
+                                local->cont.inode_wfop.prebuf  = *prebuf;
+                                local->cont.inode_wfop.postbuf = *postbuf;
+				if (xdata) {
+					if (local->xdata_rsp)
+						dict_unref (local->xdata_rsp);
+					local->xdata_rsp = dict_ref (xdata);
+				}
+                        }
+                } else {
+			local->op_errno = op_errno;
+		}
         }
         UNLOCK (&frame->lock);
 
         call_count = afr_frame_return (frame);
 
         if (call_count == 0) {
-                local->cont.fsync.prebuf.ia_ino  = local->cont.fsync.ino;
-                local->cont.fsync.postbuf.ia_ino = local->cont.fsync.ino;
-
-                AFR_STACK_UNWIND (fsync, frame, local->op_ret, local->op_errno,
-                                  &local->cont.fsync.prebuf,
-                                  &local->cont.fsync.postbuf);
+		/* 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,
+                                           local->xdata_rsp);
+		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;
@@ -1532,38 +2244,34 @@ afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 
 int
-afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
-           int32_t datasync)
+afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+	   dict_t *xdata)
 {
-        afr_private_t *priv = NULL;
+	afr_private_t *priv = NULL;
         afr_local_t *local = NULL;
-
-        int ret = -1;
-
         int i = 0;
         int32_t call_count = 0;
-        int32_t op_ret   = -1;
-        int32_t op_errno = 0;
+        int32_t op_errno = ENOMEM;
 
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
+	priv = this->private;
 
-        priv = this->private;
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-        ALLOC_OR_GOTO (local, afr_local_t, out);
+        call_count = local->call_count;
+	if (!call_count) {
+		op_errno = ENOTCONN;
+		goto out;
+	}
 
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
-                goto out;
-        }
+        local->fd = fd_ref (fd);
 
-        call_count = local->call_count;
-        frame->local = local;
+	if (afr_fd_has_witnessed_unstable_write (this, fd)) {
+		/* don't care. we only wanted to CLEAR the bit */
+	}
 
-        local->fd             = fd_ref (fd);
-        local->cont.fsync.ino = fd->inode->ino;
+	local->inode = inode_ref (fd->inode);
 
         for (i = 0; i < priv->child_count; i++) {
                 if (local->child_up[i]) {
@@ -1571,17 +2279,16 @@ afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
                                            (void *) (long) i,
                                            priv->children[i],
                                            priv->children[i]->fops->fsync,
-                                           fd, datasync);
+                                           fd, datasync, xdata);
                         if (!--call_count)
                                 break;
                 }
         }
 
-        op_ret = 0;
+	return 0;
 out:
-        if (op_ret == -1) {
-                AFR_STACK_UNWIND (fsync, frame, op_ret, op_errno, NULL, NULL);
-        }
+	AFR_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL, NULL);
+
         return 0;
 }
 
@@ -1589,22 +2296,24 @@ out:
 
 /* {{{ fsync */
 
-int32_t
-afr_fsyncdir_cbk (call_frame_t *frame, void *cookie,
-                  xlator_t *this, int32_t op_ret, int32_t op_errno)
+int
+afr_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		  int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
         afr_local_t *local = NULL;
-
         int call_count = -1;
 
         local = frame->local;
 
         LOCK (&frame->lock);
         {
-                if (op_ret == 0)
+                if (op_ret == 0) {
                         local->op_ret = 0;
-
-                local->op_errno = op_errno;
+			if (!local->xdata_rsp && xdata)
+				local->xdata_rsp = dict_ref (xdata);
+		} else {
+			local->op_errno = op_errno;
+		}
         }
         UNLOCK (&frame->lock);
 
@@ -1612,59 +2321,49 @@ afr_fsyncdir_cbk (call_frame_t *frame, void *cookie,
 
         if (call_count == 0)
                 AFR_STACK_UNWIND (fsyncdir, frame, local->op_ret,
-                                  local->op_errno);
+				  local->op_errno, local->xdata_rsp);
 
         return 0;
 }
 
 
-int32_t
-afr_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
-              int32_t datasync)
+int
+afr_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+	      dict_t *xdata)
 {
-        afr_private_t *priv = NULL;
+	afr_private_t *priv = NULL;
         afr_local_t *local = NULL;
-
-        int ret = -1;
-
         int i = 0;
         int32_t call_count = 0;
-        int32_t op_ret   = -1;
-        int32_t op_errno = 0;
-
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
+        int32_t op_errno = ENOMEM;
 
-        priv = this->private;
-
-        ALLOC_OR_GOTO (local, afr_local_t, out);
+	priv = this->private;
 
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
-                goto out;
-        }
+        local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
         call_count = local->call_count;
-        frame->local = local;
+	if (!call_count) {
+		op_errno = ENOTCONN;
+		goto out;
+	}
 
         for (i = 0; i < priv->child_count; i++) {
                 if (local->child_up[i]) {
                         STACK_WIND (frame, afr_fsyncdir_cbk,
                                     priv->children[i],
                                     priv->children[i]->fops->fsyncdir,
-                                    fd, datasync);
+                                    fd, datasync, xdata);
                         if (!--call_count)
                                 break;
                 }
         }
 
-        op_ret = 0;
+	return 0;
 out:
-        if (op_ret == -1) {
-                AFR_STACK_UNWIND (fsyncdir, frame, op_ret, op_errno);
-        }
+	AFR_STACK_UNWIND (fsyncdir, frame, -1, op_errno, NULL);
+
         return 0;
 }
 
@@ -1675,18 +2374,24 @@ out:
 int32_t
 afr_xattrop_cbk (call_frame_t *frame, void *cookie,
                  xlator_t *this, int32_t op_ret, int32_t op_errno,
-                 dict_t *xattr)
+                 dict_t *xattr, dict_t *xdata)
 {
         afr_local_t *local = NULL;
-
         int call_count = -1;
 
         local = frame->local;
 
         LOCK (&frame->lock);
         {
-                if (op_ret == 0)
+                if (op_ret == 0) {
+                        if (!local->cont.xattrop.xattr)
+                                local->cont.xattrop.xattr = dict_ref (xattr);
+
+			if (!local->xdata_rsp && xdata)
+				local->xdata_rsp = dict_ref (xdata);
+
                         local->op_ret = 0;
+                }
 
                 local->op_errno = op_errno;
         }
@@ -1696,7 +2401,7 @@ afr_xattrop_cbk (call_frame_t *frame, void *cookie,
 
         if (call_count == 0)
                 AFR_STACK_UNWIND (xattrop, frame, local->op_ret, local->op_errno,
-                                  xattr);
+                local->cont.xattrop.xattr, local->xdata_rsp);
 
         return 0;
 }
@@ -1704,51 +2409,41 @@ afr_xattrop_cbk (call_frame_t *frame, void *cookie,
 
 int32_t
 afr_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
-             gf_xattrop_flags_t optype, dict_t *xattr)
+             gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
 {
         afr_private_t *priv = NULL;
         afr_local_t *local  = NULL;
-
-        int ret = -1;
-
         int i = 0;
         int32_t call_count = 0;
-        int32_t op_ret   = -1;
-        int32_t op_errno = 0;
-
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
+        int32_t op_errno = ENOMEM;
 
         priv = this->private;
 
-        ALLOC_OR_GOTO (local, afr_local_t, out);
-
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
-                goto out;
-        }
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
         call_count = local->call_count;
-        frame->local = local;
+	if (!call_count) {
+		op_errno = ENOTCONN;
+		goto out;
+	}
 
         for (i = 0; i < priv->child_count; i++) {
                 if (local->child_up[i]) {
                         STACK_WIND (frame, afr_xattrop_cbk,
                                     priv->children[i],
                                     priv->children[i]->fops->xattrop,
-                                    loc, optype, xattr);
+                                    loc, optype, xattr, xdata);
                         if (!--call_count)
                                 break;
                 }
         }
 
-        op_ret = 0;
+	return 0;
 out:
-        if (op_ret == -1) {
-                AFR_STACK_UNWIND (xattrop, frame, op_ret, op_errno, NULL);
-        }
+	AFR_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL, NULL);
+
         return 0;
 }
 
@@ -1759,7 +2454,7 @@ out:
 int32_t
 afr_fxattrop_cbk (call_frame_t *frame, void *cookie,
                   xlator_t *this, int32_t op_ret, int32_t op_errno,
-                  dict_t *xattr)
+                  dict_t *xattr, dict_t *xdata)
 {
         afr_local_t *local = NULL;
 
@@ -1769,8 +2464,14 @@ afr_fxattrop_cbk (call_frame_t *frame, void *cookie,
 
         LOCK (&frame->lock);
         {
-                if (op_ret == 0)
+                if (op_ret == 0) {
+                        if (!local->cont.fxattrop.xattr)
+                                local->cont.fxattrop.xattr = dict_ref (xattr);
+
+			if (!local->xdata_rsp && xdata)
+				local->xdata_rsp = dict_ref (xdata);
                         local->op_ret = 0;
+                }
 
                 local->op_errno = op_errno;
         }
@@ -1780,7 +2481,7 @@ afr_fxattrop_cbk (call_frame_t *frame, void *cookie,
 
         if (call_count == 0)
                 AFR_STACK_UNWIND (fxattrop, frame, local->op_ret, local->op_errno,
-                                  xattr);
+                                  local->cont.fxattrop.xattr, local->xdata_rsp);
 
         return 0;
 }
@@ -1788,51 +2489,41 @@ afr_fxattrop_cbk (call_frame_t *frame, void *cookie,
 
 int32_t
 afr_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
-              gf_xattrop_flags_t optype, dict_t *xattr)
+              gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
 {
         afr_private_t *priv = NULL;
         afr_local_t *local  = NULL;
-
-        int ret = -1;
-
         int i = 0;
         int32_t call_count = 0;
-        int32_t op_ret   = -1;
         int32_t op_errno = 0;
 
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
-
         priv = this->private;
 
-        ALLOC_OR_GOTO (local, afr_local_t, out);
-
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
                 goto out;
-        }
 
         call_count = local->call_count;
-        frame->local = local;
+	if (!call_count) {
+		op_errno = ENOTCONN;
+		goto out;
+	}
 
         for (i = 0; i < priv->child_count; i++) {
                 if (local->child_up[i]) {
                         STACK_WIND (frame, afr_fxattrop_cbk,
                                     priv->children[i],
                                     priv->children[i]->fops->fxattrop,
-                                    fd, optype, xattr);
+                                    fd, optype, xattr, xdata);
                         if (!--call_count)
                                 break;
                 }
         }
 
-        op_ret = 0;
+	return 0;
 out:
-        if (op_ret == -1) {
-                AFR_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, NULL);
-        }
+	AFR_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL, NULL);
+
         return 0;
 }
 
@@ -1840,12 +2531,11 @@ out:
 
 
 int32_t
-afr_inodelk_cbk (call_frame_t *frame, void *cookie,
-                 xlator_t *this, int32_t op_ret, int32_t op_errno)
+afr_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		 int32_t op_ret, int32_t op_errno, dict_t *xdata)
 
 {
         afr_local_t *local = NULL;
-
         int call_count = -1;
 
         local = frame->local;
@@ -1863,7 +2553,7 @@ afr_inodelk_cbk (call_frame_t *frame, void *cookie,
 
         if (call_count == 0)
                 AFR_STACK_UNWIND (inodelk, frame, local->op_ret,
-                                  local->op_errno);
+                                  local->op_errno, xdata);
 
         return 0;
 }
@@ -1871,63 +2561,53 @@ afr_inodelk_cbk (call_frame_t *frame, void *cookie,
 
 int32_t
 afr_inodelk (call_frame_t *frame, xlator_t *this,
-             const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *flock)
+             const char *volume, loc_t *loc, int32_t cmd,
+             struct gf_flock *flock, dict_t *xdata)
 {
         afr_private_t *priv = NULL;
         afr_local_t *local  = NULL;
-
-        int ret = -1;
-
         int i = 0;
         int32_t call_count = 0;
-        int32_t op_ret   = -1;
-        int32_t op_errno = 0;
-
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
+        int32_t op_errno = ENOMEM;
 
         priv = this->private;
 
-        ALLOC_OR_GOTO (local, afr_local_t, out);
-
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
+        local = AFR_FRAME_INIT (frame, op_errno);
+        if (!local)
                 goto out;
-        }
 
         call_count = local->call_count;
-        frame->local = local;
+	if (!call_count) {
+		op_errno = ENOMEM;
+		goto out;
+	}
 
         for (i = 0; i < priv->child_count; i++) {
                 if (local->child_up[i]) {
                         STACK_WIND (frame, afr_inodelk_cbk,
                                     priv->children[i],
                                     priv->children[i]->fops->inodelk,
-                                    volume, loc, cmd, flock);
+                                    volume, loc, cmd, flock, xdata);
 
                         if (!--call_count)
                                 break;
                 }
         }
 
-        op_ret = 0;
+	return 0;
 out:
-        if (op_ret == -1) {
-                AFR_STACK_UNWIND (inodelk, frame, op_ret, op_errno);
-        }
+	AFR_STACK_UNWIND (inodelk, frame, -1, op_errno, NULL);
+
         return 0;
 }
 
 
 int32_t
-afr_finodelk_cbk (call_frame_t *frame, void *cookie,
-                  xlator_t *this, int32_t op_ret, int32_t op_errno)
+afr_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		  int32_t op_ret, int32_t op_errno, dict_t *xdata)
 
 {
         afr_local_t *local = NULL;
-
         int call_count = -1;
 
         local = frame->local;
@@ -1945,71 +2625,59 @@ afr_finodelk_cbk (call_frame_t *frame, void *cookie,
 
         if (call_count == 0)
                 AFR_STACK_UNWIND (finodelk, frame, local->op_ret,
-                                  local->op_errno);
+                                  local->op_errno, xdata);
 
         return 0;
 }
 
 
 int32_t
-afr_finodelk (call_frame_t *frame, xlator_t *this,
-              const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *flock)
+afr_finodelk (call_frame_t *frame, xlator_t *this, const char *volume, fd_t *fd,
+	      int32_t cmd, struct gf_flock *flock, dict_t *xdata)
 {
         afr_private_t *priv = NULL;
         afr_local_t *local  = NULL;
-
-        int ret = -1;
-
         int i = 0;
         int32_t call_count = 0;
-        int32_t op_ret   = -1;
-        int32_t op_errno = 0;
-
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
+        int32_t op_errno = ENOMEM;
 
         priv = this->private;
 
-        ALLOC_OR_GOTO (local, afr_local_t, out);
-
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
-                goto out;
-        }
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
         call_count = local->call_count;
-        frame->local = local;
+	if (!call_count) {
+		op_errno = ENOTCONN;
+		goto out;
+	}
 
         for (i = 0; i < priv->child_count; i++) {
                 if (local->child_up[i]) {
                         STACK_WIND (frame, afr_finodelk_cbk,
                                     priv->children[i],
                                     priv->children[i]->fops->finodelk,
-                                    volume, fd, cmd, flock);
+                                    volume, fd, cmd, flock, xdata);
 
                         if (!--call_count)
                                 break;
                 }
         }
 
-        op_ret = 0;
+	return 0;
 out:
-        if (op_ret == -1) {
-                AFR_STACK_UNWIND (finodelk, frame, op_ret, op_errno);
-        }
+	AFR_STACK_UNWIND (finodelk, frame, -1, op_errno, NULL);
+
         return 0;
 }
 
 
 int32_t
-afr_entrylk_cbk (call_frame_t *frame, void *cookie,
-                 xlator_t *this, int32_t op_ret, int32_t op_errno)
-
+afr_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
         afr_local_t *local = NULL;
-
         int call_count = -1;
 
         local = frame->local;
@@ -2027,73 +2695,62 @@ afr_entrylk_cbk (call_frame_t *frame, void *cookie,
 
         if (call_count == 0)
                 AFR_STACK_UNWIND (entrylk, frame, local->op_ret,
-                                  local->op_errno);
+                                  local->op_errno, xdata);
 
         return 0;
 }
 
 
-int32_t
-afr_entrylk (call_frame_t *frame, xlator_t *this,
-             const char *volume, loc_t *loc,
-             const char *basename, entrylk_cmd cmd, entrylk_type type)
+int
+afr_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)
 {
         afr_private_t *priv = NULL;
         afr_local_t *local  = NULL;
-
-        int ret = -1;
-
         int i = 0;
         int32_t call_count = 0;
-        int32_t op_ret   = -1;
         int32_t op_errno = 0;
 
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
-
         priv = this->private;
 
-        ALLOC_OR_GOTO (local, afr_local_t, out);
-
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
-                goto out;
-        }
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
         call_count = local->call_count;
-        frame->local = local;
+	if (!call_count) {
+		op_errno = ENOTCONN;
+		goto out;
+	}
 
         for (i = 0; i < priv->child_count; i++) {
                 if (local->child_up[i]) {
                         STACK_WIND (frame, afr_entrylk_cbk,
                                     priv->children[i],
                                     priv->children[i]->fops->entrylk,
-                                    volume, loc, basename, cmd, type);
+                                    volume, loc, basename, cmd, type, xdata);
 
                         if (!--call_count)
                                 break;
                 }
         }
 
-        op_ret = 0;
+	return 0;
 out:
-        if (op_ret == -1) {
-                AFR_STACK_UNWIND (entrylk, frame, op_ret, op_errno);
-        }
+	AFR_STACK_UNWIND (entrylk, frame, -1, op_errno, NULL);
+
         return 0;
 }
 
 
 
-int32_t
-afr_fentrylk_cbk (call_frame_t *frame, void *cookie,
-                 xlator_t *this, int32_t op_ret, int32_t op_errno)
+int
+afr_fentrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		  int32_t op_ret, int32_t op_errno, dict_t *xdata)
 
 {
         afr_local_t *local = NULL;
-
         int call_count = -1;
 
         local = frame->local;
@@ -2111,163 +2768,150 @@ afr_fentrylk_cbk (call_frame_t *frame, void *cookie,
 
         if (call_count == 0)
                 AFR_STACK_UNWIND (fentrylk, frame, local->op_ret,
-                                  local->op_errno);
+                                  local->op_errno, xdata);
 
         return 0;
 }
 
 
-int32_t
-afr_fentrylk (call_frame_t *frame, xlator_t *this,
-              const char *volume, fd_t *fd,
-              const char *basename, entrylk_cmd cmd, entrylk_type type)
+int
+afr_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)
 {
         afr_private_t *priv = NULL;
         afr_local_t *local  = NULL;
-
-        int ret = -1;
-
         int i = 0;
         int32_t call_count = 0;
-        int32_t op_ret   = -1;
-        int32_t op_errno = 0;
-
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
+        int32_t op_errno = ENOMEM;
 
         priv = this->private;
 
-        ALLOC_OR_GOTO (local, afr_local_t, out);
-
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
-                goto out;
-        }
+        local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
         call_count = local->call_count;
-        frame->local = local;
+	if (!call_count) {
+		op_errno = ENOTCONN;
+		goto out;
+	}
 
         for (i = 0; i < priv->child_count; i++) {
                 if (local->child_up[i]) {
                         STACK_WIND (frame, afr_fentrylk_cbk,
                                     priv->children[i],
                                     priv->children[i]->fops->fentrylk,
-                                    volume, fd, basename, cmd, type);
+                                    volume, fd, basename, cmd, type, xdata);
 
                         if (!--call_count)
                                 break;
                 }
         }
 
-        op_ret = 0;
+	return 0;
 out:
-        if (op_ret == -1) {
-                AFR_STACK_UNWIND (fentrylk, frame, op_ret, op_errno);
-        }
+	AFR_STACK_UNWIND (fentrylk, frame, -1, op_errno, NULL);
+
         return 0;
 }
 
-int32_t
-afr_statfs_cbk (call_frame_t *frame, void *cookie,
-                xlator_t *this, int32_t op_ret, int32_t op_errno,
-                struct statvfs *statvfs)
+
+int
+afr_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+		int op_errno, struct statvfs *statvfs, dict_t *xdata)
 {
         afr_local_t *local = NULL;
-
         int call_count = 0;
+	struct statvfs *buf = NULL;
 
         LOCK (&frame->lock);
         {
                 local = frame->local;
 
-                if (op_ret == 0) {
-                        local->op_ret   = op_ret;
-
-                        if (local->cont.statfs.buf_set) {
-                                if (statvfs->f_bavail < local->cont.statfs.buf.f_bavail)
-                                        local->cont.statfs.buf = *statvfs;
-                        } else {
-                                local->cont.statfs.buf = *statvfs;
-                                local->cont.statfs.buf_set = 1;
-                        }
-                }
-
-                if (op_ret == -1)
+                if (op_ret != 0) {
                         local->op_errno = op_errno;
-
+			goto unlock;
+		}
+
+		local->op_ret = op_ret;
+
+		buf = &local->cont.statfs.buf;
+		if (local->cont.statfs.buf_set) {
+			if (statvfs->f_bavail < buf->f_bavail) {
+				*buf = *statvfs;
+				if (xdata) {
+					if (local->xdata_rsp)
+						dict_unref (local->xdata_rsp);
+					local->xdata_rsp = dict_ref (xdata);
+				}
+			}
+		} else {
+			*buf = *statvfs;
+			local->cont.statfs.buf_set = 1;
+			if (xdata)
+				local->xdata_rsp = dict_ref (xdata);
+		}
         }
+unlock:
         UNLOCK (&frame->lock);
 
         call_count = afr_frame_return (frame);
 
         if (call_count == 0)
                 AFR_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno,
-                                  &local->cont.statfs.buf);
+                                  &local->cont.statfs.buf, local->xdata_rsp);
 
         return 0;
 }
 
 
-int32_t
-afr_statfs (call_frame_t *frame, xlator_t *this,
-            loc_t *loc)
+int
+afr_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
-        afr_private_t *  priv        = NULL;
-        int              child_count = 0;
         afr_local_t   *  local       = NULL;
+	afr_private_t   *priv        = NULL;
         int              i           = 0;
-
-        int ret = -1;
         int              call_count = 0;
-        int32_t          op_ret      = -1;
-        int32_t          op_errno    = 0;
+        int32_t          op_errno    = ENOMEM;
 
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
-        VALIDATE_OR_GOTO (loc, out);
-
-        priv = this->private;
-        child_count = priv->child_count;
+	priv = this->private;
 
-        ALLOC_OR_GOTO (local, afr_local_t, out);
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
-                goto out;
-        }
-
-        frame->local = local;
         call_count = local->call_count;
+	if (!call_count) {
+		op_errno = ENOTCONN;
+		goto out;
+	}
 
-        for (i = 0; i < child_count; i++) {
+        for (i = 0; i < priv->child_count; i++) {
                 if (local->child_up[i]) {
                         STACK_WIND (frame, afr_statfs_cbk,
                                     priv->children[i],
                                     priv->children[i]->fops->statfs,
-                                    loc);
+                                    loc, xdata);
                         if (!--call_count)
                                 break;
                 }
         }
 
-        op_ret = 0;
+	return 0;
 out:
-        if (op_ret == -1) {
-                AFR_STACK_UNWIND (statfs, frame, op_ret, op_errno, NULL);
-        }
+	AFR_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
+
         return 0;
 }
 
 
 int32_t
 afr_lk_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct gf_flock *lock)
+                   int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+                   dict_t *xdata)
 {
         afr_local_t * local = NULL;
-
         int call_count = -1;
 
         local = frame->local;
@@ -2275,7 +2919,7 @@ afr_lk_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         if (call_count == 0)
                 AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
-                                  lock);
+                                  lock, xdata);
 
         return 0;
 }
@@ -2286,8 +2930,7 @@ afr_lk_unlock (call_frame_t *frame, xlator_t *this)
 {
         afr_local_t   * local = NULL;
         afr_private_t * priv  = NULL;
-
-        int i;
+        int i = 0;
         int call_count = 0;
 
         local = frame->local;
@@ -2298,7 +2941,7 @@ afr_lk_unlock (call_frame_t *frame, xlator_t *this)
 
         if (call_count == 0) {
                 AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
-                                  &local->cont.lk.ret_flock);
+                                  &local->cont.lk.ret_flock, NULL);
                 return 0;
         }
 
@@ -2312,7 +2955,7 @@ afr_lk_unlock (call_frame_t *frame, xlator_t *this)
                                     priv->children[i],
                                     priv->children[i]->fops->lk,
                                     local->fd, F_SETLK,
-                                    &local->cont.lk.user_flock);
+                                    &local->cont.lk.user_flock, NULL);
 
                         if (!--call_count)
                                 break;
@@ -2325,13 +2968,13 @@ afr_lk_unlock (call_frame_t *frame, xlator_t *this)
 
 int32_t
 afr_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-            int32_t op_ret, int32_t op_errno, struct gf_flock *lock)
+            int32_t op_ret, int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int child_index = -1;
 /*        int            ret  = 0; */
 
-        int child_index = -1;
 
         local = frame->local;
         priv  = this->private;
@@ -2360,30 +3003,15 @@ afr_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                    priv->children[child_index],
                                    priv->children[child_index]->fops->lk,
                                    local->fd, local->cont.lk.cmd,
-                                   &local->cont.lk.user_flock);
+                                   &local->cont.lk.user_flock, xdata);
         } else if (local->op_ret == -1) {
                 /* all nodes have gone down */
 
                 AFR_STACK_UNWIND (lk, frame, -1, ENOTCONN,
-                                  &local->cont.lk.ret_flock);
+                                  &local->cont.lk.ret_flock, NULL);
         } else {
-                /* locking has succeeded on all nodes that are up */
-
-                /* temporarily
-                ret = afr_mark_locked_nodes (this, local->fd,
-                                             local->cont.lk.locked_nodes);
-                if (ret)
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "Could not save locked nodes info in fdctx");
-
-                ret = afr_save_locked_fd (this, local->fd);
-                if (ret)
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "Could not save locked fd");
-
-                */
-		AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
-                                  &local->cont.lk.ret_flock);
+                AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
+                                  &local->cont.lk.ret_flock, NULL);
         }
 
         return 0;
@@ -2392,34 +3020,24 @@ afr_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 int
 afr_lk (call_frame_t *frame, xlator_t *this,
-        fd_t *fd, int32_t cmd,
-        struct gf_flock *flock)
+        fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
 {
         afr_private_t *priv = NULL;
         afr_local_t *local = NULL;
-
         int i = 0;
-
-        int32_t op_ret   = -1;
-        int32_t op_errno = 0;
-
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
+        int32_t op_errno = ENOMEM;
 
         priv = this->private;
 
-        ALLOC_OR_GOTO (local, afr_local_t, out);
-        AFR_LOCAL_INIT (local, priv);
-
-        frame->local  = local;
+        local = AFR_FRAME_INIT (frame, op_errno);
+        if (!local)
+                goto out;
 
         local->cont.lk.locked_nodes = GF_CALLOC (priv->child_count,
-                                              sizeof (*local->cont.lk.locked_nodes),
-                                              gf_afr_mt_char);
+                                                 sizeof (*local->cont.lk.locked_nodes),
+                                                 gf_afr_mt_char);
 
         if (!local->cont.lk.locked_nodes) {
-                gf_log (this->name, GF_LOG_ERROR, "Out of memory");
                 op_errno = ENOMEM;
                 goto out;
         }
@@ -2432,13 +3050,18 @@ afr_lk (call_frame_t *frame, xlator_t *this,
         STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) 0,
                            priv->children[i],
                            priv->children[i]->fops->lk,
-                           fd, cmd, flock);
+                           fd, cmd, flock, xdata);
 
-        op_ret = 0;
+	return 0;
 out:
-        if (op_ret == -1) {
-                AFR_STACK_UNWIND (lk, frame, op_ret, op_errno, NULL);
-        }
+	AFR_STACK_UNWIND (lk, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+
+int
+afr_forget (xlator_t *this, inode_t *inode)
+{
         return 0;
 }
 
@@ -2457,41 +3080,22 @@ afr_priv_dump (xlator_t *this)
         GF_ASSERT (priv);
         snprintf(key_prefix, GF_DUMP_MAX_BUF_LEN, "%s.%s", this->type, this->name);
         gf_proc_dump_add_section(key_prefix);
-        gf_proc_dump_build_key(key, key_prefix, "child_count");
-        gf_proc_dump_write(key, "%u", priv->child_count);
-        gf_proc_dump_build_key(key, key_prefix, "read_child_rr");
-        gf_proc_dump_write(key, "%u", priv->read_child_rr);
+        gf_proc_dump_write("child_count", "%u", priv->child_count);
         for (i = 0; i < priv->child_count; i++) {
-                gf_proc_dump_build_key(key, key_prefix, "child_up[%d]", i);
+                sprintf (key, "child_up[%d]", i);
                 gf_proc_dump_write(key, "%d", priv->child_up[i]);
-                gf_proc_dump_build_key(key, key_prefix,
-                                        "pending_key[%d]", i);
+                sprintf (key, "pending_key[%d]", i);
                 gf_proc_dump_write(key, "%s", priv->pending_key[i]);
         }
-        gf_proc_dump_build_key(key, key_prefix, "data_self_heal");
-        gf_proc_dump_write(key, "%d", priv->data_self_heal);
-        gf_proc_dump_build_key(key, key_prefix, "metadata_self_heal");
-        gf_proc_dump_write(key, "%d", priv->metadata_self_heal);
-        gf_proc_dump_build_key(key, key_prefix, "entry_self_heal");
-        gf_proc_dump_write(key, "%d", priv->entry_self_heal);
-        gf_proc_dump_build_key(key, key_prefix, "data_change_log");
-        gf_proc_dump_write(key, "%d", priv->data_change_log);
-        gf_proc_dump_build_key(key, key_prefix, "metadata_change_log");
-        gf_proc_dump_write(key, "%d", priv->metadata_change_log);
-        gf_proc_dump_build_key(key, key_prefix, "entry_change_log");
-        gf_proc_dump_write(key, "%d", priv->entry_change_log);
-        gf_proc_dump_build_key(key, key_prefix, "read_child");
-        gf_proc_dump_write(key, "%d", priv->read_child);
-        gf_proc_dump_build_key(key, key_prefix, "favorite_child");
-        gf_proc_dump_write(key, "%u", priv->favorite_child);
-        gf_proc_dump_build_key(key, key_prefix, "data_lock_server_count");
-        gf_proc_dump_write(key, "%u", priv->data_lock_server_count);
-        gf_proc_dump_build_key(key, key_prefix, "metadata_lock_server_count");
-        gf_proc_dump_write(key, "%u", priv->metadata_lock_server_count);
-        gf_proc_dump_build_key(key, key_prefix, "entry_lock_server_count");
-        gf_proc_dump_write(key, "%u", priv->entry_lock_server_count);
-        gf_proc_dump_build_key(key, key_prefix, "wait_count");
-        gf_proc_dump_write(key, "%u", priv->wait_count);
+        gf_proc_dump_write("data_self_heal", "%s", priv->data_self_heal);
+        gf_proc_dump_write("metadata_self_heal", "%d", priv->metadata_self_heal);
+        gf_proc_dump_write("entry_self_heal", "%d", priv->entry_self_heal);
+        gf_proc_dump_write("data_change_log", "%d", priv->data_change_log);
+        gf_proc_dump_write("metadata_change_log", "%d", priv->metadata_change_log);
+        gf_proc_dump_write("entry-change_log", "%d", priv->entry_change_log);
+        gf_proc_dump_write("read_child", "%d", priv->read_child);
+        gf_proc_dump_write("favorite_child", "%d", priv->favorite_child);
+        gf_proc_dump_write("wait_count", "%u", priv->wait_count);
 
         return 0;
 }
@@ -2507,7 +3111,6 @@ static int
 find_child_index (xlator_t *this, xlator_t *child)
 {
         afr_private_t *priv = NULL;
-
         int i = -1;
 
         priv = this->private;
@@ -2522,90 +3125,501 @@ find_child_index (xlator_t *this, xlator_t *child)
 
 int32_t
 afr_notify (xlator_t *this, int32_t event,
-            void *data, ...)
+            void *data, void *data2)
 {
-        afr_private_t *     priv     = NULL;
-        unsigned char *     child_up = NULL;
-
-        int i           = -1;
-        int up_children = 0;
+        afr_private_t   *priv               = NULL;
+        int             i                   = -1;
+        int             up_children         = 0;
+        int             down_children       = 0;
+        int             propagate           = 0;
+        int             had_heard_from_all  = 0;
+        int             have_heard_from_all = 0;
+        int             idx                 = -1;
+        int             ret                 = -1;
+        int             call_psh            = 0;
+        int             up_child            = -1;
+        dict_t          *input              = NULL;
+        dict_t          *output             = NULL;
 
         priv = this->private;
 
         if (!priv)
                 return 0;
 
-        child_up = priv->child_up;
+        /*
+         * We need to reset this in case children come up in "staggered"
+         * fashion, so that we discover a late-arriving local subvolume.  Note
+         * that we could end up issuing N lookups to the first subvolume, and
+         * O(N^2) overall, but N is small for AFR so it shouldn't be an issue.
+         */
+        priv->did_discovery = _gf_false;
+
+        had_heard_from_all = 1;
+        for (i = 0; i < priv->child_count; i++) {
+                if (!priv->last_event[i]) {
+                        had_heard_from_all = 0;
+                }
+        }
+
+        /* parent xlators dont need to know about every child_up, child_down
+         * because of afr ha. If all subvolumes go down, child_down has
+         * to be triggered. In that state when 1 subvolume comes up child_up
+         * needs to be triggered. dht optimizes revalidate lookup by sending
+         * it only to one of its subvolumes. When child up/down happens
+         * for afr's subvolumes dht should be notified by child_modified. The
+         * subsequent revalidate lookup happens on all the dht's subvolumes
+         * which triggers afr self-heals if any.
+         */
+        idx = find_child_index (this, data);
+        if (idx < 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Received child_up "
+                        "from invalid subvolume");
+                goto out;
+        }
 
         switch (event) {
         case GF_EVENT_CHILD_UP:
-                i = find_child_index (this, data);
+                LOCK (&priv->lock);
+                {
+                        /*
+                         * This only really counts if the child was never up
+                         * (value = -1) or had been down (value = 0).  See
+                         * comment at GF_EVENT_CHILD_DOWN for a more detailed
+                         * explanation.
+                         */
+                        if (priv->child_up[idx] != 1) {
+                                priv->up_count++;
+				priv->event_generation++;
+                        }
+                        priv->child_up[idx] = 1;
+
+                        call_psh = 1;
+                        up_child = idx;
+                        for (i = 0; i < priv->child_count; i++)
+                                if (priv->child_up[i] == 1)
+                                        up_children++;
+                        if (up_children == 1) {
+                                gf_log (this->name, GF_LOG_INFO,
+                                        "Subvolume '%s' came back up; "
+                                        "going online.", ((xlator_t *)data)->name);
+                        } else {
+                                event = GF_EVENT_CHILD_MODIFIED;
+                        }
 
-                /* temporarily
-                afr_attempt_lock_recovery (this, i);
-                */
+                        priv->last_event[idx] = event;
+                }
+                UNLOCK (&priv->lock);
 
-                child_up[i] = 1;
+                break;
 
+        case GF_EVENT_CHILD_DOWN:
                 LOCK (&priv->lock);
                 {
-                        priv->up_count++;
+                        /*
+                         * If a brick is down when we start, we'll get a
+                         * CHILD_DOWN to indicate its initial state.  There
+                         * was never a CHILD_UP in this case, so if we
+                         * increment "down_count" the difference between than
+                         * and "up_count" will no longer be the number of
+                         * children that are currently up.  This has serious
+                         * implications e.g. for quorum enforcement, so we
+                         * don't increment these values unless the event
+                         * represents an actual state transition between "up"
+                         * (value = 1) and anything else.
+                         */
+                        if (priv->child_up[idx] == 1) {
+                                priv->down_count++;
+				priv->event_generation++;
+                        }
+                        priv->child_up[idx] = 0;
+
+                        for (i = 0; i < priv->child_count; i++)
+                                if (priv->child_up[i] == 0)
+                                        down_children++;
+                        if (down_children == priv->child_count) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "All subvolumes are down. Going offline "
+                                        "until atleast one of them comes back up.");
+                        } else {
+                                event = GF_EVENT_CHILD_MODIFIED;
+                        }
+
+                        priv->last_event[idx] = event;
                 }
                 UNLOCK (&priv->lock);
 
-                /*
-                   if all the children were down, and one child came up,
-                   send notify to parent
-                */
+                break;
 
-                for (i = 0; i < priv->child_count; i++)
-                        if (child_up[i])
-                                up_children++;
+        case GF_EVENT_CHILD_CONNECTING:
+                LOCK (&priv->lock);
+                {
+                        priv->last_event[idx] = event;
+                }
+                UNLOCK (&priv->lock);
 
-                if (up_children == 1) {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "Subvolume '%s' came back up; "
-                                "going online.", ((xlator_t *)data)->name);
+                break;
 
-                        default_notify (this, event, data);
+        case GF_EVENT_TRANSLATOR_OP:
+                input = data;
+                output = data2;
+                if (!had_heard_from_all) {
+                        ret = -1;
+                        goto out;
                 }
+                ret = afr_xl_op (this, input, output);
+                goto out;
+                break;
 
+        default:
+                propagate = 1;
                 break;
+        }
 
-        case GF_EVENT_CHILD_DOWN:
-                i = find_child_index (this, data);
+        /* have all subvolumes reported status once by now? */
+        have_heard_from_all = 1;
+        for (i = 0; i < priv->child_count; i++) {
+                if (!priv->last_event[i])
+                        have_heard_from_all = 0;
+        }
+
+        /* if all subvols have reported status, no need to hide anything
+           or wait for anything else. Just propagate blindly */
+        if (have_heard_from_all)
+                propagate = 1;
 
-                child_up[i] = 0;
+        if (!had_heard_from_all && have_heard_from_all) {
+                /* This is the first event which completes aggregation
+                   of events from all subvolumes. If at least one subvol
+                   had come up, propagate CHILD_UP, but only this time
+                */
+                event = GF_EVENT_CHILD_DOWN;
 
                 LOCK (&priv->lock);
                 {
-                        priv->down_count++;
+                        up_children = AFR_COUNT (priv->child_up, priv->child_count);
+                        for (i = 0; i < priv->child_count; i++) {
+                                if (priv->last_event[i] == GF_EVENT_CHILD_UP) {
+                                        event = GF_EVENT_CHILD_UP;
+                                        break;
+                                }
+
+                                if (priv->last_event[i] ==
+                                                GF_EVENT_CHILD_CONNECTING) {
+                                        event = GF_EVENT_CHILD_CONNECTING;
+                                        /* continue to check other events for CHILD_UP */
+                                }
+                        }
                 }
                 UNLOCK (&priv->lock);
+        }
 
-                /*
-                   if all children are down, and this was the last to go down,
-                   send notify to parent
-                */
+        ret = 0;
+        if (propagate)
+                ret = default_notify (this, event, data);
 
-                for (i = 0; i < priv->child_count; i++)
-                        if (child_up[i])
-                                up_children++;
+        if (call_psh && priv->shd.iamshd) {
+                afr_selfheal_childup (this, up_child);
+	}
+out:
+        return ret;
+}
+
+
+int
+afr_local_init (afr_local_t *local, afr_private_t *priv, int32_t *op_errno)
+{
+        local->op_ret = -1;
+        local->op_errno = EUCLEAN;
+
+	syncbarrier_init (&local->barrier);
+
+        local->child_up = GF_CALLOC (priv->child_count,
+                                     sizeof (*local->child_up),
+                                     gf_afr_mt_char);
+        if (!local->child_up) {
+                if (op_errno)
+                        *op_errno = ENOMEM;
+                goto out;
+        }
+
+        memcpy (local->child_up, priv->child_up,
+                sizeof (*local->child_up) * priv->child_count);
+        local->call_count = AFR_COUNT (local->child_up, priv->child_count);
+        if (local->call_count == 0) {
+                gf_log (THIS->name, GF_LOG_INFO, "no subvolumes up");
+                if (op_errno)
+                        *op_errno = ENOTCONN;
+                goto out;
+        }
+	local->event_generation = priv->event_generation;
+
+	local->read_attempted = GF_CALLOC (priv->child_count, sizeof (char),
+					   gf_afr_mt_char);
+	if (!local->read_attempted) {
+		if (op_errno)
+			*op_errno = ENOMEM;
+		goto out;
+	}
+
+	local->readable = GF_CALLOC (priv->child_count, sizeof (char),
+				     gf_afr_mt_char);
+	if (!local->readable) {
+		if (op_errno)
+			*op_errno = ENOMEM;
+		goto out;
+	}
+
+	local->replies = GF_CALLOC(priv->child_count, sizeof(*local->replies),
+				   gf_afr_mt_reply_t);
+	if (!local->replies) {
+		if (op_errno)
+			*op_errno = ENOMEM;
+		goto out;
+	}
+
+	return 0;
+out:
+        return -1;
+}
+
+int
+afr_internal_lock_init (afr_internal_lock_t *lk, size_t child_count,
+                        transaction_lk_type_t lk_type)
+{
+        int             ret = -ENOMEM;
+
+        lk->locked_nodes = GF_CALLOC (sizeof (*lk->locked_nodes),
+                                      child_count, gf_afr_mt_char);
+        if (NULL == lk->locked_nodes)
+                goto out;
+
+        lk->lower_locked_nodes = GF_CALLOC (sizeof (*lk->lower_locked_nodes),
+                                            child_count, gf_afr_mt_char);
+        if (NULL == lk->lower_locked_nodes)
+                goto out;
+
+        lk->lock_op_ret   = -1;
+        lk->lock_op_errno = EUCLEAN;
+        lk->transaction_lk_type = lk_type;
+
+        ret = 0;
+out:
+        return ret;
+}
+
+void
+afr_matrix_cleanup (int32_t **matrix, unsigned int m)
+{
+        int             i         = 0;
+
+        if (!matrix)
+                goto out;
+        for (i = 0; i < m; i++) {
+                GF_FREE (matrix[i]);
+        }
+
+        GF_FREE (matrix);
+out:
+        return;
+}
+
+int32_t**
+afr_matrix_create (unsigned int m, unsigned int n)
+{
+        int32_t         **matrix = NULL;
+        int             i       = 0;
+
+        matrix = GF_CALLOC (sizeof (*matrix), m, gf_afr_mt_int32_t);
+        if (!matrix)
+                goto out;
+
+        for (i = 0; i < m; i++) {
+                matrix[i] = GF_CALLOC (sizeof (*matrix[i]), n,
+                                       gf_afr_mt_int32_t);
+                if (!matrix[i])
+                        goto out;
+        }
+        return matrix;
+out:
+        afr_matrix_cleanup (matrix, m);
+        return NULL;
+}
+
+int
+afr_inodelk_init (afr_inodelk_t *lk, char *dom, size_t child_count)
+{
+        int             ret = -ENOMEM;
+
+        lk->domain = dom;
+        lk->locked_nodes = GF_CALLOC (sizeof (*lk->locked_nodes),
+                                      child_count, gf_afr_mt_char);
+        if (NULL == lk->locked_nodes)
+                goto out;
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+afr_transaction_local_init (afr_local_t *local, xlator_t *this)
+{
+        int            child_up_count = 0;
+        int            ret = -ENOMEM;
+        afr_private_t *priv = NULL;
+
+        priv = this->private;
+        ret = afr_internal_lock_init (&local->internal_lock, priv->child_count,
+                                      AFR_TRANSACTION_LK);
+        if (ret < 0)
+                goto out;
+
+        if ((local->transaction.type == AFR_DATA_TRANSACTION) ||
+            (local->transaction.type == AFR_METADATA_TRANSACTION)) {
+                ret = afr_inodelk_init (&local->internal_lock.inodelk[0],
+                                        this->name, priv->child_count);
+                if (ret < 0)
+                        goto out;
+        }
+
+        ret = -ENOMEM;
+        child_up_count = AFR_COUNT (local->child_up, priv->child_count);
+        if (priv->optimistic_change_log && child_up_count == priv->child_count)
+                local->optimistic_change_log = 1;
+
+	local->pre_op_compat = priv->pre_op_compat;
+
+        local->transaction.eager_lock =
+                GF_CALLOC (sizeof (*local->transaction.eager_lock),
+                           priv->child_count,
+                           gf_afr_mt_int32_t);
+
+        if (!local->transaction.eager_lock)
+                goto out;
 
-                if (up_children == 0) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "All subvolumes are down. Going offline "
-                                "until atleast one of them comes back up.");
+        local->transaction.pre_op = GF_CALLOC (sizeof (*local->transaction.pre_op),
+                                               priv->child_count,
+                                               gf_afr_mt_char);
+        if (!local->transaction.pre_op)
+                goto out;
 
-                        default_notify (this, event, data);
+        local->transaction.fop_subvols = GF_CALLOC (sizeof (*local->transaction.fop_subvols),
+						    priv->child_count,
+						    gf_afr_mt_char);
+        if (!local->transaction.fop_subvols)
+                goto out;
+
+        local->transaction.failed_subvols = GF_CALLOC (sizeof (*local->transaction.failed_subvols),
+						       priv->child_count,
+						       gf_afr_mt_char);
+        if (!local->transaction.failed_subvols)
+                goto out;
+
+        local->pending = afr_matrix_create (priv->child_count,
+                                            AFR_NUM_CHANGE_LOGS);
+        if (!local->pending)
+                goto out;
+
+	INIT_LIST_HEAD (&local->transaction.eager_locked);
+
+        ret = 0;
+out:
+        return ret;
+}
+
+
+void
+afr_set_low_priority (call_frame_t *frame)
+{
+        frame->root->pid = LOW_PRIO_PROC_PID;
+}
+
+
+gf_boolean_t
+afr_have_quorum (char *logname, afr_private_t *priv)
+{
+        unsigned int        quorum = 0;
+
+        GF_VALIDATE_OR_GOTO(logname,priv,out);
+
+        quorum = priv->quorum_count;
+        if (quorum != AFR_QUORUM_AUTO) {
+                return (priv->up_count >= (priv->down_count + quorum));
+        }
+
+        quorum = priv->child_count / 2 + 1;
+        if (priv->up_count >= (priv->down_count + quorum)) {
+                return _gf_true;
+        }
+
+        /*
+         * Special case for even numbers of nodes: if we have exactly half
+         * and that includes the first ("senior-most") node, then that counts
+         * as quorum even if it wouldn't otherwise.  This supports e.g. N=2
+         * while preserving the critical property that there can only be one
+         * such group.
+         */
+        if ((priv->child_count % 2) == 0) {
+                quorum = priv->child_count / 2;
+                if (priv->up_count >= (priv->down_count + quorum)) {
+                        if (priv->child_up[0]) {
+                                return _gf_true;
+                        }
                 }
+        }
 
-                break;
+out:
+        return _gf_false;
+}
 
-        default:
-                default_notify (this, event, data);
+void
+afr_priv_destroy (afr_private_t *priv)
+{
+        int            i           = 0;
+
+        if (!priv)
+                goto out;
+        inode_unref (priv->root_inode);
+        GF_FREE (priv->last_event);
+        if (priv->pending_key) {
+                for (i = 0; i < priv->child_count; i++)
+                        GF_FREE (priv->pending_key[i]);
         }
+        GF_FREE (priv->pending_key);
+        GF_FREE (priv->children);
+        GF_FREE (priv->child_up);
+        LOCK_DESTROY (&priv->lock);
 
-        return 0;
+        GF_FREE (priv);
+out:
+        return;
+}
+
+int
+xlator_subvolume_count (xlator_t *this)
+{
+        int i = 0;
+        xlator_list_t *list = NULL;
+
+        for (list = this->children; list; list = list->next)
+                i++;
+        return i;
+}
+
+
+void
+afr_handle_open_fd_count (call_frame_t *frame, xlator_t *this)
+{
+        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)
+		return;
 
+	fd_ctx->open_fd_count = local->open_fd_count;
 }
diff --git a/xlators/cluster/afr/src/afr-dir-read.c b/xlators/cluster/afr/src/afr-dir-read.c
index f52054eaa..fa1da3958 100644
--- a/xlators/cluster/afr/src/afr-dir-read.c
+++ b/xlators/cluster/afr/src/afr-dir-read.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -46,699 +37,384 @@
 #include "checksum.h"
 
 #include "afr.h"
-#include "afr-self-heal.h"
-#include "afr-self-heal-common.h"
-
-
-int
-afr_examine_dir_sh_unwind (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t *local  = NULL;
-
-        local = frame->local;
-
-        afr_set_opendir_done (this, local->fd->inode);
-
-        AFR_STACK_UNWIND (opendir, frame, local->op_ret,
-                          local->op_errno, local->fd);
-
-        return 0;
-}
-
-
-gf_boolean_t
-__checksums_differ (uint32_t *checksum, int child_count,
-                    unsigned char *child_up)
-{
-        int ret = _gf_false;
-        int i = 0;
-
-        uint32_t cksum;
-
-        cksum = checksum[0];
-
-        for (i = 0; i < child_count; i++) {
-                if (!child_up[i])
-                        continue;
-
-                if (cksum != checksum[i]) {
-                        ret = _gf_true;
-                        break;
-                }
-
-                cksum = checksum[i];
-        }
-
-        return ret;
-}
+#include "afr-transaction.h"
 
 
 int32_t
-afr_examine_dir_readdir_cbk (call_frame_t *frame, void *cookie,
-                             xlator_t *this, int32_t op_ret, int32_t op_errno,
-                             gf_dirent_t *entries)
+afr_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)
 {
-	afr_private_t *    priv  = NULL;
-	afr_local_t *      local = NULL;
-        afr_self_heal_t *  sh    = NULL;
+        afr_local_t   *local             = NULL;
+        int            call_count        = -1;
+        int32_t        child_index       = 0;
+	afr_fd_ctx_t  *fd_ctx = NULL;
 
-        gf_dirent_t * entry = NULL;
-        gf_dirent_t * tmp   = NULL;
-
-        int child_index = 0;
-
-        uint32_t entry_cksum;
-
-        int   call_count    = 0;
-        off_t last_offset = 0;
-        char  sh_type_str[256] = {0,};
-
-        priv  = this->private;
         local = frame->local;
-        sh    = &local->self_heal;
-
+	fd_ctx = local->fd_ctx;
         child_index = (long) cookie;
 
-        if (op_ret == -1) {
-                local->op_ret = -1;
-                local->op_ret = op_errno;
-                goto out;
-        }
-
-        if (op_ret == 0)
-                goto out;
-
-        list_for_each_entry_safe (entry, tmp, &entries->list, list) {
-                entry_cksum = gf_rsync_weak_checksum (entry->d_name,
-                                                      strlen (entry->d_name));
-                local->cont.opendir.checksum[child_index] ^= entry_cksum;
-        }
-
-        list_for_each_entry (entry, &entries->list, list) {
-                last_offset = entry->d_off;
-        }
-
-        /* read more entries */
-
-        STACK_WIND_COOKIE (frame, afr_examine_dir_readdir_cbk,
-                           (void *) (long) child_index,
-                           priv->children[child_index],
-                           priv->children[child_index]->fops->readdir,
-                           local->fd, 131072, last_offset);
-
-out:
-        if ((op_ret == 0) || (op_ret == -1)) {
-                call_count = afr_frame_return (frame);
-
-                if (call_count == 0) {
-                        if (__checksums_differ (local->cont.opendir.checksum,
-                                                priv->child_count,
-                                                local->child_up)) {
-
-                                sh->need_entry_self_heal  = _gf_true;
-                                sh->forced_merge          = _gf_true;
-                                sh->type                  = local->fd->inode->ia_type;
-                                sh->background            = _gf_false;
-                                sh->unwind                = afr_examine_dir_sh_unwind;
-
-                                afr_self_heal_type_str_get(&local->self_heal,
-                                                           sh_type_str,
-                                                           sizeof(sh_type_str));
-                                gf_log (this->name, GF_LOG_NORMAL,
-                                        "%s self-heal triggered. path: %s, "
-                                         "reason: checksums of directory differ,"
-                                        " forced merge option set",
-                                        sh_type_str, local->loc.path);
-
-                                afr_self_heal (frame, this);
-                        } else {
-                                afr_set_opendir_done (this, local->fd->inode);
-
-                                AFR_STACK_UNWIND (opendir, frame, local->op_ret,
-                                                  local->op_errno, local->fd);
-                        }
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+			fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
+                } else {
+                        local->op_ret = op_ret;
+			fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
+			if (!local->xdata_rsp && xdata)
+				local->xdata_rsp = dict_ref (xdata);
                 }
         }
+        UNLOCK (&frame->lock);
+
+        call_count = afr_frame_return (frame);
 
+        if (call_count == 0)
+		AFR_STACK_UNWIND (opendir, frame, local->op_ret,
+				  local->op_errno, local->fd, NULL);
         return 0;
 }
 
 
 int
-afr_examine_dir (call_frame_t *frame, xlator_t *this)
+afr_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
 {
-        afr_private_t * priv  = NULL;
-        afr_local_t *   local = NULL;
+        afr_private_t * priv        = NULL;
+        afr_local_t   * local       = NULL;
+        int             i           = 0;
+        int             call_count  = -1;
+        int32_t         op_errno    = ENOMEM;
+	afr_fd_ctx_t *fd_ctx = NULL;
 
-        int i;
-        int call_count = 0;
+        priv = this->private;
 
-        local = frame->local;
-        priv  = this->private;
+        local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
+
+	fd_ctx = afr_fd_ctx_get (fd, this);
+	if (!fd_ctx)
+		goto out;
 
-        local->cont.opendir.checksum = GF_CALLOC (priv->child_count,
-                                        sizeof (*local->cont.opendir.checksum),
-                                        gf_afr_mt_int32_t);
+        loc_copy (&local->loc, loc);
 
-        call_count = afr_up_children_count (priv->child_count, local->child_up);
+        local->fd    = fd_ref (fd);
+	local->fd_ctx = fd_ctx;
 
-        local->call_count = call_count;
+        call_count = local->call_count;
 
         for (i = 0; i < priv->child_count; i++) {
                 if (local->child_up[i]) {
-                        STACK_WIND_COOKIE (frame, afr_examine_dir_readdir_cbk,
-                                           (void *) (long) i,
+                        STACK_WIND_COOKIE (frame, afr_opendir_cbk,
+                                           (void*) (long) i,
                                            priv->children[i],
-                                           priv->children[i]->fops->readdir,
-                                           local->fd, 131072, 0);
+                                           priv->children[i]->fops->opendir,
+                                           loc, fd, NULL);
 
                         if (!--call_count)
                                 break;
                 }
         }
 
+	return 0;
+out:
+	AFR_STACK_UNWIND (opendir, frame, -1, op_errno, fd, NULL);
         return 0;
 }
 
 
-int32_t
-afr_opendir_cbk (call_frame_t *frame, void *cookie,
-		 xlator_t *this, int32_t op_ret, int32_t op_errno,
-		 fd_t *fd)
-{
-        afr_private_t *priv              = NULL;
-	afr_local_t   *local             = NULL;
-        int32_t        up_children_count = 0;
-        int           ret                = -1;
+#define BACKEND_D_OFF_BITS 63
+#define PRESENT_D_OFF_BITS 63
 
-	int call_count = -1;
+#define ONE 1ULL
+#define MASK (~0ULL)
+#define PRESENT_MASK (MASK >> (64 - PRESENT_D_OFF_BITS))
+#define BACKEND_MASK (MASK >> (64 - BACKEND_D_OFF_BITS))
 
-        priv  = this->private;
-        local = frame->local;
-
-        up_children_count = afr_up_children_count (priv->child_count,
-                                                   local->child_up);
+#define TOP_BIT (ONE << (PRESENT_D_OFF_BITS - 1))
+#define SHIFT_BITS (max (0, (BACKEND_D_OFF_BITS - PRESENT_D_OFF_BITS + 1)))
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret >= 0)
-			local->op_ret = op_ret;
+static uint64_t
+afr_bits_for (uint64_t num)
+{
+	uint64_t bits = 0, ctrl = 1;
 
-		local->op_errno = op_errno;
+	while (ctrl < num) {
+		ctrl *= 2;
+		bits ++;
 	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-                if (local->op_ret == 0) {
-
-                        ret = afr_fd_ctx_set (this, local->fd);
-
-                        if (ret) {
-                                local->op_ret = -1;
-                                local->op_errno = -1;
-                                gf_log (this->name, GF_LOG_ERROR, " failed to "
-                                        "set fd ctx for fd %d", local->fd);
-                                goto out;
-                        }
-                        if (!afr_is_opendir_done (this, local->fd->inode) &&
-                            up_children_count > 1) {
-
-                                /*
-                                 * This is the first opendir on this inode. We need
-                                 * to check if the directory's entries are the same
-                                 * on all subvolumes. This is needed in addition
-                                 * to regular entry self-heal because the readdir
-                                 * call is sent only to the first subvolume, and
-                                 * thus files that exist only there will never be healed
-                                 * otherwise (assuming changelog shows no anamolies).
-                                 */
-
-                                gf_log (this->name, GF_LOG_TRACE,
-                                        "reading contents of directory %s looking for mismatch",
-                                        local->loc.path);
-
-                                afr_examine_dir (frame, this);
-
-                        } else {
-                                AFR_STACK_UNWIND (opendir, frame, local->op_ret,
-                                                  local->op_errno, local->fd);
-                        }
-                } else {
-out:
-                        AFR_STACK_UNWIND (opendir, frame, local->op_ret,
-                                          local->op_errno, local->fd);
-                }
-        }
 
-	return 0;
+	return bits;
 }
 
-
-int32_t 
-afr_opendir (call_frame_t *frame, xlator_t *this,
-	     loc_t *loc, fd_t *fd)
+int
+afr_itransform (xlator_t *this, int subvol, uint64_t x, uint64_t *y_p)
 {
-	afr_private_t * priv        = NULL;
-	afr_local_t   * local       = NULL;
-
-	int             child_count = 0;
-	int             i           = 0;
-
-	int ret = -1;
-	int call_count = -1;
-
-	int32_t         op_ret   = -1;
-	int32_t         op_errno = 0;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        afr_private_t *conf = NULL;
+        int         cnt = 0;
+        int         max = 0;
+        uint64_t    y = 0;
+        uint64_t    hi_mask = 0;
+        uint64_t    off_mask = 0;
+        int         max_bits = 0;
+
+        if (x == ((uint64_t) -1)) {
+                y = (uint64_t) -1;
+                goto out;
+        }
 
-	priv = this->private;
+        conf = this->private;
+        if (!conf)
+                goto out;
 
-	child_count = priv->child_count;
+        max = conf->child_count;
+        cnt = subvol;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
+	if (max == 1) {
+		y = x;
 		goto out;
 	}
 
-        loc_copy (&local->loc, loc);
-
-	frame->local = local;
-	local->fd    = fd_ref (fd);
-
-	call_count = local->call_count;
-	
-	for (i = 0; i < child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_opendir_cbk, 
-				    priv->children[i],
-				    priv->children[i]->fops->opendir,
-				    loc, fd);
-
-			if (!--call_count)
-				break;
-		}
-	}
-
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (opendir, frame, op_ret, op_errno, fd);
-	}
-
-	return 0;
-}
-
-
-/**
- * Common algorithm for directory read calls:
- * 
- * - Try the fop on the first child that is up
- * - if we have failed due to ENOTCONN:
- *     try the next child
- *
- * Applicable to: readdir
- */
-
+        max_bits = afr_bits_for (max);
 
-struct entry_name {
-        char *name;
-        struct list_head list;
-};
+        hi_mask = ~(PRESENT_MASK >> (max_bits + 1));
 
-
-static gf_boolean_t
-remembered_name (const char *name, struct list_head *entries)
-{
-        struct entry_name *e;
-        gf_boolean_t ret = _gf_false;
-
-        list_for_each_entry (e, entries, list) {
-                if (!strcmp (name, e->name)) {
-                        ret = _gf_true;
-                        goto out;
-                }
+        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:
-        return ret;
-}
-
-
-static void
-afr_remember_entries (gf_dirent_t *entries, fd_t *fd)
-{
-	struct entry_name *n       = NULL;
-	gf_dirent_t *      entry   = NULL;
-
-	int ret = 0;
-
-	uint64_t      ctx;
-	afr_fd_ctx_t *fd_ctx;
-
-	ret = fd_ctx_get (fd, THIS, &ctx);
-	if (ret < 0) {
-		gf_log (THIS->name, GF_LOG_DEBUG,
-			"could not get fd ctx for fd=%p", fd);
-		return;
-	}
-
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+        if (y_p)
+                *y_p = y;
 
-	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);
-	}
+        return 0;
 }
 
 
-static off_t
-afr_filter_entries (gf_dirent_t *entries, fd_t *fd)
+int
+afr_deitransform (xlator_t *this, uint64_t y, int *subvol_p,
+                  uint64_t *x_p)
 {
-	gf_dirent_t *entry, *tmp;
-	int ret = 0;
-
-	uint64_t      ctx;
-	afr_fd_ctx_t *fd_ctx;
-
-	off_t offset = 0;
-
-	ret = fd_ctx_get (fd, THIS, &ctx);
-	if (ret < 0) {
-		gf_log (THIS->name, GF_LOG_DEBUG,
-			"could not get fd ctx for fd=%p", fd);
-		return -1;
-	}
-
-        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);
-		}
+        afr_private_t *conf = NULL;
+        int         cnt = 0;
+        int         max = 0;
+        uint64_t    x = 0;
+        int         subvol = 0;
+        int         max_bits = 0;
+        uint64_t    off_mask = 0;
+        uint64_t    host_mask = 0;
+
+        if (!this->private)
+                return -1;
+
+        conf = this->private;
+        max = conf->child_count;
+
+	if (max == 1) {
+		x = y;
+		cnt = 0;
+		goto out;
 	}
 
-	return offset;
-}
+        if (y & TOP_BIT) {
+                /* HUGE d_off */
+                max_bits = afr_bits_for (max);
+                off_mask = (MASK << max_bits);
+                host_mask = ~(off_mask);
 
+                x = ((y & ~TOP_BIT) & off_mask) << SHIFT_BITS;
 
-static void
-afr_forget_entries (fd_t *fd)
-{
-	struct entry_name *entry, *tmp;
-	int ret = 0;
+                cnt = y & host_mask;
+	} else {
+                /* small d_off */
+                cnt = y % max;
+                x = y / max;
+        }
 
-	uint64_t      ctx;
-	afr_fd_ctx_t *fd_ctx;
+out:
+        subvol = cnt;
 
-	ret = fd_ctx_get (fd, THIS, &ctx);
-	if (ret < 0) {
-		gf_log (THIS->name, GF_LOG_DEBUG,
-			"could not get fd ctx for fd=%p", fd);
-		return;
-	}
+        if (subvol_p)
+                *subvol_p = subvol;
 
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+        if (x_p)
+                *x_p = x;
 
-	list_for_each_entry_safe (entry, tmp, &fd_ctx->entries, list) {
-		GF_FREE (entry->name);
-		list_del (&entry->list);
-		GF_FREE (entry);
-	}
+        return 0;
 }
 
 
-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)
+static void
+afr_readdir_transform_entries (gf_dirent_t *subvol_entries, int subvol,
+			       gf_dirent_t *entries, fd_t *fd)
 {
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
+	afr_private_t *priv = NULL;
+        gf_dirent_t *entry = NULL;
+        gf_dirent_t *tmp = NULL;
+	unsigned char *data_readable = NULL;
+	unsigned char *metadata_readable = NULL;
+	int gen = 0;
+
+	priv = THIS->private;
+
+	data_readable = alloca0 (priv->child_count);
+	metadata_readable = alloca0 (priv->child_count);
+
+        list_for_each_entry_safe (entry, tmp, &subvol_entries->list, list) {
+                if (__is_root_gfid (fd->inode->gfid) &&
+                    !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR)) {
+			continue;
+                }
 
-        gf_dirent_t * entry = NULL;
-        gf_dirent_t * tmp   = NULL;
+		list_del_init (&entry->list);
+		afr_itransform (THIS, subvol, entry->d_off, &entry->d_off);
+		list_add_tail (&entry->list, &entries->list);
 
-        int child_index = -1;
+		if (entry->inode) {
+			gen = 0;
+			afr_inode_read_subvol_get (entry->inode, THIS,
+						   data_readable,
+						   metadata_readable, &gen);
 
-	priv     = this->private;
-	local = frame->local;
-        child_index = (long) cookie;
+			if (gen != priv->event_generation ||
+				!data_readable[subvol] ||
+				!metadata_readable[subvol]) {
 
-	if (op_ret != -1) {
-                list_for_each_entry_safe (entry, tmp, &entries->list, list) {
-                        entry->d_ino = afr_itransform (entry->d_ino,
-                                                       priv->child_count,
-                                                       child_index);
-
-                        if ((local->fd->inode == local->fd->inode->table->root)
-                            && !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR)) {
-                                list_del_init (&entry->list);
-                                GF_FREE (entry);
-                        }
-                }
-    	}
-
-        AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, entries);
-
-	return 0;
+				inode_unref (entry->inode);
+				entry->inode = NULL;
+			}
+		}
+        }
 }
 
 
 int32_t
-afr_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                  int32_t op_ret, int32_t op_errno, gf_dirent_t *entries)
+afr_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		 int32_t op_ret, int32_t op_errno, gf_dirent_t *subvol_entries,
+		 dict_t *xdata)
 {
-        afr_private_t * priv     = NULL;
-        afr_local_t *   local    = NULL;
-        xlator_t **     children = NULL;
-        ino_t           inum = 0;
-
-        int call_child = 0;
-        int ret        = 0;
-
-        gf_dirent_t * entry = NULL;
-        gf_dirent_t * tmp   = NULL;
-
-        int child_index = -1;
+        afr_local_t *local = NULL;
+	gf_dirent_t  entries;
 
-        uint64_t      ctx    = 0;
-        afr_fd_ctx_t *fd_ctx = NULL;
-
-	off_t offset = 0;
-
-        priv     = this->private;
-        children = priv->children;
+	INIT_LIST_HEAD (&entries.list);
 
         local = frame->local;
 
-        child_index = (long) cookie;
-
-	if (priv->strict_readdir) {
-		ret = fd_ctx_get (local->fd, this, &ctx);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"could not get fd ctx for fd=%p", local->fd);
-			op_ret   = -1;
-			op_errno = -ret;
-			goto out;
-		}
-
-		fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-		if (child_went_down (op_ret, op_errno)) {
-			if (all_tried (child_index, priv->child_count)) {
-				goto out;
-			}
-
-			call_child = ++child_index;
-
-			gf_log (this->name, GF_LOG_TRACE,
-				"starting readdir afresh on child %d, offset %"PRId64,
-				call_child, (uint64_t) 0);
-
-			fd_ctx->failed_over = _gf_true;
-
-			STACK_WIND_COOKIE (frame, afr_readdirp_cbk,
-					   (void *) (long) call_child,
-					   children[call_child],
-					   children[call_child]->fops->readdirp, local->fd,
-					   local->cont.readdir.size, 0);
-			return 0;
-		}
-	}
+        if (op_ret < 0 && !local->cont.readdir.offset) {
+		/* failover only if this was first readdir, detected
+		   by offset == 0 */
+		local->op_ret = op_ret;
+		local->op_errno = op_errno;
 
-	if (op_ret != -1) {
-		list_for_each_entry_safe (entry, tmp, &entries->list, list) {
-			inum = afr_itransform (entry->d_ino, priv->child_count,
-					       child_index);
-			entry->d_ino = inum;
-			inum  = afr_itransform (entry->d_stat.ia_ino,
-						priv->child_count, child_index);
-			entry->d_stat.ia_ino = inum;
-
-			if ((local->fd->inode == local->fd->inode->table->root)
-			    && !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR)) {
-				list_del_init (&entry->list);
-				GF_FREE (entry);
-			}
-		}
+		afr_read_txn_continue (frame, this, (long) cookie);
+		return 0;
 	}
 
-	if (priv->strict_readdir) {
-		if (fd_ctx->failed_over) {
-			if (list_empty (&entries->list)) {
-				goto out;
-			}
-
-			offset = afr_filter_entries (entries, local->fd);
+	if (op_ret >= 0)
+		afr_readdir_transform_entries (subvol_entries, (long) cookie,
+					       &entries, local->fd);
 
-			afr_remember_entries (entries, local->fd);
-
-			if (list_empty (&entries->list)) {
-				/* All the entries we got were duplicate. We
-				   shouldn't send an empty list now, because
-				   that'll make the application stop reading. So
-				   try to get more entries */
-
-				gf_log (this->name, GF_LOG_TRACE,
-					"trying to fetch non-duplicate entries from offset %"PRId64", child %s",
-					offset, children[child_index]->name);
-
-				STACK_WIND_COOKIE (frame, afr_readdirp_cbk,
-						   (void *) (long) child_index,
-						   children[child_index],
-						   children[child_index]->fops->readdirp,
-						   local->fd, local->cont.readdir.size, offset);
-				return 0;
-			}
-		} else {
-			afr_remember_entries (entries, local->fd);
-		}
-	}
-
-out:
-        AFR_STACK_UNWIND (readdirp, frame, op_ret, op_errno, entries);
+        AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, &entries, xdata);
 
         return 0;
 }
 
 
-int32_t
-afr_do_readdir (call_frame_t *frame, xlator_t *this,
-	        fd_t *fd, size_t size, off_t offset, int whichop)
+int
+afr_readdir_wind (call_frame_t *frame, xlator_t *this, int subvol)
 {
-	afr_private_t * priv       = NULL;
-	xlator_t **     children   = NULL;
-	int             call_child = 0;
-	afr_local_t     *local     = NULL;
-
-        uint64_t      ctx;
-        afr_fd_ctx_t *fd_ctx;
-
-	int ret = -1;
-
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-
-	priv     = this->private;
-	children = priv->children;
-
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
 
-	frame->local = local;
+	priv = this->private;
+	local = frame->local;
 
-	call_child = afr_first_up_child (priv);
-	if (call_child == -1) {
-		op_errno = ENOTCONN;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no child is up");
-		goto out;
+	if (subvol == -1) {
+		AFR_STACK_UNWIND (readdir, frame, local->op_ret,
+				  local->op_errno, 0, 0);
+		return 0;
 	}
 
-        local->fd                  = fd_ref (fd);
-        local->cont.readdir.size   = size;
-
-	if (priv->strict_readdir) {
-		ret = fd_ctx_get (fd, this, &ctx);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"could not get fd ctx for fd=%p", fd);
-			op_errno = -ret;
-			goto out;
-		}
+        if (local->op == GF_FOP_READDIR)
+                STACK_WIND_COOKIE (frame, afr_readdir_cbk,
+                                   (void *) (long) subvol,
+                                   priv->children[subvol],
+                                   priv->children[subvol]->fops->readdir,
+				   local->fd, local->cont.readdir.size,
+                                   local->cont.readdir.offset,
+				   local->xdata_req);
+        else
+                STACK_WIND_COOKIE (frame, afr_readdir_cbk,
+                                   (void *) (long) subvol,
+                                   priv->children[subvol],
+                                   priv->children[subvol]->fops->readdirp,
+				   local->fd, local->cont.readdir.size,
+				   local->cont.readdir.offset,
+				   local->xdata_req);
+	return 0;
+}
 
-		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);
+int
+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_local_t   *local     = NULL;
+        int32_t       op_errno   = 0;
+	int           subvol = -1;
 
-			fd_ctx->failed_over = _gf_true;
-			offset = 0;
-		}
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-		fd_ctx->last_tried = call_child;
+	local->op = whichop;
+        local->fd = fd_ref (fd);
+        local->cont.readdir.size = size;
+	local->cont.readdir.offset = offset;
+        local->xdata_req = (dict)? dict_ref (dict) : NULL;
+
+	if (offset == 0) {
+		/* First readdir has option of failing over and selecting
+		   an appropriate read subvolume */
+		afr_read_txn (frame, this, fd->inode, afr_readdir_wind,
+			      AFR_DATA_TRANSACTION);
+	} else {
+		/* But continued readdirs MUST stick to the same subvolume
+		   without an option to failover */
+		afr_deitransform (this, offset, &subvol,
+				  (uint64_t *)&local->cont.readdir.offset);
+		afr_readdir_wind (frame, this, subvol);
 	}
 
-        if (whichop == GF_FOP_READDIR)
-                STACK_WIND_COOKIE (frame, afr_readdir_cbk,
-                                   (void *) (long) call_child,
-                                   children[call_child],
-                                   children[call_child]->fops->readdir, fd,
-                                   size, offset);
-        else
-                STACK_WIND_COOKIE (frame, afr_readdirp_cbk,
-                                   (void *) (long) call_child,
-                                   children[call_child],
-                                   children[call_child]->fops->readdirp, fd,
-                                   size, offset);
-
-	op_ret = 0;
+        return 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, NULL);
-	}
-	return 0;
+        AFR_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
+        return 0;
 }
 
 
 int32_t
 afr_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-             off_t offset)
+             off_t offset, dict_t *xdata)
 {
-        afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIR);
+	afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIR, xdata);
+
         return 0;
 }
 
 
 int32_t
 afr_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-              off_t offset)
+              off_t offset, dict_t *dict)
 {
-        afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIRP);
+        afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIRP, dict);
+
         return 0;
 }
 
@@ -746,8 +422,7 @@ afr_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
 int32_t
 afr_releasedir (xlator_t *this, fd_t *fd)
 {
-	afr_forget_entries (fd);
         afr_cleanup_fd_ctx (this, fd);
 
-	return 0;
+        return 0;
 }
diff --git a/xlators/cluster/afr/src/afr-dir-read.h b/xlators/cluster/afr/src/afr-dir-read.h
index 40c7b6aef..09456d159 100644
--- a/xlators/cluster/afr/src/afr-dir-read.h
+++ b/xlators/cluster/afr/src/afr-dir-read.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __DIR_READ_H__
@@ -23,28 +14,23 @@
 
 int32_t
 afr_opendir (call_frame_t *frame, xlator_t *this,
-	     loc_t *loc, fd_t *fd);
+	     loc_t *loc, fd_t *fd, dict_t *xdata);
 
 int32_t
 afr_releasedir (xlator_t *this, fd_t *fd);
 
 int32_t
 afr_readdir (call_frame_t *frame, xlator_t *this,
-	     fd_t *fd, size_t size, off_t offset);
+	     fd_t *fd, size_t size, off_t offset, dict_t *xdata);
 
 
 int32_t
 afr_readdirp (call_frame_t *frame, xlator_t *this,
-	     fd_t *fd, size_t size, off_t offset);
-
-int32_t
-afr_getdents (call_frame_t *frame, xlator_t *this,
-	      fd_t *fd, size_t size, off_t offset, int32_t flag);
-
+              fd_t *fd, size_t size, off_t offset, dict_t *dict);
 
 int32_t
 afr_checksum (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, int32_t flags);
+	      loc_t *loc, int32_t flags, dict_t *xdata);
 
 
 #endif /* __DIR_READ_H__ */
diff --git a/xlators/cluster/afr/src/afr-dir-write.c b/xlators/cluster/afr/src/afr-dir-write.c
index af42e7e06..465dde54f 100644
--- a/xlators/cluster/afr/src/afr-dir-write.c
+++ b/xlators/cluster/afr/src/afr-dir-write.c
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -43,820 +34,753 @@
 #include "common-utils.h"
 #include "compat-errno.h"
 #include "compat.h"
+#include "byte-order.h"
 
 #include "afr.h"
 #include "afr-transaction.h"
 
-
 void
-afr_build_parent_loc (loc_t *parent, loc_t *child)
-{
-	char *tmp = NULL;
-
-	if (!child->parent) {
-		loc_copy (parent, child);
-		return;
-	}
-
-	tmp = gf_strdup (child->path);
-	parent->path   = gf_strdup (dirname (tmp));
-	GF_FREE (tmp);
-
-        parent->name   = strrchr (parent->path, '/');
-	if (parent->name)
-		parent->name++;
-
-	parent->inode  = inode_ref (child->parent);
-	parent->parent = inode_parent (parent->inode, 0, NULL);
-	parent->ino    = parent->inode->ino;
-}
-
-/* {{{ create */
+afr_mark_entry_pending_changelog (call_frame_t *frame, xlator_t *this);
 
 int
-afr_create_unwind (call_frame_t *frame, xlator_t *this)
+afr_build_parent_loc (loc_t *parent, loc_t *child, int32_t *op_errno)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
-        struct iatt  *unwind_buf = NULL;
+        int     ret = -1;
+        char    *child_path = NULL;
 
-	local = frame->local;
+        if (!child->parent) {
+                if (op_errno)
+                        *op_errno = EINVAL;
+                goto out;
+        }
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
+        child_path = gf_strdup (child->path);
+        if (!child_path) {
+                if (op_errno)
+                        *op_errno = ENOMEM;
+                goto out;
+        }
 
-	if (main_frame) {
-                if (local->cont.create.read_child_buf.ia_ino) {
-                        unwind_buf = &local->cont.create.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.create.buf;
-                }
+        parent->path = gf_strdup (dirname (child_path));
+        if (!parent->path) {
+                if (op_errno)
+                        *op_errno = ENOMEM;
+                goto out;
+        }
 
-                unwind_buf->ia_ino = local->cont.create.ino;
+        parent->inode = inode_ref (child->parent);
+	uuid_copy (parent->gfid, child->pargfid);
 
-                local->cont.create.preparent.ia_ino  = local->cont.create.parent_ino;
-                local->cont.create.postparent.ia_ino = local->cont.create.parent_ino;
+        ret = 0;
+out:
+        GF_FREE (child_path);
 
-		AFR_STACK_UNWIND (create, main_frame,
-                                  local->op_ret, local->op_errno,
-				  local->cont.create.fd,
-				  local->cont.create.inode,
-				  unwind_buf, &local->cont.create.preparent,
-                                  &local->cont.create.postparent);
-        }
-        
-	return 0;
+        return ret;
 }
 
 
-int
-afr_create_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		     int32_t op_ret, int32_t op_errno, 
-		     fd_t *fd, inode_t *inode, struct iatt *buf,
-                     struct iatt *preparent, struct iatt *postparent)
+static void
+__afr_dir_write_finalize (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-        uint64_t      ctx;
-        afr_fd_ctx_t *fd_ctx;
-
-        int ret = 0;
-
-	int call_count = -1;
-	int child_index = -1;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	int inode_read_subvol = -1;
+	int parent_read_subvol = -1;
+	int parent2_read_subvol = -1;
+	int i = 0;
 
 	local = frame->local;
-	priv  = this->private;
-
-	child_index = (long) cookie;
+	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) {
-			local->op_ret = op_ret;
-
-                        ret = afr_fd_ctx_set (this, fd);
-
-                        if (ret < 0) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "could not set ctx on fd=%p", fd);
-
-                                local->op_ret   = -1;
-                                local->op_errno = -ret;
-                                goto unlock;
-                        }
-
-                        ret = fd_ctx_get (fd, this, &ctx);
-
-                        if (ret < 0) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "could not get fd ctx for fd=%p", fd);
-                                local->op_ret   = -1;
-                                local->op_errno = -ret;
-                                goto unlock;
-                        }
-
-                        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                        fd_ctx->opened_on[child_index] = 1;
-                        fd_ctx->flags                  = local->cont.create.flags;
-
-                        if (local->success_count == 0) {
-				local->cont.create.buf        = *buf;
-
-                                local->cont.create.ino =
-                                        afr_itransform (buf->ia_ino,
-                                                        priv->child_count,
-                                                        child_index);
-
-                                if (priv->read_child >= 0) {
-                                        afr_set_read_child (this, inode, 
-                                                            priv->read_child);
-                                } else {
-                                        afr_set_read_child (this, inode, 
-                                                            local->read_child_index);
-                                }
-			}
-                        
-                        if (child_index == local->first_up_child) {
-                                local->cont.create.ino =
-                                        afr_itransform (buf->ia_ino,
-                                                        priv->child_count,
-                                                        local->first_up_child);
-                        }
-                        
-                        if (child_index == local->read_child_index) {
-                                local->cont.create.read_child_buf = *buf;
-                                local->cont.create.preparent      = *preparent;
-                                local->cont.create.postparent     = *postparent;
-                        }
-
-			local->cont.create.inode = inode;
-
-			local->success_count++;
+	if (local->inode) {
+		afr_replies_interpret (frame, this, local->inode);
+		inode_read_subvol = afr_data_subvol_get (local->inode, this,
+							 NULL, NULL);
+	}
+	if (local->parent)
+		parent_read_subvol = afr_data_subvol_get (local->parent, this,
+							  NULL, NULL);
+	if (local->parent2)
+		parent2_read_subvol = afr_data_subvol_get (local->parent2, this,
+							   NULL, NULL);
+
+	local->op_ret = -1;
+	local->op_errno = afr_final_errno (local, priv);
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (!local->replies[i].valid)
+			continue;
+		if (local->replies[i].op_ret < 0) {
+			if (local->inode)
+				afr_inode_read_subvol_reset (local->inode,
+							     this);
+			if (local->parent)
+				afr_inode_read_subvol_reset (local->parent,
+							     this);
+			if (local->parent2)
+				afr_inode_read_subvol_reset (local->parent2,
+							     this);
+			continue;
 		}
 
-		local->op_errno = op_errno;
-	}
-
-unlock:
-	UNLOCK (&frame->lock);
+		if (local->op_ret == -1) {
+			local->op_ret = local->replies[i].op_ret;
+			local->op_errno = local->replies[i].op_errno;
+
+			local->cont.dir_fop.buf =
+				local->replies[i].poststat;
+			local->cont.dir_fop.preparent =
+				local->replies[i].preparent;
+			local->cont.dir_fop.postparent =
+				local->replies[i].postparent;
+			local->cont.dir_fop.prenewparent =
+				local->replies[i].preparent2;
+			local->cont.dir_fop.postnewparent =
+				local->replies[i].postparent2;
+			if (local->replies[i].xdata)
+				local->xdata_rsp =
+					dict_ref (local->replies[i].xdata);
+			continue;
+		}
 
-	call_count = afr_frame_return (frame);
+		if (i == inode_read_subvol) {
+			local->cont.dir_fop.buf =
+				local->replies[i].poststat;
+			if (local->replies[i].xdata) {
+				if (local->xdata_rsp)
+					dict_unref (local->xdata_rsp);
+				local->xdata_rsp =
+					dict_ref (local->replies[i].xdata);
+			}
+		}
 
-	if (call_count == 0) {
-		local->transaction.unwind (frame, this);
+		if (i == parent_read_subvol) {
+			local->cont.dir_fop.preparent =
+				local->replies[i].preparent;
+			local->cont.dir_fop.postparent =
+				local->replies[i].postparent;
+		}
 
-		local->transaction.resume (frame, this);
+		if (i == parent2_read_subvol) {
+			local->cont.dir_fop.prenewparent =
+				local->replies[i].preparent2;
+			local->cont.dir_fop.postnewparent =
+				local->replies[i].postparent2;
+		}
 	}
-	
-	return 0;
 }
 
 
-int
-afr_create_wind (call_frame_t *frame, xlator_t *this)
+static void
+__afr_dir_write_fill (call_frame_t *frame, xlator_t *this, int child_index,
+		      int op_ret, int op_errno, struct iatt *poststat,
+		      struct iatt *preparent, struct iatt *postparent,
+		      struct iatt *preparent2, struct iatt *postparent2,
+		      dict_t *xdata)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+	afr_fd_ctx_t *fd_ctx = NULL;
+
+        local = frame->local;
+	fd_ctx = local->fd_ctx;
+
+	local->replies[child_index].valid = 1;
+	local->replies[child_index].op_ret = op_ret;
+	local->replies[child_index].op_errno = op_errno;
+
+	if (op_ret >= 0) {
+		if (poststat)
+			local->replies[child_index].poststat = *poststat;
+		if (preparent)
+			local->replies[child_index].preparent = *preparent;
+		if (postparent)
+			local->replies[child_index].postparent = *postparent;
+		if (preparent2)
+			local->replies[child_index].preparent2 = *preparent2;
+		if (postparent2)
+			local->replies[child_index].postparent2 = *postparent2;
+		if (xdata)
+			local->replies[child_index].xdata = dict_ref (xdata);
+
+		if (fd_ctx)
+			fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
+	} else {
+		if (op_errno != ENOTEMPTY)
+			afr_transaction_fop_failed (frame, this, child_index);
+		if (fd_ctx)
+			fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
+	}
 
-	int call_count = -1;
-	int i = 0;
+        return;
+}
 
-	local = frame->local;
-	priv = this->private;
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+static int
+__afr_dir_write_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		     int op_ret, int op_errno, struct iatt *buf,
+		     struct iatt *preparent, struct iatt *postparent,
+		     struct iatt *preparent2, struct iatt *postparent2,
+                     dict_t *xdata)
+{
+        afr_local_t *local = NULL;
+        int child_index = (long) cookie;
+        int call_count = -1;
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
+        local = frame->local;
 
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_create_wind_cbk,
-					   (void *) (long) i,
-					   priv->children[i], 
-					   priv->children[i]->fops->create,
-					   &local->loc, 
-					   local->cont.create.flags, 
-					   local->cont.create.mode, 
-					   local->cont.create.fd,
-                                           local->cont.create.params);
-			if (!--call_count)
-				break;
-		}
+	LOCK (&frame->lock);
+	{
+		__afr_dir_write_fill (frame, this, child_index, op_ret,
+				      op_errno, buf, preparent, postparent,
+				      preparent2, postparent2, xdata);
 	}
-	
-	return 0;
-}
-
+	UNLOCK (&frame->lock);
+        call_count = afr_frame_return (frame);
 
-int
-afr_create_done (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t * local = NULL;
+        if (call_count == 0) {
+		__afr_dir_write_finalize (frame, this);
 
-	local = frame->local;
+		if (afr_txn_nothing_failed (frame, this))
+			local->transaction.unwind (frame, this);
 
-	local->transaction.unwind (frame, this);
+		afr_mark_entry_pending_changelog (frame, this);
 
-	AFR_STACK_DESTROY (frame);
+                local->transaction.resume (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
-afr_create (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, int32_t flags, mode_t mode,
-            fd_t *fd, dict_t *params)
+afr_mark_new_entry_changelog_cbk (call_frame_t *frame, void *cookie,
+				  xlator_t *this, int op_ret, int op_errno,
+                                  dict_t *xattr, dict_t *xdata)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t  * transaction_frame = NULL;
+        int call_count = 0;
 
-	int ret = -1;
+        call_count = afr_frame_return (frame);
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        if (call_count == 0)
+                AFR_STACK_DESTROY (frame);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        return 0;
+}
 
-	priv = this->private;
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
+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;
+	int           idx         = 0;
+        int           op_errno    = ENOMEM;
+	unsigned char *pending    = NULL;
+	int           call_count   = 0;
+
+        local = frame->local;
+        priv = this->private;
+
+        new_frame = copy_frame (frame);
+        if (!new_frame)
+                goto out;
+
+	new_local = AFR_FRAME_INIT (new_frame, op_errno);
+	if (!new_local)
 		goto out;
-	}
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        changelog = afr_matrix_create (priv->child_count, AFR_NUM_CHANGE_LOGS);
+        if (!changelog)
+                goto out;
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
+        xattr = dict_new ();
+	if (!xattr)
 		goto out;
-	}
-
-	transaction_frame->local = local;
-
-	loc_copy (&local->loc, loc);
-
-        LOCK (&priv->read_child_lock);
-        {
-                local->read_child_index = (++priv->read_child_rr) 
-                        % (priv->child_count);
-        }
-        UNLOCK (&priv->read_child_lock);
-
-	local->cont.create.flags = flags;
-	local->cont.create.mode  = mode;
-	local->cont.create.fd    = fd_ref (fd);
-        if (params)
-                local->cont.create.params = dict_ref (params);
-
-        if (loc->parent)
-                local->cont.create.parent_ino = loc->parent->ino;
 
-	local->transaction.fop    = afr_create_wind;
-	local->transaction.done   = afr_create_done;
-	local->transaction.unwind = afr_create_unwind;
+	idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
 
-	afr_build_parent_loc (&local->transaction.parent_loc, loc);
+	pending = alloca0 (priv->child_count);
 
-	local->transaction.main_frame = frame;
-	local->transaction.basename = AFR_BASENAME (loc->path);
-
-	afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+	for (i = 0; i < priv->child_count; i++) {
+		if (local->transaction.pre_op[i] &&
+		    !local->transaction.failed_subvols[i]) {
+			call_count ++;
+			continue;
+		}
 
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (create, frame, op_ret, op_errno,
-                                  NULL, NULL, NULL, NULL, NULL);
+		changelog[i][idx] = hton32(1);
+		pending[i] = 1;
 	}
 
-	return 0;
-}
-
-/* }}} */
-
-/* {{{ mknod */
-
-int
-afr_mknod_unwind (call_frame_t *frame, xlator_t *this)
-{
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
+        new_local->pending = changelog;
+        uuid_copy (new_local->loc.gfid, local->cont.dir_fop.buf.ia_gfid);
+        new_local->loc.inode = inode_ref (local->inode);
 
-        struct iatt *unwind_buf = NULL;
 
-	local = frame->local;
+	afr_set_pending_dict (priv, xattr, changelog);
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
+        new_local->call_count = call_count;
 
-	if (main_frame) {
-                if (local->cont.mknod.read_child_buf.ia_ino) {
-                        unwind_buf = &local->cont.mknod.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.mknod.buf;
-                }
+        for (i = 0; i < priv->child_count; i++) {
+		if (pending[i])
+                        continue;
 
-                unwind_buf->ia_ino = local->cont.mknod.ino;
-
-                local->cont.mknod.preparent.ia_ino  = local->cont.mknod.parent_ino;
-                local->cont.mknod.postparent.ia_ino = local->cont.mknod.parent_ino;
-
-		AFR_STACK_UNWIND (mknod, main_frame,
-                                  local->op_ret, local->op_errno,
-				  local->cont.mknod.inode,
-				  unwind_buf, &local->cont.mknod.preparent,
-                                  &local->cont.mknod.postparent);
+                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, NULL);
+		if (!--call_count)
+			break;
         }
 
-	return 0;
+        new_frame = NULL;
+out:
+        if (new_frame)
+                AFR_STACK_DESTROY (new_frame);
+	if (xattr)
+		dict_unref (xattr);
+        return;
 }
 
 
-int
-afr_mknod_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, inode_t *inode,
-                    struct iatt *buf, struct iatt *preparent,
-                    struct iatt *postparent)
+void
+afr_mark_entry_pending_changelog (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+	int pre_op_count = 0;
+	int failed_count = 0;
 
-	int call_count = -1;
-	int child_index = -1;
+        local = frame->local;
+        priv  = this->private;
 
-	local = frame->local;
-	priv = this->private;
-
-	child_index = (long) cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
-		
-		if (op_ret != -1) {
-			local->op_ret = op_ret;
-
-			if (local->success_count == 0){
-				local->cont.mknod.buf   = *buf;
-				local->cont.mknod.ino   = 
-					afr_itransform (buf->ia_ino,
-							priv->child_count,
-							child_index);
-
-                                if (priv->read_child >= 0) {
-                                        afr_set_read_child (this, inode,
-                                                            priv->read_child);
-                                } else {
-                                        afr_set_read_child (this, inode,
-                                                            local->read_child_index);
-                                }
-			}
+        if (local->op_ret < 0)
+		return;
 
-                        if (child_index == local->first_up_child) {
-                                local->cont.mknod.ino =
-                                        afr_itransform (buf->ia_ino,
-                                                        priv->child_count,
-                                                        local->first_up_child);
-                        }
-
-                        if (child_index == local->read_child_index) {
-                                local->cont.mknod.read_child_buf = *buf;
-                                local->cont.mknod.preparent      = *preparent;
-                                local->cont.mknod.postparent     = *postparent;
-                        }
-                        
-			local->cont.mknod.inode = inode;
-
-			local->success_count++;
-		}
+	if (local->op != GF_FOP_CREATE && local->op != GF_FOP_MKNOD)
+		return;
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+	pre_op_count = AFR_COUNT (local->transaction.pre_op, priv->child_count);
+	failed_count = AFR_COUNT (local->transaction.failed_subvols,
+				  priv->child_count);
 
-	call_count = afr_frame_return (frame);
+	if (pre_op_count == priv->child_count && !failed_count)
+		return;
 
-	if (call_count == 0) {
-		local->transaction.unwind (frame, this);
+        afr_mark_new_entry_changelog (frame, this);
 
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
+        return;
 }
 
 
-int32_t
-afr_mknod_wind (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
+/* {{{ create */
 
-	int call_count = -1;
-	int i = 0;
+int
+afr_create_unwind (call_frame_t *frame, xlator_t *this)
+{
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-	local = frame->local;
-	priv  = this->private;
+        local = frame->local;
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+	main_frame = afr_transaction_detach_fop_frame (frame);
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
+        if (!main_frame)
 		return 0;
-	}
 
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_mknod_wind_cbk, (void *) (long) i,
-					   priv->children[i], 
-					   priv->children[i]->fops->mknod,
-					   &local->loc, local->cont.mknod.mode,
-					   local->cont.mknod.dev,
-                                           local->cont.mknod.params);
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+	AFR_STACK_UNWIND (create, main_frame, local->op_ret, local->op_errno,
+			  local->cont.create.fd, local->inode,
+			  &local->cont.dir_fop.buf,
+			  &local->cont.dir_fop.preparent,
+			  &local->cont.dir_fop.postparent, local->xdata_rsp);
+        return 0;
 }
 
 
 int
-afr_mknod_done (call_frame_t *frame, xlator_t *this)
+afr_create_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno,
+                     fd_t *fd, inode_t *inode, struct iatt *buf,
+                     struct iatt *preparent, struct iatt *postparent,
+                     dict_t *xdata)
 {
-	afr_local_t * local = NULL;
-
-	local = frame->local;
+        return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
+				    preparent, postparent, NULL, NULL, xdata);
+}
 
-	local->transaction.unwind (frame, this);
-	AFR_STACK_DESTROY (frame);
 
-	return 0;
+int
+afr_create_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+
+        local = frame->local;
+        priv = this->private;
+
+	STACK_WIND_COOKIE (frame, afr_create_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->create,
+			   &local->loc, local->cont.create.flags,
+			   local->cont.create.mode, local->umask,
+			   local->cont.create.fd, local->xdata_req);
+        return 0;
 }
 
 
 int
-afr_mknod (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, mode_t mode, dev_t dev, dict_t *params)
+afr_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)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t  * transaction_frame = NULL;
+        afr_private_t           *priv                   = NULL;
+        afr_local_t             *local                  = NULL;
+        afr_internal_lock_t     *int_lock               = NULL;
+        call_frame_t            *transaction_frame      = NULL;
+        int                     ret                     = -1;
+        int                     op_errno                = ENOMEM;
 
-	int ret = -1;
+        priv = this->private;
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        QUORUM_CHECK(create,out);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-
-	priv = this->private;
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto out;
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
 		goto out;
-	}
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        loc_copy (&local->loc, loc);
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
+	local->fd_ctx = afr_fd_ctx_get (fd, this);
+	if (!local->fd_ctx)
 		goto out;
-	}
-
-	transaction_frame->local = local;
-
-	loc_copy (&local->loc, loc);
 
-        LOCK (&priv->read_child_lock);
-        {
-                local->read_child_index = (++priv->read_child_rr) 
-                        % (priv->child_count);
-        }
-        UNLOCK (&priv->read_child_lock);
-
-	local->cont.mknod.mode  = mode;
-	local->cont.mknod.dev   = dev;
-        if (params)
-                local->cont.mknod.params = dict_ref (params);
-
-        if (loc->parent)
-                local->cont.mknod.parent_ino = loc->parent->ino;
+	local->inode = inode_ref (loc->inode);
+	local->parent = inode_ref (loc->parent);
 
-	local->transaction.fop    = afr_mknod_wind;
-	local->transaction.done   = afr_mknod_done;
-	local->transaction.unwind = afr_mknod_unwind;
+        local->op                = GF_FOP_CREATE;
+        local->cont.create.flags = flags;
+        local->cont.create.mode  = mode;
+        local->cont.create.fd    = fd_ref (fd);
+        local->umask  = umask;
 
-	afr_build_parent_loc (&local->transaction.parent_loc, loc);
+        if (xdata)
+                local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	local->transaction.main_frame = frame;
-	local->transaction.basename = AFR_BASENAME (loc->path);
+	if (!local->xdata_req)
+		goto out;
 
-	afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        local->transaction.wind   = afr_create_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_create_unwind;
+
+        ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+
+        local->transaction.main_frame = frame;
+        local->transaction.basename = AFR_BASENAME (loc->path);
+        int_lock = &local->internal_lock;
+
+        int_lock->lockee_count = 0;
+        ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+                                     &local->transaction.parent_loc,
+                                     local->transaction.basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        int_lock->lockee_count++;
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
 
-	op_ret = 0;
+	return 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (mknod, frame, op_ret, op_errno,
-                                  NULL, NULL, NULL, NULL);
-	}
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
 
-	return 0;
+	AFR_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+			  NULL, NULL);
+        return 0;
 }
 
 /* }}} */
 
-/* {{{ mkdir */
-
+/* {{{ mknod */
 
 int
-afr_mkdir_unwind (call_frame_t *frame, xlator_t *this)
+afr_mknod_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
-
-        struct iatt *unwind_buf = NULL;
-        
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
-
-	if (main_frame) {
-                if (local->cont.mkdir.read_child_buf.ia_ino) {
-                        unwind_buf = &local->cont.mkdir.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.mkdir.buf;
-                }
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-                unwind_buf->ia_ino = local->cont.mkdir.ino;
+        local = frame->local;
 
-                local->cont.mkdir.preparent.ia_ino  = local->cont.mkdir.parent_ino;
-                local->cont.mkdir.postparent.ia_ino = local->cont.mkdir.parent_ino;
-
-		AFR_STACK_UNWIND (mkdir, main_frame,
-                                  local->op_ret, local->op_errno,
-				  local->cont.mkdir.inode,
-				  unwind_buf, &local->cont.mkdir.preparent,
-                        &local->cont.mkdir.postparent);
-        }
+	main_frame = afr_transaction_detach_fop_frame (frame);
+        if (!main_frame)
+		return 0;
 
-	return 0;
+	AFR_STACK_UNWIND (mknod, main_frame, local->op_ret, local->op_errno,
+			  local->inode, &local->cont.dir_fop.buf,
+			  &local->cont.dir_fop.preparent,
+			  &local->cont.dir_fop.postparent, local->xdata_rsp);
+        return 0;
 }
 
 
 int
-afr_mkdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+afr_mknod_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno, inode_t *inode,
                     struct iatt *buf, struct iatt *preparent,
-                    struct iatt *postparent)
+                    struct iatt *postparent, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
+	return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
+				    preparent, postparent, NULL, NULL, xdata);
+}
 
-	int call_count = -1;
-	int child_index = -1;
 
-	local = frame->local;
-	priv = this->private;
+int
+afr_mknod_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+
+        local = frame->local;
+        priv  = this->private;
+
+	STACK_WIND_COOKIE (frame, afr_mknod_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->mknod,
+			   &local->loc, local->cont.mknod.mode,
+			   local->cont.mknod.dev, local->umask,
+			   local->xdata_req);
+        return 0;
+}
 
-	child_index = (long) cookie;
+int
+afr_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+           dev_t dev, mode_t umask, dict_t *xdata)
+{
+        afr_private_t           *priv                   = NULL;
+        afr_local_t             *local                  = NULL;
+        afr_internal_lock_t     *int_lock               = NULL;
+        call_frame_t            *transaction_frame      = NULL;
+        int                     ret                     = -1;
+        int                     op_errno                = ENOMEM;
 
-	LOCK (&frame->lock);
-	{
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
+        priv = this->private;
 
-		if (op_ret != -1) {
-			local->op_ret           = op_ret;
+        QUORUM_CHECK(mknod,out);
 
-			if (local->success_count == 0) {
-				local->cont.mkdir.buf   = *buf;
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto out;
 
-                                local->cont.mkdir.ino =
-                                        afr_itransform (buf->ia_ino,
-                                                        priv->child_count,
-                                                        child_index);
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+		goto out;
 
-                                if (priv->read_child >= 0) {
-                                        afr_set_read_child (this, inode,
-                                                            priv->read_child);
-                                } else {
-                                        afr_set_read_child (this, inode,
-                                                            local->read_child_index);
-                                }
-			}
+        loc_copy (&local->loc, loc);
+	local->inode = inode_ref (loc->inode);
+	local->parent = inode_ref (loc->parent);
 
-                        if (child_index == local->first_up_child) {
-                                local->cont.mkdir.ino =
-                                        afr_itransform (buf->ia_ino,
-                                                        priv->child_count,
-                                                        local->first_up_child);
-                        }
-                        
-                        if (child_index == local->read_child_index) {
-                                local->cont.mkdir.read_child_buf = *buf;
-                                local->cont.mkdir.preparent      = *preparent;
-                                local->cont.mkdir.postparent     = *postparent;
-                        }
-
-			local->cont.mkdir.inode = inode;
-
-			local->success_count++;
-		}
+        local->op               = GF_FOP_MKNOD;
+        local->cont.mknod.mode  = mode;
+        local->cont.mknod.dev   = dev;
+        local->umask = umask;
 
-		local->op_errno         = op_errno;
-	}
-	UNLOCK (&frame->lock);
+        if (xdata)
+                local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	call_count = afr_frame_return (frame);
+	if (!local->xdata_req)
+		goto out;
 
-	if (call_count == 0) {
-		local->transaction.unwind (frame, this);
+        local->transaction.wind   = afr_mknod_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_mknod_unwind;
+
+        ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+
+        local->transaction.main_frame = frame;
+        local->transaction.basename = AFR_BASENAME (loc->path);
+        int_lock = &local->internal_lock;
+
+        int_lock->lockee_count = 0;
+        ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+                                     &local->transaction.parent_loc,
+                                     local->transaction.basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        int_lock->lockee_count++;
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
+		op_errno = -ret;
+		goto out;
+        }
 
-		local->transaction.resume (frame, this);
-	}
-	
 	return 0;
+out:
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
+
+	AFR_STACK_UNWIND (mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+			  NULL);
+        return 0;
 }
 
+/* }}} */
+
+/* {{{ mkdir */
+
 
 int
-afr_mkdir_wind (call_frame_t *frame, xlator_t *this)
+afr_mkdir_unwind (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_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+        local = frame->local;
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
+	main_frame = afr_transaction_detach_fop_frame (frame);
+        if (!main_frame)
 		return 0;
-	}
 
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_mkdir_wind_cbk,
-					   (void *) (long) i,	
-					   priv->children[i], 
-					   priv->children[i]->fops->mkdir,
-					   &local->loc, local->cont.mkdir.mode,
-                                           local->cont.mkdir.params);
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+	AFR_STACK_UNWIND (mkdir, main_frame, local->op_ret, local->op_errno,
+			  local->inode, &local->cont.dir_fop.buf,
+			  &local->cont.dir_fop.preparent,
+			  &local->cont.dir_fop.postparent, local->xdata_rsp);
+        return 0;
 }
 
 
 int
-afr_mkdir_done (call_frame_t *frame, xlator_t *this)
+afr_mkdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, inode_t *inode,
+                    struct iatt *buf, struct iatt *preparent,
+                    struct iatt *postparent, dict_t *xdata)
 {
-	afr_local_t * local = NULL;
-
-	local = frame->local;
-
-	local->transaction.unwind (frame, this);
+	return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
+				    preparent, postparent, NULL, NULL, xdata);
+}
 
-	AFR_STACK_DESTROY (frame);
 
-	return 0;
+int
+afr_mkdir_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+
+        local = frame->local;
+        priv  = this->private;
+
+	STACK_WIND_COOKIE (frame, afr_mkdir_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->mkdir, &local->loc,
+			   local->cont.mkdir.mode, local->umask,
+			   local->xdata_req);
+        return 0;
 }
 
 
 int
-afr_mkdir (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, mode_t mode, dict_t *params)
+afr_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+	   mode_t umask, dict_t *xdata)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t  * transaction_frame = NULL;
+        afr_private_t           *priv                   = NULL;
+        afr_local_t             *local                  = NULL;
+        afr_internal_lock_t     *int_lock               = NULL;
+        call_frame_t            *transaction_frame      = NULL;
+        int                     ret                     = -1;
+        int                     op_errno                = ENOMEM;
 
-	int ret = -1;
+        priv = this->private;
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        QUORUM_CHECK(mkdir,out);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto out;
 
-	priv = this->private;
-
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
 		goto out;
-	}
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        loc_copy (&local->loc, loc);
+	local->inode = inode_ref (loc->inode);
+	local->parent = inode_ref (loc->parent);
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        local->cont.mkdir.mode  = mode;
+        local->umask = umask;
 
-	transaction_frame->local = local;
+        if (xdata)
+                local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	loc_copy (&local->loc, loc);
+	if (!local->xdata_req)
+		goto out;
 
-        LOCK (&priv->read_child_lock);
-        {
-                local->read_child_index = (++priv->read_child_rr) 
-                        % (priv->child_count);
+        local->op = GF_FOP_MKDIR;
+        local->transaction.wind   = afr_mkdir_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_mkdir_unwind;
+
+        ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+
+        local->transaction.main_frame = frame;
+        local->transaction.basename = AFR_BASENAME (loc->path);
+        int_lock = &local->internal_lock;
+
+        int_lock->lockee_count = 0;
+        ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+                                     &local->transaction.parent_loc,
+                                     local->transaction.basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        int_lock->lockee_count++;
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
+		op_errno = -ret;
+		goto out;
         }
-        UNLOCK (&priv->read_child_lock);
-
-	local->cont.mkdir.mode  = mode;
-        if (params)
-                local->cont.mkdir.params = dict_ref (params);
-
-        if (loc->parent)
-                local->cont.mkdir.parent_ino = loc->parent->ino;
-
-	local->transaction.fop    = afr_mkdir_wind;
-	local->transaction.done   = afr_mkdir_done;
-	local->transaction.unwind = afr_mkdir_unwind;
-
-	afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
-	local->transaction.main_frame = frame;
-	local->transaction.basename = AFR_BASENAME (loc->path);
 
-	afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
-
-	op_ret = 0;
+	return 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
 
-		AFR_STACK_UNWIND (mkdir, frame, op_ret, op_errno,
-                                  NULL, NULL, NULL, NULL);
-	}
-
-	return 0;
+	AFR_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+			  NULL);
+        return 0;
 }
 
 /* }}} */
@@ -867,233 +791,127 @@ out:
 int
 afr_link_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-        struct iatt *unwind_buf = NULL;
+        local = frame->local;
 
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
-
-	if (main_frame) {
-                if (local->cont.link.read_child_buf.ia_ino) {
-                        unwind_buf = &local->cont.link.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.link.buf;
-                }
-
-		unwind_buf->ia_ino = local->cont.link.ino;
-
-                local->cont.link.preparent.ia_ino  = local->cont.link.parent_ino;
-                local->cont.link.postparent.ia_ino = local->cont.link.parent_ino;
-
-		AFR_STACK_UNWIND (link, main_frame,
-                                  local->op_ret, local->op_errno, 
-				  local->cont.link.inode,
-				  unwind_buf, &local->cont.link.preparent,
-                                  &local->cont.link.postparent);
-	}
+	main_frame = afr_transaction_detach_fop_frame (frame);
+        if (!main_frame)
+		return 0;
 
-	return 0;
+	AFR_STACK_UNWIND (link, main_frame, local->op_ret, local->op_errno,
+			  local->inode, &local->cont.dir_fop.buf,
+			  &local->cont.dir_fop.preparent,
+			  &local->cont.dir_fop.postparent, local->xdata_rsp);
+        return 0;
 }
 
 
 int
-afr_link_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		   int32_t op_ret, int32_t op_errno, inode_t *inode,
+afr_link_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, inode_t *inode,
                    struct iatt *buf, struct iatt *preparent,
-                   struct iatt *postparent)
+                   struct iatt *postparent, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int call_count = -1;
-	int child_index = -1;
-
-	local = frame->local;
-	priv = this->private;
-
-	child_index = (long) cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
-
-		if (op_ret != -1) {
-			local->op_ret   = op_ret;
-
-			if (local->success_count == 0) {
-				local->cont.link.buf        = *buf;
-
-                                if (priv->read_child >= 0) {
-                                        afr_set_read_child (this, inode,
-                                                            priv->read_child);
-                                } else {
-                                        afr_set_read_child (this, inode,
-                                                            local->read_child_index);
-                                }
-			}
-
-                        if (child_index == local->read_child_index) {
-                                local->cont.link.read_child_buf = *buf;
-                                local->cont.link.preparent      = *preparent;
-                                local->cont.link.postparent     = *postparent;
-                        }
-
-			local->cont.link.inode    = inode;
-
-			local->success_count++;
-		}
-
-		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);
-	}
-	
-	return 0;
+        return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
+				    preparent, postparent, NULL, NULL, xdata);
 }
 
 
 int
-afr_link_wind (call_frame_t *frame, xlator_t *this)
+afr_link_wind (call_frame_t *frame, xlator_t *this, int subvol)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-
-	int call_count = -1;
-	int i = 0;
-
-	local = frame->local;
-	priv  = this->private;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+        local = frame->local;
+        priv  = this->private;
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_link_wind_cbk, (void *) (long) i,
-					   priv->children[i], 
-					   priv->children[i]->fops->link,
-					   &local->loc,
-					   &local->newloc);
-			
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+	STACK_WIND_COOKIE (frame, afr_link_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->link,
+			   &local->loc, &local->newloc, local->xdata_req);
+        return 0;
 }
 
 
 int
-afr_link_done (call_frame_t *frame, xlator_t *this)
+afr_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+	  dict_t *xdata)
 {
-	afr_local_t * local = frame->local;
+        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                = ENOMEM;
 
-	local->transaction.unwind (frame, this);
+        priv = this->private;
 
-	AFR_STACK_DESTROY (frame);
+        QUORUM_CHECK(link,out);
 
-	return 0;
-}
-
-
-int
-afr_link (call_frame_t *frame, xlator_t *this,
-	  loc_t *oldloc, loc_t *newloc)
-{
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t  * transaction_frame = NULL;
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto out;
 
-	int ret = -1;
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+		goto out;
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        loc_copy (&local->loc,    oldloc);
+        loc_copy (&local->newloc, newloc);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+	local->inode = inode_ref (oldloc->inode);
+	local->parent = inode_ref (newloc->parent);
 
-	priv = this->private;
+        if (xdata)
+                local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
+	if (!local->xdata_req)
 		goto out;
-	}
-
-	ALLOC_OR_GOTO (local, afr_local_t, out);
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
+        local->op = GF_FOP_LINK;
+
+        local->transaction.wind   = afr_link_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_link_unwind;
+
+        ret = afr_build_parent_loc (&local->transaction.parent_loc, newloc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+
+        local->transaction.main_frame   = frame;
+        local->transaction.basename     = AFR_BASENAME (newloc->path);
+        int_lock = &local->internal_lock;
+
+        int_lock->lockee_count = 0;
+        ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+                                     &local->transaction.parent_loc,
+                                     local->transaction.basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        int_lock->lockee_count++;
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
 		op_errno = -ret;
 		goto out;
-	}
-
-	transaction_frame->local = local;
-
-	loc_copy (&local->loc,    oldloc);
-	loc_copy (&local->newloc, newloc);
-
-        LOCK (&priv->read_child_lock);
-        {
-                local->read_child_index = (++priv->read_child_rr) 
-                        % (priv->child_count);
         }
-        UNLOCK (&priv->read_child_lock);
-
-	local->cont.link.ino = oldloc->inode->ino;
-
-        if (oldloc->parent)
-                local->cont.link.parent_ino = newloc->parent->ino;
-
-	local->transaction.fop    = afr_link_wind;
-	local->transaction.done   = afr_link_done;
-	local->transaction.unwind = afr_link_unwind;
-
-	afr_build_parent_loc (&local->transaction.parent_loc, oldloc);
-
-	local->transaction.main_frame   = frame;
-	local->transaction.basename     = AFR_BASENAME (oldloc->path);
-	local->transaction.new_basename = AFR_BASENAME (newloc->path);
 
-	afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
-
-	op_ret = 0;
+	return 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (link, frame, op_ret, op_errno,
-                                  NULL, NULL, NULL, NULL);
-	}
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
 
-	return 0;
+	AFR_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+			  NULL);
+        return 0;
 }
 
 /* }}} */
@@ -1104,246 +922,128 @@ out:
 int
 afr_symlink_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
-        
-        struct iatt *unwind_buf = NULL;
-        
-	local = frame->local;
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
+        local = frame->local;
 
-	if (main_frame) {
-                if (local->cont.symlink.read_child_buf.ia_ino) {
-                        unwind_buf = &local->cont.symlink.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.symlink.buf;
-                }
-
-                unwind_buf->ia_ino = local->cont.symlink.ino;
-
-                local->cont.symlink.preparent.ia_ino  = local->cont.symlink.parent_ino;
-                local->cont.symlink.postparent.ia_ino = local->cont.symlink.parent_ino;
-
-		AFR_STACK_UNWIND (symlink, main_frame,
-                                  local->op_ret, local->op_errno,
-				  local->cont.symlink.inode,
-				  unwind_buf, &local->cont.symlink.preparent,
-                                  &local->cont.symlink.postparent);
-        }
+	main_frame = afr_transaction_detach_fop_frame (frame);
+        if (!main_frame)
+		return 0;
 
-	return 0;
+	AFR_STACK_UNWIND (symlink, main_frame, local->op_ret, local->op_errno,
+			  local->inode, &local->cont.dir_fop.buf,
+			  &local->cont.dir_fop.preparent,
+			  &local->cont.dir_fop.postparent, local->xdata_rsp);
+        return 0;
 }
 
 
 int
-afr_symlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		      int32_t op_ret, int32_t op_errno, inode_t *inode,
+afr_symlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, inode_t *inode,
                       struct iatt *buf, struct iatt *preparent,
-                      struct iatt *postparent)
+                      struct iatt *postparent, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int call_count = -1;
-	int child_index = -1;
-
-	local = frame->local;
-	priv = this->private;
-
-	child_index = (long) cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
-		
-		if (op_ret != -1) {
-			local->op_ret   = op_ret;
-
-			if (local->success_count == 0) {
-				local->cont.symlink.buf        = *buf;
-				local->cont.symlink.ino = 
-					afr_itransform (buf->ia_ino, priv->child_count,
-							child_index);
-
-                                if (priv->read_child >= 0) {
-                                        afr_set_read_child (this, inode,
-                                                            priv->read_child);
-                                } else {
-                                        afr_set_read_child (this, inode,
-                                                            local->read_child_index);
-                                }
-			}
-
-                        if (child_index == local->first_up_child) {
-                                local->cont.symlink.ino =
-                                        afr_itransform (buf->ia_ino,
-                                                        priv->child_count,
-                                                        local->first_up_child);
-                        }
-
-                        if (child_index == local->read_child_index) {
-                                local->cont.symlink.read_child_buf = *buf;
-                                local->cont.symlink.preparent      = *preparent;
-                                local->cont.symlink.postparent     = *postparent;
-                        }
-
-			local->cont.symlink.inode    = inode;
-
-			local->success_count++;
-		}
-
-		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);
-	}
-	
-	return 0;
+        return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
+				    preparent, postparent, NULL, NULL, xdata);
 }
 
 
 int
-afr_symlink_wind (call_frame_t *frame, xlator_t *this)
+afr_symlink_wind (call_frame_t *frame, xlator_t *this, int subvol)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-
-	int call_count = -1;
-	int i = 0;
-
-	local = frame->local;
-	priv = this->private;
-
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_symlink_wind_cbk,
-					   (void *) (long) i,	
-					   priv->children[i], 
-					   priv->children[i]->fops->symlink,
-					   local->cont.symlink.linkpath,
-					   &local->loc,
-                                           local->cont.symlink.params);
-
-			if (!--call_count)
-				break;
-
-		}
-	}
-	
-	return 0;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+
+        local = frame->local;
+        priv = this->private;
+
+	STACK_WIND_COOKIE (frame, afr_symlink_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->symlink,
+			   local->cont.symlink.linkpath, &local->loc,
+			   local->umask, local->xdata_req);
+        return 0;
 }
 
 
 int
-afr_symlink_done (call_frame_t *frame, xlator_t *this)
+afr_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
+	     loc_t *loc, mode_t umask, dict_t *xdata)
 {
-	afr_local_t * local = frame->local;
+        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                = ENOMEM;
 
-	local->transaction.unwind (frame, this);
+        priv = this->private;
 
-	AFR_STACK_DESTROY (frame);
-	
-	return 0;
-}
+        QUORUM_CHECK(symlink,out);
 
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto out;
 
-int
-afr_symlink (call_frame_t *frame, xlator_t *this,
-	     const char *linkpath, loc_t *loc, dict_t *params)
-{
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t  * transaction_frame = NULL;
-
-	int ret = -1;
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+		goto out;
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        loc_copy (&local->loc, loc);
+	local->inode = inode_ref (loc->inode);
+	local->parent = inode_ref (loc->parent);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        local->cont.symlink.linkpath = gf_strdup (linkpath);
+        local->umask = umask;
 
-	priv = this->private;
+        if (xdata)
+                local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
+	if (!local->xdata_req)
 		goto out;
-	}
-
-	ALLOC_OR_GOTO (local, afr_local_t, out);
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
+        local->op = GF_FOP_SYMLINK;
+        local->transaction.wind   = afr_symlink_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_symlink_unwind;
+
+        ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+
+        local->transaction.main_frame   = frame;
+        local->transaction.basename     = AFR_BASENAME (loc->path);
+        int_lock = &local->internal_lock;
+
+        int_lock->lockee_count = 0;
+        ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+                                     &local->transaction.parent_loc,
+                                     local->transaction.basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        int_lock->lockee_count++;
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
 		op_errno = -ret;
 		goto out;
-	}
-
-	transaction_frame->local = local;
-	
-	loc_copy (&local->loc, loc);
-
-        LOCK (&priv->read_child_lock);
-        {
-                local->read_child_index = (++priv->read_child_rr) 
-                        % (priv->child_count);
         }
-        UNLOCK (&priv->read_child_lock);
-
-	local->cont.symlink.linkpath = gf_strdup (linkpath);
-        if (params)
-                local->cont.symlink.params = dict_ref (params);
-
-        if (loc->parent)
-                local->cont.symlink.parent_ino = loc->parent->ino;
-
-	local->transaction.fop    = afr_symlink_wind;
-	local->transaction.done   = afr_symlink_done;
-	local->transaction.unwind = afr_symlink_unwind;
-
-	afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
-	local->transaction.main_frame   = frame;
-	local->transaction.basename     = AFR_BASENAME (loc->path);
-
-	afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
 
-	op_ret = 0;
+	return 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (symlink, frame, op_ret, op_errno,
-                                  NULL, NULL, NULL, NULL);
-	}
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
 
-	return 0;
+	AFR_STACK_UNWIND (symlink, frame, -1, op_errno, NULL, NULL, NULL,
+			  NULL, NULL);
+        return 0;
 }
 
 /* }}} */
@@ -1353,228 +1053,159 @@ out:
 int
 afr_rename_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-        struct iatt *unwind_buf = NULL;
-
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
-
-	if (main_frame) {
-                if (local->cont.rename.read_child_buf.ia_ino) {
-                        unwind_buf = &local->cont.rename.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.rename.buf;
-                }
-
-		unwind_buf->ia_ino = local->cont.rename.ino;
-
-                local->cont.rename.preoldparent.ia_ino  = local->cont.rename.oldparent_ino;
-                local->cont.rename.postoldparent.ia_ino = local->cont.rename.oldparent_ino;
-                local->cont.rename.prenewparent.ia_ino  = local->cont.rename.newparent_ino;
-                local->cont.rename.postnewparent.ia_ino = local->cont.rename.newparent_ino;
-
-		AFR_STACK_UNWIND (rename, main_frame,
-                                  local->op_ret, local->op_errno, 
-				  unwind_buf,
-                                  &local->cont.rename.preoldparent,
-                                  &local->cont.rename.postoldparent,
-                                  &local->cont.rename.prenewparent,
-                                  &local->cont.rename.postnewparent);
-	}
+	main_frame = afr_transaction_detach_fop_frame (frame);
+        if (!main_frame)
+		return 0;
 
-	return 0;
+	AFR_STACK_UNWIND (rename, main_frame, local->op_ret, local->op_errno,
+			  &local->cont.dir_fop.buf,
+			  &local->cont.dir_fop.preparent,
+			  &local->cont.dir_fop.postparent,
+			  &local->cont.dir_fop.prenewparent,
+			  &local->cont.dir_fop.postnewparent, local->xdata_rsp);
+        return 0;
 }
 
 
 int
-afr_rename_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		     int32_t op_ret, int32_t op_errno, struct iatt *buf,
+afr_rename_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *buf,
                      struct iatt *preoldparent, struct iatt *postoldparent,
-                     struct iatt *prenewparent, struct iatt *postnewparent)
+                     struct iatt *prenewparent, struct iatt *postnewparent,
+                     dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-
-	int call_count = -1;
-	int child_index = -1;
-
-	local = frame->local;
-
-	child_index = (long) cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (afr_fop_failed (op_ret, op_errno) && op_errno != ENOTEMPTY)
-			afr_transaction_fop_failed (frame, this, child_index);
-
-		if (op_ret != -1) {
-                        if (local->success_count == 0) {
-                                local->op_ret = op_ret;
-		
-                                if (buf) {
-                                        local->cont.rename.buf = *buf;
-                                }
-
-                                local->success_count++;
-                        }
-
-                        if (child_index == local->read_child_index) {
-                                local->cont.rename.read_child_buf = *buf;
-
-                                local->cont.rename.preoldparent  = *preoldparent;
-                                local->cont.rename.postoldparent = *postoldparent;
-                                local->cont.rename.prenewparent  = *prenewparent;
-                                local->cont.rename.postnewparent = *postnewparent;
-                        }
-                }
-
-		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);
-	}
-	
-	return 0;
+        return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, buf,
+				    preoldparent, postoldparent, prenewparent,
+				    postnewparent, xdata);
 }
 
 
-int32_t
-afr_rename_wind (call_frame_t *frame, xlator_t *this)
+int
+afr_rename_wind (call_frame_t *frame, xlator_t *this, int subvol)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-
-	int call_count = -1;
-	int i = 0;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
 	local = frame->local;
 	priv = this->private;
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_rename_wind_cbk, 
-					   (void *) (long) i,	
-					   priv->children[i], 
-					   priv->children[i]->fops->rename,
-					   &local->loc,
-					   &local->newloc);
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+	STACK_WIND_COOKIE (frame, afr_rename_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->rename,
+			   &local->loc, &local->newloc, local->xdata_req);
+        return 0;
 }
 
 
 int
-afr_rename_done (call_frame_t *frame, xlator_t *this)
+afr_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+	    dict_t *xdata)
 {
-	afr_local_t * local = frame->local;
+        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                = ENOMEM;
+        int                     nlockee                 = 0;
 
-	local->transaction.unwind (frame, this);
-
-	AFR_STACK_DESTROY (frame);
-	
-	return 0;
-}
+        priv = this->private;
 
+        QUORUM_CHECK(rename,out);
 
-int
-afr_rename (call_frame_t *frame, xlator_t *this,
-	    loc_t *oldloc, loc_t *newloc)
-{
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t  * transaction_frame = NULL;
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                op_errno = ENOMEM;
 
-	int ret = -1;
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+		goto out;
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        loc_copy (&local->loc,    oldloc);
+        loc_copy (&local->newloc, newloc);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+	local->inode = inode_ref (oldloc->inode);
+	local->parent = inode_ref (oldloc->parent);
+	local->parent2 = inode_ref (newloc->parent);
 
-	priv = this->private;
+        if (xdata)
+                local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
+	if (!local->xdata_req)
 		goto out;
-	}
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        local->op = GF_FOP_RENAME;
+        local->transaction.wind   = afr_rename_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_rename_unwind;
+
+        ret = afr_build_parent_loc (&local->transaction.parent_loc, oldloc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+        ret = afr_build_parent_loc (&local->transaction.new_parent_loc, newloc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+
+        local->transaction.main_frame   = frame;
+        local->transaction.basename     = AFR_BASENAME (oldloc->path);
+        local->transaction.new_basename = AFR_BASENAME (newloc->path);
+        int_lock = &local->internal_lock;
+
+        int_lock->lockee_count = nlockee = 0;
+        ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+                                     &local->transaction.new_parent_loc,
+                                     local->transaction.new_basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        nlockee++;
+        ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+                                     &local->transaction.parent_loc,
+                                     local->transaction.basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        nlockee++;
+        if (local->newloc.inode && IA_ISDIR (local->newloc.inode->ia_type)) {
+                ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+                                             &local->newloc,
+                                             NULL,
+                                             priv->child_count);
+                if (ret)
+                        goto out;
+
+                nlockee++;
+        }
+        qsort (int_lock->lockee, nlockee, sizeof (*int_lock->lockee),
+               afr_entry_lockee_cmp);
+        int_lock->lockee_count = nlockee;
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_RENAME_TRANSACTION);
+        if (ret < 0) {
 		op_errno = -ret;
 		goto out;
-	}
-
-	transaction_frame->local = local;
-
-	loc_copy (&local->loc,    oldloc);
-	loc_copy (&local->newloc, newloc);
-
-        local->read_child_index = afr_read_child (this, oldloc->inode);
-
-	local->cont.rename.ino = oldloc->inode->ino;
-
-        if (oldloc->parent)
-                local->cont.rename.oldparent_ino = oldloc->parent->ino;
-        if (newloc->parent)
-                local->cont.rename.newparent_ino = newloc->parent->ino;
-
-	local->transaction.fop    = afr_rename_wind;
-	local->transaction.done   = afr_rename_done;
-	local->transaction.unwind = afr_rename_unwind;
-
-	afr_build_parent_loc (&local->transaction.parent_loc, oldloc);
-	afr_build_parent_loc (&local->transaction.new_parent_loc, newloc);
-
-	local->transaction.main_frame   = frame;
-	local->transaction.basename     = AFR_BASENAME (oldloc->path);
-	local->transaction.new_basename = AFR_BASENAME (newloc->path);
-
-	afr_transaction (transaction_frame, this, AFR_ENTRY_RENAME_TRANSACTION);
+        }
 
-	op_ret = 0;
+	return 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
 
-		AFR_STACK_UNWIND (rename, frame, op_ret, op_errno,
-                                  NULL, NULL, NULL, NULL, NULL);
-	}
-
-	return 0;
+	AFR_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+			  NULL, NULL);
+        return 0;
 }
 
 /* }}} */
@@ -1584,201 +1215,123 @@ out:
 int
 afr_unlink_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = 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);
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-	if (main_frame) {
-                local->cont.unlink.preparent.ia_ino  = local->cont.unlink.parent_ino;
-                local->cont.unlink.postparent.ia_ino = local->cont.unlink.parent_ino;
+        local = frame->local;
 
-		AFR_STACK_UNWIND (unlink, main_frame,
-                                  local->op_ret, local->op_errno,
-                                  &local->cont.unlink.preparent,
-                                  &local->cont.unlink.postparent);
-        }
+	main_frame = afr_transaction_detach_fop_frame (frame);
+        if (!main_frame)
+		return 0;
 
-	return 0;
+	AFR_STACK_UNWIND (unlink, main_frame, local->op_ret, local->op_errno,
+			  &local->cont.dir_fop.preparent,
+			  &local->cont.dir_fop.postparent, local->xdata_rsp);
+        return 0;
 }
 
 
 int
-afr_unlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		     int32_t op_ret, int32_t op_errno, struct iatt *preparent,
-                     struct iatt *postparent)
+afr_unlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                     struct iatt *postparent, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int call_count  = -1;
-	int child_index = (long) cookie;
-
-	local = frame->local;
-	priv  = this->private;
-
-	LOCK (&frame->lock);
-	{
-                if (child_index == local->read_child_index) {
-                        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.unlink.preparent  = *preparent;
-                                local->cont.unlink.postparent = *postparent;
-			}
-
-                        if (child_index == local->read_child_index) {
-                                local->cont.unlink.preparent  = *preparent;
-                                local->cont.unlink.postparent = *postparent;
-                        }
-
-			local->success_count++;
-		}
-
-		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);
-	}
-	
-	return 0;
+        return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, NULL,
+				    preparent, postparent, NULL, NULL, xdata);
 }
 
 
-int32_t
-afr_unlink_wind (call_frame_t *frame, xlator_t *this)
+int
+afr_unlink_wind (call_frame_t *frame, xlator_t *this, int subvol)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
-	int call_count = -1;
-	int i = 0;
-
-	local = frame->local;
-	priv  = this->private;
+        local = frame->local;
+        priv  = this->private;
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_unlink_wind_cbk,	
-					   (void *) (long) i,
-					   priv->children[i], 
-					   priv->children[i]->fops->unlink,
-					   &local->loc);
-			
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+	STACK_WIND_COOKIE (frame, afr_unlink_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->unlink,
+			   &local->loc, local->xflag, local->xdata_req);
+        return 0;
 }
 
 
-int32_t
-afr_unlink_done (call_frame_t *frame, xlator_t *this)
+int
+afr_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+	    dict_t *xdata)
 {
-	afr_local_t * local = frame->local;
+        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                = ENOMEM;
 
-	local->transaction.unwind (frame, this);
+        priv = this->private;
 
-	AFR_STACK_DESTROY (frame);
-	
-	return 0;
-}
+        QUORUM_CHECK(unlink,out);
 
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto out;
 
-int32_t
-afr_unlink (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc)
-{
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t  * transaction_frame = NULL;
-
-	int ret = -1;
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+		goto out;
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        loc_copy (&local->loc, loc);
+        local->xflag = xflag;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+	local->inode = inode_ref (loc->inode);
+	local->parent = inode_ref (loc->parent);
 
-	priv = this->private;
+        if (xdata)
+                local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
+	if (!local->xdata_req)
 		goto out;
-	}
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
+        local->op = GF_FOP_UNLINK;
+        local->transaction.wind   = afr_unlink_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_unlink_unwind;
+
+        ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+
+        local->transaction.main_frame = frame;
+        local->transaction.basename = AFR_BASENAME (loc->path);
+        int_lock = &local->internal_lock;
+
+        int_lock->lockee_count = 0;
+        ret = afr_init_entry_lockee (&int_lock->lockee[0], local,
+                                     &local->transaction.parent_loc,
+                                     local->transaction.basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        int_lock->lockee_count++;
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
 		op_errno = -ret;
 		goto out;
-	}
-
-	transaction_frame->local = local;
-
-	loc_copy (&local->loc, loc);
-
-        if (loc->parent)
-                local->cont.unlink.parent_ino = loc->parent->ino;
-
-	local->transaction.fop    = afr_unlink_wind;
-	local->transaction.done   = afr_unlink_done;
-	local->transaction.unwind = afr_unlink_unwind;
-
-	afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
-	local->transaction.main_frame = frame;
-	local->transaction.basename = AFR_BASENAME (loc->path);
-
-	afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        }
 
-	op_ret = 0;
+	return 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (unlink, frame, op_ret, op_errno,
-                                  NULL, NULL);
-	}
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
 
-	return 0;
+	AFR_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
+        return 0;
 }
 
 /* }}} */
@@ -1790,204 +1343,137 @@ out:
 int
 afr_rmdir_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
-
-	if (main_frame) {
-                local->cont.rmdir.preparent.ia_ino  = local->cont.rmdir.parent_ino;
-                local->cont.rmdir.postparent.ia_ino = local->cont.rmdir.parent_ino;
-
-		AFR_STACK_UNWIND (rmdir, main_frame,
-                                  local->op_ret, local->op_errno,
-                                  &local->cont.rmdir.preparent,
-                                  &local->cont.rmdir.postparent);
-        }
-
-	return 0;
-}
-
-
-int
-afr_rmdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		    int32_t op_ret, int32_t op_errno, struct iatt *preparent,
-                    struct iatt *postparent)
-{
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int call_count  = -1;
-	int child_index = (long) cookie;
-        int read_child  = 0;
-
-	local = frame->local;
-	priv = this->private;
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-	LOCK (&frame->lock);
-	{
-                if (child_index == read_child) {
-                        local->read_child_returned = _gf_true;
-                }
+        local = frame->local;
 
-		if (afr_fop_failed (op_ret, op_errno) && (op_errno != ENOTEMPTY))
-			afr_transaction_fop_failed (frame, this, child_index);
-
-		if (op_ret != -1) {
-			if (local->success_count == 0) {
-				local->op_ret = op_ret;
-                                local->cont.rmdir.preparent  = *preparent;
-                                local->cont.rmdir.postparent = *postparent;
-
-			}
-
-                        if (child_index == read_child) {
-                                local->cont.rmdir.preparent  = *preparent;
-                                local->cont.rmdir.postparent = *postparent;
-                        }
-
-			local->success_count++;
-		}
-
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
+	main_frame = afr_transaction_detach_fop_frame (frame);
+	if (!main_frame)
+		return 0;
 
-	if (call_count == 0) {
-		local->transaction.unwind (frame, this);
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
+	AFR_STACK_UNWIND (rmdir, main_frame, local->op_ret, local->op_errno,
+			  &local->cont.dir_fop.preparent,
+			  &local->cont.dir_fop.postparent, local->xdata_rsp);
+        return 0;
 }
 
 
 int
-afr_rmdir_wind (call_frame_t *frame, xlator_t *this)
+afr_rmdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                    struct iatt *postparent, dict_t *xdata)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-
-	int call_count = -1;
-	int i = 0;
-
-	local = frame->local;
-	priv  = this->private;
-
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_rmdir_wind_cbk,	
-					   (void *) (long) i,
-					   priv->children[i], 
-					   priv->children[i]->fops->rmdir,
-					   &local->loc, local->cont.rmdir.flags);
-
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+        return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, NULL,
+				    preparent, postparent, NULL, NULL, xdata);
 }
 
 
 int
-afr_rmdir_done (call_frame_t *frame, xlator_t *this)
+afr_rmdir_wind (call_frame_t *frame, xlator_t *this, int subvol)
 {
-	afr_local_t * local = frame->local;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
-	local->transaction.unwind (frame, this);
+        local = frame->local;
+        priv  = this->private;
 
-	AFR_STACK_DESTROY (frame);
-	
-	return 0;
+	STACK_WIND_COOKIE (frame, afr_rmdir_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->rmdir,
+			   &local->loc, local->cont.rmdir.flags, local->xdata_req);
+        return 0;
 }
 
 
 int
-afr_rmdir (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, int flags)
+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;
+        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                = ENOMEM;
+        int                     nlockee                 = 0;
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        priv = this->private;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        QUORUM_CHECK(rmdir,out);
 
-	priv = this->private;
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto out;
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
 		goto out;
-	}
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
-
-	transaction_frame->local = local;
+        loc_copy (&local->loc, loc);
+	local->inode = inode_ref (loc->inode);
+	local->parent = inode_ref (loc->parent);
 
         local->cont.rmdir.flags = flags;
-	loc_copy (&local->loc, loc);
-
-        if (loc->parent)
-                local->cont.rmdir.parent_ino = loc->parent->ino;
 
-	local->transaction.fop    = afr_rmdir_wind;
-	local->transaction.done   = afr_rmdir_done;
-	local->transaction.unwind = afr_rmdir_unwind;
+        if (xdata)
+                local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
-	local->transaction.main_frame = frame;
-	local->transaction.basename = AFR_BASENAME (loc->path);
+	if (!local->xdata_req)
+		goto out;
 
-	afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        local->op = GF_FOP_RMDIR;
+        local->transaction.wind   = afr_rmdir_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_rmdir_unwind;
+
+        ret = afr_build_parent_loc (&local->transaction.parent_loc, loc,
+                                    &op_errno);
+        if (ret)
+                goto out;
+
+        local->transaction.main_frame = frame;
+        local->transaction.basename = AFR_BASENAME (loc->path);
+        int_lock = &local->internal_lock;
+
+        int_lock->lockee_count = nlockee = 0;
+        ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+                                     &local->transaction.parent_loc,
+                                     local->transaction.basename,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        nlockee++;
+        ret = afr_init_entry_lockee (&int_lock->lockee[nlockee], local,
+                                     &local->loc,
+                                     NULL,
+                                     priv->child_count);
+        if (ret)
+                goto out;
+
+        nlockee++;
+        qsort (int_lock->lockee, nlockee, sizeof (*int_lock->lockee),
+               afr_entry_lockee_cmp);
+        int_lock->lockee_count = nlockee;
+
+        ret = afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (ret < 0) {
+		op_errno = -ret;
+		goto out;
+        }
 
-	op_ret = 0;
+	return 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (rmdir, frame, op_ret, op_errno,
-                                  NULL, NULL);
-	}
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
 
-	return 0;
+	AFR_STACK_UNWIND (rmdir, frame, -1, op_errno, NULL, NULL, NULL);
+        return 0;
 }
 
 /* }}} */
-
diff --git a/xlators/cluster/afr/src/afr-dir-write.h b/xlators/cluster/afr/src/afr-dir-write.h
index e589efa37..02f0a3682 100644
--- a/xlators/cluster/afr/src/afr-dir-write.h
+++ b/xlators/cluster/afr/src/afr-dir-write.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __DIR_WRITE_H__
@@ -23,38 +14,34 @@
 int32_t
 afr_create (call_frame_t *frame, xlator_t *this,
             loc_t *loc, int32_t flags, mode_t mode,
-            fd_t *fd, dict_t *params);
+            mode_t umask, fd_t *fd, dict_t *xdata);
 
 int32_t
 afr_mknod (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, mode_t mode, dev_t dev, dict_t *params);
+	   loc_t *loc, mode_t mode, dev_t dev, mode_t umask, dict_t *xdata);
 
 int32_t
 afr_mkdir (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, mode_t mode, dict_t *params);
+	   loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata);
 
 int32_t
 afr_unlink (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc);
+	    loc_t *loc, int xflag, dict_t *xdata);
 
 int32_t
 afr_rmdir (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, int flags);
+	   loc_t *loc, int flags, dict_t *xdata);
 
 int32_t
 afr_link (call_frame_t *frame, xlator_t *this,
-	  loc_t *oldloc, loc_t *newloc);
+	  loc_t *oldloc, loc_t *newloc, dict_t *xdata);
 
 int32_t
 afr_rename (call_frame_t *frame, xlator_t *this,
-	    loc_t *oldloc, loc_t *newloc);
+	    loc_t *oldloc, loc_t *newloc, dict_t *xdata);
 
 int
 afr_symlink (call_frame_t *frame, xlator_t *this,
-	     const char *linkpath, loc_t *oldloc, dict_t *params);
-
-int32_t
-afr_setdents (call_frame_t *frame, xlator_t *this,
-	      fd_t *fd, int32_t flags, dir_entry_t *entries, int32_t count);
+	     const char *linkpath, loc_t *oldloc, mode_t umask, dict_t *params);
 
 #endif /* __DIR_WRITE_H__ */
diff --git a/xlators/cluster/afr/src/afr-inode-read.c b/xlators/cluster/afr/src/afr-inode-read.c
index 399860767..01e078c13 100644
--- a/xlators/cluster/afr/src/afr-inode-read.c
+++ b/xlators/cluster/afr/src/afr-inode-read.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -44,843 +35,1630 @@
 #include "compat-errno.h"
 #include "compat.h"
 
-#include "afr.h"
+#include "afr-transaction.h"
 
 
-/**
- * Common algorithm for inode read calls:
- *
- * - Try the fop on the first child that is up
- * - if we have failed due to ENOTCONN:
- *     try the next child
- *
- * Applicable to: access, stat, fstat, readlink, getxattr
- */
-
 /* {{{ access */
 
-int32_t
-afr_access_cbk (call_frame_t *frame, void *cookie,
-		xlator_t *this, int32_t op_ret, int32_t op_errno)
+int
+afr_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		int op_ret, int op_errno, dict_t *xdata)
 {
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-	xlator_t **     children = NULL;
+        afr_local_t     *local = NULL;
 
-	int unwind     = 1;
-	int last_tried = -1;
-	int this_try = -1;
-        int read_child = -1;
+        local = frame->local;
 
-	priv     = this->private;
-	children = priv->children;
+	if (op_ret < 0) {
+		local->op_ret = op_ret;
+		local->op_errno = op_errno;
 
-	local = frame->local;
+		afr_read_txn_continue (frame, this, (long) cookie);
+		return 0;
+	}
 
-        read_child = (long) cookie;
+	AFR_STACK_UNWIND (access, frame, op_ret, op_errno, xdata);
 
-	if (op_ret == -1) {
-        retry:
-		last_tried = local->cont.access.last_tried;
+        return 0;
+}
 
-		if (all_tried (last_tried, priv->child_count)) {
-			goto out;
-		}
-		this_try    = ++local->cont.access.last_tried;
 
-                if (this_try == read_child) {
-                        goto retry;
-                }
+int
+afr_access_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+	afr_private_t *priv = NULL;
+	afr_local_t *local = NULL;
 
-		unwind = 0;
+	priv = this->private;
+	local = frame->local;
 
-		STACK_WIND_COOKIE (frame, afr_access_cbk,
-				   (void *) (long) read_child,
-				   children[this_try],
-				   children[this_try]->fops->access,
-				   &local->loc, local->cont.access.mask);
+	if (subvol == -1) {
+		AFR_STACK_UNWIND (access, frame, local->op_ret,
+				  local->op_errno, 0);
+		return 0;
 	}
 
-out:
-	if (unwind) {
-		AFR_STACK_UNWIND (access, frame, op_ret, op_errno);
-	}
+        STACK_WIND_COOKIE (frame, afr_access_cbk, (void *) (long) subvol,
+                           priv->children[subvol],
+                           priv->children[subvol]->fops->access,
+                           &local->loc, local->cont.access.mask,
+			   local->xdata_req);
+	return 0;
+}
+
+int
+afr_access (call_frame_t *frame, xlator_t *this, loc_t *loc,
+	    int mask, dict_t *xdata)
+{
+        afr_local_t *local = NULL;
+	int op_errno = 0;
+
+        local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
+
+	local->op = GF_FOP_ACCESS;
+	loc_copy (&local->loc, loc);
+	local->cont.access.mask = mask;
+	if (xdata)
+		local->xdata_req = dict_ref (xdata);
+
+	afr_read_txn (frame, this, loc->inode, afr_access_wind,
+		      AFR_METADATA_TRANSACTION);
 
 	return 0;
+out:
+	AFR_STACK_UNWIND (access, frame, -1, op_errno, NULL);
+
+        return 0;
 }
 
+/* }}} */
 
-int32_t
-afr_access (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, int32_t mask)
+/* {{{ stat */
+
+int
+afr_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)
 {
-	afr_private_t * priv       = NULL;
-	xlator_t **     children   = NULL;
-	int             call_child = 0;
-	afr_local_t     *local     = NULL;
+        afr_local_t     *local = NULL;
 
-        int32_t         read_child = -1;
+        local = frame->local;
 
+	if (op_ret < 0) {
+		local->op_ret = op_ret;
+		local->op_errno = op_errno;
 
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
+		afr_read_txn_continue (frame, this, (long) cookie);
+		return 0;
+	}
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+	AFR_STACK_UNWIND (stat, frame, op_ret, op_errno, buf, xdata);
 
-	priv     = this->private;
-	VALIDATE_OR_GOTO (priv->children, out);
+        return 0;
+}
 
-	children = priv->children;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+int
+afr_stat_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+	afr_private_t *priv = NULL;
+	afr_local_t *local = NULL;
 
-        read_child = afr_read_child (this, loc->inode);
+	priv = this->private;
+	local = frame->local;
 
-        if ((read_child >= 0) && (priv->child_up[read_child])) {
-                call_child = read_child;
+	if (subvol == -1) {
+		AFR_STACK_UNWIND (stat, frame, local->op_ret, local->op_errno,
+				  0, 0);
+		return 0;
+	}
 
-                local->cont.access.last_tried = -1;
+        STACK_WIND_COOKIE (frame, afr_stat_cbk, (void *) (long) subvol,
+                           priv->children[subvol],
+                           priv->children[subvol]->fops->stat,
+                           &local->loc, local->xdata_req);
+	return 0;
+}
 
-        } else {
-                call_child = afr_first_up_child (priv);
-                if (call_child == -1) {
-                        op_errno = ENOTCONN;
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "no child is up");
-                        goto out;
-                }
+int
+afr_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+        afr_local_t *local = NULL;
+	int op_errno = 0;
 
-                local->cont.access.last_tried = call_child;
-        }
+        local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
+	local->op = GF_FOP_STAT;
 	loc_copy (&local->loc, loc);
-	local->cont.access.mask       = mask;
+	if (xdata)
+		local->xdata_req = dict_ref (xdata);
 
-	STACK_WIND_COOKIE (frame, afr_access_cbk,
-			   (void *) (long) call_child,
-			   children[call_child], children[call_child]->fops->access,
-			   loc, mask);
+	afr_read_txn (frame, this, loc->inode, afr_stat_wind,
+		      AFR_DATA_TRANSACTION);
 
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (access, frame, op_ret, op_errno);
-	}
 	return 0;
+out:
+	AFR_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
 }
 
 
 /* }}} */
 
-/* {{{ stat */
+/* {{{ fstat */
 
-int32_t
-afr_stat_cbk (call_frame_t *frame, void *cookie,
-	      xlator_t *this, int32_t op_ret, int32_t op_errno,
-	      struct iatt *buf)
+int
+afr_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno, struct iatt *buf,
+               dict_t *xdata)
 {
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-	xlator_t **     children = NULL;
-
-	int unwind     = 1;
-	int last_tried = -1;
-	int this_try = -1;
-	int read_child = -1;
+        afr_local_t     *local = NULL;
 
-	priv     = this->private;
-	children = priv->children;
+        local = frame->local;
 
-	read_child = (long) cookie;
-
-	local = frame->local;
+	if (op_ret < 0) {
+		local->op_ret = op_ret;
+		local->op_errno = op_errno;
 
-	if (op_ret == -1) {
-	retry:
-		last_tried = local->cont.stat.last_tried;
+		afr_read_txn_continue (frame, this, (long) cookie);
+		return 0;
+	}
 
-		if (all_tried (last_tried, priv->child_count)) {
-			goto out;
-		}
-		this_try = ++local->cont.stat.last_tried;
+	AFR_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf, xdata);
 
-		if (this_try == read_child) {
-			goto retry;
-		}
+        return 0;
+}
 
-		unwind = 0;
 
-		STACK_WIND_COOKIE (frame, afr_stat_cbk,
-				   (void *) (long) read_child,
-				   children[this_try],
-				   children[this_try]->fops->stat,
-				   &local->loc);
-	}
+int
+afr_fstat_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+	afr_private_t *priv = NULL;
+	afr_local_t *local = NULL;
 
-out:
-	if (unwind) {
-		if (buf)
-			buf->ia_ino = local->cont.stat.ino;
+	priv = this->private;
+	local = frame->local;
 
-		AFR_STACK_UNWIND (stat, frame, op_ret, op_errno, buf);
+	if (subvol == -1) {
+		AFR_STACK_UNWIND (fstat, frame, local->op_ret, local->op_errno,
+				  0, 0);
+		return 0;
 	}
 
+        STACK_WIND_COOKIE (frame, afr_fstat_cbk, (void *) (long) subvol,
+                           priv->children[subvol],
+                           priv->children[subvol]->fops->fstat,
+                           local->fd, local->xdata_req);
 	return 0;
 }
 
 
 int32_t
-afr_stat (call_frame_t *frame, xlator_t *this,
-	  loc_t *loc)
+afr_fstat (call_frame_t *frame, xlator_t *this,
+           fd_t *fd, dict_t *xdata)
 {
-	afr_private_t * priv       = NULL;
-	afr_local_t   * local      = NULL;
-	xlator_t **     children   = NULL;
+        afr_local_t *local = NULL;
+	int op_errno = 0;
 
-        int32_t         read_child = -1;
-	int             call_child = 0;
+        local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-	int32_t         op_ret     = -1;
-	int32_t         op_errno   = 0;
+	local->op = GF_FOP_FSTAT;
+	local->fd = fd_ref (fd);
+	if (xdata)
+		local->xdata_req = dict_ref (xdata);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        afr_fix_open (fd, this);
 
-	priv     = this->private;
-	VALIDATE_OR_GOTO (priv->children, out);
+	afr_read_txn (frame, this, fd->inode, afr_fstat_wind,
+		      AFR_DATA_TRANSACTION);
 
-	children = priv->children;
+	return 0;
+out:
+	AFR_STACK_UNWIND (fstat, frame, -1, op_errno, NULL, NULL);
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        return 0;
+}
 
-	frame->local = local;
+/* }}} */
 
-        read_child = afr_read_child (this, loc->inode);
+/* {{{ readlink */
 
-        if ((read_child >= 0) && (priv->child_up[read_child])) {
-                call_child = read_child;
+int
+afr_readlink_cbk (call_frame_t *frame, void *cookie,
+                  xlator_t *this, int32_t op_ret, int32_t op_errno,
+                  const char *buf, struct iatt *sbuf, dict_t *xdata)
+{
+	afr_local_t *local = NULL;
 
-                local->cont.stat.last_tried = -1;
+	local = frame->local;
 
-        } else {
-		call_child = afr_first_up_child (priv);
-		if (call_child == -1) {
-			op_errno = ENOTCONN;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"no child is up");
-			goto out;
-		}
+	if (op_ret < 0) {
+		local->op_ret = -1;
+		local->op_errno = op_errno;
 
-		local->cont.stat.last_tried = call_child;
+		afr_read_txn_continue (frame, this, (long) cookie);
+		return 0;
 	}
 
-	loc_copy (&local->loc, loc);
+	AFR_STACK_UNWIND (readlink, frame, op_ret, op_errno,
+			  buf, sbuf, xdata);
+	return 0;
+}
 
-	local->cont.stat.ino = loc->inode->ino;
+int
+afr_readlink_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
 
-	STACK_WIND_COOKIE (frame, afr_stat_cbk, (void *) (long) call_child,
-			   children[call_child],
-			   children[call_child]->fops->stat,
-			   loc);
+	local = frame->local;
+	priv = this->private;
 
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (stat, frame, op_ret, op_errno, NULL);
+	if (subvol == -1) {
+		AFR_STACK_UNWIND (readlink, frame, local->op_ret,
+				  local->op_errno, 0, 0, 0);
+		return 0;
 	}
 
+	STACK_WIND_COOKIE (frame, afr_readlink_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->readlink,
+			   &local->loc, local->cont.readlink.size,
+			   local->xdata_req);
 	return 0;
 }
 
 
-/* }}} */
+int
+afr_readlink (call_frame_t *frame, xlator_t *this,
+              loc_t *loc, size_t size, dict_t *xdata)
+{
+        afr_local_t   * local      = NULL;
+        int32_t         op_errno   = 0;
 
-/* {{{ fstat */
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-int32_t
-afr_fstat_cbk (call_frame_t *frame, void *cookie,
-	       xlator_t *this, int32_t op_ret, int32_t op_errno,
-	       struct iatt *buf)
-{
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-	xlator_t **     children = NULL;
+	local->op = GF_FOP_READLINK;
+        loc_copy (&local->loc, loc);
+        local->cont.readlink.size = size;
+	if (xdata)
+		local->xdata_req = dict_ref (xdata);
 
-	int unwind     = 1;
-	int last_tried = -1;
-	int this_try = -1;
-        int read_child = -1;
+	afr_read_txn (frame, this, loc->inode, afr_readlink_wind,
+		      AFR_DATA_TRANSACTION);
 
-	priv     = this->private;
-	children = priv->children;
+        return 0;
+out:
+	AFR_STACK_UNWIND(readlink, frame, -1, op_errno, 0, 0, 0);
 
-	local = frame->local;
+	return 0;
+}
 
-	read_child = (long) cookie;
 
-	if (op_ret == -1) {
-	retry:
-		last_tried = local->cont.fstat.last_tried;
+/* }}} */
 
-		if (all_tried (last_tried, priv->child_count)) {
-			goto out;
-		}
-		this_try   = ++local->cont.fstat.last_tried;
+/* {{{ getxattr */
 
-		if (this_try == read_child) {
-			goto retry;
-		}
+struct _xattr_key {
+        char *key;
+        struct list_head list;
+};
 
-		unwind = 0;
 
-		STACK_WIND_COOKIE (frame, afr_fstat_cbk,
-				   (void *) (long) read_child,
-				   children[this_try],
-				   children[this_try]->fops->fstat,
-				   local->fd);
-	}
+int
+__gather_xattr_keys (dict_t *dict, char *key, data_t *value,
+                     void *data)
+{
+        struct list_head *  list  = data;
+        struct _xattr_key * xkey  = NULL;
 
-out:
-	if (unwind) {
-		if (buf)
-			buf->ia_ino = local->cont.fstat.ino;
+        if (!strncmp (key, AFR_XATTR_PREFIX,
+                      strlen (AFR_XATTR_PREFIX))) {
 
-		AFR_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf);
-	}
+                xkey = GF_CALLOC (1, sizeof (*xkey), gf_afr_mt_xattr_key);
+                if (!xkey)
+                        return -1;
 
-	return 0;
+                xkey->key = key;
+                INIT_LIST_HEAD (&xkey->list);
+
+                list_add_tail (&xkey->list, list);
+        }
+        return 0;
 }
 
 
-int32_t
-afr_fstat (call_frame_t *frame, xlator_t *this,
-	   fd_t *fd)
+void
+afr_filter_xattrs (dict_t *dict)
 {
-	afr_private_t * priv       = NULL;
-	afr_local_t   * local      = NULL;
-	xlator_t **     children   = NULL;
+        struct list_head   keys = {0,};
+        struct _xattr_key *key  = NULL;
+        struct _xattr_key *tmp  = NULL;
 
-	int             call_child = 0;
-        int32_t         read_child = -1;
+        INIT_LIST_HEAD (&keys);
 
-	int32_t         op_ret     = -1;
-	int32_t         op_errno   = 0;
+        dict_foreach (dict, __gather_xattr_keys,
+                      (void *) &keys);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (fd, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        list_for_each_entry_safe (key, tmp, &keys, list) {
+                dict_del (dict, key->key);
 
-	priv     = this->private;
-	VALIDATE_OR_GOTO (priv->children, out);
+                list_del_init (&key->list);
 
-	children = priv->children;
+                GF_FREE (key);
+        }
+}
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
 
-	frame->local = local;
+int
+afr_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)
+{
+        afr_local_t *local = NULL;
 
-	VALIDATE_OR_GOTO (fd->inode, out);
+        local = frame->local;
 
-        read_child = afr_read_child (this, fd->inode);
+	if (op_ret < 0) {
+		local->op_ret = op_ret;
+		local->op_errno = op_errno;
 
-        if ((read_child >= 0) && (priv->child_up[read_child])) {
-                call_child = read_child;
+		afr_read_txn_continue (frame, this, (long) cookie);
+		return 0;
+	}
 
-                local->cont.fstat.last_tried = -1;
-        } else {
-                call_child = afr_first_up_child (priv);
+	if (dict)
+		afr_filter_xattrs (dict);
 
-                if (call_child == -1) {
-                        op_errno = ENOTCONN;
-                        gf_log (this->name, GF_LOG_DEBUG,
-				"no child is up");
-			goto out;
-		}
+	AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
 
-                local->cont.fstat.last_tried = call_child;
-        }
+        return 0;
+}
 
-	local->cont.fstat.ino = fd->inode->ino;
-	local->fd = fd_ref (fd);
 
-	STACK_WIND_COOKIE (frame, afr_fstat_cbk, (void *) (long) call_child,
-			   children[call_child],
-			   children[call_child]->fops->fstat,
-			   fd);
+int
+afr_getxattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
 
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (fstat, frame, op_ret, op_errno, NULL);
+	local = frame->local;
+	priv = this->private;
+
+	if (subvol == -1) {
+                AFR_STACK_UNWIND (getxattr, frame, local->op_ret,
+				  local->op_errno, NULL, NULL);
+		return 0;
 	}
 
+	STACK_WIND_COOKIE (frame, afr_getxattr_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->getxattr,
+			   &local->loc, local->cont.getxattr.name,
+			   local->xdata_req);
 	return 0;
 }
 
-/* }}} */
 
-/* {{{ readlink */
+int32_t
+afr_getxattr_unwind (call_frame_t *frame, int op_ret, int op_errno,
+                     dict_t *dict, dict_t *xdata)
+
+{
+        AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+}
 
 int32_t
-afr_readlink_cbk (call_frame_t *frame, void *cookie,
-		  xlator_t *this, int32_t op_ret, int32_t op_errno,
-		  const char *buf, struct iatt *sbuf)
+afr_fgetxattr_clrlk_cbk (call_frame_t *frame, void *cookie,
+                         xlator_t *this, int32_t op_ret, int32_t op_errno,
+                         dict_t *dict, dict_t *xdata)
 {
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-	xlator_t **     children = NULL;
+        afr_local_t     *local                  = NULL;
+        afr_private_t   *priv                   = NULL;
+        xlator_t        **children              = NULL;
+        dict_t          *xattr                  = NULL;
+        char            *tmp_report             = NULL;
+        char            lk_summary[1024]        = {0,};
+        int             serz_len                = 0;
+        int32_t         callcnt                 = 0;
+        long int        cky                     = 0;
+        int             ret                     = 0;
+
+        priv     = this->private;
+        children = priv->children;
+
+        local = frame->local;
+        cky = (long) cookie;
+
+        LOCK (&frame->lock);
+        {
+                callcnt = --local->call_count;
+                if (op_ret == -1)
+                        local->replies[cky].op_errno = 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;
+                }
 
-	int unwind     = 1;
-	int last_tried = -1;
-	int this_try = -1;
-        int read_child = -1;
+        unwind:
+                // Updating child_errno with more recent 'events'
+                op_errno = afr_final_errno (local, priv);
 
-	priv     = this->private;
-	children = priv->children;
+                AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, xattr,
+                                  xdata);
+                if (xattr)
+                        dict_unref (xattr);
+        }
 
-	local = frame->local;
+        return ret;
+}
 
-        read_child = (long) cookie;
+int32_t
+afr_getxattr_clrlk_cbk (call_frame_t *frame, void *cookie,
+                        xlator_t *this, int32_t op_ret, int32_t op_errno,
+                        dict_t *dict, dict_t *xdata)
+{
+        afr_local_t     *local                  = NULL;
+        afr_private_t   *priv                   = NULL;
+        xlator_t        **children              = NULL;
+        dict_t          *xattr                  = NULL;
+        char            *tmp_report             = NULL;
+        char            lk_summary[1024]        = {0,};
+        int             serz_len                = 0;
+        int32_t         callcnt                 = 0;
+        long int        cky                     = 0;
+        int             ret                     = 0;
+
+        priv     = this->private;
+        children = priv->children;
+
+        local = frame->local;
+        cky = (long) cookie;
+
+        LOCK (&frame->lock);
+        {
+                callcnt = --local->call_count;
+                if (op_ret == -1)
+                        local->replies[cky].op_errno = 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;
+                }
 
-	if (op_ret == -1) {
-        retry:
-		last_tried = local->cont.readlink.last_tried;
+        unwind:
+                // Updating child_errno with more recent 'events'
+                op_errno = afr_final_errno (local, priv);
 
-		if (all_tried (last_tried, priv->child_count)) {
-			goto out;
-		}
-		this_try = ++local->cont.readlink.last_tried;
+                AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr, xdata);
 
-                if (this_try == read_child) {
-                        goto retry;
-                }
+                if (xattr)
+                        dict_unref (xattr);
+        }
 
-		unwind = 0;
-		STACK_WIND_COOKIE (frame, afr_readlink_cbk,
-				   (void *) (long) read_child,
-				   children[this_try],
-				   children[this_try]->fops->readlink,
-				   &local->loc,
-				   local->cont.readlink.size);
-	}
+        return ret;
+}
 
-out:
-	if (unwind) {
-                if (sbuf)
-                        sbuf->ia_ino = local->cont.readlink.ino;
+/**
+ * node-uuid cbk uses next child querying mechanism
+ */
+int32_t
+afr_getxattr_node_uuid_cbk (call_frame_t *frame, void *cookie,
+                            xlator_t *this, int32_t op_ret, int32_t op_errno,
+                            dict_t *dict, dict_t *xdata)
+{
+        afr_private_t  *priv            = NULL;
+        afr_local_t    *local           = NULL;
+        xlator_t      **children        = NULL;
+        int             unwind          = 1;
+        int             curr_call_child = 0;
+
+        priv = this->private;
+        children = priv->children;
+
+        local = frame->local;
+
+        if (op_ret == -1) { /** query the _next_ child */
+
+                /**
+                 * _current_ becomes _next_
+                 * If done with all childs and yet no success; give up !
+                 */
+                curr_call_child = (int) ((long)cookie);
+                if (++curr_call_child == priv->child_count)
+                        goto unwind;
+
+                gf_log (this->name, GF_LOG_WARNING,
+                        "op_ret (-1): Re-querying afr-child (%d/%d)",
+                        curr_call_child, priv->child_count);
+
+                unwind = 0;
+                STACK_WIND_COOKIE (frame, afr_getxattr_node_uuid_cbk,
+                                   (void *) (long) curr_call_child,
+                                   children[curr_call_child],
+                                   children[curr_call_child]->fops->getxattr,
+                                   &local->loc,
+                                   local->cont.getxattr.name,
+                                   NULL);
+        }
 
-		AFR_STACK_UNWIND (readlink, frame, op_ret, op_errno, buf, sbuf);
-	}
+ unwind:
+        if (unwind)
+                AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict,
+                                  NULL);
 
-	return 0;
+        return 0;
 }
 
-
 int32_t
-afr_readlink (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, size_t size)
+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)
 {
-	afr_private_t * priv       = NULL;
-	xlator_t **     children   = NULL;
-	int             call_child = 0;
-	afr_local_t     *local     = NULL;
+        int          call_cnt     = 0, len = 0;
+        char        *lockinfo_buf = NULL;
+        dict_t      *lockinfo     = NULL, *newdict = NULL;
+        afr_local_t *local        = NULL;
 
-        int32_t         read_child = -1;
+        LOCK (&frame->lock);
+        {
+                local = frame->local;
 
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
+                call_cnt = --local->call_count;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+                if ((op_ret < 0) || (!dict && !xdata)) {
+                        goto unlock;
+                }
 
-	priv     = this->private;
-	VALIDATE_OR_GOTO (priv->children, out);
+                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;
+                                }
+                        }
+                }
 
-	children = priv->children;
+                if (!dict) {
+                        goto unlock;
+                }
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+                op_ret = dict_get_ptr_and_len (dict, GF_XATTR_LOCKINFO_KEY,
+                                               (void **)&lockinfo_buf, &len);
 
-	frame->local = local;
+                if (!lockinfo_buf) {
+                        goto unlock;
+                }
 
-        read_child = afr_read_child (this, loc->inode);
+                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 ((read_child >= 0) && (priv->child_up[read_child])) {
-                call_child = read_child;
+        if (xdata && local->xdata_rsp) {
+                dict_copy (xdata, local->xdata_rsp);
+        }
 
-                local->cont.readlink.last_tried = -1;
+        if (!call_cnt) {
+                newdict = dict_new ();
+                if (!newdict) {
+                        local->op_ret = -1;
+                        local->op_errno = ENOMEM;
+                        goto unwind;
+                }
 
-        } else {
-                call_child = afr_first_up_child (priv);
+                len = dict_serialized_length (local->dict);
+                if (len == 0) {
+                        goto unwind;
+                }
 
-                if (call_child == -1) {
-                        op_errno = ENOTCONN;
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "no child is up");
-                        goto out;
+                lockinfo_buf = GF_CALLOC (1, len, gf_common_mt_char);
+                if (!lockinfo_buf) {
+                        local->op_ret = -1;
+                        local->op_errno = ENOMEM;
+                        goto unwind;
                 }
 
-                local->cont.readlink.last_tried = call_child;
-        }
+                op_ret = dict_serialize (local->dict, lockinfo_buf);
+                if (op_ret < 0) {
+                        local->op_ret = -1;
+                        local->op_errno = -op_ret;
+                }
 
-	loc_copy (&local->loc, loc);
+                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;
+                }
 
-	local->cont.readlink.size       = size;
-        local->cont.readlink.ino        = loc->inode->ino;
+        unwind:
+                AFR_STACK_UNWIND (getxattr, frame, op_ret,
+                                  op_errno, newdict,
+                                  local->xdata_rsp);
+        }
 
-	STACK_WIND_COOKIE (frame, afr_readlink_cbk,
-			   (void *) (long) call_child,
-			   children[call_child], children[call_child]->fops->readlink,
-			   loc, size);
+        dict_unref (lockinfo);
 
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (readlink, frame, op_ret, op_errno, NULL, NULL);
-	}
-	return 0;
+        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;
 
-/* {{{ getxattr */
+                call_cnt = --local->call_count;
 
-struct _xattr_key {
-        char *key;
-        struct list_head list;
-};
+                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;
+                                }
+                        }
+                }
 
-void
-__gather_xattr_keys (dict_t *dict, char *key, data_t *value,
-                     void *data)
-{
-        struct list_head *  list  = data;
-        struct _xattr_key * xkey  = NULL;
+                if (!dict) {
+                        goto unlock;
+                }
 
-        if (!strncmp (key, AFR_XATTR_PREFIX,
-                      strlen (AFR_XATTR_PREFIX))) {
+                op_ret = dict_get_ptr_and_len (dict, GF_XATTR_LOCKINFO_KEY,
+                                               (void **)&lockinfo_buf, &len);
 
-                xkey = GF_CALLOC (1, sizeof (*xkey), gf_afr_mt_xattr_key);
-                if (!xkey)
-                        return;
+                if (!lockinfo_buf) {
+                        goto unlock;
+                }
 
-                xkey->key = key;
-                INIT_LIST_HEAD (&xkey->list);
+                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);
+                        }
+                }
+        }
 
-                list_add_tail (&xkey->list, list);
+        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;
+                }
 
-void
-__filter_xattrs (dict_t *dict)
+                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)
 {
-        struct list_head keys;
+        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;
+        }
 
-        struct _xattr_key *key;
-        struct _xattr_key *tmp;
+        local = frame->local;
+        cky = (long) cookie;
 
-        INIT_LIST_HEAD (&keys);
+        LOCK (&frame->lock);
+        {
+                callcnt = --local->call_count;
 
-        dict_foreach (dict, __gather_xattr_keys,
-                      (void *) &keys);
+                if (op_ret < 0) {
+                        local->op_errno = op_errno;
+                } else {
+                        local->op_ret = op_ret;
+                        if (!local->xdata_rsp && xdata)
+                                local->xdata_rsp = dict_ref (xdata);
+                }
 
-        list_for_each_entry_safe (key, tmp, &keys, list) {
-                dict_del (dict, key->key);
+                if (!dict || (op_ret < 0))
+                        goto out;
 
-                list_del_init (&key->list);
+                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;
+                }
 
-                GF_FREE (key);
+                /* 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, local->op_ret,
+                                  local->op_errno, nxattr, local->xdata_rsp);
+
+                if (nxattr)
+                        dict_unref (nxattr);
         }
+
+        return ret;
 }
 
+int32_t
+afr_getxattr_pathinfo_cbk (call_frame_t *frame, void *cookie,
+                           xlator_t *this, int32_t op_ret, int32_t op_errno,
+                           dict_t *dict, dict_t *xdata)
+{
+        afr_local_t *local          = NULL;
+        int32_t      callcnt        = 0;
+        int          ret            = 0;
+        char        *xattr          = NULL;
+        char        *xattr_serz     = NULL;
+        char        xattr_cky[1024] = {0,};
+        dict_t      *nxattr         = NULL;
+        long         cky            = 0;
+        int32_t      padding        = 0;
+        int32_t      tlen           = 0;
+
+        if (!frame || !frame->local || !this) {
+                gf_log ("", GF_LOG_ERROR, "possible NULL deref");
+                goto out;
+        }
 
+        local = frame->local;
+        cky = (long) cookie;
+
+        LOCK (&frame->lock);
+                {
+                        callcnt = --local->call_count;
+
+                        if (op_ret < 0) {
+                                local->op_errno = op_errno;
+                        } else {
+                                local->op_ret = op_ret;
+                                if (!local->xdata_rsp && xdata)
+                                        local->xdata_rsp = dict_ref (xdata);
+                        }
+
+                        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;
+                }
 
-int32_t
-afr_getxattr_cbk (call_frame_t *frame, void *cookie,
-		  xlator_t *this, int32_t op_ret, int32_t op_errno,
-		  dict_t *dict)
+                /* closing part */
+                *(xattr_serz + padding + tlen) = ')';
+                *(xattr_serz + padding + tlen + 1) = '\0';
+
+                ret = dict_set_dynstr (nxattr, local->cont.getxattr.name,
+                                       xattr_serz);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR, "Cannot set pathinfo"
+                                " key in dict");
+
+        unwind:
+                AFR_STACK_UNWIND (getxattr, frame, local->op_ret,
+                                  local->op_errno, nxattr, local->xdata_rsp);
+
+                if (nxattr)
+                        dict_unref (nxattr);
+        }
+
+        return ret;
+}
+
+static int
+afr_aggregate_stime_xattr (dict_t *this, char *key, data_t *value, void *data)
 {
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-	xlator_t **     children = NULL;
+        int ret = 0;
 
-	int unwind     = 1;
-	int last_tried = -1;
-	int this_try = -1;
-        int read_child = -1;
+        if (fnmatch (GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0)
+                ret = gf_get_max_stime (THIS, data, key, value);
 
-	priv     = this->private;
-	children = priv->children;
+        return ret;
+}
 
-	local = frame->local;
+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;
 
-        read_child = (long) cookie;
+        if (!frame || !frame->local || !this) {
+                gf_log ("", GF_LOG_ERROR, "possible NULL deref");
+                goto out;
+        }
 
-	if (op_ret == -1) {
-        retry:
-		last_tried = local->cont.getxattr.last_tried;
+        local = frame->local;
 
-		if (all_tried (last_tried, priv->child_count)) {
-			goto out;
-		}
-		this_try = ++local->cont.getxattr.last_tried;
+        LOCK (&frame->lock);
+        {
+                callcnt = --local->call_count;
 
-                if (this_try == read_child) {
-                        goto retry;
+                if (!dict || (op_ret < 0)) {
+                        local->op_errno = op_errno;
+                        goto cleanup;
                 }
 
-		unwind = 0;
-		STACK_WIND_COOKIE (frame, afr_getxattr_cbk,
-				   (void *) (long) read_child,
-				   children[this_try],
-				   children[this_try]->fops->getxattr,
-				   &local->loc,
-				   local->cont.getxattr.name);
-	}
+                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:
-	if (unwind) {
-                if (op_ret >= 0 && dict)
-                        __filter_xattrs (dict);
+        return 0;
+}
 
-		AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
-	}
 
-	return 0;
+static gf_boolean_t
+afr_is_special_xattr (const char *name, fop_getxattr_cbk_t *cbk,
+                      gf_boolean_t is_fgetxattr)
+{
+        gf_boolean_t    is_spl = _gf_true;
+
+        GF_ASSERT (cbk);
+        if (!cbk || !name) {
+                is_spl = _gf_false;
+                goto out;
+        }
+
+        if (!strcmp (name, GF_XATTR_PATHINFO_KEY) ||
+	    !strcmp (name, GF_XATTR_USER_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;
+}
+
+static void
+afr_getxattr_all_subvols (xlator_t *this, call_frame_t *frame,
+			  const char *name, loc_t *loc,
+			  fop_getxattr_cbk_t cbk)
+{
+        afr_private_t   *priv           = NULL;
+        afr_local_t     *local          = NULL;
+        int             i               = 0;
+        int             call_count      = 0;
+
+        priv     = this->private;
+
+        local = frame->local;
+        //local->call_count set in afr_local_init
+        call_count = local->call_count;
+
+        //If up-children count is 0, afr_local_init would have failed already
+        //and the call would have unwound so not handling it here.
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND_COOKIE (frame, cbk,
+                                           (void *) (long) i, priv->children[i],
+                                           priv->children[i]->fops->getxattr,
+                                           loc, name, NULL);
+                        if (!--call_count)
+                                break;
+                }
+        }
+        return;
 }
 
 int32_t
 afr_getxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, const char *name)
+              loc_t *loc, const char *name, dict_t *xdata)
 {
-	afr_private_t *   priv       = NULL;
-	xlator_t **       children   = NULL;
-	int               call_child = 0;
-	afr_local_t     * local      = NULL;
+        afr_private_t           *priv         = NULL;
+        xlator_t                **children    = NULL;
+        afr_local_t             *local        = NULL;
+        xlator_list_t           *trav         = NULL;
+        xlator_t                **sub_volumes = NULL;
+        int                     i             = 0;
+        int32_t                 op_errno      = 0;
+        int                     ret           = -1;
+        fop_getxattr_cbk_t      cbk           = NULL;
+        int                     afr_xtime_gauge[MCNT_MAX] = {0,};
 
-        int               read_child = -1;
 
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
+        priv     = this->private;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        children = priv->children;
 
-	priv     = this->private;
-	VALIDATE_OR_GOTO (priv->children, out);
+        loc_copy (&local->loc, loc);
 
-	children = priv->children;
+	local->op = GF_FOP_GETXATTR;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-	frame->local = local;
+	if (xdata)
+		local->xdata_req = dict_ref (xdata);
 
-        if (name) {
-                if (!strncmp (name, AFR_XATTR_PREFIX,
-                              strlen (AFR_XATTR_PREFIX))) {
+        if (!name)
+                goto no_name;
 
-                        op_errno = ENODATA;
-                        goto out;
-                }
+        local->cont.getxattr.name = gf_strdup (name);
+
+	if (!local->cont.getxattr.name) {
+		op_errno = ENOMEM;
+		goto out;
+	}
 
+        if (!strncmp (name, AFR_XATTR_PREFIX,
+                      strlen (AFR_XATTR_PREFIX))) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "%s: no data present for key %s",
+                        loc->path, name);
+                op_errno = ENODATA;
+                goto out;
         }
+        if ((strcmp (GF_XATTR_MARKER_KEY, name) == 0)
+            && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
 
-        read_child = afr_read_child (this, loc->inode);
+                local->marker.call_count = priv->child_count;
 
-        if ((read_child >= 0) && (priv->child_up[read_child])) {
-                call_child = read_child;
+                sub_volumes = alloca ( priv->child_count * sizeof (xlator_t *));
+                for (i = 0, trav = this->children; trav ;
+                     trav = trav->next, i++) {
 
-                local->cont.getxattr.last_tried = -1;
-        } else {
-                call_child = afr_first_up_child (priv);
+                        *(sub_volumes + i)  = trav->xlator;
+                }
 
-                if (call_child == -1) {
-                        op_errno = ENOTCONN;
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "no child is up");
+                if (cluster_getmarkerattr (frame, this, loc, name,
+                                           local, afr_getxattr_unwind,
+                                           sub_volumes,
+                                           priv->child_count,
+                                           MARKER_UUID_TYPE,
+                                           marker_uuid_default_gauge,
+                                           priv->vol_uuid)) {
+
+                        gf_log (this->name, GF_LOG_INFO,
+                                "%s: failed to get marker attr (%s)",
+                                loc->path, name);
+                        op_errno = EINVAL;
                         goto out;
                 }
 
-                local->cont.getxattr.last_tried = call_child;
+                return 0;
         }
 
-	loc_copy (&local->loc, loc);
-	if (name)
-	  local->cont.getxattr.name       = gf_strdup (name);
+        /*
+         * if we are doing getxattr with pathinfo as the key then we
+         * collect information from all childs
+         */
+        if (afr_is_special_xattr (name, &cbk, 0)) {
+                afr_getxattr_all_subvols (this, frame, name, loc, cbk);
+                return 0;
+        }
 
-	STACK_WIND_COOKIE (frame, afr_getxattr_cbk,
-			   (void *) (long) call_child,
-			   children[call_child], children[call_child]->fops->getxattr,
-			   loc, name);
+        if (XATTR_IS_NODE_UUID (name)) {
+                i = 0;
+                STACK_WIND_COOKIE (frame, afr_getxattr_node_uuid_cbk,
+                                   (void *) (long) i,
+                                   children[i],
+                                   children[i]->fops->getxattr,
+                                   loc, name, xdata);
+                return 0;
+        }
 
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, NULL);
-	}
-	return 0;
-}
+        if (*priv->vol_uuid) {
+                if ((match_uuid_local (name, priv->vol_uuid) == 0)
+                    && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+                        local->marker.call_count = priv->child_count;
+
+                        sub_volumes = alloca ( priv->child_count
+                                               * sizeof (xlator_t *));
+                        for (i = 0, trav = this->children; trav ;
+                             trav = trav->next, i++) {
+
+                                *(sub_volumes + i)  = trav->xlator;
+
+                        }
+
+                        /* don't err out on getting ENOTCONN (brick down)
+                         * from a subset of the bricks
+                         */
+                        memcpy (afr_xtime_gauge, marker_xtime_default_gauge,
+                                sizeof (afr_xtime_gauge));
+                        afr_xtime_gauge[MCNT_NOTFOUND] = 0;
+                        afr_xtime_gauge[MCNT_ENOTCONN] = 0;
+                        if (cluster_getmarkerattr (frame, this, loc,
+                                                   name, local,
+                                                   afr_getxattr_unwind,
+                                                   sub_volumes,
+                                                   priv->child_count,
+                                                   MARKER_XTIME_TYPE,
+                                                   afr_xtime_gauge,
+                                                   priv->vol_uuid)) {
+                                gf_log (this->name, GF_LOG_INFO,
+                                        "%s: failed to get marker attr (%s)",
+                                        loc->path, name);
+                                op_errno = EINVAL;
+                                goto out;
+                        }
+
+                        return 0;
+                }
+        }
 
+no_name:
 
-/* }}} */
+	afr_read_txn (frame, this, local->loc.inode, afr_getxattr_wind,
+		      AFR_METADATA_TRANSACTION);
 
-/* {{{ readv */
+        ret = 0;
+out:
+        if (ret < 0)
+                AFR_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
+        return 0;
+}
+
+/* {{{ fgetxattr */
 
-/**
- * read algorithm:
- *
- * if the user has specified a read subvolume, use it
- * otherwise -
- *   use the inode number to hash it to one of the subvolumes, and
- *   read from there (to balance read load)
- *
- * if any of the above read's fail, try the children in sequence
- * beginning at the beginning
- */
 
 int32_t
-afr_readv_cbk (call_frame_t *frame, void *cookie,
-	       xlator_t *this, int32_t op_ret, int32_t op_errno,
-	       struct iovec *vector, int32_t count, struct iatt *buf,
-               struct iobref *iobref)
+afr_fgetxattr_cbk (call_frame_t *frame, void *cookie,
+                   xlator_t *this, int32_t op_ret, int32_t op_errno,
+                   dict_t *dict, dict_t *xdata)
 {
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-	xlator_t **     children = NULL;
+	afr_local_t *local = NULL;
 
-	int unwind     = 1;
-	int last_tried = -1;
-	int this_try   = -1;
-        int read_child = -1;
+	local = frame->local;
+
+	if (op_ret < 0) {
+		local->op_ret = -1;
+		local->op_errno = op_errno;
+
+		afr_read_txn_continue (frame, this, (long) cookie);
+		return 0;
+	}
+
+	if (dict)
+		afr_filter_xattrs (dict);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+	AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, dict, xdata);
 
-	priv     = this->private;
-	VALIDATE_OR_GOTO (priv->children, out);
+	return 0;
+}
 
-	children = priv->children;
+int
+afr_fgetxattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
 
 	local = frame->local;
+	priv = this->private;
 
-        read_child = (long) cookie;
+	if (subvol == -1) {
+		AFR_STACK_UNWIND (fgetxattr, frame, local->op_ret,
+				  local->op_errno, NULL, NULL);
+		return 0;
+	}
 
-	if (op_ret == -1) {
-	retry:
-		last_tried = local->cont.readv.last_tried;
+	STACK_WIND_COOKIE (frame, (void *) (long) subvol, afr_fgetxattr_cbk,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->fgetxattr,
+			   local->fd, local->cont.getxattr.name,
+			   local->xdata_req);
+	return 0;
+}
 
-		if (all_tried (last_tried, priv->child_count)) {
-			goto out;
-		}
-		this_try = ++local->cont.readv.last_tried;
-
-		if (this_try == read_child) {
-			/*
-			   skip the read child since if we are here
-			   we must have already tried that child
-			*/
-			goto retry;
-		}
 
-		unwind = 0;
+static void
+afr_fgetxattr_all_subvols (xlator_t *this, call_frame_t *frame,
+			   fop_fgetxattr_cbk_t cbk)
+{
+        afr_private_t   *priv           = NULL;
+        afr_local_t     *local          = NULL;
+        int             i               = 0;
+        int             call_count      = 0;
+
+        priv     = this->private;
+
+        local = frame->local;
+        //local->call_count set in afr_local_init
+        call_count = local->call_count;
+
+        //If up-children count is 0, afr_local_init would have failed already
+        //and the call would have unwound so not handling it here.
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND_COOKIE (frame, cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->fgetxattr,
+                                           local->fd, local->cont.getxattr.name,
+					   NULL);
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-		STACK_WIND_COOKIE (frame, afr_readv_cbk,
-				   (void *) (long) read_child,
-				   children[this_try],
-				   children[this_try]->fops->readv,
-				   local->fd, local->cont.readv.size,
-				   local->cont.readv.offset);
-	}
+        return;
+}
 
-out:
-	if (unwind) {
-                if (buf && local)
-                        buf->ia_ino = local->cont.readv.ino;
 
-		AFR_STACK_UNWIND (readv, frame, op_ret, op_errno,
-                                  vector, count, buf, iobref);
+int
+afr_fgetxattr (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, const char *name, dict_t *xdata)
+{
+        afr_local_t          *local      = NULL;
+        int32_t               op_errno   = 0;
+        fop_fgetxattr_cbk_t   cbk        = NULL;
+
+        local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
+
+	local->op = GF_FOP_FGETXATTR;
+        local->fd = fd_ref (fd);
+        if (name) {
+                local->cont.getxattr.name = gf_strdup (name);
+		if (!local->cont.getxattr.name) {
+			op_errno = ENOMEM;
+			goto out;
+		}
 	}
+	if (xdata)
+		local->xdata_req = dict_ref (xdata);
+
+        /* 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_all_subvols (this, frame, cbk);
+                return 0;
+        }
+
+        afr_fix_open (fd, this);
+
+	afr_read_txn (frame, this, fd->inode, afr_fgetxattr_wind,
+		      AFR_METADATA_TRANSACTION);
 
 	return 0;
+out:
+	AFR_STACK_UNWIND (fgetxattr, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
 }
 
 
-int32_t
-afr_readv (call_frame_t *frame, xlator_t *this,
-	   fd_t *fd, size_t size, off_t offset)
+/* }}} */
+
+/* {{{ readv */
+
+int
+afr_readv_cbk (call_frame_t *frame, void *cookie,
+               xlator_t *this, int32_t op_ret, int32_t op_errno,
+               struct iovec *vector, int32_t count, struct iatt *buf,
+               struct iobref *iobref, dict_t *xdata)
 {
-	afr_private_t * priv       = NULL;
-	afr_local_t   * local      = NULL;
-	xlator_t **     children   = NULL;
+	afr_local_t *local = NULL;
 
-        int32_t         read_child = -1;
-	int             call_child = 0;
+	local = frame->local;
+
+	if (op_ret < 0) {
+		local->op_ret = -1;
+		local->op_errno = op_errno;
 
-	int32_t         op_ret     = -1;
-	int32_t         op_errno   = 0;
+		afr_read_txn_continue (frame, this, (long) cookie);
+		return 0;
+	}
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-	VALIDATE_OR_GOTO (fd, out);
+	AFR_STACK_UNWIND (readv, frame, op_ret, op_errno,
+			  vector, count, buf, iobref, xdata);
+	return 0;
+}
 
-	priv     = this->private;
-	children = priv->children;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+int
+afr_readv_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
 
-	frame->local = local;
+	local = frame->local;
+	priv = this->private;
 
-        read_child = afr_read_child (this, fd->inode);
+	if (subvol == -1) {
+		AFR_STACK_UNWIND (readv, frame, local->op_ret, local->op_errno,
+				  0, 0, 0, 0, 0);
+		return 0;
+	}
 
-        if ((read_child >= 0) && (priv->child_up[read_child])) {
-                call_child = read_child;
+	STACK_WIND_COOKIE (frame, afr_readv_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->readv,
+			   local->fd, local->cont.readv.size,
+			   local->cont.readv.offset, local->cont.readv.flags,
+			   local->xdata_req);
+	return 0;
+}
 
-		/*
-		   if read fails from the read child, we try
-		   all children starting with the first one
-		*/
-                local->cont.readv.last_tried = -1;
 
-        } else {
-		call_child = afr_first_up_child (priv);
-		if (call_child == -1) {
-			op_errno = ENOTCONN;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"no child is up");
-			goto out;
-		}
+int
+afr_readv (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+	   off_t offset, uint32_t flags, dict_t *xdata)
+{
+        afr_local_t   * local      = NULL;
+        int32_t         op_errno   = 0;
 
-		local->cont.readv.last_tried = call_child;
-	}
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-	local->fd                    = fd_ref (fd);
+	local->op = GF_FOP_READ;
+        local->fd = fd_ref (fd);
+        local->cont.readv.size = size;
+        local->cont.readv.offset = offset;
+        local->cont.readv.flags = flags;
+	if (xdata)
+		local->xdata_req = dict_ref (xdata);
 
-        local->cont.readv.ino        = fd->inode->ino;
-	local->cont.readv.size       = size;
-	local->cont.readv.offset     = offset;
+        afr_fix_open (fd, this);
 
-	STACK_WIND_COOKIE (frame, afr_readv_cbk,
-			   (void *) (long) call_child,
-			   children[call_child],
-			   children[call_child]->fops->readv,
-			   fd, size, offset);
+	afr_read_txn (frame, this, fd->inode, afr_readv_wind,
+		      AFR_DATA_TRANSACTION);
 
-	op_ret = 0;
+        return 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (readv, frame, op_ret, op_errno, NULL, 0, NULL,
-                                  NULL);
-	}
+	AFR_STACK_UNWIND(readv, frame, -1, op_errno, 0, 0, 0, 0, 0);
+
 	return 0;
 }
 
diff --git a/xlators/cluster/afr/src/afr-inode-read.h b/xlators/cluster/afr/src/afr-inode-read.h
index acc814fb7..e4091a793 100644
--- a/xlators/cluster/afr/src/afr-inode-read.h
+++ b/xlators/cluster/afr/src/afr-inode-read.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __INODE_READ_H__
@@ -22,26 +13,30 @@
 
 int32_t
 afr_access (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, int32_t mask);
+	    loc_t *loc, int32_t mask, dict_t *xdata);
 
 int32_t
 afr_stat (call_frame_t *frame, xlator_t *this,
-	  loc_t *loc);
+	  loc_t *loc, dict_t *xdata);
 
 int32_t
 afr_fstat (call_frame_t *frame, xlator_t *this,
-	   fd_t *fd);
+	   fd_t *fd, dict_t *xdata);
 
 int32_t
 afr_readlink (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, size_t size);
+	      loc_t *loc, size_t size, dict_t *xdata);
 
 int32_t
 afr_readv (call_frame_t *frame, xlator_t *this,
-	   fd_t *fd, size_t size, off_t offset);
+	   fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata);
 
 int32_t
 afr_getxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, const char *name);
+	      loc_t *loc, const char *name, dict_t *xdata);
+
+int32_t
+afr_fgetxattr (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, const char *name, dict_t *xdata);
 
 #endif /* __INODE_READ_H__ */
diff --git a/xlators/cluster/afr/src/afr-inode-write.c b/xlators/cluster/afr/src/afr-inode-write.c
index 050a4f0e9..3dacfc8dd 100644
--- a/xlators/cluster/afr/src/afr-inode-write.c
+++ b/xlators/cluster/afr/src/afr-inode-write.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -46,146 +37,303 @@
 
 #include "afr.h"
 #include "afr-transaction.h"
+//#include "afr-self-heal-common.h"
 
-/* {{{ writev */
 
-int
-afr_writev_unwind (call_frame_t *frame, xlator_t *this)
+static void
+__afr_inode_write_finalize (call_frame_t *frame, xlator_t *this)
 {
-        afr_local_t *   local = NULL;
-        call_frame_t   *main_frame = NULL;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	int read_subvol = 0;
+	int i = 0;
+
+	local = frame->local;
+	priv = this->private;
+
+	if (local->inode) {
+		if (local->transaction.type == AFR_METADATA_TRANSACTION)
+			read_subvol = afr_metadata_subvol_get (local->inode, this,
+							       NULL, NULL);
+		else
+			read_subvol = afr_data_subvol_get (local->inode, this,
+							   NULL, NULL);
+	}
+
+	local->op_ret = -1;
+	local->op_errno = afr_final_errno (local, priv);
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (!local->replies[i].valid)
+			continue;
+		if (local->replies[i].op_ret < 0) {
+			afr_inode_read_subvol_reset (local->inode, this);
+			continue;
+		}
+
+		/* Order of checks in the compound conditional
+		   below is important.
+
+		   - Highest precedence: largest op_ret
+		   - Next precendence: if all op_rets are equal, read subvol
+		   - Least precedence: any succeeded subvol
+		*/
+		if ((local->op_ret < local->replies[i].op_ret) ||
+		    ((local->op_ret == local->replies[i].op_ret) &&
+		     (i == read_subvol))) {
+
+			local->op_ret = local->replies[i].op_ret;
+			local->op_errno = local->replies[i].op_errno;
+
+			local->cont.inode_wfop.prebuf =
+				local->replies[i].prestat;
+			local->cont.inode_wfop.postbuf =
+				local->replies[i].poststat;
+
+			if (local->replies[i].xdata) {
+				if (local->xdata_rsp)
+					dict_unref (local->xdata_rsp);
+				local->xdata_rsp =
+					dict_ref (local->replies[i].xdata);
+			}
+		}
+	}
+}
 
-        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);
+static void
+__afr_inode_write_fill (call_frame_t *frame, xlator_t *this, int child_index,
+			int op_ret, int op_errno,
+			struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+{
+        afr_local_t *local = NULL;
 
-        if (main_frame) {
-                local->cont.writev.prebuf.ia_ino  = local->cont.writev.ino;
-                local->cont.writev.postbuf.ia_ino = local->cont.writev.ino;
+        local = frame->local;
 
-                AFR_STACK_UNWIND (writev, main_frame,
-                                  local->op_ret, local->op_errno,
-                                  &local->cont.writev.prebuf,
-                                  &local->cont.writev.postbuf);
-        }
-        return 0;
+	local->replies[child_index].valid = 1;
+	local->replies[child_index].op_ret = op_ret;
+	local->replies[child_index].op_errno = op_errno;
+
+	if (op_ret >= 0) {
+		if (prebuf)
+			local->replies[child_index].prestat = *prebuf;
+		if (postbuf)
+			local->replies[child_index].poststat = *postbuf;
+		if (xdata)
+			local->replies[child_index].xdata = dict_ref (xdata);
+	} else {
+		afr_transaction_fop_failed (frame, this, child_index);
+	}
+
+        return;
 }
 
 
-int
-afr_writev_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                     struct iatt *postbuf)
+static int
+__afr_inode_write_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_local_t *local = NULL;
         int child_index = (long) cookie;
-        int call_count  = -1;
-        int read_child  = 0;
+        int call_count = -1;
 
         local = frame->local;
 
-        read_child = afr_read_child (this, local->fd->inode);
-
         LOCK (&frame->lock);
         {
-                if (child_index == read_child) {
-                        local->read_child_returned = _gf_true;
-                }
-
-                if (afr_fop_failed (op_ret, op_errno))
-                        afr_transaction_fop_failed (frame, this, child_index);
-
-                if (op_ret != -1) {
-                        if (local->success_count == 0) {
-                                local->op_ret              = op_ret;
-                                local->cont.writev.prebuf  = *prebuf;
-                                local->cont.writev.postbuf = *postbuf;
-                        }
-
-                        if (child_index == read_child) {
-                                local->cont.writev.prebuf  = *prebuf;
-                                local->cont.writev.postbuf = *postbuf;
-                        }
-                }
-
-                local->op_errno = op_errno;
+                __afr_inode_write_fill (frame, this, child_index, op_ret,
+					op_errno, prebuf, postbuf, xdata);
         }
         UNLOCK (&frame->lock);
 
         call_count = afr_frame_return (frame);
 
         if (call_count == 0) {
-                local->transaction.unwind (frame, this);
+		__afr_inode_write_finalize (frame, this);
+
+		if (afr_txn_nothing_failed (frame, this))
+			local->transaction.unwind (frame, this);
 
                 local->transaction.resume (frame, this);
         }
+
         return 0;
 }
 
+/* {{{ writev */
+
+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;
+	if (src_local->xdata_rsp)
+		dst_local->xdata_rsp = dict_ref (src_local->xdata_rsp);
+}
+
+void
+afr_writev_unwind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *   local = 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,
+                          local->xdata_rsp);
+}
+
 
 int
-afr_writev_wind (call_frame_t *frame, xlator_t *this)
+afr_transaction_writev_unwind (call_frame_t *frame, xlator_t *this)
 {
-        afr_local_t *local = NULL;
-        afr_private_t *priv = NULL;
+        call_frame_t *fop_frame = NULL;
 
-        int i = 0;
-        int call_count = -1;
+        fop_frame = afr_transaction_detach_fop_frame (frame);
+
+        if (fop_frame) {
+                afr_writev_copy_outvars (frame, fop_frame);
+                afr_writev_unwind (fop_frame, this);
+        }
+        return 0;
+}
+
+static void
+afr_writev_handle_short_writes (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t   *local = NULL;
+        afr_private_t *priv  = NULL;
+        int           i      = 0;
 
         local = frame->local;
         priv = this->private;
+        /*
+         * We already have the best case result of the writev calls staged
+         * as the return value. Any writev that returns some value less
+         * than the best case is now out of sync, so mark the fop as
+         * failed. Note that fops that have returned with errors have
+         * already been marked as failed.
+         */
+        for (i = 0; i < priv->child_count; i++) {
+                if ((!local->replies[i].valid) ||
+                    (local->replies[i].op_ret == -1))
+                        continue;
 
-        call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-        if (call_count == 0) {
-                local->transaction.resume (frame, this);
-                return 0;
+                if (local->replies[i].op_ret < local->op_ret)
+                        afr_transaction_fop_failed(frame, this, i);
         }
+}
 
-        local->call_count = call_count;
+int
+afr_writev_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                     struct iatt *postbuf, dict_t *xdata)
+{
+        afr_local_t *   local = NULL;
+        call_frame_t    *fop_frame = NULL;
+        int child_index = (long) cookie;
+        int call_count  = -1;
+        int ret = 0;
+        uint32_t open_fd_count = 0;
+        uint32_t write_is_append = 0;
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (local->child_up[i]) {
-                        STACK_WIND_COOKIE (frame, afr_writev_wind_cbk,
-                                           (void *) (long) i,
-                                           priv->children[i],
-                                           priv->children[i]->fops->writev,
-                                           local->fd,
-                                           local->cont.writev.vector,
-                                           local->cont.writev.count,
-                                           local->cont.writev.offset,
-                                           local->cont.writev.iobref);
-
-                        if (!--call_count)
-                                break;
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                __afr_inode_write_fill (frame, this, child_index, op_ret,
+					op_errno, prebuf, postbuf, xdata);
+		if (op_ret == -1 || !xdata)
+			goto unlock;
+
+		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;
+
+		ret = dict_get_uint32 (xdata, GLUSTERFS_OPEN_FD_COUNT,
+				       &open_fd_count);
+		if (ret == -1)
+			goto unlock;
+		if ((open_fd_count > local->open_fd_count)) {
+			local->open_fd_count = open_fd_count;
+			local->update_open_fd_count = _gf_true;
+		}
+        }
+unlock:
+        UNLOCK (&frame->lock);
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0) {
+		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_inode_write_finalize (frame, this);
+
+                afr_writev_handle_short_writes (frame, this);
+
+                if (local->update_open_fd_count)
+                        afr_handle_open_fd_count (frame, this);
+
+                if (!afr_txn_nothing_failed (frame, this)) {
+                        //Don't unwind until post-op is complete
+                        local->transaction.resume (frame, this);
+                } else {
+                /*
+                 * Generally inode-write fops do transaction.unwind then
+                 * transaction.resume, but writev needs to make sure that
+                 * delayed post-op frame is placed in fdctx before unwind
+                 * happens. This prevents the race of flush doing the
+                 * changelog wakeup first in fuse thread and then this
+                 * writev placing its delayed post-op frame in fdctx.
+                 * This helps flush make sure all the delayed post-ops are
+                 * completed.
+                 */
+
+                        fop_frame = afr_transaction_detach_fop_frame (frame);
+                        afr_writev_copy_outvars (frame, fop_frame);
+                        local->transaction.resume (frame, this);
+                        afr_writev_unwind (fop_frame, this);
                 }
         }
-
         return 0;
 }
 
 
 int
-afr_writev_done (call_frame_t *frame, xlator_t *this)
+afr_writev_wind (call_frame_t *frame, xlator_t *this, int subvol)
 {
         afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
         local = frame->local;
+        priv = this->private;
 
-        iobref_unref (local->cont.writev.iobref);
-        local->cont.writev.iobref = NULL;
-
-        local->transaction.unwind (frame, this);
-
-        AFR_STACK_DESTROY (frame);
-
+	STACK_WIND_COOKIE (frame, afr_writev_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->writev,
+			   local->fd, local->cont.writev.vector,
+			   local->cont.writev.count, local->cont.writev.offset,
+			   local->cont.writev.flags, local->cont.writev.iobref,
+			   local->xdata_req);
         return 0;
 }
 
@@ -193,35 +341,39 @@ afr_writev_done (call_frame_t *frame, xlator_t *this)
 int
 afr_do_writev (call_frame_t *frame, xlator_t *this)
 {
-        call_frame_t * transaction_frame = NULL;
-        afr_local_t *  local             = NULL;
-
-        int op_ret   = -1;
-        int op_errno = 0;
-
-        local = frame->local;
+        call_frame_t    *transaction_frame = NULL;
+        afr_local_t     *local             = NULL;
+        int             ret   = -1;
+        int             op_errno = ENOMEM;
 
         transaction_frame = copy_frame (frame);
-        if (!transaction_frame) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
-                op_errno = ENOMEM;
+        if (!transaction_frame)
                 goto out;
-        }
 
+        local = frame->local;
         transaction_frame->local = local;
-        frame->local = NULL;
+	frame->local = NULL;
 
-        local->op = GF_FOP_WRITE;
+	if (!AFR_FRAME_INIT (frame, op_errno))
+		goto out;
 
-        local->success_count      = 0;
+        local->op = GF_FOP_WRITE;
 
-        local->transaction.fop    = afr_writev_wind;
-        local->transaction.done   = afr_writev_done;
-        local->transaction.unwind = afr_writev_unwind;
+        local->transaction.wind   = afr_writev_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_transaction_writev_unwind;
 
         local->transaction.main_frame = frame;
+
         if (local->fd->flags & O_APPEND) {
+               /*
+                * Backend vfs ignores the 'offset' for append mode fd so
+                * locking just the region provided for the writev does not
+                * give consistency gurantee. The actual write may happen at a
+                * completely different range than the one provided by the
+                * offset, len in the fop. So lock the entire file.
+                */
                 local->transaction.start   = 0;
                 local->transaction.len     = 0;
         } else {
@@ -230,16 +382,18 @@ afr_do_writev (call_frame_t *frame, xlator_t *this)
                                                          local->cont.writev.count);
         }
 
-        afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+        ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+        if (ret < 0) {
+            op_errno = -ret;
+            goto out;
+        }
 
-        op_ret = 0;
+	return 0;
 out:
-        if (op_ret == -1) {
-                if (transaction_frame)
-                        AFR_STACK_DESTROY (transaction_frame);
-                AFR_STACK_UNWIND (writev, frame, op_ret, op_errno, NULL, NULL);
-        }
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
 
+	AFR_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
@@ -247,62 +401,67 @@ out:
 int
 afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
             struct iovec *vector, int32_t count, off_t offset,
-            struct iobref *iobref)
+            uint32_t flags, struct iobref *iobref, dict_t *xdata)
 {
-        afr_private_t * priv  = NULL;
-        afr_local_t   * local = NULL;
+        afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        int op_errno = ENOMEM;
 
-        int ret = -1;
+        priv = this->private;
 
-        int op_ret   = -1;
-        int op_errno = 0;
+        QUORUM_CHECK(writev,out);
 
-        uint64_t ctx;
-        afr_fd_ctx_t *fd_ctx = NULL;
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
+        local->cont.writev.vector = iov_dup (vector, count);
+	if (!local->cont.writev.vector)
+		goto out;
+        local->cont.writev.count      = count;
+        local->cont.writev.offset     = offset;
+        local->cont.writev.flags      = flags;
+        local->cont.writev.iobref     = iobref_ref (iobref);
 
-        priv = this->private;
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-        ALLOC_OR_GOTO (local, afr_local_t, out);
+	if (!local->xdata_req)
+		goto out;
 
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
-                goto out;
-        }
+        local->fd = fd_ref (fd);
+	local->inode = inode_ref (fd->inode);
 
-        frame->local = local;
+	if (dict_set_uint32 (local->xdata_req, GLUSTERFS_OPEN_FD_COUNT, 4)) {
+		op_errno = ENOMEM;
+		goto out;
+	}
 
-        local->cont.writev.vector     = iov_dup (vector, count);
-        local->cont.writev.count      = count;
-        local->cont.writev.offset     = offset;
-        local->cont.writev.ino        = fd->inode->ino;
-        local->cont.writev.iobref     = iobref_ref (iobref);
+	if (dict_set_uint32 (local->xdata_req, GLUSTERFS_WRITE_IS_APPEND, 4)) {
+		op_errno = ENOMEM;
+		goto out;
+	}
 
-        local->fd                = fd_ref (fd);
+	/* 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;
 
-        ret = fd_ctx_get (fd, this, &ctx);
-        if (ret < 0) {
-                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));
 
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+        afr_fix_open (fd, this);
 
-        if (fd_ctx->up_count < priv->up_count) {
-                local->openfd_flush_cbk = afr_do_writev;
-                afr_openfd_flush (frame, this, fd);
-        } else {
-                afr_do_writev (frame, this);
-        }
+        afr_do_writev (frame, this);
 
-        op_ret = 0;
+	return 0;
 out:
-        if (op_ret == -1) {
-                AFR_STACK_UNWIND (writev, frame, op_ret, op_errno, NULL, NULL);
-        }
+	AFR_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
 
         return 0;
 }
@@ -320,24 +479,13 @@ afr_truncate_unwind (call_frame_t *frame, xlator_t *this)
 
         local = frame->local;
 
-        LOCK (&frame->lock);
-        {
-                if (local->transaction.main_frame)
-                        main_frame = local->transaction.main_frame;
-                local->transaction.main_frame = NULL;
-        }
-        UNLOCK (&frame->lock);
-
-        if (main_frame) {
-                local->cont.truncate.prebuf.ia_ino  = local->cont.truncate.ino;
-                local->cont.truncate.postbuf.ia_ino = local->cont.truncate.ino;
-
-                AFR_STACK_UNWIND (truncate, main_frame, local->op_ret,
-                                  local->op_errno,
-                                  &local->cont.truncate.prebuf,
-                                  &local->cont.truncate.postbuf);
-        }
+	main_frame = afr_transaction_detach_fop_frame (frame);
+	if (!main_frame)
+		return 0;
 
+	AFR_STACK_UNWIND (truncate, main_frame, local->op_ret, local->op_errno,
+			  &local->cont.inode_wfop.prebuf,
+			  &local->cont.inode_wfop.postbuf, local->xdata_rsp);
         return 0;
 }
 
@@ -345,181 +493,99 @@ afr_truncate_unwind (call_frame_t *frame, xlator_t *this)
 int
 afr_truncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                        int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                       struct iatt *postbuf)
+                       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_read_child (this, local->loc.inode);
-
-        LOCK (&frame->lock);
-        {
-                if (child_index == read_child) {
-                        local->read_child_returned = _gf_true;
-                }
-
-                if (afr_fop_failed (op_ret, op_errno) && op_errno != EFBIG)
-                        afr_transaction_fop_failed (frame, this, child_index);
+	afr_local_t *local = NULL;
 
-                if (op_ret != -1) {
-                        if (local->success_count == 0) {
-                                local->op_ret = op_ret;
-                                local->cont.truncate.prebuf  = *prebuf;
-                                local->cont.truncate.postbuf = *postbuf;
-                        }
+	local = frame->local;
 
-                        if (child_index == read_child) {
-                                local->cont.truncate.prebuf  = *prebuf;
-                                local->cont.truncate.postbuf = *postbuf;
-                        }
+	if (op_ret == 0 && prebuf->ia_size != postbuf->ia_size)
+		local->stable_write = _gf_false;
 
-                        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;
+	return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      prebuf, postbuf, xdata);
 }
 
 
-int32_t
-afr_truncate_wind (call_frame_t *frame, xlator_t *this)
+int
+afr_truncate_wind (call_frame_t *frame, xlator_t *this, int subvol)
 {
         afr_local_t *local = NULL;
         afr_private_t *priv = NULL;
 
-        int call_count = -1;
-        int i = 0;
-
         local = frame->local;
         priv = this->private;
 
-        call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-        if (call_count == 0) {
-                local->transaction.resume (frame, this);
-                return 0;
-        }
-
-        local->call_count = call_count;
-
-        for (i = 0; i < priv->child_count; i++) {
-                if (local->child_up[i]) {
-                        STACK_WIND_COOKIE (frame, afr_truncate_wind_cbk,
-                                           (void *) (long) i,
-                                           priv->children[i],
-                                           priv->children[i]->fops->truncate,
-                                           &local->loc,
-                                           local->cont.truncate.offset);
-
-                        if (!--call_count)
-                                break;
-                }
-        }
-
-        return 0;
-}
-
-
-int
-afr_truncate_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);
-
+	STACK_WIND_COOKIE (frame, afr_truncate_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->truncate,
+			   &local->loc, local->cont.truncate.offset,
+			   local->xdata_req);
         return 0;
 }
 
 
 int
 afr_truncate (call_frame_t *frame, xlator_t *this,
-              loc_t *loc, off_t offset)
+              loc_t *loc, off_t offset, dict_t *xdata)
 {
         afr_private_t * priv  = NULL;
         afr_local_t   * local = NULL;
         call_frame_t   *transaction_frame = NULL;
-
-        int ret = -1;
-
-        int op_ret   = -1;
-        int op_errno = 0;
-
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
+	int ret = -1;
+        int op_errno = ENOMEM;
 
         priv = this->private;
 
-        transaction_frame = copy_frame (frame);
-        if (!transaction_frame) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
-                goto out;
-        }
-
-        ALLOC_OR_GOTO (local, afr_local_t, out);
+        QUORUM_CHECK(truncate,out);
 
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
                 goto out;
-        }
 
-        transaction_frame->local = local;
-
-        local->op_ret = -1;
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+		goto out;
 
         local->cont.truncate.offset  = offset;
-        local->cont.truncate.ino     = loc->inode->ino;
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
+
+	if (!local->xdata_req)
+		goto out;
 
-        local->transaction.fop    = afr_truncate_wind;
-        local->transaction.done   = afr_truncate_done;
+	local->transaction.wind   = afr_truncate_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
         local->transaction.unwind = afr_truncate_unwind;
 
         loc_copy (&local->loc, loc);
+	local->inode = inode_ref (loc->inode);
+
+        local->op = GF_FOP_TRUNCATE;
 
         local->transaction.main_frame = frame;
-        local->transaction.start   = 0;
-        local->transaction.len     = offset;
+        local->transaction.start   = offset;
+        local->transaction.len     = 0;
 
-        afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+	/* Set it true speculatively, will get reset in afr_truncate_wind_cbk
+	   if truncate was not a NOP */
+	local->stable_write = _gf_true;
 
-        op_ret = 0;
-out:
-        if (op_ret == -1) {
-                if (transaction_frame)
-                        AFR_STACK_DESTROY (transaction_frame);
-                AFR_STACK_UNWIND (truncate, frame, op_ret, op_errno, NULL, NULL);
+        ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+        if (ret < 0) {
+		op_errno = -ret;
+		goto out;
         }
 
+	return 0;
+out:
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
+
+	AFR_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
@@ -537,23 +603,13 @@ afr_ftruncate_unwind (call_frame_t *frame, xlator_t *this)
 
         local = frame->local;
 
-        LOCK (&frame->lock);
-        {
-                if (local->transaction.main_frame)
-                        main_frame = local->transaction.main_frame;
-                local->transaction.main_frame = NULL;
-        }
-        UNLOCK (&frame->lock);
-
-        if (main_frame) {
-                local->cont.ftruncate.prebuf.ia_ino  = local->cont.ftruncate.ino;
-                local->cont.ftruncate.postbuf.ia_ino = local->cont.ftruncate.ino;
+	main_frame = afr_transaction_detach_fop_frame (frame);
+	if (!main_frame)
+		return 0;
 
-                AFR_STACK_UNWIND (ftruncate, main_frame, local->op_ret,
-                                  local->op_errno,
-                                  &local->cont.ftruncate.prebuf,
-                                  &local->cont.ftruncate.postbuf);
-        }
+	AFR_STACK_UNWIND (ftruncate, main_frame, local->op_ret, local->op_errno,
+			  &local->cont.inode_wfop.prebuf,
+			  &local->cont.inode_wfop.postbuf, local->xdata_rsp);
         return 0;
 }
 
@@ -561,1034 +617,1138 @@ afr_ftruncate_unwind (call_frame_t *frame, xlator_t *this)
 int
 afr_ftruncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                        struct iatt *postbuf)
+                        struct iatt *postbuf, dict_t *xdata)
 {
-        afr_local_t *   local = NULL;
-        afr_private_t * priv  = NULL;
+	afr_local_t *local = NULL;
 
-        int child_index = (long) cookie;
-        int call_count  = -1;
-        int need_unwind = 0;
-        int read_child  = 0;
+	local = frame->local;
 
-        local = frame->local;
-        priv  = this->private;
+	if (op_ret == 0 && prebuf->ia_size != postbuf->ia_size)
+		local->stable_write = _gf_false;
 
-        read_child = afr_read_child (this, local->fd->inode);
+	return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      prebuf, postbuf, xdata);
+}
 
-        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);
+int
+afr_ftruncate_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
-                if (op_ret != -1) {
-                        if (local->success_count == 0) {
-                                local->op_ret = op_ret;
-                                local->cont.ftruncate.prebuf  = *prebuf;
-                                local->cont.ftruncate.postbuf = *postbuf;
-                        }
+	local = frame->local;
+	priv = this->private;
 
-                        if (child_index == read_child) {
-                                local->cont.ftruncate.prebuf  = *prebuf;
-                                local->cont.ftruncate.postbuf = *postbuf;
-                        }
+	STACK_WIND_COOKIE (frame, afr_ftruncate_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->ftruncate,
+			   local->fd, local->cont.ftruncate.offset,
+			   local->xdata_req);
+        return 0;
+}
 
-                        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);
+int
+afr_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+	       dict_t *xdata)
+{
+        afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+	call_frame_t *transaction_frame = NULL;
+	int ret = -1;
+        int op_errno = ENOMEM;
 
-        if (need_unwind)
-                local->transaction.unwind (frame, this);
+        priv = this->private;
 
-        call_count = afr_frame_return (frame);
+        QUORUM_CHECK(ftruncate,out);
 
-        if (call_count == 0) {
-                local->transaction.resume (frame, this);
-        }
+	transaction_frame = copy_frame (frame);
+	if (!frame)
+		goto out;
 
-        return 0;
-}
+        local = AFR_FRAME_INIT (transaction_frame, op_errno);
+        if (!local)
+		goto out;
 
+        local->cont.ftruncate.offset  = offset;
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-int
-afr_ftruncate_wind (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t *local = NULL;
-        afr_private_t *priv = NULL;
+	if (!local->xdata_req)
+		goto out;
 
-        int call_count = -1;
-        int i = 0;
+        local->fd = fd_ref (fd);
+	local->inode = inode_ref (fd->inode);
 
-        local = frame->local;
-        priv = this->private;
+        local->op = GF_FOP_FTRUNCATE;
 
-        call_count = afr_up_children_count (priv->child_count, local->child_up);
+	local->transaction.wind   = afr_ftruncate_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_ftruncate_unwind;
 
-        if (call_count == 0) {
-                local->transaction.resume (frame, this);
-                return 0;
-        }
+        local->transaction.main_frame = frame;
 
-        local->call_count = call_count;
+        local->transaction.start   = local->cont.ftruncate.offset;
+        local->transaction.len     = 0;
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (local->child_up[i]) {
-                        STACK_WIND_COOKIE (frame, afr_ftruncate_wind_cbk,
-                                           (void *) (long) i,
-                                           priv->children[i],
-                                           priv->children[i]->fops->ftruncate,
-                                           local->fd, local->cont.ftruncate.offset);
-
-                        if (!--call_count)
-                                break;
-                }
+        afr_fix_open (fd, this);
+
+	/* Set it true speculatively, will get reset in afr_ftruncate_wind_cbk
+	   if truncate was not a NOP */
+	local->stable_write = _gf_true;
+
+        ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+        if (ret < 0) {
+		op_errno = -ret;
+		goto out;
         }
 
+	return 0;
+out:
+	AFR_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
+
         return 0;
 }
 
+/* }}} */
+
+/* {{{ setattr */
 
 int
-afr_ftruncate_done (call_frame_t *frame, xlator_t *this)
+afr_setattr_unwind (call_frame_t *frame, xlator_t *this)
 {
         afr_local_t *local = NULL;
+        call_frame_t *main_frame = NULL;
 
         local = frame->local;
 
-        local->transaction.unwind (frame, this);
-
-        AFR_STACK_DESTROY (frame);
+	main_frame = afr_transaction_detach_fop_frame (frame);
+	if (!main_frame)
+		return 0;
 
+	AFR_STACK_UNWIND (setattr, main_frame, local->op_ret, local->op_errno,
+			  &local->cont.inode_wfop.prebuf,
+			  &local->cont.inode_wfop.postbuf,
+			  local->xdata_rsp);
         return 0;
 }
 
 
 int
-afr_do_ftruncate (call_frame_t *frame, xlator_t *this)
+afr_setattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int op_ret, int op_errno,
+                      struct iatt *preop, struct iatt *postop, dict_t *xdata)
 {
-        call_frame_t * transaction_frame = NULL;
-        afr_local_t *  local             = NULL;
-
-        int op_ret   = -1;
-        int op_errno = 0;
-
-        local = frame->local;
-
-        transaction_frame = copy_frame (frame);
-        if (!transaction_frame) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
-                goto out;
-        }
-
-        transaction_frame->local = local;
-        frame->local = NULL;
-
-        local->op = GF_FOP_FTRUNCATE;
-
-        local->transaction.fop    = afr_ftruncate_wind;
-        local->transaction.done   = afr_ftruncate_done;
-        local->transaction.unwind = afr_ftruncate_unwind;
-
-        local->transaction.main_frame = frame;
+	return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      preop, postop, xdata);
+}
 
-        local->transaction.start   = 0;
-        local->transaction.len     = local->cont.ftruncate.offset;
 
-        afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+int
+afr_setattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
-        op_ret = 0;
-out:
-        if (op_ret == -1) {
-                if (transaction_frame)
-                        AFR_STACK_DESTROY (transaction_frame);
-                AFR_STACK_UNWIND (ftruncate, frame, op_ret, op_errno, NULL, NULL);
-        }
+        local = frame->local;
+        priv = this->private;
 
+	STACK_WIND_COOKIE (frame, afr_setattr_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->setattr,
+			   &local->loc, &local->cont.setattr.in_buf,
+			   local->cont.setattr.valid, local->xdata_req);
         return 0;
 }
 
 
 int
-afr_ftruncate (call_frame_t *frame, xlator_t *this,
-               fd_t *fd, off_t offset)
+afr_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *buf,
+	     int32_t valid, dict_t *xdata)
 {
-        afr_private_t * priv  = NULL;
-        afr_local_t   * local = NULL;
-        call_frame_t   *transaction_frame = NULL;
-
+        afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        call_frame_t *transaction_frame = NULL;
         int ret = -1;
-
-        int op_ret   = -1;
-        int op_errno = 0;
-
-        uint64_t ctx;
-        afr_fd_ctx_t *fd_ctx = NULL;
-
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
+        int op_errno = ENOMEM;
 
         priv = this->private;
 
-        ALLOC_OR_GOTO (local, afr_local_t, out);
+        QUORUM_CHECK(setattr,out);
 
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
                 goto out;
-        }
 
-        frame->local = local;
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+		goto out;
 
-        local->cont.ftruncate.offset  = offset;
-        local->cont.ftruncate.ino     = fd->inode->ino;
+        local->cont.setattr.in_buf = *buf;
+        local->cont.setattr.valid  = valid;
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-        local->fd = fd_ref (fd);
+	if (!local->xdata_req)
+		goto out;
 
-        ret = fd_ctx_get (fd, this, &ctx);
-        if (ret < 0) {
-                goto out;
-        }
+        local->transaction.wind   = afr_setattr_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_setattr_unwind;
+
+        loc_copy (&local->loc, loc);
+	local->inode = inode_ref (loc->inode);
 
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+	local->op = GF_FOP_SETATTR;
 
-        if (fd_ctx->up_count < priv->up_count) {
-                local->openfd_flush_cbk = afr_do_ftruncate;
-                afr_openfd_flush (frame, this, fd);
-        } else {
-                afr_do_ftruncate (frame, this);
+        local->transaction.main_frame = frame;
+        local->transaction.start   = LLONG_MAX - 1;
+        local->transaction.len     = 0;
+
+        ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        if (ret < 0) {
+		op_errno = -ret;
+		goto out;
         }
 
-        op_ret = 0;
+	return 0;
 out:
-        if (op_ret == -1) {
-                if (transaction_frame)
-                        AFR_STACK_DESTROY (transaction_frame);
-                AFR_STACK_UNWIND (ftruncate, frame, op_ret, op_errno, NULL, NULL);
-        }
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
 
+	AFR_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
-/* }}} */
-
-/* {{{ setattr */
+/* {{{ fsetattr */
 
 int
-afr_setattr_unwind (call_frame_t *frame, xlator_t *this)
+afr_fsetattr_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);
+	main_frame = afr_transaction_detach_fop_frame (frame);
+        if (!main_frame)
+		return 0;
+
+	AFR_STACK_UNWIND (fsetattr, main_frame, local->op_ret, local->op_errno,
+			  &local->cont.inode_wfop.prebuf,
+			  &local->cont.inode_wfop.postbuf, local->xdata_rsp);
+        return 0;
+}
 
-        if (main_frame) {
-                local->cont.setattr.preop_buf.ia_ino  = local->cont.setattr.ino;
-                local->cont.setattr.postop_buf.ia_ino = local->cont.setattr.ino;
 
-                AFR_STACK_UNWIND (setattr, main_frame, local->op_ret,
-                                  local->op_errno,
-                                  &local->cont.setattr.preop_buf,
-                                  &local->cont.setattr.postop_buf);
-        }
+int
+afr_fsetattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno,
+                       struct iatt *preop, struct iatt *postop, dict_t *xdata)
+{
+	return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      preop, postop, xdata);
+}
 
+
+int
+afr_fsetattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+
+        local = frame->local;
+        priv = this->private;
+
+	STACK_WIND_COOKIE (frame, afr_fsetattr_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->fsetattr,
+			   local->fd, &local->cont.fsetattr.in_buf,
+			   local->cont.fsetattr.valid, local->xdata_req);
         return 0;
 }
 
 
 int
-afr_setattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno,
-                      struct iatt *preop, struct iatt *postop)
+afr_fsetattr (call_frame_t *frame, xlator_t *this,
+              fd_t *fd, struct iatt *buf, int32_t valid, dict_t *xdata)
 {
-        afr_local_t *   local = NULL;
-        afr_private_t * priv  = NULL;
+        afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        call_frame_t *transaction_frame = NULL;
+        int ret = -1;
+        int op_errno = ENOMEM;
 
-        int child_index = (long) cookie;
-        int read_child  = 0;
-        int call_count  = -1;
-        int need_unwind = 0;
+        priv = this->private;
 
-        local = frame->local;
-        priv  = this->private;
+        QUORUM_CHECK(fsetattr,out);
 
-        read_child = afr_read_child (this, local->loc.inode);
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto out;
 
-        LOCK (&frame->lock);
-        {
-                if (child_index == read_child) {
-                        local->read_child_returned = _gf_true;
-                }
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+		goto out;
 
-                if (afr_fop_failed (op_ret, op_errno))
-                        afr_transaction_fop_failed (frame, this, child_index);
+        local->cont.fsetattr.in_buf = *buf;
+        local->cont.fsetattr.valid  = valid;
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-                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 (!local->xdata_req)
+		goto out;
 
-                        if (child_index == read_child) {
-                                local->cont.setattr.preop_buf  = *preop;
-                                local->cont.setattr.postop_buf = *postop;
-                        }
+        local->transaction.wind   = afr_fsetattr_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_fsetattr_unwind;
 
-                        local->success_count++;
+        local->fd                 = fd_ref (fd);
+	local->inode = inode_ref (fd->inode);
 
-                        if ((local->success_count >= priv->wait_count)
-                            && local->read_child_returned) {
-                                need_unwind = 1;
-                        }
-                }
-                local->op_errno = op_errno;
-        }
-        UNLOCK (&frame->lock);
+	local->op = GF_FOP_FSETATTR;
 
-        if (need_unwind)
-                local->transaction.unwind (frame, this);
+        afr_fix_open (fd, this);
 
-        call_count = afr_frame_return (frame);
+        local->transaction.main_frame = frame;
+        local->transaction.start   = LLONG_MAX - 1;
+        local->transaction.len     = 0;
 
-        if (call_count == 0) {
-                local->transaction.resume (frame, this);
+        ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        if (ret < 0) {
+		op_errno = -ret;
+		goto out;
         }
 
+	return 0;
+out:
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
+
+	AFR_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 
-int32_t
-afr_setattr_wind (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t *local = NULL;
-        afr_private_t *priv = NULL;
-
-        int call_count = -1;
-        int i = 0;
-
-        local = frame->local;
-        priv = this->private;
+/* {{{ setxattr */
 
-        call_count = afr_up_children_count (priv->child_count, local->child_up);
 
-        if (call_count == 0) {
-                local->transaction.resume (frame, this);
-                return 0;
-        }
+int
+afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
 
-        local->call_count = call_count;
+        local = frame->local;
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (local->child_up[i]) {
-                        STACK_WIND_COOKIE (frame, afr_setattr_wind_cbk,
-                                           (void *) (long) i,
-                                           priv->children[i],
-                                           priv->children[i]->fops->setattr,
-                                           &local->loc,
-                                           &local->cont.setattr.in_buf,
-                                           local->cont.setattr.valid);
-
-                        if (!--call_count)
-                                break;
-                }
-        }
+	main_frame = afr_transaction_detach_fop_frame (frame);
+	if (!main_frame)
+		return 0;
 
+	AFR_STACK_UNWIND (setxattr, main_frame, local->op_ret, local->op_errno,
+			  local->xdata_rsp);
         return 0;
 }
 
 
 int
-afr_setattr_done (call_frame_t *frame, xlator_t *this)
+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;
+        return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      NULL, NULL, xdata);
+}
 
-        local = frame->local;
 
-        local->transaction.unwind (frame, this);
+int
+afr_setxattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t      *local         = NULL;
+        afr_private_t    *priv          = NULL;
 
-        AFR_STACK_DESTROY (frame);
+        local = frame->local;
+        priv = this->private;
 
+	STACK_WIND_COOKIE (frame, afr_setxattr_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->setxattr,
+			   &local->loc, local->cont.setxattr.dict,
+			   local->cont.setxattr.flags, local->xdata_req);
         return 0;
 }
 
 
 int
-afr_setattr (call_frame_t *frame, xlator_t *this,
-             loc_t *loc, struct iatt *buf, int32_t valid)
+afr_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+	      int32_t flags, dict_t *xdata)
 {
-        afr_private_t * priv  = NULL;
-        afr_local_t   * local = NULL;
+        afr_private_t  *priv              = NULL;
+        afr_local_t    *local             = NULL;
         call_frame_t   *transaction_frame = NULL;
+        int             ret               = -1;
+        int             op_errno          = EINVAL;
 
-        int ret = -1;
-
-        int op_ret   = -1;
-        int op_errno = 0;
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.afr.*", dict,
+                                   op_errno, out);
 
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.afr.*", dict,
+                                   op_errno, out);
 
         priv = this->private;
 
+        QUORUM_CHECK(setxattr,out);
+
         transaction_frame = copy_frame (frame);
-        if (!transaction_frame) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
+        if (!transaction_frame)
                 goto out;
-        }
-
-        ALLOC_OR_GOTO (local, afr_local_t, out);
 
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
                 goto out;
-        }
-
-        transaction_frame->local = local;
 
-        local->op_ret = -1;
-
-        local->cont.setattr.ino     = loc->inode->ino;
+        local->cont.setxattr.dict  = dict_ref (dict);
+        local->cont.setxattr.flags = flags;
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-        local->cont.setattr.in_buf = *buf;
-        local->cont.setattr.valid  = valid;
+	if (!local->xdata_req)
+		goto out;
 
-        local->transaction.fop    = afr_setattr_wind;
-        local->transaction.done   = afr_setattr_done;
-        local->transaction.unwind = afr_setattr_unwind;
+        local->transaction.wind   = afr_setxattr_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_setxattr_unwind;
 
         loc_copy (&local->loc, loc);
+	local->inode = inode_ref (loc->inode);
 
         local->transaction.main_frame = frame;
         local->transaction.start   = LLONG_MAX - 1;
         local->transaction.len     = 0;
 
-        afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+	local->op = GF_FOP_SETXATTR;
 
-        op_ret = 0;
-out:
-        if (op_ret == -1) {
-                if (transaction_frame)
-                        AFR_STACK_DESTROY (transaction_frame);
-                AFR_STACK_UNWIND (setattr, frame, op_ret, op_errno, NULL, NULL);
+        ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        if (ret < 0) {
+		op_errno = -ret;
+		goto out;
         }
 
+	return 0;
+out:
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
+
+	AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
+
         return 0;
 }
 
-/* {{{ fsetattr */
+/* {{{ fsetxattr */
+
 
 int
-afr_fsetattr_unwind (call_frame_t *frame, xlator_t *this)
+afr_fsetxattr_unwind (call_frame_t *frame, xlator_t *this)
 {
-        afr_local_t *   local = NULL;
-        call_frame_t   *main_frame = NULL;
+        afr_local_t    *local         = NULL;
+        call_frame_t   *main_frame    = NULL;
 
         local = frame->local;
 
-        LOCK (&frame->lock);
-        {
-                if (local->transaction.main_frame)
-                        main_frame = local->transaction.main_frame;
-                local->transaction.main_frame = NULL;
-        }
-        UNLOCK (&frame->lock);
+	main_frame = afr_transaction_detach_fop_frame (frame);
+        if (!main_frame)
+		return 0;
 
-        if (main_frame) {
-                local->cont.fsetattr.preop_buf.ia_ino  =
-                        local->cont.fsetattr.ino;
-                local->cont.fsetattr.postop_buf.ia_ino =
-                        local->cont.fsetattr.ino;
+	AFR_STACK_UNWIND (fsetxattr, main_frame, local->op_ret, local->op_errno,
+			  local->xdata_rsp);
+        return 0;
+}
 
-                AFR_STACK_UNWIND (fsetattr, main_frame, local->op_ret,
-                                  local->op_errno,
-                                  &local->cont.fsetattr.preop_buf,
-                                  &local->cont.fsetattr.postop_buf);
-        }
 
