0 files changed, 0 insertions, 0 deletions
diff --git a/xlators/Makefile.am b/xlators/Makefile.am
index 2abb52194..f60fa85ce 100644
--- a/xlators/Makefile.am
+++ b/xlators/Makefile.am
@@ -1,3 +1,4 @@
-SUBDIRS = cluster storage protocol performance debug features encryption mount
+SUBDIRS = cluster storage protocol performance debug features encryption mount nfs mgmt system \
+          playground
 
-CLEANFILES = 
+CLEANFILES =
diff --git a/xlators/bindings/Makefile.am b/xlators/bindings/Makefile.am
deleted file mode 100644
index f77665802..000000000
--- a/xlators/bindings/Makefile.am
+++ /dev/null
@@ -1 +0,0 @@
-SUBDIRS = $(BINDINGS_SUBDIRS)
diff --git a/xlators/bindings/python/src/Makefile.am b/xlators/bindings/python/src/Makefile.am
deleted file mode 100644
index c0b9141c6..000000000
--- a/xlators/bindings/python/src/Makefile.am
+++ /dev/null
@@ -1,19 +0,0 @@
-
-xlator_PROGRAMS = python.so
-
-xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/bindings
-
-python_PYTHON = gluster.py glustertypes.py glusterstack.py
-
-pythondir = $(xlatordir)/python
-
-python_so_SOURCES = python.c
-
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall \
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles \
-	$(PYTHON_CPPLAGS) -DGLUSTER_PYTHON_PATH=\"$(pythondir)\"
-
-AM_LDFLAGS = $(PYTHON_LDFLAGS)
-
-CLEANFILES = 
-
diff --git a/xlators/bindings/python/src/gluster.py b/xlators/bindings/python/src/gluster.py
deleted file mode 100644
index ee0eb1310..000000000
--- a/xlators/bindings/python/src/gluster.py
+++ /dev/null
@@ -1,47 +0,0 @@
-#   Copyright (c) 2007 Chris AtLee <chris@atlee.ca>
-#   This file is part of GlusterFS.
-#
-#   GlusterFS is free software; you can redistribute it and/or modify
-#   it under the terms of the GNU General Public License as published
-#   by the Free Software Foundation; either version 3 of the License,
-#   or (at your option) any later version.
-#
-#   GlusterFS is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-#   General Public License for more details.
-#
-#   You should have received a copy of the GNU General Public License
-#   along with this program.  If not, see
-#   <http://www.gnu.org/licenses/>.
-from ctypes import *
-from glustertypes import *
-from glusterstack import *
-import sys
-import inspect
-
-libglusterfs = CDLL("libglusterfs.so")
-_gf_log = libglusterfs._gf_log
-_gf_log.restype = c_int32
-_gf_log.argtypes = [c_char_p, c_char_p, c_char_p, c_int32, c_int, c_char_p]
-
-gf_log_loglevel = c_int.in_dll(libglusterfs, "gf_log_loglevel")
-
-GF_LOG_NONE = 0
-GF_LOG_CRITICAL = 1
-GF_LOG_ERROR = 2
-GF_LOG_WARNING = 3
-GF_LOG_DEBUG = 4
-
-def gf_log(module, level, fmt, *params):
-    if level <= gf_log_loglevel:
-        frame = sys._getframe(1)
-        _gf_log(module, frame.f_code.co_filename, frame.f_code.co_name,
-                frame.f_lineno, level, fmt, *params)
-
-class ComplexTranslator(object):
-    def __init__(self, xlator):
-        self.xlator = xlator_t.from_address(xlator)
-
-    def __getattr__(self, item):
-        return getattr(self.xlator, item)
diff --git a/xlators/bindings/python/src/glusterstack.py b/xlators/bindings/python/src/glusterstack.py
deleted file mode 100644
index ba24c8165..000000000
--- a/xlators/bindings/python/src/glusterstack.py
+++ /dev/null
@@ -1,55 +0,0 @@
-#   Copyright (c) 2007 Chris AtLee <chris@atlee.ca>
-#   This file is part of GlusterFS.
-#
-#   GlusterFS is free software; you can redistribute it and/or modify
-#   it under the terms of the GNU General Public License as published
-#   by the Free Software Foundation; either version 3 of the License,
-#   or (at your option) any later version.
-#
-#   GlusterFS is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-#   General Public License for more details.
-#
-#   You should have received a copy of the GNU General Public License
-#   along with this program.  If not, see
-#   <http://www.gnu.org/licenses/>.
-from ctypes import *
-from glustertypes import *
-
-libc = CDLL("libc.so.6")
-calloc = libc.calloc
-calloc.argtypes = [c_int, c_int]
-calloc.restype = c_void_p
-
-# TODO: Can these be done in C somehow?
-def stack_wind(frame, rfn, obj, fn, *params):
-    """Frame is a frame object"""
-    _new = cast(calloc(1, sizeof(call_frame_t)), POINTER(call_frame_t))
-    _new[0].root = frame.root
-    _new[0].next = frame.root[0].frames.next
-    _new[0].prev = pointer(frame.root[0].frames)
-    if frame.root[0].frames.next:
-        frame.root[0].frames.next[0].prev = _new
-    frame.root[0].frames.next = _new
-    _new[0].this = obj
-    # TODO: Type checking like tmp_cbk?
-    _new[0].ret = rfn
-    _new[0].parent = pointer(frame)
-    _new[0].cookie = cast(_new, c_void_p)
-    # TODO: Initialize lock
-    #_new.lock.init()
-    frame.ref_count += 1
-    fn(_new, obj, *params)
-
-def stack_unwind(frame, *params):
-    """Frame is a frame object"""
-    fn = frame[0].ret
-    parent = frame[0].parent[0]
-    parent.ref_count -= 1
-
-    op_ret = params[0]
-    op_err = params[1]
-    params = params[2:]
-    fn(parent, call_frame_t.from_address(frame[0].cookie), parent.this,
-            op_ret, op_err, *params)
diff --git a/xlators/bindings/python/src/glustertypes.py b/xlators/bindings/python/src/glustertypes.py
deleted file mode 100644
index e9069d07c..000000000
--- a/xlators/bindings/python/src/glustertypes.py
+++ /dev/null
@@ -1,167 +0,0 @@
-#   Copyright (c) 2007 Chris AtLee <chris@atlee.ca>
-#   This file is part of GlusterFS.
-#
-#   GlusterFS is free software; you can redistribute it and/or modify
-#   it under the terms of the GNU General Public License as published
-#   by the Free Software Foundation; either version 3 of the License,
-#   or (at your option) any later version.
-#
-#   GlusterFS is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-#   General Public License for more details.
-#
-#   You should have received a copy of the GNU General Public License
-#   along with this program.  If not, see
-#   <http://www.gnu.org/licenses/>.
-from ctypes import *
-import collections
-
-#
-# Forward declaration of some gluster types
-#
-class call_frame_t(Structure):
-    pass
-
-class call_ctx_t(Structure):
-    pass
-
-class call_pool_t(Structure):
-    pass
-
-class xlator_t(Structure):
-    def _getFirstChild(self):
-        return self.children[0].xlator
-    firstChild = property(_getFirstChild)
-
-class xlator_list_t(Structure):
-    pass
-
-class xlator_fops(Structure):
-    pass
-
-class xlator_mops(Structure):
-    pass
-
-class glusterfs_ctx_t(Structure):
-    pass
-
-class list_head(Structure):
-    pass
-
-class dict_t(Structure):
-    pass
-
-class inode_table_t(Structure):
-    pass
-
-class fd_t(Structure):
-    pass
-
-class iovec(Structure):
-    _fields_ = [
-            ("iov_base", c_void_p),
-            ("iov_len", c_size_t),
-            ]
-
-    def __init__(self, s):
-        self.iov_base = cast(c_char_p(s), c_void_p)
-        self.iov_len = len(s)
-
-    def getBytes(self):
-        return string_at(self.iov_base, self.iov_len)
-
-# This is a pthread_spinlock_t
-# TODO: what happens to volatile-ness?
-gf_lock_t = c_int
-
-uid_t = c_uint32
-gid_t = c_uint32
-pid_t = c_int32
-
-off_t = c_int64
-
-#
-# Function pointer types
-#
-ret_fn_t = CFUNCTYPE(c_int32, POINTER(call_frame_t), POINTER(call_frame_t),
-                              POINTER(xlator_t), c_int32, c_int32)
-
-fini_fn_t = CFUNCTYPE(None, POINTER(xlator_t))
-init_fn_t = CFUNCTYPE(c_int32, POINTER(xlator_t))
-event_notify_fn_t = CFUNCTYPE(c_int32, POINTER(xlator_t), c_int32, c_void_p)
-
-list_head._fields_ = [
-        ("next", POINTER(list_head)),
-        ("prev", POINTER(list_head)),
-        ]
-
-call_frame_t._fields_ = [
-        ("root", POINTER(call_ctx_t)),
-        ("parent", POINTER(call_frame_t)),
-        ("next", POINTER(call_frame_t)),
-        ("prev", POINTER(call_frame_t)),
-        ("local", c_void_p),
-        ("this", POINTER(xlator_t)),
-        ("ret", ret_fn_t),
-        ("ref_count", c_int32),
-        ("lock", gf_lock_t),
-        ("cookie", c_void_p),
-        ("op", c_int32),
-        ("type", c_int8),
-        ]
-
-call_ctx_t._fields_ = [
-        ("all_frames", list_head),
-        ("trans", c_void_p),
-        ("pool", call_pool_t),
-        ("unique", c_uint64),
-        ("state", c_void_p),
-        ("uid", uid_t),
-        ("gid", gid_t),
-        ("pid", pid_t),
-        ("frames", call_frame_t),
-        ("req_refs", POINTER(dict_t)),
-        ("rsp_refs", POINTER(dict_t)),
-        ]
-
-xlator_t._fields_ = [
-        ("name", c_char_p),
-        ("type", c_char_p),
-        ("next", POINTER(xlator_t)),
-        ("prev", POINTER(xlator_t)),
-        ("parent", POINTER(xlator_t)),
-        ("children", POINTER(xlator_list_t)),
-        ("fops", POINTER(xlator_fops)),
-        ("mops", POINTER(xlator_mops)),
-        ("fini", fini_fn_t),
-        ("init", init_fn_t),
-        ("notify", event_notify_fn_t),
-        ("options", POINTER(dict_t)),
-        ("ctx", POINTER(glusterfs_ctx_t)),
-        ("itable", POINTER(inode_table_t)),
-        ("ready", c_char),
-        ("private", c_void_p),
-        ]
-
-xlator_list_t._fields_ = [
-        ("xlator", POINTER(xlator_t)),
-        ("next", POINTER(xlator_list_t)),
-        ]
-
-fop_functions = collections.defaultdict(lambda: c_void_p)
-fop_function_names = ['lookup', 'forget', 'stat', 'fstat', 'chmod', 'fchmod',
-        'chown', 'fchown', 'truncate', 'ftruncate', 'utimens', 'access',
-        'readlink', 'mknod', 'mkdir', 'unlink', 'rmdir', 'symlink',
-        'rename', 'link', 'create', 'open', 'readv', 'writev', 'flush',
-        'close', 'fsync', 'opendir', 'readdir', 'closedir', 'fsyncdir',
-        'statfs', 'setxattr', 'getxattr', 'removexattr', 'lk', 'writedir',
-        # TODO: Call backs?
-        ]
-
-fop_writev_t = CFUNCTYPE(c_int32, POINTER(call_frame_t), POINTER(xlator_t),
-                                  POINTER(fd_t), POINTER(iovec), c_int32,
-                                  off_t)
-
-fop_functions['writev'] = fop_writev_t
-xlator_fops._fields_ = [(f, fop_functions[f]) for f in fop_function_names]
diff --git a/xlators/bindings/python/src/python.c b/xlators/bindings/python/src/python.c
deleted file mode 100644
index 7f32d7f29..000000000
--- a/xlators/bindings/python/src/python.c
+++ /dev/null
@@ -1,235 +0,0 @@
-/*
-   Copyright (c) 2007-2009 Chris AtLee <chris@atlee.ca>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
-*/
-
-#include <Python.h>
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include "glusterfs.h"
-#include "xlator.h"
-#include "logging.h"
-#include "defaults.h"
-
-typedef struct
-{
-    char        *scriptname;
-    PyObject    *pXlator;
-    PyObject    *pScriptModule;
-    PyObject    *pGlusterModule;
-    PyThreadState *pInterp;
-
-    PyObject    *pFrameType, *pVectorType, *pFdType;
-} python_private_t;
-
-int32_t
-python_writev (call_frame_t *frame,
-              xlator_t *this,
-              fd_t *fd,
-              struct iovec *vector,
-              int32_t count, 
-              off_t offset)
-{
-  python_private_t *priv = (python_private_t *)this->private;
-  gf_log("python", GF_LOG_DEBUG, "In writev");
-  if (PyObject_HasAttrString(priv->pXlator, "writev"))
-  {
-
-      PyObject *retval = PyObject_CallMethod(priv->pXlator, "writev",
-              "O O O i l",
-              PyObject_CallMethod(priv->pFrameType, "from_address", "O&", PyLong_FromVoidPtr, frame),
-              PyObject_CallMethod(priv->pFdType, "from_address", "O&", PyLong_FromVoidPtr, fd),
-              PyObject_CallMethod(priv->pVectorType, "from_address", "O&", PyLong_FromVoidPtr, vector),
-              count,
-              offset);
-      if (PyErr_Occurred())
-      {
-          PyErr_Print();
-      }
-      Py_XDECREF(retval);
-  }
-  else
-  {
-      return default_writev(frame, this, fd, vector, count, offset);
-  }
-  return 0;
-}
-
-struct xlator_fops fops = {
-    .writev       = python_writev
-};
-
-struct xlator_mops mops = {
-};
-
-static PyObject *
-AnonModule_FromFile (const char* fname)
-{
-    // Get the builtins
-    PyThreadState* pThread = PyThreadState_Get();
-    PyObject *pBuiltins = pThread->interp->builtins;
-
-    if (PyErr_Occurred())
-    {
-        PyErr_Print();
-        return NULL;
-    }
-
-    // Create a new dictionary for running code in
-    PyObject *pModuleDict = PyDict_New();
-    PyDict_SetItemString(pModuleDict, "__builtins__", pBuiltins);
-    Py_INCREF(pBuiltins);
-
-    // Run the file in the new context
-    FILE* fp = fopen(fname, "r");
-    PyRun_File(fp, fname, Py_file_input, pModuleDict, pModuleDict);
-    fclose(fp);
-    if (PyErr_Occurred())
-    {
-        PyErr_Print();
-        Py_DECREF(pModuleDict);
-        Py_DECREF(pBuiltins);
-        return NULL;
-    }
-
-    // Create an object to hold the new context
-    PyRun_String("class ModuleWrapper(object):\n\tpass\n", Py_single_input, pModuleDict, pModuleDict);
-    if (PyErr_Occurred())
-    {
-        PyErr_Print();
-        Py_DECREF(pModuleDict);
-        Py_DECREF(pBuiltins);
-        return NULL;
-    }
-    PyObject *pModule = PyRun_String("ModuleWrapper()", Py_eval_input, pModuleDict, pModuleDict);
-    if (PyErr_Occurred())
-    {
-        PyErr_Print();
-        Py_DECREF(pModuleDict);
-        Py_DECREF(pBuiltins);
-        Py_XDECREF(pModule);
-        return NULL;
-    }
-
-    // Set the new context's dictionary to the one we used to run the code
-    // inside
-    PyObject_SetAttrString(pModule, "__dict__", pModuleDict);
-    if (PyErr_Occurred())
-    {
-        PyErr_Print();
-        Py_DECREF(pModuleDict);
-        Py_DECREF(pBuiltins);
-        Py_DECREF(pModule);
-        return NULL;
-    }
-
-    return pModule;
-}
-
-int32_t
-init (xlator_t *this)
-{
-  // This is ok to call more than once per process
-  Py_InitializeEx(0);
-
-  if (!this->children) {
-    gf_log ("python", GF_LOG_ERROR, 
-            "FATAL: python should have exactly one child");
-    return -1;
-  }
-
-  python_private_t *priv = CALLOC (sizeof (python_private_t), 1);
-  ERR_ABORT (priv);
-
-  data_t *scriptname = dict_get (this->options, "scriptname");
-  if (scriptname) {
-      priv->scriptname = data_to_str(scriptname);
-  } else {
-      gf_log("python", GF_LOG_ERROR,
-              "FATAL: python requires the scriptname parameter");
-      return -1;
-  }
-
-  priv->pInterp = Py_NewInterpreter();
-    
-  // Adjust python's path
-  PyObject *syspath = PySys_GetObject("path");
-  PyObject *path = PyString_FromString(GLUSTER_PYTHON_PATH);
-  PyList_Append(syspath, path);
-  Py_DECREF(path);
-
-  gf_log("python", GF_LOG_DEBUG,
-          "Loading gluster module");
-
-  priv->pGlusterModule = PyImport_ImportModule("gluster");
-  if (PyErr_Occurred())
-  {
-      PyErr_Print();
-      return -1;
-  }
-
-  priv->pFrameType = PyObject_GetAttrString(priv->pGlusterModule, "call_frame_t");
-  priv->pFdType = PyObject_GetAttrString(priv->pGlusterModule, "fd_t");
-  priv->pVectorType = PyObject_GetAttrString(priv->pGlusterModule, "iovec");
-
-  gf_log("python", GF_LOG_DEBUG, "Loading script...%s", priv->scriptname);
-  
-  priv->pScriptModule = AnonModule_FromFile(priv->scriptname);
-  if (!priv->pScriptModule || PyErr_Occurred())
-  {
-      gf_log("python", GF_LOG_ERROR, "Error loading %s", priv->scriptname);
-      PyErr_Print();
-      return -1;
-  }
-
-  if (!PyObject_HasAttrString(priv->pScriptModule, "xlator"))
-  {
-      gf_log("python", GF_LOG_ERROR, "%s does not have a xlator attribute", priv->scriptname);
-      return -1;
-  }
-  gf_log("python", GF_LOG_DEBUG, "Instantiating translator");
-  priv->pXlator = PyObject_CallMethod(priv->pScriptModule, "xlator", "O&",
-          PyLong_FromVoidPtr, this);
-  if (PyErr_Occurred() || !priv->pXlator)
-  {
-      PyErr_Print();
-      return -1;
-  }
-
-  this->private = priv;
-
-  gf_log ("python", GF_LOG_DEBUG, "python xlator loaded");
-  return 0;
-}
-
-void 
-fini (xlator_t *this)
-{
-  python_private_t *priv = (python_private_t*)(this->private);
-  Py_DECREF(priv->pXlator);
-  Py_DECREF(priv->pScriptModule);
-  Py_DECREF(priv->pGlusterModule);
-  Py_DECREF(priv->pFrameType);
-  Py_DECREF(priv->pFdType);
-  Py_DECREF(priv->pVectorType);
-  Py_EndInterpreter(priv->pInterp);
-  return;
-}
diff --git a/xlators/bindings/python/src/testxlator.py b/xlators/bindings/python/src/testxlator.py
deleted file mode 100644
index 507455c85..000000000
--- a/xlators/bindings/python/src/testxlator.py
+++ /dev/null
@@ -1,56 +0,0 @@
-#   Copyright (c) 2007 Chris AtLee <chris@atlee.ca>
-#   This file is part of GlusterFS.
-#
-#   GlusterFS is free software; you can redistribute it and/or modify
-#   it under the terms of the GNU General Public License as published
-#   by the Free Software Foundation; either version 3 of the License,
-#   or (at your option) any later version.
-#
-#   GlusterFS is distributed in the hope that it will be useful, but
-#   WITHOUT ANY WARRANTY; without even the implied warranty of
-#   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-#   General Public License for more details.
-#
-#   You should have received a copy of the GNU General Public License
-#   along with this program.  If not, see
-#   <http://www.gnu.org/licenses/>.
-
-"""
-This is a test translator written in python.
-
-Important things to note:
-    This file must be import-able from glusterfsd.  This probably means
-    setting PYTHONPATH to where this file is located.
-
-    This file must have a top-level xlator class object that will be
-    used to instantiate individual translators.
-"""
-from gluster import *
-
-class MyXlator(ComplexTranslator):
-    name = "MyXlator"
-    def writev_cbk(self, frame, cookie, op_ret, op_errno, buf):
-        stack_unwind(frame, op_ret, op_errno, buf)
-        return 0
-
-    def writev(self, frame, fd, vector, count, offset):
-        gf_log(self.name, GF_LOG_WARNING, "writev %i bytes", vector.iov_len)
-        # TODO: Use cookie to pass this to writev_cbk
-        old_count = vector.iov_len
-
-        data = vector.getBytes().encode("zlib")
-
-        vector = iovec(data)
-        gf_log(self.name, GF_LOG_WARNING, "writev %i bytes", vector.iov_len)
-
-        @ret_fn_t
-        def rfn(frame, prev, this, op_ret, op_errno, *params):
-            if len(params) == 0:
-                params = [0]
-            return self.writev_cbk(frame, prev, old_count, op_errno, *params)
-
-        stack_wind(frame, rfn, self.firstChild,
-                self.firstChild[0].fops[0].writev, fd, vector, count, offset)
-        return 0
-
-xlator = MyXlator
diff --git a/xlators/cluster/Makefile.am b/xlators/cluster/Makefile.am
index a6ddb3564..0990822a7 100644
--- a/xlators/cluster/Makefile.am
+++ b/xlators/cluster/Makefile.am
@@ -1,3 +1,3 @@
-SUBDIRS = unify stripe afr dht ha map
+SUBDIRS = stripe afr dht
 
 CLEANFILES = 
diff --git a/xlators/cluster/afr/src/Makefile.am b/xlators/cluster/afr/src/Makefile.am
index 1a8ddadb7..ea5a90abb 100644
--- a/xlators/cluster/afr/src/Makefile.am
+++ b/xlators/cluster/afr/src/Makefile.am
@@ -1,20 +1,38 @@
-xlator_LTLIBRARIES = afr.la
+xlator_LTLIBRARIES = afr.la pump.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/cluster
 
-afr_la_LDFLAGS = -module -avoidversion 
+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_SOURCES = afr.c 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_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
 