+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)
+{
+        return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      NULL, NULL, xdata);
+}
+
+
+int
+afr_fsetxattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t        *local       = NULL;
+        afr_private_t      *priv        = NULL;
+
+        local = frame->local;
+        priv = this->private;
+
+	STACK_WIND_COOKIE (frame, afr_fsetxattr_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->fsetxattr,
+			   local->fd, local->cont.fsetxattr.dict,
+			   local->cont.fsetxattr.flags, local->xdata_req);
         return 0;
 }
 
 
 int
-afr_fsetattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno,
-                      struct iatt *preop, struct iatt *postop)
+afr_fsetxattr (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, dict_t *dict, int32_t flags, dict_t *xdata)
 {
-        afr_local_t *   local = NULL;
-        afr_private_t * priv  = NULL;
+        afr_private_t    *priv              = NULL;
+        afr_local_t      *local             = NULL;
+        call_frame_t     *transaction_frame = NULL;
+        int               ret               = -1;
+        int               op_errno          = ENOMEM;
 
-        int child_index = (long) cookie;
-        int read_child  = 0;
-        int call_count  = -1;
-        int need_unwind = 0;
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.afr.*", dict,
+                                   op_errno, out);
 
-        local = frame->local;
-        priv  = this->private;
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.afr.*", dict,
+                                   op_errno, out);
 
-        read_child = afr_read_child (this, local->fd->inode);
+        priv = this->private;
 
-        LOCK (&frame->lock);
-        {
-                if (child_index == read_child) {
-                        local->read_child_returned = _gf_true;
-                }
+        QUORUM_CHECK(fsetxattr,out);
 
-                if (afr_fop_failed (op_ret, op_errno))
-                        afr_transaction_fop_failed (frame, this, child_index);
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto out;
 
-                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;
-                        }
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+		goto out;
 
-                        if (child_index == read_child) {
-                                local->cont.fsetattr.preop_buf  = *preop;
-                                local->cont.fsetattr.postop_buf = *postop;
-                        }
+        local->cont.fsetxattr.dict  = dict_ref (dict);
+        local->cont.fsetxattr.flags = flags;
 
-                        local->success_count++;
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-                        if ((local->success_count >= priv->wait_count)
-                            && local->read_child_returned) {
-                                need_unwind = 1;
-                        }
-                }
-                local->op_errno = op_errno;
-        }
-        UNLOCK (&frame->lock);
+	if (!local->xdata_req)
+		goto out;
 
-        if (need_unwind)
-                local->transaction.unwind (frame, this);
+        local->transaction.wind   = afr_fsetxattr_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_fsetxattr_unwind;
 
-        call_count = afr_frame_return (frame);
+        local->fd                 = fd_ref (fd);
+	local->inode = inode_ref (fd->inode);
 
-        if (call_count == 0) {
-                local->transaction.resume (frame, this);
+	local->op = GF_FOP_FSETXATTR;
+
+        local->transaction.main_frame = frame;
+        local->transaction.start  = LLONG_MAX - 1;
+        local->transaction.len    = 0;
+
+        ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        if (ret < 0) {
+		op_errno = -ret;
+		goto out;
         }
 
+	return 0;
+out:
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
+
+	AFR_STACK_UNWIND (fsetxattr, frame, -1, op_errno, NULL);
         return 0;
 }
 
+/* }}} */
 
-int32_t
-afr_fsetattr_wind (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t *local = NULL;
-        afr_private_t *priv = NULL;
-
-        int call_count = -1;
-        int i = 0;
 
-        local = frame->local;
-        priv = this->private;
+/* {{{ removexattr */
 
-        call_count = afr_up_children_count (priv->child_count, local->child_up);
 
-        if (call_count == 0) {
-                local->transaction.resume (frame, this);
-                return 0;
-        }
+int
+afr_removexattr_unwind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
 
-        local->call_count = call_count;
+        local = frame->local;
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (local->child_up[i]) {
-                        STACK_WIND_COOKIE (frame, afr_fsetattr_wind_cbk,
-                                           (void *) (long) i,
-                                           priv->children[i],
-                                           priv->children[i]->fops->fsetattr,
-                                           local->fd,
-                                           &local->cont.fsetattr.in_buf,
-                                           local->cont.fsetattr.valid);
-
-                        if (!--call_count)
-                                break;
-                }
-        }
+	main_frame = afr_transaction_detach_fop_frame (frame);
+	if (!main_frame)
+		return 0;
 
+	AFR_STACK_UNWIND (removexattr, main_frame, local->op_ret, local->op_errno,
+			  local->xdata_rsp);
         return 0;
 }
 
 
 int
-afr_fsetattr_done (call_frame_t *frame, xlator_t *this)
+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;
+        return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      NULL, NULL, xdata);
+}
 
-        local = frame->local;
 
-        local->transaction.unwind (frame, this);
+int
+afr_removexattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
-        AFR_STACK_DESTROY (frame);
+	local = frame->local;
+	priv = this->private;
 
+	STACK_WIND_COOKIE (frame, afr_removexattr_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->removexattr,
+			   &local->loc, local->cont.removexattr.name,
+			   local->xdata_req);
         return 0;
 }
 
 
 int
-afr_fsetattr (call_frame_t *frame, xlator_t *this,
-              fd_t *fd, struct iatt *buf, int32_t valid)
+afr_removexattr (call_frame_t *frame, xlator_t *this,
+                 loc_t *loc, const char *name, dict_t *xdata)
 {
-        afr_private_t * priv  = NULL;
-        afr_local_t   * local = NULL;
-        call_frame_t   *transaction_frame = NULL;
+        afr_private_t   *priv              = NULL;
+        afr_local_t     *local             = NULL;
+        call_frame_t    *transaction_frame = NULL;
+        int              ret               = -1;
+        int              op_errno          = ENOMEM;
 
-        int ret = -1;
-
-        int op_ret   = -1;
-        int op_errno = 0;
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.afr.*",
+                                 name, op_errno, out);
 
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.afr.*",
+                                 name, op_errno, out);
 
         priv = this->private;
 
+        QUORUM_CHECK(removexattr,out);
+
         transaction_frame = copy_frame (frame);
-        if (!transaction_frame) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
+        if (!transaction_frame)
                 goto out;
-        }
-
-        ALLOC_OR_GOTO (local, afr_local_t, out);
 
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
-                goto out;
-        }
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+		goto out;
 
-        transaction_frame->local = local;
+        local->cont.removexattr.name = gf_strdup (name);
 
-        local->op_ret = -1;
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-        local->cont.fsetattr.ino     = fd->inode->ino;
+	if (!local->xdata_req)
+		goto out;
 
-        local->cont.fsetattr.in_buf = *buf;
-        local->cont.fsetattr.valid  = valid;
+        local->transaction.wind   = afr_removexattr_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_removexattr_unwind;
 
-        local->transaction.fop    = afr_fsetattr_wind;
-        local->transaction.done   = afr_fsetattr_done;
-        local->transaction.unwind = afr_fsetattr_unwind;
+        loc_copy (&local->loc, loc);
+	local->inode = inode_ref (loc->inode);
 
-        local->fd                 = fd_ref (fd);
+	local->op = GF_FOP_REMOVEXATTR;
 
         local->transaction.main_frame = frame;
         local->transaction.start   = LLONG_MAX - 1;
         local->transaction.len     = 0;
 
-        afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        if (ret < 0) {
+		op_errno = -ret;
+		goto out;
+        }
 
-        op_ret = 0;
+	return 0;
 out:
-        if (op_ret == -1) {
-                if (transaction_frame)
-                        AFR_STACK_DESTROY (transaction_frame);
-                AFR_STACK_UNWIND (fsetattr, frame, op_ret, op_errno, NULL, NULL);
-        }
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
 
+	AFR_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
         return 0;
 }
 
-
-/* {{{ setxattr */
-
-
+/* ffremovexattr */
 int
-afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
+afr_fremovexattr_unwind (call_frame_t *frame, xlator_t *this)
 {
         afr_local_t *   local = NULL;
         call_frame_t   *main_frame = NULL;
 
         local = frame->local;
 
-        LOCK (&frame->lock);
-        {
-                if (local->transaction.main_frame)
-                        main_frame = local->transaction.main_frame;
-                local->transaction.main_frame = NULL;
-        }
-        UNLOCK (&frame->lock);
+	main_frame = afr_transaction_detach_fop_frame (frame);
+        if (!main_frame)
+		return 0;
 
-        if (main_frame) {
-                AFR_STACK_UNWIND (setxattr, main_frame,
-                                  local->op_ret, local->op_errno)
-        }
+	AFR_STACK_UNWIND (fremovexattr, main_frame, local->op_ret, local->op_errno,
+			  local->xdata_rsp);
         return 0;
 }
 
 
 int
-afr_setxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                       int32_t op_ret, int32_t op_errno)
+afr_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;
+	return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      NULL, NULL, xdata);
+}
 
-        int call_count  = -1;
-        int need_unwind = 0;
+
+int
+afr_fremovexattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
         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;
-                        }
-                }
-
-                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);
-        }
-
+	STACK_WIND_COOKIE (frame, afr_fremovexattr_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->fremovexattr,
+			   local->fd, local->cont.removexattr.name,
+			   local->xdata_req);
         return 0;
 }
 
 
 int
-afr_setxattr_wind (call_frame_t *frame, xlator_t *this)
+afr_fremovexattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+		  const char *name, dict_t *xdata)
 {
+	afr_private_t *priv = NULL;
         afr_local_t *local = NULL;
-        afr_private_t *priv = NULL;
+        call_frame_t *transaction_frame = NULL;
+        int ret = -1;
+        int op_errno = ENOMEM;
 
-        int call_count = -1;
-        int i = 0;
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.afr.*",
+                                 name, op_errno, out);
 
-        local = frame->local;
-        priv = this->private;
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.afr.*",
+                                 name, op_errno, out);
 
-        call_count = afr_up_children_count (priv->child_count, local->child_up);
+	priv = this->private;
 
-        if (call_count == 0) {
-                local->transaction.resume (frame, this);
-                return 0;
-        }
+        QUORUM_CHECK(fremovexattr, out);
 
-        local->call_count = call_count;
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto out;
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (local->child_up[i]) {
-                        STACK_WIND_COOKIE (frame, afr_setxattr_wind_cbk,
-                                           (void *) (long) i,
-                                           priv->children[i],
-                                           priv->children[i]->fops->setxattr,
-                                           &local->loc,
-                                           local->cont.setxattr.dict,
-                                           local->cont.setxattr.flags);
-
-                        if (!--call_count)
-                                break;
-                }
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+                goto out;
+
+        local->cont.removexattr.name = gf_strdup (name);
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
+
+	if (!local->xdata_req)
+		goto out;
+
+        local->transaction.wind   = afr_fremovexattr_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_fremovexattr_unwind;
+
+        local->fd = fd_ref (fd);
+	local->inode = inode_ref (fd->inode);
+
+	local->op = GF_FOP_FREMOVEXATTR;
+
+        local->transaction.main_frame = frame;
+        local->transaction.start   = LLONG_MAX - 1;
+        local->transaction.len     = 0;
+
+        ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        if (ret < 0) {
+		op_errno = -ret;
+		goto out;
         }
 
+	return 0;
+out:
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
+
+	AFR_STACK_UNWIND (fremovexattr, frame, -1, op_errno, NULL);
+
         return 0;
 }
 
 
 int
-afr_setxattr_done (call_frame_t *frame, xlator_t *this)
+afr_fallocate_unwind (call_frame_t *frame, xlator_t *this)
 {
-        afr_local_t * local = frame->local;
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
 
-        local->transaction.unwind (frame, this);
+        local = frame->local;
 
-        AFR_STACK_DESTROY (frame);
+	main_frame = afr_transaction_detach_fop_frame (frame);
+	if (!main_frame)
+		return 0;
 
+	AFR_STACK_UNWIND (fallocate, main_frame, local->op_ret, local->op_errno,
+			  &local->cont.inode_wfop.prebuf,
+			  &local->cont.inode_wfop.postbuf, local->xdata_rsp);
         return 0;
 }
 
+
 int
-afr_setxattr (call_frame_t *frame, xlator_t *this,
-              loc_t *loc, dict_t *dict, int32_t flags)
+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_private_t * priv  = NULL;
-        afr_local_t   * local = NULL;
-        call_frame_t   *transaction_frame = NULL;
-
-        int ret = -1;
+        return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      prebuf, postbuf, xdata);
+}
 
-        int op_ret   = -1;
-        int op_errno = 0;
 
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
+int
+afr_fallocate_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
+        local = frame->local;
         priv = this->private;
 
-        ALLOC_OR_GOTO (local, afr_local_t, out);
+	STACK_WIND_COOKIE (frame, afr_fallocate_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->fallocate,
+			   local->fd, local->cont.fallocate.mode,
+			   local->cont.fallocate.offset,
+			   local->cont.fallocate.len, local->xdata_req);
+        return 0;
+}
 
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
-                goto out;
-        }
+
+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;
+        call_frame_t *transaction_frame = NULL;
+        afr_local_t *local = NULL;
+        int ret = -1;
+        int op_errno = ENOMEM;
+
+	priv = this->private;
+
+        QUORUM_CHECK(fallocate,out);
 
         transaction_frame = copy_frame (frame);
-        if (!transaction_frame) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
+        if (!transaction_frame)
                 goto out;
-        }
 
-        transaction_frame->local = local;
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+		goto out;
 
-        local->op_ret = -1;
+        local->cont.fallocate.mode = mode;
+        local->cont.fallocate.offset  = offset;
+        local->cont.fallocate.len = len;
 
-        local->cont.setxattr.dict  = dict_ref (dict);
-        local->cont.setxattr.flags = flags;
+        local->fd = fd_ref (fd);
+	local->inode = inode_ref (fd->inode);
 
-        local->transaction.fop    = afr_setxattr_wind;
-        local->transaction.done   = afr_setxattr_done;
-        local->transaction.unwind = afr_setxattr_unwind;
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-        loc_copy (&local->loc, loc);
+	if (!local->xdata_req)
+		goto out;
+
+        local->op = GF_FOP_FALLOCATE;
+
+        local->transaction.wind   = afr_fallocate_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_fallocate_unwind;
 
         local->transaction.main_frame = frame;
-        local->transaction.start   = LLONG_MAX - 1;
+
+        local->transaction.start   = local->cont.fallocate.offset;
         local->transaction.len     = 0;
 
-        afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        afr_fix_open (fd, this);
 
-        op_ret = 0;
-out:
-        if (op_ret == -1) {
-                if (transaction_frame)
-                        AFR_STACK_DESTROY (transaction_frame);
-                AFR_STACK_UNWIND (setxattr, frame, op_ret, op_errno);
+        ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+        if (ret < 0) {
+		op_errno = -ret;
+		goto out;
         }
 
+	return 0;
+out:
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
+
+	AFR_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
-/* }}} */
 
-/* {{{ removexattr */
+/* }}} */
 
+/* {{{ discard */
 
 int
-afr_removexattr_unwind (call_frame_t *frame, xlator_t *this)
+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);
+	main_frame = afr_transaction_detach_fop_frame (frame);
+        if (!main_frame)
+		return 0;
 
-        if (main_frame) {
-                AFR_STACK_UNWIND (removexattr, main_frame,
-                                  local->op_ret, local->op_errno)
-        }
+	AFR_STACK_UNWIND (discard, main_frame, local->op_ret, local->op_errno,
+			  &local->cont.inode_wfop.prebuf,
+			  &local->cont.inode_wfop.postbuf, local->xdata_rsp);
         return 0;
 }
 
 
 int
-afr_removexattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                          int32_t op_ret, int32_t op_errno)
+afr_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;
+	return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      prebuf, postbuf, xdata);
+}
 
-        int call_count  = -1;
-        int need_unwind = 0;
+
+int
+afr_discard_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
         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;
-                        }
-                }
+	STACK_WIND_COOKIE (frame, afr_discard_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->discard,
+			   local->fd, local->cont.discard.offset,
+			   local->cont.discard.len, local->xdata_req);
+        return 0;
+}
 
-                local->op_errno = op_errno;
-        }
-        UNLOCK (&frame->lock);
 
-        if (need_unwind)
-                local->transaction.unwind (frame, this);
+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 = ENOMEM;
 
-        call_count = afr_frame_return (frame);
+        priv = this->private;
 
-        if (call_count == 0) {
-                local->transaction.resume (frame, this);
-        }
+        QUORUM_CHECK(discard, out);
 
-        return 0;
-}
+	transaction_frame = copy_frame (frame);
+	if (!transaction_frame)
+		goto out;
 
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-int32_t
-afr_removexattr_wind (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t *local = NULL;
-        afr_private_t *priv = NULL;
+        local->cont.discard.offset  = offset;
+        local->cont.discard.len = len;
 
-        int call_count = -1;
-        int i = 0;
+        local->fd = fd_ref (fd);
+	local->inode = inode_ref (fd->inode);
 
-        local = frame->local;
-        priv = this->private;
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-        call_count = afr_up_children_count (priv->child_count, local->child_up);
+	if (!local->xdata_req)
+		goto out;
 
-        if (call_count == 0) {
-                local->transaction.resume (frame, this);
-                return 0;
-        }
+        local->op = GF_FOP_DISCARD;
 
-        local->call_count = call_count;
+        local->transaction.wind   = afr_discard_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_discard_unwind;
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (local->child_up[i]) {
-                        STACK_WIND_COOKIE (frame, afr_removexattr_wind_cbk,
-                                           (void *) (long) i,
-                                           priv->children[i],
-                                           priv->children[i]->fops->removexattr,
-                                           &local->loc,
-                                           local->cont.removexattr.name);
-
-                        if (!--call_count)
-                                break;
-                }
+        local->transaction.main_frame = frame;
+
+        local->transaction.start   = local->cont.discard.offset;
+        local->transaction.len     = 0;
+
+        afr_fix_open (fd, this);
+
+        ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+        if (ret < 0) {
+		op_errno = -ret;
+		goto out;
         }
 
+	return 0;
+out:
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
+
+	AFR_STACK_UNWIND (discard, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 
+/* {{{ zerofill */
+
 int
-afr_removexattr_done (call_frame_t *frame, xlator_t *this)
+afr_zerofill_unwind (call_frame_t *frame, xlator_t *this)
 {
-        afr_local_t * local = frame->local;
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
 
-        local->transaction.unwind (frame, this);
+        local = frame->local;
 
-        AFR_STACK_DESTROY (frame);
+	main_frame = afr_transaction_detach_fop_frame (frame);
+        if (!main_frame)
+		return 0;
 
+	AFR_STACK_UNWIND (discard, main_frame, local->op_ret, local->op_errno,
+			  &local->cont.inode_wfop.prebuf,
+			  &local->cont.inode_wfop.postbuf, local->xdata_rsp);
         return 0;
 }
 
 
 int
-afr_removexattr (call_frame_t *frame, xlator_t *this,
-                 loc_t *loc, const char *name)
+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_private_t * priv  = NULL;
-        afr_local_t   * local = NULL;
-        call_frame_t   *transaction_frame = NULL;
+	return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      prebuf, postbuf, xdata);
+}
 
-        int ret = -1;
 
-        int op_ret   = -1;
-        int op_errno = 0;
+int
+afr_zerofill_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
-        VALIDATE_OR_GOTO (frame, out);
-        VALIDATE_OR_GOTO (this, out);
-        VALIDATE_OR_GOTO (this->private, out);
-        VALIDATE_OR_GOTO (loc, out);
+        local = frame->local;
+        priv = this->private;
+
+	STACK_WIND_COOKIE (frame, afr_zerofill_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->zerofill,
+			   local->fd, local->cont.zerofill.offset,
+			   local->cont.zerofill.len, local->xdata_req);
+        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 = ENOMEM;
 
         priv = this->private;
 
-        transaction_frame = copy_frame (frame);
-        if (!transaction_frame) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
-                goto out;
-        }
+        QUORUM_CHECK(discard, out);
 
-        ALLOC_OR_GOTO (local, afr_local_t, out);
+	transaction_frame = copy_frame (frame);
+	if (!transaction_frame)
+		goto out;
 
-        ret = AFR_LOCAL_INIT (local, priv);
-        if (ret < 0) {
-                op_errno = -ret;
-                goto out;
-        }
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-        transaction_frame->local = local;
+        local->cont.zerofill.offset  = offset;
+        local->cont.zerofill.len = len;
 
-        local->op_ret = -1;
+        local->fd = fd_ref (fd);
+	local->inode = inode_ref (fd->inode);
 
-        local->cont.removexattr.name = gf_strdup (name);
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-        local->transaction.fop    = afr_removexattr_wind;
-        local->transaction.done   = afr_removexattr_done;
-        local->transaction.unwind = afr_removexattr_unwind;
+	if (!local->xdata_req)
+		goto out;
 
-        loc_copy (&local->loc, loc);
+        local->op = GF_FOP_ZEROFILL;
+
+        local->transaction.wind   = afr_zerofill_wind;
+        local->transaction.fop    = __afr_txn_write_fop;
+        local->transaction.done   = __afr_txn_write_done;
+        local->transaction.unwind = afr_zerofill_unwind;
 
         local->transaction.main_frame = frame;
-        local->transaction.start   = LLONG_MAX - 1;
-        local->transaction.len     = 0;
 
-        afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        local->transaction.start   = local->cont.discard.offset;
+        local->transaction.len     = len;
 
-        op_ret = 0;
-out:
-        if (op_ret == -1) {
-                if (transaction_frame)
-                        AFR_STACK_DESTROY (transaction_frame);
-                AFR_STACK_UNWIND (removexattr, frame, op_ret, op_errno);
+        afr_fix_open (fd, this);
+
+        ret = afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+        if (ret < 0) {
+		op_errno = -ret;
+		goto out;
         }
 
+	return 0;
+out:
+	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 475898722..7b1fc5528 100644
--- a/xlators/cluster/afr/src/afr-inode-write.h
+++ b/xlators/cluster/afr/src/afr-inode-write.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __INODE_WRITE_H__
@@ -22,51 +13,70 @@
 
 int32_t
 afr_chmod (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, mode_t mode);
+	   loc_t *loc, mode_t mode, dict_t *xdata);
 
 int32_t
 afr_chown (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, uid_t uid, gid_t gid);
+	   loc_t *loc, uid_t uid, gid_t gid, dict_t *xdata);
 
 int
 afr_fchown (call_frame_t *frame, xlator_t *this,
-	    fd_t *fd, uid_t uid, gid_t gid);
+	    fd_t *fd, uid_t uid, gid_t gid, dict_t *xdata);
 
 int32_t
 afr_fchmod (call_frame_t *frame, xlator_t *this,
-	    fd_t *fd, mode_t mode);
+	    fd_t *fd, mode_t mode, dict_t *xdata);
 
 int32_t
-afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, 
+afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
 	    struct iovec *vector, int32_t count, off_t offset,
-            struct iobref *iobref);
+            uint32_t flags, struct iobref *iobref, dict_t *xdata);
 
 int32_t
 afr_truncate (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, off_t offset);
+	      loc_t *loc, off_t offset, dict_t *xdata);
 
 int32_t
 afr_ftruncate (call_frame_t *frame, xlator_t *this,
-	       fd_t *fd, off_t offset);
+	       fd_t *fd, off_t offset, dict_t *xdata);
 
 int32_t
 afr_utimens (call_frame_t *frame, xlator_t *this,
-	     loc_t *loc, struct timespec tv[2]);
+	     loc_t *loc, struct timespec tv[2], dict_t *xdata);
 
 int
 afr_setattr (call_frame_t *frame, xlator_t *this,
-             loc_t *loc, struct iatt *buf, int32_t valid);
+             loc_t *loc, struct iatt *buf, int32_t valid, dict_t *xdata);
 
 int
 afr_fsetattr (call_frame_t *frame, xlator_t *this,
-              fd_t *fd, struct iatt *buf, int32_t valid);
+              fd_t *fd, struct iatt *buf, int32_t valid, dict_t *xdata);
 
 int32_t
 afr_setxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, dict_t *dict, int32_t flags);
+              loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata);
+
+int32_t
+afr_fsetxattr (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, dict_t *dict, int32_t flags, dict_t *xdata);
 
 int32_t
 afr_removexattr (call_frame_t *frame, xlator_t *this,
-		 loc_t *loc, const char *name);
+		 loc_t *loc, const char *name, dict_t *xdata);
+
+int32_t
+afr_fremovexattr (call_frame_t *frame, xlator_t *this,
+                  fd_t *fd, const char *name, dict_t *xdata);
 
+int
+afr_discard (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+             size_t len, dict_t *xdata);
+
+int
+afr_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+               off_t offset, size_t len, dict_t *xdata);
+
+int
+afr_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+             off_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 c8b0b6600..a2a758f35 100644
--- a/xlators/cluster/afr/src/afr-lk-common.c
+++ b/xlators/cluster/afr/src/afr-lk-common.c
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include "dict.h"
@@ -31,8 +22,69 @@
 #define LOCKED_YES      0x1        /* for DATA, METADATA, ENTRY and higher_path */
 #define LOCKED_LOWER    0x2        /* for lower path */
 
+#define AFR_TRACE_INODELK_IN(frame, this, params ...)           \
+        do {                                                    \
+                afr_private_t *_priv = this->private;           \
+                if (!_priv->inodelk_trace)                      \
+                        break;                                  \
+                afr_trace_inodelk_in (frame, this, params);     \
+        } while (0);
+
+#define AFR_TRACE_INODELK_OUT(frame, this, params ...)          \
+        do {                                                    \
+                afr_private_t *_priv = this->private;           \
+                if (!_priv->inodelk_trace)                      \
+                        break;                                  \
+                afr_trace_inodelk_out (frame, this, params);    \
+        } while (0);
+
+#define AFR_TRACE_ENTRYLK_IN(frame, this, params ...)           \
+        do {                                                    \
+                afr_private_t *_priv = this->private;           \
+                if (!_priv->entrylk_trace)                      \
+                        break;                                  \
+                afr_trace_entrylk_in (frame, this, params);     \
+        } while (0);
+
+#define AFR_TRACE_ENTRYLK_OUT(frame, this, params ...)          \
+        do {                                                    \
+                afr_private_t *_priv = this->private;           \
+                if (!_priv->entrylk_trace)                      \
+                        break;                                  \
+                afr_trace_entrylk_out (frame, this, params);    \
+        } while (0);
+
 int
-afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index);
+afr_entry_lockee_cmp (const void *l1, const void *l2)
+{
+        const afr_entry_lockee_t       *r1 = l1;
+        const afr_entry_lockee_t       *r2 = l2;
+        int                            ret = 0;
+        uuid_t                         gfid1 = {0};
+        uuid_t                         gfid2 = {0};
+
+        loc_gfid ((loc_t*)&r1->loc, gfid1);
+        loc_gfid ((loc_t*)&r2->loc, gfid2);
+        ret = uuid_compare (gfid1, gfid2);
+        /*Entrylks with NULL basename are the 'smallest'*/
+        if (ret == 0) {
+                if (!r1->basename)
+                        return -1;
+                if (!r2->basename)
+                        return 1;
+                ret = strcmp (r1->basename, r2->basename);
+        }
+
+        if (ret <= 0)
+                return -1;
+        else
+                return 1;
+}
+
+int afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index);
+
+static int
+afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this);
 
 static uint64_t afr_lock_number = 1;
 
@@ -57,14 +109,13 @@ afr_set_lock_number (call_frame_t *frame, xlator_t *this)
 }
 
 void
-afr_set_lk_owner (call_frame_t *frame, xlator_t *this)
+afr_set_lk_owner (call_frame_t *frame, xlator_t *this, void *lk_owner)
 {
-        if (!frame->root->lk_owner) {
-                gf_log (this->name, GF_LOG_TRACE,
-                        "Setting lk-owner=%llu",
-                        (unsigned long long) (unsigned long)frame->root);
-                frame->root->lk_owner = (uint64_t) (unsigned long)frame->root;
-        }
+        gf_log (this->name, GF_LOG_TRACE,
+                "Setting lk-owner=%llu",
+                (unsigned long long) (unsigned long)lk_owner);
+
+        set_lk_owner_from_ptr (&frame->root->lk_owner, lk_owner);
 }
 
 static int
@@ -90,29 +141,19 @@ is_afr_lock_selfheal (afr_local_t *local)
 }
 
 int32_t
-internal_lock_count (call_frame_t *frame, xlator_t *this,
-                     afr_fd_ctx_t *fd_ctx)
+internal_lock_count (call_frame_t *frame, xlator_t *this)
 {
         afr_local_t   *local = NULL;
         afr_private_t *priv  = NULL;
-
         int32_t call_count = 0;
         int i = 0;
 
         local = frame->local;
         priv  = this->private;
 
-        if (fd_ctx) {
-                GF_ASSERT (local->fd);
-                for (i = 0; i < priv->child_count; i++) {
-                        if (local->child_up[i] && fd_ctx->opened_on[i])
-                                ++call_count;
-                }
-        } else {
-                for (i = 0; i < priv->child_count; i++) {
-                        if (local->child_up[i])
-                                ++call_count;
-                }
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i])
+                        ++call_count;
         }
 
         return call_count;
@@ -120,7 +161,7 @@ internal_lock_count (call_frame_t *frame, xlator_t *this,
 
 static void
 afr_print_inodelk (char *str, int size, int cmd,
-                   struct gf_flock *flock, uint64_t owner)
+                   struct gf_flock *flock, gf_lkowner_t *owner)
 {
         char *cmd_str = NULL;
         char *type_str = NULL;
@@ -168,11 +209,11 @@ afr_print_inodelk (char *str, int size, int cmd,
         }
 
         snprintf (str, size, "lock=INODELK, cmd=%s, type=%s, "
-                  "start=%llu, len=%llu, pid=%llu, lk-owner=%llu",
+                  "start=%llu, len=%llu, pid=%llu, lk-owner=%s",
                   cmd_str, type_str, (unsigned long long) flock->l_start,
                   (unsigned long long) flock->l_len,
                   (unsigned long long) flock->l_pid,
-                  (unsigned long long) owner);
+                  lkowner_utoa (owner));
 
 }
 
@@ -188,11 +229,11 @@ afr_print_lockee (char *str, int size, loc_t *loc, fd_t *fd,
 
 void
 afr_print_entrylk (char *str, int size, const char *basename,
-                   uint64_t owner)
+                   gf_lkowner_t *owner)
 {
-        snprintf (str, size, "Basename=%s, lk-owner=%llu",
+        snprintf (str, size, "Basename=%s, lk-owner=%s",
                   basename ? basename : "<nul>",
-                  (unsigned long long)owner);
+                  lkowner_utoa (owner));
 }
 
 static void
@@ -246,27 +287,20 @@ afr_set_lock_call_type (afr_lock_call_type_t lock_call_type,
 }
 
 static void
-afr_trace_inodelk_out (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
+afr_trace_inodelk_out (call_frame_t *frame, xlator_t *this,
+                       afr_lock_call_type_t lock_call_type,
                        afr_lock_op_type_t lk_op_type, struct gf_flock *flock,
                        int op_ret, int op_errno, int32_t child_index)
 {
-        xlator_t            *this     = NULL;
         afr_internal_lock_t *int_lock = NULL;
         afr_local_t         *local    = NULL;
-        afr_private_t       *priv     = NULL;
 
         char lockee[256];
         char lock_call_type_str[256];
         char verdict[16];
 
-        this     = THIS;
         local    = frame->local;
         int_lock = &local->internal_lock;
-        priv     = this->private;
-
-        if (!priv->inodelk_trace) {
-                return;
-        }
 
         afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
 
@@ -274,45 +308,37 @@ afr_trace_inodelk_out (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
 
         afr_print_verdict (op_ret, op_errno, verdict);
 
-        gf_log (this->name, GF_LOG_NORMAL,
-                "[%s %s] [%s] Lockee={%s} Number={%llu}",
+        gf_log (this->name, GF_LOG_INFO,
+                "[%s %s] [%s] lk-owner=%s Lockee={%s} Number={%llu}",
                 lock_call_type_str,
                 lk_op_type == AFR_LOCK_OP ? "LOCK REPLY" : "UNLOCK REPLY",
-                verdict,
-                lockee,
+                verdict, lkowner_utoa (&frame->root->lk_owner), lockee,
                 (unsigned long long) int_lock->lock_number);
 
 }
 
 static void
-afr_trace_inodelk_in (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
+afr_trace_inodelk_in (call_frame_t *frame, xlator_t *this,
+                      afr_lock_call_type_t lock_call_type,
                       afr_lock_op_type_t lk_op_type, struct gf_flock *flock,
                       int32_t cmd, int32_t child_index)
 {
-        xlator_t            *this     = NULL;
         afr_local_t         *local    = NULL;
         afr_internal_lock_t *int_lock = NULL;
-        afr_private_t       *priv     = NULL;
 
         char lock[256];
         char lockee[256];
         char lock_call_type_str[256];
 
-        this     = THIS;
         local    = frame->local;
         int_lock = &local->internal_lock;
-        priv     = this->private;
 
-        if (!priv->inodelk_trace) {
-                return;
-        }
-
-        afr_print_inodelk (lock, 256, cmd, flock, frame->root->lk_owner);
+        afr_print_inodelk (lock, 256, cmd, flock, &frame->root->lk_owner);
         afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
 
         afr_set_lock_call_type (lock_call_type, lock_call_type_str, int_lock);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[%s %s] Lock={%s} Lockee={%s} Number={%llu}",
                 lock_call_type_str,
                 lk_op_type == AFR_LOCK_OP ? "LOCK REQUEST" : "UNLOCK REQUEST",
@@ -322,20 +348,21 @@ afr_trace_inodelk_in (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
 }
 
 static void
-afr_trace_entrylk_in (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
+afr_trace_entrylk_in (call_frame_t *frame, xlator_t *this,
+                      afr_lock_call_type_t lock_call_type,
                       afr_lock_op_type_t lk_op_type, const char *basename,
-                      int32_t child_index)
+                      int32_t cookie)
 {
-        xlator_t            *this     = NULL;
         afr_local_t         *local    = NULL;
         afr_internal_lock_t *int_lock = NULL;
         afr_private_t       *priv     = NULL;
+        int                 child_index = 0;
+        int                 lockee_no = 0;
 
         char lock[256];
         char lockee[256];
         char lock_call_type_str[256];
 
-        this     = THIS;
         local    = frame->local;
         int_lock = &local->internal_lock;
         priv     = this->private;
@@ -343,36 +370,41 @@ afr_trace_entrylk_in (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
         if (!priv->entrylk_trace) {
                 return;
         }
+        lockee_no = cookie / priv->child_count;
+        child_index = cookie % priv->child_count;
 
-        afr_print_entrylk (lock, 256, basename, frame->root->lk_owner);
-        afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
+        afr_print_entrylk (lock, 256, basename, &frame->root->lk_owner);
+        afr_print_lockee (lockee, 256, &int_lock->lockee[lockee_no].loc, local->fd,
+                          child_index);
 
         afr_set_lock_call_type (lock_call_type, lock_call_type_str, int_lock);
 
-        gf_log (this->name, GF_LOG_NORMAL,
-                "[%s %s] Lock={%s} Lockee={%s} Number={%llu}",
+        gf_log (this->name, GF_LOG_INFO,
+                "[%s %s] Lock={%s} Lockee={%s} Number={%llu}, Cookie={%d}",
                 lock_call_type_str,
                 lk_op_type == AFR_LOCK_OP ? "LOCK REQUEST" : "UNLOCK REQUEST",
                 lock, lockee,
-                (unsigned long long) int_lock->lock_number);
+                (unsigned long long) int_lock->lock_number,
+                cookie);
 }
 
 static void
-afr_trace_entrylk_out (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
-                       afr_lock_op_type_t lk_op_type, const char *basename, int op_ret,
-                       int op_errno, int32_t child_index)
+afr_trace_entrylk_out (call_frame_t *frame, xlator_t *this,
+                       afr_lock_call_type_t lock_call_type,
+                       afr_lock_op_type_t lk_op_type, const char *basename,
+                       int op_ret, int op_errno, int32_t cookie)
 {
-        xlator_t            *this     = NULL;
         afr_internal_lock_t *int_lock = NULL;
         afr_local_t         *local    = NULL;
         afr_private_t       *priv     = NULL;
+        int                 lockee_no = 0;
+        int                 child_index = 0;
 
         char lock[256];
         char lockee[256];
         char lock_call_type_str[256];
         char verdict[16];
 
-        this     = THIS;
         local    = frame->local;
         int_lock = &local->internal_lock;
         priv     = this->private;
@@ -380,20 +412,25 @@ afr_trace_entrylk_out (call_frame_t *frame, afr_lock_call_type_t lock_call_type,
         if (!priv->entrylk_trace) {
                 return;
         }
+        lockee_no = cookie / priv->child_count;
+        child_index = cookie % priv->child_count;
 
-        afr_print_lockee (lockee, 256, &local->loc, local->fd, child_index);
+        afr_print_entrylk (lock, 256, basename, &frame->root->lk_owner);
+        afr_print_lockee (lockee, 256, &int_lock->lockee[lockee_no].loc, local->fd,
+                          child_index);
 
         afr_set_lock_call_type (lock_call_type, lock_call_type_str, int_lock);
 
         afr_print_verdict (op_ret, op_errno, verdict);
 
-        gf_log (this->name, GF_LOG_NORMAL,
-                "[%s %s] [%s] Lock={%s} Lockee={%s} Number={%llu}",
+        gf_log (this->name, GF_LOG_INFO,
+                "[%s %s] [%s] Lock={%s} Lockee={%s} Number={%llu} Cookie={%d}",
                 lock_call_type_str,
                 lk_op_type == AFR_LOCK_OP ? "LOCK REPLY" : "UNLOCK REPLY",
                 verdict,
                 lock, lockee,
-                (unsigned long long) int_lock->lock_number);
+                (unsigned long long) int_lock->lock_number,
+                cookie);
 
 }
 
@@ -431,8 +468,8 @@ is_afr_lock_transaction (afr_local_t *local)
         int ret = 0;
 
         switch (local->transaction.type) {
-	case AFR_DATA_TRANSACTION:
-	case AFR_METADATA_TRANSACTION:
+        case AFR_DATA_TRANSACTION:
+        case AFR_METADATA_TRANSACTION:
                 ret = 1;
                 break;
 
@@ -446,6 +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)
 {
@@ -463,8 +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;
@@ -476,45 +559,44 @@ 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;
 }
 
-loc_t *
-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 ret = 0;
-
-	ret = strcmp (l1->path, l2->path);
+        int call_count  = 0;
+        int i           = 0;
 
-	if (ret == 0)
-		ret = strcmp (b1, b2);
+        for (i = 0; i < int_lock->lockee_count; i++)
+                call_count += int_lock->lockee[i].locked_count;
 
-	if (ret <= 0)
-		return l1;
-	else
-		return l2;
+        return call_count;
 }
 
 int
 afr_locked_nodes_count (unsigned char *locked_nodes, int child_count)
 
 {
-        int i;
+        int i = 0;
         int call_count = 0;
 
         for (i = 0; i < child_count; i++) {
@@ -528,44 +610,63 @@ afr_locked_nodes_count (unsigned char *locked_nodes, int child_count)
 /* FIXME: What if UNLOCK fails */
 static int32_t
 afr_unlock_common_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                       int32_t op_ret, int32_t op_errno)
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	afr_local_t         *local    = NULL;
+        afr_local_t         *local    = NULL;
         afr_internal_lock_t *int_lock = NULL;
-	int call_count = 0;
+        int call_count = 0;
 
-	local    = frame->local;
+        local    = frame->local;
         int_lock = &local->internal_lock;
 
-	LOCK (&frame->lock);
-	{
-		call_count = --int_lock->lk_call_count;
-	}
-	UNLOCK (&frame->lock);
+        LOCK (&frame->lock);
+        {
+                call_count = --int_lock->lk_call_count;
+        }
+        UNLOCK (&frame->lock);
 
-	if (call_count == 0) {
+        if (call_count == 0) {
                 gf_log (this->name, GF_LOG_TRACE,
                         "All internal locks unlocked");
                 int_lock->lock_cbk (frame, this);
         }
 
-	return 0;
+        return 0;
 }
 
 static int32_t
 afr_unlock_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int32_t op_ret, int32_t op_errno)
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        afr_trace_inodelk_out (frame, AFR_INODELK_TRANSACTION,
+        afr_local_t         *local = NULL;
+        afr_internal_lock_t *int_lock = NULL;
+        afr_inodelk_t       *inodelk = NULL;
+        int32_t             child_index = (long)cookie;
+        afr_private_t       *priv = NULL;
+
+        local = frame->local;
+        int_lock = &local->internal_lock;
+
+        AFR_TRACE_INODELK_OUT (frame, this, AFR_INODELK_TRANSACTION,
                                AFR_UNLOCK_OP, NULL, op_ret,
-                               op_errno, (long) cookie);
+                               op_errno, child_index);
+
+        priv = this->private;
 
         if (op_ret < 0 && op_errno != ENOTCONN && op_errno != EBADFD) {
-                gf_log (this->name, GF_LOG_TRACE,
-                        "Unlock failed for some reason");
+                gf_log (this->name, GF_LOG_INFO, "%s: unlock failed on subvolume %s "
+                        "with lock owner %s", local->loc.path,
+                        priv->children[child_index]->name,
+                        lkowner_utoa (&frame->root->lk_owner));
         }
 
-        afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno);
+
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+        inodelk->locked_nodes[child_index] &= LOCKED_NO;
+        if (local->transaction.eager_lock)
+                local->transaction.eager_lock[child_index] = 0;
+
+        afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno, xdata);
 
         return 0;
 
@@ -575,22 +676,30 @@ static int
 afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
 {
         afr_internal_lock_t *int_lock = NULL;
+        afr_inodelk_t       *inodelk  = NULL;
         afr_local_t         *local    = NULL;
         afr_private_t       *priv     = NULL;
-
-        struct gf_flock flock;
+        struct gf_flock flock = {0,};
+        struct gf_flock full_flock = {0,};
+        struct gf_flock *flock_use = NULL;
         int call_count = 0;
         int i = 0;
+        int piggyback = 0;
+        afr_fd_ctx_t        *fd_ctx      = NULL;
+
 
         local    = frame->local;
         int_lock = &local->internal_lock;
         priv     = this->private;
 
-        flock.l_start = int_lock->lk_flock.l_start;
-        flock.l_len   = int_lock->lk_flock.l_len;
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+
+        flock.l_start = inodelk->flock.l_start;
+        flock.l_len   = inodelk->flock.l_len;
         flock.l_type  = F_UNLCK;
 
-        call_count = afr_locked_nodes_count (int_lock->inode_locked_nodes,
+        full_flock.l_type = F_UNLCK;
+        call_count = afr_locked_nodes_count (inodelk->locked_nodes,
                                              priv->child_count);
 
         int_lock->lk_call_count = call_count;
@@ -602,55 +711,107 @@ afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
                 goto out;
         }
 
+        if (local->fd)
+                fd_ctx = afr_fd_ctx_get (local->fd, this);
+
         for (i = 0; i < priv->child_count; i++) {
-                if (int_lock->inode_locked_nodes[i] & LOCKED_YES) {
-                        if (local->fd) {
-                                afr_trace_inodelk_in (frame, AFR_INODELK_TRANSACTION,
-                                                      AFR_UNLOCK_OP, &flock, F_SETLK, i);
-
-                                STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
-                                                   (void *) (long)i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->finodelk,
-                                                   this->name, local->fd,
-                                                   F_SETLK, &flock);
+                if ((inodelk->locked_nodes[i] & LOCKED_YES) != LOCKED_YES)
+                        continue;
 
-                                if (!--call_count)
-                                        break;
+                if (local->fd) {
+                        flock_use = &flock;
+                        if (!local->transaction.eager_lock[i]) {
+                                goto wind;
+                        }
 
-                        } else {
-                                afr_trace_inodelk_in (frame, AFR_INODELK_TRANSACTION,
-                                                      AFR_UNLOCK_OP, &flock, F_SETLK, i);
+                        piggyback = 0;
 
-                                STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
-                                                   (void *) (long)i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->inodelk,
-                                                   this->name, &local->loc,
-                                                   F_SETLK, &flock);
+                        LOCK (&local->fd->lock);
+                        {
+                                if (fd_ctx->lock_piggyback[i]) {
+                                        fd_ctx->lock_piggyback[i]--;
+                                        piggyback = 1;
+                                } else {
+                                        fd_ctx->lock_acquired[i]--;
+                                }
+                        }
+                        UNLOCK (&local->fd->lock);
 
+                        if (piggyback) {
+                                afr_unlock_inodelk_cbk (frame, (void *) (long) i,
+                                                        this, 1, 0, NULL);
                                 if (!--call_count)
                                         break;
-
+                                continue;
                         }
 
-                }
+                        flock_use = &full_flock;
+                wind:
+                        AFR_TRACE_INODELK_IN (frame, this,
+                                              AFR_INODELK_TRANSACTION,
+                                              AFR_UNLOCK_OP, flock_use, F_SETLK,
+                                              i);
 
-        }
+                        STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
+                                           (void *) (long)i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->finodelk,
+                                           int_lock->domain, local->fd,
+                                           F_SETLK, flock_use, NULL);
+
+                        if (!--call_count)
+                                break;
+
+                } else {
+                        AFR_TRACE_INODELK_IN (frame, this,
+                                              AFR_INODELK_TRANSACTION,
+                                              AFR_UNLOCK_OP, &flock, F_SETLK, i);
 
+                        STACK_WIND_COOKIE (frame, afr_unlock_inodelk_cbk,
+                                           (void *) (long)i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->inodelk,
+                                           int_lock->domain, &local->loc,
+                                           F_SETLK, &flock, NULL);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
 out:
         return 0;
 }
 
 static int32_t
 afr_unlock_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int32_t op_ret, int32_t op_errno)
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        afr_trace_entrylk_out (frame, AFR_ENTRYLK_TRANSACTION,
-                              AFR_UNLOCK_OP, NULL, op_ret,
-                               op_errno, (long) cookie);
+        afr_local_t         *local = NULL;
+        afr_private_t       *priv  = NULL;
+        afr_internal_lock_t *int_lock = NULL;
+        int32_t             child_index = 0;
+        int                 lockee_no   = 0;
+
+        priv = this->private;
+        lockee_no = (int)((long) cookie) / priv->child_count;
+        child_index = (int) ((long) cookie) % priv->child_count;
+
+        local = frame->local;
+        int_lock = &local->internal_lock;
 
-        afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno);
+        AFR_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
+                               AFR_UNLOCK_OP,
+                               int_lock->lockee[lockee_no].basename, op_ret,
+                               op_errno, (int) ((long)cookie));
+
+        if (op_ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: unlock failed on %d, reason: %s",
+                        local->loc.path, child_index, strerror (op_errno));
+        }
+
+        int_lock->lockee[lockee_no].locked_nodes[child_index] &= LOCKED_NO;
+        afr_unlock_common_cbk (frame, cookie, this, op_ret, op_errno, NULL);
 
         return 0;
 }
@@ -658,25 +819,22 @@ 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){
@@ -686,18 +844,23 @@ afr_unlock_entrylk (call_frame_t *frame, xlator_t *this)
                 goto out;
         }
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (int_lock->entry_locked_nodes[i] & LOCKED_YES) {
-                        afr_trace_entrylk_in (frame, AFR_ENTRYLK_NB_TRANSACTION,
-                                              AFR_UNLOCK_OP, basename, i);
+        for (i = 0; i < int_lock->lockee_count * priv->child_count; i++) {
+                lockee_no = i / copies;
+                index     = i % copies;
+                if (int_lock->lockee[lockee_no].locked_nodes[index] & LOCKED_YES) {
+                        AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_NB_TRANSACTION,
+                                              AFR_UNLOCK_OP,
+                                              int_lock->lockee[lockee_no].basename,
+                                              i);
 
                         STACK_WIND_COOKIE (frame, afr_unlock_entrylk_cbk,
                                            (void *) (long) i,
-                                           priv->children[i],
-                                           priv->children[i]->fops->entrylk,
-                                           this->name,
-                                           loc, basename,
-                                           ENTRYLK_UNLOCK, ENTRYLK_WRLCK);
+                                           priv->children[index],
+                                           priv->children[index]->fops->entrylk,
+                                           int_lock->domain,
+                                           &int_lock->lockee[lockee_no].loc,
+                                           int_lock->lockee[lockee_no].basename,
+                                           ENTRYLK_UNLOCK, ENTRYLK_WRLCK, NULL);
 
                         if (!--call_count)
                                 break;
@@ -711,155 +874,85 @@ out:
 
 static int32_t
 afr_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	      int32_t op_ret, int32_t op_errno)
+              int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        afr_internal_lock_t *int_lock = NULL;
-	afr_local_t         *local    = NULL;
-	afr_private_t       *priv     = NULL;
-
-	int done = 0;
-	int child_index = (long) cookie;
-
-	local    = frame->local;
-        int_lock = &local->internal_lock;
-	priv     = this->private;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			if (op_errno == ENOSYS) {
-				/* return ENOTSUP */
-				gf_log (this->name, GF_LOG_ERROR,
-					"subvolume does not support locking. "
-					"please load features/posix-locks xlator on server");
-				local->op_ret = op_ret;
-                                int_lock->lock_op_ret = op_ret;
-				done = 1;
-			}
-
-			local->child_up[child_index] = 0;
-			local->op_errno              = op_errno;
-                        int_lock->lock_op_errno      = op_errno;
-		}
-	}
-	UNLOCK (&frame->lock);
-
-        if ((op_ret == -1) &&
-            (op_errno == ENOSYS)) {
-                afr_unlock (frame, this);
-        } else {
-                if (op_ret == 0) {
-                        int_lock->locked_nodes[child_index]
-                                |= LOCKED_YES;
-                        int_lock->lock_count++;
-                }
-                afr_lock_blocking (frame, this, child_index + 1);
-        }
-
-	return 0;
-}
-
-static int32_t
-afr_blocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                          int32_t op_ret, int32_t op_errno)
-{
-        afr_trace_inodelk_out (frame, AFR_INODELK_TRANSACTION,
-                               AFR_LOCK_OP, NULL, op_ret,
-                               op_errno, (long) cookie);
-
-        afr_lock_cbk (frame, cookie, this, op_ret, op_errno);
-        return 0;
-
-}
-
-static int32_t
-afr_lock_lower_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno)
-{
-        afr_internal_lock_t *int_lock    = NULL;
-        afr_private_t       *priv        = NULL;
-        afr_local_t         *local       = NULL;
-        loc_t               *lower       = NULL;
-        loc_t               *higher      = NULL;
-        const char          *lower_name  = NULL;
-        const char          *higher_name = NULL;
-
-        int child_index = (long) cookie;
+        afr_internal_lock_t     *int_lock       = NULL;
+        afr_local_t             *local          = NULL;
+        afr_private_t           *priv           = NULL;
+        int                     cky             = (long) cookie;
+        int                     child_index     = 0;
+        int                     lockee_no       = 0;
 
         priv     = this->private;
         local    = frame->local;
         int_lock = &local->internal_lock;
 
+        child_index = ((int)cky) % priv->child_count;
+        lockee_no   = ((int)cky) / priv->child_count;
+
         LOCK (&frame->lock);
         {
                 if (op_ret == -1) {
                         if (op_errno == ENOSYS) {
                                 /* return ENOTSUP */
-
                                 gf_log (this->name, GF_LOG_ERROR,
                                         "subvolume does not support locking. "
-                                        "please load features/posix-locks xlator on server");
-
-                                local->op_ret   = op_ret;
+                                        "please load features/locks xlator on server");
+                                local->op_ret = op_ret;
+                                int_lock->lock_op_ret = op_ret;
                         }
 
-                        local->child_up[child_index] = 0;
-                        local->op_errno = op_errno;
+                        local->op_errno              = op_errno;
+                        int_lock->lock_op_errno      = op_errno;
                 }
+
+		int_lock->lk_attempted_count++;
         }
         UNLOCK (&frame->lock);
 
-        if (op_ret != 0) {
+        if ((op_ret == -1) &&
+            (op_errno == ENOSYS)) {
                 afr_unlock (frame, this);
-                goto out;
         } else {
-                int_lock->lower_locked_nodes[child_index] |= LOCKED_LOWER;
-                int_lock->lock_count++;
+                if (op_ret == 0) {
+                        if (local->transaction.type == AFR_ENTRY_TRANSACTION ||
+                            local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
+                                int_lock->lockee[lockee_no].locked_nodes[child_index] |= LOCKED_YES;
+                                int_lock->lockee[lockee_no].locked_count++;
+                                int_lock->entrylk_lock_count++;
+                        } else {
+                                int_lock->locked_nodes[child_index] |= LOCKED_YES;
+                                int_lock->lock_count++;
+                        }
+                }
+                afr_lock_blocking (frame, this, cky + 1);
         }
 
-        /* The lower path has been locked. Now lock the higher path */
-
-        lower = lower_path (&local->transaction.parent_loc,
-                            local->transaction.basename,
-                            &local->transaction.new_parent_loc,
-                            local->transaction.new_basename);
-
-        lower_name = (lower == &local->transaction.parent_loc ?
-                      local->transaction.basename :
-                      local->transaction.new_basename);
-
-        higher = (lower == &local->transaction.parent_loc ?
-                  &local->transaction.new_parent_loc :
-                  &local->transaction.parent_loc);
-
-        higher_name = (higher == &local->transaction.parent_loc ?
-                       local->transaction.basename :
-                       local->transaction.new_basename);
-
-        afr_trace_entrylk_in (frame, AFR_ENTRYLK_TRANSACTION,
-                              AFR_LOCK_OP, higher_name, child_index);
-
+        return 0;
+}
 
-        STACK_WIND_COOKIE (frame, afr_lock_cbk,
-                           (void *) (long) child_index,
-                           priv->children[child_index],
-                           priv->children[child_index]->fops->entrylk,
-                           this->name, higher, higher_name,
-                           ENTRYLK_LOCK, ENTRYLK_WRLCK);
+static int32_t
+afr_blocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                          int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        AFR_TRACE_INODELK_OUT (frame, this, AFR_INODELK_TRANSACTION,
+                               AFR_LOCK_OP, NULL, op_ret,
+                               op_errno, (long) cookie);
 
-out:
+        afr_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
         return 0;
+
 }
 
 static int32_t
 afr_blocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                          int32_t op_ret, int32_t op_errno)
+                          int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        afr_trace_entrylk_out (frame, AFR_ENTRYLK_TRANSACTION,
+        AFR_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
                                AFR_LOCK_OP, NULL, op_ret,
                                op_errno, (long)cookie);
 
-        afr_lock_cbk (frame, cookie, this, op_ret, op_errno);
+        afr_lock_cbk (frame, cookie, this, op_ret, op_errno, xdata);
         return 0;
 }
 
@@ -867,6 +960,7 @@ 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;
 
@@ -875,20 +969,18 @@ afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this)
         int_lock = &local->internal_lock;
 
         switch (local->transaction.type) {
-	case AFR_DATA_TRANSACTION:
-	case AFR_METADATA_TRANSACTION:
-                memcpy (int_lock->inode_locked_nodes,
-                        int_lock->locked_nodes,
-                        priv->child_count);
-                int_lock->inodelk_lock_count = int_lock->lock_count;
+        case AFR_DATA_TRANSACTION:
+        case AFR_METADATA_TRANSACTION:
+                inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+                memcpy (inodelk->locked_nodes, int_lock->locked_nodes,
+                        sizeof (*inodelk->locked_nodes) * priv->child_count);
+                inodelk->lock_count = int_lock->lock_count;
                 break;
 
         case AFR_ENTRY_RENAME_TRANSACTION:
         case AFR_ENTRY_TRANSACTION:
-                memcpy (int_lock->entry_locked_nodes,
-                        int_lock->locked_nodes,
-                        priv->child_count);
-                int_lock->entrylk_lock_count = int_lock->lock_count;
+                /*entrylk_count is being used in both non-blocking and blocking
+                 * modes */
                 break;
         }
 
@@ -896,42 +988,78 @@ afr_copy_locked_nodes (call_frame_t *frame, xlator_t *this)
 
 }
 
+static inline gf_boolean_t
+afr_is_entrylk (afr_internal_lock_t *int_lock,
+                afr_transaction_type trans_type)
+{
+        gf_boolean_t is_entrylk = _gf_false;
+
+        if ((int_lock->transaction_lk_type == AFR_SELFHEAL_LK) &&
+            int_lock->selfheal_lk_type == AFR_ENTRY_SELF_HEAL_LK) {
+
+                is_entrylk = _gf_true;
+
+        } else if ((int_lock->transaction_lk_type == AFR_TRANSACTION_LK) &&
+                 (trans_type == AFR_ENTRY_TRANSACTION ||
+                  trans_type == AFR_ENTRY_RENAME_TRANSACTION)) {
+
+                is_entrylk = _gf_true;
+
+        } else {
+                is_entrylk = _gf_false;
+        }
+
+        return is_entrylk;
+}
+
+static gf_boolean_t
+_is_lock_wind_needed (afr_local_t *local, int child_index)
+{
+        if (!local->child_up[child_index])
+                return _gf_false;
+
+        return _gf_true;
+}
+
 int
-afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
+afr_lock_blocking (call_frame_t *frame, xlator_t *this, int cookie)
 {
         afr_internal_lock_t *int_lock    = NULL;
+        afr_inodelk_t       *inodelk     = NULL;
         afr_local_t         *local       = NULL;
-	afr_private_t       *priv        = NULL;
-        afr_fd_ctx_t        *fd_ctx      = NULL;
-	loc_t               *lower       = NULL;
-	loc_t               *higher      = NULL;
-	const char          *lower_name  = NULL;
-	const char          *higher_name = NULL;
-
-	struct gf_flock flock;
-        uint64_t ctx;
+        afr_private_t       *priv        = NULL;
+        struct gf_flock flock = {0,};
+        uint64_t ctx = 0;
         int ret = 0;
-
-	local    = frame->local;
-        int_lock = &local->internal_lock;
-	priv     = this->private;
-
-	flock.l_start = int_lock->lk_flock.l_start;
-	flock.l_len   = int_lock->lk_flock.l_len;
-	flock.l_type  = int_lock->lk_flock.l_type;
+        int child_index = 0;
+        int lockee_no   = 0;
+        gf_boolean_t is_entrylk = _gf_false;
+
+        local         = frame->local;
+        int_lock      = &local->internal_lock;
+        priv          = this->private;
+        child_index   = cookie % priv->child_count;
+        lockee_no     = cookie / priv->child_count;
+        is_entrylk    = afr_is_entrylk (int_lock, local->transaction.type);
+
+
+        if (!is_entrylk) {
+                inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+                flock.l_start = inodelk->flock.l_start;
+                flock.l_len   = inodelk->flock.l_len;
+                flock.l_type  = inodelk->flock.l_type;
+        }
 
         if (local->fd) {
                 ret = fd_ctx_get (local->fd, this, &ctx);
 
                 if (ret < 0) {
-                        gf_log (this->name, GF_LOG_DEBUG,
+                        gf_log (this->name, GF_LOG_INFO,
                                 "unable to get fd ctx for fd=%p",
                                 local->fd);
 
                         local->op_ret           = -1;
                         int_lock->lock_op_ret   = -1;
-                        local->op_errno         = EINVAL;
-                        int_lock->lock_op_errno = EINVAL;
 
                         afr_copy_locked_nodes (frame, this);
 
@@ -939,49 +1067,27 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
 
                         return 0;
                 }
-
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                /* skip over children that or down
-                   or don't have the fd open */
-
-                while ((child_index < priv->child_count)
-                       && (!local->child_up[child_index]
-                           || !fd_ctx->opened_on[child_index]))
-
-                        child_index++;
-        } else {
-                /* skip over children that are down */
-                while ((child_index < priv->child_count)
-                       && !local->child_up[child_index])
-                        child_index++;
         }
 
-	if ((child_index == priv->child_count) &&
-	    int_lock->lock_count == 0) {
-
-		gf_log (this->name, GF_LOG_DEBUG,
-			"unable to lock on even one child");
-
-		local->op_ret           = -1;
-                int_lock->lock_op_ret   = -1;
-		local->op_errno         = EAGAIN;
-                int_lock->lock_op_errno = EAGAIN;
+        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);
-
-                afr_unlock(frame, this);
+                        local->op_ret           = -1;
+                        int_lock->lock_op_ret   = -1;
 
-		return 0;
+                        afr_copy_locked_nodes (frame, this);
 
-	}
+                        afr_unlock(frame, this);
 
-	if ((child_index == priv->child_count)
-	    || (int_lock->lock_count ==
-		afr_up_children_count (priv->child_count,
-                                       local->child_up))) {
+                        return 0;
+                }
+        }
 
-		/* we're done locking */
+        if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
+                /* we're done locking */
 
                 gf_log (this->name, GF_LOG_DEBUG,
                         "we're done locking");
@@ -990,107 +1096,85 @@ afr_lock_blocking (call_frame_t *frame, xlator_t *this, int child_index)
 
                 int_lock->lock_op_ret = 0;
                 int_lock->lock_cbk (frame, this);
-		return 0;
-	}
+                return 0;
+        }
+
+        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:
+        switch (local->transaction.type) {
+        case AFR_DATA_TRANSACTION:
+        case AFR_METADATA_TRANSACTION:
 
-		if (local->fd) {
-                        afr_trace_inodelk_in (frame, AFR_INODELK_TRANSACTION,
+                if (local->fd) {
+                        AFR_TRACE_INODELK_IN (frame, this,
+                                              AFR_INODELK_TRANSACTION,
                                               AFR_LOCK_OP, &flock, F_SETLKW,
                                               child_index);
 
-			STACK_WIND_COOKIE (frame, afr_blocking_inodelk_cbk,
-					   (void *) (long) child_index,
-					   priv->children[child_index],
-					   priv->children[child_index]->fops->finodelk,
-					   this->name, local->fd,
-                                           F_SETLKW, &flock);
+                        STACK_WIND_COOKIE (frame, afr_blocking_inodelk_cbk,
+                                           (void *) (long) child_index,
+                                           priv->children[child_index],
+                                           priv->children[child_index]->fops->finodelk,
+                                           int_lock->domain, local->fd,
+                                           F_SETLKW, &flock, NULL);
 
-		} else {
-                        afr_trace_inodelk_in (frame, AFR_INODELK_TRANSACTION,
+                } else {
+                        AFR_TRACE_INODELK_IN (frame, this,
+                                              AFR_INODELK_TRANSACTION,
                                               AFR_LOCK_OP, &flock, F_SETLKW,
                                               child_index);
 
-			STACK_WIND_COOKIE (frame, afr_blocking_inodelk_cbk,
-					   (void *) (long) child_index,
-					   priv->children[child_index],
-					   priv->children[child_index]->fops->inodelk,
-					   this->name, &local->loc,
-                                           F_SETLKW, &flock);
-		}
-
-		break;
-
-	case AFR_ENTRY_RENAME_TRANSACTION:
-	{
-		lower = lower_path (&local->transaction.parent_loc,
-				    local->transaction.basename,
-				    &local->transaction.new_parent_loc,
-				    local->transaction.new_basename);
-
-		lower_name = (lower == &local->transaction.parent_loc ?
-			      local->transaction.basename :
-			      local->transaction.new_basename);
-
-		higher = (lower == &local->transaction.parent_loc ?
-			  &local->transaction.new_parent_loc :
-			  &local->transaction.parent_loc);
-
-		higher_name = (higher == &local->transaction.parent_loc ?
-			       local->transaction.basename :
-			       local->transaction.new_basename);
-
-                afr_trace_entrylk_in (frame, AFR_ENTRYLK_TRANSACTION,
-                                      AFR_LOCK_OP, lower_name, child_index);
-
-
-		STACK_WIND_COOKIE (frame, afr_lock_lower_cbk,
-				   (void *) (long) child_index,
-				   priv->children[child_index],
-				   priv->children[child_index]->fops->entrylk,
-				   this->name, lower, lower_name,
-				   ENTRYLK_LOCK, ENTRYLK_WRLCK);
-
-		break;
-	}
+                        STACK_WIND_COOKIE (frame, afr_blocking_inodelk_cbk,
+                                           (void *) (long) child_index,
+                                           priv->children[child_index],
+                                           priv->children[child_index]->fops->inodelk,
+                                           int_lock->domain, &local->loc,
+                                           F_SETLKW, &flock, NULL);
+                }
 
-	case AFR_ENTRY_TRANSACTION:
-		if (local->fd) {
-                        afr_trace_entrylk_in (frame, AFR_ENTRYLK_TRANSACTION,
-                                              AFR_LOCK_OP, local->transaction.basename,
-                                              child_index);
+                break;
 
-			STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
-					   (void *) (long) child_index,
-					   priv->children[child_index],
-					   priv->children[child_index]->fops->fentrylk,
-					   this->name, local->fd,
-					   local->transaction.basename,
-					   ENTRYLK_LOCK, ENTRYLK_WRLCK);
-		} else {
-                        afr_trace_entrylk_in (frame, AFR_ENTRYLK_TRANSACTION,
+        case AFR_ENTRY_RENAME_TRANSACTION:
+        case AFR_ENTRY_TRANSACTION:
+                /*Accounting for child_index increments on 'down'
+                 *and 'fd-less' children */
+
+                if (local->fd) {
+                        AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_TRANSACTION,
+                                              AFR_LOCK_OP,
+                                              int_lock->lockee[lockee_no].basename,
+                                              cookie);
+
+                        STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
+                                           (void *) (long) cookie,
+                                           priv->children[child_index],
+                                           priv->children[child_index]->fops->fentrylk,
+                                           int_lock->domain, local->fd,
+                                           int_lock->lockee[lockee_no].basename,
+                                           ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
+                } else {
+                        AFR_TRACE_ENTRYLK_IN (frame, this,
+                                              AFR_ENTRYLK_TRANSACTION,
                                               AFR_LOCK_OP, local->transaction.basename,
                                               child_index);
 
-			STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
-					   (void *) (long) child_index,
-					   priv->children[child_index],
-					   priv->children[child_index]->fops->entrylk,
-					   this->name,
-                                           &local->transaction.parent_loc,
-					   local->transaction.basename,
-					   ENTRYLK_LOCK, ENTRYLK_WRLCK);
-		}
-
-		break;
-	}
-
-	return 0;
+                        STACK_WIND_COOKIE (frame, afr_blocking_entrylk_cbk,
+                                           (void *) (long) cookie,
+                                           priv->children[child_index],
+                                           priv->children[child_index]->fops->entrylk,
+                                           int_lock->domain,
+                                           &int_lock->lockee[lockee_no].loc,
+                                           int_lock->lockee[lockee_no].basename,
+                                           ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
+                }
 
+                break;
+        }
 
+        return 0;
 }
 
 int32_t
@@ -1099,19 +1183,24 @@ afr_blocking_lock (call_frame_t *frame, xlator_t *this)
         afr_internal_lock_t *int_lock = NULL;
         afr_local_t         *local    = NULL;
         afr_private_t       *priv     = NULL;
+        int                  up_count = 0;
 
         priv     = this->private;
         local    = frame->local;
         int_lock = &local->internal_lock;
 
         switch (local->transaction.type) {
-	case AFR_DATA_TRANSACTION:
-	case AFR_METADATA_TRANSACTION:
+        case AFR_DATA_TRANSACTION:
+        case AFR_METADATA_TRANSACTION:
                 initialize_inodelk_variables (frame, this);
                 break;
 
         case AFR_ENTRY_RENAME_TRANSACTION:
         case AFR_ENTRY_TRANSACTION:
+                up_count = AFR_COUNT (local->child_up, priv->child_count);
+                int_lock->lk_call_count = int_lock->lk_expected_count
+                                        = (int_lock->lockee_count *
+                                           up_count);
                 initialize_entrylk_variables (frame, this);
                 break;
         }
@@ -1123,60 +1212,68 @@ afr_blocking_lock (call_frame_t *frame, xlator_t *this)
 
 static int32_t
 afr_nonblocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                             int32_t op_ret, int32_t op_errno)
+                             int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
         afr_internal_lock_t *int_lock = NULL;
-	afr_local_t         *local    = NULL;
-	afr_private_t       *priv     = NULL;
-
+        afr_local_t         *local    = NULL;
         int call_count          = 0;
-	int child_index         = (long) cookie;
+        int child_index         = (long) cookie;
+        int copies              = 0;
+        int index               = 0;
+        int lockee_no           = 0;
+        afr_private_t       *priv = NULL;
+
+        priv = this->private;
+
+        copies = priv->child_count;
+        index = child_index % copies;
+        lockee_no = child_index / copies;
 
-	local    = frame->local;
+        local    = frame->local;
         int_lock = &local->internal_lock;
-	priv     = this->private;
 
-        afr_trace_entrylk_out (frame, AFR_ENTRYLK_TRANSACTION,
-                               AFR_LOCK_OP, NULL, op_ret,
+        AFR_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
+                               AFR_LOCK_OP,
+                               int_lock->lockee[lockee_no].basename, op_ret,
                                op_errno, (long) cookie);
 
-        LOCK (&frame->lock);
-        {
-                call_count = --int_lock->lk_call_count;
-        }
-        UNLOCK (&frame->lock);
-
-        if (op_ret < 0 ) {
+	LOCK (&frame->lock);
+	{
+		if (op_ret < 0 ) {
 			if (op_errno == ENOSYS) {
-				/* return ENOTSUP */
+                        /* return ENOTSUP */
 				gf_log (this->name, GF_LOG_ERROR,
 					"subvolume does not support locking. "
-					"please load features/posix-locks xlator on server");
+					"please load features/locks xlator on server");
 				local->op_ret         = op_ret;
-                                int_lock->lock_op_ret = op_ret;
+				int_lock->lock_op_ret = op_ret;
 
-                                local->child_up[child_index] = 0;
-                                int_lock->lock_op_errno      = op_errno;
-                                local->op_errno              = op_errno;
-                        }
-        } else if (op_ret == 0) {
-                        int_lock->entry_locked_nodes[child_index]
-                                |= LOCKED_YES;
-                        int_lock->entrylk_lock_count++;
+				int_lock->lock_op_errno      = op_errno;
+				local->op_errno              = op_errno;
+			}
+		} else if (op_ret == 0) {
+			int_lock->lockee[lockee_no].locked_nodes[index] |= \
+				LOCKED_YES;
+			int_lock->lockee[lockee_no].locked_count++;
+			int_lock->entrylk_lock_count++;
+		}
+
+                call_count = --int_lock->lk_call_count;
         }
+        UNLOCK (&frame->lock);
 
         if (call_count == 0) {
                 gf_log (this->name, GF_LOG_TRACE,
                         "Last locking reply received");
-                /* all locks successfull. Proceed to call FOP */
+                /* all locks successful. Proceed to call FOP */
                 if (int_lock->entrylk_lock_count ==
-                    afr_up_children_count (priv->child_count, local->child_up)) {
+                                int_lock->lk_expected_count) {
                         gf_log (this->name, GF_LOG_TRACE,
                                 "All servers locked. Calling the cbk");
                         int_lock->lock_op_ret = 0;
                         int_lock->lock_cbk (frame, this);
                 }
-                /* Not all locks were successfull. Unlock and try locking
+                /* Not all locks were successful. Unlock and try locking
                    again, this time with serially blocking locks */
                 else {
                         gf_log (this->name, GF_LOG_TRACE,
@@ -1193,33 +1290,27 @@ afr_nonblocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 int
 afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this)
 {
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-        afr_private_t       *priv     = NULL;
-        afr_fd_ctx_t        *fd_ctx   = NULL;
-        const char          *basename = NULL;
-        loc_t               *loc      = NULL;
-
-        int32_t call_count = 0;
+        afr_internal_lock_t *int_lock   = NULL;
+        afr_local_t         *local      = NULL;
+        afr_private_t       *priv       = NULL;
+        afr_fd_ctx_t        *fd_ctx     = NULL;
+        int                 copies      = 0;
+        int                 index       = 0;
+        int                 lockee_no   = 0;
+        int32_t             call_count  = 0;
         int i = 0;
-        uint64_t  ctx;
-        int ret = 0;
 
         local    = frame->local;
         int_lock = &local->internal_lock;
         priv     = this->private;
 
+        copies = priv->child_count;
         initialize_entrylk_variables (frame, this);
 
-        basename = int_lock->lk_basename;
-        if (int_lock->lk_loc)
-                loc = int_lock->lk_loc;
-
         if (local->fd) {
-                ret = fd_ctx_get (local->fd, this, &ctx);
-
-                if (ret < 0) {
-                        gf_log (this->name, GF_LOG_DEBUG,
+                fd_ctx = afr_fd_ctx_get (local->fd, this);
+                if (!fd_ctx) {
+                        gf_log (this->name, GF_LOG_INFO,
                                 "unable to get fd ctx for fd=%p",
                                 local->fd);
 
@@ -1228,16 +1319,16 @@ afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this)
                         local->op_errno         = EINVAL;
                         int_lock->lock_op_errno = EINVAL;
 
+			afr_unlock (frame, this);
                         return -1;
                 }
 
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                call_count = internal_lock_count (frame, this, fd_ctx);
+                call_count = int_lock->lockee_count * internal_lock_count (frame, this);
                 int_lock->lk_call_count = call_count;
+                int_lock->lk_expected_count = call_count;
 
                 if (!call_count) {
-                        gf_log (this->name, GF_LOG_DEBUG,
+                        gf_log (this->name, GF_LOG_INFO,
                                 "fd not open on any subvolumes. aborting.");
                         afr_unlock (frame, this);
                         goto out;
@@ -1245,41 +1336,52 @@ 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] && fd_ctx->opened_on[i]) {
-                                afr_trace_entrylk_in (frame, AFR_ENTRYLK_NB_TRANSACTION,
-                                                      AFR_LOCK_OP, basename, i);
+                for (i = 0; i < int_lock->lockee_count*priv->child_count; i++) {
+                        index = i%copies;
+                        lockee_no = i/copies;
+                        if (local->child_up[index]) {
+                                AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_NB_TRANSACTION,
+                                                      AFR_LOCK_OP,
+                                                      int_lock->lockee[lockee_no].basename,
+                                                      i);
 
                                 STACK_WIND_COOKIE (frame, afr_nonblocking_entrylk_cbk,
                                                    (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->fentrylk,
+                                                   priv->children[index],
+                                                   priv->children[index]->fops->fentrylk,
                                                    this->name, local->fd,
-                                                   basename,
-                                                   ENTRYLK_LOCK_NB, ENTRYLK_WRLCK);
+                                                   int_lock->lockee[lockee_no].basename,
+                                                   ENTRYLK_LOCK_NB, ENTRYLK_WRLCK,
+                                                   NULL);
+                                if (!--call_count)
+                                        break;
                         }
                 }
         } else {
-                GF_ASSERT (loc);
-
-                call_count = internal_lock_count (frame, this, NULL);
+                call_count = int_lock->lockee_count * internal_lock_count (frame, this);
                 int_lock->lk_call_count = call_count;
+                int_lock->lk_expected_count = call_count;
 
-                for (i = 0; i < priv->child_count; i++) {
-                        if (local->child_up[i]) {
-                                afr_trace_entrylk_in (frame, AFR_ENTRYLK_NB_TRANSACTION,
-                                                      AFR_LOCK_OP, basename, i);
+                for (i = 0; i < int_lock->lockee_count*priv->child_count; i++) {
+                        index = i%copies;
+                        lockee_no = i/copies;
+                        if (local->child_up[index]) {
+                                AFR_TRACE_ENTRYLK_IN (frame, this, AFR_ENTRYLK_NB_TRANSACTION,
+                                                      AFR_LOCK_OP,
+                                                      int_lock->lockee[lockee_no].basename,
+                                                      i);
 
                                 STACK_WIND_COOKIE (frame, afr_nonblocking_entrylk_cbk,
                                                    (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->entrylk,
-                                                   this->name, loc, basename,
-                                                   ENTRYLK_LOCK, ENTRYLK_WRLCK);
+                                                   priv->children[index],
+                                                   priv->children[index]->fops->entrylk,
+                                                   this->name, &int_lock->lockee[lockee_no].loc,
+                                                   int_lock->lockee[lockee_no].basename,
+                                                   ENTRYLK_LOCK_NB, ENTRYLK_WRLCK,
+                                                   NULL);
 
                                 if (!--call_count)
                                         break;
-
                         }
                 }
         }
@@ -1289,59 +1391,75 @@ out:
 
 int32_t
 afr_nonblocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                             int32_t op_ret, int32_t op_errno)
+                             int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
         afr_internal_lock_t *int_lock = NULL;
-	afr_local_t         *local    = NULL;
-	afr_private_t       *priv     = NULL;
-
+        afr_inodelk_t       *inodelk  = NULL;
+        afr_local_t         *local    = NULL;
         int call_count  = 0;
-	int child_index = (long) cookie;
+        int child_index = (long) cookie;
+        afr_fd_ctx_t        *fd_ctx = NULL;
+
 
-	local    = frame->local;
+        local    = frame->local;
         int_lock = &local->internal_lock;
-	priv     = this->private;
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
 
-        afr_trace_inodelk_out (frame, AFR_INODELK_NB_TRANSACTION,
+        AFR_TRACE_INODELK_OUT (frame, this, AFR_INODELK_NB_TRANSACTION,
                                AFR_LOCK_OP, NULL, op_ret,
                                op_errno, (long) cookie);
 
+	if (local->fd)
+		fd_ctx = afr_fd_ctx_get (local->fd, this);
+
         LOCK (&frame->lock);
         {
-                call_count = --int_lock->lk_call_count;
-        }
-        UNLOCK (&frame->lock);
-
-        if (op_ret < 0 ) {
+		if (op_ret < 0) {
 			if (op_errno == ENOSYS) {
 				/* return ENOTSUP */
 				gf_log (this->name, GF_LOG_ERROR,
 					"subvolume does not support locking. "
-					"please load features/posix-locks xlator on server");
+					"please load features/locks xlator on "
+					"server");
 				local->op_ret                = op_ret;
-                                int_lock->lock_op_ret        = op_ret;
-                                local->child_up[child_index] = 0;
-                                int_lock->lock_op_errno      = op_errno;
-                                local->op_errno              = op_errno;
-                        }
-        } else if (op_ret == 0) {
-                        int_lock->inode_locked_nodes[child_index]
-                                |= LOCKED_YES;
-                        int_lock->inodelk_lock_count++;
+				int_lock->lock_op_ret        = op_ret;
+				int_lock->lock_op_errno      = op_errno;
+				local->op_errno              = op_errno;
+			}
+			if (local->transaction.eager_lock)
+				local->transaction.eager_lock[child_index] = 0;
+		} else {
+			inodelk->locked_nodes[child_index] |= LOCKED_YES;
+			inodelk->lock_count++;
+
+			if (local->transaction.eager_lock &&
+			    local->transaction.eager_lock[child_index] &&
+			    local->fd) {
+				/* piggybacked */
+				if (op_ret == 1) {
+					/* piggybacked */
+				} else if (op_ret == 0) {
+					/* lock acquired from server */
+                                        fd_ctx->lock_acquired[child_index]++;
+				}
+			}
+		}
+
+                call_count = --int_lock->lk_call_count;
         }
+        UNLOCK (&frame->lock);
 
         if (call_count == 0) {
                 gf_log (this->name, GF_LOG_TRACE,
                         "Last inode locking reply received");
-                /* all locks successfull. Proceed to call FOP */
-                if (int_lock->inodelk_lock_count ==
-                    afr_up_children_count (priv->child_count, local->child_up)) {
+                /* all locks successful. Proceed to call FOP */
+                if (inodelk->lock_count == int_lock->lk_expected_count) {
                         gf_log (this->name, GF_LOG_TRACE,
                                 "All servers locked. Calling the cbk");
                         int_lock->lock_op_ret = 0;
                         int_lock->lock_cbk (frame, this);
                 }
-                /* Not all locks were successfull. Unlock and try locking
+                /* Not all locks were successful. Unlock and try locking
                    again, this time with serially blocking locks */
                 else {
                         gf_log (this->name, GF_LOG_TRACE,
@@ -1359,31 +1477,36 @@ int
 afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
 {
         afr_internal_lock_t *int_lock = NULL;
+        afr_inodelk_t       *inodelk  = NULL;
         afr_local_t         *local    = NULL;
         afr_private_t       *priv     = NULL;
         afr_fd_ctx_t        *fd_ctx   = NULL;
-
-        int32_t  call_count = 0;
-        uint64_t ctx        = 0;
-        int      i          = 0;
-        int      ret        = 0;
-        struct gf_flock flock;
+        int32_t             call_count = 0;
+        int                 i          = 0;
+        int                 ret        = 0;
+        struct              gf_flock flock = {0,};
+        struct              gf_flock full_flock = {0,};
+        struct              gf_flock *flock_use = NULL;
+        int                 piggyback = 0;
 
         local    = frame->local;
         int_lock = &local->internal_lock;
         priv     = this->private;
 
-	flock.l_start = int_lock->lk_flock.l_start;
-	flock.l_len   = int_lock->lk_flock.l_len;
-	flock.l_type  = int_lock->lk_flock.l_type;
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+
+        flock.l_start = inodelk->flock.l_start;
+        flock.l_len   = inodelk->flock.l_len;
+        flock.l_type  = inodelk->flock.l_type;
+
+        full_flock.l_type = inodelk->flock.l_type;
 
         initialize_inodelk_variables (frame, this);
 
         if (local->fd) {
-                ret = fd_ctx_get (local->fd, this, &ctx);
-
-                if (ret < 0) {
-                        gf_log (this->name, GF_LOG_DEBUG,
+                fd_ctx = afr_fd_ctx_get (local->fd, this);
+                if (!fd_ctx) {
+                        gf_log (this->name, GF_LOG_INFO,
                                 "unable to get fd ctx for fd=%p",
                                 local->fd);
 
@@ -1392,17 +1515,17 @@ afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this)
                         local->op_errno         = EINVAL;
                         int_lock->lock_op_errno = EINVAL;
 
+			afr_unlock (frame, this);
                         ret = -1;
                         goto out;
                 }
 
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                call_count = internal_lock_count (frame, this, fd_ctx);
+                call_count = internal_lock_count (frame, this);
                 int_lock->lk_call_count = call_count;
+                int_lock->lk_expected_count = call_count;
 
                 if (!call_count) {
-                        gf_log (this->name, GF_LOG_DEBUG,
+                        gf_log (this->name, GF_LOG_INFO,
                                 "fd not open on any subvolumes. aborting.");
                         afr_unlock (frame, this);
                         goto out;
@@ -1411,261 +1534,77 @@ 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] && fd_ctx->opened_on[i]) {
-                                afr_trace_inodelk_in (frame, AFR_INODELK_NB_TRANSACTION,
-                                                      AFR_LOCK_OP, &flock, F_SETLK, i);
-
-                                STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
-                                                   (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->finodelk,
-                                                   this->name, local->fd,
-                                                   F_SETLK, &flock);
-
-                                if (!--call_count)
-                                        break;
+                        if (!local->child_up[i])
+                                continue;
 
+                        flock_use = &flock;
+                        if (!local->transaction.eager_lock_on) {
+                                goto wind;
                         }
 
-                }
-        } else {
-                call_count = internal_lock_count (frame, this, NULL);
-                int_lock->lk_call_count = call_count;
+                        piggyback = 0;
+                        local->transaction.eager_lock[i] = 1;
 
-                for (i = 0; i < priv->child_count; i++) {
-                        if (local->child_up[i]) {
-                                afr_trace_inodelk_in (frame, AFR_INODELK_NB_TRANSACTION,
-                                                      AFR_LOCK_OP, &flock, F_SETLK, i);
+			afr_set_delayed_post_op (frame, this);
 
-                                STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
-                                                   (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->inodelk,
-                                                   this->name, &local->loc,
-                                                   F_SETLK, &flock);
+                        LOCK (&local->fd->lock);
+                        {
+                                if (fd_ctx->lock_acquired[i]) {
+                                        fd_ctx->lock_piggyback[i]++;
+                                        piggyback = 1;
+                                }
+                        }
+                        UNLOCK (&local->fd->lock);
 
+                        if (piggyback) {
+                                /* (op_ret == 1) => indicate piggybacked lock */
+                                afr_nonblocking_inodelk_cbk (frame, (void *) (long) i,
+                                                             this, 1, 0, NULL);
                                 if (!--call_count)
                                         break;
-
+                                continue;
                         }
-                }
-        }
-
-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;
+                        flock_use = &full_flock;
+                wind:
+                        AFR_TRACE_INODELK_IN (frame, this,
+                                              AFR_INODELK_NB_TRANSACTION,
+                                              AFR_LOCK_OP, flock_use, F_SETLK, i);
 
-        call_count = __is_lower_locked (frame, this);
-        int_lock->lk_call_count = call_count;
+                        STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->finodelk,
+                                           int_lock->domain, local->fd,
+                                           F_SETLK, flock_use, NULL);
 
-        if (!call_count){
-                gf_log (this->name, GF_LOG_TRACE,
-                        "No internal locks unlocked");
-                int_lock->lock_cbk (frame, this);
-                goto out;
-        }
+                        if (!--call_count)
+                                break;
+                }
+        } else {
+                call_count = internal_lock_count (frame, this);
+                int_lock->lk_call_count = call_count;
+                int_lock->lk_expected_count = call_count;
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (int_lock->lower_locked_nodes[i] & LOCKED_LOWER) {
-                        afr_trace_entrylk_in (frame, AFR_ENTRYLK_NB_TRANSACTION,
-                                              AFR_UNLOCK_OP, basename, i);
+                for (i = 0; i < priv->child_count; i++) {
+                        if (!local->child_up[i])
+                                continue;
+                        AFR_TRACE_INODELK_IN (frame, this,
+                                              AFR_INODELK_NB_TRANSACTION,
+                                              AFR_LOCK_OP, &flock, F_SETLK, i);
 
-                        STACK_WIND_COOKIE (frame, afr_unlock_entrylk_cbk,
+                        STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
                                            (void *) (long) i,
                                            priv->children[i],
-                                           priv->children[i]->fops->entrylk,
-                                           this->name,
-                                           loc, basename,
-                                           ENTRYLK_UNLOCK, ENTRYLK_WRLCK);
+                                           priv->children[i]->fops->inodelk,
+                                           int_lock->domain, &local->loc,
+                                           F_SETLK, &flock, 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          *lower_name  = 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);
-
-        lower_name = (lower == &local->transaction.parent_loc ?
-                      local->transaction.basename :
-                      local->transaction.new_basename);
-
-        higher = (lower == &local->transaction.parent_loc ?
-                  &local->transaction.new_parent_loc :
-                  &local->transaction.parent_loc);
-
-        higher_name = (higher == &local->transaction.parent_loc ?
-                       local->transaction.basename :
-                       local->transaction.new_basename);
-
-        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;
-	loc_t               *higher      = NULL;
-	const char          *lower_name  = 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);
-
-        lower_name = (lower == &local->transaction.parent_loc ?
-                      local->transaction.basename :
-                      local->transaction.new_basename);
-
-        higher = (lower == &local->transaction.parent_loc ?
-                  &local->transaction.new_parent_loc :
-                  &local->transaction.parent_loc);
-
-        higher_name = (higher == &local->transaction.parent_loc ?
-                       local->transaction.basename :
-                       local->transaction.new_basename);
-
-
-        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);
-
+        return ret;
 }
 
 int32_t
@@ -1679,10 +1618,8 @@ 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);
@@ -1694,490 +1631,37 @@ afr_unlock (call_frame_t *frame, xlator_t *this)
 }
 
 int
-afr_mark_locked_nodes (xlator_t *this, fd_t *fd,
-                       unsigned char *locked_nodes)
+afr_lk_transfer_datalock (call_frame_t *dst, call_frame_t *src, char *dom,
+                          unsigned int child_count)
 {
-        afr_private_t *priv  = NULL;
-        afr_fd_ctx_t  *fdctx = NULL;
-        uint64_t       tmp   = 0;
-        int            ret   = 0;
-
-        priv = this->private;
-
-        afr_fd_ctx_set (this, fd);
-        if (ret < 0)
+        afr_local_t         *dst_local   = NULL;
+        afr_local_t         *src_local   = NULL;
+        afr_internal_lock_t *dst_lock    = NULL;
+        afr_internal_lock_t *src_lock    = NULL;
+        afr_inodelk_t       *dst_inodelk = NULL;
+        afr_inodelk_t       *src_inodelk = NULL;
+        int                 ret = -1;
+
+        src_local = src->local;
+        src_lock  = &src_local->internal_lock;
+        src_inodelk = afr_get_inodelk (src_lock, dom);
+        dst_local = dst->local;
+        dst_lock  = &dst_local->internal_lock;
+        dst_inodelk = afr_get_inodelk (dst_lock, dom);
+        if (!dst_inodelk || !src_inodelk)
                 goto out;
-
-        ret = fd_ctx_get (fd, this, &tmp);
-        fdctx = (afr_fd_ctx_t *) (long) tmp;
-
-        GF_ASSERT (fdctx->locked_on);
-
-        memcpy (fdctx->locked_on, locked_nodes,
-                priv->child_count);
-
-out:
-        return ret;
-}
-
-static int
-__is_fd_saved (xlator_t *this, fd_t *fd)
-{
-        afr_locked_fd_t *locked_fd = NULL;
-        afr_private_t   *priv      = NULL;
-        int              found     = 0;
-
-        priv = this->private;
-
-        list_for_each_entry (locked_fd, &priv->saved_fds, list) {
-                if (locked_fd->fd == fd) {
-                        found = 1;
-                        break;
-                }
-        }
-
-        return found;
-}
-
-static int
-__afr_save_locked_fd (xlator_t *this, fd_t *fd)
-{
-        afr_private_t   *priv      = NULL;
-        afr_locked_fd_t *locked_fd = NULL;
-        int              ret       = 0;
-
-        priv = this->private;
-
-        locked_fd = GF_CALLOC (1, sizeof (*locked_fd),
-                               gf_afr_mt_locked_fd);
-        if (!locked_fd) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                ret = -1;
-                goto out;
-        }
-
-        locked_fd->fd = fd;
-        INIT_LIST_HEAD (&locked_fd->list);
-
-        list_add_tail (&locked_fd->list, &priv->saved_fds);
-
-out:
-        return ret;
-}
-
-int
-afr_save_locked_fd (xlator_t *this, fd_t *fd)
-{
-        afr_private_t   *priv      = NULL;
-        int              ret       = 0;
-
-        priv = this->private;
-
-        pthread_mutex_lock (&priv->mutex);
-        {
-                if (__is_fd_saved (this, fd)) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "fd=%p already saved", fd);
-                        goto unlock;
-                }
-
-                ret = __afr_save_locked_fd (this, fd);
-                if (ret) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "fd=%p could not be saved");
-                        goto unlock;
-                }
-        }
-unlock:
-        pthread_mutex_unlock (&priv->mutex);
-
-        return ret;
-}
-
-static int
-afr_lock_recovery_cleanup (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t     *local     = NULL;
-        afr_locked_fd_t *locked_fd = NULL;
-
-        local = frame->local;
-
-        locked_fd = local->locked_fd;
-
-        STACK_DESTROY (frame->root);
-        afr_local_cleanup (local, this);
-
-        afr_save_locked_fd (this, locked_fd->fd);
-
-        return 0;
-
-}
-
-static int
-afr_get_source_lock_recovery (xlator_t *this, fd_t *fd)
-{
-        afr_fd_ctx_t  *fdctx        = NULL;
-        afr_private_t *priv         = NULL;
-        uint64_t      tmp           = 0;
-        int           i             = 0;
-        int           source_child  = -1;
-        int           ret           = 0;
-
-        priv = this->private;
-
-        ret = fd_ctx_get (fd, this, &tmp);
-        if (ret)
-                goto out;
-
-        fdctx = (afr_fd_ctx_t *) (long) tmp;
-
-        for (i = 0; i < priv->child_count; i++) {
-                if (fdctx->locked_on[i]) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "Found lock recovery source=%d",
-                                i);
-                        source_child = i;
-                        break;
-                }
-
-        }
-
-out:
-        return source_child;
-
-}
-
-int32_t
-afr_get_locks_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno, struct gf_flock *lock);
-int32_t
-afr_recover_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno, struct gf_flock *lock)
-{
-        afr_local_t   *local = NULL;
-        afr_private_t *priv  = NULL;
-        int32_t source_child = 0;
-        struct gf_flock flock   = {0,};
-
-        local = frame->local;
-        priv  = this->private;
-
-        if (op_ret) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "lock recovery failed");
-                goto cleanup;
-        }
-
-        source_child = local->source_child;
-
-        memcpy (&flock, lock, sizeof (*lock));
-
-        STACK_WIND_COOKIE (frame, afr_get_locks_fd_cbk,
-                           (void *) (long) source_child,
-                           priv->children[source_child],
-                           priv->children[source_child]->fops->lk,
-                           local->fd, F_GETLK_FD, &flock);
-
-        return 0;
-
-cleanup:
-        afr_lock_recovery_cleanup (frame, this);
-        return 0;
-}
-
-int
-afr_recover_lock (call_frame_t *frame, xlator_t *this,
-                  struct gf_flock *flock)
-{
-        afr_local_t   *local             = NULL;
-        afr_private_t *priv              = NULL;
-        int32_t      lock_recovery_child = 0;
-
-        priv  = this->private;
-        local = frame->local;
-
-        lock_recovery_child = local->lock_recovery_child;
-
-        frame->root->lk_owner = flock->l_owner;
-
-        STACK_WIND_COOKIE (frame, afr_recover_lock_cbk,
-                           (void *) (long) lock_recovery_child,
-                           priv->children[lock_recovery_child],
-                           priv->children[lock_recovery_child]->fops->lk,
-                           local->fd, F_SETLK, flock);
-
-        return 0;
-}
-
-static int
-is_afr_lock_eol (struct gf_flock *lock)
-{
-        int ret = 0;
-
-        if ((lock->l_type == GF_LK_EOL))
-                ret = 1;
-
-        return ret;
-}
-
-int32_t
-afr_get_locks_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno, struct gf_flock *lock)
-{
-        if (op_ret) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Failed to get locks on fd");
-                goto cleanup;
-        }
-
-        gf_log (this->name, GF_LOG_DEBUG,
-                "Got a lock on fd");
-
-        if (is_afr_lock_eol (lock)) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Reached EOL on locks on fd");
-                goto cleanup;
-        }
-
-        afr_recover_lock (frame, this, lock);
-
-        return 0;
-
-cleanup:
-        afr_lock_recovery_cleanup (frame, this);
-
-        return 0;
-}
-
-static int
-afr_lock_recovery (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t   *local        = NULL;
-        afr_private_t *priv         = NULL;
-        fd_t          *fd           = NULL;
-        int            ret          = 0;
-        int32_t        source_child = 0;
-        struct gf_flock   flock        = {0,};
-
-        priv  = this->private;
-        local = frame->local;
-
-        fd = local->fd;
-
-        source_child = afr_get_source_lock_recovery (this, fd);
-        if (source_child < 0) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Could not recover locks due to lock "
-                        "split brain");
-                ret = -1;
-                goto out;
-        }
-
-        local->source_child = source_child;
-
-        /* the flock can be zero filled as we're querying incrementally
-           the locks held on the fd.
-        */
-        STACK_WIND_COOKIE (frame, afr_get_locks_fd_cbk,
-                           (void *) (long) source_child,
-                           priv->children[source_child],
-                           priv->children[source_child]->fops->lk,
-                           local->fd, F_GETLK_FD, &flock);
-
-out:
-        return ret;
-}
-
-
-static int
-afr_mark_fd_opened (xlator_t *this, fd_t *fd, int32_t child_index)
-{
-        afr_fd_ctx_t *fdctx = NULL;
-        uint64_t      tmp   = 0;
-        int           ret   = 0;
-
-        ret = fd_ctx_get (fd, this, &tmp);
-        if (ret)
-                goto out;
-
-        fdctx = (afr_fd_ctx_t *) (long) tmp;
-
-        fdctx->opened_on[child_index] = 1;
-
-out:
-        return ret;
-}
-
-int32_t
-afr_lock_recovery_preopen_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                               int32_t op_ret, int32_t op_errno, fd_t *fd)
-{
-        int32_t child_index = (long )cookie;
-        int ret = 0;
-
-        if (op_ret) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Reopen during lock-recovery failed");
-                goto cleanup;
-        }
-
-        gf_log (this->name, GF_LOG_DEBUG,
-                "Open succeeded => proceed to recover locks");
-
-        ret = afr_lock_recovery (frame, this);
-        if (ret) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Lock recovery failed");
-                goto cleanup;
-        }
-
-        ret = afr_mark_fd_opened (this, fd, child_index);
-        if (ret) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Marking fd open failed");
-                goto cleanup;
-        }
-
-        return 0;
-
-cleanup:
-        afr_lock_recovery_cleanup (frame, this);
-        return 0;
-}
-
-static int
-afr_lock_recovery_preopen (call_frame_t *frame, xlator_t *this)
-{
-        afr_private_t *priv  = NULL;
-        afr_local_t   *local = NULL;
-        uint64_t       tmp   = 0;
-        afr_fd_ctx_t  *fdctx = NULL;
-        loc_t          loc   = {0,};
-        int32_t        child_index = 0;
-        int            ret   = 0;
-
-        priv  = this->private;
-        local = frame->local;
-
-        GF_ASSERT (local && local->fd);
-
-        ret = fd_ctx_get (local->fd, this, &tmp);
-        fdctx = (afr_fd_ctx_t *) (long) tmp;
-        GF_ASSERT (fdctx);
-
-        child_index = local->lock_recovery_child;
-
-        inode_path (local->fd->inode, NULL, (char **)&loc.path);
-        loc.name   = strrchr (loc.path, '/');
-        loc.inode  = inode_ref (local->fd->inode);
-        loc.parent = inode_parent (local->fd->inode, 0, NULL);
-
-
-        STACK_WIND_COOKIE (frame, afr_lock_recovery_preopen_cbk,
-                           (void *)(long) child_index,
-                           priv->children[child_index],
-                           priv->children[child_index]->fops->open,
-                           &loc, fdctx->flags, local->fd,
-                           fdctx->wbflags);
-
-        return 0;
-}
-
-static int
-is_fd_opened (fd_t *fd, int32_t child_index)
-{
-        afr_fd_ctx_t *fdctx = NULL;
-        uint64_t      tmp = 0;
-        int           ret = 0;
-
-        ret = fd_ctx_get (fd, THIS, &tmp);
-        if (ret)
-                goto out;
-
-        fdctx = (afr_fd_ctx_t *) (long) tmp;
-
-        if (fdctx->opened_on[child_index])
-                ret = 1;
-
-out:
-        return ret;
-}
-
-int
-afr_attempt_lock_recovery (xlator_t *this, int32_t child_index)
-{
-        call_frame_t    *frame     = NULL;
-        afr_private_t   *priv      = NULL;
-        afr_local_t     *local     = NULL;
-        afr_locked_fd_t *locked_fd = NULL;
-       afr_locked_fd_t  *tmp       = NULL;
-        int              ret       = 0;
-        struct list_head locks_list;
-
-
-        priv = this->private;
-
-        if (list_empty (&priv->saved_fds))
-                goto out;
-
-        frame = create_frame (this, this->ctx->pool);
-        if (!frame) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                ret = -1;
-                goto out;
-        }
-
-        local = GF_CALLOC (1, sizeof (*local),
-                           gf_afr_mt_afr_local_t);
-        if (!local) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Out of memory");
-                ret = -1;
-                goto out;
-        }
-
-        AFR_LOCAL_INIT (local, priv);
-        if (!local) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                ret = -1;
-                goto out;
-        }
-
-        frame->local = local;
-
-        INIT_LIST_HEAD (&locks_list);
-
-        pthread_mutex_lock (&priv->mutex);
-        {
-                list_splice_init (&priv->saved_fds, &locks_list);
-        }
-        pthread_mutex_unlock (&priv->mutex);
-
-        list_for_each_entry_safe (locked_fd, tmp,
-                                  &locks_list, list) {
-
-                list_del_init (&locked_fd->list);
-
-                local->fd                  = fd_ref (locked_fd->fd);
-                local->lock_recovery_child = child_index;
-                local->locked_fd           = locked_fd;
-
-                if (!is_fd_opened (locked_fd->fd, child_index)) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "attempting open before lock "
-                                "recovery");
-                        afr_lock_recovery_preopen (frame, this);
-                } else {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "attempting lock recovery "
-                                "without a preopen");
-                        afr_lock_recovery (frame, this);
-                }
+        if (src_inodelk->locked_nodes) {
+                memcpy (dst_inodelk->locked_nodes, src_inodelk->locked_nodes,
+                        sizeof (*dst_inodelk->locked_nodes) * child_count);
+                memset (src_inodelk->locked_nodes, 0,
+                        sizeof (*src_inodelk->locked_nodes) * child_count);
         }
 
+        dst_lock->transaction_lk_type = src_lock->transaction_lk_type;
+        dst_lock->selfheal_lk_type    = src_lock->selfheal_lk_type;
+        dst_inodelk->lock_count = src_inodelk->lock_count;
+        src_inodelk->lock_count = 0;
+        ret = 0;
 out:
         return ret;
 }
diff --git a/xlators/cluster/afr/src/afr-mem-types.h b/xlators/cluster/afr/src/afr-mem-types.h
index 14064ebcd..05df90cc0 100644
--- a/xlators/cluster/afr/src/afr-mem-types.h
+++ b/xlators/cluster/afr/src/afr-mem-types.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -26,7 +17,6 @@
 enum gf_afr_mem_types_ {
         gf_afr_mt_iovec  = gf_common_mt_end + 1,
         gf_afr_mt_afr_fd_ctx_t,
-        gf_afr_mt_afr_local_t,
         gf_afr_mt_afr_private_t,
         gf_afr_mt_int32_t,
         gf_afr_mt_char,
@@ -42,6 +32,17 @@ enum gf_afr_mem_types_ {
         gf_afr_mt_entry_name,
         gf_afr_mt_pump_priv,
         gf_afr_mt_locked_fd,
+        gf_afr_mt_inode_ctx_t,
+        gf_afr_fd_paused_call_t,
+        gf_afr_mt_crawl_data_t,
+        gf_afr_mt_brick_pos_t,
+        gf_afr_mt_shd_bool_t,
+        gf_afr_mt_shd_timer_t,
+        gf_afr_mt_shd_event_t,
+        gf_afr_mt_time_t,
+        gf_afr_mt_pos_data_t,
+	gf_afr_mt_reply_t,
+	gf_afr_mt_subvol_healer_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 44ba22ee7..f86aa7fd8 100644
--- a/xlators/cluster/afr/src/afr-open.c
+++ b/xlators/cluster/afr/src/afr-open.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include <libgen.h>
@@ -52,591 +43,295 @@
 #include "afr-dir-read.h"
 #include "afr-dir-write.h"
 #include "afr-transaction.h"
-#include "afr-self-heal.h"
-#include "afr-self-heal-common.h"
 
 
-int
-afr_open_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-			int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                        struct iatt *postbuf)
+gf_boolean_t
+afr_is_fd_fixable (fd_t *fd)
 {
-	afr_local_t * local = frame->local;
-
-	AFR_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
-			  local->fd);
-	return 0;
+        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;
 }
 
 
 int
-afr_open_cbk (call_frame_t *frame, void *cookie,
-	      xlator_t *this, int32_t op_ret, int32_t op_errno,
-	      fd_t *fd)
+afr_open_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                        struct iatt *postbuf, dict_t *xdata)
 {
-	afr_local_t *  local = NULL;
-
-        int child_index = (long) cookie;
-
-        uint64_t      ctx;
-        afr_fd_ctx_t *fd_ctx;
-
-        int ret = 0;
-
-	int call_count = -1;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-		}
-
-		if (op_ret >= 0) {
-			local->op_ret = op_ret;
-                        local->success_count++;
-
-                        ret = afr_fd_ctx_set (this, fd);
-
-                        if (ret < 0) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "could not set fd ctx for fd=%p",
-                                        fd);
-
-                                local->op_ret   = -1;
-                                local->op_errno = -ret;
-                                goto unlock;
-                        }
-
-                        ret = fd_ctx_get (fd, this, &ctx);
-
-                        if (ret < 0) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "could not get fd ctx for fd=%p", fd);
-                                local->op_ret   = -1;
-                                local->op_errno = -ret;
-                                goto unlock;
-                        }
-
-                        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                        fd_ctx->opened_on[child_index] = 1;
-                        fd_ctx->flags                  = local->cont.open.flags;
-                        fd_ctx->wbflags                = local->cont.open.wbflags;
-		}
-	}
-unlock:
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-		if ((local->cont.open.flags & O_TRUNC)
-		    && (local->op_ret >= 0)) {
-			STACK_WIND (frame, afr_open_ftruncate_cbk,
-				    this, this->fops->ftruncate,
-				    fd, 0);
-		} else {
-                        AFR_STACK_UNWIND (open, frame, local->op_ret,
-                                          local->op_errno, local->fd);
-		}
-	}
-
-	return 0;
-}
-
+        afr_local_t * local = frame->local;
 
-int
-afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
-          fd_t *fd, int32_t wbflags)
-{
-	afr_private_t * priv  = NULL;
-	afr_local_t *   local = NULL;
-
-	int     i = 0;
-	int   ret = -1;
-
-	int32_t call_count = 0;	
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
-	int32_t wind_flags = flags & (~O_TRUNC);
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-	VALIDATE_OR_GOTO (loc, out);
-
-	priv = this->private;
-
-        if (afr_is_split_brain (this, loc->inode)) {
-		/* self-heal failed */
-		op_errno = EIO;
-		goto out;
-	}
-
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
-
-	frame->local = local;
-	call_count   = local->call_count;
-
-        loc_copy (&local->loc, loc);
-
-	local->cont.open.flags   = flags;
-        local->cont.open.wbflags = wbflags;
-
-	local->fd = fd_ref (fd);
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_open_cbk, (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->open,
-					   loc, wind_flags, fd, wbflags);
-
-			if (!--call_count)
-				break;
-		}
-	}
-
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (open, frame, op_ret, op_errno, fd);
-	}
-
-	return 0;
+        AFR_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
+                          local->fd, xdata);
+        return 0;
 }
 
 
 int
-afr_openfd_sh_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int32_t op_ret, int32_t op_errno, fd_t *fd)
+afr_open_cbk (call_frame_t *frame, void *cookie,
+              xlator_t *this, int32_t op_ret, int32_t op_errno,
+              fd_t *fd, dict_t *xdata)
 {
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-        afr_private_t       *priv     = NULL;
-
-        int ret = 0;
-
-        uint64_t      ctx;
-        afr_fd_ctx_t *fd_ctx;
-
-        int call_count  = 0;
-        int child_index = (long) cookie;
+        afr_local_t *  local       = NULL;
+        int            call_count  = -1;
+        int            child_index = (long) cookie;
+	afr_fd_ctx_t  *fd_ctx = NULL;
 
-        priv     = this->private;
-        local    = frame->local;
-        int_lock = &local->internal_lock;
+        local = frame->local;
+	fd_ctx = local->fd_ctx;
 
         LOCK (&frame->lock);
         {
-                if (op_ret >= 0) {
-                        ret = fd_ctx_get (fd, this, &ctx);
-
-                        if (ret < 0) {
-                                goto out;
-                        }
-
-                        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                        fd_ctx->opened_on[child_index] = 1;
-
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "fd for %s opened successfully on subvolume %s",
-                                local->loc.path, priv->children[child_index]->name);
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+			fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
+                } else {
+                        local->op_ret = op_ret;
+			fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
+			if (!local->xdata_rsp && xdata)
+				local->xdata_rsp = dict_ref (xdata);
                 }
         }
-out:
         UNLOCK (&frame->lock);
 
         call_count = afr_frame_return (frame);
 
         if (call_count == 0) {
-                int_lock->lock_cbk = local->transaction.done;
-                local->transaction.resume (frame, this);
+                if ((fd_ctx->flags & O_TRUNC) && (local->op_ret >= 0)) {
+                        STACK_WIND (frame, afr_open_ftruncate_cbk,
+                                    this, this->fops->ftruncate,
+                                    fd, 0, NULL);
+                } else {
+                        AFR_STACK_UNWIND (open, frame, local->op_ret,
+                                          local->op_errno, local->fd,
+					  local->xdata_rsp);
+                }
         }
 
         return 0;
 }
 
-
-static int
-__unopened_count (int child_count, unsigned int *opened_on, unsigned char *child_up)
-{
-        int i;
-        int count = 0;
-
-        for (i = 0; i < child_count; i++) {
-                if (!opened_on[i] && child_up[i])
-                        count++;
-        }
-
-        return count;
-}
-
-
 int
-afr_openfd_sh_unwind (call_frame_t *frame, xlator_t *this)
+afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+          fd_t *fd, dict_t *xdata)
 {
-        afr_local_t *local  = NULL;
-        afr_private_t *priv = NULL;
-
-        uint64_t      ctx;
-        afr_fd_ctx_t *fd_ctx;
-
-        int abandon = 0;
-        int ret = 0;
-        int i;
-        int call_count = 0;
-
-        priv  = this->private;
-        local = frame->local;
+        afr_private_t * priv       = NULL;
+        afr_local_t *   local      = NULL;
+        int             i          = 0;
+        int32_t         call_count = 0;
+        int32_t         op_errno   = 0;
+	afr_fd_ctx_t   *fd_ctx = NULL;
 
-        /*
-         * Some subvolumes might have come up on which we never
-         * opened this fd in the first place. Re-open fd's on those
-         * subvolumes now.
-         */
+        //We can't let truncation to happen outside transaction.
 
-        ret = fd_ctx_get (local->fd, this, &ctx);
+        priv = this->private;
 
-        if (ret < 0) {
-                abandon = 1;
-                goto out;
+        if (flags & (O_CREAT|O_TRUNC)) {
+                QUORUM_CHECK(open,out);
         }
 
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-        LOCK (&local->fd->lock);
-        {
-                call_count = __unopened_count (priv->child_count,
-                                               fd_ctx->opened_on,
-                                               local->child_up);
-                for (i = 0; i < priv->child_count; i++) {
-                        fd_ctx->pre_op_done[i] = 0;
-                        fd_ctx->pre_op_piggyback[i] = 0;
-                }
-        }
-        UNLOCK (&local->fd->lock);
+	fd_ctx = afr_fd_ctx_get (fd, this);
+	if (!fd_ctx) {
+		op_errno = ENOMEM;
+		goto out;
+	}
 
-        if (call_count == 0) {
-                abandon = 1;
-                goto out;
-        }
+        local->fd = fd_ref (fd);
+	local->fd_ctx = fd_ctx;
+	fd_ctx->flags = flags;
 
-        local->call_count = call_count;
+        call_count = local->call_count;
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (!fd_ctx->opened_on[i] && local->child_up[i]) {
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "opening fd for %s on subvolume %s",
-                                local->loc.path, priv->children[i]->name);
+        local->cont.open.flags = flags;
 
-                        STACK_WIND_COOKIE (frame, afr_openfd_sh_open_cbk,
-                                           (void *)(long) i,
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND_COOKIE (frame, afr_open_cbk, (void *) (long) i,
                                            priv->children[i],
                                            priv->children[i]->fops->open,
-                                           &local->loc, fd_ctx->flags, local->fd,
-                                           fd_ctx->wbflags);
-
+                                           loc, (flags & ~O_TRUNC), fd, xdata);
                         if (!--call_count)
                                 break;
                 }
         }
 
+	return 0;
 out:
-        if (abandon)
-                local->transaction.resume (frame, this);
+	AFR_STACK_UNWIND (open, frame, -1, op_errno, fd, NULL);
 
         return 0;
 }
 
-
-static int
-afr_prepare_loc (call_frame_t *frame, fd_t *fd)
-{
-        afr_local_t    *local = NULL;
-        char           *name = NULL;
-        char           *path = NULL;
-        int             ret = 0;
-
-        if (!fd)
-                return -1;
-
-        local = frame->local;
-        ret = inode_path (fd->inode, NULL, (char **)&path);
-        if (ret == -1)
-                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);
-        }
-
-        local->loc.parent = inode_parent (local->loc.inode, 0, NULL);
-
-        return 0;
-}
-
-
 int
-afr_openfd_sh (call_frame_t *frame, xlator_t *this)
+afr_openfd_fix_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int32_t op_ret, int32_t op_errno, fd_t *fd,
+                         dict_t *xdata)
 {
-        afr_local_t     *local  = NULL;
-        afr_self_heal_t *sh = NULL;
-        char            sh_type_str[256] = {0,};
-
-        local = frame->local;
-        sh    = &local->self_heal;
-
-        afr_prepare_loc (frame, local->fd);
-
-        /* forcibly trigger missing-entries self-heal */
+        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;
 
-        local->success_count    = 1;
-        local->enoent_count     = 1;
-
-        sh->data_lock_held      = _gf_true;
-        sh->need_data_self_heal = _gf_true;
-        sh->type                = local->fd->inode->ia_type;
-        sh->background          = _gf_false;
-        sh->unwind              = afr_openfd_sh_unwind;
-
-        afr_self_heal_type_str_get(&local->self_heal,
-                                   sh_type_str,
-                                   sizeof(sh_type_str));
-        gf_log (this->name, GF_LOG_NORMAL, "%s self-heal triggered. "
-                "path: %s, reason: Replicate up down flush, data lock is held",
-                sh_type_str, local->loc.path);
-
-        afr_self_heal (frame, this);
-
-        return 0;
-}
-
-
-int
-afr_openfd_flush_done (call_frame_t *frame, xlator_t *this)
-{
-        afr_private_t *priv = NULL;
-	afr_local_t *local  = NULL;
-
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
+        priv     = this->private;
+        local    = frame->local;
 
-        int _ret = -1;
+        if (op_ret >= 0) {
+                gf_log (this->name, GF_LOG_DEBUG, "fd for %s opened "
+                        "successfully on subvolume %s", local->loc.path,
+                        priv->children[child_index]->name);
+        } else {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to open %s "
+                        "on subvolume %s", local->loc.path,
+                        priv->children[child_index]->name);
+        }
 
-        priv  = this->private;
-	local = frame->local;
+        fd_ctx = local->fd_ctx;
 
         LOCK (&local->fd->lock);
         {
-                _ret = __fd_ctx_get (local->fd, this, &ctx);
-
-                if (_ret < 0) {
-                        goto out;
+                if (op_ret >= 0) {
+                        fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
+                } else {
+                        fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
                 }
-
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                fd_ctx->down_count = priv->down_count;
-                fd_ctx->up_count   = priv->up_count;
         }
-out:
         UNLOCK (&local->fd->lock);
 
-        afr_local_transaction_cleanup (local, this);
-
-        gf_log (this->name, GF_LOG_TRACE,
-                "The up/down flush is over");
-
-        fd_unref (local->fd);
-        local->openfd_flush_cbk (frame, this);
-
-	return 0;
-}
-
-
-
-int
-afr_openfd_xaction (call_frame_t *frame, xlator_t *this, fd_t *fd)
-{
-	afr_local_t   * local = NULL;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-
-        local = frame->local;
-
-        local->op = GF_FOP_FLUSH;
-
-        local->transaction.fop          = afr_openfd_sh;
-        local->transaction.done         = afr_openfd_flush_done;
-
-        local->transaction.start  = 0;
-        local->transaction.len    = 0;
-
-        gf_log (this->name, GF_LOG_TRACE,
-                "doing up/down flush on fd=%p",
-                fd);
-
-        afr_transaction (frame, this, AFR_DATA_TRANSACTION);
+        call_count = afr_frame_return (frame);
+        if (call_count == 0)
+                AFR_STACK_DESTROY (frame);
 
-out:
-	return 0;
+        return 0;
 }
 
 
-
-int
-afr_openfd_xaction_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                             int32_t op_ret, int32_t op_errno, fd_t *fd)
+static int
+afr_fd_ctx_need_open (fd_t *fd, xlator_t *this, unsigned char *need_open)
 {
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-        afr_private_t       *priv     = NULL;
-
-        int ret = 0;
-
-        uint64_t      ctx = 0;
-        afr_fd_ctx_t *fd_ctx = NULL;
-
-        int call_count  = 0;
-        int child_index = (long) cookie;
-
-        priv     = this->private;
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-
-        LOCK (&frame->lock);
-        {
-                if (op_ret >= 0) {
-                        ret = fd_ctx_get (fd, this, &ctx);
-
-                        if (ret < 0) {
-                                goto out;
-                        }
-
-                        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                        fd_ctx->opened_on[child_index] = 1;
+	afr_fd_ctx_t *fd_ctx = NULL;
+	afr_private_t *priv = NULL;
+	int i = 0;
+	int count = 0;
 
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "fd for %s opened successfully on subvolume %s",
-                                local->loc.path, priv->children[child_index]->name);
-                }
-        }
-out:
-        UNLOCK (&frame->lock);
+	priv = this->private;
 
-        call_count = afr_frame_return (frame);
+	fd_ctx = afr_fd_ctx_get (fd, this);
+	if (!fd_ctx)
+		return 0;
 
-        if (call_count == 0) {
-                afr_openfd_xaction (frame, this, local->fd);
-        }
+	LOCK (&fd->lock);
+	{
+		for (i = 0; i < priv->child_count; i++) {
+			if (fd_ctx->opened_on[i] == AFR_FD_NOT_OPENED &&
+			    priv->child_up[i]) {
+				fd_ctx->opened_on[i] = AFR_FD_OPENING;
+				need_open[i] = 1;
+				count++;
+			} else {
+				need_open[i] = 0;
+			}
+		}
+	}
+	UNLOCK (&fd->lock);
 
-        return 0;
+	return count;
 }
 
 
-int
-afr_openfd_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
+void
+afr_fix_open (fd_t *fd, xlator_t *this)
 {
-        afr_local_t *local  = NULL;
-        afr_private_t *priv = NULL;
-
-        uint64_t      ctx;
-        afr_fd_ctx_t *fd_ctx;
-
-        int no_open = 0;
-        int ret = 0;
-        int i;
-        int call_count = 0;
+        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;
+	unsigned char *need_open = NULL;
+	int call_count = 0;
 
         priv  = this->private;
-        local = frame->local;
 
-        /*
-         * Some subvolumes might have come up on which we never
-         * opened this fd in the first place. Re-open fd's on those
-         * subvolumes now.
-         */
+        if (!afr_is_fd_fixable (fd))
+                goto out;
 
-        local->fd = fd_ref (fd);
+        fd_ctx = afr_fd_ctx_get (fd, this);
+        if (!fd_ctx)
+                goto out;
 
-        ret = fd_ctx_get (fd, this, &ctx);
+	need_open = alloca0 (priv->child_count);
 
-        if (ret < 0) {
-                no_open = 1;
-                goto out;
-        }
+	call_count = afr_fd_ctx_need_open (fd, this, need_open);
+	if (!call_count)
+		goto out;
 
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame)
+                goto out;
 
-        LOCK (&local->fd->lock);
-        {
-                call_count = __unopened_count (priv->child_count,
-                                               fd_ctx->opened_on,
-                                               local->child_up);
-        }
-        UNLOCK (&local->fd->lock);
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-        if (call_count == 0) {
-                no_open = 1;
+        local->loc.inode = inode_ref (fd->inode);
+        ret = loc_path (&local->loc, NULL);
+        if (ret < 0)
                 goto out;
-        }
 
-        afr_prepare_loc (frame, fd);
+        local->fd = fd_ref (fd);
+	local->fd_ctx = fd_ctx;
 
         local->call_count = call_count;
 
+        gf_log (this->name, GF_LOG_DEBUG, "need open count: %d",
+                call_count);
+
         for (i = 0; i < priv->child_count; i++) {
-                if (!fd_ctx->opened_on[i] && local->child_up[i]) {
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "opening fd for %s on subvolume %s",
+                if (!need_open[i])
+                        continue;
+
+                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 (frame, afr_openfd_fix_open_cbk,
+                                           (void*) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->opendir,
+                                           &local->loc, local->fd,
+                                           NULL);
+                } else {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "opening fd for file %s on subvolume %s",
                                 local->loc.path, priv->children[i]->name);
 
-                        STACK_WIND_COOKIE (frame, afr_openfd_xaction_open_cbk,
+                        STACK_WIND_COOKIE (frame, afr_openfd_fix_open_cbk,
                                            (void *)(long) i,
                                            priv->children[i],
                                            priv->children[i]->fops->open,
-                                           &local->loc, fd_ctx->flags, fd,
-                                           fd_ctx->wbflags);
-
-                        if (!--call_count)
-                                break;
+                                           &local->loc,
+                                           fd_ctx->flags & (~O_TRUNC),
+                                           local->fd, NULL);
                 }
+
+		if (!--call_count)
+			break;
         }
 
+	return;
 out:
-        if (no_open)
-                afr_openfd_xaction (frame, this, fd);
-
-        return 0;
+        if (frame)
+                AFR_STACK_DESTROY (frame);
 }
diff --git a/xlators/cluster/afr/src/afr-read-txn.c b/xlators/cluster/afr/src/afr-read-txn.c
new file mode 100644
index 000000000..186f68c33
--- /dev/null
+++ b/xlators/cluster/afr/src/afr-read-txn.c
@@ -0,0 +1,239 @@
+/*
+  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 "afr.h"
+#include "afr-transaction.h"
+
+int
+afr_read_txn_next_subvol (call_frame_t *frame, xlator_t *this)
+{
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	int i = 0;
+	int subvol = -1;
+
+	local = frame->local;
+	priv = this->private;
+
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (!local->readable[i]) {
+			/* don't even bother trying here.
+			   just mark as attempted and move on. */
+			local->read_attempted[i] = 1;
+			continue;
+		}
+
+		if (!local->read_attempted[i]) {
+			subvol = i;
+			break;
+		}
+	}
+
+	/* If no more subvols were available for reading, we leave
+	   @subvol as -1, which is an indication we have run out of
+	   readable subvols. */
+	if (subvol != -1)
+		local->read_attempted[subvol] = 1;
+	local->readfn (frame, this, subvol);
+
+	return 0;
+}
+
+
+int
+afr_read_txn_refresh_done (call_frame_t *frame, xlator_t *this, int err)
+{
+	afr_local_t *local = NULL;
+	int read_subvol = 0;
+	int event_generation = 0;
+	inode_t *inode = NULL;
+	int ret = -1;
+
+	local = frame->local;
+	inode = local->inode;
+
+	if (err) {
+		local->op_errno = -err;
+		local->op_ret = -1;
+		read_subvol = -1;
+		goto readfn;
+	}
+
+	ret = afr_inode_read_subvol_type_get (inode, this, local->readable,
+					      &event_generation,
+					      local->transaction.type);
+
+	if (ret == -1 || !event_generation) {
+		/* Even after refresh, we don't have a good
+		   read subvolume. Time to bail */
+		local->op_ret = -1;
+		local->op_errno = EIO;
+		read_subvol = -1;
+		goto readfn;
+	}
+
+	read_subvol = afr_read_subvol_select_by_policy (inode, this,
+							local->readable);
+
+	if (read_subvol == -1) {
+		local->op_ret = -1;
+		local->op_errno = EIO;
+		goto readfn;
+	}
+
+	if (local->read_attempted[read_subvol]) {
+		afr_read_txn_next_subvol (frame, this);
+		return 0;
+	}
+
+	local->read_attempted[read_subvol] = 1;
+readfn:
+	local->readfn (frame, this, read_subvol);
+
+	return 0;
+}
+
+
+int
+afr_read_txn_continue (call_frame_t *frame, xlator_t *this, int subvol)
+{
+	afr_local_t *local = NULL;
+
+	local = frame->local;
+
+	if (!local->refreshed) {
+		local->refreshed = _gf_true;
+		afr_inode_refresh (frame, this, local->inode,
+				   afr_read_txn_refresh_done);
+	} else {
+		afr_read_txn_next_subvol (frame, this);
+	}
+
+	return 0;
+}
+
+
+/* afr_read_txn_wipe:
+
+   clean internal variables in @local in order to make
+   it possible to call afr_read_txn() multiple times from
+   the same frame
+*/
+
+void
+afr_read_txn_wipe (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;
+
+	local->readfn = NULL;
+
+	if (local->inode)
+		inode_unref (local->inode);
+
+	for (i = 0; i < priv->child_count; i++) {
+		local->read_attempted[i] = 0;
+		local->readable[i] = 0;
+	}
+}
+
+
+/*
+  afr_read_txn:
+
+  This is the read transaction function. The way it works:
+
+  - Determine read-subvolume from inode ctx.
+
+  - If read-subvolume's generation was stale, refresh ctx once by
+    calling afr_inode_refresh()
+
+    Else make an attempt to read on read-subvolume.
+
+  - If attempted read on read-subvolume fails, refresh ctx once
+    by calling afr_inode_refresh()
+
+  - After ctx refresh, query read-subvolume freshly and attempt
+    read once.
+
+  - If read fails, try every other readable[] subvolume before
+    finally giving up. readable[] elements are set by afr_inode_refresh()
+    based on dirty and pending flags.
+
+  - If file is in split brain in the backend, generation will be
+    kept 0 by afr_inode_refresh() and readable[] will be set 0 for
+    all elements. Therefore reads always fail.
+*/
+
+int
+afr_read_txn (call_frame_t *frame, xlator_t *this, inode_t *inode,
+	      afr_read_txn_wind_t readfn, afr_transaction_type type)
+{
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	int read_subvol = -1;
+	int event_generation = 0;
+	int ret = -1;
+
+	priv = this->private;
+	local = frame->local;
+
+	afr_read_txn_wipe (frame, this);
+
+	local->readfn = readfn;
+	local->inode = inode_ref (inode);
+
+	local->transaction.type = type;
+	ret = afr_inode_read_subvol_type_get (inode, this, local->readable,
+					      &event_generation, type);
+	if (ret == -1)
+		/* very first transaction on this inode */
+		goto refresh;
+
+	if (local->event_generation != event_generation)
+		/* servers have disconnected / reconnected, and possibly
+		   rebooted, very likely changing the state of freshness
+		   of copies */
+		goto refresh;
+
+	read_subvol = afr_read_subvol_select_by_policy (inode, this,
+							local->readable);
+
+	if (read_subvol < 0 || read_subvol > priv->child_count) {
+		gf_log (this->name, GF_LOG_WARNING, "Unreadable subvolume %d "
+			"found with event generation %d", read_subvol,
+			event_generation);
+		goto refresh;
+	}
+
+	if (!local->child_up[read_subvol]) {
+		/* should never happen, just in case */
+		gf_log (this->name, GF_LOG_WARNING, "subvolume %d is the "
+			"read subvolume in this generation, but is not up",
+			read_subvol);
+		goto refresh;
+	}
+
+	local->read_attempted[read_subvol] = 1;
+
+	local->readfn (frame, this, read_subvol);
+
+	return 0;
+
+refresh:
+	afr_inode_refresh (frame, this, inode, afr_read_txn_refresh_done);
+
+	return 0;
+}
diff --git a/xlators/cluster/afr/src/afr-self-heal-algorithm.c b/xlators/cluster/afr/src/afr-self-heal-algorithm.c
deleted file mode 100644
index e61f6defa..000000000
--- a/xlators/cluster/afr/src/afr-self-heal-algorithm.c
+++ /dev/null
@@ -1,1072 +0,0 @@
-/*
-   Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
-*/
-
-
-#include "glusterfs.h"
-#include "afr.h"
-#include "xlator.h"
-#include "dict.h"
-#include "xlator.h"
-#include "hashfn.h"
-#include "logging.h"
-#include "stack.h"
-#include "list.h"
-#include "call-stub.h"
-#include "defaults.h"
-#include "common-utils.h"
-#include "compat-errno.h"
-#include "compat.h"
-#include "byte-order.h"
-#include "md5.h"
-
-#include "afr-transaction.h"
-#include "afr-self-heal.h"
-#include "afr-self-heal-common.h"
-#include "afr-self-heal-algorithm.h"
-
-/*
-  This file contains the various self-heal algorithms
-*/
-
-
-/*
-  The "full" algorithm. Copies the entire file from
-  source to sinks.
-*/
-
-
-static void
-sh_full_private_cleanup (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t *               local   = NULL;
-        afr_self_heal_t *           sh      = NULL;
-        afr_sh_algo_full_private_t *sh_priv = NULL;
-
-        local = frame->local;
-        sh    = &local->self_heal;
-
-        sh_priv = sh->private;
-
-        if (sh_priv)
-                GF_FREE (sh_priv);
-}
-
-
-static int
-sh_full_loop_driver (call_frame_t *frame, xlator_t *this);
-
-static int
-sh_full_loop_return (call_frame_t *rw_frame, xlator_t *this, off_t offset)
-{
-        afr_local_t *               rw_local   = NULL;
-        afr_self_heal_t *           rw_sh      = NULL;
-
-        call_frame_t *sh_frame              = NULL;
-	afr_local_t * sh_local              = NULL;
-	afr_self_heal_t *sh                 = NULL;
-        afr_sh_algo_full_private_t *sh_priv = NULL;
-
-        rw_local = rw_frame->local;
-        rw_sh    = &rw_local->self_heal;
-
-        sh_frame = rw_sh->sh_frame;
-        sh_local = sh_frame->local;
-        sh       = &sh_local->self_heal;
-        sh_priv  = sh->private;
-
-        LOCK (&sh_priv->lock);
-        {
-                sh_priv->loops_running--;
-        }
-        UNLOCK (&sh_priv->lock);
-
-        gf_log (this->name, GF_LOG_TRACE,
-                "loop for offset %"PRId64" returned", offset);
-
-        AFR_STACK_DESTROY (rw_frame);
-
-        sh_full_loop_driver (sh_frame, this);
-
-        return 0;
-}
-
-
-static int
-sh_full_write_cbk (call_frame_t *rw_frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                   struct iatt *postbuf)
-{
-	afr_private_t * priv    = NULL;
-	afr_local_t * rw_local  = NULL;
-	afr_self_heal_t *rw_sh  = NULL;
-
-        call_frame_t *sh_frame  = NULL;
-	afr_local_t * sh_local  = NULL;
-	afr_self_heal_t *sh     = NULL;
-
-	int child_index = (long) cookie;
-	int call_count = 0;
-
-	priv = this->private;
-
-	rw_local = rw_frame->local;
-	rw_sh    = &rw_local->self_heal;
-
-        sh_frame = rw_sh->sh_frame;
-        sh_local = sh_frame->local;
-        sh       = &sh_local->self_heal;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"wrote %d bytes of data from %s to child %d, offset %"PRId64"",
-		op_ret, sh_local->loc.path, child_index,
-                rw_sh->offset - op_ret);
-
-	LOCK (&sh_frame->lock);
-	{
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"write to %s failed on subvolume %s (%s)",
-				sh_local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-
-			sh->op_failed = 1;
-		}
-	}
-	UNLOCK (&sh_frame->lock);
-
-	call_count = afr_frame_return (rw_frame);
-
-	if (call_count == 0) {
-		sh_full_loop_return (rw_frame, this, rw_sh->offset - op_ret);
-	}
-
-	return 0;
-}
-
-
-static int
-sh_full_read_cbk (call_frame_t *rw_frame, void *cookie,
-                  xlator_t *this, int32_t op_ret, int32_t op_errno,
-                  struct iovec *vector, int32_t count, struct iatt *buf,
-                  struct iobref *iobref)
-{
-	afr_private_t * priv    = NULL;
-	afr_local_t * rw_local  = NULL;
-	afr_self_heal_t *rw_sh  = NULL;
-
-        call_frame_t *sh_frame  = NULL;
-	afr_local_t * sh_local  = NULL;
-	afr_self_heal_t *sh     = NULL;
-
-	int i = 0;
-	int call_count = 0;
-
-	off_t offset = (long) cookie;
-
-	priv = this->private;
-	rw_local = rw_frame->local;
-	rw_sh    = &rw_local->self_heal;
-
-        sh_frame = rw_sh->sh_frame;
-        sh_local = sh_frame->local;
-        sh       = &sh_local->self_heal;
-
-	call_count = sh->active_sinks;
-
-        rw_local->call_count = call_count;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"read %d bytes of data from %s, offset %"PRId64"",
-		op_ret, sh_local->loc.path, offset);
-
-	if (op_ret <= 0) {
-                sh->op_failed = 1;
-
-                sh_full_loop_return (rw_frame, this, offset);
-		return 0;
-	}
-
-	rw_sh->offset += op_ret;
-
-	if (sh->file_has_holes) {
-		if (iov_0filled (vector, count) == 0) {
-			/* the iter function depends on the
-			   sh->offset already being updated
-			   above
-			*/
-
-                        sh_full_loop_return (rw_frame, this, offset);
-			goto out;
-		}
-	}
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->sources[i] || !sh_local->child_up[i])
-			continue;
-
-		/* this is a sink, so write to it */
-
-		STACK_WIND_COOKIE (rw_frame, sh_full_write_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->writev,
-				   sh->healing_fd, vector, count, offset,
-                                   iobref);
-
-		if (!--call_count)
-			break;
-	}
-
-out:
-	return 0;
-}
-
-
-static int
-sh_full_read_write (call_frame_t *frame, xlator_t *this, off_t offset)
-{
-	afr_private_t * priv    = NULL;
-	afr_local_t * local     = NULL;
-	afr_local_t * rw_local  = NULL;
-	afr_self_heal_t *rw_sh  = NULL;
-	afr_self_heal_t *sh     = NULL;
-
-        call_frame_t *rw_frame = NULL;
-
-        int32_t op_errno = 0;
-
-	priv  = this->private;
-        local = frame->local;
-        sh    = &local->self_heal;
-
-        rw_frame = copy_frame (frame);
-        if (!rw_frame)
-                goto out;
-
-        ALLOC_OR_GOTO (rw_local, afr_local_t, out);
-
-        rw_frame->local = rw_local;
-	rw_sh           = &rw_local->self_heal;
-
-        rw_sh->offset       = offset;
-        rw_sh->sh_frame     = frame;
-
-	STACK_WIND_COOKIE (rw_frame, sh_full_read_cbk,
-			   (void *) (long) offset,
-			   priv->children[sh->source],
-			   priv->children[sh->source]->fops->readv,
-			   sh->healing_fd, sh->block_size,
-			   offset);
-        return 0;
-
-out:
-        sh->op_failed = 1;
-
-        sh_full_loop_driver (frame, this);
-
-	return 0;
-}
-
-
-static int
-sh_full_loop_driver (call_frame_t *frame, xlator_t *this)
-{
-	afr_private_t * priv = NULL;
-	afr_local_t * local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-        afr_sh_algo_full_private_t *sh_priv = NULL;
-
-        int   loop    = 0;
-        int   recurse = 0;
-
-        off_t offset  = 0;
-
-	priv    = this->private;
-	local   = frame->local;
-	sh      = &local->self_heal;
-        sh_priv = sh->private;
-
-	if (sh->op_failed) {
-                if (sh_priv->loops_running == 0) {
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "full self-heal aborting on %s",
-                                local->loc.path);
-
-                        sh_full_private_cleanup (frame, this);
-                        local->self_heal.algo_abort_cbk (frame, this);
-                }
-
-		goto out;
-	}
-
-	if (sh_priv->offset >= sh->file_size) {
-                if (sh_priv->loops_running == 0) {
-
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "full self-heal completed on %s",
-                                local->loc.path);
-
-                        sh_full_private_cleanup (frame, this);
-                        local->self_heal.algo_completion_cbk (frame, this);
-                }
-
-		goto out;
-	}
-
-spawn:
-        loop    = 0;
-        recurse = 0;
-
-        LOCK (&sh_priv->lock);
-        {
-                if ((sh_priv->loops_running < priv->data_self_heal_window_size)
-                    && (sh_priv->offset < sh->file_size)) {
-
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "spawning a loop for offset %"PRId64,
-                                sh_priv->offset);
-
-                        offset           = sh_priv->offset;
-                        sh_priv->offset += sh->block_size;
-
-                        sh_priv->loops_running++;
-
-                        loop = 1;
-
-                        if (sh_priv->offset < sh->file_size)
-                                recurse = 1;
-                }
-        }
-        UNLOCK (&sh_priv->lock);
-
-        if (loop) {
-                sh_full_read_write (frame, this, offset);
-                if (recurse)
-                        goto spawn;
-        }
-
-out:
-	return 0;
-}
-
-
-int
-afr_sh_algo_full (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t *               local   = NULL;
-        afr_self_heal_t *           sh      = NULL;
-        afr_sh_algo_full_private_t *sh_priv = NULL;
-
-        local = frame->local;
-        sh    = &local->self_heal;
-
-        sh_priv = GF_CALLOC (1, sizeof (*sh_priv),
-                             gf_afr_mt_afr_private_t);
-
-        LOCK_INIT (&sh_priv->lock);
-
-        sh->private = sh_priv;
-
-        local->call_count = 0;
-
-        sh_full_loop_driver (frame, this);
-        return 0;
-}
-
-
-/*
- * The "diff" algorithm. Copies only those blocks whose checksums
- * don't match with those of source.
- */
-
-
-static void
-sh_diff_private_cleanup (call_frame_t *frame, xlator_t *this)
-{
-        afr_private_t *             priv    = NULL;
-        afr_local_t *               local   = NULL;
-        afr_self_heal_t *           sh      = NULL;
-        afr_sh_algo_diff_private_t *sh_priv = NULL;
-
-        int i;
-
-        priv  = this->private;
-        local = frame->local;
-        sh    = &local->self_heal;
-
-        sh_priv = sh->private;
-
-        for (i = 0; i < priv->data_self_heal_window_size; i++) {
-                if (sh_priv->loops[i]) {
-                        if (sh_priv->loops[i]->write_needed)
-                                GF_FREE (sh_priv->loops[i]->write_needed);
-
-                        if (sh_priv->loops[i]->checksum)
-                                GF_FREE (sh_priv->loops[i]->checksum);
-
-                        GF_FREE (sh_priv->loops[i]);
-                }
-        }
-
-        if (sh_priv) {
-                if (sh_priv->loops)
-                        GF_FREE (sh_priv->loops);
-
-                GF_FREE (sh_priv);
-        }
-
-
-}
-
-
-static uint32_t
-__make_cookie (int loop_index, int child_index)
-{
-        uint32_t ret = (loop_index << 16) | child_index;
-        return ret;
-}
-
-
-static int
-__loop_index (uint32_t cookie)
-{
-        return (cookie & 0xFFFF0000) >> 16;
-}
-
-
-static int
-__child_index (uint32_t cookie)
-{
-        return (cookie & 0x0000FFFF);
-}
-
-
-static void
-sh_diff_loop_state_reset (struct sh_diff_loop_state *loop_state, int child_count)
-{
-        loop_state->active = _gf_false;
-//        loop_state->offset = 0;
-
-        memset (loop_state->write_needed,
-                0, sizeof (*loop_state->write_needed) * child_count);
-
-        memset (loop_state->checksum,
-                0, MD5_DIGEST_LEN * child_count);
-}
-
-
-static int
-sh_diff_number_of_writes_needed (unsigned char *write_needed, int child_count)
-{
-        int writes = 0;
-        int i;
-
-        for (i = 0; i < child_count; i++) {
-                if (write_needed[i])
-                        writes++;
-        }
-
-        return writes;
-}
-
-
-static int
-sh_diff_loop_driver (call_frame_t *frame, xlator_t *this);
-
-
-static int
-sh_diff_loop_return (call_frame_t *rw_frame, xlator_t *this,
-                     struct sh_diff_loop_state *loop_state)
-{
-        afr_private_t *             priv       = NULL;
-        afr_local_t *               rw_local   = NULL;
-        afr_self_heal_t *           rw_sh      = NULL;
-
-        call_frame_t *sh_frame              = NULL;
-	afr_local_t * sh_local              = NULL;
-	afr_self_heal_t *sh                 = NULL;
-        afr_sh_algo_diff_private_t *sh_priv = NULL;
-
-        priv  = this->private;
-
-        rw_local = rw_frame->local;
-        rw_sh    = &rw_local->self_heal;
-
-        sh_frame = rw_sh->sh_frame;
-        sh_local = sh_frame->local;
-        sh       = &sh_local->self_heal;
-        sh_priv  = sh->private;
-
-        gf_log (this->name, GF_LOG_TRACE,
-                "loop for offset %"PRId64" returned", loop_state->offset);
-
-        LOCK (&sh_priv->lock);
-        {
-                sh_priv->loops_running--;
-                sh_diff_loop_state_reset (loop_state, priv->child_count);
-        }
-        UNLOCK (&sh_priv->lock);
-
-        AFR_STACK_DESTROY (rw_frame);
-
-        sh_diff_loop_driver (sh_frame, this);
-
-        return 0;
-}
-
-
-static int
-sh_diff_write_cbk (call_frame_t *rw_frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct iatt *buf,
-                   struct iatt *postbuf)
-{
-	afr_private_t *   priv     = NULL;
-	afr_local_t *     rw_local = NULL;
-	afr_self_heal_t * rw_sh    = NULL;
-
-        call_frame_t *sh_frame  = NULL;
-	afr_local_t * sh_local  = NULL;
-	afr_self_heal_t *sh     = NULL;
-
-        afr_sh_algo_diff_private_t *sh_priv;
-        struct sh_diff_loop_state *loop_state;
-
-	int call_count  = 0;
-        int child_index = 0;
-        int loop_index  = 0;
-
-	priv     = this->private;
-	rw_local = rw_frame->local;
-	rw_sh    = &rw_local->self_heal;
-
-        sh_frame = rw_sh->sh_frame;
-        sh_local = sh_frame->local;
-        sh       = &sh_local->self_heal;
-        sh_priv  = sh->private;
-
-        child_index = __child_index ((uint32_t) (long) cookie);
-        loop_index  = __loop_index ((uint32_t) (long) cookie);
-        loop_state  = sh_priv->loops[loop_index];
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"wrote %d bytes of data from %s to child %d, offset %"PRId64"",
-		op_ret, sh_local->loc.path, child_index,
-                loop_state->offset);
-
-	LOCK (&sh_frame->lock);
-	{
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"write to %s failed on subvolume %s (%s)",
-				sh_local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-
-			sh->op_failed = 1;
-		}
-	}
-	UNLOCK (&sh_frame->lock);
-
-	call_count = afr_frame_return (rw_frame);
-
-	if (call_count == 0) {
-		sh_diff_loop_return (rw_frame, this, loop_state);
-	}
-
-	return 0;
-}
-
-
-static int
-sh_diff_read_cbk (call_frame_t *rw_frame, void *cookie,
-                  xlator_t *this, int32_t op_ret, int32_t op_errno,
-                  struct iovec *vector, int32_t count, struct iatt *buf,
-                  struct iobref *iobref)
-{
-	afr_private_t *   priv     = NULL;
-	afr_local_t *     rw_local = NULL;
-	afr_self_heal_t * rw_sh    = NULL;
-
-        afr_sh_algo_diff_private_t * sh_priv = NULL;
-
-        call_frame_t *sh_frame  = NULL;
-	afr_local_t * sh_local  = NULL;
-	afr_self_heal_t *sh     = NULL;
-
-        int loop_index;
-        struct sh_diff_loop_state *loop_state;
-
-        uint32_t wcookie;
-
-	int i = 0;
-	int call_count = 0;
-
-	priv     = this->private;
-	rw_local = rw_frame->local;
-	rw_sh    = &rw_local->self_heal;
-
-        sh_frame = rw_sh->sh_frame;
-        sh_local = sh_frame->local;
-        sh       = &sh_local->self_heal;
-        sh_priv  = sh->private;
-
-        loop_index = __loop_index ((uint32_t) (long) cookie);
-        loop_state = sh_priv->loops[loop_index];
-
-	call_count = sh_diff_number_of_writes_needed (loop_state->write_needed,
-                                                      priv->child_count);
-
-	rw_local->call_count = call_count;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"read %d bytes of data from %s, offset %"PRId64"",
-		op_ret, sh_local->loc.path, loop_state->offset);
-
-	if ((op_ret <= 0) ||
-            (call_count == 0)) {
-                sh_diff_loop_return (rw_frame, this, loop_state);
-
-		return 0;
-	}
-
-	if (sh->file_has_holes) {
-		if (iov_0filled (vector, count) == 0) {
-
-                        sh_diff_loop_return (rw_frame, this, loop_state);
-			goto out;
-		}
-	}
-
-	for (i = 0; i < priv->child_count; i++) {
-                if (loop_state->write_needed[i]) {
-                        wcookie = __make_cookie (loop_index, i);
-
-                        STACK_WIND_COOKIE (rw_frame, sh_diff_write_cbk,
-                                           (void *) (long) wcookie,
-                                           priv->children[i],
-                                           priv->children[i]->fops->writev,
-                                           sh->healing_fd, vector, count,
-                                           loop_state->offset, iobref);
-
-                        if (!--call_count)
-                                break;
-                }
-        }
-
-out:
-	return 0;
-}
-
-
-static int
-sh_diff_read (call_frame_t *rw_frame, xlator_t *this,
-              int loop_index)
-{
-	afr_private_t *   priv     = NULL;
-	afr_local_t *     rw_local = NULL;
-	afr_self_heal_t * rw_sh    = NULL;
-
-        afr_sh_algo_diff_private_t * sh_priv = NULL;
-        struct sh_diff_loop_state *loop_state;
-
-        call_frame_t *sh_frame  = NULL;
-	afr_local_t * sh_local  = NULL;
-	afr_self_heal_t *sh     = NULL;
-
-        uint32_t cookie;
-
-	priv     = this->private;
-	rw_local = rw_frame->local;
-	rw_sh    = &rw_local->self_heal;
-
-        sh_frame = rw_sh->sh_frame;
-        sh_local = sh_frame->local;
-        sh       = &sh_local->self_heal;
-        sh_priv  = sh->private;
-
-        loop_state = sh_priv->loops[loop_index];
-
-        cookie = __make_cookie (loop_index, sh->source);
-
-	STACK_WIND_COOKIE (rw_frame, sh_diff_read_cbk,
-			   (void *) (long) cookie,
-			   priv->children[sh->source],
-			   priv->children[sh->source]->fops->readv,
-			   sh->healing_fd, sh_priv->block_size,
-			   loop_state->offset);
-
-	return 0;
-}
-
-
-static int
-sh_diff_checksum_cbk (call_frame_t *rw_frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno,
-                      uint32_t weak_checksum, uint8_t *strong_checksum)
-{
-	afr_private_t * priv    = NULL;
-	afr_local_t * rw_local  = NULL;
-	afr_self_heal_t *rw_sh  = NULL;
-
-        call_frame_t *sh_frame  = NULL;
-	afr_local_t * sh_local  = NULL;
-	afr_self_heal_t *sh     = NULL;
-
-        afr_sh_algo_diff_private_t * sh_priv = NULL;
-
-        int loop_index = 0;
-        int child_index = 0;
-        struct sh_diff_loop_state *loop_state;
-
-        int call_count   = 0;
-        int i            = 0;
-        int write_needed = 0;
-
-	priv  = this->private;
-
-	rw_local = rw_frame->local;
-	rw_sh    = &rw_local->self_heal;
-
-        sh_frame = rw_sh->sh_frame;
-        sh_local = sh_frame->local;
-        sh       = &sh_local->self_heal;
-
-        sh_priv = sh->private;
-
-        child_index = __child_index ((uint32_t) (long) cookie);
-        loop_index  = __loop_index ((uint32_t) (long) cookie);
-
-        loop_state  = sh_priv->loops[loop_index];
-
-        if (op_ret < 0) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "checksum on %s failed on subvolume %s (%s)",
-                        sh_local->loc.path, priv->children[child_index]->name,
-                        strerror (op_errno));
-
-                sh->op_failed = 1;
-        } else {
-                memcpy (loop_state->checksum + child_index * MD5_DIGEST_LEN,
-                        strong_checksum,
-                        MD5_DIGEST_LEN);
-        }
-
-        call_count = afr_frame_return (rw_frame);
-
-        if (call_count == 0) {
-                for (i = 0; i < priv->child_count; i++) {
-                        if (sh->sources[i] || !sh_local->child_up[i])
-                                continue;
-
-                        if (memcmp (loop_state->checksum + (i * MD5_DIGEST_LEN),
-                                    loop_state->checksum + (sh->source * MD5_DIGEST_LEN),
-                                    MD5_DIGEST_LEN)) {
-                                /*
-                                   Checksums differ, so this block
-                                   must be written to this sink
-                                */
-
-                                gf_log (this->name, GF_LOG_TRACE,
-                                        "checksum on subvolume %s at offset %"
-                                        PRId64" differs from that on source",
-                                        priv->children[i]->name, loop_state->offset);
-
-                                write_needed = loop_state->write_needed[i] = 1;
-                        }
-                }
-
-                LOCK (&sh_priv->lock);
-                {
-                        sh_priv->total_blocks++;
-                        if (write_needed)
-                                sh_priv->diff_blocks++;
-                }
-                UNLOCK (&sh_priv->lock);
-
-                if (write_needed && !sh->op_failed) {
-                        sh_diff_read (rw_frame, this, loop_index);
-                } else {
-                        sh->offset += sh_priv->block_size;
-
-                        sh_diff_loop_return (rw_frame, this, loop_state);
-                }
-        }
-
-        return 0;
-}
-
-
-static int
-sh_diff_find_unused_loop (afr_sh_algo_diff_private_t *sh_priv, int max)
-{
-        int i;
-
-        LOCK (&sh_priv->lock);
-        {
-                for (i = 0; i < max; i++) {
-                        if (sh_priv->loops[i]->active == _gf_false) {
-                                sh_priv->loops[i]->active = _gf_true;
-                                break;
-                        }
-                }
-        }
-        UNLOCK (&sh_priv->lock);
-
-        if (i == max) {
-                gf_log ("[sh-diff]", GF_LOG_ERROR,
-                        "no free loops found! This shouldn't happen. Please"
-                        " report this to gluster-devel@nongnu.org");
-        }
-
-        return i;
-}
-
-
-static int
-sh_diff_checksum (call_frame_t *frame, xlator_t *this, off_t offset)
-{
-	afr_private_t *   priv     = NULL;
-	afr_local_t *     local    = NULL;
-	afr_local_t *     rw_local = NULL;
-	afr_self_heal_t * sh       = NULL;
-	afr_self_heal_t * rw_sh    = NULL;
-
-        afr_sh_algo_diff_private_t * sh_priv = NULL;
-
-        call_frame_t *rw_frame = NULL;
-
-        uint32_t cookie;
-        int loop_index = 0;
-        struct sh_diff_loop_state *loop_state = NULL;
-
-        int32_t op_errno = 0;
-
-        int call_count = 0;
-        int i          = 0;
-
-	priv    = this->private;
-	local   = frame->local;
-	sh      = &local->self_heal;
-
-        sh_priv = sh->private;
-
-        rw_frame = copy_frame (frame);
-        if (!rw_frame)
-                goto out;
-
-        ALLOC_OR_GOTO (rw_local, afr_local_t, out);
-
-        rw_frame->local = rw_local;
-	rw_sh           = &rw_local->self_heal;
-
-        rw_sh->offset       = sh->offset;
-        rw_sh->sh_frame     = frame;
-
-        call_count = sh->active_sinks + 1;  /* sinks and source */
-
-        rw_local->call_count = call_count;
-
-        loop_index = sh_diff_find_unused_loop (sh_priv, priv->data_self_heal_window_size);
-
-        loop_state = sh_priv->loops[loop_index];
-        loop_state->offset       = offset;
-
-        /* we need to send both the loop index and child index,
-           so squeeze them both into a 32-bit number */
-
-        cookie = __make_cookie (loop_index, sh->source);
-
-        STACK_WIND_COOKIE (rw_frame, sh_diff_checksum_cbk,
-                           (void *) (long) cookie,
-                           priv->children[sh->source],
-                           priv->children[sh->source]->fops->rchecksum,
-                           sh->healing_fd,
-                           offset, sh_priv->block_size);
-
-        for (i = 0; i < priv->child_count; i++) {
-                if (sh->sources[i] || !local->child_up[i])
-                        continue;
-
-                cookie = __make_cookie (loop_index, i);
-
-                STACK_WIND_COOKIE (rw_frame, sh_diff_checksum_cbk,
-                                   (void *) (long) cookie,
-                                   priv->children[i],
-                                   priv->children[i]->fops->rchecksum,
-                                   sh->healing_fd,
-                                   offset, sh_priv->block_size);
-
-                if (!--call_count)
-                        break;
-        }
-
-        return 0;
-
-out:
-        sh->op_failed = 1;
-
-        sh_diff_loop_driver (frame, this);
-
-        return 0;
-}
-
-
-static int
-sh_diff_loop_driver (call_frame_t *frame, xlator_t *this)
-{
-	afr_private_t *   priv  = NULL;
-	afr_local_t *     local = NULL;
-	afr_self_heal_t * sh    = NULL;
-        afr_sh_algo_diff_private_t *sh_priv = NULL;
-
-        int   loop    = 0;
-        int   recurse = 0;
-
-        off_t offset = 0;
-        char  sh_type_str[256] = {0,};
- 
-	priv    = this->private;
-	local   = frame->local;
-	sh      = &local->self_heal;
-        sh_priv = sh->private;
-
-        afr_self_heal_type_str_get(sh, sh_type_str, sizeof(sh_type_str));
-
-	if (sh->op_failed) {
-                if (sh_priv->loops_running == 0) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "diff %s self-heal aborting on %s",
-                                sh_type_str, local->loc.path);
-
-                        sh_diff_private_cleanup (frame, this);
-                        local->self_heal.algo_abort_cbk (frame, this);
-                }
-
-		goto out;
-	}
-
-	if (sh_priv->offset >= sh->file_size) {
-                if (sh_priv->loops_running == 0) {
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "diff %s self-heal completed on %s",
-                                sh_type_str, local->loc.path);
-
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "diff %s self-heal on %s: %d blocks of %d were different (%.2f%%)",
-                                sh_type_str, local->loc.path,
-                                sh_priv->diff_blocks, sh_priv->total_blocks,
-                                ((sh_priv->diff_blocks * 1.0)/sh_priv->total_blocks) * 100);
-
-                        sh_diff_private_cleanup (frame, this);
-                        local->self_heal.algo_completion_cbk (frame, this);
-                }
-
-		goto out;
-	}
-
-spawn:
-        loop    = 0;
-        recurse = 0;
-
-        LOCK (&sh_priv->lock);
-        {
-                if ((sh_priv->loops_running < priv->data_self_heal_window_size)
-                    && (sh_priv->offset < sh->file_size)) {
-
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "spawning a loop for offset %"PRId64,
-                                sh_priv->offset);
-
-                        offset           = sh_priv->offset;
-                        sh_priv->offset += sh_priv->block_size;
-
-                        sh_priv->loops_running++;
-
-                        loop = 1;
-
-                        if (sh_priv->offset < sh->file_size)
-                                recurse = 1;
-                }
-        }
-        UNLOCK (&sh_priv->lock);
-
-        if (loop) {
-                sh_diff_checksum (frame, this, offset);
-                if (recurse)
-                    goto spawn;
-        }
-
-out:
-        return 0;
-}
-
-
-int
-afr_sh_algo_diff (call_frame_t *frame, xlator_t *this)
-{
-        afr_private_t *             priv    = NULL;
-        afr_local_t *               local   = NULL;
-        afr_self_heal_t *           sh      = NULL;
-        afr_sh_algo_diff_private_t *sh_priv = NULL;
-
-        int i;
-
-        priv  = this->private;
-        local = frame->local;
-        sh    = &local->self_heal;
-
-        sh_priv = GF_CALLOC (1, sizeof (*sh_priv),
-                             gf_afr_mt_afr_private_t);
-
-        sh_priv->block_size = this->ctx->page_size;
-
-        sh->private = sh_priv;
-
-        LOCK_INIT (&sh_priv->lock);
-
-        local->call_count = 0;
-
-        sh_priv->loops = GF_CALLOC (priv->data_self_heal_window_size,
-                                    sizeof (*sh_priv->loops),
-                                    gf_afr_mt_sh_diff_loop_state);
-
-        for (i = 0; i < priv->data_self_heal_window_size; i++) {
-                sh_priv->loops[i]               = GF_CALLOC (1, sizeof (*sh_priv->loops[i]),
-                                                             gf_afr_mt_sh_diff_loop_state);
-
-                sh_priv->loops[i]->checksum     = GF_CALLOC (priv->child_count,
-                                                             MD5_DIGEST_LEN, gf_afr_mt_uint8_t);
-                sh_priv->loops[i]->write_needed = GF_CALLOC (priv->child_count,
-                                                             sizeof (*sh_priv->loops[i]->write_needed),
-                                                             gf_afr_mt_char);
-        }
-
-        sh_diff_loop_driver (frame, this);
-
-        return 0;
-}
-
-
-struct afr_sh_algorithm afr_self_heal_algorithms[] = {
-        {.name = "full",  .fn = afr_sh_algo_full},
-        {.name = "diff",  .fn = afr_sh_algo_diff},
-        {0, 0},
-};
diff --git a/xlators/cluster/afr/src/afr-self-heal-algorithm.h b/xlators/cluster/afr/src/afr-self-heal-algorithm.h
deleted file mode 100644
index e45621b0e..000000000
--- a/xlators/cluster/afr/src/afr-self-heal-algorithm.h
+++ /dev/null
@@ -1,60 +0,0 @@
-/*
-   Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
-*/
-
-#ifndef __AFR_SELF_HEAL_ALGORITHM_H__
-#define __AFR_SELF_HEAL_ALGORITHM_H__
-
-
-typedef int (*afr_sh_algo_fn) (call_frame_t *frame,
-                               xlator_t *this);
-
-struct afr_sh_algorithm {
-        const char *name;
-        afr_sh_algo_fn fn;
-};
-
-extern struct afr_sh_algorithm afr_self_heal_algorithms[3];
-
-typedef struct {
-        gf_lock_t lock;
-        unsigned int loops_running;
-        off_t offset;
-} afr_sh_algo_full_private_t;
-
-struct sh_diff_loop_state {
-        off_t   offset;
-        unsigned char *write_needed;
-        uint8_t *checksum;
-        gf_boolean_t active;
-};
-
-typedef struct {
-        size_t block_size;
-
-        gf_lock_t lock;
-        unsigned int loops_running;
-        off_t offset;
-
-        int32_t total_blocks;
-        int32_t diff_blocks;
-
-        struct sh_diff_loop_state **loops;
-} afr_sh_algo_diff_private_t;
-
-#endif /* __AFR_SELF_HEAL_ALGORITHM_H__ */
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.c b/xlators/cluster/afr/src/afr-self-heal-common.c
index 1bb395cc3..4dac83113 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.c
+++ b/xlators/cluster/afr/src/afr-self-heal-common.c
@@ -1,1657 +1,1009 @@
 /*
-  Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
-#include "glusterfs.h"
-#include "xlator.h"
-#include "byte-order.h"
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
 
 #include "afr.h"
-#include "afr-transaction.h"
-#include "afr-self-heal-common.h"
 #include "afr-self-heal.h"
-#include "pump.h"
+#include "byte-order.h"
 
-/**
- * select_source - select a source and return it
- */
 
 int
-afr_sh_select_source (int sources[], int child_count)
+afr_selfheal_post_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+			  int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
 {
-	int i;
-	for (i = 0; i < child_count; i++)
-		if (sources[i])
-			return i;
-
-	return -1;
-}
+	afr_local_t *local = NULL;
 
+	local = frame->local;
 
-/**
- * sink_count - return number of sinks in sources array
- */
+	syncbarrier_wake (&local->barrier);
 
-int
-afr_sh_sink_count (int sources[], int child_count)
-{
-	int i;
-	int sinks = 0;
-	for (i = 0; i < child_count; i++)
-		if (!sources[i])
-			sinks++;
-	return sinks;
+	return 0;
 }
 
+
 int
-afr_sh_source_count (int sources[], int child_count)
+afr_selfheal_post_op (call_frame_t *frame, xlator_t *this, inode_t *inode,
+		      int subvol, dict_t *xattr)
 {
-	int i;
-	int nsource = 0;
+	afr_private_t *priv = NULL;
+	afr_local_t *local = NULL;
+	loc_t loc = {0, };
 
-	for (i = 0; i < child_count; i++)
-		if (sources[i])
-			nsource++;
-	return nsource;
-}
+	priv = this->private;
+	local = frame->local;
 
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-int
-afr_sh_supress_errenous_children (int sources[], int child_errno[],
-				  int child_count)
-{
-	int i = 0;
+	STACK_WIND (frame, afr_selfheal_post_op_cbk, priv->children[subvol],
+		    priv->children[subvol]->fops->xattrop, &loc,
+		    GF_XATTROP_ADD_ARRAY, xattr, NULL);
 
-	for (i = 0; i < child_count; i++) {
-		if (child_errno[i] && sources[i]) {
-			sources[i] = 0;
-		}
-	}
+	syncbarrier_wait (&local->barrier, 1);
 
 	return 0;
 }
 
 
-void
-afr_sh_print_pending_matrix (int32_t *pending_matrix[], xlator_t *this)
+dict_t *
+afr_selfheal_output_xattr (xlator_t *this, afr_transaction_type type,
+			   int *output_dirty, int **output_matrix, int subvol)
 {
-	afr_private_t * priv = this->private;
+	dict_t *xattr = NULL;
+	afr_private_t *priv = NULL;
+	int j = 0;
+	int idx = 0;
+	int ret = 0;
+	int *raw = 0;
+
+	priv = this->private;
+	idx = afr_index_for_transaction_type (type);
+
+	xattr = dict_new ();
+	if (!xattr)
+		return NULL;
+
+	if (output_dirty[subvol]) {
+		/* clear dirty */
+		raw = GF_CALLOC (sizeof(int), AFR_NUM_CHANGE_LOGS, gf_afr_mt_int32_t);
+		if (!raw)
+			goto err;
+
+		raw[idx] = hton32 (output_dirty[subvol]);
+		ret = dict_set_bin (xattr, AFR_DIRTY, raw,
+				    sizeof(int) * AFR_NUM_CHANGE_LOGS);
+		if (ret)
+			goto err;
+	}
 
-	char *buf = NULL;
-	char *ptr = NULL;
+	/* clear/set pending */
+	for (j = 0; j < priv->child_count; j++) {
+		if (!output_matrix[subvol][j])
+			continue;
 
-	int i, j;
+		raw = GF_CALLOC (sizeof(int), AFR_NUM_CHANGE_LOGS,
+				 gf_afr_mt_int32_t);
+		if (!raw)
+			goto err;
 
-        /* 10 digits per entry + 1 space + '[' and ']' */
-	buf = GF_MALLOC (priv->child_count * 11 + 8, gf_afr_mt_char);
+		raw[idx] = hton32 (output_matrix[subvol][j]);
 
-	for (i = 0; i < priv->child_count; i++) {
-		ptr = buf;
-		ptr += sprintf (ptr, "[ ");
-		for (j = 0; j < priv->child_count; j++) {
-			ptr += sprintf (ptr, "%d ", pending_matrix[i][j]);
-		}
-		sprintf (ptr, "]");
-		gf_log (this->name, GF_LOG_TRACE,
-			"pending_matrix: %s", buf);
+		ret = dict_set_bin (xattr, priv->pending_key[j],
+				    raw, sizeof(int) * AFR_NUM_CHANGE_LOGS);
+		if (ret)
+			goto err;
 	}
 
-	GF_FREE (buf);
+	return xattr;
+err:
+	if (xattr)
+		dict_unref (xattr);
+	return NULL;
 }
 
 
-void
-afr_sh_build_pending_matrix (afr_private_t *priv,
-                             int32_t *pending_matrix[], dict_t *xattr[],
-			     int child_count, afr_transaction_type type)
+int
+afr_selfheal_undo_pending (call_frame_t *frame, xlator_t *this, inode_t *inode,
+			   unsigned char *sources, unsigned char *sinks,
+			   unsigned char *healed_sinks, afr_transaction_type type,
+			   struct afr_reply *replies, unsigned char *locked_on)
 {
-	int i, j, k;
+	afr_private_t *priv = NULL;
+	int i = 0;
+	int j = 0;
+	unsigned char *pending = NULL;
+	int *input_dirty = NULL;
+	int **input_matrix = NULL;
+	int *output_dirty = NULL;
+	int **output_matrix = NULL;
+	dict_t *xattr = NULL;
 
-	/* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
-	int32_t pending[3];
-	void *pending_raw = NULL;
-	int ret = -1;
+	priv = this->private;
+
+	pending = alloca0 (priv->child_count);
+
+	input_dirty = alloca0 (priv->child_count * sizeof (int));
+	input_matrix = ALLOC_MATRIX (priv->child_count, int);
+	output_dirty = alloca0 (priv->child_count * sizeof (int));
+	output_matrix = ALLOC_MATRIX (priv->child_count, int);
 
-        unsigned char *ignorant_subvols = NULL;
+	afr_selfheal_extract_xattr (this, replies, type, input_dirty,
+				    input_matrix);
 
-        ignorant_subvols = GF_CALLOC (sizeof (*ignorant_subvols), child_count,
-                                      gf_afr_mt_char);
+	for (i = 0; i < priv->child_count; i++)
+		if (sinks[i] && !healed_sinks[i])
+			pending[i] = 1;
 
-	/* start clean */
-	for (i = 0; i < child_count; i++) {
-		for (j = 0; j < child_count; j++) {
-			pending_matrix[i][j] = 0;
+	for (i = 0; i < priv->child_count; i++) {
+		for (j = 0; j < priv->child_count; j++) {
+			if (pending[j])
+				output_matrix[i][j] = 1;
+			else
+				output_matrix[i][j] = -input_matrix[i][j];
 		}
 	}
 
-	for (i = 0; i < child_count; i++) {
-		pending_raw = NULL;
-
-                for (j = 0; j < child_count; j++) {
-                        ret = dict_get_ptr (xattr[i], priv->pending_key[j],
-                                            &pending_raw);
-                        
-                        if (ret != 0) {
-                                /*
-                                 * There is no xattr present. This means this
-                                 * subvolume should be considered an 'ignorant'
-                                 * subvolume.
-                                 */
-
-                                ignorant_subvols[i] = 1;
-                                continue;
-                        }
-
-			memcpy (pending, pending_raw, sizeof(pending));
-                        k = afr_index_for_transaction_type (type);
-                        
-                        pending_matrix[i][j] = ntoh32 (pending[k]);
-                }
+	for (i = 0; i < priv->child_count; i++) {
+		if (!pending[i])
+			output_dirty[i] = -input_dirty[i];
 	}
 
-        /*
-         * Make all non-ignorant subvols point towards the ignorant
-         * subvolumes.
-         */
-
-        for (i = 0; i < child_count; i++) {
-                if (ignorant_subvols[i]) {
-                        for (j = 0; j < child_count; j++) {
-                                if (!ignorant_subvols[j])
-                                        pending_matrix[j][i] += 1;
-                        }
-                }
-        }
-
-        GF_FREE (ignorant_subvols);
-}
-
-
-/**
- * mark_sources: Mark all 'source' nodes and return number of source
- * nodes found
- *
- * A node (a row in the pending matrix) belongs to one of
- * three categories:
- *
- * M is the pending matrix.
- *
- * 'innocent' - M[i] is all zeroes
- * 'fool'     - M[i] has i'th element = 1 (self-reference)
- * 'wise'     - M[i] has i'th element = 0, others are 1 or 0.
- *
- * All 'innocent' nodes are sinks. If all nodes are innocent, no self-heal is
- * needed.
- *
- * A 'wise' node can be a source. If two 'wise' nodes conflict, it is 
- * a split-brain. If one wise node refers to the other but the other doesn't
- * refer back, the referrer is a source.
- *
- * All fools are sinks, unless there are no 'wise' nodes. In that case,
- * one of the fools is made a source.
- */
-
-typedef enum {
-        AFR_NODE_INNOCENT,
-        AFR_NODE_FOOL,
-        AFR_NODE_WISE
-} afr_node_type;
-
-typedef struct {
-        afr_node_type type;
-        int           wisdom;
-} afr_node_character;
-
-
-static int
-afr_sh_is_innocent (int32_t *array, int child_count)
-{
-        int i   = 0;
-        int ret = 1;   /* innocent until proven guilty */
+	for (i = 0; i < priv->child_count; i++) {
+		if (!locked_on[i])
+			/* perform post-op only on subvols we had locked
+			   and inspected on.
+			*/
+			continue;
 
-        for (i = 0; i < child_count; i++) {
-                if (array[i]) {
-                        ret = 0;
-                        break;
-                }
-        }
+		xattr = afr_selfheal_output_xattr (this, type, output_dirty,
+						   output_matrix, i);
+		if (!xattr) {
+			gf_log (this->name, GF_LOG_ERROR,
+				"unable to allocate xdata for subvol %d", i);
+			continue;
+		}
 
-        return ret;
-}
+		afr_selfheal_post_op (frame, this, inode, i, xattr);
 
+		dict_unref (xattr);
+	}
 
-static int
-afr_sh_is_fool (int32_t *array, int i, int child_count)
-{
-        return array[i];   /* fool if accuses itself */
+	return 0;
 }
 
 
-static int
-afr_sh_is_wise (int32_t *array, int i, int child_count)
+void
+afr_replies_copy (struct afr_reply *dst, struct afr_reply *src, int count)
 {
-        return !array[i];  /* wise if does not accuse itself */
+	int i = 0;
+	dict_t *xdata = NULL;
+
+	if (dst == src)
+		return;
+
+	for (i = 0; i < count; i++) {
+		dst[i].valid = src[i].valid;
+		dst[i].op_ret = src[i].op_ret;
+		dst[i].op_errno = src[i].op_errno;
+		dst[i].prestat = src[i].prestat;
+		dst[i].poststat = src[i].poststat;
+		dst[i].preparent = src[i].preparent;
+		dst[i].postparent = src[i].postparent;
+		dst[i].preparent2 = src[i].preparent2;
+		dst[i].postparent2 = src[i].postparent2;
+		if (src[i].xdata)
+			xdata = dict_ref (src[i].xdata);
+		else
+			xdata = NULL;
+		if (dst[i].xdata)
+			dict_unref (dst[i].xdata);
+		dst[i].xdata = xdata;
+		memcpy (dst[i].checksum, src[i].checksum,
+			MD5_DIGEST_LENGTH);
+	}
 }
 
 
-static int
-afr_sh_all_nodes_innocent (afr_node_character *characters, 
-                           int child_count)
+int
+afr_selfheal_fill_dirty (xlator_t *this, int *dirty, int subvol,
+			 int idx, dict_t *xdata)
 {
-        int i   = 0;
-        int ret = 1;
-
-        for (i = 0; i < child_count; i++) {
-                if (characters[i].type != AFR_NODE_INNOCENT) {
-                        ret = 0;
-                        break;
-                }
-        }
+	void *pending_raw = NULL;
+	int pending[3] = {0, };
 
-        return ret;
-}
+	if (dict_get_ptr (xdata, AFR_DIRTY, &pending_raw))
+		return -1;
 
+	if (!pending_raw)
+		return -1;
 
-static int
-afr_sh_wise_nodes_exist (afr_node_character *characters, int child_count)
-{
-        int i   = 0;
-        int ret = 0;
+	memcpy (pending, pending_raw, sizeof(pending));
 
-        for (i = 0; i < child_count; i++) {
-                if (characters[i].type == AFR_NODE_WISE) {
-                        ret = 1;
-                        break;
-                }
-        }
+	dirty[subvol] = ntoh32 (pending[idx]);
 
-        return ret;
+	return 0;
 }
 
 
-/*
- * The 'wisdom' of a wise node is 0 if any other wise node accuses it.
- * It is 1 if no other wise node accuses it. 
- * Only wise nodes with wisdom 1 are sources.
- *
- * If no nodes with wisdom 1 exist, a split-brain has occured.
- */
-
-static void
-afr_sh_compute_wisdom (int32_t *pending_matrix[],
-                       afr_node_character characters[], int child_count)
+int
+afr_selfheal_fill_matrix (xlator_t *this, int **matrix, int subvol,
+			  int idx, dict_t *xdata)
 {
-        int i = 0;
-        int j = 0;
-
-        for (i = 0; i < child_count; i++) {
-                if (characters[i].type == AFR_NODE_WISE) {
-                        characters[i].wisdom = 1;
-
-                        for (j = 0; j < child_count; j++) {
-                                if ((characters[j].type == AFR_NODE_WISE)
-                                    && pending_matrix[j][i]) {
-                                        
-                                        characters[i].wisdom = 0;
-                                }
-                        }
-                }
-        }
-}
-
+	int i = 0;
+	void *pending_raw = NULL;
+	int pending[3] = {0, };
+	afr_private_t *priv = NULL;
 
-static int
-afr_sh_wise_nodes_conflict (afr_node_character *characters, 
-                            int child_count)
-{
-        int i   = 0;
-        int ret = 1;
+	priv = this->private;
 
-        for (i = 0; i < child_count; i++) {
-                if ((characters[i].type == AFR_NODE_WISE)
-                    && characters[i].wisdom == 1) {
+	for (i = 0; i < priv->child_count; i++) {
+		if (dict_get_ptr (xdata, priv->pending_key[i], &pending_raw))
+			continue;
 
-                        /* There is atleast one bona-fide wise node */
-                        ret = 0;
-                        break;
-                }
-        }
+		if (!pending_raw)
+			continue;
 
-        return ret;
-}
+		memcpy (pending, pending_raw, sizeof(pending));
 
+		matrix[subvol][i] = ntoh32 (pending[idx]);
+	}
 
-static int
-afr_sh_mark_wisest_as_sources (int sources[], 
-                               afr_node_character *characters, 
-                               int child_count)
-{
-        int nsources = 0;
-        
-        int i = 0;
-
-        for (i = 0; i < child_count; i++) {
-                if (characters[i].wisdom == 1) {
-                        sources[i] = 1;
-                        nsources++;
-                }
-        }
-
-        return nsources;
+	return 0;
 }
 
 
-static int
-afr_sh_mark_if_size_differs (afr_self_heal_t *sh, int child_count)
+int
+afr_selfheal_extract_xattr (xlator_t *this, struct afr_reply *replies,
+			    afr_transaction_type type, int *dirty, int **matrix)
 {
-        int32_t ** pending_matrix;
-        int i, j;
-
-        int size_differs = 0;
+	afr_private_t *priv = NULL;
+	int i = 0;
+	dict_t *xdata = NULL;
+	int idx = -1;
 
-        pending_matrix = sh->pending_matrix;
+	idx = afr_index_for_transaction_type (type);
 
-        for (i = 0; i < child_count; i++) {
-                for (j = 0; j < child_count; j++) {
-                        if (!sh->buf)
-                                break;
+	priv = this->private;
 
-                        if (SIZE_DIFFERS (&sh->buf[i], &sh->buf[j])
-                            && (pending_matrix[i][j] == 0)
-                            && (pending_matrix[j][i] == 0)) {
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].xdata)
+			continue;
 
-                                pending_matrix[i][j] = 1;
-                                pending_matrix[j][i] = 1;
+		xdata = replies[i].xdata;
 
-                                size_differs = 1;
-                        }
-                }
-        }
+		afr_selfheal_fill_dirty (this, dirty, i, idx, xdata);
+		afr_selfheal_fill_matrix (this, matrix, i, idx, xdata);
+	}
 
-        return size_differs;
+	return 0;
 }
 
-        
-static int
-afr_sh_mark_biggest_fool_as_source (afr_self_heal_t *sh,
-                                    afr_node_character *characters, 
-                                    int child_count)
-{
-        int i = 0;
-        int biggest = 0;
-
-        for (i = 0; i < child_count; i++) {
-                if (characters[i].type == AFR_NODE_FOOL) {
-                        biggest = i;
-                        break;
-                }
-        }
-
-        for (i = 0; i < child_count; i++) {
-                if (characters[i].type != AFR_NODE_FOOL)
-                        continue;
 
-                if (!sh->buf)
-                        break;
-
-                if (SIZE_GREATER (&sh->buf[i], &sh->buf[biggest])) {
-                        biggest = i;
-                }
-        }
-
-        sh->sources[biggest] = 1;
-
-        return 1;
-}
 
+/*
+ * This function determines if a self-heal is required for a given inode,
+ * and if needed, in what direction.
+ *
+ * locked_on[] is the array representing servers which have been locked and
+ * from which xattrs have been fetched for analysis.
+ *
+ * The output of the function is by filling the arrays sources[] and sinks[].
+ *
+ * sources[i] is set if i'th server is an eligible source for a selfheal.
+ *
+ * sinks[i] is set if i'th server needs to be healed.
+ *
+ * if sources[0..N] are all set, there is no need for a selfheal.
+ *
+ * if sinks[0..N] are all set, the inode is in split brain.
+ *
+ */
 
-static int
-afr_sh_mark_biggest_as_source (afr_self_heal_t *sh, int child_count)
+int
+afr_selfheal_find_direction (call_frame_t *frame, xlator_t *this,
+			     struct afr_reply *replies,
+			     afr_transaction_type type, unsigned char *locked_on,
+			     unsigned char *sources, unsigned char *sinks)
 {
-        int biggest = 0;
-        int i;
-
-        for (i = 0; i < child_count; i++) {
-                if (!sh->buf)
-                        break;
+	afr_private_t *priv = NULL;
+	int i = 0;
+	int j = 0;
+	int *dirty = NULL;
+	int **matrix = NULL;
+	char *accused = NULL;
 
-                if (SIZE_GREATER (&sh->buf[i], &sh->buf[biggest])) {
-                        biggest = i;
-                }
-        }
+	priv = this->private;
 
-        sh->sources[biggest] = 1;
+	dirty = alloca0 (priv->child_count * sizeof (int));
+	accused = alloca0 (priv->child_count);
+	matrix = ALLOC_MATRIX(priv->child_count, int);
 
-        return 1;
-}
+	/* First construct the pending matrix for further analysis */
+	afr_selfheal_extract_xattr (this, replies, type, dirty, matrix);
 
+	/* Next short list all accused to exclude them from being sources */
+	for (i = 0; i < priv->child_count; i++) {
+		for (j = 0; j < priv->child_count; j++) {
+			if (matrix[i][j])
+				accused[j] = 1;
+		}
+	}
 
-static int
-afr_sh_mark_loweia_uid_as_source (afr_self_heal_t *sh, int child_count)
-{
-        uid_t smallest = 0;
-        int i;
+	/* Short list all non-accused as sources */
+	memset (sources, 0, priv->child_count);
+	for (i = 0; i < priv->child_count; i++) {
+		if (!accused[i] && locked_on[i])
+			sources[i] = 1;
+	}
 
-        for (i = 0; i < child_count; i++) {
-                if (!sh->buf)
-                        break;
+	/* Everyone accused by sources are sinks */
+	memset (sinks, 0, priv->child_count);
+	for (i = 0; i < priv->child_count; i++) {
+		if (!sources[i])
+			continue;
+		for (j = 0; j < priv->child_count; j++) {
+			if (matrix[i][j])
+				sinks[j] = 1;
+		}
+	}
 
-                if (sh->buf[i].ia_uid < sh->buf[smallest].ia_uid) {
-                        smallest = i;
-                }
-        }
+	/* If any source has 'dirty' bit, pick first
+	   'dirty' source and make everybody else sinks */
+	for (i = 0; i < priv->child_count; i++) {
+		if (sources[i] && dirty[i]) {
+			for (j = 0; j < priv->child_count; j++) {
+				if (j != i) {
+					sources[j] = 0;
+					sinks[j] = 1;
+				}
+			}
+			break;
+		}
+	}
 
-        sh->sources[smallest] = 1;
+	/* If no sources, all locked nodes are sinks - split brain */
+	if (AFR_COUNT (sources, priv->child_count) == 0) {
+		for (i = 0; i < priv->child_count; i++) {
+			if (locked_on[i])
+				sinks[i] = 1;
+		}
+	}
 
-        return 1;
+	return 0;
 }
 
 
 int
-afr_sh_mark_sources (afr_self_heal_t *sh, int child_count,
-                     afr_self_heal_type type)
+afr_selfheal_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+			   int op_ret, int op_errno, inode_t *inode,
+			   struct iatt *buf, dict_t *xdata, struct iatt *parbuf)
 {
-	int i = 0;
-
-        int32_t ** pending_matrix;
-        int *      sources;
+	afr_local_t *local = NULL;
+	int i = -1;
 
-        int size_differs = 0;
-
-        pending_matrix = sh->pending_matrix;
-        sources        = sh->sources;
-
-	int nsources = 0;
-
-        /* stores the 'characters' (innocent, fool, wise) of the nodes */
-        afr_node_character *
-                characters = GF_CALLOC (sizeof (afr_node_character), 
-                                        child_count,
-                                        gf_afr_mt_afr_node_character) ;
+	local = frame->local;
+	i = (long) cookie;
 
-	/* start clean */
-	for (i = 0; i < child_count; i++) {
-		sources[i] = 0;
-	}
-        
-        for (i = 0; i < child_count; i++) {
-                if (afr_sh_is_innocent (pending_matrix[i], child_count)) {
-                        characters[i].type = AFR_NODE_INNOCENT;
-
-                } else if (afr_sh_is_fool (pending_matrix[i], i, child_count)) {
-                        characters[i].type = AFR_NODE_FOOL;
-
-                } else if (afr_sh_is_wise (pending_matrix[i], i, child_count)) {
-                        characters[i].type = AFR_NODE_WISE;
-
-                } else {
-                        gf_log ("[module:replicate]", GF_LOG_ERROR,
-                                "Could not determine the state of subvolume %d!"
-                                " (This message should never appear."
-                                " Please file a bug report to "
-                                "<gluster-devel@nongnu.org>.)", i);
-                }
-        }
-
-        if (type == AFR_SELF_HEAL_DATA) {
-                size_differs = afr_sh_mark_if_size_differs (sh, child_count);
-        }
-
-        if ((type == AFR_SELF_HEAL_METADATA)
-            && afr_sh_all_nodes_innocent (characters, child_count)) {
-
-                nsources = afr_sh_mark_loweia_uid_as_source (sh, child_count);
-                goto out;
-        }
-
-        if (afr_sh_all_nodes_innocent (characters, child_count)) {
-                if (size_differs) {
-                        nsources = afr_sh_mark_biggest_as_source (sh,
-                                                                  child_count);
-                }
-
-        } else if (afr_sh_wise_nodes_exist (characters, child_count)) {
-                afr_sh_compute_wisdom (pending_matrix, characters, child_count);
-
-                if (afr_sh_wise_nodes_conflict (characters, child_count)) {
-                        /* split-brain */
-
-                        nsources = -1;
-                        goto out;
-
-                } else {
-                        nsources = afr_sh_mark_wisest_as_sources (sources, 
-                                                                  characters,
-                                                                  child_count);
-                }
-        } else {
-                nsources = afr_sh_mark_biggest_fool_as_source (sh, characters,
-                                                               child_count);
-        }
+	local->replies[i].valid = 1;
+	local->replies[i].op_ret = op_ret;
+	local->replies[i].op_errno = op_errno;
+	if (buf)
+		local->replies[i].poststat = *buf;
+	if (parbuf)
+		local->replies[i].postparent = *parbuf;
+	if (xdata)
+		local->replies[i].xdata = dict_ref (xdata);
 
-out:
-        GF_FREE (characters);
+	syncbarrier_wake (&local->barrier);
 
-	return nsources;
+	return 0;
 }
 
 
-void
-afr_sh_pending_to_delta (afr_private_t *priv, dict_t **xattr,
-                         int32_t *delta_matrix[], int success[],
-                         int child_count, afr_transaction_type type)
+inode_t *
+afr_selfheal_unlocked_lookup_on (call_frame_t *frame, inode_t *parent,
+				 const char *name, struct afr_reply *replies,
+				 unsigned char *lookup_on)
 {
-	int i = 0;
-	int j = 0;
-        int k = 0;
+	loc_t loc = {0, };
+	dict_t *xattr_req = NULL;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	inode_t *inode = NULL;
 
-	/* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
-	int32_t   pending[3];
-	void    * pending_raw = NULL;
-        int       ret         = 0;
+	local = frame->local;
+	priv = frame->this->private;
 
-	/* start clean */
-	for (i = 0; i < child_count; i++) {
-		for (j = 0; j < child_count; j++) {
-			delta_matrix[i][j] = 0;
-		}
+	xattr_req = dict_new ();
+	if (!xattr_req)
+		return NULL;
+
+	if (afr_xattr_req_prepare (frame->this, xattr_req) != 0) {
+		dict_destroy (xattr_req);
+		return NULL;
 	}
 
-	for (i = 0; i < child_count; i++) {
-                pending_raw = NULL;
+	inode = inode_new (parent->table);
+	if (!inode) {
+		dict_destroy (xattr_req);
+		return NULL;
+	}
 
-                for (j = 0; j < child_count; j++) {
-                        ret = dict_get_ptr (xattr[i], priv->pending_key[j],
-                                            &pending_raw);
-                        if (ret < 0)
-                                gf_log ("afr_sh_pending_to_delta",
-                                        GF_LOG_WARNING,
-                                        "Unable to get dict value.");
+	loc.parent = inode_ref (parent);
+	uuid_copy (loc.pargfid, parent->gfid);
+	loc.name = name;
+	loc.inode = inode_ref (inode);
 
-                        if (!success[j])
-                                continue;
+	AFR_ONLIST (lookup_on, frame, afr_selfheal_discover_cbk, lookup, &loc,
+		    xattr_req);
 
-                        k = afr_index_for_transaction_type (type);
+	afr_replies_copy (replies, local->replies, priv->child_count);
 
-                        if (pending_raw) {
-				memcpy (pending, pending_raw, sizeof(pending));
-                                delta_matrix[i][j] = -(ntoh32 (pending[k]));
-                        } else {
-                                delta_matrix[i][j]  = 0;
-                        }
+	loc_wipe (&loc);
+	dict_unref (xattr_req);
 
-                }
-	}
+	return inode;
 }
 
 
 int
-afr_sh_delta_to_xattr (afr_private_t *priv,
-                       int32_t *delta_matrix[], dict_t *xattr[],
-		       int child_count, afr_transaction_type type)
+afr_selfheal_unlocked_discover_on (call_frame_t *frame, inode_t *inode,
+				   uuid_t gfid, struct afr_reply *replies,
+				   unsigned char *discover_on)
 {
-	int i = 0;
-	int j = 0;
-        int k = 0;
+	loc_t loc = {0, };
+	dict_t *xattr_req = NULL;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
 
-	int ret = 0;
+	local = frame->local;
+	priv = frame->this->private;
 
-	int32_t *pending = 0;
+	xattr_req = dict_new ();
+	if (!xattr_req)
+		return -ENOMEM;
 
-	for (i = 0; i < child_count; i++) {
-		if (!xattr[i])
-			continue;
+	if (afr_xattr_req_prepare (frame->this, xattr_req) != 0) {
+		dict_destroy (xattr_req);
+		return -ENOMEM;
+	}
 
-		for (j = 0; j < child_count; j++) {
-                        pending = GF_CALLOC (sizeof (int32_t), 3,
-                                             gf_afr_mt_int32_t);
-                        /* 3 = data+metadata+entry */
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, gfid);
 
-                        k = afr_index_for_transaction_type (type);
+	AFR_ONLIST (discover_on, frame, afr_selfheal_discover_cbk, lookup, &loc,
+		    xattr_req);
 
-			pending[k] = hton32 (delta_matrix[i][j]);
+	afr_replies_copy (replies, local->replies, priv->child_count);
+
+	loc_wipe (&loc);
+	dict_unref (xattr_req);
 
-                        ret = dict_set_bin (xattr[i], priv->pending_key[j],
-                                            pending,
-                                            3 * sizeof (int32_t));
-                        if (ret < 0)
-                                gf_log ("afr_sh_delta_to_xattr",
-                                        GF_LOG_WARNING,
-                                        "Unable to set dict value.");
-		}
-	}
 	return 0;
 }
 
-
 int
-afr_sh_has_metadata_pending (dict_t *xattr, int child_count, xlator_t *this)
+afr_selfheal_unlocked_discover (call_frame_t *frame, inode_t *inode,
+				uuid_t gfid, struct afr_reply *replies)
 {
 	afr_private_t *priv = NULL;
-	/* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
-	int32_t       pending[3];
-	void          *pending_raw = NULL;
 
-	int           ret = -1;
-	int            i  = 0;
-        int            j  = 0;
+	priv = frame->this->private;
 
-	priv = this->private;
+	return afr_selfheal_unlocked_discover_on (frame, inode, gfid, replies,
+						  priv->child_up);
+}
 
-        for (i = 0; i < priv->child_count; i++) {
-                ret = dict_get_ptr (xattr, priv->pending_key[i],
-                                    &pending_raw);
 
-                if (ret != 0)
-                        return 0;
+int
+afr_selfheal_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		       int op_ret, int op_errno, dict_t *xdata)
+{
+	afr_local_t *local = NULL;
+	int i = 0;
 
-		memcpy (pending, pending_raw, sizeof(pending));
-                j = afr_index_for_transaction_type (AFR_METADATA_TRANSACTION);
+	local = frame->local;
+	i = (long) cookie;
+
+	local->replies[i].valid = 1;
+	local->replies[i].op_ret = op_ret;
+	local->replies[i].op_errno = op_errno;
 
-                if (pending[j])
-                        return 1;
-        }
+	syncbarrier_wake (&local->barrier);
 
 	return 0;
 }
 
 
 int
-afr_sh_has_data_pending (dict_t *xattr, int child_count, xlator_t *this)
+afr_selfheal_locked_fill (call_frame_t *frame, xlator_t *this,
+			  unsigned char *locked_on)
 {
+	int i = 0;
 	afr_private_t *priv = NULL;
-	/* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
-	int32_t       pending[3];
-	void          *pending_raw = NULL;
-
-	int           ret = -1;
-	int            i  = 0;
-        int            j  = 0;
+	afr_local_t *local = NULL;
+	int count = 0;
 
+	local = frame->local;
 	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;
-        }
+	for (i = 0; i < priv->child_count; i++) {
+		if (local->replies[i].valid && local->replies[i].op_ret == 0) {
+			locked_on[i] = 1;
+			count++;
+		} else {
+			locked_on[i] = 0;
+		}
+	}
 
-	return 0;
+	return count;
 }
 
 
 int
-afr_sh_has_entry_pending (dict_t *xattr, int child_count, xlator_t *this)
+afr_selfheal_tryinodelk (call_frame_t *frame, xlator_t *this, inode_t *inode,
+			 char *dom, off_t off, size_t size,
+			 unsigned char *locked_on)
 {
-        afr_private_t *priv = NULL;
-	/* Indexable by result of afr_index_for_transaction_type(): 0 -- 2. */
-	int32_t       pending[3];
-	void          *pending_raw = NULL;
-
-	int           ret = -1;
-	int            i  = 0;
-        int            j  = 0;
+	loc_t loc = {0,};
+	struct gf_flock flock = {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;
-}
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
+	flock.l_type = F_WRLCK;
+	flock.l_start = off;
+	flock.l_len = size;
 
-/**
- * is_matrix_zero - return true if pending matrix is all zeroes
- */
+	AFR_ONALL (frame, afr_selfheal_lock_cbk, inodelk, dom,
+		    &loc, F_SETLK, &flock, NULL);
 
-int
-afr_sh_is_matrix_zero (int32_t *pending_matrix[], int child_count)
-{
-	int i, j;
+	loc_wipe (&loc);
 
-	for (i = 0; i < child_count; i++) 
-		for (j = 0; j < child_count; j++) 
-			if (pending_matrix[i][j]) 
-				return 0;
-	return 1;
+	return afr_selfheal_locked_fill (frame, this, locked_on);
 }
 
 
 int
-afr_sh_missing_entries_done (call_frame_t *frame, xlator_t *this)
+afr_selfheal_inodelk (call_frame_t *frame, xlator_t *this, inode_t *inode,
+		      char *dom, off_t off, size_t size,
+		      unsigned char *locked_on)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              i = 0;
+	loc_t loc = {0,};
+	struct gf_flock flock = {0, };
+	afr_local_t *local = NULL;
+	int i = 0;
+	afr_private_t *priv = NULL;
 
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
+	local = frame->local;
 
-//	memset (sh->child_errno, 0, sizeof (int) * priv->child_count);
-	memset (sh->buf, 0, sizeof (struct iatt) * priv->child_count);
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-        for (i = 0; i < priv->child_count; i++) {
-                sh->locked_nodes[i] = 0;
-        }
+	flock.l_type = F_WRLCK;
+	flock.l_start = off;
+	flock.l_len = size;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->xattr[i])
-			dict_unref (sh->xattr[i]);
-		sh->xattr[i] = NULL;
-	}
+	AFR_ONALL (frame, afr_selfheal_lock_cbk, inodelk, dom,
+		    &loc, F_SETLK, &flock, NULL);
 
-	if (local->govinda_gOvinda) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"aborting selfheal of %s",
-			local->loc.path);
-		sh->completion_cbk (frame, this);
-	} else {
-		gf_log (this->name, GF_LOG_TRACE,
-			"proceeding to metadata check on %s",
-			local->loc.path);
-		afr_self_heal_metadata (frame, this);
+	for (i = 0; i < priv->child_count; i++) {
+		if (local->replies[i].op_ret == -1 &&
+		    local->replies[i].op_errno == EAGAIN) {
+			afr_selfheal_locked_fill (frame, this, locked_on);
+			afr_selfheal_uninodelk (frame, this, inode, dom, off,
+						size, locked_on);
+
+			AFR_SEQ (frame, afr_selfheal_lock_cbk, inodelk, dom,
+				 &loc, F_SETLKW, &flock, NULL);
+			break;
+		}
 	}
 
-	return 0;
-}
-
-
-static int
-sh_missing_entries_finish (call_frame_t *frame, xlator_t *this)
-{
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-
-	local = frame->local;
-        int_lock = &local->internal_lock;
-
-        int_lock->lock_cbk = afr_sh_missing_entries_done;
-        afr_unlock (frame, this);
+	loc_wipe (&loc);
 
-	return 0;
+	return afr_selfheal_locked_fill (frame, this, locked_on);
 }
 
 
-static int
-sh_destroy_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		int32_t op_ret, int op_errno,
-                struct iatt *preop, struct iatt *postop)
+int
+afr_selfheal_uninodelk (call_frame_t *frame, xlator_t *this, inode_t *inode,
+			char *dom, off_t off, size_t size,
+			const unsigned char *locked_on)
 {
-        afr_local_t *local = NULL;
+	loc_t loc = {0,};
+	struct gf_flock flock = {0, };
 
-        loc_t *parent_loc = cookie;
 
-        int call_count = 0;
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-        local = frame->local;
+	flock.l_type = F_UNLCK;
+	flock.l_start = off;
+	flock.l_len = size;
 
-        if (op_ret == -1) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "setattr on %s failed: %s",
-                        local->loc.path, strerror (op_errno));
-        }
+	AFR_ONLIST (locked_on, frame, afr_selfheal_lock_cbk, inodelk,
+		    dom, &loc, F_SETLK, &flock, NULL);
 
-        if (parent_loc) {
-                loc_wipe (parent_loc);
-                GF_FREE (parent_loc);
-        }
+	loc_wipe (&loc);
 
-        call_count = afr_frame_return (frame);
-        
-        if (call_count == 0) {
-                STACK_DESTROY (frame->root);
-        }
-        
 	return 0;
 }
 
 
-static int
-sh_missing_entries_newentry_cbk (call_frame_t *frame, void *cookie,
-				 xlator_t *this,
-				 int32_t op_ret, int32_t op_errno,
-				 inode_t *inode, struct iatt *buf,
-                                 struct iatt *preparent,
-                                 struct iatt *postparent)
+int
+afr_selfheal_tryentrylk (call_frame_t *frame, xlator_t *this, inode_t *inode,
+			 char *dom, const char *name, unsigned char *locked_on)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	call_frame_t    *setattr_frame = NULL;
-	int              call_count = 0;
-	int              child_index = 0;
+	loc_t loc = {0,};
 
-        loc_t *parent_loc = NULL;
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-	struct iatt     stbuf;
-        int32_t valid;
+	AFR_ONALL (frame, afr_selfheal_lock_cbk, entrylk, dom,
+		   &loc, name, ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL);
 
-	local = frame->local;
-	sh    = &local->self_heal;
-	priv  = this->private;
-
-	child_index = (long) cookie;
-
-	stbuf.ia_atime = sh->buf[sh->source].ia_atime;
-	stbuf.ia_atime_nsec = sh->buf[sh->source].ia_atime_nsec;
-	stbuf.ia_mtime = sh->buf[sh->source].ia_mtime;
-	stbuf.ia_mtime_nsec = sh->buf[sh->source].ia_mtime_nsec;
-
-        stbuf.ia_uid = sh->buf[sh->source].ia_uid;
-        stbuf.ia_gid = sh->buf[sh->source].ia_gid;
-        
-        valid = GF_SET_ATTR_UID   | GF_SET_ATTR_GID |
-                GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
-        
-	if (op_ret == 0) {
-		setattr_frame = copy_frame (frame);
-                
-                setattr_frame->local = GF_CALLOC (1, sizeof (afr_local_t),
-                                                  gf_afr_mt_afr_local_t);
-
-                ((afr_local_t *)setattr_frame->local)->call_count = 2;
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"setattr (%s) on subvolume %s",
-			local->loc.path, priv->children[child_index]->name);
-
-		STACK_WIND_COOKIE (setattr_frame, sh_destroy_cbk,
-                                   (void *) (long) 0,
-                                   priv->children[child_index],
-                                   priv->children[child_index]->fops->setattr,
-                                   &local->loc, &stbuf, valid);
-
-                valid      = GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
-                parent_loc = GF_CALLOC (1, sizeof (*parent_loc), 
-                                        gf_afr_mt_loc_t);
-                afr_build_parent_loc (parent_loc, &local->loc);
-
-                STACK_WIND_COOKIE (setattr_frame, sh_destroy_cbk,
-                                   (void *) (long) parent_loc,
-                                   priv->children[child_index],
-                                   priv->children[child_index]->fops->setattr,
-                                   parent_loc, &sh->parentbuf, valid);
-	}
-
-	call_count = afr_frame_return (frame);
+	loc_wipe (&loc);
 
-	if (call_count == 0) {
-		sh_missing_entries_finish (frame, this);
-	}
-
-	return 0;
+	return afr_selfheal_locked_fill (frame, this, locked_on);
 }
 
 
-static int
-sh_missing_entries_mknod (call_frame_t *frame, xlator_t *this)
+int
+afr_selfheal_entrylk (call_frame_t *frame, xlator_t *this, inode_t *inode,
+		      char *dom, const char *name, unsigned char *locked_on)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              i = 0;
-        int              ret = 0;
-	int              enoent_count = 0;
-	int              call_count = 0;
-	mode_t           st_mode = 0;
-	dev_t            ia_dev = 0;
-        dict_t          *dict = NULL;
+	loc_t loc = {0,};
+	afr_local_t *local = NULL;
+	int i = 0;
+	afr_private_t *priv = NULL;
 
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
+	local = frame->local;
 
-	for (i = 0; i < priv->child_count; i++)
-		if (sh->child_errno[i] == ENOENT)
-			enoent_count++;
-
-	call_count = enoent_count;
-	local->call_count = call_count;
-
-	st_mode = st_mode_from_ia (sh->buf[sh->source].ia_prot,
-                                   sh->buf[sh->source].ia_type);
-	ia_dev  = sh->buf[sh->source].ia_dev;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"mknod %s mode 0%o on %d subvolumes",
-		local->loc.path, st_mode, enoent_count);
-
-        dict = dict_new ();
-        if (!dict)
-                gf_log (this->name, GF_LOG_ERROR, "out of memory");
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-        ret = afr_set_dict_gfid (dict, sh->buf[sh->source].ia_gfid);
-        if (ret)
-                gf_log (this->name, GF_LOG_DEBUG, "gfid set failed");
+	AFR_ONALL (frame, afr_selfheal_lock_cbk, entrylk, dom, &loc,
+		   name, ENTRYLK_LOCK_NB, ENTRYLK_WRLCK, NULL);
 
 	for (i = 0; i < priv->child_count; i++) {
-		if (sh->child_errno[i] == ENOENT) {
-			STACK_WIND_COOKIE (frame,
-					   sh_missing_entries_newentry_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->mknod,
-					   &local->loc, st_mode, ia_dev, dict);
-			if (!--call_count)
-				break;
+		if (local->replies[i].op_ret == -1 &&
+		    local->replies[i].op_errno == EAGAIN) {
+			afr_selfheal_locked_fill (frame, this, locked_on);
+			afr_selfheal_unentrylk (frame, this, inode, dom, name,
+						locked_on);
+
+			AFR_SEQ (frame, afr_selfheal_lock_cbk, entrylk, dom,
+				 &loc, name, ENTRYLK_LOCK, ENTRYLK_WRLCK, NULL);
+			break;
 		}
 	}
 
-        if (dict)
-                dict_unref (dict);
+	loc_wipe (&loc);
 
-	return 0;
+	return afr_selfheal_locked_fill (frame, this, locked_on);
 }
 
 
-static int
-sh_missing_entries_mkdir (call_frame_t *frame, xlator_t *this)
+int
+afr_selfheal_unentrylk (call_frame_t *frame, xlator_t *this, inode_t *inode,
+			char *dom, const char *name, unsigned char *locked_on)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-        dict_t          *dict = NULL;
-	int              i = 0;
-        int              ret = 0;
-	int              enoent_count = 0;
-	int              call_count = 0;
-	mode_t           st_mode = 0;
+	loc_t loc = {0,};
 
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	for (i = 0; i < priv->child_count; i++)
-		if (sh->child_errno[i] == ENOENT)
-			enoent_count++;
-
-	call_count = enoent_count;
-	local->call_count = call_count;
-
-	st_mode = st_mode_from_ia (sh->buf[sh->source].ia_prot,
-                                   sh->buf[sh->source].ia_type);
+	AFR_ONLIST (locked_on, frame, afr_selfheal_lock_cbk, entrylk,
+		    dom, &loc, name, ENTRYLK_UNLOCK, ENTRYLK_WRLCK, NULL);
 
-        dict = dict_new ();
-        if (!dict) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                sh_missing_entries_finish (frame, this);
-                return 0;
-        }
-
-        ret = afr_set_dict_gfid (dict, sh->buf[sh->source].ia_gfid);
-        if (ret)
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "inode gfid set failed");
-
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"mkdir %s mode 0%o on %d subvolumes",
-		local->loc.path, st_mode, enoent_count);
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->child_errno[i] == ENOENT) {
-                        if (!strcmp (local->loc.path, "/")) {
-                                /* We shouldn't try to create "/" */
-
-                                sh_missing_entries_finish (frame, this);
-
-                                return 0;
-                        } else {
-                                STACK_WIND_COOKIE (frame,
-                                                   sh_missing_entries_newentry_cbk,
-                                                   (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->mkdir,
-                                                   &local->loc, st_mode, dict);
-                                if (!--call_count)
-                                        break;
-                        }
-		}
-	}
-
-        if (dict)
-                dict_unref (dict);
+	loc_wipe (&loc);
 
 	return 0;
 }
 
 
-static int
-sh_missing_entries_symlink (call_frame_t *frame, xlator_t *this,
-			    const char *link, struct iatt *buf)
+gf_boolean_t
+afr_is_pending_set (xlator_t *this, dict_t *xdata, int type)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-        dict_t          *dict = NULL;
-	int              i = 0;
-        int              ret = 0;
-	int              enoent_count = 0;
-	int              call_count = 0;
-
+	int idx = -1;
+	afr_private_t *priv = NULL;
+	void *pending_raw = NULL;
+	int *pending_int = NULL;
+	int i = 0;
 
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
+	idx = afr_index_for_transaction_type (type);
 
-	for (i = 0; i < priv->child_count; i++)
-		if (sh->child_errno[i] == ENOENT)
-			enoent_count++;
-
-	call_count = enoent_count;
-	local->call_count = call_count;
+	if (dict_get_ptr (xdata, AFR_DIRTY, &pending_raw) == 0) {
+		if (pending_raw) {
+			pending_int = pending_raw;
 
-        dict = dict_new ();
-        if (!dict) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                sh_missing_entries_finish (frame, this);
-                return 0;
-        }
-
-        ret = afr_set_dict_gfid (dict, buf->ia_gfid);
-        if (ret)
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "dict gfid set failed");
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"symlink %s -> %s on %d subvolumes",
-		local->loc.path, link, enoent_count);
+			if (ntoh32 (pending_int[idx]))
+				return _gf_true;
+		}
+	}
 
 	for (i = 0; i < priv->child_count; i++) {
-		if (sh->child_errno[i] == ENOENT) {
-			STACK_WIND_COOKIE (frame,
-					   sh_missing_entries_newentry_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->symlink,
-					   link, &local->loc, dict);
-			if (!--call_count)
-				break;
-		}
+		if (dict_get_ptr (xdata, priv->pending_key[i],
+				  &pending_raw))
+			continue;
+		if (!pending_raw)
+			continue;
+		pending_int = pending_raw;
+
+		if (ntoh32 (pending_int[idx]))
+			return _gf_true;
 	}
 
-	return 0;
+	return _gf_false;
 }
 
 
-static int
-sh_missing_entries_readlink_cbk (call_frame_t *frame, void *cookie,
-				 xlator_t *this,
-				 int32_t op_ret, int32_t op_errno,
-				 const char *link, struct iatt *sbuf)
+gf_boolean_t
+afr_is_data_set (xlator_t *this, dict_t *xdata)
 {
-	if (op_ret > 0)
-		sh_missing_entries_symlink (frame, this, link, sbuf);
-	else
-		sh_missing_entries_finish (frame, this);
-
-	return 0;
+	return afr_is_pending_set (this, xdata, AFR_DATA_TRANSACTION);
 }
 
-
-static int
-sh_missing_entries_readlink (call_frame_t *frame, xlator_t *this)
+gf_boolean_t
+afr_is_metadata_set (xlator_t *this, dict_t *xdata)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	STACK_WIND (frame, sh_missing_entries_readlink_cbk,
-		    priv->children[sh->source],
-		    priv->children[sh->source]->fops->readlink,
-		    &local->loc, 4096);
-
-	return 0;
+	return afr_is_pending_set (this, xdata, AFR_METADATA_TRANSACTION);
 }
 
-
-static int
-sh_missing_entries_create (call_frame_t *frame, xlator_t *this)
+gf_boolean_t
+afr_is_entry_set (xlator_t *this, dict_t *xdata)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	int              type = 0;
-	int              i = 0;
-	afr_private_t   *priv = NULL;
-	int              enoent_count = 0;
-	int              govinda_gOvinda = 0;
-
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	for (i = 0; i < priv->child_count; i++) {
-                if (!local->child_up[i])
-                        continue;
-
-		if (sh->child_errno[i]) {
-			if (sh->child_errno[i] == ENOENT)
-				enoent_count++;
-		} else {
-			if (type) {
-				if (type != sh->buf[i].ia_type) {
-                                        gf_log (this->name, GF_LOG_TRACE,
-                                                "file %s is govinda!",
-                                                local->loc.path);
-
-					govinda_gOvinda = 1;
-                                }
-			} else {
-				sh->source = i;
-				type = sh->buf[i].ia_type;
-			}
-		}
-	}
+	return afr_is_pending_set (this, xdata, AFR_ENTRY_TRANSACTION);
+}
 
-	if (govinda_gOvinda) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"conflicting filetypes exist for path %s. returning.",
-			local->loc.path);
 
-		local->govinda_gOvinda = 1;
-		sh_missing_entries_finish (frame, this);
-		return 0;
-	}
+void
+afr_inode_link (inode_t *inode, struct iatt *iatt)
+{
+	inode_t *linked_inode = NULL;
 
-	if (!type) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"no source found for %s. all nodes down?. returning.",
-			local->loc.path);
-		/* subvolumes down and/or file does not exist */
-		sh_missing_entries_finish (frame, this);
-		return 0;
-	}
+	linked_inode = inode_link (inode, NULL, NULL, iatt);
 
-	if (enoent_count == 0) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"no missing files - %s. proceeding to metadata check",
-			local->loc.path);
-		/* proceed to next step - metadata self-heal */
-		sh_missing_entries_finish (frame, this);
-		return 0;
-	}
+	uuid_copy (inode->gfid, iatt->ia_gfid);
+	inode->ia_type = iatt->ia_type;
 
-	switch (type) {
-	case IA_IFSOCK:
-	case IA_IFREG:
-	case IA_IFBLK:
-	case IA_IFCHR:
-	case IA_IFIFO:
-		sh_missing_entries_mknod (frame, this);
-		break;
-	case IA_IFLNK:
-		sh_missing_entries_readlink (frame, this);
-		break;
-	case IA_IFDIR:
-		sh_missing_entries_mkdir (frame, this);
-		break;
-	default:
-		gf_log (this->name, GF_LOG_ERROR,
-			"unknown file type: 0%o", type);
-		local->govinda_gOvinda = 1;
-		sh_missing_entries_finish (frame, this);
+	if (linked_inode) {
+		inode_lookup (linked_inode);
+		inode_unref (linked_inode);
 	}
-
-	return 0;
 }
 
 
-static int
-sh_missing_entries_lookup_cbk (call_frame_t *frame, void *cookie,
-			       xlator_t *this,
-			       int32_t op_ret, int32_t op_errno,
-                               inode_t *inode, struct iatt *buf, dict_t *xattr,
-                               struct iatt *postparent)
-{
-	int              child_index = 0;
-	afr_local_t     *local = NULL;
-	int              call_count = 0;
-	afr_private_t   *priv = NULL;
+/*
+ * This function inspects the looked up replies (in an unlocked manner)
+ * and decides whether a locked verification and possible healing is
+ * required or not. It updates the three booleans for each type
+ * of healing. If the boolean flag gets set to FALSE, then we are sure
+ * no healing is required. If the boolean flag gets set to TRUE then
+ * we have to proceed with locked reinspection.
+ */
 
+int
+afr_selfheal_unlocked_inspect (call_frame_t *frame, xlator_t *this,
+			       inode_t *inode, uuid_t gfid,
+			       gf_boolean_t *data_selfheal,
+			       gf_boolean_t *metadata_selfheal,
+			       gf_boolean_t *entry_selfheal)
+{
+	afr_private_t *priv = NULL;
+	int i = 0;
+	int valid_cnt = 0;
+	struct iatt first = {0, };
+	struct afr_reply *replies = NULL;
+	int ret = -1;
 
-	local = frame->local;
 	priv = this->private;
 
-	child_index = (long) cookie;
+	replies = alloca0 (sizeof (*replies) * priv->child_count);
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"path %s on subvolume %s is of mode 0%o",
-				local->loc.path,
-				priv->children[child_index]->name,
-				buf->ia_type);
+	ret = afr_selfheal_unlocked_discover (frame, inode, gfid, replies);
+	if (ret)
+		return ret;
 
-			local->self_heal.buf[child_index] = *buf;
-                        local->self_heal.parentbuf        = *postparent;
-		} else {
-			gf_log (this->name, GF_LOG_TRACE,
-				"path %s on subvolume %s => -1 (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].valid)
+			continue;
+		if (replies[i].op_ret == -1)
+			continue;
 
-			local->self_heal.child_errno[child_index] = op_errno;
-		}
+		if (afr_is_data_set (this, replies[i].xdata))
+			*data_selfheal = _gf_true;
 
-	}
-	UNLOCK (&frame->lock);
+		if (afr_is_metadata_set (this, replies[i].xdata))
+			*metadata_selfheal = _gf_true;
 
-	call_count = afr_frame_return (frame);
+		if (afr_is_entry_set (this, replies[i].xdata))
+			*entry_selfheal = _gf_true;
 
-	if (call_count == 0) {
-		sh_missing_entries_create (frame, this);
-	}
+		valid_cnt ++;
+		if (valid_cnt == 1) {
+			first = replies[i].poststat;
+			continue;
+		}
 
-	return 0;
-}
+		if (!IA_EQUAL (first, replies[i].poststat, type)) {
+			gf_log (this->name, GF_LOG_ERROR,
+				"TYPE mismatch %d vs %d on %s for gfid:%s",
+				(int) first.ia_type,
+				(int) replies[i].poststat.ia_type,
+				priv->children[i]->name,
+				uuid_utoa (replies[i].poststat.ia_gfid));
+			return -EIO;
+		}
 
+		if (!IA_EQUAL (first, replies[i].poststat, uid)) {
+			gf_log (this->name, GF_LOG_DEBUG,
+				"UID mismatch %d vs %d on %s for gfid:%s",
+				(int) first.ia_uid,
+				(int) replies[i].poststat.ia_uid,
+				priv->children[i]->name,
+				uuid_utoa (replies[i].poststat.ia_gfid));
 
-static int
-sh_missing_entries_lookup (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         *xattr_req = NULL;
-	int             ret = -1;
+			*metadata_selfheal = _gf_true;
+		}
 
-	local = frame->local;
-	priv  = this->private;
+		if (!IA_EQUAL (first, replies[i].poststat, gid)) {
+			gf_log (this->name, GF_LOG_DEBUG,
+				"GID mismatch %d vs %d on %s for gfid:%s",
+				(int) first.ia_uid,
+				(int) replies[i].poststat.ia_uid,
+				priv->children[i]->name,
+				uuid_utoa (replies[i].poststat.ia_gfid));
 
-	call_count = afr_up_children_count (priv->child_count,
-                                            local->child_up);
+			*metadata_selfheal = _gf_true;
+		}
 
-	local->call_count = call_count;
+		if (!IA_EQUAL (first, replies[i].poststat, prot)) {
+			gf_log (this->name, GF_LOG_DEBUG,
+				"MODE mismatch %d vs %d on %s for gfid:%s",
+				(int) st_mode_from_ia (first.ia_prot, 0),
+				(int) st_mode_from_ia (replies[i].poststat.ia_prot, 0),
+				priv->children[i]->name,
+				uuid_utoa (replies[i].poststat.ia_gfid));
 
-	xattr_req = dict_new();
+			*metadata_selfheal = _gf_true;
+		}
 
-	if (xattr_req) {
-                for (i = 0; i < priv->child_count; i++) {
-                        ret = dict_set_uint64 (xattr_req,
-                                               priv->pending_key[i],
-                                               3 * sizeof(int32_t));
-                        if (ret < 0)
-                                gf_log (this->name, GF_LOG_WARNING,
-                                        "Unable to set dict value.");
-                }
-        }
+		if (IA_ISREG(first.ia_type) &&
+		    !IA_EQUAL (first, replies[i].poststat, size)) {
+			gf_log (this->name, GF_LOG_DEBUG,
+				"SIZE mismatch %lld vs %lld on %s for gfid:%s",
+				(long long) first.ia_size,
+				(long long) replies[i].poststat.ia_size,
+				priv->children[i]->name,
+				uuid_utoa (replies[i].poststat.ia_gfid));
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"looking up %s on subvolume %s",
-				local->loc.path, priv->children[i]->name);
-
-			STACK_WIND_COOKIE (frame,
-					   sh_missing_entries_lookup_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->lookup,
-					   &local->loc, xattr_req);
-
-			if (!--call_count)
-				break;
+			*data_selfheal = _gf_true;
 		}
 	}
 
-	if (xattr_req)
-		dict_unref (xattr_req);
+	if (valid_cnt > 0)
+		afr_inode_link (inode, &first);
+
+	if (valid_cnt < 2)
+		return -ENOTCONN;
 
 	return 0;
 }
 
 
-
-int
-afr_sh_post_nonblocking_entrylk_cbk (call_frame_t *frame, xlator_t *this)
-{
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-
-        if (int_lock->lock_op_ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "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");
-                sh_missing_entries_lookup (frame, this);
-        }
-
-        return 0;
-}
-
-static int
-afr_sh_entrylk (call_frame_t *frame, xlator_t *this)
+inode_t *
+afr_inode_find (xlator_t *this, uuid_t gfid)
 {
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-        afr_self_heal_t     *sh       = NULL;
-
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-        sh       = &local->self_heal;
+	inode_table_t *table = NULL;
+	inode_t *inode = NULL;
 
-        int_lock->transaction_lk_type = AFR_SELFHEAL_LK;
-        int_lock->selfheal_lk_type    = AFR_ENTRY_SELF_HEAL_LK;
+	table = this->itable;
+	if (!table)
+		return NULL;
 
-        afr_set_lock_number (frame, this);
+	inode = inode_find (table, gfid);
+	if (inode)
+		return inode;
 
-        int_lock->lk_basename = local->loc.name;
-        int_lock->lk_loc      = &sh->parent_loc;
-        int_lock->lock_cbk    = afr_sh_post_nonblocking_entrylk_cbk;
+	inode = inode_new (table);
+	if (!inode)
+		return NULL;
 
-        afr_nonblocking_entrylk (frame, this);
+	uuid_copy (inode->gfid, gfid);
 
-        return 0;
+	return inode;
 }
 
-static int
-afr_self_heal_missing_entries (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;
-	afr_private_t       *priv     = NULL;
-
-	local    = frame->local;
-        int_lock = &local->internal_lock;
-	sh       = &local->self_heal;
-	priv     = this->private;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"attempting to recreate missing entries for path=%s",
-		local->loc.path);
 
-	afr_build_parent_loc (&sh->parent_loc, &local->loc);
-
-        afr_sh_entrylk (frame, this);
-	return 0;
-}
-
-afr_local_t *afr_local_copy (afr_local_t *l, xlator_t *this)
+call_frame_t *
+afr_frame_create (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;
-
-
-        priv = this->private;
-
-        sh = &l->self_heal;
-
-        lc = GF_CALLOC (1, sizeof (afr_local_t),
-                        gf_afr_mt_afr_local_t);
-
-        shc = &lc->self_heal;
-
-        shc->unwind = sh->unwind;
-        shc->need_data_self_heal = sh->need_data_self_heal;
-        shc->need_metadata_self_heal = sh->need_metadata_self_heal;
-        shc->need_entry_self_heal = sh->need_entry_self_heal;
-        shc->forced_merge = sh->forced_merge;
-        shc->healing_fd_opened = sh->healing_fd_opened;
-        shc->data_lock_held = sh->data_lock_held;
-        if (sh->healing_fd && !sh->healing_fd_opened)
-                shc->healing_fd = fd_ref (sh->healing_fd);
-        else
-                shc->healing_fd = sh->healing_fd;
-        shc->background = sh->background;
-        shc->type = sh->type;
-
-        if (l->loc.path)
-                loc_copy (&lc->loc, &l->loc);
-
-        lc->child_up  = memdup (l->child_up, priv->child_count);
-        if (l->xattr_req)
-                lc->xattr_req = dict_ref (l->xattr_req);
-
-        if (l->cont.lookup.inode)
-                lc->cont.lookup.inode = inode_ref (l->cont.lookup.inode);
-        if (l->cont.lookup.xattr)
-                lc->cont.lookup.xattr = dict_ref (l->cont.lookup.xattr);
-        if (l->internal_lock.inode_locked_nodes)
-                lc->internal_lock.inode_locked_nodes =
-                        memdup (l->internal_lock.inode_locked_nodes,
-                                priv->child_count);
-        else
-                lc->internal_lock.inode_locked_nodes =
-                        GF_CALLOC (sizeof (*l->internal_lock.inode_locked_nodes),
-                                   priv->child_count,
-                                   gf_afr_mt_char);
-        if (l->internal_lock.entry_locked_nodes)
-                lc->internal_lock.entry_locked_nodes =
-                        memdup (l->internal_lock.entry_locked_nodes,
-                                priv->child_count);
-        else
-                lc->internal_lock.entry_locked_nodes =
-                        GF_CALLOC (sizeof (*l->internal_lock.entry_locked_nodes),
-                                   priv->child_count,
-                                   gf_afr_mt_char);
-        if (l->internal_lock.locked_nodes)
-                lc->internal_lock.locked_nodes =
-                        memdup (l->internal_lock.locked_nodes,
-                                priv->child_count);
-        else
-                lc->internal_lock.locked_nodes =
-                        GF_CALLOC (sizeof (*l->internal_lock.locked_nodes),
-                                   priv->child_count,
-                                   gf_afr_mt_char);
-
-        lc->internal_lock.inodelk_lock_count =
-                l->internal_lock.inodelk_lock_count;
-        lc->internal_lock.entrylk_lock_count =
-                l->internal_lock.entrylk_lock_count;
-
-        return lc;
-}
+	call_frame_t *frame = NULL;
+	afr_local_t *local = NULL;
+	int op_errno = 0;
+	pid_t pid = -1;
 
+	frame = create_frame (this, this->ctx->pool);
+	if (!frame)
+		return NULL;
 
-int
-afr_self_heal_completion_cbk (call_frame_t *bgsh_frame, xlator_t *this)
-{
-        afr_private_t *   priv  = NULL;
-	afr_local_t *     local = NULL;
-        afr_self_heal_t * sh    = NULL;
-        char              sh_type_str[256] = {0,};
-
-        priv  = this->private;
-	local = bgsh_frame->local;
-        sh    = &local->self_heal;
-
-	if (local->govinda_gOvinda) {
-                afr_set_split_brain (this, local->cont.lookup.inode, _gf_true);
-	} else {
-                afr_set_split_brain (this, local->cont.lookup.inode, _gf_false);
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local) {
+		STACK_DESTROY (frame->root);
+		return NULL;
 	}
 
-        afr_self_heal_type_str_get(sh, sh_type_str,
-                                   sizeof(sh_type_str));
-        gf_log (this->name, GF_LOG_NORMAL,
-                "background %s self-heal completed on %s", sh_type_str,
-                local->loc.path);
-
-        if (!sh->unwound) {
-                sh->unwind (sh->orig_frame, this);
-        }
+	syncopctx_setfspid (&pid);
 
-        if (sh->background) {
-                LOCK (&priv->lock);
-                {
-                        priv->background_self_heals_started--;
-                }
-                UNLOCK (&priv->lock);
-        }
+	frame->root->pid = pid;
 
-        AFR_STACK_DESTROY (bgsh_frame);
+	afr_set_lk_owner (frame, this, frame->root);
 
-	return 0;
+	return frame;
 }
 
-int
-afr_self_heal (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              i = 0;
-
-        call_frame_t *sh_frame = NULL;
-        afr_local_t  *sh_local = NULL;
 
+/*
+ * This is the entry point for healing a given GFID
+ */
 
-	local = frame->local;
-	priv  = this->private;
-
-        afr_set_lk_owner (frame, this);
-
-        if (local->self_heal.background) {
-                LOCK (&priv->lock);
-                {
-                        if (priv->background_self_heals_started
-                            > priv->background_self_heal_count) {
-
-                                local->self_heal.background = _gf_false;
-
-                        } else {
-                                priv->background_self_heals_started++;
-                        }
-                }
-                UNLOCK (&priv->lock);
-        }
-
-        gf_log (this->name, GF_LOG_TRACE,
-                "performing self heal on %s (metadata=%d data=%d entry=%d)",
-                local->loc.path,
-                local->self_heal.need_metadata_self_heal,
-                local->self_heal.need_data_self_heal,
-                local->self_heal.need_entry_self_heal);
-
-        sh_frame        = copy_frame (frame);
-        sh_local        = afr_local_copy (local, this);
-        sh_frame->local = sh_local;
-        sh              = &sh_local->self_heal;
-
-        sh->orig_frame  = frame;
-
-        sh->completion_cbk = afr_self_heal_completion_cbk;
-
-
-	sh->buf = GF_CALLOC (priv->child_count, sizeof (struct iatt),
-                             gf_afr_mt_iatt);
-	sh->child_errno = GF_CALLOC (priv->child_count, sizeof (int),
-                                     gf_afr_mt_int);
-	sh->success = GF_CALLOC (priv->child_count, sizeof (int),
-                                gf_afr_mt_int);
-	sh->xattr = GF_CALLOC (priv->child_count, sizeof (dict_t *),
-                                gf_afr_mt_dict_t);
-	sh->sources = GF_CALLOC (sizeof (*sh->sources), priv->child_count,
-                                gf_afr_mt_int);
-	sh->locked_nodes = GF_CALLOC (sizeof (*sh->locked_nodes), 
-                                      priv->child_count,
-                                      gf_afr_mt_int);
-
-	sh->pending_matrix = GF_CALLOC (sizeof (int32_t *), priv->child_count,
-                                        gf_afr_mt_int32_t);
-
-	for (i = 0; i < priv->child_count; i++) {
-		sh->pending_matrix[i] = GF_CALLOC (sizeof (int32_t),
-					  	   priv->child_count,
-                                                   gf_afr_mt_int32_t);
-	}
+int
+afr_selfheal (xlator_t *this, uuid_t gfid)
+{
+	inode_t *inode = NULL;
+	call_frame_t *frame = NULL;
+	int ret = -1;
+	gf_boolean_t data_selfheal = _gf_false;
+	gf_boolean_t metadata_selfheal = _gf_false;
+	gf_boolean_t entry_selfheal = _gf_false;
 
-	sh->delta_matrix = GF_CALLOC (sizeof (int32_t *), priv->child_count,
-                                      gf_afr_mt_int32_t);
-	for (i = 0; i < priv->child_count; i++) {
-		sh->delta_matrix[i] = GF_CALLOC (sizeof (int32_t),
-					         priv->child_count,
-                                                 gf_afr_mt_int32_t);
-	}
+	inode = afr_inode_find (this, gfid);
+	if (!inode)
+		goto out;
 
-	if (local->success_count && local->enoent_count) {
-		afr_self_heal_missing_entries (sh_frame, this);
-	} else {
-		gf_log (this->name, GF_LOG_TRACE,
-			"proceeding to metadata check on %s",
-			local->loc.path);
+	frame = afr_frame_create (this);
+	if (!frame)
+		goto out;
 
-		afr_sh_missing_entries_done (sh_frame, this);
-	}
+	ret = afr_selfheal_unlocked_inspect (frame, this, inode, gfid,
+					     &data_selfheal,
+					     &metadata_selfheal,
+					     &entry_selfheal);
+	if (ret)
+		goto out;
 
-	return 0;
-}
+	if (data_selfheal)
+		afr_selfheal_data (frame, this, inode);
 
-void
-afr_self_heal_type_str_get (afr_self_heal_t *self_heal_p, char *str,
-                            size_t size)
-{
-        GF_ASSERT (str && (size > 0));
+	if (metadata_selfheal)
+		afr_selfheal_metadata (frame, this, inode);
 
-        if (self_heal_p->need_metadata_self_heal) {
-                snprintf(str, size, " meta-data");
-        }
+	if (entry_selfheal)
+		afr_selfheal_entry (frame, this, inode);
 
-        if (self_heal_p->need_data_self_heal) {
-                snprintf(str + strlen(str), size - strlen(str),
-                         " data");
-        }
+	inode_forget (inode, 1);
+out:
+	if (inode)
+		inode_unref (inode);
+	if (frame)
+		AFR_STACK_DESTROY (frame);
 
-        if (self_heal_p->need_entry_self_heal) {
-                snprintf(str + strlen(str), size - strlen(str),
-                         " entry");
-        }
+	return ret;
 }
diff --git a/xlators/cluster/afr/src/afr-self-heal-common.h b/xlators/cluster/afr/src/afr-self-heal-common.h
deleted file mode 100644
index 6431feaff..000000000
--- a/xlators/cluster/afr/src/afr-self-heal-common.h
+++ /dev/null
@@ -1,73 +0,0 @@
-/*
-  Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#ifndef __AFR_SELF_HEAL_COMMON_H__
-#define __AFR_SELF_HEAL_COMMON_H__
-
-#define FILE_HAS_HOLES(buf) (((buf)->ia_size) > ((buf)->ia_blocks * 512))
-
-typedef enum {
-        AFR_SELF_HEAL_ENTRY,
-        AFR_SELF_HEAL_METADATA,
-        AFR_SELF_HEAL_DATA,
-} afr_self_heal_type;
-
-int
-afr_sh_select_source (int sources[], int child_count);
-
-int
-afr_sh_sink_count (int sources[], int child_count);
-
-int
-afr_sh_source_count (int sources[], int child_count);
-
-int
-afr_sh_supress_errenous_children (int sources[], int child_errno[],
-				  int child_count);
-
-void
-afr_sh_print_pending_matrix (int32_t *pending_matrix[], xlator_t *this);
-
-void
-afr_sh_build_pending_matrix (afr_private_t *priv,
-                             int32_t *pending_matrix[], dict_t *xattr[],
-			     int child_count, afr_transaction_type type);
-
-void
-afr_sh_pending_to_delta (afr_private_t *priv, dict_t **xattr,
-                         int32_t *delta_matrix[], int success[],
-                         int child_count, afr_transaction_type type);
-
-int
-afr_sh_mark_sources (afr_self_heal_t *sh, int child_count,
-                     afr_self_heal_type type);
-
-int
-afr_sh_delta_to_xattr (afr_private_t *priv,
-                       int32_t *delta_matrix[], dict_t *xattr[],
-		       int child_count, afr_transaction_type type);
-
-int
-afr_sh_is_matrix_zero (int32_t *pending_matrix[], int child_count);
-
-void
-afr_self_heal_type_str_get (afr_self_heal_t *self_heal_p, char *str,
-                            size_t size);
-
-#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 0fd8dae69..c0548d995 100644
--- a/xlators/cluster/afr/src/afr-self-heal-data.c
+++ b/xlators/cluster/afr/src/afr-self-heal-data.c
@@ -1,1113 +1,635 @@
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 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 <libgen.h>
-#include <unistd.h>
-#include <fnmatch.h>
-#include <sys/time.h>
-#include <stdlib.h>
-#include <signal.h>
 
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
 #endif
 
-#include "glusterfs.h"
 #include "afr.h"
-#include "dict.h"
-#include "xlator.h"
-#include "hashfn.h"
-#include "logging.h"
-#include "stack.h"
-#include "list.h"
-#include "call-stub.h"
-#include "defaults.h"
-#include "common-utils.h"
-#include "compat-errno.h"
-#include "compat.h"
+#include "afr-self-heal.h"
 #include "byte-order.h"
 
-#include "afr-transaction.h"
-#include "afr-self-heal.h"
-#include "afr-self-heal-common.h"
-#include "afr-self-heal-algorithm.h"
+enum {
+	AFR_SELFHEAL_DATA_FULL = 0,
+	AFR_SELFHEAL_DATA_DIFF,
+};
 
 
-int
-afr_sh_data_done (call_frame_t *frame, xlator_t *this)
+#define HAS_HOLES(i) ((i->ia_blocks * 512) < (i->ia_size))
+static int
+__checksum_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		int op_ret, int op_errno, uint32_t weak, uint8_t *strong,
+		dict_t *xdata)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
+	afr_local_t *local = NULL;
+	int i = (long) cookie;
 
 	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	/* 
-	   TODO: cleanup sh->* 
-	 */
-
-        if (sh->healing_fd && !sh->healing_fd_opened) {
-                /* unref only if we created the fd ourselves */
-
-		fd_unref (sh->healing_fd);
-		sh->healing_fd = NULL;
-        }
-
-/*         for (i = 0; i < priv->child_count; i++) */
-/*                 sh->locked_nodes[i] = 0; */
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"self heal of %s completed",
-		local->loc.path);
 
-	sh->completion_cbk (frame, this);
+	local->replies[i].valid = 1;
+	local->replies[i].op_ret = op_ret;
+	local->replies[i].op_errno = op_errno;
+	if (strong)
+		memcpy (local->replies[i].checksum, strong, MD5_DIGEST_LENGTH);
 
+	syncbarrier_wake (&local->barrier);
 	return 0;
 }
 
 
-int
-afr_sh_data_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int32_t op_ret, int32_t op_errno)
+static int
+attr_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;
-	int              call_count = 0;
-
-        int child_index = (long) cookie;
+	int i = (long) cookie;
+	afr_local_t *local = NULL;
 
 	local = frame->local;
-	priv = this->private;
 
-	LOCK (&frame->lock);
-	{
-                if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "flush or setattr failed on %s on subvolume %s: %s",
-                                local->loc.path, priv->children[child_index]->name,
-                                strerror (op_errno));
-                }
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
+	local->replies[i].valid = 1;
+	local->replies[i].op_ret = op_ret;
+	local->replies[i].op_errno = op_errno;
+	if (pre)
+		local->replies[i].prestat = *pre;
+	if (post)
+		local->replies[i].poststat = *post;
+	if (xdata)
+		local->replies[i].xdata = dict_ref (xdata);
 
-	if (call_count == 0) {
-		afr_sh_data_done (frame, this);
-	}
+	syncbarrier_wake (&local->barrier);
 
 	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)
+static gf_boolean_t
+__afr_selfheal_data_checksums_match (call_frame_t *frame, xlator_t *this,
+				     fd_t *fd, int source,
+				     unsigned char *healed_sinks,
+				     off_t offset, size_t size)
 {
-        afr_sh_data_flush_cbk (frame, cookie, this, op_ret, op_errno);
-
-        return 0;
-}
-
-
-int
-afr_sh_data_close (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_private_t   *priv  = NULL;
-	afr_self_heal_t *sh    = NULL;
-
-	int  i            = 0;
-	int  call_count   = 0;
-        int  source       = 0;
-        int32_t valid     = 0;
-
-        struct iatt stbuf = {0,};
-
-        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;
-
-        if (sh->healing_fd_opened) {
-                /* not our job to close the fd */
+	afr_private_t *priv = NULL;
+	afr_local_t *local = NULL;
+	unsigned char *wind_subvols = NULL;
+	int i = 0;
 
-                afr_sh_data_done (frame, this);
-                return 0;
-        }
+	priv = this->private;
+	local = frame->local;
 
-	if (!sh->healing_fd) {
-		afr_sh_data_done (frame, this);
-		return 0;
+	wind_subvols = alloca0 (priv->child_count);
+	for (i = 0; i < priv->child_count; i++) {
+		if (i == source || healed_sinks[i])
+			wind_subvols[i] = 1;
 	}
 
-	call_count        = (sh->active_sinks + 1) * 2;
-	local->call_count = call_count;
+	AFR_ONLIST (wind_subvols, frame, __checksum_cbk, rchecksum, fd,
+		    offset, size, NULL);
 
-	/* closed source */
-	gf_log (this->name, GF_LOG_TRACE,
-		"closing fd of %s on %s",
-		local->loc.path, priv->children[sh->source]->name);
-
-	STACK_WIND_COOKIE (frame, afr_sh_data_flush_cbk,
-			   (void *) (long) sh->source,
-			   priv->children[sh->source],
-			   priv->children[sh->source]->fops->flush,
-			   sh->healing_fd);
-	call_count--;
-
-        STACK_WIND_COOKIE (frame, afr_sh_data_setattr_cbk,
-                           (void *) (long) sh->source,
-                           priv->children[sh->source],
-                           priv->children[sh->source]->fops->setattr,
-                           &local->loc, &stbuf, valid);
-
-        call_count--;
-
-        if (call_count == 0)
-                return 0;
+	if (!local->replies[source].valid || local->replies[source].op_ret != 0)
+		return _gf_false;
 
 	for (i = 0; i < priv->child_count; i++) {
-		if (sh->sources[i] || !local->child_up[i])
+		if (i == source)
 			continue;
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"closing fd of %s on %s",
-			local->loc.path, priv->children[i]->name);
-
-		STACK_WIND_COOKIE (frame, afr_sh_data_flush_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->flush,
-				   sh->healing_fd);
-
-                call_count--;
-
-                STACK_WIND_COOKIE (frame, afr_sh_data_setattr_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->setattr,
-				   &local->loc, &stbuf, valid);
-
-		if (!--call_count)
-			break;
-	}
-
-	return 0;
-}
-
-
-int
-afr_sh_data_unlck_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int32_t op_ret, int32_t op_errno)
-{
-	afr_local_t * local = NULL;
-	int           call_count = 0;
-	int           child_index = (long) cookie;
-
-	
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"locking inode of %s on child %d failed: %s",
-				local->loc.path, child_index,
-				strerror (op_errno));
-		} else {
-			gf_log (this->name, GF_LOG_TRACE,
-				"inode of %s on child %d locked",
-				local->loc.path, child_index);
-		}
+		if (memcmp (local->replies[source].checksum,
+			    local->replies[i].checksum,
+			    MD5_DIGEST_LENGTH))
+			return _gf_false;
 	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-		afr_sh_data_close (frame, this);
-	}
-
-	return 0;
-}
-
-
-int
-afr_sh_data_unlock (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t         *local    = NULL;
-        afr_internal_lock_t *int_lock = NULL;
-	afr_self_heal_t     *sh       = NULL;
-
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-	sh       = &local->self_heal;
-
-        GF_ASSERT (!sh->data_lock_held);
-
-        int_lock->lock_cbk = afr_sh_data_close;
-        afr_unlock (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;
-
-	local = frame->local;
-	sh = &local->self_heal;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"finishing data selfheal of %s", local->loc.path);
-
-        if (!sh->data_lock_held)
-                afr_sh_data_unlock (frame, this);
-        else
-                afr_sh_data_close (frame, this);
-
-	return 0;
+	return _gf_true;
 }
 
 
-int
-afr_sh_data_erase_pending_cbk (call_frame_t *frame, void *cookie,
-			       xlator_t *this, int32_t op_ret,
-			       int32_t op_errno, dict_t *xattr)
-{
-	int             call_count = 0;
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		afr_sh_data_finish (frame, this);
-
-	return 0;
-}
-
-
-int
-afr_sh_data_erase_pending (call_frame_t *frame, xlator_t *this)
+static int
+__afr_selfheal_data_read_write (call_frame_t *frame, xlator_t *this, fd_t *fd,
+				int source, unsigned char *healed_sinks,
+				off_t offset, size_t size,
+				struct afr_reply *replies)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              call_count = 0;
-	int              i = 0;
-	dict_t          **erase_xattr = NULL;
-
+	struct iovec *iovec = NULL;
+	int count = 0;
+	struct iobref *iobref = NULL;
+	int ret = 0;
+	int i = 0;
+	afr_private_t *priv = NULL;
 
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-	afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix, sh->success,
-                                 priv->child_count, AFR_DATA_TRANSACTION);
-
-	erase_xattr = GF_CALLOC (sizeof (*erase_xattr), priv->child_count,
-                              gf_afr_mt_dict_t);
+	ret = syncop_readv (priv->children[source], fd, size, offset, 0,
+			    &iovec, &count, &iobref);
+	if (ret <= 0)
+		return ret;
 
 	for (i = 0; i < priv->child_count; i++) {
-		if (sh->xattr[i]) {
-			call_count++;
-
-			erase_xattr[i] = get_new_dict();
-			dict_ref (erase_xattr[i]);
-		}
-	}
-
-	afr_sh_delta_to_xattr (priv, sh->delta_matrix, erase_xattr,
-			       priv->child_count, AFR_DATA_TRANSACTION);
-
-	local->call_count = call_count;
-	for (i = 0; i < priv->child_count; i++) {
-		if (!erase_xattr[i])
+		if (!healed_sinks[i])
 			continue;
 
-		gf_log (this->name, GF_LOG_TRACE,
-			"erasing pending flags from %s on %s",
-			local->loc.path, priv->children[i]->name);
-
-		STACK_WIND_COOKIE (frame, afr_sh_data_erase_pending_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->fxattrop,
-				   sh->healing_fd,
-				   GF_XATTROP_ADD_ARRAY, erase_xattr[i]);
-		if (!--call_count)
-			break;
-	}
+		/*
+		 * TODO: Use fiemap() and discard() to heal holes
+		 * in the future.
+		 *
+		 * For now,
+		 *
+		 * - if the source had any holes at all,
+		 * AND
+		 * - if we are writing past the original file size
+		 *   of the sink
+		 * AND
+		 * - is NOT the last block of the source file. if
+		 *   the block contains EOF, it has to be written
+		 *   in order to set the file size even if the
+		 *   last block is 0-filled.
+		 * AND
+		 * - if the read buffer is filled with only 0's
+		 *
+		 * then, skip writing to this source. We don't depend
+		 * on the write to happen to update the size as we
+		 * have performed an ftruncate() upfront anyways.
+		 */
+#define is_last_block(o,b,s) ((s >= o) && (s <= (o + b)))
+		if (HAS_HOLES ((&replies[source].poststat)) &&
+		    offset >= replies[i].poststat.ia_size &&
+		    !is_last_block (offset, size,
+				    replies[source].poststat.ia_size) &&
+		    (iov_0filled (iovec, count) == 0))
+			continue;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (erase_xattr[i]) {
-			dict_unref (erase_xattr[i]);
+		ret = syncop_writev (priv->children[i], fd, iovec, count,
+				     offset, iobref, 0);
+		if (ret != iov_length (iovec, count)) {
+			/* write() failed on this sink. unset the corresponding
+			   member in sinks[] (which is healed_sinks[] in the
+			   caller) so that this server does NOT get considered
+			   as successfully healed.
+			*/
+			healed_sinks[i] = 0;
 		}
 	}
-	GF_FREE (erase_xattr);
+	if (iobref)
+		iobref_unref (iobref);
 
-	return 0;
+	return ret;
 }
 
 
-int
-afr_sh_data_trim_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		      int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                      struct iatt *postbuf)
+static int
+afr_selfheal_data_block (call_frame_t *frame, xlator_t *this, fd_t *fd,
+			 int source, unsigned char *healed_sinks, off_t offset,
+			 size_t size, int type, struct afr_reply *replies)
 {
-	afr_private_t * priv = NULL;
-	afr_local_t * local  = NULL;
-	int              call_count = 0;
-	int              child_index = 0;
+	int ret = -1;
+	int sink_count = 0;
+	afr_private_t *priv = NULL;
+	unsigned char *data_lock = NULL;
 
 	priv = this->private;
-	local = frame->local;
-
-	child_index = (long) cookie;
+	sink_count = AFR_COUNT (healed_sinks, priv->child_count);
+	data_lock = alloca0 (priv->child_count);
 
-	LOCK (&frame->lock);
+	ret = afr_selfheal_inodelk (frame, this, fd->inode, this->name,
+				    offset, size, data_lock);
 	{
-		if (op_ret == -1)
-			gf_log (this->name, GF_LOG_DEBUG,
-				"ftruncate of %s on subvolume %s failed (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-		else
-			gf_log (this->name, GF_LOG_TRACE,
-				"ftruncate of %s on subvolume %s completed",
-				local->loc.path,
-				priv->children[child_index]->name);
-	}
-	UNLOCK (&frame->lock);
+		if (ret < sink_count) {
+			ret = -ENOTCONN;
+			goto unlock;
+		}
 
-	call_count = afr_frame_return (frame);
+		if (type == AFR_SELFHEAL_DATA_DIFF &&
+		    __afr_selfheal_data_checksums_match (frame, this, fd, source,
+							 healed_sinks, offset, size)) {
+			ret = 0;
+			goto unlock;
+		}
 
-	if (call_count == 0) {
-		afr_sh_data_erase_pending (frame, this);
+		ret = __afr_selfheal_data_read_write (frame, this, fd, source,
+						      healed_sinks, offset, size,
+						      replies);
 	}
-
-	return 0;
+unlock:
+	afr_selfheal_uninodelk (frame, this, fd->inode, this->name,
+				offset, size, data_lock);
+	return ret;
 }
 
 
-int
-afr_sh_data_trim_sinks (call_frame_t *frame, xlator_t *this)
-{
-	afr_private_t * priv = NULL;
-	afr_local_t * local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int             *sources = NULL;
-	int              call_count = 0;
-	int              i = 0;
 
+static int
+afr_selfheal_data_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
+			 unsigned char *healed_sinks)
+{
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	int i = 0;
 
-	priv = this->private;
 	local = frame->local;
-	sh = &local->self_heal;
-
-	sources = sh->sources;
-	call_count = sh->active_sinks;
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sources[i] || !local->child_up[i])
-			continue;
-
-		STACK_WIND_COOKIE (frame, afr_sh_data_trim_cbk,
-				   (void *) (long) i,
-				   priv->children[i], 
-				   priv->children[i]->fops->ftruncate,
-				   sh->healing_fd, sh->file_size); 
+	priv = this->private;
 
-		if (!--call_count)
-			break;
-	}
+	AFR_ONLIST (healed_sinks, frame, attr_cbk, fsync, fd, 0, NULL);
 
+	for (i = 0; i < priv->child_count; i++)
+		if (healed_sinks[i] && local->replies[i].op_ret != 0)
+			/* fsync() failed. Do NOT consider this server
+			   as successfully healed. Mark it so.
+			*/
+			healed_sinks[i] = 0;
 	return 0;
 }
 
 
-static struct afr_sh_algorithm *
-sh_algo_from_name (xlator_t *this, char *name)
+static int
+afr_selfheal_data_restore_time (call_frame_t *frame, xlator_t *this,
+				inode_t *inode, int source,
+				unsigned char *healed_sinks,
+				struct afr_reply *replies)
 {
-        int i = 0;
+	loc_t loc = {0, };
 
-        while (afr_self_heal_algorithms[i].name) {
-                if (!strcmp (name, afr_self_heal_algorithms[i].name)) {
-                        return &afr_self_heal_algorithms[i];
-                }
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-                i++;
-        }
+	AFR_ONLIST (healed_sinks, frame, attr_cbk, setattr, &loc,
+		    &replies[source].poststat,
+		    (GF_SET_ATTR_ATIME|GF_SET_ATTR_MTIME), NULL);
 
-        return NULL;
-}
+	loc_wipe (&loc);
 
+	return 0;
+}
 
 static int
-sh_zero_byte_files_exist (afr_self_heal_t *sh, int child_count)
+afr_data_self_heal_type_get (afr_private_t *priv, unsigned char *healed_sinks,
+                             int source, struct afr_reply *replies)
 {
-        int i;
-        int ret = 0;
-
-        for (i = 0; i < child_count; i++) {
-                if (sh->buf[i].ia_size == 0) {
-                        ret = 1;
-                        break;
-                }
-        }
-
-        return ret;
-}
-
+        int type = AFR_SELFHEAL_DATA_FULL;
+        int i = 0;
 
-struct afr_sh_algorithm *
-afr_sh_data_pick_algo (call_frame_t *frame, xlator_t *this)
-{
-        afr_private_t *           priv  = NULL;
-        struct afr_sh_algorithm * algo  = NULL;
-        afr_local_t *             local = NULL;
-        afr_self_heal_t *         sh    = NULL;
-
-        priv  = this->private;
-        local = frame->local;
-        sh    = &local->self_heal;
-        algo  = sh_algo_from_name (this, priv->data_self_heal_algorithm);
-
-        if (algo == NULL) {
-                /* option not set, so fall back on heuristics */
-
-                if ((local->enoent_count != 0)
-                    || sh_zero_byte_files_exist (sh, priv->child_count)
-                    || (sh->file_size <= (priv->data_self_heal_window_size * this->ctx->page_size))) {
-
-                        /*
-                         * If the file does not exist on one of the subvolumes,
-                         * or a zero-byte file exists (created by entry self-heal)
-                         * the entire content has to be copied anyway, so there
-                         * is no benefit from using the "diff" algorithm.
-                         *
-                         * If the file size is about the same as page size,
-                         * the entire file can be read and written with a few
-                         * (pipelined) STACK_WINDs, which will be faster
-                         * than "diff" which has to read checksums and then
-                         * read and write.
-                         */
-
-                        algo = sh_algo_from_name (this, "full");
-
-                } else {
-                        algo = sh_algo_from_name (this, "diff");
+        if (priv->data_self_heal_algorithm == NULL) {
+                type = AFR_SELFHEAL_DATA_FULL;
+                for (i = 0; i < priv->child_count; i++) {
+                        if (!healed_sinks[i] && i != source)
+                                continue;
+                        if (replies[i].poststat.ia_size) {
+                                type = AFR_SELFHEAL_DATA_DIFF;
+                                break;
+                        }
                 }
+        } else if (strcmp (priv->data_self_heal_algorithm, "full") == 0) {
+                type = AFR_SELFHEAL_DATA_FULL;
+        } else if (strcmp (priv->data_self_heal_algorithm, "diff") == 0) {
+                type = AFR_SELFHEAL_DATA_DIFF;
         }
-
-        return algo;
-}
-
-
-int
-afr_sh_data_sync_prepare (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              active_sinks = 0;
-	int              source = 0;
-	int              i = 0;
-
-        struct afr_sh_algorithm *sh_algo = NULL;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	source = sh->source;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->sources[i] == 0 && local->child_up[i] == 1) {
-			active_sinks++;
-			sh->success[i] = 1;
-		}
-	}
-	sh->success[source] = 1;
-
-	if (active_sinks == 0) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"no active sinks for performing self-heal on file %s",
-			local->loc.path);
-		afr_sh_data_finish (frame, this);
-		return 0;
-	}
-	sh->active_sinks = active_sinks;
-
-	gf_log (this->name, GF_LOG_DEBUG,
-		"self-healing file %s from subvolume %s to %d other",
-		local->loc.path, priv->children[source]->name, active_sinks);
-
-        sh->algo_completion_cbk = afr_sh_data_trim_sinks;
-        sh->algo_abort_cbk      = afr_sh_data_finish;
-
-        sh_algo = afr_sh_data_pick_algo (frame, this);
-
-        sh_algo->fn (frame, this);
-
-	return 0;
+        return type;
 }
 
-
-int
-afr_sh_data_fix (call_frame_t *frame, xlator_t *this)
+static int
+afr_selfheal_data_do (call_frame_t *frame, xlator_t *this, fd_t *fd,
+		      int source, unsigned char *healed_sinks,
+		      struct afr_reply *replies)
 {
-	afr_local_t     *local      = NULL;
-        afr_local_t *    orig_local = NULL;
-
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              nsources = 0;
-	int              source = 0;
-	int              i = 0;
+	afr_private_t *priv = NULL;
+	int i = 0;
+	off_t off = 0;
+	size_t block = 128 * 1024;
+	int type = AFR_SELFHEAL_DATA_FULL;
+	int ret = -1;
+	call_frame_t *iter_frame = NULL;
+	char *sinks_str = NULL;
+	char *p = NULL;
 
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-	afr_sh_build_pending_matrix (priv, sh->pending_matrix, sh->xattr, 
-				     priv->child_count, AFR_DATA_TRANSACTION);
-
-	afr_sh_print_pending_matrix (sh->pending_matrix, this);
-
-	nsources = afr_sh_mark_sources (sh, priv->child_count,
-                                        AFR_SELF_HEAL_DATA);
-
-	afr_sh_supress_errenous_children (sh->sources, sh->child_errno,
-					  priv->child_count);
-
-        if (nsources == 0) {
-                gf_log (this->name, GF_LOG_TRACE,
-                        "No self-heal needed for %s",
-                        local->loc.path);
-
-                afr_sh_data_finish (frame, this);
-                return 0;
-        }
-
-	if ((nsources == -1)
-	    && (priv->favorite_child != -1)
-	    && (sh->child_errno[priv->favorite_child] == 0)) {
-
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Picking favorite child %s as authentic source to resolve conflicting data of %s",
-			priv->children[priv->favorite_child]->name,
-			local->loc.path);
-
-		sh->sources[priv->favorite_child] = 1;
-
-		nsources = afr_sh_source_count (sh->sources,
-						priv->child_count);
-	}
-
-	if (nsources == -1) {
-		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);
-
-		local->govinda_gOvinda = 1;
-
-		afr_sh_data_finish (frame, this);
-		return 0;
-	}
-
-	source = afr_sh_select_source (sh->sources, priv->child_count);
-        
-        if (source == -1) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "No active sources found.");
-
-                afr_sh_data_finish (frame, this);
-                return 0;
-        }
-
-	sh->source     = source;
-	sh->block_size = 65536;
-	sh->file_size  = sh->buf[source].ia_size;
-
-	if (FILE_HAS_HOLES (&sh->buf[source]))
-		sh->file_has_holes = 1;
-
-        orig_local = sh->orig_frame->local;
-        orig_local->cont.lookup.buf.ia_size = sh->buf[source].ia_size;
-
-	/* detect changes not visible through pending flags -- JIC */
+	sinks_str = alloca0 (priv->child_count * 8);
+	p = sinks_str;
 	for (i = 0; i < priv->child_count; i++) {
-		if (i == source || sh->child_errno[i])
+		if (!healed_sinks[i])
 			continue;
-
-		if (SIZE_DIFFERS (&sh->buf[i], &sh->buf[source]))
-			sh->sources[i] = 0;
+		p += sprintf (p, "%d ", i);
 	}
 
-        afr_set_read_child (this, local->loc.inode, sh->source);
-
-	/*
-	   quick-read might have read the file, so send xattr from
-	   the source subvolume (http://bugs.gluster.com/cgi-bin/bugzilla3/show_bug.cgi?id=815)
-	*/
-
-	dict_unref (orig_local->cont.lookup.xattr);
-        if (orig_local->cont.lookup.xattrs)
-                orig_local->cont.lookup.xattr = dict_ref (orig_local->cont.lookup.xattrs[sh->source]);
-
-        if (sh->background) {
-                sh->unwind (sh->orig_frame, this);
-                sh->unwound = _gf_true;
-        }
-
-	afr_sh_data_sync_prepare (frame, this);
-
-	return 0;
-}
-
+	gf_log (this->name, GF_LOG_INFO, "performing data selfheal on %s. "
+		"source=%d sinks=%s",
+		uuid_utoa (fd->inode->gfid), source, sinks_str);
 
-int
-afr_self_heal_get_source (xlator_t *this, afr_local_t *local, dict_t **xattr)
-{
-	afr_self_heal_t *sh   = NULL;
-	afr_private_t   *priv = NULL;
+        type = afr_data_self_heal_type_get (priv, healed_sinks, source,
+                                            replies);
 
-	int              source = 0;
-	int              i = 0;
+	iter_frame = afr_copy_frame (frame);
+	if (!iter_frame)
+		return -ENOMEM;
 
-	sh   = &local->self_heal;
-	priv = this->private;
+	for (off = 0; off < replies[source].poststat.ia_size; off += block) {
+		ret = afr_selfheal_data_block (iter_frame, this, fd, source,
+					       healed_sinks, off, block, type,
+					       replies);
+		if (ret < 0)
+			goto out;
 
-	sh->pending_matrix = GF_CALLOC (sizeof (int32_t *), priv->child_count,
-                                        gf_afr_mt_int32_t);
-	for (i = 0; i < priv->child_count; i++) {
-		sh->pending_matrix[i] = GF_CALLOC (sizeof (int32_t),
-						priv->child_count,
-                                                gf_afr_mt_int32_t);
+		AFR_STACK_RESET (iter_frame);
 	}
 
-	sh->sources = GF_CALLOC (priv->child_count, sizeof (*sh->sources),
-                                 gf_afr_mt_int32_t);
-
-	afr_sh_build_pending_matrix (priv, sh->pending_matrix, xattr,
-				     priv->child_count, AFR_DATA_TRANSACTION);
+	afr_selfheal_data_restore_time (frame, this, fd->inode, source,
+					healed_sinks, replies);
 
-	(void)afr_sh_mark_sources (sh, priv->child_count, AFR_SELF_HEAL_DATA);
+	ret = afr_selfheal_data_fsync (frame, this, fd, healed_sinks);
 
-	source = afr_sh_select_source (sh->sources, priv->child_count);
-
-	return source;
+out:
+	if (iter_frame)
+		AFR_STACK_DESTROY (iter_frame);
+	return ret;
 }
 
 
-int
-afr_sh_data_fstat_cbk (call_frame_t *frame, void *cookie,
-                       xlator_t *this, int32_t op_ret, int32_t op_errno,
-                       struct iatt *buf)
+static int
+__afr_selfheal_truncate_sinks (call_frame_t *frame, xlator_t *this,
+			       fd_t *fd, unsigned char *healed_sinks,
+			       struct afr_reply *replies, uint64_t size)
 {
-	afr_private_t   *priv  = NULL;
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-
-	int call_count  = -1;
-	int child_index = (long) cookie;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	unsigned char *larger_sinks = 0;
+	int i = 0;
 
 	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret != -1) {
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "fstat of %s on %s succeeded",
-                                local->loc.path,
-                                priv->children[child_index]->name);
-
-			sh->buf[child_index] = *buf;
-		}
+	larger_sinks = alloca0 (priv->child_count);
+	for (i = 0; i < priv->child_count; i++) {
+		if (healed_sinks[i] && replies[i].poststat.ia_size > size)
+			larger_sinks[i] = 1;
 	}
-	UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-		afr_sh_data_fix (frame, this);
-	}
+	AFR_ONLIST (larger_sinks, frame, attr_cbk, ftruncate, fd, size, NULL);
 
+	for (i = 0; i < priv->child_count; i++)
+		if (healed_sinks[i] && local->replies[i].op_ret == -1)
+			/* truncate() failed. Do NOT consider this server
+			   as successfully healed. Mark it so.
+			*/
+			healed_sinks[i] = 0;
 	return 0;
 }
 
-
-int
-afr_sh_data_fstat (call_frame_t *frame, xlator_t *this)
+/*
+ * If by chance there are multiple sources with differing sizes, select
+ * the largest file as the source.
+ *
+ * This can only happen if data was directly modified in the backend.
+ */
+static int
+__afr_selfheal_data_finalize_source (xlator_t *this, unsigned char *sources,
+				     unsigned char *sinks,
+				     unsigned char *healed_sinks,
+				     unsigned char *locked_on,
+				     struct afr_reply *replies)
 {
-	afr_self_heal_t *sh    = NULL; 
-	afr_local_t     *local = NULL;
-	afr_private_t   *priv  = NULL;
-
-	int call_count = 0;
 	int i = 0;
+	afr_private_t *priv = NULL;
+	uint64_t size = 0;
+	int source = -1;
+	int locked_count = 0;
+	int sources_count = 0;
+	int healed_sinks_count = 0;
 
-	priv  = this->private;
-	local = frame->local;
-	sh    = &local->self_heal;
-
-	call_count = afr_up_children_count (priv->child_count,
-                                            local->child_up);
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_sh_data_fstat_cbk,
-					   (void *) (long) i,
-					   priv->children[i], 
-					   priv->children[i]->fops->fstat,
-					   sh->healing_fd);
-
-			if (!--call_count)
-				break;
-		}
-	}
-
-	return 0;
-}
-
-
-int
-afr_sh_data_fxattrop_cbk (call_frame_t *frame, void *cookie,
-                          xlator_t *this, int32_t op_ret, int32_t op_errno,
-                          dict_t *xattr)
-{
-	afr_private_t   *priv  = NULL;
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-
-	int call_count  = -1;
-	int child_index = (long) cookie;
-
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret != -1) {
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "fxattrop of %s on %s succeeded",
-                                local->loc.path,
-                                priv->children[child_index]->name);
+	locked_count = AFR_COUNT (locked_on, priv->child_count);
+	sources_count = AFR_COUNT (sources, priv->child_count);
+	healed_sinks_count = AFR_COUNT (healed_sinks, priv->child_count);
 
-			sh->xattr[child_index] = dict_ref (xattr);
-		}
+	if (locked_count == healed_sinks_count || !sources_count) {
+		/* split brain */
+		return -EIO;
 	}
-	UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-		afr_sh_data_fstat (frame, this);
+	for (i = 0; i < priv->child_count; i++) {
+		if (!sources[i])
+			continue;
+		if (size <= replies[i].poststat.ia_size) {
+			size = replies[i].poststat.ia_size;
+			source = i;
+		}
 	}
 
-	return 0;
-}
-
-
-int
-afr_sh_data_fxattrop (call_frame_t *frame, xlator_t *this)
-{
-	afr_self_heal_t *sh    = NULL;
-	afr_local_t     *local = NULL;
-	afr_private_t   *priv  = NULL;
-	dict_t          *xattr_req = NULL;
-
-        int32_t zero_pending[3] = {0, 0, 0};
-
-	int call_count = 0;
-	int i = 0;
-	int ret = 0;
-
-	priv  = this->private;
-	local = frame->local;
-	sh    = &local->self_heal;
-
-	call_count = afr_up_children_count (priv->child_count,
-                                            local->child_up);
-
-	local->call_count = call_count;
-
-	xattr_req = dict_new();
-	if (xattr_req) {
-                for (i = 0; i < priv->child_count; i++) {
-                        ret = dict_set_static_bin (xattr_req, priv->pending_key[i],
-                                                   zero_pending, 3 * sizeof(int32_t));
-                        if (ret < 0)
-				gf_log (this->name, GF_LOG_WARNING,
-					"Unable to set dict value");
-                }
-        }
-
 	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_sh_data_fxattrop_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->fxattrop,
-					   sh->healing_fd, GF_XATTROP_ADD_ARRAY,
-                                           xattr_req);
-
-			if (!--call_count)
-				break;
+		if (!sources[i])
+			continue;
+		if (replies[i].poststat.ia_size < size) {
+			sources[i] = 0;
+			sinks[i] = 1;
 		}
 	}
 
-	if (xattr_req)
-		dict_unref (xattr_req);
-
-	return 0;
-}
-
-
-int
-afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this);
-
-int
-afr_sh_data_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
-{
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-
-        if (int_lock->lock_op_ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Non Blocking inodelks failed.");
-                afr_sh_data_done (frame, this);
-        } else {
-
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Non Blocking inodelks done. Proceeding to FOP");
-                afr_sh_data_fxattrop (frame, this);
-        }
-
-        return 0;
-}
-
-int
-afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this)
-{
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-        afr_self_heal_t     *sh       = NULL;
-
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-        sh       = &local->self_heal;
-
-        int_lock->transaction_lk_type = AFR_SELFHEAL_LK;
-        int_lock->selfheal_lk_type    = AFR_DATA_SELF_HEAL_LK;
-
-        afr_set_lock_number (frame, this);
-
-        int_lock->lk_flock.l_start = 0;
-        int_lock->lk_flock.l_len   = 0;
-        int_lock->lk_flock.l_type  = F_WRLCK;
-        int_lock->lock_cbk         = afr_sh_data_post_nonblocking_inodelk_cbk;
-
-        afr_nonblocking_inodelk (frame, this);
-
-
-	return 0;
+	return source;
 }
 
-
-int
-afr_sh_data_lock (call_frame_t *frame, xlator_t *this)
+/*
+ * __afr_selfheal_data_prepare:
+ *
+ * This function inspects the on-disk xattrs and determines which subvols
+ * are sources and sinks.
+ *
+ * The return value is the index of the subvolume to be used as the source
+ * for self-healing, or -1 if no healing is necessary/split brain.
+ */
+static int
+__afr_selfheal_data_prepare (call_frame_t *frame, xlator_t *this, fd_t *fd,
+			     unsigned char *locked_on, unsigned char *sources,
+			     unsigned char *sinks, unsigned char *healed_sinks,
+			     struct afr_reply *replies)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	afr_self_heal_t * sh  = NULL;
-
-
-	local = frame->local;
-	sh    = &local->self_heal;
-	priv  = this->private;
+	int ret = -1;
+	int source = -1;
+	afr_private_t *priv = NULL;
+	int i = 0;
 
-        if (sh->data_lock_held) {
-                /* caller has held the lock already,
-                   so skip locking */
+	priv = this->private;
 
-                afr_sh_data_fxattrop (frame, this);
-                return 0;
-        }
+	ret = afr_selfheal_unlocked_discover (frame, fd->inode, fd->inode->gfid,
+					      replies);
+	if (ret)
+		return ret;
+
+	ret = afr_selfheal_find_direction (frame, this, replies,
+					   AFR_DATA_TRANSACTION,
+					   locked_on, sources, sinks);
+	if (ret)
+		return ret;
+
+	source = __afr_selfheal_data_finalize_source (this, sources, sinks,
+						      healed_sinks, locked_on,
+						      replies);
+	if (source < 0)
+		return -EIO;
+
+	for (i = 0; i < priv->child_count; i++)
+		/* Initialize the healed_sinks[] array optimistically to
+		   the intersection of to-be-healed (i.e sinks[]) and
+		   the list of servers which are up (i.e locked_on[]).
+
+		   As we encounter failures in the healing process, we
+		   will unmark the respective servers in the healed_sinks[]
+		   array.
+		*/
+		healed_sinks[i] = sinks[i] && locked_on[i];
 
-        return afr_sh_data_lock_rec (frame, this);
+	return source;
 }
 
 
-int
-afr_sh_data_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		      int32_t op_ret, int32_t op_errno, fd_t *fd)
+static int
+__afr_selfheal_data (call_frame_t *frame, xlator_t *this, fd_t *fd,
+		     unsigned char *locked_on)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              call_count = 0;
-	int              child_index = 0;
+	afr_private_t *priv = NULL;
+	int ret = -1;
+	unsigned char *sources = NULL;
+	unsigned char *sinks = NULL;
+	unsigned char *data_lock = NULL;
+	unsigned char *healed_sinks = NULL;
+	struct afr_reply *locked_replies = NULL;
+	int source = -1;
+	gf_boolean_t compat = _gf_false;
+	unsigned char *compat_lock = NULL;
 
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-	child_index = (long) cookie;
+	sources = alloca0 (priv->child_count);
+	sinks = alloca0 (priv->child_count);
+	healed_sinks = alloca0 (priv->child_count);
+	data_lock = alloca0 (priv->child_count);
+	compat_lock = alloca0 (priv->child_count);
 
-	/* TODO: some of the open's might fail.
-	   In that case, modify cleanup fn to send flush on those 
-	   fd's which are already open */
+	locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
 
-	LOCK (&frame->lock);
+	ret = afr_selfheal_inodelk (frame, this, fd->inode, this->name, 0, 0,
+				    data_lock);
 	{
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"open of %s failed on child %s (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-			sh->op_failed = 1;
+		if (ret < 2) {
+			ret = -ENOTCONN;
+			goto unlock;
 		}
 
-                gf_log (this->name, GF_LOG_TRACE,
-                        "open of %s succeeded on child %s",
-                        local->loc.path,
-                        priv->children[child_index]->name);
+		ret = __afr_selfheal_data_prepare (frame, this, fd, data_lock,
+						   sources, sinks, healed_sinks,
+						   locked_replies);
+		if (ret < 0)
+			goto unlock;
+
+		source = ret;
+
+		ret = __afr_selfheal_truncate_sinks (frame, this, fd, healed_sinks,
+						     locked_replies,
+						     locked_replies[source].poststat.ia_size);
+		if (ret < 0)
+			goto unlock;
+
+		ret = 0;
+
+		/* Locking from (LLONG_MAX - 2) to (LLONG_MAX - 1) is for
+		   compatibility with older self-heal clients which do not
+		   hold a lock in the @priv->sh_domain domain to guard
+		   against concurrent ongoing self-heals
+		*/
+		afr_selfheal_inodelk (frame, this, fd->inode, this->name,
+				      LLONG_MAX - 2, 1, compat_lock);
+		compat = _gf_true;
 	}
-	UNLOCK (&frame->lock);
+unlock:
+	afr_selfheal_uninodelk (frame, this, fd->inode, this->name, 0, 0,
+				data_lock);
+	if (ret < 0)
+		goto out;
+
+	ret = afr_selfheal_data_do (frame, this, fd, source, healed_sinks,
+				    locked_replies);
+	if (ret)
+		goto out;
+
+	ret = afr_selfheal_undo_pending (frame, this, fd->inode, sources, sinks,
+					 healed_sinks, AFR_DATA_TRANSACTION,
+					 locked_replies, data_lock);
+out:
+	if (compat)
+		afr_selfheal_uninodelk (frame, this, fd->inode, this->name,
+					LLONG_MAX - 2, 1, compat_lock);
+	return ret;
+}
 
-	call_count = afr_frame_return (frame);
 
-	if (call_count == 0) {
-		if (sh->op_failed) {
-			afr_sh_data_finish (frame, this);
-			return 0;
-		}
+static fd_t *
+afr_selfheal_data_open (xlator_t *this, inode_t *inode)
+{
+	loc_t loc = {0,};
+	int ret = 0;
+	fd_t *fd = NULL;
+
+	fd = fd_create (inode, 0);
+	if (!fd)
+		return NULL;
 
-		gf_log (this->name, GF_LOG_TRACE,
-			"fd for %s opened, commencing sync",
-			local->loc.path);
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-                afr_sh_data_lock (frame, this);
+	ret = syncop_open (this, &loc, O_RDWR|O_LARGEFILE, fd);
+	if (ret) {
+		fd_unref (fd);
+		fd = NULL;
+	} else {
+		fd_bind (fd);
 	}
 
-	return 0;
+	loc_wipe (&loc);
+
+	return fd;
 }
 
 
 int
-afr_sh_data_open (call_frame_t *frame, xlator_t *this)
+afr_selfheal_data (call_frame_t *frame, xlator_t *this, inode_t *inode)
 {
-	int i = 0;
-	int call_count = 0;
-
+	afr_private_t *priv = NULL;
+	unsigned char *locked_on = NULL;
+	int ret = 0;
 	fd_t *fd = NULL;
 
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	afr_self_heal_t *sh = NULL;
-
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-        if (sh->healing_fd_opened) {
-                /* caller has opened the fd for us already, so skip open */
+	fd = afr_selfheal_data_open (this, inode);
+	if (!fd)
+		return -EIO;
 
-                afr_sh_data_lock (frame, this);
-                return 0;
-        }
-
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
-	local->call_count = call_count;
-
-	fd = fd_create (local->loc.inode, frame->root->pid);
-	sh->healing_fd = fd;
-
-	/* open sinks */
-	for (i = 0; i < priv->child_count; i++) {
-		if(!local->child_up[i])
-			continue;
+	locked_on = alloca0 (priv->child_count);
 
-		STACK_WIND_COOKIE (frame, afr_sh_data_open_cbk,
-				   (void *) (long) i,
-				   priv->children[i], 
-				   priv->children[i]->fops->open,
-				   &local->loc, 
-				   O_RDWR|O_LARGEFILE, fd, 0); 
+	ret = afr_selfheal_tryinodelk (frame, this, inode, priv->sh_domain, 0, 0,
+				       locked_on);
+	{
+		if (ret < 2) {
+			/* Either less than two subvols available, or another
+			   selfheal (from another server) is in progress. Skip
+			   for now in any case there isn't anything to do.
+			*/
+			ret = -ENOTCONN;
+			goto unlock;
+		}
 
-		if (!--call_count)
-			break;
+		ret = __afr_selfheal_data (frame, this, fd, locked_on);
 	}
+unlock:
+	afr_selfheal_uninodelk (frame, this, inode, priv->sh_domain, 0, 0, locked_on);
 
-	return 0;
-}
-
+	if (fd)
+		fd_unref (fd);
 
-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;
-
-
-	local = frame->local;
-	sh = &local->self_heal;
-
-	if (sh->need_data_self_heal && priv->data_self_heal) {
-		afr_sh_data_open (frame, this);
-	} else {
-		gf_log (this->name, GF_LOG_TRACE,
-			"not doing data self heal on %s",
-			local->loc.path);
-		afr_sh_data_done (frame, this);
-	}
-
-	return 0;
+	return ret;
 }
-
diff --git a/xlators/cluster/afr/src/afr-self-heal-entry.c b/xlators/cluster/afr/src/afr-self-heal-entry.c
index 2010ae648..9605d69f4 100644
--- a/xlators/cluster/afr/src/afr-self-heal-entry.c
+++ b/xlators/cluster/afr/src/afr-self-heal-entry.c
@@ -1,2273 +1,631 @@
 /*
-  Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
-#include <libgen.h>
-#include <unistd.h>
-#include <fnmatch.h>
-#include <sys/time.h>
-#include <stdlib.h>
-#include <signal.h>
 
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
 #endif
 
-#include "glusterfs.h"
-#include "inode.h"
 #include "afr.h"
-#include "dict.h"
-#include "xlator.h"
-#include "hashfn.h"
-#include "logging.h"
-#include "stack.h"
-#include "list.h"
-#include "call-stub.h"
-#include "defaults.h"
-#include "common-utils.h"
-#include "compat-errno.h"
-#include "compat.h"
+#include "afr-self-heal.h"
 #include "byte-order.h"
-
 #include "afr-transaction.h"
-#include "afr-self-heal.h"
-#include "afr-self-heal-common.h"
-
-int
-afr_sh_post_nonblocking_entrylk_cbk (call_frame_t *frame, xlator_t *this);
-
-int
-afr_sh_entry_done (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	/* 
-	   TODO: cleanup sh->* 
-	*/
-
-        if (sh->healing_fd)
-                fd_unref (sh->healing_fd);
-        sh->healing_fd = NULL;
-
-/*         for (i = 0; i < priv->child_count; i++) { */
-/*                 sh->locked_nodes[i] = 0; */
-/*         } */
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"self heal of %s completed",
-		local->loc.path);
-
-	sh->completion_cbk (frame, this);
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_unlock (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t         *local    = NULL;
-        afr_internal_lock_t *int_lock = NULL;
-
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-
-        int_lock->lock_cbk = afr_sh_entry_done;
-        afr_unlock (frame, this);
-
-        return 0;
-}
-
-
-int
-afr_sh_entry_finish (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t   *local = NULL;
-
-	local = frame->local;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"finishing entry selfheal of %s", local->loc.path);
-
-	afr_sh_entry_unlock (frame, this);
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_erase_pending_cbk (call_frame_t *frame, void *cookie,
-				xlator_t *this, int32_t op_ret,
-				int32_t op_errno, dict_t *xattr)
-{
-	int             call_count = 0;
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		afr_sh_entry_finish (frame, this);
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_erase_pending (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              call_count = 0;
-	int              i = 0;
-	dict_t          **erase_xattr = NULL;
-        int              need_unwind = 0;
-
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix, sh->success,
-                                 priv->child_count, AFR_ENTRY_TRANSACTION);
-
-	erase_xattr = GF_CALLOC (sizeof (*erase_xattr), priv->child_count,
-                                 gf_afr_mt_dict_t);
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->xattr[i]) {
-			call_count++;
-
-			erase_xattr[i] = get_new_dict();
-			dict_ref (erase_xattr[i]);
-		}
-	}
-
-        if (call_count == 0)
-                need_unwind = 1;
-
-	afr_sh_delta_to_xattr (priv, sh->delta_matrix, erase_xattr,
-			       priv->child_count, AFR_ENTRY_TRANSACTION);
-
-	local->call_count = call_count;
-	for (i = 0; i < priv->child_count; i++) {
-		if (!erase_xattr[i])
-			continue;
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"erasing pending flags from %s on %s",
-			local->loc.path, priv->children[i]->name);
-
-		STACK_WIND_COOKIE (frame, afr_sh_entry_erase_pending_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->xattrop,
-				   &local->loc,
-				   GF_XATTROP_ADD_ARRAY, erase_xattr[i]);
-		if (!--call_count)
-			break;
-	}
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (erase_xattr[i]) {
-			dict_unref (erase_xattr[i]);
-		}
-	}
-	GF_FREE (erase_xattr);
-
-        if (need_unwind)
-                afr_sh_entry_finish (frame, this);
-
-	return 0;
-}
-
 
 
 static int
-next_active_source (call_frame_t *frame, xlator_t *this,
-		    int current_active_source)
+afr_selfheal_entry_delete (call_frame_t *frame, xlator_t *this, inode_t *dir,
+			   const char *name, inode_t *inode, int child,
+			   struct afr_reply *replies)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              source = -1;
-	int              next_active_source = -1;
-	int              i = 0;
+	afr_private_t *priv = NULL;
+	xlator_t *subvol = NULL;
+	int ret = 0;
+	loc_t loc = {0, };
+	char g[64];
 
 	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	source = sh->source;
-
-	if (source != -1) {
-		if (current_active_source != source)
-			next_active_source = source;
-		goto out;
-	}
 
-	/*
-	  the next active sink becomes the source for the
-	  'conservative decision' of merging all entries
-	*/
+	subvol = priv->children[child];
 
-	for (i = 0; i < priv->child_count; i++) {
-		if ((sh->sources[i] == 0)
-		    && (local->child_up[i] == 1)
-		    && (i > current_active_source)) {
+	loc.parent = inode_ref (dir);
+	uuid_copy (loc.pargfid, dir->gfid);
+	loc.name = name;
+	loc.inode = inode_ref (inode);
 
-			next_active_source = i;
+	if (replies[child].valid && replies[child].op_ret == 0) {
+		switch (replies[child].poststat.ia_type) {
+		case IA_IFDIR:
+			gf_log (this->name, GF_LOG_WARNING,
+				"expunging dir %s/%s (%s) on %s",
+				uuid_utoa (dir->gfid), name,
+				uuid_utoa_r (replies[child].poststat.ia_gfid, g),
+				subvol->name);
+			ret = syncop_rmdir (subvol, &loc, 1);
 			break;
-		}
-	}
-out:
-	return next_active_source;
-}
-
-
-
-static int
-next_active_sink (call_frame_t *frame, xlator_t *this,
-		  int current_active_sink)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              next_active_sink = -1;
-	int              i = 0;
-
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	/*
-	  the next active sink becomes the source for the
-	  'conservative decision' of merging all entries
-	*/
-
-	for (i = 0; i < priv->child_count; i++) {
-		if ((sh->sources[i] == 0)
-		    && (local->child_up[i] == 1)
-		    && (i > current_active_sink)) {
-
-			next_active_sink = i;
+		default:
+			gf_log (this->name, GF_LOG_WARNING,
+				"expunging file %s/%s (%s) on %s",
+				uuid_utoa (dir->gfid), name,
+				uuid_utoa_r (replies[child].poststat.ia_gfid, g),
+				subvol->name);
+			ret = syncop_unlink (subvol, &loc);
 			break;
 		}
 	}
 
-	return next_active_sink;
-}
-
-
-int
-build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name)
-{
-	int   ret = -1;
-
-	if (!child) {
-		goto out;
-	}
-
-	if (strcmp (parent->path, "/") == 0)
-		ret = gf_asprintf ((char **)&child->path, "/%s", name);
-	else
-		ret = gf_asprintf ((char **)&child->path, "%s/%s", parent->path, 
-                                name);
-
-        if (-1 == ret) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "asprintf failed while setting child path");
-        }
-
-	if (!child->path) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-	child->name = strrchr (child->path, '/');
-	if (child->name)
-		child->name++;
-
-	child->parent = inode_ref (parent->inode);
-	child->inode = inode_new (parent->inode->table);
-
-	if (!child->inode) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-	ret = 0;
-out:
-	if (ret == -1)
-		loc_wipe (child);
+	loc_wipe (&loc);
 
 	return ret;
 }
 
 
 int
-afr_sh_entry_expunge_all (call_frame_t *frame, xlator_t *this);
-
-int
-afr_sh_entry_expunge_subvol (call_frame_t *frame, xlator_t *this,
-			     int active_src);
-
-int
-afr_sh_entry_expunge_entry_done (call_frame_t *frame, xlator_t *this,
-				 int active_src)
-{
-	int              call_count = 0;
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		afr_sh_entry_expunge_subvol (frame, this, active_src);
-
-	return 0;
-}
-
-int
-afr_sh_entry_expunge_parent_setattr_cbk (call_frame_t *expunge_frame,
-                                         void *cookie, xlator_t *this,
-                                         int32_t op_ret, int32_t op_errno,
-                                         struct iatt *preop, struct iatt *postop)
+afr_selfheal_recreate_entry (call_frame_t *frame, xlator_t *this, int dst,
+			     int source, inode_t *dir, const char *name,
+			     inode_t *inode, struct afr_reply *replies)
 {
-	afr_private_t   *priv          = NULL;
-	afr_local_t     *expunge_local = NULL;
-	afr_self_heal_t *expunge_sh    = NULL;
-        call_frame_t    *frame         = NULL;
-
-        int active_src = (long) cookie;
-
-        priv          = this->private;
-        expunge_local = expunge_frame->local;
-	expunge_sh    = &expunge_local->self_heal;
-	frame         = expunge_sh->sh_frame;
-
-        if (op_ret != 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "setattr on parent directory of %s on subvolume %s failed: %s",
-                        expunge_local->loc.path,
-                        priv->children[active_src]->name, strerror (op_errno));
-        }
-
-	AFR_STACK_DESTROY (expunge_frame);
-	afr_sh_entry_expunge_entry_done (frame, this, active_src);
-
-        return 0;
-}
-
-
-int
-afr_sh_entry_expunge_remove_cbk (call_frame_t *expunge_frame, void *cookie,
-				 xlator_t *this,
-				 int32_t op_ret, int32_t op_errno,
-                                 struct iatt *preparent,
-                                 struct iatt *postparent)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
-	afr_self_heal_t *expunge_sh = NULL;
-	int              active_src = 0;
-	call_frame_t    *frame = NULL;
-
-        int32_t valid = 0;
+	int ret = 0;
+	loc_t loc = {0,};
+	loc_t srcloc = {0,};
+	afr_private_t *priv = NULL;
+	dict_t *xdata = NULL;
+	struct iatt *iatt = NULL;
+	char *linkname = NULL;
+	mode_t mode = 0;
+	struct iatt newent = {0,};
 
 	priv = this->private;
-	expunge_local = expunge_frame->local;
-	expunge_sh = &expunge_local->self_heal;
-	frame = expunge_sh->sh_frame;
-
-	active_src = (long) cookie;
-
-	if (op_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"removed %s on %s",
-			expunge_local->loc.path,
-			priv->children[active_src]->name);
-	} else {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"removing %s on %s failed (%s)",
-			expunge_local->loc.path,
-			priv->children[active_src]->name,
-			strerror (op_errno));
-	}
 
-        valid = GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
-        afr_build_parent_loc (&expunge_sh->parent_loc, &expunge_local->loc);
+	xdata = dict_new();
+	if (!xdata)
+		return -ENOMEM;
 
-        STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_parent_setattr_cbk,
-                           (void *) (long) active_src,
-                           priv->children[active_src],
-                           priv->children[active_src]->fops->setattr,
-                           &expunge_sh->parent_loc,
-                           &expunge_sh->parentbuf,
-                           valid);
-
-        return 0;
-}
-
-
-int
-afr_sh_entry_expunge_unlink (call_frame_t *expunge_frame, xlator_t *this,
-                             int active_src)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
-
-	priv          = this->private;
-	expunge_local = expunge_frame->local;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"expunging file %s on %s",
-		expunge_local->loc.path, priv->children[active_src]->name);
-
-	STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_remove_cbk,
-			   (void *) (long) active_src,
-			   priv->children[active_src],
-			   priv->children[active_src]->fops->unlink,
-			   &expunge_local->loc);
-
-	return 0;
-}
+	loc.parent = inode_ref (dir);
+	uuid_copy (loc.pargfid, dir->gfid);
+	loc.name = name;
+	loc.inode = inode_ref (inode);
 
+	ret = afr_selfheal_entry_delete (frame, this, dir, name, inode, dst,
+					 replies);
+	if (ret)
+		goto out;
 
+	ret = dict_set_static_bin (xdata, "gfid-req",
+				   replies[source].poststat.ia_gfid, 16);
+	if (ret)
+		goto out;
 
-int
-afr_sh_entry_expunge_rmdir (call_frame_t *expunge_frame, xlator_t *this,
-                            int active_src)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
-
-	priv          = this->private;
-	expunge_local = expunge_frame->local;
-
-	gf_log (this->name, GF_LOG_DEBUG,
-		"expunging directory %s on %s",
-		expunge_local->loc.path, priv->children[active_src]->name);
-
-	STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_remove_cbk,
-			   (void *) (long) active_src,
-			   priv->children[active_src],
-			   priv->children[active_src]->fops->rmdir,
-			   &expunge_local->loc, 1);
-
-	return 0;
-}
+	iatt = &replies[source].poststat;
 
+	srcloc.inode = inode_ref (inode);
+	uuid_copy (srcloc.gfid, iatt->ia_gfid);
 
-int
-afr_sh_entry_expunge_remove (call_frame_t *expunge_frame, xlator_t *this,
-			     int active_src, struct iatt *buf)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
-	afr_self_heal_t *expunge_sh = NULL;
-	int              source = 0;
-	call_frame_t    *frame = NULL;
-	int              type = 0;
+	mode = st_mode_from_ia (iatt->ia_prot, iatt->ia_type);
 
-	priv = this->private;
-	expunge_local = expunge_frame->local;
-	expunge_sh = &expunge_local->self_heal;
-	frame = expunge_sh->sh_frame;
-	source = expunge_sh->source;
-
-	type = buf->ia_type;
-
-	switch (type) {
-	case IA_IFSOCK:
-	case IA_IFREG:
-	case IA_IFBLK:
-	case IA_IFCHR:
-	case IA_IFIFO:
-	case IA_IFLNK:
-		afr_sh_entry_expunge_unlink (expunge_frame, this, active_src);
-                break;
+	switch (iatt->ia_type) {
 	case IA_IFDIR:
-		afr_sh_entry_expunge_rmdir (expunge_frame, this, active_src);
+		ret = syncop_mkdir (priv->children[dst], &loc, mode, xdata, 0);
+		break;
+	case IA_IFLNK:
+		ret = syncop_lookup (priv->children[dst], &srcloc, 0, 0, 0, 0);
+		if (ret == 0) {
+			ret = syncop_link (priv->children[dst], &srcloc, &loc);
+		} else {
+			ret = syncop_readlink (priv->children[source], &srcloc,
+					       &linkname, 4096);
+			if (ret <= 0)
+				goto out;
+			ret = syncop_symlink (priv->children[dst], &loc, linkname,
+					      xdata, NULL);
+		}
 		break;
 	default:
-		gf_log (this->name, GF_LOG_ERROR,
-			"%s has unknown file type on %s: 0%o",
-			expunge_local->loc.path,
-			priv->children[source]->name, type);
-		goto out;
+		ret = dict_set_int32 (xdata, GLUSTERFS_INTERNAL_FOP_KEY, 1);
+		if (ret)
+			goto out;
+		ret = syncop_mknod (priv->children[dst], &loc, mode,
+				    iatt->ia_rdev, xdata, &newent);
+		if (ret == 0 && iatt->ia_size && !newent.ia_size) {
+			/* New entry created. Mark @dst pending on all sources */
+			ret = 1;
+		}
 		break;
 	}
 
-	return 0;
 out:
-	AFR_STACK_DESTROY (expunge_frame);
-	afr_sh_entry_expunge_entry_done (frame, this, active_src);
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_expunge_lookup_cbk (call_frame_t *expunge_frame, void *cookie,
-				xlator_t *this,
-				int32_t op_ret,	int32_t op_errno,
-                                inode_t *inode, struct iatt *buf, dict_t *x,
-                                struct iatt *postparent)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
-	afr_self_heal_t *expunge_sh = NULL;
-	call_frame_t    *frame = NULL;
-	int              active_src = 0;
-
-	priv = this->private;
-	expunge_local = expunge_frame->local;
-	expunge_sh = &expunge_local->self_heal;
-	frame = expunge_sh->sh_frame;
-	active_src = (long) cookie;
-
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"lookup of %s on %s failed (%s)",
-			expunge_local->loc.path,
-			priv->children[active_src]->name,
-			strerror (op_errno));
-		goto out;
-	}
-
-	afr_sh_entry_expunge_remove (expunge_frame, this, active_src, buf);
-
-	return 0;
-out:
-	AFR_STACK_DESTROY (expunge_frame);
-	afr_sh_entry_expunge_entry_done (frame, this, active_src);
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_expunge_purge (call_frame_t *expunge_frame, xlator_t *this,
-			    int active_src)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
-
-	priv = this->private;
-	expunge_local = expunge_frame->local;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"looking up %s on %s",
-		expunge_local->loc.path, priv->children[active_src]->name);
-	
-	STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_lookup_cbk,
-			   (void *) (long) active_src,
-			   priv->children[active_src],
-			   priv->children[active_src]->fops->lookup,
-			   &expunge_local->loc, 0);
-
-	return 0;
+	if (xdata)
+		dict_unref (xdata);
+	loc_wipe (&loc);
+	loc_wipe (&srcloc);
+	return ret;
 }
 
 
-int
-afr_sh_entry_expunge_entry_cbk (call_frame_t *expunge_frame, void *cookie,
-				xlator_t *this,
-				int32_t op_ret,	int32_t op_errno,
-                                inode_t *inode, struct iatt *buf, dict_t *x,
-                                struct iatt *postparent)
+static int
+afr_selfheal_newentry_mark (call_frame_t *frame, xlator_t *this, inode_t *inode,
+			    int source, struct afr_reply *replies,
+			    unsigned char *sources, unsigned char *newentry)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
-	afr_self_heal_t *expunge_sh = NULL;
-	int              source = 0;
-	call_frame_t    *frame = NULL;
-	int              active_src = 0;
-
+	int ret = 0;
+	int i = 0;
+	afr_private_t *priv = NULL;
+	dict_t *xattr = NULL;
+	int **changelog = NULL;
+	int idx = 0;
 
 	priv = this->private;
-	expunge_local = expunge_frame->local;
-	expunge_sh = &expunge_local->self_heal;
-	frame = expunge_sh->sh_frame;
-	active_src = expunge_sh->active_source;
-	source = (long) cookie;
 
-	if (op_ret == -1 && op_errno == ENOENT && postparent) {
+	idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
 
-		gf_log (this->name, GF_LOG_TRACE,
-			"missing entry %s on %s",
-			expunge_local->loc.path,
-			priv->children[source]->name);
+	uuid_copy (inode->gfid, replies[source].poststat.ia_gfid);
 
-                expunge_sh->parentbuf = *postparent;
+	changelog = afr_matrix_create (priv->child_count, AFR_NUM_CHANGE_LOGS);
 
-		afr_sh_entry_expunge_purge (expunge_frame, this, active_src);
+	xattr = dict_new();
+	if (!xattr)
+		return -ENOMEM;
 
-		return 0;
+	for (i = 0; i < priv->child_count; i++) {
+		if (!newentry[i])
+			continue;
+		changelog[i][idx] = hton32(1);
 	}
 
-	if (op_ret == 0) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"%s exists under %s",
-			expunge_local->loc.path,
-			priv->children[source]->name);
-	} else {
-		gf_log (this->name, GF_LOG_TRACE,
-			"looking up %s under %s failed (%s)",
-			expunge_local->loc.path,
-			priv->children[source]->name,
-			strerror (op_errno));
-	}
+	afr_set_pending_dict (priv, xattr, changelog);
 
-	AFR_STACK_DESTROY (expunge_frame);
-	afr_sh_entry_expunge_entry_done (frame, this, active_src);
+	for (i = 0; i < priv->child_count; i++) {
+		if (!sources[i])
+			continue;
+		afr_selfheal_post_op (frame, this, inode, i, xattr);
+	}
 
-	return 0;
+	dict_unref (xattr);
+	return ret;
 }
 
 
-int
-afr_sh_entry_expunge_entry (call_frame_t *frame, xlator_t *this,
-			    char *name)
+static int
+__afr_selfheal_heal_dirent (call_frame_t *frame, xlator_t *this, fd_t *fd,
+			    char *name, inode_t *inode, int source,
+			    unsigned char *sources, unsigned char *healed_sinks,
+			    unsigned char *locked_on, struct afr_reply *replies)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              ret = -1;
-	call_frame_t    *expunge_frame = NULL;
-	afr_local_t     *expunge_local = NULL;
-	afr_self_heal_t *expunge_sh = NULL;
-	int              active_src = 0;
-	int              source = 0;
-	int              op_errno = 0;
+	int ret = 0;
+	afr_private_t *priv = NULL;
+	int i = 0;
+	unsigned char *newentry = NULL;
 
 	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	active_src = sh->active_source;
-	source = sh->source;
+	newentry = alloca0 (priv->child_count);
 
-	if ((strcmp (name, ".") == 0)
-	    || (strcmp (name, "..") == 0)
-            || ((strcmp (local->loc.path, "/") == 0)
-                && (strcmp (name, GF_REPLICATE_TRASH_DIR) == 0))) {
+	if (!replies[source].valid)
+		return -EIO;
 
-                gf_log (this->name, GF_LOG_TRACE,
-			"skipping inspection of %s under %s",
-			name, local->loc.path);
-		goto out;
-	}
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"inspecting existance of %s under %s",
-		name, local->loc.path);
-
-	expunge_frame = copy_frame (frame);
-	if (!expunge_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-	ALLOC_OR_GOTO (expunge_local, afr_local_t, out);
-
-	expunge_frame->local = expunge_local;
-	expunge_sh = &expunge_local->self_heal;
-	expunge_sh->sh_frame = frame;
-	expunge_sh->active_source = active_src;
-
-	ret = build_child_loc (this, &expunge_local->loc, &local->loc, name);
-	if (ret != 0) {
-		goto out;
-	}
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"looking up %s on %s", expunge_local->loc.path,
-		priv->children[source]->name);
-
-	STACK_WIND_COOKIE (expunge_frame,
-			   afr_sh_entry_expunge_entry_cbk,
-			   (void *) (long) source,
-			   priv->children[source],
-			   priv->children[source]->fops->lookup,
-			   &expunge_local->loc, 0);
-
-	ret = 0;
-out:
-	if (ret == -1)
-		afr_sh_entry_expunge_entry_done (frame, this, active_src);
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_expunge_readdir_cbk (call_frame_t *frame, void *cookie,
-				  xlator_t *this,
-				  int32_t op_ret, int32_t op_errno,
-				  gf_dirent_t *entries)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	gf_dirent_t     *entry = NULL;
-	off_t            last_offset = 0;
-	int              active_src = 0;
-	int              entry_count = 0;
-
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	active_src = sh->active_source;
-
-	if (op_ret <= 0) {
-		if (op_ret < 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"readdir of %s on subvolume %s failed (%s)",
-				local->loc.path,
-				priv->children[active_src]->name,
-				strerror (op_errno));
+	for (i = 0; i < priv->child_count; i++) {
+		if (!healed_sinks[i])
+			continue;
+		if (replies[source].op_ret == -1 &&
+		    replies[source].op_errno == ENOENT) {
+			ret = afr_selfheal_entry_delete (frame, this, fd->inode,
+							 name, inode, i, replies);
 		} else {
-			gf_log (this->name, GF_LOG_TRACE,
-				"readdir of %s on subvolume %s complete",
-				local->loc.path,
-				priv->children[active_src]->name);
+			if (!uuid_compare (replies[i].poststat.ia_gfid,
+					   replies[source].poststat.ia_gfid))
+				continue;
+
+			ret = afr_selfheal_recreate_entry (frame, this, i, source,
+							   fd->inode, name, inode,
+							   replies);
+			if (ret > 0) {
+				newentry[i] = 1;
+				ret = 0;
+			}
 		}
-
-		afr_sh_entry_expunge_all (frame, this);
-		return 0;
-	}
-
-	list_for_each_entry (entry, &entries->list, list) {
-		last_offset = entry->d_off;
-                entry_count++;
-	}
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"readdir'ed %d entries from %s",
-		entry_count, priv->children[active_src]->name);
-
-	sh->offset = last_offset;
-	local->call_count = entry_count;
-
-	list_for_each_entry (entry, &entries->list, list) {
-                afr_sh_entry_expunge_entry (frame, this, entry->d_name);
-	}
-
-	return 0;
-}
-
-int
-afr_sh_entry_expunge_subvol (call_frame_t *frame, xlator_t *this,
-			     int active_src)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	STACK_WIND (frame, afr_sh_entry_expunge_readdir_cbk,
-		    priv->children[active_src],
-		    priv->children[active_src]->fops->readdir,
-		    sh->healing_fd, sh->block_size, sh->offset);
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_expunge_all (call_frame_t *frame, xlator_t *this)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              active_src = -1;
-
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	sh->offset = 0;
-
-	if (sh->source == -1) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"no active sources for %s to expunge entries",
-			local->loc.path);
-		goto out;
-	}
-
-	active_src = next_active_sink (frame, this, sh->active_source);
-	sh->active_source = active_src;
-
-	if (sh->op_failed) {
-		goto out;
-	}
-
-	if (active_src == -1) {
-		/* completed creating missing files on all subvolumes */
-		goto out;
-	}
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"expunging entries of %s on %s to other sinks",
-		local->loc.path, priv->children[active_src]->name);
-
-	afr_sh_entry_expunge_subvol (frame, this, active_src);
-
-	return 0;
-out:
-	afr_sh_entry_erase_pending (frame, this);
-	return 0;
-
-}
-
-
-int
-afr_sh_entry_impunge_all (call_frame_t *frame, xlator_t *this);
-
-int
-afr_sh_entry_impunge_subvol (call_frame_t *frame, xlator_t *this,
-			     int active_src);
-
-int
-afr_sh_entry_impunge_entry_done (call_frame_t *frame, xlator_t *this,
-				 int active_src)
-{
-	int              call_count = 0;
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		afr_sh_entry_impunge_subvol (frame, this, active_src);
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_impunge_setattr_cbk (call_frame_t *impunge_frame, void *cookie,
-				  xlator_t *this,
-                                  int32_t op_ret, int32_t op_errno,
-                                  struct iatt *preop, struct iatt *postop)
-{
-	int              call_count = 0;
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	call_frame_t    *frame = NULL;
-	int              active_src = 0;
-	int              child_index = 0;
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-	frame = impunge_sh->sh_frame;
-        local = frame->local;
-        sh    = &local->self_heal;
-        active_src = sh->active_source;
-	child_index = (long) cookie;
-
-	if (op_ret == 0) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"setattr done for %s on %s",
-			impunge_local->loc.path,
-			priv->children[child_index]->name);
-	} else {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"setattr (%s) on %s failed (%s)",
-			impunge_local->loc.path,
-			priv->children[child_index]->name,
-			strerror (op_errno));
-	}
-
-	LOCK (&impunge_frame->lock);
-	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
-
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
-	}
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_impunge_xattrop_cbk (call_frame_t *impunge_frame, void *cookie,
-                                  xlator_t *this,
-                                  int32_t op_ret, int32_t op_errno,
-                                  dict_t *xattr)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              child_index = 0;
-
-        struct iatt stbuf;
-        int32_t     valid = 0;
-
-	priv          = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh    = &impunge_local->self_heal;
-
-	child_index = (long) cookie;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"setting ownership of %s on %s to %d/%d",
-		impunge_local->loc.path,
-		priv->children[child_index]->name,
-		impunge_local->cont.lookup.buf.ia_uid,
-		impunge_local->cont.lookup.buf.ia_gid);
-
-	stbuf.ia_atime = impunge_local->cont.lookup.buf.ia_atime;
-	stbuf.ia_atime_nsec = impunge_local->cont.lookup.buf.ia_atime_nsec;
-	stbuf.ia_mtime = impunge_local->cont.lookup.buf.ia_mtime;
-	stbuf.ia_mtime_nsec = impunge_local->cont.lookup.buf.ia_mtime_nsec;
-
-        stbuf.ia_uid = impunge_local->cont.lookup.buf.ia_uid;
-        stbuf.ia_gid = impunge_local->cont.lookup.buf.ia_gid;
-
-        valid = GF_SET_ATTR_UID   | GF_SET_ATTR_GID |
-                GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
-
-	STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_setattr_cbk,
-			   (void *) (long) child_index,
-			   priv->children[child_index],
-			   priv->children[child_index]->fops->setattr,
-			   &impunge_local->loc,
-                           &stbuf, valid);
-        return 0;
-}
-
-
-int
-afr_sh_entry_impunge_parent_setattr_cbk (call_frame_t *setattr_frame,
-                                         void *cookie, xlator_t *this,
-                                         int32_t op_ret, int32_t op_errno,
-                                         struct iatt *preop, struct iatt *postop)
-{
-        loc_t *parent_loc = cookie;
-
-        if (op_ret != 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "setattr on parent directory failed: %s",
-                        strerror (op_errno));
-        }
-
-        loc_wipe (parent_loc);
-
-        GF_FREE (parent_loc);
-
-        AFR_STACK_DESTROY (setattr_frame);
-        return 0;
-}
-
-
-int
-afr_sh_entry_impunge_newfile_cbk (call_frame_t *impunge_frame, void *cookie,
-				  xlator_t *this,
-				  int32_t op_ret, int32_t op_errno,
-                                  inode_t *inode, struct iatt *stbuf,
-                                  struct iatt *preparent,
-                                  struct iatt *postparent)
-{
-	int              call_count = 0;
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	call_frame_t    *frame = NULL;
-	int              active_src = 0;
-	int              child_index = 0;
-        int              pending_array[3] = {0, };
-        dict_t          *xattr = NULL;
-        int              ret = 0;
-        int              idx = 0;
-        afr_local_t     *local = NULL;
-        afr_self_heal_t *sh = NULL;
-
-        call_frame_t *setattr_frame = NULL;
-        int32_t valid = 0;
-        loc_t *parent_loc = NULL;
-        struct iatt parentbuf;
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-	frame = impunge_sh->sh_frame;
-        local = frame->local;
-        sh    = &local->self_heal;
-        active_src = sh->active_source;
-
-	child_index = (long) cookie;
-
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"creation of %s on %s failed (%s)",
-			impunge_local->loc.path,
-			priv->children[child_index]->name,
-			strerror (op_errno));
-		goto out;
-	}
-
-	inode->ia_type = stbuf->ia_type;
-
-        xattr = get_new_dict ();
-        dict_ref (xattr);
-
-        idx = afr_index_for_transaction_type (AFR_METADATA_TRANSACTION);
-        pending_array[idx] = hton32 (1);
-        if (IA_ISDIR (stbuf->ia_type))
-                idx = afr_index_for_transaction_type (AFR_ENTRY_TRANSACTION);
-        else
-                idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
-        pending_array[idx] = hton32 (1);
-
-        ret = dict_set_static_bin (xattr, priv->pending_key[child_index],
-                                   pending_array, sizeof (pending_array));
-        if (ret < 0)
-                gf_log (this->name, GF_LOG_WARNING,
-                        "Unable to set dict value.");
-
-        valid         = GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
-        parentbuf     = impunge_sh->parentbuf;
-        setattr_frame = copy_frame (impunge_frame);
-
-        parent_loc = GF_CALLOC (1, sizeof (*parent_loc),
-                                gf_afr_mt_loc_t);
-        afr_build_parent_loc (parent_loc, &impunge_local->loc);
-
-	STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_xattrop_cbk,
-			   (void *) (long) child_index,
-			   priv->children[active_src],
-			   priv->children[active_src]->fops->xattrop,
-			   &impunge_local->loc, GF_XATTROP_ADD_ARRAY, xattr);
-
-        STACK_WIND_COOKIE (setattr_frame, afr_sh_entry_impunge_parent_setattr_cbk,
-                           (void *) (long) parent_loc,
-                           priv->children[child_index],
-                           priv->children[child_index]->fops->setattr,
-                           parent_loc, &parentbuf, valid);
-
-        dict_unref (xattr);
-
-	return 0;
-
-out:
-	LOCK (&impunge_frame->lock);
-	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
-
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
+		if (ret < 0)
+			break;
 	}
 
-	return 0;
-}
-
-
-int
-afr_sh_entry_impunge_mknod (call_frame_t *impunge_frame, xlator_t *this,
-			    int child_index, struct iatt *stbuf)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-        dict_t          *dict = NULL;
-
-        int ret = 0;
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-
-	gf_log (this->name, GF_LOG_DEBUG,
-		"creating missing file %s on %s",
-		impunge_local->loc.path,
-		priv->children[child_index]->name);
-
-        dict = dict_new ();
-        if (!dict)
-                gf_log (this->name, GF_LOG_ERROR, "Out of memory");
-
-        ret = afr_set_dict_gfid (dict, stbuf->ia_gfid);
-        if (ret)
-                gf_log (this->name, GF_LOG_DEBUG, "gfid set failed");
-
-	STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_newfile_cbk,
-			   (void *) (long) child_index,
-			   priv->children[child_index],
-			   priv->children[child_index]->fops->mknod,
-			   &impunge_local->loc,
-			   st_mode_from_ia (stbuf->ia_prot, stbuf->ia_type),
-                           stbuf->ia_rdev, dict);
-
-        if (dict)
-                dict_unref (dict);
-
-	return 0;
-}
-
-
-
-int
-afr_sh_entry_impunge_mkdir (call_frame_t *impunge_frame, xlator_t *this,
-			    int child_index, struct iatt *stbuf)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-        dict_t          *dict = NULL;
-
-        int ret = 0;
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-
-        dict = dict_new ();
-        if (!dict) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                return 0;
-        }
-
-        ret = afr_set_dict_gfid (dict, stbuf->ia_gfid);
-        if (ret)
-                gf_log (this->name, GF_LOG_DEBUG, "gfid set failed");
-
-	gf_log (this->name, GF_LOG_DEBUG,
-		"creating missing directory %s on %s",
-		impunge_local->loc.path,
-		priv->children[child_index]->name);
-
-	STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_newfile_cbk,
-			   (void *) (long) child_index,
-			   priv->children[child_index],
-			   priv->children[child_index]->fops->mkdir,
-			   &impunge_local->loc,
-                           st_mode_from_ia (stbuf->ia_prot, stbuf->ia_type),
-                           dict);
-
-        if (dict)
-                dict_unref (dict);
-
-	return 0;
+	if (AFR_COUNT (newentry, priv->child_count))
+		afr_selfheal_newentry_mark (frame, this, inode, source, replies,
+					    sources, newentry);
+	return ret;
 }
 
 
-int
-afr_sh_entry_impunge_symlink (call_frame_t *impunge_frame, xlator_t *this,
-			      int child_index, const char *linkname)
+static int
+__afr_selfheal_merge_dirent (call_frame_t *frame, xlator_t *this, fd_t *fd,
+			     char *name, inode_t *inode, unsigned char *sources,
+			     unsigned char *healed_sinks, unsigned char *locked_on,
+			     struct afr_reply *replies)
 {
-	afr_private_t   *priv          = NULL;
-	afr_local_t     *impunge_local = NULL;
-        dict_t          *dict          = NULL;
-        struct iatt     *buf           = NULL;
-
-        int ret = 0;
+	int ret = 0;
+	afr_private_t *priv = NULL;
+	int i = 0;
+	int source = -1;
 
 	priv = this->private;
-	impunge_local = impunge_frame->local;
-
-        buf = &impunge_local->cont.symlink.buf;
-
-        dict = dict_new ();
-        if (!dict) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                afr_sh_entry_impunge_entry_done (impunge_frame, this, 0);
-        }
-
-        ret = afr_set_dict_gfid (dict, buf->ia_gfid);
-        if (ret)
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "dict set gfid failed");
-
-	gf_log (this->name, GF_LOG_DEBUG,
-		"creating missing symlink %s -> %s on %s",
-		impunge_local->loc.path, linkname,
-		priv->children[child_index]->name);
-
-	STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_newfile_cbk,
-			   (void *) (long) child_index,
-			   priv->children[child_index],
-			   priv->children[child_index]->fops->symlink,
-			   linkname, &impunge_local->loc, dict);
-
-        if (dict)
-                dict_unref (dict);
 
-	return 0;
-}
-
-
-int
-afr_sh_entry_impunge_symlink_unlink_cbk (call_frame_t *impunge_frame,
-                                         void *cookie, xlator_t *this,
-                                         int32_t op_ret, int32_t op_errno,
-                                         struct iatt *preparent,
-                                         struct iatt *postparent)
-{
-        afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              child_index = -1;
-	call_frame_t    *frame = NULL;
-	int              call_count = -1;
-	int              active_src = -1;
-
-	priv          = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh    = &impunge_local->self_heal;
-	frame         = impunge_sh->sh_frame;
-	active_src    = impunge_sh->active_source;
-
-	child_index = (long) cookie;
-
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"unlink of %s on %s failed (%s)",
-			impunge_local->loc.path,
-			priv->children[child_index]->name,
-			strerror (op_errno));
-		goto out;
-	}
-
-        afr_sh_entry_impunge_symlink (impunge_frame, this, child_index,
-                                      impunge_sh->linkname);
-
-        return 0;
-out:
-	LOCK (&impunge_frame->lock);
-	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
-
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
+	for (i = 0; i < priv->child_count; i++) {
+		if (replies[i].valid && replies[i].op_ret == 0) {
+			source = i;
+			break;
+		}
 	}
 
-	return 0;
-}
-
-
-int
-afr_sh_entry_impunge_symlink_unlink (call_frame_t *impunge_frame, xlator_t *this,
-                                     int child_index)
-{
-	afr_private_t   *priv          = NULL;
-	afr_local_t     *impunge_local = NULL;
-
-	priv          = this->private;
-	impunge_local = impunge_frame->local;
-
-	gf_log (this->name, GF_LOG_DEBUG,
-		"unlinking symlink %s with wrong target on %s",
-		impunge_local->loc.path,
-		priv->children[child_index]->name);
-
-        STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_symlink_unlink_cbk,
-                           (void *) (long) child_index,
-                           priv->children[child_index],
-                           priv->children[child_index]->fops->unlink,
-                           &impunge_local->loc);
-
-        return 0;
-}
-
-
-int
-afr_sh_entry_impunge_readlink_sink_cbk (call_frame_t *impunge_frame, void *cookie,
-                                        xlator_t *this,
-                                        int32_t op_ret, int32_t op_errno,
-                                        const char *linkname, struct iatt *sbuf)
-{
-        afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              child_index = -1;
-	call_frame_t    *frame = NULL;
-	int              call_count = -1;
-	int              active_src = -1;
-
-	priv          = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh    = &impunge_local->self_heal;
-	frame         = impunge_sh->sh_frame;
-	active_src    = impunge_sh->active_source;
-
-	child_index = (long) cookie;
-
-	if ((op_ret == -1) && (op_errno != ENOENT)) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"readlink of %s on %s failed (%s)",
-			impunge_local->loc.path,
-			priv->children[active_src]->name,
-			strerror (op_errno));
-		goto out;
+	if (source == -1) {
+		/* entry got deleted in the mean time? */
+		return 0;
 	}
 
-        /* symlink doesn't exist on the sink */
-
-        if ((op_ret == -1) && (op_errno == ENOENT)) {
-                afr_sh_entry_impunge_symlink (impunge_frame, this,
-                                              child_index, impunge_sh->linkname);
-                return 0;
-        }
-
-
-        /* symlink exists on the sink, so check if targets match */
-
-        if (strcmp (linkname, impunge_sh->linkname) == 0) {
-                /* targets match, nothing to do */
-
-                goto out;
-        } else {
-                /*
-                 * Hah! Sneaky wolf in sheep's clothing!
-                 */
-                afr_sh_entry_impunge_symlink_unlink (impunge_frame, this,
-                                                     child_index);
-                return 0;
-        }
+	for (i = 0; i < priv->child_count; i++) {
+		if (i == source || !healed_sinks[i])
+			continue;
 
-out:
-	LOCK (&impunge_frame->lock);
-	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
+		if (replies[i].op_errno != ENOENT)
+			continue;
 
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
+		ret = afr_selfheal_recreate_entry (frame, this, i, source,
+						   fd->inode, name, inode,
+						   replies);
 	}
 
-	return 0;
+	return ret;
 }
 
 
-int
-afr_sh_entry_impunge_readlink_sink (call_frame_t *impunge_frame, xlator_t *this,
-                                    int child_index)
+static int
+__afr_selfheal_entry_dirent (call_frame_t *frame, xlator_t *this, fd_t *fd,
+			     char *name, inode_t *inode, int source,
+			     unsigned char *sources, unsigned char *healed_sinks,
+			     unsigned char *locked_on, struct afr_reply *replies)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-
-	gf_log (this->name, GF_LOG_DEBUG,
-		"checking symlink target of %s on %s",
-		impunge_local->loc.path, priv->children[child_index]->name);
+	int ret = -1;
 
-	STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_readlink_sink_cbk,
-			   (void *) (long) child_index,
-			   priv->children[child_index],
-			   priv->children[child_index]->fops->readlink,
-			   &impunge_local->loc, 4096);
-
-	return 0;
+	if (source < 0)
+		ret = __afr_selfheal_merge_dirent (frame, this, fd, name, inode,
+						   sources, healed_sinks,
+						   locked_on, replies);
+	else
+		ret = __afr_selfheal_heal_dirent (frame, this, fd, name, inode,
+						  source, sources, healed_sinks,
+						  locked_on, replies);
+	return ret;
 }
 
 
-int
-afr_sh_entry_impunge_readlink_cbk (call_frame_t *impunge_frame, void *cookie,
-				   xlator_t *this,
-				   int32_t op_ret, int32_t op_errno,
-				   const char *linkname, struct iatt *sbuf)
+static int
+afr_selfheal_entry_dirent (call_frame_t *frame, xlator_t *this, fd_t *fd,
+			   int source, unsigned char *sources,
+			   unsigned char *healed_sinks, char *name)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              child_index = -1;
-	call_frame_t    *frame = NULL;
-	int              call_count = -1;
-	int              active_src = -1;
+	afr_private_t *priv = NULL;
+	int ret = 0;
+	unsigned char *locked_on = NULL;
+	struct afr_reply *replies = NULL;
+	inode_t *inode = NULL;
 
 	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-	frame = impunge_sh->sh_frame;
-	active_src = impunge_sh->active_source;
-
-	child_index = (long) cookie;
-
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"readlink of %s on %s failed (%s)",
-			impunge_local->loc.path,
-			priv->children[active_src]->name,
-			strerror (op_errno));
-		goto out;
-	}
 
-        impunge_sh->linkname = gf_strdup (linkname);
-	afr_sh_entry_impunge_readlink_sink (impunge_frame, this, child_index);
+	locked_on = alloca0 (priv->child_count);
 
-	return 0;
+	replies = alloca0 (priv->child_count * sizeof(*replies));
 
-out:
-	LOCK (&impunge_frame->lock);
+	ret = afr_selfheal_entrylk (frame, this, fd->inode, this->name,
+				    name, locked_on);
 	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
-
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
-	}
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_impunge_readlink (call_frame_t *impunge_frame, xlator_t *this,
-			       int child_index, struct iatt *stbuf)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              active_src = -1;
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-	active_src = impunge_sh->active_source;
-        impunge_local->cont.symlink.buf = *stbuf;
-
-	STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_readlink_cbk,
-			   (void *) (long) child_index,
-			   priv->children[active_src],
-			   priv->children[active_src]->fops->readlink,
-			   &impunge_local->loc, 4096);
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_impunge_recreate_lookup_cbk (call_frame_t *impunge_frame,
-					  void *cookie, xlator_t *this,
-					  int32_t op_ret, int32_t op_errno,
-					  inode_t *inode, struct iatt *buf,
-					  dict_t *xattr,struct iatt *postparent)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              active_src = 0;
-	int              type = 0;
-	int              child_index = 0;
-	call_frame_t    *frame = NULL;
-	int              call_count = 0;
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-	frame = impunge_sh->sh_frame;
-
-	child_index = (long) cookie;
-
-	active_src = impunge_sh->active_source;
-
-	if (op_ret != 0) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"looking up %s on %s (for %s) failed (%s)",
-			impunge_local->loc.path,
-			priv->children[active_src]->name,
-			priv->children[child_index]->name,
-			strerror (op_errno));
-		goto out;
-	}
-
-        impunge_sh->parentbuf = *postparent;
-
-	impunge_local->cont.lookup.buf = *buf;
-	type = buf->ia_type;
-
-	switch (type) {
-	case IA_IFSOCK:
-	case IA_IFREG:
-	case IA_IFBLK:
-	case IA_IFCHR:
-	case IA_IFIFO:
-		afr_sh_entry_impunge_mknod (impunge_frame, this,
-					    child_index, buf);
-		break;
-	case IA_IFLNK:
-		afr_sh_entry_impunge_readlink (impunge_frame, this,
-					       child_index, buf);
-		break;
-	case IA_IFDIR:
-		afr_sh_entry_impunge_mkdir (impunge_frame, this,
-					    child_index, buf);
-		break;
-	default:
-		gf_log (this->name, GF_LOG_ERROR,
-			"%s has unknown file type on %s: 0%o",
-			impunge_local->loc.path,
-			priv->children[active_src]->name, type);
-		goto out;
-		break;
-	}
-
-	return 0;
+		if (ret < 2) {
+			ret = -ENOTCONN;
+			goto unlock;
+		}
 
-out:
-	LOCK (&impunge_frame->lock);
-	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
+		inode = afr_selfheal_unlocked_lookup_on (frame, fd->inode, name,
+							 replies, locked_on);
+		if (!inode) {
+			ret = -ENOMEM;
+			goto unlock;
+		}
 
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
+		ret = __afr_selfheal_entry_dirent (frame, this, fd, name, inode,
+						   source, sources, healed_sinks,
+						   locked_on, replies);
 	}
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_impunge_recreate (call_frame_t *impunge_frame, xlator_t *this,
-			       int child_index)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              active_src = 0;
-
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-
-	active_src = impunge_sh->active_source;
-
-	STACK_WIND_COOKIE (impunge_frame,
-			   afr_sh_entry_impunge_recreate_lookup_cbk,
-			   (void *) (long) child_index,
-			   priv->children[active_src],
-			   priv->children[active_src]->fops->lookup,
-			   &impunge_local->loc, 0);
-
-	return 0;
+unlock:
+	afr_selfheal_unentrylk (frame, this, fd->inode, this->name, name,
+				locked_on);
+	if (inode)
+		inode_unref (inode);
+	return ret;
 }
 
 
-int
-afr_sh_entry_impunge_entry_cbk (call_frame_t *impunge_frame, void *cookie,
-				xlator_t *this,
-				int32_t op_ret,	int32_t op_errno,
-                                inode_t *inode, struct iatt *buf, dict_t *x,
-                                struct iatt *postparent)
+static int
+afr_selfheal_entry_do_subvol (call_frame_t *frame, xlator_t *this, fd_t *fd,
+			      int child, int source, unsigned char *sources,
+			      unsigned char *healed_sinks)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              call_count = 0;
-	int              child_index = 0;
-	call_frame_t    *frame = NULL;
-	int              active_src = 0;
+	int ret = 0;
+	gf_dirent_t entries;
+	gf_dirent_t *entry = NULL;
+	off_t offset = 0;
+	call_frame_t *iter_frame = NULL;
+	xlator_t *subvol = NULL;
+	afr_private_t *priv = NULL;
 
 	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-	frame = impunge_sh->sh_frame;
-	child_index = (long) cookie;
-	active_src = impunge_sh->active_source;
+	subvol = priv->children[child];
 
-	if ((op_ret == -1 && op_errno == ENOENT)
-            || (IA_ISLNK (impunge_sh->impunging_entry_mode))) {
+	INIT_LIST_HEAD (&entries.list);
 
-                /*
-                 * A symlink's target might have changed, so
-                 * always go down the recreate path for them.
-                 */
+	iter_frame = afr_copy_frame (frame);
+	if (!iter_frame)
+		return -ENOMEM;
 
-		/* decrease call_count in recreate-callback */
+	while ((ret = syncop_readdir (subvol, fd, 131072, offset, &entries))) {
+		if (ret > 0)
+			ret = 0;
+		list_for_each_entry (entry, &entries.list, list) {
+			offset = entry->d_off;
 
-		gf_log (this->name, GF_LOG_TRACE,
-			"missing entry %s on %s",
-			impunge_local->loc.path,
-			priv->children[child_index]->name);
+			if (!strcmp (entry->d_name, ".") ||
+			    !strcmp (entry->d_name, ".."))
+				continue;
 
-		afr_sh_entry_impunge_recreate (impunge_frame, this,
-					       child_index);
-		return 0;
-	}
-
-	if (op_ret == 0) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"%s exists under %s",
-			impunge_local->loc.path,
-			priv->children[child_index]->name);
+			if (__is_root_gfid (fd->inode->gfid) &&
+			    !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR))
+				continue;
 
-                impunge_sh->parentbuf = *postparent;
-	} else {
-		gf_log (this->name, GF_LOG_TRACE,
-			"looking up %s under %s failed (%s)",
-			impunge_local->loc.path,
-			priv->children[child_index]->name,
-			strerror (op_errno));
-	}
+			ret = afr_selfheal_entry_dirent (iter_frame, this, fd,
+							 source, sources,
+							 healed_sinks,
+							 entry->d_name);
+			AFR_STACK_RESET (iter_frame);
 
-	LOCK (&impunge_frame->lock);
-	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
+			if (ret)
+				break;
+		}
 
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
+		gf_dirent_free (&entries);
+		if (ret)
+			break;
 	}
 
-	return 0;
+	AFR_STACK_DESTROY (iter_frame);
+	return ret;
 }
 
-
-int
-afr_sh_entry_impunge_entry (call_frame_t *frame, xlator_t *this,
-			    gf_dirent_t *entry)
+static int
+afr_selfheal_entry_do (call_frame_t *frame, xlator_t *this, fd_t *fd,
+		       int source, unsigned char *sources,
+		       unsigned char *healed_sinks,
+		       struct afr_reply *locked_replies)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              ret = -1;
-	call_frame_t    *impunge_frame = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              active_src = 0;
-	int              i = 0;
-	int              call_count = 0;
-	int              op_errno = 0;
+	int i = 0;
+	afr_private_t *priv = NULL;
+	int ret = 0;
 
 	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	active_src = sh->active_source;
-
-	if ((strcmp (entry->d_name, ".") == 0)
-	    || (strcmp (entry->d_name, "..") == 0)
-            || ((strcmp (local->loc.path, "/") == 0)
-                && (strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR) == 0))) {
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"skipping inspection of %s under %s",
-			entry->d_name, local->loc.path);
-		goto out;
-	}
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"inspecting existance of %s under %s",
-		entry->d_name, local->loc.path);
-
-	impunge_frame = copy_frame (frame);
-	if (!impunge_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-	ALLOC_OR_GOTO (impunge_local, afr_local_t, out);
 
-	impunge_frame->local = impunge_local;
-	impunge_sh = &impunge_local->self_heal;
-	impunge_sh->sh_frame = frame;
-	impunge_sh->active_source = active_src;
-
-        impunge_sh->impunging_entry_mode =
-                st_mode_from_ia (entry->d_stat.ia_prot, entry->d_stat.ia_type);
-
-	ret = build_child_loc (this, &impunge_local->loc, &local->loc, entry->d_name);
-	if (ret != 0) {
-		goto out;
-	}
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (i == active_src)
-			continue;
-		if (local->child_up[i] == 0)
-			continue;
-		if (sh->sources[i] == 1)
-			continue;
-		call_count++;
-	}
-
-	impunge_local->call_count = call_count;
+	gf_log (this->name, GF_LOG_INFO, "performing entry selfheal on %s",
+		uuid_utoa (fd->inode->gfid));
 
 	for (i = 0; i < priv->child_count; i++) {
-		if (i == active_src)
+		if (i != source && !healed_sinks[i])
 			continue;
-		if (local->child_up[i] == 0)
-			continue;
-		if (sh->sources[i] == 1)
-			continue;
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"looking up %s on %s", impunge_local->loc.path,
-			priv->children[i]->name);
-
-		STACK_WIND_COOKIE (impunge_frame,
-				   afr_sh_entry_impunge_entry_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->lookup,
-				   &impunge_local->loc, 0);
-
-		if (!--call_count)
+		ret = afr_selfheal_entry_do_subvol (frame, this, fd, i, source,
+						    sources, healed_sinks);
+		if (ret)
 			break;
 	}
-
-	ret = 0;
-out:
-	if (ret == -1)
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
-
-	return 0;
+	return ret;
 }
 
 
-int
-afr_sh_entry_impunge_readdir_cbk (call_frame_t *frame, void *cookie,
-				  xlator_t *this,
-				  int32_t op_ret, int32_t op_errno,
-				  gf_dirent_t *entries)
+static int
+__afr_selfheal_entry_finalize_source (xlator_t *this, unsigned char *sources,
+				      unsigned char *sinks,
+				      unsigned char *locked_on,
+				      struct afr_reply *replies)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	gf_dirent_t     *entry = NULL;
-	off_t            last_offset = 0;
-	int              active_src = 0;
-	int              entry_count = 0;
+	int i = 0;
+	afr_private_t *priv = NULL;
+	int source = -1;
+	int locked_count = 0;
+	int sources_count = 0;
+	int sinks_count = 0;
 
 	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	active_src = sh->active_source;
-
-	if (op_ret <= 0) {
-		if (op_ret < 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"readdir of %s on subvolume %s failed (%s)",
-				local->loc.path,
-				priv->children[active_src]->name,
-				strerror (op_errno));
-		} else {
-			gf_log (this->name, GF_LOG_TRACE,
-				"readdir of %s on subvolume %s complete",
-				local->loc.path,
-				priv->children[active_src]->name);
-		}
 
-		afr_sh_entry_impunge_all (frame, this);
-		return 0;
-	}
+	locked_count = AFR_COUNT (locked_on, priv->child_count);
+	sources_count = AFR_COUNT (sources, priv->child_count);
+	sinks_count = AFR_COUNT (sinks, priv->child_count);
 
-	list_for_each_entry (entry, &entries->list, list) {
-		last_offset = entry->d_off;
-                entry_count++;
+	if (locked_count == sinks_count || !sources_count) {
+		return -1;
 	}
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"readdir'ed %d entries from %s",
-		entry_count, priv->children[active_src]->name);
-
-	sh->offset = last_offset;
-	local->call_count = entry_count;
-
-	list_for_each_entry (entry, &entries->list, list) {
-                afr_sh_entry_impunge_entry (frame, this, entry);
+	for (i = 0; i < priv->child_count; i++) {
+		if (sources[i]) {
+			source = i;
+			break;
+		}
 	}
 
-	return 0;
+	return source;
 }
-				  
 
-int
-afr_sh_entry_impunge_subvol (call_frame_t *frame, xlator_t *this,
-			     int active_src)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	STACK_WIND (frame, afr_sh_entry_impunge_readdir_cbk,
-		    priv->children[active_src],
-		    priv->children[active_src]->fops->readdirp,
-		    sh->healing_fd, sh->block_size, sh->offset);
 
-	return 0;
-}
-
-
-int
-afr_sh_entry_impunge_all (call_frame_t *frame, xlator_t *this)
+static int
+__afr_selfheal_entry_prepare (call_frame_t *frame, xlator_t *this, fd_t *fd,
+			      unsigned char *locked_on, unsigned char *sources,
+			      unsigned char *sinks, unsigned char *healed_sinks,
+			      struct afr_reply *replies, int *source_p)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              active_src = -1;
+	int ret = -1;
+	int source = -1;
+	afr_private_t *priv = NULL;
+	int i = 0;
 
 	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	sh->offset = 0;
 
-	active_src = next_active_source (frame, this, sh->active_source);
-	sh->active_source = active_src;
+	ret = afr_selfheal_unlocked_discover (frame, fd->inode, fd->inode->gfid,
+					      replies);
+	if (ret)
+		return ret;
 
-	if (sh->op_failed) {
-		afr_sh_entry_finish (frame, this);
-		return 0;
-	}
+	ret = afr_selfheal_find_direction (frame, this, replies,
+					   AFR_ENTRY_TRANSACTION,
+					   locked_on, sources, sinks);
+	if (ret)
+		return ret;
 
-	if (active_src == -1) {
-		/* completed creating missing files on all subvolumes */
-		afr_sh_entry_expunge_all (frame, this);
-		return 0;
+	source = __afr_selfheal_entry_finalize_source (this, sources, sinks,
+						       locked_on, replies);
+	if (source < 0) {
+		/* If source is < 0 (typically split-brain), we perform a
+		   conservative merge of entries rather than erroring out */
 	}
+	*source_p = source;
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"impunging entries of %s on %s to other sinks",
-		local->loc.path, priv->children[active_src]->name);
+	for (i = 0; i < priv->child_count; i++)
+		/* Initialize the healed_sinks[] array optimistically to
+		   the intersection of to-be-healed (i.e sinks[]) and
+		   the list of servers which are up (i.e locked_on[]).
 
-	afr_sh_entry_impunge_subvol (frame, this, active_src);
+		   As we encounter failures in the healing process, we
+		   will unmark the respective servers in the healed_sinks[]
+		   array.
+		*/
+		healed_sinks[i] = sinks[i] && locked_on[i];
 
-	return 0;
+	return ret;
 }
 
 
-int
-afr_sh_entry_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			  int32_t op_ret, int32_t op_errno, fd_t *fd)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              call_count = 0;
-	int              child_index = 0;
-
-	local = frame->local;
-	sh = &local->self_heal;
+static int
+__afr_selfheal_entry (call_frame_t *frame, xlator_t *this, fd_t *fd,
+		      unsigned char *locked_on)
+{
+	afr_private_t *priv = NULL;
+	int ret = -1;
+	unsigned char *sources = NULL;
+	unsigned char *sinks = NULL;
+	unsigned char *data_lock = NULL;
+	unsigned char *healed_sinks = NULL;
+	struct afr_reply *locked_replies = NULL;
+	int source = -1;
+
 	priv = this->private;
 
-	child_index = (long) cookie;
+	sources = alloca0 (priv->child_count);
+	sinks = alloca0 (priv->child_count);
+	healed_sinks = alloca0 (priv->child_count);
+	data_lock = alloca0 (priv->child_count);
 
-	/* TODO: some of the open's might fail.
-	   In that case, modify cleanup fn to send flush on those 
-	   fd's which are already open */
+	locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
 
-	LOCK (&frame->lock);
+	ret = afr_selfheal_entrylk (frame, this, fd->inode, this->name, NULL,
+				    data_lock);
 	{
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"opendir of %s failed on child %s (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-			sh->op_failed = 1;
+		if (ret < 2) {
+			ret = -ENOTCONN;
+			goto unlock;
 		}
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-		if (sh->op_failed) {
-			afr_sh_entry_finish (frame, this);
-			return 0;
-		}
-		gf_log (this->name, GF_LOG_TRACE,
-			"fd for %s opened, commencing sync",
-			local->loc.path);
 
-		sh->active_source = -1;
-		afr_sh_entry_impunge_all (frame, this);
+		ret = __afr_selfheal_entry_prepare (frame, this, fd, data_lock,
+						    sources, sinks, healed_sinks,
+						    locked_replies, &source);
 	}
+unlock:
+	afr_selfheal_unentrylk (frame, this, fd->inode, this->name, NULL,
+				data_lock);
+	if (ret < 0)
+		goto out;
 
-	return 0;
+	ret = afr_selfheal_entry_do (frame, this, fd, source, sources,
+				     healed_sinks, locked_replies);
+	if (ret)
+		goto out;
+
+	ret = afr_selfheal_undo_pending (frame, this, fd->inode, sources, sinks,
+					 healed_sinks, AFR_ENTRY_TRANSACTION,
+					 locked_replies, data_lock);
+out:
+	return ret;
 }
 
 
-int
-afr_sh_entry_open (call_frame_t *frame, xlator_t *this)
+static fd_t *
+afr_selfheal_data_opendir (xlator_t *this, inode_t *inode)
 {
-	int i = 0;				
-	int call_count = 0;		     
-
-	int source = -1;
-	int *sources = NULL;
-
+	loc_t loc = {0,};
+	int ret = 0;
 	fd_t *fd = NULL;
 
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	afr_self_heal_t *sh = NULL;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	source  = local->self_heal.source;
-	sources = local->self_heal.sources;
+	fd = fd_create (inode, 0);
+	if (!fd)
+		return NULL;
 
-	sh->block_size = 65536; //131072
-	sh->offset = 0;
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-	call_count = sh->active_sinks;
-	if (source != -1)
-		call_count++;
-
-	local->call_count = call_count;
-
-	fd = fd_create (local->loc.inode, frame->root->pid);
-	sh->healing_fd = fd;
-
-	if (source != -1) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"opening directory %s on subvolume %s (source)",
-			local->loc.path, priv->children[source]->name);
-
-		/* open source */
-		STACK_WIND_COOKIE (frame, afr_sh_entry_opendir_cbk,
-				   (void *) (long) source,
-				   priv->children[source],
-				   priv->children[source]->fops->opendir,
-				   &local->loc, fd);
-		call_count--;
+	ret = syncop_opendir (this, &loc, fd);
+	if (ret) {
+		fd_unref (fd);
+		fd = NULL;
+	} else {
+		fd_bind (fd);
 	}
 
-	/* open sinks */
-	for (i = 0; i < priv->child_count; i++) {
-		if (sources[i] || !local->child_up[i])
-			continue;
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"opening directory %s on subvolume %s (sink)",
-			local->loc.path, priv->children[i]->name);
-
-		STACK_WIND_COOKIE (frame, afr_sh_entry_opendir_cbk,
-				   (void *) (long) i,
-				   priv->children[i], 
-				   priv->children[i]->fops->opendir,
-				   &local->loc, fd);
-
-		if (!--call_count)
-			break;
-	}
+	loc_wipe (&loc);
 
-	return 0;
+	return fd;
 }
 
 
-int
-afr_sh_entry_sync_prepare (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              active_sinks = 0;
-	int              source = 0;
-	int              i = 0;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	source = sh->source;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->sources[i] == 0 && local->child_up[i] == 1) {
-			active_sinks++;
-			sh->success[i] = 1;
-		}
-	}
-	if (source != -1)
-		sh->success[source] = 1;
-
-	if (active_sinks == 0) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"no active sinks for self-heal on dir %s",
-			local->loc.path);
-		afr_sh_entry_finish (frame, this);
-		return 0;
-	}
-	if (source == -1 && active_sinks < 2) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"cannot sync with 0 sources and 1 sink on dir %s",
-			local->loc.path);
-		afr_sh_entry_finish (frame, this);
-		return 0;
-	}
-	sh->active_sinks = active_sinks;
-
-	if (source != -1)
-		gf_log (this->name, GF_LOG_DEBUG,
-			"self-healing directory %s from subvolume %s to "
-                        "%d other",
-			local->loc.path, priv->children[source]->name,
-			active_sinks);
-	else
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no active sources for %s found. "
-			"merging all entries as a conservative decision",
-			local->loc.path);
-
-	afr_sh_entry_open (frame, this);
-
-	return 0;
-}
-
 
 int
-afr_sh_entry_fix (call_frame_t *frame, xlator_t *this)
+afr_selfheal_entry (call_frame_t *frame, xlator_t *this, inode_t *inode)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              source = 0;
-
-        int nsources = 0;
+	afr_private_t *priv = NULL;
+	unsigned char *locked_on = NULL;
+	fd_t *fd = NULL;
+	int ret = 0;
 
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-        if (sh->forced_merge) {
-                sh->source = -1;
-                goto heal;
-        }
-
-	afr_sh_build_pending_matrix (priv, sh->pending_matrix, sh->xattr, 
-				     priv->child_count, AFR_ENTRY_TRANSACTION);
-
-	afr_sh_print_pending_matrix (sh->pending_matrix, this);
-
-	nsources = afr_sh_mark_sources (sh, priv->child_count,
-                                        AFR_SELF_HEAL_ENTRY);
-
-        if (nsources == 0) {
-                gf_log (this->name, GF_LOG_TRACE,
-                        "No self-heal needed for %s",
-                        local->loc.path);
-
-                afr_sh_entry_finish (frame, this);
-                return 0;
-        }
+	fd = afr_selfheal_data_opendir (this, inode);
+	if (!fd)
+		return -EIO;
 
-	afr_sh_supress_errenous_children (sh->sources, sh->child_errno,
-					  priv->child_count);
+	locked_on = alloca0 (priv->child_count);
 
-	source = afr_sh_select_source (sh->sources, priv->child_count);
-
-        sh->source = source;
-
-heal:
-	afr_sh_entry_sync_prepare (frame, this);
-
-	return 0;
-}
-
-
-
-int
-afr_sh_entry_lookup_cbk (call_frame_t *frame, void *cookie,
-			 xlator_t *this, int32_t op_ret, int32_t op_errno,
-                         inode_t *inode, struct iatt *buf, dict_t *xattr,
-                         struct iatt *postparent)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-
-	int call_count  = -1;
-	int child_index = (long) cookie;
-
-	local = frame->local;
-	sh = &local->self_heal;
-
-	LOCK (&frame->lock);
+	ret = afr_selfheal_tryentrylk (frame, this, inode, priv->sh_domain, NULL,
+				       locked_on);
 	{
-		if (op_ret != -1) {
-			sh->xattr[child_index] = dict_ref (xattr);
-			sh->buf[child_index] = *buf;
+		if (ret < 2) {
+			/* Either less than two subvols available, or another
+			   selfheal (from another server) is in progress. Skip
+			   for now in any case there isn't anything to do.
+			*/
+			ret = -ENOTCONN;
+			goto unlock;
 		}
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
 
-	if (call_count == 0) {
-		afr_sh_entry_fix (frame, this);
+		ret = __afr_selfheal_entry (frame, this, fd, locked_on);
 	}
+unlock:
+	afr_selfheal_unentrylk (frame, this, inode, priv->sh_domain, NULL, locked_on);
 
-	return 0;
-}
-
-
-
-int
-afr_sh_entry_lookup (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t    *  local = NULL;
-	afr_private_t  *  priv  = NULL;
-	dict_t         *xattr_req = NULL;
-	int ret = 0;
-	int call_count = 0;
-	int i = 0;
+	if (fd)
+		fd_unref (fd);
 
-	priv  = this->private;
-	local = frame->local;
-
-	call_count = afr_up_children_count (priv->child_count,
-                                            local->child_up);
-
-	local->call_count = call_count;
-
-	xattr_req = dict_new();
-	if (xattr_req) {
-                for (i = 0; i < priv->child_count; i++) {
-                        ret = dict_set_uint64 (xattr_req,
-                                               priv->pending_key[i],
-                                               3 * sizeof(int32_t));
-                        if (ret < 0)
-                                gf_log (this->name, GF_LOG_WARNING,
-                                        "Unable to set dict value.");
-                }
-        }
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame,
-					   afr_sh_entry_lookup_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->lookup,
-					   &local->loc, xattr_req);
-			if (!--call_count)
-				break;
-		}
-	}
-
-	if (xattr_req)
-		dict_unref (xattr_req);
-
-	return 0;
-}
-
-int
-afr_sh_post_nonblocking_entry_cbk (call_frame_t *frame, xlator_t *this)
-{
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-
-        if (int_lock->lock_op_ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Non Blocking entrylks failed.");
-                afr_sh_entry_done (frame, this);
-        } else {
-
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Non Blocking entrylks done. Proceeding to FOP");
-                afr_sh_entry_lookup(frame, this);
-        }
-
-        return 0;
-}
-
-int
-afr_sh_entry_lock (call_frame_t *frame, xlator_t *this)
-{
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-
-        int_lock->transaction_lk_type = AFR_SELFHEAL_LK;
-        int_lock->selfheal_lk_type    = AFR_ENTRY_SELF_HEAL_LK;
-
-        afr_set_lock_number (frame, this);
-
-        int_lock->lk_basename = NULL;
-        int_lock->lk_loc      = &local->loc;
-        int_lock->lock_cbk    = afr_sh_post_nonblocking_entry_cbk;
-
-        afr_nonblocking_entrylk (frame, this);
-
-
-	return 0;
-}
-
-
-int
-afr_self_heal_entry (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t   *local = NULL;
-	afr_private_t   *priv = NULL;
-
-
-	priv = this->private;
-	local = frame->local;
-
-	if (local->self_heal.need_entry_self_heal && priv->entry_self_heal) {
-		afr_sh_entry_lock (frame, this);
-	} else {
-		gf_log (this->name, GF_LOG_TRACE,
-			"proceeding to completion on %s",
-			local->loc.path);
-		afr_sh_entry_done (frame, this);
-	}
-
-	return 0;
+	return ret;
 }
-
diff --git a/xlators/cluster/afr/src/afr-self-heal-metadata.c b/xlators/cluster/afr/src/afr-self-heal-metadata.c
index e76d58850..b31a33237 100644
--- a/xlators/cluster/afr/src/afr-self-heal-metadata.c
+++ b/xlators/cluster/afr/src/afr-self-heal-metadata.c
@@ -1,731 +1,281 @@
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 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 <libgen.h>
-#include <unistd.h>
-#include <fnmatch.h>
-#include <sys/time.h>
-#include <stdlib.h>
-#include <signal.h>
 
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
 #endif
 
-#include "glusterfs.h"
 #include "afr.h"
-#include "dict.h"
-#include "xlator.h"
-#include "hashfn.h"
-#include "logging.h"
-#include "stack.h"
-#include "list.h"
-#include "call-stub.h"
-#include "defaults.h"
-#include "common-utils.h"
-#include "compat-errno.h"
-#include "compat.h"
+#include "afr-self-heal.h"
 #include "byte-order.h"
 
-#include "afr-transaction.h"
-#include "afr-self-heal.h"
-#include "afr-self-heal-common.h"
 
+#define AFR_HEAL_ATTR (GF_SET_ATTR_UID|GF_SET_ATTR_GID|GF_SET_ATTR_MODE)
 
 int
-afr_sh_metadata_done (call_frame_t *frame, xlator_t *this)
+afr_selfheal_metadata_do (call_frame_t *frame, xlator_t *this, inode_t *inode,
+			  int source, unsigned char *healed_sinks,
+			  struct afr_reply *locked_replies)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              i = 0;
+	int ret = -1;
+	loc_t loc = {0,};
+	dict_t *xattr = NULL;
+	dict_t *old_xattr = NULL;
+	afr_private_t *priv = NULL;
+	int i = 0;
 
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-//	memset (sh->child_errno, 0, sizeof (int) * priv->child_count);
-	memset (sh->buf, 0, sizeof (struct iatt) * priv->child_count);
-	memset (sh->success, 0, sizeof (int) * priv->child_count);
-
-/*         for (i = 0; i < priv->child_count; i++) { */
-/*                 sh->locked_nodes[i] = 1; */
-/*         } */
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->xattr[i])
-			dict_unref (sh->xattr[i]);
-		sh->xattr[i] = NULL;
-	}
-
-	if (local->govinda_gOvinda) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"aborting selfheal of %s",
-			local->loc.path);
-		sh->completion_cbk (frame, this);
-	} else {
-		if (IA_ISREG (sh->type)) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"proceeding to data check on %s",
-				local->loc.path);
-			afr_self_heal_data (frame, this);
-			return 0;
-		}
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-		if (IA_ISDIR (sh->type)) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"proceeding to entry check on %s",
-				local->loc.path);
-			afr_self_heal_entry (frame, this);
-			return 0;
-		}
-		gf_log (this->name, GF_LOG_DEBUG,
-			"completed self heal of %s",
-			local->loc.path);
+	gf_log (this->name, GF_LOG_INFO, "performing metadata selfheal on %s",
+		uuid_utoa (inode->gfid));
 
-		sh->completion_cbk (frame, this);
+	ret = syncop_getxattr (priv->children[source], &loc, &xattr, NULL);
+	if (ret < 0) {
+		loc_wipe (&loc);
+		return -EIO;
 	}
 
-	return 0;
-}
-
-
-int
-afr_sh_metadata_unlck_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			   int32_t op_ret, int32_t op_errno)
-{
-	int               call_count = 0;
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		afr_sh_metadata_done (frame, this);
-
-	return 0;
-}
-
-int
-afr_sh_inode_unlock (call_frame_t *frame, xlator_t *this)
-{
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-
-        int_lock->lock_cbk = afr_sh_metadata_done;
-        afr_unlock (frame, this);
-
-        return 0;
-}
-
-int
-afr_sh_metadata_finish (call_frame_t *frame, xlator_t *this)
-{
-        afr_sh_inode_unlock (frame, this);
-
-	return 0;
-}
+	afr_filter_xattrs (xattr);
+	dict_del (xattr, GF_SELINUX_XATTR_KEY);
 
+	for (i = 0; i < priv->child_count; i++) {
+		if (!healed_sinks[i])
+			continue;
 
-int
-afr_sh_metadata_erase_pending_cbk (call_frame_t *frame, void *cookie,
-				   xlator_t *this, int32_t op_ret,
-				   int32_t op_errno, dict_t *xattr)
-{
-	afr_local_t     *local = NULL;
-	int             call_count = 0;
-
-	local = frame->local;
+		ret = syncop_setattr (priv->children[i], &loc,
+				      &locked_replies[source].poststat,
+				      AFR_HEAL_ATTR, NULL, NULL);
+		if (ret)
+			healed_sinks[i] = 0;
+
+		old_xattr = NULL;
+		ret = syncop_getxattr (priv->children[i], &loc, &old_xattr, 0);
+		if (old_xattr) {
+			dict_del (old_xattr, GF_SELINUX_XATTR_KEY);
+			afr_filter_xattrs (old_xattr);
+			ret = syncop_removexattr (priv->children[i], &loc, "",
+						  old_xattr);
+		}
 
-	LOCK (&frame->lock);
-	{
+		ret = syncop_setxattr (priv->children[i], &loc, xattr, 0);
+		if (ret)
+			healed_sinks[i] = 0;
 	}
-	UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		afr_sh_metadata_finish (frame, this);
+	loc_wipe (&loc);
+	if (xattr)
+		dict_unref (xattr);
 
 	return 0;
 }
 
 
-int
-afr_sh_metadata_erase_pending (call_frame_t *frame, xlator_t *this)
+/*
+ * Look for mismatching uid/gid or mode even if xattrs don't say so, and
+ * pick one arbitrarily as winner.
+ */
+
+static int
+__afr_selfheal_metadata_finalize_source (xlator_t *this, unsigned char *sources,
+					 unsigned char *sinks,
+					 unsigned char *locked_on,
+					 struct afr_reply *replies)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              call_count = 0;
-	int              i = 0;
-	dict_t          **erase_xattr = NULL;
+	int i = 0;
+	afr_private_t *priv = NULL;
+	struct iatt first = {0, };
+	int source = -1;
+	int locked_count = 0;
+	int sources_count = 0;
+	int sinks_count = 0;
 
-
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-	afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix, 
-                                 sh->success, priv->child_count,
-                                 AFR_METADATA_TRANSACTION);
-
-	erase_xattr = GF_CALLOC (sizeof (*erase_xattr), priv->child_count,
-                                 gf_afr_mt_dict_t);
+	locked_count = AFR_COUNT (locked_on, priv->child_count);
+	sources_count = AFR_COUNT (sources, priv->child_count);
+	sinks_count = AFR_COUNT (sinks, priv->child_count);
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->xattr[i]) {
-			call_count++;
-
-			erase_xattr[i] = get_new_dict();
-			dict_ref (erase_xattr[i]);
+	if (locked_count == sinks_count || !sources_count) {
+		if (!priv->metadata_splitbrain_forced_heal) {
+			return -EIO;
+		}
+		/* Metadata split brain, select one subvol
+		   arbitrarily */
+		for (i = 0; i < priv->child_count; i++) {
+			if (locked_on[i] && sinks[i]) {
+				sources[i] = 1;
+				sinks[i] = 0;
+				break;
+			}
 		}
-	}
-
-	afr_sh_delta_to_xattr (priv, sh->delta_matrix, erase_xattr,
-			       priv->child_count, AFR_METADATA_TRANSACTION);
-
-	local->call_count = call_count;
-
-	if (call_count == 0) {
-		gf_log (this->name, GF_LOG_WARNING,
-			"metadata of %s not healed on any subvolume",
-			local->loc.path);
-
-		afr_sh_metadata_finish (frame, this);
 	}
 
 	for (i = 0; i < priv->child_count; i++) {
-		if (!erase_xattr[i])
+		if (!sources[i])
 			continue;
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"erasing pending flags from %s on %s",
-			local->loc.path, priv->children[i]->name);
-
-		STACK_WIND_COOKIE (frame, afr_sh_metadata_erase_pending_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->xattrop,
-				   &local->loc,
-				   GF_XATTROP_ADD_ARRAY, erase_xattr[i]);
-		if (!--call_count)
-			break;
-	}
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (erase_xattr[i]) {
-			dict_unref (erase_xattr[i]);
+		if (source == -1) {
+			source = i;
+			first = replies[i].poststat;
 		}
 	}
-	GF_FREE (erase_xattr);
-
-	return 0;
-}
-
-
-int
-afr_sh_metadata_sync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			  int32_t op_ret, int32_t op_errno)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              call_count = 0;
-	int              child_index = 0;
-
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	child_index = (long) cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"setting attributes failed for %s on %s (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-
-			sh->success[child_index] = 0;
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		afr_sh_metadata_erase_pending (frame, this);
-
-	return 0;
-}
-
-
-int
-afr_sh_metadata_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                             int32_t op_ret, int32_t op_errno,
-                             struct iatt *preop, struct iatt *postop)
-{
-	afr_sh_metadata_sync_cbk (frame, cookie, this, op_ret, op_errno);
-
-	return 0;
-}
-
-
-int
-afr_sh_metadata_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			   int32_t op_ret, int32_t op_errno)
-{
-	afr_sh_metadata_sync_cbk (frame, cookie, this, op_ret, op_errno);
-
-	return 0;
-}
-
-
-int
-afr_sh_metadata_sync (call_frame_t *frame, xlator_t *this, dict_t *xattr)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              source = 0;
-	int              active_sinks = 0;
-	int              call_count = 0;
-	int              i = 0;
-
-        struct iatt      stbuf;
-        int32_t          valid = 0;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	source = sh->source;
-	active_sinks = sh->active_sinks;
-
-	/*
-	 * 2 calls per sink - setattr, setxattr
-	 */
-	if (xattr)
-		call_count = active_sinks * 2;
-	else
-		call_count = active_sinks;
-
-	local->call_count = call_count;
-
-	stbuf.ia_atime = sh->buf[source].ia_atime;
-	stbuf.ia_atime_nsec = sh->buf[source].ia_atime_nsec;
-	stbuf.ia_mtime = sh->buf[source].ia_mtime;
-	stbuf.ia_mtime_nsec = sh->buf[source].ia_mtime_nsec;
-
-        stbuf.ia_uid = sh->buf[source].ia_uid;
-        stbuf.ia_gid = sh->buf[source].ia_gid;
-
-        stbuf.ia_type = sh->buf[source].ia_type;
-        stbuf.ia_prot = sh->buf[source].ia_prot;
-        
-        valid = GF_SET_ATTR_MODE  | 
-                GF_SET_ATTR_UID   | GF_SET_ATTR_GID |
-                GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
 
 	for (i = 0; i < priv->child_count; i++) {
-		if (call_count == 0) {
-			break;
-		}
-		if (sh->sources[i] || !local->child_up[i])
+		if (!sources[i])
 			continue;
-
-		gf_log (this->name, GF_LOG_DEBUG,
-			"self-healing metadata of %s from %s to %s",
-			local->loc.path, priv->children[source]->name,
-			priv->children[i]->name);
-
-		STACK_WIND_COOKIE (frame, afr_sh_metadata_setattr_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->setattr,
-				   &local->loc, &stbuf, valid);
-
-		call_count--;
-
-		if (!xattr)
-			continue;
-
-		STACK_WIND_COOKIE (frame, afr_sh_metadata_xattr_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->setxattr,
-				   &local->loc, xattr, 0);
-		call_count--;
+		if (!IA_EQUAL (first, replies[i].poststat, type) ||
+		    !IA_EQUAL (first, replies[i].poststat, uid) ||
+		    !IA_EQUAL (first, replies[i].poststat, gid) ||
+		    !IA_EQUAL (first, replies[i].poststat, prot)) {
+			sources[i] = 0;
+			sinks[i] = 1;
+		}
 	}
 
-	return 0;
+	return source;
 }
 
 
-int
-afr_sh_metadata_getxattr_cbk (call_frame_t *frame, void *cookie,
-			      xlator_t *this,
-			      int32_t op_ret, int32_t op_errno, dict_t *xattr)
+static int
+__afr_selfheal_metadata_prepare (call_frame_t *frame, xlator_t *this, inode_t *inode,
+				 unsigned char *locked_on, unsigned char *sources,
+				 unsigned char *sinks, unsigned char *healed_sinks,
+				 struct afr_reply *replies)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              source = 0;
+	int ret = -1;
+	int source = -1;
+	afr_private_t *priv = NULL;
+	int i = 0;
 
-        int i;
-
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-	source = sh->source;
-
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"getxattr of %s failed on subvolume %s (%s). proceeding without xattr",
-			local->loc.path, priv->children[source]->name,
-			strerror (op_errno));
-
-		afr_sh_metadata_sync (frame, this, NULL);
-	} else {
-                for (i = 0; i < priv->child_count; i++) {
-                        dict_del (xattr, priv->pending_key[i]);
-                }
-
-		afr_sh_metadata_sync (frame, this, xattr);
-	}
-
-	return 0;
+	ret = afr_selfheal_unlocked_discover (frame, inode, inode->gfid,
+					      replies);
+	if (ret)
+		return ret;
+
+	ret = afr_selfheal_find_direction (frame, this, replies,
+					   AFR_METADATA_TRANSACTION,
+					   locked_on, sources, sinks);
+	if (ret)
+		return ret;
+
+	source = __afr_selfheal_metadata_finalize_source (this, sources, sinks,
+							  locked_on, replies);
+	if (source < 0)
+		return -EIO;
+
+	for (i = 0; i < priv->child_count; i++)
+		/* Initialize the healed_sinks[] array optimistically to
+		   the intersection of to-be-healed (i.e sinks[]) and
+		   the list of servers which are up (i.e locked_on[]).
+
+		   As we encounter failures in the healing process, we
+		   will unmark the respective servers in the healed_sinks[]
+		   array.
+		*/
+		healed_sinks[i] = sinks[i] && locked_on[i];
+
+	return source;
 }
 
 
-int
-afr_sh_metadata_sync_prepare (call_frame_t *frame, xlator_t *this)
+static int
+__afr_selfheal_metadata (call_frame_t *frame, xlator_t *this, inode_t *inode,
+			 unsigned char *locked_on)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              active_sinks = 0;
-	int              source = 0;
-	int              i = 0;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+	afr_private_t *priv = NULL;
+	int ret = -1;
+	unsigned char *sources = NULL;
+	unsigned char *sinks = NULL;
+	unsigned char *data_lock = NULL;
+	unsigned char *healed_sinks = NULL;
+	struct afr_reply *locked_replies = NULL;
+	int source = -1;
 
-	source = sh->source;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->sources[i] == 0 && local->child_up[i] == 1) {
-			active_sinks++;
-			sh->success[i] = 1;
-		}
-	}
-	sh->success[source] = 1;
-
-	if (active_sinks == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no active sinks for performing self-heal on file %s",
-			local->loc.path);
-		afr_sh_metadata_finish (frame, this);
-		return 0;
-	}
-	sh->active_sinks = active_sinks;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"syncing metadata of %s from subvolume %s to %d active sinks",
-		local->loc.path, priv->children[source]->name, active_sinks);
-
-	STACK_WIND (frame, afr_sh_metadata_getxattr_cbk,
-		    priv->children[source],
-		    priv->children[source]->fops->getxattr,
-		    &local->loc, NULL);
-
-	return 0;
-}
-
-
-int
-afr_sh_metadata_fix (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              nsources = 0;
-	int              source = 0;
-	int              i = 0;
-
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-	afr_sh_build_pending_matrix (priv, sh->pending_matrix, sh->xattr, 
-				     priv->child_count, 
-                                     AFR_METADATA_TRANSACTION);
-
-	afr_sh_print_pending_matrix (sh->pending_matrix, this);
-
-	nsources = afr_sh_mark_sources (sh, priv->child_count,
-                                        AFR_SELF_HEAL_METADATA);
-
-	afr_sh_supress_errenous_children (sh->sources, sh->child_errno,
-					  priv->child_count);
-
-        if (nsources == 0) {
-                gf_log (this->name, GF_LOG_TRACE,
-                        "No self-heal needed for %s",
-                        local->loc.path);
-
-                afr_sh_metadata_finish (frame, this);
-                return 0;
-        }
-
-	if ((nsources == -1)
-	    && (priv->favorite_child != -1)
-	    && (sh->child_errno[priv->favorite_child] == 0)) {
-
-		gf_log (this->name, GF_LOG_WARNING,
-			"Picking favorite child %s as authentic source to resolve conflicting metadata of %s",
-			priv->children[priv->favorite_child]->name,
-			local->loc.path);
-
-		sh->sources[priv->favorite_child] = 1;
+	sources = alloca0 (priv->child_count);
+	sinks = alloca0 (priv->child_count);
+	healed_sinks = alloca0 (priv->child_count);
+	data_lock = alloca0 (priv->child_count);
 
-		nsources = afr_sh_source_count (sh->sources,
-						priv->child_count);
-	}
-
-	if (nsources == -1) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Unable to self-heal permissions/ownership of '%s' "
-                        "(possible split-brain). Please fix the file on "
-                        "all backend volumes", local->loc.path);
-
-		local->govinda_gOvinda = 1;
-
-		afr_sh_metadata_finish (frame, this);
-		return 0;
-	}
-
-	source = afr_sh_select_source (sh->sources, priv->child_count);
-
-        if (source == -1) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "No active sources found.");
+	locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
 
-                afr_sh_metadata_finish (frame, this);
-                return 0;
-        }
-
-	sh->source = source;
-
-	/* detect changes not visible through pending flags -- JIC */
-	for (i = 0; i < priv->child_count; i++) {
-		if (i == source || sh->child_errno[i])
-			continue;
+	ret = afr_selfheal_inodelk (frame, this, inode, this->name,
+				    LLONG_MAX - 1, 0, data_lock);
+	{
+		if (ret < 2) {
+			ret = -ENOTCONN;
+			goto unlock;
+		}
 
-		if (PERMISSION_DIFFERS (&sh->buf[i], &sh->buf[source]))
-			sh->sources[i] = 0;
+		ret = __afr_selfheal_metadata_prepare (frame, this, inode, data_lock,
+						       sources, sinks, healed_sinks,
+						       locked_replies);
+		if (ret < 0)
+			goto unlock;
 
-		if (OWNERSHIP_DIFFERS (&sh->buf[i], &sh->buf[source]))
-			sh->sources[i] = 0;
+		source = ret;
+		ret = 0;
 	}
-
-	afr_sh_metadata_sync_prepare (frame, this);
-
-	return 0;
+unlock:
+	afr_selfheal_uninodelk (frame, this, inode, this->name,
+				LLONG_MAX -1, 0, data_lock);
+	if (ret < 0)
+		goto out;
+
+	ret = afr_selfheal_metadata_do (frame, this, inode, source, healed_sinks,
+					locked_replies);
+	if (ret)
+		goto out;
+
+	ret = afr_selfheal_undo_pending (frame, this, inode, sources, sinks,
+					 healed_sinks, AFR_METADATA_TRANSACTION,
+					 locked_replies, data_lock);
+out:
+	return ret;
 }
 
 
 int
-afr_sh_metadata_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			    int32_t op_ret, int32_t op_errno,
-                            inode_t *inode, struct iatt *buf, dict_t *xattr,
-                            struct iatt *postparent)
+afr_selfheal_metadata (call_frame_t *frame, xlator_t *this, inode_t *inode)
 {
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              call_count = 0;
-	int              child_index = 0;
-
+	afr_private_t *priv = NULL;
+	unsigned char *locked_on = NULL;
+	int ret = 0;
 
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-	child_index = (long) cookie;
+	locked_on = alloca0 (priv->child_count);
 
-	LOCK (&frame->lock);
+	ret = afr_selfheal_tryinodelk (frame, this, inode, priv->sh_domain, 0, 0,
+				       locked_on);
 	{
-		if (op_ret == 0) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"path %s on subvolume %s is of mode 0%o",
-				local->loc.path,
-				priv->children[child_index]->name,
-				buf->ia_type);
-
-			sh->buf[child_index] = *buf;
-			if (xattr)
-				sh->xattr[child_index] = dict_ref (xattr);
-		} else {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"path %s on subvolume %s => -1 (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-
-			sh->child_errno[child_index] = op_errno;
+		if (ret < 2) {
+			/* Either less than two subvols available, or another
+			   selfheal (from another server) is in progress. Skip
+			   for now in any case there isn't anything to do.
+			*/
+			ret = -ENOTCONN;
+			goto unlock;
 		}
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		afr_sh_metadata_fix (frame, this);
 
-	return 0;
-}
-
-
-int
-afr_sh_metadata_lookup (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_private_t   *priv = NULL;
-	int              i = 0;
-	int              call_count = 0;
-	dict_t          *xattr_req = NULL;
-	int              ret = 0;
-
-	local = frame->local;
-	priv = this->private;
-
-	call_count = afr_up_children_count (priv->child_count,
-                                            local->child_up);
-	local->call_count = call_count;
-
-	xattr_req = dict_new();
-
-	if (xattr_req) {
-                for (i = 0; i < priv->child_count; i++) {
-                        ret = dict_set_uint64 (xattr_req,
-                                               priv->pending_key[i],
-                                               3 * sizeof(int32_t));
-                        if (ret < 0)
-                                gf_log (this->name, GF_LOG_WARNING,
-                                        "Unable to set dict value.");
-                }
-        }
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"looking up %s on %s",
-				local->loc.path, priv->children[i]->name);
-
-			STACK_WIND_COOKIE (frame, afr_sh_metadata_lookup_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->lookup,
-					   &local->loc, xattr_req);
-			if (!--call_count)
-				break;
-		}
+		ret = __afr_selfheal_metadata (frame, this, inode, locked_on);
 	}
-	
-	if (xattr_req)
-		dict_unref (xattr_req);
+unlock:
+	afr_selfheal_uninodelk (frame, this, inode, priv->sh_domain, 0, 0, locked_on);
 
-	return 0;
+	return ret;
 }
-
-int
-afr_sh_post_nonblocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
-{
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-
-        if (int_lock->lock_op_ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Non Blocking inodelks failed.");
-                afr_sh_metadata_done (frame, this);
-        } else {
-
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Non Blocking inodelks done. Proceeding to FOP");
-                afr_sh_metadata_lookup (frame, this);
-        }
-
-        return 0;
-}
-
-int
-afr_sh_metadata_lock (call_frame_t *frame, xlator_t *this)
-{
-        afr_internal_lock_t *int_lock = NULL;
-        afr_local_t         *local    = NULL;
-
-        local    = frame->local;
-        int_lock = &local->internal_lock;
-
-        int_lock->transaction_lk_type = AFR_SELFHEAL_LK;
-        int_lock->selfheal_lk_type    = AFR_METADATA_SELF_HEAL_LK;
-
-        afr_set_lock_number (frame, this);
-
-        int_lock->lk_flock.l_start = 0;
-        int_lock->lk_flock.l_len   = 0;
-        int_lock->lk_flock.l_type  = F_WRLCK;
-        int_lock->lock_cbk         = afr_sh_post_nonblocking_inodelk_cbk;
-
-        afr_nonblocking_inodelk (frame, this);
-
-	return 0;
-}
-
-
-int
-afr_self_heal_metadata (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t   *local = NULL;
-	afr_private_t *priv = this->private;
-
-
-	local = frame->local;
-
-	if (local->self_heal.need_metadata_self_heal && priv->metadata_self_heal) {
-		afr_sh_metadata_lock (frame, this);
-	} else {
-		afr_sh_metadata_done (frame, this);
-	}
-
-	return 0;
-}
-
diff --git a/xlators/cluster/afr/src/afr-self-heal-name.c b/xlators/cluster/afr/src/afr-self-heal-name.c
new file mode 100644
index 000000000..ce80b8da3
--- /dev/null
+++ b/xlators/cluster/afr/src/afr-self-heal-name.c
@@ -0,0 +1,457 @@
+/*
+  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 "afr.h"
+#include "afr-self-heal.h"
+
+
+int
+__afr_selfheal_assign_gfid (call_frame_t *frame, xlator_t *this, inode_t *parent,
+			    uuid_t pargfid, const char *bname, inode_t *inode,
+			    struct afr_reply *replies, int gfid_idx)
+{
+	int i = 0;
+	afr_private_t *priv = NULL;
+	dict_t *xdata = NULL;
+	int ret = 0;
+	loc_t loc = {0, };
+
+	priv = this->private;
+
+	uuid_copy (parent->gfid, pargfid);
+
+	xdata = dict_new ();
+	if (!xdata) {
+		return -ENOMEM;
+	}
+
+	ret = dict_set_static_bin (xdata, "gfid-req",
+				   replies[gfid_idx].poststat.ia_gfid, 16);
+	if (ret) {
+		dict_destroy (xdata);
+		return -ENOMEM;
+	}
+
+	loc.parent = inode_ref (parent);
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.pargfid, pargfid);
+	loc.name = bname;
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (replies[i].op_ret == 0 || replies[i].op_errno != ENODATA)
+			continue;
+
+		ret = syncop_lookup (priv->children[i], &loc, xdata, 0, 0, 0);
+	}
+
+	loc_wipe (&loc);
+	dict_unref (xdata);
+
+	return ret;
+}
+
+
+int
+__afr_selfheal_name_impunge (call_frame_t *frame, xlator_t *this, inode_t *parent,
+			     uuid_t pargfid, const char *bname, inode_t *inode,
+			     struct afr_reply *replies, int gfid_idx)
+{
+	int i = 0;
+	afr_private_t *priv = NULL;
+	int ret = 0;
+
+	priv = this->private;
+
+	uuid_copy (parent->gfid, pargfid);
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].valid)
+			continue;
+
+		if (uuid_compare (replies[i].poststat.ia_gfid,
+				  replies[gfid_idx].poststat.ia_gfid) == 0)
+			continue;
+
+		ret |= afr_selfheal_recreate_entry (frame, this, i, gfid_idx,
+						    parent, bname, inode, replies);
+	}
+
+	return ret;
+}
+
+
+int
+__afr_selfheal_name_expunge (call_frame_t *frame, xlator_t *this, inode_t *parent,
+			     uuid_t pargfid, const char *bname, inode_t *inode,
+			     struct afr_reply *replies)
+{
+	loc_t loc = {0, };
+	int i = 0;
+	afr_private_t *priv = NULL;
+	char g[64];
+	int ret = 0;
+
+	priv = this->private;
+
+	loc.parent = inode_ref (parent);
+	uuid_copy (loc.pargfid, pargfid);
+	loc.name = bname;
+	loc.inode = inode_ref (inode);
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].valid)
+			continue;
+
+		if (replies[i].op_ret)
+			continue;
+
+		switch (replies[i].poststat.ia_type) {
+		case IA_IFDIR:
+			gf_log (this->name, GF_LOG_WARNING,
+				"expunging dir %s/%s (%s) on %s",
+				uuid_utoa (pargfid), bname,
+				uuid_utoa_r (replies[i].poststat.ia_gfid, g),
+				priv->children[i]->name);
+			ret |= syncop_rmdir (priv->children[i], &loc, 1);
+			break;
+		default:
+			gf_log (this->name, GF_LOG_WARNING,
+				"expunging file %s/%s (%s) on %s",
+				uuid_utoa (pargfid), bname,
+				uuid_utoa_r (replies[i].poststat.ia_gfid, g),
+				priv->children[i]->name);
+			ret |= syncop_unlink (priv->children[i], &loc);
+			break;
+		}
+	}
+
+	loc_wipe (&loc);
+
+	return ret;
+
+}
+
+
+int
+__afr_selfheal_name_do (call_frame_t *frame, xlator_t *this, inode_t *parent,
+			uuid_t pargfid, const char *bname, inode_t *inode,
+			unsigned char *sources, unsigned char *sinks,
+			unsigned char *healed_sinks, int source,
+			unsigned char *locked_on, struct afr_reply *replies)
+{
+	int i = 0;
+	afr_private_t *priv = NULL;
+	uuid_t gfid = {0, };
+	int gfid_idx = -1;
+	gf_boolean_t source_is_empty = _gf_true;
+	gf_boolean_t need_heal = _gf_false;
+	int first_idx = -1;
+	char g1[64],g2[64];
+
+	priv = this->private;
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].valid)
+			continue;
+
+		if (first_idx == -1) {
+			first_idx = i;
+			continue;
+		}
+
+		if (replies[i].op_ret != replies[first_idx].op_ret)
+			need_heal = _gf_true;
+
+		if (uuid_compare (replies[i].poststat.ia_gfid,
+				  replies[first_idx].poststat.ia_gfid))
+			need_heal = _gf_true;
+	}
+
+	if (!need_heal)
+		return 0;
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].valid)
+			continue;
+
+		if (!replies[i].op_ret && (source == -1 || sources[i])) {
+			source_is_empty = _gf_false;
+			break;
+		}
+	}
+
+	if (source_is_empty) {
+		return __afr_selfheal_name_expunge (frame, this, parent, pargfid,
+						    bname, inode, replies);
+	}
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].valid)
+			continue;
+
+		if (uuid_is_null (replies[i].poststat.ia_gfid))
+			continue;
+
+		if (uuid_is_null (gfid)) {
+			uuid_copy (gfid, replies[i].poststat.ia_gfid);
+			gfid_idx = i;
+			continue;
+		}
+
+		if (sources[i] || source == -1) {
+			if (gfid_idx != -1 &&
+			    (sources[gfid_idx] || source == -1) &&
+			    uuid_compare (gfid, replies[i].poststat.ia_gfid)) {
+				gf_log (this->name, GF_LOG_WARNING,
+					"GFID mismatch for <gfid:%s>/%s "
+					"%s on %s and %s on %s",
+					uuid_utoa (pargfid), bname,
+					uuid_utoa_r (replies[i].poststat.ia_gfid, g1),
+					priv->children[i]->name,
+					uuid_utoa_r (replies[gfid_idx].poststat.ia_gfid, g2),
+					priv->children[gfid_idx]->name);
+				return -1;
+			}
+
+			uuid_copy (gfid, replies[i].poststat.ia_gfid);
+			gfid_idx = i;
+			continue;
+		}
+	}
+
+	if (gfid_idx == -1)
+		return -1;
+
+	__afr_selfheal_assign_gfid (frame, this, parent, pargfid, bname, inode,
+				    replies, gfid_idx);
+
+	return __afr_selfheal_name_impunge (frame, this, parent, pargfid,
+					    bname, inode, replies, gfid_idx);
+}
+
+
+int
+__afr_selfheal_name_finalize_source (xlator_t *this, unsigned char *sources,
+				     unsigned char *sinks, unsigned char *locked_on,
+				     struct afr_reply *replies)
+{
+	int i = 0;
+	afr_private_t *priv = NULL;
+	int source = -1;
+	int locked_count = 0;
+	int sources_count = 0;
+	int sinks_count = 0;
+
+	priv = this->private;
+
+	locked_count = AFR_COUNT (locked_on, priv->child_count);
+	sources_count = AFR_COUNT (sources, priv->child_count);
+	sinks_count = AFR_COUNT (sinks, priv->child_count);
+
+	if (locked_count == sinks_count || !sources_count) {
+		return -1;
+	}
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (sources[i]) {
+			source = i;
+			break;
+		}
+	}
+
+	return source;
+}
+
+
+int
+__afr_selfheal_name_prepare (call_frame_t *frame, xlator_t *this, inode_t *parent,
+			     uuid_t pargfid, unsigned char *locked_on,
+			     unsigned char *sources, unsigned char *sinks,
+			     unsigned char *healed_sinks, struct afr_reply *replies,
+			     int *source_p)
+{
+	int ret = -1;
+	int source = -1;
+	afr_private_t *priv = NULL;
+	int i = 0;
+
+	priv = this->private;
+
+	ret = afr_selfheal_unlocked_discover (frame, parent, pargfid, replies);
+	if (ret)
+		return ret;
+
+	ret = afr_selfheal_find_direction (frame, this, replies,
+					   AFR_ENTRY_TRANSACTION,
+					   locked_on, sources, sinks);
+	if (ret)
+		return ret;
+
+	source = __afr_selfheal_name_finalize_source (this, sources, sinks,
+						      locked_on, replies);
+	if (source < 0) {
+		/* If source is < 0 (typically split-brain), we perform a
+		   conservative merge of entries rather than erroring out */
+	}
+	*source_p = source;
+
+	for (i = 0; i < priv->child_count; i++)
+		/* Initialize the healed_sinks[] array optimistically to
+		   the intersection of to-be-healed (i.e sinks[]) and
+		   the list of servers which are up (i.e locked_on[]).
+
+		   As we encounter failures in the healing process, we
+		   will unmark the respective servers in the healed_sinks[]
+		   array.
+		*/
+		healed_sinks[i] = sinks[i] && locked_on[i];
+
+	return ret;
+}
+
+
+int
+afr_selfheal_name_do (call_frame_t *frame, xlator_t *this, inode_t *parent,
+		      uuid_t pargfid, const char *bname)
+{
+	afr_private_t *priv = NULL;
+	unsigned char *sources = NULL;
+	unsigned char *sinks = NULL;
+	unsigned char *healed_sinks = NULL;
+	unsigned char *locked_on = NULL;
+	int source = -1;
+	struct afr_reply *replies = NULL;
+	int ret = -1;
+	inode_t *inode = NULL;
+
+	priv = this->private;
+
+	locked_on = alloca0 (priv->child_count);
+	sources = alloca0 (priv->child_count);
+	sinks = alloca0 (priv->child_count);
+	healed_sinks = alloca0 (priv->child_count);
+
+	replies = alloca0 (priv->child_count * sizeof(*replies));
+
+	ret = afr_selfheal_entrylk (frame, this, parent, this->name, bname,
+				    locked_on);
+	{
+		if (ret < 2) {
+			ret = -ENOTCONN;
+			goto unlock;
+		}
+
+		ret = __afr_selfheal_name_prepare (frame, this, parent, pargfid,
+						   locked_on, sources, sinks,
+						   healed_sinks, replies,
+						   &source);
+		if (ret)
+			goto unlock;
+
+		inode = afr_selfheal_unlocked_lookup_on (frame, parent, bname,
+							 replies, locked_on);
+		if (!inode) {
+			ret = -ENOMEM;
+			goto unlock;
+		}
+
+		ret = __afr_selfheal_name_do (frame, this, parent, pargfid, bname,
+					      inode, sources, sinks, healed_sinks,
+					      source, locked_on, replies);
+	}
+unlock:
+	afr_selfheal_unentrylk (frame, this, parent, this->name, bname,
+				locked_on);
+	if (inode)
+		inode_unref (inode);
+
+	return ret;
+}
+
+
+int
+afr_selfheal_name_unlocked_inspect (call_frame_t *frame, xlator_t *this,
+				    inode_t *parent, uuid_t pargfid,
+				    const char *bname, gf_boolean_t *need_heal)
+{
+	afr_private_t *priv = NULL;
+	int i = 0;
+	struct afr_reply *replies = NULL;
+	inode_t *inode = NULL;
+	int first_idx = -1;
+
+	priv = this->private;
+
+	replies = alloca0 (sizeof (*replies) * priv->child_count);
+
+	inode = afr_selfheal_unlocked_lookup_on (frame, parent, bname,
+						 replies, priv->child_up);
+	if (!inode)
+		return -ENOMEM;
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].valid)
+			continue;
+
+		if (first_idx == -1) {
+			first_idx = i;
+			continue;
+		}
+
+		if (replies[i].op_ret != replies[first_idx].op_ret)
+			*need_heal = _gf_true;
+
+		if (uuid_compare (replies[i].poststat.ia_gfid,
+				  replies[first_idx].poststat.ia_gfid))
+			*need_heal = _gf_true;
+	}
+
+	if (inode)
+		inode_unref (inode);
+	return 0;
+}
+
+int
+afr_selfheal_name (xlator_t *this, uuid_t pargfid, const char *bname)
+{
+	inode_t *parent = NULL;
+	call_frame_t *frame = NULL;
+	int ret = -1;
+	gf_boolean_t need_heal = _gf_false;
+
+	parent = afr_inode_find (this, pargfid);
+	if (!parent)
+		goto out;
+
+	frame = afr_frame_create (this);
+	if (!frame)
+		goto out;
+
+	ret = afr_selfheal_name_unlocked_inspect (frame, this, parent, pargfid,
+						  bname, &need_heal);
+	if (ret)
+		goto out;
+
+	if (need_heal)
+		afr_selfheal_name_do (frame, this, parent, pargfid, bname);
+out:
+	if (parent)
+		inode_unref (parent);
+	if (frame)
+		AFR_STACK_DESTROY (frame);
+
+	return ret;
+}
diff --git a/xlators/cluster/afr/src/afr-self-heal.h b/xlators/cluster/afr/src/afr-self-heal.h
index b10ae3fc0..a1b972ac3 100644
--- a/xlators/cluster/afr/src/afr-self-heal.h
+++ b/xlators/cluster/afr/src/afr-self-heal.h
@@ -1,54 +1,167 @@
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 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 __AFR_SELF_HEAL_H__
-#define __AFR_SELF_HEAL_H__
 
-#include <sys/stat.h>
+#ifndef _AFR_SELFHEAL_H
+#define _AFR_SELFHEAL_H
+
+
+/* Perform fop on all UP subvolumes and wait for all callbacks to return */
+
+#define AFR_ONALL(frame, rfn, fop, args ...) do {			\
+	afr_local_t *__local = frame->local;				\
+	afr_private_t *__priv = frame->this->private;			\
+	int __i = 0, __count = 0;					\
+									\
+	afr_replies_wipe (__local, __priv);				\
+									\
+	for (__i = 0; __i < __priv->child_count; __i++) {		\
+		if (!__priv->child_up[__i]) continue;			\
+		STACK_WIND_COOKIE (frame, rfn, (void *)(long) __i,	\
+				   __priv->children[__i],		\
+				   __priv->children[__i]->fops->fop, args); \
+		__count++;						\
+	}								\
+	syncbarrier_wait (&__local->barrier, __count);			\
+	} while (0)
+
+
+/* Perform fop on all subvolumes represented by list[] array and wait
+   for all callbacks to return */
+
+#define AFR_ONLIST(list, frame, rfn, fop, args ...) do {		\
+	afr_local_t *__local = frame->local;				\
+	afr_private_t *__priv = frame->this->private;			\
+	int __i = 0, __count = 0;					\
+									\
+	afr_replies_wipe (__local, __priv);				\
+									\
+	for (__i = 0; __i < __priv->child_count; __i++) {		\
+		if (!list[__i]) continue;				\
+		STACK_WIND_COOKIE (frame, rfn, (void *)(long) __i,	\
+				   __priv->children[__i],		\
+				   __priv->children[__i]->fops->fop, args); \
+		__count++;						\
+	}								\
+	syncbarrier_wait (&__local->barrier, __count);			\
+	} while (0)
+
+
+#define AFR_SEQ(frame, rfn, fop, args ...) do {				\
+	afr_local_t *__local = frame->local;				\
+	afr_private_t *__priv = frame->this->private;			\
+	int __i = 0;							\
+									\
+	afr_replies_wipe (__local, __priv);				\
+									\
+	for (__i = 0; __i < __priv->child_count; __i++) {		\
+		if (!__priv->child_up[__i]) continue;			\
+		STACK_WIND_COOKIE (frame, rfn, (void *)(long) __i,	\
+				   __priv->children[__i],		\
+				   __priv->children[__i]->fops->fop, args); \
+		syncbarrier_wait (&__local->barrier, 1);		\
+	}								\
+	} while (0)
+
+
+#define ALLOC_MATRIX(n, type) ({type **__ptr = NULL; \
+	int __i; \
+	__ptr = alloca0 (n * sizeof(type *)); \
+	for (__i = 0; __i < n; __i++) __ptr[__i] = alloca0 (n * sizeof(type)); \
+	__ptr;})
+
+
+#define IA_EQUAL(f,s,field) (memcmp (&(f.ia_##field), &(s.ia_##field), sizeof (s.ia_##field)) == 0)
+
+
+int
+afr_selfheal (xlator_t *this, uuid_t gfid);
+
+int
+afr_selfheal_name (xlator_t *this, uuid_t gfid, const char *name);
+
+int
+afr_selfheal_data (call_frame_t *frame, xlator_t *this, inode_t *inode);
+
+int
+afr_selfheal_metadata (call_frame_t *frame, xlator_t *this, inode_t *inode);
+
+int
+afr_selfheal_entry (call_frame_t *frame, xlator_t *this, inode_t *inode);
+
+
+int
+afr_selfheal_inodelk (call_frame_t *frame, xlator_t *this, inode_t *inode,
+		      char *dom, off_t off, size_t size,
+		      unsigned char *locked_on);
+
+int
+afr_selfheal_tryinodelk (call_frame_t *frame, xlator_t *this, inode_t *inode,
+			 char *dom, off_t off, size_t size,
+			 unsigned char *locked_on);
 
-#define FILETYPE_DIFFERS(buf1,buf2) ((buf1)->ia_type != (buf2)->ia_type)
-#define PERMISSION_DIFFERS(buf1,buf2) (st_mode_from_ia ((buf1)->ia_prot, (buf1)->ia_type) != st_mode_from_ia ((buf2)->ia_prot, (buf2)->ia_type))
-#define OWNERSHIP_DIFFERS(buf1,buf2) (((buf1)->ia_uid != (buf2)->ia_uid) || ((buf1)->ia_gid != (buf2)->ia_gid))
-#define SIZE_DIFFERS(buf1,buf2) ((buf1)->ia_size != (buf2)->ia_size)
+int
+afr_selfheal_uninodelk (call_frame_t *frame, xlator_t *this, inode_t *inode,
+			char *dom, off_t off, size_t size,
+			const unsigned char *locked_on);
 
-#define SIZE_GREATER(buf1,buf2) ((buf1)->ia_size > (buf2)->ia_size)
+int
+afr_selfheal_entrylk (call_frame_t *frame, xlator_t *this, inode_t *inode,
+		      char *dom, const char *name, unsigned char *locked_on);
 
 int
-afr_sh_has_metadata_pending (dict_t *xattr, int child_count, xlator_t *this);
+afr_selfheal_tryentrylk (call_frame_t *frame, xlator_t *this, inode_t *inode,
+			 char *dom, const char *name, unsigned char *locked_on);
+
 int
-afr_sh_has_entry_pending (dict_t *xattr, int child_count, xlator_t *this);
+afr_selfheal_unentrylk (call_frame_t *frame, xlator_t *this, inode_t *inode,
+			char *dom, const char *name, unsigned char *locked_on);
+
 int
-afr_sh_has_data_pending (dict_t *xattr, int child_count, xlator_t *this);
+afr_selfheal_unlocked_discover (call_frame_t *frame, inode_t *inode,
+				uuid_t gfid, struct afr_reply *replies);
+
+inode_t *
+afr_selfheal_unlocked_lookup_on (call_frame_t *frame, inode_t *parent,
+				 const char *name, struct afr_reply *replies,
+				 unsigned char *lookup_on);
 
 int
-afr_self_heal_entry (call_frame_t *frame, xlator_t *this);
+afr_selfheal_find_direction (call_frame_t *frame, xlator_t *this,
+			     struct afr_reply *replies,
+			     afr_transaction_type type, unsigned char *locked_on,
+			     unsigned char *sources, unsigned char *sinks);
 
 int
-afr_self_heal_data (call_frame_t *frame, xlator_t *this);
+afr_selfheal_extract_xattr (xlator_t *this, struct afr_reply *replies,
+			    afr_transaction_type type, int *dirty, int **matrix);
 
 int
-afr_self_heal_metadata (call_frame_t *frame, xlator_t *this);
+afr_selfheal_undo_pending (call_frame_t *frame, xlator_t *this, inode_t *inode,
+			   unsigned char *sources, unsigned char *sinks,
+			   unsigned char *healed_sinks, afr_transaction_type type,
+			   struct afr_reply *replies, unsigned char *locked_on);
 
 int
-afr_self_heal_get_source (xlator_t *this, afr_local_t *local, dict_t **xattr);
+afr_selfheal_recreate_entry (call_frame_t *frame, xlator_t *this, int dst,
+			     int source, inode_t *dir, const char *name,
+			     inode_t *inode, struct afr_reply *replies);
 
 int
-afr_self_heal (call_frame_t *frame, xlator_t *this);
+afr_selfheal_post_op (call_frame_t *frame, xlator_t *this, inode_t *inode,
+		      int subvol, dict_t *xattr);
+
+call_frame_t *
+afr_frame_create (xlator_t *this);
+
+inode_t *
+afr_inode_find (xlator_t *this, uuid_t gfid);
 
-#endif /* __AFR_SELF_HEAL_H__ */
+#endif /* !_AFR_SELFHEAL_H */
diff --git a/xlators/cluster/afr/src/afr-self-heald.c b/xlators/cluster/afr/src/afr-self-heald.c
new file mode 100644
index 000000000..4bfe909bc
--- /dev/null
+++ b/xlators/cluster/afr/src/afr-self-heald.c
@@ -0,0 +1,1256 @@
+/*
+  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 "afr.h"
+#include "afr-self-heal.h"
+#include "afr-self-heald.h"
+#include "protocol-common.h"
+
+#define SHD_INODE_LRU_LIMIT          2048
+#define AFR_EH_HEALED_LIMIT          1024
+#define AFR_EH_HEAL_FAIL_LIMIT       1024
+#define AFR_EH_SPLIT_BRAIN_LIMIT     1024
+#define AFR_STATISTICS_HISTORY_SIZE    50
+
+
+#define ASSERT_LOCAL(this, healer)				\
+	if (!afr_shd_is_subvol_local(this, healer->subvol)) {	\
+		healer->local = _gf_false;			\
+		if (safe_break (healer)) {			\
+			break;					\
+		} else {					\
+			continue;				\
+		}						\
+	} else {						\
+		healer->local = _gf_true;			\
+	}
+
+
+#define NTH_INDEX_HEALER(this, n) &((((afr_private_t *)this->private))->shd.index_healers[n])
+#define NTH_FULL_HEALER(this, n) &((((afr_private_t *)this->private))->shd.full_healers[n])
+
+int afr_shd_gfid_to_path (xlator_t *this, xlator_t *subvol, uuid_t gfid, char **path_p);
+
+char *
+afr_subvol_name (xlator_t *this, int subvol)
+{
+	afr_private_t *priv = NULL;
+
+	priv = this->private;
+	if (subvol < 0 || subvol > priv->child_count)
+		return NULL;
+
+	return priv->children[subvol]->name;
+}
+
+
+void
+afr_destroy_crawl_event_data (void *data)
+{
+        return;
+}
+
+
+void
+afr_destroy_shd_event_data (void *data)
+{
+	shd_event_t *shd_event = data;
+
+	if (!shd_event)
+		return;
+	GF_FREE (shd_event->path);
+
+        return;
+}
+
+
+gf_boolean_t
+afr_shd_is_subvol_local (xlator_t *this, int subvol)
+{
+	char *pathinfo = NULL;
+	afr_private_t *priv = NULL;
+	dict_t *xattr = NULL;
+	int ret = 0;
+	gf_boolean_t is_local = _gf_false;
+	loc_t loc = {0, };
+
+	priv = this->private;
+
+	loc.inode = this->itable->root;
+	uuid_copy (loc.gfid, loc.inode->gfid);
+
+	ret = syncop_getxattr (priv->children[subvol], &loc, &xattr,
+			       GF_XATTR_PATHINFO_KEY);
+	if (ret)
+		return _gf_false;
+	if (!xattr)
+		return _gf_false;
+
+	ret = dict_get_str (xattr, GF_XATTR_PATHINFO_KEY, &pathinfo);
+	if (ret)
+		return _gf_false;
+
+	afr_local_pathinfo (pathinfo, &is_local);
+
+	gf_log (this->name, GF_LOG_DEBUG, "subvol %s is %slocal",
+		priv->children[subvol]->name, is_local? "" : "not ");
+
+	return is_local;
+}
+
+
+int
+__afr_shd_healer_wait (struct subvol_healer *healer)
+{
+	afr_private_t *priv = NULL;
+	struct timespec wait_till = {0, };
+	int ret = 0;
+
+	priv = healer->this->private;
+
+disabled_loop:
+	wait_till.tv_sec = time (NULL) + 60;
+
+	while (!healer->rerun) {
+		ret = pthread_cond_timedwait (&healer->cond,
+					      &healer->mutex,
+					      &wait_till);
+		if (ret == ETIMEDOUT)
+			break;
+	}
+
+	ret = healer->rerun;
+	healer->rerun = 0;
+
+	if (!priv->shd.enabled)
+		goto disabled_loop;
+
+	return ret;
+}
+
+
+int
+afr_shd_healer_wait (struct subvol_healer *healer)
+{
+	int ret = 0;
+
+	pthread_mutex_lock (&healer->mutex);
+	{
+		ret = __afr_shd_healer_wait (healer);
+	}
+	pthread_mutex_unlock (&healer->mutex);
+
+	return ret;
+}
+
+
+gf_boolean_t
+safe_break (struct subvol_healer *healer)
+{
+	gf_boolean_t ret = _gf_false;
+
+	pthread_mutex_lock (&healer->mutex);
+	{
+		if (healer->rerun)
+			goto unlock;
+
+		healer->running = _gf_false;
+		ret = _gf_true;
+	}
+unlock:
+	pthread_mutex_unlock (&healer->mutex);
+
+	return ret;
+}
+
+
+inode_t *
+afr_shd_inode_find (xlator_t *this, xlator_t *subvol, uuid_t gfid)
+{
+	inode_t *inode = NULL;
+	int ret = 0;
+	loc_t loc = {0, };
+	struct iatt iatt = {0, };
+
+	inode = inode_find (this->itable, gfid);
+	if (inode)
+		goto out;
+
+	loc.inode = inode_new (this->itable);
+	if (!loc.inode)
+		goto out;
+	uuid_copy (loc.gfid, gfid);
+
+	ret = syncop_lookup (subvol, &loc, NULL, &iatt, NULL, NULL);
+	if (ret < 0)
+		goto out;
+
+	inode = inode_link (loc.inode, NULL, NULL, &iatt);
+	if (inode)
+		inode_lookup (inode);
+out:
+	loc_wipe (&loc);
+	return inode;
+}
+
+
+fd_t *
+afr_shd_index_opendir (xlator_t *this, int child)
+{
+	fd_t *fd = NULL;
+	afr_private_t *priv = NULL;
+	xlator_t *subvol = NULL;
+	loc_t rootloc = {0, };
+	inode_t *inode = NULL;
+	int ret = 0;
+	dict_t *xattr = NULL;
+	void *index_gfid = NULL;
+
+	priv = this->private;
+	subvol = priv->children[child];
+
+	rootloc.inode = inode_ref (this->itable->root);
+	uuid_copy (rootloc.gfid, rootloc.inode->gfid);
+
+	ret = syncop_getxattr (subvol, &rootloc, &xattr,
+			       GF_XATTROP_INDEX_GFID);
+	if (ret || !xattr) {
+		errno = -ret;
+		goto out;
+	}
+
+	ret = dict_get_ptr (xattr, GF_XATTROP_INDEX_GFID, &index_gfid);
+	if (ret)
+		goto out;
+
+	gf_log (this->name, GF_LOG_DEBUG, "index-dir gfid for %s: %s",
+		subvol->name, uuid_utoa (index_gfid));
+
+	inode = afr_shd_inode_find (this, subvol, index_gfid);
+	if (!inode)
+		goto out;
+	fd = fd_anonymous (inode);
+out:
+	loc_wipe (&rootloc);
+	if (xattr)
+		dict_unref (xattr);
+	return fd;
+}
+
+
+int
+afr_shd_index_purge (xlator_t *subvol, inode_t *inode, char *name)
+{
+	loc_t loc = {0, };
+	int ret = 0;
+
+	loc.parent = inode_ref (inode);
+	loc.name = name;
+
+	ret = syncop_unlink (subvol, &loc);
+
+	loc_wipe (&loc);
+	return ret;
+}
+
+
+int
+afr_shd_selfheal_name (struct subvol_healer *healer, int child, uuid_t parent,
+		       const char *bname)
+{
+	int ret = -1;
+
+	ret = afr_selfheal_name (THIS, parent, bname);
+
+	return ret;
+}
+
+int
+afr_shd_selfheal (struct subvol_healer *healer, int child, uuid_t gfid)
+{
+	int ret = 0;
+	eh_t *eh = NULL;
+	afr_private_t *priv = NULL;
+	afr_self_heald_t *shd = NULL;
+	shd_event_t *shd_event = NULL;
+	char *path = NULL;
+	xlator_t *subvol = NULL;
+	xlator_t *this = NULL;
+	crawl_event_t *crawl_event = NULL;
+
+	this = healer->this;
+	priv = this->private;
+	shd = &priv->shd;
+	crawl_event = &healer->crawl_event;
+
+	subvol = priv->children[child];
+
+	ret = afr_selfheal (this, gfid);
+
+	if (ret == -EIO) {
+		eh = shd->split_brain;
+		crawl_event->split_brain_count++;
+	} else if (ret < 0) {
+		eh = shd->heal_failed;
+		crawl_event->heal_failed_count++;
+	} else if (ret == 0) {
+		eh = shd->healed;
+		crawl_event->healed_count++;
+	}
+
+	afr_shd_gfid_to_path (this, subvol, gfid, &path);
+	if (!path)
+		return ret;
+
+	if (eh) {
+		shd_event = GF_CALLOC (1, sizeof(*shd_event),
+				       gf_afr_mt_shd_event_t);
+		if (!shd_event) {
+			GF_FREE (path);
+			return ret;
+		}
+
+		shd_event->child = child;
+		shd_event->path = path;
+
+		if (eh_save_history (eh, shd_event) < 0) {
+			GF_FREE (shd_event);
+			GF_FREE (path);
+		}
+	}
+	return ret;
+}
+
+
+void
+afr_shd_sweep_prepare (struct subvol_healer *healer)
+{
+	crawl_event_t *event = NULL;
+
+	event = &healer->crawl_event;
+
+	event->healed_count = 0;
+	event->split_brain_count = 0;
+	event->heal_failed_count = 0;
+
+	time (&event->start_time);
+	event->end_time = 0;
+}
+
+
+void
+afr_shd_sweep_done (struct subvol_healer *healer)
+{
+	crawl_event_t *event = NULL;
+	crawl_event_t *history = NULL;
+	afr_self_heald_t *shd = NULL;
+
+	event = &healer->crawl_event;
+	shd = &(((afr_private_t *)healer->this->private)->shd);
+
+	time (&event->end_time);
+	history = memdup (event, sizeof (*event));
+	event->start_time = 0;
+
+	if (!history)
+		return;
+
+	if (eh_save_history (shd->statistics[healer->subvol], history) < 0)
+		GF_FREE (history);
+}
+
+
+int
+afr_shd_index_sweep (struct subvol_healer *healer)
+{
+	xlator_t *this = NULL;
+	int child = -1;
+	fd_t *fd = NULL;
+	xlator_t *subvol = NULL;
+	afr_private_t *priv = NULL;
+	off_t offset = 0;
+	gf_dirent_t entries;
+	gf_dirent_t *entry = NULL;
+	uuid_t gfid;
+	int ret = 0;
+	int count = 0;
+
+	this = healer->this;
+	child = healer->subvol;
+	priv = this->private;
+	subvol = priv->children[child];
+
+	fd = afr_shd_index_opendir (this, child);
+	if (!fd) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"unable to opendir index-dir on %s", subvol->name);
+		return -errno;
+	}
+
+	INIT_LIST_HEAD (&entries.list);
+
+	while ((ret = syncop_readdir (subvol, fd, 131072, offset, &entries))) {
+		if (ret > 0)
+			ret = 0;
+		list_for_each_entry (entry, &entries.list, list) {
+			offset = entry->d_off;
+
+			if (!priv->shd.enabled) {
+				ret = -EBUSY;
+				break;
+			}
+
+			if (!strcmp (entry->d_name, ".") ||
+			    !strcmp (entry->d_name, ".."))
+				continue;
+
+			gf_log (this->name, GF_LOG_DEBUG, "got entry: %s",
+				entry->d_name);
+
+			ret = uuid_parse (entry->d_name, gfid);
+			if (ret)
+				continue;
+
+			ret = afr_shd_selfheal (healer, child, gfid);
+			if (ret == 0)
+				count++;
+
+			if (ret == -ENOENT || ret == -ESTALE) {
+				afr_shd_index_purge (subvol, fd->inode,
+						     entry->d_name);
+				ret = 0;
+			}
+		}
+
+		gf_dirent_free (&entries);
+		if (ret)
+			break;
+	}
+
+	if (fd)
+		fd_unref (fd);
+	if (!ret)
+		ret = count;
+	return ret;
+}
+
+
+int
+afr_shd_full_sweep (struct subvol_healer *healer, inode_t *inode)
+{
+	fd_t *fd = NULL;
+	xlator_t *this = NULL;
+	xlator_t *subvol = NULL;
+	afr_private_t *priv = NULL;
+	off_t offset = 0;
+	gf_dirent_t entries;
+	gf_dirent_t *entry = NULL;
+	int ret = 0;
+
+	this = healer->this;
+	priv = this->private;
+	subvol = priv->children[healer->subvol];
+
+	fd = fd_anonymous (inode);
+	if (!fd)
+		return -errno;
+
+	INIT_LIST_HEAD (&entries.list);
+
+	while ((ret = syncop_readdirp (subvol, fd, 131072, offset, 0, &entries))) {
+		if (ret < 0)
+			break;
+
+		ret = gf_link_inodes_from_dirent (this, fd->inode, &entries);
+		if (ret)
+			break;
+
+		list_for_each_entry (entry, &entries.list, list) {
+			offset = entry->d_off;
+
+			if (!priv->shd.enabled) {
+				ret = -EBUSY;
+				break;
+			}
+
+			if (!strcmp (entry->d_name, ".") ||
+			    !strcmp (entry->d_name, ".."))
+				continue;
+
+			afr_shd_selfheal_name (healer, healer->subvol,
+					       inode->gfid, entry->d_name);
+
+			afr_shd_selfheal (healer, healer->subvol,
+					  entry->d_stat.ia_gfid);
+
+			if (entry->d_stat.ia_type == IA_IFDIR) {
+				ret = afr_shd_full_sweep (healer, entry->inode);
+				if (ret)
+					break;
+			}
+		}
+
+		gf_dirent_free (&entries);
+		if (ret)
+			break;
+	}
+
+	if (fd)
+		fd_unref (fd);
+	return ret;
+}
+
+
+void *
+afr_shd_index_healer (void *data)
+{
+	struct subvol_healer *healer = NULL;
+	xlator_t *this = NULL;
+	int ret = 0;
+
+	healer = data;
+	THIS = this = healer->this;
+
+	for (;;) {
+		afr_shd_healer_wait (healer);
+
+		ASSERT_LOCAL(this, healer);
+
+		do {
+			gf_log (this->name, GF_LOG_DEBUG,
+				"starting index sweep on subvol %s",
+				afr_subvol_name (this, healer->subvol));
+
+			afr_shd_sweep_prepare (healer);
+
+			ret = afr_shd_index_sweep (healer);
+
+			afr_shd_sweep_done (healer);
+			/*
+			  As long as at least one gfid was
+			  healed, keep retrying. We may have
+			  just healed a directory and thereby
+			  created entries for other gfids which
+			  could not be healed thus far.
+			*/
+
+			gf_log (this->name, GF_LOG_DEBUG,
+				"finished index sweep on subvol %s",
+				afr_subvol_name (this, healer->subvol));
+			/*
+			  Give a pause before retrying to avoid a busy loop
+			  in case the only entry in index is because of
+			  an ongoing I/O.
+			*/
+			sleep (1);
+		} while (ret > 0);
+	}
+
+	return NULL;
+}
+
+
+void *
+afr_shd_full_healer (void *data)
+{
+	struct subvol_healer *healer = NULL;
+	xlator_t *this = NULL;
+	int run = 0;
+
+	healer = data;
+	THIS = this = healer->this;
+
+	for (;;) {
+		pthread_mutex_lock (&healer->mutex);
+		{
+			run = __afr_shd_healer_wait (healer);
+			if (!run)
+				healer->running = _gf_false;
+		}
+		pthread_mutex_unlock (&healer->mutex);
+
+		if (!run)
+			break;
+
+		ASSERT_LOCAL(this, healer);
+
+		gf_log (this->name, GF_LOG_INFO,
+			"starting full sweep on subvol %s",
+			afr_subvol_name (this, healer->subvol));
+
+		afr_shd_sweep_prepare (healer);
+
+		afr_shd_full_sweep (healer, this->itable->root);
+
+		afr_shd_sweep_done (healer);
+
+		gf_log (this->name, GF_LOG_INFO,
+			"finished full sweep on subvol %s",
+			afr_subvol_name (this, healer->subvol));
+	}
+
+	return NULL;
+}
+
+
+int
+afr_shd_healer_init (xlator_t *this, struct subvol_healer *healer)
+{
+	int ret = 0;
+
+	ret = pthread_mutex_init (&healer->mutex, NULL);
+	if (ret)
+		goto out;
+
+	ret = pthread_cond_init (&healer->cond, NULL);
+	if (ret)
+		goto out;
+
+	healer->this = this;
+	healer->running = _gf_false;
+	healer->rerun = _gf_false;
+	healer->local = _gf_false;
+out:
+	return ret;
+}
+
+
+int
+afr_shd_healer_spawn (xlator_t *this, struct subvol_healer *healer,
+		      void *(threadfn)(void *))
+{
+	int ret = 0;
+
+	pthread_mutex_lock (&healer->mutex);
+	{
+		if (healer->running) {
+			pthread_cond_signal (&healer->cond);
+		} else {
+			ret = gf_thread_create (&healer->thread, NULL,
+						threadfn, healer);
+			if (ret)
+				goto unlock;
+			healer->running = 1;
+		}
+
+		healer->rerun = 1;
+	}
+unlock:
+	pthread_mutex_unlock (&healer->mutex);
+
+	return ret;
+}
+
+
+int
+afr_shd_full_healer_spawn (xlator_t *this, int subvol)
+{
+	return afr_shd_healer_spawn (this, NTH_FULL_HEALER (this, subvol),
+				     afr_shd_full_healer);
+}
+
+
+int
+afr_shd_index_healer_spawn (xlator_t *this, int subvol)
+{
+	return afr_shd_healer_spawn (this, NTH_INDEX_HEALER (this, subvol),
+				     afr_shd_index_healer);
+}
+
+
+int
+afr_shd_dict_add_crawl_event (xlator_t *this, dict_t *output,
+			      crawl_event_t *crawl_event)
+{
+        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 = 0;
+        char            *end_time_str = NULL;
+        char            *crawl_type = NULL;
+        int             progress = -1;
+	int             child = -1;
+
+	child = crawl_event->child;
+        healed_count = crawl_event->healed_count;
+        split_brain_count = crawl_event->split_brain_count;
+        heal_failed_count = crawl_event->heal_failed_count;
+        crawl_type = crawl_event->crawl_type;
+
+	if (!crawl_event->start_time)
+		goto out;
+
+        start_time_str = gf_strdup (ctime (&crawl_event->start_time));
+
+	if (crawl_event->end_time)
+		end_time_str = gf_strdup (ctime (&crawl_event->end_time));
+
+        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_str (output, key, 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, start_time_str);
+	if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+			"Could not add statistics_crawl_start_time to outout");
+                goto out;
+        } else {
+		start_time_str = NULL;
+	}
+
+	if (!end_time_str)
+                progress = 1;
+        else
+                progress = 0;
+
+        snprintf (key, sizeof (key), "statistics_end_time-%d-%d-%"PRIu64,
+                  xl_id, child, count);
+        if (!end_time_str)
+                end_time_str = gf_strdup ("Could not determine the end time");
+        ret = dict_set_dynstr (output, key, end_time_str);
+	if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+			"Could not add statistics_crawl_end_time to outout");
+                goto out;
+        } else {
+		end_time_str = NULL;
+	}
+
+        snprintf (key, sizeof (key), "statistics_inprogress-%d-%d-%"PRIu64,
+                  xl_id, child, count);
+
+        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:
+	GF_FREE (start_time_str);
+	GF_FREE (end_time_str);
+        return ret;
+}
+
+
+int
+afr_shd_dict_add_path (xlator_t *this, dict_t *output, int child, char *path,
+		       struct timeval *tv)
+{
+        int             ret = -1;
+        uint64_t        count = 0;
+        char            key[256] = {0};
+        int             xl_id = 0;
+
+        ret = dict_get_int32 (output, this->name, &xl_id);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "xl does not have id");
+                goto out;
+        }
+
+        snprintf (key, sizeof (key), "%d-%d-count", xl_id, child);
+        ret = dict_get_uint64 (output, key, &count);
+
+        snprintf (key, sizeof (key), "%d-%d-%"PRIu64, xl_id, child, count);
+	ret = dict_set_dynstr (output, key, path);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "%s: Could not add to output",
+                        path);
+                goto out;
+        }
+
+	if (tv) {
+		snprintf (key, sizeof (key), "%d-%d-%"PRIu64"-time", xl_id,
+			  child, count);
+		ret = dict_set_uint32 (output, key, tv->tv_sec);
+		if (ret) {
+			gf_log (this->name, GF_LOG_ERROR, "%s: Could not set time",
+				path);
+			goto out;
+		}
+	}
+
+        snprintf (key, sizeof (key), "%d-%d-count", xl_id, child);
+
+        ret = dict_set_uint64 (output, key, count + 1);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Could not increment count");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+
+int
+afr_shd_gfid_to_path (xlator_t *this, xlator_t *subvol, uuid_t gfid, char **path_p)
+{
+	loc_t loc = {0,};
+	char *path = NULL;
+	dict_t *xattr = NULL;
+	int ret = 0;
+
+	uuid_copy (loc.gfid, gfid);
+	loc.inode = inode_new (this->itable);
+
+	ret = syncop_getxattr (subvol, &loc, &xattr, GFID_TO_PATH_KEY);
+	loc_wipe (&loc);
+	if (ret)
+		return ret;
+
+	ret = dict_get_str (xattr, GFID_TO_PATH_KEY, &path);
+	if (ret || !path)
+		return -EINVAL;
+
+	*path_p = gf_strdup (path);
+	if (!*path_p)
+		return -ENOMEM;
+	return 0;
+}
+
+
+int
+afr_shd_gather_index_entries (xlator_t *this, int child, dict_t *output)
+{
+	fd_t *fd = NULL;
+	xlator_t *subvol = NULL;
+	afr_private_t *priv = NULL;
+	off_t offset = 0;
+	gf_dirent_t entries;
+	gf_dirent_t *entry = NULL;
+	uuid_t gfid;
+	int ret = 0;
+	int count = 0;
+	char *path = NULL;
+
+	priv = this->private;
+	subvol = priv->children[child];
+
+	fd = afr_shd_index_opendir (this, child);
+	if (!fd) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"unable to opendir index-dir on %s", subvol->name);
+		return -errno;
+	}
+
+	INIT_LIST_HEAD (&entries.list);
+
+	while ((ret = syncop_readdir (subvol, fd, 131072, offset, &entries))) {
+		if (ret > 0)
+			ret = 0;
+		list_for_each_entry (entry, &entries.list, list) {
+			offset = entry->d_off;
+
+			if (!strcmp (entry->d_name, ".") ||
+			    !strcmp (entry->d_name, ".."))
+				continue;
+
+			gf_log (this->name, GF_LOG_DEBUG, "got entry: %s",
+				entry->d_name);
+
+			ret = uuid_parse (entry->d_name, gfid);
+			if (ret)
+				continue;
+
+			path = NULL;
+			ret = afr_shd_gfid_to_path (this, subvol, gfid, &path);
+
+			if (ret == -ENOENT || ret == -ESTALE) {
+				afr_shd_index_purge (subvol, fd->inode,
+						     entry->d_name);
+				ret = 0;
+				continue;
+			}
+
+			ret = afr_shd_dict_add_path (this, output, child, path,
+						     NULL);
+		}
+
+		gf_dirent_free (&entries);
+		if (ret)
+			break;
+	}
+
+	if (fd)
+		fd_unref (fd);
+	if (!ret)
+		ret = count;
+	return ret;
+}
+
+
+int
+afr_add_shd_event (circular_buffer_t *cb, void *data)
+{
+	dict_t *output = NULL;
+	xlator_t *this = THIS;
+	afr_private_t *priv = NULL;
+	afr_self_heald_t *shd = NULL;
+	shd_event_t *shd_event = NULL;
+	char *path = NULL;
+
+	output = data;
+	priv = this->private;
+	shd = &priv->shd;
+	shd_event = cb->data;
+
+	if (!shd->index_healers[shd_event->child].local)
+		return 0;
+
+	path = gf_strdup (shd_event->path);
+	if (!path)
+		return -ENOMEM;
+
+	afr_shd_dict_add_path (this, output, shd_event->child, path,
+			       &cb->tv);
+	return 0;
+}
+
+int
+afr_add_crawl_event (circular_buffer_t *cb, void *data)
+{
+	dict_t *output = NULL;
+	xlator_t *this = THIS;
+	afr_private_t *priv = NULL;
+	afr_self_heald_t *shd = NULL;
+	crawl_event_t *crawl_event = NULL;
+
+	output = data;
+	priv = this->private;
+	shd = &priv->shd;
+	crawl_event = cb->data;
+
+	if (!shd->index_healers[crawl_event->child].local)
+		return 0;
+
+	afr_shd_dict_add_crawl_event (this, output, crawl_event);
+
+	return 0;
+}
+
+
+int
+afr_selfheal_daemon_init (xlator_t *this)
+{
+	afr_private_t *priv = NULL;
+	afr_self_heald_t *shd = NULL;
+	int ret = -1;
+	int i = 0;
+
+	priv = this->private;
+	shd = &priv->shd;
+
+	this->itable = inode_table_new (SHD_INODE_LRU_LIMIT, this);
+	if (!this->itable)
+		goto out;
+
+	shd->index_healers = GF_CALLOC (sizeof(*shd->index_healers),
+					priv->child_count,
+					gf_afr_mt_subvol_healer_t);
+	if (!shd->index_healers)
+		goto out;
+
+	for (i = 0; i < priv->child_count; i++) {
+		shd->index_healers[i].subvol = i;
+		ret = afr_shd_healer_init (this, &shd->index_healers[i]);
+		if (ret)
+			goto out;
+	}
+
+	shd->full_healers = GF_CALLOC (sizeof(*shd->full_healers),
+				       priv->child_count,
+				       gf_afr_mt_subvol_healer_t);
+	if (!shd->full_healers)
+		goto out;
+	for (i = 0; i < priv->child_count; i++) {
+		shd->full_healers[i].subvol = i;
+		ret = afr_shd_healer_init (this, &shd->full_healers[i]);
+		if (ret)
+			goto out;
+	}
+
+	shd->healed = eh_new (AFR_EH_HEALED_LIMIT, _gf_false,
+			      afr_destroy_shd_event_data);
+        if (!shd->healed)
+		goto out;
+
+	shd->heal_failed = eh_new (AFR_EH_HEAL_FAIL_LIMIT, _gf_false,
+				   afr_destroy_shd_event_data);
+	if (!shd->heal_failed)
+		goto out;
+
+	shd->split_brain = eh_new (AFR_EH_SPLIT_BRAIN_LIMIT, _gf_false,
+				   afr_destroy_shd_event_data);
+	if (!shd->split_brain)
+		goto out;
+
+        shd->statistics = GF_CALLOC (sizeof(eh_t *), priv->child_count,
+				     gf_common_mt_eh_t);
+        if (!shd->statistics)
+                goto out;
+
+        for (i = 0; i < priv->child_count ; i++) {
+                shd->statistics[i] = eh_new (AFR_STATISTICS_HISTORY_SIZE,
+					     _gf_false,
+					     afr_destroy_crawl_event_data);
+                if (!shd->statistics[i])
+                        goto out;
+		shd->full_healers[i].crawl_event.child = i;
+		shd->full_healers[i].crawl_event.crawl_type = "FULL";
+		shd->index_healers[i].crawl_event.child = i;
+		shd->index_healers[i].crawl_event.crawl_type = "INDEX";
+        }
+
+	ret = 0;
+out:
+	return ret;
+}
+
+
+int
+afr_selfheal_childup (xlator_t *this, int subvol)
+{
+	afr_shd_index_healer_spawn (this, subvol);
+
+	return 0;
+}
+
+
+int64_t
+afr_shd_get_index_count (xlator_t *this, int i)
+{
+	afr_private_t *priv = NULL;
+	xlator_t *subvol = NULL;
+	uint64_t count = 0;
+	loc_t rootloc = {0, };
+	dict_t *xattr = NULL;
+	int ret = -1;
+
+	priv = this->private;
+	subvol = priv->children[i];
+
+	rootloc.inode = inode_ref (this->itable->root);
+	uuid_copy (rootloc.gfid, rootloc.inode->gfid);
+
+	ret = syncop_getxattr (subvol, &rootloc, &xattr,
+			       GF_XATTROP_INDEX_COUNT);
+	loc_wipe (&rootloc);
+
+	if (ret < 0)
+		return -1;
+
+	ret = dict_get_uint64 (xattr, GF_XATTROP_INDEX_COUNT, &count);
+	if (ret)
+		return -1;
+	return count;
+}
+
+
+int
+afr_xl_op (xlator_t *this, dict_t *input, dict_t *output)
+{
+        gf_xl_afr_op_t   op = GF_AFR_OP_INVALID;
+        int              ret = 0;
+        int              xl_id = 0;
+	afr_private_t   *priv = NULL;
+	afr_self_heald_t *shd = NULL;
+	struct subvol_healer *healer = NULL;
+	int i = 0;
+	char key[64];
+	int op_ret = 0;
+	int64_t cnt = 0;
+
+	priv = this->private;
+	shd = &priv->shd;
+
+	for (i = 0; i < priv->child_count; i++)
+		if (priv->child_up[i] == -1)
+			goto out;
+
+        ret = dict_get_int32 (input, "xl-op", (int32_t*)&op);
+        if (ret)
+                goto out;
+        ret = dict_get_int32 (input, this->name, &xl_id);
+        if (ret)
+                goto out;
+        ret = dict_set_int32 (output, this->name, xl_id);
+        if (ret)
+                goto out;
+        switch (op) {
+        case GF_AFR_OP_HEAL_INDEX:
+		op_ret = -1;
+
+		for (i = 0; i < priv->child_count; i++) {
+			healer = &shd->index_healers[i];
+			snprintf (key, 64, "%d-%d-status", xl_id, i);
+
+			if (!priv->child_up[i]) {
+				ret = dict_set_str (output, key,
+						    "Brick is not connected");
+			} else if (AFR_COUNT (priv->child_up,
+					      priv->child_count) < 2) {
+				ret = dict_set_str (output, key,
+						    "< 2 bricks in replica are up");
+			} else if (!afr_shd_is_subvol_local (this, healer->subvol)) {
+				ret = dict_set_str (output, key,
+						    "Brick is remote");
+			} else {
+				ret = dict_set_str (output, key,
+						    "Started self-heal");
+				afr_shd_index_healer_spawn (this, i);
+				op_ret = 0;
+			}
+		}
+                break;
+        case GF_AFR_OP_HEAL_FULL:
+		op_ret = -1;
+
+		for (i = 0; i < priv->child_count; i++) {
+			healer = &shd->full_healers[i];
+			snprintf (key, 64, "%d-%d-status", xl_id, i);
+
+			if (!priv->child_up[i]) {
+				ret = dict_set_str (output, key,
+						    "Brick is not connected");
+			} else if (AFR_COUNT (priv->child_up,
+					      priv->child_count) < 2) {
+				ret = dict_set_str (output, key,
+						    "< 2 bricks in replica are up");
+			} else if (!afr_shd_is_subvol_local (this, healer->subvol)) {
+				ret = dict_set_str (output, key,
+						    "Brick is remote");
+			} else {
+				ret = dict_set_str (output, key,
+						    "Started self-heal");
+				afr_shd_full_healer_spawn (this, i);
+				op_ret = 0;
+			}
+		}
+                break;
+        case GF_AFR_OP_INDEX_SUMMARY:
+		for (i = 0; i < priv->child_count; i++)
+			if (shd->index_healers[i].local)
+				afr_shd_gather_index_entries (this, i, output);
+                break;
+        case GF_AFR_OP_HEALED_FILES:
+		eh_dump (shd->healed, output, afr_add_shd_event);
+                break;
+        case GF_AFR_OP_HEAL_FAILED_FILES:
+		eh_dump (shd->heal_failed, output, afr_add_shd_event);
+                break;
+        case GF_AFR_OP_SPLIT_BRAIN_FILES:
+		eh_dump (shd->split_brain, output, afr_add_shd_event);
+                break;
+        case GF_AFR_OP_STATISTICS:
+		for (i = 0; i < priv->child_count; i++) {
+			eh_dump (shd->statistics[i], output,
+				 afr_add_crawl_event);
+			afr_shd_dict_add_crawl_event (this, output,
+						      &shd->index_healers[i].crawl_event);
+			afr_shd_dict_add_crawl_event (this, output,
+						      &shd->full_healers[i].crawl_event);
+		}
+                break;
+        case GF_AFR_OP_STATISTICS_HEAL_COUNT:
+        case GF_AFR_OP_STATISTICS_HEAL_COUNT_PER_REPLICA:
+		op_ret = -1;
+
+		for (i = 0; i < priv->child_count; i++) {
+			if (!priv->child_up[i]) {
+				snprintf (key, 64, "%d-%d-status", xl_id, i);
+				ret = dict_set_str (output, key,
+						    "Brick is not connected");
+			} else {
+				snprintf (key, 64, "%d-%d-hardlinks", xl_id, i);
+				cnt = afr_shd_get_index_count (this, i);
+				if (cnt >= 0) {
+					ret = dict_set_uint64 (output, key, cnt);
+				}
+				op_ret = 0;
+			}
+		}
+
+//                ret = _do_crawl_op_on_local_subvols (this, INDEX_TO_BE_HEALED,
+//                                                     STATISTICS_TO_BE_HEALED,
+//                                                     output);
+                break;
+
+        default:
+                gf_log (this->name, GF_LOG_ERROR, "Unknown set op %d", op);
+                break;
+        }
+out:
+        dict_del (output, this->name);
+        return op_ret;
+}
diff --git a/xlators/cluster/afr/src/afr-self-heald.h b/xlators/cluster/afr/src/afr-self-heald.h
new file mode 100644
index 000000000..10e229ee7
--- /dev/null
+++ b/xlators/cluster/afr/src/afr-self-heald.h
@@ -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 _AFR_SELF_HEALD_H
+#define _AFR_SELF_HEALD_H
+
+#include <pthread.h>
+
+
+typedef struct {
+	int child;
+	char *path;
+} shd_event_t;
+
+typedef struct {
+	int      child;
+	uint64_t healed_count;
+        uint64_t split_brain_count;
+        uint64_t heal_failed_count;
+
+	/* If start_time is 0, it means crawler is not in progress
+	   and stats are not valid */
+	time_t   start_time;
+	/* If start_time is NOT 0 and end_time is 0, it means
+	   cralwer is in progress */
+        time_t   end_time;
+        char     *crawl_type;
+} crawl_event_t;
+
+struct subvol_healer {
+	xlator_t        *this;
+	int              subvol;
+	gf_boolean_t     local;
+	gf_boolean_t     running;
+	gf_boolean_t     rerun;
+	crawl_event_t    crawl_event;
+	pthread_mutex_t  mutex;
+	pthread_cond_t   cond;
+	pthread_t        thread;
+};
+
+typedef struct {
+	gf_boolean_t            iamshd;
+	gf_boolean_t            enabled;
+	struct subvol_healer   *index_healers;
+	struct subvol_healer   *full_healers;
+
+	eh_t                    *healed;
+        eh_t                    *heal_failed;
+        eh_t                    *split_brain;
+        eh_t                    **statistics;
+} afr_self_heald_t;
+
+
+int
+afr_selfheal_childup (xlator_t *this, int subvol);
+
+int
+afr_selfheal_daemon_init (xlator_t *this);
+
+int
+afr_xl_op (xlator_t *this, dict_t *input, dict_t *output);
+
+#endif /* !_AFR_SELF_HEALD_H */
diff --git a/xlators/cluster/afr/src/afr-transaction.c b/xlators/cluster/afr/src/afr-transaction.c
index ff9c88bad..205ff759e 100644
--- a/xlators/cluster/afr/src/afr-transaction.c
+++ b/xlators/cluster/afr/src/afr-transaction.c
@@ -1,265 +1,219 @@
 /*
-  Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include "dict.h"
 #include "byte-order.h"
 #include "common-utils.h"
+#include "timer.h"
 
 #include "afr.h"
 #include "afr-transaction.h"
 
 #include <signal.h>
 
+gf_boolean_t
+afr_changelog_pre_op_uninherit (call_frame_t *frame, xlator_t *this);
 
-#define LOCKED_NO       0x0        /* no lock held */
-#define LOCKED_YES      0x1        /* for DATA, METADATA, ENTRY and higher_path
-                                      of RENAME */
-#define LOCKED_LOWER    0x2        /* for lower_path of RENAME */
-
-
-afr_fd_ctx_t *
-afr_fd_ctx_get (fd_t *fd, xlator_t *this)
-{
-        uint64_t       ctx = 0;
-        afr_fd_ctx_t  *fd_ctx = NULL;
-        int            ret = 0;
-
-        ret = fd_ctx_get (fd, this, &ctx);
+gf_boolean_t
+afr_changelog_pre_op_update (call_frame_t *frame, xlator_t *this);
 
-        if (ret < 0)
-                goto out;
-
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-out:
-        return fd_ctx;
-}
+int
+afr_changelog_do (call_frame_t *frame, xlator_t *this, dict_t *xattr,
+		  afr_changelog_resume_t changelog_resume);
 
 
-static void
-afr_pid_save (call_frame_t *frame)
+int
+__afr_txn_write_fop (call_frame_t *frame, xlator_t *this)
 {
-        afr_local_t * local = NULL;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int call_count = -1;
+        int i = 0;
 
         local = frame->local;
+        priv = this->private;
 
-        local->saved_pid = frame->root->pid;
-}
+        call_count = AFR_COUNT (local->transaction.pre_op, priv->child_count);
 
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
+        }
 
-static void
-afr_pid_restore (call_frame_t *frame)
-{
-        afr_local_t * local = NULL;
+        local->call_count = call_count;
 
-        local = frame->local;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+			local->transaction.wind (frame, this, i);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
 
-        frame->root->pid = local->saved_pid;
+        return 0;
 }
 
 
-static void
-__mark_all_pending (int32_t *pending[], int child_count,
-                    afr_transaction_type type)
+int
+__afr_txn_write_done (call_frame_t *frame, xlator_t *this)
 {
-	int i;
-        int j;
-
-        for (i = 0; i < child_count; i++) {
-                j = afr_index_for_transaction_type (type);
-		pending[i][j] = hton32 (1);
-        }
-}
+        afr_local_t *local = NULL;
 
+        local = frame->local;
 
-static void
-__mark_child_dead (int32_t *pending[], int child_count, int child,
-                   afr_transaction_type type)
-{
-        int j;
+        local->transaction.unwind (frame, this);
 
-        j = afr_index_for_transaction_type (type);
+        AFR_STACK_DESTROY (frame);
 
-	pending[child][j] = 0;
+        return 0;
 }
 
 
-static void
-__mark_pre_op_done_on_fd (call_frame_t *frame, xlator_t *this, int child_index)
+call_frame_t*
+afr_transaction_detach_fop_frame (call_frame_t *frame)
 {
-        afr_local_t   *local = NULL;
-        afr_fd_ctx_t  *fd_ctx = NULL;
+        afr_local_t *   local = NULL;
+        call_frame_t   *fop_frame = 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);
+        LOCK (&frame->lock);
         {
-                if (local->transaction.type == AFR_DATA_TRANSACTION)
-                        fd_ctx->pre_op_done[child_index]++;
+                fop_frame = local->transaction.main_frame;
+                local->transaction.main_frame = NULL;
         }
-        UNLOCK (&local->fd->lock);
-out:
-        return;
+        UNLOCK (&frame->lock);
+
+        return fop_frame;
 }
 
 
 static void
-__mark_pre_op_undone_on_fd (call_frame_t *frame, xlator_t *this, int child_index)
+afr_save_lk_owner (call_frame_t *frame)
 {
-        afr_local_t   *local = NULL;
-        afr_fd_ctx_t  *fd_ctx = NULL;
+        afr_local_t * local = 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;
+        local->saved_lk_owner = frame->root->lk_owner;
 }
 
 
 static void
-__mark_down_children (int32_t *pending[], int child_count,
-                      unsigned char *child_up, afr_transaction_type type)
+afr_restore_lk_owner (call_frame_t *frame)
 {
-	int i;
-	int j;
+        afr_local_t * local = NULL;
 
-	for (i = 0; i < child_count; i++) {
-                j = afr_index_for_transaction_type (type);
+        local = frame->local;
 
-		if (!child_up[i])
-			pending[i][j] = 0;
-        }
+        frame->root->lk_owner = local->saved_lk_owner;
 }
 
-
-static void
-__mark_all_success (int32_t *pending[], int child_count,
-                    afr_transaction_type type)
+void
+__mark_all_success (call_frame_t *frame, xlator_t *this)
 {
+	afr_private_t *priv = NULL;
+	afr_local_t *local = NULL;
 	int i;
-        int j;
 
-	for (i = 0; i < child_count; i++) {
-                j = afr_index_for_transaction_type (type);
-		pending[i][j] = hton32 (-1);
-        }
+	local = frame->local;
+	priv = this->private;
+
+	for (i = 0; i < priv->child_count; i++) {
+		local->transaction.failed_subvols[i] = 0;
+	}
 }
 
 
-static int
-__changelog_enabled (afr_private_t *priv, afr_transaction_type type)
+int
+afr_transaction_perform_fop (call_frame_t *frame, xlator_t *this)
 {
-	int ret = 0;
+        afr_local_t     *local = NULL;
+        fd_t            *fd   = NULL;
 
-	switch (type) {
-	case AFR_DATA_TRANSACTION:
-		if (priv->data_change_log)
-			ret = 1;
-
-		break;
-
-	case AFR_METADATA_TRANSACTION:
-		if (priv->metadata_change_log)
-			ret = 1;
-
-		break;
-
-	case AFR_ENTRY_TRANSACTION:
-	case AFR_ENTRY_RENAME_TRANSACTION:
-		if (priv->entry_change_log)
-			ret = 1;
-
-		break;
-	}
+        local = frame->local;
+        fd    = local->fd;
+
+        /*  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;
+
+	if (local->pre_op_compat)
+		/* old mode, pre-op was done as afr_changelog_do()
+		   just now, before OP */
+		afr_changelog_pre_op_update (frame, this);
+
+        /* 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);
 
-	return ret;
+	return 0;
 }
 
 
 static int
-__changelog_needed_pre_op (call_frame_t *frame, xlator_t *this)
+__changelog_enabled (afr_private_t *priv, afr_transaction_type type)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
+        int ret = 0;
 
-	int op_ret   = 0;
+        switch (type) {
+        case AFR_DATA_TRANSACTION:
+                if (priv->data_change_log)
+                        ret = 1;
 
-	priv  = this->private;
-	local = frame->local;
+                break;
 
-	if (__changelog_enabled (priv, local->transaction.type)) {
-		switch (local->op) {
+        case AFR_METADATA_TRANSACTION:
+                if (priv->metadata_change_log)
+                        ret = 1;
 
-		case GF_FOP_WRITE:
-		case GF_FOP_FTRUNCATE:
-                        op_ret = 1;
-			break;
+                break;
 
-		case GF_FOP_FLUSH:
-			op_ret = 0;
-			break;
+        case AFR_ENTRY_TRANSACTION:
+        case AFR_ENTRY_RENAME_TRANSACTION:
+                if (priv->entry_change_log)
+                        ret = 1;
 
-		default:
-			op_ret = 1;
-		}
-	}
+                break;
+        }
 
-	return op_ret;
+        return ret;
 }
 
 
 static int
-__changelog_needed_post_op (call_frame_t *frame, xlator_t *this)
+__fop_changelog_needed (call_frame_t *frame, xlator_t *this)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-
-	int op_ret = 0;
-	afr_transaction_type type = -1;
+        afr_private_t * priv  = NULL;
+        afr_local_t   * local = NULL;
+        int op_ret = 0;
+        afr_transaction_type type = -1;
 
-	priv  = this->private;
-	local = frame->local;
-	type  = local->transaction.type;
+        priv  = this->private;
+        local = frame->local;
+        type  = local->transaction.type;
 
-	if (__changelog_enabled (priv, type)) {
+        if (__changelog_enabled (priv, type)) {
                 switch (local->op) {
 
                 case GF_FOP_WRITE:
@@ -276,617 +230,706 @@ __changelog_needed_post_op (call_frame_t *frame, xlator_t *this)
                 }
         }
 
-	return op_ret;
+        return op_ret;
 }
 
 
-static int
-afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending)
+int
+afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int **pending)
 {
-        int i;
+        int i = 0;
         int ret = 0;
+	int pending_zero[AFR_NUM_CHANGE_LOGS] = {0, };
 
         for (i = 0; i < priv->child_count; i++) {
+		if (!memcmp (pending_zero, pending[i], sizeof (pending_zero)))
+			/* don't set xattrs for non-pending servers */
+			continue;
+
                 ret = dict_set_static_bin (xattr, priv->pending_key[i],
-                                           pending[i], 3 * sizeof (int32_t));
+					   pending[i],
+					   AFR_NUM_CHANGE_LOGS * sizeof (int));
                 /* 3 = data+metadata+entry */
 
-                if (ret < 0)
-                        goto out;
+                if (ret)
+                        break;
         }
 
-out:
         return ret;
 }
 
-
-static int
-afr_set_piggyback_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending,
-                        afr_transaction_type type)
+int
+afr_lock_server_count (afr_private_t *priv, afr_transaction_type type)
 {
-        int i;
         int ret = 0;
-        int *arr = NULL;
-        int index = 0;
-
-        index = afr_index_for_transaction_type (type);
-
-        for (i = 0; i < priv->child_count; i++) {
-                arr = GF_CALLOC (3 * sizeof (int32_t), priv->child_count,
-                                 gf_afr_mt_char);
-                if (!arr) {
-                        ret = -1;
-                        goto out;
-                }
-
-                memcpy (arr, pending[i], 3 * sizeof (int32_t));
 
-                arr[index]++;
+        switch (type) {
+        case AFR_DATA_TRANSACTION:
+                ret = priv->child_count;
+                break;
 
-                ret = dict_set_bin (xattr, priv->pending_key[i],
-                                    arr, 3 * sizeof (int32_t));
-                /* 3 = data+metadata+entry */
+        case AFR_METADATA_TRANSACTION:
+                ret = priv->child_count;
+                break;
 
-                if (ret < 0)
-                        goto out;
+        case AFR_ENTRY_TRANSACTION:
+        case AFR_ENTRY_RENAME_TRANSACTION:
+                ret = priv->child_count;
+                break;
         }
 
-out:
         return ret;
 }
 
+/* {{{ pending */
+
 
 int
-afr_lock_server_count (afr_private_t *priv, afr_transaction_type type)
+afr_changelog_post_op_done (call_frame_t *frame, xlator_t *this)
 {
-	int ret = 0;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+        afr_internal_lock_t *int_lock = NULL;
 
-	switch (type) {
-	case AFR_DATA_TRANSACTION:
-		ret = priv->child_count;
-		break;
+	local = frame->local;
+	priv = this->private;
+        int_lock = &local->internal_lock;
 
-	case AFR_METADATA_TRANSACTION:
-		ret = priv->child_count;
-		break;
+	if (local->transaction.resume_stub) {
+		call_resume (local->transaction.resume_stub);
+		local->transaction.resume_stub = NULL;
+	}
 
-	case AFR_ENTRY_TRANSACTION:
-	case AFR_ENTRY_RENAME_TRANSACTION:
-		ret = priv->child_count;
-		break;
+	if (afr_lock_server_count (priv, local->transaction.type) == 0) {
+		local->transaction.done (frame, this);
+	} else {
+		int_lock->lock_cbk = local->transaction.done;
+		afr_unlock (frame, this);
 	}
 
-	return ret;
+	return 0;
 }
 
-/* {{{ pending */
 
-int32_t
-afr_changelog_post_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			   int32_t op_ret, int32_t op_errno, dict_t *xattr)
+afr_inodelk_t*
+afr_get_inodelk (afr_internal_lock_t *int_lock, char *dom)
 {
-        afr_internal_lock_t *int_lock = NULL;
-	afr_private_t       *priv     = NULL;
-	afr_local_t         *local    = NULL;
-        int                  child_index = 0;
-
-	int call_count = -1;
-
-	priv     = this->private;
-	local    = frame->local;
-        int_lock = &local->internal_lock;
-
-        child_index = (long) cookie;
+        afr_inodelk_t *inodelk = NULL;
+        int           i = 0;
 
-        if (op_ret == 1) {
+        for (i = 0; int_lock->inodelk[i].domain; i++) {
+                inodelk = &int_lock->inodelk[i];
+                if (strcmp (dom, inodelk->domain) == 0)
+                        return inodelk;
         }
+        return NULL;
+}
 
-        if (op_ret == 0) {
-                __mark_pre_op_undone_on_fd (frame, this, child_index);
+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:
+                inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+                locked_nodes = inodelk->locked_nodes;
+        break;
+
+        case AFR_ENTRY_TRANSACTION:
+        case AFR_ENTRY_RENAME_TRANSACTION:
+                /*Because same set of subvols participate in all lockee
+                 * entities*/
+                locked_nodes = int_lock->lockee[0].locked_nodes;
+        break;
         }
+        return locked_nodes;
+}
 
-	LOCK (&frame->lock);
-	{
-		call_count = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
 
-	if (call_count == 0) {
-                if (afr_lock_server_count (priv, local->transaction.type) == 0) {
-                        local->transaction.done (frame, this);
-                } else {
-                        int_lock->lock_cbk = local->transaction.done;
-                        afr_unlock (frame, this);
-                }
-	}
+int
+afr_changelog_call_count (afr_transaction_type type,
+			  unsigned char *pre_op_subvols,
+			  unsigned int child_count)
+{
+        int call_count = 0;
 
-	return 0;
+	call_count = AFR_COUNT(pre_op_subvols, child_count);
+
+        if (type == AFR_ENTRY_RENAME_TRANSACTION)
+                call_count *= 2;
+
+        return call_count;
 }
 
 
-void
-afr_update_read_child (call_frame_t *frame, xlator_t *this, inode_t *inode,
-                       afr_transaction_type type)
+gf_boolean_t
+afr_txn_nothing_failed (call_frame_t *frame, xlator_t *this)
 {
-        int           curr_read_child = -1;
-        int           new_read_child = -1;
-        afr_private_t   *priv = NULL;
-        afr_local_t  *local = NULL;
-        int         **pending = NULL;
-        int           idx = 0;
-
-        idx = afr_index_for_transaction_type (type);
+        afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        int i = 0;
 
-        priv = this->private;
         local = frame->local;
-        curr_read_child = afr_read_child (this, inode);
-        pending = local->pending;
+	priv = this->private;
 
-        if (pending[curr_read_child][idx] != 0)
-                return;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.failed_subvols[i])
+                        return _gf_false;
+        }
 
-        /* need to set new read_child */
-        for (new_read_child = 0; new_read_child < priv->child_count;
-             new_read_child++) {
+        return _gf_true;
+}
 
-                if (!priv->child_up[new_read_child])
-                        /* child is down */
-                        continue;
 
-                if (pending[new_read_child][idx] == 0)
-                        /* op just failed */
-                        continue;
+void
+afr_handle_symmetric_errors (call_frame_t *frame, xlator_t *this)
+{
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	int op_errno = 0;
+	int i_errno = 0;
+	gf_boolean_t matching_errors = _gf_true;
+	int i = 0;
 
-                break;
-        }
+	priv = this->private;
+	local = frame->local;
 
-        if (new_read_child == priv->child_count)
-                /* all children uneligible. leave as-is */
-                return;
+	for (i = 0; i < priv->child_count; i++) {
+		if (!local->replies[i].valid)
+			continue;
+		if (local->replies[i].op_ret != -1) {
+			/* Operation succeeded on at least on subvol,
+			   so it is not a failed-everywhere situation.
+			*/
+			matching_errors = _gf_false;
+			break;
+		}
+		i_errno = local->replies[i].op_errno;
+
+		if (i_errno == ENOTCONN) {
+			/* ENOTCONN is not a symmetric error. We do not
+			   know if the operation was performed on the
+			   backend or not.
+			*/
+			matching_errors = _gf_false;
+			break;
+		}
+
+		if (!op_errno) {
+			op_errno = i_errno;
+		} else if (op_errno != i_errno) {
+			/* Mismatching op_errno's */
+			matching_errors = _gf_false;
+			break;
+		}
+	}
 
-        afr_set_read_child (this, inode, new_read_child);
+	if (matching_errors)
+		__mark_all_success (frame, this);
 }
 
 
-int 
-afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
+int
+afr_changelog_post_op_now (call_frame_t *frame, xlator_t *this)
 {
-	afr_private_t * priv = this->private;
-        afr_internal_lock_t *int_lock = NULL;
-	int ret        = 0;
-	int i          = 0;
-	int call_count = 0;
-
-	afr_local_t *  local = NULL;
-        afr_fd_ctx_t  *fdctx = NULL;
-	dict_t        **xattr = NULL;
-        int            piggyback = 0;
-        int            index = 0;
+        afr_private_t * priv = this->private;
+        int i          = 0;
+	int ret = 0;
+	int idx = 0;
+        afr_local_t *  local = NULL;
+        dict_t        *xattr = NULL;
         int            nothing_failed = 1;
+	gf_boolean_t   need_undirty = _gf_false;
 
-	local    = frame->local;
-        int_lock = &local->internal_lock;
-
-	__mark_down_children (local->pending, priv->child_count,
-                              local->child_up, local->transaction.type);
+        local = frame->local;
+	idx = afr_index_for_transaction_type (local->transaction.type);
 
-        if (local->fd)
-                afr_update_read_child (frame, this, local->fd->inode,
-                                       local->transaction.type);
+        nothing_failed = afr_txn_nothing_failed (frame, this);
 
-        xattr = alloca (priv->child_count * sizeof (*xattr));
-        memset (xattr, 0, (priv->child_count * sizeof (*xattr)));
-	for (i = 0; i < priv->child_count; i++) {
-                xattr[i] = get_new_dict ();
-                dict_ref (xattr[i]);
-        }
+	if (afr_changelog_pre_op_uninherit (frame, this))
+		need_undirty = _gf_false;
+	else
+		need_undirty = _gf_true;
 
-        call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-        if (local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
-                call_count *= 2;
-        }
+	if (nothing_failed && !need_undirty) {
+		afr_changelog_post_op_done (frame, this);
+                goto out;
+	}
 
-	local->call_count = call_count;
+	xattr = dict_new ();
+	if (!xattr) {
+		local->op_ret = -1;
+		local->op_errno = ENOMEM;
+		afr_changelog_post_op_done (frame, this);
+		goto out;
+	}
 
-        if (local->fd)
-                fdctx = afr_fd_ctx_get (local->fd, this);
+	if (need_undirty) {
+		local->dirty[idx] = hton32(-1);
 
-	if (call_count == 0) {
-		/* no child is up */
-                for (i = 0; i < priv->child_count; i++) {
-                        dict_unref (xattr[i]);
-                }
+		ret = dict_set_static_bin (xattr, AFR_DIRTY, local->dirty,
+					   sizeof(int) * AFR_NUM_CHANGE_LOGS);
+		if (ret) {
+			local->op_ret = -1;
+			local->op_errno = ENOMEM;
+			afr_changelog_post_op_done (frame, this);
+			goto out;
+		}
 
-                int_lock->lock_cbk = local->transaction.done;
-		afr_unlock (frame, this);
-		return 0;
 	}
 
-        /* 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;
-                }
-        }
+	if (!nothing_failed) {
+		for (i = 0; i < priv->child_count; i++) {
+			if (local->transaction.failed_subvols[i])
+				local->pending[i][idx] = hton32(1);
+		}
+		ret = afr_set_pending_dict (priv, xattr, local->pending);
+		if (ret < 0) {
+			local->op_ret = -1;
+			local->op_errno = ENOMEM;
+			afr_changelog_post_op_done (frame, this);
+			goto out;
+		}
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (!local->child_up[i])
-                        continue;
+	}
 
-                ret = afr_set_pending_dict (priv, xattr[i],
-                                            local->pending);
+	afr_changelog_do (frame, this, xattr, afr_changelog_post_op_done);
+out:
+	if (xattr)
+                dict_unref (xattr);
 
-                if (ret < 0)
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "failed to set pending entry");
+        return 0;
+}
 
 
-                switch (local->transaction.type) {
-                case AFR_DATA_TRANSACTION:
-                {
-                        if (!fdctx) {
-                                STACK_WIND (frame, afr_changelog_post_op_cbk,
-                                            priv->children[i],
-                                            priv->children[i]->fops->xattrop,
-                                            &local->loc,
-                                            GF_XATTROP_ADD_ARRAY, xattr[i]);
-                                break;
-                        }
+gf_boolean_t
+afr_changelog_pre_op_uninherit (call_frame_t *frame, xlator_t *this)
+{
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	fd_t *fd = NULL;
+	int i = 0;
+	gf_boolean_t ret = _gf_false;
+	afr_fd_ctx_t *fd_ctx = NULL;
+	int type = 0;
 
-                        LOCK (&local->fd->lock);
-                        {
-                                piggyback = 0;
-                                if (fdctx->pre_op_piggyback[i]) {
-                                        fdctx->pre_op_piggyback[i]--;
-                                        piggyback = 1;
-                                }
-                        }
-                        UNLOCK (&local->fd->lock);
+	local = frame->local;
+	priv = this->private;
+	fd = local->fd;
 
-                        if (piggyback && !nothing_failed)
-                                ret = afr_set_piggyback_dict (priv, xattr[i],
-                                                              local->pending,
-                                                              local->transaction.type);
+	type = afr_index_for_transaction_type (local->transaction.type);
+	if (type != AFR_DATA_TRANSACTION)
+		return !local->transaction.dirtied;
 
-                        if (nothing_failed && piggyback) {
-                                afr_changelog_post_op_cbk (frame, (void *)(long)i,
-                                                          this, 1, 0, xattr[i]);
-                        } else {
-                                STACK_WIND_COOKIE (frame,
-                                                   afr_changelog_post_op_cbk,
-                                                   (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->fxattrop,
-                                                   local->fd,
-                                                   GF_XATTROP_ADD_ARRAY, xattr[i]);
-                        }
-                }
-                break;
-                case AFR_METADATA_TRANSACTION:
-                {
-                        if (local->fd)
-                                STACK_WIND (frame, afr_changelog_post_op_cbk,
-                                            priv->children[i],
-                                            priv->children[i]->fops->fxattrop,
-                                            local->fd,
-                                            GF_XATTROP_ADD_ARRAY, xattr[i]);
-                        else
-                                STACK_WIND (frame, afr_changelog_post_op_cbk,
-                                            priv->children[i],
-                                            priv->children[i]->fops->xattrop,
-                                            &local->loc,
-                                            GF_XATTROP_ADD_ARRAY, xattr[i]);
-                }
-                break;
+	if (!fd)
+		return !local->transaction.dirtied;
 
-                case AFR_ENTRY_RENAME_TRANSACTION:
-                {
-                        STACK_WIND_COOKIE (frame, afr_changelog_post_op_cbk,
-                                           (void *) (long) i,
-                                           priv->children[i],
-                                           priv->children[i]->fops->xattrop,
-                                           &local->transaction.new_parent_loc,
-                                           GF_XATTROP_ADD_ARRAY, xattr[i]);
-                        call_count--;
-                }
+	fd_ctx = afr_fd_ctx_get (fd, this);
+	if (!fd_ctx)
+		return _gf_false;
 
-                /*
-                  set it again because previous stack_wind
-                  might have already returned (think of case
-                  where subvolume is posix) and would have
-                  used the dict as placeholder for return
-                  value
-                */
+	if (local->transaction.no_uninherit)
+		return _gf_false;
 
-                ret = afr_set_pending_dict (priv, xattr[i],
-                                            local->pending);
+	/* This function must be idempotent. So check if we
+	   were called before and return the same answer again.
 