-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_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-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 -shared -nostartfiles $(GF_CFLAGS)
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
-CLEANFILES = 
+CLEANFILES =
 
 uninstall-local:
 	rm -f $(DESTDIR)$(xlatordir)/replicate.so
 
 install-data-hook:
-	ln -sf afr.so $(DESTDIR)$(xlatordir)/replicate.so
-\ No newline at end of file
+	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
new file mode 100644
index 000000000..6bd231600
--- /dev/null
+++ b/xlators/cluster/afr/src/afr-common.c
@@ -0,0 +1,3625 @@
+/*
+  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>
+#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 "byte-order.h"
+#include "statedump.h"
+#include "inode.h"
+
+#include "fd.h"
+
+#include "afr-inode-read.h"
+#include "afr-inode-write.h"
+#include "afr-dir-read.h"
+#include "afr-dir-write.h"
+#include "afr-transaction.h"
+#include "afr-self-heal.h"
+#include "afr-self-heald.h"
+
+
+call_frame_t *
+afr_copy_frame (call_frame_t *base)
+{
+	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;
+}
+
+/*
+ * 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.
+ */
+
+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;
+}
+
+
+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);
+}
+
+
+int
+__afr_inode_read_subvol_reset_small (inode_t *inode, xlator_t *this)
+{
+	int ret = -1;
+	uint16_t datamap = 0;
+	uint16_t metadatamap = 0;
+	uint32_t event = 0;
+	uint64_t val = 0;
+
+	ret = __inode_ctx_get (inode, this, &val);
+	(void) ret;
+
+	metadatamap = (val & 0x000000000000ffff) >> 0;
+	datamap =     (val & 0x00000000ffff0000) >> 16;
+	event = 0;
+
+	val = ((uint64_t) metadatamap) |
+		(((uint64_t) datamap) << 16) |
+		(((uint64_t) event) << 32);
+
+	return __inode_ctx_set (inode, this, &val);
+}
+
+
+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;
+}
+
+
+int
+__afr_inode_read_subvol_set (inode_t *inode, xlator_t *this, unsigned char *data,
+			     unsigned char *metadata, int event)
+{
+	afr_private_t *priv = NULL;
+	int ret = -1;
+
+	priv = this->private;
+
+	if (priv->child_count <= 16)
+		ret = __afr_inode_read_subvol_set_small (inode, this, data,
+							 metadata, event);
+	else
+		ret = -1;
+
+	return ret;
+}
+
+
+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;
+}
+
+
+int
+afr_inode_read_subvol_get (inode_t *inode, xlator_t *this, unsigned char *data,
+			   unsigned char *metadata, int *event_p)
+{
+	int ret = -1;
+
+	LOCK(&inode->lock);
+	{
+		ret = __afr_inode_read_subvol_get (inode, this, data,
+						   metadata, event_p);
+	}
+	UNLOCK(&inode->lock);
+
+	return ret;
+}
+
+
+int
+afr_inode_read_subvol_set (inode_t *inode, xlator_t *this, unsigned char *data,
+			   unsigned char *metadata, int event)
+{
+	int ret = -1;
+
+	LOCK(&inode->lock);
+	{
+		ret = __afr_inode_read_subvol_set (inode, this, data, metadata,
+						   event);
+	}
+	UNLOCK(&inode->lock);
+
+	return ret;
+}
+
+
+int
+afr_inode_read_subvol_reset (inode_t *inode, xlator_t *this)
+{
+	int ret = -1;
+
+	LOCK(&inode->lock);
+	{
+		ret = __afr_inode_read_subvol_reset (inode, this);
+	}
+	UNLOCK(&inode->lock);
+
+	return ret;
+}
+
+
+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;
+}
+
+
+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;
+}
+
+
+int
+afr_replies_interpret (call_frame_t *frame, xlator_t *this, inode_t *inode)
+{
+	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;
+}
+
+
+
+int
+afr_refresh_selfheal_done (int ret, call_frame_t *heal, void *opaque)
+{
+	if (heal)
+		STACK_DESTROY (heal->root);
+	return 0;
+}
+
+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;
+}
+
+
+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;
+}
+
+
+gf_boolean_t
+afr_selfheal_enabled (xlator_t *this)
+{
+	afr_private_t *priv = NULL;
+	gf_boolean_t data = _gf_false;
+
+	priv = this->private;
+
+	gf_string2boolean (priv->data_self_heal, &data);
+
+	return data || priv->metadata_self_heal || priv->entry_self_heal;
+}
+
+
+
+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;
+}
+
+
+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;
+
+	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)
+{
+	loc_t loc = {0, };
+	afr_private_t *priv = NULL;
+
+	priv = this->private;
+
+	loc.inode = inode;
+	uuid_copy (loc.gfid, inode->gfid);
+
+	STACK_WIND_COOKIE (frame, afr_inode_refresh_subvol_cbk,
+			   (void *) (long) i, priv->children[i],
+			   priv->children[i]->fops->lookup, &loc, xdata);
+	return 0;
+}
+
+
+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");
+        }
+
+	return ret;
+}
+
+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);
+        }
+
+        ret = dict_set_uint32 (local->xattr_req, GLUSTERFS_PARENT_ENTRYLK, 0);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: Unable to set dict value for %s",
+                        loc->path, GLUSTERFS_PARENT_ENTRYLK);
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+
+int
+afr_hash_child (inode_t *inode, int32_t child_count, int hashmode)
+{
+        uuid_t gfid_copy = {0,};
+        pid_t pid;
+
+        if (!hashmode) {
+                return -1;
+        }
+
+        if (inode) {
+               uuid_copy (gfid_copy, inode->gfid);
+        }
+
+        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)
+{
+        afr_private_t *priv    = NULL;
+        int           i        = 0;
+
+        priv = this->private;
+
+        afr_matrix_cleanup (local->pending, priv->child_count);
+
+        GF_FREE (local->internal_lock.locked_nodes);
+
+        for (i = 0; local->internal_lock.inodelk[i].domain; i++) {
+                GF_FREE (local->internal_lock.inodelk[i].locked_nodes);
+        }
+
+        GF_FREE (local->internal_lock.lower_locked_nodes);
+
+        afr_entry_lockee_cleanup (&local->internal_lock);
+
+        GF_FREE (local->transaction.pre_op);
+        GF_FREE (local->transaction.eager_lock);
+
+        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)
+{
+        afr_private_t * priv = NULL;
+
+        if (!local)
+                return;
+
+	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);
+
+        priv = this->private;
+
+        loc_wipe (&local->loc);
+        loc_wipe (&local->newloc);
+
+        if (local->fd)
+                fd_unref (local->fd);
+
+        if (local->xattr_req)
+                dict_unref (local->xattr_req);
+
+        if (local->dict)
+                dict_unref (local->dict);
+
+	afr_replies_wipe (local, priv);
+	GF_FREE(local->replies);
+
+        GF_FREE (local->child_up);
+
+        GF_FREE (local->read_attempted);
+
+        GF_FREE (local->readable);
+
+	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 */
+                GF_FREE (local->cont.getxattr.name);
+        }
+
+        { /* lk */
+                GF_FREE (local->cont.lk.locked_nodes);
+        }
+
+        { /* create */
+                if (local->cont.create.fd)
+                        fd_unref (local->cont.create.fd);
+                if (local->cont.create.params)
+                        dict_unref (local->cont.create.params);
+        }
+
+        { /* mknod */
+                if (local->cont.mknod.params)
+                        dict_unref (local->cont.mknod.params);
+        }
+
+        { /* mkdir */
+                if (local->cont.mkdir.params)
+                        dict_unref (local->cont.mkdir.params);
+        }
+
+        { /* symlink */
+                if (local->cont.symlink.params)
+                        dict_unref (local->cont.symlink.params);
+        }
+
+        { /* writev */
+                GF_FREE (local->cont.writev.vector);
+        }
+
+        { /* setxattr */
+                if (local->cont.setxattr.dict)
+                        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 */
+                GF_FREE (local->cont.opendir.checksum);
+        }
+
+        { /* readdirp */
+                if (local->cont.readdir.dict)
+                        dict_unref (local->cont.readdir.dict);
+        }
+
+        if (local->xdata_req)
+                dict_unref (local->xdata_req);
+
+        if (local->xdata_rsp)
+                dict_unref (local->xdata_rsp);
+}
+
+
+int
+afr_frame_return (call_frame_t *frame)
+{
+        afr_local_t *local = NULL;
+        int          call_count = 0;
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                call_count = --local->call_count;
+        }
+        UNLOCK (&frame->lock);
+
+        return call_count;
+}
+
+
+gf_boolean_t
+afr_is_entry_possibly_under_txn (afr_local_t *local, xlator_t *this)
+{
+	int i = 0;
+	int tmp = 0;
+	afr_private_t *priv = NULL;
+
+	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;
+	}
+
+	return _gf_false;
+}
+
+
+/*
+ * Quota size xattrs are not maintained by afr. There is a
+ * possibility that they differ even when both the directory changelog xattrs
+ * suggest everything is fine. So if there is at least one 'source' check among
+ * the sources which has the maximum quota size. Otherwise check among all the
+ * available ones for maximum quota size. This way if there is a source and
+ * stale copies it always votes for the 'source'.
+ * */
+
+static void
+afr_handle_quota_size (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;
+	}
+}
+
+
+static void
+afr_lookup_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;
+	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;
+
+	locked_entry = afr_is_entry_possibly_under_txn (local, this);
+
+	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_higher_errno (int32_t old_errno, int32_t new_errno)
+{
+	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_final_errno (afr_local_t *local, afr_private_t *priv)
+{
+	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;
+}
+
+static int
+get_pathinfo_host (char *pathinfo, char *hostname, size_t size)
+{
+        char    *start = NULL;
+        char    *end = NULL;
+        int     ret  = -1;
+        int     i    = 0;
+
+        if (!pathinfo)
+                goto out;
+
+        start = strchr (pathinfo, ':');
+        if (!start)
+                goto out;
+        end = strrchr (pathinfo, ':');
+        if (start == end)
+                goto out;
+
+        memset (hostname, 0, size);
+        i = 0;
+        while (++start != end)
+                hostname[i++] = *start;
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+afr_local_pathinfo (char *pathinfo, gf_boolean_t *local)
+{
+        int             ret   = 0;
+        char            pathinfohost[1024] = {0};
+        char            localhost[1024] = {0};
+        xlator_t        *this = THIS;
+
+        *local = _gf_false;
+        ret = get_pathinfo_host (pathinfo, pathinfohost, sizeof (pathinfohost));
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Invalid pathinfo: %s",
+                        pathinfo);
+                goto out;
+        }
+
+        ret = gethostname (localhost, sizeof (localhost));
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "gethostname() failed, "
+                        "reason: %s", strerror (errno));
+                goto out;
+        }
+
+        if (!strcmp (localhost, pathinfohost))
+                *local = _gf_true;
+out:
+        return ret;
+}
+
+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)
+{
+        int              ret            = 0;
+        char            *pathinfo       = NULL;
+        gf_boolean_t     is_local        = _gf_false;
+        afr_private_t   *priv           = NULL;
+        int32_t          child_index    = -1;
+
+        if (op_ret != 0) {
+                goto out;
+        }
+
+	priv = this->private;
+	child_index = (int32_t)(long)cookie;
+
+        ret = dict_get_str (dict, GF_XATTR_PATHINFO_KEY, &pathinfo);
+        if (ret != 0) {
+                goto out;
+        }
+
+        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_attempt_local_discovery (xlator_t *this, int32_t child_index)
+{
+        call_frame_t    *newframe = NULL;
+        loc_t            tmploc = {0,};
+        afr_private_t   *priv = this->private;
+
+        newframe = create_frame(this,this->ctx->pool);
+        if (!newframe) {
+                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);
+}
+
+
+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;
+
+	local = frame->local;
+	this = frame->this;
+
+	afr_selfheal_name (frame->this, local->loc.pargfid, local->loc.name);
+
+	afr_replies_wipe (local, this->private);
+
+	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);
+
+	return 0;
+}
+
+
+int
+afr_lookup_entry_heal (call_frame_t *frame, xlator_t *this)
+{
+	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_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;
+        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);
+	}
+
+        call_count = afr_frame_return (frame);
+        if (call_count == 0) {
+		afr_lookup_entry_heal (frame, this);
+        }
+
+	return 0;
+}
+
+
+
+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;
+
+        priv  = this->private;
+        local = frame->local;
+
+	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;
+	}
+
+	op_errno = afr_final_errno (frame->local, this->private);
+
+        if (local->op_ret < 0) {
+		local->op_errno = op_errno;
+		local->op_ret = -1;
+                goto unwind;
+	}
+
+	afr_replies_interpret (frame, this, local->inode);
+
+	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);
+
+		for (i = 0; i < priv->child_count; i++) {
+			if (!local->replies[i].valid ||
+			    local->replies[i].op_ret == -1)
+				continue;
+			read_subvol = i;
+			break;
+		}
+	}
+
+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);
+}
+
+
+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_discover_done (frame, this);
+        }
+
+	return 0;
+}
+
+
+int
+afr_discover_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;
+
+	local = frame->local;
+	priv = this->private;
+
+	if (err) {
+		local->op_errno = -err;
+		ret = -1;
+		goto out;
+	}
+
+	call_count = local->call_count = AFR_COUNT (local->child_up,
+						    priv->child_count);
+
+        ret = afr_lookup_xattr_req_prepare (local, this, local->xattr_req,
+					    &local->loc);
+        if (ret) {
+                local->op_errno = -ret;
+		ret = -1;
+                goto out;
+        }
+
+        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;
+                }
+        }
+
+	return 0;
+out:
+	AFR_STACK_UNWIND (lookup, frame, -1, local->op_errno, 0, 0, 0, 0);
+	return 0;
+}
+
+
+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;
+
+	priv = this->private;
+
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
+
+        if (!local->call_count) {
+                op_errno = ENOTCONN;
+                goto out;
+        }
+
+	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);
+
+		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;
+                }
+	}
+
+        local->op = GF_FOP_LOOKUP;
+
+        loc_copy (&local->loc, loc);
+
+	local->inode = inode_ref (loc->inode);
+
+	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 (uuid_is_null (loc->inode->gfid)) {
+		afr_discover_do (frame, this, 0);
+		return 0;
+	}
+
+	afr_read_subvol_get (loc->inode, this, NULL, &event,
+			     AFR_DATA_TRANSACTION);
+
+	if (event != local->event_generation)
+		afr_inode_refresh (frame, this, loc->inode, afr_discover_do);
+	else
+		afr_discover_do (frame, this, 0);
+
+	return 0;
+out:
+	AFR_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+	return 0;
+}
+
+
+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;
+
+	local = frame->local;
+	priv = this->private;
+
+	if (err < 0) {
+		local->op_errno = -err;
+		ret = -1;
+		goto out;
+	}
+
+	call_count = local->call_count = AFR_COUNT (local->child_up,
+						    priv->child_count);
+
+        ret = afr_lookup_xattr_req_prepare (local, this, local->xattr_req,
+					    &local->loc);
+        if (ret) {
+                local->op_errno = -ret;
+		ret = -1;
+                goto out;
+        }
+
+        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;
+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_local_t   *local = NULL;
+        int32_t        op_errno = 0;
+	int            event = 0;
+
+	if (!loc->parent) {
+		afr_discover (frame, this, loc, xattr_req);
+		return 0;
+	}
+
+	if (__is_root_gfid (loc->parent->gfid)) {
+		if (!strcmp (loc->name, GF_REPLICATE_TRASH_DIR)) {
+			op_errno = EPERM;
+			goto out;
+		}
+	}
+
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
+
+        if (!local->call_count) {
+                op_errno = ENOTCONN;
+                goto out;
+        }
+
+        local->op = GF_FOP_LOOKUP;
+
+        loc_copy (&local->loc, loc);
+
+	local->inode = inode_ref (loc->inode);
+
+	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);
+
+	afr_read_subvol_get (loc->parent, this, NULL, &event,
+			     AFR_DATA_TRANSACTION);
+
+	if (event != local->event_generation)
+		afr_inode_refresh (frame, this, loc->parent, afr_lookup_do);
+	else
+		afr_lookup_do (frame, this, 0);
+
+	return 0;
+out:
+	AFR_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+/* {{{ open */
+
+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;
+
+        ret = __fd_ctx_get (fd, this, &ctx);
+
+        if (ret < 0) {
+                ret = __afr_fd_ctx_set (this, fd);
+                if (ret < 0)
+                        goto out;
+
+                ret = __fd_ctx_get (fd, this, &ctx);
+                if (ret < 0)
+                        goto out;
+        }
+
+        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+out:
+        return fd_ctx;
+}
+
+
+afr_fd_ctx_t *
+afr_fd_ctx_get (fd_t *fd, xlator_t *this)
+{
+        afr_fd_ctx_t  *fd_ctx = NULL;
+
+        LOCK(&fd->lock);
+        {
+                fd_ctx = __afr_fd_ctx_get (fd, this);
+        }
+        UNLOCK(&fd->lock);
+
+        return fd_ctx;
+}
+
+
+int
+__afr_fd_ctx_set (xlator_t *this, fd_t *fd)
+{
+        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;
+
+        ret = __fd_ctx_get (fd, this, &ctx);
+
+        if (ret == 0)
+                goto out;
+
+        fd_ctx = GF_CALLOC (1, sizeof (afr_fd_ctx_t),
+                            gf_afr_mt_afr_fd_ctx_t);
+        if (!fd_ctx) {
+                ret = -ENOMEM;
+                goto out;
+        }
+
+	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;
+        }
+
+	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->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->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;
+        }
+
+	pthread_mutex_init (&fd_ctx->delay_lock, NULL);
+
+        INIT_LIST_HEAD (&fd_ctx->eager_locked);
+
+        ret = __fd_ctx_set (fd, this, (uint64_t)(long) fd_ctx);
+        if (ret)
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "failed to set fd ctx (%p)", fd);
+out:
+        return ret;
+}
+
+
+int
+afr_fd_ctx_set (xlator_t *this, fd_t *fd)
+{
+        int ret = -1;
+
+        LOCK (&fd->lock);
+        {
+                ret = __afr_fd_ctx_set (this, fd);
+        }
+        UNLOCK (&fd->lock);
+
+        return ret;
+}
+
+/* {{{ flush */
+
+int
+afr_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        afr_local_t *local = NULL;
+        int call_count = -1;
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret != -1) {
+			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);
+
+	call_count = afr_frame_return (frame);
+
+	if (call_count == 0)
+		AFR_STACK_UNWIND (flush, frame, local->op_ret,
+				  local->op_errno, local->xdata_rsp);
+
+        return 0;
+}
+
+static int
+afr_flush_wrapper (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+{
+        int           i      = 0;
+        afr_local_t   *local = NULL;
+        afr_private_t *priv  = NULL;
+        int call_count       = -1;
+
+        priv = this->private;
+        local = frame->local;
+        call_count = local->call_count;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND_COOKIE (frame, afr_flush_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->flush,
+                                           local->fd, xdata);
+                        if (!--call_count)
+                                break;
+
+                }
+        }
+
+        return 0;
+}
+
+int
+afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+{
+        afr_local_t   *local = NULL;
+        call_stub_t   *stub = NULL;
+        int            op_errno   = ENOMEM;
+
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
+
+	if (!local->call_count) {
+		op_errno = ENOTCONN;
+		goto out;
+	}
+
+	local->fd = fd_ref(fd);
+
+        stub = fop_flush_stub (frame, afr_flush_wrapper, fd, xdata);
+        if (!stub)
+                goto out;
+
+        afr_delayed_changelog_wake_resume (this, fd, stub);
+
+	return 0;
+out:
+	AFR_STACK_UNWIND (flush, frame, -1, op_errno, NULL);
+        return 0;
+}
+
+/* }}} */
+
+
+int
+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) {
+                //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));
+
+		for (i = 0; i < AFR_NUM_CHANGE_LOGS; i++)
+			GF_FREE (fd_ctx->pre_op_done[i]);
+
+                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);
+        }
+
+out:
+        return 0;
+}
+
+
+int
+afr_release (xlator_t *this, fd_t *fd)
+{
+        afr_cleanup_fd_ctx (this, fd);
+
+        return 0;
+}
+
+
+/* {{{ fsync */
+
+int
+afr_fsync_unwind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                      struct iatt *postbuf, dict_t *xdata)
+{
+        AFR_STACK_UNWIND (fsync, frame, op_ret, op_errno, prebuf, postbuf,
+                          xdata);
+        return 0;
+}
+
+int
+afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+               struct iatt *postbuf, dict_t *xdata)
+{
+        afr_local_t *local = NULL;
+        int call_count = -1;
+        int child_index = (long) cookie;
+	int read_subvol = 0;
+	call_stub_t *stub = NULL;
+
+        local = frame->local;
+
+	read_subvol = afr_data_subvol_get (local->inode, this, 0, 0);
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == 0) {
+                        if (local->op_ret == -1) {
+				local->op_ret = 0;
+
+                                local->cont.inode_wfop.prebuf  = *prebuf;
+                                local->cont.inode_wfop.postbuf = *postbuf;
+
+				if (xdata)
+					local->xdata_rsp = dict_ref (xdata);
+                        }
+
+                        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) {
+		/* 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;
+}
+
+
+int
+afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t datasync,
+	   dict_t *xdata)
+{
+	afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        int i = 0;
+        int32_t call_count = 0;
+        int32_t op_errno = ENOMEM;
+
+	priv = this->private;
+
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
+
+        call_count = local->call_count;
+	if (!call_count) {
+		op_errno = ENOTCONN;
+		goto out;
+	}
+
+        local->fd = fd_ref (fd);
+
+	if (afr_fd_has_witnessed_unstable_write (this, fd)) {
+		/* don't care. we only wanted to CLEAR the bit */
+	}
+
+	local->inode = inode_ref (fd->inode);
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i]) {
+                        STACK_WIND_COOKIE (frame, afr_fsync_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->fsync,
+                                           fd, datasync, xdata);
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+	return 0;
+out:
+	AFR_STACK_UNWIND (fsync, frame, -1, op_errno, NULL, NULL, NULL);
+
+        return 0;
+}
+
+/* }}} */
+
+/* {{{ fsync */
+
+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) {
+                        local->op_ret = 0;
+			if (!local->xdata_rsp && xdata)
+				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)
+                AFR_STACK_UNWIND (fsyncdir, frame, local->op_ret,
+				  local->op_errno, local->xdata_rsp);
+
+        return 0;
+}
+
+
+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_local_t *local = NULL;
+        int i = 0;
+        int32_t call_count = 0;
+        int32_t op_errno = ENOMEM;
+
+	priv = this->private;
+
+        local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
+
+        call_count = local->call_count;
+	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, xdata);
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+	return 0;
+out:
+	AFR_STACK_UNWIND (fsyncdir, frame, -1, op_errno, NULL);
+
+        return 0;
+}
+
+/* }}} */
+
+/* {{{ xattrop */
+
+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 *xdata)
+{
+        afr_local_t *local = NULL;
+        int call_count = -1;
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                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;
+        }
+        UNLOCK (&frame->lock);
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0)
+                AFR_STACK_UNWIND (xattrop, frame, local->op_ret, local->op_errno,
+                local->cont.xattrop.xattr, local->xdata_rsp);
+
+        return 0;
+}
+
+
+int32_t
+afr_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
+             gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
+{
+        afr_private_t *priv = NULL;
+        afr_local_t *local  = NULL;
+        int i = 0;
+        int32_t call_count = 0;
+        int32_t op_errno = ENOMEM;
+
+        priv = this->private;
+
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
+
+        call_count = local->call_count;
+	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, xdata);
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+	return 0;
+out:
+	AFR_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+
+/* }}} */
+
+/* {{{ fxattrop */
+
+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 *xdata)
+{
+        afr_local_t *local = NULL;
+
+        int call_count = -1;
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                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;
+        }
+        UNLOCK (&frame->lock);
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0)
+                AFR_STACK_UNWIND (fxattrop, frame, local->op_ret, local->op_errno,
+                                  local->cont.fxattrop.xattr, local->xdata_rsp);
+
+        return 0;
+}
+
+
+int32_t
+afr_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
+              gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
+{
+        afr_private_t *priv = NULL;
+        afr_local_t *local  = NULL;
+        int i = 0;
+        int32_t call_count = 0;
+        int32_t op_errno = 0;
+
+        priv = this->private;
+
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+                goto out;
+
+        call_count = local->call_count;
+	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, xdata);
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+	return 0;
+out:
+	AFR_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+
+/* }}} */
+
+
+int32_t
+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;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == 0)
+                        local->op_ret = 0;
+
+                local->op_errno = op_errno;
+        }
+        UNLOCK (&frame->lock);
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0)
+                AFR_STACK_UNWIND (inodelk, frame, local->op_ret,
+                                  local->op_errno, xdata);
+
+        return 0;
+}
+
+
+int32_t
+afr_inodelk (call_frame_t *frame, xlator_t *this,
+             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 i = 0;
+        int32_t call_count = 0;
+        int32_t op_errno = ENOMEM;
+
+        priv = this->private;
+
+        local = AFR_FRAME_INIT (frame, op_errno);
+        if (!local)
+                goto out;
+
+        call_count = local->call_count;
+	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, xdata);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+	return 0;
+out:
+	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, dict_t *xdata)
+
+{
+        afr_local_t *local = NULL;
+        int call_count = -1;
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == 0)
+                        local->op_ret = 0;
+
+                local->op_errno = op_errno;
+        }
+        UNLOCK (&frame->lock);
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0)
+                AFR_STACK_UNWIND (finodelk, frame, local->op_ret,
+                                  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, dict_t *xdata)
+{
+        afr_private_t *priv = NULL;
+        afr_local_t *local  = NULL;
+        int i = 0;
+        int32_t call_count = 0;
+        int32_t op_errno = ENOMEM;
+
+        priv = this->private;
+
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
+
+        call_count = local->call_count;
+	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, xdata);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+	return 0;
+out:
+	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, dict_t *xdata)
+{
+        afr_local_t *local = NULL;
+        int call_count = -1;
+
+        local = frame->local;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == 0)
+                        local->op_ret = 0;
+
+                local->op_errno = op_errno;
+        }
+        UNLOCK (&frame->lock);
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0)
+                AFR_STACK_UNWIND (entrylk, frame, local->op_ret,
+                                  local->op_errno, xdata);
+
+        return 0;
+}
+
+
+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 i = 0;
+        int32_t call_count = 0;
+        int32_t op_errno = 0;
+
+        priv = this->private;
+
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
+
+        call_count = local->call_count;
+	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, xdata);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+	return 0;
+out:
+	AFR_STACK_UNWIND (entrylk, frame, -1, op_errno, NULL);
+
+        return 0;
+}
+
+
+
+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;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == 0)
+                        local->op_ret = 0;
+
+                local->op_errno = op_errno;
+        }
+        UNLOCK (&frame->lock);
+
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0)
+                AFR_STACK_UNWIND (fentrylk, frame, local->op_ret,
+                                  local->op_errno, xdata);
+
+        return 0;
+}
+
+
+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 i = 0;
+        int32_t call_count = 0;
+        int32_t op_errno = ENOMEM;
+
+        priv = this->private;
+
+        local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
+
+        call_count = local->call_count;
+	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, xdata);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+	return 0;
+out:
+	AFR_STACK_UNWIND (fentrylk, frame, -1, op_errno, NULL);
+
+        return 0;
+}
+
+
+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_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->xdata_rsp);
+
+        return 0;
+}
+
+
+int
+afr_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+        afr_local_t   *  local       = NULL;
+	afr_private_t   *priv        = NULL;
+        int              i           = 0;
+        int              call_count = 0;
+        int32_t          op_errno    = ENOMEM;
+
+	priv = this->private;
+
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
+
+        call_count = local->call_count;
+	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_statfs_cbk,
+                                    priv->children[i],
+                                    priv->children[i]->fops->statfs,
+                                    loc, xdata);
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+	return 0;
+out:
+	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,
+                   dict_t *xdata)
+{
+        afr_local_t * local = NULL;
+        int call_count = -1;
+
+        local = frame->local;
+        call_count = afr_frame_return (frame);
+
+        if (call_count == 0)
+                AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
+                                  lock, xdata);
+
+        return 0;
+}
+
+
+int32_t
+afr_lk_unlock (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t   * local = NULL;
+        afr_private_t * priv  = NULL;
+        int i = 0;
+        int call_count = 0;
+
+        local = frame->local;
+        priv  = this->private;
+
+        call_count = afr_locked_nodes_count (local->cont.lk.locked_nodes,
+                                             priv->child_count);
+
+        if (call_count == 0) {
+                AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
+                                  &local->cont.lk.ret_flock, NULL);
+                return 0;
+        }
+
+        local->call_count = call_count;
+
+        local->cont.lk.user_flock.l_type = F_UNLCK;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->cont.lk.locked_nodes[i]) {
+                        STACK_WIND (frame, afr_lk_unlock_cbk,
+                                    priv->children[i],
+                                    priv->children[i]->fops->lk,
+                                    local->fd, F_SETLK,
+                                    &local->cont.lk.user_flock, NULL);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+        return 0;
+}
+
+
+int32_t
+afr_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+            int32_t op_ret, int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+        int child_index = -1;
+/*        int            ret  = 0; */
+
+
+        local = frame->local;
+        priv  = this->private;
+
+        child_index = (long) cookie;
+
+        if (!child_went_down (op_ret, op_errno) && (op_ret == -1)) {
+                local->op_ret   = -1;
+                local->op_errno = op_errno;
+
+                afr_lk_unlock (frame, this);
+                return 0;
+        }
+
+        if (op_ret == 0) {
+                local->op_ret        = 0;
+                local->op_errno      = 0;
+                local->cont.lk.locked_nodes[child_index] = 1;
+                local->cont.lk.ret_flock = *lock;
+        }
+
+        child_index++;
+
+        if (child_index < priv->child_count) {
+                STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) child_index,
+                                   priv->children[child_index],
+                                   priv->children[child_index]->fops->lk,
+                                   local->fd, local->cont.lk.cmd,
+                                   &local->cont.lk.user_flock, xdata);
+        } else if (local->op_ret == -1) {
+                /* all nodes have gone down */
+
+                AFR_STACK_UNWIND (lk, frame, -1, ENOTCONN,
+                                  &local->cont.lk.ret_flock, NULL);
+        } else {
+                AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
+                                  &local->cont.lk.ret_flock, NULL);
+        }
+
+        return 0;
+}
+
+
+int
+afr_lk (call_frame_t *frame, xlator_t *this,
+        fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
+{
+        afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        int i = 0;
+        int32_t op_errno = ENOMEM;
+
+        priv = this->private;
+
+        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);
+
+        if (!local->cont.lk.locked_nodes) {
+                op_errno = ENOMEM;
+                goto out;
+        }
+
+        local->fd            = fd_ref (fd);
+        local->cont.lk.cmd   = cmd;
+        local->cont.lk.user_flock = *flock;
+        local->cont.lk.ret_flock = *flock;
+
+        STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) 0,
+                           priv->children[i],
+                           priv->children[i]->fops->lk,
+                           fd, cmd, flock, xdata);
+
+	return 0;
+out:
+	AFR_STACK_UNWIND (lk, frame, -1, op_errno, NULL, NULL);
+
+        return 0;
+}
+
+int
+afr_forget (xlator_t *this, inode_t *inode)
+{
+        return 0;
+}
+
+int
+afr_priv_dump (xlator_t *this)
+{
+        afr_private_t *priv = NULL;
+        char  key_prefix[GF_DUMP_MAX_BUF_LEN];
+        char  key[GF_DUMP_MAX_BUF_LEN];
+        int   i = 0;
+
+
+        GF_ASSERT (this);
+        priv = this->private;
+
+        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_write("child_count", "%u", priv->child_count);
+        for (i = 0; i < priv->child_count; i++) {
+                sprintf (key, "child_up[%d]", i);
+                gf_proc_dump_write(key, "%d", priv->child_up[i]);
+                sprintf (key, "pending_key[%d]", i);
+                gf_proc_dump_write(key, "%s", priv->pending_key[i]);
+        }
+        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;
+}
+
+
+/**
+ * find_child_index - find the child's index in the array of subvolumes
+ * @this: AFR
+ * @child: child
+ */
+
+static int
+find_child_index (xlator_t *this, xlator_t *child)
+{
+        afr_private_t *priv = NULL;
+        int i = -1;
+
+        priv = this->private;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if ((xlator_t *) child == priv->children[i])
+                        break;
+        }
+
+        return i;
+}
+
+int32_t
+afr_notify (xlator_t *this, int32_t event,
+            void *data, void *data2)
+{
+        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;
+
+        /*
+         * 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:
+                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;
+                        }
+
+                        priv->last_event[idx] = event;
+                }
+                UNLOCK (&priv->lock);
+
+                break;
+
+        case GF_EVENT_CHILD_DOWN:
+                LOCK (&priv->lock);
+                {
+                        /*
+                         * If a brick is down when we start, we'll get a
+                         * CHILD_DOWN to indicate its initial state.  There
+                         * was never a CHILD_UP in this case, so if we
+                         * increment "down_count" the difference between than
+                         * and "up_count" will no longer be the number of
+                         * children that are currently up.  This has serious
+                         * implications e.g. for quorum enforcement, so we
+                         * don't increment these values unless the event
+                         * represents an actual state transition between "up"
+                         * (value = 1) and anything else.
+                         */
+                        if (priv->child_up[idx] == 1) {
+                                priv->down_count++;
+				priv->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);
+
+                break;
+
+        case GF_EVENT_CHILD_CONNECTING:
+                LOCK (&priv->lock);
+                {
+                        priv->last_event[idx] = event;
+                }
+                UNLOCK (&priv->lock);
+
+                break;
+
+        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;
+        }
+
+        /* 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;
+
+        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);
+                {
+                        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);
+        }
+
+        ret = 0;
+        if (propagate)
+                ret = default_notify (this, event, data);
+
+        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;
+
+        local->transaction.pre_op = GF_CALLOC (sizeof (*local->transaction.pre_op),
+                                               priv->child_count,
+                                               gf_afr_mt_char);
+        if (!local->transaction.pre_op)
+                goto out;
+
+        local->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;
+                        }
+                }
+        }
+
+out:
+        return _gf_false;
+}
+
+void
+afr_priv_destroy (afr_private_t *priv)
+{
+        int            i           = 0;
+
+        if (!priv)
+                goto out;
+        inode_unref (priv->root_inode);
+        GF_FREE (priv->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);
+
+        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 b48488526..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) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -24,6 +15,7 @@
 #include <sys/time.h>
 #include <stdlib.h>
 #include <signal.h>
+#include <string.h>
 