-                if (ret < 0)
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "failed to set pending entry");
+	   It is important to keep this function idempotent for
+	   the call in afr_changelog_post_op_safe() to not have
+	   side effects on the call from afr_changelog_post_op_now()
+	*/
+	if (local->transaction.uninherit_done)
+		return local->transaction.uninherit_value;
 
-                /* fall through */
+	LOCK(&fd->lock);
+	{
+		for (i = 0; i < priv->child_count; i++) {
+			if (local->transaction.pre_op[i] !=
+			    fd_ctx->pre_op_done[type][i]) {
+				ret = !local->transaction.dirtied;
+				goto unlock;
+			}
+		}
 
-                case AFR_ENTRY_TRANSACTION:
-                {
-                        if (local->fd)
-                                STACK_WIND (frame, afr_changelog_post_op_cbk,
-                                            priv->children[i],
-                                            priv->children[i]->fops->fxattrop,
-                                            local->fd,
-                                            GF_XATTROP_ADD_ARRAY, xattr[i]);
-                        else
-                                STACK_WIND (frame, afr_changelog_post_op_cbk,
-                                            priv->children[i],
-                                            priv->children[i]->fops->xattrop,
-                                            &local->transaction.parent_loc,
-                                            GF_XATTROP_ADD_ARRAY, xattr[i]);
-                }
-                break;
-                }
+		if (fd_ctx->inherited[type]) {
+			ret = _gf_true;
+			fd_ctx->inherited[type]--;
+		} else if (fd_ctx->on_disk[type]) {
+			ret = _gf_false;
+			fd_ctx->on_disk[type]--;
+		} else {
+			/* ASSERT */
+			ret = _gf_false;
+		}
 
-                if (!--call_count)
-                        break;
+		if (!fd_ctx->inherited[type] && !fd_ctx->on_disk[type]) {
+			for (i = 0; i < priv->child_count; i++)
+				fd_ctx->pre_op_done[type][i] = 0;
+		}
 	}
+unlock:
+	UNLOCK(&fd->lock);
 
-        for (i = 0; i < priv->child_count; i++) {
-                dict_unref (xattr[i]);
-        }
+	local->transaction.uninherit_done = _gf_true;
+	local->transaction.uninherit_value = ret;
 
-	return 0;
+	return ret;
 }
 
 
-int32_t
-afr_changelog_pre_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                          int32_t op_ret, int32_t op_errno, dict_t *xattr)
+gf_boolean_t
+afr_changelog_pre_op_inherit (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = this->private;
-	loc_t       *   loc   = NULL;
-
-	int call_count  = -1;
-	int child_index = (long) cookie;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	fd_t *fd = NULL;
+	int i = 0;
+	gf_boolean_t ret = _gf_false;
+	afr_fd_ctx_t *fd_ctx = NULL;
+	int type = 0;
 
 	local = frame->local;
-	loc   = &local->loc;
+	priv = this->private;
+	fd = local->fd;
 
-	LOCK (&frame->lock);
-	{
-                if (op_ret == 1) {
-                        /* special op_ret for piggyback */
-                }
+	if (local->transaction.type != AFR_DATA_TRANSACTION)
+		return _gf_false;
 
-                if (op_ret == 0) {
-                        __mark_pre_op_done_on_fd (frame, this, child_index);
-                }
+	type = afr_index_for_transaction_type (local->transaction.type);
+
+	if (!fd)
+		return _gf_false;
 
-		if (op_ret == -1) {
-			local->child_up[child_index] = 0;
+	fd_ctx = afr_fd_ctx_get (fd, this);
+	if (!fd_ctx)
+		return _gf_false;
 
-			if (op_errno == ENOTSUP) {
-				gf_log (this->name, GF_LOG_ERROR,
-					"xattrop not supported by %s",
-					priv->children[child_index]->name);
-				local->op_ret = -1;
+	LOCK(&fd->lock);
+	{
+		if (!fd_ctx->on_disk[type]) {
+			/* nothing to inherit yet */
+			ret = _gf_false;
+			goto unlock;
+		}
 
-			} else if (!child_went_down (op_ret, op_errno)) {
-				gf_log (this->name, GF_LOG_ERROR,
-					"xattrop failed on child %s: %s",
-					priv->children[child_index]->name,
-					strerror (op_errno));
+		for (i = 0; i < priv->child_count; i++) {
+			if (local->transaction.pre_op[i] !=
+			    fd_ctx->pre_op_done[type][i]) {
+				/* either inherit exactly, or don't */
+				ret = _gf_false;
+				goto unlock;
 			}
-			local->op_errno = op_errno;
 		}
 
-		call_count = --local->call_count;
+		fd_ctx->inherited[type]++;
+
+		ret = _gf_true;
+
+		local->transaction.inherited = _gf_true;
 	}
-	UNLOCK (&frame->lock);
+unlock:
+	UNLOCK(&fd->lock);
 
-	if (call_count == 0) {
-		if ((local->op_ret == -1) &&
-		    (local->op_errno == ENOTSUP)) {
-			local->transaction.resume (frame, this);
-		} else {
-                        __mark_all_success (local->pending, priv->child_count,
-                                            local->transaction.type);
+	return ret;
+}
 
-                        afr_pid_restore (frame);
 
-			local->transaction.fop (frame, this);
+gf_boolean_t
+afr_changelog_pre_op_update (call_frame_t *frame, xlator_t *this)
+{
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	fd_t *fd = NULL;
+	afr_fd_ctx_t *fd_ctx = NULL;
+	int i = 0;
+	gf_boolean_t ret = _gf_false;
+	int type = 0;
+
+	local = frame->local;
+	priv = this->private;
+	fd = local->fd;
+
+	if (!fd)
+		return _gf_false;
+
+	fd_ctx = afr_fd_ctx_get (fd, this);
+	if (!fd_ctx)
+		return _gf_false;
+
+	if (local->transaction.inherited)
+		/* was already inherited in afr_changelog_pre_op */
+		return _gf_false;
+
+	if (!local->transaction.dirtied)
+		return _gf_false;
+
+        if (!afr_txn_nothing_failed (frame, this))
+		return _gf_false;
+
+	type = afr_index_for_transaction_type (local->transaction.type);
+
+	ret = _gf_false;
+
+	LOCK(&fd->lock);
+	{
+		if (!fd_ctx->on_disk[type]) {
+			for (i = 0; i < priv->child_count; i++)
+				fd_ctx->pre_op_done[type][i] =
+					local->transaction.pre_op[i];
+		} else {
+			for (i = 0; i < priv->child_count; i++)
+				if (fd_ctx->pre_op_done[type][i] !=
+				    local->transaction.pre_op[i]) {
+					local->transaction.no_uninherit = 1;
+					goto unlock;
+				}
 		}
+		fd_ctx->on_disk[type]++;
+
+		ret = _gf_true;
 	}
+unlock:
+	UNLOCK(&fd->lock);
 
-	return 0;
+	return ret;
 }
 
 
 int
-afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
+afr_changelog_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		   int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
 {
-	afr_private_t * priv = this->private;
-	int i = 0;
-	int ret = 0;
-	int call_count = 0;
-	dict_t **xattr = NULL;
-        afr_fd_ctx_t *fdctx = NULL;
+        afr_local_t *local = NULL;
+        int call_count = -1;
 
-	afr_local_t *local = NULL;
-        int          piggyback = 0;
+        local = frame->local;
 
-	local = frame->local;
+	if (op_ret == -1)
+		afr_transaction_fop_failed (frame, this, (long) cookie);
 
-        xattr = alloca (priv->child_count * sizeof (*xattr));
-        memset (xattr, 0, (priv->child_count * sizeof (*xattr)));
+	call_count = afr_frame_return (frame);
 
-	for (i = 0; i < priv->child_count; i++) {
-                xattr[i] = get_new_dict ();
-                dict_ref (xattr[i]);
-        }
+        if (call_count == 0)
+		local->transaction.changelog_resume (frame, this);
 
-	call_count = afr_up_children_count (priv->child_count,
-					    local->child_up);
+        return 0;
+}
 
-	if (local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
-		call_count *= 2;
-	}
 
-	if (call_count == 0) {
-		/* no child is up */
-                for (i = 0; i < priv->child_count; i++) {
-                        dict_unref (xattr[i]);
-                }
+int
+afr_changelog_do (call_frame_t *frame, xlator_t *this, dict_t *xattr,
+		  afr_changelog_resume_t changelog_resume)
+{
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	int i = 0;
+	int call_count = 0;
 
-                local->internal_lock.lock_cbk =
-                        local->transaction.done;
-		afr_unlock (frame, this);
+	local = frame->local;
+	priv = this->private;
+
+        call_count = afr_changelog_call_count (local->transaction.type,
+					       local->transaction.pre_op,
+					       priv->child_count);
+
+	if (call_count == 0) {
+		changelog_resume (frame, this);
 		return 0;
 	}
 
 	local->call_count = call_count;
 
-	__mark_all_pending (local->pending, priv->child_count,
-                            local->transaction.type);
-
-        if (local->fd)
-                fdctx = afr_fd_ctx_get (local->fd, this);
+	local->transaction.changelog_resume = changelog_resume;
 
-	for (i = 0; i < priv->child_count; i++) {
-                if (!local->child_up[i])
+        for (i = 0; i < priv->child_count; i++) {
+                if (!local->transaction.pre_op[i])
                         continue;
-                ret = afr_set_pending_dict (priv, xattr[i],
-                                            local->pending);
-
-                if (ret < 0)
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "failed to set pending entry");
-
 
                 switch (local->transaction.type) {
                 case AFR_DATA_TRANSACTION:
-                {
-                        if (!fdctx) {
-                                STACK_WIND_COOKIE (frame,
-                                                   afr_changelog_pre_op_cbk,
-                                                   (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->xattrop,
-                                                   &(local->loc),
-                                                   GF_XATTROP_ADD_ARRAY, xattr[i]);
-                                break;
+                case AFR_METADATA_TRANSACTION:
+                        if (!local->fd) {
+                                STACK_WIND_COOKIE (frame, afr_changelog_cbk,
+						   (void *) (long) i,
+						   priv->children[i],
+						   priv->children[i]->fops->xattrop,
+						   &local->loc,
+						   GF_XATTROP_ADD_ARRAY, xattr,
+						   NULL);
+                        } else {
+                                STACK_WIND_COOKIE (frame, afr_changelog_cbk,
+						   (void *) (long) i,
+						   priv->children[i],
+						   priv->children[i]->fops->fxattrop,
+						   local->fd,
+						   GF_XATTROP_ADD_ARRAY, xattr,
+						   NULL);
                         }
+			break;
+                case AFR_ENTRY_RENAME_TRANSACTION:
 
-                        LOCK (&local->fd->lock);
-                        {
-                                piggyback = 0;
-                                if (fdctx->pre_op_done[i]) {
-                                        fdctx->pre_op_piggyback[i]++;
-                                        piggyback = 1;
-                                        fdctx->hit++;
-                                } else {
-                                        fdctx->miss++;
-                                }
-                        }
-                        UNLOCK (&local->fd->lock);
+			STACK_WIND_COOKIE (frame, afr_changelog_cbk,
+					   (void *) (long) i,
+					   priv->children[i],
+					   priv->children[i]->fops->xattrop,
+					   &local->transaction.new_parent_loc,
+					   GF_XATTROP_ADD_ARRAY, xattr,
+					   NULL);
+                        call_count--;
 
-                        if (piggyback)
-                                afr_changelog_pre_op_cbk (frame, (void *)(long)i,
-                                                          this, 1, 0, xattr[i]);
-                        else
-                                STACK_WIND_COOKIE (frame,
-                                                   afr_changelog_pre_op_cbk,
-                                                   (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->fxattrop,
-                                                   local->fd,
-                                                   GF_XATTROP_ADD_ARRAY, xattr[i]);
-                }
-                break;
-                case AFR_METADATA_TRANSACTION:
-                {
+                /* fall through */
+
+                case AFR_ENTRY_TRANSACTION:
                         if (local->fd)
-                                STACK_WIND_COOKIE (frame,
-                                                   afr_changelog_pre_op_cbk,
-                                                   (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->fxattrop,
-                                                   local->fd,
-                                                   GF_XATTROP_ADD_ARRAY, xattr[i]);
+                                STACK_WIND_COOKIE (frame, afr_changelog_cbk,
+						   (void *) (long) i,
+						   priv->children[i],
+						   priv->children[i]->fops->fxattrop,
+						   local->fd,
+						   GF_XATTROP_ADD_ARRAY, xattr,
+						   NULL);
                         else
-                                STACK_WIND_COOKIE (frame,
-                                                   afr_changelog_pre_op_cbk,
-                                                   (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->xattrop,
-                                                   &(local->loc),
-                                                   GF_XATTROP_ADD_ARRAY, xattr[i]);
-                }
-                break;
+                                STACK_WIND_COOKIE (frame, afr_changelog_cbk,
+						   (void *) (long) i,
+						   priv->children[i],
+						   priv->children[i]->fops->xattrop,
+						   &local->transaction.parent_loc,
+						   GF_XATTROP_ADD_ARRAY, xattr,
+						   NULL);
+			break;
+		}
 
-                case AFR_ENTRY_RENAME_TRANSACTION:
-                {
-                        STACK_WIND_COOKIE (frame,
-                                           afr_changelog_pre_op_cbk,
-                                           (void *) (long) i,
-                                           priv->children[i],
-                                           priv->children[i]->fops->xattrop,
-                                           &local->transaction.new_parent_loc,
-                                           GF_XATTROP_ADD_ARRAY, xattr[i]);
+                if (!--call_count)
+                        break;
+        }
 
-                        call_count--;
-                }
+	return 0;
+}
 
 
-                /*
-                  set it again because previous stack_wind
-                  might have already returned (think of case
-                  where subvolume is posix) and would have
-                  used the dict as placeholder for return
-                  value
-                */
+int
+afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
+{
+        afr_private_t * priv = this->private;
+        int i = 0;
+        int ret = 0;
+        int call_count = 0;
+	int op_errno = 0;
+        afr_local_t *local = NULL;
+        afr_internal_lock_t *int_lock = NULL;
+        unsigned char       *locked_nodes = NULL;
+	unsigned char       *pending_subvols = NULL;
+	int idx = -1;
+	gf_boolean_t pre_nop = _gf_true;
+	dict_t *xdata_req = NULL;
 
-                ret = afr_set_pending_dict (priv, xattr[i],
-                                            local->pending);
+        local = frame->local;
+        int_lock = &local->internal_lock;
+	idx = afr_index_for_transaction_type (local->transaction.type);
 
-                if (ret < 0)
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "failed to set pending entry");
+        locked_nodes = afr_locked_nodes_get (local->transaction.type, int_lock);
 
-                /* fall through */
+	pending_subvols = alloca0 (priv->child_count);
 
-                case AFR_ENTRY_TRANSACTION:
-                {
-                        if (local->fd)
-                                STACK_WIND_COOKIE (frame,
-                                                   afr_changelog_pre_op_cbk,
-                                                   (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->fxattrop,
-                                                   local->fd,
-                                                   GF_XATTROP_ADD_ARRAY, xattr[i]);
-                        else
-                                STACK_WIND_COOKIE (frame,
-                                                   afr_changelog_pre_op_cbk,
-                                                   (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->xattrop,
-                                                   &local->transaction.parent_loc,
-                                                   GF_XATTROP_ADD_ARRAY, xattr[i]);
-                }
-                break;
-                }
+	for (i = 0; i < priv->child_count; i++) {
+		if (locked_nodes[i]) {
+			local->transaction.pre_op[i] = 1;
+			call_count++;
+		} else {
+			pending_subvols[i] = 1;
+		}
+	}
 
-                if (!--call_count)
-                        break;
+	/* TBD: quorum check w/ call_count */
+
+        if (call_count == 0) {
+		op_errno = ENOTCONN;
+		goto err;
 	}
 
-        for (i = 0; i < priv->child_count; i++) {
-                dict_unref (xattr[i]);
-        }
+	xdata_req = dict_new();
+	if (!xdata_req) {
+		op_errno = ENOMEM;
+		goto err;
+	}
+
+	pre_nop = _gf_true;
+
+	if (afr_changelog_pre_op_inherit (frame, this))
+		goto next;
+
+	if (call_count < priv->child_count) {
+		/* For subvols we are not performing operation on,
+		   mark them as pending up-front along with the FOP
+		   so that we can safely defer unmarking dirty until
+		   later.
+		*/
+		for (i = 0; i < priv->child_count; i++) {
+			if (pending_subvols[i])
+				local->pending[i][idx] = hton32(1);
+		}
+		ret = afr_set_pending_dict (priv, xdata_req,
+					    local->pending);
+		if (ret < 0) {
+			op_errno = ENOMEM;
+			goto err;
+		}
+		pre_nop = _gf_false;
+	}
+
+	if (call_count > 1 &&
+	    (local->transaction.type == AFR_DATA_TRANSACTION ||
+	     !local->optimistic_change_log)) {
+
+		/* If we are performing change on only one subvol, no
+		   need to mark dirty, because we are setting the pending
+		   counts already anyways
+		*/
+		local->dirty[idx] = hton32(1);
+
+		ret = dict_set_static_bin (xdata_req, AFR_DIRTY, local->dirty,
+					   sizeof(int) * AFR_NUM_CHANGE_LOGS);
+		if (ret) {
+			op_errno = ENOMEM;
+			goto err;
+		}
+
+		pre_nop = _gf_false;
+		local->transaction.dirtied = 1;
+	}
+
+	if (pre_nop)
+		goto next;
+
+	if (!local->pre_op_compat) {
+		dict_copy (xdata_req, local->xdata_req);
+		goto next;
+	}
+
+	afr_changelog_do (frame, this, xdata_req, afr_transaction_perform_fop);
+
+	if (xdata_req)
+		dict_unref (xdata_req);
+
+	return 0;
+next:
+	afr_transaction_perform_fop (frame, this);
+
+	if (xdata_req)
+		dict_unref (xdata_req);
+
+        return 0;
+err:
+	local->internal_lock.lock_cbk = local->transaction.done;
+	local->op_ret = -1;
+	local->op_errno = op_errno;
+
+	afr_unlock (frame, this);
+
+	if (xdata_req)
+		dict_unref (xdata_req);
 
 	return 0;
 }
@@ -902,9 +945,9 @@ afr_post_blocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
         int_lock = &local->internal_lock;
 
         if (int_lock->lock_op_ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
+                gf_log (this->name, GF_LOG_INFO,
                         "Blocking inodelks failed.");
-		local->transaction.done (frame, this);
+                local->transaction.done (frame, this);
         } else {
 
                 gf_log (this->name, GF_LOG_DEBUG,
@@ -952,9 +995,9 @@ afr_post_blocking_entrylk_cbk (call_frame_t *frame, xlator_t *this)
         int_lock = &local->internal_lock;
 
         if (int_lock->lock_op_ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
+                gf_log (this->name, GF_LOG_INFO,
                         "Blocking entrylks failed.");
-		local->transaction.done (frame, this);
+                local->transaction.done (frame, this);
         } else {
 
                 gf_log (this->name, GF_LOG_DEBUG,
@@ -1002,9 +1045,9 @@ afr_post_blocking_rename_cbk (call_frame_t *frame, xlator_t *this)
         int_lock = &local->internal_lock;
 
         if (int_lock->lock_op_ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
+                gf_log (this->name, GF_LOG_INFO,
                         "Blocking entrylks failed.");
-		local->transaction.done (frame, this);
+                local->transaction.done (frame, this);
         } else {
 
                 gf_log (this->name, GF_LOG_DEBUG,
@@ -1037,12 +1080,14 @@ 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;
 }
@@ -1057,24 +1102,25 @@ afr_lock_rec (call_frame_t *frame, xlator_t *this)
         int_lock = &local->internal_lock;
 
         int_lock->transaction_lk_type = AFR_TRANSACTION_LK;
+        int_lock->domain = this->name;
 
-	switch (local->transaction.type) {
-	case AFR_DATA_TRANSACTION:
-	case AFR_METADATA_TRANSACTION:
+        switch (local->transaction.type) {
+        case AFR_DATA_TRANSACTION:
+        case AFR_METADATA_TRANSACTION:
                 afr_set_transaction_flock (local);
 
                 int_lock->lock_cbk = afr_post_nonblocking_inodelk_cbk;
 
                 afr_nonblocking_inodelk (frame, this);
-		break;
+                break;
 
         case AFR_ENTRY_RENAME_TRANSACTION:
 
-                int_lock->lock_cbk = afr_post_blocking_rename_cbk;
-                afr_blocking_lock (frame, this);
+                int_lock->lock_cbk = afr_post_nonblocking_entrylk_cbk;
+                afr_nonblocking_entrylk (frame, this);
                 break;
 
-	case AFR_ENTRY_TRANSACTION:
+        case AFR_ENTRY_TRANSACTION:
                 int_lock->lk_basename = local->transaction.basename;
                 if (&local->transaction.parent_loc)
                         int_lock->lk_loc = &local->transaction.parent_loc;
@@ -1093,15 +1139,9 @@ afr_lock_rec (call_frame_t *frame, xlator_t *this)
 int
 afr_lock (call_frame_t *frame, xlator_t *this)
 {
-        afr_pid_save (frame);
-
-        frame->root->pid = (long) frame->root;
-
-        afr_set_lk_owner (frame, this);
-
         afr_set_lock_number (frame, this);
 
-	return afr_lock_rec (frame, this);
+        return afr_lock_rec (frame, this);
 }
 
 
@@ -1110,90 +1150,573 @@ afr_lock (call_frame_t *frame, xlator_t *this)
 int
 afr_internal_lock_finish (call_frame_t *frame, xlator_t *this)
 {
-        afr_local_t   *local = NULL;
-        afr_private_t *priv  = NULL;
+        if (__fop_changelog_needed (frame, this)) {
+                afr_changelog_pre_op (frame, this);
+        } else {
+                afr_transaction_perform_fop (frame, this);
+        }
+
+        return 0;
+}
+
+
+void
+afr_set_delayed_post_op (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t    *local = NULL;
+        afr_private_t  *priv = NULL;
+
+        /* call this function from any of the related optimizations
+           which benefit from delaying post op are enabled, namely:
+
+           - changelog piggybacking
+           - eager locking
+        */
+
+        priv = this->private;
+        if (!priv)
+                return;
+
+        if (!priv->post_op_delay_secs)
+                return;
 
-        priv  = this->private;
         local = frame->local;
+        if (!local->transaction.eager_lock_on)
+                return;
 
-        if (__changelog_needed_pre_op (frame, this)) {
-                afr_changelog_pre_op (frame, this);
-        } else {
-                __mark_all_success (local->pending, priv->child_count,
-                                    local->transaction.type);
+        if (!local)
+                return;
 
-                afr_pid_restore (frame);
+        if (!local->fd)
+                return;
+
+        if (local->op == GF_FOP_WRITE)
+                local->delayed_post_op = _gf_true;
+}
 
-                local->transaction.fop (frame, this);
+gf_boolean_t
+afr_are_multiple_fds_opened (fd_t *fd, xlator_t *this)
+{
+        afr_fd_ctx_t *fd_ctx = NULL;
+
+        if (!fd) {
+                /* 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 fd");
+                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
+         */
+
+        fd_ctx = afr_fd_ctx_get (fd, this);
+        if (!fd_ctx)
+                return _gf_true;
+
+        if (fd_ctx->open_fd_count > 1)
+                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;
+
+        local = frame->local;
+        if (!local)
+                goto out;
+
+        if (!local->delayed_post_op)
+                goto out;
+
+        //Mark pending changelog ASAP
+        if (!afr_txn_nothing_failed (frame, this))
+                goto out;
+
+        if (local->fd && afr_are_multiple_fds_opened (local->fd, this))
+                goto out;
+
+        res = _gf_true;
+out:
+        return res;
+}
+
+
+void
+afr_delayed_changelog_post_op (xlator_t *this, call_frame_t *frame, fd_t *fd,
+                               call_stub_t *stub);
+
+void
+afr_delayed_changelog_wake_up_cbk (void *data)
+{
+        fd_t           *fd = NULL;
+
+        fd = data;
+
+        afr_delayed_changelog_wake_up (THIS, fd);
+}
+
+
+/* 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_transaction_resume (call_frame_t *frame, xlator_t *this)
+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_internal_lock_t *int_lock = NULL;
-	afr_local_t         *local    = NULL;
-	afr_private_t       *priv     = NULL;
+        afr_private_t *priv = NULL;
+        int child_index = (long) cookie;
+        int call_count = -1;
+        afr_local_t *local = NULL;
 
-	local    = frame->local;
-        int_lock = &local->internal_lock;
-	priv     = this->private;
+        priv = this->private;
+        local = frame->local;
 
-	if (__changelog_needed_post_op (frame, this)) {
-		afr_changelog_post_op (frame, this);
-	} else {
-		if (afr_lock_server_count (priv, local->transaction.type) == 0) {
-			local->transaction.done (frame, this);
-		} else {
-                        int_lock->lock_cbk = local->transaction.done;
-			afr_unlock (frame, this);
-		}
+        if (op_ret != 0) {
+                /* 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_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 (afr_changelog_pre_op_uninherit (frame, this) &&
+	    afr_txn_nothing_failed (frame, this)) {
+                /* just detected that this post-op is about to
+                   be optimized away as a new write() has
+                   already piggybacked on this frame's changelog.
+                   */
+                afr_changelog_post_op_now (frame, this);
+                return 0;
+        }
+
+        /* Calling afr_changelog_post_op_now() now will result in
+           issuing ->[f]xattrop().
+
+           Performing a hard POST-OP (->[f]xattrop() FOP) is a more
+           responsible operation that what it might appear on the surface.
+
+           The changelog of a file (in the xattr of the file on the server)
+           stores information (pending count) about the state of the file
+           on the OTHER server. This changelog is blindly trusted, and must
+           therefore be updated in such a way it remains trustworthy. This
+           implies that decrementing the pending count (essentially "clearing
+           the dirty flag") must be done STRICTLY after we are sure that the
+           operation on the other server has reached stable storage.
+
+           While the backend filesystem on that server will eventually flush
+           it to stable storage, we (being in userspace) have no mechanism
+           to get notified when the write became "stable".
+
+           This means we need take matter into our own hands and issue an
+           fsync() EVEN IF THE APPLICATION WAS PERFORMING UNSTABLE WRITES,
+           and get an acknowledgement for it. And we need to wait for the
+           fsync() acknowledgement before initiating the hard POST-OP.
+
+           However if the FD itself was opened in O_SYNC or O_DSYNC then
+           we are already guaranteed that the writes were made stable as
+           part of the FOP itself. The same holds true for NFS stable
+           writes which happen on an anonymous FD with O_DSYNC or O_SYNC
+           flag set in the writev() @flags param. For all other write types,
+           mark a flag in the fdctx whenever an unstable write is witnessed.
+           */
+
+        if (!afr_fd_has_witnessed_unstable_write (this, local->fd)) {
+                afr_changelog_post_op_now (frame, this);
+                return 0;
+        }
+
+        /* Check whether users want durability and perform fsync/post-op
+         * accordingly.
+         */
+        if (priv->ensure_durability) {
+                /* Time to fsync() */
+                afr_changelog_fsync (frame, this);
+        } else {
+                afr_changelog_post_op_now (frame, this);
+        }
+
+        return 0;
+}
+
+
+void
+afr_delayed_changelog_post_op (xlator_t *this, call_frame_t *frame, fd_t *fd,
+                               call_stub_t *stub)
+{
+	afr_fd_ctx_t      *fd_ctx = NULL;
+	call_frame_t      *prev_frame = NULL;
+	struct timespec    delta = {0, };
+	afr_private_t     *priv = NULL;
+	afr_local_t       *local = NULL;
+
+	priv = this->private;
+
+	fd_ctx = afr_fd_ctx_get (fd, this);
+	if (!fd_ctx)
+                goto out;
+
+	delta.tv_sec = priv->post_op_delay_secs;
+	delta.tv_nsec = 0;
+
+	pthread_mutex_lock (&fd_ctx->delay_lock);
+	{
+		prev_frame = fd_ctx->delay_frame;
+		fd_ctx->delay_frame = NULL;
+		if (fd_ctx->delay_timer)
+			gf_timer_call_cancel (this->ctx, fd_ctx->delay_timer);
+		fd_ctx->delay_timer = NULL;
+		if (!frame)
+			goto unlock;
+		fd_ctx->delay_timer = gf_timer_call_after (this->ctx, delta,
+							   afr_delayed_changelog_wake_up_cbk,
+							   fd);
+		fd_ctx->delay_frame = frame;
+	}
+unlock:
+	pthread_mutex_unlock (&fd_ctx->delay_lock);
+
+out:
+	if (prev_frame) {
+		local = prev_frame->local;
+		local->transaction.resume_stub = stub;
+		afr_changelog_post_op_now (prev_frame, this);
+	} else if (stub) {
+		call_resume (stub);
+	}
+}
+
+
+void
+afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t  *local = NULL;
+
+        local = frame->local;
+
+        if (is_afr_delayed_changelog_post_op_needed (frame, this))
+                afr_delayed_changelog_post_op (this, frame, local->fd, NULL);
+        else
+                afr_changelog_post_op_safe (frame, this);
+}
+
+
+
+/* Wake up the sleeping/delayed post-op, and also register
+   a stub to have it resumed after this transaction
+   completely finishes.
+
+   The @stub gets saved in @local and gets resumed in
+   afr_local_cleanup()
+   */
+void
+afr_delayed_changelog_wake_resume (xlator_t *this, fd_t *fd, call_stub_t *stub)
+{
+        afr_delayed_changelog_post_op (this, NULL, fd, stub);
+}
+
+
+void
+afr_delayed_changelog_wake_up (xlator_t *this, fd_t *fd)
+{
+        afr_delayed_changelog_post_op (this, NULL, fd, NULL);
+}
+
+
+int
+afr_transaction_resume (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t         *local    = NULL;
+
+        local    = frame->local;
+
+        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 */
+                afr_remove_eager_lock_stub (local);
+        }
+
+        afr_restore_lk_owner (frame);
+
+        afr_handle_symmetric_errors (frame, this);
+
+	if (!local->pre_op_compat)
+		/* new mode, pre-op was done along
+		   with OP */
+		afr_changelog_pre_op_update (frame, this);
+
+        if (__fop_changelog_needed (frame, this)) {
+                afr_changelog_post_op (frame, this);
+        } else {
+		afr_changelog_post_op_done (frame, this);
+        }
+
+        return 0;
+}
+
+
 /**
  * afr_transaction_fop_failed - inform that an fop failed
  */
 
 void
-afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this, int child_index)
+afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this,
+                            int child_index)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
+        afr_local_t *   local = NULL;
+
+        local = frame->local;
+
+	local->transaction.failed_subvols[child_index] = 1;
+}
+
 
-	local = frame->local;
-	priv  = this->private;
 
-        __mark_child_dead (local->pending, priv->child_count,
-                           child_index, local->transaction.type);
+        static gf_boolean_t
+afr_locals_overlap (afr_local_t *local1, afr_local_t *local2)
+{
+        uint64_t start1 = local1->transaction.start;
+        uint64_t start2 = local2->transaction.start;
+        uint64_t end1 = 0;
+        uint64_t end2 = 0;
+
+        if (local1->transaction.len)
+                end1 = start1 + local1->transaction.len - 1;
+        else
+                end1 = ULLONG_MAX;
+
+        if (local2->transaction.len)
+                end2 = start2 + local2->transaction.len - 1;
+        else
+                end2 = ULLONG_MAX;
+
+        return ((end1 >= start2) && (end2 >= start1));
+}
+
+void
+afr_transaction_eager_lock_init (afr_local_t *local, xlator_t *this)
+{
+        afr_private_t *priv = NULL;
+        afr_fd_ctx_t  *fdctx = NULL;
+        afr_local_t   *each = NULL;
+
+        priv = this->private;
+
+        if (!local->fd)
+                return;
+
+        if (local->transaction.type != AFR_DATA_TRANSACTION)
+                return;
+
+        if (!priv->eager_lock)
+                return;
+
+        fdctx = afr_fd_ctx_get (local->fd, this);
+        if (!fdctx)
+                return;
+
+        if (afr_are_multiple_fds_opened (local->fd, 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;
+        afr_local_t *   local = NULL;
+        afr_private_t * priv  = NULL;
+        fd_t            *fd   = NULL;
+        int             ret   = -1;
 
-	local = frame->local;
-	priv  = this->private;
+        local = frame->local;
+        priv  = this->private;
 
-	afr_transaction_local_init (local, priv);
+        local->transaction.resume = afr_transaction_resume;
+        local->transaction.type   = type;
 
-	local->transaction.resume = afr_transaction_resume;
-	local->transaction.type   = type;
+        ret = afr_transaction_local_init (local, this);
+        if (ret < 0)
+            goto out;
 
-	if (afr_lock_server_count (priv, local->transaction.type) == 0) {
-                afr_internal_lock_finish (frame, this);
-	} else {
-		afr_lock (frame, this);
-	}
+        afr_transaction_eager_lock_init (local, this);
 
-	return 0;
+        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);
+        }
+        ret = 0;
+out:
+        return ret;
 }
diff --git a/xlators/cluster/afr/src/afr-transaction.h b/xlators/cluster/afr/src/afr-transaction.h
index 84cf31d63..77cc8eed0 100644
--- a/xlators/cluster/afr/src/afr-transaction.h
+++ b/xlators/cluster/afr/src/afr-transaction.h
@@ -1,25 +1,18 @@
 /*
-   Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __TRANSACTION_H__
 #define __TRANSACTION_H__
 
+#include "afr.h"
+
 void
 afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this,
 			    int child_index);
@@ -27,7 +20,34 @@ 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);
 
+int
+afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending);
+
+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 (call_frame_t *frame, xlator_t *this);
+
+gf_boolean_t
+afr_txn_nothing_failed (call_frame_t *frame, xlator_t *this);
+
+int afr_read_txn (call_frame_t *frame, xlator_t *this, inode_t *inode,
+		  afr_read_txn_wind_t readfn, afr_transaction_type type);
+
+int afr_read_txn_continue (call_frame_t *frame, xlator_t *this, int subvol);
+
+int __afr_txn_write_fop (call_frame_t *frame, xlator_t *this);
+int __afr_txn_write_done (call_frame_t *frame, xlator_t *this);
+call_frame_t *afr_transaction_detach_fop_frame (call_frame_t *frame);
+
 #endif /* __TRANSACTION_H__ */
diff --git a/xlators/cluster/afr/src/afr.c b/xlators/cluster/afr/src/afr.c
index 775a53a8f..5e12910b7 100644
--- a/xlators/cluster/afr/src/afr.c
+++ b/xlators/cluster/afr/src/afr.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include <libgen.h>
@@ -30,13 +21,20 @@
 #endif
 #include "afr-common.c"
 
+struct volume_options options[];
+
 int32_t
 notify (xlator_t *this, int32_t event,
-	void *data, ...)
+        void *data, ...)
 {
         int ret = -1;
+        va_list         ap;
+        void *data2 = NULL;
 
-        ret = afr_notify (this, event, data);
+        va_start (ap, data);
+        data2 = va_arg (ap, dict_t*);
+        va_end (ap);
+        ret = afr_notify (this, event, data, data2);
 
         return ret;
 }
@@ -53,896 +51,449 @@ mem_acct_init (xlator_t *this)
 
         if (ret != 0) {
                 gf_log(this->name, GF_LOG_ERROR, "Memory accounting init"
-                                "failed");
+                       "failed");
                 return ret;
         }
 
         return ret;
 }
+
+
 int
-validate_options (xlator_t *this, dict_t *options, char **op_errstr)
+xlator_subvolume_index (xlator_t *this, xlator_t *subvol)
 {
+        int index = -1;
+        int i = 0;
+        xlator_list_t *list = NULL;
 
- 
-        gf_boolean_t metadata_self_heal;   
-        gf_boolean_t entry_self_heal;
-        gf_boolean_t data_self_heal;
-        gf_boolean_t data_change_log;       
-        gf_boolean_t metadata_change_log; 
-        gf_boolean_t entry_change_log;      
-        gf_boolean_t strict_readdir;
-
-        xlator_list_t * trav        = NULL;
-        
-        char * read_subvol     = NULL;
-        char * self_heal       = NULL;
-        char * change_log      = NULL;
-        char * str_readdir     = NULL;
-
-        int32_t background_count  = 0;
-        int32_t window_size       = 0;
-
-        int    read_ret      = -1;
-        int    dict_ret      = -1;
-        int    flag          = 1;
-        int    ret           = 0;
-        int    temp_ret      = -1;
-        
-
-        
-        dict_ret = dict_get_int32 (options, "background-self-heal-count",
-                                   &background_count);
-        if (dict_ret == 0) {
-                if (background_count < 0) {
-                        *op_errstr = gf_strdup ("Error, option should be >= 0");
-                        ret = -1;
-                        goto out;
-                }
+        list = this->children;
 
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "validated background self-heal count to %d",
-                        background_count);
+        while (list) {
+                if (subvol == list->xlator ||
+                    strcmp (subvol->name, list->xlator->name) == 0) {
+                        index = i;
+                        break;
+                }
+                list = list->next;
+                i++;
         }
 
-        dict_ret = dict_get_str (options, "metadata-self-heal",
-                                 &self_heal);
-        if (dict_ret == 0) {
-                temp_ret = gf_string2boolean (self_heal, &metadata_self_heal);
-                if (temp_ret < 0) {
-                        gf_log (this->name, GF_LOG_WARNING,
-                                "validation failed 'option metadata"
-                                                "-self-heal %s'.not correct.", 
-                                                self_heal);
-                        *op_errstr = gf_strdup ("Error, option should be boolean");
-                        ret = -1;
-                        goto out;
-                } 
-        
+        return index;
+}
 
-                
+void
+fix_quorum_options (xlator_t *this, afr_private_t *priv, char *qtype)
+{
+        if (priv->quorum_count && strcmp(qtype,"fixed")) {
+                gf_log(this->name,GF_LOG_WARNING,
+                       "quorum-type %s overriding quorum-count %u",
+                       qtype, priv->quorum_count);
         }
-
-        dict_ret = dict_get_str (options, "data-self-heal",
-                                 &self_heal);
-        if (dict_ret == 0) {
-                temp_ret = gf_string2boolean (self_heal, &data_self_heal);
-                if (temp_ret < 0) {
-                        gf_log (this->name, GF_LOG_WARNING,
-                                "Validation failed for data self heal", 
-                                self_heal);
-                        *op_errstr = gf_strdup ("Error, option should be boolean");
-                        ret = -1;
-                        goto out;
-                } 
-        
-
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Reconfiguring 'option data"
-                                        "-self-heal %s'.", self_heal);
+        if (!strcmp(qtype,"none")) {
+                priv->quorum_count = 0;
         }
-
-        dict_ret = dict_get_str (options, "entry-self-heal",
-                                 &self_heal);
-        if (dict_ret == 0) {
-                temp_ret = gf_string2boolean (self_heal, &entry_self_heal);
-                if (temp_ret < 0) {
-                        gf_log (this->name, GF_LOG_WARNING,
-                                "Validation faled for entry-self-heal", 
-                                self_heal);
-                        *op_errstr = gf_strdup ("Error, option should be boolean");
-                        ret = -1;
-                        goto out;
-                } 
-        
-
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Validated 'option entry"
-                                        "-self-heal %s'.", self_heal);
+        else if (!strcmp(qtype,"auto")) {
+                priv->quorum_count = AFR_QUORUM_AUTO;
         }
+}
 
+int
+reconfigure (xlator_t *this, dict_t *options)
+{
+        afr_private_t *priv        = NULL;
+        xlator_t      *read_subvol = NULL;
+        int            read_subvol_index = -1;
+        int            ret         = -1;
+        int            index       = -1;
+        char          *qtype       = NULL;
 
-        dict_ret = dict_get_str (options, "strict-readdir",
-                                 &str_readdir);
-        if (dict_ret == 0) {
-                temp_ret = gf_string2boolean (str_readdir, &strict_readdir);
-                if (temp_ret < 0) {
-                        gf_log (this->name, GF_LOG_WARNING,
-                                "Validation faled for strict_readdir",
-                                str_readdir);
-                        *op_errstr = gf_strdup ("Error, option should be boolean");
-                        ret = -1;
-                        goto out;
-                }
-                
+        priv = this->private;
 
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Validated 'option strict"
-                                        "-readdir %s'.", str_readdir);
-        }
+	GF_OPTION_RECONF ("afr-dirty-xattr",
+			  priv->afr_dirty, options, str,
+			  out);
 
-        dict_ret = dict_get_int32 (options, "data-self-heal-window-size",
-                                   &window_size);
-        if (dict_ret == 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "validated data self-heal window size to %d",
-                        window_size);
+	GF_OPTION_RECONF ("metadata-splitbrain-forced-heal",
+			  priv->metadata_splitbrain_forced_heal, options, bool,
+			  out);
 
-                if (window_size < 0) {
-                        *op_errstr = gf_strdup ("Error, option should be >= 0");
-                        ret = -1;
-                        goto out;
-                }
-
-                if (window_size > 1024) {
-                        *op_errstr = gf_strdup ("Error, option should be <= 1024");
-                        ret = -1;
-                        goto out;
-                }
+        GF_OPTION_RECONF ("background-self-heal-count",
+                          priv->background_self_heal_count, options, uint32,
+                          out);
 
+        GF_OPTION_RECONF ("metadata-self-heal",
+                          priv->metadata_self_heal, options, bool, out);
 
-        }
+        GF_OPTION_RECONF ("data-self-heal", priv->data_self_heal, options, str,
+                          out);
 
-        dict_ret = dict_get_str (options, "data-change-log",
-                                 &change_log);
-        if (dict_ret == 0) {
-                temp_ret = gf_string2boolean (change_log, &data_change_log);
-                if (temp_ret < 0) {
-                        gf_log (this->name, GF_LOG_WARNING,
-                                "Validation faled for data-change-log");
-                        *op_errstr = gf_strdup ("Error, option should be boolean");
-                        ret = -1;
-                        goto out;
-                }
- 
+        GF_OPTION_RECONF ("entry-self-heal", priv->entry_self_heal, options,
+                          bool, out);
 
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Validated 'option data-"
-                                        "change-log %s'.", change_log);
-        }
+        GF_OPTION_RECONF ("data-self-heal-window-size",
+                          priv->data_self_heal_window_size, options,
+                          uint32, out);
 
-        dict_ret = dict_get_str (options, "metadata-change-log",
-                                 &change_log);
-        if (dict_ret == 0) {
-                temp_ret = gf_string2boolean (change_log,
-                                              &metadata_change_log);
-                if (temp_ret < 0) {
-                        gf_log (this->name, GF_LOG_WARNING,
-                                "Validation faild for metadata-change-log");
-                        *op_errstr = gf_strdup ("Error, option should be boolean");
-                        ret = -1;
-                        goto out;
-                } 
-                
+        GF_OPTION_RECONF ("data-change-log", priv->data_change_log, options,
+                          bool, out);
 
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Validated 'option metadata-"
-                                        "change-log %s'.", change_log);
-        }
+        GF_OPTION_RECONF ("metadata-change-log",
+                          priv->metadata_change_log, options, bool, out);
 
-        dict_ret = dict_get_str (options, "entry-change-log",
-                                 &change_log);
-        if (dict_ret == 0) {
-                temp_ret = gf_string2boolean (change_log, &entry_change_log);
-                if (temp_ret < 0) {
-                        gf_log (this->name, GF_LOG_WARNING,
-                                "Validation faild for entr-change-log");
-                        *op_errstr = gf_strdup ("Error, option should be boolean");
-                        ret = -1;
-                        goto out;
-                } 
+        GF_OPTION_RECONF ("entry-change-log", priv->entry_change_log, options,
+                          bool, out);
 
+        GF_OPTION_RECONF ("data-self-heal-algorithm",
+                          priv->data_self_heal_algorithm, options, str, out);
 
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Validated 'option entry-"
-                                        "change-log %s'.", change_log);
-        }
+        GF_OPTION_RECONF ("read-subvolume", read_subvol, options, xlator, out);
 
-        read_ret = dict_get_str (options, "read-subvolume", &read_subvol);
+        GF_OPTION_RECONF ("read-hash-mode", priv->hash_mode,
+                          options, uint32, out);
 
-        if (read_ret) 
-                goto next;// No need to traverse, hence set the next option
-        
-        trav = this->children;
-        flag = 0;
-        while (trav) {
-                if (!read_ret && !strcmp (read_subvol, trav->xlator->name)) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "Validated Subvolume '%s'  as read child.",
-                                trav->xlator->name);
-
-                        flag = 1;
-                        ret = 0; 
+        if (read_subvol) {
+                index = xlator_subvolume_index (this, read_subvol);
+                if (index == -1) {
+                        gf_log (this->name, GF_LOG_ERROR, "%s not a subvolume",
+                                read_subvol->name);
                         goto out;
                 }
-
-                
-                trav = trav->next;
+                priv->read_child = index;
         }
 
-        if (flag == 0 ) {
+        GF_OPTION_RECONF ("read-subvolume-index",read_subvol_index, options,int32,out);
 
-                gf_log (this->name, GF_LOG_WARNING,
-                        "Invalid 'option read-subvolume %s', no such subvolume"
-                                        , read_subvol);
-                *op_errstr = gf_strdup ("Error, the sub-volume is not right");
-                ret = -1;
-                goto out;
+        if (read_subvol_index >-1) {
+                index=read_subvol_index;
+                if (index >= priv->child_count) {
+                        gf_log (this->name, GF_LOG_ERROR, "%d not a subvolume-index",
+                                index);
+                        goto out;
+                }
+                priv->read_child = index;
         }
 
-next:
-out:
-                return ret;
-
-
-}
-
+        GF_OPTION_RECONF ("pre-op-compat", priv->pre_op_compat, options, bool, out);
 
-int
-reconfigure (xlator_t *this, dict_t *options)
-{
+        GF_OPTION_RECONF ("eager-lock", priv->eager_lock, options, bool, out);
+        GF_OPTION_RECONF ("quorum-type", qtype, options, str, out);
+        GF_OPTION_RECONF ("quorum-count", priv->quorum_count, options,
+                          uint32, out);
+        fix_quorum_options(this,priv,qtype);
 
-	gf_boolean_t metadata_self_heal;   /* on/off */
-	gf_boolean_t entry_self_heal;
-	gf_boolean_t data_self_heal;
-	gf_boolean_t data_change_log;       /* on/off */
-	gf_boolean_t metadata_change_log;   /* on/off */
-	gf_boolean_t entry_change_log;      /* on/off */
-	gf_boolean_t strict_readdir;
-
-	afr_private_t * priv        = NULL;
-	xlator_list_t * trav        = NULL;
-	
-	char * read_subvol     = NULL;
-	char * self_heal       = NULL;
-	char * change_log      = NULL;
-	char * str_readdir     = NULL;
-
-        int32_t background_count  = 0;
-        int32_t window_size       = 0;
-
-	int    read_ret      = -1;
-	int    dict_ret      = -1;
-	int    flag	     = 1;
-	int    ret           = 0;
-	int    temp_ret	     = -1;
-	
-	priv = this->private;
-	
-	dict_ret = dict_get_int32 (options, "background-self-heal-count",
-				   &background_count);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring background self-heal count to %d",
-			background_count);
-
-		priv->background_self_heal_count = background_count;
-	}
+	GF_OPTION_RECONF ("post-op-delay-secs", priv->post_op_delay_secs, options,
+			  uint32, out);
 
-	dict_ret = dict_get_str (options, "metadata-self-heal",
-				 &self_heal);
-	if (dict_ret == 0) {
-		temp_ret = gf_string2boolean (self_heal, &metadata_self_heal);
-		if (temp_ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Reconfiguration Invalid 'option metadata"
-				"-self-heal %s'. Defaulting to old value.", 
-				self_heal);
-			ret = -1;
-			goto out;
-		} 
-	
-		priv->metadata_self_heal = metadata_self_heal;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring 'option metadata"
-			"-self-heal %s'.", 
-			self_heal);
-	}
+        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 ("ensure-durability", priv->ensure_durability, options,
+                          bool, out);
 
-	dict_ret = dict_get_str (options, "data-self-heal",
-				 &self_heal);
-	if (dict_ret == 0) {
-		temp_ret = gf_string2boolean (self_heal, &data_self_heal);
-		if (temp_ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Reconfiguration Invalid 'option data"
-				"-self-heal %s'. Defaulting to old value.", 
-				self_heal);
-			ret = -1;
-			goto out;
-		} 
-	
-		priv->data_self_heal = data_self_heal;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring 'option data"
-			"-self-heal %s'.", self_heal);
-	}
+	GF_OPTION_RECONF ("self-heal-daemon", priv->shd.enabled, options,
+			  bool, out);
 
-	dict_ret = dict_get_str (options, "entry-self-heal",
-				 &self_heal);
-	if (dict_ret == 0) {
-		temp_ret = gf_string2boolean (self_heal, &entry_self_heal);
-		if (temp_ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Reconfiguration Invalid 'option data"
-				"-self-heal %s'. Defaulting to old value.", 
-				self_heal);
-			ret = -1;
-			goto out;
-		} 
-	
-		priv->entry_self_heal = entry_self_heal;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring 'option entry"
-			"-self-heal %s'.", self_heal);
-	}
-
-
-	dict_ret = dict_get_str (options, "strict-readdir",
-				 &str_readdir);
-	if (dict_ret == 0) {
-		temp_ret = gf_string2boolean (str_readdir, &strict_readdir);
-		if (temp_ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option strict-readdir %s'. "
-				"Defaulting to old value.",
-				str_readdir);
-			ret = -1;
-			goto out;
-		}
-		
-		priv->strict_readdir = strict_readdir;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring 'option strict"
-			"-readdir %s'.", str_readdir);
-	}
-
-	dict_ret = dict_get_int32 (options, "data-self-heal-window-size",
-				   &window_size);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring, Setting data self-heal window size to %d",
-			window_size);
+	GF_OPTION_RECONF ("iam-self-heal-daemon", priv->shd.iamshd, options,
+			  bool, out);
 
-		priv->data_self_heal_window_size = window_size;
-	}
-        else {
-                priv->data_self_heal_window_size = 16;
-        }
-        
-
-	dict_ret = dict_get_str (options, "data-change-log",
-				 &change_log);
-	if (dict_ret == 0) {
-		temp_ret = gf_string2boolean (change_log, &data_change_log);
-		if (temp_ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Reconfiguration Invalid 'option data-"
-				"change-log %s'. Defaulting to old value.", 
-				change_log);
-			ret = -1;
-			goto out;
-		}
- 
-		priv->data_change_log = data_change_log;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring 'option data-"
-			"change-log %s'.", change_log);
-	}
-
-	dict_ret = dict_get_str (options, "metadata-change-log",
-				 &change_log);
-	if (dict_ret == 0) {
-		temp_ret = gf_string2boolean (change_log,
-					 &metadata_change_log);
-		if (temp_ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option metadata-change-log %s'. "
-				"Defaulting to metadata-change-log as 'off'.",
-				change_log);
-			ret = -1;
-			goto out;
-		} 
-		
-		priv->metadata_change_log = metadata_change_log;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring 'option metadata-"
-			"change-log %s'.", change_log);
-	}
-
-	dict_ret = dict_get_str (options, "entry-change-log",
-				 &change_log);
-	if (dict_ret == 0) {
-		temp_ret = gf_string2boolean (change_log, &entry_change_log);
-		if (temp_ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option entry-change-log %s'. "
-				"Defaulting to entry-change-log as 'on'.", 
-				change_log);
-			ret = -1;
-			goto out;
-		} 
-
-		priv->entry_change_log = entry_change_log;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Reconfiguring 'option entry-"
-			"change-log %s'.", change_log);
-	}
-
-	read_ret = dict_get_str (options, "read-subvolume", &read_subvol);
-
-	if (read_ret < 0) 
-		goto next;// No need to traverse, hence set the next option
-	
-	trav = this->children;
-	flag = 0;
-	while (trav) {
-		if (!read_ret && !strcmp (read_subvol, trav->xlator->name)) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"Subvolume '%s' specified as read child.",
-				trav->xlator->name);
-
-			flag = 1;
-			ret = -1; 
-			goto out;
-		}
-
-		
-		trav = trav->next;
-	}
-
-	if (flag == 0 ) {
-
-		gf_log (this->name, GF_LOG_ERROR,
-			"Invalid 'option read-subvolume %s', no such subvolume"
-			, read_subvol);
-		ret = -1;
-		goto out;
-	}
+        priv->did_discovery = _gf_false;
 
-next:
+        ret = 0;
 out:
-	return ret;
+        return ret;
 
 }
 
 
 static const char *favorite_child_warning_str = "You have specified subvolume '%s' "
-	"as the 'favorite child'. This means that if a discrepancy in the content "
-	"or attributes (ownership, permission, etc.) of a file is detected among "
-	"the subvolumes, the file on '%s' will be considered the definitive "
-	"version and its contents will OVERWRITE the contents of the file on other "
-	"subvolumes. All versions of the file except that on '%s' "
-	"WILL BE LOST.";
-
-static const char *no_lock_servers_warning_str = "You have set lock-server-count = 0. "
-	"This means correctness is NO LONGER GUARANTEED in all cases. If two or more "
-	"applications write to the same region of a file, there is a possibility that "
-	"its copies will be INCONSISTENT. Set it to a value greater than 0 unless you "
-	"are ABSOLUTELY SURE of what you are doing and WILL NOT HOLD GlusterFS "
-	"RESPONSIBLE for inconsistent data. If you are in doubt, set it to a value "
-	"greater than 0.";
-
-int32_t 
+        "as the 'favorite child'. This means that if a discrepancy in the content "
+        "or attributes (ownership, permission, etc.) of a file is detected among "
+        "the subvolumes, the file on '%s' will be considered the definitive "
+        "version and its contents will OVERWRITE the contents of the file on other "
+        "subvolumes. All versions of the file except that on '%s' "
+        "WILL BE LOST.";
+
+
+int32_t
 init (xlator_t *this)
 {
-	afr_private_t * priv        = NULL;
-	int             child_count = 0;
-	xlator_list_t * trav        = NULL;
-	int             i           = 0;
-	int             ret         = -1;
-	int             op_errno    = 0;
-
-	char * read_subvol     = NULL;
-	char * fav_child       = NULL;
-	char * self_heal       = NULL;
-        char * algo            = NULL;
-	char * change_log      = NULL;
-	char * strict_readdir  = NULL;
-        char * inodelk_trace   = NULL;
-        char * entrylk_trace   = NULL;
-
-        int32_t background_count  = 0;
-	int32_t lock_server_count = 1;
-        int32_t window_size       = 0;
-
-	int    fav_ret       = -1;
-	int    read_ret      = -1;
-	int    dict_ret      = -1;
-
-
-	if (!this->children) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"replicate translator needs more than one "
+        afr_private_t *priv        = NULL;
+        int            child_count = 0;
+        xlator_list_t *trav        = NULL;
+        int            i           = 0;
+        int            ret         = -1;
+        GF_UNUSED int  op_errno    = 0;
+        xlator_t      *read_subvol = NULL;
+        int            read_subvol_index = -1;
+        xlator_t      *fav_child   = NULL;
+        char          *qtype       = NULL;
+
+        if (!this->children) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "replicate translator needs more than one "
                         "subvolume defined.");
-		return -1;
-	}
-  
-	if (!this->parents) {
-		gf_log (this->name, GF_LOG_WARNING,
-			"Volume is dangling.");
-	}
-
-
-	ALLOC_OR_GOTO (this->private, afr_private_t, out);
-
-	priv = this->private;
-
-	read_ret = dict_get_str (this->options, "read-subvolume", &read_subvol);
-	priv->read_child = -1;
-
-	fav_ret = dict_get_str (this->options, "favorite-child", &fav_child);
-	priv->favorite_child = -1;
-
-        priv->background_self_heal_count = 16;
-
-	dict_ret = dict_get_int32 (this->options, "background-self-heal-count",
-				   &background_count);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Setting background self-heal count to %d",
-			background_count);
-
-		priv->background_self_heal_count = background_count;
-	}
-
-	/* Default values */
-
-	priv->data_self_heal     = 1;
-	priv->metadata_self_heal = 1;
-	priv->entry_self_heal    = 1;
-
-	dict_ret = dict_get_str (this->options, "data-self-heal", &self_heal);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (self_heal, &priv->data_self_heal);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option data-self-heal %s'. "
-				"Defaulting to data-self-heal as 'on'",
-				self_heal);
-			priv->data_self_heal = 1;
-		}
-	}
-
-        priv->data_self_heal_algorithm = "";
-
-        dict_ret = dict_get_str (this->options, "data-self-heal-algorithm",
-                                 &algo);
-        if (dict_ret == 0) {
-                priv->data_self_heal_algorithm = gf_strdup (algo);
+                return -1;
         }
 
+        if (!this->parents) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Volume is dangling.");
+        }
 
-        priv->data_self_heal_window_size = 16;
-
-	dict_ret = dict_get_int32 (this->options, "data-self-heal-window-size",
-				   &window_size);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Setting data self-heal window size to %d",
-			window_size);
+	this->private = GF_CALLOC (1, sizeof (afr_private_t),
+                                   gf_afr_mt_afr_private_t);
+        if (!this->private)
+                goto out;
 
-		priv->data_self_heal_window_size = window_size;
-	}
+        priv = this->private;
+        LOCK_INIT (&priv->lock);
 
-	dict_ret = dict_get_str (this->options, "metadata-self-heal",
-				 &self_heal);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (self_heal, &priv->metadata_self_heal);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option metadata-self-heal %s'. "
-				"Defaulting to metadata-self-heal as 'on'.", 
-				self_heal);
-			priv->metadata_self_heal = 1;
-		} 
-	}
+        child_count = xlator_subvolume_count (this);
 
-	dict_ret = dict_get_str (this->options, "entry-self-heal", &self_heal);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (self_heal, &priv->entry_self_heal);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option entry-self-heal %s'. "
-				"Defaulting to entry-self-heal as 'on'.", 
-				self_heal);
-			priv->entry_self_heal = 1;
-		} 
-	}
+        priv->child_count = child_count;
 
-	/* Change log options */
-
-	priv->data_change_log     = 1;
-	priv->metadata_change_log = 1;
-	priv->entry_change_log    = 1;
-
-	dict_ret = dict_get_str (this->options, "data-change-log",
-				 &change_log);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (change_log, &priv->data_change_log);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option data-change-log %s'. "
-				"Defaulting to data-change-log as 'on'.", 
-				change_log);
-			priv->data_change_log = 1;
-		} 
-	}
+        priv->read_child = -1;
 
-	dict_ret = dict_get_str (this->options, "metadata-change-log",
-				 &change_log);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (change_log,
-					 &priv->metadata_change_log);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option metadata-change-log %s'. "
-				"Defaulting to metadata-change-log as 'off'.",
-				change_log);
-			priv->metadata_change_log = 0;
-		} 
-	}
+	GF_OPTION_INIT ("afr-dirty-xattr", priv->afr_dirty, str, out);
 
-	dict_ret = dict_get_str (this->options, "entry-change-log",
-				 &change_log);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (change_log, &priv->entry_change_log);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option entry-change-log %s'. "
-				"Defaulting to entry-change-log as 'on'.", 
-				change_log);
-			priv->entry_change_log = 1;
-		} 
-	}
+	GF_OPTION_INIT ("metadata-splitbrain-forced-heal",
+			priv->metadata_splitbrain_forced_heal, bool, out);
 
-	/* Locking options */
+        GF_OPTION_INIT ("read-subvolume", read_subvol, xlator, out);
+        if (read_subvol) {
+                priv->read_child = xlator_subvolume_index (this, read_subvol);
+                if (priv->read_child == -1) {
+                        gf_log (this->name, GF_LOG_ERROR, "%s not a subvolume",
+                                read_subvol->name);
+                        goto out;
+                }
+        }
+        GF_OPTION_INIT ("read-subvolume-index",read_subvol_index,int32,out);
+        if (read_subvol_index > -1) {
+                if (read_subvol_index >= priv->child_count) {
+                        gf_log (this->name, GF_LOG_ERROR, "%d not a subvolume-index",
+                                read_subvol_index);
+                        goto out;
+                }
+                priv->read_child = read_subvol_index;
+        }
+        GF_OPTION_INIT ("choose-local", priv->choose_local, bool, out);
 
-        priv->inodelk_trace = 0;
-        priv->entrylk_trace = 0;
+        GF_OPTION_INIT ("read-hash-mode", priv->hash_mode, uint32, out);
 
-	dict_ret = dict_get_str (this->options, "inodelk-trace",
-				 &inodelk_trace);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (inodelk_trace, &priv->inodelk_trace);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option inodelk-trace %s' ",
-                                inodelk_trace);
+        priv->favorite_child = -1;
+        GF_OPTION_INIT ("favorite-child", fav_child, xlator, out);
+        if (fav_child) {
+                priv->favorite_child = xlator_subvolume_index (this, fav_child);
+                if (priv->favorite_child == -1) {
+                        gf_log (this->name, GF_LOG_ERROR, "%s not a subvolume",
+                                fav_child->name);
+                        goto out;
+                }
+                gf_log (this->name, GF_LOG_WARNING,
+                        favorite_child_warning_str, fav_child->name,
+                        fav_child->name, fav_child->name);
+        }
 
-			priv->inodelk_trace = 0;
-		}
-	}
 
+        GF_OPTION_INIT ("background-self-heal-count",
+                        priv->background_self_heal_count, uint32, out);
 
-	dict_ret = dict_get_str (this->options, "entrylk-trace",
-				 &entrylk_trace);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (entrylk_trace, &priv->entrylk_trace);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option entrylk-trace %s' ",
-                                inodelk_trace);
+        GF_OPTION_INIT ("data-self-heal", priv->data_self_heal, str, out);
 
-			priv->entrylk_trace = 0;
-		}
-	}
+        GF_OPTION_INIT ("data-self-heal-algorithm",
+                        priv->data_self_heal_algorithm, str, out);
 
+        GF_OPTION_INIT ("data-self-heal-window-size",
+                        priv->data_self_heal_window_size, uint32, out);
 
-	priv->data_lock_server_count = 1;
-	priv->metadata_lock_server_count = 0;
-	priv->entry_lock_server_count = 1;
+        GF_OPTION_INIT ("metadata-self-heal", priv->metadata_self_heal, bool,
+                        out);
 
-	dict_ret = dict_get_int32 (this->options, "data-lock-server-count", 
-				   &lock_server_count);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Setting data lock server count to %d.",
-			lock_server_count);
+        GF_OPTION_INIT ("entry-self-heal", priv->entry_self_heal, bool, out);
 
-		if (lock_server_count == 0) 
-			gf_log (this->name, GF_LOG_WARNING, "%s",
-                                no_lock_servers_warning_str);
+        GF_OPTION_INIT ("data-change-log", priv->data_change_log, bool, out);
 
-		priv->data_lock_server_count = lock_server_count;
-	}
+        GF_OPTION_INIT ("metadata-change-log", priv->metadata_change_log, bool,
+                        out);
 
+        GF_OPTION_INIT ("entry-change-log", priv->entry_change_log, bool, out);
 
-	dict_ret = dict_get_int32 (this->options,
-				   "metadata-lock-server-count", 
-				   &lock_server_count);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Setting metadata lock server count to %d.",
-			lock_server_count);
-		priv->metadata_lock_server_count = lock_server_count;
-	}
+        GF_OPTION_INIT ("optimistic-change-log", priv->optimistic_change_log,
+                        bool, out);
 
+        GF_OPTION_INIT ("inodelk-trace", priv->inodelk_trace, bool, out);
 
-	dict_ret = dict_get_int32 (this->options, "entry-lock-server-count", 
-				   &lock_server_count);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Setting entry lock server count to %d.",
-			lock_server_count);
+        GF_OPTION_INIT ("entrylk-trace", priv->entrylk_trace, bool, out);
 
-		priv->entry_lock_server_count = lock_server_count;
-	}
+        GF_OPTION_INIT ("pre-op-compat", priv->pre_op_compat, bool, out);
 
-	priv->strict_readdir = _gf_false;
-
-	dict_ret = dict_get_str (this->options, "strict-readdir",
-				 &strict_readdir);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (strict_readdir, &priv->strict_readdir);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option strict-readdir %s'. "
-				"Defaulting to strict-readdir as 'off'.",
-				strict_readdir);
-		}
-	}
+        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);
 
-	trav = this->children;
-	while (trav) {
-		if (!read_ret && !strcmp (read_subvol, trav->xlator->name)) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"Subvolume '%s' specified as read child.",
-				trav->xlator->name);
-
-			priv->read_child = child_count;
-		}
-
-		if (fav_ret == 0 && !strcmp (fav_child, trav->xlator->name)) {
-			gf_log (this->name, GF_LOG_WARNING,
-				favorite_child_warning_str, trav->xlator->name,
-				trav->xlator->name, trav->xlator->name);
-			priv->favorite_child = child_count;
-		}
-
-		child_count++;
-		trav = trav->next;
-	}
+	GF_OPTION_INIT ("post-op-delay-secs", priv->post_op_delay_secs, uint32, out);
+        GF_OPTION_INIT ("ensure-durability", priv->ensure_durability, bool,
+                        out);
 
-	priv->wait_count = 1;
+	GF_OPTION_INIT ("self-heal-daemon", priv->shd.enabled, bool, out);
 
-	priv->child_count = child_count;
+	GF_OPTION_INIT ("iam-self-heal-daemon", priv->shd.iamshd, bool, out);
 
-	LOCK_INIT (&priv->lock);
-        LOCK_INIT (&priv->read_child_lock);
+        priv->wait_count = 1;
 
-	priv->child_up = GF_CALLOC (sizeof (unsigned char), child_count,
+        priv->child_up = GF_CALLOC (sizeof (unsigned char), child_count,
                                     gf_afr_mt_char);
-	if (!priv->child_up) {
-		gf_log (this->name, GF_LOG_ERROR,	
-			"Out of memory.");		
-		ret = -ENOMEM;
-		goto out;
-	}
+        if (!priv->child_up) {
+                ret = -ENOMEM;
+                goto out;
+        }
+
+        for (i = 0; i < child_count; i++)
+                priv->child_up[i] = -1; /* start with unknown state.
+                                           this initialization needed
+                                           for afr_notify() to work
+                                           reliably
+                                        */
 
-	priv->children = GF_CALLOC (sizeof (xlator_t *), child_count,
+        priv->children = GF_CALLOC (sizeof (xlator_t *), child_count,
                                     gf_afr_mt_xlator_t);
-	if (!priv->children) {
-		gf_log (this->name, GF_LOG_ERROR,	
-			"Out of memory.");		
-		ret = -ENOMEM;
-		goto out;
-	}
+        if (!priv->children) {
+                ret = -ENOMEM;
+                goto out;
+        }
 
-        priv->pending_key = GF_CALLOC (sizeof (*priv->pending_key), 
-                                        child_count,
-                                        gf_afr_mt_char);
+        priv->pending_key = GF_CALLOC (sizeof (*priv->pending_key),
+                                       child_count,
+                                       gf_afr_mt_char);
         if (!priv->pending_key) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
                 ret = -ENOMEM;
                 goto out;
         }
 
-	trav = this->children;
-	i = 0;
-	while (i < child_count) {
-		priv->children[i] = trav->xlator;
+        trav = this->children;
+        i = 0;
+        while (i < child_count) {
+                priv->children[i] = trav->xlator;
 
-                ret = gf_asprintf (&priv->pending_key[i], "%s.%s", 
+                ret = gf_asprintf (&priv->pending_key[i], "%s.%s",
                                    AFR_XATTR_PREFIX,
                                    trav->xlator->name);
                 if (-1 == ret) {
-                        gf_log (this->name, GF_LOG_ERROR, 
-                                "asprintf failed to set pending key");
                         ret = -ENOMEM;
                         goto out;
                 }
 
-		trav = trav->next;
-		i++;
+                trav = trav->next;
+                i++;
+        }
+
+        ret = gf_asprintf (&priv->sh_domain, AFR_SH_DATA_DOMAIN_FMT,
+                           this->name);
+        if (-1 == ret) {
+                ret = -ENOMEM;
+                goto out;
+        }
+
+        priv->last_event = GF_CALLOC (child_count, sizeof (*priv->last_event),
+                                      gf_afr_mt_int32_t);
+        if (!priv->last_event) {
+                ret = -ENOMEM;
+                goto out;
+        }
+
+	ret = afr_selfheal_daemon_init (this);
+	if (ret) {
+		ret = -ENOMEM;
+		goto out;
 	}
 
-        LOCK_INIT (&priv->root_inode_lk);
-        priv->first_lookup = 1;
-        priv->root_inode = NULL;
+        /* keep more local here as we may need them for self-heal etc */
+        this->local_pool = mem_pool_new (afr_local_t, 512);
+        if (!this->local_pool) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create local_t's memory pool");
+                goto out;
+        }
 
-        pthread_mutex_init (&priv->mutex, NULL);
-        INIT_LIST_HEAD (&priv->saved_fds);
+        priv->root_inode = NULL;
 
-	ret = 0;
+        ret = 0;
 out:
-	return ret;
+        return ret;
 }
 
 
 int
 fini (xlator_t *this)
 {
-	return 0;
+        afr_private_t *priv = NULL;
+
+        priv = this->private;
+        this->private = NULL;
+        afr_priv_destroy (priv);
+        if (this->itable);//I dont see any destroy func
+
+        return 0;
 }
 
 
 struct xlator_fops fops = {
-	.lookup      = afr_lookup,
-	.open        = afr_open,
-	.lk          = afr_lk,
-	.flush       = afr_flush,
-	.statfs      = afr_statfs,
-	.fsync       = afr_fsync,
-	.fsyncdir    = afr_fsyncdir,
-	.xattrop     = afr_xattrop,
-	.fxattrop    = afr_fxattrop,
-	.inodelk     = afr_inodelk,
-	.finodelk    = afr_finodelk,
-	.entrylk     = afr_entrylk,
-	.fentrylk    = afr_fentrylk,
-
-	/* inode read */
-	.access      = afr_access,
-	.stat        = afr_stat,
-	.fstat       = afr_fstat,
-	.readlink    = afr_readlink,
-	.getxattr    = afr_getxattr,
-	.readv       = afr_readv,
-
-	/* inode write */
-	.writev      = afr_writev,
-	.truncate    = afr_truncate,
-	.ftruncate   = afr_ftruncate,
-	.setxattr    = afr_setxattr,
+        .lookup      = afr_lookup,
+        .open        = afr_open,
+        .lk          = afr_lk,
+        .flush       = afr_flush,
+        .statfs      = afr_statfs,
+        .fsync       = afr_fsync,
+        .fsyncdir    = afr_fsyncdir,
+        .xattrop     = afr_xattrop,
+        .fxattrop    = afr_fxattrop,
+        .inodelk     = afr_inodelk,
+        .finodelk    = afr_finodelk,
+        .entrylk     = afr_entrylk,
+        .fentrylk    = afr_fentrylk,
+	.fallocate   = afr_fallocate,
+	.discard     = afr_discard,
+        .zerofill    = afr_zerofill,
+
+        /* inode read */
+        .access      = afr_access,
+        .stat        = afr_stat,
+        .fstat       = afr_fstat,
+        .readlink    = afr_readlink,
+        .getxattr    = afr_getxattr,
+        .fgetxattr   = afr_fgetxattr,
+        .readv       = afr_readv,
+
+        /* inode write */
+        .writev      = afr_writev,
+        .truncate    = afr_truncate,
+        .ftruncate   = afr_ftruncate,
+        .setxattr    = afr_setxattr,
+        .fsetxattr   = afr_fsetxattr,
         .setattr     = afr_setattr,
-	.fsetattr    = afr_fsetattr,
-	.removexattr = afr_removexattr,
-
-	/* dir read */
-	.opendir     = afr_opendir,
-	.readdir     = afr_readdir,
-	.readdirp    = afr_readdirp,
-
-	/* dir write */
-	.create      = afr_create,
-	.mknod       = afr_mknod,
-	.mkdir       = afr_mkdir,
-	.unlink      = afr_unlink,
-	.rmdir       = afr_rmdir,
-	.link        = afr_link,
-	.symlink     = afr_symlink,
-	.rename      = afr_rename,
+        .fsetattr    = afr_fsetattr,
+        .removexattr = afr_removexattr,
+        .fremovexattr = afr_fremovexattr,
+
+        /* dir read */
+        .opendir     = afr_opendir,
+        .readdir     = afr_readdir,
+        .readdirp    = afr_readdirp,
+
+        /* dir write */
+        .create      = afr_create,
+        .mknod       = afr_mknod,
+        .mkdir       = afr_mkdir,
+        .unlink      = afr_unlink,
+        .rmdir       = afr_rmdir,
+        .link        = afr_link,
+        .symlink     = afr_symlink,
+        .rename      = afr_rename,
 };
 
 
@@ -952,62 +503,247 @@ struct xlator_dumpops dumpops = {
 
 
 struct xlator_cbks cbks = {
-	.release     = afr_release,
-	.releasedir  = afr_releasedir,
+        .release     = afr_release,
+        .releasedir  = afr_releasedir,
+        .forget      = afr_forget,
 };
 
 
 struct volume_options options[] = {
-	{ .key  = {"read-subvolume" }, 
-	  .type = GF_OPTION_TYPE_XLATOR
-	},
-	{ .key  = {"favorite-child"}, 
-	  .type = GF_OPTION_TYPE_XLATOR
-	},
+        { .key  = {"read-subvolume" },
+          .type = GF_OPTION_TYPE_XLATOR,
+          .description = "inode-read fops happen only on one of the bricks in "
+                         "replicate. Afr will prefer the one specified using "
+                         "this option if it is not stale. Option value must be "
+                         "one of the xlator names of the children. "
+                         "Ex: <volname>-client-0 till "
+                         "<volname>-client-<number-of-bricks - 1>"
+        },
+        { .key  = {"read-subvolume-index" },
+          .type = GF_OPTION_TYPE_INT,
+          .default_value = "-1",
+          .description = "inode-read fops happen only on one of the bricks in "
+                         "replicate. AFR will prefer the one specified using "
+                         "this option if it is not stale. allowed options"
+                         " include -1 till replica-count - 1"
+        },
+        { .key = {"read-hash-mode" },
+          .type = GF_OPTION_TYPE_INT,
+          .min = 0,
+          .max = 2,
+          .default_value = "1",
+          .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 up server, "
+                         "1 = hash by GFID of file (all clients use "
+                                                    "same subvolume), "
+                         "2 = hash by GFID of file and client PID",
+        },
+        { .key  = {"choose-local" },
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "true",
+          .description = "Choose a local subvolume (i.e. Brick) to read from"
+	                 " if read-subvolume is not explicitly set.",
+        },
+        { .key  = {"favorite-child"},
+          .type = GF_OPTION_TYPE_XLATOR,
+          .description = "If a split-brain happens choose subvol/brick set by "
+                         "this option as source."
+        },
         { .key  = {"background-self-heal-count"},
           .type = GF_OPTION_TYPE_INT,
-          .min  = 0
+          .min  = 0,
+          .default_value = "16",
+          .validate = GF_OPT_VALIDATE_MIN,
+          .description = "This specifies the number of self-heals that can be "
+                         " performed in background without blocking the fop"
+        },
+        { .key  = {"data-self-heal"},
+          .type = GF_OPTION_TYPE_STR,
+          .value = {"1", "on", "yes", "true", "enable",
+                    "0", "off", "no", "false", "disable",
+                    "open"},
+          .default_value = "on",
+          .description   = "Using this option we can enable/disable data "
+                           "self-heal on the file. \"open\" means data "
+                           "self-heal action will only be triggered by file "
+                           "open operations."
         },
-	{ .key  = {"data-self-heal"},  
-	  .type = GF_OPTION_TYPE_BOOL 
-	},
         { .key  = {"data-self-heal-algorithm"},
-          .type = GF_OPTION_TYPE_STR
+          .type = GF_OPTION_TYPE_STR,
+          .description   = "Select between \"full\", \"diff\". The "
+                           "\"full\" algorithm copies the entire file from "
+                           "source to sink. The \"diff\" algorithm copies to "
+                           "sink only those blocks whose checksums don't match "
+                           "with those of source. If no option is configured "
+                           "the option is chosen dynamically as follows: "
+                           "If the file does not exist on one of the sinks "
+                           "or empty file exists or if the source file size is "
+                           "about the same as page size the entire file will "
+                           "be read and written i.e \"full\" algo, "
+                           "otherwise \"diff\" algo is chosen.",
+          .value = { "diff", "full"}
         },
         { .key  = {"data-self-heal-window-size"},
           .type = GF_OPTION_TYPE_INT,
           .min  = 1,
-          .max  = 1024
+          .max  = 1024,
+          .default_value = "1",
+          .description = "Maximum number blocks per file for which self-heal "
+                         "process would be applied simultaneously."
         },
-	{ .key  = {"metadata-self-heal"},  
-	  .type = GF_OPTION_TYPE_BOOL
-	},
-	{ .key  = {"entry-self-heal"},  
-	  .type = GF_OPTION_TYPE_BOOL 
-	},
-	{ .key  = {"data-change-log"},  
-	  .type = GF_OPTION_TYPE_BOOL 
-	},
-	{ .key  = {"metadata-change-log"},  
-	  .type = GF_OPTION_TYPE_BOOL
-	},
-	{ .key  = {"entry-change-log"},  
-	  .type = GF_OPTION_TYPE_BOOL
-	},
-	{ .key  = {"data-lock-server-count"},  
-	  .type = GF_OPTION_TYPE_INT, 
-	  .min  = 0
-	},
-	{ .key  = {"metadata-lock-server-count"},  
-	  .type = GF_OPTION_TYPE_INT, 
-	  .min  = 0
+        { .key  = {"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 = {"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 = {"pre-op-compat"},
+	  .type = GF_OPTION_TYPE_BOOL,
+	  .default_value = "on",
+	  .description = "Use separate pre-op xattrop() FOP rather than "
+	                 "overloading xdata of the OP"
 	},
-	{ .key  = {"entry-lock-server-count"},  
-	  .type = GF_OPTION_TYPE_INT,
-	  .min  = 0
+        { .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 "
+                         "acquisition 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 \"optimized\" transaction."
+
+        },
+        { .key = {"self-heal-daemon"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "on",
+          .description = "This option applies to only self-heal-daemon. "
+                         "Index directory crawl and automatic healing of files "
+                         "will not be performed if this option is turned off."
+        },
+        { .key = {"iam-self-heal-daemon"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "off",
+          .description = "This option differentiates if the replicate "
+                         "translator is running as part of self-heal-daemon "
+                         "or not."
+        },
+        { .key = {"quorum-type"},
+          .type = GF_OPTION_TYPE_STR,
+          .value = { "none", "auto", "fixed"},
+          .default_value = "none",
+          .description = "If value is \"fixed\" only allow writes if "
+                         "quorum-count bricks are present.  If value is "
+                         "\"auto\" only allow writes if more than half of "
+                         "bricks, or exactly half including the first, are "
+                         "present.",
+        },
+        { .key = {"quorum-count"},
+          .type = GF_OPTION_TYPE_INT,
+          .min = 1,
+          .max = INT_MAX,
+          .default_value = 0,
+          .description = "If quorum-type is \"fixed\" only allow writes if "
+                         "this many bricks or present.  Other quorum types "
+                         "will OVERWRITE this value.",
+        },
+        { .key  = {"node-uuid"},
+          .type = GF_OPTION_TYPE_STR,
+          .description = "Local glusterd uuid string, used in starting "
+                         "self-heal-daemon so that it can crawl only on "
+                         "local index directories.",
+        },
+        { .key  = {"post-op-delay-secs"},
+          .type = GF_OPTION_TYPE_INT,
+          .min  = 0,
+          .max  = INT_MAX,
+          .default_value = "1",
+          .description = "Time interval induced artificially before "
+	                 "post-operation phase of the transaction to "
+                         "enhance overlap of adjacent write operations.",
+        },
+        { .key = {AFR_SH_READDIR_SIZE_KEY},
+          .type = GF_OPTION_TYPE_SIZET,
+          .description = "readdirp size for performing entry self-heal",
+          .min = 1024,
+          .max = 131072,
+          .default_value = "1KB",
+        },
+        { .key = {"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 = {"afr-dirty-xattr"},
+	  .type = GF_OPTION_TYPE_STR,
+	  .default_value = AFR_DIRTY_DEFAULT,
 	},
-	{ .key  = {"strict-readdir"},
+	{ .key = {"metadata-splitbrain-forced-heal"},
 	  .type = GF_OPTION_TYPE_BOOL,
+	  .default_value = "off",
 	},
-	{ .key  = {NULL} },
+        { .key  = {NULL} },
 };
diff --git a/xlators/cluster/afr/src/afr.h b/xlators/cluster/afr/src/afr.h
index 758ac789a..36042f7b2 100644
--- a/xlators/cluster/afr/src/afr.h
+++ b/xlators/cluster/afr/src/afr.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 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.
 */
 
 
@@ -30,153 +21,115 @@
 #include "compat-errno.h"
 #include "afr-mem-types.h"
 
+#include "libxlator.h"
+#include "timer.h"
+#include "syncop.h"
+
+#include "afr-self-heald.h"
+
 #define AFR_XATTR_PREFIX "trusted.afr"
+#define AFR_PATHINFO_HEADER "REPLICATE:"
+#define AFR_SH_READDIR_SIZE_KEY "self-heal-readdir-size"
+#define AFR_SH_DATA_DOMAIN_FMT "%s:self-heal"
+#define AFR_DIRTY_DEFAULT AFR_XATTR_PREFIX ".dirty"
+#define AFR_DIRTY (((afr_private_t *) (THIS->private))->afr_dirty)
 
-struct _pump_private;
+#define AFR_LOCKEE_COUNT_MAX    3
+#define AFR_DOM_COUNT_MAX    3
+#define AFR_NUM_CHANGE_LOGS            3 /*data + metadata + entry*/
 
-typedef struct _afr_private {
-	gf_lock_t lock;               /* to guard access to child_count, etc */
-	unsigned int child_count;     /* total number of children   */
+typedef int (*afr_lock_cbk_t) (call_frame_t *frame, xlator_t *this);
+
+typedef int (*afr_read_txn_wind_t) (call_frame_t *frame, xlator_t *this, int subvol);
 
-        unsigned int read_child_rr;   /* round-robin index of the read_child */
-        gf_lock_t read_child_lock;    /* lock to protect above */
+typedef int (*afr_inode_refresh_cbk_t) (call_frame_t *frame, xlator_t *this, int err);
+
+typedef int (*afr_changelog_resume_t) (call_frame_t *frame, xlator_t *this);
+
+#define alloca0(size) ({void *__ptr; __ptr = alloca(size); memset(__ptr, 0, size); __ptr;})
+#define AFR_COUNT(array,max) ({int __i; int __res = 0; for (__i = 0; __i < max; __i++) if (array[__i]) __res++; __res;})
+#define AFR_INTERSECT(dst,src1,src2,max) ({int __i; for (__i = 0; __i < max; __i++) dst[__i] = src1[__i] && src2[__i];})
+
+typedef struct _afr_private {
+        gf_lock_t lock;               /* to guard access to child_count, etc */
+        unsigned int child_count;     /* total number of children   */
 
-	xlator_t **children;
+        xlator_t **children;
 
-        gf_lock_t root_inode_lk;
-        int first_lookup;
         inode_t *root_inode;
 
-	unsigned char *child_up;
+        unsigned char *child_up;
 
         char **pending_key;
 
-	gf_boolean_t data_self_heal;              /* on/off */
+        char         *data_self_heal;              /* on/off/open */
         char *       data_self_heal_algorithm;    /* name of algorithm */
         unsigned int data_self_heal_window_size;  /* max number of pipelined
                                                      read/writes */
 
         unsigned int background_self_heal_count;
         unsigned int background_self_heals_started;
-	gf_boolean_t metadata_self_heal;   /* on/off */
-	gf_boolean_t entry_self_heal;      /* on/off */
+        gf_boolean_t metadata_self_heal;   /* on/off */
+        gf_boolean_t entry_self_heal;      /* on/off */
 
-	gf_boolean_t data_change_log;       /* on/off */
-	gf_boolean_t metadata_change_log;   /* on/off */
-	gf_boolean_t entry_change_log;      /* on/off */
+        gf_boolean_t data_change_log;       /* on/off */
+        gf_boolean_t metadata_change_log;   /* on/off */
+        gf_boolean_t entry_change_log;      /* on/off */
 
-	int read_child;               /* read-subvolume */
-	unsigned int favorite_child;  /* subvolume to be preferred in resolving
-					 split-brain cases */
+	gf_boolean_t metadata_splitbrain_forced_heal; /* on/off */
+        int read_child;               /* read-subvolume */
+        unsigned int hash_mode;       /* for when read_child is not set */
+        int favorite_child;  /* subvolume to be preferred in resolving
+                                         split-brain cases */
 
-	unsigned int data_lock_server_count;
-	unsigned int metadata_lock_server_count;
-	unsigned int entry_lock_server_count;
+        gf_boolean_t inodelk_trace;
+        gf_boolean_t entrylk_trace;
 
-	gf_boolean_t inodelk_trace;
-	gf_boolean_t entrylk_trace;
-
-	gf_boolean_t strict_readdir;
-
-	unsigned int wait_count;      /* # of servers to wait for success */
+        unsigned int wait_count;      /* # of servers to wait for success */
 
         uint64_t up_count;      /* number of CHILD_UPs we have seen */
         uint64_t down_count;    /* number of CHILD_DOWNs we have seen */
 
-	struct _pump_private *pump_private; /* Set if we are loaded as pump */
-        int                   use_afr_in_pump;
-
-        pthread_mutex_t  mutex;
-        struct list_head saved_fds;   /* list of fds on which locks have succeeded */
-} afr_private_t;
-
-typedef struct {
-        /* External interface: These are variables (some optional) that
-           are set by whoever has triggered self-heal */
-
-        gf_boolean_t need_data_self_heal;
-        gf_boolean_t need_metadata_self_heal;
-        gf_boolean_t need_entry_self_heal;
-
-        gf_boolean_t forced_merge;        /* Is this a self-heal triggered to
-                                             forcibly merge the directories? */
-
-        gf_boolean_t healing_fd_opened;   /* true if caller has already
-                                             opened fd */
-
-        gf_boolean_t data_lock_held;      /* true if caller has already
-                                             acquired 0-0 lock */
-
-	fd_t *healing_fd;                 /* set if callers has opened fd */
-
-        gf_boolean_t background;          /* do self-heal in background
-                                             if possible */
-
-        ia_type_t type;                   /* st_mode of the entry we're doing
-                                             self-heal on */
-
-        /* Function to call to unwind. If self-heal is being done in the
-           background, this function will be called as soon as possible. */
-
-        int (*unwind) (call_frame_t *frame, xlator_t *this);
-
-        /* End of external interface members */
-
-
-	/* array of stat's, one for each child */
-	struct iatt *buf;
-        struct iatt parentbuf;
-
-	/* array of xattr's, one for each child */
-	dict_t **xattr;
-
-	/* array of errno's, one for each child */
-	int *child_errno;
-
-	int32_t **pending_matrix;
-	int32_t **delta_matrix;
-
-	int *sources;
-	int source;
-	int active_source;
-	int active_sinks;
-	int *success;
-	unsigned char *locked_nodes;
-        int lock_count;
-
-        mode_t impunging_entry_mode;
-        const char *linkname;
-
-	int   op_failed;
-
-	int   file_has_holes;
-	blksize_t block_size;
-	off_t file_size;
-	off_t offset;
-
-	loc_t parent_loc;
-
-        call_frame_t *orig_frame;
-        gf_boolean_t unwound;
-
-        /* private data for the particular self-heal algorithm */
-        void *private;
+        gf_boolean_t      optimistic_change_log;
+        gf_boolean_t      eager_lock;
+        gf_boolean_t      pre_op_compat;      /* on/off */
+	uint32_t          post_op_delay_secs;
+        unsigned int      quorum_count;
+
+        char                   vol_uuid[UUID_SIZE + 1];
+        int32_t                *last_event;
+
+	/* @event_generation: Keeps count of number of events received which can
+	   potentially impact consistency decisions. The events are CHILD_UP
+	   and CHILD_DOWN, when we have to recalculate the freshness/staleness
+	   of copies to detect if changes had happened while the other server
+	   was down. CHILD_DOWN and CHILD_UP can also be received on network
+	   disconnect/reconnects and not necessarily server going down/up.
+	   Recalculating freshness/staleness on network events is equally
+	   important as we might have had a network split brain.
+	*/
+	uint32_t               event_generation;
 
-        int (*flush_self_heal_cbk) (call_frame_t *frame, xlator_t *this);
+        gf_boolean_t           choose_local;
+        gf_boolean_t           did_discovery;
+        uint64_t               sh_readdir_size;
+        gf_boolean_t           ensure_durability;
+        char                   *sh_domain;
+	char                   *afr_dirty;
 
-	int (*completion_cbk) (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);
+	afr_self_heald_t       shd;
 
-	call_frame_t *sh_frame;
-} afr_self_heal_t;
+	/* pump dependencies */
+	void                   *pump_private;
+	gf_boolean_t           use_afr_in_pump;
+} afr_private_t;
 
 
 typedef enum {
-	AFR_DATA_TRANSACTION,          /* truncate, write, ... */
-	AFR_METADATA_TRANSACTION,      /* chmod, chown, ... */
-	AFR_ENTRY_TRANSACTION,         /* create, rmdir, ... */
-	AFR_ENTRY_RENAME_TRANSACTION,  /* rename */
+        AFR_DATA_TRANSACTION,          /* truncate, write, ... */
+        AFR_METADATA_TRANSACTION,      /* chmod, chown, ... */
+        AFR_ENTRY_TRANSACTION,         /* create, rmdir, ... */
+        AFR_ENTRY_RENAME_TRANSACTION,  /* rename */
 } afr_transaction_type;
 
 typedef enum {
@@ -232,11 +185,31 @@ 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;
@@ -245,451 +218,564 @@ typedef struct {
 
         unsigned char *locked_nodes;
         unsigned char *lower_locked_nodes;
-        unsigned char *inode_locked_nodes;
-        unsigned char *entry_locked_nodes;
 
         selfheal_lk_type_t selfheal_lk_type;
         transaction_lk_type_t transaction_lk_type;
 
         int32_t lock_count;
-        int32_t inodelk_lock_count;
         int32_t entrylk_lock_count;
 
         uint64_t lock_number;
         int32_t lk_call_count;
+        int32_t lk_expected_count;
+        int32_t lk_attempted_count;
 
         int32_t lock_op_ret;
         int32_t lock_op_errno;
+        afr_lock_cbk_t lock_cbk;
+        char *domain; /* Domain on which inode/entry lock/unlock in progress.*/
+} afr_internal_lock_t;
 
-        int (*lock_cbk) (call_frame_t *, xlator_t *);
+struct afr_reply {
+	int	valid;
+	int32_t	op_ret;
+	int32_t	op_errno;
+	dict_t *xdata;
+	struct iatt poststat;
+	struct iatt postparent;
+	struct iatt prestat;
+	struct iatt preparent;
+	struct iatt preparent2;
+	struct iatt postparent2;
+	uint8_t checksum[MD5_DIGEST_LENGTH];
+};
 
-} afr_internal_lock_t;
+typedef enum {
+        AFR_FD_NOT_OPENED,
+        AFR_FD_OPENED,
+        AFR_FD_OPENING
+} afr_fd_open_status_t;
+
+typedef struct {
+        unsigned int *pre_op_done[AFR_NUM_CHANGE_LOGS];
+	int inherited[AFR_NUM_CHANGE_LOGS];
+	int on_disk[AFR_NUM_CHANGE_LOGS];
+        afr_fd_open_status_t *opened_on; /* which subvolumes the fd is open on */
+
+        unsigned int *lock_piggyback;
+        unsigned int *lock_acquired;
+
+        int flags;
+
+	/* used for delayed-post-op optimization */
+	pthread_mutex_t    delay_lock;
+	gf_timer_t        *delay_timer;
+	call_frame_t      *delay_frame;
+
+	/* 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;
+
+	/* @open_fd_count:
+	   Number of open FDs queried from the server, as queried through
+	   xdata in FOPs. Currently, used to decide if eager-locking must be
+	   temporarily disabled.
+	*/
+        uint32_t        open_fd_count;
+
+
+	/* list of frames currently in progress */
+	struct list_head  eager_locked;
+} afr_fd_ctx_t;
 
-typedef struct _afr_locked_fd {
-        fd_t  *fd;
-        struct list_head list;
-} afr_locked_fd_t;
 
 typedef struct _afr_local {
-	unsigned int call_count;
-	unsigned int success_count;
-	unsigned int enoent_count;
+	glusterfs_fop_t  op;
+        unsigned int call_count;
 
+	/* @event_generation: copy of priv->event_generation taken at the
+	   time of starting the transaction. The copy is made so that we
+	   have a stable value through the various phases of the transaction.
+	*/
+	unsigned int event_generation;
 
-	unsigned int govinda_gOvinda;
+        uint32_t     open_fd_count;
+        gf_boolean_t update_open_fd_count;
 
-	unsigned int read_child_index;
-        unsigned char read_child_returned;
-        unsigned int first_up_child;
+	gf_lkowner_t  saved_lk_owner;
 
-        pid_t saved_pid;
+        int32_t op_ret;
+        int32_t op_errno;
 
-	int32_t op_ret;
-	int32_t op_errno;
+        int32_t **pending;
 
-	int32_t **pending;
+	int dirty[AFR_NUM_CHANGE_LOGS];
 
-	loc_t loc;
-	loc_t newloc;
+        loc_t loc;
+        loc_t newloc;
 
-	fd_t *fd;
+        fd_t *fd;
+	afr_fd_ctx_t *fd_ctx;
 
-	glusterfs_fop_t fop;
+	/* @child_up: copy of priv->child_up taken at the time of transaction
+	   start. The copy is taken so that we have a stable child_up array
+	   through the phases of the transaction as priv->child_up[i] can keep
+	   changing through time.
+	*/
+        unsigned char *child_up;
+
+	/* @read_attempted:
+	   array of flags representing subvolumes where read operations of
+	   the read transaction have already been attempted. The array is
+	   first pre-filled with down subvolumes, and as reads are performed
+	   on other subvolumes, those are set as well. This way if the read
+	   operation fails we do not retry on that subvolume again.
+	*/
+	unsigned char *read_attempted;
 
-	unsigned char *child_up;
+	/* @readfn:
 
-	int32_t *child_errno;
+	   pointer to function which will perform the read operation on a given
+	   subvolume. Used in read transactions.
+	*/
 
-	dict_t  *xattr_req;
+	afr_read_txn_wind_t readfn;
 
-	int32_t  inodelk_count;
-	int32_t  entrylk_count;
+	/* @refreshed:
 
-        afr_internal_lock_t internal_lock;
+	   the inode was "refreshed" (i.e, pending xattrs from all subvols
+	   freshly inspected and inode ctx updated accordingly) as part of
+	   this transaction already.
+	*/
+	gf_boolean_t refreshed;
 
-        afr_locked_fd_t *locked_fd;
-        int32_t          source_child;
-        int32_t          lock_recovery_child;
+	/* @inode:
 
-        dict_t  *dict;
+	   the inode on which the read txn is performed on. ref'ed and copied
+	   from either fd->inode or loc.inode
+	*/
 
-        int (*openfd_flush_cbk) (call_frame_t *frame, xlator_t *this);
+	inode_t *inode;
 
-	/*
-	   This struct contains the arguments for the "continuation"
-	   (scheme-like) of fops
+	/* @parent[2]:
+
+	   parent inode[s] on which directory transactions are performed.
 	*/
 
-	int   op;
-	struct {
-		struct {
-			unsigned char buf_set;
-			struct statvfs buf;
-		} statfs;
+	inode_t *parent;
+	inode_t *parent2;
 
-		struct {
-			inode_t *inode;
-			struct iatt buf;
-                        struct iatt read_child_buf;
-                        struct iatt postparent;
-                        ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			dict_t *xattr;
-			dict_t **xattrs;
-                        gf_boolean_t is_revalidate;
-		} lookup;
+	/* @readable:
 
-		struct {
-			int32_t flags;
-                        int32_t wbflags;
-		} open;
+	   array of flags representing servers from which a read can be
+	   performed. This is the output of afr_inode_refresh()
+	*/
+	unsigned char *readable;
 
-		struct {
-			int32_t cmd;
-                        struct gf_flock user_flock;
-                        struct gf_flock ret_flock;
-			unsigned char *locked_nodes;
-		} lk;
+	afr_inode_refresh_cbk_t refreshfn;
 
-		/* inode read */
+	/* @refreshinode:
 
-		struct {
-			int32_t mask;
-			int last_tried;  /* index of the child we tried previously */
-		} access;
+	   Inode currently getting refreshed.
+	*/
+	inode_t *refreshinode;
 
-		struct {
-			int last_tried;
-			ino_t ino;
-		} stat;
+	/*
+	  @pre_op_compat:
 
-		struct {
-			int last_tried;
-			ino_t ino;
-		} fstat;
+	  compatibility mode of pre-op. send a separate pre-op and
+	  op operations as part of transaction, rather than combining
+	*/
 
-		struct {
-			size_t size;
-			int last_tried;
-                        ino_t ino;
-		} readlink;
+	gf_boolean_t pre_op_compat;
 
-		struct {
-			char *name;
-			int last_tried;
-		} getxattr;
+        dict_t  *xattr_req;
 
-		struct {
-                        ino_t ino;
-			size_t size;
-			off_t offset;
-			int last_tried;
-		} readv;
+        afr_internal_lock_t internal_lock;
 
-		/* dir read */
+        dict_t  *dict;
 
-		struct {
-			int success_count;
-			int32_t op_ret;
-			int32_t op_errno;
+        int      optimistic_change_log;
+	gf_boolean_t      delayed_post_op;
 
-                        uint32_t *checksum;
-		} opendir;
+	/* 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;
 
-		struct {
-			int32_t op_ret;
-			int32_t op_errno;
-			size_t size;
-			off_t offset;
+	/* 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;
 
-                        gf_boolean_t failed;
-			int last_tried;
-		} readdir;
+        /*
+          This struct contains the arguments for the "continuation"
+          (scheme-like) of fops
+        */
 
-		struct {
-			int32_t op_ret;
-			int32_t op_errno;
+        struct {
+                struct {
+                        unsigned char buf_set;
+                        struct statvfs buf;
+                } statfs;
 
-			size_t size;
-			off_t offset;
-			int32_t flag;
+                struct {
+                        int32_t flags;
+                } open;
 
-			int last_tried;
-		} getdents;
+                struct {
+                        int32_t cmd;
+                        struct gf_flock user_flock;
+                        struct gf_flock ret_flock;
+                        unsigned char *locked_nodes;
+                } lk;
 
-		/* inode write */
+                /* inode read */
 
-		struct {
-			ino_t ino;
-			struct iatt prebuf;
-			struct iatt postbuf;
+                struct {
+                        int32_t mask;
+                        int last_index;  /* index of the child we tried previously */
+                } access;
 
-			int32_t op_ret;
+                struct {
+                        int last_index;
+                } stat;
 
-			struct iovec *vector;
-			struct iobref *iobref;
-			int32_t count;
-			off_t offset;
-		} writev;
+                struct {
+                        int last_index;
+                } fstat;
+
+                struct {
+                        size_t size;
+                        int last_index;
+                } readlink;
+
+                struct {
+                        char *name;
+                        int last_index;
+                        long xattr_len;
+                } getxattr;
+
+                struct {
+                        size_t size;
+                        off_t offset;
+                        int last_index;
+                        uint32_t flags;
+                } readv;
+
+                /* dir read */
+
+                struct {
+                        int success_count;
+                        int32_t op_ret;
+                        int32_t op_errno;
+
+                        uint32_t *checksum;
+                } opendir;
+
+                struct {
+                        int32_t op_ret;
+                        int32_t op_errno;
+                        size_t size;
+                        off_t offset;
+                        dict_t *dict;
+                        gf_boolean_t failed;
+                        int last_index;
+                } readdir;
+                /* inode write */
 
                 struct {
-                        ino_t ino;
                         struct iatt prebuf;
                         struct iatt postbuf;
-                } fsync;
+                } inode_wfop; //common structure for all inode-write-fops
 
-		struct {
-			ino_t ino;
-			off_t offset;
-			struct iatt prebuf;
-                        struct iatt postbuf;
-		} truncate;
+                struct {
+                        int32_t op_ret;
 
-		struct {
-			ino_t ino;
-			off_t offset;
-			struct iatt prebuf;
-                        struct iatt postbuf;
-		} ftruncate;
+                        struct iovec *vector;
+                        struct iobref *iobref;
+                        int32_t count;
+                        off_t offset;
+                        uint32_t flags;
+                } writev;
 
-		struct {
-			ino_t ino;
-			struct iatt in_buf;
+                struct {
+                        off_t offset;
+                } truncate;
+
+                struct {
+                        off_t offset;
+                } ftruncate;
+
+                struct {
+                        struct iatt in_buf;
                         int32_t valid;
-                        struct iatt preop_buf;
-                        struct iatt postop_buf;
-		} setattr;
+                } setattr;
 
-		struct {
-			ino_t ino;
-			struct iatt in_buf;
+                struct {
+                        struct iatt in_buf;
                         int32_t valid;
-                        struct iatt preop_buf;
-                        struct iatt postop_buf;
-		} fsetattr;
+                } fsetattr;
 
-		struct {
-			dict_t *dict;
-			int32_t flags;
-		} setxattr;
+                struct {
+                        dict_t *dict;
+                        int32_t flags;
+                } setxattr;
 
-		struct {
-			char *name;
-		} removexattr;
+                struct {
+                        dict_t *dict;
+                        int32_t flags;
+                } fsetxattr;
 
-		/* dir write */
+                struct {
+                        char *name;
+                } removexattr;
 
-		struct {
-			ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			fd_t *fd;
-                        dict_t *params;
-			int32_t flags;
-			mode_t mode;
-			inode_t *inode;
-			struct iatt buf;
+                struct {
+                        dict_t *xattr;
+                } xattrop;
+
+                struct {
+                        dict_t *xattr;
+                } fxattrop;
+
+                /* dir write */
+
+                struct {
+                        inode_t *inode;
+                        struct iatt buf;
                         struct iatt preparent;
                         struct iatt postparent;
-                        struct iatt read_child_buf;
-		} create;
+                        struct iatt prenewparent;
+                        struct iatt postnewparent;
+                } dir_fop; //common structure for all dir fops
 
-		struct {
-			ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			dev_t dev;
-			mode_t mode;
+                struct {
+                        fd_t *fd;
                         dict_t *params;
-			inode_t *inode;
-			struct iatt buf;
-                        struct iatt preparent;
-                        struct iatt postparent;
-                        struct iatt read_child_buf;
-		} mknod;
+                        int32_t flags;
+                        mode_t mode;
+                } create;
 
-		struct {
-			ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			int32_t mode;
+                struct {
+                        dev_t dev;
+                        mode_t mode;
                         dict_t *params;
-			inode_t *inode;
-			struct iatt buf;
-                        struct iatt read_child_buf;
-                        struct iatt preparent;
-                        struct iatt postparent;
-		} mkdir;
+                } mknod;
 
-		struct {
-                        ino_t parent_ino;
-			int32_t op_ret;
-			int32_t op_errno;
-                        struct iatt preparent;
-                        struct iatt postparent;
-		} unlink;
+                struct {
+                        int32_t mode;
+                        dict_t *params;
+                } mkdir;
 
-		struct {
-                        int   flags;
-                        ino_t parent_ino;
-			int32_t op_ret;
-			int32_t op_errno;
-                        struct iatt preparent;
-                        struct iatt postparent;
-		} rmdir;
+                struct {
+                        int flags;
+                } rmdir;
 
-		struct {
-                        ino_t oldparent_ino;
-                        ino_t newparent_ino;
-			ino_t ino;
-			struct iatt buf;
-                        struct iatt read_child_buf;
-                        struct iatt preoldparent;
-                        struct iatt prenewparent;
-                        struct iatt postoldparent;
-                        struct iatt postnewparent;
-		} rename;
+                struct {
+                        dict_t *params;
+                        char *linkpath;
+                } symlink;
 
 		struct {
-			ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			inode_t *inode;
-			struct iatt buf;
-                        struct iatt read_child_buf;
-                        struct iatt preparent;
-                        struct iatt postparent;
-		} link;
+			int32_t mode;
+			off_t offset;
+			size_t len;
+		} fallocate;
 
 		struct {
-			ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			inode_t *inode;
-                        dict_t *params;
-			struct iatt buf;
-                        struct iatt read_child_buf;
-			char *linkpath;
-                        struct iatt preparent;
-                        struct iatt postparent;
-		} symlink;
+			off_t offset;
+			size_t len;
+		} discard;
 
-		struct {
-			int32_t flags;
-			dir_entry_t *entries;
-			int32_t count;
-		} setdents;
-	} cont;
+                struct {
+                        off_t offset;
+                        off_t len;
+                        struct iatt prebuf;
+                        struct iatt postbuf;
+                } zerofill;
 
-	struct {
-		off_t start, len;
 
-		char *basename;
-		char *new_basename;
+        } cont;
 
-		loc_t parent_loc;
-		loc_t new_parent_loc;
+        struct {
+                off_t start, len;
 
-		afr_transaction_type type;
+                gf_boolean_t    eager_lock_on;
+                int *eager_lock;
 
-		int success_count;
-		int erase_pending;
-		int failure_count;
+                char *basename;
+                char *new_basename;
 
-		int last_tried;
-		int32_t *child_errno;
+                loc_t parent_loc;
+                loc_t new_parent_loc;
 
-		call_frame_t *main_frame;
+                afr_transaction_type type;
 
-		int (*fop) (call_frame_t *frame, xlator_t *this);
+		/* stub to resume on destruction
+		   of the transaction frame */
+		call_stub_t      *resume_stub;
 
-		int (*done) (call_frame_t *frame, xlator_t *this);
+		struct list_head  eager_locked;
 
-		int (*resume) (call_frame_t *frame, xlator_t *this);
+                unsigned char   *pre_op;
 
-		int (*unwind) (call_frame_t *frame, xlator_t *this);
+		/* @fop_subvols: subvolumes on which FOP will be attempted */
+                unsigned char   *fop_subvols;
 
-                /* post-op hook */
-	} transaction;
+		/* @failed_subvols: subvolumes on which FOP failed. Always
+		   a subset of @fop_subvols */
+                unsigned char   *failed_subvols;
 
-	afr_self_heal_t self_heal;
-} afr_local_t;
+		/* @dirtied: flag which indicates whether we set dirty flag
+		   in the OP. Typically true when we are performing operation
+		   on more than one subvol and optimistic changelog is disabled
 
+		   A 'true' value set in @dirtied flag means an 'undirtying'
+		   has to be done in POST-OP phase.
+		*/
+		gf_boolean_t  dirtied;
 
-typedef struct {
-        unsigned int *pre_op_done;
-        unsigned int *opened_on;     /* which subvolumes the fd is open on */
-        unsigned int *pre_op_piggyback;
+		/* @inherited: flag which indicates that the dirty flags
+		   of the previous transaction were inherited
+		*/
+		gf_boolean_t  inherited;
 
-        int flags;
-        int32_t wbflags;
-        uint64_t up_count;   /* number of CHILD_UPs this fd has seen */
-        uint64_t down_count; /* number of CHILD_DOWNs this fd has seen */
+		/*
+		  @no_uninherit: flag which indicates that a pre_op_uninherit()
+		  must _not_ be attempted (and returned as failure) always. This
+		  flag is set when a hard pre-op is performed, but not accounted
+		  for it in fd_ctx->on_disk[]. Such transactions are "isolated"
+		  from the pre-op piggybacking entirely and therefore uninherit
+		  must not be attempted.
+		*/
+		gf_boolean_t no_uninherit;
 
-        int32_t last_tried;
+		/* @uninherit_done:
+		   @uninherit_value:
 
-        int  hit, miss;
-        gf_boolean_t failed_over;
-        struct list_head entries; /* needed for readdir failover */
+		   The above pair variables make pre_op_uninherit() idempotent.
+		   Both are FALSE initially. The first call to pre_op_uninherit
+		   sets @uninherit_done to TRUE and the return value to
+		   @uninherit_value. Further calls will check for @uninherit_done
+		   to be TRUE and if so will simply return @uninherit_value.
+		*/
+		gf_boolean_t uninherit_done;
+		gf_boolean_t uninherit_value;
 
-        unsigned char *locked_on; /* which subvolumes locks have been successful */
-} afr_fd_ctx_t;
+		/* @changelog_resume: function to be called after changlogging
+		   (either pre-op or post-op) is done
+		*/
+
+		afr_changelog_resume_t changelog_resume;
+
+                call_frame_t *main_frame;
+
+                int (*wind) (call_frame_t *frame, xlator_t *this, int subvol);
+
+                int (*fop) (call_frame_t *frame, xlator_t *this);
 
+                int (*done) (call_frame_t *frame, xlator_t *this);
+
+                int (*resume) (call_frame_t *frame, xlator_t *this);
+
+                int (*unwind) (call_frame_t *frame, xlator_t *this);
+
+                /* post-op hook */
+        } transaction;
 
-/* try alloc and if it fails, goto label */
-#define ALLOC_OR_GOTO(var, type, label) do {                     \
-		var = GF_CALLOC (sizeof (type), 1,               \
-                                 gf_afr_mt_##type);              \
-		if (!var) {                                      \
-			gf_log (this->name, GF_LOG_ERROR,        \
-				"out of memory :(");             \
-			op_errno = ENOMEM;                       \
-			goto label;                              \
-		}                                                \
-	} while (0);
+	syncbarrier_t barrier;
+
+        struct marker_str     marker;
+
+        /* extra data for fops */
+        dict_t         *xdata_req;
+        dict_t         *xdata_rsp;
+
+        mode_t          umask;
+        int             xflag;
+        gf_boolean_t    do_discovery;
+	struct afr_reply *replies;
+} afr_local_t;
 
 
 /* did a call fail due to a child failing? */
-#define child_went_down(op_ret, op_errno) (((op_ret) < 0) &&	      \
-					   ((op_errno == ENOTCONN) || \
-					    (op_errno == EBADFD)))
+#define child_went_down(op_ret, op_errno) (((op_ret) < 0) &&            \
+                                           ((op_errno == ENOTCONN) ||   \
+                                            (op_errno == EBADFD)))
+
+int
+afr_inode_read_subvol_get (inode_t *inode, xlator_t *this,
+			   unsigned char *data_subvols,
+			   unsigned char *metadata_subvols,
+			   int *event_generation);
+int
+__afr_inode_read_subvol_get (inode_t *inode, xlator_t *this,
+			     unsigned char *data_subvols,
+			     unsigned char *metadata_subvols,
+			     int *event_generation);
+
+int
+__afr_inode_read_subvol_set (inode_t *inode, xlator_t *this,
+			     unsigned char *data_subvols,
+			     unsigned char *metadata_subvol,
+			     int event_generation);
+int
+afr_inode_read_subvol_set (inode_t *inode, xlator_t *this,
+			   unsigned char *data_subvols,
+			   unsigned char *metadata_subvols,
+			   int event_generation);
 
-#define afr_fop_failed(op_ret, op_errno) ((op_ret) == -1)
+int
+afr_inode_read_subvol_reset (inode_t *inode, xlator_t *this);
 
-/* have we tried all children? */
-#define all_tried(i, count)  ((i) == (count) - 1)
+int
+afr_read_subvol_select_by_policy (inode_t *inode, xlator_t *this,
+				  unsigned char *readable);
 
-int32_t
-afr_set_dict_gfid (dict_t *dict, uuid_t gfid);
+int
+afr_inode_read_subvol_type_get (inode_t *inode, xlator_t *this,
+				unsigned char *readable, int *event_p,
+				int type);
+int
+afr_read_subvol_get (inode_t *inode, xlator_t *this, int *subvol_p,
+		     int *event_p, afr_transaction_type type);
+
+#define afr_data_subvol_get(i, t, s, e) \
+	afr_read_subvol_get(i, t, s, e, AFR_DATA_TRANSACTION)
+
+#define afr_metadata_subvol_get(i, t, s, e) \
+	afr_read_subvol_get(i, t, s, e, AFR_METADATA_TRANSACTION)
 
 int
-pump_command_reply (call_frame_t *frame, xlator_t *this);
+afr_inode_refresh (call_frame_t *frame, xlator_t *this, inode_t *inode,
+		   afr_inode_refresh_cbk_t cbk);
 
 int32_t
-afr_notify (xlator_t *this, int32_t event,
-            void *data, ...);
+afr_notify (xlator_t *this, int32_t event, void *data, void *data2);
 
 int
-afr_attempt_lock_recovery (xlator_t *this, int32_t child_index);
+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_save_locked_fd (xlator_t *this, fd_t *fd);
+afr_attempt_lock_recovery (xlator_t *this, int32_t child_index);
 
 int
 afr_mark_locked_nodes (xlator_t *this, fd_t *fd,
                        unsigned char *locked_nodes);
 
 void
-afr_set_lk_owner (call_frame_t *frame, xlator_t *this);
+afr_set_lk_owner (call_frame_t *frame, xlator_t *this, void *lk_owner);
 
 int
 afr_set_lock_number (call_frame_t *frame, xlator_t *this);
 
-
-loc_t *
-lower_path (loc_t *l1, const char *b1, loc_t *l2, const char *b2);
-
 int32_t
 afr_unlock (call_frame_t *frame, xlator_t *this);
 
@@ -705,32 +791,30 @@ afr_blocking_lock (call_frame_t *frame, xlator_t *this);
 int
 afr_internal_lock_finish (call_frame_t *frame, xlator_t *this);
 
+int
+afr_lk_transfer_datalock (call_frame_t *dst, call_frame_t *src, char *dom,
+                          unsigned int child_count);
 
-int pump_start (call_frame_t *frame, xlator_t *this);
+int
+__afr_fd_ctx_set (xlator_t *this, fd_t *fd);
 
 int
 afr_fd_ctx_set (xlator_t *this, fd_t *fd);
 
-uint64_t
-afr_read_child (xlator_t *this, inode_t *inode);
-
-void
-afr_set_read_child (xlator_t *this, inode_t *inode, int32_t read_child);
-
-void
-afr_build_parent_loc (loc_t *parent, loc_t *child);
+afr_fd_ctx_t *
+afr_fd_ctx_get (fd_t *fd, xlator_t *this);
 
 int
-afr_up_children_count (int child_count, unsigned char *child_up);
+afr_build_parent_loc (loc_t *parent, loc_t *child, int32_t *op_errno);
 
 int
 afr_locked_nodes_count (unsigned char *locked_nodes, int child_count);
 
-ino64_t
-afr_itransform (ino64_t ino, int child_count, int child_index);
-
 int
-afr_deitransform (ino64_t ino, int child_count);
+afr_replies_interpret (call_frame_t *frame, xlator_t *this, inode_t *inode);
+
+void
+afr_replies_wipe (afr_local_t *local, afr_private_t *priv);
 
 void
 afr_local_cleanup (afr_local_t *local, xlator_t *this);
@@ -738,21 +822,9 @@ afr_local_cleanup (afr_local_t *local, xlator_t *this);
 int
 afr_frame_return (call_frame_t *frame);
 
-uint64_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);
-
 int
 afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
-          fd_t *fd, int32_t wbflags);
-
-void
-afr_set_opendir_done (xlator_t *this, inode_t *inode);
-
-uint64_t
-afr_is_opendir_done (xlator_t *this, inode_t *inode);
+          fd_t *fd, dict_t *xdata);
 
 void
 afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this);
@@ -760,163 +832,145 @@ afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this);
 int
 afr_cleanup_fd_ctx (xlator_t *this, fd_t *fd);
 
-int
-afr_openfd_flush (call_frame_t *frame, xlator_t *this, fd_t *fd);
-
-#define AFR_STACK_UNWIND(fop, frame, params ...)        \
-	do {						\
-		afr_local_t *__local = NULL;		\
-		xlator_t    *__this = NULL;		\
-                if (frame) {                            \
-                        __local = frame->local;		\
-                        __this = frame->this;           \
-                        frame->local = NULL;            \
+#define AFR_STACK_UNWIND(fop, frame, params ...)                \
+        do {                                                    \
+                afr_local_t *__local = NULL;                    \
+                xlator_t    *__this = NULL;                     \
+                if (frame) {                                    \
+                        __local = frame->local;                 \
+                        __this = frame->this;                   \
+                        frame->local = NULL;                    \
+                }                                               \
+                STACK_UNWIND_STRICT (fop, frame, params);       \
+                if (__local) {                                  \
+                        afr_local_cleanup (__local, __this);    \
+                        mem_put (__local);                      \
+                }                                               \
+        } while (0)
+
+#define AFR_STACK_DESTROY(frame)                                \
+        do {                                                    \
+                afr_local_t *__local = NULL;                    \
+                xlator_t    *__this = NULL;                     \
+                __local = frame->local;                         \
+                __this = frame->this;                           \
+                frame->local = NULL;                            \
+                STACK_DESTROY (frame->root);                    \
+                if (__local) {                                  \
+                        afr_local_cleanup (__local, __this);    \
+                        mem_put (__local);                      \
                 }                                               \
-		STACK_UNWIND_STRICT (fop, frame, params);       \
-		afr_local_cleanup (__local, __this);            \
-		GF_FREE (__local);				\
         } while (0);
 
-#define AFR_STACK_DESTROY(frame)			\
-	do {						\
-		afr_local_t *__local = NULL;		\
-		xlator_t    *__this = NULL;		\
-		__local = frame->local;			\
-		__this = frame->this;			\
-		frame->local = NULL;                    \
-		STACK_DESTROY (frame->root);		\
-		afr_local_cleanup (__local, __this);	\
-		GF_FREE (__local);			\
-        } while (0);
+#define AFR_FRAME_INIT(frame, op_errno)				       \
+	({frame->local = mem_get0 (THIS->local_pool);		       \
+	if (afr_local_init (frame->local, THIS->private, &op_errno)) { \
+		afr_local_cleanup (frame->local, THIS);		       \
+		mem_put (frame->local);				       \
+		frame->local = NULL; };				       \
+	frame->local;})
+
+#define AFR_STACK_RESET(frame) do { int opr; STACK_RESET (frame->root); AFR_FRAME_INIT(frame, opr);} while (0)
 
 /* allocate and return a string that is the basename of argument */
 static inline char *
 AFR_BASENAME (const char *str)
 {
-	char *__tmp_str = NULL;
-	char *__basename_str = NULL;
-	__tmp_str = gf_strdup (str);
-	__basename_str = gf_strdup (basename (__tmp_str));
-	GF_FREE (__tmp_str);
-	return __basename_str;
+        char *__tmp_str = NULL;
+        char *__basename_str = NULL;
+        __tmp_str = gf_strdup (str);
+        __basename_str = gf_strdup (basename (__tmp_str));
+        GF_FREE (__tmp_str);
+        return __basename_str;
 }
 
-/* initialize local_t */
-static inline int
-AFR_LOCAL_INIT (afr_local_t *local, afr_private_t *priv)
-{
-	local->child_up = GF_CALLOC (sizeof (*local->child_up),
-                                     priv->child_count,
-                                     gf_afr_mt_char);
-	if (!local->child_up) {
-		return -ENOMEM;
-	}
+call_frame_t *
+afr_copy_frame (call_frame_t *base);
 
-	memcpy (local->child_up, priv->child_up,
-		sizeof (*local->child_up) * priv->child_count);
+int
+afr_transaction_local_init (afr_local_t *local, xlator_t *this);
 
+int32_t
+afr_marker_getxattr (call_frame_t *frame, xlator_t *this,
+                     loc_t *loc, const char *name,afr_local_t *local, afr_private_t *priv );
 
-	local->call_count = afr_up_children_count (priv->child_count, local->child_up);
-	if (local->call_count == 0)
-		return -ENOTCONN;
+int
+afr_local_init (afr_local_t *local, afr_private_t *priv, int32_t *op_errno);
 
-	local->transaction.erase_pending = 1;
+int
+afr_internal_lock_init (afr_internal_lock_t *lk, size_t child_count,
+                        transaction_lk_type_t lk_type);
 
-	local->op_ret = -1;
-	local->op_errno = EUCLEAN;
+int
+afr_higher_errno (int32_t old_errno, int32_t new_errno);
 
-        local->internal_lock.lock_op_ret   = -1;
-        local->internal_lock.lock_op_errno = EUCLEAN;
+int
+afr_final_errno (afr_local_t *local, afr_private_t *priv);
 
+int
+afr_xattr_req_prepare (xlator_t *this, dict_t *xattr_req);
 
-	return 0;
-}
+void
+afr_fix_open (fd_t *fd, xlator_t *this);
 
+afr_fd_ctx_t *
+afr_fd_ctx_get (fd_t *fd, xlator_t *this);
 
-/**
- * first_up_child - return the index of the first child that is up
- */
+void
+afr_set_low_priority (call_frame_t *frame);
+int
+afr_child_fd_ctx_set (xlator_t *this, fd_t *fd, int32_t child,
+                      int flags);
 
-static inline int
-afr_first_up_child (afr_private_t *priv)
-{
-	xlator_t ** children = NULL;
-	int         ret      = -1;
-	int         i        = 0;
-
-	LOCK (&priv->lock);
-	{
-		children = priv->children;
-		for (i = 0; i < priv->child_count; i++) {
-			if (priv->child_up[i]) {
-				ret = i;
-				break;
-			}
-		}
-	}
-	UNLOCK (&priv->lock);
-
-	return ret;
-}
+gf_boolean_t
+afr_have_quorum (char *logname, afr_private_t *priv);
 
+void
+afr_matrix_cleanup (int32_t **pending, unsigned int m);
 
-static inline int
-afr_transaction_local_init (afr_local_t *local, afr_private_t *priv)
-{
-        int i;
-
-        local->first_up_child = afr_first_up_child (priv);
-
-	local->child_errno = GF_CALLOC (sizeof (*local->child_errno),
-				        priv->child_count,
-                                        gf_afr_mt_int32_t);
-	if (!local->child_errno) {
-		return -ENOMEM;
-	}
-
-	local->pending = GF_CALLOC (sizeof (*local->pending),
-                                    priv->child_count,
-                                    gf_afr_mt_int32_t);
-
-	if (!local->pending) {
-		return -ENOMEM;
-	}
-
-        for (i = 0; i < priv->child_count; i++) {
-                local->pending[i] = GF_CALLOC (sizeof (*local->pending[i]),
-                                               3, /* data + metadata + entry */
-                                               gf_afr_mt_int32_t);
-                if (!local->pending[i])
-                        return -ENOMEM;
-        }
+int32_t**
+afr_matrix_create (unsigned int m, unsigned int n);
 
-	local->internal_lock.inode_locked_nodes =
-                GF_CALLOC (sizeof (*local->internal_lock.inode_locked_nodes),
-                           priv->child_count,
-                           gf_afr_mt_char);
+void
+afr_filter_xattrs (dict_t *xattr);
 
-	local->internal_lock.entry_locked_nodes =
-                GF_CALLOC (sizeof (*local->internal_lock.entry_locked_nodes),
-                           priv->child_count,
-                           gf_afr_mt_char);
+/*
+ * Special value indicating we should use the "auto" quorum method instead of
+ * a fixed value (including zero to turn off quorum enforcement).
+ */
+#define AFR_QUORUM_AUTO INT_MAX
 
-	local->internal_lock.locked_nodes =
-                GF_CALLOC (sizeof (*local->internal_lock.locked_nodes),
-                           priv->child_count,
-                           gf_afr_mt_char);
+/*
+ * Having this as a macro will make debugging a bit weirder, but does reduce
+ * the probability of functions handling this check inconsistently.
+ */
+#define QUORUM_CHECK(_func,_label) do {                                  \
+        if (priv->quorum_count && !afr_have_quorum(this->name,priv)) { \
+                gf_log(this->name,GF_LOG_WARNING,                        \
+                       "failing "#_func" due to lack of quorum");        \
+                op_errno = EROFS;                                        \
+                goto _label;                                             \
+        }                                                                \
+} while (0);
 
-	local->internal_lock.lower_locked_nodes
-                = GF_CALLOC (sizeof (*local->internal_lock.lower_locked_nodes),
-                             priv->child_count,
-                             gf_afr_mt_char);
+int
+afr_fd_report_unstable_write (xlator_t *this, fd_t *fd);
 
-	local->transaction.child_errno = GF_CALLOC (sizeof (*local->transaction.child_errno),
-					            priv->child_count,
-                                                    gf_afr_mt_int32_t);
+gf_boolean_t
+afr_fd_has_witnessed_unstable_write (xlator_t *this, fd_t *fd);
 
-        local->internal_lock.transaction_lk_type = AFR_TRANSACTION_LK;
+void
+afr_delayed_changelog_wake_resume (xlator_t *this, fd_t *fd, call_stub_t *stub);
 
-	return 0;
-}
+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);
+
+int
+afr_local_pathinfo (char *pathinfo, gf_boolean_t *is_local);
 
+void
+afr_remove_eager_lock_stub (afr_local_t *local);
 #endif /* __AFR_H__ */
diff --git a/xlators/cluster/afr/src/pump.c b/xlators/cluster/afr/src/pump.c
index a162d3fa1..eed509956 100644
--- a/xlators/cluster/afr/src/pump.c
+++ b/xlators/cluster/afr/src/pump.c
@@ -1,25 +1,17 @@
 /*
-   Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include <unistd.h>
 #include <sys/time.h>
 #include <stdlib.h>
+#include <fnmatch.h>
 
 #ifndef _CONFIG_H
 #define _CONFIG_H
@@ -28,6 +20,129 @@
 
 #include "afr-common.c"
 #include "defaults.c"
+#include "glusterfs.h"
+#include "pump.h"
+
+
+static int
+afr_set_dict_gfid (dict_t *dict, uuid_t gfid)
+{
+        int ret       = 0;
+        uuid_t *pgfid = NULL;
+
+        GF_ASSERT (gfid);
+
+        pgfid = GF_CALLOC (1, sizeof (uuid_t), gf_common_mt_char);
+        if (!pgfid) {
+                ret = -1;
+                goto out;
+        }
+
+        uuid_copy (*pgfid, gfid);
+
+        ret = dict_set_dynptr (dict, "gfid-req", pgfid, sizeof (uuid_t));
+        if (ret)
+                gf_log (THIS->name, GF_LOG_ERROR, "gfid set failed");
+
+out:
+        if (ret && pgfid)
+                GF_FREE (pgfid);
+	return ret;
+}
+
+static int
+afr_set_root_gfid (dict_t *dict)
+{
+        uuid_t gfid;
+        int ret = 0;
+
+        memset (gfid, 0, 16);
+        gfid[15] = 1;
+
+        ret = afr_set_dict_gfid (dict, gfid);
+
+        return ret;
+}
+
+static int
+afr_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name)
+{
+        int   ret = -1;
+        uuid_t pargfid = {0};
+
+        if (!child)
+                goto out;
+
+        if (!uuid_is_null (parent->inode->gfid))
+                uuid_copy (pargfid, parent->inode->gfid);
+        else if (!uuid_is_null (parent->gfid))
+                uuid_copy (pargfid, parent->gfid);
+
+        if (uuid_is_null (pargfid))
+                goto out;
+
+        if (strcmp (parent->path, "/") == 0)
+                ret = gf_asprintf ((char **)&child->path, "/%s", name);
+        else
+                ret = gf_asprintf ((char **)&child->path, "%s/%s", parent->path,
+                                   name);
+
+        if (-1 == ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "asprintf failed while setting child path");
+        }
+
+        child->name = strrchr (child->path, '/');
+        if (child->name)
+                child->name++;
+
+        child->parent = inode_ref (parent->inode);
+        child->inode = inode_new (parent->inode->table);
+        uuid_copy (child->pargfid, pargfid);
+
+        if (!child->inode) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = 0;
+out:
+        if ((ret == -1) && child)
+                loc_wipe (child);
+
+        return ret;
+}
+
+static void
+afr_build_root_loc (xlator_t *this, loc_t *loc)
+{
+        afr_private_t   *priv = NULL;
+
+        priv = this->private;
+        loc->path = gf_strdup ("/");
+        loc->name = "";
+        loc->inode = inode_ref (priv->root_inode);
+        uuid_copy (loc->gfid, loc->inode->gfid);
+}
+
+static void
+afr_update_loc_gfids (loc_t *loc, struct iatt *buf, struct iatt *postparent)
+{
+        GF_ASSERT (loc);
+        GF_ASSERT (buf);
+
+        uuid_copy (loc->gfid, buf->ia_gfid);
+        if (postparent)
+                uuid_copy (loc->pargfid, postparent->ia_gfid);
+}
+
+static uint64_t pump_pid = 0;
+static inline void
+pump_fill_loc_info (loc_t *loc, struct iatt *iatt, struct iatt *parent)
+{
+        afr_update_loc_gfids (loc, iatt, parent);
+        uuid_copy (loc->inode->gfid, iatt->ia_gfid);
+}
 
 static int
 pump_mark_start_pending (xlator_t *this)
@@ -139,88 +254,20 @@ pump_set_resume_path (xlator_t *this, const char *path)
 
         LOCK (&pump_priv->resume_path_lock);
         {
-                pump_priv->resume_path = strdup (path);
-                if (!pump_priv->resume_path)
-                        ret = -1;
+                strncpy (pump_priv->resume_path, path, strlen (path) + 1);
         }
         UNLOCK (&pump_priv->resume_path_lock);
 
         return ret;
 }
 
-static void
-build_child_loc (loc_t *parent, loc_t *child, char *path, char *name)
-{
-        child->path = path;
-        child->name = name;
-
-        child->parent = inode_ref (parent->inode);
-        child->inode = inode_new (parent->inode->table);
-}
-
-static char *
-build_file_path (loc_t *loc, gf_dirent_t *entry)
-{
-        xlator_t *this = NULL;
-        char *file_path = NULL;
-        int pathlen = 0;
-        int total_size = 0;
-
-        this = THIS;
-
-        pathlen = STRLEN_0 (loc->path);
-
-        if (IS_ROOT_PATH (loc->path)) {
-                total_size = pathlen + entry->d_len;
-                file_path = GF_CALLOC (1, total_size, gf_afr_mt_char);
-                if (!file_path) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        return NULL;
-                }
-
-                gf_log (this->name, GF_LOG_TRACE,
-                        "constructing file path of size=%d"
-                        "pathlen=%d, d_len=%d",
-                        total_size, pathlen,
-                        entry->d_len);
-
-                snprintf(file_path, total_size, "%s%s", loc->path, entry->d_name);
-
-        } else {
-                total_size = pathlen + entry->d_len + 1; /* for the extra '/' in the path */
-                file_path = GF_CALLOC (1, total_size + 1, gf_afr_mt_char);
-                if (!file_path) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        return NULL;
-                }
-
-                gf_log (this->name, GF_LOG_TRACE,
-                        "constructing file path of size=%d"
-                        "pathlen=%d, d_len=%d",
-                        total_size, pathlen,
-                        entry->d_len);
-
-                snprintf(file_path, total_size, "%s/%s", loc->path, entry->d_name);
-        }
-
-        gf_log (this->name, GF_LOG_TRACE,
-                "path=%s and d_name=%s", loc->path, entry->d_name);
-        gf_log (this->name, GF_LOG_TRACE,
-                "constructed file_path=%s of size=%d", file_path, total_size);
-
-        return file_path;
-}
-
 static int
 pump_save_path (xlator_t *this, const char *path)
 {
         afr_private_t *priv = NULL;
-        pump_private_t *pump_priv = NULL;
         pump_state_t state;
         dict_t *dict = NULL;
-        loc_t  loc;
+        loc_t  loc = {0};
         int dict_ret = 0;
         int ret = -1;
 
@@ -229,29 +276,30 @@ pump_save_path (xlator_t *this, const char *path)
                 return 0;
 
         priv = this->private;
-        pump_priv = priv->pump_private;
 
         GF_ASSERT (priv->root_inode);
 
-        build_root_loc (priv->root_inode, &loc);
+        afr_build_root_loc (this, &loc);
 
         dict = dict_new ();
         dict_ret = dict_set_str (dict, PUMP_PATH, (char *)path);
+        if (dict_ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to set the key %s", path, PUMP_PATH);
 
         ret = syncop_setxattr (PUMP_SOURCE_CHILD (this), &loc, dict, 0);
 
         if (ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
+                gf_log (this->name, GF_LOG_INFO,
                         "setxattr failed - could not save path=%s", path);
         } else {
                 gf_log (this->name, GF_LOG_DEBUG,
                         "setxattr succeeded - saved path=%s", path);
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Saving path for status info");
         }
 
         dict_unref (dict);
 
+        loc_wipe (&loc);
         return 0;
 }
 
@@ -314,15 +362,9 @@ pump_get_resume_path (xlator_t *this)
 static int
 pump_update_resume_state (xlator_t *this, const char *path)
 {
-        afr_private_t *priv = NULL;
-        pump_private_t *pump_priv = NULL;
-
         pump_state_t state;
         const char *resume_path = NULL;
 
-        priv = this->private;
-        pump_priv = priv->pump_private;
-
         state = pump_get_state ();
 
         if (state == PUMP_STATE_RESUME) {
@@ -350,16 +392,10 @@ pump_update_resume_state (xlator_t *this, const char *path)
 static gf_boolean_t
 is_pump_traversal_allowed (xlator_t *this, const char *path)
 {
-        afr_private_t *priv = NULL;
-        pump_private_t *pump_priv = NULL;
-
         pump_state_t state;
         const char *resume_path = NULL;
         gf_boolean_t ret = _gf_true;
 
-        priv = this->private;
-        pump_priv = priv->pump_private;
-
         state = pump_get_state ();
 
         if (state == PUMP_STATE_RESUME) {
@@ -402,38 +438,33 @@ pump_save_file_stats (xlator_t *this, const char *path)
 static int
 gf_pump_traverse_directory (loc_t *loc)
 {
-        xlator_t *this = NULL;
-        afr_private_t *priv = NULL;
-        fd_t     *fd   = NULL;
-
-        off_t       offset   = 0;
-        loc_t       entry_loc;
-        gf_dirent_t *entry = NULL;
-        gf_dirent_t *tmp = NULL;
-        gf_dirent_t entries;
-
-	struct iatt iatt, parent;
-	dict_t *xattr_rsp;
-
-        int source = 0;
-
-        char *file_path = NULL;
-        int ret = 0;
+        xlator_t        *this              = NULL;
+        fd_t            *fd                = NULL;
+        off_t           offset             = 0;
+        loc_t           entry_loc          = {0};
+        gf_dirent_t     *entry             = NULL;
+        gf_dirent_t     *tmp               = NULL;
+        gf_dirent_t     entries;
+	struct iatt     iatt               = {0};
+        struct iatt     parent             = {0};
+	dict_t          *xattr_rsp         = NULL;
+        int             ret                = 0;
+        gf_boolean_t    is_directory_empty = _gf_true;
+        gf_boolean_t    free_entries       = _gf_false;
 
         INIT_LIST_HEAD (&entries.list);
         this = THIS;
-        priv = this->private;
 
         GF_ASSERT (loc->inode);
 
-	fd = fd_create (loc->inode, PUMP_PID);
+	fd = fd_create (loc->inode, pump_pid);
         if (!fd) {
                 gf_log (this->name, GF_LOG_ERROR,
                         "Failed to create fd for %s", loc->path);
                 goto out;
         }
 
-        ret = syncop_opendir (priv->children[source], loc, fd);
+        ret = syncop_opendir (this, loc, fd);
         if (ret < 0) {
                 gf_log (this->name, GF_LOG_DEBUG,
                         "opendir failed on %s", loc->path);
@@ -444,7 +475,8 @@ gf_pump_traverse_directory (loc_t *loc)
                 "pump opendir on %s returned=%d",
                 loc->path, ret);
 
-        while (syncop_readdirp (priv->children[source], fd, 131072, offset, &entries)) {
+        while (syncop_readdirp (this, fd, 131072, offset, NULL, &entries)) {
+                free_entries = _gf_true;
 
                 if (list_empty (&entries.list)) {
                         gf_log (this->name, GF_LOG_TRACE,
@@ -456,105 +488,108 @@ gf_pump_traverse_directory (loc_t *loc)
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "found readdir entry=%s", entry->d_name);
 
-                        file_path = build_file_path (loc, entry);
-                        if (!file_path) {
-                                gf_log (this->name, GF_LOG_DEBUG,
-                                        "file path construction failed");
-                                goto out;
+                        offset = entry->d_off;
+                        if (uuid_is_null (entry->d_stat.ia_gfid)) {
+                                gf_log (this->name, GF_LOG_WARNING, "%s/%s: No "
+                                        "gfid present skipping",
+                                        loc->path, entry->d_name);
+                                continue;
                         }
+                        loc_wipe (&entry_loc);
+                        ret = afr_build_child_loc (this, &entry_loc, loc,
+                                                   entry->d_name);
+                        if (ret)
+                                goto out;
 
-                        build_child_loc (loc, &entry_loc, file_path, entry->d_name);
-
-                        if (!IS_ENTRY_CWD (entry->d_name) &&
-                                           !IS_ENTRY_PARENT (entry->d_name)) {
-
-                                    ret = syncop_lookup (this, &entry_loc, NULL,
-                                                         &iatt, &xattr_rsp, &parent);
+			if ((strcmp (entry->d_name, ".") == 0) ||
+			    (strcmp (entry->d_name, "..") == 0))
+				continue;
 
-                                    entry_loc.ino = iatt.ia_ino;
-                                    entry_loc.inode->ino = iatt.ia_ino;
-                                    memcpy (entry_loc.inode->gfid, iatt.ia_gfid, 16);
+			is_directory_empty = _gf_false;
+			gf_log (this->name, GF_LOG_DEBUG,
+				"lookup %s => %"PRId64,
+				entry_loc.path,
+				iatt.ia_ino);
 
-                                    gf_log (this->name, GF_LOG_DEBUG,
-                                            "lookup %s => %"PRId64,
-                                            entry_loc.path,
-                                            iatt.ia_ino);
+			ret = syncop_lookup (this, &entry_loc, NULL, &iatt,
+					     &xattr_rsp, &parent);
 
-                                    ret = syncop_lookup (this, &entry_loc, NULL,
-                                                         &iatt, &xattr_rsp, &parent);
+			if (ret) {
+				gf_log (this->name, GF_LOG_ERROR,
+					"%s: lookup failed", entry_loc.path);
+				continue;
+			}
 
+			ret = afr_selfheal_name (this, loc->gfid, entry->d_name);
+			if (ret) {
+				gf_log (this->name, GF_LOG_ERROR,
+					"%s: name self-heal failed (%s/%s)",
+					entry_loc.path, uuid_utoa (loc->gfid),
+					entry->d_name);
+				continue;
+			}
 
-                                    gf_log (this->name, GF_LOG_DEBUG,
-                                            "second lookup ret=%d: %s => %"PRId64,
-                                            ret,
-                                            entry_loc.path,
-                                            iatt.ia_ino);
+			ret = afr_selfheal (this, iatt.ia_gfid);
+			if (ret) {
+				gf_log (this->name, GF_LOG_ERROR,
+					"%s: self-heal failed (%s)",
+					entry_loc.path, uuid_utoa (iatt.ia_gfid));
+				continue;
+			}
 
-                                    pump_update_resume_state (this, entry_loc.path);
+			pump_fill_loc_info (&entry_loc, &iatt, &parent);
 
-                                    pump_save_path (this, entry_loc.path);
-                                    pump_save_file_stats (this, entry_loc.path);
+			pump_update_resume_state (this, entry_loc.path);
 
-                                    ret = pump_check_and_update_status (this);
-                                    if (ret < 0) {
-                                            gf_log (this->name, GF_LOG_DEBUG,
-                                                    "Pump beginning to exit out");
-                                            goto out;
-                                    }
+			pump_save_path (this, entry_loc.path);
+			pump_save_file_stats (this, entry_loc.path);
 
-                                    gf_log (this->name, GF_LOG_TRACE,
-                                            "type of file=%d, IFDIR=%d",
-                                            iatt.ia_type, IA_IFDIR);
+			ret = pump_check_and_update_status (this);
+			if (ret < 0) {
+				gf_log (this->name, GF_LOG_DEBUG,
+					"Pump beginning to exit out");
+				goto out;
+			}
 
-                                    if (IA_ISDIR (iatt.ia_type)) {
-                                            if (is_pump_traversal_allowed (this, entry_loc.path)) {
-                                                    gf_log (this->name, GF_LOG_TRACE,
-                                                            "entering dir=%s",
-                                                            entry->d_name);
-                                                    gf_pump_traverse_directory (&entry_loc);
-                                            }
-                                    }
-                            }
-                        offset = entry->d_off;
-                        loc_wipe (&entry_loc);
+			if (IA_ISDIR (iatt.ia_type)) {
+				if (is_pump_traversal_allowed (this, entry_loc.path)) {
+					gf_log (this->name, GF_LOG_TRACE,
+						"entering dir=%s", entry->d_name);
+					gf_pump_traverse_directory (&entry_loc);
+				}
+                        }
                 }
 
                 gf_dirent_free (&entries);
-                gf_log (this->name, GF_LOG_TRACE,
-                        "offset incremented to %d",
+                free_entries = _gf_false;
+                gf_log (this->name, GF_LOG_TRACE, "offset incremented to %d",
                         (int32_t ) offset);
 
         }
 
-out:
-        return 0;
-
-}
+        ret = syncop_close (fd);
+        if (ret < 0)
+                gf_log (this->name, GF_LOG_DEBUG, "closing the fd failed");
 
-void
-build_root_loc (inode_t *inode, loc_t *loc)
-{
-        loc->path = "/";
-        loc->name = "";
-        loc->inode = inode;
-        loc->ino = 1;
-        loc->inode->ino = 1;
-        memset (loc->inode->gfid, 0, 16);
-        loc->inode->gfid[15] = 1;
+        if (is_directory_empty && (strcmp (loc->path, "/") == 0)) {
+               pump_change_state (this, PUMP_STATE_RUNNING);
+               gf_log (this->name, GF_LOG_INFO, "Empty source brick. "
+                                "Nothing to be done.");
+        }
 
+out:
+        if (entry_loc.path)
+                loc_wipe (&entry_loc);
+        if (free_entries)
+                gf_dirent_free (&entries);
+        return 0;
 }
 
 static int
 pump_update_resume_path (xlator_t *this)
 {
-        afr_private_t *priv = NULL;
-        pump_private_t *pump_priv = NULL;
-
         const char *resume_path = NULL;
 
-        priv = this->private;
-        pump_priv = priv->pump_private;
-
         resume_path = pump_get_resume_path (this);
 
         if (resume_path) {
@@ -573,6 +608,40 @@ pump_update_resume_path (xlator_t *this)
         return 0;
 }
 
+static int32_t
+pump_xattr_cleaner (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        afr_private_t  *priv      = NULL;
+        loc_t           loc       = {0};
+        int             i         = 0;
+        int             ret       = 0;
+        int             source    = 0;
+        int             sink      = 1;
+
+        priv      = this->private;
+
+        afr_build_root_loc (this, &loc);
+
+        ret = syncop_removexattr (priv->children[source], &loc,
+				  PUMP_PATH, 0);
+
+        ret = syncop_removexattr (priv->children[sink], &loc,
+                                  PUMP_SINK_COMPLETE, 0);
+
+        for (i = 0; i < priv->child_count; i++) {
+                ret = syncop_removexattr (priv->children[i], &loc,
+                                          PUMP_SOURCE_COMPLETE, 0);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_DEBUG, "removexattr "
+                                "failed with %s", strerror (-ret));
+                }
+        }
+
+        loc_wipe (&loc);
+        return pump_command_reply (frame, this);
+}
+
 static int
 pump_complete_migration (xlator_t *this)
 {
@@ -580,7 +649,7 @@ pump_complete_migration (xlator_t *this)
         pump_private_t *pump_priv = NULL;
         dict_t *dict = NULL;
         pump_state_t state;
-        loc_t  loc;
+        loc_t  loc = {0};
         int dict_ret = 0;
         int ret = -1;
 
@@ -589,7 +658,7 @@ pump_complete_migration (xlator_t *this)
 
         GF_ASSERT (priv->root_inode);
 
-        build_root_loc (priv->root_inode, &loc);
+        afr_build_root_loc (this, &loc);
 
         dict = dict_new ();
 
@@ -601,6 +670,10 @@ pump_complete_migration (xlator_t *this)
                 pump_priv->pump_finished = _gf_true;
 
                 dict_ret = dict_set_str (dict, PUMP_SOURCE_COMPLETE, "jargon");
+                if (dict_ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to set the key %s",
+                                loc.path, PUMP_SOURCE_COMPLETE);
 
                 ret = syncop_setxattr (PUMP_SOURCE_CHILD (this), &loc, dict, 0);
                 if (ret < 0) {
@@ -608,29 +681,27 @@ pump_complete_migration (xlator_t *this)
                                 "setxattr failed - while  notifying source complete");
                 }
                 dict_ret = dict_set_str (dict, PUMP_SINK_COMPLETE, "jargon");
+                if (dict_ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to set the key %s",
+                                loc.path, PUMP_SINK_COMPLETE);
 
                 ret = syncop_setxattr (PUMP_SINK_CHILD (this), &loc, dict, 0);
                 if (ret < 0) {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "setxattr failed - while notifying sink complete");
                 }
-        }
 
-        return 0;
-}
-
-static int
-pump_set_root_gfid (dict_t *dict)
-{
-        uuid_t gfid;
-        int ret = 0;
-
-        memset (gfid, 0, 16);
-        gfid[15] = 1;
+                pump_save_path (this, "/");
 
-        ret = afr_set_dict_gfid (dict, gfid);
+        } else if (state == PUMP_STATE_ABORT) {
+                gf_log (this->name, GF_LOG_DEBUG, "Starting cleanup "
+                        "of pump internal xattrs");
+                call_resume (pump_priv->cleaner);
+        }
 
-        return ret;
+        loc_wipe (&loc);
+        return 0;
 }
 
 static int
@@ -646,7 +717,7 @@ pump_lookup_sink (loc_t *loc)
 
         xattr_req = dict_new ();
 
-        ret = pump_set_root_gfid (xattr_req);
+        ret = afr_set_root_gfid (xattr_req);
         if (ret)
                 goto out;
 
@@ -656,6 +727,7 @@ pump_lookup_sink (loc_t *loc)
         if (ret) {
                 gf_log (this->name, GF_LOG_DEBUG,
                         "Lookup on sink child failed");
+                ret = -1;
                 goto out;
         }
 
@@ -673,7 +745,7 @@ pump_task (void *data)
         afr_private_t *priv = NULL;
 
 
-        loc_t loc;
+        loc_t loc = {0};
 	struct iatt iatt, parent;
 	dict_t *xattr_rsp = NULL;
         dict_t *xattr_req = NULL;
@@ -685,7 +757,7 @@ pump_task (void *data)
 
         GF_ASSERT (priv->root_inode);
 
-        build_root_loc (priv->root_inode, &loc);
+        afr_build_root_loc (this, &loc);
         xattr_req = dict_new ();
         if (!xattr_req) {
                 gf_log (this->name, GF_LOG_DEBUG,
@@ -694,14 +766,13 @@ pump_task (void *data)
                 goto out;
         }
 
-        pump_set_root_gfid (xattr_req);
+        afr_set_root_gfid (xattr_req);
         ret = syncop_lookup (this, &loc, xattr_req,
                              &iatt, &xattr_rsp, &parent);
 
         gf_log (this->name, GF_LOG_TRACE,
-                "lookup: ino=%"PRId64", path=%s",
-                loc.ino,
-                loc.path);
+                "lookup: path=%s gfid=%s",
+                loc.path, uuid_utoa (loc.inode->gfid));
 
         ret = pump_check_and_update_status (this);
         if (ret < 0) {
@@ -710,7 +781,7 @@ pump_task (void *data)
 
         pump_update_resume_path (this);
 
-        pump_set_root_gfid (xattr_req);
+        afr_set_root_gfid (xattr_req);
         ret = pump_lookup_sink (&loc);
         if (ret) {
                 pump_update_resume_path (this);
@@ -724,26 +795,23 @@ out:
         if (xattr_req)
                 dict_unref (xattr_req);
 
+        loc_wipe (&loc);
 	return 0;
 }
 
 
 static int
-pump_task_completion (int ret, void *data)
+pump_task_completion (int ret, call_frame_t *sync_frame, void *data)
 {
         xlator_t *this = NULL;
-        call_frame_t *frame = NULL;
         afr_private_t *priv = NULL;
-        pump_private_t *pump_priv = NULL;
 
         this = THIS;
 
-        frame = (call_frame_t *) data;
-
         priv = this->private;
-        pump_priv = priv->pump_private;
 
         inode_unref (priv->root_inode);
+        STACK_DESTROY (sync_frame->root);
 
         gf_log (this->name, GF_LOG_DEBUG,
                 "Pump xlator exiting");
@@ -761,21 +829,22 @@ pump_start (call_frame_t *pump_frame, xlator_t *this)
 	priv = this->private;
         pump_priv = priv->pump_private;
 
-        if (!pump_frame->root->lk_owner)
-                pump_frame->root->lk_owner = PUMP_LK_OWNER;
+        afr_set_lk_owner (pump_frame, this, pump_frame->root);
+	pump_pid = (uint64_t) (unsigned long)pump_frame->root;
 
 	ret = synctask_new (pump_priv->env, pump_task,
                             pump_task_completion,
-                            pump_frame);
+                            pump_frame, NULL);
         if (ret == -1) {
-                gf_log (this->name, GF_LOG_DEBUG,
+                gf_log (this->name, GF_LOG_ERROR,
                         "starting pump failed");
                 pump_change_state (this, PUMP_STATE_ABORT);
                 goto out;
         }
 
-        gf_log (this->name, GF_LOG_TRACE,
-                "setting pump as started");
+        gf_log (this->name, GF_LOG_DEBUG,
+                "setting pump as started lk_owner: %s %"PRIu64,
+                lkowner_utoa (&pump_frame->root->lk_owner), pump_pid);
 
         priv->use_afr_in_pump = 1;
 out:
@@ -809,9 +878,10 @@ pump_cmd_start_setxattr_cbk (call_frame_t *frame,
                              void *cookie,
                              xlator_t *this,
                              int32_t op_ret,
-                             int32_t op_errno)
+                             int32_t op_errno, dict_t *xdata)
 
 {
+        call_frame_t *prev = NULL;
         afr_local_t *local = NULL;
         int ret = 0;
 
@@ -831,6 +901,11 @@ pump_cmd_start_setxattr_cbk (call_frame_t *frame,
 
         pump_mark_start_pending (this);
 
+        /* send the PARENT_UP as pump is ready now */
+        prev = cookie;
+        if (prev && prev->this)
+                prev->this->notify (prev->this, GF_EVENT_PARENT_UP, this);
+
 out:
         local->op_ret = ret;
         pump_command_reply (frame, this);
@@ -844,8 +919,9 @@ pump_initiate_sink_connect (call_frame_t *frame, xlator_t *this)
         afr_local_t   *local     = NULL;
         afr_private_t *priv      = NULL;
         dict_t        *dict      = NULL;
-        char          *dst_brick = NULL;
-        loc_t loc;
+        data_t        *data      = NULL;
+        char          *clnt_cmd  = NULL;
+        loc_t loc = {0};
 
         int ret = 0;
 
@@ -854,10 +930,11 @@ pump_initiate_sink_connect (call_frame_t *frame, xlator_t *this)
 
         GF_ASSERT (priv->root_inode);
 
-        build_root_loc (priv->root_inode, &loc);
+        afr_build_root_loc (this, &loc);
 
-        ret = dict_get_str (local->dict, PUMP_CMD_START, &dst_brick);
-        if (ret < 0) {
+        data = data_ref (dict_get (local->dict, RB_PUMP_CMD_START));
+        if (!data) {
+                ret = -1;
                 gf_log (this->name, GF_LOG_ERROR,
                         "Could not get destination brick value");
                 goto out;
@@ -865,17 +942,22 @@ pump_initiate_sink_connect (call_frame_t *frame, xlator_t *this)
 
         dict = dict_new ();
         if (!dict) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
                 ret = -1;
                 goto out;
         }
 
-        GF_ASSERT (dst_brick);
-        gf_log (this->name, GF_LOG_DEBUG,
-                "Got destination brick as %s", dst_brick);
+        clnt_cmd = GF_CALLOC (1, data->len+1, gf_common_mt_char);
+        if (!clnt_cmd) {
+                ret = -1;
+                goto out;
+        }
+
+        memcpy (clnt_cmd, data->data, data->len);
+        clnt_cmd[data->len] = '\0';
+        gf_log (this->name, GF_LOG_DEBUG, "Got destination brick %s\n",
+                        clnt_cmd);
 
-        ret = dict_set_str (dict, CLIENT_CMD_CONNECT, dst_brick);
+        ret = dict_set_dynstr (dict, CLIENT_CMD_CONNECT, clnt_cmd);
         if (ret < 0) {
                 gf_log (this->name, GF_LOG_ERROR,
                         "Could not inititiate destination brick "
@@ -889,12 +971,21 @@ pump_initiate_sink_connect (call_frame_t *frame, xlator_t *this)
 		    PUMP_SINK_CHILD(this)->fops->setxattr,
 		    &loc,
 		    dict,
-		    0);
+		    0, NULL);
 
         ret = 0;
 
-        dict_unref (dict);
 out:
+        if (dict)
+                dict_unref (dict);
+
+        if (data)
+                data_unref (data);
+
+        if (ret && clnt_cmd)
+                GF_FREE (clnt_cmd);
+
+        loc_wipe (&loc);
         return ret;
 }
 
@@ -914,7 +1005,7 @@ pump_cmd_start_getxattr_cbk (call_frame_t *frame,
                              xlator_t *this,
                              int32_t op_ret,
                              int32_t op_errno,
-                             dict_t *dict)
+                             dict_t *dict, dict_t *xdata)
 {
         afr_local_t *local = NULL;
         char *path = NULL;
@@ -981,6 +1072,7 @@ 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;
@@ -1009,28 +1101,35 @@ pump_execute_status (call_frame_t *frame, xlator_t *this)
         }
 
         if (pump_priv->pump_finished) {
-        snprintf (dict_str, PATH_MAX + 256, "Number of files migrated = %"PRIu64"        Migration complete ",
-                  number_files);
+                snprintf (summary, PATH_MAX+256,
+                          "no_of_files=%"PRIu64, number_files);
         } else {
-        snprintf (dict_str, PATH_MAX + 256, "Number of files migrated = %"PRIu64"       Current file= %s ",
-                  number_files, filename);
+                snprintf (summary, PATH_MAX+256,
+                          "no_of_files=%"PRIu64":current_file=%s",
+                          number_files, filename);
         }
+        snprintf (dict_str, PATH_MAX+256, "status=%d:%s",
+                  (pump_priv->pump_finished)?1:0, summary);
 
         dict = dict_new ();
 
-        ret = dict_set_str (dict, PUMP_CMD_STATUS, dict_str);
+        ret = dict_set_dynstr (dict, RB_PUMP_CMD_STATUS, dict_str);
         if (ret < 0) {
                 gf_log (this->name, GF_LOG_DEBUG,
-                        "dict_set_str returned negative value");
+                        "dict_set_dynstr returned negative value");
+        } else {
+                dict_str = NULL;
         }
 
         op_ret = 0;
 
 out:
 
-        AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
+        AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, NULL);
+
+        if (dict)
+                dict_unref (dict);
 
-        dict_unref (dict);
         GF_FREE (dict_str);
 
         return 0;
@@ -1058,7 +1157,7 @@ pump_execute_start (call_frame_t *frame, xlator_t *this)
         afr_local_t   *local = NULL;
 
         int ret = 0;
-        loc_t loc;
+        loc_t loc = {0};
 
         priv = this->private;
         local = frame->local;
@@ -1073,14 +1172,14 @@ pump_execute_start (call_frame_t *frame, xlator_t *this)
 
         GF_ASSERT (priv->root_inode);
 
-        build_root_loc (priv->root_inode, &loc);
+        afr_build_root_loc (this, &loc);
 
 	STACK_WIND (frame,
 		    pump_cmd_start_getxattr_cbk,
 		    PUMP_SOURCE_CHILD(this),
 		    PUMP_SOURCE_CHILD(this)->fops->getxattr,
 		    &loc,
-		    PUMP_PATH);
+		    PUMP_PATH, NULL);
 
         ret = 0;
 
@@ -1090,15 +1189,69 @@ out:
                 pump_command_reply (frame, this);
         }
 
+        loc_wipe (&loc);
 	return 0;
 }
 
+static int
+pump_cleanup_helper (void *data) {
+        call_frame_t *frame = data;
+
+        pump_xattr_cleaner (frame, 0, frame->this, 0, 0, NULL);
+
+        return 0;
+}
+
+static int
+pump_cleanup_done (int ret, call_frame_t *sync_frame, void *data)
+{
+        STACK_DESTROY (sync_frame->root);
+
+        return 0;
+}
+
+int
+pump_execute_commit (call_frame_t *frame, xlator_t *this)
+{
+        afr_private_t  *priv       = NULL;
+        pump_private_t *pump_priv  = NULL;
+        afr_local_t    *local      = NULL;
+        call_frame_t   *sync_frame = NULL;
+        int             ret        = 0;
+
+        priv      = this->private;
+        pump_priv = priv->pump_private;
+        local     = frame->local;
+
+        local->op_ret = 0;
+        if (pump_priv->pump_finished) {
+                pump_change_state (this, PUMP_STATE_COMMIT);
+                sync_frame = create_frame (this, this->ctx->pool);
+                ret = synctask_new (pump_priv->env, pump_cleanup_helper,
+                                    pump_cleanup_done, sync_frame, frame);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_DEBUG, "Couldn't create "
+                                "synctask for cleaning up xattrs.");
+                }
+
+        } else {
+                gf_log (this->name, GF_LOG_ERROR, "Commit can't proceed. "
+                        "Migration in progress");
+                local->op_ret = -1;
+                local->op_errno = EINPROGRESS;
+                pump_command_reply (frame, this);
+        }
+
+        return 0;
+}
 int
 pump_execute_abort (call_frame_t *frame, xlator_t *this)
 {
-        afr_private_t  *priv      = NULL;
-        pump_private_t *pump_priv = NULL;
-        afr_local_t    *local     = NULL;
+        afr_private_t  *priv       = NULL;
+        pump_private_t *pump_priv  = NULL;
+        afr_local_t    *local      = NULL;
+        call_frame_t   *sync_frame = NULL;
+        int             ret        = 0;
 
         priv      = this->private;
         pump_priv = priv->pump_private;
@@ -1114,7 +1267,20 @@ pump_execute_abort (call_frame_t *frame, xlator_t *this)
         UNLOCK (&pump_priv->resume_path_lock);
 
         local->op_ret = 0;
-        pump_command_reply (frame, this);
+        if (pump_priv->pump_finished) {
+                sync_frame = create_frame (this, this->ctx->pool);
+                ret = synctask_new (pump_priv->env, pump_cleanup_helper,
+                                    pump_cleanup_done, sync_frame, frame);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_DEBUG, "Couldn't create "
+                                "synctask for cleaning up xattrs.");
+                }
+
+        } else {
+                pump_priv->cleaner = fop_setxattr_cbk_stub (frame,
+                                                            pump_xattr_cleaner,
+                                                            0, 0, NULL);
+        }
 
         return 0;
 }
@@ -1126,7 +1292,7 @@ pump_command_status (xlator_t *this, dict_t *dict)
         int dict_ret = -1;
         int ret = _gf_true;
 
-        dict_ret = dict_get_str (dict, PUMP_CMD_STATUS, &cmd);
+        dict_ret = dict_get_str (dict, RB_PUMP_CMD_STATUS, &cmd);
         if (dict_ret < 0) {
                 gf_log (this->name, GF_LOG_DEBUG,
                         "Not a pump status command");
@@ -1150,7 +1316,7 @@ pump_command_pause (xlator_t *this, dict_t *dict)
         int dict_ret = -1;
         int ret = _gf_true;
 
-        dict_ret = dict_get_str (dict, PUMP_CMD_PAUSE, &cmd);
+        dict_ret = dict_get_str (dict, RB_PUMP_CMD_PAUSE, &cmd);
         if (dict_ret < 0) {
                 gf_log (this->name, GF_LOG_DEBUG,
                         "Not a pump pause command");
@@ -1168,22 +1334,22 @@ out:
 }
 
 gf_boolean_t
-pump_command_abort (xlator_t *this, dict_t *dict)
+pump_command_commit (xlator_t *this, dict_t *dict)
 {
         char *cmd = NULL;
         int dict_ret = -1;
         int ret = _gf_true;
 
-        dict_ret = dict_get_str (dict, PUMP_CMD_ABORT, &cmd);
+        dict_ret = dict_get_str (dict, RB_PUMP_CMD_COMMIT, &cmd);
         if (dict_ret < 0) {
                 gf_log (this->name, GF_LOG_DEBUG,
-                        "Not a pump abort command");
+                        "Not a pump commit command");
                 ret = _gf_false;
                 goto out;
         }
 
         gf_log (this->name, GF_LOG_DEBUG,
-                "Hit a pump command - abort");
+                "Hit a pump command - commit");
         ret = _gf_true;
 
 out:
@@ -1192,22 +1358,22 @@ out:
 }
 
 gf_boolean_t
-pump_command_start (xlator_t *this, dict_t *dict)
+pump_command_abort (xlator_t *this, dict_t *dict)
 {
         char *cmd = NULL;
         int dict_ret = -1;
         int ret = _gf_true;
 
-        dict_ret = dict_get_str (dict, PUMP_CMD_START, &cmd);
+        dict_ret = dict_get_str (dict, RB_PUMP_CMD_ABORT, &cmd);
         if (dict_ret < 0) {
                 gf_log (this->name, GF_LOG_DEBUG,
-                        "Not a pump start command");
+                        "Not a pump abort command");
                 ret = _gf_false;
                 goto out;
         }
 
         gf_log (this->name, GF_LOG_DEBUG,
-                "Hit a pump command - start");
+                "Hit a pump command - abort");
         ret = _gf_true;
 
 out:
@@ -1215,134 +1381,47 @@ out:
 
 }
 
-struct _xattr_key {
-        char *key;
-        struct list_head list;
-};
-
-static void
-__gather_xattr_keys (dict_t *dict, char *key, data_t *value,
-                     void *data)
-{
-        struct list_head *  list  = data;
-        struct _xattr_key * xkey  = NULL;
-
-        if (!strncmp (key, AFR_XATTR_PREFIX,
-                      strlen (AFR_XATTR_PREFIX))) {
-
-                xkey = GF_CALLOC (1, sizeof (*xkey), gf_afr_mt_xattr_key);
-                if (!xkey)
-                        return;
-
-                xkey->key = key;
-                INIT_LIST_HEAD (&xkey->list);
-
-                list_add_tail (&xkey->list, list);
-        }
-}
-
-static void
-__filter_xattrs (dict_t *dict)
+gf_boolean_t
+pump_command_start (xlator_t *this, dict_t *dict)
 {
-        struct list_head keys;
-
-        struct _xattr_key *key;
-        struct _xattr_key *tmp;
-
-        INIT_LIST_HEAD (&keys);
-
-        dict_foreach (dict, __gather_xattr_keys,
-                      (void *) &keys);
-
-        list_for_each_entry_safe (key, tmp, &keys, list) {
-                dict_del (dict, key->key);
-
-                list_del_init (&key->list);
+        char *cmd = NULL;
+        int dict_ret = -1;
+        int ret = _gf_true;
 
-                GF_FREE (key);
+        dict_ret = dict_get_str (dict, RB_PUMP_CMD_START, &cmd);
+        if (dict_ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Not a pump start command");
+                ret = _gf_false;
+                goto out;
         }
-}
-
-int32_t
-pump_getxattr_cbk (call_frame_t *frame, void *cookie,
-		  xlator_t *this, int32_t op_ret, int32_t op_errno,
-		  dict_t *dict)
-{
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-	xlator_t **     children = NULL;
-
-	int unwind     = 1;
-	int last_tried = -1;
-	int this_try = -1;
-        int read_child = -1;
-
-	priv     = this->private;
-	children = priv->children;
 
-	local = frame->local;
-
-        read_child = (long) cookie;
-
-	if (op_ret == -1) {
-        retry:
-		last_tried = local->cont.getxattr.last_tried;
-
-		if (all_tried (last_tried, priv->child_count)) {
-			goto out;
-		}
-		this_try = ++local->cont.getxattr.last_tried;
-
-                if (this_try == read_child) {
-                        goto retry;
-                }
-
-		unwind = 0;
-		STACK_WIND_COOKIE (frame, pump_getxattr_cbk,
-				   (void *) (long) read_child,
-				   children[this_try],
-				   children[this_try]->fops->getxattr,
-				   &local->loc,
-				   local->cont.getxattr.name);
-	}
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Hit a pump command - start");
+        ret = _gf_true;
 
 out:
-	if (unwind) {
-                if (op_ret >= 0 && dict)
-                        __filter_xattrs (dict);
-
-		AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
-	}
+        return ret;
 
-	return 0;
 }
 
-int32_t
-pump_getxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, const char *name)
+int
+pump_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+	       const char *name, dict_t *xdata)
 {
-	afr_private_t *   priv       = NULL;
-	xlator_t **       children   = NULL;
-	int               call_child = 0;
-	afr_local_t     * local      = NULL;
-
-        int               read_child = -1;
-
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
-
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-
-	priv     = this->private;
-	VALIDATE_OR_GOTO (priv->children, out);
+	afr_private_t *priv = NULL;
+	int op_errno = 0;
+	int ret = 0;
 
-	children = priv->children;
+	priv = this->private;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-	frame->local = local;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame, default_getxattr_cbk,
+                            FIRST_CHILD (this),
+                            (FIRST_CHILD (this))->fops->getxattr,
+                            loc, name, xdata);
+                return 0;
+        }
 
         if (name) {
                 if (!strncmp (name, AFR_XATTR_PREFIX,
@@ -1352,188 +1431,21 @@ pump_getxattr (call_frame_t *frame, xlator_t *this,
                         goto out;
                 }
 
-                if (!strcmp (name, PUMP_CMD_STATUS)) {
+                if (!strcmp (name, RB_PUMP_CMD_STATUS)) {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "Hit pump command - status");
                         pump_execute_status (frame, this);
-                        op_ret = 0;
-                        goto out;
-                }
-        }
-
-        if (!priv->use_afr_in_pump) {
-                STACK_WIND (frame, default_getxattr_cbk,
-                            FIRST_CHILD (this),
-                            (FIRST_CHILD (this))->fops->getxattr,
-                            loc, name);
-                return 0;
-        }
-
-        read_child = afr_read_child (this, loc->inode);
-
-        if (read_child >= 0) {
-                call_child = read_child;
-
-                local->cont.getxattr.last_tried = -1;
-        } else {
-                call_child = afr_first_up_child (priv);
-
-                if (call_child == -1) {
-                        op_errno = ENOTCONN;
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "no child is up");
+                        ret = 0;
                         goto out;
                 }
-
-                local->cont.getxattr.last_tried = call_child;
         }
 
-	loc_copy (&local->loc, loc);
-	if (name)
-	  local->cont.getxattr.name       = gf_strdup (name);
+	afr_getxattr (frame, this, loc, name, xdata);
 
-	STACK_WIND_COOKIE (frame, pump_getxattr_cbk,
-			   (void *) (long) call_child,
-			   children[call_child], children[call_child]->fops->getxattr,
-			   loc, name);
-
-	op_ret = 0;
+	ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, NULL);
-	}
-	return 0;
-}
-
-static int
-afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	call_frame_t   *main_frame = NULL;
-
-	local = frame->local;
-	priv  = this->private;
-
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame)
-			main_frame = local->transaction.main_frame;
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
-
-	if (main_frame) {
-		AFR_STACK_UNWIND (setxattr, main_frame,
-                                  local->op_ret, local->op_errno)
-	}
-	return 0;
-}
-
-static int
-afr_setxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int32_t op_ret, int32_t op_errno)
-{
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int call_count  = -1;
-	int need_unwind = 0;
-
-	local = frame->local;
-	priv = this->private;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret != -1) {
-			if (local->success_count == 0) {
-				local->op_ret = op_ret;
-			}
-			local->success_count++;
-
-			if (local->success_count == priv->child_count) {
-				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_setxattr_wind (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-
-	int call_count = -1;
-	int i = 0;
-
-	local = frame->local;
-	priv = this->private;
-
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_setxattr_wind_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->setxattr,
-					   &local->loc,
-					   local->cont.setxattr.dict,
-					   local->cont.setxattr.flags);
-
-			if (!--call_count)
-				break;
-		}
-	}
-
-	return 0;
-}
-
-
-static int
-afr_setxattr_done (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t * local = frame->local;
-
-	local->transaction.unwind (frame, this);
-
-	AFR_STACK_DESTROY (frame);
-
-	return 0;
-}
-
-int32_t
-pump_setxattr_cbk (call_frame_t *frame,
-		      void *cookie,
-		      xlator_t *this,
-		      int32_t op_ret,
-		      int32_t op_errno)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno);
+	if (ret < 0)
+		AFR_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
 	return 0;
 }
 
@@ -1545,121 +1457,91 @@ pump_command_reply (call_frame_t *frame, xlator_t *this)
         local = frame->local;
 
         if (local->op_ret < 0)
-                gf_log (this->name, GF_LOG_NORMAL,
+                gf_log (this->name, GF_LOG_INFO,
                         "Command failed");
         else
-                gf_log (this->name, GF_LOG_NORMAL,
+                gf_log (this->name, GF_LOG_INFO,
                         "Command succeeded");
 
-        dict_unref (local->dict);
-
         AFR_STACK_UNWIND (setxattr,
                           frame,
                           local->op_ret,
-                          local->op_errno);
+                          local->op_errno, NULL);
 
         return 0;
 }
 
 int
-pump_parse_command (call_frame_t *frame, xlator_t *this,
-                    afr_local_t *local, dict_t *dict)
+pump_parse_command (call_frame_t *frame, xlator_t *this, dict_t *dict,
+		    int *op_errno_p)
 {
-
+	afr_local_t *local = NULL;
         int ret = -1;
+	int op_errno = 0;
 
         if (pump_command_start (this, dict)) {
-                frame->local = local;
+                local = AFR_FRAME_INIT (frame, op_errno);
+		if (!local)
+			goto out;
                 local->dict = dict_ref (dict);
                 ret = pump_execute_start (frame, this);
 
         } else if (pump_command_pause (this, dict)) {
-                frame->local = local;
+		local = AFR_FRAME_INIT (frame, op_errno);
+		if (!local)
+			goto out;
                 local->dict = dict_ref (dict);
                 ret = pump_execute_pause (frame, this);
 
         } else if (pump_command_abort (this, dict)) {
-                frame->local = local;
+		local = AFR_FRAME_INIT (frame, op_errno);
+		if (!local)
+			goto out;
                 local->dict = dict_ref (dict);
                 ret = pump_execute_abort (frame, this);
+
+        } else if (pump_command_commit (this, dict)) {
+		local = AFR_FRAME_INIT (frame, op_errno);
+		if (!local)
+			goto out;
+                local->dict = dict_ref (dict);
+                ret = pump_execute_commit (frame, this);
         }
+out:
+	if (op_errno_p)
+		*op_errno_p = op_errno;
         return ret;
 }
 
 int
-pump_setxattr (call_frame_t *frame, xlator_t *this,
-               loc_t *loc, dict_t *dict, int32_t flags)
+pump_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+	       int32_t flags, dict_t *xdata)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t   *transaction_frame = NULL;
-
+	afr_private_t *priv = NULL;
 	int ret = -1;
-
-	int op_ret   = -1;
 	int op_errno = 0;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.pump*", dict, op_errno, out);
 
 	priv = this->private;
-
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
-
-        ret = pump_parse_command (frame, this,
-                                  local, dict);
-        if (ret >= 0) {
-                op_ret = 0;
-                goto out;
-        }
-
         if (!priv->use_afr_in_pump) {
                 STACK_WIND (frame, default_setxattr_cbk,
                             FIRST_CHILD (this),
                             (FIRST_CHILD (this))->fops->setxattr,
-                            loc, dict, flags);
+                            loc, dict, flags, xdata);
                 return 0;
         }
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-	transaction_frame->local = local;
-
-	local->op_ret = -1;
-
-	local->cont.setxattr.dict  = dict_ref (dict);
-	local->cont.setxattr.flags = flags;
-
-	local->transaction.fop    = afr_setxattr_wind;
-	local->transaction.done   = afr_setxattr_done;
-	local->transaction.unwind = afr_setxattr_unwind;
-
-	loc_copy (&local->loc, loc);
-
-	local->transaction.main_frame = frame;
-	local->transaction.start   = LLONG_MAX - 1;
-	local->transaction.len     = 0;
+        ret = pump_parse_command (frame, this, dict, &op_errno);
+        if (ret >= 0)
+                goto out;
 
-	afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+	afr_setxattr (frame, this, loc, dict, flags, xdata);
 
-	op_ret = 0;
+	ret = 0;
 out:
-	if (op_ret == -1) {
-		if (transaction_frame)
-			AFR_STACK_DESTROY (transaction_frame);
-		AFR_STACK_UNWIND (setxattr, frame, op_ret, op_errno);
+	if (ret < 0) {
+		AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
 	}
 
 	return 0;
@@ -1693,7 +1575,7 @@ static int32_t
 pump_truncate (call_frame_t *frame,
                xlator_t *this,
                loc_t *loc,
-               off_t offset)
+               off_t offset, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1703,11 +1585,11 @@ pump_truncate (call_frame_t *frame,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->truncate,
                             loc,
-                            offset);
+                            offset, xdata);
                 return 0;
         }
 
-        afr_truncate (frame, this, loc, offset);
+        afr_truncate (frame, this, loc, offset, xdata);
         return 0;
 }
 
@@ -1716,7 +1598,7 @@ static int32_t
 pump_ftruncate (call_frame_t *frame,
                 xlator_t *this,
                 fd_t *fd,
-                off_t offset)
+                off_t offset, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1726,11 +1608,11 @@ pump_ftruncate (call_frame_t *frame,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->ftruncate,
                             fd,
-                            offset);
+                            offset, xdata);
                 return 0;
         }
 
-        afr_ftruncate (frame, this, fd, offset);
+        afr_ftruncate (frame, this, fd, offset, xdata);
         return 0;
 }
 
@@ -1739,7 +1621,7 @@ pump_ftruncate (call_frame_t *frame,
 
 int
 pump_mknod (call_frame_t *frame, xlator_t *this,
-            loc_t *loc, mode_t mode, dev_t rdev, dict_t *parms)
+            loc_t *loc, mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1747,10 +1629,10 @@ pump_mknod (call_frame_t *frame, xlator_t *this,
                 STACK_WIND (frame, default_mknod_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->mknod,
-                            loc, mode, rdev, parms);
+                            loc, mode, rdev, umask, xdata);
                 return 0;
         }
-        afr_mknod (frame, this, loc, mode, rdev, parms);
+        afr_mknod (frame, this, loc, mode, rdev, umask, xdata);
         return 0;
 
 }
@@ -1759,7 +1641,7 @@ pump_mknod (call_frame_t *frame, xlator_t *this,
 
 int
 pump_mkdir (call_frame_t *frame, xlator_t *this,
-            loc_t *loc, mode_t mode, dict_t *params)
+            loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1767,10 +1649,10 @@ pump_mkdir (call_frame_t *frame, xlator_t *this,
                 STACK_WIND (frame, default_mkdir_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->mkdir,
-                            loc, mode, params);
+                            loc, mode, umask, xdata);
                 return 0;
         }
-        afr_mkdir (frame, this, loc, mode, params);
+        afr_mkdir (frame, this, loc, mode, umask, xdata);
         return 0;
 
 }
@@ -1779,7 +1661,7 @@ pump_mkdir (call_frame_t *frame, xlator_t *this,
 static int32_t
 pump_unlink (call_frame_t *frame,
              xlator_t *this,
-             loc_t *loc)
+             loc_t *loc, int xflag, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1788,10 +1670,10 @@ pump_unlink (call_frame_t *frame,
                             default_unlink_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->unlink,
-                            loc);
+                            loc, xflag, xdata);
                 return 0;
         }
-        afr_unlink (frame, this, loc);
+        afr_unlink (frame, this, loc, xflag, xdata);
         return 0;
 
 }
@@ -1799,7 +1681,7 @@ pump_unlink (call_frame_t *frame,
 
 static int
 pump_rmdir (call_frame_t *frame, xlator_t *this,
-            loc_t *loc, int flags)
+            loc_t *loc, int flags, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 
@@ -1809,11 +1691,11 @@ pump_rmdir (call_frame_t *frame, xlator_t *this,
                 STACK_WIND (frame, default_rmdir_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->rmdir,
-                            loc, flags);
+                            loc, flags, xdata);
                 return 0;
         }
 
-        afr_rmdir (frame, this, loc, flags);
+        afr_rmdir (frame, this, loc, flags, xdata);
         return 0;
 
 }
@@ -1822,7 +1704,7 @@ pump_rmdir (call_frame_t *frame, xlator_t *this,
 
 int
 pump_symlink (call_frame_t *frame, xlator_t *this,
-              const char *linkpath, loc_t *loc, dict_t *params)
+              const char *linkpath, loc_t *loc, mode_t umask, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1830,10 +1712,10 @@ pump_symlink (call_frame_t *frame, xlator_t *this,
                 STACK_WIND (frame, default_symlink_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->symlink,
-                            linkpath, loc, params);
+                            linkpath, loc, umask, xdata);
                 return 0;
         }
-        afr_symlink (frame, this, linkpath, loc, params);
+        afr_symlink (frame, this, linkpath, loc, umask, xdata);
         return 0;
 
 }
@@ -1843,7 +1725,7 @@ static int32_t
 pump_rename (call_frame_t *frame,
              xlator_t *this,
              loc_t *oldloc,
-             loc_t *newloc)
+             loc_t *newloc, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1852,10 +1734,10 @@ pump_rename (call_frame_t *frame,
                             default_rename_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->rename,
-                            oldloc, newloc);
+                            oldloc, newloc, xdata);
                 return 0;
         }
-        afr_rename (frame, this, oldloc, newloc);
+        afr_rename (frame, this, oldloc, newloc, xdata);
         return 0;
 
 }
@@ -1865,7 +1747,7 @@ static int32_t
 pump_link (call_frame_t *frame,
            xlator_t *this,
            loc_t *oldloc,
-           loc_t *newloc)
+           loc_t *newloc, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1874,10 +1756,10 @@ pump_link (call_frame_t *frame,
                             default_link_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->link,
-                            oldloc, newloc);
+                            oldloc, newloc, xdata);
                 return 0;
         }
-        afr_link (frame, this, oldloc, newloc);
+        afr_link (frame, this, oldloc, newloc, xdata);
         return 0;
 
 }
@@ -1886,7 +1768,7 @@ pump_link (call_frame_t *frame,
 static int32_t
 pump_create (call_frame_t *frame, xlator_t *this,
              loc_t *loc, int32_t flags, mode_t mode,
-             fd_t *fd, dict_t *params)
+             mode_t umask, fd_t *fd, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1894,10 +1776,10 @@ pump_create (call_frame_t *frame, xlator_t *this,
                 STACK_WIND (frame, default_create_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->create,
-                            loc, flags, mode, fd, params);
+                            loc, flags, mode, umask, fd, xdata);
                 return 0;
         }
-        afr_create (frame, this, loc, flags, mode, fd, params);
+        afr_create (frame, this, loc, flags, mode, umask, fd, xdata);
         return 0;
 
 }
@@ -1907,8 +1789,7 @@ static int32_t
 pump_open (call_frame_t *frame,
            xlator_t *this,
            loc_t *loc,
-           int32_t flags, fd_t *fd,
-           int32_t wbflags)
+           int32_t flags, fd_t *fd, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1917,10 +1798,10 @@ pump_open (call_frame_t *frame,
                             default_open_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->open,
-                            loc, flags, fd, wbflags);
+                            loc, flags, fd, xdata);
                 return 0;
         }
-        afr_open (frame, this, loc, flags, fd, wbflags);
+        afr_open (frame, this, loc, flags, fd, xdata);
         return 0;
 
 }
@@ -1932,8 +1813,8 @@ pump_writev (call_frame_t *frame,
              fd_t *fd,
              struct iovec *vector,
              int32_t count,
-             off_t off,
-             struct iobref *iobref)
+             off_t off, uint32_t flags,
+             struct iobref *iobref, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1945,20 +1826,20 @@ pump_writev (call_frame_t *frame,
                             fd,
                             vector,
                             count,
-                            off,
-                            iobref);
+                            off, flags,
+                            iobref, xdata);
                 return 0;
         }
-        afr_writev (frame, this, fd, vector, count, off, iobref);
-        return 0;
 
+        afr_writev (frame, this, fd, vector, count, off, flags, iobref, xdata);
+        return 0;
 }
 
 
 static int32_t
 pump_flush (call_frame_t *frame,
             xlator_t *this,
-            fd_t *fd)
+            fd_t *fd, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1967,10 +1848,10 @@ pump_flush (call_frame_t *frame,
                             default_flush_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->flush,
-                            fd);
+                            fd, xdata);
                 return 0;
         }
-        afr_flush (frame, this, fd);
+        afr_flush (frame, this, fd, xdata);
         return 0;
 
 }
@@ -1980,7 +1861,7 @@ static int32_t
 pump_fsync (call_frame_t *frame,
             xlator_t *this,
             fd_t *fd,
-            int32_t flags)
+            int32_t flags, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -1990,10 +1871,10 @@ pump_fsync (call_frame_t *frame,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->fsync,
                             fd,
-                            flags);
+                            flags, xdata);
                 return 0;
         }
-        afr_fsync (frame, this, fd, flags);
+        afr_fsync (frame, this, fd, flags, xdata);
         return 0;
 
 }
@@ -2002,7 +1883,7 @@ pump_fsync (call_frame_t *frame,
 static int32_t
 pump_opendir (call_frame_t *frame,
               xlator_t *this,
-              loc_t *loc, fd_t *fd)
+              loc_t *loc, fd_t *fd, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -2011,10 +1892,10 @@ pump_opendir (call_frame_t *frame,
                             default_opendir_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->opendir,
-                            loc, fd);
+                            loc, fd, xdata);
                 return 0;
         }
-        afr_opendir (frame, this, loc, fd);
+        afr_opendir (frame, this, loc, fd, xdata);
         return 0;
 
 }
@@ -2024,7 +1905,7 @@ static int32_t
 pump_fsyncdir (call_frame_t *frame,
                xlator_t *this,
                fd_t *fd,
-               int32_t flags)
+               int32_t flags, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -2034,10 +1915,10 @@ pump_fsyncdir (call_frame_t *frame,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->fsyncdir,
                             fd,
-                            flags);
+                            flags, xdata);
                 return 0;
         }
-        afr_fsyncdir (frame, this, fd, flags);
+        afr_fsyncdir (frame, this, fd, flags, xdata);
         return 0;
 
 }
@@ -2048,7 +1929,7 @@ pump_xattrop (call_frame_t *frame,
               xlator_t *this,
               loc_t *loc,
               gf_xattrop_flags_t flags,
-              dict_t *dict)
+              dict_t *dict, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -2059,10 +1940,10 @@ pump_xattrop (call_frame_t *frame,
                             FIRST_CHILD(this)->fops->xattrop,
                             loc,
                             flags,
-                            dict);
+                            dict, xdata);
                 return 0;
         }
-        afr_xattrop (frame, this, loc, flags, dict);
+        afr_xattrop (frame, this, loc, flags, dict, xdata);
         return 0;
 
 }
@@ -2072,7 +1953,7 @@ pump_fxattrop (call_frame_t *frame,
                xlator_t *this,
                fd_t *fd,
                gf_xattrop_flags_t flags,
-               dict_t *dict)
+               dict_t *dict, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -2083,10 +1964,10 @@ pump_fxattrop (call_frame_t *frame,
                             FIRST_CHILD(this)->fops->fxattrop,
                             fd,
                             flags,
-                            dict);
+                            dict, xdata);
                 return 0;
         }
-        afr_fxattrop (frame, this, fd, flags, dict);
+        afr_fxattrop (frame, this, fd, flags, dict, xdata);
         return 0;
 
 }
@@ -2096,9 +1977,17 @@ static int32_t
 pump_removexattr (call_frame_t *frame,
                   xlator_t *this,
                   loc_t *loc,
-                  const char *name)
+                  const char *name, dict_t *xdata)
 {
-        afr_private_t *priv  = NULL;
+        afr_private_t *priv     = NULL;
+        int            op_errno = -1;
+
+        VALIDATE_OR_GOTO (this, out);
+
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.pump*",
+                                 name, op_errno, out);
+
+        op_errno = 0;
 	priv = this->private;
         if (!priv->use_afr_in_pump) {
                 STACK_WIND (frame,
@@ -2106,10 +1995,14 @@ pump_removexattr (call_frame_t *frame,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->removexattr,
                             loc,
-                            name);
+                            name, xdata);
                 return 0;
         }
-        afr_removexattr (frame, this, loc, name);
+        afr_removexattr (frame, this, loc, name, xdata);
+
+ out:
+        if (op_errno)
+                AFR_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
         return 0;
 
 }
@@ -2121,7 +2014,7 @@ pump_readdir (call_frame_t *frame,
               xlator_t *this,
               fd_t *fd,
               size_t size,
-              off_t off)
+              off_t off, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -2130,21 +2023,18 @@ pump_readdir (call_frame_t *frame,
                             default_readdir_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->readdir,
-                            fd, size, off);
+                            fd, size, off, xdata);
                 return 0;
         }
-        afr_readdir (frame, this, fd, size, off);
+        afr_readdir (frame, this, fd, size, off, xdata);
         return 0;
 
 }
 
 
 static int32_t
-pump_readdirp (call_frame_t *frame,
-               xlator_t *this,
-               fd_t *fd,
-               size_t size,
-               off_t off)
+pump_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd,
+               size_t size, off_t off, dict_t *dict)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -2153,10 +2043,10 @@ pump_readdirp (call_frame_t *frame,
                             default_readdirp_cbk,
                             FIRST_CHILD(this),
                             FIRST_CHILD(this)->fops->readdirp,
-                            fd, size, off);
+                            fd, size, off, dict);
                 return 0;
         }
-        afr_readdirp (frame, this, fd, size, off);
+        afr_readdirp (frame, this, fd, size, off, dict);
         return 0;
 
 }
@@ -2187,13 +2077,24 @@ pump_release (xlator_t *this,
 
 }
 
+static int32_t
+pump_forget (xlator_t *this, inode_t *inode)
+{
+        afr_private_t  *priv  = NULL;
+
+        priv = this->private;
+        if (priv->use_afr_in_pump)
+                afr_forget (this, inode);
+
+        return 0;
+}
 
 static int32_t
 pump_setattr (call_frame_t *frame,
               xlator_t *this,
               loc_t *loc,
               struct iatt *stbuf,
-              int32_t valid)
+              int32_t valid, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -2202,10 +2103,10 @@ pump_setattr (call_frame_t *frame,
                             default_setattr_cbk,
                             FIRST_CHILD (this),
                             FIRST_CHILD (this)->fops->setattr,
-                            loc, stbuf, valid);
+                            loc, stbuf, valid, xdata);
                 return 0;
         }
-        afr_setattr (frame, this, loc, stbuf, valid);
+        afr_setattr (frame, this, loc, stbuf, valid, xdata);
         return 0;
 
 }
@@ -2216,7 +2117,7 @@ pump_fsetattr (call_frame_t *frame,
                xlator_t *this,
                fd_t *fd,
                struct iatt *stbuf,
-               int32_t valid)
+               int32_t valid, dict_t *xdata)
 {
         afr_private_t *priv  = NULL;
 	priv = this->private;
@@ -2225,10 +2126,10 @@ pump_fsetattr (call_frame_t *frame,
                             default_fsetattr_cbk,
                             FIRST_CHILD (this),
                             FIRST_CHILD (this)->fops->fsetattr,
-                            fd, stbuf, valid);
+                            fd, stbuf, valid, xdata);
                 return 0;
         }
-        afr_fsetattr (frame, this, fd, stbuf, valid);
+        afr_fsetattr (frame, this, fd, stbuf, valid, xdata);
         return 0;
 
 }
@@ -2277,7 +2178,7 @@ notify (xlator_t *this, int32_t event,
 
         child_xl = (xlator_t *) data;
 
-        ret = afr_notify (this, event, data);
+        ret = afr_notify (this, event, data, NULL);
 
 	switch (event) {
 	case GF_EVENT_CHILD_DOWN:
@@ -2307,12 +2208,12 @@ int32_t
 init (xlator_t *this)
 {
 	afr_private_t * priv        = NULL;
-        pump_private_t *pump_priv = NULL;
+        pump_private_t *pump_priv   = NULL;
 	int             child_count = 0;
 	xlator_list_t * trav        = NULL;
 	int             i           = 0;
 	int             ret         = -1;
-	int             op_errno    = 0;
+	GF_UNUSED int   op_errno    = 0;
 
         int source_child = 0;
 
@@ -2328,26 +2229,36 @@ init (xlator_t *this)
 			"Volume is dangling.");
 	}
 
-	ALLOC_OR_GOTO (this->private, afr_private_t, out);
+	priv = GF_CALLOC (1, sizeof (afr_private_t), gf_afr_mt_afr_private_t);
+        if (!priv)
+                goto out;
 
-	priv = this->private;
+        LOCK_INIT (&priv->lock);
+
+        child_count = xlator_subvolume_count (this);
+        if (child_count != 2) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "There should be exactly 2 children - one source "
+                        "and one sink");
+                return -1;
+        }
+	priv->child_count = child_count;
 
         priv->read_child = source_child;
         priv->favorite_child = source_child;
         priv->background_self_heal_count = 0;
 
-	priv->data_self_heal     = 1;
+	priv->data_self_heal     = "on";
 	priv->metadata_self_heal = 1;
 	priv->entry_self_heal    = 1;
 
-        priv->data_self_heal_algorithm = "";
-
         priv->data_self_heal_window_size = 16;
 
 	priv->data_change_log     = 1;
 	priv->metadata_change_log = 1;
 	priv->entry_change_log    = 1;
         priv->use_afr_in_pump = 1;
+        priv->sh_readdir_size = 65536;
 
 	/* Locking options */
 
@@ -2356,31 +2267,6 @@ init (xlator_t *this)
            and the sink.
         */
 
-	priv->data_lock_server_count = 2;
-	priv->metadata_lock_server_count = 2;
-	priv->entry_lock_server_count = 2;
-
-	priv->strict_readdir = _gf_false;
-
-        trav = this->children;
-        while (trav) {
-                child_count++;
-                trav = trav->next;
-        }
-
-	priv->wait_count = 1;
-
-        if (child_count != 2) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "There should be exactly 2 children - one source "
-                        "and one sink");
-                return -1;
-        }
-	priv->child_count = child_count;
-
-	LOCK_INIT (&priv->lock);
-        LOCK_INIT (&priv->read_child_lock);
-
 	priv->child_up = GF_CALLOC (sizeof (unsigned char), child_count,
                                  gf_afr_mt_char);
 	if (!priv->child_up) {
@@ -2414,8 +2300,9 @@ init (xlator_t *this)
 	while (i < child_count) {
 		priv->children[i] = trav->xlator;
 
-                ret = asprintf (&priv->pending_key[i], "%s.%s", AFR_XATTR_PREFIX,
-                                trav->xlator->name);
+                ret = gf_asprintf (&priv->pending_key[i], "%s.%s",
+                                   AFR_XATTR_PREFIX,
+                                   trav->xlator->name);
                 if (-1 == ret) {
                         gf_log (this->name, GF_LOG_ERROR,
                                 "asprintf failed to set pending key");
@@ -2427,9 +2314,21 @@ init (xlator_t *this)
 		i++;
 	}
 
-        priv->first_lookup = 1;
+        ret = gf_asprintf (&priv->sh_domain, "%s-self-heal", this->name);
+        if (-1 == ret) {
+                op_errno = ENOMEM;
+                goto out;
+        }
+
         priv->root_inode = NULL;
 
+        priv->last_event = GF_CALLOC (child_count, sizeof (*priv->last_event),
+                                      gf_afr_mt_int32_t);
+        if (!priv->last_event) {
+                ret = -ENOMEM;
+                goto out;
+        }
+
 	pump_priv = GF_CALLOC (1, sizeof (*pump_priv),
                             gf_afr_mt_pump_priv);
 	if (!pump_priv) {
@@ -2445,13 +2344,12 @@ init (xlator_t *this)
         pump_priv->resume_path = GF_CALLOC (1, PATH_MAX,
                                             gf_afr_mt_char);
         if (!pump_priv->resume_path) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
+                gf_log (this->name, GF_LOG_ERROR, "Out of memory");
                 ret = -1;
                 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");
@@ -2459,21 +2357,67 @@ init (xlator_t *this)
                 goto out;
         }
 
+        /* keep more local here as we may need them for self-heal etc */
+        this->local_pool = mem_pool_new (afr_local_t, 128);
+        if (!this->local_pool) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create local_t's memory pool");
+                goto out;
+        }
+
 	priv->pump_private = pump_priv;
+        pump_priv = NULL;
 
-        pthread_mutex_init (&priv->mutex, NULL);
-        INIT_LIST_HEAD (&priv->saved_fds);
+        this->private = priv;
+        priv = NULL;
 
         pump_change_state (this, PUMP_STATE_ABORT);
 
 	ret = 0;
 out:
+
+        if (pump_priv) {
+                GF_FREE (pump_priv->resume_path);
+                LOCK_DESTROY (&pump_priv->resume_path_lock);
+                LOCK_DESTROY (&pump_priv->pump_state_lock);
+                GF_FREE (pump_priv);
+        }
+
+        if (priv) {
+                GF_FREE (priv->child_up);
+                GF_FREE (priv->children);
+                GF_FREE (priv->pending_key);
+                GF_FREE (priv->last_event);
+                LOCK_DESTROY (&priv->lock);
+                GF_FREE (priv);
+        }
+
 	return ret;
 }
 
 int
 fini (xlator_t *this)
 {
+        afr_private_t * priv        = NULL;
+        pump_private_t *pump_priv = NULL;
+
+        priv      = this->private;
+        this->private = NULL;
+        if (!priv)
+                goto out;
+
+        pump_priv = priv->pump_private;
+        if (!pump_priv)
+                goto afr_priv;
+
+        GF_FREE (pump_priv->resume_path);
+        LOCK_DESTROY (&pump_priv->resume_path_lock);
+        LOCK_DESTROY (&pump_priv->pump_state_lock);
+        GF_FREE (pump_priv);
+afr_priv:
+        afr_priv_destroy (priv);
+out:
 	return 0;
 }
 
@@ -2521,6 +2465,7 @@ struct xlator_dumpops dumpops = {
 struct xlator_cbks cbks = {
 	.release     = pump_release,
 	.releasedir  = pump_releasedir,
+        .forget      = pump_forget,
 };
 
 struct volume_options options[] = {
diff --git a/xlators/cluster/afr/src/pump.h b/xlators/cluster/afr/src/pump.h
index a46f9d7a5..9d0b6db6a 100644
--- a/xlators/cluster/afr/src/pump.h
+++ b/xlators/cluster/afr/src/pump.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __PUMP_H__
@@ -26,18 +17,6 @@
 #define CLIENT_CMD_CONNECT "trusted.glusterfs.client-connect"
 #define CLIENT_CMD_DISCONNECT "trusted.glusterfs.client-disconnect"
 
-#define PUMP_PID 696969
-#define PUMP_LK_OWNER 696969
-
-#define IS_ROOT_PATH(path) (!strcmp (path, "/"))
-#define IS_ENTRY_CWD(entry) (!strcmp (entry, "."))
-#define IS_ENTRY_PARENT(entry) (!strcmp (entry, ".."))
-
-#define PUMP_CMD_START  "trusted.glusterfs.pump.start"
-#define PUMP_CMD_ABORT  "trusted.glusterfs.pump.abort"
-#define PUMP_CMD_PAUSE  "trusted.glusterfs.pump.pause"
-#define PUMP_CMD_STATUS "trusted.glusterfs.pump.status"
-
 #define PUMP_SOURCE_COMPLETE "trusted.glusterfs.pump-source-complete"
 #define PUMP_SINK_COMPLETE "trusted.glusterfs.pump-sink-complete"
 
@@ -51,11 +30,12 @@ typedef enum {
         PUMP_STATE_RESUME,              /* Pump is resuming from a previous pause */
         PUMP_STATE_PAUSE,               /* Pump is paused */
         PUMP_STATE_ABORT,               /* Pump is aborted */
+        PUMP_STATE_COMMIT,              /* Pump is commited */
 } pump_state_t;
 
 typedef struct _pump_private {
 	struct syncenv *env;            /* The env pointer to the pump synctask */
-        const char *resume_path;        /* path to resume from the last pause */
+        char *resume_path;              /* path to resume from the last pause */
         gf_lock_t resume_path_lock;     /* Synchronize resume_path changes */
         gf_lock_t pump_state_lock;      /* Synchronize pump_state changes */
         pump_state_t pump_state;        /* State of pump */
@@ -64,6 +44,7 @@ typedef struct _pump_private {
         gf_boolean_t pump_finished;     /* Boolean to indicate pump termination */
         char pump_start_pending;        /* Boolean to mark start pending until
                                            CHILD_UP */
+        call_stub_t *cleaner;
 } pump_private_t;
 
 void
@@ -94,4 +75,7 @@ pump_command_status (xlator_t *this, dict_t *dict);
 int
 pump_execute_status (call_frame_t *frame, xlator_t *this);
 
+int
+pump_command_reply (call_frame_t *frame, xlator_t *this);
+
 #endif /* __PUMP_H__ */
diff --git a/xlators/cluster/dht/src/Makefile.am b/xlators/cluster/dht/src/Makefile.am
index 4b69aa071..174bea841 100644
--- a/xlators/cluster/dht/src/Makefile.am
+++ b/xlators/cluster/dht/src/Makefile.am
@@ -2,33 +2,37 @@
 xlator_LTLIBRARIES = dht.la nufa.la switch.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/cluster
 
+dht_common_source = dht-layout.c dht-helper.c dht-linkfile.c dht-rebalance.c \
+	dht-selfheal.c dht-rename.c dht-hashfn.c dht-diskusage.c \
+	dht-common.c dht-inode-write.c dht-inode-read.c dht-shared.c \
+	$(top_builddir)/xlators/lib/src/libxlator.c
 
-dht_common_source = dht-layout.c dht-helper.c dht-linkfile.c \
-		dht-selfheal.c dht-rename.c dht-hashfn.c dht-diskusage.c
-
-dht_la_SOURCES = $(dht_common_source) dht.c 
+dht_la_SOURCES = $(dht_common_source) dht.c
 
 nufa_la_SOURCES = $(dht_common_source) nufa.c
 switch_la_SOURCES = $(dht_common_source) switch.c
 
-dht_la_LDFLAGS = -module -avoidversion
+dht_la_LDFLAGS = -module -avoid-version
 dht_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-nufa_la_LDFLAGS = -module -avoidversion
+nufa_la_LDFLAGS = -module -avoid-version
 nufa_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-switch_la_LDFLAGS = -module -avoidversion
+switch_la_LDFLAGS = -module -avoid-version
 switch_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-noinst_HEADERS = dht-common.h dht-common.c dht-mem-types.h
+noinst_HEADERS = dht-common.h dht-mem-types.h \
+	$(top_builddir)/xlators/lib/src/libxlator.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
+	-I$(top_srcdir)/xlators/lib/src
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
-CLEANFILES = 
+CLEANFILES =
 
 uninstall-local:
 	rm -f $(DESTDIR)$(xlatordir)/distribute.so
 
 install-data-hook:
-	ln -sf dht.so $(DESTDIR)$(xlatordir)/distribute.so
-\ No newline at end of file
+	ln -sf dht.so $(DESTDIR)$(xlatordir)/distribute.so
diff --git a/xlators/cluster/dht/src/dht-common.c b/xlators/cluster/dht/src/dht-common.c
index 1ab55ccdc..3868fc38f 100644
--- a/xlators/cluster/dht/src/dht-common.c
+++ b/xlators/cluster/dht/src/dht-common.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2009-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -27,12 +18,88 @@
 
 #include "glusterfs.h"
 #include "xlator.h"
+#include "libxlator.h"
 #include "dht-common.h"
 #include "defaults.h"
+#include "byte-order.h"
+#include "glusterfs-acl.h"
 
 #include <sys/time.h>
 #include <libgen.h>
 
+int
+dht_aggregate (dict_t *this, char *key, data_t *value, void *data)
+{
+        dict_t  *dst  = NULL;
+        int64_t *ptr  = 0, *size = NULL;
+        int32_t  ret  = -1;
+        data_t  *dict_data = NULL;
+
+        dst = data;
+
+        if (strcmp (key, GF_XATTR_QUOTA_SIZE_KEY) == 0) {
+                ret = dict_get_bin (dst, key, (void **)&size);
+                if (ret < 0) {
+                        size = GF_CALLOC (1, sizeof (int64_t),
+                                          gf_common_mt_char);
+                        if (size == NULL) {
+                                gf_log ("dht", GF_LOG_WARNING,
+                                        "memory allocation failed");
+                                return -1;
+                        }
+                        ret = dict_set_bin (dst, key, size, sizeof (int64_t));
+                        if (ret < 0) {
+                                gf_log ("dht", GF_LOG_WARNING,
+                                        "dht aggregate dict set failed");
+                                GF_FREE (size);
+                                return -1;
+                        }
+                }
+
+                ptr = data_to_bin (value);
+                if (ptr == NULL) {
+                        gf_log ("dht", GF_LOG_WARNING, "data to bin failed");
+                        return -1;
+                }
+
+                *size = hton64 (ntoh64 (*size) + ntoh64 (*ptr));
+
+        } else if (fnmatch (GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0) {
+                ret = gf_get_min_stime (THIS, dst, key, value);
+                if (ret < 0)
+                        return ret;
+        } else {
+                /* compare user xattrs only */
+                if (!strncmp (key, "user.", strlen ("user."))) {
+                        ret = dict_lookup (dst, key, &dict_data);
+                        if (!ret && dict_data && value) {
+                                ret = is_data_equal (dict_data, value);
+                                if (!ret)
+                                        gf_log ("dht", GF_LOG_DEBUG,
+                                                "xattr mismatch for %s", key);
+                        }
+                }
+                ret = dict_set (dst, key, value);
+                if (ret)
+                        gf_log ("dht", GF_LOG_WARNING, "xattr dict set failed");
+        }
+
+        return 0;
+}
+
+
+void
+dht_aggregate_xattr (dict_t *dst, dict_t *src)
+{
+        if ((dst == NULL) || (src == NULL)) {
+                goto out;
+        }
+
+        dict_foreach (src, dht_aggregate, dst);
+out:
+        return;
+}
+
 /* TODO:
    - use volumename in xattr instead of "dht"
    - use NS locks
@@ -43,219 +110,369 @@
 
 int
 dht_lookup_selfheal_cbk (call_frame_t *frame, void *cookie,
-			 xlator_t *this,
-			 int op_ret, int op_errno)
+                         xlator_t *this,
+                         int op_ret, int op_errno, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	dht_layout_t *layout = NULL;
-	int           ret = 0;
+        dht_local_t  *local = NULL;
+        dht_layout_t *layout = NULL;
+	int           ret = -1;
 
-	local = frame->local;
-	ret = op_ret;
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
 
-        dht_frame_su_undo (frame);
+        local = frame->local;
+        ret = op_ret;
 
-	if (ret == 0) {
-		layout = local->selfheal.layout;
-		ret = dht_layout_set (this, local->inode, layout);
+        FRAME_SU_UNDO (frame, dht_local_t);
 
-		if (local->ia_ino) {
-			local->stbuf.ia_ino = local->ia_ino;
-		} else {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"could not find hashed subvolume for %s",
-				local->loc.path);
-		}
+        if (ret == 0) {
+                layout = local->selfheal.layout;
+                ret = dht_layout_set (this, local->inode, layout);
+        }
 
-                if (local->loc.parent)
-                        local->postparent.ia_ino = local->loc.parent->ino;
-	}
+        if (local->loc.parent) {
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           &local->postparent, 1);
+        }
 
-        WIPE (&local->postparent);
+        DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
 
-	DHT_STACK_UNWIND (lookup, frame, ret, local->op_errno, local->inode,
-			  &local->stbuf, local->xattr, &local->postparent);
+        DHT_STACK_UNWIND (lookup, frame, ret, local->op_errno, local->inode,
+                          &local->stbuf, local->xattr, &local->postparent);
 
-	return 0;
+out:
+        return ret;
 }
 
 
 int
-dht_lookup_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int op_ret, int op_errno,
-                    inode_t *inode, struct iatt *stbuf, dict_t *xattr,
-                    struct iatt *postparent)
+dht_discover_complete (xlator_t *this, call_frame_t *discover_frame)
+{
+        dht_local_t     *local = NULL;
+        call_frame_t    *main_frame = NULL;
+        int              op_errno = 0;
+        int              ret = -1;
+        dht_layout_t    *layout = NULL;
+        dht_conf_t      *conf = NULL;
+
+        local = discover_frame->local;
+        layout = local->layout;
+        conf = this->private;
+
+        LOCK(&discover_frame->lock);
+        {
+                main_frame = local->main_frame;
+                local->main_frame = NULL;
+        }
+        UNLOCK(&discover_frame->lock);
+
+        if (!main_frame)
+                return 0;
+
+        if (local->file_count && local->dir_count) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "path %s exists as a file on one subvolume "
+                        "and directory on another. "
+                        "Please fix it manually",
+                        local->loc.path);
+                op_errno = EIO;
+                goto out;
+        }
+
+        if (local->cached_subvol) {
+                ret = dht_layout_preset (this, local->cached_subvol,
+                                         local->inode);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "failed to set layout for subvolume %s",
+                                local->cached_subvol ? local->cached_subvol->name : "<nil>");
+                        op_errno = EINVAL;
+                        goto out;
+                }
+        } else {
+                ret = dht_layout_normalize (this, &local->loc, layout);
+                if ((ret < 0) || ((ret > 0) && (local->op_ret != 0))) {
+                        /* either the layout is incorrect or the directory is
+                         * not found even in one subvolume.
+                         */
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "normalizing failed on %s "
+                                "(overlaps/holes present: %s, "
+                                "ENOENT errors: %d)", local->loc.path,
+                                (ret < 0) ? "yes" : "no", (ret > 0) ? ret : 0);
+                        if ((ret > 0) && (ret == conf->subvolume_cnt)) {
+                                op_errno = ESTALE;
+                                goto out;
+                        }
+                }
+
+                if (local->inode)
+                        dht_layout_set (this, local->inode, layout);
+        }
+
+        DHT_STACK_UNWIND (lookup, main_frame, local->op_ret, local->op_errno,
+                          local->inode, &local->stbuf, local->xattr,
+                          &local->postparent);
+        return 0;
+out:
+        DHT_STACK_UNWIND (lookup, main_frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL);
+
+        return ret;
+}
+
+
+int
+dht_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int op_ret, int op_errno,
+                  inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+                  struct iatt *postparent)
 {
-	dht_conf_t   *conf                    = NULL;
         dht_local_t  *local                   = NULL;
         int           this_call_cnt           = 0;
         call_frame_t *prev                    = NULL;
-	dht_layout_t *layout                  = NULL;
-	int           ret                     = 0;
-	int           is_dir                  = 0;
+        dht_layout_t *layout                  = NULL;
+        int           ret                     = -1;
+        int           is_dir                  = 0;
+        int           is_linkfile             = 0;
+        int           attempt_unwind          = 0;
+        dht_conf_t   *conf                    = 0;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
 
-	conf  = this->private;
         local = frame->local;
         prev  = cookie;
+        conf = this->private;
 
-	layout = local->layout;
+        layout = local->layout;
+
+        /* Check if the gfid is different for file from other node */
+        if (!op_ret && uuid_compare (local->gfid, stbuf->ia_gfid)) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: gfid different on %s",
+                        local->loc.path, prev->this->name);
+        }
 
-        if (!op_ret && uuid_is_null (local->gfid))
-                memcpy (local->gfid, stbuf->ia_gfid, 16);
 
         LOCK (&frame->lock);
         {
                 /* TODO: assert equal mode on stbuf->st_mode and
-		   local->stbuf->st_mode
-
-		   else mkdir/chmod/chown and fix
-		*/
-		ret = dht_layout_merge (this, layout, prev->this,
-					op_ret, op_errno, xattr);
+                   local->stbuf->st_mode
 
-		if (op_ret == -1) {
-			local->op_errno = ENOENT;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"lookup of %s on %s returned error (%s)",
-				local->loc.path, prev->this->name,
-				strerror (op_errno));
-
-			goto unlock;
-		}
+                   else mkdir/chmod/chown and fix
+                */
+                ret = dht_layout_merge (this, layout, prev->this,
+                                        op_ret, op_errno, xattr);
+                if (ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to merge layouts", local->loc.path);
 
- 		is_dir = check_is_dir (inode, stbuf, xattr);
- 		if (!is_dir) {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
                         gf_log (this->name, GF_LOG_DEBUG,
-                                "lookup of %s on %s returned non dir 0%o",
+                                "lookup of %s on %s returned error (%s)",
                                 local->loc.path, prev->this->name,
-                                stbuf->ia_type);
-                        local->need_selfheal = 1;
- 			goto unlock;
+                                strerror (op_errno));
+
+                        goto unlock;
                 }
 
- 		local->op_ret = 0;
- 		if (local->xattr == NULL)
- 			local->xattr = dict_ref (xattr);
- 		if (local->inode == NULL)
- 			local->inode = inode_ref (inode);
+                is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+                                                 conf->link_xattr_name);
+                is_dir = check_is_dir (inode, stbuf, xattr);
 
-		dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
-                dht_iatt_merge (this, &local->postparent, postparent,
-                                prev->this);
+                if (is_dir) {
+                        local->dir_count ++;
+                } else {
+                        local->file_count ++;
+
+                        if (!is_linkfile) {
+                                /* real file */
+                                local->cached_subvol = prev->this;
+                                attempt_unwind = 1;
+                        } else {
+                                goto unlock;
+                        }
+                }
+
+                local->op_ret = 0;
 
-                if (prev->this == dht_first_up_subvol (this)) {
-			local->ia_ino = local->stbuf.ia_ino;
+                if (local->xattr == NULL) {
+                        local->xattr = dict_ref (xattr);
+                } else {
+                        dht_aggregate_xattr (local->xattr, xattr);
                 }
 
+                if (local->inode == NULL)
+                        local->inode = inode_ref (inode);
+
+                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+                dht_iatt_merge (this, &local->postparent, postparent,
+                                prev->this);
         }
 unlock:
         UNLOCK (&frame->lock);
+out:
+        this_call_cnt = dht_frame_return (frame);
 
+        if (is_last_call (this_call_cnt) || attempt_unwind) {
+                dht_discover_complete (this, frame);
+        }
 
-        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt))
+                DHT_STACK_DESTROY (frame);
 
-        if (is_last_call (this_call_cnt)) {
-                if (local->need_selfheal) {
-                        local->need_selfheal = 0;
-                        dht_lookup_everywhere (frame, this, &local->loc);
-                        return 0;
-                }
+        return 0;
+}
 
-		if (local->op_ret == 0) {
-			ret = dht_layout_normalize (this, &local->loc, layout);
 
-			if (ret != 0) {
-				gf_log (this->name, GF_LOG_DEBUG,
-					"fixing assignment on %s",
-					local->loc.path);
-				goto selfheal;
-			}
+int
+dht_discover (call_frame_t *frame, xlator_t *this, loc_t *loc)
+{
+        int          ret;
+        dht_local_t *local = NULL;
+        dht_conf_t  *conf = NULL;
+        int          call_cnt = 0;
+        int          op_errno = EINVAL;
+        int          i = 0;
+        call_frame_t *discover_frame = NULL;
 
-			dht_layout_set (this, local->inode, layout);
+        conf = this->private;
+        local = frame->local;
 
-			if (local->ia_ino) {
-				local->stbuf.ia_ino = local->ia_ino;
-			} else {
-				gf_log (this->name, GF_LOG_DEBUG,
-					"could not find hashed subvol for %s",
-					local->loc.path);
-			}
+        ret = dict_set_uint32 (local->xattr_req, conf->xattr_name, 4 * 4);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to set '%s' key",
+                        loc->path, conf->xattr_name);
 
-                        if (local->loc.parent)
-                                local->postparent.ia_ino =
-                                        local->loc.parent->ino;
-		}
+        ret = dict_set_uint32 (local->xattr_req, conf->link_xattr_name, 256);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to set '%s' key",
+                        loc->path, conf->link_xattr_name);
 
-		DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
-				  local->inode, &local->stbuf, local->xattr,
-                                  &local->postparent);
+        call_cnt        = conf->subvolume_cnt;
+        local->call_cnt = call_cnt;
+
+        local->layout = dht_layout_new (this, conf->subvolume_cnt);
+
+        if (!local->layout) {
+                op_errno = ENOMEM;
+                goto err;
         }
 
-	return 0;
+        uuid_copy (local->gfid, loc->gfid);
 
-selfheal:
-        dht_frame_su_do (frame);
-	ret = dht_selfheal_directory (frame, dht_lookup_selfheal_cbk,
-				      &local->loc, layout);
+        discover_frame = copy_frame (frame);
+        if (!discover_frame) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	return 0;
+        discover_frame->local = local;
+        frame->local = NULL;
+        local->main_frame = frame;
+
+        for (i = 0; i < call_cnt; i++) {
+                STACK_WIND (discover_frame, dht_discover_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->lookup,
+                            &local->loc, local->xattr_req);
+        }
+
+        return 0;
+
+err:
+        DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL);
+
+        return 0;
 }
 
+
 int
-dht_lookup_root_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                         int op_ret, int op_errno,
-                         inode_t *inode, struct iatt *stbuf, dict_t *xattr,
-                         struct iatt *postparent)
+dht_lookup_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int op_ret, int op_errno,
+                    inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+                    struct iatt *postparent)
 {
-	dht_conf_t   *conf                    = NULL;
         dht_local_t  *local                   = NULL;
         int           this_call_cnt           = 0;
         call_frame_t *prev                    = NULL;
-	dht_layout_t *layout                  = NULL;
-	int           ret                     = 0;
-	int           is_dir                  = 0;
+        dht_layout_t *layout                  = NULL;
+        int           ret                     = -1;
+        int           is_dir                  = 0;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
 
-	conf  = this->private;
         local = frame->local;
         prev  = cookie;
 
-	layout = local->layout;
+        layout = local->layout;
+
+        if (!op_ret && uuid_is_null (local->gfid))
+                memcpy (local->gfid, stbuf->ia_gfid, 16);
+
+        /* Check if the gfid is different for file from other node */
+        if (!op_ret && uuid_compare (local->gfid, stbuf->ia_gfid)) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: gfid different on %s",
+                        local->loc.path, prev->this->name);
+        }
 
         LOCK (&frame->lock);
         {
-		ret = dht_layout_merge (this, layout, prev->this,
-					op_ret, op_errno, xattr);
-
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_ERROR,
-				"lookup of %s on %s returned error (%s)",
-				local->loc.path, prev->this->name,
-				strerror (op_errno));
-			goto unlock;
-		}
+                /* TODO: assert equal mode on stbuf->st_mode and
+                   local->stbuf->st_mode
+
+                   else mkdir/chmod/chown and fix
+                */
+                ret = dht_layout_merge (this, layout, prev->this,
+                                        op_ret, op_errno, xattr);
+
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "lookup of %s on %s returned error (%s)",
+                                local->loc.path, prev->this->name,
+                                strerror (op_errno));
+
+                        goto unlock;
+                }
 
                 is_dir = check_is_dir (inode, stbuf, xattr);
                 if (!is_dir) {
-                        gf_log (this->name, GF_LOG_CRITICAL,
+                        gf_log (this->name, GF_LOG_DEBUG,
                                 "lookup of %s on %s returned non dir 0%o",
                                 local->loc.path, prev->this->name,
                                 stbuf->ia_type);
+                        local->need_selfheal = 1;
                         goto unlock;
                 }
 
                 local->op_ret = 0;
-                if (local->xattr == NULL)
+                if (local->xattr == NULL) {
                         local->xattr = dict_ref (xattr);
+                } else {
+                        dht_aggregate_xattr (local->xattr, xattr);
+                }
+
                 if (local->inode == NULL)
                         local->inode = inode_ref (inode);
 
-		dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
-
-                if (prev->this == dht_first_up_subvol (this)) {
-			local->ia_ino = local->stbuf.ia_ino;
-                }
-
+                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+                dht_iatt_merge (this, &local->postparent, postparent,
+                                prev->this);
         }
 unlock:
         UNLOCK (&frame->lock);
@@ -264,74 +481,45 @@ unlock:
         this_call_cnt = dht_frame_return (frame);
 
         if (is_last_call (this_call_cnt)) {
-		if (local->op_ret == 0) {
-			ret = dht_layout_normalize (this, &local->loc, layout);
-			if (ret != 0) {
-				gf_log (this->name, GF_LOG_INFO,
-					"fixing assignment on %s",
-					local->loc.path);
-			}
-
-			dht_layout_set (this, local->inode, layout);
-		}
-
-		DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
-				  local->inode, &local->stbuf, local->xattr,
-                                  &local->postparent);
-        }
-
-	return 0;
-}
-
-static int
-dht_do_fresh_lookup_on_root (xlator_t *this, call_frame_t *frame)
-{
-        dht_local_t  *local    = NULL;
-	dht_conf_t   *conf     = NULL;
-	int           ret      = -1;
-        int           call_cnt = 0;
-        int           i        = 0;
-
-        local = frame->local;
-        conf = this->private;
-        if (!conf)
-                goto err;
+                if (local->need_selfheal) {
+                        local->need_selfheal = 0;
+                        dht_lookup_everywhere (frame, this, &local->loc);
+                        return 0;
+                }
 
-        if (local->layout) {
-                dht_layout_unref (this, local->layout);
-                local->layout = NULL;
-        }
+                if (local->op_ret == 0) {
+                        ret = dht_layout_normalize (this, &local->loc, layout);
 
-        ret = dict_set_uint32 (local->xattr_req,
-                               "trusted.glusterfs.dht", 4 * 4);
-        if (ret)
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "failed to set the dict entry for dht");
+                        if (ret != 0) {
+                                gf_log (this->name, GF_LOG_DEBUG,
+                                        "fixing assignment on %s",
+                                        local->loc.path);
+                                goto selfheal;
+                        }
 
-        call_cnt = local->call_cnt = conf->subvolume_cnt;
+                        dht_layout_set (this, local->inode, layout);
+                }
 
-        local->layout = dht_layout_new (this,
-                                        conf->subvolume_cnt);
-        if (!local->layout) {
-                local->op_errno = ENOMEM;
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
-        }
+                if (local->loc.parent) {
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->postparent, 1);
+                }
 
-        for (i = 0; i < call_cnt; i++) {
-                STACK_WIND (frame, dht_lookup_root_dir_cbk,
-                            conf->subvolumes[i],
-                            conf->subvolumes[i]->fops->lookup,
-                            &local->loc, local->xattr_req);
+                DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+                DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
+                                  local->inode, &local->stbuf, local->xattr,
+                                  &local->postparent);
         }
 
         return 0;
-err:
-        DHT_STACK_UNWIND (lookup, frame, -1, local->op_errno,
-                          local->inode, &local->stbuf, local->xattr,
-                          &local->postparent);
-        return 0;
+
+selfheal:
+        FRAME_SU_DO (frame, dht_local_t);
+        uuid_copy (local->loc.gfid, local->gfid);
+        ret = dht_selfheal_directory (frame, dht_lookup_selfheal_cbk,
+                                      &local->loc, layout);
+out:
+        return ret;
 }
 
 int
@@ -343,152 +531,226 @@ dht_revalidate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         dht_local_t  *local         = NULL;
         int           this_call_cnt = 0;
         call_frame_t *prev          = NULL;
-	dht_layout_t *layout        = NULL;
-	dht_conf_t   *conf          = NULL;
-	int           ret  = -1;
-	int           is_dir = 0;
-	int           is_linkfile = 0;
-        unsigned char root_gfid[16] = {0,};
+        dht_layout_t *layout        = NULL;
+        dht_conf_t   *conf          = NULL;
+        int           ret  = -1;
+        int           is_dir = 0;
+        int           is_linkfile = 0;
+        call_frame_t *copy          = NULL;
+        dht_local_t  *copy_local    = NULL;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, err);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, err);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, err);
 
         local = frame->local;
         prev  = cookie;
-	conf = this->private;
+        conf = this->private;
         if (!conf)
                 goto out;
 
         LOCK (&frame->lock);
         {
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
 
-			if ((op_errno != ENOTCONN) 
+                        if ((op_errno != ENOTCONN)
                             && (op_errno != ENOENT)
                             && (op_errno != ESTALE)) {
 				gf_log (this->name, GF_LOG_INFO,
-					"subvolume %s returned -1 (%s)",
-					prev->this->name, strerror (op_errno));
+					"subvolume %s for %s returned -1 (%s)",
+					prev->this->name, local->loc.path,
+                                        strerror (op_errno));
 			}
-                        
                         if (op_errno == ESTALE) {
-                                /* propogate the ESTALE to parent. 
-                                 * setting local->layout_mismatch would send
+                                /* propagate the ESTALE to parent.
+                                 * setting local->return_estale would send
                                  * ESTALE to parent. */
-                                local->layout_mismatch = 1;
+                                local->return_estale = 1;
                         }
 
-			goto unlock;
-		}
+                        /* if it is ENOENT, we may have to do a
+                         * 'lookup_everywhere()' to make sure
+                         * the file is not migrated */
+                        if (op_errno == ENOENT) {
+                                if (IA_ISREG (local->loc.inode->ia_type)) {
+                                        local->need_lookup_everywhere = 1;
+                                }
+                        }
+                        goto unlock;
+                }
 
-		if (stbuf->ia_type != local->inode->ia_type) {
-			gf_log (this->name, GF_LOG_INFO,
-				"mismatching filetypes 0%o v/s 0%o for %s",
-				(stbuf->ia_type), (local->inode->ia_type),
-				local->loc.path);
+                if (stbuf->ia_type != local->inode->ia_type) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "mismatching filetypes 0%o v/s 0%o for %s",
+                                (stbuf->ia_type), (local->inode->ia_type),
+                                local->loc.path);
 
-			local->op_ret = -1;
-			local->op_errno = EINVAL;
+                        local->op_ret = -1;
+                        local->op_errno = EINVAL;
 
-			goto unlock;
-		}
+                        goto unlock;
+                }
 
-		layout = local->layout;
-		
-		is_dir = check_is_dir (inode, stbuf, xattr);
-		is_linkfile = check_is_linkfile (inode, stbuf, xattr);
-		
-		if (is_linkfile) {
-			gf_log (this->name, GF_LOG_INFO,
-				"linkfile found in revalidate for %s",
-				local->loc.path);
-			local->layout_mismatch = 1;
-
-			goto unlock;
-		}
+                layout = local->layout;
 
-		if (is_dir) {
-			ret = dht_layout_dir_mismatch (this, layout,
-						       prev->this, &local->loc,
-						       xattr);
-			if (ret != 0) {
-				gf_log (this->name, GF_LOG_INFO,
-					"mismatching layouts for %s", 
-					local->loc.path);
-			
-				local->layout_mismatch = 1;
+                is_dir = check_is_dir (inode, stbuf, xattr);
+                is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+                                                 conf->link_xattr_name);
 
-				goto unlock;
-			}
-		} 
-		
-		dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+                if (is_linkfile) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "linkfile found in revalidate for %s",
+                                local->loc.path);
+                        local->return_estale = 1;
+
+                        goto unlock;
+                }
+
+                if (is_dir) {
+                        ret = dht_dir_has_layout (xattr, conf->xattr_name);
+                        if (ret >= 0) {
+                                if (is_greater_time(local->stbuf.ia_ctime,
+                                                    local->stbuf.ia_ctime_nsec,
+                                                    stbuf->ia_ctime,
+                                                    stbuf->ia_ctime_nsec)) {
+                                        local->prebuf.ia_gid = stbuf->ia_gid;
+                                        local->prebuf.ia_uid = stbuf->ia_uid;
+                                }
+                        }
+                        if (local->stbuf.ia_type != IA_INVAL)
+                        {
+                                if ((local->stbuf.ia_gid != stbuf->ia_gid) ||
+                                    (local->stbuf.ia_uid != stbuf->ia_uid)) {
+                                        local->need_selfheal = 1;
+                                }
+                        }
+                        ret = dht_layout_dir_mismatch (this, layout,
+                                                       prev->this, &local->loc,
+                                                       xattr);
+                        if (ret != 0) {
+                                gf_log (this->name, GF_LOG_INFO,
+                                        "mismatching layouts for %s",
+                                        local->loc.path);
+
+                                local->layout_mismatch = 1;
+
+                                goto unlock;
+                        }
+                }
+
+                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
                 dht_iatt_merge (this, &local->postparent, postparent,
                                 prev->this);
-		
-		local->op_ret = 0;
-		local->stbuf.ia_ino = local->ia_ino;
 
-                if (local->loc.parent)
-                        local->postparent.ia_ino = local->loc.parent->ino;
+                local->op_ret = 0;
 
-		if (!local->xattr)
-			local->xattr = dict_ref (xattr);
-	}
+                if (!local->xattr) {
+                        local->xattr = dict_ref (xattr);
+                } else if (is_dir) {
+                        dht_aggregate_xattr (local->xattr, xattr);
+                }
+        }
 unlock:
-	UNLOCK (&frame->lock);
+        UNLOCK (&frame->lock);
 out:
         this_call_cnt = dht_frame_return (frame);
 
         if (is_last_call (this_call_cnt)) {
-		if (!IA_ISDIR (local->stbuf.ia_type)
-		    && (local->hashed_subvol != local->cached_subvol)
-		    && (local->stbuf.ia_nlink == 1)
+                if (!IA_ISDIR (local->stbuf.ia_type)
+                    && (local->hashed_subvol != local->cached_subvol)
+                    && (local->stbuf.ia_nlink == 1)
                     && (conf && conf->unhashed_sticky_bit)) {
-			local->stbuf.ia_prot.sticky = 1;
-		}
-
-		if (local->layout_mismatch) {
-			local->op_ret = -1;
-			local->op_errno = ESTALE;
-
-                        /* Because for 'root' inode, there is no FRESH lookup
-                         * sent from FUSE layer upon ESTALE, we need to handle
-                         * that one case here */
-                        root_gfid[15] = 1;
-                        if (!local->loc.parent &&
-                            !uuid_compare (local->loc.inode->gfid, root_gfid)) {
-                                dht_do_fresh_lookup_on_root (this, frame);
-                                return 0;
+                        local->stbuf.ia_prot.sticky = 1;
+                }
+                if (local->need_selfheal) {
+                        local->need_selfheal = 0;
+                        uuid_copy (local->gfid, local->stbuf.ia_gfid);
+                        local->stbuf.ia_gid = local->prebuf.ia_gid;
+                        local->stbuf.ia_uid = local->prebuf.ia_uid;
+                        copy = create_frame (this, this->ctx->pool);
+                        if (copy) {
+                                copy_local = dht_local_init (copy, &local->loc,
+                                                             NULL, 0);
+                                if (!copy_local)
+                                        goto cont;
+                                copy_local->stbuf = local->stbuf;
+                                copy->local = copy_local;
+                                FRAME_SU_DO (copy, dht_local_t);
+                                ret = synctask_new (this->ctx->env,
+                                                    dht_dir_attr_heal,
+                                                    dht_dir_attr_heal_done,
+                                                    copy, copy);
                         }
-		}
+                }
+cont:
+		if (local->layout_mismatch) {
+                        /* Found layout mismatch in the directory, need to
+                           fix this in the inode context */
+                        dht_layout_unref (this, local->layout);
+                        local->layout = NULL;
+                        dht_lookup_directory (frame, this, &local->loc);
+                        return 0;
+                }
 
-                WIPE (&local->postparent);
+                if (local->need_lookup_everywhere) {
+                        /* As the current layout gave ENOENT error, we would
+                           need a new layout */
+                        dht_layout_unref (this, local->layout);
+                        local->layout = NULL;
 
+                        /* We know that current cached subvol is no more
+                           valid, get the new one */
+                        local->cached_subvol = NULL;
+                        dht_lookup_everywhere (frame, this, &local->loc);
+                        return 0;
+                }
+                if (local->return_estale) {
+                        local->op_ret = -1;
+                        local->op_errno = ESTALE;
+                }
+
+                if (local->loc.parent) {
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->postparent, 1);
+                }
+
+                DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
                 DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
-				  local->inode, &local->stbuf, local->xattr,
+                                  local->inode, &local->stbuf, local->xattr,
                                   &local->postparent);
-	}
+        }
 
-        return 0;
+err:
+        return ret;
 }
 
 
 int
 dht_lookup_linkfile_create_cbk (call_frame_t *frame, void *cookie,
-				xlator_t *this,
-				int32_t op_ret, int32_t op_errno,
+                                xlator_t *this,
+                                int32_t op_ret, int32_t op_errno,
                                 inode_t *inode, struct iatt *stbuf,
-                                struct iatt *preparent, struct iatt *postparent)
+                                struct iatt *preparent, struct iatt *postparent,
+                                dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	xlator_t     *cached_subvol = NULL;
-	dht_conf_t   *conf = NULL;
+        dht_local_t  *local = NULL;
+        xlator_t     *cached_subvol = NULL;
+        dht_conf_t   *conf = NULL;
         int           ret = -1;
 
-	local = frame->local;
-	cached_subvol = local->cached_subvol;
-	conf = this->private;
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
 
-        ret = dht_layout_preset (this, local->cached_subvol, inode);
+        local = frame->local;
+        cached_subvol = local->cached_subvol;
+        conf = this->private;
+
+        ret = dht_layout_preset (this, local->cached_subvol, local->loc.inode);
         if (ret < 0) {
                 gf_log (this->name, GF_LOG_DEBUG,
                         "failed to set layout for subvolume %s",
@@ -498,74 +760,232 @@ dht_lookup_linkfile_create_cbk (call_frame_t *frame, void *cookie,
                 goto unwind;
         }
 
-	local->op_ret = 0;
-	if ((local->stbuf.ia_nlink == 1)
-	    && (conf && conf->unhashed_sticky_bit)) {
-		local->stbuf.ia_prot.sticky = 1;
-	}
+        local->op_ret = 0;
+        if ((local->stbuf.ia_nlink == 1)
+            && (conf && conf->unhashed_sticky_bit)) {
+                local->stbuf.ia_prot.sticky = 1;
+        }
 
-        if (local->loc.parent)
-                local->postparent.ia_ino = local->loc.parent->ino;
+        if (local->loc.parent) {
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           postparent, 1);
+        }
 
 unwind:
-        WIPE (&local->postparent);
+        if (local->linked == _gf_true)
+                dht_linkfile_attr_heal (frame, this);
 
-	DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
-			  local->inode, &local->stbuf, local->xattr,
+        DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+        DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
+                          local->inode, &local->stbuf, local->xattr,
                           &local->postparent);
-	return 0;
+out:
+        return ret;
+}
+
+
+int
+dht_lookup_everywhere_done (call_frame_t *frame, xlator_t *this)
+{
+        int           ret = 0;
+        dht_local_t  *local = NULL;
+        xlator_t     *hashed_subvol = NULL;
+        xlator_t     *cached_subvol = NULL;
+        dht_layout_t *layout = NULL;
+
+        local = frame->local;
+        hashed_subvol = local->hashed_subvol;
+        cached_subvol = local->cached_subvol;
+
+        if (local->file_count && local->dir_count) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "path %s exists as a file on one subvolume "
+                        "and directory on another. "
+                        "Please fix it manually",
+                        local->loc.path);
+                DHT_STACK_UNWIND (lookup, frame, -1, EIO, NULL, NULL, NULL,
+                                  NULL);
+                return 0;
+        }
+
+        if (local->dir_count) {
+                dht_lookup_directory (frame, this, &local->loc);
+                return 0;
+        }
+
+        if (!cached_subvol) {
+                DHT_STACK_UNWIND (lookup, frame, -1, ENOENT, NULL, NULL, NULL,
+                                  NULL);
+                return 0;
+        }
+
+        if (local->need_lookup_everywhere) {
+                if (uuid_compare (local->gfid, local->inode->gfid)) {
+                        /* GFID different, return error */
+                        DHT_STACK_UNWIND (lookup, frame, -1, ENOENT, NULL,
+                                          NULL, NULL, NULL);
+                        return 0;
+                }
+                local->op_ret = 0;
+                local->op_errno = 0;
+                layout = dht_layout_for_subvol (this, cached_subvol);
+                if (!layout) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "%s: no pre-set layout for subvolume %s",
+                                local->loc.path, (cached_subvol ?
+                                                  cached_subvol->name :
+                                                  "<nil>"));
+                }
+
+                ret = dht_layout_set (this, local->inode, layout);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "%s: failed to set layout for subvol %s",
+                                local->loc.path, (cached_subvol ?
+                                                  cached_subvol->name :
+                                                  "<nil>"));
+                }
+
+                if (local->loc.parent) {
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->postparent, 1);
+                }
+
+                DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+                DHT_STACK_UNWIND (lookup, frame, local->op_ret,
+                                  local->op_errno, local->inode,
+                                  &local->stbuf, local->xattr,
+                                  &local->postparent);
+                return 0;
+        }
+
+        if (!hashed_subvol) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "cannot create linkfile file for %s on %s: "
+                        "hashed subvolume cannot be found.",
+                        local->loc.path, cached_subvol->name);
+
+                local->op_ret = 0;
+                local->op_errno = 0;
+
+                ret = dht_layout_preset (frame->this, cached_subvol,
+                                         local->inode);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "failed to set layout for subvol %s",
+                                cached_subvol ? cached_subvol->name :
+                                "<nil>");
+                        local->op_ret = -1;
+                        local->op_errno = EINVAL;
+                }
+
+                if (local->loc.parent) {
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->postparent, 1);
+                }
+
+                DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+                DHT_STACK_UNWIND (lookup, frame, local->op_ret,
+                                  local->op_errno, local->inode,
+                                  &local->stbuf, local->xattr,
+                                  &local->postparent);
+                return 0;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "linking file %s existing on %s to %s (hash)",
+                local->loc.path, cached_subvol->name,
+                hashed_subvol->name);
+
+        ret = dht_linkfile_create (frame,
+                                   dht_lookup_linkfile_create_cbk, this,
+                                   cached_subvol, hashed_subvol, &local->loc);
+
+        return ret;
+}
+
+
+int
+dht_lookup_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int op_ret, int op_errno,
+                       struct iatt *preparent, struct iatt *postparent,
+                       dict_t *xdata)
+{
+        int  this_call_cnt = 0;
+
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                dht_lookup_everywhere_done (frame, this);
+        }
+
+        return 0;
 }
 
 
 int
 dht_lookup_everywhere_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			   int32_t op_ret, int32_t op_errno,
+                           int32_t op_ret, int32_t op_errno,
                            inode_t *inode, struct iatt *buf, dict_t *xattr,
                            struct iatt *postparent)
 {
-	dht_conf_t   *conf          = NULL;
         dht_local_t  *local         = NULL;
         int           this_call_cnt = 0;
         call_frame_t *prev          = NULL;
-	int           is_linkfile   = 0;
-	int           is_dir        = 0;
-	xlator_t     *subvol        = NULL;
-	loc_t        *loc           = NULL;
-	xlator_t     *link_subvol   = NULL;
-	xlator_t     *hashed_subvol = NULL;
-	xlator_t     *cached_subvol = NULL;
-        int           ret = -1;
+        int           is_linkfile   = 0;
+        int           is_dir        = 0;
+        xlator_t     *subvol        = NULL;
+        loc_t        *loc           = NULL;
+        xlator_t     *link_subvol   = NULL;
+        int           ret           = -1;
+        int32_t       fd_count      = 0;
+        dht_conf_t   *conf          = NULL;
 
-	conf   = this->private;
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
 
-	local  = frame->local;
-	loc    = &local->loc;
+        local  = frame->local;
+        loc    = &local->loc;
+        conf   = this->private;
 
-	prev   = cookie;
-	subvol = prev->this;
+        prev   = cookie;
+        subvol = prev->this;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        if (op_errno != ENOENT)
+                                local->op_errno = op_errno;
+                        goto unlock;
+                }
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			if (op_errno != ENOENT)
-				local->op_errno = op_errno;
-			goto unlock;
-		}
                 if (uuid_is_null (local->gfid))
-                        memcpy (local->gfid, buf->ia_gfid, 16);
-
-		is_linkfile = check_is_linkfile (inode, buf, xattr);
-		is_dir = check_is_dir (inode, buf, xattr);
-
-		if (is_linkfile) {
-			link_subvol = dht_linkfile_subvol (this, inode, buf,
-							   xattr);
-			gf_log (this->name, GF_LOG_DEBUG,
-				"found on %s linkfile %s (-> %s)",
-				subvol->name, loc->path,
-				link_subvol ? link_subvol->name : "''");
-			goto unlock;
-		}
+                        uuid_copy (local->gfid, buf->ia_gfid);
+
+                if (uuid_compare (local->gfid, buf->ia_gfid)) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: gfid differs on subvolume %s",
+                                loc->path, prev->this->name);
+                }
+
+                is_linkfile = check_is_linkfile (inode, buf, xattr,
+                                                 conf->link_xattr_name);
+                is_dir = check_is_dir (inode, buf, xattr);
+
+                if (is_linkfile) {
+                        link_subvol = dht_linkfile_subvol (this, inode, buf,
+                                                           xattr);
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "found on %s linkfile %s (-> %s)",
+                                subvol->name, loc->path,
+                                link_subvol ? link_subvol->name : "''");
+                        goto unlock;
+                }
+
+                /* non linkfile GFID takes precedence */
+                uuid_copy (local->gfid, buf->ia_gfid);
 
                 if (is_dir) {
                         local->dir_count++;
@@ -585,131 +1005,82 @@ dht_lookup_everywhere_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                 gf_log (this->name, GF_LOG_DEBUG,
                                         "found on %s file %s",
                                         subvol->name, loc->path);
-                                
+
                                 dht_iatt_merge (this, &local->postparent,
                                                 postparent, subvol);
                         } else {
-                                gf_log (this->name, GF_LOG_DEBUG,
+                                /* This is where we need 'rename' both entries logic */
+                                gf_log (this->name, GF_LOG_WARNING,
                                         "multiple subvolumes (%s and %s) have "
-                                        "file %s", local->cached_subvol->name,
+                                        "file %s (preferably rename the file "
+                                        "in the backend, and do a fresh lookup)",
+                                        local->cached_subvol->name,
                                         subvol->name, local->loc.path);
                         }
                 }
-	}
+        }
 unlock:
-	UNLOCK (&frame->lock);
-
-	if (is_linkfile) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"deleting stale linkfile %s on %s",
-			loc->path, subvol->name);
-		dht_linkfile_unlink (frame, this, subvol, loc);
-	}
-
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt)) {
-		hashed_subvol = local->hashed_subvol;
-		cached_subvol = local->cached_subvol;
-
-                if (local->file_count && local->dir_count) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "path %s exists as a file on one subvolume " 
-                                "and directory on another. "
-                                "Please fix it manually",
-                                loc->path);
-                        DHT_STACK_UNWIND (lookup, frame, -1, EIO, NULL, NULL, NULL,
-                                          NULL);
-                        return 0;
-                }
-
-                if (local->dir_count) {
-                        dht_lookup_directory (frame, this, &local->loc);
-                        return 0;
-                }
-
-		if (!cached_subvol) {
-			DHT_STACK_UNWIND (lookup, frame, -1, ENOENT, NULL, NULL, NULL,
-                                          NULL);
-			return 0;
-		}
-
-                if (!hashed_subvol) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "cannot create linkfile file for %s on %s: "
-                                "hashed subvolume cannot be found.",
-                                loc->path, cached_subvol->name);
-                        
-                        local->op_ret = 0;
-                        local->op_errno = 0;
-
-                        ret = dht_layout_preset (frame->this, cached_subvol,
-                                                 local->inode);
-                        if (ret < 0) {
-                                gf_log (this->name, GF_LOG_DEBUG,
-                                        "failed to set layout for subvol %s",
-                                        cached_subvol ? cached_subvol->name :
-                                        "<nil>");
-                                local->op_ret = -1;
-                                local->op_errno = EINVAL;
-                        }
-
-                        if (local->loc.parent)
-                                local->postparent.ia_ino =
-                                        local->loc.parent->ino;
-
-                        WIPE (&local->postparent);
+        UNLOCK (&frame->lock);
 
-                        DHT_STACK_UNWIND (lookup, frame, local->op_ret,
-                                          local->op_errno, local->inode,
-                                          &local->stbuf, local->xattr,
-                                          &local->postparent);
+        if (is_linkfile) {
+                ret = dict_get_int32 (xattr, GLUSTERFS_OPEN_FD_COUNT, &fd_count);
+                /* Delete the linkfile only if there are no open fds on it.
+                   if there is a open-fd, it may be in migration */
+                if (!ret && (fd_count == 0)) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "deleting stale linkfile %s on %s",
+                                loc->path, subvol->name);
+                        STACK_WIND (frame, dht_lookup_unlink_cbk,
+                                    subvol, subvol->fops->unlink, loc, 0, NULL);
                         return 0;
                 }
+        }
 
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "linking file %s existing on %s to %s (hash)",
-                        loc->path, cached_subvol->name,
-                        hashed_subvol->name);
-                        
-                dht_linkfile_create (frame, 
-                                     dht_lookup_linkfile_create_cbk,
-                                     cached_subvol, hashed_subvol, loc);
-	}
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                dht_lookup_everywhere_done (frame, this);
+        }
 
-	return 0;
+out:
+        return ret;
 }
 
 
 int
 dht_lookup_everywhere (call_frame_t *frame, xlator_t *this, loc_t *loc)
 {
-	dht_conf_t     *conf = NULL;
-	dht_local_t    *local = NULL;
-	int             i = 0;
-	int             call_cnt = 0;
+        dht_conf_t     *conf = NULL;
+        dht_local_t    *local = NULL;
+        int             i = 0;
+        int             call_cnt = 0;
 
-	conf = this->private;
-	local = frame->local;
-        if (!conf)
-                goto out;
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+        GF_VALIDATE_OR_GOTO ("dht", loc, out);
 
-	call_cnt = conf->subvolume_cnt;
-	local->call_cnt = call_cnt;
+        conf = this->private;
+        local = frame->local;
 
-	if (!local->inode)
-		local->inode = inode_ref (loc->inode);
+        call_cnt = conf->subvolume_cnt;
+        local->call_cnt = call_cnt;
 
-	for (i = 0; i < call_cnt; i++) {
-		STACK_WIND (frame, dht_lookup_everywhere_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->lookup,
-			    loc, local->xattr_req);
-	}
+        if (!local->inode)
+                local->inode = inode_ref (loc->inode);
+
+        for (i = 0; i < call_cnt; i++) {
+                STACK_WIND (frame, dht_lookup_everywhere_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->lookup,
+                            loc, local->xattr_req);
+        }
 
         return 0;
 out:
         DHT_STACK_UNWIND (lookup, frame, -1, EINVAL, NULL, NULL, NULL, NULL);
-        return 0;
+err:
+        return -1;
 }
 
 
@@ -720,60 +1091,82 @@ dht_lookup_linkfile_cbk (call_frame_t *frame, void *cookie,
                          struct iatt *postparent)
 {
         call_frame_t *prev          = NULL;
-	dht_local_t  *local         = NULL;
-	xlator_t     *subvol        = NULL;
-	loc_t        *loc           = NULL;
-	dht_conf_t   *conf          = NULL;
+        dht_local_t  *local         = NULL;
+        xlator_t     *subvol        = NULL;
+        loc_t        *loc           = NULL;
+        dht_conf_t   *conf          = NULL;
         int           ret           = 0;
 
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, unwind);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, unwind);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, unwind);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, unwind);
+
         prev   = cookie;
-	subvol = prev->this;
-	conf   = this->private;
-	local  = frame->local;
-	loc    = &local->loc;
+        subvol = prev->this;
+        conf   = this->private;
+        local  = frame->local;
+        loc    = &local->loc;
 
         if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"lookup of %s on %s (following linkfile) failed (%s)",
-			local->loc.path, subvol->name, strerror (op_errno));
-                goto err;
-	}
+                gf_log (this->name, GF_LOG_INFO,
+                        "lookup of %s on %s (following linkfile) failed (%s)",
+                        local->loc.path, subvol->name, strerror (op_errno));
+
+                /* If cached subvol returned ENOTCONN, do not do
+                lookup_everywhere. We need to make sure linkfile does not get
+                removed, which can take away the namespace, and subvol is
+                anyways down. */
+
+                if (op_errno != ENOTCONN)
+                        goto err;
+                else
+                        goto unwind;
+        }
 
         if (check_is_dir (inode, stbuf, xattr)) {
-                gf_log (this->name, GF_LOG_DEBUG,
+                gf_log (this->name, GF_LOG_INFO,
                         "lookup of %s on %s (following linkfile) reached dir",
                         local->loc.path, subvol->name);
                 goto err;
         }
 
-        if (check_is_linkfile (inode, stbuf, xattr)) {
-                gf_log (this->name, GF_LOG_DEBUG,
+        if (check_is_linkfile (inode, stbuf, xattr, conf->link_xattr_name)) {
+                gf_log (this->name, GF_LOG_INFO,
                         "lookup of %s on %s (following linkfile) reached link",
                         local->loc.path, subvol->name);
                 goto err;
         }
 
-	if ((stbuf->ia_nlink == 1)
-	    && (conf && conf->unhashed_sticky_bit)) {
-		stbuf->ia_prot.sticky = 1;
-	}
-        dht_itransform (this, prev->this, stbuf->ia_ino, &stbuf->ia_ino);
-        if (local->loc.parent)
-                postparent->ia_ino = local->loc.parent->ino;
-        
-	ret = dht_layout_preset (this, prev->this, inode);
-	if (ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
+        if (uuid_compare (local->gfid, stbuf->ia_gfid)) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: gfid different on data file on %s",
+                        local->loc.path, subvol->name);
+                goto err;
+        }
+
+        if ((stbuf->ia_nlink == 1)
+            && (conf && conf->unhashed_sticky_bit)) {
+                stbuf->ia_prot.sticky = 1;
+        }
+
+        ret = dht_layout_preset (this, prev->this, inode);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_INFO,
                         "failed to set layout for subvolume %s",
                         prev->this->name);
-		op_ret   = -1;
-		op_errno = EINVAL;
-		goto out;
-	}
+                op_ret   = -1;
+                op_errno = EINVAL;
+        }
 
-out:
-        WIPE (postparent);
+        if (local->loc.parent) {
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           postparent, 1);
+        }
 
+unwind:
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
         DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, stbuf, xattr,
                           postparent);
 
@@ -781,7 +1174,7 @@ out:
 
 err:
         dht_lookup_everywhere (frame, this, loc);
-
+out:
         return 0;
 }
 
@@ -793,23 +1186,38 @@ dht_lookup_directory (call_frame_t *frame, xlator_t *this, loc_t *loc)
         int           i = 0;
         dht_conf_t   *conf = NULL;
         dht_local_t  *local = NULL;
+        int           ret = 0;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, unwind);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, unwind);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, unwind);
+        GF_VALIDATE_OR_GOTO ("dht", loc, unwind);
 
         conf = this->private;
         local = frame->local;
 
-        if (!conf)
-                goto unwind;
-
         call_cnt        = conf->subvolume_cnt;
         local->call_cnt = call_cnt;
 
         local->layout = dht_layout_new (this, conf->subvolume_cnt);
         if (!local->layout) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
                 goto unwind;
         }
 
+        if (local->xattr != NULL) {
+                dict_unref (local->xattr);
+                local->xattr = NULL;
+        }
+
+        if (!uuid_is_null (local->gfid)) {
+                ret = dict_set_static_bin (local->xattr_req, "gfid-req",
+                                           local->gfid, 16);
+                if (ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to set gfid", local->loc.path);
+        }
+
         for (i = 0; i < call_cnt; i++) {
                 STACK_WIND (frame, dht_lookup_dir_cbk,
                             conf->subvolumes[i],
@@ -819,6 +1227,7 @@ dht_lookup_directory (call_frame_t *frame, xlator_t *this, loc_t *loc)
         return 0;
 unwind:
         DHT_STACK_UNWIND (lookup, frame, -1, ENOMEM, NULL, NULL, NULL, NULL);
+out:
         return 0;
 
 }
@@ -838,12 +1247,15 @@ dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         loc_t        *loc           = NULL;
         call_frame_t *prev          = NULL;
         int           ret           = 0;
-        uint64_t      tmp_layout    = 0;
         dht_layout_t *parent_layout = NULL;
 
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+
         conf  = this->private;
-        if (!conf)
-                goto out;
 
         prev  = cookie;
         local = frame->local;
@@ -852,99 +1264,137 @@ dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         /* This is required for handling stale linkfile deletion,
          * or any more call which happens from this 'loc'.
          */
-        if (uuid_is_null (local->gfid) && !op_ret)
+        if (!op_ret && uuid_is_null (local->gfid))
                 memcpy (local->gfid, stbuf->ia_gfid, 16);
 
-	if (ENTRY_MISSING (op_ret, op_errno)) {
+        if (ENTRY_MISSING (op_ret, op_errno)) {
+                gf_log (this->name, GF_LOG_TRACE, "Entry %s missing on subvol"
+                        " %s", loc->path, prev->this->name);
                 if (conf->search_unhashed == GF_DHT_LOOKUP_UNHASHED_ON) {
-			local->op_errno = ENOENT;
-			dht_lookup_everywhere (frame, this, loc);
-			return 0;
-		}
+                        local->op_errno = ENOENT;
+                        dht_lookup_everywhere (frame, this, loc);
+                        return 0;
+                }
                 if ((conf->search_unhashed == GF_DHT_LOOKUP_UNHASHED_AUTO) &&
                     (loc->parent)) {
-                        ret = inode_ctx_get (loc->parent, this, &tmp_layout);
-                        parent_layout = (dht_layout_t *)(long)tmp_layout;
+                        ret = dht_inode_ctx_layout_get (loc->parent, this,
+                                                        &parent_layout);
+                        if (ret || !parent_layout)
+                                goto out;
                         if (parent_layout->search_unhashed) {
                                 local->op_errno = ENOENT;
                                 dht_lookup_everywhere (frame, this, loc);
                                 return 0;
                         }
                 }
-	}
+        }
 
- 	if (op_ret == 0) {
- 		is_dir      = check_is_dir (inode, stbuf, xattr);
- 		if (is_dir) {
- 			local->inode = inode_ref (inode);
- 			local->xattr = dict_ref (xattr);
- 		}
- 	}
+        if (op_ret == 0) {
+                is_dir      = check_is_dir (inode, stbuf, xattr);
+                if (is_dir) {
+                        local->inode = inode_ref (inode);
+                        local->xattr = dict_ref (xattr);
+                }
+        }
 
- 	if (is_dir || (op_ret == -1 && op_errno == ENOTCONN)) {
+        if (is_dir || (op_ret == -1 && op_errno == ENOTCONN)) {
                 dht_lookup_directory (frame, this, &local->loc);
                 return 0;
- 	}
- 
-        if (op_ret == -1)
+        }
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG, "Lookup of %s for subvolume"
+                       " %s failed with error %s", loc->path, prev->this->name,
+                       strerror (op_errno));
                 goto out;
+        }
 
-        is_linkfile = check_is_linkfile (inode, stbuf, xattr);
-        is_dir      = check_is_dir (inode, stbuf, xattr);
+        is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+                                         conf->link_xattr_name);
 
-        if (!is_dir && !is_linkfile) {
+        if (!is_linkfile) {
                 /* non-directory and not a linkfile */
 
-		dht_itransform (this, prev->this, stbuf->ia_ino,
-				&stbuf->ia_ino);
-                if (loc->parent)
-                        postparent->ia_ino = loc->parent->ino;
-
-		ret = dht_layout_preset (this, prev->this, inode);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"could not set pre-set layout for subvolume %s",
-				prev->this->name);
-			op_ret   = -1;
-			op_errno = EINVAL;
-			goto out;
-		}
-                goto out; 
-	}
-
-        if (is_linkfile) {
-                subvol = dht_linkfile_subvol (this, inode, stbuf, xattr);
-
-                if (!subvol) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "linkfile not having link subvolume. path=%s",
-                                loc->path);
-			dht_lookup_everywhere (frame, this, loc);
-			return 0;
+                ret = dht_layout_preset (this, prev->this, inode);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "could not set pre-set layout for subvolume %s",
+                                prev->this->name);
+                        op_ret   = -1;
+                        op_errno = EINVAL;
+                        goto out;
                 }
+                goto out;
+        }
 
-		STACK_WIND (frame, dht_lookup_linkfile_cbk,
-			    subvol, subvol->fops->lookup,
-			    &local->loc, local->xattr_req);
+        subvol = dht_linkfile_subvol (this, inode, stbuf, xattr);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "linkfile not having link subvolume. path=%s",
+                        loc->path);
+                dht_lookup_everywhere (frame, this, loc);
+                return 0;
         }
 
+        STACK_WIND (frame, dht_lookup_linkfile_cbk,
+                    subvol, subvol->fops->lookup,
+                    &local->loc, local->xattr_req);
+
         return 0;
 
 out:
-        /* 
-         * FIXME: postparent->ia_size and postparent->st_blocks do not have 
-         * correct values. since, postparent corresponds to a directory these 
+        /*
+         * FIXME: postparent->ia_size and postparent->st_blocks do not have
+         * correct values. since, postparent corresponds to a directory these
          * two members should have values equal to sum of corresponding values
          * from each of the subvolume. See dht_iatt_merge for reference.
          */
 
-        WIPE (postparent);
+        if (!op_ret && local->loc.parent) {
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           postparent, 1);
+        }
 
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
         DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, stbuf, xattr,
                           postparent);
+err:
         return 0;
 }
 
+/* For directories, check if acl xattrs have been requested (by the acl xlator),
+ * if not, request for them. These xattrs are needed for dht dir self-heal to
+ * perform proper self-healing of dirs
+ */
+void
+dht_check_and_set_acl_xattr_req (inode_t *inode, dict_t *xattr_req)
+{
+        int     ret = 0;
+
+        GF_ASSERT (inode);
+        GF_ASSERT (xattr_req);
+
+        if (inode->ia_type != IA_IFDIR)
+                return;
+
+        if (!dict_get (xattr_req, POSIX_ACL_ACCESS_XATTR)) {
+                ret = dict_set_int8 (xattr_req, POSIX_ACL_ACCESS_XATTR, 0);
+                if (ret)
+                        gf_log (THIS->name, GF_LOG_WARNING,
+                                "failed to set key %s",
+                                POSIX_ACL_ACCESS_XATTR);
+        }
+
+        if (!dict_get (xattr_req, POSIX_ACL_DEFAULT_XATTR)) {
+                ret = dict_set_int8 (xattr_req, POSIX_ACL_DEFAULT_XATTR, 0);
+                if (ret)
+                        gf_log (THIS->name, GF_LOG_WARNING,
+                                "failed to set key %s",
+                                POSIX_ACL_DEFAULT_XATTR);
+        }
+
+        return;
+}
 
 int
 dht_lookup (call_frame_t *frame, xlator_t *this,
@@ -952,36 +1402,40 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
 {
         xlator_t     *subvol = NULL;
         xlator_t     *hashed_subvol = NULL;
-        xlator_t     *cached_subvol = NULL;
         dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
+        dht_conf_t   *conf = NULL;
         int           ret    = -1;
         int           op_errno = -1;
-	dht_layout_t *layout = NULL;
-	int           i = 0;
-	int           call_cnt = 0;
-
+        dht_layout_t *layout = NULL;
+        int           i = 0;
+        int           call_cnt = 0;
+        loc_t         new_loc = {0,};
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (loc, err);
         VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
 
-	conf = this->private;
+        conf = this->private;
         if (!conf)
                 goto err;
 
-        local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-        if (!dht_filter_loc_subvol_key (this, loc, &local->loc,
-                                        &hashed_subvol)) {
-                ret = loc_dup (loc, &local->loc);
+        local = dht_local_init (frame, loc, NULL, GF_FOP_LOOKUP);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        ret = dht_filter_loc_subvol_key (this, loc, &new_loc,
+                                         &hashed_subvol);
+        if (ret) {
+                loc_wipe (&local->loc);
+                ret = loc_dup (&new_loc, &local->loc);
+
+                /* we no more need 'new_loc' entries */
+                loc_wipe (&new_loc);
+
+                /* check if loc_dup() is successful */
                 if (ret == -1) {
                         op_errno = errno;
                         gf_log (this->name, GF_LOG_DEBUG,
@@ -991,22 +1445,25 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
                 }
         }
 
-	if (xattr_req) {
-		local->xattr_req = dict_ref (xattr_req);
-	} else {
-		local->xattr_req = dict_new ();
-	}
+        if (xattr_req) {
+                local->xattr_req = dict_ref (xattr_req);
+        } else {
+                local->xattr_req = dict_new ();
+        }
+
+        if (uuid_is_null (loc->pargfid) && !uuid_is_null (loc->gfid) &&
+            !__is_root_gfid (loc->inode->gfid)) {
+                local->cached_subvol = NULL;
+                dht_discover (frame, this, loc);
+                return 0;
+        }
 
         if (!hashed_subvol)
                 hashed_subvol = dht_subvol_get_hashed (this, loc);
-	cached_subvol = dht_subvol_get_cached (this, loc->inode);
-
-	local->cached_subvol = cached_subvol;
-	local->hashed_subvol = hashed_subvol;
+        local->hashed_subvol = hashed_subvol;
 
         if (is_revalidate (loc)) {
-		local->layout = layout = dht_layout_get (this, loc->inode);
-
+                layout = local->layout;
                 if (!layout) {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "revalidate without cache. path=%s",
@@ -1015,71 +1472,93 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
                         goto err;
                 }
 
-		if (layout->gen && (layout->gen < conf->gen)) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"incomplete layout failure for path=%s",
-				loc->path);
+                if (layout->gen && (layout->gen < conf->gen)) {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "incomplete layout failure for path=%s",
+                                loc->path);
 
                         dht_layout_unref (this, local->layout);
                         local->layout = NULL;
-			goto do_fresh_lookup;
-		}
+                        local->cached_subvol = NULL;
+                        goto do_fresh_lookup;
+                }
 
-		local->inode    = inode_ref (loc->inode);
-		local->ia_ino   = loc->inode->ino;
-		
-		local->call_cnt = layout->cnt;
-		call_cnt = local->call_cnt;
-		
-		/* NOTE: we don't require 'trusted.glusterfs.dht.linkto' attribute,
-		 *       revalidates directly go to the cached-subvolume.
-		 */
-		ret = dict_set_uint32 (local->xattr_req, 
-				       "trusted.glusterfs.dht", 4 * 4);
-
-		for (i = 0; i < layout->cnt; i++) {
+                local->inode = inode_ref (loc->inode);
+
+                /* NOTE: we don't require 'trusted.glusterfs.dht.linkto' attribute,
+                 *       revalidates directly go to the cached-subvolume.
+                 */
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->xattr_name, 4 * 4);
+
+                if (IA_ISDIR (local->inode->ia_type)) {
+                        local->call_cnt = call_cnt = conf->subvolume_cnt;
+                        for (i = 0; i < call_cnt; i++) {
+                                STACK_WIND (frame, dht_revalidate_cbk,
+                                            conf->subvolumes[i],
+                                            conf->subvolumes[i]->fops->lookup,
+                                            loc, local->xattr_req);
+                        }
+                        return 0;
+                }
+
+                call_cnt = local->call_cnt = layout->cnt;
+
+                /* need it for self-healing linkfiles which is
+                   'in-migration' state */
+                ret = dict_set_uint32 (local->xattr_req,
+                                       GLUSTERFS_OPEN_FD_COUNT, 4);
+
+                /* need it for dir self-heal */
+                dht_check_and_set_acl_xattr_req (loc->inode, local->xattr_req);
+
+		for (i = 0; i < call_cnt; i++) {
 			subvol = layout->list[i].xlator;
-			
+
 			STACK_WIND (frame, dht_revalidate_cbk,
 				    subvol, subvol->fops->lookup,
 				    &local->loc, local->xattr_req);
 
-			if (!--call_cnt)
-				break;
 		}
         } else {
         do_fresh_lookup:
-		/* TODO: remove the hard-coding */
-		ret = dict_set_uint32 (local->xattr_req, 
-				       "trusted.glusterfs.dht", 4 * 4);
+                /* TODO: remove the hard-coding */
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->xattr_name, 4 * 4);
+
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->link_xattr_name, 256);
 
-		ret = dict_set_uint32 (local->xattr_req, 
-				       "trusted.glusterfs.dht.linkto", 256);
+                /* need it for self-healing linkfiles which is
+                   'in-migration' state */
+                ret = dict_set_uint32 (local->xattr_req,
+                                       GLUSTERFS_OPEN_FD_COUNT, 4);
+
+                /* need it for dir self-heal */
+                dht_check_and_set_acl_xattr_req (loc->inode, local->xattr_req);
 
                 if (!hashed_subvol) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"no subvolume in layout for path=%s, "
-				"checking on all the subvols to see if "
-				"it is a directory", loc->path);
- 			call_cnt        = conf->subvolume_cnt;
- 			local->call_cnt = call_cnt;
- 			
- 			local->layout = dht_layout_new (this,
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "no subvolume in layout for path=%s, "
+                                "checking on all the subvols to see if "
+                                "it is a directory", loc->path);
+                        call_cnt        = conf->subvolume_cnt;
+                        local->call_cnt = call_cnt;
+
+                        local->layout = dht_layout_new (this,
                                                         conf->subvolume_cnt);
- 			if (!local->layout) {
- 				op_errno = ENOMEM;
- 				gf_log (this->name, GF_LOG_ERROR,
- 					"Out of memory");
- 				goto err;
- 			}
-
-			for (i = 0; i < call_cnt; i++) {
- 				STACK_WIND (frame, dht_lookup_dir_cbk,
- 					    conf->subvolumes[i],
- 					    conf->subvolumes[i]->fops->lookup,
- 					    &local->loc, local->xattr_req);
- 			}
- 			return 0;
+                        if (!local->layout) {
+                                op_errno = ENOMEM;
+                                goto err;
+                        }
+
+                        for (i = 0; i < call_cnt; i++) {
+                                STACK_WIND (frame, dht_lookup_dir_cbk,
+                                            conf->subvolumes[i],
+                                            conf->subvolumes[i]->fops->lookup,
+                                            &local->loc, local->xattr_req);
+                        }
+                        return 0;
                 }
 
                 STACK_WIND (frame, dht_lookup_cbk,
@@ -1090,336 +1569,52 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
         return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-        DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
-        return 0;
-}
-
-
-int
-dht_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	      int op_ret, int op_errno, struct iatt *prebuf,
-              struct iatt *postbuf)
-{
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
-
-
-	local = frame->local;
-	prev = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-                        local->op_ret = -1;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
-
-                dht_iatt_merge (this, &local->prebuf, prebuf, prev->this);
-		dht_iatt_merge (this, &local->stbuf, postbuf, prev->this);
-
-		if (local->inode) {
-			local->stbuf.ia_ino = local->inode->ino;
-                        local->prebuf.ia_ino = local->inode->ino;
-                }
-
-		local->op_ret = 0;
-	}
-unlock:
-	UNLOCK (&frame->lock);
-
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (truncate, frame, local->op_ret, local->op_errno,
-				  &local->prebuf, &local->stbuf);
-
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL);
         return 0;
 }
 
 
-
-int
-dht_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	      int op_ret, int op_errno, struct iatt *stbuf)
-{
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
-
-
-	local = frame->local;
-	prev = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
-
-		dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
-		
-		if (local->inode)
-			local->stbuf.ia_ino = local->inode->ino;
-		local->op_ret = 0;
-	}
-unlock:
-	UNLOCK (&frame->lock);
-
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (stat, frame, local->op_ret, local->op_errno,
-				  &local->stbuf);
-
-        return 0;
-}
-
-
-int
-dht_stat (call_frame_t *frame, xlator_t *this,
-	  loc_t *loc)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-	dht_layout_t *layout = NULL;
-	int           i = 0;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
-
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	local->layout = layout = dht_layout_get (this, loc->inode);
-	if (!layout) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no layout for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = layout->cnt;
-
-	for (i = 0; i < layout->cnt; i++) {
-		subvol = layout->list[i].xlator;
-
-		STACK_WIND (frame, dht_attr_cbk,
-			    subvol, subvol->fops->stat,
-			    loc);
-	}
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL);
-
-	return 0;
-}
-
-
-int
-dht_fstat (call_frame_t *frame, xlator_t *this,
-	   fd_t *fd)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-	dht_layout_t *layout = NULL;
-	int           i = 0;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	local->layout = layout = dht_layout_get (this, fd->inode);
-	if (!layout) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no layout for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local->inode    = inode_ref (fd->inode);
-	local->call_cnt = layout->cnt;;
-
-	for (i = 0; i < layout->cnt; i++) {
-		subvol = layout->list[i].xlator;
-		STACK_WIND (frame, dht_attr_cbk,
-			    subvol, subvol->fops->fstat,
-			    fd);
-	}
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (fstat, frame, -1, op_errno, NULL);
-
-	return 0;
-}
-
-
-int
-dht_truncate (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, off_t offset)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
-
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = 1;
-
-	STACK_WIND (frame, dht_truncate_cbk,
-		    subvol, subvol->fops->truncate,
-		    loc, offset);
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL);
-
-	return 0;
-}
-
-
-int
-dht_ftruncate (call_frame_t *frame, xlator_t *this,
-	       fd_t *fd, off_t offset)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
-
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	local->inode = inode_ref (fd->inode);
-	local->call_cnt = 1;
-
-	STACK_WIND (frame, dht_truncate_cbk,
-		    subvol, subvol->fops->ftruncate,
-		    fd, offset);
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL);
-
-	return 0;
-}
-
-
 int
 dht_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 int op_ret, int op_errno, struct iatt *preparent,
-                struct iatt *postparent)
+                struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	call_frame_t *prev  = NULL;
+        dht_local_t  *local = NULL;
+        call_frame_t *prev  = NULL;
 
-	local = frame->local;
-	prev  = cookie;
+        local = frame->local;
+        prev  = cookie;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
                         local->op_ret   = -1;
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "subvolume %s returned -1 (%s)",
+                                prev->this->name, strerror (op_errno));
+                        goto unlock;
+                }
 
-                preparent->ia_ino = local->loc.parent->ino;
-                postparent->ia_ino = local->loc.parent->ino;
-		local->op_ret = 0;
+                local->op_ret = 0;
 
                 local->postparent = *postparent;
                 local->preparent = *preparent;
 
-                WIPE (&local->postparent);
-                WIPE (&local->preparent);
-	}
+                if (local->loc.parent) {
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->preparent, 0);
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->postparent, 1);
+                }
+        }
 unlock:
-	UNLOCK (&frame->lock);
+        UNLOCK (&frame->lock);
 
         DHT_STACK_UNWIND (unlink, frame, local->op_ret, local->op_errno,
-                          &local->preparent, &local->postparent);
+                          &local->preparent, &local->postparent, NULL);
 
         return 0;
 }
@@ -1428,32 +1623,33 @@ unlock:
 int
 dht_unlink_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                          int op_ret, int op_errno, struct iatt *preparent,
-                         struct iatt *postparent)
+                         struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	call_frame_t *prev = NULL;
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
 
         xlator_t *cached_subvol = NULL;
 
-	local = frame->local;
-	prev  = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
+        local = frame->local;
+        prev  = cookie;
 
-		local->op_ret = 0;
-	}
+        LOCK (&frame->lock);
+        {
+                if ((op_ret == -1) && !((op_errno == ENOENT) ||
+                                        (op_errno == ENOTCONN))) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "subvolume %s returned -1 (%s)",
+                                prev->this->name, strerror (op_errno));
+                        goto unlock;
+                }
+
+                local->op_ret = 0;
+        }
 unlock:
-	UNLOCK (&frame->lock);
+        UNLOCK (&frame->lock);
 
-        if (op_ret == -1)
+        if (local->op_ret == -1)
                 goto err;
 
         cached_subvol = dht_subvol_get_cached (this, local->loc.inode);
@@ -1467,321 +1663,314 @@ unlock:
 
         STACK_WIND (frame, dht_unlink_cbk,
                     cached_subvol, cached_subvol->fops->unlink,
-                    &local->loc);
+                    &local->loc, local->flags, NULL);
 
         return 0;
 
 err:
         DHT_STACK_UNWIND (unlink, frame, -1, local->op_errno,
-                          NULL, NULL);
+                          NULL, NULL, NULL);
         return 0;
 }
 
-
 int
-dht_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
-               int op_errno, struct iatt *prebuf, struct iatt *postbuf)
+dht_err_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+             int op_ret, int op_errno, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
 
+        local = frame->local;
+        prev = cookie;
 
-	local = frame->local;
-	prev = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "subvolume %s returned -1 (%s)",
+                                prev->this->name, strerror (op_errno));
+                        goto unlock;
+                }
 
-		local->op_ret = 0;
-	}
+                local->op_ret = 0;
+        }
 unlock:
-	UNLOCK (&frame->lock);
+        UNLOCK (&frame->lock);
 
-        if (local && (op_ret == 0)) {
-                prebuf->ia_ino = local->ia_ino;
-                postbuf->ia_ino = local->ia_ino;
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                DHT_STACK_UNWIND (setxattr, frame, local->op_ret,
+                                  local->op_errno, NULL);
         }
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (fsync, frame, local->op_ret, local->op_errno,
-                                  prebuf, postbuf);
-
         return 0;
 }
 
-
-
-int
-dht_err_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	     int op_ret, int op_errno)
+static void
+fill_layout_info (dht_layout_t *layout, char *buf)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
-
-
-	local = frame->local;
-	prev = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
-
-		local->op_ret = 0;
-	}
-unlock:
-	UNLOCK (&frame->lock);
+        int i = 0;
+        char tmp_buf[128] = {0,};
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt)) {
-		DHT_STACK_UNWIND (setxattr, frame, local->op_ret, local->op_errno);
+        for (i = 0; i < layout->cnt; i++) {
+                snprintf (tmp_buf, 128, "(%s %u %u)",
+                          layout->list[i].xlator->name,
+                          layout->list[i].start,
+                          layout->list[i].stop);
+                if (i)
+                        strcat (buf, " ");
+                strcat (buf, tmp_buf);
         }
-
-        return 0;
 }
 
+void
+dht_fill_pathinfo_xattr (xlator_t *this, dht_local_t *local,
+                         char *xattr_buf, int32_t alloc_len,
+                         int flag, char *layout_buf)
+{
+        if (flag && local->xattr_val)
+                snprintf (xattr_buf, alloc_len,
+                          "((<"DHT_PATHINFO_HEADER"%s> %s) (%s-layout %s))",
+                          this->name, local->xattr_val, this->name,
+                          layout_buf);
+        else if (local->xattr_val)
+                snprintf (xattr_buf, alloc_len,
+                          "(<"DHT_PATHINFO_HEADER"%s> %s)",
+                          this->name, local->xattr_val);
+        else if (flag)
+                snprintf (xattr_buf, alloc_len, "(%s-layout %s)",
+                          this->name, layout_buf);
+}
 
 int
-dht_access (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, int32_t mask)
+dht_vgetxattr_alloc_and_fill (dht_local_t *local, dict_t *xattr, xlator_t *this,
+                              int op_errno)
 {
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
+        int      ret       = -1;
+        char    *value     = NULL;
+        int32_t  plen      = 0;
 
+        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;
+        }
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
-
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        local->alloc_len += strlen(value);
 
-	local->call_cnt = 1;
+        if (!local->xattr_val) {
+                local->alloc_len += (strlen (DHT_PATHINFO_HEADER) + 10);
+                local->xattr_val = GF_CALLOC (local->alloc_len, sizeof (char),
+                                              gf_common_mt_char);
+                if (!local->xattr_val) {
+                        ret = -1;
+                        goto out;
+                }
+        }
 
-	STACK_WIND (frame, dht_err_cbk,
-		    subvol, subvol->fops->access,
-		    loc, mask);
+        if (local->xattr_val) {
+                plen = strlen (local->xattr_val);
+                if (plen) {
+                        /* extra byte(s) for \0 to be safe */
+                        local->alloc_len += (plen + 2);
+                        local->xattr_val = GF_REALLOC (local->xattr_val,
+                                                       local->alloc_len);
+                        if (!local->xattr_val) {
+                                ret = -1;
+                                goto out;
+                        }
+                }
 
-	return 0;
+                (void) strcat (local->xattr_val, value);
+                (void) strcat (local->xattr_val, " ");
+                local->op_ret = 0;
+        }
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (access, frame, -1, op_errno);
+        ret = 0;
 
-	return 0;
+ out:
+        return ret;
 }
 
-
 int
-dht_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		  int op_ret, int op_errno, const char *path, struct iatt *sbuf)
+dht_vgetxattr_fill_and_set (dht_local_t *local, dict_t **dict, xlator_t *this,
+                            gf_boolean_t flag)
 {
-        dht_local_t *local = NULL;
+        int   ret             = -1;
+        char *xattr_buf       = NULL;
+        char layout_buf[8192] = {0,};
 
-        local = frame->local;
-        if (op_ret == -1)
-                goto err;
+        if (flag)
+                fill_layout_info (local->layout, layout_buf);
+
+        *dict = dict_new ();
+        if (!*dict)
+                goto out;
 
-        if (local) {
-                sbuf->ia_ino = local->ia_ino;
+        local->xattr_val[strlen (local->xattr_val) - 1] = '\0';
+
+        /* we would need max this many bytes to create xattr string
+         * extra 40 bytes is just an estimated amount of additional
+         * space required as we include translator name and some
+         * spaces, brackets etc. when forming the pathinfo string.
+         *
+         * For node-uuid we just don't have all the pretty formatting,
+         * but since this is a generic routine for pathinfo & node-uuid
+         * we dont have conditional space allocation and try to be
+         * generic
+         */
+        local->alloc_len += (2 * strlen (this->name))
+                + strlen (layout_buf)
+                + 40;
+        xattr_buf = GF_CALLOC (local->alloc_len, sizeof (char),
+                               gf_common_mt_char);
+        if (!xattr_buf)
+                goto out;
+
+        if (XATTR_IS_PATHINFO (local->xsel)) {
+                (void) dht_fill_pathinfo_xattr (this, local, xattr_buf,
+                                                local->alloc_len, flag,
+                                                layout_buf);
+        } else if (XATTR_IS_NODE_UUID (local->xsel)) {
+                (void) snprintf (xattr_buf, local->alloc_len, "%s",
+                                 local->xattr_val);
         } else {
-                op_ret = -1;
-                op_errno = EINVAL;
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Unknown local->xsel (%s)", local->xsel);
+                GF_FREE (xattr_buf);
+                goto out;
         }
 
-err:
-        DHT_STACK_UNWIND (readlink, frame, op_ret, op_errno, path, sbuf);
+        ret = dict_set_dynstr (*dict, local->xsel, xattr_buf);
+        if (ret)
+                GF_FREE (xattr_buf);
+        GF_FREE (local->xattr_val);
 
-        return 0;
+ out:
+        return ret;
 }
 
-
 int
-dht_readlink (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, size_t size)
+dht_vgetxattr_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-        dht_local_t  *local = NULL;
+        int          ret           = 0;
+        dht_local_t *local         = NULL;
+        int          this_call_cnt = 0;
+        dict_t      *dict          = NULL;
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (frame->local, out);
 
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        local = frame->local;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        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;
+                        }
 
-        local->ia_ino = loc->inode->ino;
- 
-	STACK_WIND (frame, dht_readlink_cbk,
-		    subvol, subvol->fops->readlink,
-		    loc, size);
+                        goto unlock;
+                }
 
-	return 0;
+                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);
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (readlink, frame, -1, op_errno, NULL, NULL);
+        if (!is_last_call (this_call_cnt))
+                goto out;
 
-	return 0;
-}
+        /* -- last call: do patch ups -- */
 
+        if (local->op_ret == -1) {
+                goto unwind;
+        }
 
-int
-dht_fix_layout_cbk (call_frame_t *frame, void *cookie,
-                    xlator_t *this, int32_t op_ret, int32_t op_errno)
-{
-        DHT_STACK_UNWIND (getxattr, frame, -1, ENODATA, NULL);
+        ret = dht_vgetxattr_fill_and_set (local, &dict, this, _gf_true);
+        if (ret)
+                goto unwind;
 
-        return 0;
-}
+        DHT_STACK_UNWIND (getxattr, frame, 0, 0, dict, xdata);
+        goto cleanup;
 
-static void
-fill_layout_info (dht_layout_t *layout, char *buf)
-{
-        int i = 0;
-        char tmp_buf[128] = {0,};
-
-        for (i = 0; i < layout->cnt; i++) {
-                snprintf (tmp_buf, 128, "(%s %u %u)",
-                          layout->list[i].xlator->name,
-                          layout->list[i].start,
-                          layout->list[i].stop);
-                if (i)
-                        strcat (buf, " ");
-                strcat (buf, tmp_buf);
-        }
+ unwind:
+        DHT_STACK_UNWIND (getxattr, frame, -1, local->op_errno, NULL, NULL);
+ cleanup:
+        if (dict)
+                dict_unref (dict);
+ out:
+        return 0;
 }
 
 int
-dht_pathinfo_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                           int op_ret, int op_errno, dict_t *xattr)
+dht_vgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-        int ret = 0;
-        int flag = 0;
-        int this_call_cnt = 0;
-        char *value_got = NULL;
-        char  layout_buf[8192] = {0,};
-        char  xattr_buf[8192 + 1024] = {0,};
-        dict_t *dict = NULL;
+        dht_local_t  *local         = NULL;
+        int           ret           = 0;
+        dict_t       *dict          = NULL;
+        call_frame_t *prev          = NULL;
+        gf_boolean_t  flag          = _gf_true;
 
         local = frame->local;
+        prev = cookie;
 
-        if (op_ret != -1) {
-                ret = dict_get_str (xattr, GF_XATTR_PATHINFO_KEY, &value_got);
-                if (!ret) {
-                        if (!local->pathinfo)
-                                local->pathinfo = GF_CALLOC (8192, sizeof (char),
-                                                             gf_common_mt_char);
-                        if (local->pathinfo)
-                                strcat (local->pathinfo, value_got);
-                }
+        if (op_ret < 0) {
+                local->op_ret = -1;
+                local->op_errno = op_errno;
+                gf_log (this->name, GF_LOG_ERROR, "Subvolume %s returned -1 "
+                        "(%s)", prev->this->name, strerror (op_errno));
+                goto unwind;
         }
 
-        this_call_cnt = dht_frame_return (frame);
-        if (is_last_call (this_call_cnt)) {
-                if (local->layout->cnt > 1) {
-                        /* Set it for directory */
-                        fill_layout_info (local->layout, layout_buf);
-                        flag = 1;
-                }
-
-                dict = dict_new ();
-
-                if (flag && local->pathinfo)
-                        snprintf (xattr_buf, 9216, "((%s %s) (%s-layout %s))",
-                                  this->name, local->pathinfo, this->name,
-                                  layout_buf);
-                else if (local->pathinfo)
-                        snprintf (xattr_buf, 9216, "(%s %s)",
-                                  this->name, local->pathinfo);
-                else if (flag)
-                        snprintf (xattr_buf, 9216, "(%s-layout %s)",
-                                  this->name, layout_buf);
-
-                ret = dict_set_str (dict, GF_XATTR_PATHINFO_KEY,
-                                    xattr_buf);
-
-                if (local->pathinfo)
-                        GF_FREE (local->pathinfo);
-                GF_FREE (local->key);
-
-                DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict);
+        ret = dht_vgetxattr_alloc_and_fill (local, xattr, this,
+                                            op_errno);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "alloc or fill failure");
+                goto unwind;
+        }
 
-                if (dict)
-                        dict_unref (dict);
+        flag = (local->layout->cnt > 1) ? _gf_true : _gf_false;
 
-                return 0;
-        }
+        ret = dht_vgetxattr_fill_and_set (local, &dict, this, flag);
+        if (ret)
+                goto unwind;
 
-        if (local->pathinfo)
-                strcat (local->pathinfo, " Link: ");
-        if (local->hashed_subvol) {
-                /* This will happen if there pending */
-                STACK_WIND (frame, dht_pathinfo_getxattr_cbk, local->hashed_subvol,
-                            local->hashed_subvol->fops->getxattr,
-                            &local->loc, local->key);
+        DHT_STACK_UNWIND (getxattr, frame, 0, 0, dict, xdata);
+        goto cleanup;
 
-                return 0;
-        }
+ unwind:
+        DHT_STACK_UNWIND (getxattr, frame, -1, local->op_errno,
+                          NULL, NULL);
+ cleanup:
+        if (dict)
+                dict_unref (dict);
 
-        gf_log ("this->name", GF_LOG_ERROR, "Unable to find hashed_subvol for path"
-                " %s", local->pathinfo);
-      
-        DHT_STACK_UNWIND (getxattr, frame, -1, op_errno, dict);
         return 0;
 }
 
 int
 dht_linkinfo_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                           int op_ret, int op_errno, dict_t *xattr)
+                           int op_ret, int op_errno, dict_t *xattr,
+                           dict_t *xdata)
 {
         int   ret   = 0;
         char *value = NULL;
@@ -1796,21 +1985,24 @@ dht_linkinfo_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 }
         }
 
-        DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr);
+        DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr, xdata);
 
         return 0;
 }
 
 int
 dht_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		  int op_ret, int op_errno, dict_t *xattr)
+                  int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
 {
         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);
@@ -1818,37 +2010,114 @@ 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);
         }
+
+        if (frame->root->pid >= 0 ) {
+                GF_REMOVE_INTERNAL_XATTR("trusted.glusterfs.quota*", xattr);
+                GF_REMOVE_INTERNAL_XATTR("trusted.pgfid*", xattr);
+        }
+
         local->op_ret = 0;
 
         if (!local->xattr) {
                 local->xattr = dict_copy_with_ref (xattr, NULL);
         } else {
-                local->xattr = dict_copy (xattr, local->xattr);
+                dht_aggregate_xattr (local->xattr, xattr);
         }
 out:
         if (is_last_call (this_call_cnt)) {
-                DHT_STACK_UNWIND (getxattr, frame, local->op_ret, op_errno, local->xattr);
+                DHT_STACK_UNWIND (getxattr, frame, local->op_ret, op_errno,
+                                  local->xattr, NULL);
         }
         return 0;
 }
 
+int32_t
+dht_getxattr_unwind (call_frame_t *frame,
+                     int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
+{
+        DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+}
+
+
+int
+dht_getxattr_get_real_filename_cbk (call_frame_t *frame, void *cookie,
+				    xlator_t *this, int op_ret, int op_errno,
+				    dict_t *xattr, dict_t *xdata)
+{
+        int             this_call_cnt = 0;
+        dht_local_t     *local = NULL;
+
+
+        local = frame->local;
+
+	if (op_ret != -1) {
+		if (local->xattr)
+			dict_unref (local->xattr);
+		local->xattr = dict_ref (xattr);
+
+		if (local->xattr_req)
+			dict_unref (local->xattr_req);
+		local->xattr_req = dict_ref (xdata);
+	}
+
+	this_call_cnt = dht_frame_return (frame);
+	if (is_last_call (this_call_cnt)) {
+		DHT_STACK_UNWIND (getxattr, frame, local->op_ret, op_errno,
+				  local->xattr, local->xattr_req);
+	}
+
+	return 0;
+}
+
+
+int
+dht_getxattr_get_real_filename (call_frame_t *frame, xlator_t *this,
+				loc_t *loc, const char *key, dict_t *xdata)
+{
+	dht_local_t     *local = NULL;
+	int              i = 0;
+	dht_layout_t    *layout = NULL;
+	int              cnt = 0;
+	xlator_t        *subvol = NULL;
+
+
+	local = frame->local;
+	layout = local->layout;
+
+	cnt = local->call_cnt = layout->cnt;
+
+	local->op_ret = -1;
+	local->op_errno = ENODATA;
+
+	for (i = 0; i < cnt; i++) {
+		subvol = layout->list[i].xlator;
+		STACK_WIND (frame, dht_getxattr_get_real_filename_cbk,
+			    subvol, subvol->fops->getxattr,
+			    loc, key, xdata);
+	}
+
+	return 0;
+}
+
 
 int
 dht_getxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, const char *key)
+              loc_t *loc, const char *key, dict_t *xdata)
+#define DHT_IS_DIR(layout)  (layout->cnt > 1)
 {
-	xlator_t     *subvol        = NULL;
-	xlator_t     *hashed_subvol = NULL;
-	xlator_t     *cached_subvol = NULL;
-	dht_conf_t   *conf          = NULL;
-	dht_local_t  *local         = NULL;
+
+        xlator_t     *subvol        = NULL;
+        xlator_t     *hashed_subvol = NULL;
+        xlator_t     *cached_subvol = NULL;
+        dht_conf_t   *conf          = NULL;
+        dht_local_t  *local         = NULL;
         dht_layout_t *layout        = NULL;
+        xlator_t     **sub_volumes  = NULL;
         int           op_errno      = -1;
-        int           ret           = 0;
-        int           flag          = 0;
         int           i             = 0;
         int           cnt           = 0;
 
@@ -1856,11 +2125,18 @@ 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;
-        layout = dht_layout_get (this, loc->inode);
+
+        local = dht_local_init (frame, loc, NULL, GF_FOP_GETXATTR);
+        if (!local) {
+                op_errno = ENOMEM;
+
+                goto err;
+        }
+
+        layout = local->layout;
         if (!layout) {
                 gf_log (this->name, GF_LOG_ERROR,
                         "layout is NULL");
@@ -1868,48 +2144,75 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
                 goto err;
         }
 
-        if (key && (strcmp (key, GF_XATTR_PATHINFO_KEY) == 0)) {
-                hashed_subvol = dht_subvol_get_hashed (this, loc);
-                cached_subvol = dht_subvol_get_cached (this, loc->inode);
-
-                local = dht_local_init (frame);
-                if (!local) {
-                        op_errno = ENOMEM;
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        goto err;
-                }
-
-                ret = loc_dup (loc, &local->loc);
-                if (ret == -1) {
-                        op_errno = ENOMEM;
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        goto err;
-                }
+        if (key) {
                 local->key = gf_strdup (key);
                 if (!local->key) {
                         op_errno = ENOMEM;
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
                         goto err;
                 }
-                local->layout = layout;
+        }
 
-                local->call_cnt = 1;
-                if (hashed_subvol != cached_subvol) {
-                        local->call_cnt = 2;
-                        local->hashed_subvol = hashed_subvol;
+	if (key &&
+	    (strncmp (key, GF_XATTR_GET_REAL_FILENAME_KEY,
+		      strlen (GF_XATTR_GET_REAL_FILENAME_KEY)) == 0)
+	    && DHT_IS_DIR(layout)) {
+		dht_getxattr_get_real_filename (frame, this, loc, key, xdata);
+		return 0;
+	}
+
+        /* for file use cached subvolume (obviously!): see if {}
+         * below
+         * for directory:
+         *  wind to all subvolumes and exclude subvolumes which
+         *  return ENOTCONN (in callback)
+         *
+         * NOTE: Don't trust inode here, as that may not be valid
+         *       (until inode_link() happens)
+         */
+        if (key && DHT_IS_DIR(layout) &&
+            (XATTR_IS_PATHINFO (key)
+             || (strcmp (key, GF_XATTR_NODE_UUID_KEY) == 0))) {
+                (void) strncpy (local->xsel, key, 256);
+                cnt = local->call_cnt = layout->cnt;
+                for (i = 0; i < cnt; i++) {
+                        subvol = layout->list[i].xlator;
+                        STACK_WIND (frame, dht_vgetxattr_dir_cbk,
+                                    subvol, subvol->fops->getxattr,
+                                    loc, key, NULL);
                 }
+                return 0;
+        }
 
-                STACK_WIND (frame, dht_pathinfo_getxattr_cbk, cached_subvol,
-                            cached_subvol->fops->getxattr, loc, key);
+        /* node-uuid or pathinfo for files */
+        if (key && ((strcmp (key, GF_XATTR_NODE_UUID_KEY) == 0)
+                    || XATTR_IS_PATHINFO (key))) {
+                cached_subvol = local->cached_subvol;
+                (void) strncpy (local->xsel, key, 256);
+
+                local->call_cnt = 1;
+                STACK_WIND (frame, dht_vgetxattr_cbk, cached_subvol,
+                            cached_subvol->fops->getxattr, loc, key, NULL);
 
                 return 0;
         }
+
         if (key && (strcmp (key, GF_XATTR_LINKINFO_KEY) == 0)) {
                 hashed_subvol = dht_subvol_get_hashed (this, loc);
+                if (!hashed_subvol) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get"
+                                "hashed subvol for %s", loc->path);
+                        op_errno = EINVAL;
+                        goto err;
+                }
+
                 cached_subvol = dht_subvol_get_cached (this, loc->inode);
+                if (!cached_subvol) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get"
+                                "cached subvol for %s", loc->path);
+                        op_errno = EINVAL;
+                        goto err;
+                }
+
                 if (hashed_subvol == cached_subvol) {
                         op_errno = ENODATA;
                         goto err;
@@ -1917,62 +2220,126 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
                 if (hashed_subvol) {
                         STACK_WIND (frame, dht_linkinfo_getxattr_cbk, hashed_subvol,
                                     hashed_subvol->fops->getxattr, loc,
-                                    GF_XATTR_PATHINFO_KEY);
+                                    GF_XATTR_PATHINFO_KEY, NULL);
                         return 0;
                 }
                 op_errno = ENODATA;
                 goto err;
         }
-        if (key && (strcmp (key, GF_XATTR_FIX_LAYOUT_KEY) == 0)) {
-                for (i = 0; i < layout->cnt; i++) {
-                        if (layout->list[i].start == layout->list[i].stop) {
-                                flag = 1;
-                                break;
-                        }
+
+        if (key && (!strcmp (GF_XATTR_MARKER_KEY, key))
+            && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+                if (DHT_IS_DIR(layout)) {
+                        cnt = layout->cnt;
+                } else {
+                        cnt = 1;
                 }
-                if ((layout->cnt < conf->subvolume_cnt) || flag) {
-                        gf_log (this->name, GF_LOG_INFO,
-                                "expanding layout of %s from %d to %d",
-                                loc->path, layout->cnt, conf->subvolume_cnt);
-                        local = dht_local_init (frame);
-                        if (!local) {
-                                op_errno = ENOMEM;
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "Out of memory");
-                                goto err;
-                        }
 
-                        ret = loc_dup (loc, &local->loc);
-                        if (ret == -1) {
-                                op_errno = ENOMEM;
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "Out of memory");
+                sub_volumes = alloca ( cnt * sizeof (xlator_t *));
+                for (i = 0; i < cnt; i++)
+                        *(sub_volumes + i) = layout->list[i].xlator;
+
+                if (cluster_getmarkerattr (frame, this, loc, key,
+                                           local, dht_getxattr_unwind,
+                                           sub_volumes, cnt,
+                                           MARKER_UUID_TYPE, marker_uuid_default_gauge,
+                                           conf->vol_uuid)) {
+                        op_errno = EINVAL;
+                        goto err;
+                }
+
+                return 0;
+        }
+
+        if (key && !strcmp (GF_XATTR_QUOTA_LIMIT_LIST, key)) {
+                /* quota hardlimit and aggregated size of a directory is stored
+                 * in inode contexts of each brick. Hence its good enough that
+                 * we send getxattr for this key to any brick.
+                 */
+                local->call_cnt = 1;
+                subvol = dht_first_up_subvol (this);
+                STACK_WIND (frame, dht_getxattr_cbk, subvol,
+                            subvol->fops->getxattr, loc, key, xdata);
+                return 0;
+        }
+
+        if (key && *conf->vol_uuid) {
+                if ((match_uuid_local (key, conf->vol_uuid) == 0) &&
+                    (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+                        if (DHT_IS_DIR(layout)) {
+                                cnt = layout->cnt;
+                        } else {
+                                cnt = 1;
+                        }
+                        sub_volumes = alloca ( cnt * sizeof (xlator_t *));
+                        for (i = 0; i < cnt; i++)
+                                sub_volumes[i] = layout->list[i].xlator;
+
+                        if (cluster_getmarkerattr (frame, this, loc, key,
+                                                   local, dht_getxattr_unwind,
+                                                   sub_volumes, cnt,
+                                                   MARKER_XTIME_TYPE,
+                                                   marker_xtime_default_gauge,
+                                                   conf->vol_uuid)) {
+                                op_errno = EINVAL;
                                 goto err;
                         }
-                        local->layout = layout;
-                        //layout = dht_layout_new (this, conf->subvolume_cnt);
 
-                        dht_selfheal_new_directory (frame, dht_fix_layout_cbk,
-                                                    layout);
                         return 0;
                 }
-                op_errno = ENODATA;
-                goto err;
         }
 
-        local = dht_local_init (frame);
+        if (DHT_IS_DIR(layout)) {
+                cnt = local->call_cnt = layout->cnt;
+        } else {
+                cnt = local->call_cnt  = 1;
+        }
+
+        for (i = 0; i < cnt; i++) {
+                subvol = layout->list[i].xlator;
+                STACK_WIND (frame, dht_getxattr_cbk,
+                            subvol, subvol->fops->getxattr,
+                            loc, key, NULL);
+        }
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+#undef DHT_IS_DIR
+
+int
+dht_fgetxattr (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, const char *key, dict_t *xdata)
+{
+        xlator_t     *subvol        = NULL;
+        dht_local_t  *local         = NULL;
+        dht_layout_t *layout        = NULL;
+        int           op_errno      = -1;
+        int           i             = 0;
+        int           cnt           = 0;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (fd->inode, err);
+        VALIDATE_OR_GOTO (this->private, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FGETXATTR);
         if (!local) {
                 op_errno = ENOMEM;
-                gf_log (this->name, GF_LOG_ERROR,
-                       "Out of memory");
+
                 goto err;
         }
 
-        ret = loc_dup (loc, &local->loc);
-        if (ret == -1) {
-                op_errno = ENOMEM;
+        layout = local->layout;
+        if (!layout) {
                 gf_log (this->name, GF_LOG_ERROR,
-                       "Out of memory");
+                        "layout is NULL");
+                op_errno = ENOENT;
                 goto err;
         }
 
@@ -1980,14 +2347,14 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
                 local->key = gf_strdup (key);
                 if (!local->key) {
                         op_errno = ENOMEM;
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
                         goto err;
                 }
-        } 
-        local->layout = layout;
+        }
 
-        if (loc->inode-> ia_type == IA_IFDIR) {
+        if ((fd->inode->ia_type == IA_IFDIR)
+	    && key
+            && (strncmp (key, GF_XATTR_LOCKINFO_KEY,
+                         strlen (GF_XATTR_LOCKINFO_KEY) != 0))) {
                 cnt = local->call_cnt = layout->cnt;
         } else {
                 cnt = local->call_cnt  = 1;
@@ -1995,489 +2362,492 @@ dht_getxattr (call_frame_t *frame, xlator_t *this,
 
         for (i = 0; i < cnt; i++) {
                 subvol = layout->list[i].xlator;
-	        STACK_WIND (frame, dht_getxattr_cbk,
-		            subvol, subvol->fops->getxattr,
-		            loc, key);
+                STACK_WIND (frame, dht_getxattr_cbk,
+                            subvol, subvol->fops->fgetxattr,
+                            fd, key, NULL);
         }
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fgetxattr, frame, -1, op_errno, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
-
 int
-dht_setxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, dict_t *xattr, int flags)
+dht_fsetxattr (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, dict_t *xattr, int flags, dict_t *xdata)
 {
-	xlator_t     *subvol   = NULL;
-	dht_local_t  *local    = NULL;
-        dht_conf_t   *conf     = NULL;
-        dht_layout_t *layout   = NULL;
-        int           i        = 0;
+        xlator_t     *subvol   = NULL;
+        dht_local_t  *local    = NULL;
         int           op_errno = EINVAL;
+        dht_conf_t   *conf     = NULL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (fd->inode, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-        conf   = this->private;
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        conf = this->private;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        GF_IF_INTERNAL_XATTR_GOTO (conf->wild_xattr_name, xattr,
+                                   op_errno, err);
 
-	local->layout = layout = dht_layout_get (this, loc->inode);
-	if (!layout) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no layout for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FSETXATTR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	local->call_cnt = layout->cnt;
+        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;
+        }
 
-	for (i = 0; i < layout->cnt; i++) {
-		STACK_WIND (frame, dht_err_cbk,
-			    layout->list[i].xlator,
-			    layout->list[i].xlator->fops->setxattr,
-			    loc, xattr, flags);
-	}
+        local->call_cnt = 1;
 
-	return 0;
+        STACK_WIND (frame, dht_err_cbk, subvol, subvol->fops->fsetxattr,
+                    fd, xattr, flags, NULL);
+
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (setxattr, frame, -1, op_errno);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fsetxattr, frame, -1, op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
 
-int
-dht_removexattr (call_frame_t *frame, xlator_t *this,
-		 loc_t *loc, const char *key)
+static int
+dht_common_setxattr_cbk (call_frame_t *frame, void *cookie,
+                         xlator_t *this, int32_t op_ret, int32_t op_errno,
+                         dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
-
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	local->call_cnt = 1;
-
-	STACK_WIND (frame, dht_err_cbk,
-		    subvol, subvol->fops->removexattr,
-		    loc, key);
+        DHT_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
 
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (removexattr, frame, -1, op_errno);
-
-	return 0;
+        return 0;
 }
 
-
 int
-dht_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	      int op_ret, int op_errno, fd_t *fd)
+dht_checking_pathinfo_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                           int op_ret, int op_errno, dict_t *xattr,
+                           dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
+        int           i     = -1;
+        int           ret   = -1;
+        char         *value = NULL;
+        dht_local_t  *local = NULL;
+        dht_conf_t   *conf  = NULL;
+        call_frame_t *prev  = NULL;
+        int           this_call_cnt = 0;
 
+        local = frame->local;
+        prev = cookie;
+        conf = this->private;
 
-	local = frame->local;
-	prev = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
+        if (op_ret == -1)
+                goto out;
 
-		local->op_ret = 0;
-	}
-unlock:
-	UNLOCK (&frame->lock);
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
-				  local->fd);
+        ret = dict_get_str (xattr, GF_XATTR_PATHINFO_KEY, &value);
+        if (ret)
+                goto out;
+
+        if (!strcmp (value, local->key)) {
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (conf->subvolumes[i] == prev->this)
+                                conf->decommissioned_bricks[i] = prev->this;
+                }
+        }
 
+out:
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                DHT_STACK_UNWIND (setxattr, frame, local->op_ret, ENOTSUP, NULL);
+        }
         return 0;
-}
 
+}
 
 int
-dht_open (call_frame_t *frame, xlator_t *this,
-	  loc_t *loc, int flags, fd_t *fd, int wbflags)
+dht_setxattr (call_frame_t *frame, xlator_t *this,
+              loc_t *loc, dict_t *xattr, int flags, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
-	int           ret = -1;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
+        xlator_t     *subvol   = NULL;
+        dht_local_t  *local    = NULL;
+        dht_conf_t   *conf     = NULL;
+        dht_layout_t *layout   = NULL;
+        int           i        = 0;
+        int           op_errno = EINVAL;
+        int           ret      = -1;
+        data_t       *tmp      = NULL;
+        uint32_t      dir_spread = 0;
+        char          value[4096] = {0,};
+        gf_dht_migrate_data_type_t forced_rebalance = GF_DHT_MIGRATE_DATA;
+        int           call_cnt = 0;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+        conf   = this->private;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        GF_IF_INTERNAL_XATTR_GOTO (conf->wild_xattr_name, xattr,
+                                   op_errno, err);
 
-	local->fd = fd_ref (fd);
-	ret = loc_dup (loc, &local->loc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        local = dht_local_init (frame, loc, NULL, GF_FOP_SETXATTR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	local->call_cnt = 1;
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	STACK_WIND (frame, dht_fd_cbk,
-		    subvol, subvol->fops->open,
-		    loc, flags, fd, wbflags);
+        layout = local->layout;
+        if (!layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no layout for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	return 0;
+        local->call_cnt = call_cnt = layout->cnt;
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (open, frame, -1, op_errno, NULL);
+        tmp = dict_get (xattr, "distribute.migrate-data");
+        if (tmp) {
+                if (IA_ISDIR (loc->inode->ia_type)) {
+                        op_errno = ENOTSUP;
+                        goto err;
+                }
 
-	return 0;
-}
+                /* TODO: need to interpret the 'value' for more meaning
+                   (ie, 'target' subvolume given there, etc) */
+                memcpy (value, tmp->data, tmp->len);
+                if (strcmp (value, "force") == 0)
+                       forced_rebalance =
+                                GF_DHT_MIGRATE_DATA_EVEN_IF_LINK_EXISTS;
 
+                if (conf->decommission_in_progress)
+                        forced_rebalance = GF_DHT_MIGRATE_HARDLINK;
 
-int
-dht_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	       int op_ret, int op_errno,
-	       struct iovec *vector, int count, struct iatt *stbuf,
-               struct iobref *iobref)
-{
-        dht_local_t     *local = frame->local;
+                local->rebalance.target_node = dht_subvol_get_hashed (this, loc);
+                if (!local->rebalance.target_node) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+                                "hashed subvol for %s", loc->path);
+                        op_errno = EINVAL;
+                        goto err;
+                }
 
-        if (!local) {
-                op_ret = -1;
+                local->rebalance.from_subvol = local->cached_subvol;
+
+                if (local->rebalance.target_node == local->rebalance.from_subvol) {
+                        op_errno = EEXIST;
+                        goto err;
+                }
+                if (local->rebalance.target_node) {
+                        local->flags = forced_rebalance;
+
+                        ret = dht_start_rebalance_task (this, frame);
+                        if (!ret)
+                                return 0;
+
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "%s: failed to create a new synctask",
+                                loc->path);
+                }
                 op_errno = EINVAL;
-                goto out;
-        }
+                goto err;
 
-        if (op_ret != -1)
-                stbuf->ia_ino = local->ia_ino;
-out:
-        DHT_STACK_UNWIND (readv, frame, op_ret, op_errno, vector, count, stbuf,
-                          iobref);
+        }
 
-        return 0;
-}
+        tmp = dict_get (xattr, "decommission-brick");
+        if (tmp) {
+                /* This operation should happen only on '/' */
+                if (!__is_root_gfid (loc->inode->gfid)) {
+                        op_errno = ENOTSUP;
+                        goto err;
+                }
 
+                memcpy (value, tmp->data, ((tmp->len < 4095) ? tmp->len : 4095));
+                local->key = gf_strdup (value);
+                local->call_cnt = conf->subvolume_cnt;
 
-int
-dht_readv (call_frame_t *frame, xlator_t *this,
-	   fd_t *fd, size_t size, off_t off)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-        dht_local_t  *local = NULL;
+                for (i = 0 ; i < conf->subvolume_cnt; i++) {
+                        /* Get the pathinfo, and then compare */
+                        STACK_WIND (frame, dht_checking_pathinfo_cbk,
+                                    conf->subvolumes[i],
+                                    conf->subvolumes[i]->fops->getxattr,
+                                    loc, GF_XATTR_PATHINFO_KEY, NULL);
+                }
+                return 0;
+        }
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        tmp = dict_get (xattr, GF_XATTR_FIX_LAYOUT_KEY);
+        if (tmp) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "fixing the layout of %s", loc->path);
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+                ret = dht_fix_directory_layout (frame, dht_common_setxattr_cbk,
+                                                layout);
+                if (ret) {
+                        op_errno = ENOTCONN;
+                        goto err;
+                }
+                return ret;
+        }
 
-        local = dht_local_init (frame);
-        if (!local) {
-                gf_log (this->name, GF_LOG_ERROR, "Out of memory");
-                op_errno = ENOMEM;
+        tmp = dict_get (xattr, "distribute.directory-spread-count");
+        if (tmp) {
+                /* Setxattr value is packed as 'binary', not string */
+                memcpy (value, tmp->data, ((tmp->len < 4095)?tmp->len:4095));
+                ret = gf_string2uint32 (value, &dir_spread);
+                if (!ret && ((dir_spread <= conf->subvolume_cnt) &&
+                             (dir_spread > 0))) {
+                        layout->spread_cnt = dir_spread;
+
+                        ret = dht_fix_directory_layout (frame,
+                                                        dht_common_setxattr_cbk,
+                                                        layout);
+                        if (ret) {
+                                op_errno = ENOTCONN;
+                                goto err;
+                        }
+                        return ret;
+                }
+                gf_log (this->name, GF_LOG_ERROR,
+                        "wrong 'directory-spread-count' value (%s)", value);
+                op_errno = ENOTSUP;
                 goto err;
         }
 
-        local->ia_ino = fd->inode->ino;
-	STACK_WIND (frame, dht_readv_cbk,
-		    subvol, subvol->fops->readv,
-		    fd, size, off);
+        for (i = 0; i < call_cnt; i++) {
+                STACK_WIND (frame, dht_err_cbk,
+                            layout->list[i].xlator,
+                            layout->list[i].xlator->fops->setxattr,
+                            loc, xattr, flags, xdata);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
-dht_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		int op_ret, int op_errno, struct iatt *prebuf,
-                struct iatt *postbuf)
+dht_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int op_ret, int op_errno, dict_t *xdata)
 {
-        dht_local_t *local = NULL;
-
-        if (op_ret == -1) {
-                goto out;
-        }
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
 
         local = frame->local;
-        if (!local) {
-                op_ret = -1;
-                op_errno = EINVAL;
-                goto out;
-        } 
-        
-        prebuf->ia_ino = local->ia_ino;
-        postbuf->ia_ino = local->ia_ino;
+        prev = cookie;
 
-out:
-        DHT_STACK_UNWIND (writev, frame, op_ret, op_errno, prebuf, postbuf);
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "subvolume %s returned -1 (%s)",
+                                prev->this->name, strerror (op_errno));
+                        goto unlock;
+                }
+
+                local->op_ret = 0;
+        }
+unlock:
+        UNLOCK (&frame->lock);
+
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                DHT_STACK_UNWIND (removexattr, frame, local->op_ret,
+                                  local->op_errno, NULL);
+        }
 
         return 0;
 }
 
 
 int
-dht_writev (call_frame_t *frame, xlator_t *this,
-	    fd_t *fd, struct iovec *vector, int count, off_t off,
-            struct iobref *iobref)
+dht_removexattr (call_frame_t *frame, xlator_t *this,
+                 loc_t *loc, const char *key, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
+        xlator_t     *subvol = NULL;
         int           op_errno = -1;
         dht_local_t  *local = NULL;
+        dht_layout_t *layout = NULL;
+        int           call_cnt = 0;
+        dht_conf_t   *conf = NULL;
+
+        int i;
 
-        VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+        conf = this->private;
+
+        GF_IF_NATIVE_XATTR_GOTO (conf->wild_xattr_name, key, op_errno, err);
 
-        local = dht_local_init (frame);
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+
+        local = dht_local_init (frame, loc, NULL, GF_FOP_REMOVEXATTR);
         if (!local) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
                 op_errno = ENOMEM;
                 goto err;
         }
 
-        local->ia_ino = fd->inode->ino;
-
-	STACK_WIND (frame, dht_writev_cbk,
-		    subvol, subvol->fops->writev,
-		    fd, vector, count, off, iobref);
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL);
-
-	return 0;
-}
-
-
-int
-dht_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
-
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        layout = local->layout;
+        if (!local->layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no layout for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	local->fd = fd_ref (fd);
-	local->call_cnt = 1;
+        local->call_cnt = call_cnt = layout->cnt;
+        local->key = gf_strdup (key);
 
-	STACK_WIND (frame, dht_err_cbk,
-		    subvol, subvol->fops->flush, fd);
+        for (i = 0; i < call_cnt; i++) {
+                STACK_WIND (frame, dht_removexattr_cbk,
+                            layout->list[i].xlator,
+                            layout->list[i].xlator->fops->removexattr,
+                            loc, key, NULL);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (flush, frame, -1, op_errno);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
-
 int
-dht_fsync (call_frame_t *frame, xlator_t *this,
-	   fd_t *fd, int datasync)
+dht_fremovexattr (call_frame_t *frame, xlator_t *this,
+                  fd_t *fd, const char *key, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
+        xlator_t     *subvol = NULL;
         int           op_errno = -1;
-	dht_local_t  *local = NULL;
+        dht_local_t  *local = NULL;
+        dht_layout_t *layout = NULL;
+        int           call_cnt = 0;
+        dht_conf_t   *conf = 0;
 
+        int i;
 
-        VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+        conf = this->private;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-	local->call_cnt = 1;
+        GF_IF_NATIVE_XATTR_GOTO (conf->wild_xattr_name, key, op_errno, err);
 
-        local->ia_ino = fd->inode->ino;
+        VALIDATE_OR_GOTO (frame, err);
 
-	STACK_WIND (frame, dht_fsync_cbk,
-		    subvol, subvol->fops->fsync,
-		    fd, datasync);
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FREMOVEXATTR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	return 0;
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for inode=%s",
+                        uuid_utoa (fd->inode->gfid));
+                op_errno = EINVAL;
+                goto err;
+        }
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL);
+        layout = local->layout;
+        if (!local->layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no layout for inode=%s", uuid_utoa (fd->inode->gfid));
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	return 0;
-}
+        local->call_cnt = call_cnt = layout->cnt;
+        local->key = gf_strdup (key);
+
+        for (i = 0; i < call_cnt; i++) {
+                STACK_WIND (frame, dht_removexattr_cbk,
+                            layout->list[i].xlator,
+                            layout->list[i].xlator->fops->fremovexattr,
+                            fd, key, NULL);
+        }
 
+        return 0;
 
-int
-dht_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	    int op_ret, int op_errno, struct gf_flock *flock)
-{
-        DHT_STACK_UNWIND (lk, frame, op_ret, op_errno, flock);
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fremovexattr, frame, -1, op_errno, NULL);
 
         return 0;
 }
 
 
 int
-dht_lk (call_frame_t *frame, xlator_t *this,
-	fd_t *fd, int cmd, struct gf_flock *flock)
+dht_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+            int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+        local = frame->local;
+        prev = cookie;
 
-	STACK_WIND (frame, dht_lk_cbk,
-		    subvol, subvol->fops->lk,
-		    fd, cmd, flock);
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "subvolume %s returned -1 (%s)",
+                                prev->this->name, strerror (op_errno));
+                        goto unlock;
+                }
 
-	return 0;
+                local->op_ret = 0;
+        }
+unlock:
+        UNLOCK (&frame->lock);
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (lk, frame, -1, op_errno, NULL);
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt))
+                DHT_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
+                                  local->fd, NULL);
 
-	return 0;
+        return 0;
 }
 
 /*
@@ -2505,23 +2875,42 @@ dht_normalize_stats (struct statvfs *buf, unsigned long bsize,
 
 int
 dht_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		int op_ret, int op_errno, struct statvfs *statvfs)
+                int op_ret, int op_errno, struct statvfs *statvfs, dict_t *xdata)
 {
-	dht_local_t *local         = NULL;
-	int          this_call_cnt = 0;
-        int          bsize         = 0;
-        int          frsize        = 0;
+        dht_local_t *local              = NULL;
+        int          this_call_cnt      = 0;
+        int          bsize              = 0;
+        int          frsize             = 0;
+        int8_t       quota_deem_statfs  = 0;
+        GF_UNUSED int     ret           = 0;
+        unsigned long     new_usage     = 0;
+        unsigned long     cur_usage     = 0;
 
+        ret = dict_get_int8 (xdata, "quota-deem-statfs", &quota_deem_statfs);
 
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			goto unlock;
-		}
-		local->op_ret = 0;
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        goto unlock;
+                }
+                if (!statvfs) {
+                        op_errno = EINVAL;
+                        local->op_ret = -1;
+                        goto unlock;
+                }
+                local->op_ret = 0;
+
+                if (quota_deem_statfs) {
+                        new_usage = statvfs->f_blocks - statvfs->f_bfree;
+                        cur_usage = local->statvfs.f_blocks - local->statvfs.f_bfree;
+                        /* We take the maximux of the usage from the subvols */
+                        if (new_usage >= cur_usage)
+                                local->statvfs = *statvfs;
+                        goto unlock;
+                }
 
                 if (local->statvfs.f_bsize != 0) {
                         bsize = max(local->statvfs.f_bsize, statvfs->f_bsize);
@@ -2533,162 +2922,176 @@ dht_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->statvfs.f_frsize   = statvfs->f_frsize;
                 }
 
-		local->statvfs.f_blocks  += statvfs->f_blocks;
-		local->statvfs.f_bfree   += statvfs->f_bfree;
-		local->statvfs.f_bavail  += statvfs->f_bavail;
-		local->statvfs.f_files   += statvfs->f_files;
-		local->statvfs.f_ffree   += statvfs->f_ffree;
-		local->statvfs.f_favail  += statvfs->f_favail;
-		local->statvfs.f_fsid     = statvfs->f_fsid;
-		local->statvfs.f_flag     = statvfs->f_flag;
-		local->statvfs.f_namemax  = statvfs->f_namemax;
+                local->statvfs.f_blocks  += statvfs->f_blocks;
+                local->statvfs.f_bfree   += statvfs->f_bfree;
+                local->statvfs.f_bavail  += statvfs->f_bavail;
+                local->statvfs.f_files   += statvfs->f_files;
+                local->statvfs.f_ffree   += statvfs->f_ffree;
+                local->statvfs.f_favail  += statvfs->f_favail;
+                local->statvfs.f_fsid     = statvfs->f_fsid;
+                local->statvfs.f_flag     = statvfs->f_flag;
+                local->statvfs.f_namemax  = statvfs->f_namemax;
 
-	}
+
+        }
 unlock:
-	UNLOCK (&frame->lock);
+        UNLOCK (&frame->lock);
 
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno,
-				  &local->statvfs);
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt))
+                DHT_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno,
+                                  &local->statvfs, xdata);
 
         return 0;
 }
 
 
 int
-dht_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
+dht_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
-	dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
+        xlator_t     *subvol = NULL;
+        dht_local_t  *local  = NULL;
+        dht_conf_t   *conf = NULL;
         int           op_errno = -1;
-	int           i = -1;
-
+        int           i = -1;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (loc, err);
         VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
         VALIDATE_OR_GOTO (this->private, err);
 
-	conf = this->private;
+        conf = this->private;
 
-	local = dht_local_init (frame);
-	local->call_cnt = conf->subvolume_cnt;
+        local = dht_local_init (frame, NULL, NULL, GF_FOP_STATFS);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_statfs_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->statfs, loc);
-	}
+        if (IA_ISDIR (loc->inode->ia_type)) {
+                local->call_cnt = conf->subvolume_cnt;
 
-	return 0;
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        STACK_WIND (frame, dht_statfs_cbk,
+                                    conf->subvolumes[i],
+                                    conf->subvolumes[i]->fops->statfs, loc,
+                                    xdata);
+                }
+                return 0;
+        }
+
+        subvol = dht_subvol_get_cached (this, loc->inode);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        local->call_cnt = 1;
+
+        STACK_WIND (frame, dht_statfs_cbk,
+                    subvol, subvol->fops->statfs, loc, xdata);
+
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (statfs, frame, -1, op_errno, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
-dht_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
+dht_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+             dict_t *xdata)
 {
-	dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
-	int           ret = -1;
+        dht_local_t  *local  = NULL;
+        dht_conf_t   *conf = NULL;
         int           op_errno = -1;
-	int           i = -1;
-
+        int           i = -1;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (fd, err);
         VALIDATE_OR_GOTO (this->private, err);
 
-	conf = this->private;
+        conf = this->private;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        local = dht_local_init (frame, loc, fd, GF_FOP_OPENDIR);
+        if (!local) {
+                op_errno = ENOMEM;
 
-	local->fd = fd_ref (fd);
-	ret = loc_dup (loc, &local->loc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+                goto err;
+        }
 
-	local->call_cnt = conf->subvolume_cnt;
+        local->call_cnt = conf->subvolume_cnt;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_fd_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->opendir,
-			    loc, fd);
-	}
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                STACK_WIND (frame, dht_fd_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->opendir,
+                            loc, fd, xdata);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (opendir, frame, -1, op_errno, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (opendir, frame, -1, op_errno, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
-                  int op_errno, gf_dirent_t *orig_entries)
+                  int op_errno, gf_dirent_t *orig_entries, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	gf_dirent_t   entries;
-	gf_dirent_t  *orig_entry = NULL;
-	gf_dirent_t  *entry = NULL;
-	call_frame_t *prev = NULL;
-	xlator_t     *next_subvol = NULL;
+        dht_local_t  *local = NULL;
+        gf_dirent_t   entries;
+        gf_dirent_t  *orig_entry = NULL;
+        gf_dirent_t  *entry = NULL;
+        call_frame_t *prev = NULL;
+        xlator_t     *next_subvol = NULL;
         off_t         next_offset = 0;
-	int           count = 0;
+        int           count = 0;
         dht_layout_t *layout = 0;
         dht_conf_t   *conf   = NULL;
         xlator_t     *subvol = 0;
+        int           ret    = 0;
 
-	INIT_LIST_HEAD (&entries.list);
-	prev = cookie;
-	local = frame->local;
-	conf  = this->private;
+        INIT_LIST_HEAD (&entries.list);
+        prev = cookie;
+        local = frame->local;
+        conf  = this->private;
 
-	if (op_ret < 0)
-		goto done;
+        if (op_ret < 0)
+                goto done;
 
         if (!local->layout)
                 local->layout = dht_layout_get (this, local->fd->inode);
 
         layout = local->layout;
 
-	list_for_each_entry (orig_entry, (&orig_entries->list), list) {
+        list_for_each_entry (orig_entry, (&orig_entries->list), list) {
                 next_offset = orig_entry->d_off;
-
-                if (check_is_linkfile (NULL, (&orig_entry->d_stat), NULL)
-                    || (check_is_dir (NULL, (&orig_entry->d_stat), NULL)
-                        && (prev->this != dht_first_up_subvol (this)))) {
+                if (check_is_dir (NULL, (&orig_entry->d_stat), NULL) &&
+                    (prev->this != local->first_up_subvol)) {
+                        continue;
+                }
+                if (check_is_linkfile (NULL, (&orig_entry->d_stat),
+                                       orig_entry->dict,
+                                       conf->link_xattr_name)) {
                         continue;
                 }
 
                 entry = gf_dirent_for_name (orig_entry->d_name);
                 if (!entry) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
+
                         goto unwind;
                 }
 
@@ -2698,25 +3101,41 @@ dht_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
                                                     orig_entry->d_name);
                         if (!subvol || (subvol != prev->this)) {
                                 /* TODO: Count the number of entries which need
-                                   linkfile to prove its existance in fs */
+                                   linkfile to prove its existence in fs */
                                 layout->search_unhashed++;
                         }
                 }
-                entry->d_stat = orig_entry->d_stat;
 
-                dht_itransform (this, prev->this, orig_entry->d_ino,
-                                &entry->d_ino);
                 dht_itransform (this, prev->this, orig_entry->d_off,
                                 &entry->d_off);
 
-                entry->d_stat.ia_ino = entry->d_ino;
+                entry->d_stat = orig_entry->d_stat;
+                entry->d_ino  = orig_entry->d_ino;
                 entry->d_type = orig_entry->d_type;
                 entry->d_len  = orig_entry->d_len;
 
+                if (orig_entry->dict)
+                        entry->dict = dict_ref (orig_entry->dict);
+
+                /* making sure we set the inode ctx right with layout,
+                   currently possible only for non-directories, so for
+                   directories don't set entry inodes */
+                if (!IA_ISDIR(entry->d_stat.ia_type) && orig_entry->inode) {
+                        ret = dht_layout_preset (this, prev->this,
+                                                 orig_entry->inode);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "failed to link the layout in inode");
+                        entry->inode = inode_ref (orig_entry->inode);
+                } else if (orig_entry->inode) {
+                        dht_inode_ctx_time_update (orig_entry->inode, this,
+                                                   &entry->d_stat, 1);
+                }
+
                 list_add_tail (&entry->list, &entries.list);
                 count++;
-	}
-	op_ret = count;
+        }
+        op_ret = count;
         /* We need to ensure that only the last subvolume's end-of-directory
          * notification is respected so that directory reading does not stop
          * before all subvolumes have been read. That could happen because the
@@ -2728,7 +3147,7 @@ dht_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
                 op_errno = 0;
 
 done:
-	if (count == 0) {
+        if (count == 0) {
                 /* non-zero next_offset means that
                    EOF is not yet hit on the current subvol
                 */
@@ -2738,23 +3157,37 @@ done:
                         next_subvol = prev->this;
                 }
 
-		if (!next_subvol) {
-			goto unwind;
-		}
+                if (!next_subvol) {
+                        goto unwind;
+                }
 
-		STACK_WIND (frame, dht_readdirp_cbk,
-			    next_subvol, next_subvol->fops->readdirp,
-			    local->fd, local->size, next_offset);
-		return 0;
-	}
+		if (conf->readdir_optimize == _gf_true) {
+                        if (next_subvol != local->first_up_subvol) {
+                                ret = dict_set_int32 (local->xattr,
+                                                      GF_READDIR_SKIP_DIRS, 1);
+                                if (ret)
+                                        gf_log (this->name, GF_LOG_ERROR,
+					         "dict set failed");
+		        } else {
+                                 dict_del (local->xattr,
+                                           GF_READDIR_SKIP_DIRS);
+                        }
+                }
+
+                STACK_WIND (frame, dht_readdirp_cbk,
+                            next_subvol, next_subvol->fops->readdirp,
+                            local->fd, local->size, next_offset,
+                            local->xattr);
+                return 0;
+        }
 
 unwind:
-	if (op_ret < 0)
-		op_ret = 0;
+        if (op_ret < 0)
+                op_ret = 0;
 
-	DHT_STACK_UNWIND (readdirp, frame, op_ret, op_errno, &entries);
+        DHT_STACK_UNWIND (readdirp, frame, op_ret, op_errno, &entries, NULL);
 
-	gf_dirent_free (&entries);
+        gf_dirent_free (&entries);
 
         return 0;
 }
@@ -2763,34 +3196,33 @@ unwind:
 
 int
 dht_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		 int op_ret, int op_errno, gf_dirent_t *orig_entries)
+                 int op_ret, int op_errno, gf_dirent_t *orig_entries,
+                 dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	gf_dirent_t   entries;
-	gf_dirent_t  *orig_entry = NULL;
-	gf_dirent_t  *entry = NULL;
-	call_frame_t *prev = NULL;
-	xlator_t     *next_subvol = NULL;
+        dht_local_t  *local = NULL;
+        gf_dirent_t   entries;
+        gf_dirent_t  *orig_entry = NULL;
+        gf_dirent_t  *entry = NULL;
+        call_frame_t *prev = NULL;
+        xlator_t     *next_subvol = NULL;
         off_t         next_offset = 0;
-	int           count = 0;
+        int           count = 0;
         dht_layout_t *layout = 0;
-        dht_conf_t   *conf   = NULL;
         xlator_t     *subvol = 0;
 
-	INIT_LIST_HEAD (&entries.list);
-	prev = cookie;
-	local = frame->local;
-	conf  = this->private;
+        INIT_LIST_HEAD (&entries.list);
+        prev = cookie;
+        local = frame->local;
 
-	if (op_ret < 0)
-		goto done;
+        if (op_ret < 0)
+                goto done;
 
         if (!local->layout)
                 local->layout = dht_layout_get (this, local->fd->inode);
 
         layout = local->layout;
 
-	list_for_each_entry (orig_entry, (&orig_entries->list), list) {
+        list_for_each_entry (orig_entry, (&orig_entries->list), list) {
                 next_offset = orig_entry->d_off;
 
                 subvol = dht_layout_search (this, layout, orig_entry->d_name);
@@ -2803,11 +3235,10 @@ dht_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                 goto unwind;
                         }
 
-                        dht_itransform (this, prev->this, orig_entry->d_ino,
-                                        &entry->d_ino);
                         dht_itransform (this, prev->this, orig_entry->d_off,
                                         &entry->d_off);
 
+                        entry->d_ino  = orig_entry->d_ino;
                         entry->d_type = orig_entry->d_type;
                         entry->d_len  = orig_entry->d_len;
 
@@ -2815,7 +3246,7 @@ dht_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         count++;
                 }
         }
-	op_ret = count;
+        op_ret = count;
         /* We need to ensure that only the last subvolume's end-of-directory
          * notification is respected so that directory reading does not stop
          * before all subvolumes have been read. That could happen because the
@@ -2827,7 +3258,7 @@ dht_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 op_errno = 0;
 
 done:
-	if (count == 0) {
+        if (count == 0) {
                 /* non-zero next_offset means that
                    EOF is not yet hit on the current subvol
                 */
@@ -2837,23 +3268,23 @@ done:
                         next_subvol = prev->this;
                 }
 
-		if (!next_subvol) {
-			goto unwind;
-		}
+                if (!next_subvol) {
+                        goto unwind;
+                }
 
-		STACK_WIND (frame, dht_readdir_cbk,
-			    next_subvol, next_subvol->fops->readdir,
-			    local->fd, local->size, next_offset);
-		return 0;
-	}
+                STACK_WIND (frame, dht_readdir_cbk,
+                            next_subvol, next_subvol->fops->readdir,
+                            local->fd, local->size, next_offset, NULL);
+                return 0;
+        }
 
 unwind:
-	if (op_ret < 0)
-		op_ret = 0;
+        if (op_ret < 0)
+                op_ret = 0;
 
-	DHT_STACK_UNWIND (readdir, frame, op_ret, op_errno, &entries);
+        DHT_STACK_UNWIND (readdir, frame, op_ret, op_errno, &entries, NULL);
 
-	gf_dirent_free (&entries);
+        gf_dirent_free (&entries);
 
         return 0;
 }
@@ -2861,55 +3292,84 @@ unwind:
 
 int
 dht_do_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-                off_t yoff, int whichop)
+                off_t yoff, int whichop, dict_t *dict)
 {
-	dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
+        dht_local_t  *local  = NULL;
         int           op_errno = -1;
-	xlator_t     *xvol = NULL;
-	off_t         xoff = 0;
-
+        xlator_t     *xvol = NULL;
+        off_t         xoff = 0;
+        int           ret = 0;
+        dht_conf_t   *conf = NULL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	conf = this->private;
+        conf = this->private;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
+        local = dht_local_init (frame, NULL, NULL, whichop);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	local->fd = fd_ref (fd);
-	local->size = size;
+        local->fd = fd_ref (fd);
+        local->size = size;
+        local->xattr_req = (dict)? dict_ref (dict) : NULL;
+        local->first_up_subvol = dht_first_up_subvol (this);
 
-	dht_deitransform (this, yoff, &xvol, (uint64_t *)&xoff);
+        dht_deitransform (this, yoff, &xvol, (uint64_t *)&xoff);
+
+        /* TODO: do proper readdir */
+        if (whichop == GF_FOP_READDIRP) {
+                if (dict)
+                        local->xattr = dict_ref (dict);
+                else
+                        local->xattr = dict_new ();
+
+                if (local->xattr) {
+                        ret = dict_set_uint32 (local->xattr,
+                                               conf->link_xattr_name, 256);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "failed to set '%s' key",
+                                        conf->link_xattr_name);
+			if (conf->readdir_optimize == _gf_true) {
+                                if (xvol != local->first_up_subvol) {
+				        ret = dict_set_int32 (local->xattr,
+			                               GF_READDIR_SKIP_DIRS, 1);
+				        if (ret)
+					        gf_log (this->name,
+                                                        GF_LOG_ERROR,
+						        "Dict set failed");
+                                } else {
+                                        dict_del (local->xattr,
+                                                  GF_READDIR_SKIP_DIRS);
+                                }
+			}
+                }
 
-	/* TODO: do proper readdir */
-        if (whichop == GF_FOP_READDIR)
-                STACK_WIND (frame, dht_readdir_cbk, xvol, xvol->fops->readdir,
-                            fd, size, xoff);
-        else
                 STACK_WIND (frame, dht_readdirp_cbk, xvol, xvol->fops->readdirp,
-                            fd, size, xoff);
+                            fd, size, xoff, local->xattr);
+        } else {
+                STACK_WIND (frame, dht_readdir_cbk, xvol, xvol->fops->readdir,
+                            fd, size, xoff, local->xattr);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (readdir, frame, -1, op_errno, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-             off_t yoff)
+             off_t yoff, dict_t *xdata)
 {
         int          op = GF_FOP_READDIR;
         dht_conf_t  *conf = NULL;
@@ -2930,15 +3390,15 @@ dht_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
                 op = GF_FOP_READDIRP;
 
 out:
-        dht_do_readdir (frame, this, fd, size, yoff, op);
+        dht_do_readdir (frame, this, fd, size, yoff, op, 0);
         return 0;
 }
 
 int
 dht_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-              off_t yoff)
+              off_t yoff, dict_t *dict)
 {
-        dht_do_readdir (frame, this, fd, size, yoff, GF_FOP_READDIRP);
+        dht_do_readdir (frame, this, fd, size, yoff, GF_FOP_READDIRP, dict);
         return 0;
 }
 
@@ -2946,89 +3406,88 @@ dht_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
 
 int
 dht_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		  int op_ret, int op_errno)
+                  int op_ret, int op_errno, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
 
 
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1)
-			local->op_errno = op_errno;
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1)
+                        local->op_errno = op_errno;
 
-		if (op_ret == 0)
-			local->op_ret = 0;
-	}
-	UNLOCK (&frame->lock);
+                if (op_ret == 0)
+                        local->op_ret = 0;
+        }
+        UNLOCK (&frame->lock);
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (fsyncdir, frame, local->op_ret, local->op_errno);
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt))
+                DHT_STACK_UNWIND (fsyncdir, frame, local->op_ret,
+                                  local->op_errno, xdata);
 
         return 0;
 }
 
 
 int
-dht_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int datasync)
+dht_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
+              int datasync, dict_t *xdata)
 {
-	dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
+        dht_local_t  *local  = NULL;
+        dht_conf_t   *conf = NULL;
         int           op_errno = -1;
-	int           i = -1;
-
+        int           i = -1;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (fd, err);
         VALIDATE_OR_GOTO (this->private, err);
 
-	conf = this->private;
+        conf = this->private;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        local = dht_local_init (frame, NULL, NULL, GF_FOP_FSYNCDIR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	local->fd = fd_ref (fd);
-	local->call_cnt = conf->subvolume_cnt;
+        local->fd = fd_ref (fd);
+        local->call_cnt = conf->subvolume_cnt;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_fsyncdir_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->fsyncdir,
-			    fd, datasync);
-	}
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                STACK_WIND (frame, dht_fsyncdir_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->fsyncdir,
+                            fd, datasync, xdata);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (fsyncdir, frame, -1, op_errno);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fsyncdir, frame, -1, op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_newfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		 int op_ret, int op_errno,
+                 int op_ret, int op_errno,
                  inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
-                 struct iatt *postparent)
+                 struct iatt *postparent, dict_t *xdata)
 {
-	call_frame_t *prev = NULL;
-	int           ret = -1;
+        xlator_t     *prev = NULL;
+        int           ret = -1;
         dht_local_t  *local = NULL;
 
 
-	if (op_ret == -1)
-		goto out;
+        if (op_ret == -1)
+                goto out;
 
         local = frame->local;
         if (!local) {
@@ -3037,38 +3496,39 @@ dht_newfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 goto out;
         }
 
-	prev = cookie;
+        prev = cookie;
 
-	dht_itransform (this, prev->this, stbuf->ia_ino, &stbuf->ia_ino);
         if (local->loc.parent) {
-                preparent->ia_ino = local->loc.parent->ino;
-                postparent->ia_ino = local->loc.parent->ino;
 
-                WIPE (preparent);
-                WIPE (postparent);
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           preparent, 0);
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           postparent, 1);
         }
 
-        ret = dht_layout_preset (this, prev->this, inode);
-	if (ret < 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"could not set pre-set layout for subvolume %s",
-			prev->this->name);
-		op_ret   = -1;
-		op_errno = EINVAL;
-		goto out;
-	}
+        ret = dht_layout_preset (this, prev, inode);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "could not set pre-set layout for subvolume %s",
+                        prev? prev->name: NULL);
+                op_ret   = -1;
+                op_errno = EINVAL;
+                goto out;
+        }
+        if (local->linked == _gf_true)
+                dht_linkfile_attr_heal (frame, this);
 out:
-        /* 
-         * FIXME: ia_size and st_blocks of preparent and postparent do not have 
+        /*
+         * FIXME: ia_size and st_blocks of preparent and postparent do not have
          * correct values. since, preparent and postparent buffers correspond
          * to a directory these two members should have values equal to sum of
          * corresponding values from each of the subvolume.
          * See dht_iatt_merge for reference.
-         */ 
-
-	DHT_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode, stbuf, preparent,
-                          postparent);
-	return 0;
+         */
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
+        DHT_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode, stbuf,
+                          preparent, postparent, xdata);
+        return 0;
 }
 
 int
@@ -3076,406 +3536,376 @@ dht_mknod_linkfile_create_cbk (call_frame_t *frame, void *cookie,
                                xlator_t *this,
                                int32_t op_ret, int32_t op_errno,
                                inode_t *inode, struct iatt *stbuf,
-                               struct iatt *preparent, struct iatt *postparent)
+                               struct iatt *preparent, struct iatt *postparent,
+                               dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	xlator_t     *cached_subvol = NULL;
+        dht_local_t  *local = NULL;
+        xlator_t     *cached_subvol = NULL;
 
         if (op_ret == -1)
                 goto err;
 
-	local = frame->local;
-	cached_subvol = local->cached_subvol;
+        local = frame->local;
+        if (!local || !local->cached_subvol) {
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        cached_subvol = local->cached_subvol;
 
-        STACK_WIND (frame, dht_newfile_cbk,
-                    cached_subvol, cached_subvol->fops->mknod,
-                    &local->loc, local->mode, local->rdev,
-                    local->params);
+        STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)cached_subvol,
+                           cached_subvol, cached_subvol->fops->mknod,
+                           &local->loc, local->mode, local->rdev, local->umask,
+                           local->params);
 
         return 0;
- err:
- 	DHT_STACK_UNWIND (mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL);	
- 	return 0;
+err:
+        DHT_STACK_UNWIND (mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+                          NULL);
+        return 0;
 }
 
 int
 dht_mknod (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, mode_t mode, dev_t rdev, dict_t *params)
+           loc_t *loc, mode_t mode, dev_t rdev, mode_t umask, dict_t *params)
 {
-	xlator_t    *subvol = NULL;
-	int          op_errno = -1;
-        int          ret = -1;
+        xlator_t    *subvol = NULL;
+        int          op_errno = -1;
         xlator_t    *avail_subvol = NULL;
-	dht_conf_t  *conf = NULL;
-	dht_local_t *local = NULL;
-
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
+        dht_local_t *local = NULL;
 
-	conf = this->private;
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
         dht_get_du_info (frame, this, loc);
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	subvol = dht_subvol_get_hashed (this, loc);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			loc->path);
-		op_errno = ENOENT;
-		goto err;
-	}
-
-        ret = loc_dup (loc, &local->loc);
-        if (ret == -1) {
+        local = dht_local_init (frame, loc, NULL, GF_FOP_MKNOD);
+        if (!local) {
                 op_errno = ENOMEM;
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
+                goto err;
+        }
+
+        subvol = dht_subvol_get_hashed (this, loc);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        loc->path);
+                op_errno = ENOENT;
                 goto err;
         }
 
         if (!dht_is_subvol_filled (this, subvol)) {
                 gf_log (this->name, GF_LOG_TRACE,
                         "creating %s on %s", loc->path, subvol->name);
-                
-                STACK_WIND (frame, dht_newfile_cbk,
-                            subvol, subvol->fops->mknod,
-                            loc, mode, rdev, params);
+
+                STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol,
+                                   subvol, subvol->fops->mknod, loc, mode,
+                                   rdev, umask, params);
         } else {
-                avail_subvol = dht_free_disk_available_subvol (this, subvol);
+
+                avail_subvol = dht_free_disk_available_subvol (this, subvol,
+                                                               local);
                 if (avail_subvol != subvol) {
-                        /* Choose the minimum filled volume, and create the 
+                        /* Choose the minimum filled volume, and create the
                            files there */
 
                         local->params = dict_ref (params);
                         local->cached_subvol = avail_subvol;
-                        local->mode = mode; 
+                        local->mode = mode;
                         local->rdev = rdev;
-                        
-                        dht_linkfile_create (frame, 
+                        local->umask = umask;
+                        dht_linkfile_create (frame,
                                              dht_mknod_linkfile_create_cbk,
-                                             avail_subvol, subvol, loc);
+                                             this, avail_subvol, subvol, loc);
                 } else {
                         gf_log (this->name, GF_LOG_TRACE,
                                 "creating %s on %s", loc->path, subvol->name);
-                        
-                        STACK_WIND (frame, dht_newfile_cbk,
-                                    subvol, subvol->fops->mknod,
-                                    loc, mode, rdev, params);
+
+                        STACK_WIND_COOKIE (frame, dht_newfile_cbk,
+                                           (void *)subvol, subvol,
+                                           subvol->fops->mknod, loc, mode,
+                                           rdev, umask, params);
                 }
         }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_symlink (call_frame_t *frame, xlator_t *this,
-	     const char *linkname, loc_t *loc, dict_t *params)
+             const char *linkname, loc_t *loc, mode_t umask, dict_t *params)
 {
-	xlator_t    *subvol = NULL;
-	int          op_errno = -1;
+        xlator_t    *subvol = NULL;
+        int          op_errno = -1;
         dht_local_t *local = NULL;
-        int          ret = -1;
-
-
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
 
-	subvol = dht_subvol_get_hashed (this, loc);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			loc->path);
-		op_errno = ENOENT;
-		goto err;
-	}
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
-        ret = loc_copy (&local->loc, loc);
-        if (ret == -1) {
-                gf_log (this->name, GF_LOG_TRACE, "Failed to copy loc");
+        local = dht_local_init (frame, loc, NULL, GF_FOP_SYMLINK);
+        if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"creating %s on %s", loc->path, subvol->name);
+        subvol = dht_subvol_get_hashed (this, loc);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        loc->path);
+                op_errno = ENOENT;
+                goto err;
+        }
 
-	STACK_WIND (frame, dht_newfile_cbk,
-		    subvol, subvol->fops->symlink,
-		    linkname, loc, params);
+        gf_log (this->name, GF_LOG_TRACE,
+                "creating %s on %s", loc->path, subvol->name);
 
-	return 0;
+        STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol, subvol,
+                           subvol->fops->symlink, linkname, loc, umask,
+                           params);
+
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (link, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (link, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
-dht_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
+dht_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+            dict_t *xdata)
 {
-	xlator_t    *cached_subvol = NULL;
-	xlator_t    *hashed_subvol = NULL;
-        int          ret = -1;
-	int          op_errno = -1;
-	dht_local_t *local = NULL;
-
+        xlator_t    *cached_subvol = NULL;
+        xlator_t    *hashed_subvol = NULL;
+        int          op_errno = -1;
+        dht_local_t *local = NULL;
 
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
         if (dht_filter_loc_subvol_key (this, loc, &local->loc,
                                        &cached_subvol)) {
-                gf_log (this->name, GF_LOG_NORMAL,
+                gf_log (this->name, GF_LOG_INFO,
                         "unlinking %s on %s (given path %s)",
                         local->loc.path, cached_subvol->name, loc->path);
                 STACK_WIND (frame, dht_unlink_cbk,
                             cached_subvol, cached_subvol->fops->unlink,
-                            &local->loc);
+                            &local->loc, xflag, xdata);
                 goto done;
         }
 
-	cached_subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!cached_subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        local = dht_local_init (frame, loc, NULL, GF_FOP_UNLINK);
+        if (!local) {
+                op_errno = ENOMEM;
 
-	hashed_subvol = dht_subvol_get_hashed (this, loc);
-	if (!hashed_subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+                goto err;
+        }
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        hashed_subvol = dht_subvol_get_hashed (this, loc);
+        /* Dont fail unlink if hashed_subvol is NULL which can be the result
+         * of layout anomaly */
+        if (!hashed_subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        loc->path);
+        }
 
-	ret = loc_copy (&local->loc, loc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        cached_subvol = local->cached_subvol;
+        if (!cached_subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	if (hashed_subvol != cached_subvol) {
-		STACK_WIND (frame, dht_unlink_linkfile_cbk,
-			    hashed_subvol, hashed_subvol->fops->unlink, loc);
+        local->flags = xflag;
+        if (hashed_subvol && hashed_subvol != cached_subvol) {
+                STACK_WIND (frame, dht_unlink_linkfile_cbk,
+                            hashed_subvol, hashed_subvol->fops->unlink, loc,
+                            xflag, xdata);
         } else {
                 STACK_WIND (frame, dht_unlink_cbk,
-                            cached_subvol, cached_subvol->fops->unlink, loc);
+                            cached_subvol, cached_subvol->fops->unlink, loc,
+                            xflag, xdata);
         }
 done:
-	return 0;
+        return 0;
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	      int op_ret, int op_errno,
+              int op_ret, int op_errno,
               inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
-              struct iatt *postparent)
+              struct iatt *postparent, dict_t *xdata)
 {
         call_frame_t *prev = NULL;
-	dht_layout_t *layout = NULL;
-	dht_local_t  *local = NULL;
+        dht_layout_t *layout = NULL;
+        dht_local_t  *local = NULL;
 
         prev = cookie;
-	local = frame->local;
+
+        local = frame->local;
 
         if (op_ret == -1)
                 goto out;
 
-	layout = dht_layout_for_subvol (this, prev->this);
-	if (!layout) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no pre-set layout for subvolume %s",
-			prev->this->name);
-		op_ret   = -1;
-		op_errno = EINVAL;
-		goto out;
-	}
-
-	stbuf->ia_ino = local->loc.inode->ino;
-
-        preparent->ia_ino = local->loc2.parent->ino;
-        postparent->ia_ino = local->loc2.parent->ino;
-
-        WIPE (preparent);
-        WIPE (postparent);
+        layout = dht_layout_for_subvol (this, prev->this);
+        if (!layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no pre-set layout for subvolume %s",
+                        prev->this->name);
+                op_ret   = -1;
+                op_errno = EINVAL;
+                goto out;
+        }
 
+        if (local->loc.parent) {
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           preparent, 0);
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           postparent, 1);
+        }
+        if (local->linked == _gf_true) {
+                local->stbuf = *stbuf;
+                dht_linkfile_attr_heal (frame, this);
+        }
 out:
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
         DHT_STACK_UNWIND (link, frame, op_ret, op_errno, inode, stbuf, preparent,
-                          postparent);
+                          postparent, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_link_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int op_ret, int op_errno,
+                       int op_ret, int op_errno,
                        inode_t *inode, struct iatt *stbuf,
-                       struct iatt *preparent, struct iatt *postparent)
+                       struct iatt *preparent, struct iatt *postparent,
+                       dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	xlator_t     *srcvol = NULL;
-
+        dht_local_t  *local = NULL;
+        xlator_t     *srcvol = NULL;
 
-	if (op_ret == -1)
-		goto err;
+        if (op_ret == -1)
+                goto err;
 
-	local = frame->local;
-	srcvol = local->linkfile.srcvol;
+        local = frame->local;
+        srcvol = local->linkfile.srcvol;
 
-	STACK_WIND (frame, dht_link_cbk,
-		    srcvol, srcvol->fops->link,
-		    &local->loc, &local->loc2);
+        STACK_WIND (frame, dht_link_cbk, srcvol, srcvol->fops->link,
+                    &local->loc, &local->loc2, xdata);
 
-	return 0;
+        return 0;
 
 err:
-	DHT_STACK_UNWIND (link, frame, op_ret, op_errno, inode, stbuf, preparent,
-                          postparent);
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
+        DHT_STACK_UNWIND (link, frame, op_ret, op_errno, inode, stbuf, preparent,
+                          postparent, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_link (call_frame_t *frame, xlator_t *this,
-	  loc_t *oldloc, loc_t *newloc)
+          loc_t *oldloc, loc_t *newloc, dict_t *xdata)
 {
-	xlator_t    *cached_subvol = NULL;
-	xlator_t    *hashed_subvol = NULL;
-	int          op_errno = -1;
-	int          ret = -1;
-	dht_local_t *local = NULL;
-
-
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (oldloc, err);
-	VALIDATE_OR_GOTO (newloc, err);
-
-	cached_subvol = dht_subvol_get_cached (this, oldloc->inode);
-	if (!cached_subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", oldloc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        xlator_t    *cached_subvol = NULL;
+        xlator_t    *hashed_subvol = NULL;
+        int          op_errno = -1;
+        int          ret = -1;
+        dht_local_t *local = NULL;
 
-	hashed_subvol = dht_subvol_get_hashed (this, newloc);
-	if (!hashed_subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			newloc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (oldloc, err);
+        VALIDATE_OR_GOTO (newloc, err);
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        local = dht_local_init (frame, oldloc, NULL, GF_FOP_LINK);
+        if (!local) {
+                op_errno = ENOMEM;
 
-	ret = loc_copy (&local->loc, oldloc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+                goto err;
+        }
 
-	ret = loc_copy (&local->loc2, newloc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        cached_subvol = local->cached_subvol;
+        if (!cached_subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", oldloc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	if (hashed_subvol != cached_subvol) {
-                memcpy (local->gfid, oldloc->inode->gfid, 16);
-		dht_linkfile_create (frame, dht_link_linkfile_cbk,
-				     cached_subvol, hashed_subvol, newloc);
-	} else {
-		STACK_WIND (frame, dht_link_cbk,
-			    cached_subvol, cached_subvol->fops->link,
-			    oldloc, newloc);
-	}
+        hashed_subvol = dht_subvol_get_hashed (this, newloc);
+        if (!hashed_subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        newloc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	return 0;
+        ret = loc_copy (&local->loc2, newloc);
+        if (ret == -1) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        if (hashed_subvol != cached_subvol) {
+                uuid_copy (local->gfid, oldloc->inode->gfid);
+                dht_linkfile_create (frame, dht_link_linkfile_cbk, this,
+                                     cached_subvol, hashed_subvol, newloc);
+        } else {
+                STACK_WIND (frame, dht_link_cbk,
+                            cached_subvol, cached_subvol->fops->link,
+                            oldloc, newloc, xdata);
+        }
+
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		 int op_ret, int op_errno,
-		 fd_t *fd, inode_t *inode, struct iatt *stbuf,
-                 struct iatt *preparent, struct iatt *postparent)
+                int op_ret, int op_errno,
+                fd_t *fd, inode_t *inode, struct iatt *stbuf,
+                struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
-	call_frame_t *prev = NULL;
-	int           ret = -1;
+        call_frame_t *prev = NULL;
+        int           ret = -1;
         dht_local_t  *local = NULL;
 
-	if (op_ret == -1)
-		goto out;
+        if (op_ret == -1)
+                goto out;
 
         local = frame->local;
         if (!local) {
@@ -3484,107 +3914,99 @@ dht_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 goto out;
         }
 
-	prev = cookie;
+        prev = cookie;
 
-	dht_itransform (this, prev->this, stbuf->ia_ino, &stbuf->ia_ino);
         if (local->loc.parent) {
-                preparent->ia_ino = local->loc.parent->ino;
-                postparent->ia_ino = local->loc.parent->ino;
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           preparent, 0);
 
-                WIPE (preparent);
-                WIPE (postparent);
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           postparent, 1);
         }
 
         ret = dht_layout_preset (this, prev->this, inode);
-	if (ret != 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"could not set preset layout for subvol %s",
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "could not set preset layout for subvol %s",
                         prev->this->name);
-		op_ret   = -1;
-		op_errno = EINVAL;
-		goto out;
-	}
-
+                op_ret   = -1;
+                op_errno = EINVAL;
+                goto out;
+        }
+        if (local->linked == _gf_true) {
+                local->stbuf = *stbuf;
+                dht_linkfile_attr_heal (frame, this);
+        }
 out:
-	DHT_STACK_UNWIND (create, frame, op_ret, op_errno, fd, inode, stbuf, preparent,
-                          postparent);
-	return 0;
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
+        DHT_STACK_UNWIND (create, frame, op_ret, op_errno, fd, inode, stbuf, preparent,
+                          postparent, NULL);
+        return 0;
 }
 
 
 int
 dht_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
-				xlator_t *this,
-				int32_t op_ret, int32_t op_errno,
+                                xlator_t *this,
+                                int32_t op_ret, int32_t op_errno,
                                 inode_t *inode, struct iatt *stbuf,
-                                struct iatt *preparent, struct iatt *postparent)
+                                struct iatt *preparent, struct iatt *postparent,
+                                dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	xlator_t     *cached_subvol = NULL;
+        dht_local_t  *local = NULL;
+        xlator_t     *cached_subvol = NULL;
 
         if (op_ret == -1)
                 goto err;
 
-	local = frame->local;
-	cached_subvol = local->cached_subvol;
+        local = frame->local;
+        cached_subvol = local->cached_subvol;
 
         STACK_WIND (frame, dht_create_cbk,
                     cached_subvol, cached_subvol->fops->create,
                     &local->loc, local->flags, local->mode,
-                    local->fd, local->params);
+                    local->umask, local->fd, local->params);
 
         return 0;
- err:
- 	DHT_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
- 	return 0;
+err:
+        DHT_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL, NULL, NULL);
+        return 0;
 }
 
 int
 dht_create (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, int32_t flags, mode_t mode,
-            fd_t *fd, dict_t *params)
+            loc_t *loc, int32_t flags, mode_t mode,
+            mode_t umask, fd_t *fd, dict_t *params)
 {
-	int          op_errno = -1;
-        int          ret = -1;
-	xlator_t    *subvol = NULL;
-	dht_conf_t  *conf = NULL;
+        int          op_errno = -1;
+        xlator_t    *subvol = NULL;
         dht_local_t *local = NULL;
         xlator_t    *avail_subvol = NULL;
 
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
-
-	conf = this->private;
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
         dht_get_du_info (frame, this, loc);
 
-	local = dht_local_init (frame);
-	if (!local) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
+        local = dht_local_init (frame, loc, fd, GF_FOP_CREATE);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
         if (dht_filter_loc_subvol_key (this, loc, &local->loc,
                                        &subvol)) {
-                gf_log (this->name, GF_LOG_NORMAL,
+                gf_log (this->name, GF_LOG_INFO,
                         "creating %s on %s (got create on %s)",
                         local->loc.path, subvol->name, loc->path);
                 STACK_WIND (frame, dht_create_cbk,
                             subvol, subvol->fops->create,
-                            &local->loc, flags, mode, fd, params);
+                            &local->loc, flags, mode, umask, fd, params);
                 goto done;
         }
 
-        ret = loc_dup (loc, &local->loc);
-        if (ret == -1) {
-                op_errno = ENOMEM;
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
-        }
         subvol = dht_subvol_get_hashed (this, loc);
         if (!subvol) {
                 gf_log (this->name, GF_LOG_DEBUG,
@@ -3599,156 +4021,158 @@ dht_create (call_frame_t *frame, xlator_t *this,
                         "creating %s on %s", loc->path, subvol->name);
                 STACK_WIND (frame, dht_create_cbk,
                             subvol, subvol->fops->create,
-                            loc, flags, mode, fd, params);
+                            loc, flags, mode, umask, fd, params);
                 goto done;
         }
-        /* Choose the minimum filled volume, and create the 
+        /* Choose the minimum filled volume, and create the
            files there */
-        /* TODO */
-        avail_subvol = dht_free_disk_available_subvol (this, subvol);
+        avail_subvol = dht_free_disk_available_subvol (this, subvol, local);
         if (avail_subvol != subvol) {
-                local->fd = fd_ref (fd);
                 local->params = dict_ref (params);
                 local->flags = flags;
                 local->mode = mode;
-
+                local->umask = umask;
                 local->cached_subvol = avail_subvol;
                 local->hashed_subvol = subvol;
                 gf_log (this->name, GF_LOG_TRACE,
                         "creating %s on %s (link at %s)", loc->path,
                         avail_subvol->name, subvol->name);
-                dht_linkfile_create (frame,
-                                     dht_create_linkfile_create_cbk,
-                                     avail_subvol, subvol, loc);
+                dht_linkfile_create (frame, dht_create_linkfile_create_cbk,
+                                     this, avail_subvol, subvol, loc);
                 goto done;
         }
         gf_log (this->name, GF_LOG_TRACE,
                 "creating %s on %s", loc->path, subvol->name);
         STACK_WIND (frame, dht_create_cbk,
                     subvol, subvol->fops->create,
-                    loc, flags, mode, fd, params);
+                    loc, flags, mode, umask, fd, params);
 done:
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_mkdir_selfheal_cbk (call_frame_t *frame, void *cookie,
-			xlator_t *this,
-			int32_t op_ret, int32_t op_errno)
+                        xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	dht_local_t   *local = NULL;
-	dht_layout_t  *layout = NULL;
-
+        dht_local_t   *local = NULL;
+        dht_layout_t  *layout = NULL;
 
-	local = frame->local;
-	layout = local->selfheal.layout;
+        local = frame->local;
+        layout = local->selfheal.layout;
 
-	if (op_ret == 0) {
+        if (op_ret == 0) {
                 dht_layout_set (this, local->inode, layout);
-		local->stbuf.ia_ino = local->ia_ino;
                 if (local->loc.parent) {
-                        local->preparent.ia_ino = local->loc.parent->ino;
-                        local->postparent.ia_ino = local->loc.parent->ino;
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->preparent, 0);
 
-                        WIPE (&local->preparent);
-                        WIPE (&local->postparent);
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->postparent, 1);
                 }
-	}
+        }
 
-	DHT_STACK_UNWIND (mkdir, frame, op_ret, op_errno,
-			  local->inode, &local->stbuf, &local->preparent,
-                          &local->postparent);
+        DHT_STACK_UNWIND (mkdir, frame, op_ret, op_errno,
+                          local->inode, &local->stbuf, &local->preparent,
+                          &local->postparent, NULL);
 
-	return 0;
+        return 0;
 }
 
 int
 dht_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                int op_ret, int op_errno, inode_t *inode, struct iatt *stbuf,
-               struct iatt *preparent, struct iatt *postparent)
+               struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	int           ret = -1;
-        int           subvol_filled = 0;
-	call_frame_t *prev = NULL;
-	dht_layout_t *layout = NULL;
-	dht_conf_t   *conf = NULL;
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        int           ret = -1;
+        gf_boolean_t subvol_filled = _gf_false;
+        call_frame_t *prev = NULL;
+        dht_layout_t *layout = NULL;
 
-	conf = this->private;
-	local = frame->local;
-	prev  = cookie;
-	layout = local->layout;
+        local = frame->local;
+        prev  = cookie;
+        layout = local->layout;
 
         subvol_filled = dht_is_subvol_filled (this, prev->this);
 
-	LOCK (&frame->lock);
-	{
+        LOCK (&frame->lock);
+        {
                 if (subvol_filled && (op_ret != -1)) {
                         ret = dht_layout_merge (this, layout, prev->this,
                                                 -1, ENOSPC, NULL);
                 } else {
+			if (op_ret == -1 && op_errno == EEXIST)
+				/* Very likely just a race between mkdir and
+				   self-heal (from lookup of a concurrent mkdir
+				   attempt).
+				   Ignore error for now. layout setting will
+				   anyways fail if this was a different (old)
+				   pre-existing different directory.
+				*/
+				op_ret = 0;
                         ret = dht_layout_merge (this, layout, prev->this,
                                                 op_ret, op_errno, NULL);
                 }
+                if (ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to merge layouts", local->loc.path);
 
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			goto unlock;
-		}
-		dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        goto unlock;
+                }
+                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
                 dht_iatt_merge (this, &local->preparent, preparent, prev->this);
                 dht_iatt_merge (this, &local->postparent, postparent,
                                 prev->this);
-
-                if (prev->this == dht_first_up_subvol (this)) {
-                        local->ia_ino = local->stbuf.ia_ino;
-                }
-
-	}
+        }
 unlock:
-	UNLOCK (&frame->lock);
+        UNLOCK (&frame->lock);
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt)) {
-		dht_selfheal_new_directory (frame, dht_mkdir_selfheal_cbk,
-					    layout);
-	}
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                dht_selfheal_new_directory (frame, dht_mkdir_selfheal_cbk,
+                                            layout);
+        }
 
         return 0;
 }
 
 int
-dht_mkdir_hashed_cbk (call_frame_t *frame, void *cookie, 
-		      xlator_t *this, int op_ret, int op_errno,
+dht_mkdir_hashed_cbk (call_frame_t *frame, void *cookie,
+                      xlator_t *this, int op_ret, int op_errno,
                       inode_t *inode, struct iatt *stbuf,
-                      struct iatt *preparent, struct iatt *postparent)
+                      struct iatt *preparent, struct iatt *postparent,
+                      dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           ret = -1;
-	call_frame_t *prev = NULL;
-	dht_layout_t *layout = NULL;
-	dht_conf_t   *conf = NULL;
-	int           i = 0;
-	xlator_t     *hashed_subvol = NULL;
+        dht_local_t  *local = NULL;
+        int           ret = -1;
+        call_frame_t *prev = NULL;
+        dht_layout_t *layout = NULL;
+        dht_conf_t   *conf = NULL;
+        int           i = 0;
+        xlator_t     *hashed_subvol = NULL;
 
         VALIDATE_OR_GOTO (this->private, err);
 
-	local = frame->local;
-	prev  = cookie;
-	layout = local->layout;
-	conf = this->private;
-	hashed_subvol = local->hashed_subvol;
+        local = frame->local;
+        prev  = cookie;
+        layout = local->layout;
+        conf = this->private;
+        hashed_subvol = local->hashed_subvol;
 
-        if (uuid_is_null (local->loc.inode->gfid) && !op_ret)
-                memcpy (local->loc.inode->gfid, stbuf->ia_gfid, 16);
+        if (uuid_is_null (local->loc.gfid) && !op_ret)
+                uuid_copy (local->loc.gfid, stbuf->ia_gfid);
 
         if (dht_is_subvol_filled (this, hashed_subvol))
                 ret = dht_layout_merge (this, layout, prev->this,
@@ -3756,49 +4180,55 @@ dht_mkdir_hashed_cbk (call_frame_t *frame, void *cookie,
         else
                 ret = dht_layout_merge (this, layout, prev->this,
                                         op_ret, op_errno, NULL);
-        
-	if (op_ret == -1) {
-		local->op_errno = op_errno;
-		goto err;
-	}
-	local->op_ret = 0;
 
-	dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+        /* TODO: we may have to return from the function
+           if layout merge fails. For now, lets just log an error */
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to merge layouts", local->loc.path);
+
+        if (op_ret == -1) {
+                local->op_errno = op_errno;
+                goto err;
+        }
+        local->op_ret = 0;
+
+        dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
         dht_iatt_merge (this, &local->preparent, preparent, prev->this);
         dht_iatt_merge (this, &local->postparent, postparent, prev->this);
 
-	local->ia_ino = local->stbuf.ia_ino;
+        local->call_cnt = conf->subvolume_cnt - 1;
 
-	local->call_cnt = conf->subvolume_cnt - 1;
-	
-	if (local->call_cnt == 0) {
-		dht_selfheal_directory (frame, dht_mkdir_selfheal_cbk,
-					&local->loc, layout);
-	}
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		if (conf->subvolumes[i] == hashed_subvol)
-			continue;
-		STACK_WIND (frame, dht_mkdir_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->mkdir,
-			    &local->loc, local->mode, local->params);
-	}
-	return 0;
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, stbuf->ia_gfid);
+        if (local->call_cnt == 0) {
+                dht_selfheal_directory (frame, dht_mkdir_selfheal_cbk,
+                                        &local->loc, layout);
+        }
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (conf->subvolumes[i] == hashed_subvol)
+                        continue;
+                STACK_WIND (frame, dht_mkdir_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->mkdir, &local->loc,
+                            local->mode, local->umask, local->params);
+        }
+        return 0;
 err:
-	DHT_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        DHT_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL, NULL);
         return 0;
 }
 
 
-int
+ int
 dht_mkdir (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, mode_t mode, dict_t *params)
+           loc_t *loc, mode_t mode, mode_t umask, dict_t *params)
 {
-	dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
+        dht_local_t  *local  = NULL;
+        dht_conf_t   *conf = NULL;
         int           op_errno = -1;
-	int           ret = -1;
-	xlator_t     *hashed_subvol = NULL;
+        xlator_t     *hashed_subvol = NULL;
 
 
         VALIDATE_OR_GOTO (frame, err);
@@ -3808,148 +4238,232 @@ dht_mkdir (call_frame_t *frame, xlator_t *this,
         VALIDATE_OR_GOTO (loc->path, err);
         VALIDATE_OR_GOTO (this->private, err);
 
-	conf = this->private;
+        conf = this->private;
 
         dht_get_du_info (frame, this, loc);
 
-	local = dht_local_init (frame);
-	if (!local) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
-
-	hashed_subvol = dht_subvol_get_hashed (this, loc);
+        local = dht_local_init (frame, loc, NULL, GF_FOP_MKDIR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	if (hashed_subvol == NULL) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"hashed subvol not found for %s",
+        hashed_subvol = dht_subvol_get_hashed (this, loc);
+        if (hashed_subvol == NULL) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "hashed subvol not found for %s",
                         loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local->hashed_subvol = hashed_subvol;
-	local->inode = inode_ref (loc->inode);
-	ret = loc_copy (&local->loc, loc);
-	local->mode = mode;
-
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
+                op_errno = EINVAL;
+                goto err;
+        }
 
+        local->hashed_subvol = hashed_subvol;
+        local->mode = mode;
+        local->umask = umask;
         local->params = dict_ref (params);
+        local->inode  = inode_ref (loc->inode);
 
-	local->layout = dht_layout_new (this, conf->subvolume_cnt);
-	if (!local->layout) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
+        local->layout = dht_layout_new (this, conf->subvolume_cnt);
+        if (!local->layout) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	STACK_WIND (frame, dht_mkdir_hashed_cbk,
-		    hashed_subvol,
-		    hashed_subvol->fops->mkdir,
-		    loc, mode, params);
+        STACK_WIND (frame, dht_mkdir_hashed_cbk,
+                    hashed_subvol,
+                    hashed_subvol->fops->mkdir,
+                    loc, mode, umask, params);
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_rmdir_selfheal_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			int op_ret, int op_errno)
+                        int op_ret, int op_errno, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
+        dht_local_t  *local = NULL;
 
-	local = frame->local;
+        local = frame->local;
+
+        DHT_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
+                          &local->preparent, &local->postparent, NULL);
+
+        return 0;
+}
+
+
+int
+dht_rmdir_hashed_subvol_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+               int op_ret, int op_errno, struct iatt *preparent,
+               struct iatt *postparent, dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
+
+        local = frame->local;
+        prev  = cookie;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        local->op_ret   = -1;
+                        if (op_errno != ENOENT && op_errno != EACCES) {
+                                local->need_selfheal = 1;
+                        }
+
+
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "rmdir on %s for %s failed (%s)",
+                                prev->this->name, local->loc.path,
+                                strerror (op_errno));
+                        goto unlock;
+                }
+
+                dht_iatt_merge (this, &local->preparent, preparent, prev->this);
+                dht_iatt_merge (this, &local->postparent, postparent,
+                                prev->this);
 
-        if (local->loc.parent) {
-                local->preparent.ia_ino = local->loc.parent->ino;
-                local->postparent.ia_ino = local->loc.parent->ino;
         }
+unlock:
+        UNLOCK (&frame->lock);
+
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+               if (local->need_selfheal) {
+                        local->layout =
+                                dht_layout_get (this, local->loc.inode);
 
-	DHT_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
-                          &local->preparent, &local->postparent);
+                        /* TODO: neater interface needed below */
+                        local->stbuf.ia_type = local->loc.inode->ia_type;
 
-	return 0;
+                        uuid_copy (local->gfid, local->loc.inode->gfid);
+                        dht_selfheal_restore (frame, dht_rmdir_selfheal_cbk,
+                                              &local->loc, local->layout);
+               } else {
+
+                        if (local->loc.parent) {
+                                dht_inode_ctx_time_update (local->loc.parent,
+                                                           this,
+                                                           &local->preparent,
+                                                           0);
+
+                                dht_inode_ctx_time_update (local->loc.parent,
+                                                           this,
+                                                           &local->postparent,
+                                                           1);
+                        }
+
+                        DHT_STACK_UNWIND (rmdir, frame, local->op_ret,
+                                          local->op_errno, &local->preparent,
+                                          &local->postparent, NULL);
+               }
+        }
+
+        return 0;
 }
 
 
 int
 dht_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	       int op_ret, int op_errno, struct iatt *preparent,
-               struct iatt *postparent)
+               int op_ret, int op_errno, struct iatt *preparent,
+               struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
+        int           done = 0;
 
-	local = frame->local;
-	prev  = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			local->op_ret   = -1;
-
-			if (op_errno != ENOENT)
-				local->need_selfheal = 1;
-
-			gf_log (this->name, GF_LOG_DEBUG,
-				"rmdir on %s for %s failed (%s)",
-				prev->this->name, local->loc.path,
-				strerror (op_errno));
-			goto unlock;
-		}
+        local = frame->local;
+        prev  = cookie;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        local->op_ret   = -1;
+
+                        if (op_errno != ENOENT && op_errno != EACCES) {
+                                local->need_selfheal = 1;
+                        }
+
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "rmdir on %s for %s failed (%s)",
+                                prev->this->name, local->loc.path,
+                                strerror (op_errno));
+                        goto unlock;
+                }
 
+                /* Track if rmdir succeeded on atleast one subvol*/
+                local->fop_succeeded = 1;
                 dht_iatt_merge (this, &local->preparent, preparent, prev->this);
                 dht_iatt_merge (this, &local->postparent, postparent,
                                 prev->this);
-	}
+        }
 unlock:
-	UNLOCK (&frame->lock);
+        UNLOCK (&frame->lock);
 
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt)) {
-		if (local->need_selfheal) {
+        this_call_cnt = dht_frame_return (frame);
+
+        /* if local->hashed_subvol, we are yet to wind to hashed_subvol. */
+        if (local->hashed_subvol && (this_call_cnt == 1)) {
+                done = 1;
+        } else if (!local->hashed_subvol && !this_call_cnt) {
+                done = 1;
+        }
+
+
+        if (done) {
+                if (local->need_selfheal && local->fop_succeeded) {
                         local->layout =
                                 dht_layout_get (this, local->loc.inode);
 
-			/* TODO: neater interface needed below */
-			local->stbuf.ia_type = local->loc.inode->ia_type;
+                        /* TODO: neater interface needed below */
+                        local->stbuf.ia_type = local->loc.inode->ia_type;
+
+                        uuid_copy (local->gfid, local->loc.inode->gfid);
+                        dht_selfheal_restore (frame, dht_rmdir_selfheal_cbk,
+                                              &local->loc, local->layout);
+                } else if (this_call_cnt) {
+                        /* If non-hashed subvol's have responded, proceed */
+
+                        local->need_selfheal = 0;
+                        STACK_WIND (frame, dht_rmdir_hashed_subvol_cbk,
+                                    local->hashed_subvol,
+                                    local->hashed_subvol->fops->rmdir,
+                                    &local->loc, local->flags, NULL);
+                } else if (!this_call_cnt) {
+                        /* All subvol's have responded, proceed */
 
-			dht_selfheal_restore (frame, dht_rmdir_selfheal_cbk,
-					      &local->loc, local->layout);
-		} else {
                         if (local->loc.parent) {
-                                local->preparent.ia_ino =
-                                        local->loc.parent->ino;
-                                local->postparent.ia_ino =
-                                        local->loc.parent->ino;
 
-                                WIPE (&local->preparent);
-                                WIPE (&local->postparent);
+                                dht_inode_ctx_time_update (local->loc.parent,
+                                                           this,
+                                                           &local->preparent,
+                                                           0);
+
+                                dht_inode_ctx_time_update (local->loc.parent,
+                                                           this,
+                                                           &local->postparent,
+                                                           1);
+
                         }
 
-			DHT_STACK_UNWIND (rmdir, frame, local->op_ret,
-					  local->op_errno, &local->preparent,
-                                          &local->postparent);
-		}
-	}
+                        DHT_STACK_UNWIND (rmdir, frame, local->op_ret,
+                                          local->op_errno, &local->preparent,
+                                          &local->postparent, NULL);
+                }
+        }
 
         return 0;
 }
@@ -3958,40 +4472,64 @@ unlock:
 int
 dht_rmdir_do (call_frame_t *frame, xlator_t *this)
 {
-	dht_local_t  *local = NULL;
-	dht_conf_t   *conf = NULL;
-	int           i = 0;
+        dht_local_t  *local = NULL;
+        dht_conf_t   *conf = NULL;
+        int           i = 0;
+        xlator_t     *hashed_subvol = NULL;
 
         VALIDATE_OR_GOTO (this->private, err);
 
-	conf = this->private;
-	local = frame->local;
+        conf = this->private;
+        local = frame->local;
 
-	if (local->op_ret == -1)
-		goto err;
+        if (local->op_ret == -1)
+                goto err;
 
-	local->call_cnt = conf->subvolume_cnt;
+        local->call_cnt = conf->subvolume_cnt;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_rmdir_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->rmdir,
-			    &local->loc, local->flags);
-	}
+        /* first remove from non-hashed_subvol */
+        hashed_subvol = dht_subvol_get_hashed (this, &local->loc);
 
-	return 0;
+        if (!hashed_subvol) {
+                gf_log (this->name, GF_LOG_WARNING, "failed to get hashed "
+                        "subvol for %s",local->loc.path);
+        } else {
+                local->hashed_subvol = hashed_subvol;
+        }
+
+        /* When DHT has only 1 child */
+        if (conf->subvolume_cnt == 1) {
+                STACK_WIND (frame, dht_rmdir_hashed_subvol_cbk,
+                            conf->subvolumes[0],
+                            conf->subvolumes[0]->fops->rmdir,
+                            &local->loc, local->flags, NULL);
+                return 0;
+        }
+
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (hashed_subvol &&
+                    (hashed_subvol == conf->subvolumes[i]))
+                        continue;
+
+                STACK_WIND (frame, dht_rmdir_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->rmdir,
+                            &local->loc, local->flags, NULL);
+        }
+
+        return 0;
 
 err:
-	DHT_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
-                          &local->preparent, &local->postparent);
-	return 0;
+        DHT_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
+                          &local->preparent, &local->postparent, NULL);
+        return 0;
 }
 
 
 int
 dht_rmdir_linkfile_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                int op_ret, int op_errno, struct iatt *preparent,
-                               struct iatt *postparent)
+                               struct iatt *postparent, dict_t *xdata)
 {
         dht_local_t    *local = NULL;
         call_frame_t   *prev = NULL;
@@ -4039,6 +4577,7 @@ 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;
@@ -4050,7 +4589,7 @@ 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;
 
@@ -4061,7 +4600,74 @@ dht_rmdir_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         }
 
         STACK_WIND (frame, dht_rmdir_linkfile_unlink_cbk,
-                    src, src->fops->unlink, &local->loc);
+                    src, src->fops->unlink, &local->loc, 0, NULL);
+        return 0;
+err:
+
+        this_call_cnt = dht_frame_return (main_frame);
+        if (is_last_call (this_call_cnt))
+                dht_rmdir_do (main_frame, this);
+
+        DHT_STACK_DESTROY (frame);
+        return 0;
+}
+
+
+int
+dht_rmdir_cached_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                             int op_ret, int op_errno, inode_t *inode,
+                             struct iatt *stbuf, dict_t *xattr,
+                             struct iatt *parent)
+{
+        dht_local_t    *local         = NULL;
+        xlator_t       *src           = NULL;
+        call_frame_t   *main_frame    = NULL;
+        dht_local_t    *main_local    = NULL;
+        int             this_call_cnt = 0;
+        dht_conf_t     *conf          = this->private;
+        dict_t         *xattrs        = NULL;
+        int             ret           = 0;
+
+        local = frame->local;
+        src   = local->hashed_subvol;
+
+        main_frame = local->main_frame;
+        main_local = main_frame->local;
+
+        if (op_ret == 0) {
+                main_local->op_ret  = -1;
+                main_local->op_errno = ENOTEMPTY;
+
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s found on cached subvol %s",
+                        local->loc.path, src->name);
+                goto err;
+        } else if (op_errno != ENOENT) {
+                main_local->op_ret  = -1;
+                main_local->op_errno = op_errno;
+                goto err;
+        }
+
+        xattrs = dict_new ();
+        if (!xattrs) {
+                gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
+                goto err;
+        }
+
+        ret = dict_set_uint32 (xattrs, conf->link_xattr_name, 256);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to set linkto key"
+                        " in dict");
+                if (xattrs)
+                        dict_unref (xattrs);
+                goto err;
+        }
+
+        STACK_WIND (frame, dht_rmdir_lookup_cbk,
+                    src, src->fops->lookup, &local->loc, xattrs);
+        if (xattrs)
+                dict_unref (xattrs);
+
         return 0;
 err:
 
@@ -4078,12 +4684,15 @@ int
 dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
                            gf_dirent_t *entries, xlator_t *src)
 {
-        int                 ret = 0;
-        int                 build_ret = 0;
-        gf_dirent_t        *trav = NULL;
+        int                 ret          = 0;
+        int                 build_ret    = 0;
+        gf_dirent_t        *trav         = NULL;
         call_frame_t       *lookup_frame = NULL;
         dht_local_t        *lookup_local = NULL;
-        dht_local_t        *local = NULL;
+        dht_local_t        *local        = NULL;
+        dict_t             *xattrs       = NULL;
+        dht_conf_t         *conf         = this->private;
+        xlator_t           *subvol       = NULL;
 
         local = frame->local;
 
@@ -4092,7 +4701,8 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
                         continue;
                 if (strcmp (trav->d_name, "..") == 0)
                         continue;
-                if (check_is_linkfile (NULL, (&trav->d_stat), NULL) == 1) {
+                if (check_is_linkfile (NULL, (&trav->d_stat), trav->dict,
+                                              conf->link_xattr_name)) {
                         ret++;
                         continue;
                 }
@@ -4104,6 +4714,21 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
                 return 0;
         }
 
+        xattrs = dict_new ();
+        if (!xattrs) {
+                gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
+                return -1;
+        }
+
+        ret = dict_set_uint32 (xattrs, conf->link_xattr_name, 256);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to set linkto key"
+                        " in dict");
+                if (xattrs)
+                        dict_unref (xattrs);
+                return -1;
+        }
+
         list_for_each_entry (trav, &entries->list, list) {
                 if (strcmp (trav->d_name, ".") == 0)
                         continue;
@@ -4115,29 +4740,27 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
 
                 lookup_frame = copy_frame (frame);
                 if (!lookup_frame) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of Memory");
                         /* out of memory, let the rmdir fail
                            (as non-empty, unfortunately) */
                         goto err;
                 }
 
-                lookup_local = GF_CALLOC (sizeof (*local), 1,
-                                          gf_dht_mt_dht_local_t);
+                lookup_local = mem_get0 (this->local_pool);
                 if (!lookup_local) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of Memory");
                         goto err;
                 }
 
                 lookup_frame->local = lookup_local;
                 lookup_local->main_frame = frame;
+                lookup_local->hashed_subvol = src;
 
                 build_ret = dht_build_child_loc (this, &lookup_local->loc,
                                                  &local->loc, trav->d_name);
                 if (build_ret != 0)
                         goto err;
 
+                uuid_copy (lookup_local->loc.gfid, trav->d_stat.ia_gfid);
+
                 gf_log (this->name, GF_LOG_TRACE,
                         "looking up %s on %s",
                         lookup_local->loc.path, src->name);
@@ -4148,14 +4771,31 @@ dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
                 }
                 UNLOCK (&frame->lock);
 
-                STACK_WIND (lookup_frame, dht_rmdir_lookup_cbk,
-                            src, src->fops->lookup,
-                            &lookup_local->loc, NULL);
+                subvol = dht_linkfile_subvol (this, NULL, &trav->d_stat,
+                                              trav->dict);
+                if (!subvol) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "linkfile not having link subvolume. path=%s",
+                                lookup_local->loc.path);
+                        STACK_WIND (lookup_frame, dht_rmdir_lookup_cbk,
+                                    src, src->fops->lookup,
+                                    &lookup_local->loc, xattrs);
+                } else {
+                        STACK_WIND (lookup_frame, dht_rmdir_cached_lookup_cbk,
+                                    subvol, subvol->fops->lookup,
+                                    &lookup_local->loc, xattrs);
+                }
                 ret++;
         }
 
+        if (xattrs)
+                dict_unref (xattrs);
+
         return ret;
 err:
+        if (xattrs)
+                dict_unref (xattrs);
+
         DHT_STACK_DESTROY (lookup_frame);
         return 0;
 }
@@ -4163,19 +4803,20 @@ err:
 
 int
 dht_rmdir_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int op_ret, int op_errno, gf_dirent_t *entries)
+                        int op_ret, int op_errno, gf_dirent_t *entries,
+                        dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = -1;
-	call_frame_t *prev = NULL;
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = -1;
+        call_frame_t *prev = NULL;
         xlator_t     *src = NULL;
         int           ret = 0;
 
-	local = frame->local;
-	prev  = cookie;
+        local = frame->local;
+        prev  = cookie;
         src   = prev->this;
 
-	if (op_ret > 2) {
+        if (op_ret > 2) {
                 ret = dht_rmdir_is_subvol_empty (frame, this, entries, src);
 
                 switch (ret) {
@@ -4193,647 +4834,258 @@ dht_rmdir_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                 prev->this->name, local->loc.path, ret);
                         break;
                 }
-	}
+        }
 
-	this_call_cnt = dht_frame_return (frame);
+        this_call_cnt = dht_frame_return (frame);
 
-	if (is_last_call (this_call_cnt)) {
-		dht_rmdir_do (frame, this);
-	}
+        if (is_last_call (this_call_cnt)) {
+                dht_rmdir_do (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_rmdir_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int op_ret, int op_errno, fd_t *fd)
-{
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = -1;
-	call_frame_t *prev = NULL;
-
-
-	local = frame->local;
-	prev  = cookie;
-
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"opendir on %s for %s failed (%s)",
-			prev->this->name, local->loc.path,
-			strerror (op_errno));
-		goto err;
-	}
-
-	STACK_WIND (frame, dht_rmdir_readdirp_cbk,
-		    prev->this, prev->this->fops->readdirp,
-		    local->fd, 4096, 0);
-
-	return 0;
-
-err:
-	this_call_cnt = dht_frame_return (frame);
-
-	if (is_last_call (this_call_cnt)) {
-		dht_rmdir_do (frame, this);
-	}
-
-	return 0;
-}
-
-
-int
-dht_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags)
-{
-	dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
-        int           op_errno = -1;
-	int           i = -1;
-	int           ret = -1;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
-        VALIDATE_OR_GOTO (this->private, err);
-
-	conf = this->private;
-
-	local = dht_local_init (frame);
-	if (!local) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
-
-	local->call_cnt = conf->subvolume_cnt;
-	local->op_ret   = 0;
-
-	ret = loc_copy (&local->loc, loc);
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
-
-        local->flags = flags;
-
-	local->fd = fd_create (local->loc.inode, frame->root->pid);
-	if (!local->fd) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
-
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_rmdir_opendir_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->opendir,
-			    loc, local->fd);
-	}
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (rmdir, frame, -1, op_errno,
-                          NULL, NULL);
-
-	return 0;
-}
-
-
-int
-dht_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		 int32_t op_ret, int32_t op_errno, dict_t *dict)
+                       int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
 {
-	DHT_STACK_UNWIND (xattrop, frame, op_ret, op_errno, dict);
-	return 0;
-}
-
-
-int
-dht_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
-	     gf_xattrop_flags_t flags, dict_t *dict)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
-
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = 1;
-
-	STACK_WIND (frame,
-		    dht_xattrop_cbk,
-		    subvol, subvol->fops->xattrop,
-		    loc, flags, dict);
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL);
+        dht_local_t  *local         = NULL;
+        int           this_call_cnt = -1;
+        call_frame_t *prev          = NULL;
+        dict_t       *dict          = NULL;
+        int           ret           = 0;
+        dht_conf_t   *conf          = this->private;
+        int           i             = 0;
 
-	return 0;
-}
+        local = frame->local;
+        prev  = cookie;
 
+        this_call_cnt = dht_frame_return (frame);
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "opendir on %s for %s failed (%s)",
+                        prev->this->name, local->loc.path,
+                        strerror (op_errno));
+                if (op_errno != ENOENT) {
+                        local->op_ret = -1;
+                        local->op_errno = op_errno;
+                }
+                goto err;
+        }
 
-int
-dht_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		  int32_t op_ret, int32_t op_errno, dict_t *dict)
-{
-	DHT_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, dict);
-	return 0;
-}
+        if (!is_last_call (this_call_cnt))
+                return 0;
 
+        if (local->op_ret == -1)
+                goto err;
 
-int
-dht_fxattrop (call_frame_t *frame, xlator_t *this,
-	      fd_t *fd, gf_xattrop_flags_t flags, dict_t *dict)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
+        dict = dict_new ();
+        if (!dict) {
+                local->op_ret = -1;
+                local->op_errno = ENOMEM;
+                goto err;
+        }
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        ret = dict_set_uint32 (dict, conf->link_xattr_name, 256);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to set '%s' key",
+                        local->loc.path, conf->link_xattr_name);
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+        local->call_cnt = conf->subvolume_cnt;
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                STACK_WIND (frame, dht_rmdir_readdirp_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->readdirp,
+                            local->fd, 4096, 0, dict);
+        }
 
-	STACK_WIND (frame,
-		    dht_fxattrop_cbk,
-		    subvol, subvol->fops->fxattrop,
-		    fd, flags, dict);
+        if (dict)
+                dict_unref (dict);
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL);
+        if (is_last_call (this_call_cnt)) {
+                dht_rmdir_do (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
-dht_inodelk_cbk (call_frame_t *frame, void *cookie,
-		 xlator_t *this, int32_t op_ret, int32_t op_errno)
-
-{
-	DHT_STACK_UNWIND (inodelk, frame, op_ret, op_errno);
-	return 0;
-}
-
-
-int32_t
-dht_inodelk (call_frame_t *frame, xlator_t *this,
-	     const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *lock)
+dht_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+           dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
+        dht_local_t  *local  = NULL;
+        dht_conf_t   *conf = NULL;
         int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
+        int           i = -1;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (loc, err);
         VALIDATE_OR_GOTO (loc->inode, err);
         VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = 1;
-
-	STACK_WIND (frame,
-		    dht_inodelk_cbk,
-		    subvol, subvol->fops->inodelk,
-		    volume, loc, cmd, lock);
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (inodelk, frame, -1, op_errno);
-
-	return 0;
-}
-
-
-int
-dht_finodelk_cbk (call_frame_t *frame, void *cookie,
-		  xlator_t *this, int32_t op_ret, int32_t op_errno)
-
-{
-	DHT_STACK_UNWIND (finodelk, frame, op_ret, op_errno);
-	return 0;
-}
+        conf = this->private;
 
+        local = dht_local_init (frame, loc, NULL, GF_FOP_RMDIR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-int
-dht_finodelk (call_frame_t *frame, xlator_t *this,
-	      const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *lock)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
+        local->call_cnt = conf->subvolume_cnt;
+        local->op_ret   = 0;
+        local->fop_succeeded = 0;
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        local->flags = flags;
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+        local->fd = fd_create (local->loc.inode, frame->root->pid);
+        if (!local->fd) {
 
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	STACK_WIND (frame,
-		    dht_finodelk_cbk,
-		    subvol, subvol->fops->finodelk,
-		    volume, fd, cmd, lock);
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                STACK_WIND (frame, dht_rmdir_opendir_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->opendir,
+                            loc, local->fd, NULL);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (finodelk, frame, -1, op_errno);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (rmdir, frame, -1, op_errno,
+                          NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
-
 int
 dht_entrylk_cbk (call_frame_t *frame, void *cookie,
-		 xlator_t *this, int32_t op_ret, int32_t op_errno)
+                 xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
 
 {
-	DHT_STACK_UNWIND (entrylk, frame, op_ret, op_errno);
-	return 0;
+        DHT_STACK_UNWIND (entrylk, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
 
 int
 dht_entrylk (call_frame_t *frame, xlator_t *this,
-	     const char *volume, loc_t *loc, const char *basename,
-	     entrylk_cmd cmd, entrylk_type type)
+             const char *volume, loc_t *loc, const char *basename,
+             entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
+        xlator_t     *subvol = NULL;
         int           op_errno = -1;
-	dht_local_t  *local = NULL;
+        dht_local_t  *local = NULL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (loc, err);
         VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
 
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        local = dht_local_init (frame, loc, NULL, GF_FOP_ENTRYLK);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = 1;
+        local->call_cnt = 1;
 
-	STACK_WIND (frame, dht_entrylk_cbk,
-		    subvol, subvol->fops->entrylk,
-		    volume, loc, basename, cmd, type);
+        STACK_WIND (frame, dht_entrylk_cbk,
+                    subvol, subvol->fops->entrylk,
+                    volume, loc, basename, cmd, type, xdata);
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (entrylk, frame, -1, op_errno);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (entrylk, frame, -1, op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_fentrylk_cbk (call_frame_t *frame, void *cookie,
-		  xlator_t *this, int32_t op_ret, int32_t op_errno)
+                  xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
 
 {
-	DHT_STACK_UNWIND (fentrylk, frame, op_ret, op_errno);
-	return 0;
+        DHT_STACK_UNWIND (fentrylk, frame, op_ret, op_errno, NULL);
+        return 0;
 }
 
 
 int
 dht_fentrylk (call_frame_t *frame, xlator_t *this,
-	      const char *volume, fd_t *fd, const char *basename,
-	      entrylk_cmd cmd, entrylk_type type)
+              const char *volume, fd_t *fd, const char *basename,
+              entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
+        xlator_t     *subvol = NULL;
         int           op_errno = -1;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (fd, err);
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	STACK_WIND (frame, dht_fentrylk_cbk,
-		    subvol, subvol->fops->fentrylk,
-		    volume, fd, basename, cmd, type);
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (fentrylk, frame, -1, op_errno);
-
-	return 0;
-}
-
-
-int
-dht_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                 int op_ret, int op_errno, struct iatt *statpre,
-                 struct iatt *statpost)
-{
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
-
-
-	local = frame->local;
-	prev = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
-
-		dht_iatt_merge (this, &local->prebuf, statpre, prev->this);
-                dht_iatt_merge (this, &local->stbuf, statpost, prev->this);
-		
-		if (local->inode) {
-			local->prebuf.ia_ino = local->inode->ino;
-                        local->stbuf.ia_ino = local->inode->ino;
-                }
-
-		local->op_ret = 0;
-	}
-unlock:
-	UNLOCK (&frame->lock);
+        subvol = dht_subvol_get_cached (this, fd->inode);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (setattr, frame, local->op_ret, local->op_errno,
-				  &local->prebuf, &local->stbuf);
+        STACK_WIND (frame, dht_fentrylk_cbk,
+                    subvol, subvol->fops->fentrylk,
+                    volume, fd, basename, cmd, type, xdata);
 
         return 0;
-}
-
-
-int
-dht_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-             struct iatt *stbuf, int32_t valid)
-{
-	dht_layout_t *layout = NULL;
-	dht_local_t  *local  = NULL;
-        int           op_errno = -1;
-	int           i = -1;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"memory allocation failed :(");
-		goto err;
-	}
-
-	local->layout = layout = dht_layout_get (this, loc->inode);
-	if (!layout) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no layout for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	if (!layout_is_sane (layout)) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"layout is not sane for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = layout->cnt;
-
-	for (i = 0; i < layout->cnt; i++) {
-		STACK_WIND (frame, dht_setattr_cbk,
-			    layout->list[i].xlator,
-			    layout->list[i].xlator->fops->setattr,
-			    loc, stbuf, valid);
-	}
-
-	return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL);
-
-	return 0;
-}
-
-
-int
-dht_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
-              int32_t valid)
-{
-	dht_layout_t *layout = NULL;
-	dht_local_t  *local  = NULL;
-        int           op_errno = -1;
-	int           i = -1;
-
-
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fentrylk, frame, -1, op_errno, NULL);
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	local->layout = layout = dht_layout_get (this, fd->inode);
-	if (!layout) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no layout for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	if (!layout_is_sane (layout)) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"layout is not sane for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local->inode = inode_ref (fd->inode);
-	local->call_cnt = layout->cnt;
-
-	for (i = 0; i < layout->cnt; i++) {
-		STACK_WIND (frame, dht_setattr_cbk,
-			    layout->list[i].xlator,
-			    layout->list[i].xlator->fops->fsetattr,
-			    fd, stbuf, valid);
-	}
-
-	return 0;
-
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL);
-
-	return 0;
+        return 0;
 }
 
 
 int
 dht_forget (xlator_t *this, inode_t *inode)
 {
-	uint64_t      tmp_layout = 0;
-	dht_layout_t *layout = NULL;
-
-	inode_ctx_get (inode, this, &tmp_layout);
-
-	if (!tmp_layout)
-		return 0;
-
-	layout = (dht_layout_t *)(long)tmp_layout;
-        dht_layout_unref (this, layout);
-
-	return 0;
-}
+        uint64_t        ctx_int = 0;
+        dht_inode_ctx_t *ctx    = NULL;
+        dht_layout_t *layout = NULL;
 
+        inode_ctx_del (inode, this, &ctx_int);
 
+        if (!ctx_int)
+                return 0;
 
-int
-dht_init_subvolumes (xlator_t *this, dht_conf_t *conf)
-{
-        xlator_list_t *subvols = NULL;
-        int            cnt = 0;
-
-        if (!conf)
-                return -1;
-
-        for (subvols = this->children; subvols; subvols = subvols->next)
-                cnt++;
-
-        conf->subvolumes = GF_CALLOC (cnt, sizeof (xlator_t *),
-                                      gf_dht_mt_xlator_t);
-        if (!conf->subvolumes) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                return -1;
-        }
-        conf->subvolume_cnt = cnt;
-
-        cnt = 0;
-        for (subvols = this->children; subvols; subvols = subvols->next)
-                conf->subvolumes[cnt++] = subvols->xlator;
+        ctx = (dht_inode_ctx_t *) (long) ctx_int;
 
-	conf->subvolume_status = GF_CALLOC (cnt, sizeof (char),
-                                            gf_dht_mt_char);
-	if (!conf->subvolume_status) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		return -1;
-	}
+        layout = ctx->layout;
+        ctx->layout = NULL;
+        dht_layout_unref (this, layout);
+        GF_FREE (ctx);
 
-	conf->last_event = GF_CALLOC (cnt, sizeof (int),
-                                      gf_dht_mt_char);
-	if (!conf->last_event) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		return -1;
-	}
         return 0;
 }
 
@@ -4841,18 +5093,24 @@ dht_init_subvolumes (xlator_t *this, dht_conf_t *conf)
 int
 dht_notify (xlator_t *this, int event, void *data, ...)
 {
-	xlator_t   *subvol = NULL;
-	int         cnt    = -1;
-	int         i      = -1;
-	dht_conf_t *conf   = NULL;
-	int         ret    = -1;
-        int         propagate = 0;
+        xlator_t                *subvol = NULL;
+        int                      cnt    = -1;
+        int                      i      = -1;
+        dht_conf_t              *conf   = NULL;
+        int                      ret    = -1;
+        int                      propagate = 0;
+
+        int                      had_heard_from_all = 0;
+        int                      have_heard_from_all = 0;
+        struct timeval           time = {0,};
+        gf_defrag_info_t        *defrag = NULL;
+        dict_t                  *dict   = NULL;
+        gf_defrag_type           cmd    = 0;
+        dict_t                  *output = NULL;
+        va_list                  ap;
 
-        int         had_heard_from_all = 0;
-        int         have_heard_from_all = 0;
 
-
-	conf = this->private;
+        conf = this->private;
         if (!conf)
                 return ret;
 
@@ -4864,92 +5122,145 @@ dht_notify (xlator_t *this, int event, void *data, ...)
                 }
         }
 
-	switch (event) {
-	case GF_EVENT_CHILD_UP:
-		subvol = data;
+        switch (event) {
+        case GF_EVENT_CHILD_UP:
+                subvol = data;
 
-		conf->gen++;
+                conf->gen++;
 
-		for (i = 0; i < conf->subvolume_cnt; i++) {
-			if (subvol == conf->subvolumes[i]) {
-				cnt = i;
-				break;
-			}
-		}
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (subvol == conf->subvolumes[i]) {
+                                cnt = i;
+                                break;
+                        }
+                }
 
-		if (cnt == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"got GF_EVENT_CHILD_UP bad subvolume %s",
-				subvol->name);
-			break;
-		}
+                if (cnt == -1) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "got GF_EVENT_CHILD_UP bad subvolume %s",
+                                subvol->name);
+                        break;
+                }
 
-		LOCK (&conf->subvolume_lock);
-		{
-			conf->subvolume_status[cnt] = 1;
+                gettimeofday (&time, NULL);
+                LOCK (&conf->subvolume_lock);
+                {
+                        conf->subvolume_status[cnt] = 1;
                         conf->last_event[cnt] = event;
-		}
-		UNLOCK (&conf->subvolume_lock);
+                        conf->subvol_up_time[cnt] = time.tv_sec;
+                }
+                UNLOCK (&conf->subvolume_lock);
 
                 /* one of the node came back up, do a stat update */
                 dht_get_du_info_for_subvol (this, cnt);
 
-		break;
+                break;
 
-	case GF_EVENT_CHILD_DOWN:
-		subvol = data;
+        case GF_EVENT_CHILD_MODIFIED:
+                subvol = data;
 
-		for (i = 0; i < conf->subvolume_cnt; i++) {
-			if (subvol == conf->subvolumes[i]) {
-				cnt = i;
-				break;
-			}
-		}
+                conf->gen++;
+                propagate = 1;
 
-		if (cnt == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"got GF_EVENT_CHILD_DOWN bad subvolume %s",
-				subvol->name);
-			break;
-		}
+                break;
+
+        case GF_EVENT_CHILD_DOWN:
+                subvol = data;
 
-		LOCK (&conf->subvolume_lock);
-		{
-			conf->subvolume_status[cnt] = 0;
+                if (conf->assert_no_child_down) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "Received CHILD_DOWN. Exiting");
+                        if (conf->defrag) {
+                                gf_defrag_stop (conf->defrag,
+                                                GF_DEFRAG_STATUS_FAILED, NULL);
+                        } else {
+                                kill (getpid(), SIGTERM);
+                        }
+                }
+
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (subvol == conf->subvolumes[i]) {
+                                cnt = i;
+                                break;
+                        }
+                }
+
+                if (cnt == -1) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "got GF_EVENT_CHILD_DOWN bad subvolume %s",
+                                subvol->name);
+                        break;
+                }
+
+                LOCK (&conf->subvolume_lock);
+                {
+                        conf->subvolume_status[cnt] = 0;
                         conf->last_event[cnt] = event;
-		}
-		UNLOCK (&conf->subvolume_lock);
+                        conf->subvol_up_time[cnt] = 0;
+                }
+                UNLOCK (&conf->subvolume_lock);
 
                 break;
 
-	case GF_EVENT_CHILD_CONNECTING:
-		subvol = data;
+        case GF_EVENT_CHILD_CONNECTING:
+                subvol = data;
 
-		for (i = 0; i < conf->subvolume_cnt; i++) {
-			if (subvol == conf->subvolumes[i]) {
-				cnt = i;
-				break;
-			}
-		}
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (subvol == conf->subvolumes[i]) {
+                                cnt = i;
+                                break;
+                        }
+                }
 
-		if (cnt == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"got GF_EVENT_CHILD_CONNECTING bad subvolume %s",
-				subvol->name);
-			break;
-		}
+                if (cnt == -1) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "got GF_EVENT_CHILD_CONNECTING bad subvolume %s",
+                                subvol->name);
+                        break;
+                }
 
-		LOCK (&conf->subvolume_lock);
-		{
+                LOCK (&conf->subvolume_lock);
+                {
                         conf->last_event[cnt] = event;
-		}
-		UNLOCK (&conf->subvolume_lock);
+                }
+                UNLOCK (&conf->subvolume_lock);
+
+                break;
+        case GF_EVENT_VOLUME_DEFRAG:
+        {
+                if (!conf->defrag) {
+                        return ret;
+                }
+                defrag = conf->defrag;
+
+                dict = data;
+                va_start (ap, data);
+                output = va_arg (ap, dict_t*);
+
+                ret = dict_get_int32 (dict, "rebalance-command",
+                                      (int32_t*)&cmd);
+                if (ret)
+                        return ret;
+                LOCK (&defrag->lock);
+                {
+                        if (defrag->is_exiting)
+                                goto unlock;
+                        if (cmd == GF_DEFRAG_CMD_STATUS)
+                                gf_defrag_status_get (defrag, output);
+                        else if (cmd == GF_DEFRAG_CMD_STOP)
+                                gf_defrag_stop (defrag,
+                                                GF_DEFRAG_STATUS_STOPPED, output);
+                }
+unlock:
+                UNLOCK (&defrag->lock);
+                return 0;
+                break;
+        }
 
-		break;
         default:
                 propagate = 1;
                 break;
-	}
+        }
 
 
         /* have all subvolumes reported status once by now? */
@@ -4961,9 +5272,12 @@ dht_notify (xlator_t *this, int event, void *data, ...)
 
         /* if all subvols have reported status, no need to hide anything
            or wait for anything else. Just propagate blindly */
-        if (have_heard_from_all)
+        if (have_heard_from_all) {
                 propagate = 1;
 
+        }
+
+
         if (!had_heard_from_all && have_heard_from_all) {
                 /* This is the first event which completes aggregation
                    of events from all subvolumes. If at least one subvol
@@ -4982,11 +5296,45 @@ dht_notify (xlator_t *this, int event, void *data, ...)
                                 /* 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;
         if (propagate)
                 ret = default_notify (this, event, data);
 
-	return ret;
+        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 becc8d9c8..d391b87d5 100644
--- a/xlators/cluster/dht/src/dht-common.h
+++ b/xlators/cluster/dht/src/dht-common.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -22,283 +13,774 @@
 #include "config.h"
 #endif
 
+#include <regex.h>
+#include <signal.h>
+
 #include "dht-mem-types.h"
+#include "libxlator.h"
+#include "syncop.h"
 
 #ifndef _DHT_H
 #define _DHT_H
 
-#define GF_XATTR_FIX_LAYOUT_KEY   "trusted.distribute.fix.layout"
+#define GF_XATTR_FIX_LAYOUT_KEY     "distribute.fix.layout"
 #define GF_DHT_LOOKUP_UNHASHED_ON   1
 #define GF_DHT_LOOKUP_UNHASHED_AUTO 2
+#define DHT_PATHINFO_HEADER         "DISTRIBUTE:"
 
 #include <fnmatch.h>
 
 typedef int (*dht_selfheal_dir_cbk_t) (call_frame_t *frame, void *cookie,
-				       xlator_t *this,
-				       int32_t op_ret, int32_t op_errno);
+                                       xlator_t     *this,
+                                       int32_t       op_ret, int32_t op_errno,
+                                       dict_t *xdata);
+typedef int (*dht_defrag_cbk_fn_t) (xlator_t        *this, call_frame_t *frame,
+                                    int              ret);
 
 
 struct dht_layout {
-        int               cnt;
-	int               preset;
-        int               gen;
-	int               type;
-        int               ref;   /* use with dht_conf_t->layout_lock */
-        int               search_unhashed;
+        int                spread_cnt;  /* layout spread count per directory,
+                                           is controlled by 'setxattr()' with
+                                           special key */
+        int                cnt;
+        int                preset;
+        int                gen;
+        int                type;
+        int                ref; /* use with dht_conf_t->layout_lock */
+        int                search_unhashed;
         struct {
-		int       err;   /* 0 = normal
-				   -1 = dir exists and no xattr
-				   >0 = dir lookup failed with errno
-				 */
-                uint32_t  start;
-                uint32_t  stop;
-                xlator_t *xlator;
-        } list[0];
+                int        err;   /* 0 = normal
+                                     -1 = dir exists and no xattr
+                                     >0 = dir lookup failed with errno
+                                  */
+                uint32_t   start;
+                uint32_t   stop;
+                xlator_t  *xlator;
+        } list[];
+};
+typedef struct dht_layout  dht_layout_t;
+
+struct dht_stat_time {
+        uint32_t        atime;
+        uint32_t        atime_nsec;
+        uint32_t        ctime;
+        uint32_t        ctime_nsec;
+        uint32_t        mtime;
+        uint32_t        mtime_nsec;
+};
+
+typedef struct dht_stat_time dht_stat_time_t;
+
+struct dht_inode_ctx {
+        dht_layout_t    *layout;
+        dht_stat_time_t  time;
 };
-typedef struct dht_layout dht_layout_t;
+
+typedef struct dht_inode_ctx dht_inode_ctx_t;
 
 
 typedef enum {
-	DHT_HASH_TYPE_DM,
+        DHT_HASH_TYPE_DM,
+        DHT_HASH_TYPE_DM_USER,
 } dht_hashfn_type_t;
 
+/* rebalance related */
+struct dht_rebalance_ {
+        xlator_t            *from_subvol;
+        xlator_t            *target_node;
+        off_t                offset;
+        size_t               size;
+        int32_t              flags;
+        int                  count;
+        struct iobref       *iobref;
+        struct iovec        *vector;
+        struct iatt          stbuf;
+        dht_defrag_cbk_fn_t  target_op_fn;
+        dict_t              *xdata;
+};
 
 struct dht_local {
-	int                      call_cnt;
-	loc_t                    loc;
-	loc_t                    loc2;
-	int                      op_ret;
-	int                      op_errno;
-	int                      layout_mismatch;
+        int                      call_cnt;
+        loc_t                    loc;
+        loc_t                    loc2;
+        int                      op_ret;
+        int                      op_errno;
+        int                      layout_mismatch;
         /* Use stbuf as the postbuf, when we require both
          * pre and post attrs */
-	struct iatt              stbuf;
+        struct iatt              stbuf;
         struct iatt              prebuf;
         struct iatt              preoldparent;
         struct iatt              postoldparent;
         struct iatt              preparent;
         struct iatt              postparent;
-	struct statvfs           statvfs;
-	fd_t                    *fd;
-	inode_t                 *inode;
+        struct statvfs           statvfs;
+        fd_t                    *fd;
+        inode_t                 *inode;
         dict_t                  *params;
-	dict_t                  *xattr;
-	dict_t                  *xattr_req;
-	dht_layout_t            *layout;
-	size_t                   size;
-	ino_t                    ia_ino;
-	xlator_t                *src_hashed, *src_cached;
-	xlator_t                *dst_hashed, *dst_cached;
-	xlator_t                *cached_subvol;
-	xlator_t                *hashed_subvol;
-	char                     need_selfheal;
+        dict_t                  *xattr;
+        dict_t                  *xattr_req;
+        dht_layout_t            *layout;
+        size_t                   size;
+        ino_t                    ia_ino;
+        xlator_t                *src_hashed, *src_cached;
+        xlator_t                *dst_hashed, *dst_cached;
+        xlator_t                *cached_subvol;
+        xlator_t                *hashed_subvol;
+        char                     need_selfheal;
         int                      file_count;
         int                      dir_count;
         call_frame_t            *main_frame;
-	struct {
-		fop_mknod_cbk_t  linkfile_cbk;
-		struct iatt      stbuf;
-		loc_t            loc;
-		inode_t         *inode;
-		dict_t          *xattr;
-		xlator_t        *srcvol;
-	} linkfile;
-	struct {
-		uint32_t         hole_cnt;
-		uint32_t         overlaps_cnt;
-		uint32_t         missing;
-		uint32_t         down;
-		uint32_t         misc;
-		dht_selfheal_dir_cbk_t   dir_cbk;
-		dht_layout_t    *layout;
-	} selfheal;
+        int                      fop_succeeded;
+        struct {
+                fop_mknod_cbk_t  linkfile_cbk;
+                struct iatt      stbuf;
+                loc_t            loc;
+                inode_t         *inode;
+                dict_t          *xattr;
+                xlator_t        *srcvol;
+        } linkfile;
+        struct {
+                uint32_t         hole_cnt;
+                uint32_t         overlaps_cnt;
+                uint32_t         down;
+                uint32_t         misc;
+                dht_selfheal_dir_cbk_t   dir_cbk;
+                dht_layout_t    *layout;
+        } selfheal;
         uint32_t                 uid;
         uint32_t                 gid;
 
-	/* needed by nufa */
-	int32_t flags;
-	mode_t  mode;
-	dev_t   rdev;
+        /* needed by nufa */
+        int32_t flags;
+        mode_t  mode;
+        dev_t   rdev;
+        mode_t  umask;
 
         /* need for file-info */
-        char   *pathinfo;
+        char   *xattr_val;
         char   *key;
 
+        /* which xattr request? */
+        char xsel[256];
+        int32_t alloc_len;
+
         char   *newpath;
 
         /* gfid related */
         uuid_t  gfid;
+
+        /*Marker Related*/
+        struct marker_str    marker;
+
+        /* flag used to make sure we need to return estale in
+           {lookup,revalidate}_cbk */
+        char return_estale;
+        char need_lookup_everywhere;
+
+        glusterfs_fop_t      fop;
+
+        gf_boolean_t     linked;
+        xlator_t        *link_subvol;
+
+        struct dht_rebalance_ rebalance;
+        xlator_t        *first_up_subvol;
+
 };
 typedef struct dht_local dht_local_t;
 
 /* du - disk-usage */
 struct dht_du {
         double   avail_percent;
+	double   avail_inodes;
         uint64_t avail_space;
         uint32_t log;
 };
 typedef struct dht_du dht_du_t;
 
+enum gf_defrag_type {
+        GF_DEFRAG_CMD_START = 1,
+        GF_DEFRAG_CMD_STOP = 1 + 1,
+        GF_DEFRAG_CMD_STATUS = 1 + 2,
+        GF_DEFRAG_CMD_START_LAYOUT_FIX = 1 + 3,
+        GF_DEFRAG_CMD_START_FORCE = 1 + 4,
+};
+typedef enum gf_defrag_type gf_defrag_type;
+
+enum gf_defrag_status_t {
+        GF_DEFRAG_STATUS_NOT_STARTED,
+        GF_DEFRAG_STATUS_STARTED,
+        GF_DEFRAG_STATUS_STOPPED,
+        GF_DEFRAG_STATUS_COMPLETE,
+        GF_DEFRAG_STATUS_FAILED,
+        GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED,
+        GF_DEFRAG_STATUS_LAYOUT_FIX_STOPPED,
+        GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE,
+        GF_DEFRAG_STATUS_LAYOUT_FIX_FAILED,
+};
+typedef enum gf_defrag_status_t gf_defrag_status_t;
+
+typedef struct gf_defrag_pattern_list gf_defrag_pattern_list_t;
+
+struct gf_defrag_pattern_list {
+        char                       path_pattern[256];
+        uint64_t                   size;
+        gf_defrag_pattern_list_t  *next;
+};
+
+struct gf_defrag_info_ {
+        uint64_t                     total_files;
+        uint64_t                     total_data;
+        uint64_t                     num_files_lookedup;
+        uint64_t                     total_failures;
+        uint64_t                     skipped;
+        gf_lock_t                    lock;
+        int                          cmd;
+        pthread_t                    th;
+        gf_defrag_status_t           defrag_status;
+        struct rpc_clnt             *rpc;
+        uint32_t                     connected;
+        uint32_t                     is_exiting;
+        pid_t                        pid;
+        inode_t                     *root_inode;
+        uuid_t                       node_uuid;
+        struct timeval               start_time;
+        gf_boolean_t                 stats;
+        gf_defrag_pattern_list_t    *defrag_pattern;
+};
+
+typedef struct gf_defrag_info_ gf_defrag_info_t;
+
 struct dht_conf {
-	gf_lock_t      subvolume_lock;
+        gf_lock_t      subvolume_lock;
         int            subvolume_cnt;
         xlator_t     **subvolumes;
-	char          *subvolume_status;
+        char          *subvolume_status;
         int           *last_event;
-	dht_layout_t **file_layouts;
-	dht_layout_t **dir_layouts;
-	dht_layout_t  *default_dir_layout;
-	gf_boolean_t   search_unhashed;
-	int            gen;
+        dht_layout_t **file_layouts;
+        dht_layout_t **dir_layouts;
+        gf_boolean_t   search_unhashed;
+        int            gen;
         dht_du_t      *du_stats;
-        uint64_t       min_free_disk;
+        double         min_free_disk;
+	double         min_free_inodes;
         char           disk_unit;
         int32_t        refresh_interval;
         gf_boolean_t   unhashed_sticky_bit;
-	struct timeval last_stat_fetch;
+        struct timeval last_stat_fetch;
         gf_lock_t      layout_lock;
         void          *private;     /* Can be used by wrapper xlators over
                                        dht */
         gf_boolean_t   use_readdirp;
+        char           vol_uuid[UUID_SIZE + 1];
+        gf_boolean_t   assert_no_child_down;
+        time_t        *subvol_up_time;
+
+        /* This is the count used as the distribute layout for a directory */
+        /* Will be a global flag to control the layout spread count */
+        uint32_t       dir_spread_cnt;
+
+        /* to keep track of nodes which are decomissioned */
+        xlator_t     **decommissioned_bricks;
+        int            decommission_in_progress;
+        int            decommission_subvols_cnt;
+
+        /* defrag related */
+        gf_defrag_info_t *defrag;
+
+        /* Request to filter directory entries in readdir request */
+
+        gf_boolean_t    readdir_optimize;
+
+        /* Support regex-based name reinterpretation. */
+        regex_t         rsync_regex;
+        gf_boolean_t    rsync_regex_valid;
+        regex_t         extra_regex;
+        gf_boolean_t    extra_regex_valid;
+
+        /* Support variable xattr names. */
+        char            *xattr_name;
+        char            *link_xattr_name;
+        char            *wild_xattr_name;
 };
 typedef struct dht_conf dht_conf_t;
 
 
 struct dht_disk_layout {
-	uint32_t           cnt;
-	uint32_t           type;
-	struct {
-		uint32_t   start;
-		uint32_t   stop;
-	} list[1];
+        uint32_t           cnt;
+        uint32_t           type;
+        struct {
+                uint32_t   start;
+                uint32_t   stop;
+        } list[1];
 };
 typedef struct dht_disk_layout dht_disk_layout_t;
 
-#define WIPE(statp) do { typeof(*statp) z = {0,}; if (statp) *statp = z; } while (0)
+typedef enum {
+        GF_DHT_MIGRATE_DATA,
+        GF_DHT_MIGRATE_DATA_EVEN_IF_LINK_EXISTS,
+        GF_DHT_MIGRATE_HARDLINK,
+        GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS
+} gf_dht_migrate_data_type_t;
 
 #define ENTRY_MISSING(op_ret, op_errno) (op_ret == -1 && op_errno == ENOENT)
 
-#define is_fs_root(loc) (strcmp (loc->path, "/") == 0)
-
-#define is_revalidate(loc) (inode_ctx_get (loc->inode, this, NULL) == 0)
+#define is_revalidate(loc) (dht_inode_ctx_layout_get (loc->inode, this, NULL) == 0)
 
 #define is_last_call(cnt) (cnt == 0)
 
-#define DHT_LINKFILE_MODE (S_ISVTX)
-#define check_is_linkfile(i,s,x) (                                      \
-                ((st_mode_from_ia (s->ia_prot, s->ia_type) & ~S_IFMT)   \
-                 == DHT_LINKFILE_MODE) &&                               \
-                (s->ia_size == 0))
+#define DHT_MIGRATION_IN_PROGRESS 1
+#define DHT_MIGRATION_COMPLETED   2
+
+#define check_is_linkfile(i,s,x,n) (IS_DHT_LINKFILE_MODE (s) && dict_get (x, n))
+
+#define IS_DHT_MIGRATION_PHASE2(buf)  (                                 \
+                IA_ISREG ((buf)->ia_type) &&                            \
+                ((st_mode_from_ia ((buf)->ia_prot, (buf)->ia_type) &    \
+                  ~S_IFMT) == DHT_LINKFILE_MODE))
+
+#define IS_DHT_MIGRATION_PHASE1(buf)  (                                 \
+                IA_ISREG ((buf)->ia_type) &&                            \
+                ((buf)->ia_prot.sticky == 1) &&                         \
+                ((buf)->ia_prot.sgid == 1))
+
+#define DHT_STRIP_PHASE1_FLAGS(buf)  do {                       \
+                if ((buf) && IS_DHT_MIGRATION_PHASE1(buf)) {    \
+                        (buf)->ia_prot.sticky = 0;              \
+                        (buf)->ia_prot.sgid = 0;                \
+                }                                               \
+        } while (0)
+
+#define dht_inode_missing(op_errno) (op_errno == ENOENT || op_errno == ESTALE)
 
 #define check_is_dir(i,s,x) (IA_ISDIR(s->ia_type))
 
 #define layout_is_sane(layout) ((layout) && (layout->cnt > 0))
 
 #define DHT_STACK_UNWIND(fop, frame, params ...) do {           \
-		dht_local_t *__local = NULL;                    \
-                xlator_t *__xl = NULL;                          \
+                dht_local_t *__local = NULL;                    \
+                xlator_t    *__xl    = NULL;                    \
                 if (frame) {                                    \
-                        __xl = frame->this;                     \
-                        __local = frame->local;                 \
+                        __xl         = frame->this;             \
+                        __local      = frame->local;            \
                         frame->local = NULL;                    \
                 }                                               \
-		STACK_UNWIND_STRICT (fop, frame, params);       \
-		dht_local_wipe (__xl, __local);                 \
-	} while (0)
-
-#define DHT_STACK_DESTROY(frame) do {		       \
-		dht_local_t *__local = NULL;           \
-                xlator_t *__xl = NULL;                 \
-                __xl = frame->this;                    \
-		__local = frame->local;                \
-		frame->local = NULL;		       \
-		STACK_DESTROY (frame->root);	       \
-		dht_local_wipe (__xl, __local);	       \
-	} while (0)
-
-dht_layout_t *dht_layout_new (xlator_t *this, int cnt);
-dht_layout_t *dht_layout_get (xlator_t *this, inode_t *inode);
-dht_layout_t *dht_layout_for_subvol (xlator_t *this, xlator_t *subvol);
-xlator_t *dht_layout_search (xlator_t *this, dht_layout_t *layout,
-			     const char *name);
-int dht_layout_normalize (xlator_t *this, loc_t *loc, dht_layout_t *layout);
-int dht_layout_anomalies (xlator_t *this, loc_t *loc, dht_layout_t *layout,
-			  uint32_t *holes_p, uint32_t *overlaps_p,
-			  uint32_t *missing_p, uint32_t *down_p,
-			  uint32_t *misc_p);
-int dht_layout_dir_mismatch (xlator_t *this, dht_layout_t *layout,
-			     xlator_t *subvol, loc_t *loc, dict_t *xattr);
+                STACK_UNWIND_STRICT (fop, frame, params);       \
+                dht_local_wipe (__xl, __local);                 \
+        } while (0)
+
+#define DHT_STACK_DESTROY(frame) do {                   \
+                dht_local_t *__local = NULL;            \
+                xlator_t    *__xl    = NULL;            \
+                __xl                 = frame->this;     \
+                __local              = frame->local;    \
+                frame->local         = NULL;            \
+                STACK_DESTROY (frame->root);            \
+                dht_local_wipe (__xl, __local);         \
+        } while (0)
+
+#define DHT_UPDATE_TIME(ctx_sec, ctx_nsec, new_sec, new_nsec, inode, post) do {\
+                int32_t sec = 0;                                        \
+                sec = new_sec;                                          \
+                LOCK (&inode->lock);                                    \
+                {                                                       \
+                        new_sec = max(new_sec, ctx_sec);                \
+                        if (sec < new_sec)                              \
+                                new_nsec = ctx_nsec;                    \
+                        if (sec == new_sec)                             \
+                                new_nsec = max (new_nsec, ctx_nsec);    \
+                        if (post) {                                     \
+                                ctx_sec = new_sec;                      \
+                                ctx_nsec = new_nsec;                    \
+                        }                                               \
+                }                                                       \
+                UNLOCK (&inode->lock);                                  \
+        } while (0)
+
+#define is_greater_time(a, an, b, bn) (((a) < (b)) || (((a) == (b)) && ((an) < (bn))))
+dht_layout_t                            *dht_layout_new (xlator_t *this, int cnt);
+dht_layout_t                            *dht_layout_get (xlator_t *this, inode_t *inode);
+dht_layout_t                            *dht_layout_for_subvol (xlator_t *this, xlator_t *subvol);
+xlator_t *dht_layout_search (xlator_t   *this, dht_layout_t *layout,
+                             const char *name);
+int                                      dht_layout_normalize (xlator_t *this, loc_t *loc, dht_layout_t *layout);
+int dht_layout_anomalies (xlator_t      *this, loc_t *loc, dht_layout_t *layout,
+                          uint32_t      *holes_p, uint32_t *overlaps_p,
+                          uint32_t      *missing_p, uint32_t *down_p,
+                          uint32_t      *misc_p, uint32_t *no_space_p);
+int dht_layout_dir_mismatch (xlator_t   *this, dht_layout_t *layout,
+                             xlator_t   *subvol, loc_t *loc, dict_t *xattr);
 
 xlator_t *dht_linkfile_subvol (xlator_t *this, inode_t *inode,
-			       struct iatt *buf, dict_t *xattr);
-int dht_linkfile_unlink (call_frame_t *frame, xlator_t *this,
-			 xlator_t *subvol, loc_t *loc);
+                               struct iatt *buf, dict_t *xattr);
+int dht_linkfile_unlink (call_frame_t      *frame, xlator_t *this,
+                         xlator_t          *subvol, loc_t *loc);
 
 int dht_layouts_init (xlator_t *this, dht_conf_t *conf);
 int dht_layout_merge (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
-		      int op_ret, int op_errno, dict_t *xattr);
+                      int       op_ret, int op_errno, dict_t *xattr);
 
 int dht_disk_layout_extract (xlator_t *this, dht_layout_t *layout,
-			     int pos, int32_t **disk_layout_p);
-int dht_disk_layout_merge (xlator_t *this, dht_layout_t *layout,
-			   int pos, void *disk_layout_raw);
+                             int       pos, int32_t **disk_layout_p);
+int dht_disk_layout_merge (xlator_t   *this, dht_layout_t *layout,
+                           int         pos, void *disk_layout_raw, int disk_layout_len);
 
 
 int dht_frame_return (call_frame_t *frame);
 
-int dht_itransform (xlator_t *this, xlator_t *subvol, uint64_t x, uint64_t *y);
+int                             dht_itransform (xlator_t *this, xlator_t *subvol, uint64_t x, uint64_t *y);
 int dht_deitransform (xlator_t *this, uint64_t y, xlator_t **subvol,
-		      uint64_t *x);
+                      uint64_t *x);
 
 void dht_local_wipe (xlator_t *this, dht_local_t *local);
-dht_local_t *dht_local_init (call_frame_t *frame);
-int dht_iatt_merge (xlator_t *this, struct iatt *to, struct iatt *from,
-		    xlator_t *subvol);
+dht_local_t *dht_local_init (call_frame_t    *frame, loc_t *loc, fd_t *fd,
+                             glusterfs_fop_t  fop);
+int dht_iatt_merge (xlator_t                 *this, struct iatt *to, struct iatt *from,
+                    xlator_t                 *subvol);
 
 xlator_t *dht_subvol_get_hashed (xlator_t *this, loc_t *loc);
 xlator_t *dht_subvol_get_cached (xlator_t *this, inode_t *inode);
 xlator_t *dht_subvol_next (xlator_t *this, xlator_t *prev);
-int dht_subvol_cnt (xlator_t *this, xlator_t *subvol);
+xlator_t *dht_subvol_next_available (xlator_t *this, xlator_t *prev);
+int       dht_subvol_cnt (xlator_t *this, xlator_t *subvol);
 
-int dht_hash_compute (int type, const char *name, uint32_t *hash_p);
+int dht_hash_compute (xlator_t *this, int type, const char *name, uint32_t *hash_p);
 
-int dht_linkfile_create (call_frame_t *frame, fop_mknod_cbk_t linkfile_cbk,
-			 xlator_t *tovol, xlator_t *fromvol, loc_t *loc);
-int dht_lookup_directory (call_frame_t *frame, xlator_t *this, loc_t *loc);
-int dht_lookup_everywhere (call_frame_t *frame, xlator_t *this, loc_t *loc);
+int dht_linkfile_create (call_frame_t    *frame, fop_mknod_cbk_t linkfile_cbk,
+                         xlator_t        *this, xlator_t *tovol,
+                         xlator_t        *fromvol, loc_t *loc);
+int                                       dht_lookup_directory (call_frame_t *frame, xlator_t *this, loc_t *loc);
+int                                       dht_lookup_everywhere (call_frame_t *frame, xlator_t *this, loc_t *loc);
 int
-dht_selfheal_directory (call_frame_t *frame, dht_selfheal_dir_cbk_t cbk,
-			loc_t *loc, dht_layout_t *layout);
+dht_selfheal_directory (call_frame_t     *frame, dht_selfheal_dir_cbk_t cbk,
+                        loc_t            *loc, dht_layout_t *layout);
 int
 dht_selfheal_new_directory (call_frame_t *frame, dht_selfheal_dir_cbk_t cbk,
-			    dht_layout_t *layout);
+                            dht_layout_t *layout);
 int
-dht_selfheal_restore (call_frame_t *frame, dht_selfheal_dir_cbk_t cbk,
-		      loc_t *loc, dht_layout_t *layout);
+dht_selfheal_restore (call_frame_t       *frame, dht_selfheal_dir_cbk_t cbk,
+                      loc_t              *loc, dht_layout_t *layout);
 int
 dht_layout_sort_volname (dht_layout_t *layout);
 
-int dht_rename (call_frame_t *frame, xlator_t *this,
-		loc_t *oldloc, loc_t *newloc);
-
 int dht_get_du_info (call_frame_t *frame, xlator_t *this, loc_t *loc);
 
-int dht_is_subvol_filled (xlator_t *this, xlator_t *subvol);
-xlator_t *dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol);
-int dht_get_du_info_for_subvol (xlator_t *this, int subvol_idx);
+gf_boolean_t dht_is_subvol_filled (xlator_t *this, xlator_t *subvol);
+xlator_t *dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol,
+                                          dht_local_t *layout);
+int       dht_get_du_info_for_subvol (xlator_t *this, int subvol_idx);
 
 int dht_layout_preset (xlator_t *this, xlator_t *subvol, inode_t *inode);
-int dht_layout_set (xlator_t *this, inode_t *inode, dht_layout_t *layout);
-void dht_layout_unref (xlator_t *this, dht_layout_t *layout);
+int           dht_layout_set (xlator_t *this, inode_t *inode, dht_layout_t *layout);;
+void          dht_layout_unref (xlator_t *this, dht_layout_t *layout);
 dht_layout_t *dht_layout_ref (xlator_t *this, dht_layout_t *layout);
-xlator_t *dht_first_up_subvol (xlator_t *this);
-xlator_t *dht_last_up_subvol (xlator_t *this);
-int dht_frame_su_do (call_frame_t *frame);
-int dht_frame_su_undo (call_frame_t *frame);
+xlator_t     *dht_first_up_subvol (xlator_t *this);
+xlator_t     *dht_last_up_subvol (xlator_t *this);
 
 int dht_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name);
 
-int dht_filter_loc_subvol_key (xlator_t *this, loc_t *loc, loc_t *new_loc,
+int dht_filter_loc_subvol_key (xlator_t  *this, loc_t *loc, loc_t *new_loc,
                                xlator_t **subvol);
 
-#endif /* _DHT_H */
+int                                     dht_rename_cleanup (call_frame_t *frame);
+int dht_rename_links_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                          int32_t           op_ret, int32_t op_errno,
+                          inode_t          *inode, struct iatt *stbuf,
+                          struct iatt      *preparent, struct iatt *postparent,
+                          dict_t *xdata);
+
+int dht_fix_directory_layout (call_frame_t *frame,
+                              dht_selfheal_dir_cbk_t  dir_cbk,
+                              dht_layout_t           *layout);
+
+int dht_init_subvolumes (xlator_t *this, dht_conf_t *conf);
+
+/* migration/rebalance */
+int dht_start_rebalance_task (xlator_t *this, call_frame_t *frame);
+
+int dht_rebalance_in_progress_check (xlator_t *this, call_frame_t *frame);
+int dht_rebalance_complete_check (xlator_t *this, call_frame_t *frame);
+
+
+/* FOPS */
+int32_t dht_lookup (call_frame_t *frame,
+                    xlator_t *this,
+                    loc_t    *loc,
+                    dict_t   *xattr_req);
+
+int32_t dht_stat (call_frame_t *frame,
+                  xlator_t *this,
+                  loc_t    *loc, dict_t *xdata);
+
+int32_t dht_fstat (call_frame_t *frame,
+                   xlator_t *this,
+                   fd_t     *fd, dict_t *xdata);
+
+int32_t dht_truncate (call_frame_t *frame,
+                      xlator_t *this,
+                      loc_t    *loc,
+                      off_t     offset, dict_t *xdata);
+
+int32_t dht_ftruncate (call_frame_t *frame,
+                       xlator_t *this,
+                       fd_t     *fd,
+                       off_t     offset, dict_t *xdata);
+
+int32_t dht_access (call_frame_t *frame,
+                    xlator_t *this,
+                    loc_t    *loc,
+                    int32_t   mask, dict_t *xdata);
+
+int32_t dht_readlink (call_frame_t *frame,
+                      xlator_t *this,
+                      loc_t    *loc,
+                      size_t    size, dict_t *xdata);
+
+int32_t dht_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                   mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata);
+
+int32_t dht_mkdir (call_frame_t *frame, xlator_t *this,
+                   loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata);
+
+int32_t dht_unlink (call_frame_t *frame,
+                    xlator_t *this,
+                    loc_t    *loc, int xflag, dict_t *xdata);
+
+int32_t dht_rmdir (call_frame_t *frame, xlator_t *this,
+                   loc_t    *loc, int flags, dict_t *xdata);
+
+int32_t dht_symlink (call_frame_t   *frame, xlator_t *this,
+                     const char *linkpath, loc_t *loc, mode_t umask,
+                     dict_t *xdata);
+
+int32_t dht_rename (call_frame_t *frame,
+                    xlator_t *this,
+                    loc_t    *oldloc,
+                    loc_t    *newloc, dict_t *xdata);
+
+int32_t dht_link (call_frame_t *frame,
+                  xlator_t *this,
+                  loc_t    *oldloc,
+                  loc_t    *newloc, dict_t *xdata);
+
+int32_t dht_create (call_frame_t *frame, xlator_t *this,
+                    loc_t    *loc, int32_t flags, mode_t mode,
+                    mode_t umask, fd_t     *fd, dict_t *params);
+
+int32_t dht_open (call_frame_t *frame,
+                  xlator_t *this,
+                  loc_t    *loc,
+                  int32_t   flags, fd_t *fd, dict_t *xdata);
+
+int32_t dht_readv (call_frame_t *frame,
+                   xlator_t *this,
+                   fd_t     *fd,
+                   size_t    size,
+                   off_t     offset, uint32_t flags, dict_t *xdata);
+
+int32_t dht_writev (call_frame_t      *frame,
+                    xlator_t      *this,
+                    fd_t          *fd,
+                    struct iovec  *vector,
+                    int32_t        count,
+                    off_t          offset,
+                    uint32_t       flags,
+                    struct iobref *iobref, dict_t *xdata);
+
+int32_t dht_flush (call_frame_t *frame,
+                   xlator_t *this,
+                   fd_t     *fd, dict_t *xdata);
+
+int32_t dht_fsync (call_frame_t *frame,
+                   xlator_t *this,
+                   fd_t     *fd,
+                   int32_t   datasync, dict_t *xdata);
+
+int32_t dht_opendir (call_frame_t *frame,
+                     xlator_t *this,
+                     loc_t    *loc, fd_t *fd, dict_t *xdata);
+
+int32_t dht_fsyncdir (call_frame_t *frame,
+                      xlator_t *this,
+                      fd_t     *fd,
+                      int32_t   datasync, dict_t *xdata);
+
+int32_t dht_statfs (call_frame_t *frame,
+                    xlator_t *this,
+                    loc_t    *loc, dict_t *xdata);
+
+int32_t dht_setxattr (call_frame_t *frame,
+                      xlator_t *this,
+                      loc_t    *loc,
+                      dict_t   *dict,
+                      int32_t   flags, dict_t *xdata);
+
+int32_t dht_getxattr (call_frame_t   *frame,
+                      xlator_t   *this,
+                      loc_t      *loc,
+                      const char *name, dict_t *xdata);
+
+int32_t dht_fsetxattr (call_frame_t *frame,
+                       xlator_t *this,
+                       fd_t     *fd,
+                       dict_t   *dict,
+                       int32_t   flags, dict_t *xdata);
+
+int32_t dht_fgetxattr (call_frame_t   *frame,
+                       xlator_t   *this,
+                       fd_t       *fd,
+                       const char *name, dict_t *xdata);
+
+int32_t dht_removexattr (call_frame_t   *frame,
+                         xlator_t   *this,
+                         loc_t      *loc,
+                         const char *name, dict_t *xdata);
+int32_t dht_fremovexattr (call_frame_t   *frame,
+                          xlator_t   *this,
+                          fd_t      *fd,
+                          const char *name, dict_t *xdata);
+
+int32_t dht_lk (call_frame_t        *frame,
+                xlator_t        *this,
+                fd_t            *fd,
+                int32_t          cmd,
+                struct gf_flock *flock, dict_t *xdata);
+
+int32_t dht_inodelk (call_frame_t *frame, xlator_t *this,
+                     const char      *volume, loc_t *loc, int32_t cmd,
+                     struct gf_flock *flock, dict_t *xdata);
+
+int32_t dht_finodelk (call_frame_t        *frame, xlator_t *this,
+                      const char      *volume, fd_t *fd, int32_t cmd,
+                      struct gf_flock *flock, dict_t *xdata);
+
+int32_t dht_entrylk (call_frame_t    *frame, xlator_t *this,
+                     const char  *volume, loc_t *loc, const char *basename,
+                     entrylk_cmd  cmd, entrylk_type type, dict_t *xdata);
+
+int32_t dht_fentrylk (call_frame_t    *frame, xlator_t *this,
+                      const char  *volume, fd_t *fd, const char *basename,
+                      entrylk_cmd  cmd, entrylk_type type, dict_t *xdata);
+
+int32_t dht_readdir (call_frame_t  *frame,
+                     xlator_t *this,
+                     fd_t     *fd,
+                     size_t    size, off_t off, dict_t *xdata);
+
+int32_t dht_readdirp (call_frame_t *frame,
+                      xlator_t *this,
+                      fd_t     *fd,
+                      size_t    size, off_t off, dict_t *dict);
+
+int32_t dht_xattrop (call_frame_t           *frame,
+                     xlator_t           *this,
+                     loc_t              *loc,
+                     gf_xattrop_flags_t  flags,
+                     dict_t             *dict, dict_t *xdata);
+
+int32_t dht_fxattrop (call_frame_t *frame,
+                      xlator_t           *this,
+                      fd_t               *fd,
+                      gf_xattrop_flags_t  flags,
+                      dict_t             *dict, dict_t *xdata);
+
+int32_t dht_forget (xlator_t *this, inode_t *inode);
+int32_t dht_setattr (call_frame_t  *frame, xlator_t *this, loc_t *loc,
+                     struct iatt   *stbuf, int32_t valid, dict_t *xdata);
+int32_t dht_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                      struct iatt  *stbuf, int32_t valid, dict_t *xdata);
+int32_t dht_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd,
+		      int32_t mode, off_t offset, size_t len, dict_t *xdata);
+int32_t dht_discard(call_frame_t *frame, xlator_t *this, fd_t *fd,
+		    off_t offset, size_t len, dict_t *xdata);
+int32_t dht_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd,
+                    off_t offset, off_t len, dict_t *xdata);
+
+int32_t dht_init (xlator_t *this);
+void    dht_fini (xlator_t *this);
+int     dht_reconfigure (xlator_t *this, dict_t *options);
+int32_t dht_notify (xlator_t *this, int32_t event, void *data, ...);
+
+/* definitions for nufa/switch */
+int dht_revalidate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int op_ret, int op_errno, inode_t *inode,
+                        struct iatt *stbuf, dict_t *xattr,
+                        struct iatt *postparent);
+int dht_lookup_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int op_ret, int op_errno, inode_t *inode,
+                        struct iatt *stbuf, dict_t *xattr,
+                        struct iatt *postparent);
+int dht_lookup_linkfile_cbk (call_frame_t *frame, void *cookie,
+                             xlator_t *this, int op_ret, int op_errno,
+                             inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+                             struct iatt *postparent);
+int dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int op_ret, int op_errno,
+                    inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+                    struct iatt *postparent);
+int dht_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int op_ret, int op_errno,
+                    fd_t *fd, inode_t *inode, struct iatt *stbuf,
+                    struct iatt *preparent, struct iatt *postparent,
+                    dict_t *xdata);
+int dht_newfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int op_ret, int op_errno,
+                     inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
+                     struct iatt *postparent, dict_t *xdata);
+
+int
+gf_defrag_status_get (gf_defrag_info_t *defrag, dict_t *dict);
+
+int
+gf_defrag_stop (gf_defrag_info_t *defrag, gf_defrag_status_t status,
+                dict_t *output);
+
+void*
+gf_defrag_start (void *this);
+
+int32_t
+gf_defrag_handle_hardlink (xlator_t *this, loc_t *loc, dict_t  *xattrs,
+                           struct iatt *stbuf);
+int
+dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
+                 int flag);
+int
+dht_inode_ctx_layout_get (inode_t *inode, xlator_t *this,
+                          dht_layout_t **layout_int);
+int
+dht_inode_ctx_layout_set (inode_t *inode, xlator_t *this,
+                          dht_layout_t* layout_int);
+int
+dht_inode_ctx_time_update (inode_t *inode, xlator_t *this, struct iatt *stat,
+                           int32_t update_ctx);
+
+int dht_inode_ctx_get (inode_t *inode, xlator_t *this, dht_inode_ctx_t **ctx);
+int dht_inode_ctx_set (inode_t *inode, xlator_t *this, dht_inode_ctx_t *ctx);
+int
+dht_dir_attr_heal (void *data);
+int
+dht_dir_attr_heal_done (int ret, call_frame_t *sync_frame, void *data);
+int
+dht_dir_has_layout (dict_t *xattr, char *name);
+gf_boolean_t
+dht_is_subvol_in_layout (dht_layout_t *layout, xlator_t *xlator);
+xlator_t *
+dht_subvol_with_free_space_inodes (xlator_t *this, xlator_t *subvol,
+                                   dht_layout_t *layout);
+xlator_t *
+dht_subvol_maxspace_nonzeroinode (xlator_t *this, xlator_t *subvol,
+                                  dht_layout_t *layout);
+int
+dht_linkfile_attr_heal (call_frame_t *frame, xlator_t *this);
+
+void
+dht_layout_dump (dht_layout_t  *layout, const char *prefix);
+int32_t
+dht_priv_dump (xlator_t *this);
+int32_t
+dht_inodectx_dump (xlator_t *this, inode_t *inode);
+
+int
+dht_inode_ctx_get1 (xlator_t *this, inode_t *inode, xlator_t **subvol);
+
+#endif/* _DHT_H */
diff --git a/xlators/cluster/dht/src/dht-diskusage.c b/xlators/cluster/dht/src/dht-diskusage.c
index 14e278b12..fe3955ecb 100644
--- a/xlators/cluster/dht/src/dht-diskusage.c
+++ b/xlators/cluster/dht/src/dht-diskusage.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -33,50 +24,70 @@
 #include <sys/time.h>
 
 
-int 
+int
 dht_du_info_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                 int op_ret, int op_errno, struct statvfs *statvfs)
+		 int op_ret, int op_errno, struct statvfs *statvfs,
+                 dict_t *xdata)
 {
 	dht_conf_t    *conf         = NULL;
-        call_frame_t  *prev          = NULL;
+	call_frame_t  *prev          = NULL;
 	int            this_call_cnt = 0;
-        int            i = 0;
-        double         percent = 0;
-        uint64_t       bytes = 0;
-
-        conf = this->private;
-        prev = cookie;
-
-        if (op_ret == -1) 
-                goto out;
+	int            i = 0;
+	double         percent = 0;
+	double         percent_inodes = 0;
+	uint64_t       bytes = 0;
 
-        if (statvfs && statvfs->f_blocks) {
-                percent = (statvfs->f_bfree * 100) / statvfs->f_blocks;
-                bytes = (statvfs->f_bfree * statvfs->f_frsize);
-        }
-        
-        LOCK (&conf->subvolume_lock);
-        {
-                for (i = 0; i < conf->subvolume_cnt; i++)
-                        if (prev->this == conf->subvolumes[i]) {
-                                conf->du_stats[i].avail_percent = percent;
-                                conf->du_stats[i].avail_space   = bytes;
-                                gf_log (this->name, GF_LOG_DEBUG,
-                                        "on subvolume '%s': avail_percent is: "
-                                        "%.2f and avail_space is: %"PRIu64"",
-                                        prev->this->name, 
-                                        conf->du_stats[i].avail_percent,
-                                        conf->du_stats[i].avail_space);
-                        }
-        }
-        UNLOCK (&conf->subvolume_lock);
-
- out:
+	conf = this->private;
+	prev = cookie;
+
+	if (op_ret == -1) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"failed to get disk info from %s", prev->this->name);
+		goto out;
+	}
+
+	if (statvfs && statvfs->f_blocks) {
+		percent = (statvfs->f_bavail * 100) / statvfs->f_blocks;
+		bytes = (statvfs->f_bavail * statvfs->f_frsize);
+	}
+
+	if (statvfs && statvfs->f_files) {
+		percent_inodes = (statvfs->f_ffree * 100) / statvfs->f_files;
+	} else {
+		/* set percent inodes to 100 for dynamically allocated inode filesystems
+		   this logic holds good so that, distribute has nothing to worry about
+		   total inodes rather let the 'create()' to be scheduled on the hashed
+		   subvol regardless of the total inodes. since we have no awareness on
+		   loosing inodes this logic fits well
+		*/
+		percent_inodes = 100;
+	}
+
+	LOCK (&conf->subvolume_lock);
+	{
+		for (i = 0; i < conf->subvolume_cnt; i++)
+			if (prev->this == conf->subvolumes[i]) {
+				conf->du_stats[i].avail_percent = percent;
+				conf->du_stats[i].avail_space   = bytes;
+				conf->du_stats[i].avail_inodes  = percent_inodes;
+				gf_log (this->name, GF_LOG_DEBUG,
+					"on subvolume '%s': avail_percent is: "
+					"%.2f and avail_space is: %"PRIu64" "
+					"and avail_inodes is: %.2f",
+					prev->this->name,
+					conf->du_stats[i].avail_percent,
+					conf->du_stats[i].avail_space,
+					conf->du_stats[i].avail_inodes);
+			}
+	}
+	UNLOCK (&conf->subvolume_lock);
+
+out:
 	this_call_cnt = dht_frame_return (frame);
 	if (is_last_call (this_call_cnt))
 		DHT_STACK_DESTROY (frame);
 
-        return 0;
+	return 0;
 }
 
 int
@@ -85,179 +96,319 @@ dht_get_du_info_for_subvol (xlator_t *this, int subvol_idx)
 	dht_conf_t    *conf         = NULL;
 	call_frame_t  *statfs_frame = NULL;
 	dht_local_t   *statfs_local = NULL;
-        call_pool_t   *pool         = NULL;
+	call_pool_t   *pool         = NULL;
+	loc_t          tmp_loc      = {0,};
 
 	conf = this->private;
-        pool = this->ctx->pool;
-
-        statfs_frame = create_frame (this, pool);
-        if (!statfs_frame) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
-        }
-        
-        statfs_local = dht_local_init (statfs_frame);
-        if (!statfs_local) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
-        }
-        
-        loc_t tmp_loc = { .inode = NULL,
-                          .path = "/",
-        };
-        
-        statfs_local->call_cnt = 1;
-        STACK_WIND (statfs_frame, dht_du_info_cbk,
-                    conf->subvolumes[subvol_idx],
-                    conf->subvolumes[subvol_idx]->fops->statfs,
-                    &tmp_loc);
-
-        return 0;
- err:
+	pool = this->ctx->pool;
+
+	statfs_frame = create_frame (this, pool);
+	if (!statfs_frame) {
+		goto err;
+	}
+
+	/* local->fop value is not used in this case */
+	statfs_local = dht_local_init (statfs_frame, NULL, NULL,
+				       GF_FOP_MAXVALUE);
+	if (!statfs_local) {
+		goto err;
+	}
+
+        /* make it root gfid, should be enough to get the proper info back */
+        tmp_loc.gfid[15] = 1;
+
+	statfs_local->call_cnt = 1;
+	STACK_WIND (statfs_frame, dht_du_info_cbk,
+		    conf->subvolumes[subvol_idx],
+		    conf->subvolumes[subvol_idx]->fops->statfs,
+		    &tmp_loc, NULL);
+
+	return 0;
+err:
 	if (statfs_frame)
 		DHT_STACK_DESTROY (statfs_frame);
-        
-        return -1;
+
+	return -1;
 }
 
 int
 dht_get_du_info (call_frame_t *frame, xlator_t *this, loc_t *loc)
 {
-        int            i = 0;
+	int            i            = 0;
 	dht_conf_t    *conf         = NULL;
 	call_frame_t  *statfs_frame = NULL;
 	dht_local_t   *statfs_local = NULL;
-        struct timeval tv = {0,};
+	struct timeval tv           = {0,};
+        loc_t          tmp_loc      = {0,};
 
 	conf  = this->private;
 
 	gettimeofday (&tv, NULL);
-	if (tv.tv_sec > (conf->refresh_interval 
-			 + conf->last_stat_fetch.tv_sec)) {
 
-                statfs_frame = copy_frame (frame);
-                if (!statfs_frame) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        goto err;
-                }
+        /* make it root gfid, should be enough to get the proper
+           info back */
+        tmp_loc.gfid[15] = 1;
 
-                statfs_local = dht_local_init (statfs_frame);
-                if (!statfs_local) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        goto err;
-                }
-
-                loc_copy (&statfs_local->loc, loc);
-                loc_t tmp_loc = { .inode = NULL,
-                              .path = "/",
-                };
-                
-                statfs_local->call_cnt = conf->subvolume_cnt;
-                for (i = 0; i < conf->subvolume_cnt; i++) {
-                        STACK_WIND (statfs_frame, dht_du_info_cbk,
-                                    conf->subvolumes[i],
-                                    conf->subvolumes[i]->fops->statfs,
-                                    &tmp_loc);
-                }
+	if (tv.tv_sec > (conf->refresh_interval
+			 + conf->last_stat_fetch.tv_sec)) {
 
-                conf->last_stat_fetch.tv_sec = tv.tv_sec;
-        }
-        return 0;
+		statfs_frame = copy_frame (frame);
+		if (!statfs_frame) {
+			goto err;
+		}
+
+		/* In this case, 'local->fop' is not used */
+		statfs_local = dht_local_init (statfs_frame, loc, NULL,
+					       GF_FOP_MAXVALUE);
+		if (!statfs_local) {
+			goto err;
+		}
+
+		statfs_local->call_cnt = conf->subvolume_cnt;
+		for (i = 0; i < conf->subvolume_cnt; i++) {
+			STACK_WIND (statfs_frame, dht_du_info_cbk,
+				    conf->subvolumes[i],
+				    conf->subvolumes[i]->fops->statfs,
+				    &tmp_loc, NULL);
+		}
+
+		conf->last_stat_fetch.tv_sec = tv.tv_sec;
+	}
+	return 0;
 err:
 	if (statfs_frame)
 		DHT_STACK_DESTROY (statfs_frame);
 
-        return -1;
+	return -1;
 }
 
 
-int
+gf_boolean_t
 dht_is_subvol_filled (xlator_t *this, xlator_t *subvol)
 {
-        int         i = 0;
-        int         subvol_filled = 0;
+	int         i = 0;
 	dht_conf_t *conf = NULL;
+	gf_boolean_t subvol_filled_inodes = _gf_false;
+	gf_boolean_t subvol_filled_space = _gf_false;
+	gf_boolean_t is_subvol_filled = _gf_false;
 
-        conf = this->private;
+	conf = this->private;
 
-        /* Check for values above specified percent or free disk */
-        LOCK (&conf->subvolume_lock);
-        {
-                for (i = 0; i < conf->subvolume_cnt; i++) {
-                        if (subvol == conf->subvolumes[i]) {
-                                if (conf->disk_unit == 'p') {
-                                        if (conf->du_stats[i].avail_percent <
-                                            conf->min_free_disk) {
-                                                subvol_filled = 1;
-                                                break;
-                                        }
-                                } else {
-                                        if (conf->du_stats[i].avail_space <
-                                            conf->min_free_disk) {
-                                                subvol_filled = 1;
-                                                break;
-                                        }
-                                }
-                        }
+	/* Check for values above specified percent or free disk */
+	LOCK (&conf->subvolume_lock);
+	{
+		for (i = 0; i < conf->subvolume_cnt; i++) {
+			if (subvol == conf->subvolumes[i]) {
+				if (conf->disk_unit == 'p') {
+					if (conf->du_stats[i].avail_percent <
+					    conf->min_free_disk) {
+						subvol_filled_space = _gf_true;
+						break;
+					}
+
+				} else {
+					if (conf->du_stats[i].avail_space <
+					    conf->min_free_disk) {
+						subvol_filled_space = _gf_true;
+						break;
+					}
+				}
+				if (conf->du_stats[i].avail_inodes <
+				    conf->min_free_inodes) {
+					subvol_filled_inodes = _gf_true;
+					break;
+				}
+			}
+		}
+	}
+	UNLOCK (&conf->subvolume_lock);
+
+	if (subvol_filled_space && conf->subvolume_status[i]) {
+		if (!(conf->du_stats[i].log++ % (GF_UNIVERSAL_ANSWER * 10))) {
+			gf_log (this->name, GF_LOG_WARNING,
+				"disk space on subvolume '%s' is getting "
+				"full (%.2f %%), consider adding more nodes",
+				subvol->name,
+				(100 - conf->du_stats[i].avail_percent));
+		}
+	}
+
+	if (subvol_filled_inodes && conf->subvolume_status[i]) {
+		if (!(conf->du_stats[i].log++ % (GF_UNIVERSAL_ANSWER * 10))) {
+			gf_log (this->name, GF_LOG_CRITICAL,
+				"inodes on subvolume '%s' are at "
+				"(%.2f %%), consider adding more nodes",
+				subvol->name,
+				(100 - conf->du_stats[i].avail_inodes));
+		}
+	}
+
+	is_subvol_filled = (subvol_filled_space || subvol_filled_inodes);
+
+	return is_subvol_filled;
+}
+
+
+/*Get the best subvolume to create the file in*/
+xlator_t *
+dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol,
+                                dht_local_t *local)
+{
+	xlator_t   *avail_subvol = NULL;
+	dht_conf_t *conf = NULL;
+        dht_layout_t *layout = NULL;
+        loc_t      *loc = NULL;
+
+	conf = this->private;
+        if (!local)
+                goto out;
+        loc = &local->loc;
+        if (!local->layout) {
+                layout = dht_layout_get (this, loc->parent);
+
+                if (!layout) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "layout missing path=%s parent=%s",
+                                loc->path, uuid_utoa (loc->parent->gfid));
+                        goto out;
                 }
+        } else {
+                layout = dht_layout_ref (this, local->layout);
         }
-        UNLOCK (&conf->subvolume_lock);
-
-        if (subvol_filled && conf->subvolume_status[i]) {
-                if (!(conf->du_stats[i].log++ % (GF_UNIVERSAL_ANSWER * 10))) {
-                        gf_log (this->name, GF_LOG_WARNING,
-                                "disk space on subvolume '%s' is getting "
-                                "full (%.2f %%), consider adding more nodes", 
-                                subvol->name, 
-                                (100 - conf->du_stats[i].avail_percent));
+
+        LOCK (&conf->subvolume_lock);
+	{
+                avail_subvol = dht_subvol_with_free_space_inodes(this, subvol,
+                                                                 layout);
+                if(!avail_subvol)
+                {
+                        avail_subvol = dht_subvol_maxspace_nonzeroinode(this,
+                                                                        subvol,
+                                                                        layout);
                 }
-        }
 
-        return subvol_filled;
+	}
+	UNLOCK (&conf->subvolume_lock);
+out:
+	if (!avail_subvol) {
+		gf_log (this->name,
+                        GF_LOG_DEBUG,
+			"no subvolume has enough free space and/or inodes\
+                         to create");
+                avail_subvol = subvol;
+	}
+
+        if (layout)
+                dht_layout_unref (this, layout);
+	return avail_subvol;
 }
 
+static inline
+int32_t dht_subvol_has_err (xlator_t *this, dht_layout_t *layout)
+{
+        int ret = -1;
+        int i   = 0;
+
+        if (!this || !layout)
+                goto out;
+
+        /* check if subvol has layout errors, before selecting it */
+        for (i = 0; i < layout->cnt; i++) {
+                if (!strcmp (layout->list[i].xlator->name, this->name) &&
+                     (layout->list[i].err != 0)) {
+                        ret = -1;
+                        goto out;
+                }
+        }
+        ret = 0;
+out:
+        return ret;
+}
+
+/*Get subvolume which has both space and inodes more than the min criteria*/
 xlator_t *
-dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol) 
+dht_subvol_with_free_space_inodes(xlator_t *this, xlator_t *subvol,
+                                  dht_layout_t *layout)
 {
-        int         i = 0;
-        double      max= 0;
-        xlator_t   *avail_subvol = NULL;
-	dht_conf_t *conf = NULL;
+        int i = 0;
+        double max = 0;
+        double max_inodes = 0;
+        int    ignore_subvol = 0;
+
+        xlator_t *avail_subvol = NULL;
+        dht_conf_t *conf = NULL;
 
         conf = this->private;
 
-        LOCK (&conf->subvolume_lock);
-        {
-                for (i = 0; i < conf->subvolume_cnt; i++) {
-                        if (conf->disk_unit == 'p') {
-                                if (conf->du_stats[i].avail_percent > max) {
-                                        max = conf->du_stats[i].avail_percent;
-                                        avail_subvol = conf->subvolumes[i];
-                                }
-                        } else {
-                                if (conf->du_stats[i].avail_space > max) {
-                                        max = conf->du_stats[i].avail_space;
-                                        avail_subvol = conf->subvolumes[i];
-                                }
+        for(i=0; i < conf->subvolume_cnt; i++) {
+                /* check if subvol has layout errors, before selecting it */
+                ignore_subvol = dht_subvol_has_err (conf->subvolumes[i],
+                                                    layout);
+                if (ignore_subvol)
+                        continue;
+
+                if ((conf->disk_unit == 'p') &&
+                    (conf->du_stats[i].avail_percent > conf->min_free_disk) &&
+                    (conf->du_stats[i].avail_inodes  > conf->min_free_inodes)) {
+                        if ((conf->du_stats[i].avail_inodes > max_inodes) ||
+                            (conf->du_stats[i].avail_percent > max)) {
+                                max = conf->du_stats[i].avail_percent;
+                                max_inodes = conf->du_stats[i].avail_inodes;
+                                avail_subvol = conf->subvolumes[i];
                         }
                 }
-        }
-        UNLOCK (&conf->subvolume_lock);
 
-        if (!avail_subvol) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "no subvolume has enough free space to create");
+                if ((conf->disk_unit != 'p') &&
+                    (conf->du_stats[i].avail_space > conf->min_free_disk) &&
+                    (conf->du_stats[i].avail_inodes  > conf->min_free_inodes)) {
+                        if ((conf->du_stats[i].avail_inodes > max_inodes) ||
+                            (conf->du_stats[i].avail_space > max)) {
+                                max = conf->du_stats[i].avail_space;
+                                max_inodes = conf->du_stats[i].avail_inodes;
+                                avail_subvol = conf->subvolumes[i];
+                        }
+                }
         }
 
-        if (max < conf->min_free_disk)
-                avail_subvol = subvol;
+        return avail_subvol;
+}
+
+
+/* Get subvol which has atleast one inode and maximum space */
+xlator_t *
+dht_subvol_maxspace_nonzeroinode (xlator_t *this, xlator_t *subvol,
+                                  dht_layout_t *layout)
+{
+        int         i = 0;
+        double      max = 0;
+        int         ignore_subvol = 0;
+
+        xlator_t   *avail_subvol = NULL;
+        dht_conf_t *conf = NULL;
 
-        if (!avail_subvol)
-                avail_subvol = subvol;                
+        conf = this->private;
+
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                /* check if subvol has layout errors, before selecting it */
+                ignore_subvol = dht_subvol_has_err (conf->subvolumes[i],
+                                                    layout);
+                if (ignore_subvol)
+                        continue;
+
+                if (conf->disk_unit == 'p') {
+                        if ((conf->du_stats[i].avail_percent > max)
+                            && (conf->du_stats[i].avail_inodes > 0 )) {
+                                max = conf->du_stats[i].avail_percent;
+                                avail_subvol = conf->subvolumes[i];
+                        }
+               } else {
+                         if ((conf->du_stats[i].avail_space > max)
+                            && (conf->du_stats[i].avail_inodes > 0)) {
+                                 max = conf->du_stats[i].avail_space;
+                                 avail_subvol = conf->subvolumes[i];
+                         }
+               }
+        }
 
         return avail_subvol;
 }
diff --git a/xlators/cluster/dht/src/dht-hashfn.c b/xlators/cluster/dht/src/dht-hashfn.c
index 528ac2f90..656cf23a0 100644
--- a/xlators/cluster/dht/src/dht-hashfn.c
+++ b/xlators/cluster/dht/src/dht-hashfn.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -32,50 +23,89 @@
 int
 dht_hash_compute_internal (int type, const char *name, uint32_t *hash_p)
 {
-	int      ret = 0;
-	uint32_t hash = 0;
-
-	switch (type) {
-	case DHT_HASH_TYPE_DM:
-		hash = gf_dm_hashfn (name, strlen (name));
-		break;
-	default:
-		ret = -1;
-		break;
-	}
-
-	if (ret == 0) {
-		*hash_p = hash;
-	}
-
-	return ret;
+        int      ret = 0;
+        uint32_t hash = 0;
+
+        switch (type) {
+        case DHT_HASH_TYPE_DM:
+        case DHT_HASH_TYPE_DM_USER:
+                hash = gf_dm_hashfn (name, strlen (name));
+                break;
+        default:
+                ret = -1;
+                break;
+        }
+
+        if (ret == 0) {
+                *hash_p = hash;
+        }
+
+        return ret;
 }
 
 
-#define MAKE_RSYNC_FRIENDLY_NAME(rsync_frndly_name, name) do {          \
-                rsync_frndly_name = (char *) name;			\
-                if (name[0] == '.') {                                   \
-                        char *dot   = 0;                                \
-                        int namelen = 0;                                \
-                                                                        \
-                        dot = strrchr (name, '.');                      \
-                        if (dot && dot > (name + 1) && *(dot + 1)) {    \
-                                namelen = (dot - name);                 \
-                                rsync_frndly_name = alloca (namelen);   \
-                                strncpy (rsync_frndly_name, name + 1,   \
-                                         namelen);                      \
-                                rsync_frndly_name[namelen - 1] = 0;     \
-                        }                                               \
-                }                                                       \
-        } while (0);
-
+static inline
+gf_boolean_t
+dht_munge_name (const char *original, char *modified, size_t len, regex_t *re)
+{
+        regmatch_t      matches[2];
+        size_t          new_len;
+
+        if (regexec(re,original,2,matches,0) != REG_NOMATCH) {
+                if (matches[1].rm_so != -1) {
+                        new_len = matches[1].rm_eo - matches[1].rm_so;
+                        /* Equal would fail due to the NUL at the end. */
+                        if (new_len < len) {
+                                memcpy (modified,original+matches[1].rm_so,
+                                        new_len);
+                                modified[new_len] = '\0';
+                                return _gf_true;
+                        }
+                }
+        }
+
+        /* This is guaranteed safe because of how the dest was allocated. */
+        strcpy(modified,original);
+        return _gf_false;
+}
 
 int
-dht_hash_compute (int type, const char *name, uint32_t *hash_p)
+dht_hash_compute (xlator_t *this, int type, const char *name, uint32_t *hash_p)
 {
-	char     *rsync_friendly_name = NULL;
-
-	MAKE_RSYNC_FRIENDLY_NAME (rsync_friendly_name, name);
-
-	return dht_hash_compute_internal (type, rsync_friendly_name, hash_p);
+        char            *rsync_friendly_name    = NULL;
+        dht_conf_t      *priv                   = this->private;
+        size_t           len                    = 0;
+        gf_boolean_t     munged                 = _gf_false;
+
+        /*
+         * It wouldn't be safe to use alloca in an inline function that doesn't
+         * actually get inlined, and it wouldn't be efficient to do a real
+         * allocation, so we use alloca here (if needed) and pass that to the
+         * inline.
+         */
+
+        if (priv->extra_regex_valid) {
+                len = strlen(name) + 1;
+                rsync_friendly_name = alloca(len);
+                munged = dht_munge_name (name, rsync_friendly_name, len,
+                                         &priv->extra_regex);
+        }
+
+        if (!munged && priv->rsync_regex_valid) {
+                len = strlen(name) + 1;
+                rsync_friendly_name = alloca(len);
+                gf_log (this->name, GF_LOG_TRACE, "trying regex for %s", name);
+                munged = dht_munge_name (name, rsync_friendly_name, len,
+                                         &priv->rsync_regex);
+                if (munged) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "munged down to %s", rsync_friendly_name);
+                }
+        }
+
+        if (!munged) {
+                rsync_friendly_name = (char *)name;
+        }
+
+        return dht_hash_compute_internal (type, rsync_friendly_name, hash_p);
 }
diff --git a/xlators/cluster/dht/src/dht-helper.c b/xlators/cluster/dht/src/dht-helper.c
index dd2571647..18a501f04 100644
--- a/xlators/cluster/dht/src/dht-helper.c
+++ b/xlators/cluster/dht/src/dht-helper.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -27,56 +18,133 @@
 #include "xlator.h"
 #include "dht-common.h"
 
+static inline int
+dht_inode_ctx_set1 (xlator_t *this, inode_t *inode, xlator_t *subvol)
+{
+        uint64_t tmp_subvol = 0;
+
+        tmp_subvol = (long)subvol;
+        return inode_ctx_set1 (inode, this, &tmp_subvol);
+}
+
+int
+dht_inode_ctx_get1 (xlator_t *this, inode_t *inode, xlator_t **subvol)
+{
+        int ret = -1;
+        uint64_t tmp_subvol = 0;
+
+        ret =  inode_ctx_get1 (inode, this, &tmp_subvol);
+        if (tmp_subvol && subvol)
+                *subvol = (xlator_t *)tmp_subvol;
+
+        return ret;
+}
+
 
 int
 dht_frame_return (call_frame_t *frame)
 {
-	dht_local_t *local = NULL;
-	int          this_call_cnt = -1;
+        dht_local_t *local = NULL;
+        int          this_call_cnt = -1;
+
+        if (!frame)
+                return -1;
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                this_call_cnt = --local->call_cnt;
+        }
+        UNLOCK (&frame->lock);
+
+        return this_call_cnt;
+}
 
-	if (!frame)
-		return -1;
 
-	local = frame->local;
+static uint64_t
+dht_bits_for (uint64_t num)
+{
+	uint64_t bits = 0, ctrl = 1;
 
-	LOCK (&frame->lock);
-	{
-		this_call_cnt = --local->call_cnt;
+	while (ctrl < num) {
+		ctrl *= 2;
+		bits ++;
 	}
-	UNLOCK (&frame->lock);
 
-	return this_call_cnt;
+	return bits;
 }
 
+/*
+ * A slightly "updated" version of the algorithm described in the commit log
+ * is used here.
+ *
+ * The only enhancement is that:
+ *
+ * - The number of bits used by the backend filesystem for HUGE d_off which
+ *   is described as 63, and
+ * - The number of bits used by the d_off presented by the transformation
+ *   upwards which is described as 64, are both made "configurable."
+ */
+
+
+#define BACKEND_D_OFF_BITS 63
+#define PRESENT_D_OFF_BITS 63
+
+#define ONE 1ULL
+#define MASK (~0ULL)
+#define PRESENT_MASK (MASK >> (64 - PRESENT_D_OFF_BITS))
+#define BACKEND_MASK (MASK >> (64 - BACKEND_D_OFF_BITS))
+
+#define TOP_BIT (ONE << (PRESENT_D_OFF_BITS - 1))
+#define SHIFT_BITS (max (0, (BACKEND_D_OFF_BITS - PRESENT_D_OFF_BITS + 1)))
 
 int
 dht_itransform (xlator_t *this, xlator_t *subvol, uint64_t x, uint64_t *y_p)
 {
-	dht_conf_t *conf = NULL;
-	int         cnt = 0;
-	int         max = 0;
-	uint64_t    y = 0;
+        dht_conf_t *conf = NULL;
+        int         cnt = 0;
+        int         max = 0;
+        uint64_t    y = 0;
+        uint64_t    hi_mask = 0;
+        uint64_t    off_mask = 0;
+        int         max_bits = 0;
+
+        if (x == ((uint64_t) -1)) {
+                y = (uint64_t) -1;
+                goto out;
+        }
+
+        conf = this->private;
+        if (!conf)
+                goto out;
 
+        max = conf->subvolume_cnt;
+        cnt = dht_subvol_cnt (this, subvol);
 
-	if (x == ((uint64_t) -1)) {
-		y = (uint64_t) -1;
+	if (max == 1) {
+		y = x;
 		goto out;
 	}
 
-	conf = this->private;
-        if (!conf)
-                goto out;
+        max_bits = dht_bits_for (max);
 
-	max = conf->subvolume_cnt;
-	cnt = dht_subvol_cnt (this, subvol);
+        hi_mask = ~(PRESENT_MASK >> (max_bits + 1));
 
-	y = ((x * max) + cnt);
+        if (x & hi_mask) {
+                /* HUGE d_off */
+                off_mask = MASK << max_bits;
+                y = TOP_BIT | ((x >> SHIFT_BITS) & off_mask) | cnt;
+        } else {
+                /* small d_off */
+                y = ((x * max) + cnt);
+        }
 
 out:
-	if (y_p)
-		*y_p = y;
+        if (y_p)
+                *y_p = y;
 
-	return 0;
+        return 0;
 }
 
 int
@@ -90,7 +158,7 @@ dht_filter_loc_subvol_key (xlator_t *this, loc_t *loc, loc_t *new_loc,
         int            ret       = 0; /* not found */
 
         /* Why do other tasks if first required 'char' itself is not there */
-        if (loc->name && !strchr (loc->name, '@'))
+        if (!new_loc || !loc || !loc->name || !strchr (loc->name, '@'))
                 goto out;
 
         trav = this->children;
@@ -118,7 +186,6 @@ dht_filter_loc_subvol_key (xlator_t *this, loc_t *loc, loc_t *new_loc,
                                 new_loc->path   = ((new_path) ? new_path:
                                                    gf_strdup (loc->path));
                                 new_loc->name   = new_name;
-                                new_loc->ino    = loc->ino;
                                 new_loc->inode  = inode_ref (loc->inode);
                                 new_loc->parent = inode_ref (loc->parent);
                         }
@@ -131,162 +198,200 @@ dht_filter_loc_subvol_key (xlator_t *this, loc_t *loc, loc_t *new_loc,
 out:
         if (!ret) {
                 /* !success */
-                if (new_path)
-                        GF_FREE (new_path);
-                if (new_name)
-                        GF_FREE (new_name);
+                GF_FREE (new_path);
+                GF_FREE (new_name);
         }
         return ret;
 }
 
 int
 dht_deitransform (xlator_t *this, uint64_t y, xlator_t **subvol_p,
-		  uint64_t *x_p)
+                  uint64_t *x_p)
 {
-	dht_conf_t *conf = NULL;
-	int         cnt = 0;
-	int         max = 0;
-	uint64_t    x = 0;
-	xlator_t   *subvol = 0;
+        dht_conf_t *conf = NULL;
+        int         cnt = 0;
+        int         max = 0;
+        uint64_t    x = 0;
+        xlator_t   *subvol = 0;
+        int         max_bits = 0;
+        uint64_t    off_mask = 0;
+        uint64_t    host_mask = 0;
 
         if (!this->private)
-                goto out;
+                return -1;
 
-	conf = this->private;
-	max = conf->subvolume_cnt;
+        conf = this->private;
+        max = conf->subvolume_cnt;
 
-	cnt = y % max;
-	x   = y / max;
+	if (max == 1) {
+		x = y;
+		cnt = 0;
+		goto out;
+	}
 
-	subvol = conf->subvolumes[cnt];
+        if (y & TOP_BIT) {
+                /* HUGE d_off */
+                max_bits = dht_bits_for (max);
+                off_mask = (MASK << max_bits);
+                host_mask = ~(off_mask);
 
-	if (subvol_p)
-		*subvol_p = subvol;
+                x = ((y & ~TOP_BIT) & off_mask) << SHIFT_BITS;
 
-	if (x_p)
-		*x_p = x;
+                cnt = y & host_mask;
+	} else {
+                /* small d_off */
+                cnt = y % max;
+                x = y / max;
+        }
 
 out:
-	return 0;
+        subvol = conf->subvolumes[cnt];
+
+        if (subvol_p)
+                *subvol_p = subvol;
+
+        if (x_p)
+                *x_p = x;
+
+        return 0;
 }
 
 
 void
 dht_local_wipe (xlator_t *this, dht_local_t *local)
 {
-	if (!local)
-		return;
+        if (!local)
+                return;
 
-	loc_wipe (&local->loc);
-	loc_wipe (&local->loc2);
+        loc_wipe (&local->loc);
+        loc_wipe (&local->loc2);
 
-	if (local->xattr)
-		dict_unref (local->xattr);
+        if (local->xattr)
+                dict_unref (local->xattr);
 
-	if (local->inode)
-		inode_unref (local->inode);
+        if (local->inode)
+                inode_unref (local->inode);
 
-	if (local->layout) {
-		dht_layout_unref (this, local->layout);
+        if (local->layout) {
+                dht_layout_unref (this, local->layout);
                 local->layout = NULL;
         }
 
-	loc_wipe (&local->linkfile.loc);
+        loc_wipe (&local->linkfile.loc);
 
-	if (local->linkfile.xattr)
-		dict_unref (local->linkfile.xattr);
+        if (local->linkfile.xattr)
+                dict_unref (local->linkfile.xattr);
 
-	if (local->linkfile.inode)
-		inode_unref (local->linkfile.inode);
+        if (local->linkfile.inode)
+                inode_unref (local->linkfile.inode);
 
-	if (local->fd) {
-		fd_unref (local->fd);
-		local->fd = NULL;
-	}
+        if (local->fd) {
+                fd_unref (local->fd);
+                local->fd = NULL;
+        }
 
         if (local->params) {
                 dict_unref (local->params);
                 local->params = NULL;
         }
 
-	if (local->xattr_req)
-		dict_unref (local->xattr_req);
+        if (local->xattr_req)
+                dict_unref (local->xattr_req);
 
         if (local->selfheal.layout) {
                 dht_layout_unref (this, local->selfheal.layout);
                 local->selfheal.layout = NULL;
         }
 
-        if (local->newpath) {
-                GF_FREE (local->newpath);
-        }
+        GF_FREE (local->newpath);
 
-        if (local->key) {
-               GF_FREE (local->key);
-        }
-           
-	GF_FREE (local);
+        GF_FREE (local->key);
+
+        GF_FREE (local->rebalance.vector);
+
+        if (local->rebalance.iobref)
+                iobref_unref (local->rebalance.iobref);
+
+        mem_put (local);
 }
 
 
 dht_local_t *
-dht_local_init (call_frame_t *frame)
+dht_local_init (call_frame_t *frame, loc_t *loc, fd_t *fd, glusterfs_fop_t fop)
 {
-	dht_local_t *local = NULL;
-
-	/* TODO: use mem-pool */
-	local = GF_CALLOC (1, sizeof (*local),
-                           gf_dht_mt_dht_local_t);
+        dht_local_t *local = NULL;
+        inode_t     *inode = NULL;
+        int          ret   = 0;
 
-	if (!local)
-		return NULL;
+        local = mem_get0 (THIS->local_pool);
+        if (!local)
+                goto out;
 
-	local->op_ret = -1;
-	local->op_errno = EUCLEAN;
+        if (loc) {
+                ret = loc_copy (&local->loc, loc);
+                if (ret)
+                        goto out;
 
-	frame->local = local;
+                inode = loc->inode;
+        }
 
-	return local;
-}
+        if (fd) {
+                local->fd = fd_ref (fd);
+                if (!inode)
+                        inode = fd->inode;
+        }
 
+        local->op_ret   = -1;
+        local->op_errno = EUCLEAN;
+        local->fop      = fop;
 