 #ifndef _CONFIG_H
 #define _CONFIG_H
@@ -45,556 +37,392 @@
 #include "checksum.h"
 
 #include "afr.h"
-#include "afr-self-heal.h"
-
-
-int
-afr_examine_dir_sh_unwind (call_frame_t *frame, xlator_t *this)
-{
-        afr_local_t *local  = NULL;
-        afr_self_heal_t *sh = NULL;
-
-        local = frame->local;
-        sh    = &local->self_heal;
-
-        afr_set_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)
-{
-        int ret = _gf_false;
-        int i = 0;
-
-        uint32_t cksum;
-
-        cksum = checksum[0];
-
-        while (i < child_count) {
-                if (cksum != checksum[i]) {
-                        ret = _gf_true;
-                        break;
-                }
-
-                cksum = checksum[i];
-                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;
-
-        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;
+        afr_local_t   *local             = NULL;
+        int            call_count        = -1;
+        int32_t        child_index       = 0;
+	afr_fd_ctx_t  *fd_ctx = NULL;
 
-        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)) {
-
-                                sh->need_entry_self_heal  = _gf_true;
-                                sh->forced_merge          = _gf_true;
-                                sh->mode                  = local->fd->inode->st_mode;
-                                sh->background            = _gf_false;
-                                sh->unwind                = afr_examine_dir_sh_unwind;
-
-                                gf_log (this->name, GF_LOG_DEBUG,
-                                        "checksums of directory %s differ,"
-                                        " triggering forced merge",
-                                        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 = CALLOC (priv->child_count,
-                                               sizeof (*local->cont.opendir.checksum));
+        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_local_t * local  = NULL;
-
-	int call_count = -1;
-
-	LOCK (&frame->lock);
-	{
-		local = frame->local;
-
-		if (op_ret >= 0)
-			local->op_ret = op_ret;
-
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
+#define BACKEND_D_OFF_BITS 63
+#define PRESENT_D_OFF_BITS 63
 
-	if (call_count == 0) {
-                if ((local->op_ret == 0) &&
-                    !afr_is_opendir_done (this, fd->inode)) {
+#define ONE 1ULL
+#define MASK (~0ULL)
+#define PRESENT_MASK (MASK >> (64 - PRESENT_D_OFF_BITS))
+#define BACKEND_MASK (MASK >> (64 - BACKEND_D_OFF_BITS))
 
-                        /*
-                         * 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).
-                         */
+#define TOP_BIT (ONE << (PRESENT_D_OFF_BITS - 1))
+#define SHIFT_BITS (max (0, (BACKEND_D_OFF_BITS - PRESENT_D_OFF_BITS + 1)))
 
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "reading contents of directory %s looking for mismatch",
-                                local->loc.path);
-
-                        afr_examine_dir (frame, this);
+static uint64_t
+afr_bits_for (uint64_t num)
+{
+	uint64_t bits = 0, ctrl = 1;
 
-                } else {
-                        AFR_STACK_UNWIND (opendir, frame, local->op_ret,
-                                          local->op_errno, local->fd);
-                }
+	while (ctrl < num) {
+		ctrl *= 2;
+		bits ++;
 	}
 
-	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);
+        max_bits = afr_bits_for (max);
 
-	frame->local = local;
-	local->fd    = fd_ref (fd);
+        hi_mask = ~(PRESENT_MASK >> (max_bits + 1));
 
-	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;
-		}
-	}
+        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);
+        }
 
-	op_ret = 0;
 out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (opendir, frame, op_ret, op_errno, fd);
-	}
+        if (y_p)
+                *y_p = y;
 
-	return 0;
+        return 0;
 }
 
 
-/**
- * Common algorithm for directory read calls:
- * 
- * - Try the fop on the first child that is up
- * - if we have failed due to ENOTCONN:
- *     try the next child
- *
- * Applicable to: readdir
- */
-
-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)
+int
+afr_deitransform (xlator_t *this, uint64_t y, int *subvol_p,
+                  uint64_t *x_p)
 {
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-	xlator_t **     children = NULL;
-
-        gf_dirent_t * entry = NULL;
-        gf_dirent_t * tmp   = NULL;
+        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;
+	}
 
-        int child_index = -1;
+        if (y & TOP_BIT) {
+                /* HUGE d_off */
+                max_bits = afr_bits_for (max);
+                off_mask = (MASK << max_bits);
+                host_mask = ~(off_mask);
 
-	priv     = this->private;
-	children = priv->children;
+                x = ((y & ~TOP_BIT) & off_mask) << SHIFT_BITS;
 
-	local = frame->local;
+                cnt = y & host_mask;
+	} else {
+                /* small d_off */
+                cnt = y % max;
+                x = y / max;
+        }
 
-        child_index = (long) cookie;
+out:
+        subvol = cnt;
 
-	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);
-                                FREE (entry);
-                        }
-                }
-    	}
+        if (subvol_p)
+                *subvol_p = subvol;
 
-        AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, entries);
+        if (x_p)
+                *x_p = x;
 
-	return 0;
+        return 0;
 }
 
 
-int32_t
-afr_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                  int32_t op_ret, int32_t op_errno, gf_dirent_t *entries)
+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;
-        xlator_t **     children = NULL;
-        ino_t           inum = 0;
-
-        gf_dirent_t * entry = NULL;
-        gf_dirent_t * tmp   = NULL;
-
-        int child_index = -1;
+	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;
+                }
 
-        priv     = this->private;
-        children = priv->children;
+		list_del_init (&entry->list);
+		afr_itransform (THIS, subvol, entry->d_off, &entry->d_off);
+		list_add_tail (&entry->list, &entries->list);
 
-        local = frame->local;
+		if (entry->inode) {
+			gen = 0;
+			afr_inode_read_subvol_get (entry->inode, THIS,
+						   data_readable,
+						   metadata_readable, &gen);
 
-        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) {
-                        inum = afr_itransform (entry->d_ino, priv->child_count,
-                                               child_index);
-                        entry->d_ino = inum;
-                        inum  = afr_itransform (entry->d_stat.st_ino,
-                                                priv->child_count, child_index);
-                        entry->d_stat.st_ino = inum;
-
-                        if ((local->fd->inode == local->fd->inode->table->root)
-                            && !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR)) {
-                                list_del_init (&entry->list);
-                                FREE (entry);
-                        }
-                }
+				inode_unref (entry->inode);
+				entry->inode = NULL;
+			}
+		}
         }
-
-        AFR_STACK_UNWIND (readdirp, frame, op_ret, op_errno, entries);
-
-        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)
+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;
-	xlator_t **     children   = NULL;
-	int             call_child = 0;
-	afr_local_t     *local     = NULL;
-
-	int ret = -1;
+        afr_local_t *local = NULL;
+	gf_dirent_t  entries;
 
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
+	INIT_LIST_HEAD (&entries.list);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        local = frame->local;
 
-	priv     = this->private;
-	children = priv->children;
+        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;
 
-	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_child = afr_first_up_child (priv);
-	if (call_child == -1) {
-		op_errno = ENOTCONN;
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no child is up");
-		goto out;
+		afr_read_txn_continue (frame, this, (long) cookie);
+		return 0;
 	}
 
-        local->fd                  = fd_ref (fd);
-        local->cont.readdir.size   = size;
-        local->cont.readdir.offset = offset;
+	if (op_ret >= 0)
+		afr_readdir_transform_entries (subvol_entries, (long) cookie,
+					       &entries, local->fd);
 
-        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);
+        AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, &entries, xdata);
 
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (readdir, frame, op_ret, op_errno, NULL);
-	}
-	return 0;
-}
-
-
-int32_t
-afr_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-             off_t offset)
-{
-        afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIR);
         return 0;
 }
 
 
-int32_t
-afr_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-              off_t offset)
-{
-        afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIRP);
-        return 0;
-}
-
-int32_t
-afr_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)
+int
+afr_readdir_wind (call_frame_t *frame, xlator_t *this, int subvol)
 {
-	afr_private_t * priv     = NULL;
-	afr_local_t *   local    = NULL;
-	xlator_t **     children = NULL;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
 
-	int unwind     = 1;
-	int last_tried = -1;
-	int this_try = -1;
+	priv = this->private;
+	local = frame->local;
 
-	priv     = this->private;
-	children = priv->children;
+	if (subvol == -1) {
+		AFR_STACK_UNWIND (readdir, frame, local->op_ret,
+				  local->op_errno, 0, 0);
+		return 0;
+	}
 
-	local = frame->local;
+        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;
+}
 
-	if (op_ret == -1) {
-		last_tried = local->cont.getdents.last_tried;
 
-		if (all_tried (last_tried, priv->child_count)) {
-			goto out;
-		}
+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;
 
-		this_try = ++local->cont.getdents.last_tried;
-		unwind = 0;
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-		STACK_WIND (frame, afr_getdents_cbk,
-			    children[this_try],
-			    children[this_try]->fops->getdents,
-			    local->fd, local->cont.getdents.size,
-			    local->cont.getdents.offset, local->cont.getdents.flag);
+	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);
 	}
 
+        return 0;
 out:
-	if (unwind) {
-		AFR_STACK_UNWIND (getdents, frame, op_ret, op_errno,
-                                  entry, count);
-	}
-
-	return 0;
+        AFR_STACK_UNWIND (readdir, frame, -1, op_errno, NULL, NULL);
+        return 0;
 }
 
 
 int32_t
-afr_getdents (call_frame_t *frame, xlator_t *this,
-	      fd_t *fd, size_t size, off_t offset, int32_t flag)
+afr_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+             off_t offset, dict_t *xdata)
 {
-	afr_private_t * priv       = NULL;
-	xlator_t **     children   = NULL;
-	int             call_child = 0;
-	afr_local_t     *local     = NULL;
+	afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIR, xdata);
 
-	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);
-
-	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;
-	}
+        return 0;
+}
 
-	local->cont.getdents.last_tried = call_child;
 
-	local->fd                   = fd_ref (fd);
+int32_t
+afr_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+              off_t offset, dict_t *dict)
+{
+        afr_do_readdir (frame, this, fd, size, offset, GF_FOP_READDIRP, dict);
 
-	local->cont.getdents.size   = size;
-	local->cont.getdents.offset = offset;
-	local->cont.getdents.flag   = flag;
-	
-	frame->local = local;
+        return 0;
+}
 
-	STACK_WIND (frame, afr_getdents_cbk,
-		    children[call_child], children[call_child]->fops->getdents,
-		    fd, size, offset, flag);
 
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (getdents, frame, op_ret, op_errno,
-                                  NULL, 0);
-	}
+int32_t
+afr_releasedir (xlator_t *this, fd_t *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 98ce1ca16..09456d159 100644
--- a/xlators/cluster/afr/src/afr-dir-read.h
+++ b/xlators/cluster/afr/src/afr-dir-read.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __DIR_READ_H__
@@ -23,29 +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_closedir (call_frame_t *frame, xlator_t *this,
-	      fd_t *fd);
+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 1d7d3f195..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) 2007-2009 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -43,2108 +34,1446 @@
 #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 = strdup (child->path);
-	parent->path   = strdup (dirname (tmp));
-	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_private_t * priv     = NULL;
-	afr_local_t  *local = NULL;
-        struct stat  *unwind_buf = NULL;
-
-        priv  = this->private;
-	local = frame->local;
+        int     ret = -1;
+        char    *child_path = NULL;
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
-
-	if (main_frame) {
-                if (local->cont.create.read_child_buf.st_ino) {
-                        unwind_buf = &local->cont.create.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.create.buf;
-                }
-
-                unwind_buf->st_ino = local->cont.create.ino;
-                unwind_buf->st_dev = local->cont.create.gen;
-
-                local->cont.create.preparent.st_ino  = local->cont.create.parent_ino;
-                local->cont.create.postparent.st_ino = local->cont.create.parent_ino;
-
-		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);
+        if (!child->parent) {
+                if (op_errno)
+                        *op_errno = EINVAL;
+                goto out;
         }
-        
-	return 0;
-}
 
+        child_path = gf_strdup (child->path);
+        if (!child_path) {
+                if (op_errno)
+                        *op_errno = ENOMEM;
+                goto out;
+        }
 
-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 stat *buf,
-                     struct stat *preparent, struct stat *postparent)
-{
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-        uint64_t      ctx;
-        afr_fd_ctx_t *fd_ctx;
-
-        int ret = 0;
-
-	int call_count = -1;
-	int child_index = -1;
-
-	local = frame->local;
-	priv  = this->private;
-
-	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;
-
-                        ret = afr_fd_ctx_set (this, fd);
-
-                        if (ret < 0) {
-                                gf_log (this->name, GF_LOG_DEBUG,
-                                        "could not set ctx on fd=%p", fd);
-
-                                local->op_ret   = -1;
-                                local->op_errno = -ret;
-                        }
-
-                        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);
-                                local->op_ret   = -1;
-                                local->op_errno = -ret;
-                        }
-
-                        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                        fd_ctx->opened_on[child_index] = 1;
-                        fd_ctx->flags                  = local->cont.create.flags;
-
-                        if (local->success_count == 0) {
-				local->cont.create.buf        = *buf;
-
-                                local->cont.create.ino =
-                                        afr_itransform (buf->st_ino,
-                                                        priv->child_count,
-                                                        child_index);
-                                local->cont.create.gen = buf->st_dev;
-
-                                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->st_ino,
-                                                        priv->child_count,
-                                                        local->first_up_child);
-                                local->cont.create.gen = buf->st_dev;
-                        }
-                        
-                        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++;
-		}
-
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+        parent->path = gf_strdup (dirname (child_path));
+        if (!parent->path) {
+                if (op_errno)
+                        *op_errno = ENOMEM;
+                goto out;
+        }
 
-	call_count = afr_frame_return (frame);
+        parent->inode = inode_ref (child->parent);
+	uuid_copy (parent->gfid, child->pargfid);
 
-	if (call_count == 0) {
-		local->transaction.unwind (frame, this);
+        ret = 0;
+out:
+        GF_FREE (child_path);
 
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
+        return ret;
 }
 
 
-int
-afr_create_wind (call_frame_t *frame, xlator_t *this)
+static void
+__afr_dir_write_finalize (call_frame_t *frame, xlator_t *this)
 {
 	afr_local_t *local = NULL;
 	afr_private_t *priv = NULL;
-
-	int call_count = -1;
+	int inode_read_subvol = -1;
+	int parent_read_subvol = -1;
+	int parent2_read_subvol = -1;
 	int i = 0;
 
 	local = frame->local;
 	priv = this->private;
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+	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;
+		}
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
+		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;
+		}
 
-	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);
-			if (!--call_count)
-				break;
+		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 (i == parent_read_subvol) {
+			local->cont.dir_fop.preparent =
+				local->replies[i].preparent;
+			local->cont.dir_fop.postparent =
+				local->replies[i].postparent;
+		}
+
+		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_done (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;
-
-	local = frame->local;
-
-	local->transaction.unwind (frame, this);
-
-	AFR_STACK_DESTROY (frame);
+        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;
+	}
 
-	return 0;
+        return;
 }
 
 
-int
-afr_create (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, int32_t flags, mode_t mode, fd_t *fd)
+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_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t  * transaction_frame = NULL;
-
-	int ret = -1;
-
-	int op_ret   = -1;
-	int op_errno = 0;
+        afr_local_t *local = NULL;
+        int child_index = (long) cookie;
+        int call_count = -1;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        local = frame->local;
 
-	priv = this->private;
-
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
+	LOCK (&frame->lock);
+	{
+		__afr_dir_write_fill (frame, this, child_index, op_ret,
+				      op_errno, buf, preparent, postparent,
+				      preparent2, postparent2, xdata);
 	}
+	UNLOCK (&frame->lock);
+        call_count = afr_frame_return (frame);
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
-		goto out;
-	}
+        if (call_count == 0) {
+		__afr_dir_write_finalize (frame, this);
 
-	transaction_frame->local = local;
+		if (afr_txn_nothing_failed (frame, this))
+			local->transaction.unwind (frame, this);
 
-	loc_copy (&local->loc, loc);
+		afr_mark_entry_pending_changelog (frame, this);
 
-        LOCK (&priv->read_child_lock);
-        {
-                local->read_child_index = (++priv->read_child_rr) 
-                        % (priv->child_count);
+                local->transaction.resume (frame, this);
         }
-        UNLOCK (&priv->read_child_lock);
 
-	local->cont.create.flags = flags;
-	local->cont.create.mode  = mode;
-	local->cont.create.fd    = fd_ref (fd);
-
-        if (loc->parent)
-                local->cont.create.parent_ino = loc->parent->ino;
+        return 0;
+}
 
-	local->transaction.fop    = afr_create_wind;
-	local->transaction.done   = afr_create_done;
-	local->transaction.unwind = afr_create_unwind;
 
-	afr_build_parent_loc (&local->transaction.parent_loc, loc);
+int
+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)
+{
+        int call_count = 0;
 
-	local->transaction.main_frame = frame;
-	local->transaction.basename = AFR_BASENAME (loc->path);
+        call_count = afr_frame_return (frame);
 
-	afr_transaction (transaction_frame, this, AFR_ENTRY_TRANSACTION);
+        if (call_count == 0)
+                AFR_STACK_DESTROY (frame);
 
-	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);
-	}
-
-	return 0;
+        return 0;
 }
 
-/* }}} */
-
-/* {{{ mknod */
 
-int
-afr_mknod_unwind (call_frame_t *frame, xlator_t *this)
+void
+afr_mark_new_entry_changelog (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
+        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;
 
-        struct stat *unwind_buf = NULL;
+        changelog = afr_matrix_create (priv->child_count, AFR_NUM_CHANGE_LOGS);
+        if (!changelog)
+                goto out;
 
-	local = frame->local;
+        xattr = dict_new ();
+	if (!xattr)
+		goto out;
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
+	idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
+
+	pending = alloca0 (priv->child_count);
+
+	for (i = 0; i < priv->child_count; i++) {
+		if (local->transaction.pre_op[i] &&
+		    !local->transaction.failed_subvols[i]) {
+			call_count ++;
+			continue;
 		}
-		local->transaction.main_frame = NULL;
+
+		changelog[i][idx] = hton32(1);
+		pending[i] = 1;
 	}
-	UNLOCK (&frame->lock);
 
-	if (main_frame) {
-                if (local->cont.mknod.read_child_buf.st_ino) {
-                        unwind_buf = &local->cont.mknod.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.mknod.buf;
-                }
+        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);
+
 
-                unwind_buf->st_ino = local->cont.mknod.ino;
-                unwind_buf->st_dev = local->cont.mknod.gen;
+	afr_set_pending_dict (priv, xattr, changelog);
 
-                local->cont.mknod.preparent.st_ino  = local->cont.mknod.parent_ino;
-                local->cont.mknod.postparent.st_ino = local->cont.mknod.parent_ino;
+        new_local->call_count = call_count;
 
-		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);
+        for (i = 0; i < priv->child_count; i++) {
+		if (pending[i])
+                        continue;
+
+                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 stat *buf, struct stat *preparent,
-                    struct stat *postparent)
+void
+afr_mark_entry_pending_changelog (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int call_count = -1;
-	int child_index = -1;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+	int pre_op_count = 0;
+	int failed_count = 0;
 
-	local = frame->local;
-	priv = this->private;
-
-	child_index = (long) cookie;
+        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) {
-			local->op_ret = op_ret;
-
-			if (local->success_count == 0){
-				local->cont.mknod.buf   = *buf;
-				local->cont.mknod.ino   = 
-					afr_itransform (buf->st_ino,
-							priv->child_count,
-							child_index);
-                                local->cont.mknod.gen   = buf->st_dev;
-
-                                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->st_ino,
-                                                        priv->child_count,
-                                                        local->first_up_child);
-                                local->cont.mknod.gen = buf->st_dev;
-                        }
-
-                        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);
-			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)
+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;
-
-	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 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(create,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);
+	local->inode = inode_ref (loc->inode);
+	local->parent = inode_ref (loc->parent);
 
-        LOCK (&priv->read_child_lock);
-        {
-                local->read_child_index = (++priv->read_child_rr) 
-                        % (priv->child_count);
-        }
-        UNLOCK (&priv->read_child_lock);
+        local->op                = GF_FOP_CREATE;
+        local->cont.create.flags = flags;
+        local->cont.create.mode  = mode;
+        local->cont.create.fd    = fd_ref (fd);
+        local->umask  = umask;
 
-	local->cont.mknod.mode  = mode;
-	local->cont.mknod.dev   = dev;
+        if (xdata)
+                local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-        if (loc->parent)
-                local->cont.mknod.parent_ino = loc->parent->ino;
-
-	local->transaction.fop    = afr_mknod_wind;
-	local->transaction.done   = afr_mknod_done;
-	local->transaction.unwind = afr_mknod_unwind;
-
-	afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
-	local->transaction.main_frame = frame;
-	local->transaction.basename = AFR_BASENAME (loc->path);
+	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 stat *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);
-
-	if (main_frame) {
-                if (local->cont.mkdir.read_child_buf.st_ino) {
-                        unwind_buf = &local->cont.mkdir.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.mkdir.buf;
-                }
+        local = frame->local;
 
-                unwind_buf->st_ino = local->cont.mkdir.ino;
-                unwind_buf->st_dev = local->cont.mkdir.gen;
-
-                local->cont.mkdir.preparent.st_ino  = local->cont.mkdir.parent_ino;
-                local->cont.mkdir.postparent.st_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 stat *buf, struct stat *preparent,
-                    struct stat *postparent)
+                    struct iatt *buf, 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 = -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.mkdir.buf   = *buf;
-
-                                local->cont.mkdir.ino =
-                                        afr_itransform (buf->st_ino,
-                                                        priv->child_count,
-                                                        child_index);
-                                local->cont.mkdir.gen = buf->st_dev;
-
-                                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.mkdir.ino =
-                                        afr_itransform (buf->st_ino,
-                                                        priv->child_count,
-                                                        local->first_up_child);
-                                local->cont.mkdir.gen = buf->st_dev;
-                        }
-                        
-                        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_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_mkdir_wind (call_frame_t *frame, xlator_t *this)
+afr_mknod_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);
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
+        local = frame->local;
+        priv  = this->private;
 
-	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);
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+	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;
 }
 
-
 int
-afr_mkdir_done (call_frame_t *frame, xlator_t *this)
+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_local_t * local = 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;
 
-	local = frame->local;
+        priv = this->private;
 
-	local->transaction.unwind (frame, this);
+        QUORUM_CHECK(mknod,out);
 
-	AFR_STACK_DESTROY (frame);
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto out;
 
-	return 0;
-}
-
-
-int
-afr_mkdir (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, mode_t mode)
-{
-	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->op               = GF_FOP_MKNOD;
+        local->cont.mknod.mode  = mode;
+        local->cont.mknod.dev   = dev;
+        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->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;
-	}
-
-	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.mkdir.mode  = mode;
 
-        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 (mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+			  NULL);
+        return 0;
 }
 
 /* }}} */
 
-/* {{{ link */
+/* {{{ mkdir */
 
 
 int
-afr_link_unwind (call_frame_t *frame, xlator_t *this)
+afr_mkdir_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
-
-        struct stat *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.link.read_child_buf.st_ino) {
-                        unwind_buf = &local->cont.link.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.link.buf;
-                }
-
-		unwind_buf->st_ino = local->cont.link.ino;
-
-                local->cont.link.preparent.st_ino  = local->cont.link.parent_ino;
-                local->cont.link.postparent.st_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 (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_link_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		   int32_t op_ret, int32_t op_errno, inode_t *inode,
-                   struct stat *buf, struct stat *preparent,
-                   struct stat *postparent)
+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;
-	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_mkdir_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 = frame->local;
+        priv  = this->private;
 
-	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_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_link_done (call_frame_t *frame, xlator_t *this)
+afr_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+	   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);
+        QUORUM_CHECK(mkdir,out);
 
-	return 0;
-}
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto out;
 
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+		goto out;
 
-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;
-
-	int ret = -1;
+        loc_copy (&local->loc, loc);
+	local->inode = inode_ref (loc->inode);
+	local->parent = inode_ref (loc->parent);
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        local->cont.mkdir.mode  = mode;
+        local->umask = umask;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        if (xdata)
+                local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	priv = this->private;
-
-	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_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;
-	}
-
-	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 = oldloc->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 (mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+			  NULL);
+        return 0;
 }
 
 /* }}} */
 
-/* {{{ symlink */
+/* {{{ link */
 
 
 int
-afr_symlink_unwind (call_frame_t *frame, xlator_t *this)
+afr_link_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
-        
-        struct stat *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.st_ino) {
-                        unwind_buf = &local->cont.symlink.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.symlink.buf;
-                }
-
-                unwind_buf->st_ino = local->cont.symlink.ino;
-                unwind_buf->st_dev = local->cont.symlink.gen;
-
-                local->cont.symlink.preparent.st_ino  = local->cont.symlink.parent_ino;
-                local->cont.symlink.postparent.st_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 (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_symlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		      int32_t op_ret, int32_t op_errno, inode_t *inode,
-                      struct stat *buf, struct stat *preparent,
-                      struct stat *postparent)
+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, 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->st_ino, priv->child_count,
-							child_index);
-                                local->cont.symlink.gen = buf->st_dev;
-
-                                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->st_ino,
-                                                        priv->child_count,
-                                                        local->first_up_child);
-                                local->cont.symlink.gen = buf->st_dev;
-                        }
-
-                        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_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;
+        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_symlink_wind_cbk,
-					   (void *) (long) i,	
-					   priv->children[i], 
-					   priv->children[i]->fops->symlink,
-					   local->cont.symlink.linkpath,
-					   &local->loc);
-
-			if (!--call_count)
-				break;
+        local = frame->local;
+        priv  = this->private;
 
-		}
-	}
-	
-	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_symlink_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);
-	
-	return 0;
-}
+        QUORUM_CHECK(link,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)
-{
-	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,    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, 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 = strdup (linkpath);
 
-        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 (link, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+			  NULL);
+        return 0;
 }
 
 /* }}} */
 
-/* {{{ rename */
+/* {{{ symlink */
+
 
 int
-afr_rename_unwind (call_frame_t *frame, xlator_t *this)
+afr_symlink_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
-
-        struct stat *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.rename.read_child_buf.st_ino) {
-                        unwind_buf = &local->cont.rename.read_child_buf;
-                } else {
-                        unwind_buf = &local->cont.rename.buf;
-                }
-
-		unwind_buf->st_ino = local->cont.rename.ino;
-
-                local->cont.rename.preoldparent.st_ino  = local->cont.rename.oldparent_ino;
-                local->cont.rename.postoldparent.st_ino = local->cont.rename.oldparent_ino;
-                local->cont.rename.prenewparent.st_ino  = local->cont.rename.newparent_ino;
-                local->cont.rename.postnewparent.st_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 (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_rename_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		     int32_t op_ret, int32_t op_errno, struct stat *buf,
-                     struct stat *preoldparent, struct stat *postoldparent,
-                     struct stat *prenewparent, struct stat *postnewparent)
+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, 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) {
-                        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,
+				    preparent, postparent, NULL, NULL, xdata);
 }
 
 
-int32_t
-afr_rename_wind (call_frame_t *frame, xlator_t *this)
+int
+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_rename_wind_cbk, 
-					   (void *) (long) i,	
-					   priv->children[i], 
-					   priv->children[i]->fops->rename,
-					   &local->loc,
-					   &local->newloc);
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
+
+        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_rename_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_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;
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+		goto out;
 
-	int ret = -1;
+        loc_copy (&local->loc, loc);
+	local->inode = inode_ref (loc->inode);
+	local->parent = inode_ref (loc->parent);
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        local->cont.symlink.linkpath = gf_strdup (linkpath);
+        local->umask = umask;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        if (xdata)
+                local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	priv = this->private;
-
-	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,    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 (symlink, frame, -1, op_errno, NULL, NULL, NULL,
+			  NULL, NULL);
+        return 0;
 }
 
 /* }}} */
 
-/* {{{ unlink */
+/* {{{ rename */
 
 int
-afr_unlink_unwind (call_frame_t *frame, xlator_t *this)
+afr_rename_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.st_ino  = local->cont.unlink.parent_ino;
-                local->cont.unlink.postparent.st_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 (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_unlink_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		     int32_t op_ret, int32_t op_errno, struct stat *preparent,
-                     struct stat *postparent)
+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,
+                     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;
-        int read_child  = 0;
-
-	local = frame->local;
-	priv  = this->private;
-
-        read_child = afr_read_child (this, local->loc.inode);
-
-	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++;
-
-			if ((local->success_count == priv->wait_count)
-                            && local->read_child_returned) {
-				need_unwind = 1;
-			}
-		}
-
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (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_unlink_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;
-	}
+	priv = this->private;
 
-	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_rename_wind_cbk, (void *) (long) subvol,
+			   priv->children[subvol],
+			   priv->children[subvol]->fops->rename,
+			   &local->loc, &local->newloc, local->xdata_req);
+        return 0;
 }
 
 
-int32_t
-afr_unlink_done (call_frame_t *frame, xlator_t *this)
+int
+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);
+        priv = this->private;
 
-	AFR_STACK_DESTROY (frame);
-	
-	return 0;
-}
+        QUORUM_CHECK(rename,out);
 
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                op_errno = ENOMEM;
 
-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;
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+		goto out;
 
-	int ret = -1;
+        loc_copy (&local->loc,    oldloc);
+        loc_copy (&local->newloc, newloc);
 
-	int op_ret   = -1;
-	int op_errno = 0;
+	local->inode = inode_ref (oldloc->inode);
+	local->parent = inode_ref (oldloc->parent);
+	local->parent2 = inode_ref (newloc->parent);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        if (xdata)
+                local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	priv = this->private;
-
-	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, 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 (rename, frame, -1, op_errno, NULL, NULL, NULL, NULL,
+			  NULL, NULL);
+        return 0;
 }
 
 /* }}} */
 
-/* {{{ rmdir */
-
-
+/* {{{ unlink */
 
 int
-afr_rmdir_unwind (call_frame_t *frame, xlator_t *this)
+afr_unlink_unwind (call_frame_t *frame, xlator_t *this)
 {
-	call_frame_t *main_frame = NULL;
-	afr_local_t  *local = NULL;
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame) {
-			main_frame = local->transaction.main_frame;
-		}
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
-
-	if (main_frame) {
-                local->cont.rmdir.preparent.st_ino  = local->cont.rmdir.parent_ino;
-                local->cont.rmdir.postparent.st_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);
-        }
+	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_rmdir_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		    int32_t op_ret, int32_t op_errno, struct stat *preparent,
-                    struct stat *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;
-	int need_unwind = 0;
-        int read_child  = 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 != 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++;
-
-			if ((local->success_count == priv->wait_count)
-                            && local->read_child_returned)
-				need_unwind = 1;
-		}
-
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (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);
 }
 
 
 int
-afr_rmdir_wind (call_frame_t *frame, xlator_t *this)
+afr_unlink_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;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
-	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 = frame->local;
+        priv  = this->private;
 
-			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;
 }
 
 
 int
-afr_rmdir_done (call_frame_t *frame, xlator_t *this)
+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;
 
-int
-afr_rmdir (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.rmdir.parent_ino = loc->parent->ino;
-
-	local->transaction.fop    = afr_rmdir_wind;
-	local->transaction.done   = afr_rmdir_done;
-	local->transaction.unwind = afr_rmdir_unwind;
-
-	afr_build_parent_loc (&local->transaction.parent_loc, loc);
-
-	local->transaction.main_frame = frame;
-	local->transaction.basename = AFR_BASENAME (loc->path);
-
-	afr_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 (rmdir, 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;
 }
 
 /* }}} */
 
-/* {{{ setdents */
-
-int32_t
-afr_setdents_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		       int32_t op_ret, int32_t op_errno)
-{
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int call_count  = -1;
-	int child_index = (long) cookie;
+/* {{{ rmdir */
 
-	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) && (local->success_count == 0)) {
-			local->op_ret = op_ret;
-			local->success_count++;
-		}
+int
+afr_rmdir_unwind (call_frame_t *frame, xlator_t *this)
+{
+        call_frame_t *main_frame = NULL;
+        afr_local_t  *local = NULL;
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+        local = frame->local;
 
-	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.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;
 }
 
 
-int32_t
-afr_setdents_wind (call_frame_t *frame, xlator_t *this)
+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, 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_setdents_wind_cbk,	
-					   (void *) (long) i,
-					   priv->children[i], 
-					   priv->children[i]->fops->setdents,
-					   local->fd, local->cont.setdents.flags,
-					   local->cont.setdents.entries, 
-					   local->cont.setdents.count);
-			
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+        return __afr_dir_write_cbk (frame, cookie, this, op_ret, op_errno, NULL,
+				    preparent, postparent, NULL, NULL, xdata);
 }
 
 
-int32_t
-afr_setdents_done (call_frame_t *frame, xlator_t *this)
+int
+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;
 
-	AFR_STACK_UNWIND (setdents, frame, local->op_ret, local->op_errno);
-	
-	return 0;
+        local = frame->local;
+        priv  = this->private;
+
+	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;
 }
 
 
-int32_t
-afr_setdents (call_frame_t *frame, xlator_t *this,
-	      fd_t *fd, int32_t flags, dir_entry_t *entries, int32_t count)
+int
+afr_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+	   dict_t *xdata)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	
-	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;
 
-	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;
-	}
 
-	frame->local = local;
 
-	local->fd = fd_ref (fd);
+        loc_copy (&local->loc, loc);
+	local->inode = inode_ref (loc->inode);
+	local->parent = inode_ref (loc->parent);
 
-	local->cont.setdents.flags   = flags;
-	local->cont.setdents.entries = entries;
-	local->cont.setdents.count   = count;
+        local->cont.rmdir.flags = flags;
 
-	local->transaction.fop  = afr_setdents_wind;
-	local->transaction.done = afr_setdents_done;
+        if (xdata)
+                local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	local->transaction.basename = NULL;
+	if (!local->xdata_req)
+		goto out;
 
-	afr_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) {
-		AFR_STACK_UNWIND (setdents, frame, op_ret, op_errno);
-	}
+	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 4fa618b65..02f0a3682 100644
--- a/xlators/cluster/afr/src/afr-dir-write.h
+++ b/xlators/cluster/afr/src/afr-dir-write.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __DIR_WRITE_H__
@@ -22,38 +13,35 @@
 
 int32_t
 afr_create (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, int32_t flags, mode_t mode, fd_t *fd);
+            loc_t *loc, int32_t flags, mode_t mode,
+            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);
+	   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);
+	   loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata);
 
 int32_t
 afr_unlink (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc);
+	    loc_t *loc, int xflag, dict_t *xdata);
 
 int32_t
 afr_rmdir (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc);
+	   loc_t *loc, int flags, dict_t *xdata);
 
 int32_t
 afr_link (call_frame_t *frame, xlator_t *this,
-	  loc_t *oldloc, loc_t *newloc);
+	  loc_t *oldloc, loc_t *newloc, dict_t *xdata);
 
 int32_t
 afr_rename (call_frame_t *frame, xlator_t *this,
-	    loc_t *oldloc, loc_t *newloc);
+	    loc_t *oldloc, loc_t *newloc, dict_t *xdata);
 
-int32_t
+int
 afr_symlink (call_frame_t *frame, xlator_t *this,
-	     const char *linkpath, loc_t *oldloc);
-
-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 3b89da9ab..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) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -44,842 +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) {
-                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 stat *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;
+        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;
 
-	read_child = (long) cookie;
-
-	local = frame->local;
+		afr_read_txn_continue (frame, this, (long) cookie);
+		return 0;
+	}
 
-	if (op_ret == -1) {
-	retry:
-		last_tried = local->cont.stat.last_tried;
+	AFR_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf, xdata);
 
-		if (all_tried (last_tried, priv->child_count)) {
-			goto out;
-		}
-		this_try = ++local->cont.stat.last_tried;
-
-		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->st_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) {
-                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 stat *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->st_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) {
-                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 stat *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->st_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) {
-                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 = CALLOC (1, sizeof (*xkey));
-                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;
+                }
+
+                len = dict_serialized_length (local->dict);
+                if (len <= 0) {
+                        goto unwind;
+                }
 
-void
-__filter_xattrs (dict_t *dict)
+                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;
+                }
+
+                /* 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);
 
-                FREE (key);
+                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;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        priv     = this->private;
 
-	priv     = this->private;
-	VALIDATE_OR_GOTO (priv->children, out);
+        children = priv->children;
 
-	children = priv->children;
+        loc_copy (&local->loc, loc);
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-	frame->local = local;
+	local->op = GF_FOP_GETXATTR;
 
-        if (name) {
-                if (!strncmp (name, AFR_XATTR_PREFIX,
-                              strlen (AFR_XATTR_PREFIX))) {
+	if (xdata)
+		local->xdata_req = dict_ref (xdata);
 
-                        op_errno = ENODATA;
-                        goto out;
-                }
+        if (!name)
+                goto no_name;
+
+        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) {
-                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       = 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 stat *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;
+
+	local = frame->local;
+
+	if (op_ret < 0) {
+		local->op_ret = -1;
+		local->op_errno = op_errno;
 
-	int unwind     = 1;
-	int last_tried = -1;
-	int this_try   = -1;
-        int read_child = -1;
+		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);
+	if (dict)
+		afr_filter_xattrs (dict);
 
-	priv     = this->private;
-	VALIDATE_OR_GOTO (priv->children, out);
+	AFR_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, dict, xdata);
 
-	children = priv->children;
+	return 0;
+}
+
+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)
-                        buf->st_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) {
-                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 2b22131db..e4091a793 100644
--- a/xlators/cluster/afr/src/afr-inode-read.h
+++ b/xlators/cluster/afr/src/afr-inode-read.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __INODE_READ_H__
@@ -22,26 +13,30 @@
 
 int32_t
 afr_access (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, int32_t mask);
+	    loc_t *loc, int32_t mask, dict_t *xdata);
 
 int32_t
 afr_stat (call_frame_t *frame, xlator_t *this,
-	  loc_t *loc);
+	  loc_t *loc, dict_t *xdata);
 
 int32_t
 afr_fstat (call_frame_t *frame, xlator_t *this,
-	   fd_t *fd);
+	   fd_t *fd, dict_t *xdata);
 
 int32_t
 afr_readlink (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, size_t size);
+	      loc_t *loc, size_t size, dict_t *xdata);
 
 int32_t
 afr_readv (call_frame_t *frame, xlator_t *this,
-	   fd_t *fd, size_t size, off_t offset);
+	   fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata);
 
 int32_t
 afr_getxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, const char *name);
+	      loc_t *loc, const char *name, dict_t *xdata);
+
+int32_t
+afr_fgetxattr (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, const char *name, dict_t *xdata);
 
 #endif /* __INODE_READ_H__ */
diff --git a/xlators/cluster/afr/src/afr-inode-write.c b/xlators/cluster/afr/src/afr-inode-write.c
index 32e93e6e2..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) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -46,284 +37,433 @@
 
 #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;
-	afr_private_t * priv  = 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;
+	priv = this->private;
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame)
-			main_frame = local->transaction.main_frame;
-		local->transaction.main_frame = NULL;
+	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);
 	}
-	UNLOCK (&frame->lock);
 
-	if (main_frame) {
-                local->cont.writev.prebuf.st_ino  = local->cont.writev.ino;
-                local->cont.writev.postbuf.st_ino = local->cont.writev.ino;
+	local->op_ret = -1;
+	local->op_errno = afr_final_errno (local, priv);
 
-		AFR_STACK_UNWIND (writev, main_frame,
-                                  local->op_ret, local->op_errno,
-                                  &local->cont.writev.prebuf,
-                                  &local->cont.writev.postbuf);
+	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);
+			}
+		}
 	}
-	return 0;
 }
 
 
-int
-afr_writev_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		     int32_t op_ret, int32_t op_errno, struct stat *prebuf,
-                     struct stat *postbuf)
+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;
-	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;
+	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);
+	}
 
-        read_child = afr_read_child (this, local->fd->inode);
+        return;
+}
 
-	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);
+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;
+        int child_index = (long) cookie;
+        int call_count = -1;
 
-		if (op_ret != -1) {
-			if (local->success_count == 0) {
-				local->op_ret              = op_ret;
-				local->cont.writev.prebuf  = *prebuf;
-				local->cont.writev.postbuf = *postbuf;
-			}
+        local = frame->local;
 
-                        if (child_index == read_child) {
-                                local->cont.writev.prebuf  = *prebuf;
-                                local->cont.writev.postbuf = *postbuf;
-                        }
+        LOCK (&frame->lock);
+        {
+                __afr_inode_write_fill (frame, this, child_index, op_ret,
+					op_errno, prebuf, postbuf, xdata);
+        }
+        UNLOCK (&frame->lock);
 
-			local->success_count++;
+        call_count = afr_frame_return (frame);
 
-			if ((local->success_count >= priv->wait_count)
-                            && local->read_child_returned) {
-				need_unwind = 1;
-			}
-		}
+        if (call_count == 0) {
+		__afr_inode_write_finalize (frame, this);
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+		if (afr_txn_nothing_failed (frame, this))
+			local->transaction.unwind (frame, this);
 
-	call_count = afr_frame_return (frame);
+                local->transaction.resume (frame, this);
+        }
 
-	if (call_count == 0) {
-		local->transaction.unwind (frame, this);
+        return 0;
+}
 
-		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;
-	
-	int i = 0;
-	int call_count = -1;
+        call_frame_t *fop_frame = NULL;
 
-	local = frame->local;
-	priv = this->private;
+        fop_frame = afr_transaction_detach_fop_frame (frame);
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+        if (fop_frame) {
+                afr_writev_copy_outvars (frame, fop_frame);
+                afr_writev_unwind (fop_frame, this);
+        }
+        return 0;
+}
 
-	if (call_count == 0) {
-		local->transaction.resume (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->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_writev_wind_cbk, 
-					   (void *) (long) i,	
-					   priv->children[i], 
-					   priv->children[i]->fops->writev,
-					   local->fd, 
-					   local->cont.writev.vector,
-					   local->cont.writev.count, 
-					   local->cont.writev.offset,
-                                           local->cont.writev.iobref);
-		
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+        local = frame->local;
+        priv = this->private;
+        /*
+         * We already have the best case result of the writev calls staged
+         * as the return value. Any writev that returns some value less
+         * than the best case is now out of sync, so mark the fop as
+         * failed. Note that fops that have returned with errors have
+         * already been marked as failed.
+         */
+        for (i = 0; i < priv->child_count; i++) {
+                if ((!local->replies[i].valid) ||
+                    (local->replies[i].op_ret == -1))
+                        continue;
+
+                if (local->replies[i].op_ret < local->op_ret)
+                        afr_transaction_fop_failed(frame, this, i);
+        }
 }
 