-char *
-basestr (const char *str)
-{
-        char *basestr = NULL;
+        if (inode) {
+                local->layout   = dht_layout_get (frame->this, inode);
+                local->cached_subvol = dht_subvol_get_cached (frame->this,
+                                                              inode);
+        }
 
-        basestr = strrchr (str, '/');
-        if (basestr)
-                basestr ++;
+        frame->local = local;
 
-        return basestr;
+out:
+        if (ret) {
+                if (local)
+                        mem_put (local);
+                local = NULL;
+        }
+        return local;
 }
 
-
 xlator_t *
 dht_first_up_subvol (xlator_t *this)
 {
-	dht_conf_t *conf = NULL;
-	xlator_t   *child = NULL;
-	int         i = 0;
+        dht_conf_t *conf = NULL;
+        xlator_t   *child = NULL;
+        int         i = 0;
+        time_t      time = 0;
 
-	conf = this->private;
+        conf = this->private;
         if (!conf)
                 goto out;
 
-	LOCK (&conf->subvolume_lock);
-	{
-		for (i = 0; i < conf->subvolume_cnt; i++) {
-			if (conf->subvolume_status[i]) {
-				child = conf->subvolumes[i];
-				break;
-			}
-		}
-	}
-	UNLOCK (&conf->subvolume_lock);
+        LOCK (&conf->subvolume_lock);
+        {
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (conf->subvol_up_time[i]) {
+                                if (!time) {
+                                        time = conf->subvol_up_time[i];
+                                        child = conf->subvolumes[i];
+                                } else if (time > conf->subvol_up_time[i]) {
+                                        time  = conf->subvol_up_time[i];
+                                        child = conf->subvolumes[i];
+                                }
+                        }
+                }
+        }
+        UNLOCK (&conf->subvolume_lock);
 
 out:
-	return child;
+        return child;
 }
 
 xlator_t *
@@ -321,17 +426,23 @@ dht_subvol_get_hashed (xlator_t *this, loc_t *loc)
         dht_layout_t *layout = NULL;
         xlator_t     *subvol = NULL;
 
-        if (is_fs_root (loc)) {
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, loc, out);
+
+        if (__is_root_gfid (loc->gfid)) {
                 subvol = dht_first_up_subvol (this);
                 goto out;
         }
 
+        GF_VALIDATE_OR_GOTO (this->name, loc->parent, out);
+        GF_VALIDATE_OR_GOTO (this->name, loc->name, out);
+
         layout = dht_layout_get (this, loc->parent);
 
         if (!layout) {
                 gf_log (this->name, GF_LOG_DEBUG,
-                        "layout missing path=%s parent=%"PRId64,
-                        loc->path, loc->parent->ino);
+                        "layout missing path=%s parent=%s",
+                        loc->path, uuid_utoa (loc->parent->gfid));
                 goto out;
         }
 
@@ -359,6 +470,8 @@ dht_subvol_get_cached (xlator_t *this, inode_t *inode)
         dht_layout_t *layout = NULL;
         xlator_t     *subvol = NULL;
 
+        GF_VALIDATE_OR_GOTO (this->name, this, out);
+        GF_VALIDATE_OR_GOTO (this->name, inode, out);
 
         layout = dht_layout_get (this, inode);
 
@@ -366,7 +479,7 @@ dht_subvol_get_cached (xlator_t *this, inode_t *inode)
                 goto out;
         }
 
-	subvol = layout->list[0].xlator;
+        subvol = layout->list[0].xlator;
 
 out:
         if (layout) {
@@ -380,150 +493,686 @@ out:
 xlator_t *
 dht_subvol_next (xlator_t *this, xlator_t *prev)
 {
-	dht_conf_t *conf = NULL;
-	int         i = 0;
-	xlator_t   *next = NULL;
+        dht_conf_t *conf = NULL;
+        int         i = 0;
+        xlator_t   *next = NULL;
 
-	conf = this->private;
+        conf = this->private;
         if (!conf)
                 goto out;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		if (conf->subvolumes[i] == prev) {
-			if ((i + 1) < conf->subvolume_cnt)
-				next = conf->subvolumes[i + 1];
-			break;
-		}
-	}
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (conf->subvolumes[i] == prev) {
+                        if ((i + 1) < conf->subvolume_cnt)
+                                next = conf->subvolumes[i + 1];
+                        break;
+                }
+        }
 
 out:
-	return next;
+        return next;
 }
 
+/* This func wraps around, if prev is actually the last subvol.
+ */
+xlator_t *
+dht_subvol_next_available (xlator_t *this, xlator_t *prev)
+{
+        dht_conf_t *conf = NULL;
+        int         i = 0;
+        xlator_t   *next = NULL;
+
+        conf = this->private;
+        if (!conf)
+                goto out;
+
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (conf->subvolumes[i] == prev) {
+                        /* if prev is last in conf->subvolumes, then wrap
+                         * around.
+                         */
+                        if ((i + 1) < conf->subvolume_cnt) {
+                                next = conf->subvolumes[i + 1];
+                        } else {
+                                next = conf->subvolumes[0];
+                        }
+                        break;
+                }
+        }
 
+out:
+        return next;
+}
 int
 dht_subvol_cnt (xlator_t *this, xlator_t *subvol)
 {
-	int i = 0;
-	int ret = -1;
-	dht_conf_t *conf = NULL;
+        int i = 0;
+        int ret = -1;
+        dht_conf_t *conf = NULL;
 
-	conf = this->private;
+        conf = this->private;
         if (!conf)
                 goto out;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		if (subvol == conf->subvolumes[i]) {
-			ret = i;
-			break;
-		}
-	}
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (subvol == conf->subvolumes[i]) {
+                        ret = i;
+                        break;
+                }
+        }
 
 out:
-	return ret;
+        return ret;
 }
 
 
-#define set_if_greater(a, b) do {		\
-		if ((a) < (b))			\
-			(a) = (b);		\
-	} while (0)
+#define set_if_greater(a, b) do {               \
+                if ((a) < (b))                  \
+                        (a) = (b);              \
+        } while (0)
+
+
+#define set_if_greater_time(a, an, b, bn) do {                          \
+                if (((a) < (b)) || (((a) == (b)) && ((an) < (bn)))){    \
+                        (a) = (b);                                      \
+                        (an) = (bn);                                    \
+                }                                                       \
+        } while (0)                                                     \
+
 
 int
 dht_iatt_merge (xlator_t *this, struct iatt *to,
-		struct iatt *from, xlator_t *subvol)
+                struct iatt *from, xlator_t *subvol)
 {
         if (!from || !to)
                 return 0;
 
-	to->ia_dev      = from->ia_dev;
+        to->ia_dev      = from->ia_dev;
 
         uuid_copy (to->ia_gfid, from->ia_gfid);
 
-	dht_itransform (this, subvol, from->ia_ino, &to->ia_ino);
+        to->ia_ino      = from->ia_ino;
+        to->ia_prot     = from->ia_prot;
+        to->ia_type     = from->ia_type;
+        to->ia_nlink    = from->ia_nlink;
+        to->ia_rdev     = from->ia_rdev;
+        to->ia_size    += from->ia_size;
+        to->ia_blksize  = from->ia_blksize;
+        to->ia_blocks  += from->ia_blocks;
+
+        set_if_greater (to->ia_uid, from->ia_uid);
+        set_if_greater (to->ia_gid, from->ia_gid);
+
+        set_if_greater_time(to->ia_atime, to->ia_atime_nsec,
+                            from->ia_atime, from->ia_atime_nsec);
+        set_if_greater_time (to->ia_mtime, to->ia_mtime_nsec,
+                             from->ia_mtime, from->ia_mtime_nsec);
+        set_if_greater_time (to->ia_ctime, to->ia_ctime_nsec,
+                             from->ia_ctime, from->ia_ctime_nsec);
+
+        return 0;
+}
+
+int
+dht_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name)
+{
+        if (!child) {
+                goto err;
+        }
+
+        if (strcmp (parent->path, "/") == 0)
+                gf_asprintf ((char **)&child->path, "/%s", name);
+        else
+                gf_asprintf ((char **)&child->path, "%s/%s", parent->path, name);
+
+        if (!child->path) {
+                goto err;
+        }
 
-	to->ia_prot     = from->ia_prot;
-	to->ia_type     = from->ia_type;
-	to->ia_nlink    = from->ia_nlink;
-	to->ia_rdev     = from->ia_rdev;
-	to->ia_size    += from->ia_size;
-	to->ia_blksize  = from->ia_blksize;
-	to->ia_blocks  += from->ia_blocks;
+        child->name = strrchr (child->path, '/');
+        if (child->name)
+                child->name++;
 
-	set_if_greater (to->ia_uid, from->ia_uid);
-	set_if_greater (to->ia_gid, from->ia_gid);
+        child->parent = inode_ref (parent->inode);
+        child->inode = inode_new (parent->inode->table);
 
-	set_if_greater (to->ia_atime, from->ia_atime);
-	set_if_greater (to->ia_mtime, from->ia_mtime);
-	set_if_greater (to->ia_ctime, from->ia_ctime);
+        if (!child->inode) {
+                goto err;
+        }
 
-	return 0;
+        return 0;
+err:
+        loc_wipe (child);
+        return -1;
 }
 
+
+
 int
-dht_frame_su_do (call_frame_t *frame)
+dht_init_subvolumes (xlator_t *this, dht_conf_t *conf)
 {
-        dht_local_t     *local = NULL;
+        xlator_list_t *subvols = NULL;
+        int            cnt = 0;
 
-        local = frame->local;
+        if (!conf)
+                return -1;
+
+        for (subvols = this->children; subvols; subvols = subvols->next)
+                cnt++;
+
+        conf->subvolumes = GF_CALLOC (cnt, sizeof (xlator_t *),
+                                      gf_dht_mt_xlator_t);
+        if (!conf->subvolumes) {
+                return -1;
+        }
+        conf->subvolume_cnt = cnt;
+
+        cnt = 0;
+        for (subvols = this->children; subvols; subvols = subvols->next)
+                conf->subvolumes[cnt++] = subvols->xlator;
+
+        conf->subvolume_status = GF_CALLOC (cnt, sizeof (char),
+                                            gf_dht_mt_char);
+        if (!conf->subvolume_status) {
+                return -1;
+        }
+
+        conf->last_event = GF_CALLOC (cnt, sizeof (int),
+                                      gf_dht_mt_char);
+        if (!conf->last_event) {
+                return -1;
+        }
 
-        local->uid = frame->root->uid;
-        local->gid = frame->root->gid;
+        conf->subvol_up_time = GF_CALLOC (cnt, sizeof (time_t),
+                                          gf_dht_mt_subvol_time);
+        if (!conf->subvol_up_time) {
+                return -1;
+        }
 
-        frame->root->uid = 0;
-        frame->root->gid = 0;
+        conf->du_stats = GF_CALLOC (conf->subvolume_cnt, sizeof (dht_du_t),
+                                    gf_dht_mt_dht_du_t);
+        if (!conf->du_stats) {
+                return -1;
+        }
+
+        conf->decommissioned_bricks = GF_CALLOC (cnt, sizeof (xlator_t *),
+                                                 gf_dht_mt_xlator_t);
+        if (!conf->decommissioned_bricks) {
+                return -1;
+        }
 
         return 0;
 }
 
 
-int
-dht_frame_su_undo (call_frame_t *frame)
+
+
+static int
+dht_migration_complete_check_done (int op_ret, call_frame_t *frame, void *data)
 {
-        dht_local_t     *local = NULL;
+        dht_local_t *local = NULL;
 
         local = frame->local;
 
-        frame->root->uid = local->uid;
-        frame->root->gid = local->gid;
+        local->rebalance.target_op_fn (THIS, frame, op_ret);
 
         return 0;
 }
 
 
 int
-dht_build_child_loc (xlator_t *this, loc_t *child, loc_t *parent, char *name)
+dht_migration_complete_check_task (void *data)
 {
-        if (!child) {
-                goto err;
+        int           ret      = -1;
+        xlator_t     *src_node = NULL;
+        xlator_t     *dst_node = NULL;
+        dht_local_t  *local    = NULL;
+        dict_t       *dict     = NULL;
+        dht_layout_t *layout   = NULL;
+        struct iatt   stbuf    = {0,};
+        xlator_t     *this     = NULL;
+        call_frame_t *frame    = NULL;
+        loc_t         tmp_loc  = {0,};
+        char         *path     = NULL;
+        dht_conf_t   *conf     = NULL;
+        inode_t      *inode    = NULL;
+        fd_t         *iter_fd  = NULL;
+        uint64_t      tmp_subvol = 0;
+        int           open_failed = 0;
+
+        this  = THIS;
+        frame = data;
+        local = frame->local;
+        conf = this->private;
+
+        src_node = local->cached_subvol;
+
+        if (!local->loc.inode && !local->fd) {
+                local->op_errno = EINVAL;
+                goto out;
         }
 
-        if (strcmp (parent->path, "/") == 0)
-                gf_asprintf ((char **)&child->path, "/%s", name);
-        else
-                gf_asprintf ((char **)&child->path, "%s/%s", parent->path, name);
+        inode = (!local->fd) ? local->loc.inode : local->fd->inode;
 
-        if (!child->path) {
+        /* getxattr on cached_subvol for 'linkto' value. Do path based getxattr
+         * as root:root. If a fd is already open, access check wont be done*/
+
+        if (!local->loc.inode) {
+                ret = syncop_fgetxattr (src_node, local->fd, &dict,
+                                        conf->link_xattr_name);
+        } else {
+                SYNCTASK_SETID (0, 0);
+                ret = syncop_getxattr (src_node, &local->loc, &dict,
+                                       conf->link_xattr_name);
+                SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+        }
+
+        if (!ret)
+                dst_node = dht_linkfile_subvol (this, NULL, NULL, dict);
+
+        if (ret) {
+                if (!dht_inode_missing(-ret) || (!local->loc.inode)) {
+                        local->op_errno = -ret;
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "%s: failed to get the 'linkto' xattr %s",
+                                local->loc.path, strerror (-ret));
+                        ret = -1;
+                        goto out;
+                }
+
+                /* Need to do lookup on hashed subvol, then get the file */
+                ret = syncop_lookup (this, &local->loc, NULL, &stbuf, NULL,
+                                     NULL);
+                if (ret) {
+                        local->op_errno = -ret;
+                        ret = -1;
+                        goto out;
+                }
+
+                dst_node = dht_subvol_get_cached (this, local->loc.inode);
+        }
+
+        if (!dst_node) {
                 gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
+                        "%s: failed to get the destination node",
+                        local->loc.path);
+                ret = -1;
+                local->op_errno = EINVAL;
+                goto out;
         }
 
-        child->name = strrchr (child->path, '/');
-        if (child->name)
-                child->name++;
+        /* lookup on dst */
+        if (local->loc.inode) {
+                ret = syncop_lookup (dst_node, &local->loc, NULL, &stbuf, NULL,
+                                     NULL);
+
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "%s: failed to lookup the file on %s",
+                                local->loc.path, dst_node->name);
+                        local->op_errno = -ret;
+                        ret = -1;
+                        goto out;
+                }
 
-        child->parent = inode_ref (parent->inode);
-        child->inode = inode_new (parent->inode->table);
+                if (uuid_compare (stbuf.ia_gfid, local->loc.inode->gfid)) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "%s: gfid different on the target file on %s",
+                                local->loc.path, dst_node->name);
+                        ret = -1;
+                        local->op_errno = EIO;
+                        goto out;
+                }
+        }
 
-        if (!child->inode) {
+        /* update inode ctx (the layout) */
+        dht_layout_unref (this, local->layout);
+
+        ret = dht_layout_preset (this, dst_node, inode);
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "%s: could not set preset layout for subvol %s",
+                        local->loc.path, dst_node->name);
+                ret   = -1;
+                local->op_errno = EINVAL;
+                goto out;
+        }
+
+        layout = dht_layout_for_subvol (this, dst_node);
+        if (!layout) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "%s: no pre-set layout for subvolume %s",
+                        local->loc.path, dst_node ? dst_node->name : "<nil>");
+                ret = -1;
+                local->op_errno = EINVAL;
+                goto out;
+        }
+
+        ret = dht_layout_set (this, inode, layout);
+        if (ret) {
                 gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
+                        "%s: failed to set the new layout",
+                        local->loc.path);
+                local->op_errno = EINVAL;
+                goto out;
+        }
+
+        local->cached_subvol = dst_node;
+        ret = 0;
+
+        /* once we detect the migration complete, the inode-ctx2 is no more
+           required.. delete the ctx and also, it means, open() already
+           done on all the fd of inode */
+        ret = inode_ctx_reset1 (inode, this, &tmp_subvol);
+        if (tmp_subvol)
+                goto out;
+
+        if (list_empty (&inode->fd_list))
+                goto out;
+
+        /* perform open as root:root. There is window between linkfile
+         * creation(root:root) and setattr with the correct uid/gid
+         */
+        SYNCTASK_SETID(0, 0);
+
+        /* perform 'open()' on all the fd's present on the inode */
+        tmp_loc.inode = inode;
+        inode_path (inode, NULL, &path);
+        if (path)
+                tmp_loc.path = path;
+        list_for_each_entry (iter_fd, &inode->fd_list, inode_list) {
+                if (fd_is_anonymous (iter_fd))
+                        continue;
+
+                /* flags for open are stripped down to allow following the
+                 * new location of the file, otherwise we can get EEXIST or
+                 * truncate the file again as rebalance is moving the data */
+                ret = syncop_open (dst_node, &tmp_loc,
+                                   (iter_fd->flags &
+                                   ~(O_CREAT | O_EXCL | O_TRUNC)), iter_fd);
+                if (ret < 0) {
+                        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;
+                        local->op_errno = -ret;
+                        ret = -1;
+                }
         }
+        GF_FREE (path);
+
+        SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+
+        if (open_failed) {
+                ret = -1;
+                goto out;
+        }
+        ret = 0;
+out:
+
+        return ret;
+}
+
+int
+dht_rebalance_complete_check (xlator_t *this, call_frame_t *frame)
+{
+        int         ret     = -1;
+
+        ret = synctask_new (this->ctx->env, dht_migration_complete_check_task,
+                            dht_migration_complete_check_done,
+                            frame, frame);
+        return ret;
+}
+
+/* During 'in-progress' state, both nodes should have the file */
+static int
+dht_inprogress_check_done (int op_ret, call_frame_t *sync_frame, void *data)
+{
+        dht_local_t *local = NULL;
+
+        local = sync_frame->local;
+
+        local->rebalance.target_op_fn (THIS, sync_frame, op_ret);
 
         return 0;
-err:
-        loc_wipe (child);
-        return -1;
+}
+
+static int
+dht_rebalance_inprogress_task (void *data)
+{
+        int           ret      = -1;
+        xlator_t     *src_node = NULL;
+        xlator_t     *dst_node = NULL;
+        dht_local_t  *local    = NULL;
+        dict_t       *dict     = NULL;
+        call_frame_t *frame    = NULL;
+        xlator_t     *this     = NULL;
+        char         *path     = NULL;
+        struct iatt   stbuf    = {0,};
+        loc_t         tmp_loc  = {0,};
+        dht_conf_t   *conf     = NULL;
+        inode_t      *inode    = NULL;
+        fd_t         *iter_fd  = NULL;
+        int           open_failed = 0;
+
+        this  = THIS;
+        frame = data;
+        local = frame->local;
+        conf = this->private;
+
+        src_node = local->cached_subvol;
+
+        if (!local->loc.inode && !local->fd)
+                goto out;
+
+        inode = (!local->fd) ? local->loc.inode : local->fd->inode;
+
+        /* getxattr on cached_subvol for 'linkto' value. Do path based getxattr
+         * as root:root. If a fd is already open, access check wont be done*/
+        if (local->loc.inode) {
+                SYNCTASK_SETID (0, 0);
+                ret = syncop_getxattr (src_node, &local->loc, &dict,
+                                       conf->link_xattr_name);
+                SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+        } else {
+                ret = syncop_fgetxattr (src_node, local->fd, &dict,
+                                        conf->link_xattr_name);
+        }
+
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to get the 'linkto' xattr %s",
+                        local->loc.path, strerror (-ret));
+                ret = -1;
+                goto out;
+        }
+
+        dst_node = dht_linkfile_subvol (this, NULL, NULL, dict);
+        if (!dst_node) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to get the 'linkto' xattr from dict",
+                        local->loc.path);
+                ret = -1;
+                goto out;
+        }
+
+        local->rebalance.target_node = dst_node;
+
+        if (local->loc.inode) {
+                /* lookup on dst */
+                ret = syncop_lookup (dst_node, &local->loc, NULL,
+                                     &stbuf, NULL, NULL);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "%s: failed to lookup the file on %s",
+                                local->loc.path, dst_node->name);
+                        ret = -1;
+                        goto out;
+                }
+
+                if (uuid_compare (stbuf.ia_gfid, local->loc.inode->gfid)) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "%s: gfid different on the target file on %s",
+                                local->loc.path, dst_node->name);
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        ret = 0;
+
+        if (list_empty (&inode->fd_list))
+                goto done;
+
+        /* perform open as root:root. There is window between linkfile
+         * creation(root:root) and setattr with the correct uid/gid
+         */
+        SYNCTASK_SETID (0, 0);
+
+        tmp_loc.inode = inode;
+        inode_path (inode, NULL, &path);
+        if (path)
+                tmp_loc.path = path;
+
+        list_for_each_entry (iter_fd, &inode->fd_list, inode_list) {
+                if (fd_is_anonymous (iter_fd))
+                        continue;
+
+                /* flags for open are stripped down to allow following the
+                 * new location of the file, otherwise we can get EEXIST or
+                 * truncate the file again as rebalance is moving the data */
+                ret = syncop_open (dst_node, &tmp_loc,
+                                   (iter_fd->flags &
+                                   ~(O_CREAT | O_EXCL | O_TRUNC)), iter_fd);
+                if (ret < 0) {
+                        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);
+                        ret = -1;
+                        open_failed = 1;
+                }
+        }
+        GF_FREE (path);
+
+        SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+
+        if (open_failed) {
+                ret = -1;
+                goto out;
+        }
+
+done:
+        ret = dht_inode_ctx_set1 (this, inode, dst_node);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to set inode-ctx target file at %s",
+                        local->loc.path, dst_node->name);
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+dht_rebalance_in_progress_check (xlator_t *this, call_frame_t *frame)
+{
+
+        int         ret     = -1;
+
+        ret = synctask_new (this->ctx->env, dht_rebalance_inprogress_task,
+                            dht_inprogress_check_done,
+                            frame, frame);
+        return ret;
+}
+
+int
+dht_inode_ctx_layout_set (inode_t *inode, xlator_t *this,
+                          dht_layout_t *layout_int)
+{
+        dht_inode_ctx_t         *ctx            = NULL;
+        int                      ret            = -1;
+
+        ret = dht_inode_ctx_get (inode, this, &ctx);
+        if (!ret && ctx) {
+                ctx->layout = layout_int;
+        } else {
+                ctx = GF_CALLOC (1, sizeof (*ctx), gf_dht_mt_inode_ctx_t);
+                if (!ctx)
+                        return ret;
+                ctx->layout = layout_int;
+        }
+
+        ret = dht_inode_ctx_set (inode, this, ctx);
+
+        return ret;
+}
+
+int
+dht_inode_ctx_time_update (inode_t *inode, xlator_t *this, struct iatt *stat,
+                           int32_t post)
+{
+        dht_inode_ctx_t         *ctx            = NULL;
+        dht_stat_time_t         *time           = 0;
+        int                      ret            = -1;
+
+        GF_VALIDATE_OR_GOTO (this->name, stat, out);
+        GF_VALIDATE_OR_GOTO (this->name, inode, out);
+
+        ret = dht_inode_ctx_get (inode, this, &ctx);
+
+        if (ret) {
+                ctx = GF_CALLOC (1, sizeof (*ctx), gf_dht_mt_inode_ctx_t);
+                if (!ctx)
+                        return -1;
+        }
+
+        time = &ctx->time;
+
+        DHT_UPDATE_TIME(time->mtime, time->mtime_nsec,
+                        stat->ia_mtime, stat->ia_mtime_nsec, inode, post);
+        DHT_UPDATE_TIME(time->ctime, time->ctime_nsec,
+                        stat->ia_ctime, stat->ia_ctime_nsec, inode, post);
+        DHT_UPDATE_TIME(time->atime, time->atime_nsec,
+                        stat->ia_atime, stat->ia_atime_nsec, inode, post);
+
+        ret = dht_inode_ctx_set (inode, this, ctx);
+out:
+        return 0;
+}
+
+int
+dht_inode_ctx_get (inode_t *inode, xlator_t *this, dht_inode_ctx_t **ctx)
+{
+        int             ret     = -1;
+        uint64_t        ctx_int = 0;
+
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, inode, out);
+
+        ret = inode_ctx_get (inode, this, &ctx_int);
+
+        if (ret)
+                return ret;
+
+        if (ctx)
+                *ctx = (dht_inode_ctx_t *) ctx_int;
+out:
+        return ret;
+}
+
+int dht_inode_ctx_set (inode_t *inode, xlator_t *this, dht_inode_ctx_t *ctx)
+{
+        int             ret = -1;
+        uint64_t        ctx_int = 0;
+
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, inode, out);
+        GF_VALIDATE_OR_GOTO (this->name, ctx, out);
+
+        ctx_int = (long)ctx;
+        ret = inode_ctx_set (inode, this, &ctx_int);
+out:
+        return ret;
 }
diff --git a/xlators/cluster/dht/src/dht-inode-read.c b/xlators/cluster/dht/src/dht-inode-read.c
new file mode 100644
index 000000000..e8a9a7196
--- /dev/null
+++ b/xlators/cluster/dht/src/dht-inode-read.c
@@ -0,0 +1,1139 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "dht-common.h"
+
+int dht_access2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_readv2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_attr2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_open2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_flush2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_lk2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_fsync2 (xlator_t *this, call_frame_t *frame, int ret);
+
+int
+dht_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+              int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           ret = 0;
+
+        local = frame->local;
+        prev = cookie;
+
+        local->op_errno = op_errno;
+        if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "subvolume %s returned -1 (%s)",
+                        prev->this->name, strerror (op_errno));
+                goto out;
+        }
+
+        if (!op_ret || (local->call_cnt != 1))
+                goto out;
+
+        /* rebalance would have happened */
+        local->rebalance.target_op_fn = dht_open2;
+        ret = dht_rebalance_complete_check (this, frame);
+        if (!ret)
+                return 0;
+
+out:
+        DHT_STACK_UNWIND (open, frame, op_ret, op_errno, local->fd, xdata);
+
+        return 0;
+}
+
+int
+dht_open2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t *local  = NULL;
+        xlator_t    *subvol = NULL;
+        int          op_errno = EINVAL;
+
+        local = frame->local;
+        if (!local)
+                goto out;
+
+        op_errno = ENOENT;
+        if (op_ret)
+                goto out;
+
+        local->call_cnt = 2;
+        subvol = local->cached_subvol;
+
+        STACK_WIND (frame, dht_open_cbk, subvol, subvol->fops->open,
+                    &local->loc, local->rebalance.flags, local->fd,
+                    NULL);
+        return 0;
+
+out:
+        DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
+        return 0;
+}
+
+int
+dht_open (call_frame_t *frame, xlator_t *this,
+          loc_t *loc, int flags, fd_t *fd, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, loc, fd, GF_FOP_OPEN);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        local->rebalance.flags = flags;
+        local->call_cnt = 1;
+
+        STACK_WIND (frame, dht_open_cbk, subvol, subvol->fops->open,
+                    loc, flags, fd, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (open, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+
+int
+dht_file_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int op_ret, int op_errno, struct iatt *stbuf, dict_t *xdata)
+{
+        xlator_t     *subvol = 0;
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           ret = -1;
+        inode_t      *inode = NULL;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+        local = frame->local;
+        prev = cookie;
+
+        if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
+                local->op_errno = op_errno;
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "subvolume %s returned -1 (%s)",
+                        prev->this->name, strerror (op_errno));
+                goto out;
+        }
+
+        if (local->call_cnt != 1)
+                goto out;
+
+        local->op_errno = op_errno;
+        /* Check if the rebalance phase2 is true */
+        if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (stbuf)) {
+                inode = (local->fd) ? local->fd->inode : local->loc.inode;
+                ret = dht_inode_ctx_get1 (this, inode, &subvol);
+                if (!subvol) {
+                        /* Phase 2 of migration */
+                        local->rebalance.target_op_fn = dht_attr2;
+                        ret = dht_rebalance_complete_check (this, frame);
+                        if (!ret)
+                                return 0;
+                } else {
+                        /* value is already set in fd_ctx, that means no need
+                           to check for whether its complete or not. */
+                        dht_attr2 (this, frame, 0);
+                        return 0;
+                }
+        }
+
+out:
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
+        DHT_STACK_UNWIND (stat, frame, op_ret, op_errno, stbuf, xdata);
+err:
+        return 0;
+}
+
+int
+dht_attr2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t *local  = NULL;
+        xlator_t    *subvol = NULL;
+        int          op_errno = EINVAL;
+
+        local = frame->local;
+        if (!local)
+                goto out;
+
+        op_errno = local->op_errno;
+        if (op_ret == -1)
+                goto out;
+
+        subvol = local->cached_subvol;
+        local->call_cnt = 2;
+
+        if (local->fop == GF_FOP_FSTAT) {
+                STACK_WIND (frame, dht_file_attr_cbk, subvol,
+                            subvol->fops->fstat, local->fd, NULL);
+        } else {
+                STACK_WIND (frame, dht_file_attr_cbk, subvol,
+                            subvol->fops->stat, &local->loc, NULL);
+        }
+        return 0;
+
+out:
+        DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
+        return 0;
+}
+
+int
+dht_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+              int op_ret, int op_errno, struct iatt *stbuf, dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+        local = frame->local;
+        prev = cookie;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "subvolume %s returned -1 (%s)",
+                                prev->this->name, strerror (op_errno));
+
+                        goto unlock;
+                }
+
+                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+
+                local->op_ret = 0;
+        }
+unlock:
+        UNLOCK (&frame->lock);
+out:
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                DHT_STACK_UNWIND (stat, frame, local->op_ret, local->op_errno,
+                                  &local->stbuf, xdata);
+        }
+err:
+        return 0;
+}
+
+int
+dht_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+        dht_layout_t *layout = NULL;
+        int           i = 0;
+        int           call_cnt = 0;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+
+
+        local = dht_local_init (frame, loc, NULL, GF_FOP_STAT);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        layout = local->layout;
+        if (!layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no layout for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        if (IA_ISREG (loc->inode->ia_type)) {
+                local->call_cnt = 1;
+
+                subvol = local->cached_subvol;
+
+                STACK_WIND (frame, dht_file_attr_cbk, subvol,
+                            subvol->fops->stat, loc, xdata);
+
+                return 0;
+        }
+
+        local->call_cnt = call_cnt = layout->cnt;
+
+        for (i = 0; i < call_cnt; i++) {
+                subvol = layout->list[i].xlator;
+
+                STACK_WIND (frame, dht_attr_cbk,
+                            subvol, subvol->fops->stat,
+                            loc, xdata);
+        }
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+
+
+int
+dht_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+        dht_layout_t *layout = NULL;
+        int           i = 0;
+        int           call_cnt = 0;
+
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FSTAT);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        layout = local->layout;
+        if (!layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no layout for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        if (IA_ISREG (fd->inode->ia_type)) {
+                local->call_cnt = 1;
+
+                subvol = local->cached_subvol;
+
+                STACK_WIND (frame, dht_file_attr_cbk, subvol,
+                            subvol->fops->fstat, fd, xdata);
+
+                return 0;
+        }
+
+        local->call_cnt = call_cnt = layout->cnt;
+
+        for (i = 0; i < call_cnt; i++) {
+                subvol = layout->list[i].xlator;
+                STACK_WIND (frame, dht_attr_cbk,
+                            subvol, subvol->fops->fstat,
+                            fd, xdata);
+        }
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fstat, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+
+int
+dht_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+               int op_ret, int op_errno,
+               struct iovec *vector, int count, struct iatt *stbuf,
+               struct iobref *iobref, dict_t *xdata)
+{
+        dht_local_t *local      = NULL;
+        int          ret        = 0;
+        inode_t     *inode      = NULL;
+        xlator_t    *subvol = 0;
+
+        local = frame->local;
+        if (!local) {
+                op_ret = -1;
+                op_errno = EINVAL;
+                goto out;
+        }
+
+        /* This is already second try, no need for re-check */
+        if (local->call_cnt != 1)
+                goto out;
+
+        if ((op_ret == -1) && !dht_inode_missing(op_errno))
+                goto out;
+
+        local->op_errno = op_errno;
+        if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (stbuf)) {
+                /* File would be migrated to other node */
+                ret = dht_inode_ctx_get1 (this, inode, &subvol);
+                if (!subvol) {
+                        local->rebalance.target_op_fn = dht_readv2;
+                        ret = dht_rebalance_complete_check (this, frame);
+                        if (!ret)
+                                return 0;
+                } else {
+                        /* value is already set in fd_ctx, that means no need
+                           to check for whether its complete or not. */
+                        dht_readv2 (this, frame, 0);
+                        return 0;
+                }
+        }
+
+out:
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
+        DHT_STACK_UNWIND (readv, frame, op_ret, op_errno, vector, count, stbuf,
+                          iobref, xdata);
+
+        return 0;
+}
+
+int
+dht_readv2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t *local  = NULL;
+        xlator_t    *subvol = NULL;
+        int          op_errno = EINVAL;
+
+        local = frame->local;
+        if (!local)
+                goto out;
+
+        op_errno = local->op_errno;
+        if (op_ret == -1)
+                goto out;
+
+        local->call_cnt = 2;
+        subvol = local->cached_subvol;
+
+        STACK_WIND (frame, dht_readv_cbk, subvol, subvol->fops->readv,
+                    local->fd, local->rebalance.size, local->rebalance.offset,
+                    local->rebalance.flags, NULL);
+
+        return 0;
+
+out:
+        DHT_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, NULL, NULL);
+        return 0;
+}
+
+int
+dht_readv (call_frame_t *frame, xlator_t *this,
+           fd_t *fd, size_t size, off_t off, uint32_t flags, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_READ);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        local->rebalance.offset = off;
+        local->rebalance.size   = size;
+        local->rebalance.flags  = flags;
+        local->call_cnt = 1;
+
+        STACK_WIND (frame, dht_readv_cbk,
+                    subvol, subvol->fops->readv,
+                    fd, size, off, flags, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, NULL, NULL);
+
+        return 0;
+}
+
+int
+dht_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int op_ret, int op_errno, dict_t *xdata)
+{
+        int          ret = -1;
+        dht_local_t *local = NULL;
+        xlator_t    *subvol = NULL;
+        call_frame_t *prev = NULL;
+
+        local = frame->local;
+        prev = cookie;
+
+        if (!prev || !prev->this)
+                goto out;
+        if (local->call_cnt != 1)
+                goto out;
+        if ((op_ret == -1) && (op_errno == ENOTCONN) &&
+            IA_ISDIR(local->loc.inode->ia_type)) {
+
+                subvol = dht_subvol_next_available (this, prev->this);
+                if (!subvol)
+                        goto out;
+
+                /* check if we are done with visiting every node */
+                if (subvol == local->cached_subvol) {
+                        goto out;
+                }
+
+                STACK_WIND (frame, dht_access_cbk, subvol, subvol->fops->access,
+                            &local->loc, local->rebalance.flags, NULL);
+                return 0;
+        }
+        if ((op_ret == -1) && dht_inode_missing(op_errno)) {
+                /* File would be migrated to other node */
+                local->op_errno = op_errno;
+                local->rebalance.target_op_fn = dht_access2;
+                ret = dht_rebalance_complete_check (frame->this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+out:
+        DHT_STACK_UNWIND (access, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int
+dht_access2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t *local  = NULL;
+        xlator_t    *subvol = NULL;
+        int          op_errno = EINVAL;
+
+        local = frame->local;
+        if (!local)
+                goto out;
+
+        op_errno = local->op_errno;
+        if (op_ret == -1)
+                goto out;
+
+        local->call_cnt = 2;
+        subvol = local->cached_subvol;
+
+        STACK_WIND (frame, dht_access_cbk, subvol, subvol->fops->access,
+                    &local->loc, local->rebalance.flags, NULL);
+
+        return 0;
+
+out:
+        DHT_STACK_UNWIND (access, frame, -1, op_errno, NULL);
+        return 0;
+}
+
+
+int
+dht_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+            dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+
+        local = dht_local_init (frame, loc, NULL, GF_FOP_ACCESS);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        local->rebalance.flags = mask;
+        local->call_cnt = 1;
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        STACK_WIND (frame, dht_access_cbk, subvol, subvol->fops->access,
+                    loc, mask, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (access, frame, -1, op_errno, NULL);
+
+        return 0;
+}
+
+
+int
+dht_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+               int op_ret, int op_errno, dict_t *xdata)
+{
+        dht_local_t  *local  = NULL;
+        inode_t      *inode  = NULL;
+        xlator_t     *subvol = 0;
+
+        local = frame->local;
+
+        local->op_errno = op_errno;
+
+        if (local->call_cnt != 1)
+                goto out;
+
+        /* If context is set, then send flush() it to the destination */
+        dht_inode_ctx_get1 (this, inode, &subvol);
+        if (subvol) {
+                dht_flush2 (this, frame, 0);
+                return 0;
+        }
+
+out:
+        DHT_STACK_UNWIND (flush, frame, op_ret, op_errno, xdata);
+
+        return 0;
+}
+
+int
+dht_flush2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t  *local  = NULL;
+        xlator_t     *subvol = NULL;
+
+        local = frame->local;
+
+        dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+
+        if (!subvol)
+                subvol = local->cached_subvol;
+
+        local->call_cnt = 2; /* This is the second attempt */
+
+        STACK_WIND (frame, dht_flush_cbk,
+                    subvol, subvol->fops->flush, local->fd, NULL);
+
+        return 0;
+}
+
+
+int
+dht_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FLUSH);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        local->call_cnt = 1;
+
+        STACK_WIND (frame, dht_flush_cbk,
+                    subvol, subvol->fops->flush, fd, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (flush, frame, -1, op_errno, NULL);
+
+        return 0;
+}
+
+
+int
+dht_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
+               int op_errno, struct iatt *prebuf, struct iatt *postbuf,
+               dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           ret = -1;
+        inode_t      *inode = NULL;
+        xlator_t     *subvol = 0;
+
+        local = frame->local;
+        prev = cookie;
+
+        local->op_errno = op_errno;
+        if (op_ret == -1 && !dht_inode_missing(op_errno)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "subvolume %s returned -1 (%s)",
+                        prev->this->name, strerror (op_errno));
+                goto out;
+        }
+
+        if (local->call_cnt != 1) {
+                if (local->stbuf.ia_blocks) {
+                        dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
+                        dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
+                }
+                goto out;
+        }
+
+        local->op_errno = op_errno;
+        dht_inode_ctx_get1 (this, inode, &subvol);
+        if (!subvol) {
+                local->rebalance.target_op_fn = dht_fsync2;
+
+                /* Check if the rebalance phase1 is true */
+                if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
+                        dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
+                        dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
+
+                        ret = dht_rebalance_in_progress_check (this, frame);
+                }
+
+                /* Check if the rebalance phase2 is true */
+                if (IS_DHT_MIGRATION_PHASE2 (postbuf)) {
+                        ret = dht_rebalance_complete_check (this, frame);
+                }
+                if (!ret)
+                        return 0;
+        } else {
+                dht_fsync2 (this, frame, 0);
+                return 0;
+        }
+
+out:
+        DHT_STRIP_PHASE1_FLAGS (postbuf);
+        DHT_STRIP_PHASE1_FLAGS (prebuf);
+        DHT_STACK_UNWIND (fsync, frame, op_ret, op_errno,
+                          prebuf, postbuf, xdata);
+
+        return 0;
+}
+
+int
+dht_fsync2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t  *local  = NULL;
+        xlator_t     *subvol = NULL;
+
+        local = frame->local;
+
+        dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+        if (!subvol)
+                subvol = local->cached_subvol;
+
+        local->call_cnt = 2; /* This is the second attempt */
+
+        STACK_WIND (frame, dht_fsync_cbk, subvol, subvol->fops->fsync,
+                    local->fd, local->rebalance.flags, NULL);
+
+        return 0;
+}
+
+int
+dht_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int datasync,
+           dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FSYNC);
+        if (!local) {
+                op_errno = ENOMEM;
+
+                goto err;
+        }
+
+        local->call_cnt = 1;
+        local->rebalance.flags = datasync;
+
+        subvol = local->cached_subvol;
+
+        STACK_WIND (frame, dht_fsync_cbk, subvol, subvol->fops->fsync,
+                    fd, datasync, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+/* TODO: for 'lk()' call, we need some other special error, may be ESTALE to
+   indicate that lock migration happened on the fd, so we can consider it as
+   phase 2 of migration */
+int
+dht_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+            int op_ret, int op_errno, struct gf_flock *flock, dict_t *xdata)
+{
+        DHT_STACK_UNWIND (lk, frame, op_ret, op_errno, flock, xdata);
+
+        return 0;
+}
+
+
+int
+dht_lk (call_frame_t *frame, xlator_t *this,
+        fd_t *fd, int cmd, struct gf_flock *flock, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        subvol = dht_subvol_get_cached (this, fd->inode);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        /* TODO: for rebalance, we need to preserve the fop arguments */
+        STACK_WIND (frame, dht_lk_cbk, subvol, subvol->fops->lk, fd,
+                    cmd, flock, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (lk, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+
+/* Symlinks are currently not migrated, so no need for any check here */
+int
+dht_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int op_ret, int op_errno, const char *path,
+                  struct iatt *stbuf, dict_t *xdata)
+{
+        dht_local_t *local = NULL;
+
+        local = frame->local;
+        if (op_ret == -1)
+                goto err;
+
+        if (!local) {
+                op_ret = -1;
+                op_errno = EINVAL;
+        }
+
+err:
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
+        DHT_STACK_UNWIND (readlink, frame, op_ret, op_errno, path, stbuf, xdata);
+
+        return 0;
+}
+
+
+int
+dht_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+              dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+
+        local = dht_local_init (frame, loc, NULL, GF_FOP_READLINK);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        STACK_WIND (frame, dht_readlink_cbk,
+                    subvol, subvol->fops->readlink,
+                    loc, size, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (readlink, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+/* Currently no translators on top of 'distribute' will be using
+ * below fops, hence not implementing 'migration' related checks
+ */
+
+int
+dht_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        DHT_STACK_UNWIND (xattrop, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+}
+
+
+int
+dht_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
+             gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+
+        local = dht_local_init (frame, loc, NULL, GF_FOP_XATTROP);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        local->call_cnt = 1;
+
+        STACK_WIND (frame,
+                    dht_xattrop_cbk,
+                    subvol, subvol->fops->xattrop,
+                    loc, flags, dict, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+
+
+int
+dht_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        DHT_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+}
+
+
+int
+dht_fxattrop (call_frame_t *frame, xlator_t *this,
+              fd_t *fd, gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        subvol = dht_subvol_get_cached (this, fd->inode);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        STACK_WIND (frame,
+                    dht_fxattrop_cbk,
+                    subvol, subvol->fops->fxattrop,
+                    fd, flags, dict, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+
+
+int
+dht_inodelk_cbk (call_frame_t *frame, void *cookie,
+                 xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
+
+{
+        DHT_STACK_UNWIND (inodelk, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+
+int32_t
+dht_inodelk (call_frame_t *frame, xlator_t *this, const char *volume,
+             loc_t *loc, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+
+        local = dht_local_init (frame, loc, NULL, GF_FOP_INODELK);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        local->call_cnt = 1;
+
+        STACK_WIND (frame,
+                    dht_inodelk_cbk,
+                    subvol, subvol->fops->inodelk,
+                    volume, loc, cmd, lock, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (inodelk, frame, -1, op_errno, NULL);
+
+        return 0;
+}
+
+
+int
+dht_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, dict_t *xdata)
+
+{
+        DHT_STACK_UNWIND (finodelk, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+
+int
+dht_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
+              fd_t *fd, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        subvol = dht_subvol_get_cached (this, fd->inode);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+
+        STACK_WIND (frame, dht_finodelk_cbk, subvol, subvol->fops->finodelk,
+                    volume, fd, cmd, lock, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (finodelk, frame, -1, op_errno, NULL);
+
+        return 0;
+}
diff --git a/xlators/cluster/dht/src/dht-inode-write.c b/xlators/cluster/dht/src/dht-inode-write.c
new file mode 100644
index 000000000..363bff3bf
--- /dev/null
+++ b/xlators/cluster/dht/src/dht-inode-write.c
@@ -0,0 +1,1013 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "dht-common.h"
+
+int dht_writev2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_truncate2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_setattr2 (xlator_t *this, call_frame_t *frame, int ret);
+int dht_fallocate2(xlator_t *this, call_frame_t *frame, int op_ret);
+int dht_discard2(xlator_t *this, call_frame_t *frame, int op_ret);
+int dht_zerofill2(xlator_t *this, call_frame_t *frame, int op_ret);
+
+int
+dht_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int op_ret, int op_errno, struct iatt *prebuf,
+                struct iatt *postbuf, dict_t *xdata)
+{
+        dht_local_t *local = NULL;
+        int          ret   = -1;
+        xlator_t    *subvol = NULL;
+
+        if (op_ret == -1 && !dht_inode_missing(op_errno)) {
+                goto out;
+        }
+
+        local = frame->local;
+        if (!local) {
+                op_ret = -1;
+                op_errno = EINVAL;
+                goto out;
+        }
+
+        if (local->call_cnt != 1) {
+                /* preserve the modes of source */
+                if (local->stbuf.ia_blocks) {
+                        dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
+                        dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
+                }
+                goto out;
+        }
+
+        local->rebalance.target_op_fn = dht_writev2;
+
+        local->op_errno = op_errno;
+        /* Phase 2 of migration */
+        if (IS_DHT_MIGRATION_PHASE2 (postbuf)) {
+                ret = dht_rebalance_complete_check (this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+        /* Check if the rebalance phase1 is true */
+        if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
+                dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
+                dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
+
+                ret = dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+                if (subvol) {
+                        dht_writev2 (this, frame, 0);
+                        return 0;
+                }
+                ret = dht_rebalance_in_progress_check (this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+out:
+        DHT_STRIP_PHASE1_FLAGS (postbuf);
+        DHT_STRIP_PHASE1_FLAGS (prebuf);
+
+        DHT_STACK_UNWIND (writev, frame, op_ret, op_errno, prebuf, postbuf,
+                          xdata);
+
+        return 0;
+}
+
+int
+dht_writev2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t  *local  = NULL;
+        xlator_t     *subvol = NULL;
+
+        local = frame->local;
+
+        dht_inode_ctx_get1 (this, local->fd->inode, &subvol);
+
+        if (!subvol)
+                subvol = local->cached_subvol;
+
+        local->call_cnt = 2; /* This is the second attempt */
+
+        STACK_WIND (frame, dht_writev_cbk,
+                    subvol, subvol->fops->writev,
+                    local->fd, local->rebalance.vector, local->rebalance.count,
+                    local->rebalance.offset, local->rebalance.flags,
+                    local->rebalance.iobref, NULL);
+
+        return 0;
+}
+
+int
+dht_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
+            struct iovec *vector, int count, off_t off, uint32_t flags,
+            struct iobref *iobref, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_WRITE);
+        if (!local) {
+
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+
+        local->rebalance.vector = iov_dup (vector, count);
+        local->rebalance.offset = off;
+        local->rebalance.count = count;
+        local->rebalance.flags = flags;
+        local->rebalance.iobref = iobref_ref (iobref);
+        local->call_cnt = 1;
+
+        STACK_WIND (frame, dht_writev_cbk,
+                    subvol, subvol->fops->writev,
+                    fd, vector, count, off, flags, iobref, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+
+int
+dht_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int op_ret, int op_errno, struct iatt *prebuf,
+                  struct iatt *postbuf, dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           ret = -1;
+        xlator_t    *subvol = NULL;
+        inode_t      *inode = NULL;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+        local = frame->local;
+        prev = cookie;
+
+        if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
+                local->op_errno = op_errno;
+                local->op_ret = -1;
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "subvolume %s returned -1 (%s)",
+                        prev->this->name, strerror (op_errno));
+
+                goto out;
+        }
+
+        if (local->call_cnt != 1) {
+                if (local->stbuf.ia_blocks) {
+                        dht_iatt_merge (this, postbuf, &local->stbuf, NULL);
+                        dht_iatt_merge (this, prebuf, &local->prebuf, NULL);
+                }
+                goto out;
+        }
+
+        local->rebalance.target_op_fn = dht_truncate2;
+
+        local->op_errno = op_errno;
+        /* Phase 2 of migration */
+        if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
+                ret = dht_rebalance_complete_check (this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+        /* Check if the rebalance phase1 is true */
+        if (IS_DHT_MIGRATION_PHASE1 (postbuf)) {
+                dht_iatt_merge (this, &local->stbuf, postbuf, NULL);
+                dht_iatt_merge (this, &local->prebuf, prebuf, NULL);
+                inode = (local->fd) ? local->fd->inode : local->loc.inode;
+                dht_inode_ctx_get1 (this, inode, &subvol);
+                if (subvol) {
+                        dht_truncate2 (this, frame, 0);
+                        return 0;
+                }
+                ret = dht_rebalance_in_progress_check (this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+out:
+        DHT_STRIP_PHASE1_FLAGS (postbuf);
+        DHT_STRIP_PHASE1_FLAGS (prebuf);
+        DHT_STACK_UNWIND (truncate, frame, op_ret, op_errno,
+                          prebuf, postbuf, xdata);
+err:
+        return 0;
+}
+
+
+int
+dht_truncate2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t  *local  = NULL;
+        xlator_t     *subvol = NULL;
+        inode_t      *inode = NULL;
+
+        local = frame->local;
+
+        inode = local->fd ? local->fd->inode : local->loc.inode;
+
+        dht_inode_ctx_get1 (this, inode, &subvol);
+        if (!subvol)
+                subvol = local->cached_subvol;
+
+        local->call_cnt = 2; /* This is the second attempt */
+
+        if (local->fop == GF_FOP_TRUNCATE) {
+                STACK_WIND (frame, dht_truncate_cbk, subvol,
+                            subvol->fops->truncate, &local->loc,
+                            local->rebalance.offset, NULL);
+        } else {
+                STACK_WIND (frame, dht_truncate_cbk, subvol,
+                            subvol->fops->ftruncate, local->fd,
+                            local->rebalance.offset, NULL);
+        }
+
+        return 0;
+}
+
+int
+dht_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+              dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+
+        local = dht_local_init (frame, loc, NULL, GF_FOP_TRUNCATE);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        local->rebalance.offset = offset;
+        local->call_cnt = 1;
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        STACK_WIND (frame, dht_truncate_cbk,
+                    subvol, subvol->fops->truncate,
+                    loc, offset, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+int
+dht_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+               dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        int           op_errno = -1;
+        dht_local_t  *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FTRUNCATE);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        local->rebalance.offset = offset;
+        local->call_cnt = 1;
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        STACK_WIND (frame, dht_truncate_cbk,
+                    subvol, subvol->fops->ftruncate,
+                    fd, offset, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int
+dht_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                  int op_ret, int op_errno, struct iatt *prebuf,
+                  struct iatt *postbuf, dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           ret = -1;
+        xlator_t    *subvol = NULL;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+        local = frame->local;
+        prev = cookie;
+
+        if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
+                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) && !dht_inode_missing(op_errno)) {
+                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) && !dht_inode_missing(op_errno)) {
+                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,
+            off_t len, dict_t *xdata)
+{
+        xlator_t     *subvol       = NULL;
+        int           op_errno     = -1;
+        dht_local_t  *local        = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_ZEROFILL);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        local->rebalance.offset = offset;
+        local->rebalance.size = len;
+
+        local->call_cnt = 1;
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        STACK_WIND (frame, dht_zerofill_cbk, subvol, subvol->fops->zerofill,
+                    fd, offset, len, xdata);
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (zerofill, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+
+/* handle cases of migration here for 'setattr()' calls */
+int
+dht_file_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int op_ret, int op_errno, struct iatt *prebuf,
+                      struct iatt *postbuf, dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           ret = -1;
+
+        local = frame->local;
+        prev = cookie;
+
+        local->op_errno = op_errno;
+        if ((op_ret == -1) && !dht_inode_missing(op_errno)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "subvolume %s returned -1 (%s)",
+                        prev->this->name, strerror (op_errno));
+                goto out;
+        }
+
+        if (local->call_cnt != 1)
+                goto out;
+
+        local->rebalance.target_op_fn = dht_setattr2;
+
+        /* Phase 2 of migration */
+        if ((op_ret == -1) || IS_DHT_MIGRATION_PHASE2 (postbuf)) {
+                ret = dht_rebalance_complete_check (this, frame);
+                if (!ret)
+                        return 0;
+        }
+
+        /* At the end of the migration process, whatever 'attr' we
+           have on source file will be migrated to destination file
+           in one shot, hence we don't need to check for in progress
+           state here (ie, PHASE1) */
+out:
+        DHT_STRIP_PHASE1_FLAGS (postbuf);
+        DHT_STRIP_PHASE1_FLAGS (prebuf);
+        DHT_STACK_UNWIND (setattr, frame, op_ret, op_errno,
+                          prebuf, postbuf, xdata);
+
+        return 0;
+}
+
+int
+dht_setattr2 (xlator_t *this, call_frame_t *frame, int op_ret)
+{
+        dht_local_t  *local  = NULL;
+        xlator_t     *subvol = NULL;
+        inode_t      *inode = NULL;
+
+        local = frame->local;
+
+        inode = (local->fd) ? local->fd->inode : local->loc.inode;
+
+        dht_inode_ctx_get1 (this, inode, &subvol);
+
+        if (!subvol)
+                subvol = local->cached_subvol;
+
+        local->call_cnt = 2; /* This is the second attempt */
+
+        if (local->fop == GF_FOP_SETATTR) {
+                STACK_WIND (frame, dht_file_setattr_cbk, subvol,
+                            subvol->fops->setattr, &local->loc,
+                            &local->rebalance.stbuf, local->rebalance.flags,
+                            NULL);
+        } else {
+                STACK_WIND (frame, dht_file_setattr_cbk, subvol,
+                            subvol->fops->fsetattr, local->fd,
+                            &local->rebalance.stbuf, local->rebalance.flags,
+                            NULL);
+        }
+
+        return 0;
+}
+
+
+/* Keep the existing code same for all the cases other than regular file */
+int
+dht_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int op_ret, int op_errno, struct iatt *statpre,
+                 struct iatt *statpost, dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
+
+
+        local = frame->local;
+        prev = cookie;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "subvolume %s returned -1 (%s)",
+                                prev->this->name, strerror (op_errno));
+                        goto unlock;
+                }
+
+                dht_iatt_merge (this, &local->prebuf, statpre, prev->this);
+                dht_iatt_merge (this, &local->stbuf, statpost, prev->this);
+
+                local->op_ret = 0;
+        }
+unlock:
+        UNLOCK (&frame->lock);
+
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt))
+                DHT_STACK_UNWIND (setattr, frame, local->op_ret, local->op_errno,
+                                  &local->prebuf, &local->stbuf, xdata);
+
+        return 0;
+}
+
+
+int
+dht_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+             struct iatt *stbuf, int32_t valid, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        dht_layout_t *layout = NULL;
+        dht_local_t  *local  = NULL;
+        int           op_errno = -1;
+        int           i = -1;
+        int           call_cnt = 0;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+
+        local = dht_local_init (frame, loc, NULL, GF_FOP_SETATTR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        layout = local->layout;
+        if (!layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no layout for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        if (!layout_is_sane (layout)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "layout is not sane for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        if (IA_ISREG (loc->inode->ia_type)) {
+                /* in the regular file _cbk(), we need to check for
+                   migration possibilities */
+                local->rebalance.stbuf = *stbuf;
+                local->rebalance.flags = valid;
+                local->call_cnt = 1;
+                subvol = local->cached_subvol;
+
+                STACK_WIND (frame, dht_file_setattr_cbk, subvol,
+                            subvol->fops->setattr,
+                            loc, stbuf, valid, xdata);
+
+                return 0;
+        }
+
+        local->call_cnt = call_cnt = layout->cnt;
+
+        for (i = 0; i < call_cnt; i++) {
+                STACK_WIND (frame, dht_setattr_cbk,
+                            layout->list[i].xlator,
+                            layout->list[i].xlator->fops->setattr,
+                            loc, stbuf, valid, xdata);
+        }
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int
+dht_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
+              int32_t valid, dict_t *xdata)
+{
+        xlator_t     *subvol = NULL;
+        dht_layout_t *layout = NULL;
+        dht_local_t  *local  = NULL;
+        int           op_errno = -1;
+        int           i = -1;
+        int           call_cnt = 0;
+
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FSETATTR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        layout = local->layout;
+        if (!layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no layout for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        if (!layout_is_sane (layout)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "layout is not sane for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        if (IA_ISREG (fd->inode->ia_type)) {
+                /* in the regular file _cbk(), we need to check for
+                   migration possibilities */
+                local->rebalance.stbuf = *stbuf;
+                local->rebalance.flags = valid;
+                local->call_cnt = 1;
+                subvol = local->cached_subvol;
+
+                STACK_WIND (frame, dht_file_setattr_cbk, subvol,
+                            subvol->fops->fsetattr,
+                            fd, stbuf, valid, xdata);
+
+                return 0;
+        }
+
+        local->call_cnt = call_cnt = layout->cnt;
+
+        for (i = 0; i < call_cnt; i++) {
+                STACK_WIND (frame, dht_setattr_cbk,
+                            layout->list[i].xlator,
+                            layout->list[i].xlator->fops->fsetattr,
+                            fd, stbuf, valid, xdata);
+        }
+
+        return 0;
+
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
diff --git a/xlators/cluster/dht/src/dht-layout.c b/xlators/cluster/dht/src/dht-layout.c
index 8a5f45d88..e1a37b77c 100644
--- a/xlators/cluster/dht/src/dht-layout.c
+++ b/xlators/cluster/dht/src/dht-layout.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 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
@@ -34,32 +25,32 @@
 
 #define layout_size(cnt) (layout_base_size + (cnt * layout_entry_size))
 
-
 dht_layout_t *
 dht_layout_new (xlator_t *this, int cnt)
 {
-	dht_layout_t *layout = NULL;
+        dht_layout_t *layout = NULL;
         dht_conf_t   *conf = NULL;
 
-
         conf = this->private;
 
-	layout = GF_CALLOC (1, layout_size (cnt),
+        layout = GF_CALLOC (1, layout_size (cnt),
                             gf_dht_mt_dht_layout_t);
-	if (!layout) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto out;
-	}
+        if (!layout) {
+                goto out;
+        }
 
         layout->type = DHT_HASH_TYPE_DM;
-	layout->cnt = cnt;
-        if (conf)
+        layout->cnt = cnt;
+
+        if (conf) {
+                layout->spread_cnt = conf->dir_spread_cnt;
                 layout->gen = conf->gen;
+        }
 
         layout->ref = 1;
+
 out:
-	return layout;
+        return layout;
 }
 
 
@@ -67,9 +58,7 @@ 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)
@@ -77,9 +66,8 @@ 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++;
                 }
         }
@@ -97,7 +85,6 @@ 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)
@@ -105,16 +92,13 @@ 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);
         }
 
@@ -129,7 +113,7 @@ 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;
@@ -167,532 +151,626 @@ dht_layout_ref (xlator_t *this, dht_layout_t *layout)
 xlator_t *
 dht_layout_search (xlator_t *this, dht_layout_t *layout, const char *name)
 {
-	uint32_t   hash = 0;
+        uint32_t   hash = 0;
         xlator_t  *subvol = NULL;
-	int        i = 0;
-	int        ret = 0;
+        int        i = 0;
+        int        ret = 0;
 
 
-	ret = dht_hash_compute (layout->type, name, &hash);
-	if (ret != 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"hash computation failed for type=%d name=%s",
-			layout->type, name);
-		goto out;
-	}
+        ret = dht_hash_compute (this, layout->type, name, &hash);
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "hash computation failed for type=%d name=%s",
+                        layout->type, name);
+                goto out;
+        }
 
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].start <= hash
-		    && layout->list[i].stop >= hash) {
-			subvol = layout->list[i].xlator;
-			break;
-		}
-	}
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].start <= hash
+                    && layout->list[i].stop >= hash) {
+                        subvol = layout->list[i].xlator;
+                        break;
+                }
+        }
 
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume for hash (value) = %u", hash);
-	}
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "no subvolume for hash (value) = %u", hash);
+        }
 
 out:
-	return subvol;
+        return subvol;
 }
 
 
 dht_layout_t *
 dht_layout_for_subvol (xlator_t *this, xlator_t *subvol)
 {
-	dht_conf_t   *conf = NULL;
-	dht_layout_t *layout = NULL;
-	int           i = 0;
+        dht_conf_t   *conf = NULL;
+        dht_layout_t *layout = NULL;
+        int           i = 0;
 
-	conf = this->private;
+        conf = this->private;
         if (!conf)
                 goto out;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		if (conf->subvolumes[i] == subvol) {
-			layout = conf->file_layouts[i];
-			break;
-		}
-	}
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (conf->subvolumes[i] == subvol) {
+                        layout = conf->file_layouts[i];
+                        break;
+                }
+        }
 
 out:
-	return layout;
+        return layout;
 }
 
 
 int
 dht_layouts_init (xlator_t *this, dht_conf_t *conf)
 {
-	dht_layout_t *layout = NULL;
-	int           i = 0;
-	int           ret = -1;
+        dht_layout_t *layout = NULL;
+        int           i = 0;
+        int           ret = -1;
 
         if (!conf)
                 goto out;
 
-	conf->file_layouts = GF_CALLOC (conf->subvolume_cnt,
-				        sizeof (dht_layout_t *),
+        conf->file_layouts = GF_CALLOC (conf->subvolume_cnt,
+                                        sizeof (dht_layout_t *),
                                         gf_dht_mt_dht_layout_t);
-	if (!conf->file_layouts) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto out;
-	}
+        if (!conf->file_layouts) {
+                goto out;
+        }
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		layout = dht_layout_new (this, 1);
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                layout = dht_layout_new (this, 1);
 
-		if (!layout) {
-			goto out;
-		}
+                if (!layout) {
+                        goto out;
+                }
 
-		layout->preset = 1;
+                layout->preset = 1;
 
-		layout->list[0].xlator = conf->subvolumes[i];
+                layout->list[0].xlator = conf->subvolumes[i];
 
-		conf->file_layouts[i] = layout;
-	}
+                conf->file_layouts[i] = layout;
+        }
 
-	ret = 0;
+        ret = 0;
 out:
-	return ret;
+        return ret;
 }
 
 
 int
 dht_disk_layout_extract (xlator_t *this, dht_layout_t *layout,
-			 int pos, int32_t **disk_layout_p)
+                         int pos, int32_t **disk_layout_p)
 {
-	int      ret = -1;
-	int32_t *disk_layout = NULL;
+        int      ret = -1;
+        int32_t *disk_layout = NULL;
 
-	disk_layout = GF_CALLOC (5, sizeof (int),
+        disk_layout = GF_CALLOC (5, sizeof (int),
                                  gf_dht_mt_int32_t);
-	if (!disk_layout) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto out;
-	}
+        if (!disk_layout) {
+                goto out;
+        }
 
-	disk_layout[0] = hton32 (1);
-	disk_layout[1] = hton32 (layout->type);
-	disk_layout[2] = hton32 (layout->list[pos].start);
-	disk_layout[3] = hton32 (layout->list[pos].stop);
+        disk_layout[0] = hton32 (1);
+        disk_layout[1] = hton32 (layout->type);
+        disk_layout[2] = hton32 (layout->list[pos].start);
+        disk_layout[3] = hton32 (layout->list[pos].stop);
 
-	if (disk_layout_p)
-		*disk_layout_p = disk_layout;
-	ret = 0;
+        if (disk_layout_p)
+                *disk_layout_p = disk_layout;
+        else
+                GF_FREE (disk_layout);
+
+        ret = 0;
 
 out:
-	return ret;
+        return ret;
 }
 
 
 int
 dht_disk_layout_merge (xlator_t *this, dht_layout_t *layout,
-		       int pos, void *disk_layout_raw)
+		       int pos, void *disk_layout_raw, int disk_layout_len)
 {
-	int      cnt = 0;
-	int      type = 0;
-	int      start_off = 0;
-	int      stop_off = 0;
+        int      cnt = 0;
+        int      type = 0;
+        int      start_off = 0;
+        int      stop_off = 0;
         int      disk_layout[4];
 
-	/* TODO: assert disk_layout_ptr is of required length */
+	if (!disk_layout_raw) {
+		gf_log (this->name, GF_LOG_CRITICAL,
+                        "error no layout on disk for merge");
+		return -1;
+	}
 
-        memcpy (disk_layout, disk_layout_raw, sizeof (disk_layout));
+	GF_ASSERT (disk_layout_len == sizeof (disk_layout));
+
+        memcpy (disk_layout, disk_layout_raw, disk_layout_len);
 
-	cnt  = ntoh32 (disk_layout[0]);
-	if (cnt != 1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"disk layout has invalid count %d", cnt);
+        cnt  = ntoh32 (disk_layout[0]);
+        if (cnt != 1) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "disk layout has invalid count %d", cnt);
+                return -1;
+        }
+
+        type = ntoh32 (disk_layout[1]);
+	switch (type) {
+        case DHT_HASH_TYPE_DM_USER:
+                gf_log (this->name, GF_LOG_DEBUG, "found user-set layout");
+                layout->type = type;
+                /* Fall through. */
+	case DHT_HASH_TYPE_DM:
+		break;
+        default:
+		gf_log (this->name, GF_LOG_CRITICAL,
+			"Catastrophic error layout with unknown type found %d",
+			disk_layout[1]);
 		return -1;
 	}
 
-	/* TODO: assert type is compatible */
-	type      = ntoh32 (disk_layout[1]);
-	start_off = ntoh32 (disk_layout[2]);
-	stop_off  = ntoh32 (disk_layout[3]);
+        start_off = ntoh32 (disk_layout[2]);
+        stop_off  = ntoh32 (disk_layout[3]);
 
-	layout->list[pos].start = start_off;
-	layout->list[pos].stop  = stop_off;
+        layout->list[pos].start = start_off;
+        layout->list[pos].stop  = stop_off;
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"merged to layout: %u - %u (type %d) from %s",
-		start_off, stop_off, type,
-		layout->list[pos].xlator->name);
+        gf_log (this->name, GF_LOG_TRACE,
+                "merged to layout: %u - %u (type %d) from %s",
+                start_off, stop_off, type,
+                layout->list[pos].xlator->name);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_layout_merge (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
-		  int op_ret, int op_errno, dict_t *xattr)
+                  int op_ret, int op_errno, dict_t *xattr)
 {
-	int      i     = 0;
-	int      ret   = -1;
-	int      err   = -1;
-	void    *disk_layout_raw = NULL;
-
-
-	if (op_ret != 0) {
-		err = op_errno;
-	}
+        int         i     = 0;
+        int         ret   = -1;
+        int         err   = -1;
+        void       *disk_layout_raw = NULL;
+        int         disk_layout_len = 0;
+        dht_conf_t *conf = this->private;
+
+        if (op_ret != 0) {
+                err = op_errno;
+        }
 
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].xlator == NULL) {
-			layout->list[i].err    = err;
-			layout->list[i].xlator = subvol;
-			break;
-		}
-	}
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].xlator == NULL) {
+                        layout->list[i].err    = err;
+                        layout->list[i].xlator = subvol;
+                        break;
+                }
+        }
 
-	if (op_ret != 0) {
-		ret = 0;
-		goto out;
-	}
+        if (op_ret != 0) {
+                ret = 0;
+                goto out;
+        }
 
-	if (xattr) {
-		/* during lookup and not mkdir */
-		ret = dict_get_ptr (xattr, "trusted.glusterfs.dht",
-				    &disk_layout_raw);
-	}
+        if (xattr) {
+                /* during lookup and not mkdir */
+                ret = dict_get_ptr_and_len (xattr, conf->xattr_name,
+					    &disk_layout_raw, &disk_layout_len);
+        }
 
-	if (ret != 0) {
-		layout->list[i].err = -1;
-		gf_log (this->name, GF_LOG_TRACE,
-			"missing disk layout on %s. err = %d",
-			subvol->name, err);
-		ret = 0;
-		goto out;
-	}
+        if (ret != 0) {
+                layout->list[i].err = 0;
+                gf_log (this->name, GF_LOG_TRACE,
+                        "missing disk layout on %s. err = %d",
+                        subvol->name, err);
+                ret = 0;
+                goto out;
+        }
 
-	ret = dht_disk_layout_merge (this, layout, i, disk_layout_raw);
-	if (ret != 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"layout merge from subvolume %s failed",
-			subvol->name);
-		goto out;
-	}
-	layout->list[i].err = 0;
+        ret = dht_disk_layout_merge (this, layout, i, disk_layout_raw,
+				     disk_layout_len);
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "layout merge from subvolume %s failed",
+                        subvol->name);
+                goto out;
+        }
+        layout->list[i].err = 0;
 
 out:
-	return ret;
+        return ret;
 }
 
 
 void
 dht_layout_entry_swap (dht_layout_t *layout, int i, int j)
 {
-	uint32_t  start_swap = 0;
-	uint32_t  stop_swap = 0;
-	xlator_t *xlator_swap = 0;
-	int       err_swap = 0;
-
-
-	start_swap  = layout->list[i].start;
-	stop_swap   = layout->list[i].stop;
-	xlator_swap = layout->list[i].xlator;
-	err_swap    = layout->list[i].err;
-
-	layout->list[i].start  = layout->list[j].start;
-	layout->list[i].stop   = layout->list[j].stop;
-	layout->list[i].xlator = layout->list[j].xlator;
-	layout->list[i].err    = layout->list[j].err;
-
-	layout->list[j].start  = start_swap;
-	layout->list[j].stop   = stop_swap;
-	layout->list[j].xlator = xlator_swap;
-	layout->list[j].err    = err_swap;
+        uint32_t  start_swap = 0;
+        uint32_t  stop_swap = 0;
+        xlator_t *xlator_swap = 0;
+        int       err_swap = 0;
+
+        start_swap  = layout->list[i].start;
+        stop_swap   = layout->list[i].stop;
+        xlator_swap = layout->list[i].xlator;
+        err_swap    = layout->list[i].err;
+
+        layout->list[i].start  = layout->list[j].start;
+        layout->list[i].stop   = layout->list[j].stop;
+        layout->list[i].xlator = layout->list[j].xlator;
+        layout->list[i].err    = layout->list[j].err;
+
+        layout->list[j].start  = start_swap;
+        layout->list[j].stop   = stop_swap;
+        layout->list[j].xlator = xlator_swap;
+        layout->list[j].err    = err_swap;
+}
+
+void
+dht_layout_range_swap (dht_layout_t *layout, int i, int j)
+{
+        uint32_t  start_swap = 0;
+        uint32_t  stop_swap = 0;
+
+        start_swap  = layout->list[i].start;
+        stop_swap   = layout->list[i].stop;
+
+        layout->list[i].start  = layout->list[j].start;
+        layout->list[i].stop   = layout->list[j].stop;
+
+        layout->list[j].start  = start_swap;
+        layout->list[j].stop   = stop_swap;
 }
 
 int64_t
 dht_layout_entry_cmp_volname (dht_layout_t *layout, int i, int j)
 {
-	return (strcmp (layout->list[i].xlator->name, 
-			layout->list[j].xlator->name));
+        return (strcmp (layout->list[i].xlator->name,
+                        layout->list[j].xlator->name));
+}
+
+
+gf_boolean_t
+dht_is_subvol_in_layout (dht_layout_t *layout, xlator_t *xlator)
+{
+        int i = 0;
+
+        for (i = 0; i < layout->cnt; i++) {
+                /* Check if xlator is already part of layout, and layout is
+                 * non-zero. */
+                if (!strcmp (layout->list[i].xlator->name, xlator->name)) {
+                        if (layout->list[i].start != layout->list[i].stop)
+                                return _gf_true;
+                        break;
+                }
+        }
+        return _gf_false;
 }
 
 int64_t
 dht_layout_entry_cmp (dht_layout_t *layout, int i, int j)
 {
-	int64_t diff = 0;
+        int64_t diff = 0;
 
-	if (layout->list[i].err || layout->list[j].err)
-		diff = layout->list[i].err - layout->list[j].err;
-	else
-		diff = (int64_t) layout->list[i].start
-			- (int64_t) layout->list[j].start;
+        /* swap zero'ed out layouts to front, if needed */
+        if (!layout->list[j].start && !layout->list[j].stop) {
+                diff = (int64_t) layout->list[i].stop
+                       - (int64_t) layout->list[j].stop;
+                       goto out;
+        }
+        diff = (int64_t) layout->list[i].start
+                - (int64_t) layout->list[j].start;
 
-	return diff;
+out:
+        return diff;
 }
 
 
 int
 dht_layout_sort (dht_layout_t *layout)
 {
-	int       i = 0;
-	int       j = 0;
-	int64_t   ret = 0;
-
-	/* TODO: O(n^2) -- bad bad */
-
-	for (i = 0; i < layout->cnt - 1; i++) {
-		for (j = i + 1; j < layout->cnt; j++) {
-			ret = dht_layout_entry_cmp (layout, i, j);
-			if (ret > 0)
-				dht_layout_entry_swap (layout, i, j);
-		}
-	}
+        int       i = 0;
+        int       j = 0;
+        int64_t   ret = 0;
 
-	return 0;
+        /* TODO: O(n^2) -- bad bad */
+
+        for (i = 0; i < layout->cnt - 1; i++) {
+                for (j = i + 1; j < layout->cnt; j++) {
+                        ret = dht_layout_entry_cmp (layout, i, j);
+                        if (ret > 0)
+                                dht_layout_entry_swap (layout, i, j);
+                }
+        }
+
+        return 0;
 }
 
 int
 dht_layout_sort_volname (dht_layout_t *layout)
 {
-	int       i = 0;
-	int       j = 0;
-	int64_t   ret = 0;
-
-	/* TODO: O(n^2) -- bad bad */
-
-	for (i = 0; i < layout->cnt - 1; i++) {
-		for (j = i + 1; j < layout->cnt; j++) {
-			ret = dht_layout_entry_cmp_volname (layout, i, j);
-			if (ret > 0)
-				dht_layout_entry_swap (layout, i, j);
-		}
-	}
+        int       i = 0;
+        int       j = 0;
+        int64_t   ret = 0;
+
+        /* TODO: O(n^2) -- bad bad */
+
+        for (i = 0; i < layout->cnt - 1; i++) {
+                for (j = i + 1; j < layout->cnt; j++) {
+                        ret = dht_layout_entry_cmp_volname (layout, i, j);
+                        if (ret > 0)
+                                dht_layout_entry_swap (layout, i, j);
+                }
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_layout_anomalies (xlator_t *this, loc_t *loc, dht_layout_t *layout,
-		      uint32_t *holes_p, uint32_t *overlaps_p,
-		      uint32_t *missing_p, uint32_t *down_p, uint32_t *misc_p)
+                      uint32_t *holes_p, uint32_t *overlaps_p,
+                      uint32_t *missing_p, uint32_t *down_p, uint32_t *misc_p,
+                      uint32_t *no_space_p)
 {
-	uint32_t    overlaps = 0;
-	uint32_t    missing  = 0;
-	uint32_t    down     = 0;
-	uint32_t    misc     = 0;
-	uint32_t    hole_cnt = 0;
-	uint32_t    overlap_cnt = 0;
-	int         i = 0;
-	int         ret = 0;
-	uint32_t    prev_stop = 0;
-	uint32_t    last_stop = 0;
-	char        is_virgin = 1;
-
-	/* TODO: explain WTF is happening */
-
-	last_stop = layout->list[0].start - 1;
-	prev_stop = last_stop;
-
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].err) {
-			switch (layout->list[i].err) {
-			case -1:
-			case ENOENT:
-				missing++;
-				break;
-			case ENOTCONN:
-				down++;
-				break;
-			case ENOSPC:
-				down++;
-				break;
-			default:
-				misc++;
-			}
-			continue;
-		}
-
-		is_virgin = 0;
-
-		if ((prev_stop + 1) < layout->list[i].start) {
-			hole_cnt++;
-		}
-
-		if ((prev_stop + 1) > layout->list[i].start) {
-			overlap_cnt++;
-			overlaps += ((prev_stop + 1) - layout->list[i].start);
-		}
-		prev_stop = layout->list[i].stop;
-	}
+        uint32_t    overlaps = 0;
+        uint32_t    missing  = 0;
+        uint32_t    down     = 0;
+        uint32_t    misc     = 0;
+        uint32_t    hole_cnt = 0;
+        uint32_t    overlap_cnt = 0;
+        int         i = 0;
+        int         ret = 0;
+        uint32_t    prev_stop = 0;
+        uint32_t    last_stop = 0;
+        char        is_virgin = 1;
+        uint32_t    no_space  = 0;
+
+        /* This funtion scans through the layout spread of a directory to
+           check if there are any anomalies. Prior to calling this function
+           the layout entries should be sorted in the ascending order.
+
+           If the layout entry has err != 0
+                then increment the corresponding anomaly.
+           else
+                if (start of the current layout entry > stop + 1 of previous
+                   non erroneous layout entry)
+                        then it indicates a hole in the layout
+                if (start of the current layout entry < stop + 1 of previous
+                    non erroneous layout entry)
+                         then it indicates an overlap in the layout
+        */
+        last_stop = layout->list[0].start - 1;
+        prev_stop = last_stop;
+
+        for (i = 0; i < layout->cnt; i++) {
+                switch (layout->list[i].err) {
+                case -1:
+                case ENOENT:
+                case ESTALE:
+                        missing++;
+                        continue;
+                case ENOTCONN:
+                        down++;
+                        continue;
+                case ENOSPC:
+                        no_space++;
+                        continue;
+                case 0:
+                        /* if err == 0 and start == stop, then it is a non misc++;
+                         * participating subvolume(spread-cnt). Then, do not
+                         * check for anomalies. If start != stop, then treat it
+                         * as misc err */
+                        if (layout->list[i].start == layout->list[i].stop) {
+                                continue;
+                        }
+                        break;
+                default:
+                        misc++;
+                        continue;
+                 }
+
+                is_virgin = 0;
+
+                if ((prev_stop + 1) < layout->list[i].start) {
+                        hole_cnt++;
+                }
+
+                if ((prev_stop + 1) > layout->list[i].start) {
+                        overlap_cnt++;
+                        overlaps += ((prev_stop + 1) - layout->list[i].start);
+                }
+                prev_stop = layout->list[i].stop;
+        }
 
-	if ((last_stop - prev_stop) || is_virgin)
-	    hole_cnt++;
+        if ((last_stop - prev_stop) || is_virgin)
+                hole_cnt++;
 
-	if (holes_p)
-		*holes_p = hole_cnt;
+        if (holes_p)
+                *holes_p = hole_cnt;
 
-	if (overlaps_p)
-		*overlaps_p = overlap_cnt;
+        if (overlaps_p)
+                *overlaps_p = overlap_cnt;
 
-	if (missing_p)
-		*missing_p = missing;
+        if (missing_p)
+                *missing_p = missing;
 
-	if (down_p)
-		*down_p = down;
+        if (down_p)
+                *down_p = down;
 
-	if (misc_p)
-		*misc_p = misc;
+        if (misc_p)
+                *misc_p = misc;
 
-	return ret;
+        if (no_space_p)
+                *no_space_p = no_space;
+
+        return ret;
 }
 
 
 int
 dht_layout_normalize (xlator_t *this, loc_t *loc, dht_layout_t *layout)
 {
-	int          ret   = 0;
-	int          i = 0;
-	uint32_t     holes = 0;
-	uint32_t     overlaps = 0;
-	uint32_t     missing = 0;
-	uint32_t     down = 0;
-	uint32_t     misc = 0;
-
-
-	ret = dht_layout_sort (layout);
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"sort failed?! how the ....");
-		goto out;
-	}
+        int          ret   = 0;
+        int          i = 0;
+        uint32_t     holes = 0;
+        uint32_t     overlaps = 0;
+        uint32_t     missing = 0;
+        uint32_t     down = 0;
+        uint32_t     misc = 0;
+
+        ret = dht_layout_sort (layout);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "sort failed?! how the ....");
+                goto out;
+        }
 
-	ret = dht_layout_anomalies (this, loc, layout,
-				    &holes, &overlaps,
-				    &missing, &down, &misc);
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_WARNING,
-			"error while finding anomalies in %s -- not good news",
-			loc->path);
-		goto out;
-	}
+        ret = dht_layout_anomalies (this, loc, layout,
+                                    &holes, &overlaps,
+                                    &missing, &down, &misc, NULL);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "error while finding anomalies in %s -- not good news",
+                        loc->path);
+                goto out;
+        }
 
-	if (holes || overlaps) {
-		if (missing == layout->cnt) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"directory %s looked up first time",
-				loc->path);
-		} else {
-			gf_log (this->name, GF_LOG_INFO,
-				"found anomalies in %s. holes=%d overlaps=%d",
-				loc->path, holes, overlaps);
-		}
-		ret = 1;
-	}
+        if (holes || overlaps) {
+                if (missing == layout->cnt) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "directory %s looked up first time",
+                                loc->path);
+                } else {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "found anomalies in %s. holes=%d overlaps=%d",
+                                loc->path, holes, overlaps);
+                }
+                ret = -1;
+        }
+
+        for (i = 0; i < layout->cnt; i++) {
+                /* TODO During DHT selfheal rewrite (almost) find a better place
+                 * to detect this - probably in dht_layout_anomalies()
+                 */
+                if (layout->list[i].err > 0) {
+                        gf_log_callingfn (this->name, GF_LOG_DEBUG,
+                                          "path=%s err=%s on subvol=%s",
+                                          loc->path,
+                                          strerror (layout->list[i].err),
+                                          (layout->list[i].xlator ?
+                                           layout->list[i].xlator->name
+                                           : "<>"));
+                        if ((layout->list[i].err == ENOENT) && (ret >= 0)) {
+                                ret++;
+                        }
+                }
+        }
 
-	for (i = 0; i < layout->cnt; i++) {
-	/* TODO During DHT selfheal rewrite (almost) find a better place to 
-	 * detect this - probably in dht_layout_anomalies() 
-	 */
-		if (layout->list[i].err > 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"path=%s err=%s on subvol=%s",
-                                loc->path, strerror (layout->list[i].err),
-                                (layout->list[i].xlator ?
-                                 layout->list[i].xlator->name : "<>"));
-                        if (layout->list[i].err == ENOENT)
-                                ret = 1;
-		}
-	}
 
 out:
-	return ret;
+        return ret;
 }
 
-
 int
-dht_layout_dir_mismatch (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
-			 loc_t *loc, dict_t *xattr)
+dht_dir_has_layout (dict_t *xattr, char *name)
 {
-	int       idx = 0;
-	int       pos = -1;
-	int       ret = 0;
-        int       err = 0;
-        int       dict_ret = 0;
-	int32_t   disk_layout[4];
+
         void     *disk_layout_raw = NULL;
-	int32_t   count = -1;
-	uint32_t  start_off = -1;
-	uint32_t  stop_off = -1;
 
+        return dict_get_ptr (xattr, name, &disk_layout_raw);
+}
 
-	for (idx = 0; idx < layout->cnt; idx++) {
-		if (layout->list[idx].xlator == subvol) {
-			pos = idx;
-			break;
-		}
-	}
-	
-	if (pos == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"%s - no layout info for subvolume %s",
-			loc->path, subvol->name);
-		ret = 1;
-		goto out;
-	}
+int
+dht_layout_dir_mismatch (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
+                         loc_t *loc, dict_t *xattr)
+{
+        int         idx = 0;
+        int         pos = -1;
+        int         ret = 0;
+        int         err = 0;
+        int         dict_ret = 0;
+        int32_t     disk_layout[4];
+        void       *disk_layout_raw = NULL;
+        int32_t     count = -1;
+        uint32_t    start_off = -1;
+        uint32_t    stop_off = -1;
+        dht_conf_t *conf = this->private;
+
+
+        for (idx = 0; idx < layout->cnt; idx++) {
+                if (layout->list[idx].xlator == subvol) {
+                        pos = idx;
+                        break;
+                }
+        }
+
+        if (pos == -1) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "%s - no layout info for subvolume %s",
+                        loc->path, subvol->name);
+                ret = 1;
+                goto out;
+        }
 
         err = layout->list[pos].err;
 
-	if (!xattr) {
+        if (!xattr) {
                 if (err == 0) {
-                        gf_log (this->name, GF_LOG_DEBUG,
+                        gf_log (this->name, GF_LOG_INFO,
                                 "%s - xattr dictionary is NULL",
                                 loc->path);
                         ret = -1;
                 }
-		goto out;
-	}
+                goto out;
+        }
 
-	dict_ret = dict_get_ptr (xattr, "trusted.glusterfs.dht",
+        dict_ret = dict_get_ptr (xattr, conf->xattr_name,
                                  &disk_layout_raw);
 
-	if (dict_ret < 0) {
-                if (err == 0) {
-                        gf_log (this->name, GF_LOG_DEBUG,
+        if (dict_ret < 0) {
+                if (err == 0 && layout->list[pos].stop) {
+                        gf_log (this->name, GF_LOG_INFO,
                                 "%s - disk layout missing", loc->path);
                         ret = -1;
                 }
-		goto out;
-	}
+                goto out;
+        }
 
         memcpy (disk_layout, disk_layout_raw, sizeof (disk_layout));
 
-	count  = ntoh32 (disk_layout[0]);
-	if (count != 1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"%s - disk layout has invalid count %d",
-			loc->path, count);
-		ret = -1;
-		goto out;
-	}
+        count  = ntoh32 (disk_layout[0]);
+        if (count != 1) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s - disk layout has invalid count %d",
+                        loc->path, count);
+                ret = -1;
+                goto out;
+        }
 
-	start_off = ntoh32 (disk_layout[2]);
-	stop_off  = ntoh32 (disk_layout[3]);
-
-	if ((layout->list[pos].start != start_off)
-	    || (layout->list[pos].stop != stop_off)) {
-		gf_log (this->name, GF_LOG_INFO,
-			"subvol: %s; inode layout - %"PRIu32" - %"PRIu32"; "
-			"disk layout - %"PRIu32" - %"PRIu32,
-			layout->list[pos].xlator->name,
-			layout->list[pos].start, layout->list[pos].stop,
-			start_off, stop_off);
-		ret = 1;
-	} else {
-		ret = 0;
-	}
+        start_off = ntoh32 (disk_layout[2]);
+        stop_off  = ntoh32 (disk_layout[3]);
+
+        if ((layout->list[pos].start != start_off)
+            || (layout->list[pos].stop != stop_off)) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "subvol: %s; inode layout - %"PRIu32" - %"PRIu32"; "
+                        "disk layout - %"PRIu32" - %"PRIu32,
+                        layout->list[pos].xlator->name,
+                        layout->list[pos].start, layout->list[pos].stop,
+                        start_off, stop_off);
+                ret = 1;
+        } else {
+                ret = 0;
+        }
 out:
-	return ret;
+        return ret;
 }
 
 
@@ -707,18 +785,18 @@ dht_layout_preset (xlator_t *this, xlator_t *subvol, inode_t *inode)
         if (!conf)
                 goto out;
 
-	layout = dht_layout_for_subvol (this, subvol);
-	if (!layout) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no pre-set layout for subvolume %s",
-			subvol ? subvol->name : "<nil>");
-		ret = -1;
-		goto out;
-	}
+        layout = dht_layout_for_subvol (this, subvol);
+        if (!layout) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "no pre-set layout for subvolume %s",
+                        subvol ? subvol->name : "<nil>");
+                ret = -1;
+                goto out;
+        }
 
         LOCK (&conf->layout_lock);
         {
-                inode_ctx_put (inode, this, (uint64_t)(long)layout);
+                dht_inode_ctx_layout_set (inode, this, layout);
         }
         UNLOCK (&conf->layout_lock);
 
diff --git a/xlators/cluster/dht/src/dht-linkfile.c b/xlators/cluster/dht/src/dht-linkfile.c
index 6afba25a1..dbc9d0b3c 100644
--- a/xlators/cluster/dht/src/dht-linkfile.c
+++ b/xlators/cluster/dht/src/dht-linkfile.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -28,220 +19,310 @@
 #include "compat.h"
 #include "dht-common.h"
 
-
-
 int
-dht_linkfile_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			int op_ret, int op_errno)
+dht_linkfile_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int op_ret, int op_errno,
+                         inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+                         struct iatt *postparent)
 {
-	dht_local_t *local = NULL;
-
-
-	local = frame->local;
-	local->linkfile.linkfile_cbk (frame, cookie, this, op_ret, op_errno,
-				      local->linkfile.inode,
-				      &local->linkfile.stbuf, NULL, NULL);
-
-	return 0;
+        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,
                          struct iatt *stbuf, struct iatt *preparent,
-                         struct iatt *postparent)
+                         struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	call_frame_t *prev = NULL;
-	dict_t       *xattr = NULL;
-	data_t       *str_data = NULL;
-	int           ret = -1;
-
-	local = frame->local;
-	prev  = cookie;
-
-	if (op_ret == -1)
-		goto err;
-
-	xattr = get_new_dict ();
-	if (!xattr) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
-
-	local->linkfile.xattr = dict_ref (xattr);
-	local->linkfile.inode = inode_ref (inode);
-
-	str_data = str_to_data (local->linkfile.srcvol->name);
-	if (!str_data) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
-
-	ret = dict_set (xattr, "trusted.glusterfs.dht.linkto", str_data);
-	if (ret < 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"failed to initialize linkfile data");
-	}
-	str_data = NULL;
-
-	local->linkfile.stbuf = *stbuf;
-
-	STACK_WIND (frame, dht_linkfile_xattr_cbk,
-		    prev->this, prev->this->fops->setxattr,
-		    &local->linkfile.loc, local->linkfile.xattr, 0);
-
-	return 0;
+        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;
 
-err:
-	if (str_data) {
-		data_destroy (str_data);
-		str_data = NULL;
-	}
+        local = frame->local;
+
+        if (!op_ret)
+                local->linked = _gf_true;
+
+        FRAME_SU_UNDO (frame, dht_local_t);
+
+        if (op_ret && (op_errno == EEXIST)) {
+                conf = this->private;
+                prev = cookie;
+                subvol = prev->this;
+                if (!subvol)
+                        goto out;
+                xattrs = dict_new ();
+                if (!xattrs)
+                        goto out;
+                ret = dict_set_uint32 (xattrs, conf->link_xattr_name, 256);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to set linkto key");
+                        goto out;
+                }
 
+                STACK_WIND (frame, dht_linkfile_lookup_cbk, subvol,
+                            subvol->fops->lookup, &local->loc, xattrs);
+                if (xattrs)
+                        dict_unref (xattrs);
+                return 0;
+        }
+out:
         local->linkfile.linkfile_cbk (frame, cookie, this, op_ret, op_errno,
-                                      inode, stbuf, preparent, postparent);
-	return 0;
+                                      inode, stbuf, preparent, postparent,
+                                      xdata);
+        if (xattrs)
+                dict_unref (xattrs);
+        return 0;
 }
 
 
 int
 dht_linkfile_create (call_frame_t *frame, fop_mknod_cbk_t linkfile_cbk,
-		     xlator_t *tovol, xlator_t *fromvol, loc_t *loc)
+                     xlator_t *this,
+                     xlator_t *tovol, xlator_t *fromvol, loc_t *loc)
 {
-	dht_local_t *local = NULL;
+        dht_local_t *local = NULL;
         dict_t      *dict = NULL;
+        int          need_unref = 0;
         int          ret = 0;
+        dht_conf_t  *conf = this->private;
 
-	local = frame->local;
-	local->linkfile.linkfile_cbk = linkfile_cbk;
-	local->linkfile.srcvol = tovol;
-	loc_copy (&local->linkfile.loc, loc);
+        local = frame->local;
+        local->linkfile.linkfile_cbk = linkfile_cbk;
+        local->linkfile.srcvol = tovol;
 
-        if (!uuid_is_null (local->gfid)) {
+        local->linked = _gf_false;
+
+        dict = local->params;
+        if (!dict) {
                 dict = dict_new ();
                 if (!dict)
                         goto out;
+                need_unref = 1;
+        }
+
+        if (!uuid_is_null (local->gfid)) {
                 ret = dict_set_static_bin (dict, "gfid-req", local->gfid, 16);
                 if (ret)
-                        gf_log ("dht-linkfile", GF_LOG_DEBUG, "gfid set failed");
-        } else if (local->params) {
-                dict = dict_ref (local->params);
+                        gf_log ("dht-linkfile", GF_LOG_INFO,
+                                "%s: gfid set failed", loc->path);
         }
-        if (!dict)
-                gf_log ("", GF_LOG_DEBUG,
-                        "dict is NULL, need to make sure gfid's are same");
 
-	STACK_WIND (frame, dht_linkfile_create_cbk,
-		    fromvol, fromvol->fops->mknod, loc,
-		    S_IFREG | DHT_LINKFILE_MODE, 0, dict);
+        ret = dict_set_str (dict, GLUSTERFS_INTERNAL_FOP_KEY, "yes");
+        if (ret)
+                gf_log ("dht-linkfile", GF_LOG_INFO,
+                        "%s: internal-fop set failed", loc->path);
 
-        if (dict)
+        ret = dict_set_str (dict, conf->link_xattr_name, tovol->name);
+
+        if (ret < 0) {
+                gf_log (frame->this->name, GF_LOG_INFO,
+                        "%s: failed to initialize linkfile data",
+                        loc->path);
+                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);
+
+        if (need_unref && dict)
                 dict_unref (dict);
 
         return 0;
 out:
-	local->linkfile.linkfile_cbk (frame, NULL, frame->this, -1, ENOMEM,
-				      loc->inode, NULL, NULL, NULL);
-	return 0;
+        local->linkfile.linkfile_cbk (frame, NULL, frame->this, -1, ENOMEM,
+                                      loc->inode, NULL, NULL, NULL, NULL);
+
+        if (need_unref && dict)
+                dict_unref (dict);
+
+        return 0;
 }
 
 
 int
 dht_linkfile_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			 int32_t op_ret, int32_t op_errno,
-                         struct iatt *preparent, struct iatt *postparent)
+                         int32_t op_ret, int32_t op_errno,
+                         struct iatt *preparent, struct iatt *postparent,
+                         dict_t *xdata)
 {
-	dht_local_t   *local = NULL;
-	call_frame_t  *prev = NULL;
-	xlator_t      *subvol = NULL;
-
-	local = frame->local;
-	prev = cookie;
-	subvol = prev->this;
-
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"unlinking linkfile %s on %s failed (%s)",
-			local->loc.path, subvol->name, strerror (op_errno));
-	}
+        dht_local_t   *local = NULL;
+        call_frame_t  *prev = NULL;
+        xlator_t      *subvol = NULL;
+
+        local = frame->local;
+        prev = cookie;
+        subvol = prev->this;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "unlinking linkfile %s on %s failed (%s)",
+                        local->loc.path, subvol->name, strerror (op_errno));
+        }
 
-	DHT_STACK_DESTROY (frame);
+        DHT_STACK_DESTROY (frame);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_linkfile_unlink (call_frame_t *frame, xlator_t *this,
-		     xlator_t *subvol, loc_t *loc)
+                     xlator_t *subvol, loc_t *loc)
 {
-	call_frame_t *unlink_frame = NULL;
-	dht_local_t  *unlink_local = NULL;
-
-	unlink_frame = copy_frame (frame);
-	if (!unlink_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	unlink_local = dht_local_init (unlink_frame);
-	if (!unlink_local) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	loc_copy (&unlink_local->loc, loc);
-
-	STACK_WIND (unlink_frame, dht_linkfile_unlink_cbk,
-		    subvol, subvol->fops->unlink,
-		    &unlink_local->loc);
-
-	return 0;
+        call_frame_t *unlink_frame = NULL;
+        dht_local_t  *unlink_local = NULL;
+
+        unlink_frame = copy_frame (frame);
+        if (!unlink_frame) {
+                goto err;
+        }
+
+        /* Using non-fop value here, as anyways, 'local->fop' is not used in
+           this particular case */
+        unlink_local = dht_local_init (unlink_frame, loc, NULL,
+                                       GF_FOP_MAXVALUE);
+        if (!unlink_local) {
+                goto err;
+        }
+
+        STACK_WIND (unlink_frame, dht_linkfile_unlink_cbk,
+                    subvol, subvol->fops->unlink,
+                    &unlink_local->loc, 0, NULL);
+
+        return 0;
 err:
-	if (unlink_frame)
-		DHT_STACK_DESTROY (unlink_frame);
+        if (unlink_frame)
+                DHT_STACK_DESTROY (unlink_frame);
 
-	return -1;
+        return -1;
 }
 
 
 xlator_t *
 dht_linkfile_subvol (xlator_t *this, inode_t *inode, struct iatt *stbuf,
-		     dict_t *xattr)
+                     dict_t *xattr)
 {
-	dht_conf_t *conf = NULL;
-	xlator_t   *subvol = NULL;
-	void       *volname = NULL;
-	int         i = 0, ret = 0;
-
+        dht_conf_t *conf = NULL;
+        xlator_t   *subvol = NULL;
+        void       *volname = NULL;
+        int         i = 0, ret = 0;
 
-	conf = this->private;
+        conf = this->private;
 
-	if (!xattr)
-		goto out;
+        if (!xattr)
+                goto out;
 
-	ret = dict_get_ptr (xattr, "trusted.glusterfs.dht.linkto", &volname);
+        ret = dict_get_ptr (xattr, conf->link_xattr_name, &volname);
 
-	if ((-1 == ret) || !volname)
-		goto out;
+        if ((-1 == ret) || !volname)
+                goto out;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		if (strcmp (conf->subvolumes[i]->name, (char *)volname) == 0) {
-			subvol = conf->subvolumes[i];
-			break;
-		}
-	}
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (strcmp (conf->subvolumes[i]->name, (char *)volname) == 0) {
+                        subvol = conf->subvolumes[i];
+                        break;
+                }
+        }
 
 out:
-	return subvol;
+        return subvol;
 }
 
+int
+dht_linkfile_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                          int op_ret, int op_errno, struct iatt *statpre,
+                          struct iatt *statpost, dict_t *xdata)
+{
+        dht_local_t *local = NULL;
+        loc_t *loc = NULL;
 
+        local = frame->local;
+        loc = &local->loc;
+
+        if (op_ret)
+                gf_log (this->name, GF_LOG_ERROR, "setattr of uid/gid on %s"
+                        " :<gfid:%s> failed (%s)",
+                        (loc->path? loc->path: "NULL"),
+                        uuid_utoa(local->gfid), strerror(op_errno));
+
+        DHT_STACK_DESTROY (frame);
+
+       return 0;
+}
+
+int
+dht_linkfile_attr_heal (call_frame_t *frame, xlator_t *this)
+{
+        int     ret = -1;
+        call_frame_t *copy = NULL;
+        dht_local_t  *local = NULL;
+        dht_local_t  *copy_local = NULL;
+        xlator_t     *subvol = NULL;
+        struct iatt   stbuf = {0,};
+
+        local = frame->local;
+
+        GF_VALIDATE_OR_GOTO ("dht", local, out);
+        GF_VALIDATE_OR_GOTO ("dht", local->link_subvol, out);
+
+        if (local->stbuf.ia_type == IA_INVAL)
+                return 0;
+
+        uuid_copy (local->loc.gfid, local->stbuf.ia_gfid);
+
+        copy = copy_frame (frame);
+
+        if (!copy)
+                goto out;
+
+       copy_local = dht_local_init (copy, &local->loc, NULL, 0);
+
+       if (!copy_local)
+               goto out;
+
+        stbuf = local->stbuf;
+        subvol = local->link_subvol;
+
+        copy->local = copy_local;
+
+        FRAME_SU_DO (copy, dht_local_t);
+
+        STACK_WIND (copy, dht_linkfile_setattr_cbk, subvol,
+                    subvol->fops->setattr, &copy_local->loc,
+                    &stbuf, (GF_SET_ATTR_UID | GF_SET_ATTR_GID), NULL);
+        ret = 0;
+out:
+        return ret;
+}
diff --git a/xlators/cluster/dht/src/dht-mem-types.h b/xlators/cluster/dht/src/dht-mem-types.h
index 570f4c24b..e893eb48f 100644
--- a/xlators/cluster/dht/src/dht-mem-types.h
+++ b/xlators/cluster/dht/src/dht-mem-types.h
@@ -1,21 +1,11 @@
-
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
 
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -29,14 +19,17 @@ enum gf_dht_mem_types_ {
         gf_dht_mt_dht_conf_t,
         gf_dht_mt_char,
         gf_dht_mt_int32_t,
-        gf_dht_mt_dht_local_t,
         gf_dht_mt_xlator_t,
         gf_dht_mt_dht_layout_t,
         gf_switch_mt_dht_conf_t,
         gf_switch_mt_dht_du_t,
         gf_switch_mt_switch_sched_array,
         gf_switch_mt_switch_struct,
+        gf_dht_mt_subvol_time,
+        gf_dht_mt_loc_t,
+        gf_defrag_info_mt,
+        gf_dht_mt_inode_ctx_t,
+        gf_dht_mt_ctx_stat_time_t,
         gf_dht_mt_end
 };
 #endif
-
diff --git a/xlators/cluster/dht/src/dht-rebalance.c b/xlators/cluster/dht/src/dht-rebalance.c
new file mode 100644
index 000000000..4f78f5203
--- /dev/null
+++ b/xlators/cluster/dht/src/dht-rebalance.c
@@ -0,0 +1,1908 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "dht-common.h"
+#include "xlator.h"
+#include <signal.h>
+#include <fnmatch.h>
+
+#define GF_DISK_SECTOR_SIZE             512
+#define DHT_REBALANCE_PID               4242 /* Change it if required */
+#define DHT_REBALANCE_BLKSIZE           (128 * 1024)
+
+static int
+dht_write_with_holes (xlator_t *to, fd_t *fd, struct iovec *vec, int count,
+                      int32_t size, off_t offset, struct iobref *iobref)
+{
+        int i            = 0;
+        int ret          = -1;
+        int start_idx    = 0;
+        int tmp_offset   = 0;
+        int write_needed = 0;
+        int buf_len      = 0;
+        int size_pending = 0;
+        char *buf        = NULL;
+
+        /* loop through each vector */
+        for (i = 0; i < count; i++) {
+                buf = vec[i].iov_base;
+                buf_len = vec[i].iov_len;
+
+                for (start_idx = 0; (start_idx + GF_DISK_SECTOR_SIZE) <= buf_len;
+                     start_idx += GF_DISK_SECTOR_SIZE) {
+
+                        if (mem_0filled (buf + start_idx, GF_DISK_SECTOR_SIZE) != 0) {
+                                write_needed = 1;
+                                continue;
+                        }
+
+                        if (write_needed) {
+                                ret = syncop_write (to, fd, (buf + tmp_offset),
+                                                    (start_idx - tmp_offset),
+                                                    (offset + tmp_offset),
+                                                    iobref, 0);
+                                /* 'path' will be logged in calling function */
+                                if (ret < 0) {
+                                        gf_log (THIS->name, GF_LOG_WARNING,
+                                                "failed to write (%s)",
+                                                strerror (-ret));
+                                        ret = -1;
+                                        goto out;
+                                }
+
+                                write_needed = 0;
+                        }
+                        tmp_offset = start_idx + GF_DISK_SECTOR_SIZE;
+                }
+
+                if ((start_idx < buf_len) || write_needed) {
+                        /* This means, last chunk is not yet written.. write it */
+                        ret = syncop_write (to, fd, (buf + tmp_offset),
+                                            (buf_len - tmp_offset),
+                                            (offset + tmp_offset), iobref, 0);
+                        if (ret < 0) {
+                                /* 'path' will be logged in calling function */
+                                gf_log (THIS->name, GF_LOG_WARNING,
+                                        "failed to write (%s)",
+                                        strerror (-ret));
+                                ret = -1;
+                                goto out;
+                        }
+                }
+
+                size_pending = (size - buf_len);
+                if (!size_pending)
+                        break;
+        }
+
+        ret = size;
+out:
+        return ret;
+
+}
+
+/*
+   return values:
+   -1 : failure
+   -2 : success
+
+Hard link migration is carried out in three stages.
+
+(Say there are n hardlinks)
+Stage 1: Setting the new hashed subvol information on the 1st hardlink
+         encountered (linkto setxattr)
+
+Stage 2: Creating hardlinks on new hashed subvol for the 2nd to (n-1)th
+         hardlink
+
+Stage 3: Physical migration of the data file for nth hardlink
+
+Why to deem "-2" as success and not "0":
+
+   dht_migrate_file expects return value "0" from _is_file_migratable if
+the file has to be migrated.
+
+   _is_file_migratable returns zero only when it is called with the
+flag "GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS".
+
+   gf_defrag_handle_hardlink calls dht_migrate_file for physical migration
+of the data file with the flag "GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS"
+
+Hence, gf_defrag_handle_hardlink returning "0" for success will force
+"dht_migrate_file" to migrate each of the hardlink which is not intended.
+
+For each of the three stage mentioned above "-2" will be returned and will
+be converted to "0" in dht_migrate_file.
+
+*/
+
+int32_t
+gf_defrag_handle_hardlink (xlator_t *this, loc_t *loc, dict_t  *xattrs,
+                           struct iatt *stbuf)
+{
+        int32_t                 ret             = -1;
+        xlator_t               *cached_subvol   = NULL;
+        xlator_t               *hashed_subvol   = NULL;
+        xlator_t               *linkto_subvol   = NULL;
+        data_t                 *data            = NULL;
+        struct iatt             iatt            = {0,};
+        int32_t                 op_errno        = 0;
+        dht_conf_t             *conf            = NULL;
+
+        GF_VALIDATE_OR_GOTO ("defrag", loc, out);
+        GF_VALIDATE_OR_GOTO ("defrag", loc->name, out);
+        GF_VALIDATE_OR_GOTO ("defrag", stbuf, out);
+        GF_VALIDATE_OR_GOTO ("defrag", this, out);
+        GF_VALIDATE_OR_GOTO ("defrag", xattrs, out);
+        GF_VALIDATE_OR_GOTO ("defrag", this->private, out);
+
+        conf = this->private;
+
+        if (uuid_is_null (loc->pargfid)) {
+                gf_log ("", GF_LOG_ERROR, "loc->pargfid is NULL for "
+                        "%s", loc->path);
+                goto out;
+        }
+
+        if (uuid_is_null (loc->gfid)) {
+                gf_log ("", GF_LOG_ERROR, "loc->gfid is NULL for "
+                        "%s", loc->path);
+                goto out;
+        }
+
+        cached_subvol = dht_subvol_get_cached (this, loc->inode);
+        if (!cached_subvol) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get cached subvol"
+                        " for %s on %s", loc->name, this->name);
+                goto out;
+        }
+
+        hashed_subvol = dht_subvol_get_hashed (this, loc);
+        if (!hashed_subvol) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to get hashed subvol"
+                        " for %s on %s", loc->name, this->name);
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_INFO, "Attempting to migrate hardlink %s "
+                "with gfid %s from %s -> %s", loc->name, uuid_utoa (loc->gfid),
+                cached_subvol->name, hashed_subvol->name);
+        data = dict_get (xattrs, conf->link_xattr_name);
+        /* set linkto on cached -> hashed if not present, else link it */
+        if (!data) {
+                ret = dict_set_str (xattrs, conf->link_xattr_name,
+                                    hashed_subvol->name);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to set "
+                                "linkto xattr in dict for %s", loc->name);
+                        goto out;
+                }
+
+                ret = syncop_setxattr (cached_subvol, loc, xattrs, 0);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Linkto setxattr "
+                                "failed %s -> %s (%s)", cached_subvol->name,
+                                loc->name, strerror (-ret));
+                        ret = -1;
+                        goto out;
+                }
+                ret = -2;
+                goto out;
+        } else {
+                linkto_subvol = dht_linkfile_subvol (this, NULL, NULL, xattrs);
+                if (!linkto_subvol) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+                                "linkto subvol for %s", loc->name);
+                } else {
+                        hashed_subvol = linkto_subvol;
+                }
+
+                ret = syncop_link (hashed_subvol, loc, loc);
+                if  (ret) {
+                        op_errno = -ret;
+                        ret = -1;
+                        gf_log (this->name, GF_LOG_ERROR, "link of %s -> %s"
+                                " failed on  subvol %s (%s)", loc->name,
+                                uuid_utoa(loc->gfid),
+                                hashed_subvol->name, strerror (op_errno));
+                        if (op_errno != EEXIST)
+                                goto out;
+                }
+        }
+        ret = syncop_lookup (hashed_subvol, loc, NULL, &iatt, NULL, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed lookup %s on %s (%s)"
+                        , loc->name, hashed_subvol->name, strerror (-ret));
+                ret = -1;
+                goto out;
+        }
+
+        if (iatt.ia_nlink == stbuf->ia_nlink) {
+                ret = dht_migrate_file (this, loc, cached_subvol, hashed_subvol,
+                                        GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS);
+                if (ret)
+                        goto out;
+        }
+        ret = -2;
+out:
+        return ret;
+}
+
+/*
+     return values
+     0 : File will be migrated
+    -2 : File will not be migrated
+         (This is the return value from gf_defrag_handle_hardlink. Checkout
+         gf_defrag_handle_hardlink for description of "returning -2")
+    -1 : failure
+*/
+static inline int
+__is_file_migratable (xlator_t *this, loc_t *loc,
+                      struct iatt *stbuf, dict_t *xattrs, int flags)
+{
+        int ret = -1;
+
+        if (IA_ISDIR (stbuf->ia_type)) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: migrate-file called on directory", loc->path);
+                ret = -1;
+                goto out;
+        }
+
+        if (flags == GF_DHT_MIGRATE_HARDLINK_IN_PROGRESS) {
+                ret = 0;
+                goto out;
+        }
+        if (stbuf->ia_nlink > 1) {
+                /* support for decomission */
+                if (flags == GF_DHT_MIGRATE_HARDLINK) {
+                        ret = gf_defrag_handle_hardlink (this, loc,
+                                                         xattrs, stbuf);
+
+                   /*
+                   Returning zero will force the file to be remigrated.
+                   Checkout gf_defrag_handle_hardlink for more information.
+                   */
+                        if (ret && ret != -2) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "%s: failed to migrate file with link",
+                                        loc->path);
+                        }
+                } else {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: file has hardlinks", loc->path);
+                        ret = -ENOTSUP;
+                }
+                goto out;
+        }
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+static inline int
+__dht_rebalance_create_dst_file (xlator_t *to, xlator_t *from, loc_t *loc, struct iatt *stbuf,
+                                 dict_t *dict, fd_t **dst_fd, dict_t *xattr)
+{
+        xlator_t    *this = NULL;
+        int          ret  = -1;
+        fd_t        *fd   = NULL;
+        struct iatt  new_stbuf = {0,};
+        dht_conf_t  *conf = NULL;
+
+        this = THIS;
+        conf = this->private;
+
+        ret = dict_set_static_bin (dict, "gfid-req", stbuf->ia_gfid, 16);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to set gfid in dict for create", loc->path);
+                goto out;
+        }
+
+        ret = dict_set_str (dict, conf->link_xattr_name, from->name);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to set gfid in dict for create", loc->path);
+                goto out;
+        }
+
+        fd = fd_create (loc->inode, DHT_REBALANCE_PID);
+        if (!fd) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: fd create failed (destination) (%s)",
+                        loc->path, strerror (errno));
+                ret = -1;
+                goto out;
+        }
+
+        ret = syncop_lookup (to, loc, NULL, &new_stbuf, NULL, NULL);
+        if (!ret) {
+                /* File exits in the destination, check if gfid matches */
+                if (uuid_compare (stbuf->ia_gfid, new_stbuf.ia_gfid) != 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "file %s exits in %s with different gfid",
+                                loc->path, to->name);
+                        fd_unref (fd);
+                        goto out;
+                }
+        }
+        if ((ret < 0) && (-ret != ENOENT)) {
+                /* File exists in destination, but not accessible */
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "%s: failed to lookup file (%s)",
+                        loc->path, strerror (-ret));
+                ret = -1;
+                goto out;
+        }
+
+        /* Create the destination with LINKFILE mode, and linkto xattr,
+           if the linkfile already exists, it will just open the file */
+        ret = syncop_create (to, loc, O_RDWR, DHT_LINKFILE_MODE, fd,
+                             dict, &new_stbuf);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create %s on %s (%s)",
+                        loc->path, to->name, strerror (-ret));
+                ret = -1;
+                goto out;
+        }
+
+        ret = syncop_fsetxattr (to, fd, xattr, 0);
+        if (ret < 0)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to set xattr on %s (%s)",
+                        loc->path, to->name, strerror (-ret));
+
+        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 (-ret));
+
+        ret = syncop_fsetattr (to, fd, stbuf,
+                               (GF_SET_ATTR_UID | GF_SET_ATTR_GID),
+                                NULL, NULL);
+        if (ret < 0)
+                gf_log (this->name, GF_LOG_ERROR,
+                        "chown failed for %s on %s (%s)",
+                        loc->path, to->name, strerror (-ret));
+
+        if (dst_fd)
+                *dst_fd = fd;
+
+        /* success */
+        ret = 0;
+
+out:
+        return ret;
+}
+
+static inline int
+__dht_check_free_space (xlator_t *to, xlator_t *from, loc_t *loc,
+                        struct iatt *stbuf, int flag)
+{
+        struct statvfs  src_statfs = {0,};
+        struct statvfs  dst_statfs = {0,};
+        int             ret        = -1;
+        xlator_t       *this       = NULL;
+
+        uint64_t        src_statfs_blocks = 1;
+        uint64_t        dst_statfs_blocks = 1;
+
+        this = THIS;
+
+        ret = syncop_statfs (from, loc, &src_statfs);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to get statfs of %s on %s (%s)",
+                        loc->path, from->name, strerror (-ret));
+                ret = -1;
+                goto out;
+        }
+
+        ret = syncop_statfs (to, loc, &dst_statfs);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to get statfs of %s on %s (%s)",
+                        loc->path, to->name, strerror (-ret));
+                ret = -1;
+                goto out;
+        }
+
+        /* if force option is given, do not check for space @ dst.
+         * Check only if space is avail for the file */
+        if (flag != GF_DHT_MIGRATE_DATA)
+                goto check_avail_space;
+
+        /* Check:
+           During rebalance `migrate-data` - Destination subvol experiences
+           a `reduction` in 'blocks' of free space, at the same time source
+           subvol gains certain 'blocks' of free space. A valid check is
+           necessary here to avoid errorneous move to destination where
+           the space could be scantily available.
+         */
+        if (stbuf) {
+                dst_statfs_blocks = ((dst_statfs.f_bavail *
+                                      dst_statfs.f_bsize) /
+                                     GF_DISK_SECTOR_SIZE);
+                src_statfs_blocks = ((src_statfs.f_bavail *
+                                      src_statfs.f_bsize) /
+                                     GF_DISK_SECTOR_SIZE);
+                if ((dst_statfs_blocks - stbuf->ia_blocks) <
+                    (src_statfs_blocks + stbuf->ia_blocks)) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "data movement attempted from node (%s) with"
+                                " higher disk space to a node (%s) with "
+                                "lesser disk space (%s)", from->name,
+                                to->name, loc->path);
+
+                        /* this is not a 'failure', but we don't want to
+                           consider this as 'success' too :-/ */
+                        ret = 1;
+                        goto out;
+                }
+        }
+check_avail_space:
+        if (((dst_statfs.f_bavail * dst_statfs.f_bsize) /
+              GF_DISK_SECTOR_SIZE) < stbuf->ia_blocks) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "data movement attempted from node (%s) with "
+                        "to node (%s) which does not have required free space"
+                        " for %s", from->name, to->name, loc->path);
+                ret = 1;
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+static inline int
+__dht_rebalance_migrate_data (xlator_t *from, xlator_t *to, fd_t *src, fd_t *dst,
+                             uint64_t ia_size, int hole_exists)
+{
+        int            ret    = 0;
+        int            count  = 0;
+        off_t          offset = 0;
+        struct iovec  *vector = NULL;
+        struct iobref *iobref = NULL;
+        uint64_t       total  = 0;
+        size_t         read_size = 0;
+
+        /* if file size is '0', no need to enter this loop */
+        while (total < ia_size) {
+                read_size = (((ia_size - total) > DHT_REBALANCE_BLKSIZE) ?
+                             DHT_REBALANCE_BLKSIZE : (ia_size - total));
+                ret = syncop_readv (from, src, read_size,
+                                    offset, 0, &vector, &count, &iobref);
+                if (!ret || (ret < 0)) {
+                        break;
+                }
+
+                if (hole_exists)
+                        ret = dht_write_with_holes (to, dst, vector, count,
+                                                    ret, offset, iobref);
+                else
+                        ret = syncop_writev (to, dst, vector, count,
+                                             offset, iobref, 0);
+                if (ret < 0) {
+                        break;
+                }
+                offset += ret;
+                total += ret;
+
+                GF_FREE (vector);
+                if (iobref)
+                        iobref_unref (iobref);
+                iobref = NULL;
+                vector = NULL;
+        }
+        if (iobref)
+                iobref_unref (iobref);
+        GF_FREE (vector);
+
+        if (ret >= 0)
+                ret = 0;
+        else
+                ret = -1;
+
+        return ret;
+}
+
+
+static inline int
+__dht_rebalance_open_src_file (xlator_t *from, xlator_t *to, loc_t *loc,
+                               struct iatt *stbuf, fd_t **src_fd)
+{
+        int          ret  = 0;
+        fd_t        *fd   = NULL;
+        dict_t      *dict = NULL;
+        xlator_t    *this = NULL;
+        struct iatt  iatt = {0,};
+        dht_conf_t  *conf = NULL;
+
+        this = THIS;
+        conf = this->private;
+
+        fd = fd_create (loc->inode, DHT_REBALANCE_PID);
+        if (!fd) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: fd create failed (source)", loc->path);
+                ret = -1;
+                goto out;
+        }
+
+        ret = syncop_open (from, loc, O_RDWR, fd);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to open file %s on %s (%s)",
+                        loc->path, from->name, strerror (-ret));
+                ret = -1;
+                goto out;
+        }
+
+        ret = -1;
+        dict = dict_new ();
+        if (!dict)
+                goto out;
+
+        ret = dict_set_str (dict, conf->link_xattr_name, to->name);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to set xattr in dict for %s (linkto:%s)",
+                        loc->path, to->name);
+                goto out;
+        }
+
+        /* Once the migration starts, the source should have 'linkto' key set
+           to show which is the target, so other clients can work around it */
+        ret = syncop_setxattr (from, loc, dict, 0);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to set xattr on %s in %s (%s)",
+                        loc->path, from->name, strerror (-ret));
+                ret = -1;
+                goto out;
+        }
+
+        /* mode should be (+S+T) to indicate migration is in progress */
+        iatt.ia_prot = stbuf->ia_prot;
+        iatt.ia_type = stbuf->ia_type;
+        iatt.ia_prot.sticky = 1;
+        iatt.ia_prot.sgid = 1;
+
+        ret = syncop_setattr (from, loc, &iatt, GF_SET_ATTR_MODE, NULL, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to set mode on %s in %s (%s)",
+                        loc->path, from->name, strerror (-ret));
+                ret = -1;
+                goto out;
+        }
+
+        if (src_fd)
+                *src_fd = fd;
+
+        /* success */
+        ret = 0;
+out:
+        if (dict)
+                dict_unref (dict);
+
+        return ret;
+}
+
+int
+migrate_special_files (xlator_t *this, xlator_t *from, xlator_t *to, loc_t *loc,
+                       struct iatt *buf)
+{
+        int          ret      = -1;
+        dict_t      *rsp_dict = NULL;
+        dict_t      *dict     = NULL;
+        char        *link     = NULL;
+        struct iatt  stbuf    = {0,};
+        dht_conf_t  *conf     = this->private;
+
+        dict = dict_new ();
+        if (!dict)
+                goto out;
+
+        ret = dict_set_int32 (dict, conf->link_xattr_name, 256);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to set 'linkto' key in dict", loc->path);
+                goto out;
+        }
+
+        /* check in the destination if the file is link file */
+        ret = syncop_lookup (to, loc, dict, &stbuf, &rsp_dict, NULL);
+        if ((ret < 0) && (-ret != ENOENT)) {
+                gf_log (this->name, GF_LOG_WARNING, "%s: lookup failed (%s)",
+                        loc->path, strerror (-ret));
+                ret = -1;
+                goto out;
+        }
+
+        /* we no more require this key */
+        dict_del (dict, conf->link_xattr_name);
+
+        /* file exists in target node, only if it is 'linkfile' its valid,
+           otherwise, error out */
+        if (!ret) {
+                if (!check_is_linkfile (loc->inode, &stbuf, rsp_dict,
+                                        conf->link_xattr_name)) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: file exists in destination", loc->path);
+                        ret = -1;
+                        goto out;
+                }
+
+                /* as file is linkfile, delete it */
+                ret = syncop_unlink (to, loc);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to delete the linkfile (%s)",
+                                loc->path, strerror (-ret));
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        /* Set the gfid of the source file in dict */
+        ret = dict_set_static_bin (dict, "gfid-req", buf->ia_gfid, 16);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to set gfid in dict for create", loc->path);
+                goto out;
+        }
+
+        /* Create the file in target */
+        if (IA_ISLNK (buf->ia_type)) {
+                /* Handle symlinks separately */
+                ret = syncop_readlink (from, loc, &link, buf->ia_size);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: readlink on symlink failed (%s)",
+                                loc->path, strerror (-ret));
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = syncop_symlink (to, loc, link, dict, 0);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: creating symlink failed (%s)",
+                                loc->path, strerror (-ret));
+                        ret = -1;
+                        goto out;
+                }
+
+                goto done;
+        }
+
+        ret = syncop_mknod (to, loc, st_mode_from_ia (buf->ia_prot,
+                                                      buf->ia_type),
+                            makedev (ia_major (buf->ia_rdev),
+                                     ia_minor (buf->ia_rdev)), dict, 0);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "%s: mknod failed (%s)",
+                        loc->path, strerror (-ret));
+                ret = -1;
+                goto out;
+        }
+
+done:
+        ret = syncop_setattr (to, loc, buf,
+                              (GF_SET_ATTR_UID | GF_SET_ATTR_GID |
+                               GF_SET_ATTR_MODE), NULL, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to perform setattr on %s (%s)",
+                        loc->path, to->name, strerror (-ret));
+                ret = -1;
+        }
+
+        ret = syncop_unlink (from, loc);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "%s: unlink failed (%s)",
+                        loc->path, strerror (-ret));
+                ret = -1;
+        }
+
+out:
+        if (dict)
+                dict_unref (dict);
+
+        if (rsp_dict)
+                dict_unref (rsp_dict);
+
+        return ret;
+}
+
+/*
+  return values:
+
+   -1 : failure
+    0 : successfully migrated data
+    1 : not a failure, but we can't migrate data as of now
+*/
+int
+dht_migrate_file (xlator_t *this, loc_t *loc, xlator_t *from, xlator_t *to,
+                  int flag)
+{
+        int             ret            = -1;
+        struct iatt     new_stbuf      = {0,};
+        struct iatt     stbuf          = {0,};
+        struct iatt     empty_iatt     = {0,};
+        ia_prot_t       src_ia_prot    = {0,};
+        fd_t           *src_fd         = NULL;
+        fd_t           *dst_fd         = NULL;
+        dict_t         *dict           = NULL;
+        dict_t         *xattr          = NULL;
+        dict_t         *xattr_rsp      = NULL;
+        int             file_has_holes = 0;
+        dht_conf_t     *conf           = this->private;
+
+        gf_log (this->name, GF_LOG_INFO, "%s: attempting to move from %s to %s",
+                loc->path, from->name, to->name);
+
+        dict = dict_new ();
+        if (!dict)
+                goto out;
+
+        ret = dict_set_int32 (dict, conf->link_xattr_name, 256);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to set 'linkto' key in dict", loc->path);
+                goto out;
+        }
+
+        /* Phase 1 - Data migration is in progress from now on */
+        ret = syncop_lookup (from, loc, dict, &stbuf, &xattr_rsp, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "%s: lookup failed on %s (%s)",
+                        loc->path, from->name, strerror (-ret));
+                ret = -1;
+                goto out;
+        }
+
+        /* we no more require this key */
+        dict_del (dict, conf->link_xattr_name);
+
+        /* preserve source mode, so set the same to the destination */
+        src_ia_prot = stbuf.ia_prot;
+
+        /* Check if file can be migrated */
+        ret = __is_file_migratable (this, loc, &stbuf, xattr_rsp, flag);
+        if (ret) {
+                if (ret == -2)
+                        ret = 0;
+                goto out;
+        }
+        /* Take care of the special files */
+        if (!IA_ISREG (stbuf.ia_type)) {
+                /* Special files */
+                ret = migrate_special_files (this, from, to, loc, &stbuf);
+                goto out;
+        }
+
+        /* TODO: move all xattr related operations to fd based operations */
+        ret = syncop_listxattr (from, loc, &xattr);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to get xattr from %s (%s)",
+                        loc->path, from->name, strerror (-ret));
+                ret = -1;
+        }
+
+        /* create the destination, with required modes/xattr */
+        ret = __dht_rebalance_create_dst_file (to, from, loc, &stbuf,
+                                               dict, &dst_fd, xattr);
+        if (ret)
+                goto out;
+
+        ret = __dht_check_free_space (to, from, loc, &stbuf, flag);
+        if (ret) {
+                goto out;
+        }
+
+        /* Open the source, and also update mode/xattr */
+        ret = __dht_rebalance_open_src_file (from, to, loc, &stbuf, &src_fd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to open %s on %s",
+                        loc->path, from->name);
+                goto out;
+        }
+
+
+        ret = syncop_fstat (from, src_fd, &stbuf);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to lookup %s on %s (%s)",
+                        loc->path, from->name, strerror (-ret));
+                ret = -1;
+                goto out;
+        }
+
+        /* Try to preserve 'holes' while migrating data */
+        if (stbuf.ia_size > (stbuf.ia_blocks * GF_DISK_SECTOR_SIZE))
+                file_has_holes = 1;
+
+        /* All I/O happens in this function */
+        ret = __dht_rebalance_migrate_data (from, to, src_fd, dst_fd,
+					    stbuf.ia_size, file_has_holes);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "%s: failed to migrate data",
+                        loc->path);
+                /* reset the destination back to 0 */
+                ret = syncop_ftruncate (to, dst_fd, 0);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "%s: failed to reset target size back to 0 (%s)",
+                                loc->path, strerror (-ret));
+                }
+
+                ret = -1;
+                goto out;
+        }
+
+        /* TODO: Sync the locks */
+
+        ret = syncop_fsync (to, dst_fd, 0);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to fsync on %s (%s)",
+                        loc->path, to->name, strerror (-ret));
+                ret = -1;
+        }
+
+
+        /* Phase 2 - Data-Migration Complete, Housekeeping updates pending */
+
+        ret = syncop_fstat (from, src_fd, &new_stbuf);
+        if (ret < 0) {
+                /* Failed to get the stat info */
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to fstat file %s on %s (%s)",
+                        loc->path, from->name, strerror (-ret));
+                ret = -1;
+                goto out;
+        }
+
+        /* source would have both sticky bit and sgid bit set, reset it to 0,
+           and set the source permission on destination, if it was not set
+           prior to setting rebalance-modes in source  */
+        if (!src_ia_prot.sticky)
+                new_stbuf.ia_prot.sticky = 0;
+
+        if (!src_ia_prot.sgid)
+                new_stbuf.ia_prot.sgid = 0;
+
+        /* TODO: if the source actually had sticky bit, or sgid bit set,
+           we are not handling it */
+
+        ret = syncop_fsetattr (to, dst_fd, &new_stbuf,
+                               (GF_SET_ATTR_UID | GF_SET_ATTR_GID |
+                                GF_SET_ATTR_MODE), NULL, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to perform setattr on %s (%s)",
+                        loc->path, to->name, strerror (-ret));
+                ret = -1;
+                goto out;
+        }
+
+        /* Because 'futimes' is not portable */
+        ret = syncop_setattr (to, loc, &new_stbuf,
+                              (GF_SET_ATTR_MTIME | GF_SET_ATTR_ATIME),
+                              NULL, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to perform setattr on %s (%s)",
+                        loc->path, to->name, strerror (-ret));
+                ret = -1;
+        }
+
+        /* Make the source as a linkfile first before deleting it */
+        empty_iatt.ia_prot.sticky = 1;
+        ret = syncop_fsetattr (from, src_fd, &empty_iatt,
+                               GF_SET_ATTR_MODE, NULL, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,             \
+                        "%s: failed to perform setattr on %s (%s)",
+                        loc->path, from->name, strerror (-ret));
+                ret = -1;
+                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 (-ret));
+                ret = -1;
+        }
+
+        /* remove the 'linkto' xattr from the destination */
+        ret = syncop_fremovexattr (to, dst_fd, conf->link_xattr_name, 0);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to perform removexattr on %s (%s)",
+                        loc->path, to->name, strerror (-ret));
+                ret = -1;
+        }
+
+        /* Do a stat and check the gfid before unlink */
+        ret = syncop_stat (from, loc, &empty_iatt);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to do a stat on %s (%s)",
+                        loc->path, from->name, strerror (-ret));
+                ret = -1;
+                goto out;
+        }
+
+        if (uuid_compare (empty_iatt.ia_gfid, loc->gfid) == 0) {
+                /* take out the source from namespace */
+                ret = syncop_unlink (from, loc);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to perform unlink on %s (%s)",
+                                loc->path, from->name, strerror (-ret));
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        ret = syncop_lookup (this, loc, NULL, NULL, NULL, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "%s: failed to lookup the file on subvolumes (%s)",
+                        loc->path, strerror (-ret));
+                ret = -1;
+        }
+
+        gf_log (this->name, GF_LOG_INFO,
+                "completed migration of %s from subvolume %s to %s",
+                loc->path, from->name, to->name);
+
+        ret = 0;
+out:
+        if (dict)
+                dict_unref (dict);
+
+        if (xattr)
+                dict_unref (xattr);
+        if (xattr_rsp)
+                dict_unref (xattr_rsp);
+
+        if (dst_fd)
+                syncop_close (dst_fd);
+        if (src_fd)
+                syncop_close (src_fd);
+
+        return ret;
+}
+
+static int
+rebalance_task (void *data)
+{
+        int           ret   = -1;
+        dht_local_t  *local = NULL;
+        call_frame_t *frame = NULL;
+
+        frame = data;
+
+        local = frame->local;
+
+        /* This function is 'synchrounous', hence if it returns,
+           we are done with the task */
+        ret = dht_migrate_file (THIS, &local->loc, local->rebalance.from_subvol,
+                                local->rebalance.target_node, local->flags);
+
+        return ret;
+}
+
+static int
+rebalance_task_completion (int op_ret, call_frame_t *sync_frame, void *data)
+{
+        int           ret        = -1;
+        uint64_t      layout_int = 0;
+        dht_layout_t *layout     = 0;
+        xlator_t     *this       = NULL;
+        dht_local_t  *local      = NULL;
+        int32_t       op_errno   = EINVAL;
+
+        this = THIS;
+        local = sync_frame->local;
+
+        if (!op_ret) {
+                /* Make sure we have valid 'layout' in inode ctx
+                   after the operation */
+                ret = inode_ctx_del (local->loc.inode, this, &layout_int);
+                if (!ret && layout_int) {
+                        layout = (dht_layout_t *)(long)layout_int;
+                        dht_layout_unref (this, layout);
+                }
+
+                ret = dht_layout_preset (this, local->rebalance.target_node,
+                                         local->loc.inode);
+                if (ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to set inode ctx", local->loc.path);
+        }
+
+        if (op_ret == -1) {
+                /* Failure of migration process, mostly due to write process.
+                   as we can't preserve the exact errno, lets say there was
+                   no space to migrate-data
+                */
+                op_errno = ENOSPC;
+        }
+
+        if (op_ret == 1) {
+                /* migration didn't happen, but is not a failure, let the user
+                   understand that he doesn't have permission to migrate the
+                   file.
+                */
+                op_ret = -1;
+                op_errno = EPERM;
+        }
+
+        DHT_STACK_UNWIND (setxattr, sync_frame, op_ret, op_errno, NULL);
+        return 0;
+}
+
+int
+dht_start_rebalance_task (xlator_t *this, call_frame_t *frame)
+{
+        int         ret     = -1;
+
+        ret = synctask_new (this->ctx->env, rebalance_task,
+                            rebalance_task_completion,
+                            frame, frame);
+        return ret;
+}
+
+int
+gf_listener_stop (xlator_t *this)
+{
+        glusterfs_ctx_t  *ctx = NULL;
+        cmd_args_t       *cmd_args = NULL;
+        int              ret = 0;
+
+        ctx = this->ctx;
+        GF_ASSERT (ctx);
+        cmd_args = &ctx->cmd_args;
+        if (cmd_args->sock_file) {
+                ret = unlink (cmd_args->sock_file);
+                if (ret && (ENOENT == errno)) {
+                        ret = 0;
+                }
+        }
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to unlink listener "
+                        "socket %s, error: %s", cmd_args->sock_file,
+                        strerror (errno));
+        }
+        return ret;
+}
+
+void
+dht_build_root_inode (xlator_t *this, inode_t **inode)
+{
+        inode_table_t    *itable        = NULL;
+        uuid_t            root_gfid     = {0, };
+
+        itable = inode_table_new (0, this);
+        if (!itable)
+                return;
+
+        root_gfid[15] = 1;
+        *inode = inode_find (itable, root_gfid);
+}
+
+void
+dht_build_root_loc (inode_t *inode, loc_t *loc)
+{
+        loc->path = "/";
+        loc->inode = inode;
+        loc->inode->ia_type = IA_IFDIR;
+        memset (loc->gfid, 0, 16);
+        loc->gfid[15] = 1;
+}
+
+
+/* return values: 1 -> error, bug ignore and continue
+                  0 -> proceed
+                 -1 -> error, handle it */
+int32_t
+gf_defrag_handle_migrate_error (int32_t op_errno, gf_defrag_info_t *defrag)
+{
+        /* if errno is not ENOSPC or ENOTCONN, we can still continue
+           with rebalance process */
+        if ((op_errno != ENOSPC) || (op_errno != ENOTCONN))
+                return 1;
+
+        if (op_errno == ENOTCONN) {
+                /* Most probably mount point went missing (mostly due
+                   to a brick down), say rebalance failure to user,
+                   let him restart it if everything is fine */
+                defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
+                return -1;
+        }
+
+        if (op_errno == ENOSPC) {
+                /* rebalance process itself failed, may be
+                   remote brick went down, or write failed due to
+                   disk full etc etc.. */
+                defrag->defrag_status = GF_DEFRAG_STATUS_FAILED;
+                return -1;
+        }
+
+        return 0;
+}
+
+static gf_boolean_t
+gf_defrag_pattern_match (gf_defrag_info_t *defrag, char *name, uint64_t size)
+{
+        gf_defrag_pattern_list_t *trav = NULL;
+        gf_boolean_t               match = _gf_false;
+        gf_boolean_t               ret = _gf_false;
+
+        GF_VALIDATE_OR_GOTO ("dht", defrag, out);
+
+        trav = defrag->defrag_pattern;
+        while (trav) {
+                if (!fnmatch (trav->path_pattern, name, FNM_NOESCAPE)) {
+                        match = _gf_true;
+                        break;
+                }
+                trav = trav->next;
+        }
+
+        if ((match == _gf_true) && (size >= trav->size))
+                ret = _gf_true;
+
+ out:
+        return ret;
+}
+
+/* We do a depth first traversal of directories. But before we move into
+ * subdirs, we complete the data migration of those directories whose layouts
+ * have been fixed
+ */
+
+int
+gf_defrag_migrate_data (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
+                        dict_t *migrate_data)
+{
+        int                      ret            = -1;
+        loc_t                    entry_loc      = {0,};
+        fd_t                    *fd             = NULL;
+        gf_dirent_t              entries;
+        gf_dirent_t             *tmp            = NULL;
+        gf_dirent_t             *entry          = NULL;
+        gf_boolean_t             free_entries   = _gf_false;
+        off_t                    offset         = 0;
+        dict_t                  *dict           = NULL;
+        struct iatt              iatt           = {0,};
+        int32_t                  op_errno       = 0;
+        char                    *uuid_str       = NULL;
+        uuid_t                   node_uuid      = {0,};
+        struct timeval           dir_start      = {0,};
+        struct timeval           end            = {0,};
+        double                   elapsed        = {0,};
+        struct timeval           start          = {0,};
+        int32_t                  err            = 0;
+        int                      loglevel       = GF_LOG_TRACE;
+
+        gf_log (this->name, GF_LOG_INFO, "migrate data called on %s",
+                loc->path);
+        gettimeofday (&dir_start, NULL);
+
+        fd = fd_create (loc->inode, defrag->pid);
+        if (!fd) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to create fd");
+                goto out;
+        }
+
+        ret = syncop_opendir (this, loc, fd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to open dir %s",
+                        loc->path);
+                ret = -1;
+                goto out;
+        }
+
+        INIT_LIST_HEAD (&entries.list);
+
+        while ((ret = syncop_readdirp (this, fd, 131072, offset, NULL,
+                                       &entries)) != 0) {
+
+                if (ret < 0) {
+
+                        gf_log (this->name, GF_LOG_ERROR, "Readdir returned %s."
+                                " Aborting migrate-data",
+                                strerror(-ret));
+                        ret = -1;
+                        goto out;
+                }
+
+                if (list_empty (&entries.list))
+                        break;
+
+                free_entries = _gf_true;
+
+                list_for_each_entry_safe (entry, tmp, &entries.list, list) {
+                        if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
+                                ret = 1;
+                                goto out;
+                        }
+
+                        offset = entry->d_off;
+
+                        if (!strcmp (entry->d_name, ".") ||
+                            !strcmp (entry->d_name, ".."))
+                                continue;
+
+                        if (IA_ISDIR (entry->d_stat.ia_type))
+                                continue;
+
+                        defrag->num_files_lookedup++;
+                        if (defrag->stats == _gf_true) {
+                                gettimeofday (&start, NULL);
+                        }
+                        if (defrag->defrag_pattern &&
+                            (gf_defrag_pattern_match (defrag, entry->d_name,
+                                                      entry->d_stat.ia_size)
+                             == _gf_false)) {
+                                continue;
+                        }
+                        loc_wipe (&entry_loc);
+                        ret =dht_build_child_loc (this, &entry_loc, loc,
+                                                  entry->d_name);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Child loc"
+                                        " build failed");
+                                goto out;
+                        }
+
+                        if (uuid_is_null (entry->d_stat.ia_gfid)) {
+                                gf_log (this->name, GF_LOG_ERROR, "%s/%s"
+                                        " gfid not present", loc->path,
+                                         entry->d_name);
+                                continue;
+                        }
+
+                        uuid_copy (entry_loc.gfid, entry->d_stat.ia_gfid);
+
+                        if (uuid_is_null (loc->gfid)) {
+                                gf_log (this->name, GF_LOG_ERROR, "%s/%s"
+                                        " gfid not present", loc->path,
+                                         entry->d_name);
+                                continue;
+                        }
+
+                        uuid_copy (entry_loc.pargfid, loc->gfid);
+
+                        entry_loc.inode->ia_type = entry->d_stat.ia_type;
+
+                        ret = syncop_lookup (this, &entry_loc, NULL, &iatt,
+                                             NULL, NULL);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "%s"
+                                        " lookup failed", entry_loc.path);
+                                ret = -1;
+                                continue;
+                        }
+
+                        ret = syncop_getxattr (this, &entry_loc, &dict,
+                                               GF_XATTR_NODE_UUID_KEY);
+                        if(ret < 0) {
+                                gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                        "get node-uuid for %s", entry_loc.path);
+                                ret = -1;
+                                continue;
+                        }
+
+                        ret = dict_get_str (dict, GF_XATTR_NODE_UUID_KEY,
+                                            &uuid_str);
+                        if(ret < 0) {
+                                gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                        "get node-uuid from dict for %s",
+                                        entry_loc.path);
+                                ret = -1;
+                                continue;
+                        }
+
+                        if (uuid_parse (uuid_str, node_uuid)) {
+                                gf_log (this->name, GF_LOG_ERROR, "uuid_parse "
+                                        "failed for %s", entry_loc.path);
+                                continue;
+                        }
+
+                        /* if file belongs to different node, skip migration
+                         * the other node will take responsibility of migration
+                         */
+                        if (uuid_compare (node_uuid, defrag->node_uuid)) {
+                                gf_log (this->name, GF_LOG_TRACE, "%s does not"
+                                        "belong to this node", entry_loc.path);
+                                continue;
+                        }
+
+                        uuid_str = NULL;
+
+                        dict_del (dict, GF_XATTR_NODE_UUID_KEY);
+
+
+                        /* if distribute is present, it will honor this key.
+                         * -1, ENODATA is returned if distribute is not present
+                         * or file doesn't have a link-file. If file has
+                         * link-file, the path of link-file will be the value,
+                         * and also that guarantees that file has to be mostly
+                         * migrated */
+
+                        ret = syncop_getxattr (this, &entry_loc, &dict,
+                                               GF_XATTR_LINKINFO_KEY);
+                        if (ret < 0) {
+                                if (-ret != ENODATA) {
+                                        loglevel = GF_LOG_ERROR;
+                                        defrag->total_failures += 1;
+                                } else {
+                                        loglevel = GF_LOG_TRACE;
+                                }
+                                gf_log (this->name, loglevel, "%s: failed to "
+                                        "get "GF_XATTR_LINKINFO_KEY" key - %s",
+                                        entry_loc.path, strerror (-ret));
+                                ret = -1;
+                                continue;
+                        }
+
+                        ret = syncop_setxattr (this, &entry_loc, migrate_data,
+                                               0);
+                        if (ret) {
+                                err = op_errno;
+                                /* errno is overloaded. See
+                                 * rebalance_task_completion () */
+                                if (err != ENOSPC) {
+                                        gf_log (this->name, GF_LOG_DEBUG,
+                                                "migrate-data skipped for %s"
+                                                " due to space constraints",
+                                                entry_loc.path);
+                                        defrag->skipped +=1;
+                                } else{
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "migrate-data failed for %s",
+                                                entry_loc.path);
+                                        defrag->total_failures +=1;
+                                }
+                        }
+
+                        if (ret < 0) {
+                                op_errno = -ret;
+                                ret = gf_defrag_handle_migrate_error (op_errno,
+                                                                      defrag);
+
+                                if (!ret)
+                                        gf_log (this->name, GF_LOG_DEBUG,
+                                                "migrate-data on %s failed: %s",
+                                                entry_loc.path,
+                                                strerror (op_errno));
+                                else if (ret == 1)
+                                        continue;
+                                else if (ret == -1)
+                                        goto out;
+                        }
+
+                        LOCK (&defrag->lock);
+                        {
+                                defrag->total_files += 1;
+                                defrag->total_data += iatt.ia_size;
+                        }
+                        UNLOCK (&defrag->lock);
+                        if (defrag->stats == _gf_true) {
+                                gettimeofday (&end, NULL);
+                                elapsed = (end.tv_sec - start.tv_sec) * 1e6 +
+                                          (end.tv_usec - start.tv_usec);
+                                gf_log (this->name, GF_LOG_INFO, "Migration of "
+                                        "file:%s size:%"PRIu64" bytes took %.2f"
+                                        "secs", entry_loc.path, iatt.ia_size,
+                                         elapsed/1e6);
+                        }
+                }
+
+                gf_dirent_free (&entries);
+                free_entries = _gf_false;
+                INIT_LIST_HEAD (&entries.list);
+        }
+
+        gettimeofday (&end, NULL);
+        elapsed = (end.tv_sec - dir_start.tv_sec) * 1e6 +
+                  (end.tv_usec - dir_start.tv_usec);
+        gf_log (this->name, GF_LOG_INFO, "Migration operation on dir %s took "
+                "%.2f secs", loc->path, elapsed/1e6);
+        ret = 0;
+out:
+        if (free_entries)
+                gf_dirent_free (&entries);
+
+        loc_wipe (&entry_loc);
+
+        if (dict)
+                dict_unref(dict);
+
+        if (fd)
+                fd_unref (fd);
+        return ret;
+
+}
+
+int
+gf_defrag_fix_layout (xlator_t *this, gf_defrag_info_t *defrag, loc_t *loc,
+                  dict_t *fix_layout, dict_t *migrate_data)
+{
+        int                      ret            = -1;
+        loc_t                    entry_loc      = {0,};
+        fd_t                    *fd             = NULL;
+        gf_dirent_t              entries;
+        gf_dirent_t             *tmp            = NULL;
+        gf_dirent_t             *entry          = NULL;
+        gf_boolean_t             free_entries   = _gf_false;
+        dict_t                  *dict           = NULL;
+        off_t                    offset         = 0;
+        struct iatt              iatt           = {0,};
+
+        ret = syncop_lookup (this, loc, NULL, &iatt, NULL, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Lookup failed on %s",
+                        loc->path);
+                ret = -1;
+                goto out;
+        }
+
+        if (defrag->cmd != GF_DEFRAG_CMD_START_LAYOUT_FIX) {
+                ret = gf_defrag_migrate_data (this, defrag, loc, migrate_data);
+                if (ret)
+                        goto out;
+        }
+
+        gf_log (this->name, GF_LOG_TRACE, "fix layout called on %s", loc->path);
+
+        fd = fd_create (loc->inode, defrag->pid);
+        if (!fd) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to create fd");
+                ret = -1;
+                goto out;
+        }
+
+        ret = syncop_opendir (this, loc, fd);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to open dir %s",
+                        loc->path);
+                ret = -1;
+                goto out;
+        }
+
+        INIT_LIST_HEAD (&entries.list);
+        while ((ret = syncop_readdirp (this, fd, 131072, offset, NULL,
+                &entries)) != 0)
+        {
+
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR, "Readdir returned %s"
+                                ". Aborting fix-layout",strerror(-ret));
+                        ret = -1;
+                        goto out;
+                }
+
+                if (list_empty (&entries.list))
+                        break;
+
+                free_entries = _gf_true;
+
+                list_for_each_entry_safe (entry, tmp, &entries.list, list) {
+                        if (defrag->defrag_status != GF_DEFRAG_STATUS_STARTED) {
+                                ret = 1;
+                                goto out;
+                        }
+
+                        offset = entry->d_off;
+
+                        if (!strcmp (entry->d_name, ".") ||
+                            !strcmp (entry->d_name, ".."))
+                                continue;
+
+                        if (!IA_ISDIR (entry->d_stat.ia_type))
+                                continue;
+
+                        loc_wipe (&entry_loc);
+                        ret =dht_build_child_loc (this, &entry_loc, loc,
+                                                  entry->d_name);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Child loc"
+                                        " build failed");
+                                goto out;
+                        }
+
+                        if (uuid_is_null (entry->d_stat.ia_gfid)) {
+                                gf_log (this->name, GF_LOG_ERROR, "%s/%s"
+                                        " gfid not present", loc->path,
+                                         entry->d_name);
+                                continue;
+                        }
+
+                        entry_loc.inode->ia_type = entry->d_stat.ia_type;
+
+                        uuid_copy (entry_loc.gfid, entry->d_stat.ia_gfid);
+                        if (uuid_is_null (loc->gfid)) {
+                                gf_log (this->name, GF_LOG_ERROR, "%s/%s"
+                                        " gfid not present", loc->path,
+                                         entry->d_name);
+                                continue;
+                        }
+
+                        uuid_copy (entry_loc.pargfid, loc->gfid);
+
+                        ret = syncop_lookup (this, &entry_loc, NULL, &iatt,
+                                             NULL, NULL);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "%s"
+                                        " lookup failed", entry_loc.path);
+                                ret = -1;
+                                continue;
+                        }
+
+                        ret = syncop_setxattr (this, &entry_loc, fix_layout,
+                                               0);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Setxattr "
+                                        "failed for %s", entry_loc.path);
+                                defrag->defrag_status =
+                                GF_DEFRAG_STATUS_FAILED;
+                                defrag->total_failures ++;
+                                ret = -1;
+                                goto out;
+                        }
+                        ret = gf_defrag_fix_layout (this, defrag, &entry_loc,
+                                                    fix_layout, migrate_data);
+
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Fix layout "
+                                        "failed for %s", entry_loc.path);
+                                defrag->total_failures++;
+                                goto out;
+                        }
+
+                }
+                gf_dirent_free (&entries);
+                free_entries = _gf_false;
+                INIT_LIST_HEAD (&entries.list);
+        }
+
+        ret = 0;
+out:
+        if (free_entries)
+                gf_dirent_free (&entries);
+
+        loc_wipe (&entry_loc);
+
+        if (dict)
+                dict_unref(dict);
+
+        if (fd)
+                fd_unref (fd);
+
+        return ret;
+
+}
+
+
+int
+gf_defrag_start_crawl (void *data)
+{
+        xlator_t                *this   = NULL;
+        dht_conf_t              *conf   = NULL;
+        gf_defrag_info_t        *defrag = NULL;
+        int                      ret    = -1;
+        loc_t                    loc    = {0,};
+        struct iatt              iatt   = {0,};
+        struct iatt              parent = {0,};
+        dict_t                  *fix_layout = NULL;
+        dict_t                  *migrate_data = NULL;
+        dict_t                  *status = NULL;
+        glusterfs_ctx_t         *ctx = NULL;
+
+        this = data;
+        if (!this)
+                goto out;
+
+        ctx = this->ctx;
+        if (!ctx)
+                goto out;
+
+        conf = this->private;
+        if (!conf)
+                goto out;
+
+        defrag = conf->defrag;
+        if (!defrag)
+                goto out;
+
+        gettimeofday (&defrag->start_time, NULL);
+        dht_build_root_inode (this, &defrag->root_inode);
+        if (!defrag->root_inode)
+                goto out;
+
+        dht_build_root_loc (defrag->root_inode, &loc);
+
+        /* fix-layout on '/' first */
+
+        ret = syncop_lookup (this, &loc, NULL, &iatt, NULL, &parent);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "look up on / failed");
+                ret = -1;
+                goto out;
+        }
+
+        fix_layout = dict_new ();
+        if (!fix_layout) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_str (fix_layout, GF_XATTR_FIX_LAYOUT_KEY, "yes");
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Failed to set dict str");
+                goto out;
+        }
+
+        ret = syncop_setxattr (this, &loc, fix_layout, 0);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "fix layout on %s failed",
+                        loc.path);
+                defrag->total_failures++;
+                ret = -1;
+                goto out;
+        }
+
+        if (defrag->cmd != GF_DEFRAG_CMD_START_LAYOUT_FIX) {
+                migrate_data = dict_new ();
+                if (!migrate_data) {
+                        ret = -1;
+                        goto out;
+                }
+                if (defrag->cmd == GF_DEFRAG_CMD_START_FORCE)
+                        ret = dict_set_str (migrate_data,
+                                            "distribute.migrate-data", "force");
+                else
+                        ret = dict_set_str (migrate_data,
+                                            "distribute.migrate-data",
+                                            "non-force");
+                if (ret)
+                        goto out;
+        }
+        ret = gf_defrag_fix_layout (this, defrag, &loc, fix_layout,
+                                    migrate_data);
+        if ((defrag->defrag_status != GF_DEFRAG_STATUS_STOPPED) &&
+            (defrag->defrag_status != GF_DEFRAG_STATUS_FAILED)) {
+                defrag->defrag_status = GF_DEFRAG_STATUS_COMPLETE;
+        }
+
+
+
+out:
+        LOCK (&defrag->lock);
+        {
+                status = dict_new ();
+                gf_defrag_status_get (defrag, status);
+                if (ctx->notify)
+                        ctx->notify (GF_EN_DEFRAG_STATUS, status);
+                if (status)
+                        dict_unref (status);
+                defrag->is_exiting = 1;
+        }
+        UNLOCK (&defrag->lock);
+
+        if (defrag) {
+                GF_FREE (defrag);
+                conf->defrag = NULL;
+        }
+
+        return ret;
+}
+
+
+static int
+gf_defrag_done  (int ret, call_frame_t *sync_frame, void *data)
+{
+        gf_listener_stop (sync_frame->this);
+
+        STACK_DESTROY (sync_frame->root);
+        kill (getpid(), SIGTERM);
+        return 0;
+}
+
+void *
+gf_defrag_start (void *data)
+{
+        int                      ret    = -1;
+        call_frame_t            *frame  = NULL;
+        dht_conf_t              *conf   = NULL;
+        gf_defrag_info_t        *defrag = NULL;
+        xlator_t                *this  = NULL;
+
+        this = data;
+        conf = this->private;
+        if (!conf)
+                goto out;
+
+        defrag = conf->defrag;
+        if (!defrag)
+                goto out;
+
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame)
+                goto out;
+
+        frame->root->pid = GF_CLIENT_PID_DEFRAG;
+
+        defrag->pid = frame->root->pid;
+
+        defrag->defrag_status = GF_DEFRAG_STATUS_STARTED;
+
+        ret = synctask_new (this->ctx->env, gf_defrag_start_crawl,
+                            gf_defrag_done, frame, this);
+
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR, "Could not create"
+                        " task for rebalance");
+out:
+        return NULL;
+}
+
+int
+gf_defrag_status_get (gf_defrag_info_t *defrag, dict_t *dict)
+{
+        int      ret    = 0;
+        uint64_t files  = 0;
+        uint64_t size   = 0;
+        uint64_t lookup = 0;
+        uint64_t failures = 0;
+        uint64_t skipped = 0;
+        char     *status = "";
+        double   elapsed = 0;
+        struct timeval end = {0,};
+
+
+        if (!defrag)
+                goto out;
+
+        ret = 0;
+        if (defrag->defrag_status == GF_DEFRAG_STATUS_NOT_STARTED)
+                goto out;
+
+        files  = defrag->total_files;
+        size   = defrag->total_data;
+        lookup = defrag->num_files_lookedup;
+        failures = defrag->total_failures;
+        skipped = defrag->skipped;
+
+        gettimeofday (&end, NULL);
+
+        elapsed = end.tv_sec - defrag->start_time.tv_sec;
+
+        if (!dict)
+                goto log;
+
+        ret = dict_set_uint64 (dict, "files", files);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "failed to set file count");
+
+        ret = dict_set_uint64 (dict, "size", size);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "failed to set size of xfer");
+
+        ret = dict_set_uint64 (dict, "lookups", lookup);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "failed to set lookedup file count");
+
+
+        ret = dict_set_int32 (dict, "status", defrag->defrag_status);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "failed to set status");
+        if (elapsed) {
+                ret = dict_set_double (dict, "run-time", elapsed);
+                if (ret)
+                        gf_log (THIS->name, GF_LOG_WARNING,
+                                "failed to set run-time");
+        }
+
+        ret = dict_set_uint64 (dict, "failures", failures);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "failed to set failure count");
+
+        ret = dict_set_uint64 (dict, "skipped", skipped);
+        if (ret)
+                gf_log (THIS->name, GF_LOG_WARNING,
+                        "failed to set skipped file count");
+log:
+        switch (defrag->defrag_status) {
+        case GF_DEFRAG_STATUS_NOT_STARTED:
+                status = "not started";
+                break;
+        case GF_DEFRAG_STATUS_STARTED:
+                status = "in progress";
+                break;
+        case GF_DEFRAG_STATUS_STOPPED:
+                status = "stopped";
+                break;
+        case GF_DEFRAG_STATUS_COMPLETE:
+                status = "completed";
+                break;
+        case GF_DEFRAG_STATUS_FAILED:
+                status = "failed";
+                break;
+        default:
+                break;
+        }
+
+        gf_log (THIS->name, GF_LOG_INFO, "Rebalance is %s. Time taken is %.2f "
+                "secs", status, elapsed);
+        gf_log (THIS->name, GF_LOG_INFO, "Files migrated: %"PRIu64", size: %"
+                PRIu64", lookups: %"PRIu64", failures: %"PRIu64", skipped: "
+                "%"PRIu64, files, size, lookup, failures, skipped);
+
+
+out:
+        return 0;
+}
+
+int
+gf_defrag_stop (gf_defrag_info_t *defrag, gf_defrag_status_t status,
+                dict_t *output)
+{
+        /* TODO: set a variable 'stop_defrag' here, it should be checked
+           in defrag loop */
+        int     ret = -1;
+        GF_ASSERT (defrag);
+
+        if (defrag->defrag_status == GF_DEFRAG_STATUS_NOT_STARTED) {
+                goto out;
+        }
+
+        gf_log ("", GF_LOG_INFO, "Received stop command on rebalance");
+        defrag->defrag_status = status;
+
+        if (output)
+                gf_defrag_status_get (defrag, output);
+        ret = 0;
+out:
+        gf_log ("", GF_LOG_DEBUG, "Returning %d", ret);
+        return ret;
+}
diff --git a/xlators/cluster/dht/src/dht-rename.c b/xlators/cluster/dht/src/dht-rename.c
index 52185d2ad..925538cc8 100644
--- a/xlators/cluster/dht/src/dht-rename.c
+++ b/xlators/cluster/dht/src/dht-rename.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 /* TODO: link(oldpath, newpath) fails if newpath already exists. DHT should
@@ -33,232 +24,348 @@
 
 int
 dht_rename_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
+                    int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
                     struct iatt *preoldparent, struct iatt *postoldparent,
-                    struct iatt *prenewparent, struct iatt *postnewparent)
+                    struct iatt *prenewparent, struct iatt *postnewparent,
+                    dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
 
+        local = frame->local;
+        prev = cookie;
 
-	local = frame->local;
-	prev = cookie;
-
-	if (op_ret == -1) {
-		/* TODO: undo the damage */
-
-		gf_log (this->name, GF_LOG_DEBUG,
-			"rename %s -> %s on %s failed (%s)",
-			local->loc.path, local->loc2.path,
-			prev->this->name, strerror (op_errno));
-
-		local->op_ret   = op_ret;
-		local->op_errno = op_errno;
-	} else {
-		/* TODO: construct proper stbuf for dir */
-                /* 
-                 * FIXME: is this the correct way to build stbuf and
-                 * parent bufs?
-                */
-                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
-                dht_iatt_merge (this, &local->preoldparent, preoldparent,
-                                prev->this);
-                dht_iatt_merge (this, &local->postoldparent, postoldparent,
-                                prev->this);
-                dht_iatt_merge (this, &local->preparent, prenewparent,
-                                prev->this);
-                dht_iatt_merge (this, &local->postparent, postnewparent,
-                                prev->this);
-	}
+        if (op_ret == -1) {
+                /* TODO: undo the damage */
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt)) {
-                local->stbuf.ia_ino = local->loc.inode->ino;
+                gf_log (this->name, GF_LOG_INFO,
+                        "rename %s -> %s on %s failed (%s)",
+                        local->loc.path, local->loc2.path,
+                        prev->this->name, strerror (op_errno));
 
-                local->preoldparent.ia_ino = local->loc.parent->ino;
-                local->postoldparent.ia_ino = local->loc.parent->ino;
+                local->op_ret   = op_ret;
+                local->op_errno = op_errno;
+                goto unwind;
+        }
+        /* TODO: construct proper stbuf for dir */
+        /*
+         * FIXME: is this the correct way to build stbuf and
+         * parent bufs?
+         */
+        dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+        dht_iatt_merge (this, &local->preoldparent, preoldparent,
+                        prev->this);
+        dht_iatt_merge (this, &local->postoldparent, postoldparent,
+                        prev->this);
+        dht_iatt_merge (this, &local->preparent, prenewparent,
+                        prev->this);
+        dht_iatt_merge (this, &local->postparent, postnewparent,
+                        prev->this);
 
-                local->preparent.ia_ino = local->loc2.parent->ino;
-                local->postparent.ia_ino = local->loc2.parent->ino;
 
+unwind:
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
                 WIPE (&local->preoldparent);
                 WIPE (&local->postoldparent);
                 WIPE (&local->preparent);
                 WIPE (&local->postparent);
 
-		DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
-				  &local->stbuf, &local->preoldparent,
+                DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+                DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
+                                  &local->stbuf, &local->preoldparent,
                                   &local->postoldparent,
-                                  &local->preparent, &local->postparent);
-	}
+                                  &local->preparent, &local->postparent, xdata);
+        }
 
-	return 0;
+        return 0;
 }
 
 
+int
+dht_rename_hashed_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                           int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
+                           struct iatt *preoldparent,
+                           struct iatt *postoldparent,
+                           struct iatt *prenewparent,
+                           struct iatt *postnewparent, dict_t *xdata)
+{
+        dht_conf_t   *conf = NULL;
+        dht_local_t  *local = NULL;
+        int           call_cnt = 0;
+        call_frame_t *prev = NULL;
+        int           i = 0;
+
+        conf = this->private;
+        local = frame->local;
+        prev = cookie;
+
+        if (op_ret == -1) {
+                /* TODO: undo the damage */
+
+                gf_log (this->name, GF_LOG_INFO,
+                        "rename %s -> %s on %s failed (%s)",
+                        local->loc.path, local->loc2.path,
+                        prev->this->name, strerror (op_errno));
+
+                local->op_ret   = op_ret;
+                local->op_errno = op_errno;
+                goto unwind;
+        }
+        /* TODO: construct proper stbuf for dir */
+        /*
+         * FIXME: is this the correct way to build stbuf and
+         * parent bufs?
+         */
+        dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+        dht_iatt_merge (this, &local->preoldparent, preoldparent,
+                        prev->this);
+        dht_iatt_merge (this, &local->postoldparent, postoldparent,
+                        prev->this);
+        dht_iatt_merge (this, &local->preparent, prenewparent,
+                        prev->this);
+        dht_iatt_merge (this, &local->postparent, postnewparent,
+                        prev->this);
+
+        call_cnt = local->call_cnt = conf->subvolume_cnt - 1;
+
+        if (!local->call_cnt)
+                goto unwind;
+
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (conf->subvolumes[i] == local->dst_hashed)
+                        continue;
+                STACK_WIND (frame, dht_rename_dir_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->rename,
+                            &local->loc, &local->loc2, NULL);
+                if (!--call_cnt)
+                        break;
+        }
+
+
+        return 0;
+unwind:
+        WIPE (&local->preoldparent);
+        WIPE (&local->postoldparent);
+        WIPE (&local->preparent);
+        WIPE (&local->postparent);
+
+        DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+        DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
+                          &local->stbuf, &local->preoldparent,
+                          &local->postoldparent,
+                          &local->preparent, &local->postparent, NULL);
+
+        return 0;
+}
+
 
 int
 dht_rename_dir_do (call_frame_t *frame, xlator_t *this)
 {
-	dht_local_t  *local = NULL;
-	dht_conf_t   *conf = NULL;
-	int           i = 0;
+        dht_local_t  *local = NULL;
 
-	conf = this->private;
-	local = frame->local;
+        local = frame->local;
 
-	if (local->op_ret == -1)
-		goto err;
+        if (local->op_ret == -1)
+                goto err;
 
-	local->call_cnt = conf->subvolume_cnt;
-	local->op_ret = 0;
+        local->op_ret = 0;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_rename_dir_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->rename,
-			    &local->loc, &local->loc2);
-	}
-
-	return 0;
+        STACK_WIND (frame, dht_rename_hashed_dir_cbk,
+                    local->dst_hashed,
+                    local->dst_hashed->fops->rename,
+                    &local->loc, &local->loc2, NULL);
+        return 0;
 
 err:
-	DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno, NULL, NULL,
-                          NULL, NULL, NULL);
-	return 0;
+        DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno, NULL, NULL,
+                          NULL, NULL, NULL, NULL);
+        return 0;
 }
 
 
 int
 dht_rename_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			int op_ret, int op_errno, gf_dirent_t *entries)
+                        int op_ret, int op_errno, gf_dirent_t *entries,
+                        dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = -1;
-	call_frame_t *prev = NULL;
-
-	local = frame->local;
-	prev  = cookie;
-
-	if (op_ret > 2) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"readdir on %s for %s returned %d entries",
-			prev->this->name, local->loc.path, op_ret);
-		local->op_ret = -1;
-		local->op_errno = ENOTEMPTY;
-	}
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = -1;
+        call_frame_t *prev = NULL;
+
+        local = frame->local;
+        prev  = cookie;
+
+        if (op_ret > 2) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "readdir on %s for %s returned %d entries",
+                        prev->this->name, local->loc.path, op_ret);
+                local->op_ret = -1;
+                local->op_errno = ENOTEMPTY;
+        }
 
-	this_call_cnt = dht_frame_return (frame);
+        this_call_cnt = dht_frame_return (frame);
 
-	if (is_last_call (this_call_cnt)) {
-		dht_rename_dir_do (frame, this);
-	}
+        if (is_last_call (this_call_cnt)) {
+                dht_rename_dir_do (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_rename_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			int op_ret, int op_errno, fd_t *fd)
+                        int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = -1;
-	call_frame_t *prev = NULL;
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = -1;
+        call_frame_t *prev = NULL;
 
 
-	local = frame->local;
-	prev  = cookie;
+        local = frame->local;
+        prev  = cookie;
 
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"opendir on %s for %s failed (%s)",
-			prev->this->name, local->loc.path,
-			strerror (op_errno));
-		goto err;
-	}
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "opendir on %s for %s failed (%s)",
+                        prev->this->name, local->loc.path,
+                        strerror (op_errno));
+                goto err;
+        }
 
-	STACK_WIND (frame, dht_rename_readdir_cbk,
-		    prev->this, prev->this->fops->readdir,
-		    local->fd, 4096, 0);
+        STACK_WIND (frame, dht_rename_readdir_cbk,
+                    prev->this, prev->this->fops->readdir,
+                    local->fd, 4096, 0, NULL);
 
-	return 0;
+        return 0;
 
 err:
-	this_call_cnt = dht_frame_return (frame);
+        this_call_cnt = dht_frame_return (frame);
 
-	if (is_last_call (this_call_cnt)) {
-		dht_rename_dir_do (frame, this);
-	}
+        if (is_last_call (this_call_cnt)) {
+                dht_rename_dir_do (frame, this);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_rename_dir (call_frame_t *frame, xlator_t *this)
 {
-	dht_conf_t  *conf = NULL;
-	dht_local_t *local = NULL;
-	int          i = 0;
-	int          op_errno = -1;
+        dht_conf_t  *conf = NULL;
+        dht_local_t *local = NULL;
+        int          i = 0;
+        int          op_errno = -1;
 
 
-	conf = frame->this->private;
-	local = frame->local;
+        conf = frame->this->private;
+        local = frame->local;
 
-	local->call_cnt = conf->subvolume_cnt;
+        local->call_cnt = conf->subvolume_cnt;
 
         for (i = 0; i < conf->subvolume_cnt; i++) {
                 if (!conf->subvolume_status[i]) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "one of the subvolumes down (%s)",
+                                conf->subvolumes[i]->name);
                         op_errno = ENOTCONN;
                         goto err;
                 }
         }
 
-	local->fd = fd_create (local->loc.inode, frame->root->pid);
-	if (!local->fd) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
+        local->fd = fd_create (local->loc.inode, frame->root->pid);
+        if (!local->fd) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	local->op_ret = 0;
+        local->op_ret = 0;
 
-	if (!local->dst_cached) {
-		dht_rename_dir_do (frame, this);
-		return 0;
-	}
+        if (!local->dst_cached) {
+                dht_rename_dir_do (frame, this);
+                return 0;
+        }
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_rename_opendir_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->opendir,
-			    &local->loc2, local->fd);
-	}
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                STACK_WIND (frame, dht_rename_opendir_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->opendir,
+                            &local->loc2, local->fd, NULL);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
-	return 0;
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+                          NULL, NULL);
+        return 0;
 }
+#define DHT_MARK_FOP_INTERNAL(xattr) do {                                      \
+                int tmp = -1;                                                  \
+                if (!xattr) {                                                  \
+                        xattr = dict_new ();                                   \
+                        if (!xattr)                                            \
+                                break;                                         \
+                }                                                              \
+                tmp = dict_set_str (xattr, GLUSTERFS_INTERNAL_FOP_KEY, "yes"); \
+                if (tmp) {                                                     \
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to set"      \
+                                " internal dict key for %s", local->loc.path); \
+                }                                                              \
+        }while (0)
+
+#define DHT_MARKER_DONT_ACCOUNT(xattr) do {                             \
+                int tmp = -1;                                                  \
+                if (!xattr) {                                                  \
+                        xattr = dict_new ();                                   \
+                        if (!xattr)                                            \
+                                break;                                         \
+                }                                                              \
+                tmp = dict_set_str (xattr, GLUSTERFS_MARKER_DONT_ACCOUNT_KEY,  \
+                                    "yes");                                    \
+                if (tmp) {                                                     \
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to set"      \
+                                " marker dont account key for %s", local->loc.path); \
+                }                                                              \
+        }while (0)
+
+int
+dht_rename_done (call_frame_t *frame, xlator_t *this)
+{
+        dht_local_t     *local = NULL;
+
+        local = frame->local;
+
+        if (local->linked == _gf_true) {
+                local->linked = _gf_false;
+                dht_linkfile_attr_heal (frame, this);
+        }
+        DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+        DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
+                          &local->stbuf, &local->preoldparent,
+                          &local->postoldparent, &local->preparent,
+                          &local->postparent, NULL);
 
+        return 0;
+}
 
 int
 dht_rename_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int32_t op_ret, int32_t op_errno, struct iatt *preparent,
-                       struct iatt *postparent)
+                       int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                       struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	call_frame_t *prev = NULL;
-	int           this_call_cnt = 0;
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           this_call_cnt = 0;
 
-	local = frame->local;
-	prev  = cookie;
+        local = frame->local;
+        prev  = cookie;
 
         if (!local) {
                 gf_log (this->name, GF_LOG_ERROR,
@@ -266,82 +373,108 @@ dht_rename_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 goto out;
         }
 
-	this_call_cnt = dht_frame_return (frame);
+        this_call_cnt = dht_frame_return (frame);
 
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"unlink on %s failed (%s)",
-			prev->this->name, strerror (op_errno));
-	}
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: unlink on %s failed (%s)",
+                        local->loc.path, prev->this->name, strerror (op_errno));
+        }
 
         WIPE (&local->preoldparent);
         WIPE (&local->postoldparent);
         WIPE (&local->preparent);
         WIPE (&local->postparent);
 
-	if (is_last_call (this_call_cnt)) {
-		DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
-				  &local->stbuf, &local->preoldparent,
-                                  &local->postoldparent, &local->preparent,
-                                  &local->postparent);
+        if (is_last_call (this_call_cnt)) {
+                dht_rename_done (frame, this);
         }
 
 out:
-	return 0;
+        return 0;
 }
 
 
 int
 dht_rename_cleanup (call_frame_t *frame)
 {
-	dht_local_t *local = NULL;
-	xlator_t    *this = NULL;
-	xlator_t    *src_hashed = NULL;
-	xlator_t    *src_cached = NULL;
-	xlator_t    *dst_hashed = NULL;
-	xlator_t    *dst_cached = NULL;
-	int          call_cnt = 0;
+        dht_local_t *local = NULL;
+        xlator_t    *this = NULL;
+        xlator_t    *src_hashed = NULL;
+        xlator_t    *src_cached = NULL;
+        xlator_t    *dst_hashed = NULL;
+        xlator_t    *dst_cached = NULL;
+        int          call_cnt = 0;
+        dict_t      *xattr      = NULL;
+
+        local = frame->local;
+        this  = frame->this;
+
+        src_hashed = local->src_hashed;
+        src_cached = local->src_cached;
+        dst_hashed = local->dst_hashed;
+        dst_cached = local->dst_cached;
+
+        if (src_cached == dst_cached)
+                goto nolinks;
 
+        if (dst_hashed != src_hashed && dst_hashed != src_cached)
+                call_cnt++;
 
-	local = frame->local;
-	this  = frame->this;
+        if (src_cached != dst_hashed)
+                call_cnt++;
 
-	src_hashed = local->src_hashed;
-	src_cached = local->src_cached;
-	dst_hashed = local->dst_hashed;
-	dst_cached = local->dst_cached;
+        local->call_cnt = call_cnt;
 
-	if (src_cached == dst_cached)
-		goto nolinks;
+        if (!call_cnt)
+                goto nolinks;
 
-	if (dst_hashed != src_hashed && dst_hashed != src_cached)
-		call_cnt++;
+        DHT_MARK_FOP_INTERNAL (xattr);
 
-	if (src_cached != dst_hashed)
-		call_cnt++;
+        if (dst_hashed != src_hashed && dst_hashed != src_cached) {
+                dict_t *xattr_new = NULL;
 
-	local->call_cnt = call_cnt;
+                gf_log (this->name, GF_LOG_TRACE,
+                        "unlinking linkfile %s @ %s => %s",
+                        local->loc.path, dst_hashed->name, src_cached->name);
 
-        if (!call_cnt)
-                goto nolinks;
+                xattr_new = dict_copy_with_ref (xattr, NULL);
+
+
+                DHT_MARKER_DONT_ACCOUNT(xattr_new);
 
-	if (dst_hashed != src_hashed && dst_hashed != src_cached) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"unlinking linkfile %s @ %s => %s",
-			local->loc.path, dst_hashed->name, src_cached->name);
                 STACK_WIND (frame, dht_rename_unlink_cbk,
                             dst_hashed, dst_hashed->fops->unlink,
-                            &local->loc);
-	}
+                            &local->loc, 0, xattr_new);
 
-	if (src_cached != dst_hashed) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"unlinking link %s => %s (%s)", local->loc.path,
-			local->loc2.path, src_cached->name);
-		STACK_WIND (frame, dht_rename_unlink_cbk,
-			    src_cached, src_cached->fops->unlink,
-			    &local->loc2);
-	}
+                dict_unref (xattr_new);
+                xattr_new = NULL;
+        }
+
+        if (src_cached != dst_hashed) {
+                dict_t *xattr_new = NULL;
+
+                gf_log (this->name, GF_LOG_TRACE,
+                        "unlinking link %s => %s (%s)", local->loc.path,
+                        local->loc2.path, src_cached->name);
+
+                xattr_new = dict_copy_with_ref (xattr, NULL);
+
+                if (uuid_compare (local->loc.pargfid,
+                                  local->loc2.pargfid) == 0) {
+                        DHT_MARKER_DONT_ACCOUNT(xattr_new);
+                }
+
+                STACK_WIND (frame, dht_rename_unlink_cbk,
+                            src_cached, src_cached->fops->unlink,
+                            &local->loc2, 0, xattr_new);
+
+                dict_unref (xattr_new);
+                xattr_new = NULL;
+        }
+
+        if (xattr)
+                dict_unref (xattr);
 
         return 0;
 
@@ -351,10 +484,40 @@ nolinks:
         WIPE (&local->preparent);
         WIPE (&local->postparent);
 
-	DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
-			  &local->stbuf, &local->preoldparent,
+        DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+        DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
+                          &local->stbuf, &local->preoldparent,
                           &local->postoldparent, &local->preparent,
-                          &local->postparent);
+                          &local->postparent, NULL);
+
+        return 0;
+}
+
+
+int
+dht_rename_links_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                             int32_t op_ret, int32_t op_errno,
+                             inode_t *inode, struct iatt *stbuf,
+                             struct iatt *preparent, struct iatt *postparent,
+                             dict_t *xdata)
+{
+        call_frame_t *prev = NULL;
+        dht_local_t  *local = NULL;
+
+        prev = cookie;
+        local = frame->local;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "link/file %s on %s failed (%s)",
+                        local->loc.path, prev->this->name, strerror (op_errno));
+        }
+
+        if (local->linked == _gf_true) {
+                local->linked = _gf_false;
+                dht_linkfile_attr_heal (frame, this);
+        }
+        DHT_STACK_DESTROY (frame);
 
         return 0;
 }
@@ -362,119 +525,177 @@ nolinks:
 
 int
 dht_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
+                int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
                 struct iatt *preoldparent, struct iatt *postoldparent,
-                struct iatt *prenewparent, struct iatt *postnewparent)
+                struct iatt *prenewparent, struct iatt *postnewparent,
+                dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	call_frame_t *prev = NULL;
-	xlator_t     *src_hashed = NULL;
-	xlator_t     *src_cached = NULL;
-	xlator_t     *dst_hashed = NULL;
-	xlator_t     *dst_cached = NULL;
-	xlator_t     *rename_subvol = NULL;
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        xlator_t     *src_hashed = NULL;
+        xlator_t     *src_cached = NULL;
+        xlator_t     *dst_hashed = NULL;
+        xlator_t     *dst_cached = NULL;
+        xlator_t     *rename_subvol = NULL;
+        call_frame_t *link_frame = NULL;
+        dht_local_t *link_local = NULL;
+        dict_t       *xattr     = NULL;
+
+        local = frame->local;
+        prev = cookie;
+
+        src_hashed = local->src_hashed;
+        src_cached = local->src_cached;
+        dst_hashed = local->dst_hashed;
+        dst_cached = local->dst_cached;
+
+        if (local->linked == _gf_true)
+                FRAME_SU_UNDO (frame, dht_local_t);
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: rename on %s failed (%s)", local->loc.path,
+                        prev->this->name, strerror (op_errno));
+                local->op_ret   = op_ret;
+                local->op_errno = op_errno;
+                goto cleanup;
+        }
 
-	local = frame->local;
-	prev = cookie;
+        if ((src_cached == dst_cached) && (dst_hashed != dst_cached)) {
+                link_frame = copy_frame (frame);
+                if (!link_frame) {
+                        goto err;
+                }
 
-	src_hashed = local->src_hashed;
-	src_cached = local->src_cached;
-	dst_hashed = local->dst_hashed;
-	dst_cached = local->dst_cached;
+                /* fop value sent as maxvalue because it is not used
+                   anywhere in this case */
+                link_local = dht_local_init (link_frame, &local->loc2, NULL,
+                                             GF_FOP_MAXVALUE);
+                if (!link_local) {
+                        goto err;
+                }
 
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"rename on %s failed (%s)", prev->this->name,
-			strerror (op_errno));
-		local->op_ret   = op_ret;
-		local->op_errno = op_errno;
-		goto cleanup;
-	}
+                if (link_local->loc.inode)
+                        inode_unref (link_local->loc.inode);
+                link_local->loc.inode = inode_ref (local->loc.inode);
+                uuid_copy (link_local->gfid, local->loc.inode->gfid);
 
-        dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
-        dht_iatt_merge (this, &local->preoldparent, preoldparent, prev->this);
-        dht_iatt_merge (this, &local->postoldparent, postoldparent, prev->this);
-        dht_iatt_merge (this, &local->preparent, prenewparent, prev->this);
-        dht_iatt_merge (this, &local->postparent, postnewparent, prev->this);
+                dht_linkfile_create (link_frame, dht_rename_links_create_cbk,
+                                     this, src_cached, dst_hashed,
+                                     &link_local->loc);
+        }
+
+err:
+        /* Merge attrs only from src_cached. In case there of src_cached !=
+         * dst_hashed, this ignores linkfile attrs. */
+        if (prev->this == src_cached) {
+                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+                dht_iatt_merge (this, &local->preoldparent, preoldparent,
+                                prev->this);
+                dht_iatt_merge (this, &local->postoldparent, postoldparent,
+                                prev->this);
+                dht_iatt_merge (this, &local->preparent, prenewparent,
+                                prev->this);
+                dht_iatt_merge (this, &local->postparent, postnewparent,
+                                prev->this);
+        }
 
-        local->stbuf.ia_ino = local->loc.inode->ino;
 
-        local->preoldparent.ia_ino = local->loc.parent->ino;
-        local->postoldparent.ia_ino = local->loc.parent->ino;
+        /* NOTE: rename_subvol is the same subvolume from which dht_rename_cbk
+         *       is called. since rename has already happened on rename_subvol,
+         *       unlink should not be sent for oldpath (either linkfile or cached-file)
+         *       on rename_subvol. */
+        if (src_cached == dst_cached)
+                rename_subvol = src_cached;
+        else
+                rename_subvol = dst_hashed;
 
-        local->preparent.ia_ino = local->loc2.parent->ino;
-        local->postparent.ia_ino = local->loc2.parent->ino;
-	
-	/* NOTE: rename_subvol is the same subvolume from which dht_rename_cbk
-	 *       is called. since rename has already happened on rename_subvol,
-	 *       unlink should not be sent for oldpath (either linkfile or cached-file)
-	 *       on rename_subvol. */
-	if (src_cached == dst_cached)
-		rename_subvol = src_cached;
-	else
-		rename_subvol = dst_hashed;
+        /* TODO: delete files in background */
 
-	/* TODO: delete files in background */
+        if (src_cached != dst_hashed && src_cached != dst_cached)
+                local->call_cnt++;
 
-	if (src_cached != dst_hashed && src_cached != dst_cached)
-		local->call_cnt++;
+        if (src_hashed != rename_subvol && src_hashed != src_cached)
+                local->call_cnt++;
 
-	if (src_hashed != rename_subvol && src_hashed != src_cached)
-		local->call_cnt++;
+        if (dst_cached && dst_cached != dst_hashed && dst_cached != src_cached)
+                local->call_cnt++;
 
-	if (dst_cached && dst_cached != dst_hashed && dst_cached != src_cached)
-		local->call_cnt++;
+        if (local->call_cnt == 0)
+                goto unwind;
 
-	if (local->call_cnt == 0)
-		goto unwind;
+        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",
-			local->loc.path, src_cached->name);
+        if (src_cached != dst_hashed && src_cached != dst_cached) {
+                dict_t *xattr_new = NULL;
 
-		STACK_WIND (frame, dht_rename_unlink_cbk,
-			    src_cached, src_cached->fops->unlink,
-			    &local->loc);
-	}
+                xattr_new = dict_copy_with_ref (xattr, NULL);
 
-	if (src_hashed != rename_subvol && src_hashed != src_cached) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"deleting old src linkfile %s @ %s",
-			local->loc.path, src_hashed->name);
+                gf_log (this->name, GF_LOG_TRACE,
+                        "deleting old src datafile %s @ %s",
+                        local->loc.path, src_cached->name);
 
-		STACK_WIND (frame, dht_rename_unlink_cbk,
-			    src_hashed, src_hashed->fops->unlink,
-			    &local->loc);
-	}
+                if (uuid_compare (local->loc.pargfid,
+                                  local->loc2.pargfid) == 0) {
+                        DHT_MARKER_DONT_ACCOUNT(xattr_new);
+                }
 
-	if (dst_cached
-	    && (dst_cached != dst_hashed)
-	    && (dst_cached != src_cached)) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"deleting old dst datafile %s @ %s",
-			local->loc2.path, dst_cached->name);
+                STACK_WIND (frame, dht_rename_unlink_cbk,
+                            src_cached, src_cached->fops->unlink,
+                            &local->loc, 0, xattr_new);
 
-		STACK_WIND (frame, dht_rename_unlink_cbk,
-			    dst_cached, dst_cached->fops->unlink,
-			    &local->loc2);
-	}
-	return 0;
+                dict_unref (xattr_new);
+                xattr_new = NULL;
+        }
+
+        if (src_hashed != rename_subvol && src_hashed != src_cached) {
+                dict_t *xattr_new = NULL;
+
+                xattr_new = dict_copy_with_ref (xattr, NULL);
+
+                gf_log (this->name, GF_LOG_TRACE,
+                        "deleting old src linkfile %s @ %s",
+                        local->loc.path, src_hashed->name);
+
+                DHT_MARKER_DONT_ACCOUNT(xattr_new);
+
+                STACK_WIND (frame, dht_rename_unlink_cbk,
+                            src_hashed, src_hashed->fops->unlink,
+                            &local->loc, 0, xattr_new);
+
+                dict_unref (xattr_new);
+                xattr_new = NULL;
+        }
+
+        if (dst_cached
+            && (dst_cached != dst_hashed)
+            && (dst_cached != src_cached)) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "deleting old dst datafile %s @ %s",
+                        local->loc2.path, dst_cached->name);
+
+                STACK_WIND (frame, dht_rename_unlink_cbk,
+                            dst_cached, dst_cached->fops->unlink,
+                            &local->loc2, 0, xattr);
+        }
+        if (xattr)
+                dict_unref (xattr);
+        return 0;
 
 unwind:
         WIPE (&local->preoldparent);
         WIPE (&local->postoldparent);
         WIPE (&local->preparent);
         WIPE (&local->postparent);
+        if (xattr)
+                dict_unref (xattr);
 
-	DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
-			  &local->stbuf, &local->preoldparent,
-                          &local->postoldparent, &local->preparent,
-                          &local->postparent);
+        dht_rename_done (frame, this);
 
-	return 0;
+        return 0;
 
 cleanup:
+        if (xattr)
+                dict_unref (xattr);
         dht_rename_cleanup (frame);
 
         return 0;
@@ -484,67 +705,115 @@ cleanup:
 int
 dht_do_rename (call_frame_t *frame)
 {
-	dht_local_t *local = NULL;
-	xlator_t    *dst_hashed = NULL;
-	xlator_t    *src_cached = NULL;
-	xlator_t    *dst_cached = NULL;
-	xlator_t    *this = NULL;
-	xlator_t    *rename_subvol = NULL;
+        dht_local_t *local         = NULL;
+        xlator_t    *dst_hashed    = NULL;
+        xlator_t    *src_cached    = NULL;
+        xlator_t    *dst_cached    = NULL;
+        xlator_t    *this          = NULL;
+        xlator_t    *rename_subvol = NULL;
+        dict_t      *dict          = NULL;
 
 
-	local = frame->local;
-	this  = frame->this;
+        local = frame->local;
+        this  = frame->this;
 
-	dst_hashed = local->dst_hashed;
-	dst_cached = local->dst_cached;
-	src_cached = local->src_cached;
+        dst_hashed = local->dst_hashed;
+        dst_cached = local->dst_cached;
+        src_cached = local->src_cached;
 
-	if (src_cached == dst_cached)
-		rename_subvol = src_cached;
-	else
-		rename_subvol = dst_hashed;
+        if (src_cached == dst_cached)
+                rename_subvol = src_cached;
+        else
+                rename_subvol = dst_hashed;
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"renaming %s => %s (%s)",
-		local->loc.path, local->loc2.path, rename_subvol->name);
+        if ((src_cached != dst_hashed) && (rename_subvol == dst_hashed)) {
+                DHT_MARKER_DONT_ACCOUNT(dict);
+        }
 
-	STACK_WIND (frame, dht_rename_cbk,
-		    rename_subvol, rename_subvol->fops->rename,
-		    &local->loc, &local->loc2);
+        gf_log (this->name, GF_LOG_TRACE,
+                "renaming %s => %s (%s)",
+                local->loc.path, local->loc2.path, rename_subvol->name);
 
-	return 0;
+        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, dict);
+
+        return 0;
 }
 
 
 int
 dht_rename_links_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		      int32_t op_ret, int32_t op_errno,
+                      int32_t op_ret, int32_t op_errno,
                       inode_t *inode, struct iatt *stbuf,
-                      struct iatt *preparent, struct iatt *postparent)
+                      struct iatt *preparent, struct iatt *postparent,
+                      dict_t *xdata)
+{
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           this_call_cnt  = 0;
+
+
+        local = frame->local;
+        prev = cookie;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "link/file on %s failed (%s)",
+                        prev->this->name, strerror (op_errno));
+                local->op_ret   = -1;
+                if (op_errno != ENOENT)
+                        local->op_errno = op_errno;
+        } else if (local->src_cached == prev->this) {
+                /* merge of attr returned only from linkfile creation */
+                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+        }
+
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                if (local->op_ret == -1)
+                        goto cleanup;
+
+                dht_do_rename (frame);
+        }
+
+        return 0;
+
+cleanup:
+        dht_rename_cleanup (frame);
+
+        return 0;
+}
+
+
+int
+dht_rename_unlink_links_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                             int32_t op_ret, int32_t op_errno,
+                             struct iatt *preparent, struct iatt *postparent,
+                             dict_t *xdata)
 {
 	dht_local_t  *local = NULL;
 	call_frame_t *prev = NULL;
-	int           this_call_cnt  = 0;
 
 
 	local = frame->local;
 	prev = cookie;
-	
-	if (op_ret == -1) {
+
+	if ((op_ret == -1) && (op_errno != ENOENT)) {
 		gf_log (this->name, GF_LOG_DEBUG,
-			"link/file on %s failed (%s)",
-			prev->this->name, strerror (op_errno));
+			"unlink of %s on %s failed (%s)",
+			local->loc2.path, prev->this->name,
+                        strerror (op_errno));
 		local->op_ret   = -1;
 		local->op_errno = op_errno;
 	}
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt)) {
-		if (local->op_ret == -1)
-			goto cleanup;
-		
-		dht_do_rename (frame);
-	}
+        if (local->op_ret == -1)
+                goto cleanup;
+
+        dht_do_rename (frame);
 
 	return 0;
 
@@ -558,156 +827,183 @@ cleanup:
 int
 dht_rename_create_links (call_frame_t *frame)
 {
-	dht_local_t *local = NULL;
-	xlator_t    *this = NULL;
-	xlator_t    *src_hashed = NULL;
-	xlator_t    *src_cached = NULL;
-	xlator_t    *dst_hashed = NULL;
-	xlator_t    *dst_cached = NULL;
-	int          call_cnt = 0;
+        dht_local_t *local = NULL;
+        xlator_t    *this = NULL;
+        xlator_t    *src_hashed = NULL;
+        xlator_t    *src_cached = NULL;
+        xlator_t    *dst_hashed = NULL;
+        xlator_t    *dst_cached = NULL;
+        int          call_cnt = 0;
+        dict_t      *xattr = NULL;
 
 
-	local = frame->local;
-	this  = frame->this;
+        local = frame->local;
+        this  = frame->this;
 
-	src_hashed = local->src_hashed;
-	src_cached = local->src_cached;
-	dst_hashed = local->dst_hashed;
-	dst_cached = local->dst_cached;
+        src_hashed = local->src_hashed;
+        src_cached = local->src_cached;
+        dst_hashed = local->dst_hashed;
+        dst_cached = local->dst_cached;
 
-	if (src_cached == dst_cached)
-		goto nolinks;
+        DHT_MARK_FOP_INTERNAL (xattr);
 
-	if (dst_hashed != src_hashed && dst_hashed != src_cached)
-		call_cnt++;
+        if (src_cached == dst_cached) {
+                dict_t *xattr_new = NULL;
 
-	if (src_cached != dst_hashed)
-		call_cnt++;
+                if (dst_hashed == dst_cached)
+                        goto nolinks;
 
-	local->call_cnt = call_cnt;
+                xattr_new = dict_copy_with_ref (xattr, NULL);
 
-	if (dst_hashed != src_hashed && dst_hashed != src_cached) {
 		gf_log (this->name, GF_LOG_TRACE,
-			"linkfile %s @ %s => %s",
-			local->loc.path, dst_hashed->name, src_cached->name);
+			"unlinking dst linkfile %s @ %s",
+			local->loc2.path, dst_hashed->name);
+
+                DHT_MARKER_DONT_ACCOUNT(xattr_new);
+
+		STACK_WIND (frame, dht_rename_unlink_links_cbk,
+			    dst_hashed, dst_hashed->fops->unlink,
+			    &local->loc2, 0, xattr_new);
+
+                dict_unref (xattr_new);
+                return 0;
+        }
+
+        if (dst_hashed != src_hashed && dst_hashed != src_cached)
+                call_cnt++;
+
+        if (src_cached != dst_hashed)
+                call_cnt++;
+
+        local->call_cnt = call_cnt;
+
+        if (dst_hashed != src_hashed && dst_hashed != src_cached) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "linkfile %s @ %s => %s",
+                        local->loc.path, dst_hashed->name, src_cached->name);
                 memcpy (local->gfid, local->loc.inode->gfid, 16);
-		dht_linkfile_create (frame, dht_rename_links_cbk,
+		dht_linkfile_create (frame, dht_rename_links_cbk, this,
 				     src_cached, dst_hashed, &local->loc);
 	}
 
 	if (src_cached != dst_hashed) {
+                dict_t *xattr_new = NULL;
+
+                xattr_new = dict_copy_with_ref (xattr, NULL);
+
 		gf_log (this->name, GF_LOG_TRACE,
 			"link %s => %s (%s)", local->loc.path,
 			local->loc2.path, src_cached->name);
+                if (uuid_compare (local->loc.pargfid,
+                                  local->loc2.pargfid) == 0) {
+                        DHT_MARKER_DONT_ACCOUNT(xattr_new);
+                }
+
 		STACK_WIND (frame, dht_rename_links_cbk,
 			    src_cached, src_cached->fops->link,
-			    &local->loc, &local->loc2);
+			    &local->loc, &local->loc2, xattr_new);
+
+                dict_unref (xattr_new);
 	}
 
 nolinks:
-	if (!call_cnt) {
-		/* skip to next step */
-		dht_do_rename (frame);
-	}
+        if (!call_cnt) {
+                /* skip to next step */
+                dht_do_rename (frame);
+        }
+        if (xattr)
+                dict_unref (xattr);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_rename (call_frame_t *frame, xlator_t *this,
-	    loc_t *oldloc, loc_t *newloc)
+            loc_t *oldloc, loc_t *newloc, dict_t *xdata)
 {
-	xlator_t    *src_cached = NULL;
-	xlator_t    *src_hashed = NULL;
-	xlator_t    *dst_cached = NULL;
-	xlator_t    *dst_hashed = NULL;
-	int          op_errno = -1;
-	int          ret = -1;
-	dht_local_t *local = NULL;
-
-
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (oldloc, err);
-	VALIDATE_OR_GOTO (newloc, err);
-
-	src_hashed = dht_subvol_get_hashed (this, oldloc);
-	if (!src_hashed) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			oldloc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	src_cached = dht_subvol_get_cached (this, oldloc->inode);
-	if (!src_cached) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", oldloc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	dst_hashed = dht_subvol_get_hashed (this, newloc);
-	if (!dst_hashed) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			newloc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        xlator_t    *src_cached = NULL;
+        xlator_t    *src_hashed = NULL;
+        xlator_t    *dst_cached = NULL;
+        xlator_t    *dst_hashed = NULL;
+        int          op_errno = -1;
+        int          ret = -1;
+        dht_local_t *local = NULL;
+
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (oldloc, err);
+        VALIDATE_OR_GOTO (newloc, err);
+
+        src_hashed = dht_subvol_get_hashed (this, oldloc);
+        if (!src_hashed) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "no subvolume in layout for path=%s",
+                        oldloc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	if (newloc->inode)
-		dst_cached = dht_subvol_get_cached (this, newloc->inode);
+        src_cached = dht_subvol_get_cached (this, oldloc->inode);
+        if (!src_cached) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "no cached subvolume for path=%s", oldloc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        dst_hashed = dht_subvol_get_hashed (this, newloc);
+        if (!dst_hashed) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "no subvolume in layout for path=%s",
+                        newloc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	ret = loc_copy (&local->loc, oldloc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        if (newloc->inode)
+                dst_cached = dht_subvol_get_cached (this, newloc->inode);
 
-	ret = loc_copy (&local->loc2, newloc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        local = dht_local_init (frame, oldloc, NULL, GF_FOP_RENAME);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+        /* cached_subvol will be set from dht_local_init, reset it to NULL,
+           as the logic of handling rename is different  */
+        local->cached_subvol = NULL;
+
+        ret = loc_copy (&local->loc2, newloc);
+        if (ret == -1) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	local->src_hashed = src_hashed;
-	local->src_cached = src_cached;
-	local->dst_hashed = dst_hashed;
-	local->dst_cached = dst_cached;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"renaming %s (hash=%s/cache=%s) => %s (hash=%s/cache=%s)",
-		oldloc->path, src_hashed->name, src_cached->name,
-		newloc->path, dst_hashed->name,
-		dst_cached ? dst_cached->name : "<nul>");
-
-	if (IA_ISDIR (oldloc->inode->ia_type)) {
-		dht_rename_dir (frame, this);
-	} else {
-		local->op_ret = 0;
-		dht_rename_create_links (frame);
-	}
+        local->src_hashed = src_hashed;
+        local->src_cached = src_cached;
+        local->dst_hashed = dst_hashed;
+        local->dst_cached = dst_cached;
+
+        gf_log (this->name, GF_LOG_TRACE,
+                "renaming %s (hash=%s/cache=%s) => %s (hash=%s/cache=%s)",
+                oldloc->path, src_hashed->name, src_cached->name,
+                newloc->path, dst_hashed->name,
+                dst_cached ? dst_cached->name : "<nul>");
+
+        if (IA_ISDIR (oldloc->inode->ia_type)) {
+                dht_rename_dir (frame, this);
+        } else {
+                local->op_ret = 0;
+                dht_rename_create_links (frame);
+        }
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+                          NULL, NULL);
 
-	return 0;
+        return 0;
 }
diff --git a/xlators/cluster/dht/src/dht-selfheal.c b/xlators/cluster/dht/src/dht-selfheal.c
index 74c89a4fd..0e6527544 100644
--- a/xlators/cluster/dht/src/dht-selfheal.c
+++ b/xlators/cluster/dht/src/dht-selfheal.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -26,247 +17,489 @@
 #include "glusterfs.h"
 #include "xlator.h"
 #include "dht-common.h"
+#include "glusterfs-acl.h"
+
+#define DHT_SET_LAYOUT_RANGE(layout,i,srt,chunk,cnt,path)    do {       \
+                layout->list[i].start = srt;                            \
+                layout->list[i].stop  = srt + chunk - 1;                \
+                                                                        \
+                gf_log (this->name, GF_LOG_TRACE,                       \
+                        "gave fix: %u - %u on %s for %s",               \
+                        layout->list[i].start, layout->list[i].stop,    \
+                        layout->list[i].xlator->name, path);            \
+        } while (0)
+
+#define DHT_RESET_LAYOUT_RANGE(layout)    do {                          \
+                int cnt = 0;                                            \
+                for (cnt = 0; cnt < layout->cnt; cnt++ ) {              \
+                        layout->list[cnt].start = 0;                    \
+                        layout->list[cnt].stop  = 0;                    \
+                }                                                       \
+        } while (0)
+
+static uint32_t
+dht_overlap_calc (dht_layout_t *old, int o, dht_layout_t *new, int n)
+{
+	if (o >= old->cnt || n >= new->cnt)
+		return 0;
+
+	if (old->list[o].err > 0 || new->list[n].err > 0)
+		return 0;
+
+        if (old->list[o].start == old->list[o].stop) {
+                return 0;
+        }
+
+        if (new->list[n].start == new->list[n].stop) {
+                return 0;
+        }
+
+	if ((old->list[o].start > new->list[n].stop) ||
+	    (old->list[o].stop < new->list[n].start))
+		return 0;
+
+	return min (old->list[o].stop, new->list[n].stop) -
+	        max (old->list[o].start, new->list[n].start) + 1;
+}
 
 
 int
 dht_selfheal_dir_finish (call_frame_t *frame, xlator_t *this, int ret)
 {
-	dht_local_t  *local = NULL;
-
+        dht_local_t  *local = NULL;
 
-	local = frame->local;
-	local->selfheal.dir_cbk (frame, NULL, frame->this, ret,
-				 local->op_errno);
+        local = frame->local;
+        local->selfheal.dir_cbk (frame, NULL, frame->this, ret,
+                                 local->op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_selfheal_dir_xattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			    int op_ret, int op_errno)
+                            int op_ret, int op_errno, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	call_frame_t *prev = NULL;
-	xlator_t     *subvol = NULL;
-	int           i = 0;
-	dht_layout_t *layout = NULL;
-	int           err = 0;
-	int           this_call_cnt = 0;
-
-	local = frame->local;
-	layout = local->selfheal.layout;
-	prev = cookie;
-	subvol = prev->this;
-
-	if (op_ret == 0)
-		err = 0;
-	else
-		err = op_errno;
-
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].xlator == subvol) {
-			layout->list[i].err = err;
-			break;
-		}
-	}
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        xlator_t     *subvol = NULL;
+        int           i = 0;
+        dht_layout_t *layout = NULL;
+        int           err = 0;
+        int           this_call_cnt = 0;
+
+        local = frame->local;
+        layout = local->selfheal.layout;
+        prev = cookie;
+        subvol = prev->this;
+
+        if (op_ret == 0)
+                err = 0;
+        else
+                err = op_errno;
+
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].xlator == subvol) {
+                        layout->list[i].err = err;
+                        break;
+                }
+        }
 
-	this_call_cnt = dht_frame_return (frame);
+        this_call_cnt = dht_frame_return (frame);
 
-	if (is_last_call (this_call_cnt)) {
-		dht_selfheal_dir_finish (frame, this, 0);
-	}
+        if (is_last_call (this_call_cnt)) {
+                dht_selfheal_dir_finish (frame, this, 0);
+        }
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_selfheal_dir_xattr_persubvol (call_frame_t *frame, loc_t *loc,
-				  dht_layout_t *layout, int i)
+                                  dht_layout_t *layout, int i,
+                                  xlator_t *req_subvol)
 {
-	xlator_t          *subvol = NULL;
-	dict_t            *xattr = NULL;
-	int                ret = 0;
-	xlator_t          *this = NULL;
-	int32_t           *disk_layout = NULL;
+        xlator_t          *subvol = NULL;
+        dict_t            *xattr = NULL;
+        int                ret = 0;
+        xlator_t          *this = NULL;
+        int32_t           *disk_layout = NULL;
+        dht_local_t       *local = NULL;
+        dht_conf_t        *conf = NULL;
+        data_t            *data = NULL;
+
+        local = frame->local;
+        if (req_subvol)
+                subvol = req_subvol;
+        else
+                subvol = layout->list[i].xlator;
+        this = frame->this;
+
+        GF_VALIDATE_OR_GOTO ("", this, err);
+        GF_VALIDATE_OR_GOTO (this->name, layout, err);
+        GF_VALIDATE_OR_GOTO (this->name, local, err);
+        GF_VALIDATE_OR_GOTO (this->name, subvol, err);
+        VALIDATE_OR_GOTO (this->private, err);
+
+        conf = this->private;
+
+        xattr = get_new_dict ();
+        if (!xattr) {
+                goto err;
+        }
 
+        ret = dht_disk_layout_extract (this, layout, i, &disk_layout);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: (subvol %s) failed to extract disk layout",
+                        loc->path, subvol->name);
+                goto err;
+        }
 
-	subvol = layout->list[i].xlator;
-	this = frame->this;
+        ret = dict_set_bin (xattr, conf->xattr_name, disk_layout, 4 * 4);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: (subvol %s) failed to set xattr dictionary",
+                        loc->path, subvol->name);
+                goto err;
+        }
+        disk_layout = NULL;
+
+        gf_log (this->name, GF_LOG_TRACE,
+                "setting hash range %u - %u (type %d) on subvolume %s for %s",
+                layout->list[i].start, layout->list[i].stop,
+                layout->type, subvol->name, loc->path);
+
+        dict_ref (xattr);
+        if (local->xattr) {
+                data = dict_get (local->xattr, QUOTA_LIMIT_KEY);
+                if (data) {
+                        ret = dict_add (xattr, QUOTA_LIMIT_KEY, data);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "Failed to "
+                                        "set quota limit key on %s",loc->path);
+                        }
+                }
+        }
+        if (!uuid_is_null (local->gfid))
+                uuid_copy (loc->gfid, local->gfid);
 
-	xattr = get_new_dict ();
-	if (!xattr) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        STACK_WIND (frame, dht_selfheal_dir_xattr_cbk,
+                    subvol, subvol->fops->setxattr,
+                    loc, xattr, 0, NULL);
 
-	ret = dht_disk_layout_extract (this, layout, i, &disk_layout);
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"failed to extract disk layout");
-		goto err;
-	}
+        dict_unref (xattr);
 
-	ret = dict_set_bin (xattr, "trusted.glusterfs.dht",
-			    disk_layout, 4 * 4);
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"failed to set xattr dictionary");
-		goto err;
-	}
-	disk_layout = NULL;
+        return 0;
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"setting hash range %u - %u (type %d) on subvolume %s for %s",
-		layout->list[i].start, layout->list[i].stop,
-		layout->type, subvol->name, loc->path);
+err:
+        if (xattr)
+                dict_destroy (xattr);
 
-	dict_ref (xattr);
+        GF_FREE (disk_layout);
 
-	STACK_WIND (frame, dht_selfheal_dir_xattr_cbk,
-		    subvol, subvol->fops->setxattr,
-		    loc, xattr, 0);
+        dht_selfheal_dir_xattr_cbk (frame, subvol, frame->this,
+                                    -1, ENOMEM, NULL);
+        return 0;
+}
 
-	dict_unref (xattr);
+int
+dht_fix_dir_xattr (call_frame_t *frame, loc_t *loc, dht_layout_t *layout)
+{
+        dht_local_t *local = NULL;
+        int          i = 0;
+        int          count = 0;
+        xlator_t    *this = NULL;
+        dht_conf_t  *conf = NULL;
+        dht_layout_t *dummy = NULL;
 
-	return 0;
+        local = frame->local;
+        this = frame->this;
+        conf = this->private;
 
-err:
-	if (xattr)
-		dict_destroy (xattr);
+        gf_log (this->name, GF_LOG_DEBUG,
+                "writing the new range for all subvolumes");
 
-	if (disk_layout)
-		GF_FREE (disk_layout);
+        local->call_cnt = count = conf->subvolume_cnt;
 
-	dht_selfheal_dir_xattr_cbk (frame, subvol, frame->this,
-				    -1, ENOMEM);
-	return 0;
-}
+        for (i = 0; i < layout->cnt; i++) {
+                dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i, NULL);
+
+                if (--count == 0)
+                        goto out;
+        }
+        /* if we are here, subvolcount > layout_count. subvols-per-directory
+         * option might be set here. We need to clear out layout from the
+         * non-participating subvolumes, else it will result in overlaps */
+        dummy = dht_layout_new (this, 1);
+        if (!dummy)
+                goto out;
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (_gf_false ==
+                    dht_is_subvol_in_layout (layout, conf->subvolumes[i])) {
+                        dht_selfheal_dir_xattr_persubvol (frame, loc, dummy, 0,
+                                                          conf->subvolumes[i]);
+                        if (--count == 0)
+                                break;
+                }
+        }
 
+        dht_layout_unref (this, dummy);
+out:
+        return 0;
+}
 
 int
 dht_selfheal_dir_xattr (call_frame_t *frame, loc_t *loc, dht_layout_t *layout)
 {
-	dht_local_t *local = NULL;
-	int          missing_xattr = 0;
-	int          i = 0;
-	xlator_t    *this = NULL;
-
-	local = frame->local;
-	this = frame->this;
-
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].err != -1 || !layout->list[i].stop) {
-			/* err != -1 would mean xattr present on the directory
-			 * or the directory is itself non existant.
-			 * !layout->list[i].stop would mean layout absent
-			 */
-			continue;
-		}
-		missing_xattr++;
-	}
+        dht_local_t *local = NULL;
+        int          missing_xattr = 0;
+        int          i = 0;
+        xlator_t    *this = NULL;
+        dht_conf_t   *conf = NULL;
+        dht_layout_t *dummy = NULL;
+
+        local = frame->local;
+        this = frame->this;
+        conf = this->private;
+
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err != -1 || !layout->list[i].stop) {
+                        /* err != -1 would mean xattr present on the directory
+                         * or the directory is non existent.
+                         * !layout->list[i].stop would mean layout absent
+                         */
+
+                        continue;
+                }
+                missing_xattr++;
+        }
+        /* Also account for subvolumes with no-layout. Used for zero'ing out
+         * the layouts and for setting quota key's if present */
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (_gf_false ==
+                    dht_is_subvol_in_layout (layout, conf->subvolumes[i])) {
+                        missing_xattr++;
+                }
+        }
+        gf_log (this->name, GF_LOG_TRACE,
+                "%d subvolumes missing xattr for %s",
+                missing_xattr, loc->path);
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"%d subvolumes missing xattr for %s",
-		missing_xattr, loc->path);
+        if (missing_xattr == 0) {
+                dht_selfheal_dir_finish (frame, this, 0);
+                return 0;
+        }
 
-	if (missing_xattr == 0) {
-		dht_selfheal_dir_finish (frame, this, 0);
-		return 0;
-	}
+        local->call_cnt = missing_xattr;
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err != -1 || !layout->list[i].stop)
+                        continue;
 
-	local->call_cnt = missing_xattr;
+                dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i, NULL);
 
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].err != -1 || !layout->list[i].stop)
-			continue;
+                if (--missing_xattr == 0)
+                        break;
+        }
+        dummy = dht_layout_new (this, 1);
+        if (!dummy)
+                goto out;
+        for (i = 0; i < conf->subvolume_cnt && missing_xattr; 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]);
+                        missing_xattr--;
+                }
+        }
+
+        dht_layout_unref (this, dummy);
+out:
+        return 0;
+}
 
-		dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i);
+int
+dht_selfheal_dir_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                              int op_ret, int op_errno, struct iatt *statpre,
+                              struct iatt *statpost, dict_t *xdata)
+{
+        dht_local_t   *local = NULL;
+        dht_layout_t  *layout = NULL;
+        int            this_call_cnt = 0;
 
-		if (--missing_xattr == 0)
-			break;
-	}
-	return 0;
+        local  = frame->local;
+        layout = local->selfheal.layout;
+
+        this_call_cnt = dht_frame_return (frame);
+
+        if (is_last_call (this_call_cnt)) {
+                dht_selfheal_dir_xattr (frame, &local->loc, layout);
+        }
+
+        return 0;
 }
 
 
 int
+dht_selfheal_dir_setattr (call_frame_t *frame, loc_t *loc, struct iatt *stbuf,
+                          int32_t valid, dht_layout_t *layout)
+{
+        int           missing_attr = 0;
+        int           i     = 0;
+        dht_local_t  *local = NULL;
+        xlator_t     *this = NULL;
+
+        local = frame->local;
+        this = frame->this;
+
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err == -1)
+                        missing_attr++;
+        }
+
+        if (missing_attr == 0) {
+                dht_selfheal_dir_xattr (frame, loc, layout);
+                return 0;
+        }
+
+        if (!uuid_is_null (local->gfid))
+                uuid_copy (loc->gfid, local->gfid);
+
+        local->call_cnt = missing_attr;
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err == -1) {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "setattr for %s on subvol %s",
+                                loc->path, layout->list[i].xlator->name);
+
+                        STACK_WIND (frame, dht_selfheal_dir_setattr_cbk,
+                                    layout->list[i].xlator,
+                                    layout->list[i].xlator->fops->setattr,
+                                    loc, stbuf, valid, NULL);
+                }
+        }
+
+        return 0;
+}
+
+int
 dht_selfheal_dir_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			    int op_ret, int op_errno,
+                            int op_ret, int op_errno,
                             inode_t *inode, struct iatt *stbuf,
-                            struct iatt *preparent, struct iatt *postparent)
+                            struct iatt *preparent, struct iatt *postparent,
+                            dict_t *xdata)
 {
-	dht_local_t   *local = NULL;
-	dht_layout_t  *layout = NULL;
-	call_frame_t  *prev = NULL;
-	xlator_t      *subvol = NULL;
-	int            i = 0;
-	int            this_call_cnt = 0;
-
-
-	local  = frame->local;
-	layout = local->selfheal.layout;
-	prev   = cookie;
-	subvol = prev->this;
-
-	if ((op_ret == 0) || ((op_ret == -1) && (op_errno == EEXIST))) {
-		for (i = 0; i < layout->cnt; i++) {
-			if (layout->list[i].xlator == subvol) {
-				layout->list[i].err = -1;
-				break;
-			}
-		}
-	}
+        dht_local_t   *local = NULL;
+        dht_layout_t  *layout = NULL;
+        call_frame_t  *prev = NULL;
+        xlator_t      *subvol = NULL;
+        int            i = 0;
+        int            this_call_cnt = 0;
 
-        if (op_ret)
-                goto out;
 
+        local  = frame->local;
+        layout = local->selfheal.layout;
+        prev   = cookie;
+        subvol = prev->this;
 
-        dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
-        if (prev->this == local->hashed_subvol)
-                local->ia_ino = local->stbuf.ia_ino;
+        if ((op_ret == 0) || ((op_ret == -1) && (op_errno == EEXIST))) {
+                for (i = 0; i < layout->cnt; i++) {
+                        if (layout->list[i].xlator == subvol) {
+                                layout->list[i].err = -1;
+                                break;
+                        }
+                }
+        }
+
+        if (op_ret) {
+                gf_log (this->name, ((op_errno == EEXIST) ? GF_LOG_DEBUG :
+                                     GF_LOG_WARNING),
+                        "selfhealing directory %s failed: %s",
+                        local->loc.path, strerror (op_errno));
+                goto out;
+        }
 
+        dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
         dht_iatt_merge (this, &local->preparent, preparent, prev->this);
         dht_iatt_merge (this, &local->postparent, postparent, prev->this);
 
 out:
-	this_call_cnt = dht_frame_return (frame);
+        this_call_cnt = dht_frame_return (frame);
 
-	if (is_last_call (this_call_cnt)) {
-		dht_selfheal_dir_xattr (frame, &local->loc, layout);
-	}
+        if (is_last_call (this_call_cnt)) {
+                dht_selfheal_dir_setattr (frame, &local->loc, &local->stbuf, 0xffffff, layout);
+        }
 
-	return 0;
+        return 0;
 }
 
+void
+dht_selfheal_dir_mkdir_setacl (dict_t *xattr, dict_t *dict)
+{
+        data_t          *acl_default = NULL;
+        data_t          *acl_access = NULL;
+        xlator_t        *this = NULL;
+        int     ret = -1;
+
+        GF_ASSERT (xattr);
+        GF_ASSERT (dict);
+
+        this = THIS;
+        GF_ASSERT (this);
+
+        acl_default = dict_get (xattr, POSIX_ACL_DEFAULT_XATTR);
+
+        if (!acl_default) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "ACL_DEFAULT xattr not present");
+                goto cont;
+        }
+        ret = dict_set (dict, POSIX_ACL_DEFAULT_XATTR, acl_default);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Could not set ACL_DEFAULT xattr");
+cont:
+        acl_access = dict_get (xattr, POSIX_ACL_ACCESS_XATTR);
+        if (!acl_access) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "ACL_ACCESS xattr not present");
+                goto out;
+        }
+        ret = dict_set (dict, POSIX_ACL_ACCESS_XATTR, acl_access);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Could not set ACL_ACCESS xattr");
+
+out:
+        return;
+}
 
 int
 dht_selfheal_dir_mkdir (call_frame_t *frame, loc_t *loc,
-			dht_layout_t *layout, int force)
+                        dht_layout_t *layout, int force)
 {
-	int           missing_dirs = 0;
-	int           i     = 0;
-	int           ret   = -1;
-	dht_local_t  *local = NULL;
-	xlator_t     *this = NULL;
+        int           missing_dirs = 0;
+        int           i     = 0;
+        int           ret   = -1;
+        dht_local_t  *local = NULL;
+        xlator_t     *this = NULL;
         dict_t       *dict = NULL;
 
-	local = frame->local;
-	this = frame->this;
+        local = frame->local;
+        this = frame->this;
 
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].err == ENOENT || force)
-			missing_dirs++;
-	}
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err == ENOENT || force)
+                        missing_dirs++;
+        }
 
-	if (missing_dirs == 0) {
-		dht_selfheal_dir_xattr (frame, loc, layout);
-		return 0;
-	}
+        if (missing_dirs == 0) {
+                dht_selfheal_dir_setattr (frame, loc, &local->stbuf, 0xffffffff, layout);
+                return 0;
+        }
 
-	local->call_cnt = missing_dirs;
+        local->call_cnt = missing_dirs;
         if (!uuid_is_null (local->gfid)) {
                 dict = dict_new ();
                 if (!dict)
@@ -274,37 +507,40 @@ dht_selfheal_dir_mkdir (call_frame_t *frame, loc_t *loc,
 
                 ret = dict_set_static_bin (dict, "gfid-req", local->gfid, 16);
                 if (ret)
-                        gf_log (this->name, GF_LOG_INFO,
-                                "failed to set gfid in dict");
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to set gfid in dict", loc->path);
         } else if (local->params) {
                 /* Send the dictionary from higher layers directly */
                 dict = dict_ref (local->params);
         }
+        /* Set acls */
+        if (local->xattr && dict)
+                dht_selfheal_dir_mkdir_setacl (local->xattr, dict);
 
         if (!dict)
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "dict is NULL, need to make sure gfid's are same");
-
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].err == ENOENT || force) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"creating directory %s on subvol %s",
-				loc->path, layout->list[i].xlator->name);
-
-			STACK_WIND (frame, dht_selfheal_dir_mkdir_cbk,
-				    layout->list[i].xlator,
-				    layout->list[i].xlator->fops->mkdir,
-				    loc,
+                gf_log (this->name, GF_LOG_WARNING,
+                        "dict is NULL, need to make sure gfids are same");
+
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err == ENOENT || force) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "creating directory %s on subvol %s",
+                                loc->path, layout->list[i].xlator->name);
+
+                        STACK_WIND (frame, dht_selfheal_dir_mkdir_cbk,
+                                    layout->list[i].xlator,
+                                    layout->list[i].xlator->fops->mkdir,
+                                    loc,
                                     st_mode_from_ia (local->stbuf.ia_prot,
                                                      local->stbuf.ia_type),
-                                    dict);
-		}
-	}
+                                    0, dict);
+                }
+        }
 
         if (dict)
                 dict_unref (dict);
 
-	return 0;
+        return 0;
 }
 
 
@@ -316,7 +552,7 @@ 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);
         }
@@ -324,235 +560,489 @@ dht_selfheal_layout_alloc_start (xlator_t *this, loc_t *loc,
         return start;
 }
 
-
-void
-dht_selfheal_layout_new_directory (call_frame_t *frame, loc_t *loc,
-				   dht_layout_t *layout)
+static inline int
+dht_get_layout_count (xlator_t *this, dht_layout_t *layout, int new_layout)
 {
-	xlator_t    *this = NULL;
-	uint32_t     chunk = 0;
-	int          i = 0;
-	uint32_t     start = 0;
-	int          cnt = 0;
-	int          err = 0;
-        int          start_subvol = 0;
-
-	this = frame->this;
+        int i = 0;
+        int j = 0;
+        int err = 0;
+        int count = 0;
+        dht_conf_t *conf = NULL;
+
+        /* Gets in use only for replace-brick, remove-brick */
+        conf = this->private;
+        for (i = 0; i < layout->cnt; i++) {
+                for (j = 0; j < conf->subvolume_cnt; j++) {
+                        if (conf->decommissioned_bricks[j] &&
+                            conf->decommissioned_bricks[j] == layout->list[i].xlator) {
+                                layout->list[i].err = EINVAL;
+                                break;
+                        }
+                }
+        }
 
-	for (i = 0; i < layout->cnt; i++) {
-		err = layout->list[i].err;
-		if (err == -1 || err == 0) {
-			layout->list[i].err = -1;
-			cnt++;
-		}
-	}
+        for (i = 0; i < layout->cnt; i++) {
+                err = layout->list[i].err;
+                if (err == -1 || err == 0 || err == ENOENT) {
+			/* Setting list[i].err = -1 is an indication for
+			   dht_selfheal_layout_new_directory() to assign
+			   a range. We set it to -1 based on any one of
+			   the three criteria:
+
+			   - err == -1 already, which means directory
+			     existed but layout was not set on it.
+
+			   - err == 0, which means directory exists and
+			     has an old layout piece which will be
+			     overwritten now.
+
+			   - err == ENOENT, which means directory does
+			     not exist (possibly racing with mkdir or
+			     finishing half done mkdir). The missing
+			     directory will be attempted to be recreated.
+
+			     It is important to note that it is safe
+			     to race with mkdir() as self-heal and
+			     mkdir are idempotent operations. Both will
+			     strive to set the directory and layouts to
+			     the same final state.
+			*/
+                        count++;
+			if (!err)
+				layout->list[i].err = -1;
+                }
+        }
 
         /* no subvolume has enough space, but can't stop directory creation */
-        if (!cnt) {
+        if (!count || !new_layout) {
                 for (i = 0; i < layout->cnt; i++) {
                         err = layout->list[i].err;
                         if (err == ENOSPC) {
                                 layout->list[i].err = -1;
-                                cnt++;
+                                count++;
                         }
                 }
         }
 
-	chunk = ((unsigned long) 0xffffffff) / ((cnt) ? cnt : 1);
+        /* if layout->spread_cnt is set, check if it is <= available
+         * subvolumes (down brick and decommissioned bricks are considered
+         * un-availbale). Else return count (available up bricks) */
+        count = ((layout->spread_cnt &&
+                 (layout->spread_cnt <= count)) ?
+                 layout->spread_cnt : ((count) ? count : 1));
 
-	start_subvol = dht_selfheal_layout_alloc_start (this, loc, layout);
+        return count;
+}
 
-	for (i = start_subvol; i < layout->cnt; i++) {
-		err = layout->list[i].err;
-		if (err == -1) {
-			layout->list[i].start = start;
-			layout->list[i].stop  = start + chunk - 1;
-			
-			start = start + chunk;
 
-			gf_log (this->name, GF_LOG_TRACE,
-				"gave fix: %u - %u on %s for %s",
-				layout->list[i].start, layout->list[i].stop,
-				layout->list[i].xlator->name, loc->path);
-			if (--cnt == 0) {
-				layout->list[i].stop = 0xffffffff;
-				break;
-			}
-		}
-	}
+void dht_selfheal_layout_new_directory (call_frame_t *frame, loc_t *loc,
+					dht_layout_t *new_layout);
+
+void dht_layout_entry_swap (dht_layout_t *layout, int i, int j);
+void dht_layout_range_swap (dht_layout_t *layout, int i, int j);
+
+/*
+ * It's a bit icky using local variables in a macro, but it makes the rest
+ * of the code a lot clearer.
+ */
+#define OV_ENTRY(x,y)      table[x*new->cnt+y]
+
+void
+dht_selfheal_layout_maximize_overlap (call_frame_t *frame, loc_t *loc,
+				      dht_layout_t *new, dht_layout_t *old)
+{
+	int           i            = 0;
+	int           j            = 0;
+	uint32_t      curr_overlap = 0;
+	uint32_t      max_overlap  = 0;
+	int           max_overlap_idx = -1;
+	uint32_t      overlap      = 0;
+        uint32_t     *table = NULL;
+
+	dht_layout_sort_volname (old);
+	/* Now both old_layout->list[] and new_layout->list[]
+	   are match the same xlators/subvolumes. i.e,
+	   old_layout->[i] and new_layout->[i] are referring
+	   to the same subvolumes
+	*/
+
+        /* Build a table of overlaps between new[i] and old[j]. */
+        table = alloca(sizeof(overlap)*old->cnt*new->cnt);
+        if (!table) {
+                return;
+        }
+        memset(table,0,sizeof(overlap)*old->cnt*new->cnt);
+        for (i = 0; i < new->cnt; ++i) {
+                for (j = 0; j < old->cnt; ++j) {
+                        OV_ENTRY(i,j) = dht_overlap_calc(old,j,new,i);
+                }
+        }
+
+	for (i = 0; i < new->cnt; i++) {
+		if (new->list[i].err > 0) {
+			/* Subvol might be marked for decommission
+			   with EINVAL, or some other serious error
+			   marked with positive errno.
+			*/
+			continue;
+                }
+
+                max_overlap = 0;
+                max_overlap_idx = i;
+                for (j = (i + 1); j < new->cnt; ++j) {
+                        if (new->list[j].err > 0) {
+			        /* Subvol might be marked for decommission
+			        with EINVAL, or some other serious error
+			        marked with positive errno.
+			        */
+			        continue;
+                        }
+                        /* Calculate the overlap now. */
+                        curr_overlap = OV_ENTRY(i,i) + OV_ENTRY(j,j);
+                        /* Calculate the overlap after the proposed swap. */
+                        overlap = OV_ENTRY(i,j) + OV_ENTRY(j,i);
+                        /* Are we better than status quo? */
+                        if (overlap > curr_overlap) {
+                                overlap -= curr_overlap;
+                                /* Are we better than the previous choice? */
+                                if (overlap > max_overlap) {
+                                        max_overlap = overlap;
+                                        max_overlap_idx = j;
+                                }
+                        }
+                }
 
-	for (i = 0; i < start_subvol; i++) {
-		err = layout->list[i].err;
-		if (err == -1) {
-			layout->list[i].start = start;
-			layout->list[i].stop  = start + chunk - 1;
-			
-			start = start + chunk;
-
-			gf_log (this->name, GF_LOG_TRACE,
-				"gave fix: %u - %u on %s for %s",
-				layout->list[i].start, layout->list[i].stop,
-				layout->list[i].xlator->name, loc->path);
-			if (--cnt == 0) {
-				layout->list[i].stop = 0xffffffff;
-				break;
-			}
-		}
+		if (max_overlap_idx != i) {
+ 			dht_layout_range_swap (new, i, max_overlap_idx);
+                        /* Need to swap the table values too. */
+                        for (j = 0; j < old->cnt; ++j) {
+                                overlap = OV_ENTRY(i,j);
+                                OV_ENTRY(i,j) = OV_ENTRY(max_overlap_idx,j);
+                                OV_ENTRY(max_overlap_idx,j) = overlap;
+                        }
+                }
 	}
 }
 
 
+dht_layout_t *
+dht_fix_layout_of_directory (call_frame_t *frame, loc_t *loc,
+                             dht_layout_t *layout)
+{
+        int           i            = 0;
+        xlator_t     *this         = NULL;
+        dht_layout_t *new_layout   = NULL;
+        dht_conf_t   *priv         = NULL;
+        dht_local_t  *local        = NULL;
+        uint32_t      subvol_down  = 0;
+        int           ret          = 0;
+
+        this  = frame->this;
+        priv  = this->private;
+        local = frame->local;
+
+        if (layout->type == DHT_HASH_TYPE_DM_USER) {
+                gf_log (THIS->name, GF_LOG_DEBUG, "leaving %s alone",
+                        loc->path);
+                goto done;
+        }
+
+        new_layout = dht_layout_new (this, priv->subvolume_cnt);
+        if (!new_layout)
+                goto done;
+
+        /* If a subvolume is down, do not re-write the layout. */
+        ret = dht_layout_anomalies (this, loc, layout, NULL, NULL, NULL,
+                                    &subvol_down, NULL, NULL);
+
+        if (subvol_down || (ret == -1)) {
+                gf_log (this->name, GF_LOG_WARNING, "%u subvolume(s) are down"
+                        ". Skipping fix layout.", subvol_down);
+                GF_FREE (new_layout);
+                return NULL;
+        }
+
+        for (i = 0; i < new_layout->cnt; i++) {
+		if (layout->list[i].err != ENOSPC)
+			new_layout->list[i].err = layout->list[i].err;
+		else
+			new_layout->list[i].err = -1;
+
+		new_layout->list[i].xlator = layout->list[i].xlator;
+        }
+
+	/* First give it a layout as though it is a new directory. This
+	   ensures rotation to kick in */
+        dht_layout_sort_volname (new_layout);
+	dht_selfheal_layout_new_directory (frame, loc, new_layout);
+
+	/* Now selectively re-assign ranges only when it helps */
+	dht_selfheal_layout_maximize_overlap (frame, loc, new_layout, layout);
+
+done:
+        if (new_layout) {
+                /* Now that the new layout has all the proper layout, change the
+                   inode context */
+                dht_layout_set (this, loc->inode, new_layout);
+
+                /* Make sure the extra 'ref' for existing layout is removed */
+                dht_layout_unref (this, local->layout);
+
+                local->layout = new_layout;
+        }
+
+        return local->layout;
+}
+
+
+void
+dht_selfheal_layout_new_directory (call_frame_t *frame, loc_t *loc,
+                                   dht_layout_t *layout)
+{
+        xlator_t    *this = NULL;
+        uint32_t     chunk = 0;
+        int          i = 0;
+        uint32_t     start = 0;
+        int          cnt = 0;
+        int          err = 0;
+        int          start_subvol = 0;
+
+        this = frame->this;
+
+        cnt = dht_get_layout_count (this, layout, 1);
+
+        chunk = ((unsigned long) 0xffffffff) / ((cnt) ? cnt : 1);
+
+        start_subvol = dht_selfheal_layout_alloc_start (this, loc, layout);
+
+        /* clear out the range, as we are re-computing here */
+        DHT_RESET_LAYOUT_RANGE (layout);
+        for (i = start_subvol; i < layout->cnt; i++) {
+                err = layout->list[i].err;
+                if (err == -1 || err == ENOENT) {
+                        DHT_SET_LAYOUT_RANGE(layout, i, start, chunk,
+                                             cnt, loc->path);
+                        if (--cnt == 0) {
+                                layout->list[i].stop = 0xffffffff;
+                                goto done;
+                        }
+                        start += chunk;
+                }
+        }
+
+        for (i = 0; i < start_subvol; i++) {
+                err = layout->list[i].err;
+                if (err == -1 || err == ENOENT) {
+                        DHT_SET_LAYOUT_RANGE(layout, i, start, chunk,
+                                             cnt, loc->path);
+                        if (--cnt == 0) {
+                                layout->list[i].stop = 0xffffffff;
+                                goto done;
+                        }
+                        start += chunk;
+                }
+        }
+
+done:
+        return;
+}
+
 int
 dht_selfheal_dir_getafix (call_frame_t *frame, loc_t *loc,
-			  dht_layout_t *layout)
+                          dht_layout_t *layout)
 {
-	dht_conf_t  *conf = NULL;
-	xlator_t    *this = NULL;
-	dht_local_t *local = NULL;
-	int          missing = -1;
-	int          down = -1;
-	int          holes = -1;
-	int          ret = -1;
-	int          i = -1;
-	int          overlaps = -1;
-
-	this = frame->this;
-	conf = this->private;
-	local = frame->local;
-
-	missing = local->selfheal.missing;
-	down = local->selfheal.down;
-	holes = local->selfheal.hole_cnt;
-	overlaps = local->selfheal.overlaps_cnt;
-
-	if ((missing + down) == conf->subvolume_cnt) {
-		dht_selfheal_layout_new_directory (frame, loc, layout);
-		ret = 0;
-	}
+        dht_local_t *local = NULL;
+        uint32_t     holes = 0;
+        int          ret = -1;
+        int          i = -1;
+        uint32_t     overlaps = 0;
 
-	if (holes <= down) {
-		/* the down subvol might fill up the holes */
-		ret = 0;
-	}
+        local = frame->local;
 
-	if (holes || overlaps) {
-		dht_selfheal_layout_new_directory (frame, loc, layout);
-		ret = 0;
-	}
+        holes = local->selfheal.hole_cnt;
+        overlaps = local->selfheal.overlaps_cnt;
 
-	for (i = 0; i < layout->cnt; i++) {
-		/* directory not present */
-		if (layout->list[i].err == ENOENT) {
-			ret = 0;
-			break;
-		}
-	}
+        if (holes || overlaps) {
+                dht_selfheal_layout_new_directory (frame, loc, layout);
+                ret = 0;
+        }
+
+        for (i = 0; i < layout->cnt; i++) {
+                /* directory not present */
+                if (layout->list[i].err == ENOENT) {
+                        ret = 0;
+                        break;
+                }
+        }
 
-	/* TODO: give a fix to these non-virgins */
+        /* TODO: give a fix to these non-virgins */
 
-	return ret;
+        return ret;
 }
 
 int
-dht_selfheal_new_directory (call_frame_t *frame, 
-			    dht_selfheal_dir_cbk_t dir_cbk,
-			    dht_layout_t *layout)
+dht_selfheal_new_directory (call_frame_t *frame,
+                            dht_selfheal_dir_cbk_t dir_cbk,
+                            dht_layout_t *layout)
 {
-	dht_local_t *local = NULL;
+        dht_local_t *local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	local->selfheal.dir_cbk = dir_cbk;
-	local->selfheal.layout = dht_layout_ref (frame->this, layout);
+        local->selfheal.dir_cbk = dir_cbk;
+        local->selfheal.layout = dht_layout_ref (frame->this, layout);
 
-	dht_layout_sort_volname (layout);
-	dht_selfheal_layout_new_directory (frame, &local->loc, layout);
-	dht_selfheal_dir_xattr (frame, &local->loc, layout);
-	return 0;
+        dht_layout_sort_volname (layout);
+        dht_selfheal_layout_new_directory (frame, &local->loc, layout);
+        dht_selfheal_dir_xattr (frame, &local->loc, layout);
+        return 0;
+}
+
+int
+dht_fix_directory_layout (call_frame_t *frame,
+                          dht_selfheal_dir_cbk_t dir_cbk,
+                          dht_layout_t *layout)
+{
+        dht_local_t  *local = NULL;
+        dht_layout_t *tmp_layout = NULL;
+
+        local = frame->local;
+
+        local->selfheal.dir_cbk = dir_cbk;
+        local->selfheal.layout = dht_layout_ref (frame->this, layout);
+
+        /* No layout sorting required here */
+        tmp_layout = dht_fix_layout_of_directory (frame, &local->loc, layout);
+        if (!tmp_layout) {
+                return -1;
+        }
+        dht_fix_dir_xattr (frame, &local->loc, tmp_layout);
+
+        return 0;
 }
 
 
 int
 dht_selfheal_directory (call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
-			loc_t *loc, dht_layout_t *layout)
+                        loc_t *loc, dht_layout_t *layout)
 {
-	dht_local_t *local    = NULL;
-	uint32_t     holes    = 0;
-	uint32_t     down     = 0;
-	uint32_t     misc     = 0;
-	int          ret      = 0;
-	xlator_t    *this     = NULL;
+        dht_local_t *local    = NULL;
+        uint32_t     down     = 0;
+        uint32_t     misc     = 0;
+        int          ret      = 0;
+        xlator_t    *this     = NULL;
 
-	local = frame->local;
-	this = frame->this;
+        local = frame->local;
+        this = frame->this;
 
-	dht_layout_anomalies (this, loc, layout,
+        dht_layout_anomalies (this, loc, layout,
                               &local->selfheal.hole_cnt,
                               &local->selfheal.overlaps_cnt,
-                              &local->selfheal.missing,
-                              &local->selfheal.down,
-                              &local->selfheal.misc);
-
-	holes    = local->selfheal.hole_cnt;
-	down     = local->selfheal.down;
-	misc     = local->selfheal.misc;
-
-	local->selfheal.dir_cbk = dir_cbk;
-	local->selfheal.layout = dht_layout_ref (this, layout);
-
-	if (down) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"%d subvolumes down -- not fixing", down);
-                ret = -1;
-                local->op_errno = ESTALE;
-		goto sorry_no_fix;
-	}
+                              NULL, &local->selfheal.down,
+                              &local->selfheal.misc, NULL);
 
-	if (misc) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"%d subvolumes have unrecoverable errors", misc);
-                ret = -1;
-                local->op_errno = ESTALE;
-		goto sorry_no_fix;
-	}
+        down     = local->selfheal.down;
+        misc     = local->selfheal.misc;
 
-	dht_layout_sort_volname (layout);
-	ret = dht_selfheal_dir_getafix (frame, loc, layout);
+        local->selfheal.dir_cbk = dir_cbk;
+        local->selfheal.layout = dht_layout_ref (this, layout);
 
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"not able to form layout for the directory");
-                ret = -1;
-                local->op_errno = ESTALE;
-		goto sorry_no_fix;
-	}
+        if (down) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%d subvolumes down -- not fixing", down);
+                ret = 0;
+                goto sorry_no_fix;
+        }
+
+        if (misc) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%d subvolumes have unrecoverable errors", misc);
+                ret = 0;
+                goto sorry_no_fix;
+        }
+
+        dht_layout_sort_volname (layout);
+        ret = dht_selfheal_dir_getafix (frame, loc, layout);
+
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "not able to form layout for the directory");
+                goto sorry_no_fix;
+        }
 
-	dht_selfheal_dir_mkdir (frame, loc, layout, 0);
+        dht_selfheal_dir_mkdir (frame, loc, layout, 0);
 
-	return 0;
+        return 0;
 
 sorry_no_fix:
-	/* TODO: need to put appropriate local->op_errno */
-	dht_selfheal_dir_finish (frame, this, ret);
+        /* TODO: need to put appropriate local->op_errno */
+        dht_selfheal_dir_finish (frame, this, ret);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_selfheal_restore (call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
-		      loc_t *loc, dht_layout_t *layout)
+                      loc_t *loc, dht_layout_t *layout)
 {
-	int          ret = 0;
-	dht_local_t *local    = NULL;
+        int          ret = 0;
+        dht_local_t *local    = NULL;
 
+        local = frame->local;
 
-	local = frame->local;
+        local->selfheal.dir_cbk = dir_cbk;
+        local->selfheal.layout = dht_layout_ref (frame->this, layout);
 
-	local->selfheal.dir_cbk = dir_cbk;
-	local->selfheal.layout = dht_layout_ref (frame->this, layout);
+        ret = dht_selfheal_dir_mkdir (frame, loc, layout, 1);
 
-	ret = dht_selfheal_dir_mkdir (frame, loc, layout, 1);
+        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 (-ret));
+                }
+        }
+out:
+        return 0;
+}
 
-	return ret;
+int
+dht_dir_attr_heal_done (int ret, call_frame_t *sync_frame, void *data)
+{
+        DHT_STACK_DESTROY (sync_frame);
+        return 0;
 }
diff --git a/xlators/cluster/dht/src/dht-shared.c b/xlators/cluster/dht/src/dht-shared.c
new file mode 100644
index 000000000..36c073973
--- /dev/null
+++ b/xlators/cluster/dht/src/dht-shared.c
@@ -0,0 +1,781 @@
+/*
+  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++) {
+                snprintf (key, sizeof (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]){
+                        snprintf (key, sizeof (key), "file_layouts[%d]", i);
+                        dht_layout_dump(conf->file_layouts[i], key);
+                }
+                if (conf->dir_layouts && conf->dir_layouts[i]) {
+                        snprintf (key, sizeof (key), "dir_layouts[%d]", i);
+                        dht_layout_dump(conf->dir_layouts[i], key);
+                }
+                if (conf->subvolume_status) {
+
+                        snprintf (key, sizeof (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) {
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (!conf->subvolume_status[i])
+                                continue;
+
+                        snprintf (key, sizeof (key), "subvolumes[%d]", i);
+                        gf_proc_dump_write (key, "%s",
+                                            conf->subvolumes[i]->name);
+
+                        snprintf (key, sizeof (key),
+                                  "du_stats[%d].avail_percent", i);
+                        gf_proc_dump_write (key, "%lf",
+                                            conf->du_stats[i].avail_percent);
+
+                        snprintf (key, sizeof (key), "du_stats[%d].avail_space",
+                                  i);
+                        gf_proc_dump_write (key, "%lu",
+                                            conf->du_stats[i].avail_space);
+
+                        snprintf (key, sizeof (key),
+                                  "du_stats[%d].avail_inodes", i);
+                        gf_proc_dump_write (key, "%lf",
+                                            conf->du_stats[i].avail_inodes);
+
+                        snprintf (key, sizeof (key), "du_stats[%d].log", i);
+                        gf_proc_dump_write (key, "%lu",
+                                            conf->du_stats[i].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."DHT_LINKFILE_STR,
+                     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. Takes number "
+                         "of subvolumes as default value."
+        },
+        { .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 d0f35e0ea..fc0ca2f77 100644
--- a/xlators/cluster/dht/src/dht.c
+++ b/xlators/cluster/dht/src/dht.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -23,501 +14,65 @@
 #include "config.h"
 #endif
 
-/* TODO: add NS locking */
-
 #include "statedump.h"
-#include "dht-common.c"
-
-/* TODO:
-   - use volumename in xattr instead of "dht"
-   - use NS locks
-   - handle all cases in self heal layout reconstruction
-   - complete linkfile selfheal
-*/
-
-
-void
-dht_layout_dump (dht_layout_t  *layout, const char *prefix)
-{
-
-        char    key[GF_DUMP_MAX_BUF_LEN];
-        int     i = 0;
-
-        if (!layout)
-                return;
-
-        gf_proc_dump_build_key(key, prefix, "cnt");
-        gf_proc_dump_write(key, "%d", layout->cnt);
-        gf_proc_dump_build_key(key, prefix, "preset");
-        gf_proc_dump_write(key, "%d", layout->preset);
-        gf_proc_dump_build_key(key, prefix, "gen");
-        gf_proc_dump_write(key, "%d", layout->gen);
-        gf_proc_dump_build_key(key, prefix, "type");
-        gf_proc_dump_write(key, "%d", layout->type);
-
-        for (i = 0; i < layout->cnt; i++) {
-                gf_proc_dump_build_key(key, prefix,"list[%d].err", i);
-                gf_proc_dump_write(key, "%d", layout->list[i].err);
-                gf_proc_dump_build_key(key, prefix,"list[%d].start", i);
-                gf_proc_dump_write(key, "%u", layout->list[i].start);
-                gf_proc_dump_build_key(key, prefix,"list[%d].stop", i);
-                gf_proc_dump_write(key, "%u", layout->list[i].stop);
-                if (layout->list[i].xlator) {
-                        gf_proc_dump_build_key(key, prefix,
-                                               "list[%d].xlator.type", i);
-                        gf_proc_dump_write(key, "%s",
-                                           layout->list[i].xlator->type);
-                        gf_proc_dump_build_key(key, prefix,
-                                               "list[%d].xlator.name", i);
-                        gf_proc_dump_write(key, "%s",
-                                           layout->list[i].xlator->name);
-                }
-        }
-}
-
-
-int32_t
-dht_priv_dump (xlator_t *this)
-{
-	char            key_prefix[GF_DUMP_MAX_BUF_LEN];
-	char            key[GF_DUMP_MAX_BUF_LEN];
-        int             i = 0;
-        dht_conf_t      *conf = NULL;
-        int             ret = 0;
-
-        if (!this)
-                return -1;
-
-        conf = this->private;
-
-        if (!conf)
-                return -1;
-
-        ret = TRY_LOCK(&conf->subvolume_lock);
-
-        if (ret != 0) {
-                gf_log("", GF_LOG_WARNING, "Unable to lock dht subvolume %s",
-                                this->name);
-                return ret;
-        }
-
-        gf_proc_dump_add_section("xlator.cluster.dht.%s.priv", this->name);
-        gf_proc_dump_build_key(key_prefix,"xlator.cluster.dht","%s.priv",
-                               this->name);
-        gf_proc_dump_build_key(key, key_prefix, "subvolume_cnt");
-        gf_proc_dump_write(key,"%d", conf->subvolume_cnt);
-        for (i = 0; i < conf->subvolume_cnt; i++) {
-                gf_proc_dump_build_key(key, key_prefix, "subvolumes[%d]", i);
-                gf_proc_dump_write(key, "%s.%s", conf->subvolumes[i]->type,
-                                   conf->subvolumes[i]->name);
-                if (conf->file_layouts && conf->file_layouts[i]){
-                        gf_proc_dump_build_key(key, key_prefix,
-                                               "file_layouts[%d]",i);
-                        dht_layout_dump(conf->file_layouts[i], key);
-                }
-                if (conf->dir_layouts && conf->dir_layouts[i]) {
-                        gf_proc_dump_build_key(key, key_prefix,
-                                              "dir_layouts[%d]",i);
-                        dht_layout_dump(conf->dir_layouts[i], key);
-                }
-                if (conf->subvolume_status) {
-                        gf_proc_dump_build_key(key, key_prefix,
-                                                "subvolume_status[%d]", i);
-                        gf_proc_dump_write(key, "%d",
-                                               (int)conf->subvolume_status[i]);
-                }
-
-        }
-
-        gf_proc_dump_build_key(key, key_prefix,"default_dir_layout");
-        dht_layout_dump(conf->default_dir_layout, key);
-
-        gf_proc_dump_build_key(key, key_prefix, "search_unhashed");
-        gf_proc_dump_write(key, "%d", conf->search_unhashed);
-        gf_proc_dump_build_key(key, key_prefix, "gen");
-        gf_proc_dump_write(key, "%d", conf->gen);
-        gf_proc_dump_build_key(key, key_prefix, "min_free_disk");
-        gf_proc_dump_write(key, "%lu", conf->min_free_disk);
-        gf_proc_dump_build_key(key, key_prefix, "disk_unit");
-        gf_proc_dump_write(key, "%c", conf->disk_unit);
-        gf_proc_dump_build_key(key, key_prefix, "refresh_interval");
-        gf_proc_dump_write(key, "%d", conf->refresh_interval);
-        gf_proc_dump_build_key(key, key_prefix, "unhashed_sticky_bit");
-        gf_proc_dump_write(key, "%d", conf->unhashed_sticky_bit);
-        if (conf ->du_stats) {
-                gf_proc_dump_build_key(key, key_prefix,
-                                "du_stats.avail_percent");
-                gf_proc_dump_write(key, "%lf", conf->du_stats->avail_percent);
-                gf_proc_dump_build_key(key, key_prefix,
-                                "du_stats.avail_space");
-                gf_proc_dump_write(key, "%lu", conf->du_stats->avail_space);
-                gf_proc_dump_build_key(key, key_prefix,
-                                "du_stats.log");
-                gf_proc_dump_write(key, "%lu", conf->du_stats->log);
-        }
-        gf_proc_dump_build_key(key, key_prefix, "last_stat_fetch");
-        gf_proc_dump_write(key, "%s", ctime(&conf->last_stat_fetch.tv_sec));
-
-        UNLOCK(&conf->subvolume_lock);
-
-        return 0;
-}
-
-int32_t
-dht_inodectx_dump (xlator_t *this, inode_t *inode)
-{
-        int             ret = -1;
-        char            key_prefix[GF_DUMP_MAX_BUF_LEN];
-        dht_layout_t    *layout = NULL;
-	uint64_t        tmp_layout = 0;
-
-        if (!inode)
-                return -1;
-
-	ret = inode_ctx_get (inode, this, &tmp_layout);
-
-        if (ret != 0)
-                return ret;
-
-        layout = (dht_layout_t *)(long)tmp_layout;
-
-        if (!layout)
-                return -1;
-
-        gf_proc_dump_build_key(key_prefix, "xlator.cluster.dht",
-                               "%s.inode.%ld", this->name, inode->ino);
-        dht_layout_dump(layout, key_prefix);
-
-        return 0;
-}
-
-int
-notify (xlator_t *this, int event, void *data, ...)
-{
-	int ret = -1;
-
-	ret = dht_notify (this, event, data);
-
-	return ret;
-}
-
-void
-fini (xlator_t *this)
-{
-        int         i = 0;
-        dht_conf_t *conf = NULL;
-
-	conf = this->private;
-        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);
-                }
-
-                if (conf->default_dir_layout)
-                        GF_FREE (conf->default_dir_layout);
-
-                if (conf->subvolumes)
-                        GF_FREE (conf->subvolumes);
-
-		if (conf->subvolume_status)
-			GF_FREE (conf->subvolume_status);
-
-                GF_FREE (conf);
-        }
-
-	return;
-}
-
-int32_t
-mem_acct_init (xlator_t *this)
-{
-        int     ret = -1;
-
-        if (!this)
-                return ret;
-
-        ret = xlator_mem_acct_init (this, gf_dht_mt_end + 1);
-        
-        if (ret != 0) {
-                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
-                                "failed");
-                return ret;
-        }
-
-        return ret;
-}
-int
-validate_options (xlator_t *this, dict_t *options, char **op_errstr)
-{
-        char            *temp_str = NULL;
-        gf_boolean_t     search_unhashed;
-        int              ret = 0;
-        
-
-
-
-
-        if (dict_get_str (options, "lookup-unhashed", &temp_str) == 0) {
-                if (strcasecmp (temp_str, "auto")) {
-                        if (!gf_string2boolean (temp_str, &search_unhashed)) {
-                                gf_log(this->name, GF_LOG_DEBUG, "Validated"
-                                                " lookup-unahashed (%s)",
-                                                                temp_str);
-                        }
-                        else {
-                                gf_log(this->name, GF_LOG_ERROR, "Validation:"
-                                                " lookup-unahashed should be boolean,"
-                                                                " not (%s)",
-                                                                temp_str);
-                                *op_errstr = gf_strdup ("Error, lookup-"
-                                                "unhashed be boolean");
-                                ret = -1;
-                                goto out;
-                        }
-                
-                }
-        }
-
-        
-
-
-
-out:
-                return ret;
-}
-
-int
-reconfigure (xlator_t *this, dict_t *options)
-{
-	dht_conf_t	*conf = NULL;
-	char		*temp_str = NULL;
-	gf_boolean_t     search_unhashed;
-	uint32_t         temp_free_disk = 0;
-	int		 ret = 0;
-
-
-	conf = this->private;
-        if (!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-unahashed reconfigured (%s)",
-				       temp_str);
-				conf->search_unhashed = search_unhashed;
-			}
-			else {
-				gf_log(this->name, GF_LOG_ERROR, "Reconfigure:"
-				       " lookup-unahashed should be boolean,"
-				        " not (%s), defaulting to (%d)",
-				       temp_str, conf->search_unhashed);
-				//return -1;
-				ret = -1;
-				goto out;
-			}
-		
-		}
-                else {
-			gf_log(this->name, GF_LOG_DEBUG, "Reconfigure:"
-			       " lookup-unahashed reconfigured auto ");
-                        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
-		}
-	}
-
-	if (dict_get_str (options, "min-free-disk", &temp_str) == 0) {
-		if (gf_string2percent (temp_str, &temp_free_disk) == 0) {
-                        if (temp_free_disk > 100) {
-                                gf_string2bytesize (temp_str, 
-                                                    &conf->min_free_disk);
-                                conf->disk_unit = 'b';
-                        } else {
-                                conf->min_free_disk = (uint64_t)temp_free_disk;
-                        }
-                } else {
-                        gf_string2bytesize (temp_str, &conf->min_free_disk);
-                        conf->disk_unit = 'b';
-                }
-
-		gf_log(this->name, GF_LOG_DEBUG, "Reconfigure:"
-			       " min-free-disk reconfigured to ",
-			       temp_str);
-	}
-
-out:
-	return ret;
-}
-
-int
-init (xlator_t *this)
-{
-        dht_conf_t    *conf = NULL;
-	char          *temp_str = NULL;
-        int            ret = -1;
-        int            i = 0;
-        uint32_t       temp_free_disk = 0;
-
-
-	if (!this->children) {
-		gf_log (this->name, GF_LOG_CRITICAL,
-			"Distribute needs more than one subvolume");
-		return -1;
-	}
-  
-	if (!this->parents) {
-		gf_log (this->name, GF_LOG_WARNING,
-			"dangling volume. check volfile");
-	}
-
-        conf = GF_CALLOC (1, sizeof (*conf), gf_dht_mt_dht_conf_t);
-        if (!conf) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
-        }
-
-        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_ON;
-	if (dict_get_str (this->options, "lookup-unhashed", &temp_str) == 0) {
-                /* If option is not "auto", other options _should_ be boolean */
-                if (strcasecmp (temp_str, "auto"))
-                        gf_string2boolean (temp_str, &conf->search_unhashed);
-                else
-                        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
-	}
-
-	conf->unhashed_sticky_bit = 0;
-
-	if (dict_get_str (this->options, "unhashed-sticky-bit", 
-                          &temp_str) == 0) {
-	        gf_string2boolean (temp_str, &conf->unhashed_sticky_bit);
-	}
-
-	conf->use_readdirp = 1;
-
-	if (dict_get_str (this->options, "use-readdirp",
-                          &temp_str) == 0) {
-	        gf_string2boolean (temp_str, &conf->use_readdirp);
-	}
-
-        conf->disk_unit = 'p';
-        conf->min_free_disk = 10;
-
-	if (dict_get_str (this->options, "min-free-disk", &temp_str) == 0) {
-		if (gf_string2percent (temp_str, &temp_free_disk) == 0) {
-                        if (temp_free_disk > 100) {
-                                gf_string2bytesize (temp_str, 
-                                                    &conf->min_free_disk);
-                                conf->disk_unit = 'b';
-                        } else {
-                                conf->min_free_disk = (uint64_t)temp_free_disk;
-                        }
-                } else {
-                        gf_string2bytesize (temp_str, &conf->min_free_disk);
-                        conf->disk_unit = 'b';
-                }
-	}
-
-
-        ret = dht_init_subvolumes (this, conf);
-        if (ret == -1) {
-                goto err;
-        }
-
-        ret = dht_layouts_init (this, conf);
-        if (ret == -1) {
-                goto err;
-        }
-
-        conf->du_stats = GF_CALLOC (conf->subvolume_cnt, sizeof (dht_du_t),
-                                    gf_dht_mt_dht_du_t);
-        if (!conf->du_stats) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
-        }
-
-	LOCK_INIT (&conf->subvolume_lock);
-	LOCK_INIT (&conf->layout_lock);
-
-	conf->gen = 1;
-
-        this->private = conf;
-
-        return 0;
-
-err:
-        if (conf) {
-                if (conf->file_layouts) {
-                        for (i = 0; i < conf->subvolume_cnt; i++) {
-                                GF_FREE (conf->file_layouts[i]);
-                        }
-                        GF_FREE (conf->file_layouts);
-                }
-
-                if (conf->default_dir_layout)
-                        GF_FREE (conf->default_dir_layout);
-
-                if (conf->subvolumes)
-                        GF_FREE (conf->subvolumes);
-
-		if (conf->subvolume_status)
-			GF_FREE (conf->subvolume_status);
-
-                if (conf->du_stats)
-                        GF_FREE (conf->du_stats);
-
-                GF_FREE (conf);
-        }
-
-        return -1;
-}
+#include "dht-common.h"
 
+class_methods_t class_methods = {
+        .init           = dht_init,
+        .fini           = dht_fini,
+        .reconfigure    = dht_reconfigure,
+        .notify         = dht_notify
+};
 
 struct xlator_fops fops = {
-	.lookup      = dht_lookup,
-	.mknod       = dht_mknod,
-	.create      = dht_create,
-
-	.stat        = dht_stat,
-	.fstat       = dht_fstat,
-	.truncate    = dht_truncate,
-	.ftruncate   = dht_ftruncate,
-	.access      = dht_access,
-	.readlink    = dht_readlink,
-	.setxattr    = dht_setxattr,
-	.getxattr    = dht_getxattr,
-	.removexattr = dht_removexattr,
-	.open        = dht_open,
-	.readv       = dht_readv,
-	.writev      = dht_writev,
-	.flush       = dht_flush,
-	.fsync       = dht_fsync,
-	.statfs      = dht_statfs,
-	.lk          = dht_lk,
-	.opendir     = dht_opendir,
-	.readdir     = dht_readdir,
-	.readdirp    = dht_readdirp,
-	.fsyncdir    = dht_fsyncdir,
-	.symlink     = dht_symlink,
-	.unlink      = dht_unlink,
-	.link        = dht_link,
-	.mkdir       = dht_mkdir,
-	.rmdir       = dht_rmdir,
-	.rename      = dht_rename,
-	.inodelk     = dht_inodelk,
-	.finodelk    = dht_finodelk,
-	.entrylk     = dht_entrylk,
-	.fentrylk    = dht_fentrylk,
-	.xattrop     = dht_xattrop,
-	.fxattrop    = dht_fxattrop,
-        .setattr     = dht_setattr, 
+        .lookup      = dht_lookup,
+        .mknod       = dht_mknod,
+        .create      = dht_create,
+
+        .open        = dht_open,
+        .statfs      = dht_statfs,
+        .opendir     = dht_opendir,
+        .readdir     = dht_readdir,
+        .readdirp    = dht_readdirp,
+        .fsyncdir    = dht_fsyncdir,
+        .symlink     = dht_symlink,
+        .unlink      = dht_unlink,
+        .link        = dht_link,
+        .mkdir       = dht_mkdir,
+        .rmdir       = dht_rmdir,
+        .rename      = dht_rename,
+        .entrylk     = dht_entrylk,
+        .fentrylk    = dht_fentrylk,
+
+        /* Inode read operations */
+        .stat        = dht_stat,
+        .fstat       = dht_fstat,
+        .access      = dht_access,
+        .readlink    = dht_readlink,
+        .getxattr    = dht_getxattr,
+        .fgetxattr    = dht_fgetxattr,
+        .readv       = dht_readv,
+        .flush       = dht_flush,
+        .fsync       = dht_fsync,
+        .inodelk     = dht_inodelk,
+        .finodelk    = dht_finodelk,
+        .lk          = dht_lk,
+
+        /* Inode write operations */
+        .fremovexattr = dht_fremovexattr,
+        .removexattr = dht_removexattr,
+        .setxattr    = dht_setxattr,
+        .fsetxattr   = dht_fsetxattr,
+        .truncate    = dht_truncate,
+        .ftruncate   = dht_ftruncate,
+        .writev      = dht_writev,
+        .xattrop     = dht_xattrop,
+        .fxattrop    = dht_fxattrop,
+        .setattr     = dht_setattr,
         .fsetattr    = dht_fsetattr,
+	.fallocate   = dht_fallocate,
+	.discard     = dht_discard,
+        .zerofill    = dht_zerofill,
 };
 
 struct xlator_dumpops dumpops = {
@@ -527,26 +82,8 @@ struct xlator_dumpops dumpops = {
 
 
 struct xlator_cbks cbks = {
-//	.release    = dht_release,
+//      .release    = dht_release,
 //      .releasedir = dht_releasedir,
-	.forget     = dht_forget
-};
-
-
-struct volume_options options[] = {
-        { .key  = {"lookup-unhashed"},
-          .value = {"auto", "yes", "no", "enable", "disable", "1", "0",
-                    "on", "off"},
-	  .type = GF_OPTION_TYPE_STR
-	},
-        { .key  = {"min-free-disk"},
-          .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
-        },
-        { .key = {"unhashed-sticky-bit"},
-          .type = GF_OPTION_TYPE_BOOL
-        },
-        { .key = {"use-readdirp"},
-          .type = GF_OPTION_TYPE_BOOL
-        },
-	{ .key  = {NULL} },
+        .forget     = dht_forget
 };
+;
diff --git a/xlators/cluster/dht/src/nufa.c b/xlators/cluster/dht/src/nufa.c
index 09388cbb0..e934acdf0 100644
--- a/xlators/cluster/dht/src/nufa.c
+++ b/xlators/cluster/dht/src/nufa.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -23,13 +14,15 @@
 #include "config.h"
 #endif
 
-#include "dht-common.c"
+#include "dht-common.h"
 
 /* TODO: all 'TODO's in dht.c holds good */
 
+extern struct volume_options options[];
+
 int
 nufa_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int op_ret, int op_errno,
+                       int op_ret, int op_errno,
                        inode_t *inode, struct iatt *stbuf, dict_t *xattr,
                        struct iatt *postparent)
 {
@@ -41,67 +34,61 @@ nufa_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         loc_t        *loc         = NULL;
         int           i           = 0;
         call_frame_t *prev        = NULL;
-	int           call_cnt    = 0;
+        int           call_cnt    = 0;
         int           ret         = 0;
 
-
         conf  = this->private;
 
         prev  = cookie;
         local = frame->local;
         loc   = &local->loc;
 
-	if (ENTRY_MISSING (op_ret, op_errno)) {
-		if (conf->search_unhashed) {
-			local->op_errno = ENOENT;
-			dht_lookup_everywhere (frame, this, loc);
-			return 0;
-		}
-	}
+        if (ENTRY_MISSING (op_ret, op_errno)) {
+                if (conf->search_unhashed) {
+                        local->op_errno = ENOENT;
+                        dht_lookup_everywhere (frame, this, loc);
+                        return 0;
+                }
+        }
 
         if (op_ret == -1)
                 goto out;
 
-        is_linkfile = check_is_linkfile (inode, stbuf, xattr);
+        is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+                                         conf->link_xattr_name);
         is_dir      = check_is_dir (inode, stbuf, xattr);
 
         if (!is_dir && !is_linkfile) {
                 /* non-directory and not a linkfile */
-
-		dht_itransform (this, prev->this, stbuf->ia_ino,
-				&stbuf->ia_ino);
-
-		ret = dht_layout_preset (this, prev->this, inode);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"could not set pre-set layout for subvol %s",
-				prev->this->name);
-			op_ret   = -1;
-			op_errno = EINVAL;
-			goto err;
-		}
+                ret = dht_layout_preset (this, prev->this, inode);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "could not set pre-set layout for subvol %s",
+                                prev->this->name);
+                        op_ret   = -1;
+                        op_errno = EINVAL;
+                        goto err;
+                }
 
                 goto out;
         }
 
         if (is_dir) {
                 call_cnt        = conf->subvolume_cnt;
-		local->call_cnt = call_cnt;
+                local->call_cnt = call_cnt;
 
                 local->inode = inode_ref (inode);
                 local->xattr = dict_ref (xattr);
 
-		local->op_ret = 0;
-		local->op_errno = 0;
+                local->op_ret = 0;
+                local->op_errno = 0;
 
-		local->layout = dht_layout_new (this, conf->subvolume_cnt);
-		if (!local->layout) {
-			op_ret   = -1;
-			op_errno = ENOMEM;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"memory allocation failed :(");
-			goto err;
-		}
+                local->layout = dht_layout_new (this, conf->subvolume_cnt);
+                if (!local->layout) {
+                        op_ret   = -1;
+                        op_errno = ENOMEM;
+                        goto err;
+                }
 
                 for (i = 0; i < call_cnt; i++) {
                         STACK_WIND (frame, dht_lookup_dir_cbk,
@@ -118,53 +105,52 @@ nufa_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "linkfile not having link subvolume. path=%s",
                                 loc->path);
-			dht_lookup_everywhere (frame, this, loc);
-			return 0;
+                        dht_lookup_everywhere (frame, this, loc);
+                        return 0;
                 }
 
-		STACK_WIND (frame, dht_lookup_linkfile_cbk,
-			    subvol, subvol->fops->lookup,
-			    &local->loc, local->xattr_req);
+                STACK_WIND (frame, dht_lookup_linkfile_cbk,
+                            subvol, subvol->fops->lookup,
+                            &local->loc, local->xattr_req);
         }
 
         return 0;
 
 out:
-	if (!local->hashed_subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			local->loc.path);
+        if (!local->hashed_subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        local->loc.path);
                 local->op_errno = ENOENT;
                 dht_lookup_everywhere (frame, this, loc);
                 return 0;
-	}
+        }
 
-	STACK_WIND (frame, dht_lookup_cbk,
-		    local->hashed_subvol, local->hashed_subvol->fops->lookup,
-		    &local->loc, local->xattr_req);
+        STACK_WIND (frame, dht_lookup_cbk,
+                    local->hashed_subvol, local->hashed_subvol->fops->lookup,
+                    &local->loc, local->xattr_req);
 
-	return 0;
+        return 0;
 
- err:
+err:
         DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno,
-                          inode, stbuf, xattr, NULL);
+                          inode, stbuf, xattr, postparent);
         return 0;
 }
 
 int
 nufa_lookup (call_frame_t *frame, xlator_t *this,
-	     loc_t *loc, dict_t *xattr_req)
+             loc_t *loc, dict_t *xattr_req)
 {
         xlator_t     *hashed_subvol = NULL;
-        xlator_t     *cached_subvol = NULL;
         xlator_t     *subvol = NULL;
         dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
+        dht_conf_t   *conf = NULL;
         int           ret    = -1;
         int           op_errno = -1;
-	dht_layout_t *layout = NULL;
-	int           i = 0;
-	int           call_cnt = 0;
+        dht_layout_t *layout = NULL;
+        int           i = 0;
+        int           call_cnt = 0;
 
 
         VALIDATE_OR_GOTO (frame, err);
@@ -173,40 +159,26 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
         VALIDATE_OR_GOTO (loc->inode, err);
         VALIDATE_OR_GOTO (loc->path, err);
 
-	conf = this->private;
+        conf = this->private;
 
-        local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-        ret = loc_dup (loc, &local->loc);
-        if (ret == -1) {
-                op_errno = errno;
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "copying location failed for path=%s",
-                        loc->path);
+        local = dht_local_init (frame, loc, NULL, GF_FOP_LOOKUP);
+        if (!local) {
+                op_errno = ENOMEM;
                 goto err;
         }
 
-	if (xattr_req) {
-		local->xattr_req = dict_ref (xattr_req);
-	} else {
-		local->xattr_req = dict_new ();
-	}
+        if (xattr_req) {
+                local->xattr_req = dict_ref (xattr_req);
+        } else {
+                local->xattr_req = dict_new ();
+        }
 
-	hashed_subvol = dht_subvol_get_hashed (this, &local->loc);
-	cached_subvol = dht_subvol_get_cached (this, local->loc.inode);
+        hashed_subvol = dht_subvol_get_hashed (this, &local->loc);
 
-	local->cached_subvol = cached_subvol;
-	local->hashed_subvol = hashed_subvol;
+        local->hashed_subvol = hashed_subvol;
 
         if (is_revalidate (loc)) {
-		local->layout = layout = dht_layout_get (this, loc->inode);
-
+                layout = local->layout;
                 if (!layout) {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "revalidate without cache. path=%s",
@@ -215,25 +187,24 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
                         goto err;
                 }
 
-		if (layout->gen && (layout->gen < conf->gen)) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"incomplete layout failure for path=%s",
-				loc->path);
+                if (layout->gen && (layout->gen < conf->gen)) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "incomplete layout failure for path=%s",
+                                loc->path);
                         dht_layout_unref (this, local->layout);
-			goto do_fresh_lookup;
-		}
+                        goto do_fresh_lookup;
+                }
 
-		local->inode    = inode_ref (loc->inode);
-		local->ia_ino   = loc->inode->ino;
+                local->inode = inode_ref (loc->inode);
 
-		local->call_cnt = layout->cnt;
-		call_cnt = local->call_cnt;
+                local->call_cnt = layout->cnt;
+                call_cnt = local->call_cnt;
 
-		/* NOTE: we don't require 'trusted.glusterfs.dht.linkto' attribute,
-		 *       revalidates directly go to the cached-subvolume.
-		 */
-		ret = dict_set_uint32 (local->xattr_req,
-				       "trusted.glusterfs.dht", 4 * 4);
+                /* NOTE: we don't require 'trusted.glusterfs.dht.linkto' attribute,
+                 *       revalidates directly go to the cached-subvolume.
+                 */
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->xattr_name, 4 * 4);
                 if (ret < 0) {
                         gf_log (this->name, GF_LOG_ERROR,
                                 "Failed to set dict value.");
@@ -241,20 +212,20 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
                         goto err;
                 }
 
-		for (i = 0; i < layout->cnt; i++) {
-			subvol = layout->list[i].xlator;
+                for (i = 0; i < layout->cnt; i++) {
+                        subvol = layout->list[i].xlator;
 
-			STACK_WIND (frame, dht_revalidate_cbk,
-				    subvol, subvol->fops->lookup,
-				    loc, local->xattr_req);
+                        STACK_WIND (frame, dht_revalidate_cbk,
+                                    subvol, subvol->fops->lookup,
+                                    loc, local->xattr_req);
 
-			if (!--call_cnt)
-				break;
-		}
-	} else {
+                        if (!--call_cnt)
+                                break;
+                }
+        } else {
         do_fresh_lookup:
-		ret = dict_set_uint32 (local->xattr_req,
-				       "trusted.glusterfs.dht", 4 * 4);
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->xattr_name, 4 * 4);
                 if (ret < 0) {
                         gf_log (this->name, GF_LOG_ERROR,
                                 "Failed to set dict value.");
@@ -262,8 +233,8 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
                         goto err;
                 }
 
-		ret = dict_set_uint32 (local->xattr_req,
-				       "trusted.glusterfs.dht.linkto", 256);
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->link_xattr_name, 256);
                 if (ret < 0) {
                         gf_log (this->name, GF_LOG_ERROR,
                                 "Failed to set dict value.");
@@ -271,111 +242,100 @@ nufa_lookup (call_frame_t *frame, xlator_t *this,
                         goto err;
                 }
 
-		/* Send it to only local volume */
-		STACK_WIND (frame, nufa_local_lookup_cbk,
-			    (xlator_t *)conf->private,
-			    ((xlator_t *)conf->private)->fops->lookup,
-			    loc, local->xattr_req);
-	}
+                /* Send it to only local volume */
+                STACK_WIND (frame, nufa_local_lookup_cbk,
+                            (xlator_t *)conf->private,
+                            ((xlator_t *)conf->private)->fops->lookup,
+                            loc, local->xattr_req);
+        }
 
         return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-        DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
-	return 0;
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL);
+        return 0;
 }
 
 int
 nufa_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
-				 xlator_t *this, int op_ret, int op_errno,
+                                 xlator_t *this, int op_ret, int op_errno,
                                  inode_t *inode, struct iatt *stbuf,
                                  struct iatt *preparent,
-                                 struct iatt *postparent)
+                                 struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
+        dht_local_t  *local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	if (op_ret == -1)
-		goto err;
+        if (op_ret == -1)
+                goto err;
 
-	STACK_WIND (frame, dht_create_cbk,
-		    local->cached_subvol, local->cached_subvol->fops->create,
-		    &local->loc, local->flags, local->mode, local->fd,
-                    local->params);
+        STACK_WIND (frame, dht_create_cbk,
+                    local->cached_subvol, local->cached_subvol->fops->create,
+                    &local->loc, local->flags, local->mode, local->umask,
+                    local->fd, local->params);
 
-	return 0;
+        return 0;
 
- err:
-	DHT_STACK_UNWIND (create, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL, NULL);
-	return 0;
+err:
+        DHT_STACK_UNWIND (create, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL, NULL);
+        return 0;
 }
 
 int
 nufa_create (call_frame_t *frame, xlator_t *this,
-	     loc_t *loc, int32_t flags, mode_t mode,
-             fd_t *fd, dict_t *params)
+             loc_t *loc, int32_t flags, mode_t mode,
+             mode_t umask, fd_t *fd, dict_t *params)
 {
-	dht_local_t *local = NULL;
-	dht_conf_t  *conf  = NULL;
-	xlator_t    *subvol = NULL;
+        dht_local_t *local = NULL;
+        dht_conf_t  *conf  = NULL;
+        xlator_t    *subvol = NULL;
         xlator_t    *avail_subvol = NULL;
-	int          op_errno = -1;
-	int          ret = -1;
+        int          op_errno = -1;
 
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
-	conf  = this->private;
+        conf  = this->private;
 
         dht_get_du_info (frame, this, loc);
 
-        local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	subvol = dht_subvol_get_hashed (this, loc);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			loc->path);
-		op_errno = ENOENT;
-		goto err;
-	}
+        local = dht_local_init (frame, loc, fd, GF_FOP_CREATE);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = dht_subvol_get_hashed (this, loc);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        loc->path);
+                op_errno = ENOENT;
+                goto err;
+        }
 
         avail_subvol = conf->private;
         if (dht_is_subvol_filled (this, (xlator_t *)conf->private)) {
                 avail_subvol =
                         dht_free_disk_available_subvol (this,
-                                                        (xlator_t *)conf->private);
+                                                        (xlator_t *)conf->private,
+                                                        local);
         }
 
         if (subvol != avail_subvol) {
                 /* create a link file instead of actual file */
-                ret = loc_copy (&local->loc, loc);
-                if (ret == -1) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "Out of memory");
-                                op_errno = ENOMEM;
-                                goto err;
-                }
-
-                local->fd = fd_ref (fd);
                 local->params = dict_ref (params);
                 local->mode = mode;
                 local->flags = flags;
-
+                local->umask = umask;
                 local->cached_subvol = avail_subvol;
-                dht_linkfile_create (frame,
-                                     nufa_create_linkfile_create_cbk,
-                                     avail_subvol, subvol, loc);
+                dht_linkfile_create (frame, nufa_create_linkfile_create_cbk,
+                                     this, avail_subvol, subvol, loc);
                 return 0;
         }
 
@@ -384,379 +344,333 @@ nufa_create (call_frame_t *frame, xlator_t *this,
 
         STACK_WIND (frame, dht_create_cbk,
                     subvol, subvol->fops->create,
-                    loc, flags, mode, fd, params);
+                    loc, flags, mode, umask, fd, params);
 
         return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (create, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (create, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 int
 nufa_mknod_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                          int op_ret, int op_errno, inode_t *inode,
                          struct iatt *stbuf, struct iatt *preparent,
-                         struct iatt *postparent)
+                         struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
+        dht_local_t  *local = NULL;
 
-	local = frame->local;
-
-	if (op_ret >= 0) {
-		STACK_WIND (frame, dht_newfile_cbk,
-			    local->cached_subvol,
-			    local->cached_subvol->fops->mknod,
-			    &local->loc, local->mode, local->rdev,
-                            local->params);
+        local = frame->local;
+        if (!local || !local->cached_subvol) {
+                op_errno = EINVAL;
+                op_ret = -1;
+                goto err;
+        }
 
-		return 0;
-	}
+        if (op_ret >= 0) {
+                STACK_WIND_COOKIE (frame, dht_newfile_cbk,
+                            (void *)local->cached_subvol, local->cached_subvol,
+                            local->cached_subvol->fops->mknod,
+                            &local->loc, local->mode, local->rdev,
+                            local->umask, local->params);
 
+                return 0;
+        }
+err:
         WIPE (postparent);
         WIPE (preparent);
 
-	DHT_STACK_UNWIND (link, frame, op_ret, op_errno,
-                          inode, stbuf, preparent, postparent);
-	return 0;
+        DHT_STACK_UNWIND (link, frame, op_ret, op_errno,
+                          inode, stbuf, preparent, postparent, xdata);
+        return 0;
 }
 
 
 int
 nufa_mknod (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, mode_t mode, dev_t rdev, dict_t *params)
+            loc_t *loc, mode_t mode, dev_t rdev, mode_t umask, dict_t *params)
 {
-	dht_local_t *local = NULL;
-	dht_conf_t  *conf  = NULL;
-	xlator_t    *subvol = NULL;
+        dht_local_t *local = NULL;
+        dht_conf_t  *conf  = NULL;
+        xlator_t    *subvol = NULL;
         xlator_t    *avail_subvol = NULL;
-	int          op_errno = -1;
-	int          ret = -1;
+        int          op_errno = -1;
 
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
-	conf  = this->private;
+        conf  = this->private;
 
         dht_get_du_info (frame, this, loc);
 
-        local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	subvol = dht_subvol_get_hashed (this, loc);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			loc->path);
-		op_errno = ENOENT;
-		goto err;
-	}
+        local = dht_local_init (frame, loc, NULL, GF_FOP_MKNOD);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = dht_subvol_get_hashed (this, loc);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        loc->path);
+                op_errno = ENOENT;
+                goto err;
+        }
 
         /* Consider the disksize in consideration */
         avail_subvol = conf->private;
         if (dht_is_subvol_filled (this, (xlator_t *)conf->private)) {
                 avail_subvol =
                         dht_free_disk_available_subvol (this,
-                                                        (xlator_t *)conf->private);
+                                                        (xlator_t *)conf->private,
+                                                        local);
         }
 
-	if (avail_subvol != subvol) {
-		/* Create linkfile first */
-		ret = loc_copy (&local->loc, loc);
-		if (ret == -1) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"Out of memory");
-			op_errno = ENOMEM;
-			goto err;
-		}
+        if (avail_subvol != subvol) {
+                /* Create linkfile first */
 
                 local->params = dict_ref (params);
-		local->mode = mode;
-		local->rdev = rdev;
-		local->cached_subvol = avail_subvol;
+                local->mode = mode;
+                local->umask = umask;
+                local->rdev = rdev;
+                local->cached_subvol = avail_subvol;
 
-		dht_linkfile_create (frame, nufa_mknod_linkfile_cbk,
+                dht_linkfile_create (frame, nufa_mknod_linkfile_cbk, this,
                                      avail_subvol, subvol, loc);
-		return 0;
-	}
+                return 0;
+        }
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"creating %s on %s", loc->path, subvol->name);
+        gf_log (this->name, GF_LOG_TRACE,
+                "creating %s on %s", loc->path, subvol->name);
 
-	STACK_WIND (frame, dht_newfile_cbk,
-		    subvol, subvol->fops->mknod,
-		    loc, mode, rdev, params);
+        STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol, subvol,
+                           subvol->fops->mknod, loc, mode, rdev, umask,
+                           params);
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
-int
-notify (xlator_t *this, int event, void *data, ...)
+gf_boolean_t
+same_first_part (char *str1, char term1, char *str2, char term2)
 {
-	int ret = -1;
-
-	ret = dht_notify (this, event, data);
-
-	return ret;
+        gf_boolean_t    ended1;
+        gf_boolean_t    ended2;
+
+        for (;;) {
+                ended1 = ((*str1 == '\0') || (*str1 == term1));
+                ended2 = ((*str2 == '\0') || (*str2 == term2));
+                if (ended1 && ended2) {
+                        return _gf_true;
+                }
+                if (ended1 || ended2 || (*str1 != *str2)) {
+                        return _gf_false;
+                }
+                ++str1;
+                ++str2;
+        }
 }
 
-void
-fini (xlator_t *this)
+typedef struct nufa_args {
+        xlator_t    *this;
+        char        *volname;
+        gf_boolean_t addr_match;
+} nufa_args_t;
+
+static void
+nufa_find_local_brick (xlator_t *xl, void *data)
 {
-        int         i = 0;
-        dht_conf_t *conf = NULL;
+        nufa_args_t     *args = data;
+        xlator_t        *this = args->this;
+        char            *local_volname = args->volname;
+        gf_boolean_t    addr_match = args->addr_match;
+        char            *brick_host = NULL;
+        dht_conf_t      *conf = this->private;
+        int             ret = -1;
+
+        /*This means a local subvol was already found. We pick the first brick
+         * that is local*/
+        if (conf->private)
+                return;
+
+        if (strcmp (xl->name, local_volname) == 0) {
+                conf->private = xl;
+                gf_log (this->name, GF_LOG_INFO, "Using specified subvol %s",
+                        local_volname);
+                return;
+        }
 
-	conf = this->private;
+        if (!addr_match)
+                return;
+
+        ret = dict_get_str (xl->options, "remote-host", &brick_host);
+        if ((ret == 0) &&
+            (gf_is_same_address (local_volname, brick_host) ||
+             gf_is_local_addr (brick_host))) {
+                conf->private = xl;
+                gf_log (this->name, GF_LOG_INFO, "Using the first local "
+                        "subvol %s", xl->name);
+                return;
+        }
 
-        if (conf) {
-                if (conf->file_layouts) {
-                        for (i = 0; i < conf->subvolume_cnt; i++) {
-                                GF_FREE (conf->file_layouts[i]);
-                        }
-                        GF_FREE (conf->file_layouts);
-                }
+}
 
-                if (conf->default_dir_layout)
-                        GF_FREE (conf->default_dir_layout);
+static void
+nufa_to_dht (xlator_t *this)
+{
+        GF_ASSERT (this);
+        GF_ASSERT (this->fops);
 
-                if (conf->subvolumes)
-                        GF_FREE (conf->subvolumes);
+        this->fops->lookup = dht_lookup;
+        this->fops->create = dht_create;
+        this->fops->mknod  = dht_mknod;
+}
 
-		if (conf->subvolume_status)
-			GF_FREE (conf->subvolume_status);
+int
+nufa_find_local_subvol (xlator_t *this,
+                        void (*fn) (xlator_t *each, void* data), void *data)
+{
+        int             ret = -1;
+        dht_conf_t      *conf = this->private;
+        xlator_list_t   *trav = NULL;
+        xlator_t        *parent = NULL;
+        xlator_t        *candidate = NULL;
+
+        xlator_foreach_depth_first (this, fn, data);
+        if (!conf->private) {
+                gf_log (this->name, GF_LOG_ERROR, "Couldn't find a local "
+                        "brick");
+                return -1;
+        }
+
+        candidate = conf->private;
+        trav = candidate->parents;
+        while (trav) {
+
+                parent = trav->xlator;
+                if (strcmp (parent->type, "cluster/nufa") == 0) {
+                        gf_log (this->name, GF_LOG_INFO, "Found local subvol, "
+                                "%s", candidate->name);
+                        ret = 0;
+                        conf->private = candidate;
+                        break;
+                }
 
-                GF_FREE (conf);
+                candidate = parent;
+                trav = parent->parents;
         }
 
-	return;
+        return ret;
 }
 
 int
-init (xlator_t *this)
+nufa_init (xlator_t *this)
 {
-        dht_conf_t    *conf = NULL;
-	xlator_list_t *trav = NULL;
-	data_t        *data = NULL;
-	char          *local_volname = NULL;
-	char          *temp_str = NULL;
+        data_t        *data = NULL;
+        char          *local_volname = NULL;
         int            ret = -1;
-        int            i = 0;
-	char           my_hostname[256];
-	uint32_t       temp_free_disk = 0;
-
-	if (!this->children) {
-		gf_log (this->name, GF_LOG_CRITICAL,
-			"NUFA needs more than one subvolume");
-		return -1;
-	}
-
-	if (!this->parents) {
-		gf_log (this->name, GF_LOG_WARNING,
-			"dangling volume. check volfile");
-	}
-
-        conf = GF_CALLOC (1, sizeof (*conf),
-                          gf_dht_mt_dht_conf_t);
-        if (!conf) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
+        char           my_hostname[256];
+        gf_boolean_t   addr_match = _gf_false;
+        nufa_args_t    args = {0, };
+
+        ret = dht_init(this);
+        if (ret) {
+                return ret;
         }
 
-        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_ON;
-	if (dict_get_str (this->options, "lookup-unhashed", &temp_str) == 0) {
-                /* If option is not "auto", other options _should_ be boolean */
-                if (strcasecmp (temp_str, "auto"))
-                        gf_string2boolean (temp_str, &conf->search_unhashed);
-                else
-                        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
-	}
+        if ((data = dict_get (this->options, "local-volume-name"))) {
+                local_volname = data->data;
 
-        ret = dht_init_subvolumes (this, conf);
-        if (ret == -1) {
-                goto err;
-        }
+        } else {
+                addr_match = _gf_true;
+                local_volname = "localhost";
+                ret = gethostname (my_hostname, 256);
+                if (ret == 0)
+                        local_volname = my_hostname;
 
-        ret = dht_layouts_init (this, conf);
-        if (ret == -1) {
-                goto err;
-        }
+                else
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "could not find hostname (%s)",
+                                strerror (errno));
 
-	LOCK_INIT (&conf->subvolume_lock);
-	LOCK_INIT (&conf->layout_lock);
-
-	conf->gen = 1;
-
-	local_volname = "localhost";
-	ret = gethostname (my_hostname, 256);
-	if (ret < 0) {
-		gf_log (this->name, GF_LOG_WARNING,
-			"could not find hostname (%s)",
-			strerror (errno));
-	}
-
-	if (ret == 0)
-		local_volname = my_hostname;
-
-	data = dict_get (this->options, "local-volume-name");
-	if (data) {
-		local_volname = data->data;
-	}
-
-	trav = this->children;
-	while (trav) {
-		if (strcmp (trav->xlator->name, local_volname) == 0)
-			break;
-		trav = trav->next;
-	}
-
-	if (!trav) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Could not find subvolume named '%s'. "
-			"Please define volume with the name as the hostname "
-			"or override it with 'option local-volume-name'",
-			local_volname);
-		goto err;
-	}
-	/* The volume specified exists */
-	conf->private = trav->xlator;
-
-        conf->min_free_disk = 10;
-        conf->disk_unit = 'p';
-
-        if (dict_get_str (this->options, "min-free-disk",
-                          &temp_str) == 0) {
-                if (gf_string2percent (temp_str,
-                                       &temp_free_disk) == 0) {
-                        if (temp_free_disk > 100) {
-                                gf_string2bytesize (temp_str,
-                                                        &conf->min_free_disk);
-                                conf->disk_unit = 'b';
-                        } else {
-                                conf->min_free_disk = (uint64_t)temp_free_disk;
-                                conf->disk_unit = 'p';
-                        }
-                } else {
-                        gf_string2bytesize (temp_str,
-                                                &conf->min_free_disk);
-                        conf->disk_unit = 'b';
-                }
         }
 
-        conf->du_stats = GF_CALLOC (conf->subvolume_cnt, sizeof (dht_du_t),
-                                    gf_dht_mt_dht_du_t);
-        if (!conf->du_stats) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
+        args.this = this;
+        args.volname = local_volname;
+        args.addr_match = addr_match;
+        ret = nufa_find_local_subvol (this, nufa_find_local_brick, &args);
+        if (ret) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "Unable to find local subvolume, switching "
+                        "to dht mode");
+                nufa_to_dht (this);
         }
-
-        this->private = conf;
-
         return 0;
+}
 
-err:
-        if (conf) {
-                if (conf->file_layouts) {
-                        for (i = 0; i < conf->subvolume_cnt; i++) {
-                                GF_FREE (conf->file_layouts[i]);
-                        }
-                        GF_FREE (conf->file_layouts);
-                }
-
-                if (conf->default_dir_layout)
-                        GF_FREE (conf->default_dir_layout);
-
-                if (conf->subvolumes)
-                        GF_FREE (conf->subvolumes);
-
-		if (conf->subvolume_status)
-			GF_FREE (conf->subvolume_status);
-
-                if (conf->du_stats)
-                        GF_FREE (conf->du_stats);
-
-                GF_FREE (conf);
-        }
 
-        return -1;
-}
+class_methods_t class_methods = {
+        .init           = nufa_init,
+        .fini           = dht_fini,
+        .reconfigure    = dht_reconfigure,
+        .notify         = dht_notify
+};
 
 
 struct xlator_fops fops = {
-	.lookup      = nufa_lookup,
-	.create      = nufa_create,
-	.mknod       = nufa_mknod,
-
-	.stat        = dht_stat,
-	.fstat       = dht_fstat,
-	.truncate    = dht_truncate,
-	.ftruncate   = dht_ftruncate,
-	.access      = dht_access,
-	.readlink    = dht_readlink,
-	.setxattr    = dht_setxattr,
-	.getxattr    = dht_getxattr,
-	.removexattr = dht_removexattr,
-	.open        = dht_open,
-	.readv       = dht_readv,
-	.writev      = dht_writev,
-	.flush       = dht_flush,
-	.fsync       = dht_fsync,
-	.statfs      = dht_statfs,
-	.lk          = dht_lk,
-	.opendir     = dht_opendir,
-	.readdir     = dht_readdir,
-	.readdirp    = dht_readdirp,
-	.fsyncdir    = dht_fsyncdir,
-	.symlink     = dht_symlink,
-	.unlink      = dht_unlink,
-	.link        = dht_link,
-	.mkdir       = dht_mkdir,
-	.rmdir       = dht_rmdir,
-	.rename      = dht_rename,
-	.inodelk     = dht_inodelk,
-	.finodelk    = dht_finodelk,
-	.entrylk     = dht_entrylk,
-	.fentrylk    = dht_fentrylk,
-	.xattrop     = dht_xattrop,
-	.fxattrop    = dht_fxattrop,
+        .lookup      = nufa_lookup,
+        .create      = nufa_create,
+        .mknod       = nufa_mknod,
+
+        .stat        = dht_stat,
+        .fstat       = dht_fstat,
+        .truncate    = dht_truncate,
+        .ftruncate   = dht_ftruncate,
+        .access      = dht_access,
+        .readlink    = dht_readlink,
+        .setxattr    = dht_setxattr,
+        .getxattr    = dht_getxattr,
+        .removexattr = dht_removexattr,
+        .open        = dht_open,
+        .readv       = dht_readv,
+        .writev      = dht_writev,
+        .flush       = dht_flush,
+        .fsync       = dht_fsync,
+        .statfs      = dht_statfs,
+        .lk          = dht_lk,
+        .opendir     = dht_opendir,
+        .readdir     = dht_readdir,
+        .readdirp    = dht_readdirp,
+        .fsyncdir    = dht_fsyncdir,
+        .symlink     = dht_symlink,
+        .unlink      = dht_unlink,
+        .link        = dht_link,
+        .mkdir       = dht_mkdir,
+        .rmdir       = dht_rmdir,
+        .rename      = dht_rename,
+        .inodelk     = dht_inodelk,
+        .finodelk    = dht_finodelk,
+        .entrylk     = dht_entrylk,
+        .fentrylk    = dht_fentrylk,
+        .xattrop     = dht_xattrop,
+        .fxattrop    = dht_fxattrop,
         .setattr     = dht_setattr,
 };
 
 
 struct xlator_cbks cbks = {
-	.forget     = dht_forget
-};
-
-
-struct volume_options options[] = {
-        { .key  = {"lookup-unhashed"},
-          .value = {"auto", "yes", "no", "enable", "disable", "1", "0",
-                    "on", "off"},
-	  .type = GF_OPTION_TYPE_STR
-	},
-	{ .key  = {"local-volume-name"},
-	  .type = GF_OPTION_TYPE_XLATOR
-	},
-        { .key  = {"min-free-disk"},
-          .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
-        },
-	{ .key  = {NULL} },
+        .forget     = dht_forget
 };
diff --git a/xlators/cluster/dht/src/switch.c b/xlators/cluster/dht/src/switch.c
index 344a65789..2717ce975 100644
--- a/xlators/cluster/dht/src/switch.c
+++ b/xlators/cluster/dht/src/switch.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -23,7 +14,7 @@
 #include "config.h"
 #endif
 
-#include "dht-common.c"
+#include "dht-common.h"
 #include "dht-mem-types.h"
 
 #include <sys/time.h>
@@ -31,21 +22,23 @@
 #include <fnmatch.h>
 #include <string.h>
 
+extern struct volume_options options[];
+
 struct switch_sched_array {
-	xlator_t *xl;
-	int32_t   eligible;
-	int32_t   considered;
+        xlator_t *xl;
+        int32_t   eligible;
+        int32_t   considered;
 };
 
 /* Select one of this struct based on the path's pattern match */
 struct switch_struct {
-	struct switch_struct      *next;
+        struct switch_struct      *next;
         struct switch_sched_array *array;
-	int32_t                    node_index; /* Index of the node in
+        int32_t                    node_index; /* Index of the node in
                                                   this pattern. */
-	int32_t                    num_child;  /* Total num of child nodes
+        int32_t                    num_child;  /* Total num of child nodes
                                                   with this pattern. */
-	char                       path_pattern[256];
+        char                       path_pattern[256];
 };
 
 /* TODO: all 'TODO's in dht.c holds good */
@@ -74,31 +67,39 @@ get_switch_matching_subvol (const char *path, dht_conf_t *conf,
                             xlator_t *hashed_subvol)
 {
         struct switch_struct *cond      = NULL;
-	struct switch_struct *trav      = NULL;
-	char                 *pathname  = NULL;
-	int                   idx     = 0;
+        struct switch_struct *trav      = NULL;
+        char                 *pathname  = NULL;
+        int                   idx       = 0;
+        xlator_t             *subvol    = NULL;
 
         cond = conf->private;
+        subvol = hashed_subvol;
         if (!cond)
-                return hashed_subvol;
+                goto out;
 
-        trav = cond;
         pathname = gf_strdup (path);
-	while (trav) {
-		if (fnmatch (trav->path_pattern,
-			     pathname, FNM_NOESCAPE) == 0) {
+        if (!pathname)
+                goto out;
+
+        trav = cond;
+        while (trav) {
+                if (fnmatch (trav->path_pattern,
+                             pathname, FNM_NOESCAPE) == 0) {
                         for (idx = 0; idx < trav->num_child; idx++) {
                                 if (trav->array[idx].xl == hashed_subvol)
-                                        return hashed_subvol;
+                                        goto out;
                         }
                         idx = trav->node_index++;
                         trav->node_index %= trav->num_child;
-                        return trav->array[idx].xl;
-		}
-		trav = trav->next;
-	}
-	GF_FREE (pathname);
-	return hashed_subvol;
+                        subvol = trav->array[idx].xl;
+                        goto out;
+                }
+                trav = trav->next;
+        }
+out:
+        GF_FREE (pathname);
+
+        return subvol;
 }
 
 
@@ -116,7 +117,7 @@ switch_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         loc_t        *loc         = NULL;
         int           i           = 0;
         call_frame_t *prev        = NULL;
-	int           call_cnt    = 0;
+        int           call_cnt    = 0;
         int           ret         = 0;
 
         conf  = this->private;
@@ -125,57 +126,55 @@ switch_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         local = frame->local;
         loc   = &local->loc;
 
-	if (ENTRY_MISSING (op_ret, op_errno)) {
-		if (conf->search_unhashed) {
-			local->op_errno = ENOENT;
-			dht_lookup_everywhere (frame, this, loc);
-			return 0;
-		}
-	}
+        if (ENTRY_MISSING (op_ret, op_errno)) {
+                if (conf->search_unhashed) {
+                        local->op_errno = ENOENT;
+                        dht_lookup_everywhere (frame, this, loc);
+                        return 0;
+                }
+        }
 
         if (op_ret == -1)
                 goto out;
 
-        is_linkfile = check_is_linkfile (inode, stbuf, xattr);
+        is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+                                         conf->link_xattr_name);
         is_dir      = check_is_dir (inode, stbuf, xattr);
 
         if (!is_dir && !is_linkfile) {
                 /* non-directory and not a linkfile */
 
-		dht_itransform (this, prev->this, stbuf->ia_ino,
-				&stbuf->ia_ino);
-
-		ret = dht_layout_preset (this, prev->this, inode);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"could not set pre-set layout for subvol %s",
-				prev->this->name);
-			op_ret   = -1;
-			op_errno = EINVAL;
-			goto err;
-		}
+                ret = dht_layout_preset (this, prev->this, inode);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "could not set pre-set layout for subvol %s",
+                                prev->this->name);
+                        op_ret   = -1;
+                        op_errno = EINVAL;
+                        goto err;
+                }
 
                 goto out;
         }
 
         if (is_dir) {
                 call_cnt        = conf->subvolume_cnt;
-		local->call_cnt = call_cnt;
+                local->call_cnt = call_cnt;
 
                 local->inode = inode_ref (inode);
                 local->xattr = dict_ref (xattr);
 
-		local->op_ret = 0;
-		local->op_errno = 0;
+                local->op_ret = 0;
+                local->op_errno = 0;
 
-		local->layout = dht_layout_new (this, conf->subvolume_cnt);
-		if (!local->layout) {
-			op_ret   = -1;
-			op_errno = ENOMEM;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"memory allocation failed :(");
-			goto err;
-		}
+                local->layout = dht_layout_new (this, conf->subvolume_cnt);
+                if (!local->layout) {
+                        op_ret   = -1;
+                        op_errno = ENOMEM;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "memory allocation failed :(");
+                        goto err;
+                }
 
                 for (i = 0; i < call_cnt; i++) {
                         STACK_WIND (frame, dht_lookup_dir_cbk,
@@ -192,34 +191,34 @@ switch_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "linkfile not having link subvolume. path=%s",
                                 loc->path);
-			dht_lookup_everywhere (frame, this, loc);
-			return 0;
+                        dht_lookup_everywhere (frame, this, loc);
+                        return 0;
                 }
 
-		STACK_WIND (frame, dht_lookup_linkfile_cbk,
-			    subvol, subvol->fops->lookup,
-			    &local->loc, local->xattr_req);
+                STACK_WIND (frame, dht_lookup_linkfile_cbk,
+                            subvol, subvol->fops->lookup,
+                            &local->loc, local->xattr_req);
         }
 
         return 0;
 
 out:
-	if (!local->hashed_subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			local->loc.path);
+        if (!local->hashed_subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        local->loc.path);
                 local->op_errno = ENOENT;
                 dht_lookup_everywhere (frame, this, loc);
                 return 0;
-	}
+        }
 
-	STACK_WIND (frame, dht_lookup_cbk,
-		    local->hashed_subvol, local->hashed_subvol->fops->lookup,
-		    &local->loc, local->xattr_req);
+        STACK_WIND (frame, dht_lookup_cbk,
+                    local->hashed_subvol, local->hashed_subvol->fops->lookup,
+                    &local->loc, local->xattr_req);
 
-	return 0;
+        return 0;
 
- err:
+err:
         DHT_STACK_UNWIND (lookup, frame, op_ret, op_errno,
                           inode, stbuf, xattr, NULL);
         return 0;
@@ -227,18 +226,18 @@ out:
 
 int
 switch_lookup (call_frame_t *frame, xlator_t *this,
-	     loc_t *loc, dict_t *xattr_req)
+               loc_t *loc, dict_t *xattr_req)
 {
         xlator_t     *hashed_subvol = NULL;
         xlator_t     *cached_subvol = NULL;
         xlator_t     *subvol = NULL;
         dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
+        dht_conf_t   *conf = NULL;
         int           ret    = -1;
         int           op_errno = -1;
-	dht_layout_t *layout = NULL;
-	int           i = 0;
-	int           call_cnt = 0;
+        dht_layout_t *layout = NULL;
+        int           i = 0;
+        int           call_cnt = 0;
 
 
         VALIDATE_OR_GOTO (frame, err);
@@ -247,40 +246,27 @@ switch_lookup (call_frame_t *frame, xlator_t *this,
         VALIDATE_OR_GOTO (loc->inode, err);
         VALIDATE_OR_GOTO (loc->path, err);
 
-	conf = this->private;
+        conf = this->private;
 
-        local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-        ret = loc_dup (loc, &local->loc);
-        if (ret == -1) {
-                op_errno = errno;
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "copying location failed for path=%s",
-                        loc->path);
+        local = dht_local_init (frame, loc, NULL, GF_FOP_LOOKUP);
+        if (!local) {
+                op_errno = ENOMEM;
                 goto err;
         }
 
-	if (xattr_req) {
-		local->xattr_req = dict_ref (xattr_req);
-	} else {
-		local->xattr_req = dict_new ();
-	}
+        if (xattr_req) {
+                local->xattr_req = dict_ref (xattr_req);
+        } else {
+                local->xattr_req = dict_new ();
+        }
 
-	hashed_subvol = dht_subvol_get_hashed (this, &local->loc);
-	cached_subvol = dht_subvol_get_cached (this, local->loc.inode);
+        hashed_subvol = dht_subvol_get_hashed (this, &local->loc);
+        cached_subvol = local->cached_subvol;
 
-	local->cached_subvol = cached_subvol;
-	local->hashed_subvol = hashed_subvol;
+        local->hashed_subvol = hashed_subvol;
 
         if (is_revalidate (loc)) {
-		local->layout = layout = dht_layout_get (this, loc->inode);
-
+                layout = local->layout;
                 if (!layout) {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "revalidate without cache. path=%s",
@@ -289,83 +275,80 @@ switch_lookup (call_frame_t *frame, xlator_t *this,
                         goto err;
                 }
 
-		if (layout->gen && (layout->gen < conf->gen)) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"incomplete layout failure for path=%s",
-				loc->path);
+                if (layout->gen && (layout->gen < conf->gen)) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "incomplete layout failure for path=%s",
+                                loc->path);
                         dht_layout_unref (this, local->layout);
-			goto do_fresh_lookup;
-		}
+                        goto do_fresh_lookup;
+                }
 
-		local->inode    = inode_ref (loc->inode);
-		local->ia_ino   = loc->inode->ino;
+                local->inode    = inode_ref (loc->inode);
 
-		local->call_cnt = layout->cnt;
-		call_cnt = local->call_cnt;
+                local->call_cnt = layout->cnt;
+                call_cnt = local->call_cnt;
 
-		/* NOTE: we don't require 'trusted.glusterfs.dht.linkto'
+                /* NOTE: we don't require 'trusted.glusterfs.dht.linkto'
                  * attribute, revalidates directly go to the cached-subvolume.
-		 */
-		ret = dict_set_uint32 (local->xattr_req,
-				       "trusted.glusterfs.dht", 4 * 4);
+                 */
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->xattr_name, 4 * 4);
                 if (ret < 0)
                         gf_log (this->name, GF_LOG_WARNING,
-                                "failed to set dict value for "
-                                "trusted.glusterfs.dht");
+                                "failed to set dict value for %s",
+                                conf->xattr_name);
 
-		for (i = 0; i < layout->cnt; i++) {
-			subvol = layout->list[i].xlator;
+                for (i = 0; i < layout->cnt; i++) {
+                        subvol = layout->list[i].xlator;
 
-			STACK_WIND (frame, dht_revalidate_cbk,
-				    subvol, subvol->fops->lookup,
-				    loc, local->xattr_req);
+                        STACK_WIND (frame, dht_revalidate_cbk,
+                                    subvol, subvol->fops->lookup,
+                                    loc, local->xattr_req);
 
-			if (!--call_cnt)
-				break;
-		}
-	} else {
+                        if (!--call_cnt)
+                                break;
+                }
+        } else {
         do_fresh_lookup:
-		ret = dict_set_uint32 (local->xattr_req,
-				       "trusted.glusterfs.dht", 4 * 4);
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->xattr_name, 4 * 4);
                 if (ret < 0)
                         gf_log (this->name, GF_LOG_WARNING,
-                                "failed to set dict value for "
-                                "trusted.glusterfs.dht");
-                
-		ret = dict_set_uint32 (local->xattr_req,
-				       "trusted.glusterfs.dht.linkto", 256);
+                                "failed to set dict value for %s",
+                                conf->xattr_name);
+
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->link_xattr_name, 256);
                 if (ret < 0)
-                        gf_log (this->name, GF_LOG_WARNING, 
-                                "failed to set dict value for " 
-                                "trusted.glusterfs.dht.linkto");
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "failed to set dict value for %s",
+                                conf->link_xattr_name);
 
                 if (!hashed_subvol) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"no subvolume in layout for path=%s, "
-				"checking on all the subvols to see if "
-				"it is a directory", loc->path);
- 			call_cnt        = conf->subvolume_cnt;
- 			local->call_cnt = call_cnt;
-
- 			local->layout = dht_layout_new (this,
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "no subvolume in layout for path=%s, "
+                                "checking on all the subvols to see if "
+                                "it is a directory", loc->path);
+                        call_cnt        = conf->subvolume_cnt;
+                        local->call_cnt = call_cnt;
+
+                        local->layout = dht_layout_new (this,
                                                         conf->subvolume_cnt);
- 			if (!local->layout) {
- 				op_errno = ENOMEM;
- 				gf_log (this->name, GF_LOG_ERROR,
- 					"Out of memory");
- 				goto err;
- 			}
-
-			for (i = 0; i < call_cnt; i++) {
- 				STACK_WIND (frame, dht_lookup_dir_cbk,
- 					    conf->subvolumes[i],
- 					    conf->subvolumes[i]->fops->lookup,
- 					    &local->loc, local->xattr_req);
- 			}
- 			return 0;
+                        if (!local->layout) {
+                                op_errno = ENOMEM;
+                                goto err;
+                        }
+
+                        for (i = 0; i < call_cnt; i++) {
+                                STACK_WIND (frame, dht_lookup_dir_cbk,
+                                            conf->subvolumes[i],
+                                            conf->subvolumes[i]->fops->lookup,
+                                            &local->loc, local->xattr_req);
+                        }
+                        return 0;
                 }
 
-		/*  */
+                /*  */
                 cached_subvol = get_switch_matching_subvol (loc->path, conf,
                                                             hashed_subvol);
                 if (cached_subvol == hashed_subvol) {
@@ -384,99 +367,88 @@ switch_lookup (call_frame_t *frame, xlator_t *this,
         return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-        DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
-	return 0;
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (lookup, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL);
+        return 0;
 }
 
 int
 switch_create_linkfile_create_cbk (call_frame_t *frame, void *cookie,
-				 xlator_t *this, int op_ret, int op_errno,
-                                 inode_t *inode, struct iatt *stbuf,
-                                 struct iatt *preparent,
-                                 struct iatt *postparent)
+                                   xlator_t *this, int op_ret, int op_errno,
+                                   inode_t *inode, struct iatt *stbuf,
+                                   struct iatt *preparent,
+                                   struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
+        dht_local_t  *local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	if (op_ret == -1)
-		goto err;
+        if (op_ret == -1)
+                goto err;
 
-	STACK_WIND (frame, dht_create_cbk,
-		    local->cached_subvol, local->cached_subvol->fops->create,
-		    &local->loc, local->flags, local->mode, local->fd,
-                    local->params);
+        STACK_WIND (frame, dht_create_cbk,
+                    local->cached_subvol, local->cached_subvol->fops->create,
+                    &local->loc, local->flags, local->mode, local->umask,
+                    local->fd, local->params);
 
-	return 0;
+        return 0;
 
- err:
-	DHT_STACK_UNWIND (create, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL, NULL);
-	return 0;
+err:
+        DHT_STACK_UNWIND (create, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL, NULL);
+        return 0;
 }
 
 int
 switch_create (call_frame_t *frame, xlator_t *this,
                loc_t *loc, int32_t flags, mode_t mode,
-               fd_t *fd, dict_t *params)
+               mode_t umask, fd_t *fd, dict_t *params)
 {
-	dht_local_t *local = NULL;
-	dht_conf_t  *conf  = NULL;
-	xlator_t    *subvol = NULL;
+        dht_local_t *local = NULL;
+        dht_conf_t  *conf  = NULL;
+        xlator_t    *subvol = NULL;
         xlator_t    *avail_subvol = NULL;
-	int          op_errno = -1;
-	int          ret = -1;
+        int          op_errno = -1;
 
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
-	conf  = this->private;
+        conf  = this->private;
 
         dht_get_du_info (frame, this, loc);
 
-        local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	subvol = dht_subvol_get_hashed (this, loc);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			loc->path);
-		op_errno = ENOENT;
-		goto err;
-	}
+        local = dht_local_init (frame, loc, fd, GF_FOP_CREATE);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = dht_subvol_get_hashed (this, loc);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        loc->path);
+                op_errno = ENOENT;
+                goto err;
+        }
 
         avail_subvol = get_switch_matching_subvol (loc->path, conf, subvol);
         if (dht_is_subvol_filled (this, avail_subvol)) {
                 avail_subvol =
-                        dht_free_disk_available_subvol (this, avail_subvol);
+                        dht_free_disk_available_subvol (this, avail_subvol,
+                                                        local);
         }
 
         if (subvol != avail_subvol) {
                 /* create a link file instead of actual file */
-                ret = loc_copy (&local->loc, loc);
-                if (ret == -1) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "Out of memory");
-                                op_errno = ENOMEM;
-                                goto err;
-                }
-
-                local->fd = fd_ref (fd);
                 local->mode = mode;
                 local->flags = flags;
-
+                local->umask = umask;
                 local->cached_subvol = avail_subvol;
-                dht_linkfile_create (frame,
-                                     switch_create_linkfile_create_cbk,
-                                     avail_subvol, subvol, loc);
+                dht_linkfile_create (frame, switch_create_linkfile_create_cbk,
+                                     this, avail_subvol, subvol, loc);
                 return 0;
         }
 
@@ -485,176 +457,143 @@ switch_create (call_frame_t *frame, xlator_t *this,
 
         STACK_WIND (frame, dht_create_cbk,
                     subvol, subvol->fops->create,
-                    loc, flags, mode, fd, params);
+                    loc, flags, mode, umask, fd, params);
 
         return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (create, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (create, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 int
 switch_mknod_linkfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                         int op_ret, int op_errno, inode_t *inode,
-                         struct iatt *stbuf, struct iatt *preparent,
-                         struct iatt *postparent)
+                           int op_ret, int op_errno, inode_t *inode,
+                           struct iatt *stbuf, struct iatt *preparent,
+                           struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-
-	local = frame->local;
+        dht_local_t  *local = NULL;
 
-	if (op_ret >= 0) {
-		STACK_WIND (frame, dht_newfile_cbk,
-			    local->cached_subvol,
-			    local->cached_subvol->fops->mknod,
-			    &local->loc, local->mode, local->rdev,
-                            local->params);
+        local = frame->local;
+        if (!local || !local->cached_subvol) {
+                op_errno = EINVAL;
+                op_ret = -1;
+                goto err;
+        }
 
-		return 0;
-	}
+        if (op_ret >= 0) {
+                STACK_WIND_COOKIE (frame, dht_newfile_cbk,
+                            (void *)local->cached_subvol, local->cached_subvol,
+                            local->cached_subvol->fops->mknod,
+                            &local->loc, local->mode, local->rdev,
+                            local->umask, local->params);
 
-	DHT_STACK_UNWIND (link, frame, op_ret, op_errno,
-                          inode, stbuf, preparent, postparent);
-	return 0;
+                return 0;
+        }
+err:
+        DHT_STACK_UNWIND (link, frame, op_ret, op_errno,
+                          inode, stbuf, preparent, postparent, xdata);
+        return 0;
 }
 
 
 int
-switch_mknod (call_frame_t *frame, xlator_t *this,
-              loc_t *loc, mode_t mode, dev_t rdev, dict_t *params)
+switch_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+              dev_t rdev, mode_t umask, dict_t *params)
 {
-	dht_local_t *local = NULL;
-	dht_conf_t  *conf  = NULL;
-	xlator_t    *subvol = NULL;
+        dht_local_t *local = NULL;
+        dht_conf_t  *conf  = NULL;
+        xlator_t    *subvol = NULL;
         xlator_t    *avail_subvol = NULL;
-	int          op_errno = -1;
-	int          ret = -1;
+        int          op_errno = -1;
 
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
-	conf  = this->private;
+        conf  = this->private;
 
         dht_get_du_info (frame, this, loc);
 
-        local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	subvol = dht_subvol_get_hashed (this, loc);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			loc->path);
-		op_errno = ENOENT;
-		goto err;
-	}
+        local = dht_local_init (frame, loc, NULL, GF_FOP_MKNOD);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        subvol = dht_subvol_get_hashed (this, loc);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        loc->path);
+                op_errno = ENOENT;
+                goto err;
+        }
 
         /* Consider the disksize in consideration */
         avail_subvol = get_switch_matching_subvol (loc->path, conf, subvol);
         if (dht_is_subvol_filled (this, avail_subvol)) {
                 avail_subvol =
-                        dht_free_disk_available_subvol (this, avail_subvol);
+                        dht_free_disk_available_subvol (this, avail_subvol,
+                                                        local);
         }
 
-	if (avail_subvol != subvol) {
-		/* Create linkfile first */
-		ret = loc_copy (&local->loc, loc);
-		if (ret == -1) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"Out of memory");
-			op_errno = ENOMEM;
-			goto err;
-		}
+        if (avail_subvol != subvol) {
+                /* Create linkfile first */
 
                 local->params = dict_ref (params);
-		local->mode = mode;
-		local->rdev = rdev;
-		local->cached_subvol = avail_subvol;
+                local->mode = mode;
+                local->umask = umask;
+                local->rdev = rdev;
+                local->cached_subvol = avail_subvol;
 
-		dht_linkfile_create (frame, switch_mknod_linkfile_cbk,
-                                     avail_subvol, subvol, loc);
-		return 0;
-	}
+                dht_linkfile_create (frame, switch_mknod_linkfile_cbk,
+                                     this, avail_subvol, subvol, loc);
+                return 0;
+        }
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"creating %s on %s", loc->path, subvol->name);
+        gf_log (this->name, GF_LOG_TRACE,
+                "creating %s on %s", loc->path, subvol->name);
 
-	STACK_WIND (frame, dht_newfile_cbk,
-		    subvol, subvol->fops->mknod,
-		    loc, mode, rdev, params);
+        STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol, subvol,
+                           subvol->fops->mknod, loc, mode, rdev, umask,
+                           params);
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (mknod, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
-int
-notify (xlator_t *this, int event, void *data, ...)
-{
-	int ret = -1;
-
-	ret = dht_notify (this, event, data);
-
-	return ret;
-}
-
 void
-fini (xlator_t *this)
+switch_fini (xlator_t *this)
 {
-        int                   i = 0;
         dht_conf_t           *conf = NULL;
         struct switch_struct *trav = NULL;
         struct switch_struct *prev = NULL;
 
-	conf = this->private;
+        conf = this->private;
 
         if (conf) {
                 trav = (struct switch_struct *)conf->private;
                 conf->private = NULL;
                 while (trav) {
-                        if (trav->array)
-                                GF_FREE (trav->array);
+                        GF_FREE (trav->array);
                         prev = trav;
                         trav = trav->next;
                         GF_FREE (prev);
                 }
-
-                if (conf->file_layouts) {
-                        for (i = 0; i < conf->subvolume_cnt; i++) {
-                                GF_FREE (conf->file_layouts[i]);
-                        }
-                        GF_FREE (conf->file_layouts);
-                }
-
-                if (conf->default_dir_layout)
-                        GF_FREE (conf->default_dir_layout);
-
-                if (conf->subvolumes)
-                        GF_FREE (conf->subvolumes);
-
-		if (conf->subvolume_status)
-			GF_FREE (conf->subvolume_status);
-
-                GF_FREE (conf);
         }
 
-	return;
+        dht_fini(this);
 }
 
 int
@@ -679,32 +618,32 @@ set_switch_pattern (xlator_t *this, dht_conf_t *conf,
         struct switch_struct        *switch_buf = NULL;
         struct switch_struct        *switch_opt = NULL;
         struct switch_struct        *trav = NULL;
-	struct switch_sched_array  *switch_buf_array = NULL;
-	xlator_list_t              *trav_xl = NULL;
+        struct switch_sched_array  *switch_buf_array = NULL;
+        xlator_list_t              *trav_xl = NULL;
 
         trav_xl = this->children;
-	while (trav_xl) {
-		index++;
-		trav_xl = trav_xl->next;
-	}
-	child_count = index;
-	switch_buf_array = GF_CALLOC ((index + 1),
+        while (trav_xl) {
+                index++;
+                trav_xl = trav_xl->next;
+        }
+        child_count = index;
+        switch_buf_array = GF_CALLOC ((index + 1),
                                       sizeof (struct switch_sched_array),
                                       gf_switch_mt_switch_sched_array);
         if (!switch_buf_array)
                 goto err;
 
-	trav_xl = this->children;
-	index = 0;
+        trav_xl = this->children;
+        index = 0;
 
-	while (trav_xl) {
-		switch_buf_array[index].xl = trav_xl->xlator;
-		switch_buf_array[index].eligible = 1;
-		trav_xl = trav_xl->next;
-		index++;
-	}
+        while (trav_xl) {
+                switch_buf_array[index].xl = trav_xl->xlator;
+                switch_buf_array[index].eligible = 1;
+                trav_xl = trav_xl->next;
+                index++;
+        }
 
-	/*  *jpg:child1,child2;*mpg:child3;*:child4,child5,child6 */
+        /*  *jpg:child1,child2;*mpg:child3;*:child4,child5,child6 */
 
         /* Get the pattern for considering switch case.
            "option block-size *avi:10MB" etc */
@@ -714,20 +653,24 @@ set_switch_pattern (xlator_t *this, dht_conf_t *conf,
                 dup_str = gf_strdup (switch_str);
                 switch_opt = GF_CALLOC (1, sizeof (struct switch_struct),
                                         gf_switch_mt_switch_struct);
-                if (!switch_opt)
+                if (!switch_opt) {
+                        GF_FREE (dup_str);
                         goto err;
+                }
 
                 pattern = strtok_r (dup_str, ":", &tmp_str1);
                 childs = strtok_r (NULL, ":", &tmp_str1);
                 if (strncmp (pattern, "*", 2) == 0) {
-                        gf_log ("switch", GF_LOG_NORMAL,
+                        gf_log ("switch", GF_LOG_INFO,
                                 "'*' pattern will be taken by default "
                                 "for all the unconfigured child nodes,"
                                 " hence neglecting current option");
                         switch_str = strtok_r (NULL, ";", &tmp_str);
+                        GF_FREE (switch_opt);
                         GF_FREE (dup_str);
                         continue;
                 }
+                GF_FREE (dup_str);
                 memcpy (switch_opt->path_pattern, pattern, strlen (pattern));
                 if (childs) {
                         dup_childs = gf_strdup (childs);
@@ -749,7 +692,7 @@ set_switch_pattern (xlator_t *this, dht_conf_t *conf,
                         child = strtok_r (childs, ",", &tmp1);
                         switch_opt->num_child = idx;
                         switch_opt->array = GF_CALLOC (1, (idx *
-                                                       sizeof (struct switch_sched_array)),
+                                                           sizeof (struct switch_sched_array)),
                                                        gf_switch_mt_switch_sched_array);
                         if (!switch_opt->array)
                                 goto err;
@@ -784,8 +727,7 @@ set_switch_pattern (xlator_t *this, dht_conf_t *conf,
                                 "option in unify volume. Exiting");
                         goto err;
                 }
-                GF_FREE (dup_str);
-                
+
                 /* Link it to the main structure */
                 if (switch_buf) {
                         /* there are already few entries */
@@ -797,75 +739,77 @@ set_switch_pattern (xlator_t *this, dht_conf_t *conf,
                         /* First entry */
                         switch_buf = switch_opt;
                 }
+                switch_opt = NULL;
                 switch_str = strtok_r (NULL, ";", &tmp_str);
         }
 
-	/* Now, all the pattern based considerations done, so for all the
-	 * remaining pattern, '*' to all the remaining child nodes
-	 */
-	{
-		for (index=0; index < child_count; index++) {
-			/* check for considered flag */
-			if (switch_buf_array[index].considered)
-				continue;
-			flag++;
-		}
-		if (!flag) {
-			gf_log ("switch", GF_LOG_ERROR,
-				"No nodes left for pattern '*'. Exiting");
-			goto err;
-		}
-		switch_opt = GF_CALLOC (1, sizeof (struct switch_struct),
+        /* Now, all the pattern based considerations done, so for all the
+         * remaining pattern, '*' to all the remaining child nodes
+         */
+        {
+                for (index=0; index < child_count; index++) {
+                        /* check for considered flag */
+                        if (switch_buf_array[index].considered)
+                                continue;
+                        flag++;
+                }
+                if (!flag) {
+                        gf_log ("switch", GF_LOG_ERROR,
+                                "No nodes left for pattern '*'. Exiting");
+                        goto err;
+                }
+                switch_opt = GF_CALLOC (1, sizeof (struct switch_struct),
                                         gf_switch_mt_switch_struct);
                 if (!switch_opt)
                         goto err;
 
-		/* Add the '*' pattern to the array */
-		memcpy (switch_opt->path_pattern, "*", 2);
-		switch_opt->num_child = flag;
-		switch_opt->array =
-			GF_CALLOC (1, 
+                /* Add the '*' pattern to the array */
+                memcpy (switch_opt->path_pattern, "*", 2);
+                switch_opt->num_child = flag;
+                switch_opt->array =
+                        GF_CALLOC (1,
                                    flag * sizeof (struct switch_sched_array),
                                    gf_switch_mt_switch_sched_array);
                 if (!switch_opt->array)
                         goto err;
-		flag = 0;
-		for (index=0; index < child_count; index++) {
-			/* check for considered flag */
-			if (switch_buf_array[index].considered)
-				continue;
-			gf_log ("switch", GF_LOG_DEBUG,
-				"'%s' pattern will be scheduled to \"%s\"",
-				switch_opt->path_pattern,
-				switch_buf_array[index].xl->name);
-			switch_opt->array[flag].xl =
-				switch_buf_array[index].xl;
-			switch_buf_array[index].considered = 1;
-			flag++;
+                flag = 0;
+                for (index=0; index < child_count; index++) {
+                        /* check for considered flag */
+                        if (switch_buf_array[index].considered)
+                                continue;
+                        gf_log ("switch", GF_LOG_DEBUG,
+                                "'%s' pattern will be scheduled to \"%s\"",
+                                switch_opt->path_pattern,
+                                switch_buf_array[index].xl->name);
+                        switch_opt->array[flag].xl =
+                                switch_buf_array[index].xl;
+                        switch_buf_array[index].considered = 1;
+                        flag++;
                 }
-		if (switch_buf) {
-			/* there are already few entries */
-			trav = switch_buf;
-			while (trav->next)
-				trav = trav->next;
-			trav->next = switch_opt;
-		} else {
-			/* First entry */
-			switch_buf = switch_opt;
-		}
-	}
+                if (switch_buf) {
+                        /* there are already few entries */
+                        trav = switch_buf;
+                        while (trav->next)
+                                trav = trav->next;
+                        trav->next = switch_opt;
+                } else {
+                        /* First entry */
+                        switch_buf = switch_opt;
+                }
+                switch_opt = NULL;
+        }
         /* */
         conf->private = switch_buf;
 
         return 0;
 err:
+        GF_FREE (switch_buf_array);
+        GF_FREE (switch_opt);
+
         if (switch_buf) {
-                if (switch_buf_array)
-                        GF_FREE (switch_buf_array);
                 trav = switch_buf;
                 while (trav) {
-                        if (trav->array)
-                                GF_FREE (trav->array);
+                        GF_FREE (trav->array);
                         switch_opt = trav;
                         trav = trav->next;
                         GF_FREE (switch_opt);
@@ -875,192 +819,86 @@ err:
 }
 
 
-int
-init (xlator_t *this)
+int32_t
+switch_init (xlator_t *this)
 {
         dht_conf_t            *conf = NULL;
-	data_t                *data = NULL;
-	char                  *temp_str = NULL;
+        data_t                *data = NULL;
         int                    ret = -1;
-        int                    i = 0;
-	uint32_t               temp_free_disk = 0;
-
-	if (!this->children) {
-		gf_log (this->name, GF_LOG_CRITICAL,
-			"SWITCH needs more than one subvolume");
-		return -1;
-	}
-
-	if (!this->parents) {
-		gf_log (this->name, GF_LOG_WARNING,
-			"dangling volume. check volfile");
-	}
-
-        conf = GF_CALLOC (1, sizeof (*conf), gf_switch_mt_dht_conf_t);
-        if (!conf) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
-        }
 
-        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_ON;
-	if (dict_get_str (this->options, "lookup-unhashed", &temp_str) == 0) {
-                /* If option is not "auto", other options _should_ be boolean */
-                if (strcasecmp (temp_str, "auto"))
-                        gf_string2boolean (temp_str, &conf->search_unhashed);
-                else
-                        conf->search_unhashed = GF_DHT_LOOKUP_UNHASHED_AUTO;
-	}
-
-	conf->unhashed_sticky_bit = 0;
-	if (dict_get_str (this->options, "unhashed-sticky-bit",
-                          &temp_str) == 0) {
-	        gf_string2boolean (temp_str, &conf->unhashed_sticky_bit);
-	}
-
-        conf->min_free_disk = 10;
-        conf->disk_unit = 'p';
-
-        if (dict_get_str (this->options, "min-free-disk",
-                          &temp_str) == 0) {
-                if (gf_string2percent (temp_str,
-                                       &temp_free_disk) == 0) {
-                        if (temp_free_disk > 100) {
-                                gf_string2bytesize (temp_str,
-                                                        &conf->min_free_disk);
-                                conf->disk_unit = 'b';
-                        } else {
-                                conf->min_free_disk = (uint64_t)temp_free_disk;
-                                conf->disk_unit = 'p';
-                        }
-                } else {
-                        gf_string2bytesize (temp_str,
-                                                &conf->min_free_disk);
-                        conf->disk_unit = 'b';
-                }
+        ret = dht_init(this);
+        if (ret) {
+                return ret;
         }
+        conf = this->private;
 
-	data = dict_get (this->options, "pattern.switch.case");
-	if (data) {
+        data = dict_get (this->options, "pattern.switch.case");
+        if (data) {
                 /* TODO: */
                 ret = set_switch_pattern (this, conf, data->data);
                 if (ret) {
                         goto err;
                 }
-	}
-
-        ret = dht_init_subvolumes (this, conf);
-        if (ret == -1) {
-                goto err;
-        }
-
-        ret = dht_layouts_init (this, conf);
-        if (ret == -1) {
-                goto err;
-        }
-
-	LOCK_INIT (&conf->subvolume_lock);
-	LOCK_INIT (&conf->layout_lock);
-
-	conf->gen = 1;
-
-        conf->du_stats = GF_CALLOC (conf->subvolume_cnt, sizeof (dht_du_t),
-                                    gf_switch_mt_dht_du_t);
-        if (!conf->du_stats) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
         }
 
         this->private = conf;
-
         return 0;
 
 err:
-        if (conf) {
-                if (conf->file_layouts) {
-                        for (i = 0; i < conf->subvolume_cnt; i++) {
-                                GF_FREE (conf->file_layouts[i]);
-                        }
-                        GF_FREE (conf->file_layouts);
-                }
-
-                if (conf->default_dir_layout)
-                        GF_FREE (conf->default_dir_layout);
-
-                if (conf->subvolumes)
-                        GF_FREE (conf->subvolumes);
-
-		if (conf->subvolume_status)
-			GF_FREE (conf->subvolume_status);
-
-                if (conf->du_stats)
-                        GF_FREE (conf->du_stats);
-
-                GF_FREE (conf);
-        }
-
+        dht_fini(this);
         return -1;
 }
 
 
-struct xlator_fops fops = {
-	.lookup      = switch_lookup,
-	.create      = switch_create,
-	.mknod       = switch_mknod,
-
-	.stat        = dht_stat,
-	.fstat       = dht_fstat,
-	.truncate    = dht_truncate,
-	.ftruncate   = dht_ftruncate,
-	.access      = dht_access,
-	.readlink    = dht_readlink,
-	.setxattr    = dht_setxattr,
-	.getxattr    = dht_getxattr,
-	.removexattr = dht_removexattr,
-	.open        = dht_open,
-	.readv       = dht_readv,
-	.writev      = dht_writev,
-	.flush       = dht_flush,
-	.fsync       = dht_fsync,
-	.statfs      = dht_statfs,
-	.lk          = dht_lk,
-	.opendir     = dht_opendir,
-	.readdir     = dht_readdir,
-	.readdirp    = dht_readdirp,
-	.fsyncdir    = dht_fsyncdir,
-	.symlink     = dht_symlink,
-	.unlink      = dht_unlink,
-	.link        = dht_link,
-	.mkdir       = dht_mkdir,
-	.rmdir       = dht_rmdir,
-	.rename      = dht_rename,
-	.inodelk     = dht_inodelk,
-	.finodelk    = dht_finodelk,
-	.entrylk     = dht_entrylk,
-	.fentrylk    = dht_fentrylk,
-	.xattrop     = dht_xattrop,
-	.fxattrop    = dht_fxattrop,
-        .setattr     = dht_setattr,
+class_methods_t class_methods = {
+        .init           = switch_init,
+        .fini           = switch_fini,
+        .reconfigure    = dht_reconfigure,
+        .notify         = dht_notify
 };
 
 
-struct xlator_cbks cbks = {
-	.forget     = dht_forget
+struct xlator_fops fops = {
+        .lookup      = switch_lookup,
+        .create      = switch_create,
+        .mknod       = switch_mknod,
+
+        .stat        = dht_stat,
+        .fstat       = dht_fstat,
+        .truncate    = dht_truncate,
+        .ftruncate   = dht_ftruncate,
+        .access      = dht_access,
+        .readlink    = dht_readlink,
+        .setxattr    = dht_setxattr,
+        .getxattr    = dht_getxattr,
+        .removexattr = dht_removexattr,
+        .open        = dht_open,
+        .readv       = dht_readv,
+        .writev      = dht_writev,
+        .flush       = dht_flush,
+        .fsync       = dht_fsync,
+        .statfs      = dht_statfs,
+        .lk          = dht_lk,
+        .opendir     = dht_opendir,
+        .readdir     = dht_readdir,
+        .readdirp    = dht_readdirp,
+        .fsyncdir    = dht_fsyncdir,
+        .symlink     = dht_symlink,
+        .unlink      = dht_unlink,
+        .link        = dht_link,
+        .mkdir       = dht_mkdir,
+        .rmdir       = dht_rmdir,
+        .rename      = dht_rename,
+        .inodelk     = dht_inodelk,
+        .finodelk    = dht_finodelk,
+        .entrylk     = dht_entrylk,
+        .fentrylk    = dht_fentrylk,
+        .xattrop     = dht_xattrop,
+        .fxattrop    = dht_fxattrop,
+        .setattr     = dht_setattr,
 };
 
 
-struct volume_options options[] = {
-        { .key  = {"lookup-unhashed"},
-          .value = {"auto", "yes", "no", "enable", "disable", "1", "0",
-                    "on", "off"},
-	  .type = GF_OPTION_TYPE_STR
-	},
-	{ .key  = {"pattern.switch.case"},
-	  .type = GF_OPTION_TYPE_ANY
-	},
-        { .key  = {"min-free-disk"},
-          .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
-        },
-	{ .key  = {NULL} },
+struct xlator_cbks cbks = {
+        .forget     = dht_forget
 };
diff --git a/xlators/cluster/dht/src/unittest/dht_layout_mock.c b/xlators/cluster/dht/src/unittest/dht_layout_mock.c
new file mode 100644
index 000000000..aa19ddc57
--- /dev/null
+++ b/xlators/cluster/dht/src/unittest/dht_layout_mock.c
@@ -0,0 +1,63 @@
+/*
+  Copyright (c) 2014 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "xlator.h"
+#include "dht-common.h"
+#include "byte-order.h"
+
+int
+dht_hash_compute (xlator_t *this, int type, const char *name, uint32_t *hash_p)
+{
+    return 0;
+}
+
+int
+dht_inode_ctx_layout_get (inode_t *inode, xlator_t *this, dht_layout_t **layout)
+{
+    return 0;
+}
+
+int
+dht_inode_ctx_layout_set (inode_t *inode, xlator_t *this,
+                          dht_layout_t *layout_int)
+{
+    return 0;
+}
+
+int
+dict_get_ptr (dict_t *this, char *key, void **ptr)
+{
+    return 0;
+}
+
+int
+dict_get_ptr_and_len (dict_t *this, char *key, void **ptr, int *len)
+{
+    return 0;
+}
+
+int _gf_log (const char *domain, const char *file,
+             const char *function, int32_t line, gf_loglevel_t level,
+             const char *fmt, ...)
+{
+    return 0;
+}
+
+int _gf_log_callingfn (const char *domain, const char *file,
+                       const char *function, int32_t line, gf_loglevel_t level,
+                       const char *fmt, ...)
+{
+    return 0;
+}
diff --git a/xlators/cluster/dht/src/unittest/dht_layout_unittest.c b/xlators/cluster/dht/src/unittest/dht_layout_unittest.c
new file mode 100644
index 000000000..b5233d235
--- /dev/null
+++ b/xlators/cluster/dht/src/unittest/dht_layout_unittest.c
@@ -0,0 +1,124 @@
+/*
+  Copyright (c) 2008-2014 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#include "dht-common.h"
+#include "logging.h"
+#include "xlator.h"
+
+#include <stdarg.h>
+#include <stddef.h>
+#include <setjmp.h>
+#include <inttypes.h>
+#include <cmockery/pbc.h>
+#include <cmockery/cmockery.h>
+
+/*
+ * Helper functions
+ */
+
+static xlator_t *
+helper_xlator_init(uint32_t num_types)
+{
+    xlator_t *xl;
+    int i, ret;
+
+    REQUIRE(num_types > 0);
+
+    xl = test_calloc(1, sizeof(xlator_t));
+    assert_non_null(xl);
+    xl->mem_acct.num_types = num_types;
+    xl->mem_acct.rec = test_calloc(num_types, sizeof(struct mem_acct_rec));
+    assert_non_null(xl->mem_acct.rec);
+
+    xl->ctx = test_calloc(1, sizeof(glusterfs_ctx_t));
+    assert_non_null(xl->ctx);
+
+    for (i = 0; i < num_types; i++) {
+            ret = LOCK_INIT(&(xl->mem_acct.rec[i].lock));
+            assert_false(ret);
+    }
+
+    ENSURE(num_types == xl->mem_acct.num_types);
+    ENSURE(NULL != xl);
+
+    return xl;
+}
+
+static int
+helper_xlator_destroy(xlator_t *xl)
+{
+    int i, ret;
+
+    for (i = 0; i < xl->mem_acct.num_types; i++) {
+            ret = LOCK_DESTROY(&(xl->mem_acct.rec[i].lock));
+            assert_int_equal(ret, 0);
+    }
+
+    free(xl->mem_acct.rec);
+    free(xl->ctx);
+    free(xl);
+    return 0;
+}
+
+/*
+ * Unit tests
+ */
+static void
+test_dht_layout_new(void **state)
+{
+    xlator_t *xl;
+    dht_layout_t *layout;
+    dht_conf_t   *conf;
+    int cnt;
+
+    expect_assert_failure(dht_layout_new(NULL, 0));
+    expect_assert_failure(dht_layout_new((xlator_t *)0x12345, -1));
+    xl = helper_xlator_init(10);
+
+    // xl->private is NULL
+    assert_null(xl->private);
+    cnt = 100;
+    layout = dht_layout_new(xl, cnt);
+    assert_non_null(layout);
+    assert_int_equal(layout->type, DHT_HASH_TYPE_DM);
+    assert_int_equal(layout->cnt, cnt);
+    assert_int_equal(layout->ref, 1);
+    assert_int_equal(layout->gen, 0);
+    assert_int_equal(layout->spread_cnt, 0);
+    free(layout);
+
+    // xl->private is not NULL
+    cnt = 110;
+    conf = (dht_conf_t *)test_calloc(1, sizeof(dht_conf_t));
+    assert_non_null(conf);
+    conf->dir_spread_cnt = 12345;
+    conf->gen = -123;
+    xl->private = conf;
+
+    layout = dht_layout_new(xl, cnt);
+    assert_non_null(layout);
+    assert_int_equal(layout->type, DHT_HASH_TYPE_DM);
+    assert_int_equal(layout->cnt, cnt);
+    assert_int_equal(layout->ref, 1);
+    assert_int_equal(layout->gen, conf->gen);
+    assert_int_equal(layout->spread_cnt, conf->dir_spread_cnt);
+    free(layout);
+
+    free(conf);
+    helper_xlator_destroy(xl);
+}
+
+int main(void) {
+    const UnitTest tests[] = {
+        unit_test(test_dht_layout_new),
+    };
+
+    return run_tests(tests, "xlator_dht_layout");
+}
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 7c361547a..19be1ed27 100644
--- a/xlators/cluster/ha/src/ha-helpers.c
+++ b/xlators/cluster/ha/src/ha-helpers.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #include "xlator.h"
 #include "call-stub.h"
 #include "defaults.h"
diff --git a/xlators/cluster/ha/src/ha-mem-types.h b/xlators/cluster/ha/src/ha-mem-types.h
index b460588aa..e5e97d237 100644
--- a/xlators/cluster/ha/src/ha-mem-types.h
+++ b/xlators/cluster/ha/src/ha-mem-types.h
@@ -1,24 +1,13 @@
-
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
 
-
 #ifndef __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 7bb1824a1..3eccb516b 100644
--- a/xlators/cluster/ha/src/ha.c
+++ b/xlators/cluster/ha/src/ha.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 /* generate errors randomly, code is simple now, better alogorithm
  * can be written to decide what error to be returned and when
  */
@@ -1876,13 +1866,9 @@ err:
         }
 
         if (hafdp) {
-                if (hafdp->fdstate) {
-                        GF_FREE (hafdp->fdstate);
-                }
+                GF_FREE (hafdp->fdstate);
 
-                if (hafdp->path) {
-                        GF_FREE (hafdp->path);
-                }
+                GF_FREE (hafdp->path);
 
                 GF_FREE (hafdp);
         }
diff --git a/xlators/cluster/ha/src/ha.h b/xlators/cluster/ha/src/ha.h
index d6a519fa8..e2ed7eaa6 100644
--- a/xlators/cluster/ha/src/ha.h
+++ b/xlators/cluster/ha/src/ha.h
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef __HA_H_
 #define __HA_H_
 
diff --git a/xlators/cluster/map/src/Makefile.am b/xlators/cluster/map/src/Makefile.am
index 26e19137a..a278b05e2 100644
--- a/xlators/cluster/map/src/Makefile.am
+++ b/xlators/cluster/map/src/Makefile.am
@@ -1,15 +1,16 @@
 xlator_LTLIBRARIES = map.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/testing/cluster
 
-map_la_LDFLAGS = -module -avoidversion 
+map_la_LDFLAGS = -module -avoid-version 
 
 map_la_SOURCES = map.c map-helper.c
 map_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
 noinst_HEADERS = map.h
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	    -I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/cluster/map/src/map-helper.c b/xlators/cluster/map/src/map-helper.c
index 24ab6eb50..851397b68 100644
--- a/xlators/cluster/map/src/map-helper.c
+++ b/xlators/cluster/map/src/map-helper.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2009-2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2009-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
diff --git a/xlators/cluster/map/src/map-mem-types.h b/xlators/cluster/map/src/map-mem-types.h
index 23c798dea..3e89f4736 100644
--- a/xlators/cluster/map/src/map-mem-types.h
+++ b/xlators/cluster/map/src/map-mem-types.h
@@ -1,24 +1,13 @@
-
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
 
-
 #ifndef __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 73186e8cb..6150a33ce 100644
--- a/xlators/cluster/map/src/map.c
+++ b/xlators/cluster/map/src/map.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -2375,8 +2365,7 @@ fini (xlator_t *this)
 	priv = this->private;
 
 	if (priv) {
-		if (priv->xlarray)
-			GF_FREE (priv->xlarray);
+		GF_FREE (priv->xlarray);
 
 		trav_map = priv->map;
 		while (trav_map) {
diff --git a/xlators/cluster/map/src/map.h b/xlators/cluster/map/src/map.h
index 44ba3ee12..7703a543e 100644
--- a/xlators/cluster/map/src/map.h
+++ b/xlators/cluster/map/src/map.h
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef __MAP_H__
 #define __MAP_H__
 
diff --git a/xlators/cluster/stripe/src/Makefile.am b/xlators/cluster/stripe/src/Makefile.am
index 6d4fae4d2..2d151422a 100644
--- a/xlators/cluster/stripe/src/Makefile.am
+++ b/xlators/cluster/stripe/src/Makefile.am
@@ -2,15 +2,19 @@
 xlator_LTLIBRARIES = stripe.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/cluster
 
-stripe_la_LDFLAGS = -module -avoidversion
+stripe_la_LDFLAGS = -module -avoid-version
+
+stripe_la_SOURCES = stripe.c stripe-helpers.c \
+	$(top_builddir)/xlators/lib/src/libxlator.c
 
-stripe_la_SOURCES = stripe.c
 stripe_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-noinst_HEADERS = stripe.h stripe-mem-types.h
+noinst_HEADERS = stripe.h stripe-mem-types.h $(top_builddir)/xlators/lib/src/libxlator.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
+	-I$(top_srcdir)/xlators/lib/src
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/cluster/stripe/src/stripe-helpers.c b/xlators/cluster/stripe/src/stripe-helpers.c
new file mode 100644
index 000000000..a047d4a2e
--- /dev/null
+++ b/xlators/cluster/stripe/src/stripe-helpers.c
@@ -0,0 +1,677 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#include <fnmatch.h>
+
+#include "stripe.h"
+#include "byte-order.h"
+#include "mem-types.h"
+
+void
+stripe_local_wipe (stripe_local_t *local)
+{
+        if (!local)
+                goto out;
+
+        loc_wipe (&local->loc);
+        loc_wipe (&local->loc2);
+
+        if (local->fd)
+                fd_unref (local->fd);
+
+        if (local->inode)
+                inode_unref (local->inode);
+
+        if (local->xattr)
+                dict_unref (local->xattr);
+
+        if (local->xdata)
+                dict_unref (local->xdata);
+
+out:
+        return;
+}
+
+
+
+int
+stripe_aggregate (dict_t *this, char *key, data_t *value, void *data)
+{
+        dict_t  *dst  = NULL;
+        int64_t *ptr  = 0, *size = NULL;
+        int32_t  ret  = -1;
+
+        dst = data;
+
+        if (strcmp (key, GF_XATTR_QUOTA_SIZE_KEY) == 0) {
+                ret = dict_get_bin (dst, key, (void **)&size);
+                if (ret < 0) {
+                        size = GF_CALLOC (1, sizeof (int64_t),
+                                          gf_common_mt_char);
+                        if (size == NULL) {
+                                gf_log ("stripe", GF_LOG_WARNING,
+                                        "memory allocation failed");
+                                goto out;
+                        }
+                        ret = dict_set_bin (dst, key, size, sizeof (int64_t));
+                        if (ret < 0) {
+                                gf_log ("stripe", GF_LOG_WARNING,
+                                        "stripe aggregate dict set failed");
+                                GF_FREE (size);
+                                goto out;
+                        }
+                }
+
+                ptr = data_to_bin (value);
+                if (ptr == NULL) {
+                        gf_log ("stripe", GF_LOG_WARNING, "data to bin failed");
+                        goto out;
+                }
+
+                *size = hton64 (ntoh64 (*size) + ntoh64 (*ptr));
+        } else if (strcmp (key, GF_CONTENT_KEY)) {
+                /* No need to aggregate 'CONTENT' data */
+                ret = dict_set (dst, key, value);
+                if (ret)
+                        gf_log ("stripe", GF_LOG_WARNING, "xattr dict set failed");
+        }
+
+out:
+        return 0;
+}
+
+
+void
+stripe_aggregate_xattr (dict_t *dst, dict_t *src)
+{
+        if ((dst == NULL) || (src == NULL)) {
+                goto out;
+        }
+
+        dict_foreach (src, stripe_aggregate, dst);
+out:
+        return;
+}
+
+
+int32_t
+stripe_xattr_aggregate (char *buffer, stripe_local_t *local, int32_t *total)
+{
+        int32_t              i     = 0;
+        int32_t              ret   = -1;
+        int32_t              len   = 0;
+        char                *sbuf  = NULL;
+        stripe_xattr_sort_t *xattr = NULL;
+
+        if (!buffer || !local || !local->xattr_list)
+                goto out;
+
+        sbuf = buffer;
+
+        for (i = 0; i < local->nallocs; i++) {
+                xattr = local->xattr_list + i;
+                len = xattr->xattr_len;
+
+                if (len && xattr && xattr->xattr_value) {
+                        memcpy (buffer, xattr->xattr_value, len);
+                        buffer += len;
+                        *buffer++ = ' ';
+                }
+        }
+
+        *--buffer = '\0';
+        if (total)
+                *total = buffer - sbuf;
+        ret = 0;
+
+ out:
+        return ret;
+}
+
+int32_t
+stripe_free_xattr_str (stripe_local_t *local)
+{
+        int32_t              i     = 0;
+        int32_t              ret   = -1;
+        stripe_xattr_sort_t *xattr = NULL;
+
+        if (!local || !local->xattr_list)
+                goto out;
+
+        for (i = 0; i < local->nallocs; i++) {
+                xattr = local->xattr_list + i;
+
+                if (xattr && xattr->xattr_value)
+                        GF_FREE (xattr->xattr_value);
+        }
+
+        ret = 0;
+ out:
+        return ret;
+}
+
+
+int32_t
+stripe_fill_lockinfo_xattr (xlator_t *this, stripe_local_t *local,
+                            void **xattr_serz)
+{
+        int32_t              ret   = -1, i = 0, len = 0;
+        dict_t              *tmp1  = NULL, *tmp2 = NULL;
+        char                *buf   = NULL;
+        stripe_xattr_sort_t *xattr = NULL;
+
+        if (xattr_serz == NULL) {
+                goto out;
+        }
+
+        tmp2 = dict_new ();
+
+        if (tmp2 == NULL) {
+                goto out;
+        }
+
+        for (i = 0; i < local->nallocs; i++) {
+                xattr = local->xattr_list + i;
+                len = xattr->xattr_len;
+
+                if (len && xattr && xattr->xattr_value) {
+                        ret = dict_reset (tmp2);
+                        if (ret < 0) {
+                                gf_log (this->name, GF_LOG_DEBUG,
+                                        "dict_reset failed (%s)",
+                                        strerror (-ret));
+                        }
+
+                        ret = dict_unserialize (xattr->xattr_value,
+                                                xattr->xattr_len,
+                                                &tmp2);
+                        if (ret < 0) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "dict_unserialize failed (%s)",
+                                        strerror (-ret));
+                                ret = -1;
+                                goto out;
+                        }
+
+                        tmp1 = dict_copy (tmp2, tmp1);
+                        if (tmp1 == NULL) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "dict_copy failed (%s)",
+                                        strerror (-ret));
+                                ret = -1;
+                                goto out;
+                        }
+                }
+        }
+
+        len = dict_serialized_length (tmp1);
+        if (len > 0) {
+                buf = GF_CALLOC (1, len, gf_common_mt_dict_t);
+                if (buf == NULL) {
+                        ret = -1;
+                        goto out;
+                }
+
+                ret = dict_serialize (tmp1, buf);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "dict_serialize failed (%s)", strerror (-ret));
+                        GF_FREE(buf);
+                        ret = -1;
+                        goto out;
+                }
+
+                *xattr_serz = buf;
+        }
+
+        ret = 0;
+out:
+        if (tmp1 != NULL) {
+                dict_unref (tmp1);
+        }
+
+        if (tmp2 != NULL) {
+                dict_unref (tmp2);
+        }
+
+        return ret;
+}
+
+
+int32_t
+stripe_fill_pathinfo_xattr (xlator_t *this, stripe_local_t *local,
+                            char **xattr_serz)
+{
+        int      ret             = -1;
+        int32_t  padding         = 0;
+        int32_t  tlen            = 0;
+        char stripe_size_str[20] = {0,};
+        char    *pathinfo_serz   = NULL;
+
+        if (!local) {
+                gf_log (this->name, GF_LOG_ERROR, "Possible NULL deref");
+                goto out;
+        }
+
+        (void) snprintf (stripe_size_str, 20, "%ld",
+                         (local->fctx) ? local->fctx->stripe_size : 0);
+
+        /* extra bytes for decorations (brackets and <>'s) */
+        padding = strlen (this->name) + strlen (STRIPE_PATHINFO_HEADER)
+                + strlen (stripe_size_str) + 7;
+        local->xattr_total_len += (padding + 2);
+
+        pathinfo_serz = GF_CALLOC (local->xattr_total_len, sizeof (char),
+                                   gf_common_mt_char);
+        if (!pathinfo_serz)
+                goto out;
+
+        /* xlator info */
+        (void) sprintf (pathinfo_serz, "(<"STRIPE_PATHINFO_HEADER"%s:[%s]> ",
+                        this->name, stripe_size_str);
+
+        ret = stripe_xattr_aggregate (pathinfo_serz + padding, local, &tlen);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Cannot aggregate pathinfo list");
+                GF_FREE(pathinfo_serz);
+                goto out;
+        }
+
+        *(pathinfo_serz + padding + tlen) = ')';
+        *(pathinfo_serz + padding + tlen + 1) = '\0';
+
+        *xattr_serz = pathinfo_serz;
+
+        ret = 0;
+ out:
+        return ret;
+}
+
+/**
+ * stripe_get_matching_bs - Get the matching block size for the given path.
+ */
+int32_t
+stripe_get_matching_bs (const char *path, stripe_private_t *priv)
+{
+        struct stripe_options *trav       = NULL;
+        uint64_t               block_size = 0;
+
+        GF_VALIDATE_OR_GOTO ("stripe", priv, out);
+        GF_VALIDATE_OR_GOTO ("stripe", path, out);
+
+        LOCK (&priv->lock);
+        {
+                block_size = priv->block_size;
+                trav = priv->pattern;
+                while (trav) {
+                        if (!fnmatch (trav->path_pattern, path, FNM_NOESCAPE)) {
+                                block_size = trav->block_size;
+                                break;
+                        }
+                        trav = trav->next;
+                }
+        }
+        UNLOCK (&priv->lock);
+
+out:
+        return block_size;
+}
+
+int32_t
+stripe_ctx_handle (xlator_t *this, call_frame_t *prev, stripe_local_t *local,
+                   dict_t *dict)
+{
+        char            key[256]       = {0,};
+        data_t         *data            = NULL;
+        int32_t         index           = 0;
+        stripe_private_t *priv          = NULL;
+
+        priv = this->private;
+
+
+        if (!local->fctx) {
+                local->fctx =  GF_CALLOC (1, sizeof (stripe_fd_ctx_t),
+                                         gf_stripe_mt_stripe_fd_ctx_t);
+                if (!local->fctx) {
+                        local->op_errno = ENOMEM;
+                        local->op_ret = -1;
+                        goto out;
+                }
+
+                local->fctx->static_array = 0;
+        }
+        /* Stripe block size */
+        sprintf (key, "trusted.%s.stripe-size", this->name);
+        data = dict_get (dict, key);
+        if (!data) {
+                local->xattr_self_heal_needed = 1;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to get stripe-size");
+                goto out;
+        } else {
+                if (!local->fctx->stripe_size) {
+                        local->fctx->stripe_size =
+                                     data_to_int64 (data);
+                }
+
+                if (local->fctx->stripe_size != data_to_int64 (data)) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "stripe-size mismatch in blocks");
+                        local->xattr_self_heal_needed = 1;
+                }
+        }
+
+        /* Stripe count */
+        sprintf (key, "trusted.%s.stripe-count", this->name);
+        data = dict_get (dict, key);
+
+        if (!data) {
+                local->xattr_self_heal_needed = 1;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to get stripe-count");
+                goto out;
+        }
+        if (!local->fctx->xl_array) {
+                local->fctx->stripe_count = data_to_int32 (data);
+                if (!local->fctx->stripe_count) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "error with stripe-count xattr");
+                        local->op_ret   = -1;
+                        local->op_errno = EIO;
+                        goto out;
+                }
+
+                local->fctx->xl_array = GF_CALLOC (local->fctx->stripe_count,
+                                                   sizeof (xlator_t *),
+                                                   gf_stripe_mt_xlator_t);
+
+                if (!local->fctx->xl_array) {
+                        local->op_errno = ENOMEM;
+                        local->op_ret   = -1;
+                        goto out;
+                }
+        }
+        if (local->fctx->stripe_count != data_to_int32 (data)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "error with stripe-count xattr (%d != %d)",
+                        local->fctx->stripe_count, data_to_int32 (data));
+                local->op_ret   = -1;
+                local->op_errno = EIO;
+                goto out;
+        }
+
+        /* index */
+        sprintf (key, "trusted.%s.stripe-index", this->name);
+        data = dict_get (dict, key);
+        if (!data) {
+                local->xattr_self_heal_needed = 1;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Failed to get stripe-index");
+                goto out;
+        }
+        index = data_to_int32 (data);
+        if (index > priv->child_count) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "error with stripe-index xattr (%d)", index);
+                local->op_ret   = -1;
+                local->op_errno = EIO;
+                goto out;
+        }
+        if (local->fctx->xl_array) {
+                if (!local->fctx->xl_array[index])
+                        local->fctx->xl_array[index] = prev->this;
+        }
+
+	sprintf(key, "trusted.%s.stripe-coalesce", this->name);
+	data = dict_get(dict, key);
+	if (!data) {
+		/*
+		 * The file was probably created prior to coalesce support.
+		 * Assume non-coalesce mode for this file to maintain backwards
+		 * compatibility.
+		 */
+		gf_log(this->name, GF_LOG_DEBUG, "missing stripe-coalesce "
+			"attr, assume non-coalesce mode");
+		local->fctx->stripe_coalesce = 0;
+	} else {
+		local->fctx->stripe_coalesce = data_to_int32(data);
+	}
+
+
+out:
+        return 0;
+}
+
+int32_t
+stripe_xattr_request_build (xlator_t *this, dict_t *dict, uint64_t stripe_size,
+                            uint32_t stripe_count, uint32_t stripe_index,
+			    uint32_t stripe_coalesce)
+{
+        char            key[256]       = {0,};
+        int32_t         ret             = -1;
+
+        sprintf (key, "trusted.%s.stripe-size", this->name);
+        ret = dict_set_int64 (dict, key, stripe_size);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to set %s in xattr_req dict", key);
+                goto out;
+        }
+
+        sprintf (key, "trusted.%s.stripe-count", this->name);
+        ret = dict_set_int32 (dict, key, stripe_count);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to set %s in xattr_req dict", key);
+                goto out;
+        }
+
+        sprintf (key, "trusted.%s.stripe-index", this->name);
+        ret = dict_set_int32 (dict, key, stripe_index);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to set %s in xattr_req dict", key);
+                goto out;
+        }
+
+	sprintf(key, "trusted.%s.stripe-coalesce", this->name);
+	ret = dict_set_int32(dict, key, stripe_coalesce);
+	if (ret) {
+		gf_log(this->name, GF_LOG_WARNING,
+			"failed to set %s in xattr_req_dict", key);
+		goto out;
+	}
+out:
+        return ret;
+}
+
+
+static int
+set_default_block_size (stripe_private_t *priv, char *num)
+{
+
+        int             ret = -1;
+        GF_VALIDATE_OR_GOTO ("stripe", THIS, out);
+        GF_VALIDATE_OR_GOTO (THIS->name, priv, out);
+        GF_VALIDATE_OR_GOTO (THIS->name, num, out);
+
+
+        if (gf_string2bytesize (num, &priv->block_size) != 0) {
+                gf_log (THIS->name, GF_LOG_ERROR,
+                        "invalid number format \"%s\"", num);
+                goto out;
+        }
+
+        ret = 0;
+
+ out:
+        return ret;
+
+}
+
+
+int
+set_stripe_block_size (xlator_t *this, stripe_private_t *priv, char *data)
+{
+        int                    ret = -1;
+        char                  *tmp_str = NULL;
+        char                  *tmp_str1 = NULL;
+        char                  *dup_str = NULL;
+        char                  *stripe_str = NULL;
+        char                  *pattern = NULL;
+        char                  *num = NULL;
+        struct stripe_options *temp_stripeopt = NULL;
+        struct stripe_options *stripe_opt = NULL;
+
+        if (!this || !priv || !data)
+                goto out;
+
+        /* Get the pattern for striping.
+           "option block-size *avi:10MB" etc */
+        stripe_str = strtok_r (data, ",", &tmp_str);
+        while (stripe_str) {
+                dup_str = gf_strdup (stripe_str);
+                stripe_opt = GF_CALLOC (1, sizeof (struct stripe_options),
+                                        gf_stripe_mt_stripe_options);
+                if (!stripe_opt) {
+                        goto out;
+                }
+
+                pattern = strtok_r (dup_str, ":", &tmp_str1);
+                num = strtok_r (NULL, ":", &tmp_str1);
+                if (!num) {
+                        num = pattern;
+                        pattern = "*";
+                        ret = set_default_block_size (priv, num);
+                        if (ret)
+                                goto out;
+                }
+                if (gf_string2bytesize (num, &stripe_opt->block_size) != 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "invalid number format \"%s\"", num);
+                        goto out;
+                }
+
+                if (stripe_opt->block_size < STRIPE_MIN_BLOCK_SIZE) {
+                        gf_log (this->name, GF_LOG_ERROR, "Invalid Block-size: "
+                                "%s. Should be atleast %llu bytes", num,
+                                STRIPE_MIN_BLOCK_SIZE);
+                        goto out;
+                }
+                if (stripe_opt->block_size % 512) {
+                        gf_log (this->name, GF_LOG_ERROR, "Block-size: %s should"
+                                " be a multiple of 512 bytes", num);
+                        goto out;
+                }
+
+                memcpy (stripe_opt->path_pattern, pattern, strlen (pattern));
+
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "block-size : pattern %s : size %"PRId64,
+                        stripe_opt->path_pattern, stripe_opt->block_size);
+
+                if (priv->pattern)
+                        temp_stripeopt = NULL;
+                else
+                        temp_stripeopt = priv->pattern;
+
+                stripe_opt->next = temp_stripeopt;
+
+                priv->pattern = stripe_opt;
+                stripe_opt = NULL;
+
+                GF_FREE (dup_str);
+                dup_str = NULL;
+
+                stripe_str = strtok_r (NULL, ",", &tmp_str);
+        }
+
+        ret = 0;
+out:
+
+        GF_FREE (dup_str);
+
+        GF_FREE (stripe_opt);
+
+        return ret;
+}
+
+int32_t
+stripe_iatt_merge (struct iatt *from, struct iatt *to)
+{
+        if (to->ia_size < from->ia_size)
+                to->ia_size = from->ia_size;
+        if (to->ia_mtime < from->ia_mtime)
+                to->ia_mtime = from->ia_mtime;
+        if (to->ia_ctime < from->ia_ctime)
+                to->ia_ctime = from->ia_ctime;
+        if (to->ia_atime < from->ia_atime)
+                to->ia_atime = from->ia_atime;
+        return 0;
+}
+
+off_t
+coalesced_offset(off_t offset, uint64_t stripe_size, int stripe_count)
+{
+	size_t line_size = 0;
+	uint64_t stripe_num = 0;
+	off_t coalesced_offset = 0;
+
+	line_size = stripe_size * stripe_count;
+	stripe_num = offset / line_size;
+
+	coalesced_offset = (stripe_num * stripe_size) +
+		(offset % stripe_size);
+
+	return coalesced_offset;
+}
+
+off_t
+uncoalesced_size(off_t size, uint64_t stripe_size, int stripe_count,
+		int stripe_index)
+{
+        uint64_t nr_full_stripe_chunks = 0, mod = 0;
+
+        if (!size)
+                return size;
+
+	/*
+	 * Estimate the number of fully written stripes from the
+	 * local file size. Each stripe_size chunk corresponds to
+	 * a stripe.
+	 */
+        nr_full_stripe_chunks = (size / stripe_size) * stripe_count;
+        mod = size % stripe_size;
+
+        if (!mod) {
+                /*
+		 * There is no remainder, thus we could have overestimated
+		 * the size of the file in terms of chunks. Trim the number
+		 * of chunks by the following stripe members and leave it
+		 * up to those nodes to respond with a larger size (if
+		 * necessary).
+		 */
+                nr_full_stripe_chunks -= stripe_count -
+                        (stripe_index + 1);
+                size = nr_full_stripe_chunks * stripe_size;
+        } else {
+		/*
+		 * There is a remainder and thus we own the last chunk of the
+		 * file. Add the preceding stripe members of the final stripe
+		 * along with the remainder to calculate the exact size.
+		 */
+                nr_full_stripe_chunks += stripe_index;
+                size = nr_full_stripe_chunks * stripe_size + mod;
+        }
+
+        return size;
+}
+
diff --git a/xlators/cluster/stripe/src/stripe-mem-types.h b/xlators/cluster/stripe/src/stripe-mem-types.h
index 2d2757e86..e9ac9cf46 100644
--- a/xlators/cluster/stripe/src/stripe-mem-types.h
+++ b/xlators/cluster/stripe/src/stripe-mem-types.h
@@ -1,21 +1,11 @@
-
 /*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
 
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -25,15 +15,16 @@
 #include "mem-types.h"
 
 enum gf_stripe_mem_types_ {
-        gf_stripe_mt_stripe_local_t = gf_common_mt_end + 1,
-        gf_stripe_mt_iovec,
-        gf_stripe_mt_readv_replies,
+        gf_stripe_mt_iovec = gf_common_mt_end + 1,
+        gf_stripe_mt_stripe_replies,
         gf_stripe_mt_stripe_fd_ctx_t,
         gf_stripe_mt_char,
         gf_stripe_mt_int8_t,
+        gf_stripe_mt_int32_t,
         gf_stripe_mt_xlator_t,
         gf_stripe_mt_stripe_private_t,
         gf_stripe_mt_stripe_options,
+        gf_stripe_mt_xattr_sort_t,
         gf_stripe_mt_end
 };
 #endif
diff --git a/xlators/cluster/stripe/src/stripe.c b/xlators/cluster/stripe/src/stripe.c
index b2c70cfe5..79e80b513 100644
--- a/xlators/cluster/stripe/src/stripe.c
+++ b/xlators/cluster/stripe/src/stripe.c
@@ -1,25 +1,16 @@
 /*
-  Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 /**
  * xlators/cluster/stripe:
- *    Stripe translator, stripes the data accross its child nodes,
+ *    Stripe translator, stripes the data across its child nodes,
  *    as per the options given in the volfile. The striping works
  *    fairly simple. It writes files at different offset as per
  *    calculation. So, 'ls -l' output at the real posix level will
@@ -32,65 +23,19 @@
  *  very much necessary, or else, use it in combination with AFR, to have a
  *  backup copy.
  */
+#include <fnmatch.h>
 
 #include "stripe.h"
+#include "libxlator.h"
+#include "byte-order.h"
+#include "statedump.h"
 
-void
-stripe_local_wipe (stripe_local_t *local)
-{
-        if (!local)
-                goto out;
-
-        if (local->loc.path)
-                loc_wipe (&local->loc);
-        if (local->loc2.path)
-                loc_wipe (&local->loc2);
-out:
-        return;
-}
-
-/**
- * stripe_get_matching_bs - Get the matching block size for the given path.
- */
-int32_t
-stripe_get_matching_bs (const char *path, struct stripe_options *opts,
-                        uint64_t default_bs)
-{
-        struct stripe_options *trav       = NULL;
-        char                  *pathname   = NULL;
-        uint64_t               block_size = 0;
-
-        block_size = default_bs;
-
-        if (!path || !opts)
-                goto out;
-
-        /* FIXME: is a strdup really necessary? */
-        pathname = gf_strdup (path);
-        if (!pathname)
-                goto out;
-
-        trav = opts;
-        while (trav) {
-                if (!fnmatch (trav->path_pattern, pathname, FNM_NOESCAPE)) {
-                        block_size = trav->block_size;
-                        break;
-                }
-                trav = trav->next;
-        }
-
-        GF_FREE (pathname);
-
-out:
-        return block_size;
-}
-
-
+struct volume_options options[];
 
 int32_t
 stripe_sh_chown_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                      int32_t op_ret, int32_t op_errno,
-                     struct iatt *preop, struct iatt *postop)
+                     struct iatt *preop, struct iatt *postop, dict_t *xdata)
 {
         int             callcnt = -1;
         stripe_local_t *local   = NULL;
@@ -109,8 +54,7 @@ stripe_sh_chown_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                stripe_local_wipe (local);
-                STACK_DESTROY (frame->root);
+                STRIPE_STACK_DESTROY (frame);
         }
 out:
         return 0;
@@ -120,7 +64,7 @@ int32_t
 stripe_sh_make_entry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                           int32_t op_ret, int32_t op_errno, inode_t *inode,
                           struct iatt *buf, struct iatt *preparent,
-                          struct iatt *postparent)
+                          struct iatt *postparent, dict_t *xdata)
 {
         stripe_local_t *local = NULL;
         call_frame_t    *prev = NULL;
@@ -135,7 +79,7 @@ stripe_sh_make_entry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         STACK_WIND (frame, stripe_sh_chown_cbk, prev->this,
                     prev->this->fops->setattr, &local->loc,
-                    &local->stbuf, (GF_SET_ATTR_UID | GF_SET_ATTR_GID));
+                    &local->stbuf, (GF_SET_ATTR_UID | GF_SET_ATTR_GID), NULL);
 
 out:
         return 0;
@@ -149,7 +93,7 @@ stripe_entry_self_heal (call_frame_t *frame, xlator_t *this,
         call_frame_t     *rframe = NULL;
         stripe_local_t   *rlocal = NULL;
         stripe_private_t *priv   = NULL;
-        dict_t           *dict   = NULL;
+        dict_t           *xdata  = NULL;
         int               ret    = 0;
 
         if (!local || !this || !frame) {
@@ -167,8 +111,7 @@ stripe_entry_self_heal (call_frame_t *frame, xlator_t *this,
         if (!rframe) {
                 goto out;
         }
-        rlocal = GF_CALLOC (1, sizeof (stripe_local_t),
-                            gf_stripe_mt_stripe_local_t);
+        rlocal = mem_get0 (this->local_pool);
         if (!rlocal) {
                 goto out;
         }
@@ -177,14 +120,14 @@ stripe_entry_self_heal (call_frame_t *frame, xlator_t *this,
         loc_copy (&rlocal->loc, &local->loc);
         memcpy (&rlocal->stbuf, &local->stbuf, sizeof (struct iatt));
 
-        dict = dict_new ();
-        if (!dict)
+        xdata = dict_new ();
+        if (!xdata)
                 goto out;
 
-        ret = dict_set_static_bin (dict, "gfid-req", local->stbuf.ia_gfid, 16);
+        ret = dict_set_static_bin (xdata, "gfid-req", local->stbuf.ia_gfid, 16);
         if (ret)
                 gf_log (this->name, GF_LOG_WARNING,
-                        "failed to set gfid-req");
+                        "%s: failed to set gfid-req", local->loc.path);
 
         while (trav) {
                 if (IA_ISREG (local->stbuf.ia_type)) {
@@ -192,36 +135,43 @@ stripe_entry_self_heal (call_frame_t *frame, xlator_t *this,
                                     trav->xlator, trav->xlator->fops->mknod,
                                     &local->loc,
                                     st_mode_from_ia (local->stbuf.ia_prot,
-                                                     local->stbuf.ia_type), 0,
-                                    dict);
+                                                     local->stbuf.ia_type),
+                                    0, 0, xdata);
                 }
                 if (IA_ISDIR (local->stbuf.ia_type)) {
                         STACK_WIND (rframe, stripe_sh_make_entry_cbk,
                                     trav->xlator, trav->xlator->fops->mkdir,
-                                    &local->loc, st_mode_from_ia (local->stbuf.ia_prot,
-                                                                  local->stbuf.ia_type),
-                                    dict);
+                                    &local->loc,
+                                    st_mode_from_ia (local->stbuf.ia_prot,
+                                                     local->stbuf.ia_type),
+                                    0, xdata);
                 }
                 trav = trav->next;
         }
 
+        if (xdata)
+                dict_unref (xdata);
+        return 0;
+
 out:
-        if (dict)
-                dict_unref (dict);
+        if (rframe)
+                STRIPE_STACK_DESTROY (rframe);
+        if (xdata)
+                dict_unref (xdata);
 
         return 0;
 }
 
+
 int32_t
 stripe_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno, inode_t *inode,
-                   struct iatt *buf, dict_t *dict, struct iatt *postparent)
+                   struct iatt *buf, dict_t *xdata, struct iatt *postparent)
 {
-        int32_t         callcnt = 0;
-        dict_t         *tmp_dict = NULL;
-        inode_t        *tmp_inode = NULL;
-        stripe_local_t *local = NULL;
-        call_frame_t   *prev = NULL;
+        int32_t         callcnt     = 0;
+        stripe_local_t *local       = NULL;
+        call_frame_t   *prev        = NULL;
+        int             ret         = 0;
 
         if (!this || !frame || !frame->local || !cookie) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -243,7 +193,7 @@ stripe_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                         strerror (op_errno));
                         if (local->op_errno != ESTALE)
                                 local->op_errno = op_errno;
-                        if ((op_errno != ENOENT) ||
+                        if (((op_errno != ENOENT) && (op_errno != ENOTCONN)) ||
                             (prev->this == FIRST_CHILD (this)))
                                 local->failed = 1;
                         if (op_errno == ENOENT)
@@ -252,51 +202,81 @@ stripe_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
                 if (op_ret >= 0) {
                         local->op_ret = 0;
+                        if (IA_ISREG (buf->ia_type)) {
+                                ret = stripe_ctx_handle (this, prev, local,
+                                                         xdata);
+                                if (ret)
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                 "Error getting fctx info from"
+                                                 " dict");
+                        }
 
                         if (FIRST_CHILD(this) == prev->this) {
                                 local->stbuf      = *buf;
                                 local->postparent = *postparent;
                                 local->inode = inode_ref (inode);
-                                local->dict = dict_ref (dict);
+                                if (xdata)
+                                        local->xdata = dict_ref (xdata);
+                                if (local->xattr) {
+                                        stripe_aggregate_xattr (local->xdata,
+                                                                local->xattr);
+                                        dict_unref (local->xattr);
+                                        local->xattr = NULL;
+                                }
                         }
+
+                        if (!local->xdata && !local->xattr) {
+                                local->xattr = dict_ref (xdata);
+                        } else if (local->xdata) {
+                                stripe_aggregate_xattr (local->xdata, xdata);
+                        } else if (local->xattr) {
+                                stripe_aggregate_xattr (local->xattr, xdata);
+                        }
+
                         local->stbuf_blocks      += buf->ia_blocks;
                         local->postparent_blocks += postparent->ia_blocks;
 
+			correct_file_size(buf, local->fctx, prev);
+
                         if (local->stbuf_size < buf->ia_size)
                                 local->stbuf_size = buf->ia_size;
                         if (local->postparent_size < postparent->ia_size)
                                 local->postparent_size = postparent->ia_size;
+
+                        if (uuid_is_null (local->ia_gfid))
+                                uuid_copy (local->ia_gfid, buf->ia_gfid);
+
+                        /* Make sure the gfid on all the nodes are same */
+                        if (uuid_compare (local->ia_gfid, buf->ia_gfid)) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "%s: gfid different on subvolume %s",
+                                        local->loc.path, prev->this->name);
+                        }
                 }
         }
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                if (local->op_ret == 0 && local->entry_self_heal_needed)
+                if (local->op_ret == 0 && local->entry_self_heal_needed &&
+                    !uuid_is_null (local->loc.inode->gfid))
                         stripe_entry_self_heal (frame, this, local);
 
                 if (local->failed)
                         local->op_ret = -1;
 
-                tmp_dict  = local->dict;
-                tmp_inode = local->inode;
-
                 if (local->op_ret != -1) {
                         local->stbuf.ia_blocks      = local->stbuf_blocks;
                         local->stbuf.ia_size        = local->stbuf_size;
                         local->postparent.ia_blocks = local->postparent_blocks;
                         local->postparent.ia_size   = local->postparent_size;
+                        inode_ctx_put (local->inode, this,
+                                       (uint64_t) (long)local->fctx);
                 }
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (lookup, frame, local->op_ret,
+                STRIPE_STACK_UNWIND (lookup, frame, local->op_ret,
                                      local->op_errno, local->inode,
-                                     &local->stbuf, local->dict,
+                                     &local->stbuf, local->xdata,
                                      &local->postparent);
-
-                if (tmp_inode)
-                        inode_unref (tmp_inode);
-                if (tmp_dict)
-                        dict_unref (tmp_dict);
         }
 out:
         return 0;
@@ -304,12 +284,15 @@ out:
 
 int32_t
 stripe_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
-               dict_t *xattr_req)
+               dict_t *xdata)
 {
-        stripe_local_t   *local = NULL;
-        xlator_list_t    *trav = NULL;
-        stripe_private_t *priv = NULL;
+        stripe_local_t   *local    = NULL;
+        xlator_list_t    *trav     = NULL;
+        stripe_private_t *priv     = NULL;
         int32_t           op_errno = EINVAL;
+        int64_t           filesize = 0;
+        int               ret      = 0;
+        uint64_t          tmpctx   = 0;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -321,8 +304,7 @@ stripe_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
         trav = this->children;
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -331,26 +313,60 @@ stripe_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
         frame->local = local;
         loc_copy (&local->loc, loc);
 
+        inode_ctx_get (local->inode, this, &tmpctx);
+        if (tmpctx)
+                local->fctx = (stripe_fd_ctx_t*) (long)tmpctx;
+
+        /* quick-read friendly changes */
+        if (xdata && dict_get (xdata, GF_CONTENT_KEY)) {
+                ret = dict_get_int64 (xdata, GF_CONTENT_KEY, &filesize);
+                if (!ret && (filesize > priv->block_size))
+                        dict_del (xdata, GF_CONTENT_KEY);
+        }
+
+        /* get stripe-size xattr on lookup. This would be required for
+         * open/read/write/pathinfo calls. Hence we send down the request
+         * even when type == IA_INVAL */
+
+	/*
+	 * We aren't guaranteed to have xdata here. We need the format info for
+	 * the file, so allocate xdata if necessary.
+	 */
+	if (!xdata)
+		xdata = dict_new();
+	else
+		xdata = dict_ref(xdata);
+
+        if (xdata && (IA_ISREG (loc->inode->ia_type) ||
+            (loc->inode->ia_type == IA_INVAL))) {
+                ret = stripe_xattr_request_build (this, xdata, 8, 4, 4, 0);
+                if (ret)
+                        gf_log (this->name , GF_LOG_ERROR, "Failed to build"
+                                " xattr request for %s", loc->path);
+
+        }
+
         /* Everytime in stripe lookup, all child nodes
            should be looked up */
         local->call_count = priv->child_count;
         while (trav) {
                 STACK_WIND (frame, stripe_lookup_cbk, trav->xlator,
-                            trav->xlator->fops->lookup,
-                            loc, xattr_req);
+                            trav->xlator->fops->lookup, loc, xdata);
                 trav = trav->next;
         }
 
+	dict_unref(xdata);
+
         return 0;
 err:
-        STACK_UNWIND_STRICT (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        STRIPE_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
         return 0;
 }
 
 
 int32_t
 stripe_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                 int32_t op_ret, int32_t op_errno, struct iatt *buf)
+                 int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
@@ -385,6 +401,9 @@ stripe_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         }
 
                         local->stbuf_blocks += buf->ia_blocks;
+
+			correct_file_size(buf, local->fctx, prev);
+
                         if (local->stbuf_size < buf->ia_size)
                                 local->stbuf_size = buf->ia_size;
                 }
@@ -400,20 +419,20 @@ stripe_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->stbuf.ia_blocks = local->stbuf_blocks;
                 }
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (stat, frame, local->op_ret,
-                                     local->op_errno, &local->stbuf);
+                STRIPE_STACK_UNWIND (stat, frame, local->op_ret,
+                                     local->op_errno, &local->stbuf, NULL);
         }
 out:
         return 0;
 }
 
 int32_t
-stripe_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
+stripe_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
         xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
+	stripe_fd_ctx_t  *fctx = NULL;
         int32_t           op_errno = EINVAL;
 
         VALIDATE_OR_GOTO (frame, err);
@@ -431,8 +450,7 @@ stripe_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -441,23 +459,30 @@ stripe_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
         frame->local = local;
         local->call_count = priv->child_count;
 
+	if (IA_ISREG(loc->inode->ia_type)) {
+		inode_ctx_get(loc->inode, this, (uint64_t *) &fctx);
+		if (!fctx)
+			goto err;
+		local->fctx = fctx;
+	}
+
         while (trav) {
                 STACK_WIND (frame, stripe_stat_cbk, trav->xlator,
-                            trav->xlator->fops->stat, loc);
+                            trav->xlator->fops->stat, loc, NULL);
                 trav = trav->next;
         }
 
         return 0;
 
 err:
-        STACK_UNWIND_STRICT (stat, frame, -1, op_errno, NULL);
+        STRIPE_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
 
 int32_t
 stripe_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct statvfs *stbuf)
+                   int32_t op_ret, int32_t op_errno, struct statvfs *stbuf, dict_t *xdata)
 {
         stripe_local_t *local = NULL;
         int32_t         callcnt = 0;
@@ -494,15 +519,15 @@ stripe_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                STACK_UNWIND_STRICT (statfs, frame, local->op_ret,
-                                     local->op_errno, &local->statvfs_buf);
+                STRIPE_STACK_UNWIND (statfs, frame, local->op_ret,
+                                     local->op_errno, &local->statvfs_buf, NULL);
         }
 out:
         return 0;
 }
 
 int32_t
-stripe_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
+stripe_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         xlator_list_t    *trav = NULL;
@@ -517,8 +542,7 @@ stripe_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
         priv = this->private;
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -530,13 +554,13 @@ stripe_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
         local->call_count = priv->child_count;
         while (trav) {
                 STACK_WIND (frame, stripe_statfs_cbk, trav->xlator,
-                            trav->xlator->fops->statfs, loc);
+                            trav->xlator->fops->statfs, loc, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (statfs, frame, -1, op_errno, NULL);
+        STRIPE_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
@@ -545,7 +569,7 @@ err:
 int32_t
 stripe_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                      int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                     struct iatt *postbuf)
+                     struct iatt *postbuf, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
@@ -583,6 +607,9 @@ stripe_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->prebuf_blocks  += prebuf->ia_blocks;
                         local->postbuf_blocks += postbuf->ia_blocks;
 
+			correct_file_size(prebuf, local->fctx, prev);
+			correct_file_size(postbuf, local->fctx, prev);
+
                         if (local->prebuf_size < prebuf->ia_size)
                                 local->prebuf_size = prebuf->ia_size;
 
@@ -603,22 +630,23 @@ stripe_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->post_buf.ia_size   = local->postbuf_size;
                 }
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (truncate, frame, local->op_ret,
+                STRIPE_STACK_UNWIND (truncate, frame, local->op_ret,
                                      local->op_errno, &local->pre_buf,
-                                     &local->post_buf);
+                                     &local->post_buf, NULL);
         }
 out:
         return 0;
 }
 
 int32_t
-stripe_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset)
+stripe_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset, dict_t *xdata)
 {
-        xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
+	stripe_fd_ctx_t  *fctx = NULL;
         int32_t           op_errno = EINVAL;
+	int		  i, eof_idx;
+	off_t		  dest_offset, tmp_offset;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -627,7 +655,6 @@ stripe_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset)
         VALIDATE_OR_GOTO (loc->inode, err);
 
         priv = this->private;
-        trav = this->children;
 
         if (priv->first_child_down) {
                 op_errno = ENOTCONN;
@@ -635,8 +662,7 @@ stripe_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset)
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -645,15 +671,55 @@ stripe_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset)
         frame->local = local;
         local->call_count = priv->child_count;
 
-        while (trav) {
-                STACK_WIND (frame, stripe_truncate_cbk, trav->xlator,
-                            trav->xlator->fops->truncate, loc, offset);
-                trav = trav->next;
-        }
+	inode_ctx_get(loc->inode, this, (uint64_t *) &fctx);
+	if (!fctx) {
+		gf_log(this->name, GF_LOG_ERROR, "no stripe context");
+		op_errno = EINVAL;
+		goto err;
+	}
+
+	local->fctx = fctx;
+	eof_idx = (offset / fctx->stripe_size) % fctx->stripe_count;
+
+	for (i = 0; i < fctx->stripe_count; i++) {
+		if (!fctx->xl_array[i]) {
+			gf_log(this->name, GF_LOG_ERROR,
+				"no xlator at index %d", i);
+			op_errno = EINVAL;
+			goto err;
+		}
+
+		if (fctx->stripe_coalesce) {
+			/*
+	 		 * The node that owns EOF is truncated to the exact
+	 		 * coalesced offset. Nodes prior to this index should
+	 		 * be rounded up to the size of the complete stripe,
+	 		 * while nodes after this index should be rounded down
+			 * to the size of the previous stripe.
+			 */
+			if (i < eof_idx)
+				tmp_offset = roof(offset, fctx->stripe_size *
+						fctx->stripe_count);
+			else if (i > eof_idx)
+				tmp_offset = floor(offset, fctx->stripe_size *
+						fctx->stripe_count);
+			else
+				tmp_offset = offset;
+
+			dest_offset = coalesced_offset(tmp_offset,
+					fctx->stripe_size, fctx->stripe_count);
+		} else {
+			dest_offset = offset;
+		}
+
+		STACK_WIND(frame, stripe_truncate_cbk, fctx->xl_array[i],
+			fctx->xl_array[i]->fops->truncate, loc, dest_offset,
+			NULL);
+	}
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (truncate, frame, -1, op_errno, NULL, NULL);
+        STRIPE_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
@@ -661,7 +727,7 @@ err:
 int32_t
 stripe_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno,
-                    struct iatt *preop, struct iatt *postop)
+                    struct iatt *preop, struct iatt *postop, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
@@ -700,6 +766,9 @@ stripe_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->prebuf_blocks  += preop->ia_blocks;
                         local->postbuf_blocks += postop->ia_blocks;
 
+			correct_file_size(preop, local->fctx, prev);
+			correct_file_size(postop, local->fctx, prev);
+
                         if (local->prebuf_size < preop->ia_size)
                                 local->prebuf_size = preop->ia_size;
                         if (local->postbuf_size < postop->ia_size)
@@ -719,10 +788,9 @@ stripe_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->post_buf.ia_size   = local->postbuf_size;
                 }
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (setattr, frame, local->op_ret,
+                STRIPE_STACK_UNWIND (setattr, frame, local->op_ret,
                                      local->op_errno, &local->pre_buf,
-                                     &local->post_buf);
+                                     &local->post_buf, NULL);
         }
 out:
         return 0;
@@ -731,11 +799,12 @@ out:
 
 int32_t
 stripe_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                struct iatt *stbuf, int32_t valid)
+                struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
         xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
+	stripe_fd_ctx_t	 *fctx = NULL;
         int32_t           op_errno = EINVAL;
 
         VALIDATE_OR_GOTO (frame, err);
@@ -753,33 +822,47 @@ stripe_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         local->op_ret = -1;
         frame->local = local;
-        local->call_count = priv->child_count;
+        if (!IA_ISDIR (loc->inode->ia_type) &&
+            !IA_ISREG (loc->inode->ia_type)) {
+                local->call_count = 1;
+                STACK_WIND (frame, stripe_setattr_cbk, FIRST_CHILD (this),
+                            FIRST_CHILD (this)->fops->setattr,
+                            loc, stbuf, valid, NULL);
+                return 0;
+        }
 
+	if (IA_ISREG(loc->inode->ia_type)) {
+		inode_ctx_get(loc->inode, this, (uint64_t *) &fctx);
+		if (!fctx)
+			goto err;
+		local->fctx = fctx;
+	}
+
+        local->call_count = priv->child_count;
         while (trav) {
                 STACK_WIND (frame, stripe_setattr_cbk,
                             trav->xlator, trav->xlator->fops->setattr,
-                            loc, stbuf, valid);
+                            loc, stbuf, valid, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (setattr, frame, -1, op_errno, NULL, NULL);
+        STRIPE_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 
 int32_t
 stripe_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
-                 struct iatt *stbuf, int32_t valid)
+                 struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
@@ -795,8 +878,7 @@ stripe_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
         trav = this->children;
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -807,13 +889,13 @@ stripe_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
 
         while (trav) {
                 STACK_WIND (frame, stripe_setattr_cbk, trav->xlator,
-                            trav->xlator->fops->fsetattr, fd, stbuf, valid);
+                            trav->xlator->fops->fsetattr, fd, stbuf, valid, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (fsetattr, frame, -1, op_errno, NULL, NULL);
+        STRIPE_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
@@ -821,7 +903,8 @@ int32_t
 stripe_stack_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                          int32_t op_ret, int32_t op_errno, struct iatt *buf,
                          struct iatt *preoldparent, struct iatt *postoldparent,
-                         struct iatt *prenewparent, struct iatt *postnewparent)
+                         struct iatt *prenewparent, struct iatt *postnewparent,
+                         dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
@@ -858,6 +941,8 @@ stripe_stack_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->pre_buf.ia_blocks    += prenewparent->ia_blocks;
                         local->post_buf.ia_blocks   += postnewparent->ia_blocks;
 
+			correct_file_size(buf, local->fctx, prev);
+
                         if (local->stbuf.ia_size < buf->ia_size)
                                 local->stbuf.ia_size =  buf->ia_size;
 
@@ -880,11 +965,10 @@ stripe_stack_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (local->failed)
                         local->op_ret = -1;
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (rename, frame, local->op_ret, local->op_errno,
+                STRIPE_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
                                      &local->stbuf, &local->preparent,
                                      &local->postparent,  &local->pre_buf,
-                                     &local->post_buf);
+                                     &local->post_buf, NULL);
         }
 out:
         return 0;
@@ -894,7 +978,8 @@ int32_t
 stripe_first_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                          int32_t op_ret, int32_t op_errno, struct iatt *buf,
                          struct iatt *preoldparent, struct iatt *postoldparent,
-                         struct iatt *prenewparent, struct iatt *postnewparent)
+                         struct iatt *prenewparent, struct iatt *postnewparent,
+                         dict_t *xdata)
 {
         stripe_local_t *local = NULL;
         xlator_list_t  *trav = NULL;
@@ -925,25 +1010,25 @@ stripe_first_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         while (trav) {
                 STACK_WIND (frame, stripe_stack_rename_cbk,
                             trav->xlator, trav->xlator->fops->rename,
-                            &local->loc, &local->loc2);
+                            &local->loc, &local->loc2, NULL);
                 trav = trav->next;
         }
         return 0;
 
 unwind:
-        stripe_local_wipe (local);
-        STACK_UNWIND_STRICT (rename, frame, -1, op_errno, buf, preoldparent,
-                             postoldparent, prenewparent, postnewparent);
+        STRIPE_STACK_UNWIND (rename, frame, -1, op_errno, buf, preoldparent,
+                             postoldparent, prenewparent, postnewparent, NULL);
         return 0;
 }
 
 int32_t
 stripe_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
-               loc_t *newloc)
+               loc_t *newloc, dict_t *xdata)
 {
         stripe_private_t *priv = NULL;
         stripe_local_t   *local = NULL;
         xlator_list_t    *trav = NULL;
+	stripe_fd_ctx_t	 *fctx = NULL;
         int32_t           op_errno = EINVAL;
 
         VALIDATE_OR_GOTO (frame, err);
@@ -963,36 +1048,79 @@ stripe_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
+
+        frame->local = local;
+
         local->op_ret = -1;
         loc_copy (&local->loc, oldloc);
         loc_copy (&local->loc2, newloc);
 
         local->call_count = priv->child_count;
 
-        frame->local = local;
+	if (IA_ISREG(oldloc->inode->ia_type)) {
+		inode_ctx_get(oldloc->inode, this, (uint64_t *) &fctx);
+		if (!fctx)
+			goto err;
+		local->fctx = fctx;
+	}
 
         STACK_WIND (frame, stripe_first_rename_cbk, trav->xlator,
-                    trav->xlator->fops->rename, oldloc, newloc);
+                    trav->xlator->fops->rename, oldloc, newloc, NULL);
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (rename, frame, -1, op_errno, NULL, NULL, NULL,
-                             NULL, NULL);
+        STRIPE_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL,
+                             NULL, NULL, NULL);
+        return 0;
+}
+int32_t
+stripe_first_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                         struct iatt *postparent, dict_t *xdata)
+{
+        stripe_local_t *local   = NULL;
+        call_frame_t   *prev = NULL;
+
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+
+        prev  = cookie;
+        local = frame->local;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG, "%s returned %s",
+                        prev->this->name, strerror (op_errno));
+                goto out;
+        }
+        local->op_ret = 0;
+        local->preparent  = *preparent;
+        local->postparent = *postparent;
+        local->preparent_blocks  += preparent->ia_blocks;
+        local->postparent_blocks += postparent->ia_blocks;
+
+        STRIPE_STACK_UNWIND(unlink, frame, local->op_ret, local->op_errno,
+                            &local->preparent, &local->postparent, xdata);
+        return 0;
+out:
+        STRIPE_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
+
         return 0;
 }
 
 
 
+
 int32_t
 stripe_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno, struct iatt *preparent,
-                   struct iatt *postparent)
+                   struct iatt *postparent, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local   = NULL;
@@ -1014,50 +1142,33 @@ stripe_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         gf_log (this->name, GF_LOG_DEBUG, "%s returned %s",
                                 prev->this->name, strerror (op_errno));
                         local->op_errno = op_errno;
-                        if ((op_errno != ENOENT) ||
-                            (prev->this == FIRST_CHILD (this)))
+                        if (op_errno != ENOENT) {
                                 local->failed = 1;
-                }
-                if (op_ret >= 0) {
-                        local->op_ret = op_ret;
-                        if (FIRST_CHILD(this) == prev->this) {
-                                local->preparent  = *preparent;
-                                local->postparent = *postparent;
+                                local->op_ret = op_ret;
                         }
-                        local->preparent_blocks  += preparent->ia_blocks;
-                        local->postparent_blocks += postparent->ia_blocks;
-
-                        if (local->preparent_size < preparent->ia_size)
-                                local->preparent_size = preparent->ia_size;
-
-                        if (local->postparent_size < postparent->ia_size)
-                                local->postparent_size = postparent->ia_size;
                 }
         }
         UNLOCK (&frame->lock);
 
-        if (!callcnt) {
-                if (local->failed)
-                        local->op_ret = -1;
-
-                if (local->op_ret != -1) {
-                        local->preparent.ia_blocks  = local->preparent_blocks;
-                        local->preparent.ia_size    = local->preparent_size;
-                        local->postparent.ia_blocks = local->postparent_blocks;
-                        local->postparent.ia_size   = local->postparent_size;
+        if (callcnt == 1) {
+                if (local->failed) {
+                        op_errno = local->op_errno;
+                        goto out;
                 }
-
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (unlink, frame, local->op_ret,
-                                     local->op_errno, &local->preparent,
-                                     &local->postparent);
+                STACK_WIND(frame, stripe_first_unlink_cbk, FIRST_CHILD (this),
+                           FIRST_CHILD (this)->fops->unlink, &local->loc,
+                           local->xflag, local->xdata);
         }
+        return 0;
 out:
+        STRIPE_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
+
         return 0;
 }
 
 int32_t
-stripe_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
+stripe_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc,
+               int xflag, dict_t *xdata)
 {
         xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
@@ -1085,26 +1196,32 @@ stripe_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         local->op_ret = -1;
+        loc_copy (&local->loc, loc);
+        local->xflag = xflag;
+
+        if (xdata)
+            local->xdata = dict_ref (xdata);
+
         frame->local = local;
         local->call_count = priv->child_count;
+        trav = trav->next; /* Skip the first child */
 
         while (trav) {
                 STACK_WIND (frame, stripe_unlink_cbk,
                             trav->xlator, trav->xlator->fops->unlink,
-                            loc);
+                            loc, xflag, xdata);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (unlink, frame, -1, op_errno, NULL, NULL);
+        STRIPE_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
@@ -1112,10 +1229,8 @@ err:
 int32_t
 stripe_first_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         int32_t op_ret, int32_t op_errno,struct iatt *preparent,
-                        struct iatt *postparent)
-
+                        struct iatt *postparent, dict_t *xdata)
 {
-        xlator_list_t *trav = NULL;
         stripe_local_t *local = NULL;
 
         if (!this || !frame || !frame->local) {
@@ -1128,11 +1243,10 @@ stripe_first_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 goto err;
         }
 
-        trav = this->children;
         local = frame->local;
+        local->op_ret = 0;
 
         local->call_count--; /* First child successful */
-        trav = trav->next; /* Skip first child */
 
         local->preparent  = *preparent;
         local->postparent = *postparent;
@@ -1141,22 +1255,60 @@ stripe_first_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         local->preparent_blocks  += preparent->ia_blocks;
         local->postparent_blocks += postparent->ia_blocks;
 
-        while (trav) {
-                STACK_WIND (frame, stripe_unlink_cbk, trav->xlator,
-                            trav->xlator->fops->rmdir, &local->loc,
-                            local->flags);
-                trav = trav->next;
-        }
-
+        STRIPE_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
+                             &local->preparent, &local->postparent, xdata);
         return 0;
 err:
-        STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno, NULL, NULL);
+        STRIPE_STACK_UNWIND (rmdir, frame, op_ret, op_errno, NULL, NULL, NULL);
         return 0;
 
 }
 
 int32_t
-stripe_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags)
+stripe_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                   struct iatt *postparent, dict_t *xdata)
+{
+        int32_t         callcnt = 0;
+        stripe_local_t *local   = NULL;
+        call_frame_t   *prev = NULL;
+
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+
+        prev  = cookie;
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                callcnt = --local->call_count;
+
+                if (op_ret == -1) {
+                        gf_log (this->name, GF_LOG_DEBUG, "%s returned %s",
+                                prev->this->name, strerror (op_errno));
+                        if (op_errno != ENOENT)
+                                local->failed = 1;
+                }
+        }
+        UNLOCK (&frame->lock);
+
+        if (callcnt == 1) {
+                if (local->failed)
+                        goto out;
+                STACK_WIND (frame, stripe_first_rmdir_cbk, FIRST_CHILD (this),
+                            FIRST_CHILD (this)->fops->rmdir, &local->loc,
+                            local->flags, NULL);
+        }
+        return 0;
+out:
+        STRIPE_STACK_UNWIND (rmdir, frame, -1, op_errno, NULL, NULL, NULL);
+        return 0;
+}
+
+int32_t
+stripe_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags, dict_t *xdata)
 {
         xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
@@ -1179,8 +1331,7 @@ stripe_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags)
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -1190,13 +1341,17 @@ stripe_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags)
         loc_copy (&local->loc, loc);
         local->flags = flags;
         local->call_count = priv->child_count;
+        trav = trav->next; /* skip the first child */
 
-        STACK_WIND (frame, stripe_first_rmdir_cbk,  trav->xlator,
-                    trav->xlator->fops->rmdir, loc, flags);
+        while (trav) {
+                STACK_WIND (frame, stripe_rmdir_cbk,  trav->xlator,
+                            trav->xlator->fops->rmdir, loc, flags, NULL);
+                trav = trav->next;
+        }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (rmdir, frame, -1, op_errno, NULL, NULL);
+        STRIPE_STACK_UNWIND (rmdir, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
@@ -1205,7 +1360,7 @@ int32_t
 stripe_mknod_ifreg_fail_unlink_cbk (call_frame_t *frame, void *cookie,
                                     xlator_t *this, int32_t op_ret,
                                     int32_t op_errno, struct iatt *preparent,
-                                    struct iatt *postparent)
+                                    struct iatt *postparent, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
@@ -1224,10 +1379,9 @@ stripe_mknod_ifreg_fail_unlink_cbk (call_frame_t *frame, void *cookie,
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (mknod, frame, local->op_ret, local->op_errno,
+                STRIPE_STACK_UNWIND (mknod, frame, local->op_ret, local->op_errno,
                                      local->inode, &local->stbuf,
-                                     &local->preparent, &local->postparent);
+                                     &local->preparent, &local->postparent, NULL);
         }
 out:
         return 0;
@@ -1239,7 +1393,7 @@ out:
 int32_t
 stripe_mknod_ifreg_setxattr_cbk (call_frame_t *frame, void *cookie,
                                  xlator_t *this, int32_t op_ret,
-                                 int32_t op_errno)
+                                 int32_t op_errno, dict_t *xdata)
 {
         int32_t           callcnt = 0;
         stripe_local_t   *local = NULL;
@@ -1278,16 +1432,15 @@ stripe_mknod_ifreg_setxattr_cbk (call_frame_t *frame, void *cookie,
                                             stripe_mknod_ifreg_fail_unlink_cbk,
                                             trav->xlator,
                                             trav->xlator->fops->unlink,
-                                            &local->loc);
+                                            &local->loc, 0, NULL);
                                 trav = trav->next;
                         }
                         return 0;
                 }
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (mknod, frame, local->op_ret, local->op_errno,
+                STRIPE_STACK_UNWIND (mknod, frame, local->op_ret, local->op_errno,
                                      local->inode, &local->stbuf,
-                                     &local->preparent, &local->postparent);
+                                     &local->preparent, &local->postparent, NULL);
         }
 out:
         return 0;
@@ -1297,13 +1450,13 @@ int32_t
 stripe_mknod_ifreg_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         int32_t op_ret, int32_t op_errno, inode_t *inode,
                         struct iatt *buf, struct iatt *preparent,
-                        struct iatt *postparent)
+                        struct iatt *postparent, dict_t *xdata)
 {
-        int               ret = 0;
         int32_t           callcnt = 0;
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
         call_frame_t     *prev = NULL;
+        xlator_list_t    *trav = NULL;
 
         if (!this || !frame || !frame->local || !cookie) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -1311,7 +1464,7 @@ stripe_mknod_ifreg_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         }
 
         prev  = cookie;
-	priv  = this->private;
+        priv  = this->private;
         local = frame->local;
 
         LOCK (&frame->lock);
@@ -1327,20 +1480,24 @@ stripe_mknod_ifreg_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                 local->failed = 1;
                         local->op_errno = op_errno;
                 }
-
                 if (op_ret >= 0) {
                         local->op_ret = op_ret;
 
-                        if (FIRST_CHILD(this) == prev->this) {
-                                local->stbuf      = *buf;
-                                local->preparent  = *preparent;
-                                local->postparent = *postparent;
-                        }
+                        /* Can be used as a mechanism to understand if mknod
+                           was successful in at least one place */
+                        if (uuid_is_null (local->ia_gfid))
+                                uuid_copy (local->ia_gfid, buf->ia_gfid);
+
+			if (stripe_ctx_handle(this, prev, local, xdata))
+				gf_log(this->name, GF_LOG_ERROR,
+					"Error getting fctx info from dict");
 
                         local->stbuf_blocks += buf->ia_blocks;
                         local->preparent_blocks  += preparent->ia_blocks;
                         local->postparent_blocks += postparent->ia_blocks;
 
+			correct_file_size(buf, local->fctx, prev);
+
                         if (local->stbuf_size < buf->ia_size)
                                 local->stbuf_size = buf->ia_size;
                         if (local->preparent_size < preparent->ia_size)
@@ -1355,6 +1512,23 @@ stripe_mknod_ifreg_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (local->failed)
                         local->op_ret = -1;
 
+                if ((local->op_ret == -1) && !uuid_is_null (local->ia_gfid)) {
+                        /* ia_gfid set means, at least on one node 'mknod'
+                           is successful */
+                        local->call_count = priv->child_count;
+                        trav = this->children;
+                        while (trav) {
+                                STACK_WIND (frame,
+                                            stripe_mknod_ifreg_fail_unlink_cbk,
+                                            trav->xlator,
+                                            trav->xlator->fops->unlink,
+                                            &local->loc, 0, NULL);
+                                trav = trav->next;
+                        }
+                        return 0;
+                }
+
+
                 if (local->op_ret != -1) {
                         local->preparent.ia_blocks  = local->preparent_blocks;
                         local->preparent.ia_size    = local->preparent_size;
@@ -1362,73 +1536,109 @@ stripe_mknod_ifreg_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->postparent.ia_size   = local->postparent_size;
                         local->stbuf.ia_size        = local->stbuf_size;
                         local->stbuf.ia_blocks      = local->stbuf_blocks;
+                        inode_ctx_put (local->inode, this,
+                                       (uint64_t)(long) local->fctx);
+
                 }
+                STRIPE_STACK_UNWIND (mknod, frame, local->op_ret, local->op_errno,
+                                     local->inode, &local->stbuf,
+                                     &local->preparent, &local->postparent, NULL);
+        }
+out:
+        return 0;
+}
 
-                if ((local->op_ret != -1) && priv->xattr_supported) {
-                        /* Send a setxattr request to nodes where the
-                           files are created */
-                        int32_t  i              = 0;
-                        char     size_key[256]  = {0,};
-                        char     index_key[256] = {0,};
-                        char     count_key[256] = {0,};
-                        dict_t  *dict           = NULL;
 
-                        sprintf (size_key,
-                                 "trusted.%s.stripe-size", this->name);
-                        sprintf (count_key,
-                                 "trusted.%s.stripe-count", this->name);
-                        sprintf (index_key,
-                                 "trusted.%s.stripe-index", this->name);
+int32_t
+stripe_mknod_first_ifreg_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, inode_t *inode,
+                        struct iatt *buf, struct iatt *preparent,
+                        struct iatt *postparent, dict_t *xdata)
+{
+        stripe_local_t   *local = NULL;
+        stripe_private_t *priv = NULL;
+        call_frame_t     *prev = NULL;
+        xlator_list_t    *trav = NULL;
+        int               i    = 1;
+        dict_t           *dict           = NULL;
+        int               ret            = 0;
+        int               need_unref     = 0;
 
-                        local->call_count = priv->child_count;
-                        memcpy (local->loc.inode->gfid, local->stbuf.ia_gfid, 16);
-                        for (i = 0; i < priv->child_count; i++) {
-                                dict = get_new_dict ();
-                                if (!dict) {
-                                        gf_log (this->name, GF_LOG_ERROR,
-                                                "failed to allocate dict");
-                                }
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
 
-                                dict_ref (dict);
-                                /* TODO: check return value */
-                                ret = dict_set_int64 (dict, size_key,
-                                                      local->stripe_size);
-                                if (ret)
-                                        gf_log (this->name, GF_LOG_ERROR,
-                                                "%s: set stripe-size failed",
-                                                local->loc.path);
-                                ret = dict_set_int32 (dict, count_key,
-                                                      priv->child_count);
-                                if (ret)
-                                        gf_log (this->name, GF_LOG_ERROR,
-                                                "%s: set child_count failed",
-                                                local->loc.path);
-                                ret = dict_set_int32 (dict, index_key, i);
-                                if (ret)
-                                        gf_log (this->name, GF_LOG_ERROR,
-                                                "%s: set stripe-index failed",
-                                                local->loc.path);
+        prev  = cookie;
+        priv  = this->private;
+        local = frame->local;
+        trav = this->children;
 
-                                STACK_WIND (frame,
-                                            stripe_mknod_ifreg_setxattr_cbk,
-                                            priv->xl_array[i],
-                                            priv->xl_array[i]->fops->setxattr,
-                                            &local->loc, dict, 0);
+        local->call_count--;
 
-                                dict_unref (dict);
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG, "%s returned error %s",
+                        prev->this->name, strerror (op_errno));
+                local->failed = 1;
+                local->op_errno = op_errno;
+                goto out;
+        }
+
+        local->op_ret = op_ret;
+
+        local->stbuf      = *buf;
+        local->preparent  = *preparent;
+        local->postparent = *postparent;
+
+        if (uuid_is_null (local->ia_gfid))
+                uuid_copy (local->ia_gfid, buf->ia_gfid);
+        local->preparent.ia_blocks  = local->preparent_blocks;
+        local->preparent.ia_size    = local->preparent_size;
+        local->postparent.ia_blocks = local->postparent_blocks;
+        local->postparent.ia_size   = local->postparent_size;
+        local->stbuf.ia_size        = local->stbuf_size;
+        local->stbuf.ia_blocks      = local->stbuf_blocks;
+
+        trav = trav->next;
+        while (trav) {
+                if (priv->xattr_supported) {
+                        dict = dict_new ();
+                        if (!dict) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "failed to allocate dict %s", local->loc.path);
                         }
-                        return 0;
+                        need_unref = 1;
+
+                        dict_copy (local->xattr, dict);
+
+                        ret = stripe_xattr_request_build (this, dict,
+                                                          local->stripe_size,
+                                                          priv->child_count, i,
+							  priv->coalesce);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to build xattr request");
+
+                } else {
+                        dict = local->xattr;
                 }
 
-                /* Create itself has failed.. so return
-                   without setxattring */
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (mknod, frame, local->op_ret, local->op_errno,
-                                     local->inode, &local->stbuf,
-                                     &local->preparent, &local->postparent);
+                STACK_WIND (frame, stripe_mknod_ifreg_cbk,
+                            trav->xlator, trav->xlator->fops->mknod,
+                            &local->loc, local->mode, local->rdev, 0, dict);
+                trav = trav->next;
+                i++;
+
+                if (dict && need_unref)
+                        dict_unref (dict);
         }
-out:
+
         return 0;
+
+out:
+
+       STRIPE_STACK_UNWIND (mknod, frame, op_ret, op_errno, NULL, NULL, NULL, NULL, NULL);
+       return 0;
 }
 
 
@@ -1436,22 +1646,25 @@ int32_t
 stripe_single_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                          int32_t op_ret, int32_t op_errno, inode_t *inode,
                          struct iatt *buf, struct iatt *preparent,
-                         struct iatt *postparent)
+                         struct iatt *postparent, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (mknod, frame, op_ret, op_errno, inode, buf,
-                             preparent, postparent);
+        STRIPE_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
         return 0;
 }
 
 
 int
 stripe_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
-              dev_t rdev, dict_t *params)
+              dev_t rdev, mode_t umask, dict_t *xdata)
 {
-        stripe_private_t *priv = NULL;
-        stripe_local_t   *local = NULL;
-        xlator_list_t    *trav = NULL;
-        int32_t           op_errno = EINVAL;
+        stripe_private_t *priv           = NULL;
+        stripe_local_t   *local          = NULL;
+        int32_t           op_errno       = EINVAL;
+        int32_t           i              = 0;
+        dict_t           *dict           = NULL;
+        int               ret            = 0;
+        int               need_unref     = 0;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -1460,7 +1673,6 @@ stripe_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
         VALIDATE_OR_GOTO (loc->inode, err);
 
         priv = this->private;
-        trav = this->children;
 
         if (priv->first_child_down) {
                 op_errno = ENOTCONN;
@@ -1479,43 +1691,63 @@ stripe_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
                 }
 
                 /* Initialization */
-                local = GF_CALLOC (1, sizeof (stripe_local_t),
-                                   gf_stripe_mt_stripe_local_t);
+                local = mem_get0 (this->local_pool);
                 if (!local) {
                         op_errno = ENOMEM;
                         goto err;
                 }
                 local->op_ret = -1;
                 local->op_errno = ENOTCONN;
-                local->stripe_size = stripe_get_matching_bs (loc->path,
-                                                             priv->pattern,
-                                                             priv->block_size);
+                local->stripe_size = stripe_get_matching_bs (loc->path, priv);
                 frame->local = local;
-                local->inode = loc->inode;
+                local->inode = inode_ref (loc->inode);
                 loc_copy (&local->loc, loc);
+                local->xattr = dict_copy_with_ref (xdata, NULL);
+                local->mode = mode;
+                local->umask = umask;
+                local->rdev = rdev;
 
                 /* Everytime in stripe lookup, all child nodes should
                    be looked up */
                 local->call_count = priv->child_count;
 
-                while (trav) {
-                        STACK_WIND (frame, stripe_mknod_ifreg_cbk,
-                                    trav->xlator, trav->xlator->fops->mknod,
-                                    loc, mode, rdev, params);
-                        trav = trav->next;
+                if (priv->xattr_supported) {
+                        dict = dict_new ();
+                        if (!dict) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "failed to allocate dict %s", loc->path);
+                        }
+                        need_unref = 1;
+
+                        dict_copy (xdata, dict);
+
+                        ret = stripe_xattr_request_build (this, dict,
+                                                          local->stripe_size,
+                                                          priv->child_count,
+                                                          i, priv->coalesce);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "failed to build xattr request");
+                } else {
+                        dict = xdata;
                 }
 
-                /* This case is handled, no need to continue further. */
+                STACK_WIND (frame, stripe_mknod_first_ifreg_cbk,
+                            FIRST_CHILD (this), FIRST_CHILD (this)->fops->mknod,
+                            loc, mode, rdev, umask, dict);
+
+                        if (dict && need_unref)
+                                dict_unref (dict);
                 return 0;
         }
 
         STACK_WIND (frame, stripe_single_mknod_cbk,
                     FIRST_CHILD(this), FIRST_CHILD(this)->fops->mknod,
-                    loc, mode, rdev, params);
+                    loc, mode, rdev, umask, xdata);
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        STRIPE_STACK_UNWIND (mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
         return 0;
 }
 
@@ -1524,11 +1756,10 @@ int32_t
 stripe_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                   int32_t op_ret, int32_t op_errno, inode_t *inode,
                   struct iatt *buf, struct iatt *preparent,
-                  struct iatt *postparent)
+                  struct iatt *postparent, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t  *local   = NULL;
-        inode_t         *local_inode = NULL;
         call_frame_t    *prev = NULL;
 
         if (!this || !frame || !frame->local || !cookie) {
@@ -1556,12 +1787,6 @@ stripe_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (op_ret >= 0) {
                         local->op_ret = 0;
 
-                        if (FIRST_CHILD(this) == prev->this) {
-                                local->inode      = inode_ref (inode);
-                                local->stbuf      = *buf;
-                                local->postparent = *postparent;
-                                local->preparent  = *preparent;
-                        }
                         local->stbuf_blocks      += buf->ia_blocks;
                         local->preparent_blocks  += preparent->ia_blocks;
                         local->postparent_blocks += postparent->ia_blocks;
@@ -1577,12 +1802,7 @@ stripe_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                if (local->failed)
-                        local->op_ret = -1;
-
-                local_inode = local->inode;
-
-                if (local->op_ret != -1) {
+                if (local->failed != -1) {
                         local->preparent.ia_blocks  = local->preparent_blocks;
                         local->preparent.ia_size    = local->preparent_size;
                         local->postparent.ia_blocks = local->postparent_blocks;
@@ -1590,22 +1810,79 @@ stripe_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->stbuf.ia_size        = local->stbuf_size;
                         local->stbuf.ia_blocks      = local->stbuf_blocks;
                 }
-                STACK_UNWIND_STRICT (mkdir, frame, local->op_ret,
+                STRIPE_STACK_UNWIND (mkdir, frame, local->op_ret,
                                      local->op_errno, local->inode,
                                      &local->stbuf, &local->preparent,
-                                     &local->postparent);
+                                     &local->postparent, NULL);
+        }
+out:
+        return 0;
+}
+
 
-                if (local_inode)
-                        inode_unref (local_inode);
+int32_t
+stripe_first_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, inode_t *inode,
+                        struct iatt *buf, struct iatt *preparent,
+                        struct iatt *postparent, dict_t *xdata)
+{
+        stripe_local_t  *local   = NULL;
+        call_frame_t    *prev = NULL;
+        xlator_list_t        *trav = NULL;
+
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
         }
+
+        prev  = cookie;
+        local = frame->local;
+        trav = this->children;
+
+        local->call_count--; /* first child is successful */
+        trav = trav->next;   /* skip first child */
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG, "%s returned error %s",
+                        prev->this->name, strerror (op_errno));
+                local->op_errno = op_errno;
+                goto out;
+        }
+
+        local->op_ret = 0;
+
+        local->inode      = inode_ref (inode);
+        local->stbuf      = *buf;
+        local->postparent = *postparent;
+        local->preparent  = *preparent;
+
+        local->stbuf_blocks      += buf->ia_blocks;
+        local->preparent_blocks  += preparent->ia_blocks;
+        local->postparent_blocks += postparent->ia_blocks;
+
+        local->stbuf_size = buf->ia_size;
+        local->preparent_size = preparent->ia_size;
+        local->postparent_size = postparent->ia_size;
+
+        while (trav) {
+                STACK_WIND (frame, stripe_mkdir_cbk, trav->xlator,
+                            trav->xlator->fops->mkdir, &local->loc, local->mode,
+                            local->umask, local->xdata);
+                trav = trav->next;
+        }
+        return 0;
 out:
+        STRIPE_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL,
+                             NULL, NULL);
+
         return 0;
+
 }
 
 
 int
 stripe_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
-              dict_t *params)
+              mode_t umask, dict_t *xdata)
 {
         stripe_private_t *priv = NULL;
         stripe_local_t   *local = NULL;
@@ -1627,27 +1904,27 @@ stripe_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         local->op_ret = -1;
         local->call_count = priv->child_count;
+        if (xdata)
+                local->xdata = dict_ref (xdata);
+        local->mode  = mode;
+        local->umask = umask;
+        loc_copy (&local->loc, loc);
         frame->local = local;
 
         /* Everytime in stripe lookup, all child nodes should be looked up */
-        while (trav) {
-                STACK_WIND (frame, stripe_mkdir_cbk,
-                            trav->xlator, trav->xlator->fops->mkdir,
-                            loc, mode, params);
-                trav = trav->next;
-        }
+        STACK_WIND (frame, stripe_first_mkdir_cbk, trav->xlator,
+                    trav->xlator->fops->mkdir, loc, mode, umask, xdata);
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        STRIPE_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
         return 0;
 }
 
@@ -1656,12 +1933,12 @@ int32_t
 stripe_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                  int32_t op_ret, int32_t op_errno, inode_t *inode,
                  struct iatt *buf, struct iatt *preparent,
-                 struct iatt *postparent)
+                 struct iatt *postparent, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t  *local   = NULL;
-        inode_t         *local_inode = NULL;
         call_frame_t    *prev = NULL;
+	stripe_fd_ctx_t *fctx = NULL;
 
         if (!this || !frame || !frame->local || !cookie) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -1688,6 +1965,16 @@ stripe_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (op_ret >= 0) {
                         local->op_ret = 0;
 
+			if (IA_ISREG(inode->ia_type)) {
+				inode_ctx_get(inode, this, (uint64_t *) &fctx);
+				if (!fctx) {
+					gf_log(this->name, GF_LOG_ERROR,
+						"failed to get stripe context");
+					op_ret = -1;
+					op_errno = EINVAL;
+				}
+			}
+
                         if (FIRST_CHILD(this) == prev->this) {
                                 local->inode      = inode_ref (inode);
                                 local->stbuf      = *buf;
@@ -1698,6 +1985,8 @@ stripe_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->preparent_blocks  += preparent->ia_blocks;
                         local->postparent_blocks += postparent->ia_blocks;
 
+			correct_file_size(buf, fctx, prev);
+
                         if (local->stbuf_size < buf->ia_size)
                                 local->stbuf_size = buf->ia_size;
                         if (local->preparent_size < preparent->ia_size)
@@ -1712,8 +2001,6 @@ stripe_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (local->failed)
                         local->op_ret = -1;
 
-                local_inode = local->inode;
-
                 if (local->op_ret != -1) {
                         local->preparent.ia_blocks  = local->preparent_blocks;
                         local->preparent.ia_size    = local->preparent_size;
@@ -1722,20 +2009,17 @@ stripe_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->stbuf.ia_size        = local->stbuf_size;
                         local->stbuf.ia_blocks      = local->stbuf_blocks;
                 }
-                STACK_UNWIND_STRICT (link, frame, local->op_ret,
+                STRIPE_STACK_UNWIND (link, frame, local->op_ret,
                                      local->op_errno, local->inode,
                                      &local->stbuf, &local->preparent,
-                                     &local->postparent);
-
-                if (local_inode)
-                        inode_unref (local_inode);
+                                     &local->postparent, NULL);
         }
 out:
         return 0;
 }
 
 int32_t
-stripe_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc)
+stripe_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc, dict_t *xdata)
 {
         xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
@@ -1758,8 +2042,7 @@ stripe_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc)
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -1773,13 +2056,13 @@ stripe_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc)
         while (trav) {
                 STACK_WIND (frame, stripe_link_cbk,
                             trav->xlator, trav->xlator->fops->link,
-                            oldloc, newloc);
+                            oldloc, newloc, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (link, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        STRIPE_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
         return 0;
 }
 
@@ -1787,12 +2070,10 @@ int32_t
 stripe_create_fail_unlink_cbk (call_frame_t *frame, void *cookie,
                                xlator_t *this, int32_t op_ret,
                                int32_t op_errno, struct iatt *preparent,
-                               struct iatt *postparent)
+                               struct iatt *postparent, dict_t *xdata)
 {
         int32_t         callcnt = 0;
-        fd_t           *lfd = NULL;
         stripe_local_t *local = NULL;
-        inode_t        *local_inode = NULL;
 
         if (!this || !frame || !frame->local) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -1808,18 +2089,9 @@ stripe_create_fail_unlink_cbk (call_frame_t *frame, void *cookie,
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                local_inode = local->inode;
-                lfd = local->fd;
-
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (create, frame, local->op_ret, local->op_errno,
+                STRIPE_STACK_UNWIND (create, frame, local->op_ret, local->op_errno,
                                      local->fd, local->inode, &local->stbuf,
-                                     &local->preparent, &local->postparent);
-
-                if (local_inode)
-                        inode_unref (local_inode);
-                if (lfd)
-                        fd_unref (lfd);
+                                     &local->preparent, &local->postparent, NULL);
         }
 out:
         return 0;
@@ -1827,16 +2099,16 @@ out:
 
 
 int32_t
-stripe_create_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                            int32_t op_ret, int32_t op_errno)
+stripe_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, fd_t *fd,
+                   inode_t *inode, struct iatt *buf, struct iatt *preparent,
+                   struct iatt *postparent, dict_t *xdata)
 {
-        inode_t          *local_inode = NULL;
-        fd_t             *lfd = NULL;
+        int32_t           callcnt = 0;
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
+        call_frame_t     *prev = NULL;
         xlator_list_t    *trav = NULL;
-        int32_t           callcnt = 0;
-        call_frame_t   *prev = NULL;
 
         if (!this || !frame || !frame->local || !cookie) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -1844,7 +2116,7 @@ stripe_create_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         }
 
         prev  = cookie;
-	priv  = this->private;
+        priv  = this->private;
         local = frame->local;
 
         LOCK (&frame->lock);
@@ -1855,13 +2127,40 @@ stripe_create_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "%s returned error %s",
                                 prev->this->name, strerror (op_errno));
-                        local->op_ret = -1;
+                        local->failed = 1;
                         local->op_errno = op_errno;
                 }
+
+                if (op_ret >= 0) {
+			if (IA_ISREG(buf->ia_type)) {
+				if (stripe_ctx_handle(this, prev, local, xdata))
+					gf_log(this->name, GF_LOG_ERROR,
+						"Error getting fctx info from "
+						"dict");
+			}
+
+                        local->op_ret = op_ret;
+
+                        local->stbuf_blocks += buf->ia_blocks;
+                        local->preparent_blocks  += preparent->ia_blocks;
+                        local->postparent_blocks += postparent->ia_blocks;
+
+			correct_file_size(buf, local->fctx, prev);
+
+                        if (local->stbuf_size < buf->ia_size)
+                                local->stbuf_size = buf->ia_size;
+                        if (local->preparent_size < preparent->ia_size)
+                                local->preparent_size = preparent->ia_size;
+                        if (local->postparent_size < postparent->ia_size)
+                                local->postparent_size = postparent->ia_size;
+                }
         }
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
+                if (local->failed)
+                        local->op_ret = -1;
+
                 if (local->op_ret == -1) {
                         local->call_count = priv->child_count;
                         trav = this->children;
@@ -1870,44 +2169,57 @@ stripe_create_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                             stripe_create_fail_unlink_cbk,
                                             trav->xlator,
                                             trav->xlator->fops->unlink,
-                                            &local->loc);
+                                            &local->loc, 0, NULL);
                                 trav = trav->next;
                         }
 
                         return 0;
                 }
 
-                lfd = local->fd;
-                local_inode = local->inode;
+                if (local->op_ret >= 0) {
+                        local->preparent.ia_blocks  = local->preparent_blocks;
+                        local->preparent.ia_size    = local->preparent_size;
+                        local->postparent.ia_blocks = local->postparent_blocks;
+                        local->postparent.ia_size   = local->postparent_size;
+                        local->stbuf.ia_size        = local->stbuf_size;
+                        local->stbuf.ia_blocks      = local->stbuf_blocks;
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (create, frame, local->op_ret, local->op_errno,
-                                     local->fd, local->inode, &local->stbuf,
-                                     &local->preparent, &local->postparent);
+			stripe_copy_xl_array(local->fctx->xl_array,
+				             priv->xl_array,
+					     local->fctx->stripe_count);
+			inode_ctx_put(local->inode, this,
+					(uint64_t) local->fctx);
+                }
 
-                if (local_inode)
-                        inode_unref (local_inode);
-                if (lfd)
-                        fd_unref (lfd);
+                /* Create itself has failed.. so return
+                   without setxattring */
+                STRIPE_STACK_UNWIND (create, frame, local->op_ret,
+                                     local->op_errno, local->fd,
+                                     local->inode, &local->stbuf,
+                                     &local->preparent, &local->postparent, NULL);
         }
+
 out:
         return 0;
 }
 
+
+
 int32_t
-stripe_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+stripe_first_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno, fd_t *fd,
                    inode_t *inode, struct iatt *buf, struct iatt *preparent,
-                   struct iatt *postparent)
+                   struct iatt *postparent, dict_t *xdata)
 {
-        int32_t           callcnt = 0;
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
-        fd_t             *lfd = NULL;
-        stripe_fd_ctx_t  *fctx = NULL;
-        inode_t          *local_inode = NULL;
         call_frame_t     *prev = NULL;
-        int               ret = 0;
+        xlator_list_t    *trav = NULL;
+        int               i    = 1;
+        dict_t           *dict = NULL;
+        loc_t            *loc  = NULL;
+        int32_t           need_unref = 0;
+        int32_t           ret  = -1;
 
         if (!this || !frame || !frame->local || !cookie) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -1917,150 +2229,89 @@ stripe_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         prev  = cookie;
         priv  = this->private;
         local = frame->local;
+        trav = this->children;
+        loc = &local->loc;
 
-        LOCK (&frame->lock);
-        {
-                callcnt = --local->call_count;
-
-                if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "%s returned error %s",
-                                prev->this->name, strerror (op_errno));
-                        if ((op_errno != ENOENT) ||
-                            (prev->this == FIRST_CHILD (this)))
-                                local->failed = 1;
-                        local->op_errno = op_errno;
-                }
-
-                if (op_ret >= 0) {
-                        local->op_ret = op_ret;
-                        /* Get the mapping in inode private */
-                        /* Get the stat buf right */
-                        if (FIRST_CHILD(this) == prev->this) {
-                                local->stbuf      = *buf;
-                                local->preparent  = *preparent;
-                                local->postparent = *postparent;
-                        }
-
-                        local->stbuf_blocks += buf->ia_blocks;
-                        local->preparent_blocks  += preparent->ia_blocks;
-                        local->postparent_blocks += postparent->ia_blocks;
+        --local->call_count;
 
-                        if (local->stbuf_size < buf->ia_size)
-                                local->stbuf_size = buf->ia_size;
-                        if (local->preparent_size < preparent->ia_size)
-                                local->preparent_size = preparent->ia_size;
-                        if (local->postparent_size < postparent->ia_size)
-                                local->postparent_size = postparent->ia_size;
-                }
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG, "%s returned error %s",
+                        prev->this->name, strerror (op_errno));
+                 local->failed = 1;
+                 local->op_errno = op_errno;
         }
-        UNLOCK (&frame->lock);
-
-        if (!callcnt) {
-                if (local->failed)
-                        local->op_ret = -1;
-
-                if (local->op_ret != -1) {
-                        local->preparent.ia_blocks  = local->preparent_blocks;
-                        local->preparent.ia_size    = local->preparent_size;
-                        local->postparent.ia_blocks = local->postparent_blocks;
-                        local->postparent.ia_size   = local->postparent_size;
-                        local->stbuf.ia_size        = local->stbuf_size;
-                        local->stbuf.ia_blocks      = local->stbuf_blocks;
-                }
-
-                /* */
-                if (local->op_ret >= 0) {
-                        fctx = GF_CALLOC (1, sizeof (stripe_fd_ctx_t),
-                                          gf_stripe_mt_stripe_fd_ctx_t);
-                        if (!fctx) {
-                                local->op_ret = -1;
-                                local->op_errno = ENOMEM;
-                                goto unwind;
-                        }
 
-                        fctx->stripe_size  = local->stripe_size;
-                        fctx->stripe_count = priv->child_count;
-                        fctx->static_array = 1;
-                        fctx->xl_array = priv->xl_array;
-                        fd_ctx_set (local->fd, this,
-                                    (uint64_t)(long)fctx);
-                }
-
-                if ((local->op_ret != -1) &&
-                    local->stripe_size && priv->xattr_supported) {
-                        /* Send a setxattr request to nodes where
-                           the files are created */
-                        int32_t        i = 0;
-                        char           size_key[256] = {0,};
-                        char           index_key[256] = {0,};
-                        char           count_key[256] = {0,};
-                        dict_t        *dict = NULL;
-
-                        sprintf (size_key,
-                                 "trusted.%s.stripe-size", this->name);
-                        sprintf (count_key,
-                                 "trusted.%s.stripe-count", this->name);
-                        sprintf (index_key,
-                                 "trusted.%s.stripe-index", this->name);
+        local->op_ret = 0;
+        /* Get the mapping in inode private */
+        /* Get the stat buf right */
+        local->stbuf      = *buf;
+        local->preparent  = *preparent;
+        local->postparent = *postparent;
 
-                        local->call_count = priv->child_count;
-                        memcpy (local->loc.inode->gfid, local->stbuf.ia_gfid, 16);
-                        for (i = 0; i < priv->child_count; i++) {
-                                dict = get_new_dict ();
-                                if (!dict) {
-                                        gf_log (this->name, GF_LOG_ERROR,
-                                                "error allocating dict");
-                                }
-                                dict_ref (dict);
+        local->stbuf_blocks += buf->ia_blocks;
+        local->preparent_blocks  += preparent->ia_blocks;
+        local->postparent_blocks += postparent->ia_blocks;
 
-                                /* TODO: check return values */
-                                ret = dict_set_int64 (dict, size_key,
-                                                      local->stripe_size);
-                                if (ret)
-                                        gf_log (this->name, GF_LOG_ERROR,
-                                                "%s: set stripe-size failed",
-                                                local->loc.path);
+        if (local->stbuf_size < buf->ia_size)
+              local->stbuf_size = buf->ia_size;
+        if (local->preparent_size < preparent->ia_size)
+              local->preparent_size = preparent->ia_size;
+        if (local->postparent_size < postparent->ia_size)
+              local->postparent_size = postparent->ia_size;
 
-                                ret = dict_set_int32 (dict, count_key,
-                                                      priv->child_count);
-                                if (ret)
-                                        gf_log (this->name, GF_LOG_ERROR,
-                                                "%s: set stripe-size failed",
-                                                local->loc.path);
+        if (local->failed)
+                local->op_ret = -1;
 
-                                ret = dict_set_int32 (dict, index_key, i);
-                                if (ret)
-                                        gf_log (this->name, GF_LOG_ERROR,
-                                                "%s: set stripe-size failed",
-                                                local->loc.path);
+        if (local->op_ret == -1) {
+                local->call_count = 1;
+                STACK_WIND (frame, stripe_create_fail_unlink_cbk,
+                            FIRST_CHILD (this), FIRST_CHILD (this)->fops->unlink,
+                            &local->loc, 0, NULL);
+                return 0;
+        }
 
-                                STACK_WIND (frame, stripe_create_setxattr_cbk,
-                                            priv->xl_array[i],
-                                            priv->xl_array[i]->fops->setxattr,
-                                            &local->loc, dict, 0);
+        if (local->op_ret >= 0) {
+                local->preparent.ia_blocks  = local->preparent_blocks;
+                local->preparent.ia_size    = local->preparent_size;
+                local->postparent.ia_blocks = local->postparent_blocks;
+                local->postparent.ia_size   = local->postparent_size;
+                local->stbuf.ia_size        = local->stbuf_size;
+                local->stbuf.ia_blocks      = local->stbuf_blocks;
+        }
 
-                                dict_unref (dict);
+        /* Send a setxattr request to nodes where the
+           files are created */
+        trav = trav->next;
+        while (trav) {
+                if (priv->xattr_supported) {
+                        dict = dict_new ();
+                        if (!dict) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "failed to allocate dict %s", loc->path);
                         }
-                        return 0;
-                }
+                        need_unref = 1;
 
-unwind:
-                /* Create itself has failed.. so return
-                   without setxattring */
-                lfd = local->fd;
-                local_inode = local->inode;
+                        dict_copy (local->xattr, dict);
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (create, frame, local->op_ret,
-                                     local->op_errno, local->fd,
-                                     local->inode, &local->stbuf,
-                                     &local->preparent, &local->postparent);
+                        ret = stripe_xattr_request_build (this, dict,
+                                                          local->stripe_size,
+                                                          priv->child_count,
+                                                          i, priv->coalesce);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "failed to build xattr request");
+                } else {
+                        dict = local->xattr;
+                }
 
-                if (local_inode)
-                        inode_unref (local_inode);
-                if (lfd)
-                        fd_unref (lfd);
+                STACK_WIND (frame, stripe_create_cbk, trav->xlator,
+                            trav->xlator->fops->create, &local->loc,
+                            local->flags, local->mode, local->umask, local->fd,
+                            dict);
+                trav = trav->next;
+                if (need_unref && dict)
+                        dict_unref (dict);
+                i++;
         }
 
 out:
@@ -2068,6 +2319,7 @@ out:
 }
 
 
+
 /**
  * stripe_create - If a block-size is specified for the 'name', create the
  *    file in all the child nodes. If not, create it in only first child.
@@ -2076,12 +2328,15 @@ out:
  */
 int32_t
 stripe_create (call_frame_t *frame, xlator_t *this, loc_t *loc,
-               int32_t flags, mode_t mode, fd_t *fd, dict_t *params)
+               int32_t flags, mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
 {
         stripe_private_t *priv = NULL;
         stripe_local_t   *local = NULL;
-        xlator_list_t    *trav = NULL;
         int32_t           op_errno = EINVAL;
+        int               ret            = 0;
+        int               need_unref     = 0;
+        int               i              = 0;
+        dict_t           *dict           = NULL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -2102,46 +2357,71 @@ stripe_create (call_frame_t *frame, xlator_t *this, loc_t *loc,
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         local->op_ret = -1;
         local->op_errno = ENOTCONN;
-        local->stripe_size = stripe_get_matching_bs (loc->path,
-                                                     priv->pattern,
-                                                     priv->block_size);
+        local->stripe_size = stripe_get_matching_bs (loc->path, priv);
         frame->local = local;
         local->inode = inode_ref (loc->inode);
         loc_copy (&local->loc, loc);
         local->fd = fd_ref (fd);
+        local->flags = flags;
+        local->mode = mode;
+        local->umask = umask;
+        if (xdata)
+                local->xattr = dict_ref (xdata);
 
         local->call_count = priv->child_count;
+        /* Send a setxattr request to nodes where the
+           files are created */
 
-        trav = this->children;
-        while (trav) {
-                STACK_WIND (frame, stripe_create_cbk, trav->xlator,
-                            trav->xlator->fops->create, loc, flags,
-                            mode, fd, params);
-                trav = trav->next;
+        if (priv->xattr_supported) {
+                dict = dict_new ();
+                if (!dict) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to allocate dict %s", loc->path);
+                }
+                need_unref = 1;
+
+                dict_copy (xdata, dict);
+
+                ret = stripe_xattr_request_build (this, dict,
+                                                  local->stripe_size,
+                                                  priv->child_count,
+                                                  i, priv->coalesce);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to build xattr request");
+        } else {
+                        dict = xdata;
         }
 
+
+        STACK_WIND (frame, stripe_first_create_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->create, loc, flags, mode,
+                    umask, fd, dict);
+
+        if (need_unref && dict)
+                dict_unref (dict);
+
+
         return 0;
 err:
-        STACK_UNWIND_STRICT (create, frame, -1, op_errno, NULL, NULL, NULL,
-                             NULL, NULL);
+        STRIPE_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL,
+                             NULL, NULL, xdata);
         return 0;
 }
 
 int32_t
 stripe_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                 int32_t op_ret, int32_t op_errno, fd_t *fd)
+                 int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
-        fd_t           *lfd = NULL;
         call_frame_t   *prev = NULL;
 
         if (!this || !frame || !frame->local || !cookie) {
@@ -2176,228 +2456,20 @@ stripe_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (local->failed)
                         local->op_ret = -1;
 
-                if (local->op_ret == -1) {
-                        if (local->fctx) {
-                                if (!local->fctx->static_array)
-                                        GF_FREE (local->fctx->xl_array);
-                                GF_FREE (local->fctx);
-                        }
-                } else {
-                        fd_ctx_set (local->fd, this,
-                                    (uint64_t)(long)local->fctx);
-                }
-
-                lfd = local->fd;
-
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (open, frame, local->op_ret,
-                                     local->op_errno, local->fd);
-                if (lfd)
-                        fd_unref (lfd);
-
+                STRIPE_STACK_UNWIND (open, frame, local->op_ret,
+                                     local->op_errno, local->fd, xdata);
         }
 out:
         return 0;
 }
 
 
-int32_t
-stripe_open_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                          int32_t op_ret, int32_t op_errno, dict_t *dict)
-{
-        int32_t           index = 0;
-        int32_t           callcnt = 0;
-        char              key[256] = {0,};
-        stripe_local_t   *local = NULL;
-        xlator_list_t    *trav = NULL;
-        stripe_private_t *priv = NULL;
-        data_t           *data = NULL;
-        call_frame_t     *prev = NULL;
-        fd_t             *lfd = NULL;
-
-        if (!this || !frame || !frame->local || !cookie) {
-                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
-                goto out;
-        }
-
-        prev  = (call_frame_t *)cookie;
-        priv  = this->private;
-        local = frame->local;
-
-        LOCK (&frame->lock);
-        {
-                callcnt = --local->call_count;
-
-                if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "%s returned error %s",
-                                prev->this->name, strerror (op_errno));
-                        local->op_ret = -1;
-                        if (local->op_errno != EIO)
-                                local->op_errno = op_errno;
-                        if ((op_errno != ENOENT) ||
-                            (prev->this == FIRST_CHILD (this)))
-                                local->failed = 1;
-                        goto unlock;
-                }
-
-                if (!dict)
-                        goto unlock;
-
-                if (!local->fctx) {
-                        local->fctx =  GF_CALLOC (1, sizeof (stripe_fd_ctx_t),
-                                                  gf_stripe_mt_stripe_fd_ctx_t);
-                        if (!local->fctx) {
-                                local->op_errno = ENOMEM;
-                                local->op_ret = -1;
-                                goto unlock;
-                        }
-
-                        local->fctx->static_array = 0;
-                }
-                /* Stripe block size */
-                sprintf (key, "trusted.%s.stripe-size", this->name);
-                data = dict_get (dict, key);
-                if (!data) {
-                        local->xattr_self_heal_needed = 1;
-                } else {
-                        if (!local->fctx->stripe_size) {
-                                local->fctx->stripe_size =
-                                        data_to_int64 (data);
-                        }
-
-                        if (local->fctx->stripe_size != data_to_int64 (data)) {
-                                gf_log (this->name, GF_LOG_WARNING,
-                                        "stripe-size mismatch in blocks");
-                                local->xattr_self_heal_needed = 1;
-                        }
-                }
-                /* Stripe count */
-                sprintf (key, "trusted.%s.stripe-count", this->name);
-                data = dict_get (dict, key);
-                if (!data) {
-                        local->xattr_self_heal_needed = 1;
-                        goto unlock;
-                }
-                if (!local->fctx->xl_array) {
-                        local->fctx->stripe_count = data_to_int32 (data);
-                        if (!local->fctx->stripe_count) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "error with stripe-count xattr");
-                                local->op_ret   = -1;
-                                local->op_errno = EIO;
-                                goto unlock;
-                        }
-
-                        local->fctx->xl_array =
-                                GF_CALLOC (local->fctx->stripe_count,
-                                           sizeof (xlator_t *),
-                                           gf_stripe_mt_xlator_t);
-                        if (!local->fctx->xl_array) {
-                                local->op_errno = ENOMEM;
-                                local->op_ret   = -1;
-                                goto unlock;
-                        }
-                }
-                if (local->fctx->stripe_count != data_to_int32 (data)) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "error with stripe-count xattr (%d != %d)",
-                                local->fctx->stripe_count, data_to_int32 (data));
-                        local->op_ret   = -1;
-                        local->op_errno = EIO;
-                        goto unlock;
-                }
-
-                /* index */
-                sprintf (key, "trusted.%s.stripe-index", this->name);
-                data = dict_get (dict, key);
-                if (!data) {
-                        local->xattr_self_heal_needed = 1;
-                        goto unlock;
-                }
-                index = data_to_int32 (data);
-                if (index > priv->child_count) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "error with stripe-index xattr (%d)", index);
-                        local->op_ret   = -1;
-                        local->op_errno = EIO;
-                        goto unlock;
-                }
-                if (local->fctx->xl_array) {
-                        if (local->fctx->xl_array[index]) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "duplicate entry @ index (%d)", index);
-                                local->op_ret   = -1;
-                                local->op_errno = EIO;
-                                goto unlock;
-                        }
-                        local->fctx->xl_array[index] = prev->this;
-                }
-                local->entry_count++;
-                local->op_ret = 0;
-        }
-unlock:
-        UNLOCK (&frame->lock);
-
-        if (!callcnt) {
-                /* TODO: if self-heal flag is set, do it */
-                if (local->xattr_self_heal_needed) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "%s: stripe info need to be healed",
-                                local->loc.path);
-                }
-
-                if (local->failed)
-                        local->op_ret = -1;
-
-                if (local->op_ret)
-                        goto err;
-
-                if (local->entry_count != local->fctx->stripe_count) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "entry-count (%d) != stripe-count (%d)",
-                                local->entry_count, local->fctx->stripe_count);
-                        local->op_ret = -1;
-                        local->op_errno = EIO;
-                        goto err;
-                }
-                if (!local->fctx->stripe_size) {
-                        gf_log (this->name, GF_LOG_ERROR, "stripe size not set");
-                        local->op_ret = -1;
-                        local->op_errno = EIO;
-                        goto err;
-                }
-
-                local->call_count = local->fctx->stripe_count;
-
-                trav = this->children;
-                while (trav) {
-                        STACK_WIND (frame, stripe_open_cbk, trav->xlator,
-                                    trav->xlator->fops->open, &local->loc,
-                                    local->flags, local->fd, 0);
-                        trav = trav->next;
-                }
-        }
-
-        return 0;
-err:
-        lfd = local->fd;
-
-        stripe_local_wipe (local);
-        STACK_UNWIND_STRICT (open, frame, local->op_ret, local->op_errno,
-                             local->fd);
-        if (lfd)
-                fd_unref (lfd);
-out:
-        return 0;
-}
-
 /**
  * stripe_open -
  */
 int32_t
 stripe_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
-             int32_t flags, fd_t *fd, int32_t wbflags)
+             int32_t flags, fd_t *fd, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
@@ -2419,8 +2491,7 @@ stripe_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -2436,52 +2507,28 @@ stripe_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
         /* Striped files */
         local->flags = flags;
         local->call_count = priv->child_count;
-        local->stripe_size = stripe_get_matching_bs (loc->path,
-                                                     priv->pattern,
-                                                     priv->block_size);
-
-        if (priv->xattr_supported) {
-                while (trav) {
-                        STACK_WIND (frame, stripe_open_getxattr_cbk,
-                                    trav->xlator, trav->xlator->fops->getxattr,
-                                    loc, NULL);
-                        trav = trav->next;
-                }
-                return 0;
-        }
-        local->fctx =  GF_CALLOC (1, sizeof (stripe_fd_ctx_t),
-                                  gf_stripe_mt_stripe_fd_ctx_t);
-        if (!local->fctx) {
-                op_errno = ENOMEM;
-                goto err;
-        }
-
-        local->fctx->static_array = 1;
-        local->fctx->stripe_size  = local->stripe_size;
-        local->fctx->stripe_count = priv->child_count;
-        local->fctx->xl_array     = priv->xl_array;
+        local->stripe_size = stripe_get_matching_bs (loc->path, priv);
 
         while (trav) {
                 STACK_WIND (frame, stripe_open_cbk, trav->xlator,
                             trav->xlator->fops->open,
                             &local->loc, local->flags, local->fd,
-                            wbflags);
+                            xdata);
                 trav = trav->next;
         }
         return 0;
 err:
-        STACK_UNWIND_STRICT (open, frame, -1, op_errno, NULL);
+        STRIPE_STACK_UNWIND (open, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
 
 int32_t
 stripe_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, fd_t *fd)
+                    int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
-        fd_t           *local_fd = NULL;
         call_frame_t   *prev = NULL;
 
         if (!this || !frame || !frame->local || !cookie) {
@@ -2510,11 +2557,8 @@ stripe_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                local_fd = local->fd;
-                STACK_UNWIND_STRICT (opendir, frame, local->op_ret,
-                                     local->op_errno, local->fd);
-                if (local_fd)
-                        fd_unref (local_fd);
+                STRIPE_STACK_UNWIND (opendir, frame, local->op_ret,
+                                     local->op_errno, local->fd, NULL);
         }
 out:
         return 0;
@@ -2522,7 +2566,7 @@ out:
 
 
 int32_t
-stripe_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
+stripe_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd, dict_t *xdata)
 {
         xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
@@ -2544,8 +2588,7 @@ stripe_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
         }
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -2556,19 +2599,19 @@ stripe_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
 
         while (trav) {
                 STACK_WIND (frame, stripe_opendir_cbk, trav->xlator,
-                            trav->xlator->fops->opendir, loc, fd);
+                            trav->xlator->fops->opendir, loc, fd, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (opendir, frame, -1, op_errno, NULL);
+        STRIPE_STACK_UNWIND (opendir, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
 int32_t
 stripe_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-               int32_t op_ret, int32_t op_errno, struct gf_flock *lock)
+               int32_t op_ret, int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
@@ -2607,8 +2650,8 @@ stripe_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         if (!callcnt) {
                 if (local->failed)
                         local->op_ret = -1;
-                STACK_UNWIND_STRICT (lk, frame, local->op_ret,
-                                     local->op_errno, &local->lock);
+                STRIPE_STACK_UNWIND (lk, frame, local->op_ret,
+                                     local->op_errno, &local->lock, NULL);
         }
 out:
         return 0;
@@ -2616,7 +2659,7 @@ out:
 
 int32_t
 stripe_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
-           struct gf_flock *lock)
+           struct gf_flock *lock, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         xlator_list_t    *trav = NULL;
@@ -2632,8 +2675,7 @@ stripe_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
         priv = this->private;
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -2644,20 +2686,20 @@ stripe_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
 
         while (trav) {
                 STACK_WIND (frame, stripe_lk_cbk, trav->xlator,
-                            trav->xlator->fops->lk, fd, cmd, lock);
+                            trav->xlator->fops->lk, fd, cmd, lock, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (lk, frame, -1, op_errno, NULL);
+        STRIPE_STACK_UNWIND (lk, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
 
 int32_t
 stripe_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                  int32_t op_ret, int32_t op_errno)
+                  int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local   = NULL;
@@ -2693,16 +2735,15 @@ stripe_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (local->failed)
                         local->op_ret = -1;
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (flush, frame, local->op_ret,
-                                     local->op_errno);
+                STRIPE_STACK_UNWIND (flush, frame, local->op_ret,
+                                     local->op_errno, NULL);
         }
 out:
         return 0;
 }
 
 int32_t
-stripe_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
+stripe_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
@@ -2722,8 +2763,7 @@ stripe_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
                 goto err;
         }
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -2734,13 +2774,13 @@ stripe_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
 
         while (trav) {
                 STACK_WIND (frame, stripe_flush_cbk,  trav->xlator,
-                            trav->xlator->fops->flush, fd);
+                            trav->xlator->fops->flush, fd, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (flush, frame, -1, op_errno);
+        STRIPE_STACK_UNWIND (flush, frame, -1, op_errno, NULL);
         return 0;
 }
 
@@ -2749,7 +2789,7 @@ err:
 int32_t
 stripe_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                   int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                  struct iatt *postbuf)
+                  struct iatt *postbuf, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local   = NULL;
@@ -2785,6 +2825,9 @@ stripe_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->prebuf_blocks  += prebuf->ia_blocks;
                         local->postbuf_blocks += postbuf->ia_blocks;
 
+			correct_file_size(prebuf, local->fctx, prev);
+			correct_file_size(postbuf, local->fctx, prev);
+
                         if (local->prebuf_size < prebuf->ia_size)
                                 local->prebuf_size = prebuf->ia_size;
 
@@ -2805,21 +2848,21 @@ stripe_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->post_buf.ia_size   = local->postbuf_size;
                 }
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (fsync, frame, local->op_ret,
+                STRIPE_STACK_UNWIND (fsync, frame, local->op_ret,
                                      local->op_errno, &local->pre_buf,
-                                     &local->post_buf);
+                                     &local->post_buf, NULL);
         }
 out:
         return 0;
 }
 
 int32_t
-stripe_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
+stripe_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
         xlator_list_t    *trav = NULL;
+	stripe_fd_ctx_t  *fctx = NULL;
         int32_t           op_errno = 1;
 
         VALIDATE_OR_GOTO (frame, err);
@@ -2831,31 +2874,38 @@ stripe_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
         trav = this->children;
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
-        local->op_ret = -1;
+
         frame->local = local;
+
+	inode_ctx_get(fd->inode, this, (uint64_t *) &fctx);
+	if (!fctx) {
+		op_errno = EINVAL;
+		goto err;
+	}
+	local->fctx = fctx;
+        local->op_ret = -1;
         local->call_count = priv->child_count;
 
         while (trav) {
                 STACK_WIND (frame, stripe_fsync_cbk, trav->xlator,
-                            trav->xlator->fops->fsync, fd, flags);
+                            trav->xlator->fops->fsync, fd, flags, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (fsync, frame, -1, op_errno, NULL, NULL);
+        STRIPE_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 int32_t
 stripe_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                  int32_t op_ret, int32_t op_errno, struct iatt *buf)
+                  int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
@@ -2890,6 +2940,9 @@ stripe_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                 local->stbuf = *buf;
 
                         local->stbuf_blocks += buf->ia_blocks;
+
+			correct_file_size(buf, local->fctx, prev);
+
                         if (local->stbuf_size < buf->ia_size)
                                 local->stbuf_size = buf->ia_size;
                 }
@@ -2905,9 +2958,8 @@ stripe_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         local->stbuf.ia_blocks = local->stbuf_blocks;
                 }
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (fstat, frame, local->op_ret,
-                                     local->op_errno, &local->stbuf);
+                STRIPE_STACK_UNWIND (fstat, frame, local->op_ret,
+                                     local->op_errno, &local->stbuf, NULL);
         }
 
 out:
@@ -2917,11 +2969,12 @@ out:
 int32_t
 stripe_fstat (call_frame_t *frame,
               xlator_t *this,
-              fd_t *fd)
+              fd_t *fd, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
         xlator_list_t    *trav = NULL;
+	stripe_fd_ctx_t  *fctx = NULL;
         int32_t           op_errno = 1;
 
         VALIDATE_OR_GOTO (frame, err);
@@ -2933,8 +2986,7 @@ stripe_fstat (call_frame_t *frame,
         trav = this->children;
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -2943,26 +2995,35 @@ stripe_fstat (call_frame_t *frame,
         frame->local = local;
         local->call_count = priv->child_count;
 
+	if (IA_ISREG(fd->inode->ia_type)) {
+		inode_ctx_get(fd->inode, this, (uint64_t *) &fctx);
+		if (!fctx)
+			goto err;
+		local->fctx = fctx;
+	}
+
         while (trav) {
                 STACK_WIND (frame, stripe_fstat_cbk, trav->xlator,
-                            trav->xlator->fops->fstat, fd);
+                            trav->xlator->fops->fstat, fd, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (fstat, frame, -1, op_errno, NULL);
+        STRIPE_STACK_UNWIND (fstat, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
 
 int32_t
-stripe_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset)
+stripe_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
-        xlator_list_t    *trav = NULL;
-        int32_t           op_errno = 1;
+	stripe_fd_ctx_t  *fctx = NULL;
+	int		  i, eof_idx;
+	off_t		  dest_offset, tmp_offset;
+        int32_t		  op_errno = 1;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -2970,11 +3031,9 @@ stripe_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset)
         VALIDATE_OR_GOTO (fd->inode, err);
 
         priv = this->private;
-        trav = this->children;
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -2983,22 +3042,60 @@ stripe_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset)
         frame->local = local;
         local->call_count = priv->child_count;
 
-        while (trav) {
-                STACK_WIND (frame, stripe_truncate_cbk, trav->xlator,
-                            trav->xlator->fops->ftruncate, fd, offset);
-                trav = trav->next;
-        }
+	inode_ctx_get(fd->inode, this, (uint64_t *) &fctx);
+	if (!fctx) {
+		gf_log(this->name, GF_LOG_ERROR, "no stripe context");
+		op_errno = EINVAL;
+		goto err;
+	}
+	if (!fctx->stripe_count) {
+		gf_log(this->name, GF_LOG_ERROR, "no stripe count");
+		op_errno = EINVAL;
+		goto err;
+	}
+
+	local->fctx = fctx;
+	eof_idx = (offset / fctx->stripe_size) % fctx->stripe_count;
+
+	for (i = 0; i < fctx->stripe_count; i++) {
+		if (!fctx->xl_array[i]) {
+			gf_log(this->name, GF_LOG_ERROR, "no xlator at index "
+				"%d", i);
+			op_errno = EINVAL;
+			goto err;
+		}
+
+		if (fctx->stripe_coalesce) {
+			if (i < eof_idx)
+				tmp_offset = roof(offset, fctx->stripe_size *
+						fctx->stripe_count);
+			else if (i > eof_idx)
+				tmp_offset = floor(offset, fctx->stripe_size *
+						fctx->stripe_count);
+			else
+				tmp_offset = offset;
+
+			dest_offset = coalesced_offset(tmp_offset,
+				fctx->stripe_size, fctx->stripe_count);
+		} else {
+			dest_offset = offset;
+		}
+
+		STACK_WIND(frame, stripe_truncate_cbk, fctx->xl_array[i],
+			fctx->xl_array[i]->fops->ftruncate, fd, dest_offset,
+			NULL);
+	}
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (ftruncate, frame, -1, op_errno, NULL, NULL);
+        STRIPE_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 
 int32_t
 stripe_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                     int32_t op_ret, int32_t op_errno)
+                     int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local   = NULL;
@@ -3034,16 +3131,15 @@ stripe_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (local->failed)
                         local->op_ret = -1;
 
-                stripe_local_wipe (local);
-                STACK_UNWIND_STRICT (fsyncdir, frame, local->op_ret,
-                                     local->op_errno);
+                STRIPE_STACK_UNWIND (fsyncdir, frame, local->op_ret,
+                                     local->op_errno, NULL);
         }
 out:
         return 0;
 }
 
 int32_t
-stripe_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
+stripe_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
@@ -3059,8 +3155,7 @@ stripe_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
         trav = this->children;
 
         /* Initialization */
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -3071,20 +3166,20 @@ stripe_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
 
         while (trav) {
                 STACK_WIND (frame, stripe_fsyncdir_cbk, trav->xlator,
-                            trav->xlator->fops->fsyncdir, fd, flags);
+                            trav->xlator->fops->fsyncdir, fd, flags, NULL);
                 trav = trav->next;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (fsyncdir, frame, -1, op_errno);
+        STRIPE_STACK_UNWIND (fsyncdir, frame, -1, op_errno, NULL);
         return 0;
 }
 
 
 int32_t
 stripe_readv_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int32_t op_ret, int32_t op_errno, struct iatt *buf)
+                        int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
 {
         int32_t         i = 0;
         int32_t         callcnt = 0;
@@ -3094,6 +3189,7 @@ stripe_readv_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         struct iatt     tmp_stbuf = {0,};
         struct iobref  *tmp_iobref = NULL;
         struct iobuf   *iobuf = NULL;
+	call_frame_t   *prev = NULL;
 
         if (!this || !frame || !frame->local) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -3101,13 +3197,16 @@ stripe_readv_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         }
 
         local = frame->local;
+	prev = cookie;
 
         LOCK (&frame->lock);
         {
                 callcnt = --local->call_count;
-                if (op_ret != -1)
+                if (op_ret != -1) {
+			correct_file_size(buf, local->fctx, prev);
                         if (local->stbuf_size < buf->ia_size)
                                 local->stbuf_size = buf->ia_size;
+		}
         }
         UNLOCK (&frame->lock);
 
@@ -3136,7 +3235,8 @@ stripe_readv_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                 vec[count].iov_len  =
                                         (local->replies[i].requested_size -
                                          local->replies[i].op_ret);
-                                iobuf = iobuf_get (this->ctx->iobuf_pool);
+                                iobuf = iobuf_get2 (this->ctx->iobuf_pool,
+                                                    vec[count].iov_len);
                                 if (!iobuf) {
                                         gf_log (this->name, GF_LOG_ERROR,
                                                 "Out of memory.");
@@ -3145,9 +3245,11 @@ stripe_readv_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                                         goto done;
                                 }
                                 memset (iobuf->ptr, 0, vec[count].iov_len);
-                                iobref_add (local->iobref, iobuf);
                                 vec[count].iov_base = iobuf->ptr;
 
+                                iobref_add (local->iobref, iobuf);
+                                iobuf_unref(iobuf);
+
                                 op_ret += vec[count].iov_len;
                                 count++;
                         }
@@ -3164,13 +3266,11 @@ stripe_readv_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         done:
                 GF_FREE (local->replies);
                 tmp_iobref = local->iobref;
-                fd_unref (local->fd);
-                STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vec,
-                                     count, &tmp_stbuf, tmp_iobref);
+                STRIPE_STACK_UNWIND (readv, frame, op_ret, op_errno, vec,
+                                     count, &tmp_stbuf, tmp_iobref, NULL);
 
                 iobref_unref (tmp_iobref);
-                if (vec)
-                        GF_FREE (vec);
+                GF_FREE (vec);
         }
 out:
         return 0;
@@ -3183,7 +3283,7 @@ out:
 int32_t
 stripe_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                   int32_t op_ret, int32_t op_errno, struct iovec *vector,
-                  int32_t count, struct iatt *stbuf, struct iobref *iobref)
+                  int32_t count, struct iatt *stbuf, struct iobref *iobref, dict_t *xdata)
 {
         int32_t         index = 0;
         int32_t         callcnt = 0;
@@ -3194,8 +3294,10 @@ stripe_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         stripe_local_t *local = NULL;
         struct iovec   *final_vec = NULL;
         struct iatt     tmp_stbuf = {0,};
+        struct iatt    *tmp_stbuf_p = NULL; //need it for a warning
         struct iobref  *tmp_iobref = NULL;
         stripe_fd_ctx_t  *fctx = NULL;
+	call_frame_t	*prev = NULL;
 
         if (!this || !frame || !frame->local || !cookie) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -3204,6 +3306,7 @@ stripe_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         local  = frame->local;
         index  = local->node_index;
+	prev = cookie;
         mframe = local->orig_frame;
         if (!mframe)
                 goto out;
@@ -3224,6 +3327,12 @@ stripe_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                         mlocal->replies[index].count  = count;
                         mlocal->replies[index].vector = iov_dup (vector, count);
 
+			correct_file_size(stbuf, fctx, prev);
+
+                        if (local->stbuf_size < stbuf->ia_size)
+                                local->stbuf_size = stbuf->ia_size;
+                        local->stbuf_blocks += stbuf->ia_blocks;
+
                         if (!mlocal->iobref)
                                 mlocal->iobref = iobref_new ();
                         iobref_merge (mlocal->iobref, iobref);
@@ -3280,18 +3389,21 @@ stripe_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                  * cause any bugs at higher levels */
                 memcpy (&tmp_stbuf, &mlocal->replies[0].stbuf,
                         sizeof (struct iatt));
+                tmp_stbuf.ia_size = local->stbuf_size;
+                tmp_stbuf.ia_blocks = local->stbuf_blocks;
 
         done:
                 /* */
                 GF_FREE (mlocal->replies);
                 tmp_iobref = mlocal->iobref;
-                fd_unref (mlocal->fd);
-                STACK_UNWIND_STRICT (readv, mframe, op_ret, op_errno, final_vec,
-                                     final_count, &tmp_stbuf, tmp_iobref);
+                /* work around for nfs truncated read. Bug 3774 */
+                tmp_stbuf_p = &tmp_stbuf;
+                WIPE (tmp_stbuf_p);
+                STRIPE_STACK_UNWIND (readv, mframe, op_ret, op_errno, final_vec,
+                                     final_count, &tmp_stbuf, tmp_iobref, NULL);
 
                 iobref_unref (tmp_iobref);
-                if (final_vec)
-                        GF_FREE (final_vec);
+                GF_FREE (final_vec);
         }
 
         goto out;
@@ -3303,11 +3415,11 @@ check_size:
                 STACK_WIND (mframe, stripe_readv_fstat_cbk,
                             (fctx->xl_array[index]),
                             (fctx->xl_array[index])->fops->fstat,
-                            mlocal->fd);
+                            mlocal->fd, NULL);
         }
 
 out:
-        STACK_DESTROY (frame->root);
+        STRIPE_STACK_DESTROY (frame);
 end:
         return 0;
 }
@@ -3315,7 +3427,7 @@ end:
 
 int32_t
 stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
-              size_t size, off_t offset)
+              size_t size, off_t offset, uint32_t flags, dict_t *xdata)
 {
         int32_t           op_errno = EINVAL;
         int32_t           idx = 0;
@@ -3328,6 +3440,7 @@ stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
         uint64_t          stripe_size = 0;
         off_t             rounded_start = 0;
         off_t             frame_offset = offset;
+	off_t		  dest_offset = 0;
         stripe_local_t   *local = NULL;
         call_frame_t     *rframe = NULL;
         stripe_local_t   *rlocal = NULL;
@@ -3338,7 +3451,7 @@ stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
         VALIDATE_OR_GOTO (fd, err);
         VALIDATE_OR_GOTO (fd->inode, err);
 
-        fd_ctx_get (fd, this, &tmp_fctx);
+        inode_ctx_get (fd->inode, this, &tmp_fctx);
         if (!tmp_fctx) {
                 op_errno = EBADFD;
                 goto err;
@@ -3346,6 +3459,8 @@ stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
         fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
         stripe_size = fctx->stripe_size;
 
+        STRIPE_VALIDATE_FCTX (fctx, err);
+
         if (!stripe_size) {
                 gf_log (this->name, GF_LOG_DEBUG,
                         "Wrong stripe size for the file");
@@ -3360,8 +3475,7 @@ stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
         rounded_end = roof (offset+size, stripe_size);
         num_stripe = (rounded_end- rounded_start)/stripe_size;
 