-
 int
-afr_writev_done (call_frame_t *frame, xlator_t *this)
+afr_writev_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                     struct iatt *postbuf, dict_t *xdata)
 {
-	afr_local_t *local = NULL;
+        afr_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;
 
-	local = frame->local;
+        local = frame->local;
 
-        iobref_unref (local->cont.writev.iobref);
-	local->cont.writev.iobref = NULL;
+        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;
+}
 
-	local->transaction.unwind (frame, this);
 
-	AFR_STACK_DESTROY (frame);
+int
+afr_writev_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
-	return 0;
+        local = frame->local;
+        priv = this->private;
+
+	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;
 }
 
 
 int
 afr_do_writev (call_frame_t *frame, xlator_t *this)
 {
-        call_frame_t * transaction_frame = NULL;
-        afr_local_t *  local             = NULL;
+        call_frame_t    *transaction_frame = NULL;
+        afr_local_t     *local             = NULL;
+        int             ret   = -1;
+        int             op_errno = ENOMEM;
 
-        int op_ret   = -1;
-        int op_errno = 0;
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto out;
 
         local = frame->local;
+        transaction_frame->local = local;
+	frame->local = NULL;
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-                op_errno = ENOMEM;
+	if (!AFR_FRAME_INIT (frame, op_errno))
 		goto out;
-	}
-
-	transaction_frame->local = local;
-        frame->local = NULL;
-
-	local->op = GF_FOP_WRITE;
 
-        local->success_count      = 0;
-
-	local->transaction.fop    = afr_writev_wind;
-	local->transaction.done   = afr_writev_done;
-	local->transaction.unwind = afr_writev_unwind;
-
-	local->transaction.main_frame = frame;
-	if (local->fd->flags & O_APPEND) {
-		local->transaction.start   = 0;
-		local->transaction.len     = 0;
-	} else {
-		local->transaction.start   = local->cont.writev.offset;
-		local->transaction.len     = iov_length (local->cont.writev.vector,
+        local->op = GF_FOP_WRITE;
+
+        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 {
+                local->transaction.start   = local->cont.writev.offset;
+                local->transaction.len     = iov_length (local->cont.writev.vector,
                                                          local->cont.writev.count);
-	}
+        }
 
-	afr_transaction (transaction_frame, this, AFR_DATA_TRANSACTION);
+        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;
 }
 
 
 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)
+afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
+            struct iovec *vector, int32_t count, off_t offset,
+            uint32_t flags, struct iobref *iobref, dict_t *xdata)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
+        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;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-	priv = this->private;
+        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);
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
+	if (!local->xdata_req)
 		goto out;
-	}
 
-        frame->local = local;
+        local->fd = fd_ref (fd);
+	local->inode = inode_ref (fd->inode);
 
-	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);
-
-	local->fd                = fd_ref (fd);
+	if (dict_set_uint32 (local->xdata_req, GLUSTERFS_OPEN_FD_COUNT, 4)) {
+		op_errno = ENOMEM;
+		goto out;
+	}
 
-        ret = fd_ctx_get (fd, this, &ctx);
-        if (ret < 0) {
-                goto out;
-        }
+	if (dict_set_uint32 (local->xdata_req, GLUSTERFS_WRITE_IS_APPEND, 4)) {
+		op_errno = ENOMEM;
+		goto out;
+	}
 
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+	/* 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;
 
-        if (fd_ctx->down_count < priv->down_count) {
-                local->up_down_flush_cbk = afr_do_writev;
-                afr_up_down_flush (frame, this, fd, AFR_CHILD_DOWN_FLUSH);
+	/* 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));
 
-        } else if (fd_ctx->up_count < priv->up_count) {
-                local->up_down_flush_cbk = afr_do_writev;
-                afr_up_down_flush (frame, this, fd, AFR_CHILD_UP_FLUSH);
+        afr_fix_open (fd, this);
 
-        } 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;
+        return 0;
 }
 
 
@@ -334,214 +474,119 @@ out:
 int
 afr_truncate_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);
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
 
-	if (main_frame) {
-                local->cont.truncate.prebuf.st_ino  = local->cont.truncate.ino;
-                local->cont.truncate.postbuf.st_ino = local->cont.truncate.ino;
+        local = frame->local;
 
-                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;
 
-	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;
 }
 
 
 int
-afr_truncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		       int32_t op_ret, int32_t op_errno, struct stat *prebuf,
-                       struct stat *postbuf)
-{
-	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))
-			afr_transaction_fop_failed (frame, this, child_index);
-
-		if (op_ret != -1) {
-			if (local->success_count == 0) {
-				local->op_ret = op_ret;
-				local->cont.truncate.prebuf  = *prebuf;
-				local->cont.truncate.postbuf = *postbuf;
-			}
-
-                        if (child_index == read_child) {
-                                local->cont.truncate.prebuf  = *prebuf;
-                                local->cont.truncate.postbuf = *postbuf;
-                        }
-
-			local->success_count++;
-
-			if ((local->success_count >= priv->wait_count)
-                            && local->read_child_returned) {
-				need_unwind = 1;
-			}
-		}
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	if (need_unwind)
-		local->transaction.unwind (frame, this);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
-}
-
-
-int32_t
-afr_truncate_wind (call_frame_t *frame, xlator_t *this)
+afr_truncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                       struct iatt *postbuf, dict_t *xdata)
 {
 	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-	
-	int 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 (op_ret == 0 && prebuf->ia_size != postbuf->ia_size)
+		local->stable_write = _gf_false;
 
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
+	return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      prebuf, postbuf, xdata);
 }
 
 
 int
-afr_truncate_done (call_frame_t *frame, xlator_t *this)
+afr_truncate_wind (call_frame_t *frame, xlator_t *this, int subvol)
 {
-	afr_local_t *local = NULL;
-
-	local = frame->local;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
-	local->transaction.unwind (frame, this);
-
-	AFR_STACK_DESTROY (frame);
+        local = frame->local;
+        priv = this->private;
 
-	return 0;
+	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;
-
+        afr_private_t * priv  = NULL;
+        afr_local_t   * local = NULL;
+        call_frame_t   *transaction_frame = NULL;
 	int ret = -1;
+        int op_errno = ENOMEM;
 
-	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(truncate,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);
+        local->cont.truncate.offset  = offset;
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
+	if (!local->xdata_req)
 		goto out;
-	}
 
-	transaction_frame->local = local;
-
-	local->op_ret = -1;
+	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;
 
-	local->cont.truncate.offset  = offset;
-	local->cont.truncate.ino     = loc->inode->ino;
+        loc_copy (&local->loc, loc);
+	local->inode = inode_ref (loc->inode);
 
-	local->transaction.fop    = afr_truncate_wind;
-	local->transaction.done   = afr_truncate_done;
-	local->transaction.unwind = afr_truncate_unwind;
+        local->op = GF_FOP_TRUNCATE;
 
-	loc_copy (&local->loc, loc);
+        local->transaction.main_frame = frame;
+        local->transaction.start   = offset;
+        local->transaction.len     = 0;
 
-	local->transaction.main_frame = frame;
-	local->transaction.start   = 0;
-	local->transaction.len     = offset;
+	/* Set it true speculatively, will get reset in afr_truncate_wind_cbk
+	   if truncate was not a NOP */
+	local->stable_write = _gf_true;
 
-	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 (truncate, frame, op_ret, op_errno, NULL, NULL);
-	}
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
 
-	return 0;
+	AFR_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
+        return 0;
 }
 
 
@@ -553,1074 +598,1157 @@ out:
 int
 afr_ftruncate_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;
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
 
-	LOCK (&frame->lock);
-	{
-		if (local->transaction.main_frame)
-			main_frame = local->transaction.main_frame;
-		local->transaction.main_frame = NULL;
-	}
-	UNLOCK (&frame->lock);
+        local = frame->local;
 
-	if (main_frame) {
-                local->cont.ftruncate.prebuf.st_ino  = local->cont.ftruncate.ino;
-                local->cont.ftruncate.postbuf.st_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);
-	}
-	return 0;
+	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;
 }
 
 
 int
-afr_ftruncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-			int32_t op_ret, int32_t op_errno, struct stat *prebuf,
-                        struct stat *postbuf)
+afr_ftruncate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                        struct iatt *postbuf, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int child_index = (long) cookie;
-	int call_count  = -1;
-	int need_unwind = 0;
-        int read_child  = 0;
+	afr_local_t *local = NULL;
 
 	local = frame->local;
-	priv  = this->private;
-
-        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.ftruncate.prebuf  = *prebuf;
-				local->cont.ftruncate.postbuf = *postbuf;
-			}
+	if (op_ret == 0 && prebuf->ia_size != postbuf->ia_size)
+		local->stable_write = _gf_false;
 
-                        if (child_index == read_child) {
-                                local->cont.ftruncate.prebuf  = *prebuf;
-                                local->cont.ftruncate.postbuf = *postbuf;
-                        }
-
-			local->success_count++;
+	return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      prebuf, postbuf, xdata);
+}
 
-			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);
+int
+afr_ftruncate_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
-	call_count = afr_frame_return (frame);
+	local = frame->local;
+	priv = this->private;
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
+	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;
 }
 
 
 int
-afr_ftruncate_wind (call_frame_t *frame, xlator_t *this)
+afr_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+	       dict_t *xdata)
 {
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-	
-	int call_count = -1;
-	int i = 0;
+        afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+	call_frame_t *transaction_frame = NULL;
+	int ret = -1;
+        int op_errno = ENOMEM;
 
-	local = frame->local;
-	priv = this->private;
+        priv = this->private;
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+        QUORUM_CHECK(ftruncate,out);
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
+	transaction_frame = copy_frame (frame);
+	if (!frame)
+		goto out;
 
-	local->call_count = call_count;
+        local = AFR_FRAME_INIT (transaction_frame, op_errno);
+        if (!local)
+		goto out;
 
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_ftruncate_wind_cbk,
-					   (void *) (long) i,	
-					   priv->children[i], 
-					   priv->children[i]->fops->ftruncate,
-					   local->fd, local->cont.ftruncate.offset);
+        local->cont.ftruncate.offset  = offset;
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	return 0;
-}
+	if (!local->xdata_req)
+		goto out;
 
+        local->fd = fd_ref (fd);
+	local->inode = inode_ref (fd->inode);
 
-int
-afr_ftruncate_done (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t *local = NULL;
+        local->op = GF_FOP_FTRUNCATE;
 
-	local = frame->local;
+	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;
+
+        local->transaction.main_frame = frame;
 
-	local->transaction.unwind (frame, this);
+        local->transaction.start   = local->cont.ftruncate.offset;
+        local->transaction.len     = 0;
 
-	AFR_STACK_DESTROY (frame);
+        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_do_ftruncate (call_frame_t *frame, xlator_t *this)
+afr_setattr_unwind (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;
+        afr_local_t *local = NULL;
+        call_frame_t *main_frame = NULL;
 
         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;
+	main_frame = afr_transaction_detach_fop_frame (frame);
+	if (!main_frame)
+		return 0;
 
-	local->op = GF_FOP_FTRUNCATE;
+	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;
+}
 
-	local->transaction.fop    = afr_ftruncate_wind;
-	local->transaction.done   = afr_ftruncate_done;
-	local->transaction.unwind = afr_ftruncate_unwind;
 
-	local->transaction.main_frame = frame;
+int
+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)
+{
+	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_errno = ENOMEM;
 
-	int ret = -1;
-
-	int op_ret   = -1;
-	int op_errno = 0;
+        priv = this->private;
 
-        uint64_t ctx;
-        afr_fd_ctx_t *fd_ctx = NULL;
+        QUORUM_CHECK(setattr,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;
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+		goto out;
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        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 ();
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
-		op_errno = -ret;
+	if (!local->xdata_req)
 		goto out;
-	}
-
-        frame->local = local;
 
-	local->cont.ftruncate.offset  = offset;
-	local->cont.ftruncate.ino     = fd->inode->ino;
+        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;
 
-	local->fd = fd_ref (fd);
+        loc_copy (&local->loc, loc);
+	local->inode = inode_ref (loc->inode);
 
-        ret = fd_ctx_get (fd, this, &ctx);
-        if (ret < 0) {
-                goto out;
-        }
+	local->op = GF_FOP_SETATTR;
 
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+        local->transaction.main_frame = frame;
+        local->transaction.start   = LLONG_MAX - 1;
+        local->transaction.len     = 0;
 
-        if (fd_ctx->down_count < priv->down_count) {
-                local->up_down_flush_cbk = afr_do_ftruncate;
-                afr_up_down_flush (frame, this, fd, AFR_CHILD_DOWN_FLUSH);
-        } else {
-                afr_do_ftruncate (frame, this);
+        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);
 
-	return 0;
+	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;
-	afr_private_t * priv  = NULL;
-	call_frame_t   *main_frame = NULL;
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
 
-	local = frame->local;
-	priv  = this->private;
+        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.setattr.preop_buf.st_ino  = local->cont.setattr.ino;
-                local->cont.setattr.postop_buf.st_ino = local->cont.setattr.ino;
+	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;
+}
 
-                AFR_STACK_UNWIND (setattr, main_frame, local->op_ret,
-                                  local->op_errno,
-                                  &local->cont.setattr.preop_buf,
-                                  &local->cont.setattr.postop_buf);
-        }
 
-	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, dict_t *xdata)
+{
+	return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      preop, postop, xdata);
 }
 
 
 int
-afr_setattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                      int32_t op_ret, int32_t op_errno,
-                      struct stat *preop, struct stat *postop)
+afr_fsetattr_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 child_index = (long) cookie;
-        int read_child  = 0;
-	int call_count  = -1;
-	int need_unwind = 0;
+        local = frame->local;
+        priv = this->private;
 
-	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;
+}
 
-        read_child = afr_read_child (this, local->loc.inode);
 
-	LOCK (&frame->lock);
-	{
-                if (child_index == read_child) {
-                        local->read_child_returned = _gf_true;
-                }
+int
+afr_fsetattr (call_frame_t *frame, xlator_t *this,
+              fd_t *fd, struct iatt *buf, int32_t valid, dict_t *xdata)
+{
+        afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        call_frame_t *transaction_frame = NULL;
+        int ret = -1;
+        int op_errno = ENOMEM;
 
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
+        priv = this->private;
 
-		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;
-			}
+        QUORUM_CHECK(fsetattr,out);
 
-                        if (child_index == read_child) {
-                                local->cont.setattr.preop_buf  = *preop;
-                                local->cont.setattr.postop_buf = *postop;
-                        }
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto out;
 
-			local->success_count++;
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+		goto out;
 
-			if ((local->success_count >= priv->wait_count)
-                            && local->read_child_returned) {
-				need_unwind = 1;
-			}
-		}
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+        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 (need_unwind)
-		local->transaction.unwind (frame, this);
+	if (!local->xdata_req)
+		goto out;
 
-	call_count = afr_frame_return (frame);
+        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;
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-	}
+        local->fd                 = fd_ref (fd);
+	local->inode = inode_ref (fd->inode);
+
+	local->op = GF_FOP_FSETATTR;
+
+        afr_fix_open (fd, 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;
+        }
 
 	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;
+/* {{{ setxattr */
 
-	int call_count = -1;
-	int i = 0;
 
-	local = frame->local;
-	priv = this->private;
+int
+afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = 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_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;
-		}
-	}
 
-	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;
 
-	return 0;
+	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 stat *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;
-	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          = EINVAL;
 
-	int ret = -1;
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.afr.*", dict,
+                                   op_errno, out);
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.afr.*", dict,
+                                   op_errno, out);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        priv = this->private;
 
-	priv = this->private;
+        QUORUM_CHECK(setxattr,out);
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto out;
+
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+                goto out;
+
+        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 ();
+
+	if (!local->xdata_req)
 		goto out;
-	}
 
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        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);
 
-	ret = AFR_LOCAL_INIT (local, priv);
-	if (ret < 0) {
+        local->transaction.main_frame = frame;
+        local->transaction.start   = LLONG_MAX - 1;
+        local->transaction.len     = 0;
+
+	local->op = GF_FOP_SETXATTR;
+
+        ret = afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        if (ret < 0) {
 		op_errno = -ret;
 		goto out;
-	}
-
-	transaction_frame->local = local;
+        }
 
-	local->op_ret = -1;
+	return 0;
+out:
+	if (transaction_frame)
+		AFR_STACK_DESTROY (transaction_frame);
 
-	local->cont.setattr.ino     = loc->inode->ino;
+	AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
 
-        local->cont.setattr.in_buf = *buf;
-        local->cont.setattr.valid  = valid;
+        return 0;
+}
 
-	local->transaction.fop    = afr_setattr_wind;
-	local->transaction.done   = afr_setattr_done;
-	local->transaction.unwind = afr_setattr_unwind;
+/* {{{ fsetxattr */
 
-	loc_copy (&local->loc, loc);
 
-	local->transaction.main_frame = frame;
-	local->transaction.start   = LLONG_MAX - 1;
-	local->transaction.len     = 0;
+int
+afr_fsetxattr_unwind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t    *local         = NULL;
+        call_frame_t   *main_frame    = NULL;
 
-	afr_transaction (transaction_frame, this, AFR_METADATA_TRANSACTION);
+        local = frame->local;
 
-	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);
-	}
+	main_frame = afr_transaction_detach_fop_frame (frame);
+        if (!main_frame)
+		return 0;
 
-	return 0;
+	AFR_STACK_UNWIND (fsetxattr, main_frame, local->op_ret, local->op_errno,
+			  local->xdata_rsp);
+        return 0;
 }
 
-/* {{{ fsetattr */
 
 int
-afr_fsetattr_unwind (call_frame_t *frame, xlator_t *this)
+afr_fsetxattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	call_frame_t   *main_frame = NULL;
+        return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      NULL, NULL, xdata);
+}
 
-	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) {
-                local->cont.fsetattr.preop_buf.st_ino  =
-                        local->cont.fsetattr.ino;
-                local->cont.fsetattr.postop_buf.st_ino =
-                        local->cont.fsetattr.ino;
-
-                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 (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t        *local       = NULL;
+        afr_private_t      *priv        = NULL;
 
-	return 0;
+        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 stat *preop, struct stat *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->loc.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;
+
+        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;
 
-	if (need_unwind)
-		local->transaction.unwind (frame, this);
+        local->fd                 = fd_ref (fd);
+	local->inode = inode_ref (fd->inode);
 
-	call_count = afr_frame_return (frame);
+	local->op = GF_FOP_FSETXATTR;
 
-	if (call_count == 0) {
-		local->transaction.resume (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;
+        }
 
 	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;
+/* {{{ removexattr */
 
-	local = frame->local;
-	priv = this->private;
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+int
+afr_removexattr_unwind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
+
+        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;
+	AFR_STACK_UNWIND (removexattr, main_frame, local->op_ret, local->op_errno,
+			  local->xdata_rsp);
+        return 0;
+}
 
-	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;
-		}
-	}
 
-	return 0;
+int
+afr_removexattr_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                          int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      NULL, NULL, xdata);
 }
 
 
 int
-afr_fsetattr_done (call_frame_t *frame, xlator_t *this)
+afr_removexattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
 {
-	afr_local_t *local = NULL;
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
 	local = frame->local;
+	priv = this->private;
 
-	local->transaction.unwind (frame, this);
-
-	AFR_STACK_DESTROY (frame);
-
-	return 0;
+	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 stat *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;
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.afr.*",
+                                 name, op_errno, out);
 
-	int op_ret   = -1;
-	int op_errno = 0;
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.afr.*",
+                                 name, op_errno, out);
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        priv = this->private;
 
-	priv = this->private;
+        QUORUM_CHECK(removexattr,out);
 
-	transaction_frame = copy_frame (frame);
-	if (!transaction_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-	ALLOC_OR_GOTO (local, afr_local_t, out);
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto 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->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;
+        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);
 
-	return 0;
+	AFR_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
+        return 0;
 }
 
+/* ffremovexattr */
+int
+afr_fremovexattr_unwind (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
+
+        local = frame->local;
 
-/* {{{ setxattr */
+	main_frame = afr_transaction_detach_fop_frame (frame);
+        if (!main_frame)
+		return 0;
+
+	AFR_STACK_UNWIND (fremovexattr, main_frame, local->op_ret, local->op_errno,
+			  local->xdata_rsp);
+        return 0;
+}
 
 
 int
-afr_setxattr_unwind (call_frame_t *frame, xlator_t *this)
+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;
-	call_frame_t   *main_frame = NULL;
+	return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      NULL, NULL, xdata);
+}
 
-	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);
+int
+afr_fremovexattr_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
-	if (main_frame) {
-		AFR_STACK_UNWIND (setxattr, main_frame,
-                                  local->op_ret, local->op_errno)
-	}
-	return 0;
+        local = frame->local;
+        priv = this->private;
+
+	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_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		       int32_t op_ret, int32_t op_errno)
+afr_fremovexattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+		  const char *name, 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 call_count  = -1;
-	int need_unwind = 0;
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.afr.*",
+                                 name, op_errno, out);
+
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.glusterfs.afr.*",
+                                 name, op_errno, out);
 
-	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++;
+        QUORUM_CHECK(fremovexattr, out);
 
-			if (local->success_count == priv->wait_count) {
-				need_unwind = 1;
-			}
-		}
+        transaction_frame = copy_frame (frame);
+        if (!transaction_frame)
+                goto out;
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+	local = AFR_FRAME_INIT (transaction_frame, op_errno);
+	if (!local)
+                goto out;
 
-	if (need_unwind)
-		local->transaction.unwind (frame, this);
+        local->cont.removexattr.name = gf_strdup (name);
+	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;
+
+        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;
+        }
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-	}
-	
 	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_wind (call_frame_t *frame, xlator_t *this)
+afr_fallocate_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;
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
 
-	local = frame->local;
-	priv = this->private;
-
-	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;
+	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;
+}
 
-	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;
+int
+afr_fallocate_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                        struct iatt *postbuf, dict_t *xdata)
+{
+        return __afr_inode_write_cbk (frame, cookie, this, op_ret, op_errno,
+				      prebuf, postbuf, xdata);
 }
 
 
 int
-afr_setxattr_done (call_frame_t *frame, xlator_t *this)
+afr_fallocate_wind (call_frame_t *frame, xlator_t *this, int subvol)
 {
-	afr_local_t * local = frame->local;
-
-	local->transaction.unwind (frame, this);
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
-	AFR_STACK_DESTROY (frame);
-	
-	return 0;
+        local = frame->local;
+        priv = this->private;
+
+	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;
 }
 
 
 int
-afr_setxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, dict_t *dict, int32_t flags)
+afr_fallocate (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+               off_t offset, size_t len, dict_t *xdata)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	call_frame_t   *transaction_frame = NULL;
-
-	int ret = -1;
+	afr_private_t *priv = NULL;
+        call_frame_t *transaction_frame = NULL;
+        afr_local_t *local = NULL;
+        int ret = -1;
+        int op_errno = ENOMEM;
 
-	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(fallocate,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);
+        local->cont.fallocate.mode = mode;
+        local->cont.fallocate.offset  = offset;
+        local->cont.fallocate.len = len;
 
-	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);
 
-	transaction_frame->local = local;
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	local->op_ret = -1;
+	if (!local->xdata_req)
+		goto out;
 
-	local->cont.setxattr.dict  = dict_ref (dict);
-	local->cont.setxattr.flags = flags;
+        local->op = GF_FOP_FALLOCATE;
 
-	local->transaction.fop    = afr_setxattr_wind;
-	local->transaction.done   = afr_setxattr_done;
-	local->transaction.unwind = afr_setxattr_unwind;
+        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;
 
-	loc_copy (&local->loc, loc);
+        local->transaction.main_frame = frame;
 
-	local->transaction.main_frame = frame;
-	local->transaction.start   = LLONG_MAX - 1;
-	local->transaction.len     = 0;
+        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;
-	afr_private_t * priv  = NULL;
-	call_frame_t   *main_frame = NULL;
+        afr_local_t *   local = NULL;
+        call_frame_t   *main_frame = NULL;
 
-	local = frame->local;
-	priv  = this->private;
+        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)
-	}
-	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_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;
 
-	local = frame->local;
-	priv = this->private;
+int
+afr_discard_wind (call_frame_t *frame, xlator_t *this, int subvol)
+{
+        afr_local_t *local = NULL;
+        afr_private_t *priv = NULL;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret != -1) {
-			if (local->success_count == 0) {
-				local->op_ret = op_ret;
-			}
-			local->success_count++;
+        local = frame->local;
+        priv = this->private;
 
-			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);
-	}
-	
-	return 0;
-}
+        QUORUM_CHECK(discard, out);
 
+	transaction_frame = copy_frame (frame);
+	if (!transaction_frame)
+		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 = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-	int call_count = -1;
-	int i = 0;
+        local->cont.discard.offset  = offset;
+        local->cont.discard.len = len;
 
-	local = frame->local;
-	priv = this->private;
+        local->fd = fd_ref (fd);
+	local->inode = inode_ref (fd->inode);
 
-	call_count = afr_up_children_count (priv->child_count, local->child_up);
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-		return 0;
-	}
+	if (!local->xdata_req)
+		goto out;
 
-	local->call_count = call_count;
+        local->op = GF_FOP_DISCARD;
 
-	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);
+        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;
+
+        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;
+        }
 
-			if (!--call_count)
-				break;
-		}
-	}
-	
 	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);
-	
-	return 0;
+	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;
+
+        local = frame->local;
+        priv = this->private;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-	VALIDATE_OR_GOTO (loc, out);
+	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;
+}
 
-	priv = this->private;
+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;
+
+        QUORUM_CHECK(discard, 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 (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);
+
+	if (xdata)
+		local->xdata_req = dict_copy_with_ref (xdata, NULL);
+	else
+		local->xdata_req = dict_new ();
 
-	local->cont.removexattr.name = strdup (name);
+	if (!local->xdata_req)
+		goto out;
 
-	local->transaction.fop    = afr_removexattr_wind;
-	local->transaction.done   = afr_removexattr_done;
-	local->transaction.unwind = afr_removexattr_unwind;
+        local->op = GF_FOP_ZEROFILL;
 
-	loc_copy (&local->loc, loc);
+        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;
+        local->transaction.main_frame = frame;
 
-	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 66da777b6..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) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __INODE_WRITE_H__
@@ -22,51 +13,70 @@
 
 int32_t
 afr_chmod (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, mode_t mode);
+	   loc_t *loc, mode_t mode, dict_t *xdata);
 
 int32_t
 afr_chown (call_frame_t *frame, xlator_t *this,
-	   loc_t *loc, uid_t uid, gid_t gid);
+	   loc_t *loc, uid_t uid, gid_t gid, dict_t *xdata);
 
 int
 afr_fchown (call_frame_t *frame, xlator_t *this,
-	    fd_t *fd, uid_t uid, gid_t gid);
+	    fd_t *fd, uid_t uid, gid_t gid, dict_t *xdata);
 
 int32_t
 afr_fchmod (call_frame_t *frame, xlator_t *this,
-	    fd_t *fd, mode_t mode);
+	    fd_t *fd, mode_t mode, dict_t *xdata);
 
 int32_t
-afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, 
+afr_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
 	    struct iovec *vector, int32_t count, off_t offset,
-            struct iobref *iobref);
+            uint32_t flags, struct iobref *iobref, dict_t *xdata);
 
 int32_t
 afr_truncate (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, off_t offset);
+	      loc_t *loc, off_t offset, dict_t *xdata);
 
 int32_t
 afr_ftruncate (call_frame_t *frame, xlator_t *this,
-	       fd_t *fd, off_t offset);
+	       fd_t *fd, off_t offset, dict_t *xdata);
 
 int32_t
 afr_utimens (call_frame_t *frame, xlator_t *this,
-	     loc_t *loc, struct timespec tv[2]);
+	     loc_t *loc, struct timespec tv[2], dict_t *xdata);
 
 int
 afr_setattr (call_frame_t *frame, xlator_t *this,
-             loc_t *loc, struct stat *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 stat *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
new file mode 100644
index 000000000..a2a758f35
--- /dev/null
+++ b/xlators/cluster/afr/src/afr-lk-common.c
@@ -0,0 +1,1667 @@
+/*
+  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 "dict.h"
+#include "byte-order.h"
+#include "common-utils.h"
+
+#include "afr.h"
+#include "afr-transaction.h"
+
+#include <signal.h>
+
+
+#define LOCKED_NO       0x0        /* no lock held */
+#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_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;
+
+static uint64_t
+get_afr_lock_number ()
+{
+        return (++afr_lock_number);
+}
+
+int
+afr_set_lock_number (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_number = get_afr_lock_number ();
+
+        return 0;
+}
+
+void
+afr_set_lk_owner (call_frame_t *frame, xlator_t *this, void *lk_owner)
+{
+        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
+is_afr_lock_selfheal (afr_local_t *local)
+{
+        afr_internal_lock_t *int_lock = NULL;
+        int                  ret      = -1;
+
+        int_lock = &local->internal_lock;
+
+        switch (int_lock->selfheal_lk_type) {
+        case AFR_DATA_SELF_HEAL_LK:
+        case AFR_METADATA_SELF_HEAL_LK:
+                ret = 1;
+                break;
+        case AFR_ENTRY_SELF_HEAL_LK:
+                ret = 0;
+                break;
+        }
+
+        return ret;
+
+}
+
+int32_t
+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;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->child_up[i])
+                        ++call_count;
+        }
+
+        return call_count;
+}
+
+static void
+afr_print_inodelk (char *str, int size, int cmd,
+                   struct gf_flock *flock, gf_lkowner_t *owner)
+{
+        char *cmd_str = NULL;
+        char *type_str = NULL;
+
+        switch (cmd) {
+#if F_GETLK != F_GETLK64
+        case F_GETLK64:
+#endif
+        case F_GETLK:
+                cmd_str = "GETLK";
+                break;
+
+#if F_SETLK != F_SETLK64
+        case F_SETLK64:
+#endif
+        case F_SETLK:
+                cmd_str = "SETLK";
+                break;
+
+#if F_SETLKW != F_SETLKW64
+        case F_SETLKW64:
+#endif
+        case F_SETLKW:
+                cmd_str = "SETLKW";
+                break;
+
+        default:
+                cmd_str = "<null>";
+                break;
+        }
+
+        switch (flock->l_type) {
+        case F_RDLCK:
+                type_str = "READ";
+                break;
+        case F_WRLCK:
+                type_str = "WRITE";
+                break;
+        case F_UNLCK:
+                type_str = "UNLOCK";
+                break;
+        default:
+                type_str = "UNKNOWN";
+                break;
+        }
+
+        snprintf (str, size, "lock=INODELK, cmd=%s, type=%s, "
+                  "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,
+                  lkowner_utoa (owner));
+
+}
+
+static void
+afr_print_lockee (char *str, int size, loc_t *loc, fd_t *fd,
+                  int child_index)
+{
+        snprintf (str, size, "path=%s, fd=%p, child=%d",
+                  loc->path ? loc->path : "<nul>",
+                  fd ? fd : NULL,
+                  child_index);
+}
+
+void
+afr_print_entrylk (char *str, int size, const char *basename,
+                   gf_lkowner_t *owner)
+{
+        snprintf (str, size, "Basename=%s, lk-owner=%s",
+                  basename ? basename : "<nul>",
+                  lkowner_utoa (owner));
+}
+
+static void
+afr_print_verdict (int op_ret, int op_errno, char *str)
+{
+        if (op_ret < 0) {
+                if (op_errno == EAGAIN)
+                        strcpy (str, "EAGAIN");
+                else
+                        strcpy (str, "FAILED");
+        }
+        else
+                strcpy (str, "GRANTED");
+}
+
+static void
+afr_set_lock_call_type (afr_lock_call_type_t lock_call_type,
+                        char *lock_call_type_str,
+                        afr_internal_lock_t *int_lock)
+{
+        switch (lock_call_type) {
+        case AFR_INODELK_TRANSACTION:
+                if (int_lock->transaction_lk_type == AFR_TRANSACTION_LK)
+                        strcpy (lock_call_type_str, "AFR_INODELK_TRANSACTION");
+                else
+                        strcpy (lock_call_type_str, "AFR_INODELK_SELFHEAL");
+                break;
+        case AFR_INODELK_NB_TRANSACTION:
+                if (int_lock->transaction_lk_type == AFR_TRANSACTION_LK)
+                        strcpy (lock_call_type_str, "AFR_INODELK_NB_TRANSACTION");
+                else
+                        strcpy (lock_call_type_str, "AFR_INODELK_NB_SELFHEAL");
+                break;
+        case AFR_ENTRYLK_TRANSACTION:
+                if (int_lock->transaction_lk_type == AFR_TRANSACTION_LK)
+                        strcpy (lock_call_type_str, "AFR_ENTRYLK_TRANSACTION");
+                else
+                        strcpy (lock_call_type_str, "AFR_ENTRYLK_SELFHEAL");
+                break;
+        case AFR_ENTRYLK_NB_TRANSACTION:
+                if (int_lock->transaction_lk_type == AFR_TRANSACTION_LK)
+                        strcpy (lock_call_type_str, "AFR_ENTRYLK_NB_TRANSACTION");
+                else
+                        strcpy (lock_call_type_str, "AFR_ENTRYLK_NB_SELFHEAL");
+                break;
+        default:
+                strcpy (lock_call_type_str, "UNKNOWN");
+                break;
+        }
+
+}
+
+static void
+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)
+{
+        afr_internal_lock_t *int_lock = NULL;
+        afr_local_t         *local    = NULL;
+
+        char lockee[256];
+        char lock_call_type_str[256];
+        char verdict[16];
+
+        local    = frame->local;
+        int_lock = &local->internal_lock;
+
+        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);
+
+        afr_print_verdict (op_ret, op_errno, verdict);
+
+        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, lkowner_utoa (&frame->root->lk_owner), lockee,
+                (unsigned long long) int_lock->lock_number);
+
+}
+
+static void
+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)
+{
+        afr_local_t         *local    = NULL;
+        afr_internal_lock_t *int_lock = NULL;
+
+        char lock[256];
+        char lockee[256];
+        char lock_call_type_str[256];
+
+        local    = frame->local;
+        int_lock = &local->internal_lock;
+
+        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_INFO,
+                "[%s %s] Lock={%s} Lockee={%s} Number={%llu}",
+                lock_call_type_str,
+                lk_op_type == AFR_LOCK_OP ? "LOCK REQUEST" : "UNLOCK REQUEST",
+                lock, lockee,
+                (unsigned long long) int_lock->lock_number);
+
+}
+
+static void
+afr_trace_entrylk_in (call_frame_t *frame, xlator_t *this,
+                      afr_lock_call_type_t lock_call_type,
+                      afr_lock_op_type_t lk_op_type, const char *basename,
+                      int32_t cookie)
+{
+        afr_local_t         *local    = NULL;
+        afr_internal_lock_t *int_lock = NULL;
+        afr_private_t       *priv     = NULL;
+        int                 child_index = 0;
+        int                 lockee_no = 0;
+
+        char lock[256];
+        char lockee[256];
+        char lock_call_type_str[256];
+
+        local    = frame->local;
+        int_lock = &local->internal_lock;
+        priv     = this->private;
+
+        if (!priv->entrylk_trace) {
+                return;
+        }
+        lockee_no = cookie / priv->child_count;
+        child_index = cookie % priv->child_count;
+
+        afr_print_entrylk (lock, 256, basename, &frame->root->lk_owner);
+        afr_print_lockee (lockee, 256, &int_lock->lockee[lockee_no].loc, local->fd,
+                          child_index);
+
+        afr_set_lock_call_type (lock_call_type, lock_call_type_str, int_lock);
+
+        gf_log (this->name, GF_LOG_INFO,
+                "[%s %s] Lock={%s} Lockee={%s} Number={%llu}, 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,
+                cookie);
+}
+
+static void
+afr_trace_entrylk_out (call_frame_t *frame, xlator_t *this,
+                       afr_lock_call_type_t lock_call_type,
+                       afr_lock_op_type_t lk_op_type, const char *basename,
+                       int op_ret, int op_errno, int32_t cookie)
+{
+        afr_internal_lock_t *int_lock = NULL;
+        afr_local_t         *local    = NULL;
+        afr_private_t       *priv     = NULL;
+        int                 lockee_no = 0;
+        int                 child_index = 0;
+
+        char lock[256];
+        char lockee[256];
+        char lock_call_type_str[256];
+        char verdict[16];
+
+        local    = frame->local;
+        int_lock = &local->internal_lock;
+        priv     = this->private;
+
+        if (!priv->entrylk_trace) {
+                return;
+        }
+        lockee_no = cookie / priv->child_count;
+        child_index = cookie % priv->child_count;
+
+        afr_print_entrylk (lock, 256, basename, &frame->root->lk_owner);
+        afr_print_lockee (lockee, 256, &int_lock->lockee[lockee_no].loc, local->fd,
+                          child_index);
+
+        afr_set_lock_call_type (lock_call_type, lock_call_type_str, int_lock);
+
+        afr_print_verdict (op_ret, op_errno, verdict);
+
+        gf_log (this->name, GF_LOG_INFO,
+                "[%s %s] [%s] Lock={%s} Lockee={%s} Number={%llu} 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,
+                cookie);
+
+}
+
+static int
+transaction_lk_op (afr_local_t *local)
+{
+        afr_internal_lock_t *int_lock = NULL;
+        int ret = -1;
+
+        int_lock = &local->internal_lock;
+
+        if (int_lock->transaction_lk_type == AFR_TRANSACTION_LK) {
+                gf_log (THIS->name, GF_LOG_DEBUG,
+                        "lk op is for a transaction");
+                ret = 1;
+        }
+        else if (int_lock->transaction_lk_type == AFR_SELFHEAL_LK) {
+                gf_log (THIS->name, GF_LOG_DEBUG,
+                        "lk op is for a self heal");
+
+                ret = 0;
+        }
+
+        if (ret == -1)
+                gf_log (THIS->name, GF_LOG_DEBUG,
+                        "lk op is not set");
+
+        return ret;
+
+}
+
+static int
+is_afr_lock_transaction (afr_local_t *local)
+{
+        int ret = 0;
+
+        switch (local->transaction.type) {
+        case AFR_DATA_TRANSACTION:
+        case AFR_METADATA_TRANSACTION:
+                ret = 1;
+                break;
+
+        case AFR_ENTRY_RENAME_TRANSACTION:
+        case AFR_ENTRY_TRANSACTION:
+                ret = 0;
+                break;
+
+        }
+
+        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)
+{
+        afr_local_t         *local    = NULL;
+        afr_internal_lock_t *int_lock = NULL;
+        afr_private_t       *priv     = NULL;
+
+        int i = 0;
+
+        priv     = this->private;
+        local    = frame->local;
+        int_lock = &local->internal_lock;
+
+        int_lock->entrylk_lock_count = 0;
+        int_lock->lock_op_ret        = -1;
+        int_lock->lock_op_errno      = 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;
+}
+
+static int
+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;
+        afr_inodelk_t       *inodelk  = NULL;
+
+        priv     = this->private;
+        local    = frame->local;
+        int_lock = &local->internal_lock;
+
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+
+        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;
+}
+
+int
+afr_lockee_locked_nodes_count (afr_internal_lock_t *int_lock)
+{
+        int call_count  = 0;
+        int i           = 0;
+
+        for (i = 0; i < int_lock->lockee_count; i++)
+                call_count += int_lock->lockee[i].locked_count;
+
+        return call_count;
+}
+
+int
+afr_locked_nodes_count (unsigned char *locked_nodes, int child_count)
+
+{
+        int i = 0;
+        int call_count = 0;
+
+        for (i = 0; i < child_count; i++) {
+                if (locked_nodes[i] & LOCKED_YES)
+                        call_count++;
+        }
+
+        return call_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, dict_t *xdata)
+{
+        afr_local_t         *local    = NULL;
+        afr_internal_lock_t *int_lock = NULL;
+        int call_count = 0;
+
+        local    = frame->local;
+        int_lock = &local->internal_lock;
+
+        LOCK (&frame->lock);
+        {
+                call_count = --int_lock->lk_call_count;
+        }
+        UNLOCK (&frame->lock);
+
+        if (call_count == 0) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "All internal locks unlocked");
+                int_lock->lock_cbk (frame, this);
+        }
+
+        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, dict_t *xdata)
+{
+        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, child_index);
+
+        priv = this->private;
+
+        if (op_ret < 0 && op_errno != ENOTCONN && op_errno != EBADFD) {
+                gf_log (this->name, GF_LOG_INFO, "%s: unlock failed on subvolume %s "
+                        "with lock owner %s", local->loc.path,
+                        priv->children[child_index]->name,
+                        lkowner_utoa (&frame->root->lk_owner));
+        }
+
+
+        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;
+
+}
+
+static int
+afr_unlock_inodelk (call_frame_t *frame, xlator_t *this)
+{
+        afr_internal_lock_t *int_lock = NULL;
+        afr_inodelk_t       *inodelk  = NULL;
+        afr_local_t         *local    = NULL;
+        afr_private_t       *priv     = NULL;
+        struct gf_flock flock = {0,};
+        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;
+
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+
+        flock.l_start = inodelk->flock.l_start;
+        flock.l_len   = inodelk->flock.l_len;
+        flock.l_type  = F_UNLCK;
+
+        full_flock.l_type = F_UNLCK;
+        call_count = afr_locked_nodes_count (inodelk->locked_nodes,
+                                             priv->child_count);
+
+        int_lock->lk_call_count = call_count;
+
+        if (!call_count) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "No internal locks unlocked");
+                int_lock->lock_cbk (frame, this);
+                goto out;
+        }
+
+        if (local->fd)
+                fd_ctx = afr_fd_ctx_get (local->fd, this);
+
+        for (i = 0; i < priv->child_count; i++) {
+                if ((inodelk->locked_nodes[i] & LOCKED_YES) != LOCKED_YES)
+                        continue;
+
+                if (local->fd) {
+                        flock_use = &flock;
+                        if (!local->transaction.eager_lock[i]) {
+                                goto wind;
+                        }
+
+                        piggyback = 0;
+
+                        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, dict_t *xdata)
+{
+        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_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;
+}
+
+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;
+        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;
+
+        call_count = afr_lockee_locked_nodes_count (int_lock);
+
+        int_lock->lk_call_count = call_count;
+
+        if (!call_count){
+                gf_log (this->name, GF_LOG_TRACE,
+                        "No internal locks unlocked");
+                int_lock->lock_cbk (frame, this);
+                goto out;
+        }
+
+        for (i = 0; i < 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[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;
+                }
+        }
+
+out:
+        return 0;
+
+}
+
+static int32_t
+afr_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+              int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        afr_internal_lock_t     *int_lock       = NULL;
+        afr_local_t             *local          = NULL;
+        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/locks xlator on server");
+                                local->op_ret = op_ret;
+                                int_lock->lock_op_ret = op_ret;
+                        }
+
+                        local->op_errno              = op_errno;
+                        int_lock->lock_op_errno      = op_errno;
+                }
+
+		int_lock->lk_attempted_count++;
+        }
+        UNLOCK (&frame->lock);
+
+        if ((op_ret == -1) &&
+            (op_errno == ENOSYS)) {
+                afr_unlock (frame, this);
+        } else {
+                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);
+        }
+
+        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, dict_t *xdata)
+{
+        AFR_TRACE_INODELK_OUT (frame, this, AFR_INODELK_TRANSACTION,
+                               AFR_LOCK_OP, NULL, op_ret,
+                               op_errno, (long) cookie);
+
+        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, dict_t *xdata)
+{
+        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, xdata);
+        return 0;
+}
+
+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;
+
+        priv     = this->private;
+        local    = frame->local;
+        int_lock = &local->internal_lock;
+
+        switch (local->transaction.type) {
+        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:
+                /*entrylk_count is being used in both non-blocking and blocking
+                 * modes */
+                break;
+        }
+
+        return 0;
+
+}
+
+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 cookie)
+{
+        afr_internal_lock_t *int_lock    = NULL;
+        afr_inodelk_t       *inodelk     = NULL;
+        afr_local_t         *local       = NULL;
+        afr_private_t       *priv        = NULL;
+        struct gf_flock flock = {0,};
+        uint64_t ctx = 0;
+        int ret = 0;
+        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_INFO,
+                                "unable to get fd ctx for fd=%p",
+                                local->fd);
+
+                        local->op_ret           = -1;
+                        int_lock->lock_op_ret   = -1;
+
+                        afr_copy_locked_nodes (frame, this);
+
+                        afr_unlock (frame, this);
+
+                        return 0;
+                }
+        }
+
+        if (int_lock->lk_expected_count == int_lock->lk_attempted_count) {
+                if ((is_entrylk && int_lock->entrylk_lock_count == 0) ||
+                    (!is_entrylk && int_lock->lock_count == 0)) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "unable to lock on even one child");
+
+                        local->op_ret           = -1;
+                        int_lock->lock_op_ret   = -1;
+
+                        afr_copy_locked_nodes (frame, this);
+
+                        afr_unlock(frame, this);
+
+                        return 0;
+                }
+        }
+
+        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");
+
+                afr_copy_locked_nodes (frame, this);
+
+                int_lock->lock_op_ret = 0;
+                int_lock->lock_cbk (frame, this);
+                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:
+
+                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,
+                                           int_lock->domain, local->fd,
+                                           F_SETLKW, &flock, NULL);
+
+                } 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,
+                                           int_lock->domain, &local->loc,
+                                           F_SETLKW, &flock, NULL);
+                }
+
+                break;
+
+        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) 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
+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:
+                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;
+        }
+
+        afr_lock_blocking (frame, this, 0);
+
+        return 0;
+}
+
+static int32_t
+afr_nonblocking_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                             int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        afr_internal_lock_t *int_lock = NULL;
+        afr_local_t         *local    = NULL;
+        int call_count          = 0;
+        int child_index         = (long) cookie;
+        int copies              = 0;
+        int index               = 0;
+        int lockee_no           = 0;
+        afr_private_t       *priv = NULL;
+
+        priv = this->private;
+
+        copies = priv->child_count;
+        index = child_index % copies;
+        lockee_no = child_index / copies;
+
+        local    = frame->local;
+        int_lock = &local->internal_lock;
+
+        AFR_TRACE_ENTRYLK_OUT (frame, this, AFR_ENTRYLK_TRANSACTION,
+                               AFR_LOCK_OP,
+                               int_lock->lockee[lockee_no].basename, op_ret,
+                               op_errno, (long) cookie);
+
+	LOCK (&frame->lock);
+	{
+		if (op_ret < 0 ) {
+			if (op_errno == ENOSYS) {
+                        /* return ENOTSUP */
+				gf_log (this->name, GF_LOG_ERROR,
+					"subvolume does not support locking. "
+					"please load features/locks xlator on server");
+				local->op_ret         = op_ret;
+				int_lock->lock_op_ret = op_ret;
+
+				int_lock->lock_op_errno      = op_errno;
+				local->op_errno              = op_errno;
+			}
+		} else if (op_ret == 0) {
+			int_lock->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 successful. Proceed to call FOP */
+                if (int_lock->entrylk_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 successful. Unlock and try locking
+                   again, this time with serially blocking locks */
+                else {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "%d servers locked. Trying again with blocking calls",
+                                int_lock->lock_count);
+
+                        afr_unlock(frame, this);
+                }
+        }
+
+        return 0;
+}
+
+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;
+        int                 copies      = 0;
+        int                 index       = 0;
+        int                 lockee_no   = 0;
+        int32_t             call_count  = 0;
+        int i = 0;
+
+        local    = frame->local;
+        int_lock = &local->internal_lock;
+        priv     = this->private;
+
+        copies = priv->child_count;
+        initialize_entrylk_variables (frame, this);
+
+        if (local->fd) {
+                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);
+
+                        local->op_ret           = -1;
+                        int_lock->lock_op_ret   = -1;
+                        local->op_errno         = EINVAL;
+                        int_lock->lock_op_errno = EINVAL;
+
+			afr_unlock (frame, this);
+                        return -1;
+                }
+
+                call_count = int_lock->lockee_count * internal_lock_count (frame, this);
+                int_lock->lk_call_count = call_count;
+                int_lock->lk_expected_count = call_count;
+
+                if (!call_count) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "fd not open on any subvolumes. aborting.");
+                        afr_unlock (frame, this);
+                        goto out;
+                }
+
+                /* Send non-blocking entrylk calls only on up children
+                   and where the fd has been opened */
+                for (i = 0; i < 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[index],
+                                                   priv->children[index]->fops->fentrylk,
+                                                   this->name, local->fd,
+                                                   int_lock->lockee[lockee_no].basename,
+                                                   ENTRYLK_LOCK_NB, ENTRYLK_WRLCK,
+                                                   NULL);
+                                if (!--call_count)
+                                        break;
+                        }
+                }
+        } else {
+                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 < 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[index],
+                                                   priv->children[index]->fops->entrylk,
+                                                   this->name, &int_lock->lockee[lockee_no].loc,
+                                                   int_lock->lockee[lockee_no].basename,
+                                                   ENTRYLK_LOCK_NB, ENTRYLK_WRLCK,
+                                                   NULL);
+
+                                if (!--call_count)
+                                        break;
+                        }
+                }
+        }
+out:
+        return 0;
+}
+
+int32_t
+afr_nonblocking_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                             int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        afr_internal_lock_t *int_lock = NULL;
+        afr_inodelk_t       *inodelk  = NULL;
+        afr_local_t         *local    = NULL;
+        int call_count  = 0;
+        int child_index = (long) cookie;
+        afr_fd_ctx_t        *fd_ctx = NULL;
+
+
+        local    = frame->local;
+        int_lock = &local->internal_lock;
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
+
+        AFR_TRACE_INODELK_OUT (frame, this, AFR_INODELK_NB_TRANSACTION,
+                               AFR_LOCK_OP, NULL, op_ret,
+                               op_errno, (long) cookie);
+
+	if (local->fd)
+		fd_ctx = afr_fd_ctx_get (local->fd, this);
+
+        LOCK (&frame->lock);
+        {
+		if (op_ret < 0) {
+			if (op_errno == ENOSYS) {
+				/* return ENOTSUP */
+				gf_log (this->name, GF_LOG_ERROR,
+					"subvolume does not support locking. "
+					"please load features/locks xlator on "
+					"server");
+				local->op_ret                = op_ret;
+				int_lock->lock_op_ret        = op_ret;
+				int_lock->lock_op_errno      = op_errno;
+				local->op_errno              = op_errno;
+			}
+			if (local->transaction.eager_lock)
+				local->transaction.eager_lock[child_index] = 0;
+		} else {
+			inodelk->locked_nodes[child_index] |= LOCKED_YES;
+			inodelk->lock_count++;
+
+			if (local->transaction.eager_lock &&
+			    local->transaction.eager_lock[child_index] &&
+			    local->fd) {
+				/* piggybacked */
+				if (op_ret == 1) {
+					/* piggybacked */
+				} else if (op_ret == 0) {
+					/* lock acquired from server */
+                                        fd_ctx->lock_acquired[child_index]++;
+				}
+			}
+		}
+
+                call_count = --int_lock->lk_call_count;
+        }
+        UNLOCK (&frame->lock);
+
+        if (call_count == 0) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "Last inode locking reply received");
+                /* all locks successful. Proceed to call FOP */
+                if (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 successful. Unlock and try locking
+                   again, this time with serially blocking locks */
+                else {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "%d servers locked. Trying again with blocking calls",
+                                int_lock->lock_count);
+
+                        afr_unlock(frame, this);
+                }
+        }
+
+        return 0;
+}
+
+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;
+        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;
+
+        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) {
+                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);
+
+                        local->op_ret           = -1;
+                        int_lock->lock_op_ret   = -1;
+                        local->op_errno         = EINVAL;
+                        int_lock->lock_op_errno = EINVAL;
+
+			afr_unlock (frame, this);
+                        ret = -1;
+                        goto out;
+                }
+
+                call_count = internal_lock_count (frame, this);
+                int_lock->lk_call_count = call_count;
+                int_lock->lk_expected_count = call_count;
+
+                if (!call_count) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "fd not open on any subvolumes. aborting.");
+                        afr_unlock (frame, this);
+                        goto out;
+                }
+
+                /* Send non-blocking inodelk calls only on up children
+                   and where the fd has been opened */
+                for (i = 0; i < priv->child_count; i++) {
+                        if (!local->child_up[i])
+                                continue;
+
+                        flock_use = &flock;
+                        if (!local->transaction.eager_lock_on) {
+                                goto wind;
+                        }
+
+                        piggyback = 0;
+                        local->transaction.eager_lock[i] = 1;
+
+			afr_set_delayed_post_op (frame, this);
+
+                        LOCK (&local->fd->lock);
+                        {
+                                if (fd_ctx->lock_acquired[i]) {
+                                        fd_ctx->lock_piggyback[i]++;
+                                        piggyback = 1;
+                                }
+                        }
+                        UNLOCK (&local->fd->lock);
+
+                        if (piggyback) {
+                                /* (op_ret == 1) => indicate piggybacked lock */
+                                afr_nonblocking_inodelk_cbk (frame, (void *) (long) i,
+                                                             this, 1, 0, NULL);
+                                if (!--call_count)
+                                        break;
+                                continue;
+                        }
+                        flock_use = &full_flock;
+                wind:
+                        AFR_TRACE_INODELK_IN (frame, this,
+                                              AFR_INODELK_NB_TRANSACTION,
+                                              AFR_LOCK_OP, flock_use, F_SETLK, i);
+
+                        STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->finodelk,
+                                           int_lock->domain, local->fd,
+                                           F_SETLK, flock_use, NULL);
+
+                        if (!--call_count)
+                                break;
+                }
+        } else {
+                call_count = internal_lock_count (frame, this);
+                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])
+                                continue;
+                        AFR_TRACE_INODELK_IN (frame, this,
+                                              AFR_INODELK_NB_TRANSACTION,
+                                              AFR_LOCK_OP, &flock, F_SETLK, i);
+
+                        STACK_WIND_COOKIE (frame, afr_nonblocking_inodelk_cbk,
+                                           (void *) (long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->inodelk,
+                                           int_lock->domain, &local->loc,
+                                           F_SETLK, &flock, NULL);
+
+                        if (!--call_count)
+                                break;
+                }
+        }
+out:
+        return ret;
+}
+
+int32_t
+afr_unlock (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *local = NULL;
+
+        local = frame->local;
+
+        if (transaction_lk_op (local)) {
+                if (is_afr_lock_transaction (local))
+                        afr_unlock_inodelk (frame, this);
+                else
+                        afr_unlock_entrylk (frame, this);
+
+        } else {
+                if (is_afr_lock_selfheal (local))
+                        afr_unlock_inodelk (frame, this);
+                else
+                        afr_unlock_entrylk (frame, this);
+        }
+
+        return 0;
+}
+
+int
+afr_lk_transfer_datalock (call_frame_t *dst, call_frame_t *src, char *dom,
+                          unsigned int child_count)
+{
+        afr_local_t         *dst_local   = NULL;
+        afr_local_t         *src_local   = NULL;
+        afr_internal_lock_t *dst_lock    = NULL;
+        afr_internal_lock_t *src_lock    = NULL;
+        afr_inodelk_t       *dst_inodelk = NULL;
+        afr_inodelk_t       *src_inodelk = NULL;
+        int                 ret = -1;
+
+        src_local = src->local;
+        src_lock  = &src_local->internal_lock;
+        src_inodelk = afr_get_inodelk (src_lock, dom);
+        dst_local = dst->local;
+        dst_lock  = &dst_local->internal_lock;
+        dst_inodelk = afr_get_inodelk (dst_lock, dom);
+        if (!dst_inodelk || !src_inodelk)
+                goto out;
+        if (src_inodelk->locked_nodes) {
+                memcpy (dst_inodelk->locked_nodes, src_inodelk->locked_nodes,
+                        sizeof (*dst_inodelk->locked_nodes) * child_count);
+                memset (src_inodelk->locked_nodes, 0,
+                        sizeof (*src_inodelk->locked_nodes) * child_count);
+        }
+
+        dst_lock->transaction_lk_type = src_lock->transaction_lk_type;
+        dst_lock->selfheal_lk_type    = src_lock->selfheal_lk_type;
+        dst_inodelk->lock_count = src_inodelk->lock_count;
+        src_inodelk->lock_count = 0;
+        ret = 0;
+out:
+        return ret;
+}
diff --git a/xlators/cluster/afr/src/afr-mem-types.h b/xlators/cluster/afr/src/afr-mem-types.h
new file mode 100644
index 000000000..05df90cc0
--- /dev/null
+++ b/xlators/cluster/afr/src/afr-mem-types.h
@@ -0,0 +1,49 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef __AFR_MEM_TYPES_H__
+#define __AFR_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+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_private_t,
+        gf_afr_mt_int32_t,
+        gf_afr_mt_char,
+        gf_afr_mt_xattr_key,
+        gf_afr_mt_dict_t,
+        gf_afr_mt_xlator_t,
+        gf_afr_mt_iatt,
+        gf_afr_mt_int,
+        gf_afr_mt_afr_node_character,
+        gf_afr_mt_sh_diff_loop_state,
+        gf_afr_mt_uint8_t,
+        gf_afr_mt_loc_t,
+        gf_afr_mt_entry_name,
+        gf_afr_mt_pump_priv,
+        gf_afr_mt_locked_fd,
+        gf_afr_mt_inode_ctx_t,
+        gf_afr_fd_paused_call_t,
+        gf_afr_mt_crawl_data_t,
+        gf_afr_mt_brick_pos_t,
+        gf_afr_mt_shd_bool_t,
+        gf_afr_mt_shd_timer_t,
+        gf_afr_mt_shd_event_t,
+        gf_afr_mt_time_t,
+        gf_afr_mt_pos_data_t,
+	gf_afr_mt_reply_t,
+	gf_afr_mt_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 e60ec70c0..f86aa7fd8 100644
--- a/xlators/cluster/afr/src/afr-open.c
+++ b/xlators/cluster/afr/src/afr-open.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include <libgen.h>
@@ -53,433 +44,294 @@
 #include "afr-dir-write.h"
 #include "afr-transaction.h"
 