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -3369,8 +3483,8 @@ stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
         frame->local = local;
 
         /* This is where all the vectors should be copied. */
-        local->replies = GF_CALLOC (num_stripe, sizeof (struct readv_replies),
-                                    gf_stripe_mt_readv_replies);
+        local->replies = GF_CALLOC (num_stripe, sizeof (struct stripe_replies),
+                                    gf_stripe_mt_stripe_replies);
         if (!local->replies) {
                 op_errno = ENOMEM;
                 goto err;
@@ -3385,8 +3499,7 @@ stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
 
         for (index = off_index; index < (num_stripe + off_index); index++) {
                 rframe = copy_frame (frame);
-                rlocal = GF_CALLOC (1, sizeof (stripe_local_t),
-                                    gf_stripe_mt_stripe_local_t);
+                rlocal = mem_get0 (this->local_pool);
                 if (!rlocal) {
                         op_errno = ENOMEM;
                         goto err;
@@ -3400,19 +3513,26 @@ stripe_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
                 rlocal->readv_size = frame_size;
                 rframe->local = rlocal;
                 idx = (index % fctx->stripe_count);
+
+		if (fctx->stripe_coalesce)
+			dest_offset = coalesced_offset(frame_offset,
+				stripe_size, fctx->stripe_count);
+		else
+			dest_offset = frame_offset;
+
                 STACK_WIND (rframe, stripe_readv_cbk, fctx->xl_array[idx],
                             fctx->xl_array[idx]->fops->readv,
-                            fd, frame_size, frame_offset);
+                            fd, frame_size, dest_offset, flags, xdata);
 
                 frame_offset += frame_size;
         }
 
         return 0;
 err:
-        if (local && local->fd)
-                fd_unref (local->fd);
+        if (rframe)
+                STRIPE_STACK_DESTROY (rframe);
 
-        STACK_UNWIND_STRICT (readv, frame, -1, op_errno, NULL, 0, NULL, NULL);
+        STRIPE_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, NULL, NULL);
         return 0;
 }
 
@@ -3420,11 +3540,15 @@ err:
 int32_t
 stripe_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                    int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
-                   struct iatt *postbuf)
+                   struct iatt *postbuf, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
+	stripe_local_t *mlocal = NULL;
         call_frame_t   *prev = NULL;
+	call_frame_t   *mframe = NULL;
+	struct stripe_replies *reply = NULL;
+	int32_t		i = 0;
 
         if (!this || !frame || !frame->local || !cookie) {
                 gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
@@ -3433,39 +3557,82 @@ stripe_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         prev  = cookie;
         local = frame->local;
+	mframe = local->orig_frame;
+	mlocal = mframe->local;
 
         LOCK(&frame->lock);
         {
-                callcnt = ++local->call_count;
+                callcnt = ++mlocal->call_count;
+
+		mlocal->replies[local->node_index].op_ret = op_ret;
+		mlocal->replies[local->node_index].op_errno = op_errno;
 
-                if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "%s returned error %s",
-                                prev->this->name, strerror (op_errno));
-                        local->op_errno = op_errno;
-                        local->op_ret = -1;
-                }
                 if (op_ret >= 0) {
-                        local->op_ret += op_ret;
-                        local->post_buf = *postbuf;
-                        local->pre_buf = *prebuf;
+                        mlocal->post_buf = *postbuf;
+                        mlocal->pre_buf = *prebuf;
+
+			mlocal->prebuf_blocks  += prebuf->ia_blocks;
+			mlocal->postbuf_blocks += postbuf->ia_blocks;
+
+			correct_file_size(prebuf, mlocal->fctx, prev);
+			correct_file_size(postbuf, mlocal->fctx, prev);
+
+			if (mlocal->prebuf_size < prebuf->ia_size)
+				mlocal->prebuf_size = prebuf->ia_size;
+			if (mlocal->postbuf_size < postbuf->ia_size)
+				mlocal->postbuf_size = postbuf->ia_size;
                 }
         }
         UNLOCK (&frame->lock);
 
-        if ((callcnt == local->wind_count) && local->unwind) {
-                STACK_UNWIND_STRICT (writev, frame, local->op_ret,
-                                     local->op_errno, &local->pre_buf,
-                                     &local->post_buf);
+        if ((callcnt == mlocal->wind_count) && mlocal->unwind) {
+		mlocal->pre_buf.ia_size = mlocal->prebuf_size;
+		mlocal->pre_buf.ia_blocks = mlocal->prebuf_blocks;
+		mlocal->post_buf.ia_size = mlocal->postbuf_size;
+		mlocal->post_buf.ia_blocks = mlocal->postbuf_blocks;
+
+		/*
+		 * Only return the number of consecutively written bytes up until
+		 * the first error. Only return an error if it occurs first.
+		 *
+		 * When a short write occurs, the application should retry at the
+		 * appropriate offset, at which point we'll potentially pass back
+		 * the error.
+		 */
+		for (i = 0, reply = mlocal->replies; i < mlocal->wind_count;
+			i++, reply++) {
+			if (reply->op_ret == -1) {
+				gf_log(this->name, GF_LOG_DEBUG, "reply %d "
+					"returned error %s", i,
+					strerror(reply->op_errno));
+				if (!mlocal->op_ret) {
+					mlocal->op_ret = -1;
+					mlocal->op_errno = reply->op_errno;
+				}
+				break;
+			}
+
+			mlocal->op_ret += reply->op_ret;
+
+			if (reply->op_ret < reply->requested_size)
+				break;
+		}
+
+		GF_FREE(mlocal->replies);
+
+                STRIPE_STACK_UNWIND (writev, mframe, mlocal->op_ret,
+                                     mlocal->op_errno, &mlocal->pre_buf,
+                                     &mlocal->post_buf, NULL);
         }
 out:
+	STRIPE_STACK_DESTROY(frame);
         return 0;
 }
 
 int32_t
 stripe_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
                struct iovec *vector, int32_t count, off_t offset,
-               struct iobref *iobref)
+               uint32_t flags, struct iobref *iobref, dict_t *xdata)
 {
         struct iovec     *tmp_vec = NULL;
         stripe_local_t   *local = NULL;
@@ -3479,13 +3646,19 @@ stripe_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
         off_t             fill_size = 0;
         uint64_t          stripe_size = 0;
         uint64_t          tmp_fctx = 0;
+	off_t		  dest_offset = 0;
+	off_t		  rounded_start = 0;
+	off_t		  rounded_end = 0;
+	int32_t		  total_chunks = 0;
+	call_frame_t	  *wframe = NULL;
+	stripe_local_t	  *wlocal = NULL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (fd, err);
         VALIDATE_OR_GOTO (fd->inode, err);
 
-        fd_ctx_get (fd, this, &tmp_fctx);
+        inode_ctx_get (fd->inode, this, &tmp_fctx);
         if (!tmp_fctx) {
                 op_errno = EINVAL;
                 goto err;
@@ -3493,22 +3666,51 @@ stripe_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
         fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
         stripe_size = fctx->stripe_size;
 
+        STRIPE_VALIDATE_FCTX (fctx, err);
+
         /* File has to be stripped across the child nodes */
         for (idx = 0; idx< count; idx ++) {
                 total_size += vector[idx].iov_len;
         }
         remaining_size = total_size;
 
-        local = GF_CALLOC (1, sizeof (stripe_local_t),
-                           gf_stripe_mt_stripe_local_t);
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         frame->local = local;
         local->stripe_size = stripe_size;
+	local->fctx = fctx;
+
+        if (!stripe_size) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Wrong stripe size for the file");
+                op_errno = EINVAL;
+                goto err;
+        }
+
+	rounded_start = floor(offset, stripe_size);
+	rounded_end = roof(offset + total_size, stripe_size);
+	total_chunks = (rounded_end - rounded_start) / stripe_size;
+	local->replies = GF_CALLOC(total_chunks, sizeof(struct stripe_replies),
+				gf_stripe_mt_stripe_replies);
+	if (!local->replies) {
+		op_errno = ENOMEM;
+		goto err;
+	}
 
+	total_chunks = 0;
         while (1) {
+		wframe = copy_frame(frame);
+		wlocal = mem_get0(this->local_pool);
+		if (!wlocal) {
+			op_errno = ENOMEM;
+			goto err;
+		}
+		wlocal->orig_frame = frame;
+		wframe->local = wlocal;
+
                 /* Send striped chunk of the vector to child
                    nodes appropriately. */
                 idx = (((offset + offset_offset) /
@@ -3536,32 +3738,554 @@ stripe_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
                 if (remaining_size == 0)
                         local->unwind = 1;
 
-                STACK_WIND (frame, stripe_writev_cbk, fctx->xl_array[idx],
+		/*
+		 * Store off the request index (with respect to the chunk of the
+		 * initial offset) and the size of the request. This is required
+		 * in the callback to calculate an appropriate return value in
+		 * the event of a write failure in one or more requests.
+		 */
+		wlocal->node_index = total_chunks;
+		local->replies[total_chunks].requested_size = fill_size;
+
+		dest_offset = offset + offset_offset;
+		if (fctx->stripe_coalesce)
+			dest_offset = coalesced_offset(dest_offset,
+					local->stripe_size, fctx->stripe_count);
+
+                STACK_WIND (wframe, stripe_writev_cbk, fctx->xl_array[idx],
                             fctx->xl_array[idx]->fops->writev, fd, tmp_vec,
-                            tmp_count, offset + offset_offset, iobref);
+                            tmp_count, dest_offset, flags, iobref,
+                            xdata);
+
                 GF_FREE (tmp_vec);
                 offset_offset += fill_size;
+		total_chunks++;
                 if (remaining_size == 0)
                         break;
         }
 
         return 0;
 err:
-        STACK_UNWIND_STRICT (writev, frame, -1, op_errno, NULL, NULL);
+	if (wframe)
+		STRIPE_STACK_DESTROY(wframe);
+
+        STRIPE_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 
 int32_t
-stripe_release (xlator_t *this, fd_t *fd)
+stripe_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                     struct iatt *postbuf, dict_t *xdata)
 {
+        int32_t         callcnt = 0;
+        stripe_local_t *local = NULL;
+	stripe_local_t *mlocal = NULL;
+        call_frame_t   *prev = NULL;
+	call_frame_t   *mframe = NULL;
+
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+
+        prev  = cookie;
+        local = frame->local;
+	mframe = local->orig_frame;
+	mlocal = mframe->local;
+
+        LOCK(&frame->lock);
+        {
+                callcnt = ++mlocal->call_count;
+
+                if (op_ret == 0) {
+                        mlocal->post_buf = *postbuf;
+                        mlocal->pre_buf = *prebuf;
+
+			mlocal->prebuf_blocks  += prebuf->ia_blocks;
+			mlocal->postbuf_blocks += postbuf->ia_blocks;
+
+			correct_file_size(prebuf, mlocal->fctx, prev);
+			correct_file_size(postbuf, mlocal->fctx, prev);
+
+			if (mlocal->prebuf_size < prebuf->ia_size)
+				mlocal->prebuf_size = prebuf->ia_size;
+			if (mlocal->postbuf_size < postbuf->ia_size)
+				mlocal->postbuf_size = postbuf->ia_size;
+                }
+
+		/* return the first failure */
+		if (mlocal->op_ret == 0) {
+			mlocal->op_ret = op_ret;
+			mlocal->op_errno = op_errno;
+		}
+        }
+        UNLOCK (&frame->lock);
+
+        if ((callcnt == mlocal->wind_count) && mlocal->unwind) {
+		mlocal->pre_buf.ia_size = mlocal->prebuf_size;
+		mlocal->pre_buf.ia_blocks = mlocal->prebuf_blocks;
+		mlocal->post_buf.ia_size = mlocal->postbuf_size;
+		mlocal->post_buf.ia_blocks = mlocal->postbuf_blocks;
+
+                STRIPE_STACK_UNWIND (fallocate, mframe, mlocal->op_ret,
+                                     mlocal->op_errno, &mlocal->pre_buf,
+                                     &mlocal->post_buf, NULL);
+        }
+out:
+	STRIPE_STACK_DESTROY(frame);
+        return 0;
+}
+
+int32_t
+stripe_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+		 off_t offset, size_t len, dict_t *xdata)
+{
+        stripe_local_t   *local = NULL;
+        stripe_fd_ctx_t  *fctx = NULL;
+        int32_t           op_errno = 1;
+        int32_t           idx = 0;
+        int32_t           offset_offset = 0;
+        int32_t           remaining_size = 0;
+        off_t             fill_size = 0;
+        uint64_t          stripe_size = 0;
         uint64_t          tmp_fctx = 0;
+	off_t		  dest_offset = 0;
+	call_frame_t	  *fframe = NULL;
+	stripe_local_t	  *flocal = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (fd->inode, err);
+
+        inode_ctx_get (fd->inode, this, &tmp_fctx);
+        if (!tmp_fctx) {
+                op_errno = EINVAL;
+                goto err;
+        }
+        fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
+        stripe_size = fctx->stripe_size;
+
+        STRIPE_VALIDATE_FCTX (fctx, err);
+
+        remaining_size = len;
+
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+        frame->local = local;
+        local->stripe_size = stripe_size;
+	local->fctx = fctx;
+
+        if (!stripe_size) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Wrong stripe size for the file");
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        while (1) {
+		fframe = copy_frame(frame);
+		flocal = mem_get0(this->local_pool);
+		if (!flocal) {
+			op_errno = ENOMEM;
+			goto err;
+		}
+		flocal->orig_frame = frame;
+		fframe->local = flocal;
+
+		/* send fallocate request to the associated child node */
+                idx = (((offset + offset_offset) /
+                        local->stripe_size) % fctx->stripe_count);
+
+                fill_size = (local->stripe_size -
+                             ((offset + offset_offset) % local->stripe_size));
+                if (fill_size > remaining_size)
+                        fill_size = remaining_size;
+
+                remaining_size -= fill_size;
+
+                local->wind_count++;
+                if (remaining_size == 0)
+                        local->unwind = 1;
+
+		dest_offset = offset + offset_offset;
+		if (fctx->stripe_coalesce)
+			dest_offset = coalesced_offset(dest_offset,
+					local->stripe_size, fctx->stripe_count);
+
+		/*
+		 * TODO: Create a separate handler for coalesce mode that sends a
+		 * single fallocate per-child (since the ranges are linear).
+		 */
+		STACK_WIND(fframe, stripe_fallocate_cbk, fctx->xl_array[idx],
+			   fctx->xl_array[idx]->fops->fallocate, fd, mode,
+			   dest_offset, fill_size, xdata);
+
+                offset_offset += fill_size;
+                if (remaining_size == 0)
+                        break;
+        }
+
+        return 0;
+err:
+	if (fframe)
+		STRIPE_STACK_DESTROY(fframe);
+
+        STRIPE_STACK_UNWIND (fallocate, frame, -1, op_errno, NULL, NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+stripe_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                   struct iatt *postbuf, dict_t *xdata)
+{
+        int32_t         callcnt = 0;
+        stripe_local_t *local = NULL;
+	stripe_local_t *mlocal = NULL;
+        call_frame_t   *prev = NULL;
+	call_frame_t   *mframe = NULL;
+
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+
+        prev  = cookie;
+        local = frame->local;
+	mframe = local->orig_frame;
+	mlocal = mframe->local;
+
+        LOCK(&frame->lock);
+        {
+                callcnt = ++mlocal->call_count;
+
+                if (op_ret == 0) {
+                        mlocal->post_buf = *postbuf;
+                        mlocal->pre_buf = *prebuf;
+
+			mlocal->prebuf_blocks  += prebuf->ia_blocks;
+			mlocal->postbuf_blocks += postbuf->ia_blocks;
+
+			correct_file_size(prebuf, mlocal->fctx, prev);
+			correct_file_size(postbuf, mlocal->fctx, prev);
+
+			if (mlocal->prebuf_size < prebuf->ia_size)
+				mlocal->prebuf_size = prebuf->ia_size;
+			if (mlocal->postbuf_size < postbuf->ia_size)
+				mlocal->postbuf_size = postbuf->ia_size;
+                }
+
+		/* return the first failure */
+		if (mlocal->op_ret == 0) {
+			mlocal->op_ret = op_ret;
+			mlocal->op_errno = op_errno;
+		}
+        }
+        UNLOCK (&frame->lock);
+
+        if ((callcnt == mlocal->wind_count) && mlocal->unwind) {
+		mlocal->pre_buf.ia_size = mlocal->prebuf_size;
+		mlocal->pre_buf.ia_blocks = mlocal->prebuf_blocks;
+		mlocal->post_buf.ia_size = mlocal->postbuf_size;
+		mlocal->post_buf.ia_blocks = mlocal->postbuf_blocks;
+
+                STRIPE_STACK_UNWIND (discard, mframe, mlocal->op_ret,
+                                     mlocal->op_errno, &mlocal->pre_buf,
+                                     &mlocal->post_buf, NULL);
+        }
+out:
+	STRIPE_STACK_DESTROY(frame);
+        return 0;
+}
+
+int32_t
+stripe_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+	       size_t len, dict_t *xdata)
+{
+        stripe_local_t   *local = NULL;
         stripe_fd_ctx_t  *fctx = NULL;
+        int32_t           op_errno = 1;
+        int32_t           idx = 0;
+        int32_t           offset_offset = 0;
+        int32_t           remaining_size = 0;
+        off_t             fill_size = 0;
+        uint64_t          stripe_size = 0;
+        uint64_t          tmp_fctx = 0;
+	off_t		  dest_offset = 0;
+	call_frame_t	  *fframe = NULL;
+	stripe_local_t	  *flocal = NULL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (fd->inode, err);
+
+        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;
+
+        GF_ASSERT (frame);
+
+        if (!this || !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,
+               off_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;
+
+                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;
+}
 
-        fd_ctx_del (fd, this, &tmp_fctx);
+int32_t
+stripe_release (xlator_t *this, fd_t *fd)
+{
+	return 0;
+}
+
+int
+stripe_forget (xlator_t *this, inode_t *inode)
+{
+        uint64_t          tmp_fctx = 0;
+        stripe_fd_ctx_t  *fctx = NULL;
+
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (inode, err);
+
+        (void) inode_ctx_del (inode, this, &tmp_fctx);
         if (!tmp_fctx) {
                 goto err;
         }
@@ -3572,18 +4296,17 @@ stripe_release (xlator_t *this, fd_t *fd)
                 GF_FREE (fctx->xl_array);
 
         GF_FREE (fctx);
-
 err:
-	return 0;
+        return 0;
 }
 
-
 int32_t
 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;
@@ -3601,23 +4324,28 @@ notify (xlator_t *this, int32_t event, void *data, ...)
                         if (data == priv->xl_array[i])
                                 break;
                 }
-                priv->state[i] = 1;
-                for (i = 0; i < priv->child_count; i++) {
-                        if (!priv->state[i])
-                                down_client++;
+
+                if (priv->child_count == i) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "got GF_EVENT_CHILD_UP bad subvolume %s",
+                                data? ((xlator_t *)data)->name: NULL);
+                        break;
                 }
 
                 LOCK (&priv->lock);
                 {
-                        priv->nodes_down = down_client;
                         if (data == FIRST_CHILD (this))
                                 priv->first_child_down = 0;
-                        if (!priv->nodes_down)
-                                default_notify (this, event, data);
+                        priv->last_event[i] = event;
                 }
                 UNLOCK (&priv->lock);
         }
         break;
+        case GF_EVENT_CHILD_CONNECTING:
+        {
+                // 'CONNECTING' doesn't ensure its CHILD_UP, so do nothing
+                goto out;
+        }
         case GF_EVENT_CHILD_DOWN:
         {
                 /* get an index number to set */
@@ -3625,20 +4353,19 @@ 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);
         }
@@ -3648,67 +4375,252 @@ notify (xlator_t *this, int32_t event, void *data, ...)
         {
                 /* */
                 default_notify (this, event, data);
+                goto out;
         }
         break;
         }
 
+        // Consider child as down if it's last_event is not CHILD_UP
+        for (i = 0, down_client = 0; i < priv->child_count; i++)
+                if (priv->last_event[i] != GF_EVENT_CHILD_UP)
+                        down_client++;
+
+        LOCK (&priv->lock);
+        {
+                priv->nodes_down = down_client;
+        }
+        UNLOCK (&priv->lock);
+
+        heard_from_all_children = _gf_true;
+        for (i = 0; i < priv->child_count; i++)
+                if (!priv->last_event[i])
+                        heard_from_all_children = _gf_false;
+
+        if (heard_from_all_children)
+                default_notify (this, event, data);
+out:
         return 0;
 }
 
 int
-set_stripe_block_size (xlator_t *this, stripe_private_t *priv, char *data)
+stripe_setxattr_cbk (call_frame_t *frame, void *cookie,
+                     xlator_t *this, int op_ret, int op_errno, dict_t *xdata)
 {
-        int                    ret = -1;
-        char                  *tmp_str = NULL;
-        char                  *tmp_str1 = NULL;
-        char                  *dup_str = NULL;
-        char                  *stripe_str = NULL;
-        char                  *pattern = NULL;
-        char                  *num = NULL;
-        struct stripe_options *temp_stripeopt = NULL;
-        struct stripe_options *stripe_opt = NULL;
-
-        if (!this || !priv || !data)
-                goto out;
-
-        /* Get the pattern for striping.
-           "option block-size *avi:10MB" etc */
-        stripe_str = strtok_r (data, ",", &tmp_str);
-        while (stripe_str) {
-                dup_str = gf_strdup (stripe_str);
-                stripe_opt = CALLOC (1, sizeof (struct stripe_options));
-                if (!stripe_opt) {
-                        GF_FREE (dup_str);
-                        goto out;
-                }
+        int ret = -1;
+        int call_cnt = 0;
+        stripe_local_t *local = NULL;
+
+        if (!frame || !frame->local || !this) {
+                gf_log ("", GF_LOG_ERROR, "Possible NULL deref");
+                return ret;
+        }
 
-                pattern = strtok_r (dup_str, ":", &tmp_str1);
-                num = strtok_r (NULL, ":", &tmp_str1);
-                if (!num) {
-                        num = pattern;
-                        pattern = "*";
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                call_cnt = --local->wind_count;
+
+                /**
+                 * We overwrite ->op_* values here for subsequent faliure
+                 * conditions, hence we propogate the last errno down the
+                 * stack.
+                 */
+                if (op_ret < 0) {
+                        local->op_ret = op_ret;
+                        local->op_errno = op_errno;
+                        goto unlock;
                 }
-                if (gf_string2bytesize (num, &stripe_opt->block_size) != 0) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "invalid number format \"%s\"", num);
-                        goto out;
+        }
+
+ unlock:
+        UNLOCK (&frame->lock);
+
+        if (!call_cnt) {
+                STRIPE_STACK_UNWIND (setxattr, frame, local->op_ret,
+                                     local->op_errno, xdata);
+        }
+
+        return 0;
+}
+
+#ifdef HAVE_BD_XLATOR
+int
+stripe_is_bd (dict_t *this, char *key, data_t *value, void *data)
+{
+        gf_boolean_t *is_bd = data;
+
+        if (data == NULL)
+                return 0;
+
+        if (XATTR_IS_BD (key))
+            *is_bd = _gf_true;
+
+        return 0;
+}
+
+inline gf_boolean_t
+stripe_setxattr_is_bd (dict_t *dict)
+{
+        gf_boolean_t is_bd = _gf_false;
+
+        if (dict == NULL)
+                goto out;
+
+        dict_foreach (dict, stripe_is_bd, &is_bd);
+out:
+        return is_bd;
+}
+#else
+#define stripe_setxattr_is_bd(dict) _gf_false
+#endif
+
+int
+stripe_setxattr (call_frame_t *frame, xlator_t *this,
+                 loc_t *loc, dict_t *dict, int flags, dict_t *xdata)
+{
+        int32_t           op_errno = EINVAL;
+        xlator_list_t    *trav     = NULL;
+        stripe_private_t *priv     = NULL;
+        stripe_local_t   *local    = NULL;
+        int               i        = 0;
+        gf_boolean_t      is_bd    = _gf_false;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.*stripe*", dict,
+                                   op_errno, err);
+
+        priv = this->private;
+        trav = this->children;
+
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        frame->local = local;
+        local->wind_count = priv->child_count;
+        local->op_ret = local->op_errno = 0;
+
+        is_bd = stripe_setxattr_is_bd (dict);
+
+        /**
+         * Set xattrs for directories on all subvolumes. Additionally
+         * this power is only given to a special client. Bd xlator
+         * also needs xattrs for regular files (ie LVs)
+         */
+        if (((frame->root->pid == GF_CLIENT_PID_GSYNCD) &&
+             IA_ISDIR (loc->inode->ia_type)) || is_bd) {
+                for (i = 0; i < priv->child_count; i++, trav = trav->next) {
+                        STACK_WIND (frame, stripe_setxattr_cbk,
+                                    trav->xlator, trav->xlator->fops->setxattr,
+                                    loc, dict, flags, xdata);
                 }
-                memcpy (stripe_opt->path_pattern, pattern, strlen (pattern));
+        } else {
+                local->wind_count = 1;
+                STACK_WIND (frame, stripe_setxattr_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->setxattr,
+                            loc, dict, flags, xdata);
+        }
 
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "block-size : pattern %s : size %"PRId64,
-                        stripe_opt->path_pattern, stripe_opt->block_size);
+        return 0;
+err:
+        STRIPE_STACK_UNWIND (setxattr, frame, -1,  op_errno, NULL);
+        return 0;
+}
 
-                if (!priv->pattern) {
-                        priv->pattern = stripe_opt;
-                } else {
-                        temp_stripeopt = priv->pattern;
-                        while (temp_stripeopt->next)
-                                temp_stripeopt = temp_stripeopt->next;
-                        temp_stripeopt->next = stripe_opt;
+
+int
+stripe_fsetxattr_cbk (call_frame_t *frame, void *cookie,
+                      xlator_t *this, int op_ret, int op_errno, dict_t *xdata)
+{
+        STRIPE_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+
+int
+stripe_is_special_key (dict_t *this,
+                       char *key,
+                       data_t *value,
+                       void *data)
+{
+        gf_boolean_t *is_special = NULL;
+
+        if (data == NULL) {
+                goto out;
+        }
+
+        is_special = data;
+
+        if (XATTR_IS_LOCKINFO (key) || XATTR_IS_BD (key))
+                *is_special = _gf_true;
+
+out:
+        return 0;
+}
+
+int32_t
+stripe_fsetxattr_everyone_cbk (call_frame_t *frame, void *cookie,
+                               xlator_t *this, int32_t op_ret, int32_t op_errno,
+                               dict_t *xdata)
+{
+        int             call_count = 0;
+        stripe_local_t *local      = NULL;
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                call_count = --local->wind_count;
+
+                if (op_ret < 0) {
+                        local->op_ret = op_ret;
+                        local->op_errno = op_errno;
                 }
-                stripe_str = strtok_r (NULL, ",", &tmp_str);
-                GF_FREE (dup_str);
+        }
+        UNLOCK (&frame->lock);
+
+        if (call_count == 0) {
+                STRIPE_STACK_UNWIND (fsetxattr, frame, local->op_ret,
+                                     local->op_errno, NULL);
+        }
+        return 0;
+}
+
+int
+stripe_fsetxattr_to_everyone (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                              dict_t *dict, int flags, dict_t *xdata)
+{
+        xlator_list_t    *trav  = NULL;
+        stripe_private_t *priv  = NULL;
+        int               ret   = -1;
+        stripe_local_t   *local = NULL;
+
+        priv = this->private;
+
+        local = mem_get0 (this->local_pool);
+        if (local == NULL) {
+                goto out;
+        }
+
+        frame->local = local;
+
+        local->wind_count = priv->child_count;
+
+        trav = this->children;
+
+        while (trav) {
+                STACK_WIND (frame, stripe_fsetxattr_everyone_cbk,
+                            trav->xlator, trav->xlator->fops->fsetxattr,
+                            fd, dict, flags, xdata);
+                trav = trav->next;
         }
 
         ret = 0;
@@ -3716,97 +4628,477 @@ out:
         return ret;
 }
 
-int32_t
-mem_acct_init (xlator_t *this)
+inline gf_boolean_t
+stripe_fsetxattr_is_special (dict_t *dict)
 {
-        int     ret = -1;
+        gf_boolean_t is_spl = _gf_false;
 
-        if (!this)
+        if (dict == NULL) {
                 goto out;
+        }
 
-        ret = xlator_mem_acct_init (this, gf_stripe_mt_end + 1);
+        dict_foreach (dict, stripe_is_special_key, &is_spl);
+
+out:
+        return is_spl;
+}
+
+int
+stripe_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                  dict_t *dict, int flags, dict_t *xdata)
+{
+        int32_t      op_ret = -1, ret = -1, op_errno = EINVAL;
+        gf_boolean_t is_spl = _gf_false;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.*stripe*", dict,
+                                   op_errno, err);
+
+        is_spl = stripe_fsetxattr_is_special (dict);
+        if (is_spl) {
+                ret = stripe_fsetxattr_to_everyone (frame, this, fd, dict,
+                                                    flags, xdata);
+                if (ret < 0) {
+                        op_errno = ENOMEM;
+                        goto err;
+                }
 
-        if (ret != 0) {
-                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
-                        "failed");
                 goto out;
         }
 
+        STACK_WIND (frame, stripe_fsetxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsetxattr,
+                    fd, dict, flags, xdata);
 out:
-        return ret;
+        return 0;
+err:
+        STRIPE_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, NULL);
+        return 0;
 }
+
 int
-validate_options (xlator_t *this, dict_t *options, char **op_errstr)
+stripe_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
+        STRIPE_STACK_UNWIND (removexattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
 
+int
+stripe_removexattr (call_frame_t *frame, xlator_t *this,
+                    loc_t *loc, const char *name, dict_t *xdata)
+{
+        int32_t         op_errno = EINVAL;
 
-        data_t           *data = NULL;
-        int               ret = 0;
+        VALIDATE_OR_GOTO (this, err);
+
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.*stripe*",
+                                 name, op_errno, err);
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (loc, err);
+
+        STACK_WIND (frame, stripe_removexattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->removexattr,
+                    loc, name, xdata);
+        return 0;
+err:
+        STRIPE_STACK_UNWIND (removexattr, frame, -1,  op_errno, NULL);
+        return 0;
+}
+
+
+int
+stripe_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        STRIPE_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int
+stripe_fremovexattr (call_frame_t *frame, xlator_t *this,
+                     fd_t *fd, const char *name, dict_t *xdata)
+{
+        int32_t         op_ret   = -1;
+        int32_t         op_errno = EINVAL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.*stripe*",
+                                 name, op_errno, err);
+
+        STACK_WIND (frame, stripe_fremovexattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fremovexattr,
+                    fd, name, xdata);
+        return 0;
+ err:
+        STRIPE_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int32_t
+stripe_readdirp_lookup_cbk (call_frame_t *frame, void *cookie,
+                            xlator_t *this, int op_ret, int op_errno,
+                            inode_t *inode, struct iatt *stbuf,
+                            dict_t *xattr, struct iatt *parent)
+{
+        stripe_local_t          *local          = NULL;
+        call_frame_t            *main_frame     = NULL;
+        stripe_local_t          *main_local     = NULL;
+        gf_dirent_t             *entry          = NULL;
+        call_frame_t            *prev           = NULL;
+        int                      done           = 0;
+
+        local = frame->local;
+        prev = cookie;
+
+        entry = local->dirent;
+
+        main_frame = local->orig_frame;
+        main_local = main_frame->local;
+        LOCK (&frame->lock);
+        {
+
+                local->call_count--;
+                if (!local->call_count)
+                        done = 1;
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        local->op_ret = op_ret;
+                        goto unlock;
+                }
+
+		if (stripe_ctx_handle(this, prev, local, xattr))
+			gf_log(this->name, GF_LOG_ERROR,
+				"Error getting fctx info from dict.");
+
+		correct_file_size(stbuf, local->fctx, prev);
+
+                stripe_iatt_merge (stbuf, &entry->d_stat);
+                local->stbuf_blocks += stbuf->ia_blocks;
+        }
+unlock:
+        UNLOCK(&frame->lock);
+
+        if (done) {
+                inode_ctx_put (entry->inode, this,
+                               (uint64_t) (long)local->fctx);
+
+                done = 0;
+                LOCK (&main_frame->lock);
+                {
+                        main_local->wind_count--;
+                        if (!main_local->wind_count)
+                                done = 1;
+                        if (local->op_ret == -1) {
+                                main_local->op_errno = local->op_errno;
+                                main_local->op_ret = local->op_ret;
+                        }
+                        entry->d_stat.ia_blocks = local->stbuf_blocks;
+                }
+                UNLOCK (&main_frame->lock);
+                if (done) {
+                        main_frame->local = NULL;
+                        STRIPE_STACK_UNWIND (readdir, main_frame,
+                                             main_local->op_ret,
+                                             main_local->op_errno,
+                                             &main_local->entries, NULL);
+                        gf_dirent_free (&main_local->entries);
+                        stripe_local_wipe (main_local);
+                        mem_put (main_local);
+                }
+                frame->local = NULL;
+                stripe_local_wipe (local);
+                mem_put (local);
+                STRIPE_STACK_DESTROY (frame);
+        }
+
+        return 0;
+}
+
+int32_t
+stripe_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno,
+                     gf_dirent_t *orig_entries, dict_t *xdata)
+{
+        stripe_local_t *local = NULL;
+        call_frame_t   *prev = NULL;
+        gf_dirent_t    *local_entry = NULL;
+        gf_dirent_t    *tmp_entry = NULL;
+        xlator_list_t  *trav = NULL;
+        loc_t          loc = {0, };
+        int32_t        count = 0;
         stripe_private_t *priv = NULL;
+        int32_t        subvols = 0;
+        dict_t         *xattrs = NULL;
+        call_frame_t   *local_frame = NULL;
+        stripe_local_t *local_ent = NULL;
 
-        data = dict_get (options, "block-size");
-        if (data) {
-                gf_log (this->name, GF_LOG_TRACE,"Reconfiguring Stripe"
-                                " Block-size");
-                priv = GF_CALLOC (1, sizeof (stripe_private_t),
-                                  gf_stripe_mt_stripe_private_t);
-                if (!priv) {
-                        gf_log ("",GF_LOG_ERROR, "Unable to allocate memory");
-                        ret = -1;
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+        prev  = cookie;
+        local = frame->local;
+        trav = this->children;
+        priv = this->private;
+
+        subvols = priv->child_count;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s returned error %s",
+                                prev->this->name, strerror (op_errno));
+                        local->op_errno = op_errno;
+                        local->op_ret = op_ret;
+                        goto unlock;
+                } else {
+                        local->op_ret = op_ret;
+                        list_splice_init (&orig_entries->list,
+                                          &local->entries.list);
+                        local->wind_count = op_ret;
+                }
+
+        }
+unlock:
+        UNLOCK (&frame->lock);
+
+        if (op_ret == -1)
+                goto out;
+
+        xattrs = dict_new ();
+        if (xattrs)
+                (void) stripe_xattr_request_build (this, xattrs, 0, 0, 0, 0);
+        count = op_ret;
+        list_for_each_entry_safe (local_entry, tmp_entry,
+                                  (&local->entries.list), list) {
+
+                if (!local_entry)
+                        break;
+                if (!IA_ISREG (local_entry->d_stat.ia_type) || !local_entry->inode) {
+                        LOCK (&frame->lock);
+                        {
+                                local->wind_count--;
+                                count = local->wind_count;
+                        }
+                        UNLOCK (&frame->lock);
+                        continue;
+                }
+
+                local_frame = copy_frame (frame);
+
+                if (!local_frame) {
+                        op_errno = ENOMEM;
+                        op_ret = -1;
                         goto out;
                 }
 
-                ret = set_stripe_block_size (this, priv, data->data);
-                if (ret) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "Reconfigue: Block-Size reconfiguration failed");
-                        *op_errstr = gf_strdup ("Error, could not parse list");
-                        ret = -1;
+                local_ent = mem_get0 (this->local_pool);
+                if (!local_ent) {
+                        op_errno = ENOMEM;
+                        op_ret = -1;
                         goto out;
                 }
-                gf_log (this->name, GF_LOG_TRACE,
-                        "Reconfigue: Block-Size reconfigured Successfully");
+
+                loc.inode = inode_ref (local_entry->inode);
+
+                uuid_copy (loc.gfid, local_entry->d_stat.ia_gfid);
+
+                local_ent->orig_frame = frame;
+
+                local_ent->call_count = subvols;
+
+                local_ent->dirent = local_entry;
+
+                local_frame->local = local_ent;
+
+                trav = this->children;
+                while (trav) {
+                        STACK_WIND (local_frame, stripe_readdirp_lookup_cbk,
+                                    trav->xlator, trav->xlator->fops->lookup,
+                                    &loc, xattrs);
+                        trav = trav->next;
+                }
+                loc_wipe (&loc);
+        }
+out:
+        if (!count) {
+                /* all entries are directories */
+                frame->local = NULL;
+                STRIPE_STACK_UNWIND (readdir, frame, local->op_ret,
+                                     local->op_errno, &local->entries, NULL);
+                gf_dirent_free (&local->entries);
+                stripe_local_wipe (local);
+                mem_put (local);
+        }
+        if (xattrs)
+                dict_unref (xattrs);
+        return 0;
+
+}
+int32_t
+stripe_readdirp (call_frame_t *frame, xlator_t *this,
+                 fd_t *fd, size_t size, off_t off, dict_t *xdata)
+{
+        stripe_local_t  *local  = NULL;
+        stripe_private_t *priv = NULL;
+        xlator_list_t   *trav = NULL;
+        int             op_errno = -1;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        priv = this->private;
+        trav = this->children;
+
+        if (priv->first_child_down) {
+                op_errno = ENOTCONN;
+                goto err;
+        }
+
+        /* Initialization */
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        frame->local = local;
+
+        local->fd = fd_ref (fd);
+
+        local->wind_count = 0;
+
+        local->count = 0;
+        local->op_ret = -1;
+        INIT_LIST_HEAD(&local->entries);
+
+        if (!trav)
+                goto err;
+
+        STACK_WIND (frame, stripe_readdirp_cbk, trav->xlator,
+                    trav->xlator->fops->readdirp, fd, size, off, xdata);
+        return 0;
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        STRIPE_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+
+}
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                goto out;
+
+        ret = xlator_mem_acct_init (this, gf_stripe_mt_end + 1);
+
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
+                        "failed");
+                goto out;
         }
 
 out:
-                if (priv)
-                GF_FREE (priv);
         return ret;
+}
+
+static int
+clear_pattern_list (stripe_private_t *priv)
+{
+        struct stripe_options *prev = NULL;
+        struct stripe_options *trav = NULL;
+        int                    ret = -1;
+
+        GF_VALIDATE_OR_GOTO ("stripe", priv, out);
+
+        trav = priv->pattern;
+        priv->pattern = NULL;
+        while (trav) {
+                prev = trav;
+                trav = trav->next;
+                GF_FREE (prev);
+        }
+
+        ret = 0;
+ out:
+        return ret;
+
 
 }
 
+
 int
 reconfigure (xlator_t *this, dict_t *options)
 {
 
-	stripe_private_t *priv = NULL;
-	data_t           *data = NULL;
-	int 		  ret = 0;
+        stripe_private_t *priv = NULL;
+        data_t           *data = NULL;
+        int               ret = -1;
+        volume_option_t  *opt = NULL;
 
+        GF_ASSERT (this);
+        GF_ASSERT (this->private);
 
-	priv = this->private;
+        priv = this->private;
 
-	data = dict_get (options, "block-size");
-        if (data) {
-		gf_log (this->name, GF_LOG_TRACE,"Reconfiguring Stripe"
-			" Block-size");
-                ret = set_stripe_block_size (this, priv, data->data);
-                if (ret) {
-			gf_log (this->name, GF_LOG_ERROR,
-                        "Reconfigue: Block-Size reconfiguration failed");
-                        ret = -1;
-			goto out;
-		}
-		gf_log (this->name, GF_LOG_TRACE,
-                        "Reconfigue: Block-Size reconfigured Successfully");
-	}
-        else {
-                priv->block_size = (128 * GF_UNIT_KB);
+
+        ret = 0;
+        LOCK (&priv->lock);
+        {
+                ret = clear_pattern_list (priv);
+                if (ret)
+                         goto unlock;
+
+                data = dict_get (options, "block-size");
+                if (data) {
+                        ret = set_stripe_block_size (this, priv, data->data);
+                        if (ret)
+                                goto unlock;
+                } else {
+                        opt = xlator_volume_option_get (this, "block-size");
+                        if (!opt) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "option 'block-size' not found");
+                                ret = -1;
+                                goto unlock;
+                        }
+
+                        if (gf_string2bytesize (opt->default_value, &priv->block_size)){
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Unable to set default block-size ");
+                                ret = -1;
+                                goto unlock;
+                        }
+                }
+
+		GF_OPTION_RECONF("coalesce", priv->coalesce, options, bool,
+				unlock);
         }
-                        
+ unlock:
+        UNLOCK (&priv->lock);
+        if (ret)
+                goto out;
 
-out:
-	return ret;
+        ret = 0;
+ out:
+        return ret;
 
 }
 
@@ -3819,6 +5111,7 @@ int32_t
 init (xlator_t *this)
 {
         stripe_private_t *priv = NULL;
+        volume_option_t  *opt = NULL;
         xlator_list_t    *trav = NULL;
         data_t           *data = NULL;
         int32_t           count = 0;
@@ -3862,9 +5155,9 @@ 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;
@@ -3884,41 +5177,56 @@ init (xlator_t *this)
                 goto out;
         }
 
-        priv->block_size = (128 * GF_UNIT_KB);
-        /* option stripe-pattern *avi:1GB,*pdf:4096 */
-        data = dict_get (this->options, "block-size");
-        if (!data) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "No \"option block-size <x>\" given, defaulting "
-                        "to 128KB");
-        } else {
-                ret = set_stripe_block_size (this, priv, data->data);
-                if (ret)
-                        goto out;
-        }
-
-        priv->xattr_supported = 1;
-        data = dict_get (this->options, "use-xattr");
-        if (data) {
-                if (gf_string2boolean (data->data,
-                                       &priv->xattr_supported) == -1) {
+        ret = 0;
+        LOCK (&priv->lock);
+        {
+                opt = xlator_volume_option_get (this, "block-size");
+                if (!opt) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "option 'block-size' not found");
+                        ret = -1;
+                        goto unlock;
+                }
+                if (gf_string2bytesize (opt->default_value, &priv->block_size)){
                         gf_log (this->name, GF_LOG_ERROR,
-                                "error setting hard check for extended "
-                                "attribute");
-                        //return -1;
+                                "Unable to set default block-size ");
+                        ret = -1;
+                        goto unlock;
+                }
+                /* option stripe-pattern *avi:1GB,*pdf:16K */
+                data = dict_get (this->options, "block-size");
+                if (data) {
+                        ret = set_stripe_block_size (this, priv, data->data);
+                        if (ret)
+                                goto unlock;
                 }
         }
+ unlock:
+        UNLOCK (&priv->lock);
+        if (ret)
+                goto out;
 
+        GF_OPTION_INIT ("use-xattr", priv->xattr_supported, bool, out);
         /* notify related */
         priv->nodes_down = priv->child_count;
+
+	GF_OPTION_INIT("coalesce", priv->coalesce, bool, out);
+
+        this->local_pool = mem_pool_new (stripe_local_t, 128);
+        if (!this->local_pool) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create local_t's memory pool");
+                goto out;
+        }
+
         this->private = priv;
 
         ret = 0;
 out:
         if (ret) {
                 if (priv) {
-                        if (priv->xl_array)
-                                GF_FREE (priv->xl_array);
+                        GF_FREE (priv->xl_array);
                         GF_FREE (priv);
                 }
         }
@@ -3942,15 +5250,15 @@ fini (xlator_t *this)
         priv = this->private;
         if (priv) {
                 this->private = NULL;
-                if (priv->xl_array)
-                        GF_FREE (priv->xl_array);
+                GF_FREE (priv->xl_array);
 
                 trav = priv->pattern;
                 while (trav) {
                         prev = trav;
                         trav = trav->next;
-                        FREE (prev);
+                        GF_FREE (prev);
                 }
+                GF_FREE (priv->last_event);
                 LOCK_DESTROY (&priv->lock);
                 GF_FREE (priv);
         }
@@ -3959,44 +5267,524 @@ out:
         return;
 }
 
+int32_t
+stripe_getxattr_unwind (call_frame_t *frame,
+                        int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
+
+{
+        STRIPE_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+}
+
+int
+stripe_internal_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                              int op_ret, int op_errno, dict_t *xattr,
+                              dict_t *xdata)
+{
+
+        char        size_key[256]  = {0,};
+        char        index_key[256] = {0,};
+        char        count_key[256] = {0,};
+        char        coalesce_key[256] = {0,};
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (frame->local, out);
+
+        if (!xattr || (op_ret == -1))
+            goto out;
+
+        sprintf (size_key, "trusted.%s.stripe-size", this->name);
+        sprintf (count_key, "trusted.%s.stripe-count", this->name);
+        sprintf (index_key, "trusted.%s.stripe-index", this->name);
+	sprintf (coalesce_key, "trusted.%s.stripe-coalesce", this->name);
+
+        dict_del (xattr, size_key);
+        dict_del (xattr, count_key);
+        dict_del (xattr, index_key);
+        dict_del (xattr, coalesce_key);
+
+out:
+        STRIPE_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr, xdata);
+
+        return 0;
+
+}
+
+int
+stripe_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int op_ret, int op_errno, dict_t *xattr, dict_t *xdata)
+{
+        int                     call_cnt = 0;
+        stripe_local_t         *local = NULL;
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (frame->local, out);
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                call_cnt = --local->wind_count;
+        }
+        UNLOCK (&frame->lock);
+
+        if (!xattr || (op_ret < 0))
+                goto out;
+
+        local->op_ret = 0;
+
+        if (!local->xattr) {
+                local->xattr = dict_ref (xattr);
+        } else {
+                stripe_aggregate_xattr (local->xattr, xattr);
+        }
+
+out:
+        if (!call_cnt) {
+                STRIPE_STACK_UNWIND (getxattr, frame, local->op_ret, op_errno,
+                                     local->xattr, xdata);
+        }
+
+        return 0;
+}
+
+int32_t
+stripe_vgetxattr_cbk (call_frame_t *frame, void *cookie,
+                      xlator_t *this, int32_t op_ret, int32_t op_errno,
+                      dict_t *dict, dict_t *xdata)
+{
+        stripe_local_t      *local         = NULL;
+        int32_t              callcnt       = 0;
+        int32_t              ret           = -1;
+        long                 cky           = 0;
+        void                *xattr_val     = NULL;
+        void                *xattr_serz    = NULL;
+        stripe_xattr_sort_t *xattr         = NULL;
+        dict_t              *stripe_xattr  = NULL;
+
+        if (!frame || !frame->local || !this) {
+                gf_log ("", GF_LOG_ERROR, "Possible NULL deref");
+                return ret;
+        }
+
+        local = frame->local;
+        cky = (long) cookie;
+
+        if (local->xsel[0] == '\0') {
+                gf_log (this->name, GF_LOG_ERROR, "Empty xattr in cbk");
+                return ret;
+        }
+
+        LOCK (&frame->lock);
+        {
+                callcnt = --local->wind_count;
+
+                if (!dict || (op_ret < 0))
+                        goto out;
+
+                if (!local->xattr_list)
+                        local->xattr_list = (stripe_xattr_sort_t *)
+                                GF_CALLOC (local->nallocs,
+                                           sizeof (stripe_xattr_sort_t),
+                                           gf_stripe_mt_xattr_sort_t);
+
+                if (local->xattr_list) {
+                        xattr = local->xattr_list + (int32_t) cky;
+
+                        ret = dict_get_ptr_and_len (dict, local->xsel,
+                                                    &xattr_val,
+                                                    &xattr->xattr_len);
+                        if (xattr->xattr_len == 0)
+                                goto out;
+
+                        xattr->pos = cky;
+                        xattr->xattr_value = gf_memdup (xattr_val,
+                                                        xattr->xattr_len);
+
+                        if (xattr->xattr_value != NULL)
+                                local->xattr_total_len += xattr->xattr_len + 1;
+                }
+        }
+ out:
+        UNLOCK (&frame->lock);
+
+        if (!callcnt) {
+                if (!local->xattr_total_len)
+                        goto unwind;
+
+                stripe_xattr = dict_new ();
+                if (!stripe_xattr)
+                        goto unwind;
+
+                /* select filler based on ->xsel */
+                if (XATTR_IS_PATHINFO (local->xsel))
+                        ret = stripe_fill_pathinfo_xattr (this, local,
+                                                          (char **)&xattr_serz);
+                else if (XATTR_IS_LOCKINFO (local->xsel)) {
+                        ret = stripe_fill_lockinfo_xattr (this, local,
+                                                          &xattr_serz);
+                } else {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "Unknown xattr in xattr request");
+                        goto unwind;
+                }
+
+                if (!ret) {
+                        ret = dict_set_dynptr (stripe_xattr, local->xsel,
+                                               xattr_serz,
+                                               local->xattr_total_len);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Can't set %s key in dict",
+                                        local->xsel);
+                }
+
+        unwind:
+                STRIPE_STACK_UNWIND (getxattr, frame, op_ret, op_errno,
+                                     stripe_xattr, NULL);
+
+                ret = stripe_free_xattr_str (local);
+
+                GF_FREE (local->xattr_list);
+
+                if (stripe_xattr)
+                        dict_unref (stripe_xattr);
+        }
+
+        return ret;
+}
+
+int32_t
+stripe_getxattr (call_frame_t *frame, xlator_t *this,
+                 loc_t *loc, const char *name, dict_t *xdata)
+{
+        stripe_local_t    *local    = NULL;
+        xlator_list_t     *trav     = NULL;
+        stripe_private_t  *priv     = NULL;
+        int32_t            op_errno = EINVAL;
+        int                i        = 0;
+        xlator_t         **sub_volumes;
+        int                ret      = 0;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
+
+        priv = this->private;
+        trav = this->children;
+
+        /* Initialization */
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+        local->op_ret = -1;
+        frame->local = local;
+        loc_copy (&local->loc, loc);
+
+
+        if (name && (strcmp (GF_XATTR_MARKER_KEY, name) == 0)
+            && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+                local->marker.call_count = priv->child_count;
+
+                sub_volumes = alloca ( priv->child_count *
+                                       sizeof (xlator_t *));
+                for (i = 0, trav = this->children; trav ;
+                     trav = trav->next, i++) {
+
+                        *(sub_volumes + i)  = trav->xlator;
+
+                }
+
+                if (cluster_getmarkerattr (frame, this, loc, name,
+                                           local, stripe_getxattr_unwind,
+                                           sub_volumes, priv->child_count,
+                                           MARKER_UUID_TYPE, marker_uuid_default_gauge,
+                                           priv->vol_uuid)) {
+                        op_errno = EINVAL;
+                        goto err;
+                }
+
+                return 0;
+        }
+
+        if (name && strncmp (name, GF_XATTR_QUOTA_SIZE_KEY,
+                             strlen (GF_XATTR_QUOTA_SIZE_KEY)) == 0) {
+                local->wind_count = priv->child_count;
+
+                for (i = 0, trav=this->children; i < priv->child_count; i++,
+                             trav = trav->next) {
+                        STACK_WIND (frame, stripe_getxattr_cbk,
+                                    trav->xlator, trav->xlator->fops->getxattr,
+                                    loc, name, xdata);
+                }
+
+                return 0;
+        }
+
+        if (name && (XATTR_IS_PATHINFO (name))) {
+                if (IA_ISREG (loc->inode->ia_type)) {
+                        ret = inode_ctx_get (loc->inode, this,
+                                             (uint64_t *) &local->fctx);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "stripe size unavailable from fctx"
+                                        " relying on pathinfo could lead to"
+                                        " wrong results");
+                }
+
+                local->nallocs = local->wind_count = priv->child_count;
+                (void) strncpy (local->xsel, name, strlen (name));
+
+                /**
+                 * for xattrs that need info from all childs, fill ->xsel
+                 * as above and call the filler function in cbk based on
+                 * it
+                 */
+                for (i = 0, trav = this->children; i < priv->child_count; i++,
+                     trav = trav->next) {
+                        STACK_WIND_COOKIE (frame, stripe_vgetxattr_cbk,
+                                           (void *) (long) i, trav->xlator,
+                                           trav->xlator->fops->getxattr,
+                                           loc, name, xdata);
+                }
+
+                return 0;
+        }
+
+        if (name &&(*priv->vol_uuid)) {
+                if ((match_uuid_local (name, priv->vol_uuid) == 0)
+                    && (GF_CLIENT_PID_GSYNCD == frame->root->pid)) {
+
+                        if (!IA_FILE_OR_DIR (loc->inode->ia_type))
+                                local->marker.call_count = 1;
+                        else
+                                local->marker.call_count = priv->child_count;
+
+                        sub_volumes = alloca (local->marker.call_count *
+                                              sizeof (xlator_t *));
+
+                        for (i = 0, trav = this->children;
+                             i < local->marker.call_count;
+                             i++, trav = trav->next) {
+                                *(sub_volumes + i) = trav->xlator;
+
+                        }
+
+                        if (cluster_getmarkerattr (frame, this, loc, name,
+                                                   local,
+                                                   stripe_getxattr_unwind,
+                                                   sub_volumes,
+                                                   local->marker.call_count,
+                                                   MARKER_XTIME_TYPE,
+                                                   marker_xtime_default_gauge,
+                                                   priv->vol_uuid)) {
+                                op_errno = EINVAL;
+                                goto err;
+                        }
+
+                        return 0;
+                }
+        }
+
+
+        STACK_WIND (frame, stripe_internal_getxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+
+        return 0;
+
+err:
+        STRIPE_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
+        return 0;
+}
+
+inline gf_boolean_t
+stripe_is_special_xattr (const char *name)
+{
+        gf_boolean_t    is_spl = _gf_false;
+
+        if (!name) {
+                goto out;
+        }
+
+        if (!strncmp (name, GF_XATTR_LOCKINFO_KEY,
+                      strlen (GF_XATTR_LOCKINFO_KEY))
+            || XATTR_IS_PATHINFO (name))
+                is_spl = _gf_true;
+out:
+        return is_spl;
+}
+
+int32_t
+stripe_fgetxattr_from_everyone (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                                const char *name, dict_t *xdata)
+{
+        stripe_local_t   *local = NULL;
+        stripe_private_t *priv  = NULL;
+        int32_t           ret   = -1, op_errno = 0;
+        int               i     = 0;
+        xlator_list_t    *trav  = NULL;
+
+        priv = this->private;
+
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        local->op_ret = -1;
+        frame->local = local;
+
+        strncpy (local->xsel, name, strlen (name));
+        local->nallocs = local->wind_count = priv->child_count;
+
+        for (i = 0, trav = this->children; i < priv->child_count; i++,
+                     trav = trav->next) {
+                STACK_WIND_COOKIE (frame, stripe_vgetxattr_cbk,
+                                   (void *) (long) i, trav->xlator,
+                                   trav->xlator->fops->fgetxattr,
+                                   fd, name, xdata);
+        }
+
+        return 0;
+
+err:
+        STACK_UNWIND_STRICT (fgetxattr, frame, -1, op_errno, NULL, NULL);
+        return ret;
+}
+
+int32_t
+stripe_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                  const char *name, dict_t *xdata)
+{
+        if (stripe_is_special_xattr (name)) {
+                stripe_fgetxattr_from_everyone (frame, this, fd, name, xdata);
+                goto out;
+        }
+
+        STACK_WIND (frame, stripe_internal_getxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
+
+out:
+        return 0;
+}
+
+
+
+int32_t
+stripe_priv_dump (xlator_t *this)
+{
+        char                    key[GF_DUMP_MAX_BUF_LEN];
+        int                     i = 0;
+        stripe_private_t       *priv = NULL;
+        int                     ret = -1;
+        struct stripe_options  *options = NULL;
+
+        GF_VALIDATE_OR_GOTO ("stripe", this, out);
+
+        priv = this->private;
+        if (!priv)
+                goto out;
+
+        ret = TRY_LOCK (&priv->lock);
+        if (ret != 0)
+                goto out;
+
+        gf_proc_dump_add_section("xlator.cluster.stripe.%s.priv", this->name);
+        gf_proc_dump_write("child_count","%d", priv->child_count);
+
+        for (i = 0; i < priv->child_count; i++) {
+                sprintf (key, "subvolumes[%d]", i);
+                gf_proc_dump_write (key, "%s.%s", priv->xl_array[i]->type,
+                                    priv->xl_array[i]->name);
+        }
+
+        options = priv->pattern;
+        while (options != NULL) {
+                gf_proc_dump_write ("path_pattern", "%s", priv->pattern->path_pattern);
+                gf_proc_dump_write ("options_block_size", "%ul", options->block_size);
+
+                options = options->next;
+        }
+
+        gf_proc_dump_write ("block_size", "%ul", priv->block_size);
+        gf_proc_dump_write ("nodes-down", "%d", priv->nodes_down);
+        gf_proc_dump_write ("first-child_down", "%d", priv->first_child_down);
+        gf_proc_dump_write ("xattr_supported", "%d", priv->xattr_supported);
+
+        UNLOCK (&priv->lock);
+
+out:
+        return ret;
+}
 
 struct xlator_fops fops = {
-        .stat        = stripe_stat,
-        .unlink      = stripe_unlink,
-        .rename      = stripe_rename,
-        .link        = stripe_link,
-        .truncate    = stripe_truncate,
-        .create      = stripe_create,
-        .open        = stripe_open,
-        .readv       = stripe_readv,
-        .writev      = stripe_writev,
-        .statfs      = stripe_statfs,
-        .flush       = stripe_flush,
-        .fsync       = stripe_fsync,
-        .ftruncate   = stripe_ftruncate,
-        .fstat       = stripe_fstat,
-        .mkdir       = stripe_mkdir,
-        .rmdir       = stripe_rmdir,
-        .lk          = stripe_lk,
-        .opendir     = stripe_opendir,
-        .fsyncdir    = stripe_fsyncdir,
-        .setattr     = stripe_setattr,
-        .fsetattr    = stripe_fsetattr,
-        .lookup      = stripe_lookup,
-        .mknod       = stripe_mknod,
+        .stat           = stripe_stat,
+        .unlink         = stripe_unlink,
+        .rename         = stripe_rename,
+        .link           = stripe_link,
+        .truncate       = stripe_truncate,
+        .create         = stripe_create,
+        .open           = stripe_open,
+        .readv          = stripe_readv,
+        .writev         = stripe_writev,
+        .statfs         = stripe_statfs,
+        .flush          = stripe_flush,
+        .fsync          = stripe_fsync,
+        .ftruncate      = stripe_ftruncate,
+        .fstat          = stripe_fstat,
+        .mkdir          = stripe_mkdir,
+        .rmdir          = stripe_rmdir,
+        .lk             = stripe_lk,
+        .opendir        = stripe_opendir,
+        .fsyncdir       = stripe_fsyncdir,
+        .setattr        = stripe_setattr,
+        .fsetattr       = stripe_fsetattr,
+        .lookup         = stripe_lookup,
+        .mknod          = stripe_mknod,
+        .setxattr       = stripe_setxattr,
+        .fsetxattr      = stripe_fsetxattr,
+        .getxattr       = stripe_getxattr,
+        .fgetxattr      = stripe_fgetxattr,
+        .removexattr    = stripe_removexattr,
+        .fremovexattr   = stripe_fremovexattr,
+        .readdirp       = stripe_readdirp,
+	.fallocate	= stripe_fallocate,
+	.discard	= stripe_discard,
+        .zerofill       = stripe_zerofill,
 };
 
 struct xlator_cbks cbks = {
         .release = stripe_release,
+        .forget  = stripe_forget,
 };
 
+struct xlator_dumpops dumpops = {
+        .priv = stripe_priv_dump,
+};
 
 struct volume_options options[] = {
         { .key  = {"block-size"},
-          .type = GF_OPTION_TYPE_ANY
+          .type = GF_OPTION_TYPE_SIZE_LIST,
+          .default_value = "128KB",
+          .min = STRIPE_MIN_BLOCK_SIZE,
+          .description = "Size of the stripe unit that would be read "
+                         "from or written to the striped servers."
         },
         { .key  = {"use-xattr"},
-          .type = GF_OPTION_TYPE_BOOL
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "true"
         },
+	{ .key = {"coalesce"},
+	  .type = GF_OPTION_TYPE_BOOL,
+	  .default_value = "true",
+	  .description = "Enable/Disable coalesce mode to flatten striped "
+			 "files as stored on the server (i.e., eliminate holes "
+			 "caused by the traditional format)."
+	},
         { .key  = {NULL} },
 };
diff --git a/xlators/cluster/stripe/src/stripe.h b/xlators/cluster/stripe/src/stripe.h
index d94d82d6d..5673d18f3 100644
--- a/xlators/cluster/stripe/src/stripe.h
+++ b/xlators/cluster/stripe/src/stripe.h
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2010 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -33,9 +24,59 @@
 #include "compat.h"
 #include "compat-errno.h"
 #include "stripe-mem-types.h"
+#include "libxlator.h"
 #include <fnmatch.h>
 #include <signal.h>
 
+#define STRIPE_PATHINFO_HEADER "STRIPE:"
+#define STRIPE_MIN_BLOCK_SIZE  (16*GF_UNIT_KB)
+
+#define STRIPE_STACK_UNWIND(fop, frame, params ...) do {           \
+                stripe_local_t *__local = NULL;                    \
+                if (frame) {                                       \
+                        __local = frame->local;                    \
+                        frame->local = NULL;                       \
+                }                                                  \
+                STACK_UNWIND_STRICT (fop, frame, params);          \
+                if (__local) {                                     \
+                        stripe_local_wipe(__local);                \
+                        mem_put (__local);       \
+                }                                                  \
+        } while (0)
+
+#define STRIPE_STACK_DESTROY(frame) do {                        \
+                stripe_local_t *__local = NULL;                 \
+                __local = frame->local;                         \
+                frame->local = NULL;                            \
+                STACK_DESTROY (frame->root);                    \
+                if (__local) {                                  \
+                        stripe_local_wipe (__local);            \
+                        mem_put (__local);    \
+                }                                               \
+        } while (0)
+
+#define STRIPE_VALIDATE_FCTX(fctx, label) do {                  \
+        int     idx = 0;                                        \
+        if (!fctx) {                                            \
+                op_errno = EINVAL;                              \
+                goto label;                                     \
+        }                                                       \
+        for (idx = 0; idx < fctx->stripe_count; idx++) {        \
+                if (!fctx->xl_array[idx]) {                     \
+                        gf_log (this->name, GF_LOG_ERROR,       \
+                                "fctx->xl_array[%d] is NULL",   \
+                                idx);                           \
+                        op_errno = ESTALE;                      \
+                        goto label;                             \
+                }                                               \
+        }                                                       \
+       } while (0)
+
+typedef struct stripe_xattr_sort {
+        int   pos;
+        int   xattr_len;
+        char *xattr_value;
+} stripe_xattr_sort_t;
 
 /**
  * struct stripe_options : This keeps the pattern and the block-size
@@ -57,15 +98,17 @@ struct stripe_private {
         gf_lock_t               lock;
         uint8_t                 nodes_down;
         int8_t                  first_child_down;
+        int                    *last_event;
         int8_t                  child_count;
-        int8_t                 *state; /* Current state of child node */
         gf_boolean_t            xattr_supported;  /* default yes */
+	gf_boolean_t		coalesce;
+        char                    vol_uuid[UUID_SIZE + 1];
 };
 
 /**
- * Used to keep info about the replies received from fops->readv calls
+ * Used to keep info about the replies received from readv/writev calls
  */
-struct readv_replies {
+struct stripe_replies {
         struct iovec *vector;
         int32_t       count;    //count of vector
         int32_t       op_ret;   //op_ret of readv
@@ -77,6 +120,7 @@ struct readv_replies {
 typedef struct _stripe_fd_ctx {
         off_t      stripe_size;
         int        stripe_count;
+	int	   stripe_coalesce;
         int        static_array;
         xlator_t **xl_array;
 } stripe_fd_ctx_t;
@@ -112,9 +156,9 @@ struct stripe_local {
         blkcnt_t             preparent_blocks;
         blkcnt_t             postparent_blocks;
 
-        struct readv_replies *replies;
-        struct statvfs       statvfs_buf;
-        dir_entry_t         *entry;
+        struct stripe_replies *replies;
+        struct statvfs        statvfs_buf;
+        dir_entry_t          *entry;
 
         int8_t               revalidate;
         int8_t               failed;
@@ -136,8 +180,17 @@ struct stripe_local {
         loc_t                loc;
         loc_t                loc2;
 
+        mode_t               mode;
+        dev_t                rdev;
         /* For File I/O fops */
-        dict_t              *dict;
+        dict_t              *xdata;
+
+        stripe_xattr_sort_t *xattr_list;
+        int32_t              xattr_total_len;
+        int32_t              nallocs;
+        char xsel[256];
+
+        struct marker_str    marker;
 
         /* General usage */
         off_t                offset;
@@ -151,9 +204,85 @@ struct stripe_local {
         fd_t                *fd;
         void                *value;
         struct iobref       *iobref;
+        gf_dirent_t          entries;
+        gf_dirent_t         *dirent;
+        dict_t              *xattr;
+        uuid_t               ia_gfid;
+
+        int                  xflag;
+        mode_t               umask;
 };
 
 typedef struct stripe_local   stripe_local_t;
 typedef struct stripe_private stripe_private_t;
 
+/*
+ * Determine the stripe index of a particular frame based on the translator.
+ */
+static inline int32_t stripe_get_frame_index(stripe_fd_ctx_t *fctx,
+					     call_frame_t *prev)
+{
+	int32_t i, idx = -1;
+
+	for (i = 0; i < fctx->stripe_count; i++) {
+		if (fctx->xl_array[i] == prev->this) {
+			idx = i;
+			break;
+		}
+	}
+
+	return idx;
+}
+
+static inline void stripe_copy_xl_array(xlator_t **dst, xlator_t **src,
+					int count)
+{
+	int i;
+
+	for (i = 0; i < count; i++)
+		dst[i] = src[i];
+}
+
+void stripe_local_wipe (stripe_local_t *local);
+int32_t stripe_ctx_handle (xlator_t *this, call_frame_t *prev,
+                           stripe_local_t *local, dict_t *dict);
+void stripe_aggregate_xattr (dict_t *dst, dict_t *src);
+int32_t stripe_xattr_request_build (xlator_t *this, dict_t *dict,
+                                    uint64_t stripe_size, uint32_t stripe_count,
+                                    uint32_t stripe_index,
+				    uint32_t stripe_coalesce);
+int32_t stripe_get_matching_bs (const char *path, stripe_private_t *priv);
+int set_stripe_block_size (xlator_t *this, stripe_private_t *priv, char *data);
+int32_t stripe_iatt_merge (struct iatt *from, struct iatt *to);
+int32_t stripe_fill_pathinfo_xattr (xlator_t *this, stripe_local_t *local,
+                                    char **xattr_serz);
+int32_t stripe_free_xattr_str (stripe_local_t *local);
+int32_t stripe_xattr_aggregate (char *buffer, stripe_local_t *local,
+                                int32_t *total);
+off_t coalesced_offset(off_t offset, uint64_t stripe_size, int stripe_count);
+off_t uncoalesced_size(off_t size, uint64_t stripe_size, int stripe_count,
+			int stripe_index);
+int32_t
+stripe_fill_lockinfo_xattr (xlator_t *this, stripe_local_t *local,
+                            void **xattr_serz);
+
+/*
+ * Adjust the size attribute for files if coalesce is enabled.
+ */
+static inline void correct_file_size(struct iatt *buf, stripe_fd_ctx_t *fctx,
+	call_frame_t *prev)
+{
+	int index;
+
+	if (!IA_ISREG(buf->ia_type))
+		return;
+
+	if (!fctx || !fctx->stripe_coalesce)
+		return;
+
+	index = stripe_get_frame_index(fctx, prev);
+	buf->ia_size = uncoalesced_size(buf->ia_size, fctx->stripe_size,
+		fctx->stripe_count, index);
+}
+
 #endif /* _STRIPE_H_ */
diff --git a/xlators/cluster/unify/Makefile.am b/xlators/cluster/unify/Makefile.am
deleted file mode 100644
index d471a3f92..000000000
--- a/xlators/cluster/unify/Makefile.am
+++ /dev/null
@@ -1,3 +0,0 @@
-SUBDIRS = src
-
-CLEANFILES = 
diff --git a/xlators/cluster/unify/src/Makefile.am b/xlators/cluster/unify/src/Makefile.am
deleted file mode 100644
index 2a1fe8372..000000000
--- a/xlators/cluster/unify/src/Makefile.am
+++ /dev/null
@@ -1,16 +0,0 @@
-
-xlator_LTLIBRARIES = unify.la
-xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/legacy/cluster
-
-unify_la_LDFLAGS = -module -avoidversion
-
-unify_la_SOURCES = unify.c unify-self-heal.c
-unify_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-
-noinst_HEADERS = unify.h
-
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
-
-CLEANFILES = 
-
diff --git a/xlators/cluster/unify/src/unify-mem-types.h b/xlators/cluster/unify/src/unify-mem-types.h
deleted file mode 100644
index dcf964779..000000000
--- a/xlators/cluster/unify/src/unify-mem-types.h
+++ /dev/null
@@ -1,41 +0,0 @@
-
-/*
-   Copyright (c) 2008-2010 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU Affero General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   Affero General Public License for more details.
-
-   You should have received a copy of the GNU Affero General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
-*/
-
-
-#ifndef __UNIFY_MEM_TYPES_H__
-#define __UNIFY_MEM_TYPES_H__
-
-#include "mem-types.h"
-
-enum gf_unify_mem_types_ {
-        gf_unify_mt_char = gf_common_mt_end + 1,
-        gf_unify_mt_int16_t,
-        gf_unify_mt_xlator_t,
-        gf_unify_mt_unify_private_t,
-        gf_unify_mt_xlator_list_t,
-        gf_unify_mt_dir_entry_t,
-        gf_unify_mt_off_t,
-        gf_unify_mt_int,
-        gf_unify_mt_unify_self_heal_struct,
-        gf_unify_mt_unify_local_t,
-        gf_unify_mt_end
-};
-#endif
-
diff --git a/xlators/cluster/unify/src/unify-self-heal.c b/xlators/cluster/unify/src/unify-self-heal.c
deleted file mode 100644
index 725523a2e..000000000
--- a/xlators/cluster/unify/src/unify-self-heal.c
+++ /dev/null
@@ -1,1239 +0,0 @@
-/*
-  Copyright (c) 2007-2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-/**
- * unify-self-heal.c : 
- *   This file implements few functions which enables 'unify' translator 
- *  to be consistent in its behaviour when 
- *     > a node fails, 
- *     > a node gets added, 
- *     > a failed node comes back
- *     > a new namespace server is added (ie, an fresh namespace server).
- * 
- *  This functionality of 'unify' will enable glusterfs to support storage
- *  system failure, and maintain consistancy. This works both ways, ie, when
- *  an entry (either file or directory) is found on namespace server, and not
- *  on storage nodes, its created in storage nodes and vica-versa.
- * 
- *  The two fops, where it can be implemented are 'getdents ()' and 'lookup ()'
- *
- */
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include "glusterfs.h"
-#include "unify.h"
-#include "dict.h"
-#include "xlator.h"
-#include "hashfn.h"
-#include "logging.h"
-#include "stack.h"
-#include "common-utils.h"
-
-int32_t
-unify_sh_getdents_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno,
-		       dir_entry_t *entry,
-		       int32_t count);
-
-int32_t
-unify_sh_ns_getdents_cbk (call_frame_t *frame,
-			  void *cookie,
-			  xlator_t *this,
-			  int32_t op_ret,
-			  int32_t op_errno,
-			  dir_entry_t *entry,
-			  int32_t count);
-
-int32_t
-unify_bgsh_getdents_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno,
-			 dir_entry_t *entry,
-			 int32_t count);
-
-int32_t
-unify_bgsh_ns_getdents_cbk (call_frame_t *frame,
-			    void *cookie,
-			    xlator_t *this,
-			    int32_t op_ret,
-			    int32_t op_errno,
-			    dir_entry_t *entry,
-			    int32_t count);
-
-/**
- * unify_local_wipe - free all the extra allocation of local->* here.
- */
-static void 
-unify_local_wipe (unify_local_t *local)
-{
-	/* Free the strdup'd variables in the local structure */
-	if (local->name) {
-		GF_FREE (local->name);
-	}
-
-	if (local->sh_struct) {
-		if (local->sh_struct->offset_list)
-			GF_FREE (local->sh_struct->offset_list);
-
-		if (local->sh_struct->entry_list)
-			GF_FREE (local->sh_struct->entry_list);
-
-		if (local->sh_struct->count_list)
-			GF_FREE (local->sh_struct->count_list);
-
-		GF_FREE (local->sh_struct);
-	}
-
-	loc_wipe (&local->loc1);
-	loc_wipe (&local->loc2);
-}
-
-int32_t 
-unify_sh_setdents_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno)
-{
-	int32_t callcnt = -1;
-	unify_local_t *local = frame->local;
-	inode_t *inode = NULL;
-	dict_t *tmp_dict = NULL;
-	dir_entry_t *prev, *entry, *trav;
-
-	LOCK (&frame->lock);
-	{
-		/* if local->call_count == 0, that means, setdents on 
-		 * storagenodes is still pending.
-		 */
-		if (local->call_count)
-			callcnt = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
-
-	if (callcnt == 0) {
-		if (local->sh_struct->entry_list[0]) {
-			prev = entry = local->sh_struct->entry_list[0];
-			if (!entry)
-				return 0;
-			trav = entry->next;
-			while (trav) {
-				prev->next = trav->next;
-				GF_FREE (trav->name);
-				if (IA_ISLNK (trav->buf.ia_type))
-					GF_FREE (trav->link);
-				GF_FREE (trav);
-				trav = prev->next;
-			}
-			GF_FREE (entry);
-		}
-
-		if (!local->flags) {
-			if (local->sh_struct->count_list[0] >= 
-			    UNIFY_SELF_HEAL_GETDENTS_COUNT) {
-				/* count == size, that means, there are more entries
-				   to read from */
-				//local->call_count = 0;
-				local->sh_struct->offset_list[0] += 
-					UNIFY_SELF_HEAL_GETDENTS_COUNT;
-				STACK_WIND (frame,
-					    unify_sh_ns_getdents_cbk,
-					    NS(this),
-					    NS(this)->fops->getdents,
-					    local->fd,
-					    UNIFY_SELF_HEAL_GETDENTS_COUNT,
-					    local->sh_struct->offset_list[0],
-					    GF_GET_DIR_ONLY);
-			}		
-		} else {
-			inode = local->loc1.inode;
-			fd_unref (local->fd);
-			tmp_dict = local->dict;
-
-			unify_local_wipe (local);
-			
-			STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-				      inode, &local->stbuf, local->dict,
-                                      &local->oldpostparent);
-			if (tmp_dict)
-				dict_unref (tmp_dict);
-		}
-	}
-  
-	return 0;
-}
-
-
-int32_t
-unify_sh_ns_getdents_cbk (call_frame_t *frame,
-			  void *cookie,
-			  xlator_t *this,
-			  int32_t op_ret,
-			  int32_t op_errno,
-			  dir_entry_t *entry,
-			  int32_t count)
-{
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	long index = 0;
-	unsigned long final = 0;
-	dir_entry_t *tmp = GF_CALLOC (1, sizeof (dir_entry_t),
-                                      gf_unify_mt_dir_entry_t);
-	
-	local->sh_struct->entry_list[0] = tmp;
-	local->sh_struct->count_list[0] = count;
-	if (entry) {
-		tmp->next = entry->next;
-		entry->next = NULL;
-	}
-
-	if ((count < UNIFY_SELF_HEAL_GETDENTS_COUNT) || !entry) {
-		final = 1;
-	}
-
-	LOCK (&frame->lock);
-	{
-		/* local->call_count will be '0' till now. make it 1 so, it 
-		   can be UNWIND'ed for the last call. */
-		local->call_count = priv->child_count;
-		if (final)
-			local->flags = 1;
-	}
-	UNLOCK (&frame->lock);
-
-	for (index = 0; index < priv->child_count; index++) 
-	{
-		STACK_WIND_COOKIE (frame,
-				   unify_sh_setdents_cbk, 
-				   (void *)index,
-				   priv->xl_array[index],
-				   priv->xl_array[index]->fops->setdents,
-				   local->fd, GF_SET_DIR_ONLY,
-				   local->sh_struct->entry_list[0], count);
-	}
-
-	return 0;
-}
-
-int32_t 
-unify_sh_ns_setdents_cbk (call_frame_t *frame,
-			  void *cookie,
-			  xlator_t *this,
-			  int32_t op_ret,
-			  int32_t op_errno)
-{
-	int32_t callcnt = -1;
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	long index = (long)cookie;
-	dir_entry_t *prev, *entry, *trav;
-
-	LOCK (&frame->lock);
-	{
-		if (local->sh_struct->entry_list[index]) {
-			prev = entry = local->sh_struct->entry_list[index];
-			trav = entry->next;
-			while (trav) {
-				prev->next = trav->next;
-				GF_FREE (trav->name);
-				if (IA_ISLNK (trav->buf.ia_type))
-					GF_FREE (trav->link);
-				GF_FREE (trav);
-				trav = prev->next;
-			}
-			GF_FREE (entry);
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	if (local->sh_struct->count_list[index] < 
-	    UNIFY_SELF_HEAL_GETDENTS_COUNT) {
-		LOCK (&frame->lock);
-		{
-			callcnt = --local->call_count;
-		}
-		UNLOCK (&frame->lock);
-	} else {
-		/* count == size, that means, there are more entries 
-		   to read from */
-		local->sh_struct->offset_list[index] += 
-			UNIFY_SELF_HEAL_GETDENTS_COUNT;
-		STACK_WIND_COOKIE (frame,
-				   unify_sh_getdents_cbk,
-				   cookie,
-				   priv->xl_array[index],
-				   priv->xl_array[index]->fops->getdents,
-				   local->fd,
-				   UNIFY_SELF_HEAL_GETDENTS_COUNT,
-				   local->sh_struct->offset_list[index],
-				   GF_GET_ALL);
-    
-		gf_log (this->name, GF_LOG_DEBUG, 
-			"readdir on (%s) with offset %"PRId64"", 
-			priv->xl_array[index]->name, 
-			local->sh_struct->offset_list[index]);
-	}
-
-	if (!callcnt) {
-		/* All storage nodes have done unified setdents on NS node.
-		 * Now, do getdents from NS and do setdents on storage nodes.
-		 */
-    
-		/* sh_struct->offset_list is no longer required for
-		   storage nodes now */
-		local->sh_struct->offset_list[0] = 0; /* reset */
-
-		STACK_WIND (frame,
-			    unify_sh_ns_getdents_cbk,
-			    NS(this),
-			    NS(this)->fops->getdents,
-			    local->fd,
-			    UNIFY_SELF_HEAL_GETDENTS_COUNT,
-			    0, /* In this call, do send '0' as offset */
-			    GF_GET_DIR_ONLY);
-	}
-
-	return 0;
-}
-
-
-/**
- * unify_sh_getdents_cbk -
- */
-int32_t
-unify_sh_getdents_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno,
-		       dir_entry_t *entry,
-		       int32_t count)
-{
-	int32_t callcnt = -1;
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	long index = (long)cookie;
-	dir_entry_t *tmp = NULL; 
-
-	if (op_ret >= 0 && count > 0) {
-		/* There is some dentry found, just send the dentry to NS */
-		tmp = GF_CALLOC (1, sizeof (dir_entry_t),
-                                 gf_unify_mt_dir_entry_t);
-		local->sh_struct->entry_list[index] = tmp;
-		local->sh_struct->count_list[index] = count;
-		if (entry) {
-			tmp->next = entry->next;
-			entry->next = NULL;
-		}
-		STACK_WIND_COOKIE (frame,
-				   unify_sh_ns_setdents_cbk,
-				   cookie,
-				   NS(this),
-				   NS(this)->fops->setdents,
-				   local->fd,
-				   GF_SET_IF_NOT_PRESENT,
-				   local->sh_struct->entry_list[index],
-				   count);
-		return 0;
-	}
-  
-	if (count < UNIFY_SELF_HEAL_GETDENTS_COUNT) {
-		LOCK (&frame->lock);
-		{
-			callcnt = --local->call_count;
-		}
-		UNLOCK (&frame->lock);
-	} else {
-		/* count == size, that means, there are more entries 
-		   to read from */
-		local->sh_struct->offset_list[index] += 
-			UNIFY_SELF_HEAL_GETDENTS_COUNT;
-		STACK_WIND_COOKIE (frame,
-				   unify_sh_getdents_cbk,
-				   cookie,
-				   priv->xl_array[index],
-				   priv->xl_array[index]->fops->getdents,
-				   local->fd,
-				   UNIFY_SELF_HEAL_GETDENTS_COUNT,
-				   local->sh_struct->offset_list[index],
-				   GF_GET_ALL);
-    
-		gf_log (this->name, GF_LOG_DEBUG, 
-			"readdir on (%s) with offset %"PRId64"", 
-			priv->xl_array[index]->name, 
-			local->sh_struct->offset_list[index]);
-	}
-
-	if (!callcnt) {
-		/* All storage nodes have done unified setdents on NS node.
-		 * Now, do getdents from NS and do setdents on storage nodes.
-		 */
-    
-		/* sh_struct->offset_list is no longer required for
-		   storage nodes now */
-		local->sh_struct->offset_list[0] = 0; /* reset */
-
-		STACK_WIND (frame,
-			    unify_sh_ns_getdents_cbk,
-			    NS(this),
-			    NS(this)->fops->getdents,
-			    local->fd,
-			    UNIFY_SELF_HEAL_GETDENTS_COUNT,
-			    0, /* In this call, do send '0' as offset */
-			    GF_GET_DIR_ONLY);
-	}
-
-	return 0;
-}
-
-/**
- * unify_sh_opendir_cbk -
- *
- * @cookie: 
- */
-int32_t 
-unify_sh_opendir_cbk (call_frame_t *frame,
-		      void *cookie,
-		      xlator_t *this,
-		      int32_t op_ret,
-		      int32_t op_errno,
-		      fd_t *fd)
-{
-	int32_t callcnt = 0;
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	int16_t index = 0;
-	inode_t *inode = NULL;
-	dict_t *tmp_dict = NULL;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-    
-		if (op_ret >= 0) {
-			local->op_ret = op_ret;
-		} else {
-			gf_log (this->name, GF_LOG_WARNING, "failed");
-			local->failed = 1;
-		}
-	}
-	UNLOCK (&frame->lock);
-  
-	if (!callcnt) {
-		local->call_count = priv->child_count + 1;
-    
-		if (!local->failed) {
-			/* send getdents() namespace after finishing
-			   storage nodes */
-			local->call_count--; 
-      
-			fd_bind (fd);
-
-			if (local->call_count) {
-				/* Used as the offset index. This list keeps
-				 * track of offset sent to each node during
-				 * STACK_WIND.
-				 */
-				local->sh_struct->offset_list = 
-					GF_CALLOC (priv->child_count, 
-					           sizeof (off_t),
-                                                   gf_unify_mt_off_t);
-				ERR_ABORT (local->sh_struct->offset_list);
-	
-				local->sh_struct->entry_list = 
-					GF_CALLOC (priv->child_count, 
-						   sizeof (dir_entry_t *),
-                                                   gf_unify_mt_dir_entry_t);
-				ERR_ABORT (local->sh_struct->entry_list);
-
-				local->sh_struct->count_list = 
-					GF_CALLOC (priv->child_count, 
-						   sizeof (int),
-                                                   gf_unify_mt_int);
-				ERR_ABORT (local->sh_struct->count_list);
-
-				/* Send getdents on all the fds */
-				for (index = 0; 
-				     index < priv->child_count; index++) {
-					STACK_WIND_COOKIE (frame,
-							   unify_sh_getdents_cbk,
-							   (void *)(long)index,
-							   priv->xl_array[index],
-							   priv->xl_array[index]->fops->getdents,
-							   local->fd,
-							   UNIFY_SELF_HEAL_GETDENTS_COUNT,
-							   0, /* In this call, do send '0' as offset */
-							   GF_GET_ALL);
-				}
-
-				/* did stack wind, so no need to unwind here */
-				return 0;
-			} /* (local->call_count) */
-		} /* (!local->failed) */
-
-		/* Opendir failed on one node. */ 
-		inode = local->loc1.inode;
-		fd_unref (local->fd);
-		tmp_dict = local->dict;
-
-		unify_local_wipe (local);
-		/* Only 'self-heal' failed, lookup() was successful. */
-		local->op_ret = 0;
-
-		/* This is lookup_cbk ()'s UNWIND. */
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, inode,
-			      &local->stbuf, local->dict, &local->oldpostparent);
-		if (tmp_dict)
-			dict_unref (tmp_dict);
-	}
-
-	return 0;
-}
-
-/**
- * gf_sh_checksum_cbk - 
- * 
- * @frame: frame used in lookup. get a copy of it, and use that copy.
- * @this: pointer to unify xlator.
- * @inode: pointer to inode, for which the consistency check is required.
- *
- */
-int32_t 
-unify_sh_checksum_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno,
-		       uint8_t *file_checksum,
-		       uint8_t *dir_checksum)
-{
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	int16_t index = 0;
-	int32_t callcnt = 0;
-	inode_t *inode = NULL;
-	dict_t *tmp_dict = NULL;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-		if (op_ret >= 0) {
-			if (NS(this) == (xlator_t *)cookie) {
-				memcpy (local->sh_struct->ns_file_checksum, 
-					file_checksum, NAME_MAX);
-				memcpy (local->sh_struct->ns_dir_checksum, 
-					dir_checksum, NAME_MAX);
-			} else {
-				if (local->entry_count == 0) {
-					/* Initialize the dir_checksum to be 
-					 * used for comparision with other
-					 * storage nodes. Should be done for
-					 * the first successful call *only*. 
-					 */
-                                        /* Using 'entry_count' as a flag */
-					local->entry_count = 1;
-					memcpy (local->sh_struct->dir_checksum,
-						dir_checksum, NAME_MAX);
-				}
-
-				/* Reply from the storage nodes */
-				for (index = 0; 
-				     index < NAME_MAX; index++) {
-					/* Files should be present in
-					   only one node */
-					local->sh_struct->file_checksum[index] ^= file_checksum[index];
-	  
-					/* directory structure should be
-					   same accross */
-					if (local->sh_struct->dir_checksum[index] != dir_checksum[index])
-						local->failed = 1;
-				}
-			}
-		} 
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		for (index = 0; index < NAME_MAX ; index++) {
-			if (local->sh_struct->file_checksum[index] != 
-			    local->sh_struct->ns_file_checksum[index]) {
-				local->failed = 1;
-				break;
-			}
-			if (local->sh_struct->dir_checksum[index] != 
-			    local->sh_struct->ns_dir_checksum[index]) {
-				local->failed = 1;
-				break;
-			}
-		}
-	
-		if (local->failed) {
-			/* Log it, it should be a rare event */
-			gf_log (this->name, GF_LOG_WARNING, 
-				"Self-heal triggered on directory %s", 
-				local->loc1.path);
-
-			/* Any self heal will be done at directory level */
-			local->call_count = 0;
-			local->op_ret = -1;
-			local->failed = 0;
-      
-			local->fd = fd_create (local->loc1.inode, 
-					       frame->root->pid);
-
-			local->call_count = priv->child_count + 1;
-	
-			for (index = 0; 
-			     index < (priv->child_count + 1); index++) {
-				STACK_WIND_COOKIE (frame,
-						   unify_sh_opendir_cbk,
-						   priv->xl_array[index]->name,
-						   priv->xl_array[index],
-						   priv->xl_array[index]->fops->opendir,
-						   &local->loc1,
-						   local->fd);
-			}
-			/* opendir can be done on the directory */
-			return 0;
-		}
-
-		/* no mismatch */
-		inode = local->loc1.inode;
-		tmp_dict = local->dict;
-
-		unify_local_wipe (local);
-
-		/* This is lookup_cbk ()'s UNWIND. */
-		STACK_UNWIND (frame,
-			      local->op_ret,
-			      local->op_errno,
-			      inode,
-			      &local->stbuf,
-			      local->dict, &local->oldpostparent);
-		if (tmp_dict)
-			dict_unref (tmp_dict);
-	}
-
-	return 0;
-}
-
-/* Foreground self-heal part over */
-
-/* Background self-heal part */
-
-int32_t 
-unify_bgsh_setdents_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno)
-{
-	int32_t callcnt = -1;
-	unify_local_t *local = frame->local;
-	dir_entry_t *prev, *entry, *trav;
-
-	LOCK (&frame->lock);
-	{
-		/* if local->call_count == 0, that means, setdents 
-		   on storagenodes is still pending. */
-		if (local->call_count)
-			callcnt = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
-
-
-	if (callcnt == 0) {
-		if (local->sh_struct->entry_list[0]) {
-			prev = entry = local->sh_struct->entry_list[0];
-			trav = entry->next;
-			while (trav) {
-				prev->next = trav->next;
-				GF_FREE (trav->name);
-				if (IA_ISLNK (trav->buf.ia_type))
-					GF_FREE (trav->link);
-				GF_FREE (trav);
-				trav = prev->next;
-			}
-			GF_FREE (entry);
-		}
-
-		if (!local->flags) {
-			if (local->sh_struct->count_list[0] >= 
-			    UNIFY_SELF_HEAL_GETDENTS_COUNT) {
-				/* count == size, that means, there are more
-				   entries to read from */
-				//local->call_count = 0;
-				local->sh_struct->offset_list[0] += 
-					UNIFY_SELF_HEAL_GETDENTS_COUNT;
-				STACK_WIND (frame,
-					    unify_bgsh_ns_getdents_cbk,
-					    NS(this),
-					    NS(this)->fops->getdents,
-					    local->fd,
-					    UNIFY_SELF_HEAL_GETDENTS_COUNT,
-					    local->sh_struct->offset_list[0],
-					    GF_GET_DIR_ONLY);
-			}		
-		} else {
-			fd_unref (local->fd);
-			unify_local_wipe (local);
-			STACK_DESTROY (frame->root);
-		}
-	}
-
-	return 0;
-}
-
-
-int32_t
-unify_bgsh_ns_getdents_cbk (call_frame_t *frame,
-			    void *cookie,
-			    xlator_t *this,
-			    int32_t op_ret,
-			    int32_t op_errno,
-			    dir_entry_t *entry,
-			    int32_t count)
-{
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	long index = 0;
-	unsigned long final = 0;
-	dir_entry_t *tmp = GF_CALLOC (1, sizeof (dir_entry_t),
-                                      gf_unify_mt_dir_entry_t);
-
-	local->sh_struct->entry_list[0] = tmp;
-	local->sh_struct->count_list[0] = count;
-	if (entry) {
-		tmp->next = entry->next;
-		entry->next = NULL;
-	}
-
-	if ((count < UNIFY_SELF_HEAL_GETDENTS_COUNT) || !entry) {
-		final = 1;
-	}
-
-	LOCK (&frame->lock);
-	{
-		/* local->call_count will be '0' till now. make it 1 so, 
-		   it can be UNWIND'ed for the last call. */
-		local->call_count = priv->child_count;
-		if (final)
-			local->flags = 1;
-	}
-	UNLOCK (&frame->lock);
-
-	for (index = 0; index < priv->child_count; index++) 
-	{
-		STACK_WIND_COOKIE (frame,
-				   unify_bgsh_setdents_cbk, 
-				   (void *)index,
-				   priv->xl_array[index],
-				   priv->xl_array[index]->fops->setdents,
-				   local->fd, GF_SET_DIR_ONLY,
-				   local->sh_struct->entry_list[0], count);
-	}
-
-	return 0;
-}
-
-int32_t 
-unify_bgsh_ns_setdents_cbk (call_frame_t *frame,
-			    void *cookie,
-			    xlator_t *this,
-			    int32_t op_ret,
-			    int32_t op_errno)
-{
-	int32_t callcnt = -1;
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	long index = (long)cookie;
-	dir_entry_t *prev, *entry, *trav;
-
-	if (local->sh_struct->entry_list[index]) {
-		prev = entry = local->sh_struct->entry_list[index];
-		if (!entry)
-			return 0;
-		trav = entry->next;
-		while (trav) {
-			prev->next = trav->next;
-			GF_FREE (trav->name);
-			if (IA_ISLNK (trav->buf.ia_type))
-				GF_FREE (trav->link);
-			GF_FREE (trav);
-			trav = prev->next;
-		}
-		GF_FREE (entry);
-	}
-
-	if (local->sh_struct->count_list[index] < 
-	    UNIFY_SELF_HEAL_GETDENTS_COUNT) {
-		LOCK (&frame->lock);
-		{
-			callcnt = --local->call_count;
-		}
-		UNLOCK (&frame->lock);
-	} else {
-		/* count == size, that means, there are more entries 
-		   to read from */
-		local->sh_struct->offset_list[index] += 
-			UNIFY_SELF_HEAL_GETDENTS_COUNT;
-		STACK_WIND_COOKIE (frame,
-				   unify_bgsh_getdents_cbk,
-				   cookie,
-				   priv->xl_array[index],
-				   priv->xl_array[index]->fops->getdents,
-				   local->fd,
-				   UNIFY_SELF_HEAL_GETDENTS_COUNT,
-				   local->sh_struct->offset_list[index],
-				   GF_GET_ALL);
-    
-		gf_log (this->name, GF_LOG_DEBUG, 
-			"readdir on (%s) with offset %"PRId64"", 
-			priv->xl_array[index]->name, 
-			local->sh_struct->offset_list[index]);
-	}
-
-	if (!callcnt) {
-		/* All storage nodes have done unified setdents on NS node.
-		 * Now, do getdents from NS and do setdents on storage nodes.
-		 */
-    
-		/* sh_struct->offset_list is no longer required for
-		   storage nodes now */
-		local->sh_struct->offset_list[0] = 0; /* reset */
-
-		STACK_WIND (frame,
-			    unify_bgsh_ns_getdents_cbk,
-			    NS(this),
-			    NS(this)->fops->getdents,
-			    local->fd,
-			    UNIFY_SELF_HEAL_GETDENTS_COUNT,
-			    0, /* In this call, do send '0' as offset */
-			    GF_GET_DIR_ONLY);
-	}
-
-	return 0;
-}
-
-
-/**
- * unify_bgsh_getdents_cbk -
- */
-int32_t
-unify_bgsh_getdents_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno,
-			 dir_entry_t *entry,
-			 int32_t count)
-{
-	int32_t callcnt = -1;
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	long index = (long)cookie;
-	dir_entry_t *tmp = NULL; 
-
-	if (op_ret >= 0 && count > 0) {
-		/* There is some dentry found, just send the dentry to NS */
-		tmp = GF_CALLOC (1, sizeof (dir_entry_t),
-                                 gf_unify_mt_dir_entry_t);
-		local->sh_struct->entry_list[index] = tmp;
-		local->sh_struct->count_list[index] = count;
-		if (entry) {
-			tmp->next = entry->next;
-			entry->next = NULL;
-		}
-		STACK_WIND_COOKIE (frame,
-				   unify_bgsh_ns_setdents_cbk,
-				   cookie,
-				   NS(this),
-				   NS(this)->fops->setdents,
-				   local->fd,
-				   GF_SET_IF_NOT_PRESENT,
-				   local->sh_struct->entry_list[index],
-				   count);
-		return 0;
-	}
-  
-	if (count < UNIFY_SELF_HEAL_GETDENTS_COUNT) {
-		LOCK (&frame->lock);
-		{
-			callcnt = --local->call_count;
-		}
-		UNLOCK (&frame->lock);
-	} else {
-		/* count == size, that means, there are more entries to read from */
-		local->sh_struct->offset_list[index] += 
-			UNIFY_SELF_HEAL_GETDENTS_COUNT;
-
-		STACK_WIND_COOKIE (frame,
-				   unify_bgsh_getdents_cbk,
-				   cookie,
-				   priv->xl_array[index],
-				   priv->xl_array[index]->fops->getdents,
-				   local->fd,
-				   UNIFY_SELF_HEAL_GETDENTS_COUNT,
-				   local->sh_struct->offset_list[index],
-				   GF_GET_ALL);
-    
-		gf_log (this->name, GF_LOG_DEBUG, 
-			"readdir on (%s) with offset %"PRId64"", 
-			priv->xl_array[index]->name, 
-			local->sh_struct->offset_list[index]);
-	}
-
-	if (!callcnt) {
-		/* All storage nodes have done unified setdents on NS node.
-		 * Now, do getdents from NS and do setdents on storage nodes.
-		 */
-    
-		/* sh_struct->offset_list is no longer required for 
-		   storage nodes now */
-		local->sh_struct->offset_list[0] = 0; /* reset */
-
-		STACK_WIND (frame,
-			    unify_bgsh_ns_getdents_cbk,
-			    NS(this),
-			    NS(this)->fops->getdents,
-			    local->fd,
-			    UNIFY_SELF_HEAL_GETDENTS_COUNT,
-			    0, /* In this call, do send '0' as offset */
-			    GF_GET_DIR_ONLY);
-	}
-
-	return 0;
-}
-
-/**
- * unify_bgsh_opendir_cbk -
- *
- * @cookie: 
- */
-int32_t 
-unify_bgsh_opendir_cbk (call_frame_t *frame,
-			void *cookie,
-			xlator_t *this,
-			int32_t op_ret,
-			int32_t op_errno,
-			fd_t *fd)
-{
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	int32_t callcnt = 0;
-	int16_t index = 0;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-    
-		if (op_ret >= 0) {
-			local->op_ret = op_ret;
-		} else {
-			local->failed = 1;
-		}
-	}
-	UNLOCK (&frame->lock);
-  
-	if (!callcnt) {
-		local->call_count = priv->child_count + 1;
-    
-		if (!local->failed) {
-			/* send getdents() namespace after finishing 
-			   storage nodes */
-			local->call_count--; 
-			callcnt = local->call_count;
-      
-			fd_bind (fd);
-
-			if (local->call_count) {
-				/* Used as the offset index. This list keeps 
-				   track of offset sent to each node during 
-				   STACK_WIND. */
-				local->sh_struct->offset_list = 
-					GF_CALLOC (priv->child_count, 
-					           sizeof (off_t),
-                                                   gf_unify_mt_off_t);
-				ERR_ABORT (local->sh_struct->offset_list);
-	
-				local->sh_struct->entry_list = 
-					GF_CALLOC (priv->child_count, 
-						   sizeof (dir_entry_t *),
-                                                   gf_unify_mt_dir_entry_t);
-				ERR_ABORT (local->sh_struct->entry_list);
-
-				local->sh_struct->count_list = 
-					GF_CALLOC (priv->child_count, 
-						   sizeof (int),
-                                                   gf_unify_mt_int);
-				ERR_ABORT (local->sh_struct->count_list);
-
-				/* Send getdents on all the fds */
-				for (index = 0; 
-				     index < priv->child_count; index++) {
-					STACK_WIND_COOKIE (frame,
-							   unify_bgsh_getdents_cbk,
-							   (void *)(long)index,
-							   priv->xl_array[index],
-							   priv->xl_array[index]->fops->getdents,
-							   local->fd,
-							   UNIFY_SELF_HEAL_GETDENTS_COUNT,
-							   0, /* In this call, do send '0' as offset */
-							   GF_GET_ALL);
-				}
-				/* did a stack wind, so no need to unwind here */
-				return 0;
-			} /* (local->call_count) */
-		} /* (!local->failed) */
-
-		/* Opendir failed on one node. 	 */
-		fd_unref (local->fd);
-		
-		unify_local_wipe (local);
-		STACK_DESTROY (frame->root);
-	}
-
-	return 0;
-}
-
-/**
- * gf_bgsh_checksum_cbk - 
- * 
- * @frame: frame used in lookup. get a copy of it, and use that copy.
- * @this: pointer to unify xlator.
- * @inode: pointer to inode, for which the consistency check is required.
- *
- */
-int32_t 
-unify_bgsh_checksum_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno,
-			 uint8_t *file_checksum,
-			 uint8_t *dir_checksum)
-{
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	int16_t index = 0;
-	int32_t callcnt = 0;
-  
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-		if (op_ret >= 0) {
-			if (NS(this) == (xlator_t *)cookie) {
-				memcpy (local->sh_struct->ns_file_checksum, 
-					file_checksum, NAME_MAX);
-				memcpy (local->sh_struct->ns_dir_checksum, 
-					dir_checksum, NAME_MAX);
-			} else {
-				if (local->entry_count == 0) {
-					/* Initialize the dir_checksum to be 
-					 * used for comparision with other 
-					 * storage nodes. Should be done for
-					 * the first successful call *only*. 
-					 */
-					/* Using 'entry_count' as a flag */
-					local->entry_count = 1; 
-					memcpy (local->sh_struct->dir_checksum,
-						dir_checksum, NAME_MAX);
-				}
-
-				/* Reply from the storage nodes */
-				for (index = 0; 
-				     index < NAME_MAX; index++) {
-					/* Files should be present in only 
-					   one node */
-					local->sh_struct->file_checksum[index] ^= file_checksum[index];
-	  
-					/* directory structure should be same 
-					   accross */
-					if (local->sh_struct->dir_checksum[index] != dir_checksum[index])
-						local->failed = 1;
-				}
-			}
-		} 
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		for (index = 0; index < NAME_MAX ; index++) {
-			if (local->sh_struct->file_checksum[index] != 
-			    local->sh_struct->ns_file_checksum[index]) {
-				local->failed = 1;
-				break;
-			}
-			if (local->sh_struct->dir_checksum[index] != 
-			    local->sh_struct->ns_dir_checksum[index]) {
-				local->failed = 1;
-				break;
-			}
-		}
-	
-		if (local->failed) {
-			/* Log it, it should be a rare event */
-			gf_log (this->name, GF_LOG_WARNING, 
-				"Self-heal triggered on directory %s", 
-				local->loc1.path);
-
-			/* Any self heal will be done at the directory level */
-			local->op_ret = -1;
-			local->failed = 0;
-      
-			local->fd = fd_create (local->loc1.inode, 
-					       frame->root->pid);
-			local->call_count = priv->child_count + 1;
-	
-			for (index = 0; 
-			     index < (priv->child_count + 1); index++) {
-				STACK_WIND_COOKIE (frame,
-						   unify_bgsh_opendir_cbk,
-						   priv->xl_array[index]->name,
-						   priv->xl_array[index],
-						   priv->xl_array[index]->fops->opendir,
-						   &local->loc1,
-						   local->fd);
-			}
-      
-			/* opendir can be done on the directory */
-			return 0;
-		}
-
-		/* no mismatch */
-		unify_local_wipe (local);
-		STACK_DESTROY (frame->root);
-	}
-
-	return 0;
-}
-
-/* Background self-heal part over */
-
-
-
-
-/**
- * zr_unify_self_heal - 
- * 
- * @frame: frame used in lookup. get a copy of it, and use that copy.
- * @this: pointer to unify xlator.
- * @inode: pointer to inode, for which the consistency check is required.
- *
- */
-int32_t 
-zr_unify_self_heal (call_frame_t *frame,
-		    xlator_t *this,
-		    unify_local_t *local)
-{
-	unify_private_t *priv = this->private;
-	call_frame_t *bg_frame = NULL;
-	unify_local_t *bg_local = NULL;
-	inode_t *tmp_inode = NULL;
-	dict_t *tmp_dict = NULL;
-	int16_t index = 0;
-  
-	if (local->inode_generation < priv->inode_generation) {
-		/* Any self heal will be done at the directory level */
-		/* Update the inode's generation to the current generation
-		   value. */
-		local->inode_generation = priv->inode_generation;
-		inode_ctx_put (local->loc1.inode, this, 
-			  (uint64_t)(long)local->inode_generation);
-
-		if (priv->self_heal == ZR_UNIFY_FG_SELF_HEAL) {
-			local->op_ret = 0;
-			local->failed = 0;
-			local->call_count = priv->child_count + 1;
-			local->sh_struct = 
-	         	     GF_CALLOC (1, sizeof (struct unify_self_heal_struct),
-                                        gf_unify_mt_unify_self_heal_struct);
-      
-			/* +1 is for NS */
-			for (index = 0; 
-			     index < (priv->child_count + 1); index++) {
-				STACK_WIND_COOKIE (frame,
-						   unify_sh_checksum_cbk,
-						   priv->xl_array[index],
-						   priv->xl_array[index],
-						   priv->xl_array[index]->fops->checksum,
-						   &local->loc1,
-						   0);
-			}
-
-			/* Self-heal in foreground, hence no need 
-			   to UNWIND here */
-			return 0;
-		}
-
-		/* Self Heal done in background */
-		bg_frame = copy_frame (frame);
-		INIT_LOCAL (bg_frame, bg_local);
-		loc_copy (&bg_local->loc1, &local->loc1);
-		bg_local->op_ret = 0;
-		bg_local->failed = 0;
-		bg_local->call_count = priv->child_count + 1;
-		bg_local->sh_struct = 
-			GF_CALLOC (1, sizeof (struct unify_self_heal_struct),
-                                   gf_unify_mt_unify_self_heal_struct);
-    
-		/* +1 is for NS */
-		for (index = 0; index < (priv->child_count + 1); index++) {
-			STACK_WIND_COOKIE (bg_frame,
-					   unify_bgsh_checksum_cbk,
-					   priv->xl_array[index],
-					   priv->xl_array[index],
-					   priv->xl_array[index]->fops->checksum,
-					   &bg_local->loc1,
-					   0);
-		}
-	}
-
-	/* generation number matches, self heal already done or
-	 * self heal done in background: just do STACK_UNWIND 
-	 */
-	tmp_inode = local->loc1.inode;
-	tmp_dict = local->dict;
-
-	unify_local_wipe (local);
-
-	/* This is lookup_cbk ()'s UNWIND. */
-	STACK_UNWIND (frame,
-		      local->op_ret,
-		      local->op_errno,
-		      tmp_inode,
-		      &local->stbuf,
-		      local->dict,
-                      &local->oldpostparent);
-
-	if (tmp_dict)
-		dict_unref (tmp_dict);
-
-	return 0;
-}
-
diff --git a/xlators/cluster/unify/src/unify.c b/xlators/cluster/unify/src/unify.c
deleted file mode 100644
index 422c8d6d8..000000000
--- a/xlators/cluster/unify/src/unify.c
+++ /dev/null
@@ -1,4589 +0,0 @@
-/*
-  Copyright (c) 2006-2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-/**
- * xlators/cluster/unify:
- *     - This xlator is one of the main translator in GlusterFS, which
- *   actually does the clustering work of the file system. One need to 
- *   understand that, unify assumes file to be existing in only one of 
- *   the child node, and directories to be present on all the nodes. 
- *
- * NOTE:
- *   Now, unify has support for global namespace, which is used to keep a 
- * global view of fs's namespace tree. The stat for directories are taken
- * just from the namespace, where as for files, just 'ia_ino' is taken from
- * Namespace node, and other stat info is taken from the actual storage node.
- * Also Namespace node helps to keep consistant inode for files across 
- * glusterfs (re-)mounts.
- */
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include "glusterfs.h"
-#include "unify.h"
-#include "dict.h"
-#include "xlator.h"
-#include "hashfn.h"
-#include "logging.h"
-#include "stack.h"
-#include "defaults.h"
-#include "common-utils.h"
-#include <signal.h>
-#include <libgen.h>
-#include "compat-errno.h"
-#include "compat.h"
-
-#define UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR(_loc) do { \
-  if (!(_loc && _loc->inode)) {                            \
-    STACK_UNWIND (frame, -1, EINVAL, NULL, NULL, NULL);    \
-    return 0;                                              \
-  }                                                        \
-} while(0)
-
-
-#define UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR(_fd) do { \
-  if (!(_fd && !fd_ctx_get (_fd, this, NULL))) {       \
-    STACK_UNWIND (frame, -1, EBADFD, NULL, NULL);      \
-    return 0;                                          \
-  }                                                    \
-} while(0)
-
-#define UNIFY_CHECK_FD_AND_UNWIND_ON_ERR(_fd) do { \
-  if (!_fd) {                                      \
-    STACK_UNWIND (frame, -1, EBADFD, NULL, NULL);  \
-    return 0;                                      \
-  }                                                \
-} while(0)
-
-/**
- * unify_local_wipe - free all the extra allocation of local->* here.
- */
-static void 
-unify_local_wipe (unify_local_t *local)
-{
-	/* Free the strdup'd variables in the local structure */
-	if (local->name) {
-		GF_FREE (local->name);
-	}
-	loc_wipe (&local->loc1);
-	loc_wipe (&local->loc2);
-}
-
-
-
-/*
- * unify_normalize_stats -
- */
-void
-unify_normalize_stats (struct statvfs *buf,
-		       unsigned long bsize,
-		       unsigned long frsize)
-{
-	double factor;
-
-	if (buf->f_bsize != bsize) {
-		factor = ((double) buf->f_bsize) / bsize;
-		buf->f_bsize  = bsize;
-		buf->f_bfree  = (fsblkcnt_t) (factor * buf->f_bfree);
-		buf->f_bavail = (fsblkcnt_t) (factor * buf->f_bavail);
-	}
-  
-	if (buf->f_frsize != frsize) {
-		factor = ((double) buf->f_frsize) / frsize;
-		buf->f_frsize = frsize;
-		buf->f_blocks = (fsblkcnt_t) (factor * buf->f_blocks);
-	}
-}
-
-
-xlator_t *
-unify_loc_subvol (loc_t *loc, xlator_t *this)
-{
-	unify_private_t *priv = NULL;
-	xlator_t        *subvol = NULL;
-	int16_t         *list = NULL;
-	long             index = 0;
-	xlator_t        *subvol_i = NULL;
-	int              ret = 0;
-	uint64_t         tmp_list = 0;
-
-	priv   = this->private;
-	subvol = NS (this);
-
-	if (!IA_ISDIR (loc->inode->ia_type)) {
-		ret = inode_ctx_get (loc->inode, this, &tmp_list);
-		list = (int16_t *)(long)tmp_list;
-		if (!list)
-			goto out;
-
-		for (index = 0; list[index] != -1; index++) {
-			subvol_i = priv->xl_array[list[index]];
-			if (subvol_i != NS (this)) {
-				subvol = subvol_i;
-				break;
-			}
-		}
-	}
-out:
-	return subvol;
-}
-
-
-
-/**
- * unify_statfs_cbk -
- */
-int32_t
-unify_statfs_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-		  struct statvfs *stbuf)
-{
-	int32_t callcnt = 0;
-	struct statvfs *dict_buf = NULL;
-	unsigned long bsize;
-	unsigned long frsize;
-	unify_local_t *local = (unify_local_t *)frame->local;
-	call_frame_t *prev_frame = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret >= 0) {
-			/* when a call is successfull, add it to local->dict */
-			dict_buf = &local->statvfs_buf;
-
-			if (dict_buf->f_bsize != 0) {
-				bsize  = max (dict_buf->f_bsize, 
-					      stbuf->f_bsize);
-
-				frsize = max (dict_buf->f_frsize, 
-					      stbuf->f_frsize);
-				unify_normalize_stats(dict_buf, bsize, frsize);
-				unify_normalize_stats(stbuf, bsize, frsize);
-			} else {
-				dict_buf->f_bsize   = stbuf->f_bsize;
-				dict_buf->f_frsize  = stbuf->f_frsize;
-			}
-      
-			dict_buf->f_blocks += stbuf->f_blocks;
-			dict_buf->f_bfree  += stbuf->f_bfree;
-			dict_buf->f_bavail += stbuf->f_bavail;
-			dict_buf->f_files  += stbuf->f_files;
-			dict_buf->f_ffree  += stbuf->f_ffree;
-			dict_buf->f_favail += stbuf->f_favail;
-			dict_buf->f_fsid    = stbuf->f_fsid;
-			dict_buf->f_flag    = stbuf->f_flag;
-			dict_buf->f_namemax = stbuf->f_namemax;
-			local->op_ret = op_ret;
-		} else {
-			/* fop on storage node has failed due to some error */
-			if (op_errno != ENOTCONN) {
-				gf_log (this->name, GF_LOG_ERROR, 
-					"child(%s): %s", 
-					prev_frame->this->name, 
-					strerror (op_errno));
-			}
-			local->op_errno = op_errno;
-		}
-		callcnt = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-			      &local->statvfs_buf);
-	}
-
-	return 0;
-}
-
-/**
- * unify_statfs -
- */
-int32_t
-unify_statfs (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *loc)
-{
-	unify_local_t *local = NULL;
-	xlator_list_t *trav = this->children;
-
-	INIT_LOCAL (frame, local);
-	local->call_count = ((unify_private_t *)this->private)->child_count;
-
-	while(trav) {
-		STACK_WIND (frame,
-			    unify_statfs_cbk,
-			    trav->xlator,
-			    trav->xlator->fops->statfs,
-			    loc);
-		trav = trav->next;
-	}
-
-	return 0;
-}
-
-/**
- * unify_buf_cbk - 
- */
-int32_t
-unify_buf_cbk (call_frame_t *frame,
-	       void *cookie,
-	       xlator_t *this,
-	       int32_t op_ret,
-	       int32_t op_errno,
-	       struct iatt *buf)
-{
-	int32_t callcnt = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-    
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"%s(): child(%s): path(%s): %s", 
-				gf_fop_list[frame->root->op],
-				prev_frame->this->name, 
-				(local->loc1.path)?local->loc1.path:"", 
-				strerror (op_errno));
-
-			local->op_errno = op_errno;
-			if ((op_errno == ENOENT) && priv->optimist) 
-				local->op_ret = 0;
-		}
-
-		if (op_ret >= 0) {
-			local->op_ret = 0;
-
-			if (NS (this) == prev_frame->this) {
-				local->ia_ino = buf->ia_ino;
-				/* If the entry is directory, get the stat
-				   from NS node */
-				if (IA_ISDIR (buf->ia_type) || 
-				    !local->stbuf.ia_blksize) {
-					local->stbuf = *buf;
-				}
-			}
-
-			if ((!IA_ISDIR (buf->ia_type)) && 
-			    (NS (this) != prev_frame->this)) {
-				/* If file, take the stat info from Storage 
-				   node. */
-				local->stbuf = *buf;
-			}
-		}
-	}
-	UNLOCK (&frame->lock);
-    
-	if (!callcnt) {
-		/* If the inode number is not filled, operation should
-		   fail */
-		if (!local->ia_ino)
-			local->op_ret = -1;
-
-		local->stbuf.ia_ino = local->ia_ino;
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-			      &local->stbuf);
-	}
-
-	return 0;
-}
-
-#define check_if_dht_linkfile(s) \
-        ((st_mode_from_ia (s->ia_prot, s->ia_type) & ~S_IFMT) == S_ISVTX)
-
-/**
- * unify_lookup_cbk - 
- */
-int32_t 
-unify_lookup_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-		  inode_t *inode,
-		  struct iatt *buf,
-		  dict_t *dict,
-                  struct iatt *postparent)
-{
-	int32_t callcnt = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-	inode_t *tmp_inode = NULL;
-	dict_t *local_dict = NULL;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
- 
-		if (op_ret == -1) {
-                        if (local->revalidate &&
-                            (op_errno == ESTALE)) {
-                                /* ESTALE takes priority */
-                                local->op_errno = op_errno;
-                                local->failed = 1;
-                        }
-
-			if ((op_errno != ENOTCONN) 
-                            && (op_errno != ENOENT)
-                            && (local->op_errno != ESTALE)) {
-                                /* if local->op_errno is already ESTALE, then
-                                 * ESTALE has to propogated to the parent first.
-                                 * do not enter here.
-                                 */
-				gf_log (this->name, GF_LOG_ERROR,
-					"child(%s): path(%s): %s", 
-					priv->xl_array[(long)cookie]->name, 
-					local->loc1.path, strerror (op_errno));
-                                local->op_errno = op_errno;
-				local->failed = 1;
-
-			} else if (local->revalidate && 
-                                   (local->op_errno != ESTALE) &&
-				   !(priv->optimist && (op_errno == ENOENT))) {
-
-				gf_log (this->name, 
-					(op_errno == ENOTCONN) ? 
-					GF_LOG_DEBUG:GF_LOG_ERROR,
-					"child(%s): path(%s): %s", 
-					priv->xl_array[(long)cookie]->name, 
-					local->loc1.path, strerror (op_errno));
-				local->op_errno = op_errno;
-				local->failed = 1;
-			}
-		}
-
-		if (op_ret == 0) {
-			local->op_ret = 0; 
-			
-			if (check_if_dht_linkfile(buf)) {
-				gf_log (this->name, GF_LOG_CRITICAL,
-					"file %s may be DHT link file on %s, "
-					"make sure the backend is not shared "
-					"between unify and DHT", 
-					local->loc1.path, 
-					priv->xl_array[(long)cookie]->name);
-			}
-
-			if (local->stbuf.ia_type && local->stbuf.ia_blksize) {
-				/* make sure we already have a stbuf
-				   stored in local->stbuf */
-				if (IA_ISDIR (local->stbuf.ia_type) && 
-				    !IA_ISDIR (buf->ia_type)) {
-					gf_log (this->name, GF_LOG_CRITICAL, 
-						"[CRITICAL] '%s' is directory "
-						"on namespace, non-directory "
-						"on node '%s', returning EIO",
-						local->loc1.path, 
-						priv->xl_array[(long)cookie]->name);
-					local->return_eio = 1;
-				}
-				if (!IA_ISDIR (local->stbuf.ia_type) && 
-				    IA_ISDIR (buf->ia_type)) {
-					gf_log (this->name, GF_LOG_CRITICAL, 
-						"[CRITICAL] '%s' is directory "
-						"on node '%s', non-directory "
-						"on namespace, returning EIO",
-						local->loc1.path, 
-						priv->xl_array[(long)cookie]->name);
-					local->return_eio = 1;
-				}
-			}
-	
-			if (!local->revalidate && !IA_ISDIR (buf->ia_type)) {
-				/* This is the first time lookup on file*/
-				if (!local->list) {
-					/* list is not allocated, allocate 
-					   the max possible range */
-					local->list = GF_CALLOC (1, 2 * (priv->child_count + 2),
-                                                                 gf_unify_mt_int16_t);
-					if (!local->list) {
-						gf_log (this->name, 
-							GF_LOG_CRITICAL, 
-							"Not enough memory");
-						STACK_UNWIND (frame, -1, 
-							      ENOMEM, inode, 
-							      NULL, NULL, NULL);
-						return 0;
-					}
-				}
-				/* update the index of the list */
-				local->list [local->index++] = 
-					(int16_t)(long)cookie;
-			}
-                        
-                        if (!local->revalidate && IA_ISDIR (buf->ia_type)) {
-                                /* fresh lookup of a directory */
-                                inode_ctx_put (local->loc1.inode, this, 
-                                               priv->inode_generation);
-                        }
-
-			if ((!local->dict) && dict &&
-			    (priv->xl_array[(long)cookie] != NS(this)))	{
-				local->dict = dict_ref (dict);
-			}
-
-			/* index of NS node is == total child count */
-			if (priv->child_count == (int16_t)(long)cookie) {
-				/* Take the inode number from namespace */
-				local->ia_ino = buf->ia_ino;
-				if (IA_ISDIR (buf->ia_type) || 
-				    !(local->stbuf.ia_blksize)) {
-					local->stbuf = *buf;
-                                        local->oldpostparent = *postparent;
-				}
-			} else if (!IA_ISDIR (buf->ia_type)) {
-				/* If file, then get the stat from 
-				   storage node */
-				local->stbuf = *buf;
-			}
-
-			if (local->ia_nlink < buf->ia_nlink) {
-				local->ia_nlink = buf->ia_nlink;
-			}
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		local_dict = local->dict;
-		if (local->return_eio) {
-			gf_log (this->name, GF_LOG_CRITICAL, 
-				"[CRITICAL] Unable to fix the path (%s) with "
-				"self-heal, try manual verification. "
-				"returning EIO.", local->loc1.path);
-			unify_local_wipe (local);
-			STACK_UNWIND (frame, -1, EIO, inode, NULL, NULL);
-			if (local_dict)	{
-				dict_unref (local_dict);
-			}
-			return 0;
-		}
-
-		if (!local->stbuf.ia_blksize) {
-			/* Inode not present */
-			local->op_ret = -1;
-		} else {
-			if (!local->revalidate && 
-			    !IA_ISDIR (local->stbuf.ia_type)) { 
-				/* If its a file, big array is useless, 
-				   allocate the smaller one */
-				int16_t *list = NULL;
-				list = GF_CALLOC (1, 2 * (local->index + 1),
-                                                  gf_unify_mt_int16_t);
-				ERR_ABORT (list);
-				memcpy (list, local->list, 2 * local->index);
-				/* Make the end of the list as -1 */
-				GF_FREE (local->list);
-				local->list = list;
-				local->list [local->index] = -1;
-				/* Update the inode's ctx with proper array */
-				/* TODO: log on failure */
-				inode_ctx_put (local->loc1.inode, this, 
-					       (uint64_t)(long)local->list);
-			}
-
-			if (IA_ISDIR(local->loc1.inode->ia_type)) {
-				/* lookup is done for directory */
-				if (local->failed && priv->self_heal) {
-					/* Triggering self-heal */
-                                        /* means, self-heal required for this 
-					   inode */
-					local->inode_generation = 0; 
-					priv->inode_generation++;
-				}
-			} else {
-				local->stbuf.ia_ino = local->ia_ino;
-			}
-	  
-			local->stbuf.ia_nlink = local->ia_nlink;
-		}
-		if (local->op_ret == -1) {
-			if (!local->revalidate && local->list)
-				GF_FREE (local->list);
-		}
-
-		if ((local->op_ret >= 0) && local->failed && 
-		    local->revalidate) {
-			/* Done revalidate, but it failed */
-			if ((op_errno != ENOTCONN) 
-                            && (local->op_errno != ESTALE)) {
-				gf_log (this->name, GF_LOG_ERROR, 
-					"Revalidate failed for path(%s): %s", 
-					local->loc1.path, strerror (op_errno));
-			}
-			local->op_ret = -1;
-		}
-
-		if ((priv->self_heal && !priv->optimist) && 
-		    (!local->revalidate && (local->op_ret == 0) && 
-		     IA_ISDIR(local->stbuf.ia_type))) {
-			/* Let the self heal be done here */
-			zr_unify_self_heal (frame, this, local);
-			local_dict = NULL;
-		} else {
-                        if (local->failed) {
-                                /* NOTE: directory lookup is sent to all 
-                                 * subvolumes and success from a subvolume
-                                 * might set local->op_ret to 0 (zero) */
-                                local->op_ret = -1;
-                        }
-
-			/* either no self heal, or op_ret == -1 (failure) */
-			tmp_inode = local->loc1.inode;
-			unify_local_wipe (local);
-			STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-				      tmp_inode, &local->stbuf, local->dict,
-                                      &local->oldpostparent);
-		}
-		if (local_dict) {
-			dict_unref (local_dict);
-		}
-	}
-  
-	return 0;
-}
-
-/**
- * unify_lookup - 
- */
-int32_t 
-unify_lookup (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *loc,
-	      dict_t *xattr_req)
-{
-	unify_local_t *local = NULL;
-	unify_private_t *priv = this->private;
-	int16_t *list = NULL;
-	long index = 0;
-
-	if (!(loc && loc->inode)) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"%s: Argument not right", loc?loc->path:"(null)");
-		STACK_UNWIND (frame, -1, EINVAL, NULL, NULL, NULL, NULL);
-		return 0;
-	}
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	loc_copy (&local->loc1, loc);
-	if (local->loc1.path == NULL) {
-		gf_log (this->name, GF_LOG_CRITICAL, "Not enough memory :O");
-		STACK_UNWIND (frame, -1, ENOMEM, loc->inode, NULL, NULL, NULL);
-		return 0;
-	}
-        
-        if (inode_ctx_get (loc->inode, this, NULL)
-            && IA_ISDIR (loc->inode->ia_type)) {
-                local->revalidate = 1;
-        }
-
-	if (!inode_ctx_get (loc->inode, this, NULL) && 
-	    loc->inode->ia_type && 
-	    !IA_ISDIR (loc->inode->ia_type)) {
-		uint64_t tmp_list = 0;
-		/* check if revalidate or fresh lookup */
-		inode_ctx_get (loc->inode, this, &tmp_list);
-		local->list = (int16_t *)(long)tmp_list;
-	}
-
-	if (local->list) {
-		list = local->list;
-		for (index = 0; list[index] != -1; index++);
-		if (index != 2) {
-			if (index < 2) {
-				gf_log (this->name, GF_LOG_ERROR,
-					"returning ESTALE for %s: file "
-					"count is %ld", loc->path, index);
-				/* Print where all the file is present */
-				for (index = 0; 
-				     local->list[index] != -1; index++) {
-					gf_log (this->name, GF_LOG_ERROR, 
-						"%s: found on %s", loc->path, 
-						priv->xl_array[list[index]]->name);
-				}
-				unify_local_wipe (local);
-				STACK_UNWIND (frame, -1, ESTALE, 
-					      NULL, NULL, NULL, NULL);
-				return 0;  
-			} else {
-				/* There are more than 2 presences */
-				/* Just log and continue */
-				gf_log (this->name, GF_LOG_ERROR,
-					"%s: file count is %ld", 
-					loc->path, index);
-				/* Print where all the file is present */
-				for (index = 0; 
-				     local->list[index] != -1; index++) {
-					gf_log (this->name, GF_LOG_ERROR, 
-						"%s: found on %s", loc->path, 
-						priv->xl_array[list[index]]->name);
-				}
-			}
-		}
-      
-		/* is revalidate */
-		local->revalidate = 1;
-      
-		for (index = 0; list[index] != -1; index++)
-			local->call_count++;
-      
-		for (index = 0; list[index] != -1; index++) {
-			char need_break = (list[index+1] == -1);
-			STACK_WIND_COOKIE (frame,
-					   unify_lookup_cbk,
-					   (void *)(long)list[index], //cookie
-					   priv->xl_array [list[index]],
-					   priv->xl_array [list[index]]->fops->lookup,
-					   loc,
-					   xattr_req);
-			if (need_break)
-				break;
-		}
-	} else {
-		if (loc->inode->ia_type) {
-			if (inode_ctx_get (loc->inode, this, NULL)) {
-				inode_ctx_get (loc->inode, this, 
-					       &local->inode_generation);
-			}
-		}
-		/* This is first call, there is no list */
-		/* call count should be all child + 1 namespace */
-		local->call_count = priv->child_count + 1;
-      
-		for (index = 0; index <= priv->child_count; index++) {
-			STACK_WIND_COOKIE (frame,
-					   unify_lookup_cbk,
-					   (void *)index, //cookie
-					   priv->xl_array[index],
-					   priv->xl_array[index]->fops->lookup,
-					   loc,
-					   xattr_req);
-		}
-	}
-
-	return 0;
-}
-
-/**
- * unify_stat - if directory, get the stat directly from NameSpace child.
- *     if file, check for a hint and send it only there (also to NS).
- *     if its a fresh stat, then do it on all the nodes.
- *
- * NOTE: for all the call, sending cookie as xlator pointer, which will be 
- *       used in cbk.
- */
-int32_t
-unify_stat (call_frame_t *frame,
-	    xlator_t *this,
-	    loc_t *loc)
-{
-	unify_local_t *local = NULL;
-	unify_private_t *priv = this->private;
-	int16_t index = 0;
-	int16_t *list = NULL;
-	uint64_t tmp_list = 0;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	loc_copy (&local->loc1, loc);
-	if (local->loc1.path == NULL) {
-		gf_log (this->name, GF_LOG_CRITICAL, "Not enough memory :O");
-		STACK_UNWIND (frame, -1, ENOMEM, NULL);
-		return 0;
-	}
-	local->ia_ino = loc->inode->ino;
-	if (IA_ISDIR (loc->inode->ia_type)) {
-		/* Directory */
-		local->call_count = 1;
-		STACK_WIND (frame, unify_buf_cbk, NS(this),
-			    NS(this)->fops->stat, loc);
-	} else {
-		/* File */
-		inode_ctx_get (loc->inode, this, &tmp_list);
-    		list = (int16_t *)(long)tmp_list;
-
-		for (index = 0; list[index] != -1; index++)
-			local->call_count++;
-    
-		for (index = 0; list[index] != -1; index++) {
-			char need_break = (list[index+1] == -1);
-			STACK_WIND (frame,
-				    unify_buf_cbk,
-				    priv->xl_array[list[index]],
-				    priv->xl_array[list[index]]->fops->stat,
-				    loc);
-			if (need_break)
-				break;
-		}
-	}
-
-	return 0;
-}
-
-/**
- * unify_access_cbk -
- */
-int32_t
-unify_access_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-
-/**
- * unify_access - Send request to only namespace, which has all the 
- *      attributes set for the file.
- */
-int32_t
-unify_access (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *loc,
-	      int32_t mask)
-{
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	STACK_WIND (frame,
-		    unify_access_cbk,
-		    NS(this),
-		    NS(this)->fops->access,
-		    loc,
-		    mask);
-
-	return 0;
-}
-
-int32_t
-unify_mkdir_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-		 inode_t *inode,
-                 struct iatt *buf,
-                 struct iatt *preparent,
-                 struct iatt *postparent)
-{
-	int32_t callcnt = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-	inode_t *tmp_inode = NULL;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-  
-		if ((op_ret == -1) && !(priv->optimist && 
-					(op_errno == ENOENT || 
-					 op_errno == EEXIST))) {
-			/* TODO: Decrement the inode_generation of 
-			 * this->inode's parent inode, hence the missing 
-			 * directory is created properly by self-heal. 
-			 * Currently, there is no way to get the parent 
-			 * inode directly.
-			 */
-			gf_log (this->name, GF_LOG_ERROR,
-				"child(%s): path(%s): %s", 
-				priv->xl_array[(long)cookie]->name, 
-				local->loc1.path, strerror (op_errno));
-			if (op_errno != EEXIST)
-				local->failed = 1;
-			local->op_errno = op_errno;
-		}
-  
-		if (op_ret >= 0)
-			local->op_ret = 0;
-
-	}
-	UNLOCK (&frame->lock);
-  
-	if (!callcnt) {
-		if (!local->failed) {
-			inode_ctx_put (local->loc1.inode, this, 
-				       priv->inode_generation);
-		}
-		
-		tmp_inode = local->loc1.inode;
-		unify_local_wipe (local);
-
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-			      tmp_inode, &local->stbuf,
-                              &local->oldpreparent, &local->oldpostparent);
-	}
-
-	return 0;
-}
-
-/**
- * unify_ns_mkdir_cbk -
- */
-int32_t
-unify_ns_mkdir_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    inode_t *inode,
-                    struct iatt *buf,
-                    struct iatt *preparent,
-                    struct iatt *postparent)
-{
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-	long index = 0;
-
-	if (op_ret == -1) {
-		/* No need to send mkdir request to other servers, 
-		 * as namespace action failed 
-		 */
-		gf_log (this->name, GF_LOG_ERROR,
-			"namespace: path(%s): %s", 
-			local->name, strerror (op_errno));
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, op_ret, op_errno, inode, NULL,
-                              NULL, NULL);
-		return 0;
-	}
-  
-	/* Create one inode for this entry */
-	local->op_ret = 0;
-	local->stbuf = *buf;
-
-        local->oldpreparent = *preparent;
-        local->oldpostparent = *postparent;
-
-	local->call_count = priv->child_count;
-
-	/* Send mkdir request to all the nodes now */
-	for (index = 0; index < priv->child_count; index++) {
-		STACK_WIND_COOKIE (frame,
-				   unify_mkdir_cbk,
-				   (void *)index, //cookie
-				   priv->xl_array[index],
-				   priv->xl_array[index]->fops->mkdir,
-				   &local->loc1,
-				   local->mode);
-	}
-  
-	return 0;
-}
-
-
-/**
- * unify_mkdir -
- */
-int32_t
-unify_mkdir (call_frame_t *frame,
-	     xlator_t *this,
-	     loc_t *loc,
-	     mode_t mode)
-{
-	unify_local_t *local = NULL;
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	local->mode = mode;
-
-	loc_copy (&local->loc1, loc);
-
-	if (local->loc1.path == NULL) {
-		gf_log (this->name, GF_LOG_CRITICAL, "Not enough memory :O");
-		STACK_UNWIND (frame, -1, ENOMEM, NULL, NULL);
-		return 0;
-	}
-
-	STACK_WIND (frame,
-		    unify_ns_mkdir_cbk,
-		    NS(this),
-		    NS(this)->fops->mkdir,
-		    loc,
-		    mode);
-	return 0;
-}
-
-/**
- * unify_rmdir_cbk -
- */
-int32_t
-unify_rmdir_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-                 struct iatt *preparent,
-                 struct iatt *postparent)
-{
-	int32_t callcnt = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-		if (op_ret == 0 || (priv->optimist && (op_errno == ENOENT)))
-			local->op_ret = 0;
-		if (op_ret == -1)
-			local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno,
-                              &local->oldpreparent, &local->oldpostparent);
-	}
-
-	return 0;
-}
-
-/**
- * unify_ns_rmdir_cbk -
- */
-int32_t
-unify_ns_rmdir_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-                    struct iatt *preparent,
-                    struct iatt *postparent)
-{
-	int16_t index = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-  
-	if (op_ret == -1) {
-		/* No need to send rmdir request to other servers, 
-		 * as namespace action failed 
-		 */
-		gf_log (this->name, 
-			((op_errno != ENOTEMPTY) ? 
-			 GF_LOG_ERROR : GF_LOG_DEBUG),
-			"namespace: path(%s): %s", 
-			local->loc1.path, strerror (op_errno));
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, op_ret, op_errno, NULL, NULL);
-		return 0;
-	}
-
-	local->call_count = priv->child_count;
-
-        local->oldpreparent = *preparent;
-        local->oldpostparent = *postparent;
-
-	for (index = 0; index < priv->child_count; index++) {
-		STACK_WIND (frame,
-			    unify_rmdir_cbk,
-			    priv->xl_array[index],
-			    priv->xl_array[index]->fops->rmdir,
-			    &local->loc1);
-	}
-
-	return 0;
-}
-
-/**
- * unify_rmdir -
- */
-int32_t
-unify_rmdir (call_frame_t *frame,
-	     xlator_t *this,
-	     loc_t *loc)
-{
-	unify_local_t *local = NULL;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-
-	loc_copy (&local->loc1, loc);
-	if (local->loc1.path == NULL) {
-		gf_log (this->name, GF_LOG_CRITICAL, "Not enough memory :O");
-		STACK_UNWIND (frame, -1, ENOMEM, NULL, NULL);
-		return 0;
-	}
-
-	STACK_WIND (frame,
-		    unify_ns_rmdir_cbk,
-		    NS(this),
-		    NS(this)->fops->rmdir,
-		    loc);
-
-	return 0;
-}
-
-/**
- * unify_open_cbk -
- */
-int32_t
-unify_open_cbk (call_frame_t *frame,
-		void *cookie,
-		xlator_t *this,
-		int32_t op_ret,
-		int32_t op_errno,
-		fd_t *fd)
-{
-	int32_t callcnt = 0;
-	unify_local_t *local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret >= 0) {
-			local->op_ret = op_ret;
-			if (NS(this) != (xlator_t *)cookie) {
-				/* Store child node's ptr, used in 
-				   all the f*** / FileIO calls */
-				fd_ctx_set (fd, this, (uint64_t)(long)cookie);
-			}
-		}
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			local->failed = 1;
-		}
-		callcnt = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
-  
-	if (!callcnt) {
-		if ((local->failed == 1) && (local->op_ret >= 0)) {
-			local->call_count = 1;
-			/* return -1 to user */
-			local->op_ret = -1;
-			//local->op_errno = EIO; 
-      
-			if (!fd_ctx_get (local->fd, this, NULL)) {
-				gf_log (this->name, GF_LOG_ERROR, 
-					"Open success on child node, "
-					"failed on namespace");
-			} else {
-				gf_log (this->name, GF_LOG_ERROR, 
-					"Open success on namespace, "
-					"failed on child node");
-			}
-		}
-
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, 
-			      local->op_errno, local->fd);
-	}
-
-	return 0;
-}
-
-#ifdef GF_DARWIN_HOST_OS
-/**
- * unify_create_lookup_cbk - 
- */
-int32_t 
-unify_open_lookup_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno,
-		       inode_t *inode,
-		       struct iatt *buf,
-		       dict_t *dict,
-		       struct iatt *postparent)
-{
-	int32_t callcnt = 0;
-	int16_t index = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-		if ((op_ret == -1) && (op_errno != ENOENT)) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"child(%s): path(%s): %s", 
-				priv->xl_array[(long)cookie]->name, 
-				local->loc1.path, strerror (op_errno));
-			local->op_errno = op_errno;
-		}
-    
-		if (op_ret >= 0) {
-			local->op_ret = op_ret; 
-			local->index++;
-			if (NS(this) == priv->xl_array[(long)cookie]) {
-				local->list[0] = (int16_t)(long)cookie;
-			} else {
-				local->list[1] = (int16_t)(long)cookie;
-			}
-			if (IA_ISDIR (buf->ia_type))
-				local->failed = 1;
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		int16_t file_list[3] = {0,};
-		local->op_ret = -1;
-
-		file_list[0] = local->list[0];
-		file_list[1] = local->list[1];
-		file_list[2] = -1;
-
-		if (local->index != 2) {
-			/* Lookup failed, can't do open */
-			gf_log (this->name, GF_LOG_ERROR,
-				"%s: present on %d nodes", 
-				local->name, local->index);
-
-			if (local->index < 2) {
-				unify_local_wipe (local);
-				gf_log (this->name, GF_LOG_ERROR,
-					"returning as file found on less "
-					"than 2 nodes");
-				STACK_UNWIND (frame, local->op_ret, 
-					      local->op_errno, local->fd);
-				return 0;
-			}
-		}
-
-		if (local->failed) {
-			/* Open on directory, return EISDIR */
-			unify_local_wipe (local);
-			STACK_UNWIND (frame, -1, EISDIR, local->fd);
-			return 0;
-		}
-
-		/* Everything is perfect :) */    
-		local->call_count = 2;
-    
-		for (index = 0; file_list[index] != -1; index++) {
-			char need_break = (file_list[index+1] == -1);
-			STACK_WIND_COOKIE (frame,
-					   unify_open_cbk,
-					   priv->xl_array[file_list[index]],
-					   priv->xl_array[file_list[index]],
-					   priv->xl_array[file_list[index]]->fops->open,
-					   &local->loc1,
-					   local->flags,
-					   local->fd, local->wbflags);
-			if (need_break)
-				break;
-		}
-	}
-
-	return 0;
-}
-
-
-int32_t
-unify_open_readlink_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno,
-			 const char *path,
-			 struct iatt *sbuf)
-{
-	int16_t index = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-
-	if (op_ret == -1) {
-		STACK_UNWIND (frame, -1, ENOENT);
-		return 0;
-	}
-
-	if (path[0] == '/') {
-		local->name = gf_strdup (path);
-		ERR_ABORT (local->name);
-	} else {
-		char *tmp_str = gf_strdup (local->loc1.path);
-		char *tmp_base = dirname (tmp_str);
-		local->name = GF_CALLOC (1, ZR_PATH_MAX, gf_unify_mt_char);
-		strcpy (local->name, tmp_base);
-		strncat (local->name, "/", 1);
-		strcat (local->name, path);
-		GF_FREE (tmp_str);
-	}
-  
-	local->list = GF_CALLOC (1, sizeof (int16_t) * 3,
-                                 gf_unify_mt_int16_t);
-	ERR_ABORT (local->list);
-	local->call_count = priv->child_count + 1;
-	local->op_ret = -1;
-	for (index = 0; index <= priv->child_count; index++) {
-		/* Send the lookup to all the nodes including namespace */
-		STACK_WIND_COOKIE (frame,
-				   unify_open_lookup_cbk,
-				   (void *)(long)index,
-				   priv->xl_array[index],
-				   priv->xl_array[index]->fops->lookup,
-				   &local->loc1,
-				   NULL);
-	}
-
-	return 0;
-}
-#endif /* GF_DARWIN_HOST_OS */
-
-/**
- * unify_open - 
- */
-int32_t
-unify_open (call_frame_t *frame,
-	    xlator_t *this,
-	    loc_t *loc,
-	    int32_t flags,
-	    fd_t *fd,
-            int32_t wbflags)
-{
-	unify_private_t *priv = this->private;
-	unify_local_t *local = NULL;
-	int16_t *list = NULL;
-	int16_t index = 0;
-	int16_t file_list[3] = {0,};
-	uint64_t tmp_list = 0;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	/* Init */
-	INIT_LOCAL (frame, local);
-	loc_copy (&local->loc1, loc);
-	local->fd    = fd;
-	local->flags = flags;
-        local->wbflags = wbflags;
-	inode_ctx_get (loc->inode, this, &tmp_list);
-	list = (int16_t *)(long)tmp_list;
-
-	local->list = list;
-	file_list[0] = priv->child_count; /* Thats namespace */
-	file_list[2] = -1;
-	for (index = 0; list[index] != -1; index++) {
-		local->call_count++;
-		if (list[index] != priv->child_count)
-			file_list[1] = list[index];
-	}
-
-	if (local->call_count != 2) {
-		/* If the lookup was done for file */
-		gf_log (this->name, GF_LOG_ERROR,
-			"%s: entry_count is %d",
-			loc->path, local->call_count);
-		for (index = 0; local->list[index] != -1; index++)
-			gf_log (this->name, GF_LOG_ERROR, "%s: found on %s",
-				loc->path, priv->xl_array[list[index]]->name);
-
-		if (local->call_count < 2) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"returning EIO as file found on onlyone node");
-			STACK_UNWIND (frame, -1, EIO, fd);
-			return 0;
-		}
-	}
-
-#ifdef GF_DARWIN_HOST_OS
-	/* Handle symlink here */
-	if (IA_ISLNK (loc->inode->ia_type)) {
-		/* Callcount doesn't matter here */
-		STACK_WIND (frame,
-			    unify_open_readlink_cbk,
-			    NS(this),
-			    NS(this)->fops->readlink,
-			    loc, ZR_PATH_MAX);
-		return 0;
-	}
-#endif /* GF_DARWIN_HOST_OS */
-
-	local->call_count = 2;
-	for (index = 0; file_list[index] != -1; index++) {
-		char need_break = (file_list[index+1] == -1);
-		STACK_WIND_COOKIE (frame,
-				   unify_open_cbk,
-				   priv->xl_array[file_list[index]], //cookie
-				   priv->xl_array[file_list[index]],
-				   priv->xl_array[file_list[index]]->fops->open,
-				   loc,
-				   flags,
-				   fd, wbflags);
-		if (need_break)
-			break;
-	}
-
-	return 0;
-}
-
-
-int32_t 
-unify_create_unlink_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno,
-                         struct iatt *preparent,
-                         struct iatt *postparent)
-{
-	unify_local_t *local = frame->local;
-	inode_t *inode = local->loc1.inode;
-
-	unify_local_wipe (local);
-
-	STACK_UNWIND (frame, local->op_ret, local->op_errno, local->fd, 
-		      inode, &local->stbuf,
-                      &local->oldpreparent, &local->oldpostparent);
-  
-	return 0;
-}
-
-/**
- * unify_create_open_cbk -
- */
-int32_t
-unify_create_open_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno,
-		       fd_t *fd)
-{
-	int ret = 0;
-	int32_t callcnt = 0;
-	unify_local_t *local = frame->local;
-	inode_t *inode = NULL;
-	xlator_t *child = NULL;
-	uint64_t tmp_value = 0;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret >= 0) {
-			local->op_ret = op_ret;
-			if (NS(this) != (xlator_t *)cookie) {
-				/* Store child node's ptr, used in all 
-				   the f*** / FileIO calls */
-				/* TODO: log on failure */
-				ret = fd_ctx_get (fd, this, &tmp_value);
-				cookie = (void *)(long)tmp_value;
-			} else {
-				/* NOTE: open successful on namespace.
-				 *       fd's ctx can be used to identify open 
-				 *       failure on storage subvolume. cool 
-				 *       ide ;) */
-				local->failed = 0;
-			}
-		} else {
-			gf_log (this->name, GF_LOG_ERROR,
-				"child(%s): path(%s): %s", 
-				((xlator_t *)cookie)->name,
-				local->loc1.path, strerror (op_errno));
-			local->op_errno = op_errno;
-			local->failed = 1;
-		}
-		callcnt = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
-  
-	if (!callcnt) {
-		if (local->failed == 1 && (local->op_ret >= 0)) {
-			local->call_count = 1;
-			/* return -1 to user */
-			local->op_ret = -1;
-			local->op_errno = EIO;
-			local->fd = fd;
-			local->call_count = 1;
-
-			if (!fd_ctx_get (local->fd, this, &tmp_value)) {
-				child = (xlator_t *)(long)tmp_value;
-
-				gf_log (this->name, GF_LOG_ERROR, 
-					"Create success on child node, "
-					"failed on namespace");
-
-				STACK_WIND (frame,
-					    unify_create_unlink_cbk,
-					    child,
-					    child->fops->unlink,
-					    &local->loc1);
-			} else {
-				gf_log (this->name, GF_LOG_ERROR, 
-					"Create success on namespace, "
-					"failed on child node");
-
-				STACK_WIND (frame,
-					    unify_create_unlink_cbk,
-					    NS(this),
-					    NS(this)->fops->unlink,
-					    &local->loc1);
-			}
-			return 0;
-		}
-		inode = local->loc1.inode;
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, fd,
-			      inode, &local->stbuf,
-                              &local->oldpreparent, &local->oldpostparent);
-	}
-	return 0;
-}
-
-/**
- * unify_create_lookup_cbk - 
- */
-int32_t 
-unify_create_lookup_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno,
-			 inode_t *inode,
-			 struct iatt *buf,
-			 dict_t *dict,
-                         struct iatt *postparent)
-{
-	int32_t callcnt = 0;
-	int16_t index = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"child(%s): path(%s): %s", 
-				priv->xl_array[(long)cookie]->name, 
-				local->loc1.path, strerror (op_errno));
-			local->op_errno = op_errno;
-			local->failed = 1;
-		}
-
-		if (op_ret >= 0) {
-			local->op_ret = op_ret; 
-			local->list[local->index++] = (int16_t)(long)cookie;
-			if (NS(this) == priv->xl_array[(long)cookie]) {
-				local->ia_ino = buf->ia_ino;
-			} else {
-				local->stbuf = *buf;
-			}
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		int16_t *list = local->list;
-		int16_t file_list[3] = {0,};
-		local->op_ret = -1;
-
-		local->list [local->index] = -1;
-		file_list[0] = list[0];
-		file_list[1] = list[1];
-		file_list[2] = -1;
-
-		local->stbuf.ia_ino = local->ia_ino;
-		/* TODO: log on failure */
-		inode_ctx_put (local->loc1.inode, this, 
-			       (uint64_t)(long)local->list);
-
-		if (local->index != 2) {
-			/* Lookup failed, can't do open */
-			gf_log (this->name, GF_LOG_ERROR,
-				"%s: present on %d nodes", 
-				local->loc1.path, local->index);
-			file_list[0] = priv->child_count;
-			for (index = 0; list[index] != -1; index++) {
-				gf_log (this->name, GF_LOG_ERROR, 
-					"%s: found on %s", local->loc1.path, 
-					priv->xl_array[list[index]]->name);
-				if (list[index] != priv->child_count)
-					file_list[1] = list[index];
-			}
-
-			if (local->index < 2) {
-				unify_local_wipe (local);
-				gf_log (this->name, GF_LOG_ERROR,
-					"returning EIO as file found on "
-					"only one node");
-				STACK_UNWIND (frame, -1, EIO, 
-					      local->fd, inode, NULL,
-                                              NULL, NULL);
-				return 0;
-			}
-		}
-		/* Everything is perfect :) */    
-		local->call_count = 2;
-    
-		for (index = 0; file_list[index] != -1; index++) {
-			char need_break = (file_list[index+1] == -1);
-			STACK_WIND_COOKIE (frame,
-					   unify_create_open_cbk,
-					   priv->xl_array[file_list[index]],
-					   priv->xl_array[file_list[index]],
-					   priv->xl_array[file_list[index]]->fops->open,
-					   &local->loc1,
-					   local->flags,
-					   local->fd, 0);
-			if (need_break)
-				break;
-		}
-	}
-
-	return 0;
-}
-
-
-/**
- * unify_create_cbk -
- */
-int32_t
-unify_create_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-		  fd_t *fd,
-		  inode_t *inode,
-		  struct iatt *buf,
-                  struct iatt *preparent,
-                  struct iatt *postparent)
-{
-	int ret = 0;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-	inode_t *tmp_inode = NULL;
-
-	if (op_ret == -1) {
-		/* send unlink () on Namespace */
-		local->op_errno = op_errno;
-		local->op_ret = -1;
-		local->call_count = 1;
-		gf_log (this->name, GF_LOG_ERROR,
-			"create failed on %s (file %s, error %s), "
-			"sending unlink to namespace", 
-			prev_frame->this->name, 
-			local->loc1.path, strerror (op_errno));
-
-		STACK_WIND (frame,
-			    unify_create_unlink_cbk,
-			    NS(this),
-			    NS(this)->fops->unlink,
-			    &local->loc1);
-
-		return 0;
-	}
-
-	if (op_ret >= 0) {
-		local->op_ret = op_ret;
-		local->stbuf = *buf;
-		/* Just inode number should be from NS node */
-		local->stbuf.ia_ino = local->ia_ino;
-
-		/* TODO: log on failure */
-		ret = fd_ctx_set (fd, this, (uint64_t)(long)prev_frame->this);
-	}
-  
-	tmp_inode = local->loc1.inode;
-	unify_local_wipe (local);
-	STACK_UNWIND (frame, local->op_ret, local->op_errno, local->fd, 
-		      tmp_inode, &local->stbuf,
-                      &local->oldpreparent, &local->oldpostparent);
-
-	return 0;
-}
-
-/**
- * unify_ns_create_cbk -
- * 
- */
-int32_t
-unify_ns_create_cbk (call_frame_t *frame,
-		     void *cookie,
-		     xlator_t *this,
-		     int32_t op_ret,
-		     int32_t op_errno,
-		     fd_t *fd,
-		     inode_t *inode,
-		     struct iatt *buf,
-                     struct iatt *preparent,
-                     struct iatt *postparent)
-{
-	struct sched_ops *sched_ops = NULL;
-	xlator_t *sched_xl = NULL;
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	int16_t *list = NULL;
-	int16_t index = 0;
-
-	if (op_ret == -1) {
-		/* No need to send create request to other servers, as 
-		   namespace action failed. Handle exclusive create here. */
-		if ((op_errno != EEXIST) || 
-		    ((op_errno == EEXIST) && 
-		     ((local->flags & O_EXCL) == O_EXCL))) {
-			/* If its just a create call without O_EXCL, 
-			   don't do this */
-			gf_log (this->name, GF_LOG_ERROR,
-				"namespace: path(%s): %s", 
-				local->loc1.path, strerror (op_errno));
-			unify_local_wipe (local);
-			STACK_UNWIND (frame, op_ret, op_errno, fd, inode, buf,
-                                      preparent, postparent);
-			return 0;
-		}
-	}
-  
-	if (op_ret >= 0) {
-		/* Get the inode number from the NS node */
-		local->ia_ino = buf->ia_ino;
-  
-                local->oldpreparent = *preparent;
-                local->oldpostparent = *postparent;
-
-		local->op_ret = -1;
-
-		/* Start the mapping list */
-		list = GF_CALLOC (1, sizeof (int16_t) * 3,
-                                  gf_unify_mt_int16_t);
-		ERR_ABORT (list);
-		inode_ctx_put (inode, this, (uint64_t)(long)list);
-		list[0] = priv->child_count;
-		list[2] = -1;
-
-		/* This means, file doesn't exist anywhere in the Filesystem */
-		sched_ops = priv->sched_ops;
-
-		/* Send create request to the scheduled node now */
-		sched_xl = sched_ops->schedule (this, local->loc1.path);
-		if (sched_xl == NULL)
-		{
-			/* send unlink () on Namespace */
-			local->op_errno = ENOTCONN;
-			local->op_ret = -1;
-			local->call_count = 1;
-			gf_log (this->name, GF_LOG_ERROR,
-				"no node online to schedule create:(file %s) "
-				"sending unlink to namespace", 
-				(local->loc1.path)?local->loc1.path:"");
-
-			STACK_WIND (frame,
-				    unify_create_unlink_cbk,
-				    NS(this),
-				    NS(this)->fops->unlink,
-				    &local->loc1);
-	
-			return 0;
-		}
-
-		for (index = 0; index < priv->child_count; index++)
-			if (sched_xl == priv->xl_array[index])
-				break;
-		list[1] = index;
-
-		STACK_WIND (frame, unify_create_cbk,
-			    sched_xl, sched_xl->fops->create,
-			    &local->loc1, local->flags, local->mode, fd);
-	} else {
-		/* File already exists, and there is no O_EXCL flag */
-
-		gf_log (this->name, GF_LOG_DEBUG, 
-			"File(%s) already exists on namespace, sending "
-			"open instead", local->loc1.path);
-
-		local->list = GF_CALLOC (1, sizeof (int16_t) * 3,
-                                         gf_unify_mt_int16_t);
-		ERR_ABORT (local->list);
-		local->call_count = priv->child_count + 1;
-		local->op_ret = -1;
-		for (index = 0; index <= priv->child_count; index++) {
-			/* Send lookup() to all nodes including namespace */
-			STACK_WIND_COOKIE (frame,
-					   unify_create_lookup_cbk,
-					   (void *)(long)index,
-					   priv->xl_array[index],
-					   priv->xl_array[index]->fops->lookup,
-					   &local->loc1,
-					   NULL);
-		}
-	}
-	return 0;
-}
-
-/**
- * unify_create - create a file in global namespace first, so other 
- *    clients can see them. Create the file in storage nodes in background.
- */
-int32_t
-unify_create (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *loc,
-	      int32_t flags,
-	      mode_t mode,
-	      fd_t *fd)
-{
-	unify_local_t *local = NULL;
-  
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	local->mode = mode;
-	local->flags = flags;
-	local->fd = fd;
-
-	loc_copy (&local->loc1, loc);
-	if (local->loc1.path == NULL) {
-		gf_log (this->name, GF_LOG_CRITICAL, "Not enough memory :O");
-		STACK_UNWIND (frame, -1, ENOMEM, fd, loc->inode, NULL,
-                              NULL, NULL);
-		return 0;
-	}
-
-	STACK_WIND (frame,
-		    unify_ns_create_cbk,
-		    NS(this),
-		    NS(this)->fops->create,
-		    loc,
-		    flags | O_EXCL,
-		    mode,
-		    fd);
-  
-	return 0;
-}
-
-
-/**
- * unify_opendir_cbk - 
- */
-int32_t
-unify_opendir_cbk (call_frame_t *frame,
-		   void *cookie,
-		   xlator_t *this,
-		   int32_t op_ret,
-		   int32_t op_errno,
-		   fd_t *fd)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, fd);
-
-	return 0;
-}
-
-/** 
- * unify_opendir -
- */
-int32_t
-unify_opendir (call_frame_t *frame,
-	       xlator_t *this,
-	       loc_t *loc,
-	       fd_t *fd)
-{
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	STACK_WIND (frame, unify_opendir_cbk,
-		    NS(this), NS(this)->fops->opendir, loc, fd);
-
-	return 0;
-}
-
-
-int32_t
-unify_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct iatt *statpre,
-                   struct iatt *statpost)
-{
-	int32_t callcnt = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-    
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"%s(): child(%s): path(%s): %s", 
-				gf_fop_list[frame->root->op],
-				prev_frame->this->name, 
-				(local->loc1.path)?local->loc1.path:"", 
-				strerror (op_errno));
-
-			local->op_errno = op_errno;
-			if ((op_errno == ENOENT) && priv->optimist) 
-				local->op_ret = 0;
-		}
-
-		if (op_ret >= 0) {
-			local->op_ret = 0;
-
-			if (NS (this) == prev_frame->this) {
-				local->ia_ino = statpost->ia_ino;
-				/* If the entry is directory, get the stat
-				   from NS node */
-				if (IA_ISDIR (statpost->ia_type) || 
-				    !local->stpost.ia_blksize) {
-					local->stpre = *statpre;
-                                        local->stpost = *statpost;
-				}
-			}
-
-			if ((!IA_ISDIR (statpost->ia_type)) && 
-			    (NS (this) != prev_frame->this)) {
-				/* If file, take the stat info from Storage 
-				   node. */
-				local->stpre = *statpre;
-                                local->stpost = *statpost;
-			}
-		}
-	}
-	UNLOCK (&frame->lock);
-    
-	if (!callcnt) {
-		/* If the inode number is not filled, operation should
-		   fail */
-		if (!local->ia_ino)
-			local->op_ret = -1;
-
-                local->stpre.ia_ino = local->ia_ino;
-		local->stpost.ia_ino = local->ia_ino;
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-			      &local->stpre, &local->stpost);
-	}
-
-	return 0;
-}
-
-
-int32_t
-unify_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-               struct iatt *stbuf, int32_t valid)
-{
-	unify_local_t *local = NULL;
-	unify_private_t *priv = this->private;
-	int32_t index = 0;
-	int32_t callcnt = 0;
-  	uint64_t tmp_list = 0;
-
-        if (!(loc && loc->inode)) {                     
-                STACK_UNWIND (frame, -1, EINVAL, NULL, NULL);
-                return 0;
-        }                                                    
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	loc_copy (&local->loc1, loc);
-
-	if (IA_ISDIR (loc->inode->ia_type)) {
-		local->call_count = 1;
-      
-                STACK_WIND (frame,
-                            unify_setattr_cbk,
-                            NS (this),
-                            NS (this)->fops->setattr,
-                            loc, stbuf, valid);
-        } else {
-		inode_ctx_get (loc->inode, this, &tmp_list);
-		local->list = (int16_t *)(long)tmp_list;
-
-		for (index = 0; local->list[index] != -1; index++) {
-			local->call_count++;
-			callcnt++;
-		}
-      
-		for (index = 0; local->list[index] != -1; index++) {
-			STACK_WIND (frame,
-				    unify_setattr_cbk,
-				    priv->xl_array[local->list[index]],
-				    priv->xl_array[local->list[index]]->fops->setattr,
-				    loc, stbuf, valid);
-
-			if (!--callcnt)
-				break;
-		}
-	}
-
-	return 0;
-}
-
-
-int32_t
-unify_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
-                struct iatt *stbuf, int32_t valid)
-{
-	unify_local_t *local = NULL;
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	UNIFY_CHECK_FD_AND_UNWIND_ON_ERR(fd);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-
-	if (!fd_ctx_get (fd, this, &tmp_child)) {
-		/* If its set, then its file */
-		child = (xlator_t *)(long)tmp_child;		     
-
-		local->call_count = 2;
-
-		STACK_WIND (frame, unify_setattr_cbk, child,
-			    child->fops->fsetattr, fd, stbuf, valid);
-
-		STACK_WIND (frame, unify_setattr_cbk, NS(this),
-			    NS(this)->fops->fsetattr, fd, stbuf, valid);
-	} else {
-		local->call_count = 1;
-    
-		STACK_WIND (frame, unify_setattr_cbk,
-			    NS(this), NS(this)->fops->fsetattr,
-			    fd, stbuf, valid);
-	}
-  
-	return 0;
-}
-
-
-/**
- * unify_truncate_cbk - 
- */
-int32_t
-unify_truncate_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    struct iatt *prebuf,
-                    struct iatt *postbuf)
-{
-	int32_t callcnt = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-    
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"child(%s): path(%s): %s", 
-				prev_frame->this->name, 
-				(local->loc1.path)?local->loc1.path:"", 
-				strerror (op_errno));
-			local->op_errno = op_errno;
-			if (!((op_errno == ENOENT) && priv->optimist))
-				local->op_ret = -1;
-		}
-
-		if (op_ret >= 0) {
-			if (NS (this) == prev_frame->this) {
-				local->ia_ino = postbuf->ia_ino;
-				/* If the entry is directory, get the 
-				   stat from NS node */
-				if (IA_ISDIR (postbuf->ia_type) ||
-				    !local->stbuf.ia_blksize) {
-					local->stbuf = *prebuf;
-                                        local->poststbuf = *postbuf;
-				}
-			}
-
-			if ((!IA_ISDIR (postbuf->ia_type)) &&
-			    (NS (this) != prev_frame->this)) {
-				/* If file, take the stat info from 
-				   Storage node. */
-				local->stbuf = *prebuf;
-                                local->poststbuf = *postbuf;
-			}
-		}
-	}
-	UNLOCK (&frame->lock);
-    
-	if (!callcnt) {
-		if (local->ia_ino) {
-			local->stbuf.ia_ino = local->ia_ino;
-                        local->poststbuf.ia_ino = local->ia_ino;
-                } else {
-			local->op_ret = -1;
-                }
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-			      &local->stbuf, &local->poststbuf);
-	}
-
-	return 0;
-}
-
-
-/**
- * unify_truncate - 
- */
-int32_t
-unify_truncate (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		off_t offset)
-{
-	unify_local_t *local = NULL;
-	unify_private_t *priv = this->private;
-	int32_t index = 0;
-	int32_t callcnt = 0;
-  	uint64_t tmp_list = 0;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	loc_copy (&local->loc1, loc);
-	local->ia_ino = loc->inode->ino;
-
-	if (IA_ISDIR (loc->inode->ia_type)) {
-		local->call_count = 1;
-      
-		STACK_WIND (frame,
-			    unify_truncate_cbk,
-			    NS(this),
-			    NS(this)->fops->truncate,
-			    loc,
-                            0);
-	} else {
-		local->op_ret = 0;
-		inode_ctx_get (loc->inode, this, &tmp_list);
-		local->list = (int16_t *)(long)tmp_list;
-
-		for (index = 0; local->list[index] != -1; index++) {
-			local->call_count++;
-			callcnt++;
-		}
-      
-		/* Don't send offset to NS truncate */
-		STACK_WIND (frame, unify_truncate_cbk, NS(this),
-			    NS(this)->fops->truncate, loc, 0);
-		callcnt--;
-
-		for (index = 0; local->list[index] != -1; index++) {
-			if (NS(this) != priv->xl_array[local->list[index]]) {
-				STACK_WIND (frame,
-					    unify_truncate_cbk,
-					    priv->xl_array[local->list[index]],
-					    priv->xl_array[local->list[index]]->fops->truncate,
-					    loc,
-					    offset);
-				if (!--callcnt)
-					break;
-			}
-		}
-	}
-
-	return 0;
-}
-
-/**
- * unify_readlink_cbk - 
- */
-int32_t
-unify_readlink_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    const char *path,
-                    struct iatt *sbuf)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, path, sbuf);
-	return 0;
-}
-
-/**
- * unify_readlink - Read the link only from the storage node.
- */
-int32_t
-unify_readlink (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		size_t size)
-{
-	unify_private_t *priv = this->private;
-	int32_t entry_count = 0;
-	int16_t *list = NULL;
-	int16_t index = 0;
-  	uint64_t tmp_list = 0;
-  
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-	
-	inode_ctx_get (loc->inode, this, &tmp_list);
-	list = (int16_t *)(long)tmp_list;
-
-	for (index = 0; list[index] != -1; index++)
-		entry_count++;
-
-	if (entry_count >= 2) {
-		for (index = 0; list[index] != -1; index++) {
-			if (priv->xl_array[list[index]] != NS(this)) {
-				STACK_WIND (frame,
-					    unify_readlink_cbk,
-					    priv->xl_array[list[index]],
-					    priv->xl_array[list[index]]->fops->readlink,
-					    loc,
-					    size);
-				break;
-			}
-		}
-	} else {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"returning ENOENT, no softlink files found "
-			"on storage node");
-		STACK_UNWIND (frame, -1, ENOENT, NULL);
-	}
-
-	return 0;
-}
-
-
-/**
- * unify_unlink_cbk - 
- */
-int32_t
-unify_unlink_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-                  struct iatt *preparent,
-                  struct iatt *postparent)
-{
-	int32_t callcnt = 0;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-		if (op_ret == 0  || ((op_errno == ENOENT) && priv->optimist))
-			local->op_ret = 0;
-		if (op_ret == -1)
-			local->op_errno = op_errno;
-
-                if (((call_frame_t *)cookie)->this == NS(this)) {
-                        local->oldpreparent = *preparent;
-                        local->oldpostparent = *postparent;
-                }
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno,
-                              &local->oldpreparent, &local->oldpostparent);
-	}
-
-	return 0;
-}
-
-
-/**
- * unify_unlink - 
- */
-int32_t
-unify_unlink (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *loc)
-{
-	unify_private_t *priv = this->private;
-	unify_local_t *local = NULL;
-	int16_t *list = NULL;
-	int16_t index = 0;
-  	uint64_t tmp_list = 0;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	loc_copy (&local->loc1, loc);
-
-	inode_ctx_get (loc->inode, this, &tmp_list);
-	list = (int16_t *)(long)tmp_list;
-
-	for (index = 0; list[index] != -1; index++)
-		local->call_count++;
-
-	if (local->call_count) {
-		for (index = 0; list[index] != -1; index++) {
-			char need_break = (list[index+1] == -1);
-			STACK_WIND (frame,
-				    unify_unlink_cbk,
-				    priv->xl_array[list[index]],
-				    priv->xl_array[list[index]]->fops->unlink,
-				    loc);
-			if (need_break)
-				break;
-		}
-	} else {
-		gf_log (this->name, GF_LOG_ERROR,
-			"%s: returning ENOENT", loc->path);
-		STACK_UNWIND (frame, -1, ENOENT, NULL, NULL);
-	}
-
-	return 0;
-}
-
-
-/**
- * unify_readv_cbk - 
- */
-int32_t
-unify_readv_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-		 struct iovec *vector,
-		 int32_t count,
-		 struct iatt *stbuf,
-                 struct iobref *iobref)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, vector, count, stbuf, iobref);
-	return 0;
-}
-
-/**
- * unify_readv - 
- */
-int32_t
-unify_readv (call_frame_t *frame,
-	     xlator_t *this,
-	     fd_t *fd,
-	     size_t size,
-	     off_t offset)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame,
-		    unify_readv_cbk,
-		    child,
-		    child->fops->readv,
-		    fd,
-		    size,
-		    offset);
-
-
-	return 0;
-}
-
-/**
- * unify_writev_cbk - 
- */
-int32_t
-unify_writev_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-                  struct iatt *prebuf,
-		  struct iatt *postbuf)
-{
-	unify_local_t *local = NULL;
-
-        local = frame->local;
-
-        local->stbuf = *prebuf;
-        local->stbuf.ia_ino = local->ia_ino;
-
-        local->poststbuf = *postbuf;
-        local->poststbuf.ia_ino = local->ia_ino;
-
-	STACK_UNWIND (frame, op_ret, op_errno,
-                      &local->stbuf, &local->poststbuf);
-	return 0;
-}
-
-/**
- * unify_writev - 
- */
-int32_t
-unify_writev (call_frame_t *frame,
-	      xlator_t *this,
-	      fd_t *fd,
-	      struct iovec *vector,
-	      int32_t count,
-	      off_t off,
-              struct iobref *iobref)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-	unify_local_t *local = NULL;
-
-	INIT_LOCAL (frame, local);
-        local->ia_ino = fd->inode->ino;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame,
-		    unify_writev_cbk,
-		    child,
-		    child->fops->writev,
-		    fd,
-		    vector,
-		    count,
-		    off,
-                    iobref);
-
-	return 0;
-}
-
-/**
- * unify_ftruncate -
- */
-int32_t
-unify_ftruncate (call_frame_t *frame,
-		 xlator_t *this,
-		 fd_t *fd,
-		 off_t offset)
-{
-	xlator_t *child = NULL;
-	unify_local_t *local = NULL;
-	uint64_t tmp_child = 0;
-
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR(fd);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	local->op_ret = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	local->call_count = 2;
-  
-	STACK_WIND (frame, unify_truncate_cbk,
-		    child, child->fops->ftruncate,
-		    fd, offset);
-  
-	STACK_WIND (frame, unify_truncate_cbk,
-		    NS(this), NS(this)->fops->ftruncate,
-		    fd, 0);
-  
-	return 0;
-}
-
-
-/**
- * unify_flush_cbk - 
- */
-int32_t
-unify_flush_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-/**
- * unify_flush -
- */
-int32_t
-unify_flush (call_frame_t *frame,
-	     xlator_t *this,
-	     fd_t *fd)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame, unify_flush_cbk, child, 
-		    child->fops->flush, fd);
-
-	return 0;
-}
-
-
-/**
- * unify_fsync_cbk - 
- */
-int32_t
-unify_fsync_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-                 struct iatt *prebuf,
-                 struct iatt *postbuf)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, prebuf, postbuf);
-	return 0;
-}
-
-/**
- * unify_fsync - 
- */
-int32_t
-unify_fsync (call_frame_t *frame,
-	     xlator_t *this,
-	     fd_t *fd,
-	     int32_t flags)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame, unify_fsync_cbk, child,
-		    child->fops->fsync, fd, flags);
-
-	return 0;
-}
-
-/**
- * unify_fstat - Send fstat FOP to Namespace only if its directory, and to 
- *     both namespace and the storage node if its a file.
- */
-int32_t
-unify_fstat (call_frame_t *frame,
-	     xlator_t *this,
-	     fd_t *fd)
-{
-	unify_local_t *local = NULL;
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	UNIFY_CHECK_FD_AND_UNWIND_ON_ERR(fd);
-
-	INIT_LOCAL (frame, local);
-	local->ia_ino = fd->inode->ino;
-
-	if (!fd_ctx_get (fd, this, &tmp_child)) {
-		/* If its set, then its file */
-		child = (xlator_t *)(long)tmp_child;		     
-		local->call_count = 2;
-
-		STACK_WIND (frame, unify_buf_cbk, child,
-			    child->fops->fstat, fd);
-
-		STACK_WIND (frame, unify_buf_cbk, NS(this),
-			    NS(this)->fops->fstat, fd);
-
-	} else {
-		/* this is an directory */
-		local->call_count = 1;
-		STACK_WIND (frame, unify_buf_cbk, NS(this),
-			    NS(this)->fops->fstat, fd);
-	}
-
-	return 0;
-}
-
-/**
- * unify_getdents_cbk - 
- */
-int32_t
-unify_getdents_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    dir_entry_t *entry,
-		    int32_t count)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, entry, count);
-	return 0;
-}
-
-/**
- * unify_getdents - send the FOP request to all the nodes.
- */
-int32_t
-unify_getdents (call_frame_t *frame,
-		xlator_t *this,
-		fd_t *fd,
-		size_t size,
-		off_t offset,
-		int32_t flag)
-{
-	UNIFY_CHECK_FD_AND_UNWIND_ON_ERR (fd);
-
-	STACK_WIND (frame, unify_getdents_cbk, NS(this),
-		    NS(this)->fops->getdents, fd, size, offset, flag);
-
-	return 0;
-}
-
-
-/**
- * unify_readdir_cbk - 
- */
-int32_t
-unify_readdir_cbk (call_frame_t *frame,
-		   void *cookie,
-		   xlator_t *this,
-		   int32_t op_ret,
-		   int32_t op_errno,
-		   gf_dirent_t *buf)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, buf);
-
-	return 0;
-}
-
-/**
- * unify_readdir - send the FOP request to all the nodes.
- */
-int32_t
-unify_readdir (call_frame_t *frame,
-	       xlator_t *this,
-	       fd_t *fd,
-	       size_t size,
-	       off_t offset)
-{
-	UNIFY_CHECK_FD_AND_UNWIND_ON_ERR (fd);
-
-	STACK_WIND (frame, unify_readdir_cbk, NS(this),
-		    NS(this)->fops->readdir, fd, size, offset);
-
-	return 0;
-}
-
-
-int32_t
-unify_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, gf_dirent_t *buf)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, buf);
-
-	return 0;
-}
-
-
-int32_t
-unify_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-                off_t offset)
-{
-	UNIFY_CHECK_FD_AND_UNWIND_ON_ERR (fd);
-
-	STACK_WIND (frame, unify_readdirp_cbk, NS(this),
-		    NS(this)->fops->readdirp, fd, size, offset);
-
-	return 0;
-}
-
-
-/**
- * unify_fsyncdir_cbk - 
- */
-int32_t
-unify_fsyncdir_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-
-	return 0;
-}
-
-/**
- * unify_fsyncdir -
- */
-int32_t
-unify_fsyncdir (call_frame_t *frame,
-		xlator_t *this,
-		fd_t *fd,
-		int32_t flags)
-{
-	UNIFY_CHECK_FD_AND_UNWIND_ON_ERR (fd);
-
-	STACK_WIND (frame, unify_fsyncdir_cbk,
-		    NS(this), NS(this)->fops->fsyncdir, fd, flags);
-
-	return 0;
-}
-
-/**
- * unify_lk_cbk - UNWIND frame with the proper return arguments.
- */
-int32_t
-unify_lk_cbk (call_frame_t *frame,
-	      void *cookie,
-	      xlator_t *this,
-	      int32_t op_ret,
-	      int32_t op_errno,
-	      struct gf_flock *lock)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, lock);
-	return 0;
-}
-
-/**
- * unify_lk - Send it to all the storage nodes, (should be 1) which has file.
- */
-int32_t
-unify_lk (call_frame_t *frame,
-	  xlator_t *this,
-	  fd_t *fd,
-	  int32_t cmd,
-	  struct gf_flock *lock)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame, unify_lk_cbk, child,
-		    child->fops->lk, fd, cmd, lock);
-
-	return 0;
-}
-
-
-int32_t
-unify_setxattr_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno);
-
-static int32_t
-unify_setxattr_file_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno)
-{
-	unify_private_t *private = this->private;
-	unify_local_t *local = frame->local;
-	xlator_t *sched_xl = NULL;
-	struct sched_ops *sched_ops = NULL;
-
-	if (op_ret == -1) {
-		if (!ENOTSUP)
-			gf_log (this->name, GF_LOG_ERROR,
-				"setxattr with XATTR_CREATE on ns: "
-				"path(%s) key(%s): %s",
-				local->loc1.path, local->name, 
-				strerror (op_errno));
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, op_ret, op_errno);
-		return 0;
-	} 
-
-	LOCK (&frame->lock);
-	{
-		local->failed = 0;
-		local->op_ret = 0;
-		local->op_errno = 0;
-		local->call_count = 1;
-	}
-	UNLOCK (&frame->lock);
-
-	/* schedule XATTR_CREATE on one of the child node */
-	sched_ops = private->sched_ops;
-    
-	/* Send create request to the scheduled node now */
-	sched_xl = sched_ops->schedule (this, local->name); 
-	if (!sched_xl) {
-		STACK_UNWIND (frame, -1, ENOTCONN);
-		return 0;
-	}
-
-	STACK_WIND (frame,
-		    unify_setxattr_cbk,
-		    sched_xl,
-		    sched_xl->fops->setxattr,
-		    &local->loc1,
-		    local->dict,
-		    local->flags);
-	return 0;
-}
-
-/**
- * unify_setxattr_cbk - When all the child nodes return, UNWIND frame.
- */
-int32_t
-unify_setxattr_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno)
-{
-	int32_t callcnt = 0;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-	dict_t *dict = NULL;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-    
-		if (op_ret == -1) {
-			gf_log (this->name, (((op_errno == ENOENT) || 
-					      (op_errno == ENOTSUP))? 
-					     GF_LOG_DEBUG : GF_LOG_ERROR), 
-				"child(%s): path(%s): %s", 
-				prev_frame->this->name, 
-				(local->loc1.path)?local->loc1.path:"", 
-				strerror (op_errno));
-			if (local->failed == -1) {
-				local->failed = 1;
-			}
-			local->op_errno = op_errno;
-		} else {
-			local->failed = 0;
-			local->op_ret = op_ret;
-		}
-	}
-	UNLOCK (&frame->lock);
-  
-	if (!callcnt) {
-		if (local->failed && local->name && 
-		    ZR_FILE_CONTENT_REQUEST(local->name)) {      
-			dict = get_new_dict ();
-			dict_set (dict, local->dict->members_list->key, 
-				  data_from_dynptr(NULL, 0));
-			dict_ref (dict);
-
-			local->call_count = 1;
-
-			STACK_WIND (frame,
-				    unify_setxattr_file_cbk,
-				    NS(this),
-				    NS(this)->fops->setxattr,
-				    &local->loc1,
-				    dict,
-				    XATTR_CREATE);
-
-			dict_unref (dict);
-			return 0;
-		}
-    
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno);
-	}
-
-	return 0;
-}
-
-/**
- * unify_sexattr - This function should be sent to all the storage nodes, 
- *       which contains the file, (excluding namespace).
- */
-int32_t
-unify_setxattr (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		dict_t *dict,
-		int32_t flags)
-{
-	unify_private_t *priv = this->private;
-	unify_local_t *local = NULL;
-	int16_t *list = NULL;
-	int16_t index = 0;
-	int32_t call_count = 0;
-  	uint64_t tmp_list = 0;
-	data_pair_t *trav = dict->members_list;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	local->failed = -1;
-	loc_copy (&local->loc1, loc);
-
-	if (IA_ISDIR (loc->inode->ia_type)) {
-
-		if (trav && trav->key && ZR_FILE_CONTENT_REQUEST(trav->key)) {
-			/* direct the storage xlators to change file 
-			   content only if file exists */
-			local->flags = flags;
-			local->dict = dict;
-			local->name = gf_strdup (trav->key);
-			flags |= XATTR_REPLACE;
-		}
-
-		local->call_count = priv->child_count;
-		for (index = 0; index < priv->child_count; index++) {
-			STACK_WIND (frame,
-				    unify_setxattr_cbk,
-				    priv->xl_array[index],
-				    priv->xl_array[index]->fops->setxattr,
-				    loc, dict, flags);
-		}
-		return 0;
-	}
-
-	inode_ctx_get (loc->inode, this, &tmp_list);
-	list = (int16_t *)(long)tmp_list;
-
-	for (index = 0; list[index] != -1; index++) {
-		if (NS(this) != priv->xl_array[list[index]]) {
-			local->call_count++;
-			call_count++;
-		}
-	}
-  
-	if (local->call_count) {
-		for (index = 0; list[index] != -1; index++) {
-			if (priv->xl_array[list[index]] != NS(this)) {
-				STACK_WIND (frame,
-					    unify_setxattr_cbk,
-					    priv->xl_array[list[index]],
-					    priv->xl_array[list[index]]->fops->setxattr,
-					    loc,
-					    dict,
-					    flags);
-				if (!--call_count)
-					break;
-			}
-		}
-		return 0;
-	}
-
-	/* No entry in storage nodes */
-	gf_log (this->name, GF_LOG_DEBUG, 
-		"returning ENOENT, file not found on storage node.");
-	STACK_UNWIND (frame, -1, ENOENT);
-
-	return 0;
-}
-
-
-/**
- * unify_getxattr_cbk - This function is called from only one child, so, no
- *     need of any lock or anything else, just send it to above layer 
- */
-int32_t
-unify_getxattr_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    dict_t *value)
-{
-	int32_t callcnt = 0;
-	dict_t *local_value = NULL;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-  
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-    
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, 
-				(((op_errno == ENOENT) || 
-				  (op_errno == ENODATA) || 
-				  (op_errno == ENOTSUP)) ? 
-				 GF_LOG_DEBUG : GF_LOG_ERROR), 
-				"child(%s): path(%s): %s", 
-				prev_frame->this->name, 
-				(local->loc1.path)?local->loc1.path:"", 
-				strerror (op_errno));
-		} else {
-			if (!local->dict)
-				local->dict = dict_ref (value);
-			local->op_ret = op_ret;
-		}
-	}
-	UNLOCK (&frame->lock);
-  
-	if (!callcnt) {
-		local_value = local->dict;
-		local->dict = NULL;
-      
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-			      local_value);
-      
-		if (local_value)
-			dict_unref (local_value);
-	} 
-
-	return 0;
-}
-
-
-/** 
- * unify_getxattr - This FOP is sent to only the storage node.
- */
-int32_t
-unify_getxattr (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		const char *name)
-{
-	unify_private_t *priv = this->private;
-	int16_t *list = NULL;
-	int16_t index = 0;
-	int16_t count = 0;
-	unify_local_t *local = NULL;
-  	uint64_t tmp_list = 0;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-	INIT_LOCAL (frame, local);
-
-	if (IA_ISDIR (loc->inode->ia_type)) {
-		local->call_count = priv->child_count;
-		for (index = 0; index < priv->child_count; index++)
-			STACK_WIND (frame,
-				    unify_getxattr_cbk,
-				    priv->xl_array[index],
-				    priv->xl_array[index]->fops->getxattr,
-				    loc,
-				    name);
-		return 0;
-	}
-
-	inode_ctx_get (loc->inode, this, &tmp_list);
-	list = (int16_t *)(long)tmp_list;
-
-	for (index = 0; list[index] != -1; index++) {
-		if (NS(this) != priv->xl_array[list[index]]) {
-			local->call_count++;
-			count++;
-		}
-	}
-
-	if (count) {
-		for (index = 0; list[index] != -1; index++) {
-			if (priv->xl_array[list[index]] != NS(this)) {
-				STACK_WIND (frame,
-					    unify_getxattr_cbk,
-					    priv->xl_array[list[index]],
-					    priv->xl_array[list[index]]->fops->getxattr,
-					    loc,
-					    name);
-				if (!--count)
-					break;
-			}
-		}
-	} else {
-		dict_t *tmp_dict = get_new_dict ();
-		gf_log (this->name, GF_LOG_DEBUG, 
-			"%s: returning ENODATA, no file found on storage node",
-			loc->path);
-		STACK_UNWIND (frame, -1, ENODATA, tmp_dict);
-		dict_destroy (tmp_dict);
-	}
-
-	return 0;
-}
-
-/**
- * unify_removexattr_cbk - Wait till all the child node returns the call
- *      and then UNWIND to above layer.
- */
-int32_t
-unify_removexattr_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno)
-{
-	int32_t callcnt = 0;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-
-	LOCK (&frame->lock);
-	{ 
-		callcnt = --local->call_count;
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			if (op_errno != ENOTSUP)
-				gf_log (this->name, GF_LOG_ERROR, 
-					"child(%s): path(%s): %s", 
-					prev_frame->this->name, 
-					local->loc1.path, strerror (op_errno));
-		} else {
-			local->op_ret = op_ret;
-		}
-	}
-	UNLOCK (&frame->lock);  
-
-	if (!callcnt) {
-		STACK_UNWIND (frame, local->op_ret, local->op_errno);
-	}
-
-	return 0;
-}
-
-/**
- * unify_removexattr - Send it to all the child nodes which has the files.
- */
-int32_t
-unify_removexattr (call_frame_t *frame,
-		   xlator_t *this,
-		   loc_t *loc,
-		   const char *name)
-{
-	unify_private_t *priv = this->private;
-	unify_local_t *local = NULL;
-	int16_t *list = NULL;
-	int16_t index = 0;
-	int32_t call_count = 0;
-  	uint64_t tmp_list = 0;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-
-	if (IA_ISDIR (loc->inode->ia_type)) {
-		local->call_count = priv->child_count;
-		for (index = 0; index < priv->child_count; index++)
-			STACK_WIND (frame, 		    
-				    unify_removexattr_cbk,
-				    priv->xl_array[index],
-				    priv->xl_array[index]->fops->removexattr,
-				    loc,
-				    name);
-
-		return 0;
-	}
-
-	inode_ctx_get (loc->inode, this, &tmp_list);
-	list = (int16_t *)(long)tmp_list;
-
-	for (index = 0; list[index] != -1; index++) {
-		if (NS(this) != priv->xl_array[list[index]]) {
-			local->call_count++;
-			call_count++;
-		}
-	}
-
-	if (local->call_count) {
-		for (index = 0; list[index] != -1; index++) {
-			if (priv->xl_array[list[index]] != NS(this)) {
-				STACK_WIND (frame,
-					    unify_removexattr_cbk,
-					    priv->xl_array[list[index]],
-					    priv->xl_array[list[index]]->fops->removexattr,
-					    loc,
-					    name);
-				if (!--call_count)
-					break;
-			}
-		}
-		return 0;
-	} 
-
-	gf_log (this->name, GF_LOG_DEBUG, 
-		"%s: returning ENOENT, not found on storage node.", loc->path);
-	STACK_UNWIND (frame, -1, ENOENT);
-
-	return 0;
-}
-
-
-int32_t 
-unify_mknod_unlink_cbk (call_frame_t *frame,
-			void *cookie,
-			xlator_t *this,
-			int32_t op_ret,
-			int32_t op_errno,
-                        struct iatt *preparent,
-                        struct iatt *postparent)
-{
-	unify_local_t *local = frame->local;
-
-	if (op_ret == -1)
-		gf_log (this->name, GF_LOG_ERROR, 
-			"%s: %s", local->loc1.path, strerror (op_errno));
-  
-	unify_local_wipe (local);
-	/* No log required here as this -1 is for mknod call */
-	STACK_UNWIND (frame, -1, local->op_errno, NULL, NULL);
-	return 0;
-}
-
-/**
- * unify_mknod_cbk - 
- */
-int32_t
-unify_mknod_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-		 inode_t *inode,
-                 struct iatt *buf,
-                 struct iatt *preparent,
-                 struct iatt *postparent)
-{
-	unify_local_t *local = frame->local;
-
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"mknod failed on storage node, sending unlink to "
-			"namespace");
-		local->op_errno = op_errno;
-		STACK_WIND (frame,
-			    unify_mknod_unlink_cbk,
-			    NS(this),
-			    NS(this)->fops->unlink,
-			    &local->loc1);
-		return 0;
-	}
-  
-	local->stbuf = *buf;
-	local->stbuf.ia_ino = local->ia_ino;
-	unify_local_wipe (local);
-	STACK_UNWIND (frame, op_ret, op_errno, inode, &local->stbuf,
-                      &local->oldpreparent, &local->oldpostparent);
-	return 0;
-}
-
-/**
- * unify_ns_mknod_cbk - 
- */
-int32_t
-unify_ns_mknod_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    inode_t *inode,
-                    struct iatt *buf,
-                    struct iatt *preparent,
-                    struct iatt *postparent)
-{
-	struct sched_ops *sched_ops = NULL;
-	xlator_t *sched_xl = NULL;
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	int16_t *list = NULL;
-	int16_t index = 0;
-	call_frame_t *prev_frame = cookie;
-
-	if (op_ret == -1) {
-		/* No need to send mknod request to other servers, 
-		 * as namespace action failed 
-		 */
-		gf_log (this->name, GF_LOG_ERROR, 
-			"child(%s): path(%s): %s", 
-			prev_frame->this->name, local->loc1.path, 
-			strerror (op_errno));
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, op_ret, op_errno, inode, buf,
-                              preparent, postparent);
-		return 0;
-	}
-  
-	/* Create one inode for this entry */
-	local->op_ret = 0;
-	local->stbuf = *buf;
-	local->ia_ino = buf->ia_ino;
-
-        local->oldpreparent = *preparent;
-        local->oldpostparent = *postparent;
-
-	list = GF_CALLOC (1, sizeof (int16_t) * 3, gf_unify_mt_int16_t);
-	ERR_ABORT (list);
-	list[0] = priv->child_count;
-	list[2] = -1;
-	inode_ctx_put (inode, this, (uint64_t)(long)list);
-
-	sched_ops = priv->sched_ops;
-
-	/* Send mknod request to scheduled node now */
-	sched_xl = sched_ops->schedule (this, local->loc1.path); 
-	if (!sched_xl) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"mknod failed on storage node, no node online "
-			"at the moment, sending unlink to NS");
-		local->op_errno = ENOTCONN;
-		STACK_WIND (frame,
-			    unify_mknod_unlink_cbk,
-			    NS(this),
-			    NS(this)->fops->unlink,
-			    &local->loc1);
-      
-		return 0;
-	}
-
-	for (index = 0; index < priv->child_count; index++)
-		if (sched_xl == priv->xl_array[index])
-			break;
-	list[1] = index;
-  
-	STACK_WIND (frame,  unify_mknod_cbk,
-		    sched_xl,  sched_xl->fops->mknod,
-		    &local->loc1, local->mode, local->dev);
-
-	return 0;
-}
-
-/**
- * unify_mknod - Create a device on namespace first, and later create on 
- *       the storage node.
- */
-int32_t
-unify_mknod (call_frame_t *frame,
-	     xlator_t *this,
-	     loc_t *loc,
-	     mode_t mode,
-	     dev_t rdev)
-{
-	unify_local_t *local = NULL;
-  
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	local->mode = mode;
-	local->dev = rdev;
-	loc_copy (&local->loc1, loc);
-	if (local->loc1.path == NULL) {
-		gf_log (this->name, GF_LOG_CRITICAL, "Not enough memory :O");
-		STACK_UNWIND (frame, -1, ENOMEM, loc->inode, NULL);
-		return 0;
-	}
-
-	STACK_WIND (frame,
-		    unify_ns_mknod_cbk,
-		    NS(this),
-		    NS(this)->fops->mknod,
-		    loc,
-		    mode,
-		    rdev);
-
-	return 0;
-}
-
-int32_t 
-unify_symlink_unlink_cbk (call_frame_t *frame,
-			  void *cookie,
-			  xlator_t *this,
-			  int32_t op_ret,
-			  int32_t op_errno,
-                          struct iatt *preparent,
-                          struct iatt *postparent)
-{
-	unify_local_t *local = frame->local;
-	if (op_ret == -1)
-		gf_log (this->name, GF_LOG_ERROR, 
-			"%s: %s", local->loc1.path, strerror (op_errno));
-
-	unify_local_wipe (local);
-	STACK_UNWIND (frame, -1, local->op_errno, NULL, NULL);
-	return 0;
-}
-
-/**
- * unify_symlink_cbk - 
- */
-int32_t
-unify_symlink_cbk (call_frame_t *frame,
-		   void *cookie,
-		   xlator_t *this,
-		   int32_t op_ret,
-		   int32_t op_errno,
-		   inode_t *inode,
-                   struct iatt *buf,
-                   struct iatt *preparent,
-                   struct iatt *postparent)
-{
-	unify_local_t *local = frame->local;
-
-	if (op_ret == -1) {
-		/* Symlink on storage node failed, hence send unlink 
-		   to the NS node */
-		local->op_errno = op_errno;
-		gf_log (this->name, GF_LOG_ERROR, 
-			"symlink on storage node failed, sending unlink "
-			"to namespace");
-
-		STACK_WIND (frame,
-			    unify_symlink_unlink_cbk,
-			    NS(this),
-			    NS(this)->fops->unlink,
-			    &local->loc1);
-    
-		return 0;
-	}
-  
-	local->stbuf = *buf;
-	local->stbuf.ia_ino = local->ia_ino;
-	unify_local_wipe (local);
-	STACK_UNWIND (frame, op_ret, op_errno, inode, &local->stbuf,
-                      &local->oldpreparent, &local->oldpostparent);
-
-	return 0;
-}
-
-/**
- * unify_ns_symlink_cbk - 
- */
-int32_t
-unify_ns_symlink_cbk (call_frame_t *frame,
-		      void *cookie,
-		      xlator_t *this,
-		      int32_t op_ret,
-		      int32_t op_errno,
-		      inode_t *inode,
-                      struct iatt *buf,
-                      struct iatt *preparent,
-                      struct iatt *postparent)
-{
-
-	struct sched_ops *sched_ops = NULL;
-	xlator_t *sched_xl = NULL;
-	int16_t *list = NULL;
-	unify_local_t *local = frame->local;
-	unify_private_t *priv = this->private;
-	int16_t index = 0;
-
-	if (op_ret == -1) {
-		/* No need to send symlink request to other servers, 
-		 * as namespace action failed 
-		 */
-		gf_log (this->name, GF_LOG_ERROR, 
-			"namespace: path(%s): %s", 
-			local->loc1.path, strerror (op_errno));
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, op_ret, op_errno, NULL, buf,
-                              preparent, postparent);
-		return 0;
-	}
-  
-	/* Create one inode for this entry */
-	local->op_ret = 0;
-	local->ia_ino = buf->ia_ino;
-  
-        local->oldpreparent = *preparent;
-        local->oldpostparent = *postparent;
-
-	/* Start the mapping list */
-
-	list = GF_CALLOC (1, sizeof (int16_t) * 3, gf_unify_mt_int16_t);
-	ERR_ABORT (list);
-	list[0] = priv->child_count; //namespace's index
-	list[2] = -1;
-	inode_ctx_put (inode, this, (uint64_t)(long)list);
-
-	sched_ops = priv->sched_ops;
-
-	/* Send symlink request to all the nodes now */
-	sched_xl = sched_ops->schedule (this, local->loc1.path); 
-	if (!sched_xl) {
-		/* Symlink on storage node failed, hence send unlink 
-		   to the NS node */
-		local->op_errno = ENOTCONN;
-		gf_log (this->name, GF_LOG_ERROR, 
-			"symlink on storage node failed, no node online, "
-			"sending unlink to namespace");
-      
-		STACK_WIND (frame,
-			    unify_symlink_unlink_cbk,
-			    NS(this),
-			    NS(this)->fops->unlink,
-			    &local->loc1);
-      
-		return 0;
-	}
-
-	for (index = 0; index < priv->child_count; index++)
-		if (sched_xl == priv->xl_array[index])
-			break;
-	list[1] = index;
-	
-	STACK_WIND (frame,
-		    unify_symlink_cbk,
-		    sched_xl,
-		    sched_xl->fops->symlink,
-		    local->name,
-		    &local->loc1);
-
-	return 0;
-}
-
-/**
- * unify_symlink - 
- */
-int32_t
-unify_symlink (call_frame_t *frame,
-	       xlator_t *this,
-	       const char *linkpath,
-	       loc_t *loc)
-{
-	unify_local_t *local = NULL;
-  
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	loc_copy (&local->loc1, loc);
-	local->name = gf_strdup (linkpath);
-
-	if ((local->name == NULL) || 
-	    (local->loc1.path == NULL)) {
-		gf_log (this->name, GF_LOG_CRITICAL, "Not enough memory :O");
-		STACK_UNWIND (frame, -1, ENOMEM, loc->inode, NULL);
-		return 0;
-	}
-
-	STACK_WIND (frame,
-		    unify_ns_symlink_cbk,
-		    NS(this),
-		    NS(this)->fops->symlink,
-		    linkpath,
-		    loc);
-
-	return 0;
-}
-
-
-int32_t 
-unify_rename_unlink_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno,
-                         struct iatt *preparent,
-                         struct iatt *postparent)
-{
-	int32_t callcnt = 0;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-  
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"child(%s): path(%s -> %s): %s", 
-			prev_frame->this->name, 
-			local->loc1.path, local->loc2.path, 
-			strerror (op_errno));
-      
-	}
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		local->stbuf.ia_ino = local->ia_ino;
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-			      &local->stbuf);
-	}
-	return 0;
-}
-
-int32_t 
-unify_ns_rename_undo_cbk (call_frame_t *frame,
-			  void *cookie,
-			  xlator_t *this,
-			  int32_t op_ret,
-			  int32_t op_errno,
-			  struct iatt *buf,
-                          struct iatt *preoldparent,
-                          struct iatt *postoldparent,
-                          struct iatt *prenewparent,
-                          struct iatt *postnewparent)
-{
-	unify_local_t *local = frame->local;
-
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"namespace: path(%s -> %s): %s", 
-			local->loc1.path, local->loc2.path, 
-			strerror (op_errno));
-	}
-
-	local->stbuf.ia_ino = local->ia_ino;
-	unify_local_wipe (local);
-	STACK_UNWIND (frame, local->op_ret, local->op_errno, &local->stbuf);
-	return 0;
-}
-
-int32_t 
-unify_rename_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-		  struct iatt *buf,
-                  struct iatt *preoldparent,
-                  struct iatt *postoldparent,
-                  struct iatt *prenewparent,
-                  struct iatt *postnewparent)
-{
-	int32_t index = 0;
-	int32_t callcnt = 0;
-	int16_t *list = NULL;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-  
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-		if (op_ret >= 0) {
-			if (!IA_ISDIR (buf->ia_type))
-				local->stbuf = *buf;
-			local->op_ret = op_ret;
-		} else {
-			gf_log (this->name, GF_LOG_ERROR, 
-				"child(%s): path(%s -> %s): %s", 
-				prev_frame->this->name, 
-				local->loc1.path, local->loc2.path, 
-				strerror (op_errno));
-			local->op_errno = op_errno;
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	if (!callcnt) {
-		local->stbuf.ia_ino = local->ia_ino;
-		if (IA_ISDIR (local->loc1.inode->ia_type)) {
-			unify_local_wipe (local);
-			STACK_UNWIND (frame, local->op_ret, local->op_errno,
-                                      &local->stbuf, &local->oldpreparent,
-                                      &local->oldpostparent, &local->newpreparent,
-                                      &local->newpostparent);
-			return 0;
-		}
-
-		if (local->op_ret == -1) {
-			/* TODO: check this logic */
-
-			/* Rename failed in storage node, successful on NS, 
-			 * hence, rename back the entries in NS */
-			/* NOTE: this will be done only if the destination 
-			 * doesn't exists, if  the destination exists, the 
-			 * job of correcting NS is left to self-heal
-			 */
-			if (!local->index) {
-				loc_t tmp_oldloc = {
-                                        /* its actual 'newloc->path' */
-					.path = local->loc2.path, 
-					.inode = local->loc1.inode,
-					.parent = local->loc2.parent
-				};
-	
-				loc_t tmp_newloc = {
-					/* Actual 'oldloc->path' */
-					.path = local->loc1.path,
-					.parent = local->loc1.parent
-				};
-
-				gf_log (this->name, GF_LOG_ERROR, 
-					"rename succussful on namespace, on "
-					"stroage node failed, reverting back");
-
-				STACK_WIND (frame,
-					    unify_ns_rename_undo_cbk,
-					    NS(this),
-					    NS(this)->fops->rename,
-					    &tmp_oldloc,
-					    &tmp_newloc);
-				return 0;
-			}
-		} else {
-			/* Rename successful on storage nodes */
-
-			int32_t idx = 0;
-			int16_t *tmp_list = NULL;
-			uint64_t tmp_list_int64 = 0;
-			if (local->loc2.inode) {
-				inode_ctx_get (local->loc2.inode, 
-					       this, &tmp_list_int64);
-				list = (int16_t *)(long)tmp_list_int64;
-
-			}
-
-			if (list) {				
-				for (index = 0; list[index] != -1; index++);
-				tmp_list = GF_CALLOC (1, index * 2, 
-                                                      gf_unify_mt_int16_t);
-				memcpy (tmp_list, list, index * 2);
-
-				for (index = 0; list[index] != -1; index++) {
-					/* TODO: Check this logic. */
-					/* If the destination file exists in 
-					 * the same storage node where we sent
-					 * 'rename' call, no need to send 
-					 * unlink 
-					 */
-					for (idx = 0; 
-					     local->list[idx] != -1; idx++) {
-						if (tmp_list[index] == local->list[idx]) {
-							tmp_list[index] = priv->child_count;
-							continue;
-						}
-					}
-	  
-					if (NS(this) != priv->xl_array[tmp_list[index]]) {
-						local->call_count++;
-						callcnt++;
-					}
-				}
-
-				if (local->call_count) {
-					if (callcnt > 1)
-						gf_log (this->name, 
-							GF_LOG_ERROR, 
-							"%s->%s: more (%d) "
-							"subvolumes have the "
-							"newloc entry", 
-							local->loc1.path, 
-							local->loc2.path, 
-							callcnt);
-
-					for (index=0; 
-					     tmp_list[index] != -1; index++) {
-						if (NS(this) != priv->xl_array[tmp_list[index]]) {		    
-							STACK_WIND (frame,
-								    unify_rename_unlink_cbk,
-								    priv->xl_array[tmp_list[index]],
-								    priv->xl_array[tmp_list[index]]->fops->unlink,
-								    &local->loc2);
-							if (!--callcnt)
-								break;
-						}
-					}
-
-					GF_FREE (tmp_list);
-					return 0;
-				}
-				if (tmp_list)
-					GF_FREE (tmp_list);
-			}
-		}
-    
-		/* Need not send 'unlink' to storage node */
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, 
-			      local->op_errno, &local->stbuf,
-                              &local->oldpreparent, &local->oldpostparent,
-                              &local->newpreparent, &local->newpostparent);
-	}
-
-	return 0;
-}
-
-int32_t 
-unify_ns_rename_cbk (call_frame_t *frame,
-		     void *cookie,
-		     xlator_t *this,
-		     int32_t op_ret,
-		     int32_t op_errno,
-		     struct iatt *buf,
-                     struct iatt *preoldparent,
-                     struct iatt *postoldparent,
-                     struct iatt *prenewparent,
-                     struct iatt *postnewparent)
-{
-	int32_t index = 0;
-	int32_t callcnt = 0;
-	int16_t *list = NULL;
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-
-	if (op_ret == -1) {
-		/* Free local->new_inode */
-		gf_log (this->name, GF_LOG_ERROR, 
-			"namespace: path(%s -> %s): %s", 
-			local->loc1.path, local->loc2.path, 
-			strerror (op_errno));
-
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, op_ret, op_errno, buf,
-                              preoldparent, postoldparent,
-                              prenewparent, postnewparent);
-		return 0;
-	}
-
-	local->stbuf = *buf;
-	local->ia_ino = buf->ia_ino;
-
-        local->oldpreparent = *preoldparent;
-        local->oldpostparent = *postoldparent;
-        local->newpreparent = *prenewparent;
-        local->newpostparent = *postnewparent;
-
-	/* Everything is fine. */
-	if (IA_ISDIR (buf->ia_type)) {
-		local->call_count = priv->child_count;
-		for (index=0; index < priv->child_count; index++) {
-			STACK_WIND (frame,
-				    unify_rename_cbk,
-				    priv->xl_array[index],
-				    priv->xl_array[index]->fops->rename,
-				    &local->loc1,
-				    &local->loc2);
-		}
-
-		return 0;
-	}
-
-	local->call_count = 0;  
-	/* send rename */
-	list = local->list;
-	for (index=0; list[index] != -1; index++) {
-		if (NS(this) != priv->xl_array[list[index]]) {
-			local->call_count++;
-			callcnt++;
-		}
-	}
-
-	if (local->call_count) {
-		for (index=0; list[index] != -1; index++) {
-			if (NS(this) != priv->xl_array[list[index]]) {
-				STACK_WIND (frame,
-					    unify_rename_cbk,
-					    priv->xl_array[list[index]],
-					    priv->xl_array[list[index]]->fops->rename,
-					    &local->loc1,
-					    &local->loc2);
-				if (!--callcnt)
-					break;
-			}
-		}
-	} else {
-		/* file doesn't seem to be present in storage nodes */
-		gf_log (this->name, GF_LOG_CRITICAL,
-			"CRITICAL: source file not in storage node, "
-			"rename successful on namespace :O");
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, -1, EIO, NULL,
-                              NULL, NULL, /* preoldparent, postoldparent */
-                              NULL, NULL); /* prenewparent, postnewparent */
-	}
-	return 0;
-}
-
-
-/**
- * unify_rename - One of the tricky function. The deadliest of all :O
- */
-int32_t
-unify_rename (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *oldloc,
-	      loc_t *newloc)
-{
-	unify_local_t *local = NULL;
-  	uint64_t tmp_list = 0;
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	loc_copy (&local->loc1, oldloc);
-	loc_copy (&local->loc2, newloc);
-
-	if ((local->loc1.path == NULL) || 
-	    (local->loc2.path == NULL)) {
-		gf_log (this->name, GF_LOG_CRITICAL, "Not enough memory :O");
-		STACK_UNWIND (frame, -1, ENOMEM, NULL,
-                              NULL, NULL, /* preoldparent, postoldparent */
-                              NULL, NULL); /* prenewparent, postnewparent */
-		return 0;
-	}
-  
-	inode_ctx_get (oldloc->inode, this, &tmp_list);
-	local->list = (int16_t *)(long)tmp_list;
-
-	STACK_WIND (frame,
-		    unify_ns_rename_cbk,
-		    NS(this),
-		    NS(this)->fops->rename,
-		    oldloc,
-		    newloc);
-	return 0;
-}
-
-/**
- * unify_link_cbk -
- */
-int32_t
-unify_link_cbk (call_frame_t *frame,
-		void *cookie,
-		xlator_t *this,
-		int32_t op_ret,
-		int32_t op_errno,
-		inode_t *inode,
-                struct iatt *buf,
-                struct iatt *preparent,
-                struct iatt *postparent)
-{
-	unify_local_t *local = frame->local;
-
-	if (op_ret >= 0) 
-		local->stbuf = *buf;
-	local->stbuf.ia_ino = local->ia_ino;
-
-	unify_local_wipe (local);
-	STACK_UNWIND (frame, op_ret, op_errno, inode, &local->stbuf,
-                      &local->oldpreparent, &local->oldpostparent);
-
-	return 0;
-}
-
-/**
- * unify_ns_link_cbk - 
- */
-int32_t
-unify_ns_link_cbk (call_frame_t *frame,
-		   void *cookie,
-		   xlator_t *this,
-		   int32_t op_ret,
-		   int32_t op_errno,
-		   inode_t *inode,
-                   struct iatt *buf,
-                   struct iatt *preparent,
-                   struct iatt *postparent)
-{
-	unify_private_t *priv = this->private;
-	unify_local_t *local = frame->local;
-	int16_t *list = local->list;
-	int16_t index = 0;
-
-	if (op_ret == -1) {
-		/* No need to send link request to other servers, 
-		 * as namespace action failed 
-		 */
-		gf_log (this->name, GF_LOG_ERROR, 
-			"namespace: path(%s -> %s): %s", 
-			local->loc1.path, local->loc2.path, 
-			strerror (op_errno));
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, op_ret, op_errno, inode, buf,
-                              preparent, postparent);
-		return 0;
-	}
-
-	/* Update inode for this entry */
-	local->op_ret = 0;
-	local->ia_ino = buf->ia_ino;
-
-        local->oldpreparent = *preparent;
-        local->oldpostparent = *postparent;
-
-	/* Send link request to the node now */
-	for (index = 0; list[index] != -1; index++) {
-		char need_break = (list[index+1] == -1);
-		if (priv->xl_array[list[index]] != NS (this)) {
-			STACK_WIND (frame,
-				    unify_link_cbk,
-				    priv->xl_array[list[index]],
-				    priv->xl_array[list[index]]->fops->link,
-				    &local->loc1,
-				    &local->loc2);
-                        break;
-		}
-		if (need_break)
-			break;
-	}
-
-	return 0;
-}
-
-/**
- * unify_link - 
- */
-int32_t
-unify_link (call_frame_t *frame,
-	    xlator_t *this,
-	    loc_t *oldloc,
-	    loc_t *newloc)
-{
-	unify_local_t *local = NULL;
-  	uint64_t tmp_list = 0;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (oldloc);
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (newloc);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-
-	loc_copy (&local->loc1, oldloc);
-	loc_copy (&local->loc2, newloc);
-
-	inode_ctx_get (oldloc->inode, this, &tmp_list);
-	local->list = (int16_t *)(long)tmp_list;
-
-	STACK_WIND (frame,
-		    unify_ns_link_cbk,
-		    NS(this),
-		    NS(this)->fops->link,
-		    oldloc,
-		    newloc);
-
-	return 0;
-}
-
-
-/**
- * unify_checksum_cbk - 
- */
-int32_t
-unify_checksum_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    uint8_t *fchecksum,
-		    uint8_t *dchecksum)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, fchecksum, dchecksum);
-
-	return 0;
-}
-
-/**
- * unify_checksum - 
- */
-int32_t
-unify_checksum (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		int32_t flag)
-{
-	STACK_WIND (frame,
-		    unify_checksum_cbk,
-		    NS(this),
-		    NS(this)->fops->checksum,
-		    loc,
-		    flag);
-
-	return 0;
-}
-
-
-/**
- * unify_finodelk_cbk - 
- */
-int
-unify_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-/**
- * unify_finodelk
- */
-int
-unify_finodelk (call_frame_t *frame, xlator_t *this,
-		const char *volume, fd_t *fd, int cmd, struct gf_flock *flock)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame, unify_finodelk_cbk,
-		    child, child->fops->finodelk,
-		    volume, fd, cmd, flock);
-
-	return 0;
-}
-
-
-
-/**
- * unify_fentrylk_cbk - 
- */
-int
-unify_fentrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-/**
- * unify_fentrylk
- */
-int
-unify_fentrylk (call_frame_t *frame, xlator_t *this,
-		const char *volume, fd_t *fd, const char *basename,
-		entrylk_cmd cmd, entrylk_type type)
-		
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame, unify_fentrylk_cbk,
-		    child, child->fops->fentrylk,
-		    volume, fd, basename, cmd, type);
-
-	return 0;
-}
-
-
-
-/**
- * unify_fxattrop_cbk - 
- */
-int
-unify_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno, dict_t *xattr)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, xattr);
-	return 0;
-}
-
-/**
- * unify_fxattrop
- */
-int
-unify_fxattrop (call_frame_t *frame, xlator_t *this,
-		fd_t *fd, gf_xattrop_flags_t optype, dict_t *xattr)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame, unify_fxattrop_cbk,
-		    child, child->fops->fxattrop,
-		    fd, optype, xattr);
-
-	return 0;
-}
-
-
-/**
- * unify_inodelk_cbk - 
- */
-int
-unify_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		   int32_t op_ret, int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-
-/**
- * unify_inodelk
- */
-int
-unify_inodelk (call_frame_t *frame, xlator_t *this,
-	       const char *volume, loc_t *loc, int cmd, struct gf_flock *flock)
-{
-	xlator_t *child = NULL;
-
-	child = unify_loc_subvol (loc, this);
-
-	STACK_WIND (frame, unify_inodelk_cbk,
-		    child, child->fops->inodelk,
-		    volume, loc, cmd, flock);
-
-	return 0;
-}
-
-
-
-/**
- * unify_entrylk_cbk - 
- */
-int
-unify_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		   int32_t op_ret, int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-/**
- * unify_entrylk
- */
-int
-unify_entrylk (call_frame_t *frame, xlator_t *this,
-	       const char *volume, loc_t *loc, const char *basename,
-	       entrylk_cmd cmd, entrylk_type type)
-		
-{
-	xlator_t *child = NULL;
-
-	child = unify_loc_subvol (loc, this);
-
-	STACK_WIND (frame, unify_entrylk_cbk,
-		    child, child->fops->entrylk,
-		    volume, loc, basename, cmd, type);
-
-	return 0;
-}
-
-
-
-/**
- * unify_xattrop_cbk - 
- */
-int
-unify_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		   int32_t op_ret, int32_t op_errno, dict_t *xattr)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, xattr);
-	return 0;
-}
-
-/**
- * unify_xattrop
- */
-int
-unify_xattrop (call_frame_t *frame, xlator_t *this,
-		loc_t *loc, gf_xattrop_flags_t optype, dict_t *xattr)
-{
-	xlator_t *child = NULL;
-
-	child = unify_loc_subvol (loc, this);
-
-	STACK_WIND (frame, unify_xattrop_cbk,
-		    child, child->fops->xattrop,
-		    loc, optype, xattr);
-
-	return 0;
-}
-
-int
-unify_forget (xlator_t *this,
-              inode_t *inode)
-{
-        int16_t *list = NULL;
-        uint64_t tmp_list = 0;
-
-        if (inode->ia_type && (!IA_ISDIR(inode->ia_type))) {
-                inode_ctx_get (inode, this, &tmp_list);
-                if (tmp_list) {
-                        list = (int16_t *)(long)tmp_list;
-                        GF_FREE (list);
-                }
-        }
-
-        return 0;
-}
-
-/**
- * notify
- */
-int32_t
-notify (xlator_t *this,
-        int32_t event,
-        void *data,
-        ...)
-{
-	unify_private_t *priv = this->private;
-	struct sched_ops *sched = NULL;
-
-	if (!priv) {
-		return 0;
-	}
-
-	sched = priv->sched_ops;    
-	if (!sched) {
-		gf_log (this->name, GF_LOG_CRITICAL, "No scheduler :O");
-		raise (SIGTERM);
-		return 0;
-	}
-	if (priv->namespace == data) {
-		if (event == GF_EVENT_CHILD_UP) {
-			sched->notify (this, event, data);
-		}
-		return 0;
-	}
-
-	switch (event)
-	{
-	case GF_EVENT_CHILD_UP:
-	{
-		/* Call scheduler's update () to enable it for scheduling */
-		sched->notify (this, event, data);
-	
-		LOCK (&priv->lock);
-		{
-			/* Increment the inode's generation, which is 
-			   used for self_heal */
-			++priv->inode_generation;
-			++priv->num_child_up;
-		}
-		UNLOCK (&priv->lock);
-
-		if (!priv->is_up) {
-			default_notify (this, event, data);
-			priv->is_up = 1;
-		}
-	}
-	break;
-	case GF_EVENT_CHILD_DOWN:
-	{
-		/* Call scheduler's update () to disable the child node 
-		 * for scheduling
-		 */
-		sched->notify (this, event, data);
-		LOCK (&priv->lock);
-		{
-			--priv->num_child_up;
-		}
-		UNLOCK (&priv->lock);
-
-		if (priv->num_child_up == 0) {
-			/* Send CHILD_DOWN to upper layer */
-			default_notify (this, event, data);
-			priv->is_up = 0;
-		}
-	}
-	break;
-
-	default:
-	{
-		default_notify (this, event, data);
-	}
-	break;
-	}
-
-	return 0;
-}
-
-int32_t
-mem_acct_init (xlator_t *this)
-{
-        int     ret = -1;
-
-        if (!this)
-                return ret;
-
-        ret = xlator_mem_acct_init (this, gf_unify_mt_end + 1);
-        
-        if (ret != 0) {
-                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
-                                "failed");
-                return ret;
-        }
-
-        return ret;
-}
-
-/** 
- * init - This function is called first in the xlator, while initializing.
- *   All the config file options are checked and appropriate flags are set.
- *
- * @this - 
- */
-int32_t 
-init (xlator_t *this)
-{
-	int32_t          ret       = 0;
-	int32_t          count     = 0;
-	data_t          *scheduler = NULL;
-	data_t          *data      = NULL;
-	xlator_t        *ns_xl     = NULL;
-	xlator_list_t   *trav      = NULL;
-	xlator_list_t   *xlparent  = NULL;
-	xlator_list_t   *parent    = NULL;
-	unify_private_t *_private  = NULL; 
-
-
-	/* Check for number of child nodes, if there is no child nodes, exit */
-	if (!this->children) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"No child nodes specified. check \"subvolumes \" "
-			"option in volfile");
-		return -1;
-	}
-
-  	if (!this->parents) {
-		gf_log (this->name, GF_LOG_WARNING,
-			"dangling volume. check volfile ");
-	}
-  
-	/* Check for 'scheduler' in volume */
-	scheduler = dict_get (this->options, "scheduler");
-	if (!scheduler) {
-		gf_log (this->name, GF_LOG_ERROR, 
-			"\"option scheduler <x>\" is missing in volfile");
-		return -1;
-	}
-
-	/* Setting "option namespace <node>" */
-	data = dict_get (this->options, "namespace");
-	if(!data) {
-		gf_log (this->name, GF_LOG_CRITICAL, 
-			"namespace option not specified, Exiting");
-		return -1;
-	}
-	/* Search namespace in the child node, if found, exit */
-	trav = this->children;
-	while (trav) {
-		if (strcmp (trav->xlator->name, data->data) == 0)
-			break;
-		trav = trav->next;
-	}
-	if (trav) {
-		gf_log (this->name, GF_LOG_CRITICAL, 
-			"namespace node used as a subvolume, Exiting");
-		return -1;
-	}
-	
-	/* Search for the namespace node, if found, continue */
-	ns_xl = this->next;
-	while (ns_xl) {
-		if (strcmp (ns_xl->name, data->data) == 0)
-			break;
-		ns_xl = ns_xl->next;
-	}
-	if (!ns_xl) {
-		gf_log (this->name, GF_LOG_CRITICAL, 
-			"namespace node not found in volfile, Exiting");
-		return -1;
-	}
-	
-	gf_log (this->name, GF_LOG_DEBUG, 
-		"namespace node specified as %s", data->data);
-	
-	_private = GF_CALLOC (1, sizeof (*_private), 
-                              gf_unify_mt_unify_private_t);
-	ERR_ABORT (_private);
-	_private->sched_ops = get_scheduler (this, scheduler->data);
-	if (!_private->sched_ops) {
-		gf_log (this->name, GF_LOG_CRITICAL, 
-			"Error while loading scheduler. Exiting");
-		GF_FREE (_private);
-		return -1;
-	}
-	
-	if (ns_xl->parents) {
-		gf_log (this->name, GF_LOG_CRITICAL,
-			"Namespace node should not be a child of any other node. Exiting");
-		GF_FREE (_private);
-		return -1;
-	}
-
-	_private->namespace = ns_xl;
-	
-	/* update _private structure */
-	{
-		count = 0;
-		trav = this->children;
-		/* Get the number of child count */
-		while (trav) {
-			count++;
-			trav = trav->next;
-		}
-		
-		gf_log (this->name, GF_LOG_DEBUG, 
-			"Child node count is %d", count);    
-
-		_private->child_count = count;
-		if (count == 1) {
-			/* TODO: Should I error out here? */
-			gf_log (this->name, GF_LOG_CRITICAL, 
-				"WARNING: You have defined only one "
-				"\"subvolumes\" for unify volume. It may not "
-				"be the desired config, review your volume "
-				"volfile. If this is how you are testing it,"
-				" you may hit some performance penalty");
-		}
-		
-		_private->xl_array = GF_CALLOC (1, 
-					        sizeof (xlator_t) * (count + 1),
-                                                gf_unify_mt_xlator_t);
-		ERR_ABORT (_private->xl_array);
-		
-		count = 0;
-		trav = this->children;
-		while (trav) {
-			_private->xl_array[count++] = trav->xlator;
-			trav = trav->next;
-		}
-		_private->xl_array[count] = _private->namespace;
-		
-		/* self-heal part, start with generation '1' */
-		_private->inode_generation = 1; 
-                /* Because, Foreground part is tested well */
-		_private->self_heal = ZR_UNIFY_FG_SELF_HEAL; 
-		data = dict_get (this->options, "self-heal");
-		if (data) {
-			if (strcasecmp (data->data, "off") == 0) 
-				_private->self_heal = ZR_UNIFY_SELF_HEAL_OFF;
-
-			if (strcasecmp (data->data, "foreground") == 0)
-				_private->self_heal = ZR_UNIFY_FG_SELF_HEAL;
-
-			if (strcasecmp (data->data, "background") == 0)
-				_private->self_heal = ZR_UNIFY_BG_SELF_HEAL;
-		}
-    
-		/* optimist - ask bulde for more about it */
-		data = dict_get (this->options, "optimist");
-		if (data) {
-			if (gf_string2boolean (data->data, 
-					       &_private->optimist) == -1) {
-				gf_log (this->name, GF_LOG_ERROR, 
-					"optimist excepts only boolean "
-					"options");
-			}
-		}
-
-		LOCK_INIT (&_private->lock);
-	}
-
-	/* Now that everything is fine. */
-	this->private = (void *)_private;
-	{
-                ret = _private->sched_ops->mem_acct_init (this);
-
-                if (ret == -1) {
-                        return -1;
-                }
-
-		/* Initialize scheduler, if everything else is successful */
-		ret = _private->sched_ops->init (this); 
-		if (ret == -1) {
-			gf_log (this->name, GF_LOG_CRITICAL,
-				"Initializing scheduler failed, Exiting");
-			GF_FREE (_private);
-			return -1;
-		}
-
-
-		ret = 0;
-
-		/* This section is required because some fops may look 
-		 * for 'xl->parent' variable 
-		 */
-		xlparent = GF_CALLOC (1, sizeof (*xlparent),
-                                      gf_unify_mt_xlator_list_t);
-		xlparent->xlator = this;
-		if (!ns_xl->parents) {
-			ns_xl->parents = xlparent;
-		} else {
-			parent = ns_xl->parents;
-			while (parent->next)
-				parent = parent->next;
-			parent->next = xlparent;
-		}
-	}
-
-	/* Tell namespace node that init is done */
-	xlator_notify (ns_xl, GF_EVENT_PARENT_UP, this);
-
-	return 0;
-}
-
-/** 
- * fini  - Free all the allocated memory 
- */
-void
-fini (xlator_t *this)
-{
-	unify_private_t *priv = this->private;
-	priv->sched_ops->fini (this);
-	this->private = NULL;
-	LOCK_DESTROY (&priv->lock);
-	GF_FREE (priv->xl_array);
-	GF_FREE (priv);
-	return;
-}
-
-
-struct xlator_fops fops = {
-	.stat        = unify_stat,
-	.readlink    = unify_readlink,
-	.mknod       = unify_mknod,
-	.mkdir       = unify_mkdir,
-	.unlink      = unify_unlink,
-	.rmdir       = unify_rmdir,
-	.symlink     = unify_symlink,
-	.rename      = unify_rename,
-	.link        = unify_link,
-	.truncate    = unify_truncate,
-	.create      = unify_create,
-	.open        = unify_open,
-	.readv       = unify_readv,
-	.writev      = unify_writev,
-	.statfs      = unify_statfs,
-	.flush       = unify_flush,
-	.fsync       = unify_fsync,
-	.setxattr    = unify_setxattr,
-	.getxattr    = unify_getxattr,
-	.removexattr = unify_removexattr,
-	.opendir     = unify_opendir,
-	.readdir     = unify_readdir,
-	.readdirp    = unify_readdirp,
-	.fsyncdir    = unify_fsyncdir,
-	.access      = unify_access,
-	.ftruncate   = unify_ftruncate,
-	.fstat       = unify_fstat,
-	.lk          = unify_lk,
-	.lookup      = unify_lookup,
-	.getdents    = unify_getdents,
-	.checksum    = unify_checksum,
-	.inodelk     = unify_inodelk,
-	.finodelk    = unify_finodelk,
-	.entrylk     = unify_entrylk,
-	.fentrylk    = unify_fentrylk,
-	.xattrop     = unify_xattrop,
-	.fxattrop    = unify_fxattrop,
-        .setattr     = unify_setattr, 
-        .fsetattr    = unify_fsetattr,
-};
-
-
-struct xlator_cbks cbks = {
-        .forget  = unify_forget,
-};
-
-struct volume_options options[] = {
-	{ .key   = { "namespace" },  
-	  .type  = GF_OPTION_TYPE_XLATOR 
-	},
-	{ .key   = { "scheduler" },  
-	  .value = { "alu", "rr", "random", "nufa", "switch" },
-	  .type  = GF_OPTION_TYPE_STR
-	},
-	{ .key   = {"self-heal"},  
-	  .value = { "foreground", "background", "off" },
-	  .type  = GF_OPTION_TYPE_STR
-	},
-	/* TODO: remove it some time later */
-	{ .key   = {"optimist"},  
-	  .type  = GF_OPTION_TYPE_BOOL 
-	},
-
-	{ .key   = {NULL} },
-};
diff --git a/xlators/cluster/unify/src/unify.h b/xlators/cluster/unify/src/unify.h
deleted file mode 100644
index 3cfe725f4..000000000
--- a/xlators/cluster/unify/src/unify.h
+++ /dev/null
@@ -1,146 +0,0 @@
-/*
-  Copyright (c) 2006-2010 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU Affero General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  Affero General Public License for more details.
-
-  You should have received a copy of the GNU Affero General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#ifndef _UNIFY_H
-#define _UNIFY_H
-
-#include "scheduler.h"
-#include "list.h"
-#include "unify-mem-types.h"
-
-#define MAX_DIR_ENTRY_STRING     (32 * 1024)
-
-#define ZR_UNIFY_SELF_HEAL_OFF 0
-#define ZR_UNIFY_FG_SELF_HEAL  1
-#define ZR_UNIFY_BG_SELF_HEAL  2
-
-/* Sometimes one should use completely random numbers.. its good :p */
-#define UNIFY_SELF_HEAL_GETDENTS_COUNT 512
-
-#define NS(xl)          (((unify_private_t *)xl->private)->namespace)
-
-/* This is used to allocate memory for local structure */
-#define INIT_LOCAL(fr, loc)                   \
-do {                                          \
-  loc = GF_CALLOC (1, sizeof (unify_local_t), gf_unify_mt_unify_local_t);   \
-  ERR_ABORT (loc);			      \
-  if (!loc) {                                 \
-    STACK_UNWIND (fr, -1, ENOMEM);            \
-    return 0;                                 \
-  }                                           \
-  fr->local = loc;                            \
-  loc->op_ret = -1;                           \
-  loc->op_errno = ENOENT;                     \
-} while (0)
-
-
-
-struct unify_private {
-	/* Update this structure depending on requirement */
-	void *scheduler;               /* THIS SHOULD BE THE FIRST VARIABLE, 
-					  if xlator is using scheduler */
-	struct sched_ops *sched_ops;   /* Scheduler options  */
-	xlator_t *namespace;           /* ptr to namespace xlator */
-	xlator_t **xl_array;
-	gf_boolean_t optimist;
-	int16_t child_count;
-	int16_t num_child_up;
-	uint8_t self_heal;
-	uint8_t is_up;
-	uint64_t inode_generation;
-	gf_lock_t lock;
-};
-typedef struct unify_private unify_private_t;
-
-struct unify_self_heal_struct {
-	uint8_t dir_checksum[NAME_MAX];
-	uint8_t ns_dir_checksum[NAME_MAX];
-	uint8_t file_checksum[NAME_MAX];
-	uint8_t ns_file_checksum[NAME_MAX];
-	off_t *offset_list;
-	int   *count_list;
-	dir_entry_t **entry_list;
-};
-
-
-struct _unify_local_t {
-	int32_t call_count;
-	int32_t op_ret;
-	int32_t op_errno;
-	mode_t mode;
-	off_t offset;
-	dev_t dev;
-	uid_t uid;
-	gid_t gid;
-	int32_t flags;
-	int32_t entry_count;
-	int32_t count;    // dir_entry_t count;
-	fd_t *fd;
-	struct iatt stbuf;
-        struct iatt stpre;
-        struct iatt stpost;
-	struct statvfs statvfs_buf;
-	struct timespec tv[2];
-	char *name;
-	int32_t revalidate;
-
-	ino_t ia_ino;
-	nlink_t ia_nlink;
-  
-	dict_t *dict;
-
-	int16_t *list;
-	int16_t *new_list; /* Used only in case of rename */
-	int16_t index;
-
-	int32_t failed;
-	int32_t return_eio;  /* Used in case of different st-mode 
-				present for a given path */
-
-	uint64_t inode_generation; /* used to store the per directory 
-				    * inode_generation. Got from inode's ctx 
-				    * of directory inodes
-				    */
-
-	struct unify_self_heal_struct *sh_struct;
-	loc_t loc1, loc2;
-
-        struct iatt poststbuf;
-        /* When not used for rename, old*
-         * are used as the attrs for the current
-         * parent directory.
-         */
-        struct iatt oldpreparent;
-        struct iatt oldpostparent;
-        struct iatt newpreparent;
-        struct iatt newpostparent;
-        int32_t wbflags;
-};
-typedef struct _unify_local_t unify_local_t;
-
-int32_t zr_unify_self_heal (call_frame_t *frame,
-			    xlator_t *this,
-			    unify_local_t *local);
-
-#endif /* _UNIFY_H */
author	Raghavendra Bhat <raghavendra@redhat.com>	2014-06-03 00:28:08 +0530
committer	Raghavendra G <rgowdapp@redhat.com>	2014-06-13 01:49:10 -0700
commit	6ba178fd9ebf9fc98415c30bcd338a68ee5eb601 (patch)
tree	b050a02506263651ba24603f25672bfd13d1f0b3 /cli
parent	4d656f9008747172db52f6d532b610c487528bfb (diff)