-#include "afr-self-heal.h"
-
-
-int
-afr_open_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-			int32_t op_ret, int32_t op_errno, struct stat *prebuf,
-                        struct stat *postbuf)
-{
-	afr_local_t * local = frame->local;
 
-	AFR_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
-			  local->fd);
-	return 0;
-}
-
-
-int
-afr_open_cbk (call_frame_t *frame, void *cookie,
-	      xlator_t *this, int32_t op_ret, int32_t op_errno,
-	      fd_t *fd)
+gf_boolean_t
+afr_is_fd_fixable (fd_t *fd)
 {
-	afr_local_t *  local = NULL;
-	afr_private_t * priv = NULL;
-
-        int child_index = (long) cookie;
-
-        uint64_t      ctx;
-        afr_fd_ctx_t *fd_ctx;
-
-        int ret = 0;
-
-	int call_count = -1;
-
-	priv  = this->private;
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-		}
-
-		if (op_ret >= 0) {
-			local->op_ret = op_ret;
-                        local->success_count++;
-
-                        ret = afr_fd_ctx_set (this, fd);
-
-                        if (ret < 0) {
-                                gf_log (this->name, GF_LOG_DEBUG,
-                                        "could not set fd ctx for fd=%p",
-                                        fd);
-
-                                local->op_ret   = -1;
-                                local->op_errno = -ret;
-                        }
-
-                        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);
-                                local->op_ret   = -1;
-                                local->op_errno = -ret;
-                        }
-
-                        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                        fd_ctx->opened_on[child_index] = 1;
-                        fd_ctx->flags                  = local->cont.open.flags;
-                        fd_ctx->wbflags                = local->cont.open.wbflags;
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (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;
+        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 (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
-          fd_t *fd, int32_t wbflags)
+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_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;
+        afr_local_t * local = frame->local;
 
-        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_up_down_flush_open_cbk (call_frame_t *frame, void *cookie,
-                            xlator_t *this, int32_t op_ret, int32_t op_errno,
-                            fd_t *fd)
+afr_open_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_private_t *priv = NULL;
+        afr_local_t *  local       = NULL;
+        int            call_count  = -1;
+        int            child_index = (long) cookie;
+	afr_fd_ctx_t  *fd_ctx = NULL;
 
-        int ret = 0;
-
-        uint64_t      ctx;
-        afr_fd_ctx_t *fd_ctx;
-
-        int call_count  = 0;
-        int child_index = (long) cookie;
-
-        priv  = this->private;
         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) {
-                local->transaction.post_post_op (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 char *opened_on, unsigned char *child_up)
-{
-        int i;
-        int count = 0;
-
-        for (i = 0; i < child_count; i++) {
-                if (!opened_on[i] && child_up[i])
-                        count++;
-        }
-
-        return count;
-}
-
-
 int
-afr_up_down_flush_sh_unwind (call_frame_t *frame, xlator_t *this)
+afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+          fd_t *fd, dict_t *xdata)
 {
-        afr_local_t *local  = NULL;
-        afr_private_t *priv = NULL;
-
-        uint64_t      ctx;
-        afr_fd_ctx_t *fd_ctx;
+        afr_private_t * priv       = NULL;
+        afr_local_t *   local      = NULL;
+        int             i          = 0;
+        int32_t         call_count = 0;
+        int32_t         op_errno   = 0;
+	afr_fd_ctx_t   *fd_ctx = NULL;
 
-        int abandon = 0;
-        int ret = 0;
-        int i;
-        int call_count = 0;
-
-        priv  = this->private;
-        local = frame->local;
+        //We can't let truncation to happen outside transaction.
 
-        /*
-         * 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.
-         */
+        priv = this->private;
 
-        ret = fd_ctx_get (local->fd, this, &ctx);
-
-        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;
 
-        call_count = __unopened_count (priv->child_count, fd_ctx->opened_on,
-                                       local->child_up);
+	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_up_down_flush_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.post_post_op (frame, this);
+	AFR_STACK_UNWIND (open, frame, -1, op_errno, fd, NULL);
 
         return 0;
 }
 
-
 int
-afr_up_down_flush_post_post_op (call_frame_t *frame, xlator_t *this)
+afr_openfd_fix_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int32_t op_ret, int32_t op_errno, fd_t *fd,
+                         dict_t *xdata)
 {
-        afr_private_t *priv = NULL;
-        afr_local_t *local  = NULL;
-        afr_self_heal_t *sh = NULL;
-
-        priv  = this->private;
-        local = frame->local;
-        sh    = &local->self_heal;
-
-        inode_path (local->fd->inode, NULL, (char **)&local->loc.path);
-        local->loc.name   = strrchr (local->loc.path, '/');
-        local->loc.inode  = inode_ref (local->fd->inode);
-        local->loc.parent = inode_parent (local->fd->inode, 0, NULL);
-
-        /* forcibly trigger missing-entries self-heal */
+        afr_local_t   *local      = NULL;
+        afr_private_t *priv       = NULL;
+        afr_fd_ctx_t  *fd_ctx     = NULL;
+        int           call_count  = 0;
+        int           child_index = (long) cookie;
+
+        priv     = this->private;
+        local    = frame->local;
+
+        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);
+        }
 
-        local->success_count    = 1;
-        local->enoent_count     = 1;
+        fd_ctx = local->fd_ctx;
 
-        sh->data_lock_held      = _gf_true;
-        sh->need_data_self_heal = _gf_true;
-        sh->mode                = local->fd->inode->st_mode;
-        sh->background          = _gf_false;
-        sh->unwind              = afr_up_down_flush_sh_unwind;
+        LOCK (&local->fd->lock);
+        {
+                if (op_ret >= 0) {
+                        fd_ctx->opened_on[child_index] = AFR_FD_OPENED;
+                } else {
+                        fd_ctx->opened_on[child_index] = AFR_FD_NOT_OPENED;
+                }
+        }
+        UNLOCK (&local->fd->lock);
 
-        afr_self_heal (frame, this);
+        call_count = afr_frame_return (frame);
+        if (call_count == 0)
+                AFR_STACK_DESTROY (frame);
 
         return 0;
 }
 
 
-int
-afr_up_down_flush_wind (call_frame_t *frame, xlator_t *this)
+static int
+afr_fd_ctx_need_open (fd_t *fd, xlator_t *this, unsigned char *need_open)
 {
-	afr_local_t *local = NULL;
+	afr_fd_ctx_t *fd_ctx = NULL;
 	afr_private_t *priv = NULL;
+	int i = 0;
+	int count = 0;
 
-	local = frame->local;
-	priv  = this->private;
+	priv = this->private;
 
-        local->transaction.resume (frame, this);
-	return 0;
-}
+	fd_ctx = afr_fd_ctx_get (fd, this);
+	if (!fd_ctx)
+		return 0;
 
+	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);
 
-int
-afr_up_down_flush_done (call_frame_t *frame, xlator_t *this)
-{
-        afr_private_t *priv = NULL;
-	afr_local_t *local  = NULL;
+	return count;
+}
 
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
 
-        int _ret = -1;
-        int i    = 0;
+void
+afr_fix_open (fd_t *fd, xlator_t *this)
+{
+        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;
-
-        LOCK (&local->fd->lock);
-        {
-                _ret = __fd_ctx_get (local->fd, this, &ctx);
-
-                if (_ret < 0) {
-                        goto out;
-                }
 
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                fd_ctx->down_count = priv->down_count;
-                fd_ctx->up_count   = priv->up_count;
+        if (!afr_is_fd_fixable (fd))
+                goto out;
 
-                for (i = 0; i < priv->child_count; i++) {
-                        if (local->child_up[i])
-                                fd_ctx->pre_op_done[i] = 0;
-                }
-        }
-out:
-        UNLOCK (&local->fd->lock);
+        fd_ctx = afr_fd_ctx_get (fd, this);
+        if (!fd_ctx)
+                goto out;
 
-        local->up_down_flush_cbk (frame, this);
+	need_open = alloca0 (priv->child_count);
 
-	return 0;
-}
+	call_count = afr_fd_ctx_need_open (fd, this, need_open);
+	if (!call_count)
+		goto out;
 
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame)
+                goto out;
 
-int
-afr_up_down_flush (call_frame_t *frame, xlator_t *this, fd_t *fd,
-                   afr_flush_type type)
-{
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local)
+		goto out;
 
-	int op_ret   = -1;
+        local->loc.inode = inode_ref (fd->inode);
+        ret = loc_path (&local->loc, NULL);
+        if (ret < 0)
+                goto out;
 
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
+        local->fd = fd_ref (fd);
+	local->fd_ctx = fd_ctx;
 
-	priv = this->private;
+        local->call_count = call_count;
 
-        local = frame->local;
+        gf_log (this->name, GF_LOG_DEBUG, "need open count: %d",
+                call_count);
 
-        local->op = GF_FOP_FLUSH;
+        for (i = 0; i < priv->child_count; i++) {
+                if (!need_open[i])
+                        continue;
 
-//        local->fd = fd_ref (local->fd);
+                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);
 
-        local->transaction.fop          = afr_up_down_flush_wind;
-        local->transaction.done         = afr_up_down_flush_done;
+                        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);
 
-        switch (type) {
-        case AFR_CHILD_UP_FLUSH:
-                local->transaction.post_post_op = afr_up_down_flush_post_post_op;
-                break;
+                        STACK_WIND_COOKIE (frame, afr_openfd_fix_open_cbk,
+                                           (void *)(long) i,
+                                           priv->children[i],
+                                           priv->children[i]->fops->open,
+                                           &local->loc,
+                                           fd_ctx->flags & (~O_TRUNC),
+                                           local->fd, NULL);
+                }
 
-        case AFR_CHILD_DOWN_FLUSH:
-                local->transaction.post_post_op = NULL;
-                break;
+		if (!--call_count)
+			break;
         }
 
-        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_FLUSH_TRANSACTION);
-
-	op_ret = 0;
+	return;
 out:
-	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 73a878247..000000000
--- a/xlators/cluster/afr/src/afr-self-heal-algorithm.c
+++ /dev/null
@@ -1,941 +0,0 @@
-/*
-   Copyright (c) 2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
-*/
-
-
-#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_private_t *             priv    = NULL;
-        afr_local_t *               local   = NULL;
-        afr_self_heal_t *           sh      = NULL;
-        afr_sh_algo_full_private_t *sh_priv = NULL;
-
-        priv  = this->private;
-        local = frame->local;
-        sh    = &local->self_heal;
-
-        sh_priv = sh->private;
-
-        if (sh_priv)
-                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_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_full_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;
-
-        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 stat *prebuf,
-                   struct stat *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 stat *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       = sh->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_private_t *             priv    = NULL;
-        afr_local_t *               local   = NULL;
-        afr_self_heal_t *           sh      = NULL;
-        afr_sh_algo_full_private_t *sh_priv = NULL;
-
-        priv  = this->private;
-        local = frame->local;
-        sh    = &local->self_heal;
-
-        sh_priv = CALLOC (1, sizeof (*sh_priv));
-
-        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;
-
-        priv  = this->private;
-        local = frame->local;
-        sh    = &local->self_heal;
-
-        sh_priv = sh->private;
-
-        if (sh_priv) {
-                if (sh_priv->checksum)
-                        FREE (sh_priv->checksum);
-
-                FREE (sh_priv);
-        }
-}
-
-
-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;
-}
-
-
-struct sh_diff_loop_state {
-        off_t   offset;
-        int32_t child_index;
-        unsigned char *write_needed;
-};
-
-
-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;
-
-        LOCK (&sh_priv->lock);
-        {
-                sh_priv->loops_running--;
-        }
-        UNLOCK (&sh_priv->lock);
-
-        gf_log (this->name, GF_LOG_TRACE,
-                "loop for offset %"PRId64" returned", loop_state->offset);
-
-        AFR_STACK_DESTROY (rw_frame);
-
-        if (loop_state) {
-                if (loop_state->write_needed)
-                        FREE (loop_state->write_needed);
-
-                FREE (loop_state);
-        }
-
-        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 stat *buf,
-                   struct stat *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;
-
-        struct sh_diff_loop_state *loop_state = (struct sh_diff_loop_state *) 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, loop_state->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[loop_state->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 stat *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;
-
-        struct sh_diff_loop_state *loop_state = (struct sh_diff_loop_state *) cookie;
-
-	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;
-
-	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, sh->offset);
-
-	if (op_ret <= 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]) {
-                        STACK_WIND_COOKIE (rw_frame, sh_diff_write_cbk,
-                                           (void *) (long) loop_state,
-                                           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,
-              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;
-
-        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;
-
-	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;
-
-	STACK_WIND_COOKIE (rw_frame, sh_diff_read_cbk,
-			   (void *) (long) loop_state,
-			   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;
-
-        struct sh_diff_loop_state *loop_state = (struct sh_diff_loop_state *) cookie;
-
-        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;
-
-        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[loop_state->child_index]->name,
-                        strerror (op_errno));
-
-                sh->op_failed = 1;
-        } else {
-                memcpy ((void *) sh_priv->checksum + loop_state->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 ((const void *) sh_priv->checksum + (i * MD5_DIGEST_LEN),
-                                    (const void *) sh_priv->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;
-                        }
-                }
-
-                if (write_needed && !sh->op_failed) {
-                        sh_diff_read (rw_frame, this, loop_state);
-                } else {
-                        sh->offset += sh_priv->block_size;
-
-                        sh_diff_loop_return (rw_frame, this, loop_state);
-                }
-        }
-
-        return 0;
-}
-
-
-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;
-
-        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_state = CALLOC (1, sizeof (*loop_state));
-        loop_state->child_index  = sh->source;
-        loop_state->offset       = offset;
-        loop_state->write_needed = CALLOC (priv->child_count,
-                                           sizeof (*loop_state->write_needed));
-
-        STACK_WIND_COOKIE (rw_frame, sh_diff_checksum_cbk,
-                           (void *) (long) loop_state,
-                           priv->children[sh->source],
-                           priv->children[sh->source]->fops->rchecksum,
-                           sh->healing_fd,
-                           sh->offset, sh_priv->block_size);
-
-        for (i = 0; i < priv->child_count; i++) {
-                if (sh->sources[i] || !local->child_up[i])
-                        continue;
-
-                STACK_WIND_COOKIE (rw_frame, sh_diff_checksum_cbk,
-                                   (void *) (long) loop_state,
-                                   priv->children[i],
-                                   priv->children[i]->fops->rchecksum,
-                                   sh->healing_fd,
-                                   sh->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;
- 
-	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,
-                                "diff self-heal aborting on %s",
-                                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,
-                                "full self-heal completed on %s",
-                                local->loc.path);
-
-
-                        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;
-
-        priv  = this->private;
-        local = frame->local;
-        sh    = &local->self_heal;
-
-        sh_priv = CALLOC (1, sizeof (*sh_priv));
-
-        sh_priv->checksum = CALLOC (priv->child_count, MD5_DIGEST_LEN);
-
-        sh_priv->block_size = this->ctx->page_size;
-
-        sh->private = sh_priv;
-
-        LOCK_INIT (&sh_priv->lock);
-
-        local->call_count = 0;
-
-        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 7ef324477..000000000
--- a/xlators/cluster/afr/src/afr-self-heal-algorithm.h
+++ /dev/null
@@ -1,51 +0,0 @@
-/*
-   Copyright (c) 2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
-*/
-
-#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;
-
-typedef struct {
-        uint8_t *checksum;     /* array of MD5 checksums for each child
-                                  Each checksum is MD5_DIGEST_LEN bytes long */
-
-        size_t block_size;
-
-        gf_lock_t lock;
-        unsigned int loops_running;
-        off_t offset;
-} 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 ef36be145..4dac83113 100644
--- a/xlators/cluster/afr/src/afr-self-heal-common.c
+++ b/xlators/cluster/afr/src/afr-self-heal-common.c
@@ -1,593 +1,383 @@
 /*
-  Copyright (c) 2008-2009 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 General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
-#include "glusterfs.h"
-#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 "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;
 
-	char *buf = NULL;
-	char *ptr = NULL;
+	priv = this->private;
+	idx = afr_index_for_transaction_type (type);
 
-	int i, j;
+	xattr = dict_new ();
+	if (!xattr)
+		return NULL;
 
-        /* 10 digits per entry + 1 space + '[' and ']' */
-	buf = MALLOC (priv->child_count * 11 + 8); 
+	if (output_dirty[subvol]) {
+		/* clear dirty */
+		raw = GF_CALLOC (sizeof(int), AFR_NUM_CHANGE_LOGS, gf_afr_mt_int32_t);
+		if (!raw)
+			goto err;
 
-	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]);
-		}
-		ptr += sprintf (ptr, "]");
-		gf_log (this->name, GF_LOG_TRACE,
-			"pending_matrix: %s", buf);
+		raw[idx] = hton32 (output_dirty[subvol]);
+		ret = dict_set_bin (xattr, AFR_DIRTY, raw,
+				    sizeof(int) * AFR_NUM_CHANGE_LOGS);
+		if (ret)
+			goto err;
 	}
 
-	FREE (buf);
-}
-
-
-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 i, j, k;
-
-	int32_t *pending = NULL;
-	int ret = -1;
-
-        unsigned char *ignorant_subvols = NULL;
+	/* clear/set pending */
+	for (j = 0; j < priv->child_count; j++) {
+		if (!output_matrix[subvol][j])
+			continue;
 
-        ignorant_subvols = CALLOC (sizeof (*ignorant_subvols), child_count);
+		raw = GF_CALLOC (sizeof(int), AFR_NUM_CHANGE_LOGS,
+				 gf_afr_mt_int32_t);
+		if (!raw)
+			goto err;
 
-	/* start clean */
-	for (i = 0; i < child_count; i++) {
-		for (j = 0; j < child_count; j++) {
-			pending_matrix[i][j] = 0;
-		}
-	}
+		raw[idx] = hton32 (output_matrix[subvol][j]);
 
-	for (i = 0; i < child_count; i++) {
-		pending = NULL;
-
-                for (j = 0; j < child_count; j++) {
-                        ret = dict_get_ptr (xattr[i], priv->pending_key[j],
-                                            VOID(&pending));
-                        
-                        if (ret != 0) {
-                                /*
-                                 * There is no xattr present. This means this
-                                 * subvolume should be considered an 'ignorant'
-                                 * subvolume.
-                                 */
-
-                                ignorant_subvols[i] = 1;
-                                continue;
-                        }
-
-                        k = afr_index_for_transaction_type (type);
-                        
-                        pending_matrix[i][j] = ntoh32 (pending[k]);
-                }
+		ret = dict_set_bin (xattr, priv->pending_key[j],
+				    raw, sizeof(int) * AFR_NUM_CHANGE_LOGS);
+		if (ret)
+			goto err;
 	}
 
-        /*
-         * 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;
-                        }
-                }
-        }
-
-        FREE (ignorant_subvols);
+	return xattr;
+err:
+	if (xattr)
+		dict_unref (xattr);
+	return NULL;
 }
 
 
-/**
- * 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
+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   = 0;
-        int ret = 1;   /* innocent until proven guilty */
-
-        for (i = 0; i < child_count; i++) {
-                if (array[i]) {
-                        ret = 0;
-                        break;
-                }
-        }
-
-        return ret;
-}
+	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;
 
+	priv = this->private;
 
-static int
-afr_sh_is_fool (int32_t *array, int i, int child_count)
-{
-        return array[i];   /* fool if accuses itself */
-}
+	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);
 
-static int
-afr_sh_is_wise (int32_t *array, int i, int child_count)
-{
-        return !array[i];  /* wise if does not accuse itself */
-}
+	afr_selfheal_extract_xattr (this, replies, type, input_dirty,
+				    input_matrix);
 
+	for (i = 0; i < priv->child_count; i++)
+		if (sinks[i] && !healed_sinks[i])
+			pending[i] = 1;
 
-static int
-afr_sh_all_nodes_innocent (afr_node_character *characters, 
-                           int child_count)
-{
-        int i   = 0;
-        int ret = 1;
+	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++) {
-                if (characters[i].type != AFR_NODE_INNOCENT) {
-                        ret = 0;
-                        break;
-                }
-        }
+	for (i = 0; i < priv->child_count; i++) {
+		if (!pending[i])
+			output_dirty[i] = -input_dirty[i];
+	}
 
-        return ret;
-}
+	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;
 
+		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;
+		}
 
-static int
-afr_sh_wise_nodes_exist (afr_node_character *characters, int child_count)
-{
-        int i   = 0;
-        int ret = 0;
+		afr_selfheal_post_op (frame, this, inode, i, xattr);
 
-        for (i = 0; i < child_count; i++) {
-                if (characters[i].type == AFR_NODE_WISE) {
-                        ret = 1;
-                        break;
-                }
-        }
+		dict_unref (xattr);
+	}
 
-        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)
+void
+afr_replies_copy (struct afr_reply *dst, struct afr_reply *src, int count)
 {
-        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;
+	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_wise_nodes_conflict (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;
+	void *pending_raw = NULL;
+	int pending[3] = {0, };
 
-        for (i = 0; i < child_count; i++) {
-                if ((characters[i].type == AFR_NODE_WISE)
-                    && characters[i].wisdom == 1) {
+	if (dict_get_ptr (xdata, AFR_DIRTY, &pending_raw))
+		return -1;
 
-                        /* There is atleast one bona-fide wise node */
-                        ret = 0;
-                        break;
-                }
-        }
+	if (!pending_raw)
+		return -1;
 
-        return ret;
-}
+	memcpy (pending, pending_raw, sizeof(pending));
 
+	dirty[subvol] = 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)
-{
-        int32_t ** pending_matrix;
-        int i, j;
-
-        int size_differs = 0;
-
-        pending_matrix = sh->pending_matrix;
-
-        for (i = 0; i < child_count; i++) {
-                for (j = 0; j < child_count; j++) {
-                        if (!sh->buf)
-                                break;
-
-                        if (SIZE_DIFFERS (&sh->buf[i], &sh->buf[j])
-                            && (pending_matrix[i][j] == 0)
-                            && (pending_matrix[j][i] == 0)) {
-
-                                pending_matrix[i][j] = 1;
-                                pending_matrix[j][i] = 1;
-
-                                size_differs = 1;
-                        }
-                }
-        }
-
-        return size_differs;
-}
-
-        
-static int
-afr_sh_mark_biggest_fool_as_source (afr_self_heal_t *sh,
-                                    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 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;
-}
+	int i = 0;
+	void *pending_raw = NULL;
+	int pending[3] = {0, };
+	afr_private_t *priv = NULL;
 
+	priv = this->private;
 
-static int
-afr_sh_mark_biggest_as_source (afr_self_heal_t *sh, int child_count)
-{
-        int biggest = 0;
-        int i;
+	for (i = 0; i < priv->child_count; i++) {
+		if (dict_get_ptr (xdata, priv->pending_key[i], &pending_raw))
+			continue;
 
-        for (i = 0; i < child_count; i++) {
-                if (!sh->buf)
-                        break;
+		if (!pending_raw)
+			continue;
 
-                if (SIZE_GREATER (&sh->buf[i], &sh->buf[biggest])) {
-                        biggest = i;
-                }
-        }
+		memcpy (pending, pending_raw, sizeof(pending));
 
-        sh->sources[biggest] = 1;
+		matrix[subvol][i] = ntoh32 (pending[idx]);
+	}
 
-        return 1;
+	return 0;
 }
 
 
 int
-afr_sh_mark_sources (afr_self_heal_t *sh, int child_count,
-                     afr_self_heal_type type)
+afr_selfheal_extract_xattr (xlator_t *this, struct afr_reply *replies,
+			    afr_transaction_type type, int *dirty, int **matrix)
 {
+	afr_private_t *priv = NULL;
 	int i = 0;
+	dict_t *xdata = NULL;
+	int idx = -1;
 
-        int32_t ** pending_matrix;
-        int *      sources;
+	idx = afr_index_for_transaction_type (type);
 
-        int size_differs = 0;
-
-        pending_matrix = sh->pending_matrix;
-        sources        = sh->sources;
+	priv = this->private;
 
-	int nsources = 0;
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].xdata)
+			continue;
 
-        /* stores the 'characters' (innocent, fool, wise) of the nodes */
-        afr_node_character *
-                characters = CALLOC (sizeof (afr_node_character), 
-                                     child_count);
+		xdata = replies[i].xdata;
 
-	/* start clean */
-	for (i = 0; i < child_count; i++) {
-		sources[i] = 0;
+		afr_selfheal_fill_dirty (this, dirty, i, idx, xdata);
+		afr_selfheal_fill_matrix (this, matrix, i, idx, xdata);
 	}
-        
-        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 (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);
-        }
-
-out:
-        FREE (characters);
 
-	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)
-{
-	int i = 0;
-	int j = 0;
-        int k = 0;
-
-        int32_t * pending = NULL;
-        int       ret     = 0;
-
-	/* start clean */
-	for (i = 0; i < child_count; i++) {
-		for (j = 0; j < child_count; j++) {
-			delta_matrix[i][j] = 0;
-		}
-	}
-
-	for (i = 0; i < child_count; i++) {
-                pending = NULL;
-
-                for (j = 0; j < child_count; j++) {
-                        ret = dict_get_ptr (xattr[i], priv->pending_key[j],
-                                            VOID(&pending));
-                        
-                        if (!success[j])
-                                continue;
-
-                        k = afr_index_for_transaction_type (type);
-                        
-                        if (pending) {
-                                delta_matrix[i][j] = -(ntoh32 (pending[k]));
-                        } else {
-                                delta_matrix[i][j]  = 0;
-                        }
-
-                }
-	}
-}
 
+/*
+ * 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.
+ *
+ */
 
 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_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)
 {
+	afr_private_t *priv = NULL;
 	int i = 0;
 	int j = 0;
-        int k = 0;
+	int *dirty = NULL;
+	int **matrix = NULL;
+	char *accused = NULL;
 
-	int ret = 0;
+	priv = this->private;
 
-	int32_t *pending = 0;
+	dirty = alloca0 (priv->child_count * sizeof (int));
+	accused = alloca0 (priv->child_count);
+	matrix = ALLOC_MATRIX(priv->child_count, int);
 
-	for (i = 0; i < child_count; i++) {
-		if (!xattr[i])
-			continue;
+	/* 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;
+		}
+	}
 
-		for (j = 0; j < child_count; j++) {
-                        pending = CALLOC (sizeof (int32_t), 3);
-                        /* 3 = data+metadata+entry */
+	/* 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;
+	}
 
-                        k = afr_index_for_transaction_type (type);
+	/* 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;
+		}
+	}
 
-			pending[k] = hton32 (delta_matrix[i][j]);
+	/* 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;
+		}
+	}
 
-                        ret = dict_set_bin (xattr[i], priv->pending_key[j], 
-                                            pending,
-                                            3 * sizeof (int32_t));
+	/* 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;
 		}
 	}
 
@@ -596,954 +386,624 @@ afr_sh_delta_to_xattr (afr_private_t *priv,
 
 
 int
-afr_sh_has_metadata_pending (dict_t *xattr, int child_count, xlator_t *this)
+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)
 {
-	afr_private_t *priv = NULL;
-	int32_t       *pending = NULL;
-	void          *tmp_pending = NULL; /* This is required to remove 'type-punned' warnings from gcc */
-
-	int           ret = -1;
-	int            i  = 0;
-        int            j  = 0;
-
-	priv = this->private;
-
-        for (i = 0; i < priv->child_count; i++) {
-                ret = dict_get_ptr (xattr, priv->pending_key[i],
-                                    &tmp_pending);
+	afr_local_t *local = NULL;
+	int i = -1;
 
-                if (ret != 0)
-                        return 0;
-                
-                pending = tmp_pending;
+	local = frame->local;
+	i = (long) cookie;
 
-                j = afr_index_for_transaction_type (AFR_METADATA_TRANSACTION);
+	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);
 
-                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)
+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)
 {
+	loc_t loc = {0, };
+	dict_t *xattr_req = NULL;
+	afr_local_t *local = NULL;
 	afr_private_t *priv = NULL;
-	int32_t       *pending = NULL;
-	void          *tmp_pending = NULL; /* This is required to remove 'type-punned' warnings from gcc */
+	inode_t *inode = NULL;
 
-	int           ret = -1;
-	int            i  = 0;
-        int            j  = 0;
+	local = frame->local;
+	priv = frame->this->private;
 
-	priv = this->private;
+	xattr_req = dict_new ();
+	if (!xattr_req)
+		return NULL;
 
-        for (i = 0; i < priv->child_count; i++) {
-                ret = dict_get_ptr (xattr, priv->pending_key[i],
-                                    &tmp_pending);
+	if (afr_xattr_req_prepare (frame->this, xattr_req) != 0) {
+		dict_destroy (xattr_req);
+		return NULL;
+	}
 
-                if (ret != 0)
-                        return 0;
-                
-                pending = tmp_pending;
+	inode = inode_new (parent->table);
+	if (!inode) {
+		dict_destroy (xattr_req);
+		return NULL;
+	}
 
-                j = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
+	loc.parent = inode_ref (parent);
+	uuid_copy (loc.pargfid, parent->gfid);
+	loc.name = name;
+	loc.inode = inode_ref (inode);
 
-                if (pending[j])
-                        return 1;
-        }
+	AFR_ONLIST (lookup_on, frame, afr_selfheal_discover_cbk, lookup, &loc,
+		    xattr_req);
 
-	return 0;
+	afr_replies_copy (replies, local->replies, priv->child_count);
+
+	loc_wipe (&loc);
+	dict_unref (xattr_req);
+
+	return inode;
 }
 
 
 int
-afr_sh_has_entry_pending (dict_t *xattr, int child_count, xlator_t *this)
+afr_selfheal_unlocked_discover_on (call_frame_t *frame, inode_t *inode,
+				   uuid_t gfid, struct afr_reply *replies,
+				   unsigned char *discover_on)
 {
-        afr_private_t *priv = NULL;
-	int32_t       *pending = NULL;
-	void          *tmp_pending = NULL; /* This is required to remove 'type-punned' warnings from gcc */
+	loc_t loc = {0, };
+	dict_t *xattr_req = NULL;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
 
-	int           ret = -1;
-	int            i  = 0;
-        int            j  = 0;
+	local = frame->local;
+	priv = frame->this->private;
 
-	priv = this->private;
+	xattr_req = dict_new ();
+	if (!xattr_req)
+		return -ENOMEM;
+
+	if (afr_xattr_req_prepare (frame->this, xattr_req) != 0) {
+		dict_destroy (xattr_req);
+		return -ENOMEM;
+	}
 
-        for (i = 0; i < priv->child_count; i++) {
-                ret = dict_get_ptr (xattr, priv->pending_key[i],
-                                    &tmp_pending);
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, gfid);
 
-                if (ret != 0)
-                        return 0;
-                
-                pending = tmp_pending;
+	AFR_ONLIST (discover_on, frame, afr_selfheal_discover_cbk, lookup, &loc,
+		    xattr_req);
 
-                j = afr_index_for_transaction_type (AFR_ENTRY_TRANSACTION);
+	afr_replies_copy (replies, local->replies, priv->child_count);
 
-                if (pending[j])
-                        return 1;
-        }
+	loc_wipe (&loc);
+	dict_unref (xattr_req);
 
 	return 0;
 }
 
-
-/**
- * is_matrix_zero - return true if pending matrix is all zeroes
- */
-
 int
-afr_sh_is_matrix_zero (int32_t *pending_matrix[], int child_count)
+afr_selfheal_unlocked_discover (call_frame_t *frame, inode_t *inode,
+				uuid_t gfid, struct afr_reply *replies)
 {
-	int i, j;
+	afr_private_t *priv = NULL;
+
+	priv = frame->this->private;
 
-	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_unlocked_discover_on (frame, inode, gfid, replies,
+						  priv->child_up);
 }
 
 
 int
-afr_sh_missing_entries_done (call_frame_t *frame, xlator_t *this)
+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;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              i = 0;
+	afr_local_t *local = 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 stat) * priv->child_count);
+	i = (long) cookie;
 
-        for (i = 0; i < priv->child_count; i++) {
-                sh->locked_nodes[i] = 0;
-        }
+	local->replies[i].valid = 1;
+	local->replies[i].op_ret = op_ret;
+	local->replies[i].op_errno = op_errno;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->xattr[i])
-			dict_unref (sh->xattr[i]);
-		sh->xattr[i] = NULL;
-	}
-
-	if (local->govinda_gOvinda) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"aborting selfheal of %s",
-			local->loc.path);
-		sh->completion_cbk (frame, this);
-	} else {
-		gf_log (this->name, GF_LOG_TRACE,
-			"proceeding to metadata check on %s",
-			local->loc.path);
-		afr_self_heal_metadata (frame, this);
-	}
+	syncbarrier_wake (&local->barrier);
 
 	return 0;
 }
 
 
 int
-sh_missing_entries_unlck_cbk (call_frame_t *frame, void *cookie,
-			      xlator_t *this,
-			      int32_t op_ret, int32_t op_errno)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              call_count = 0;
-
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	LOCK (&frame->lock);
-	{
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-		afr_sh_missing_entries_done (frame, this);
-	}
-
-	return 0;
-}
-			      
-
-static int
-sh_missing_entries_finish (call_frame_t *frame, xlator_t *this)
+afr_selfheal_locked_fill (call_frame_t *frame, xlator_t *this,
+			  unsigned char *locked_on)
 {
-	afr_private_t      *priv = NULL;
-	afr_local_t        *local = NULL;
-	int                 i = 0;
-	int                 call_count = 0;
-	afr_self_heal_t    *sh = NULL;
-
+	int i = 0;
+	afr_private_t *priv = NULL;
+	afr_local_t *local = NULL;
+	int count = 0;
 
 	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (sh->locked_nodes[i])
-                        call_count++;
-        }
-
-        if (call_count == 0) {
-                afr_sh_missing_entries_done (frame, this);
-                return 0;
-        }
-
-	local->call_count = call_count;
-
 	for (i = 0; i < priv->child_count; i++) {
-		if (sh->locked_nodes[i]) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"unlocking %"PRId64"/%s on subvolume %s",
-				sh->parent_loc.inode->ino, local->loc.name,
-				priv->children[i]->name);
-
-			STACK_WIND (frame, sh_missing_entries_unlck_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->entrylk,
-                                    this->name,
-				    &sh->parent_loc, local->loc.name,
-				    ENTRYLK_UNLOCK, ENTRYLK_WRLCK);
-
-			if (!--call_count)
-				break;
+		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;
 }
 
 
-static int
-sh_destroy_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		int32_t op_ret, int op_errno,
-                struct stat *preop, struct stat *postop)
+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)
 {
-        afr_local_t *local = NULL;
+	loc_t loc = {0,};
+	struct gf_flock flock = {0, };
 
-        int call_count = 0;
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-        local = frame->local;
+	flock.l_type = F_WRLCK;
+	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_ONALL (frame, afr_selfheal_lock_cbk, inodelk, dom,
+		    &loc, F_SETLK, &flock, NULL);
 
-        call_count = afr_frame_return (frame);
-        
-        if (call_count == 0) {
-                STACK_DESTROY (frame->root);
-        }
-        
-	return 0;
+	loc_wipe (&loc);
+
+	return afr_selfheal_locked_fill (frame, this, locked_on);
 }
 
 
-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 stat *buf,
-                                 struct stat *preparent,
-                                 struct stat *postparent)
+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)
 {
-	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;
-
-	struct stat     stbuf;
-        int32_t valid;
+	loc_t loc = {0,};
+	struct gf_flock flock = {0, };
+	afr_local_t *local = NULL;
+	int i = 0;
+	afr_private_t *priv = NULL;
 
+	priv = this->private;
 	local = frame->local;
-	sh    = &local->self_heal;
-	priv  = this->private;
-
-	child_index = (long) cookie;
-
-#ifdef HAVE_STRUCT_STAT_ST_ATIM_TV_NSEC
-	stbuf.st_atim = sh->buf[sh->source].st_atim;
-	stbuf.st_mtim = sh->buf[sh->source].st_mtim;
-        
-#elif HAVE_STRUCT_STAT_ST_ATIMESPEC_TV_NSEC
-	stbuf.st_atimespec = sh->buf[sh->source].st_atimespec;
-	stbuf.st_mtimespec = sh->buf[sh->source].st_mtimespec;
-#else
-	stbuf.st_atime = sh->buf[sh->source].st_atime;
-	stbuf.st_mtime = sh->buf[sh->source].st_mtime;
-#endif
 
-        stbuf.st_uid = sh->buf[sh->source].st_uid;
-        stbuf.st_gid = sh->buf[sh->source].st_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 = CALLOC (1, sizeof (afr_local_t));
-
-                ((afr_local_t *)setattr_frame->local)->call_count = 1;
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"setattr (%s) on subvolume %s",
-			local->loc.path, priv->children[child_index]->name);
-
-		STACK_WIND (setattr_frame, sh_destroy_cbk,
-			    priv->children[child_index],
-			    priv->children[child_index]->fops->setattr,
-			    &local->loc, &stbuf, valid);
-	}
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
+
+	flock.l_type = F_WRLCK;
+	flock.l_start = off;
+	flock.l_len = size;
 
-	call_count = afr_frame_return (frame);
+	AFR_ONALL (frame, afr_selfheal_lock_cbk, inodelk, dom,
+		    &loc, F_SETLK, &flock, NULL);
 
-	if (call_count == 0) {
-		sh_missing_entries_finish (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;
+	loc_wipe (&loc);
+
+	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_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;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              i = 0;
-	int              enoent_count = 0;
-	int              call_count = 0;
-	mode_t           st_mode = 0;
-	dev_t            st_dev = 0;
-
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+	loc_t loc = {0,};
+	struct gf_flock flock = {0, };
 
-	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;
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-	st_mode = sh->buf[sh->source].st_mode;
-	st_dev  = sh->buf[sh->source].st_dev;
+	flock.l_type = F_UNLCK;
+	flock.l_start = off;
+	flock.l_len = size;
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"mknod %s mode 0%o on %d subvolumes",
-		local->loc.path, st_mode, enoent_count);
+	AFR_ONLIST (locked_on, frame, afr_selfheal_lock_cbk, inodelk,
+		    dom, &loc, F_SETLK, &flock, 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, st_dev);
-			if (!--call_count)
-				break;
-		}
-	}
+	loc_wipe (&loc);
 
 	return 0;
 }
 
 
-static int
-sh_missing_entries_mkdir (call_frame_t *frame, xlator_t *this)
+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;
-	int              i = 0;
-	int              enoent_count = 0;
-	int              call_count = 0;
-	mode_t           st_mode = 0;
-
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	for (i = 0; i < priv->child_count; i++)
-		if (sh->child_errno[i] == ENOENT)
-			enoent_count++;
-
-	call_count = enoent_count;
-	local->call_count = call_count;
+	loc_t loc = {0,};
 
-	st_mode = sh->buf[sh->source].st_mode;
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"mkdir %s mode 0%o on %d subvolumes",
-		local->loc.path, st_mode, enoent_count);
+	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) {
-                        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);
-                                if (!--call_count)
-                                        break;
-                        }
-		}
-	}
+	loc_wipe (&loc);
 
-	return 0;
+	return afr_selfheal_locked_fill (frame, this, locked_on);
 }
 
 
-static int
-sh_missing_entries_symlink (call_frame_t *frame, xlator_t *this,
-			    const char *link)
+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              enoent_count = 0;
-	int              call_count = 0;
-
+	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;
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"symlink %s -> %s on %d subvolumes",
-		local->loc.path, link, enoent_count);
+	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->symlink,
-					   link, &local->loc);
-			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;
 		}
 	}
 
-	return 0;
-}
-
-
-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 stat *sbuf)
-{
-	if (op_ret > 0)
-		sh_missing_entries_symlink (frame, this, link);
-	else
-		sh_missing_entries_finish (frame, this);
+	loc_wipe (&loc);
 
-	return 0;
+	return afr_selfheal_locked_fill (frame, this, locked_on);
 }
 
 
-static int
-sh_missing_entries_readlink (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;
+	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;
+	AFR_ONLIST (locked_on, frame, afr_selfheal_lock_cbk, entrylk,
+		    dom, &loc, name, ENTRYLK_UNLOCK, ENTRYLK_WRLCK, NULL);
 
-	STACK_WIND (frame, sh_missing_entries_readlink_cbk,
-		    priv->children[sh->source],
-		    priv->children[sh->source]->fops->readlink,
-		    &local->loc, 4096);
+	loc_wipe (&loc);
 
 	return 0;
 }
 
 
-static int
-sh_missing_entries_create (call_frame_t *frame, xlator_t *this)
+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;
-	int              type = 0;
-	int              i = 0;
-	afr_private_t   *priv = NULL;
-	int              enoent_count = 0;
-	int              govinda_gOvinda = 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]) {
-			if (sh->child_errno[i] == ENOENT)
-				enoent_count++;
-		} else {
-			if (type) {
-				if (type != (sh->buf[i].st_mode & S_IFMT)) {
-                                        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].st_mode & S_IFMT;
-			}
-		}
-	}
-
-	if (govinda_gOvinda) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"conflicing filetypes exist for path %s. returning.",
-			local->loc.path);
+	if (dict_get_ptr (xdata, AFR_DIRTY, &pending_raw) == 0) {
+		if (pending_raw) {
+			pending_int = pending_raw;
 
-		local->govinda_gOvinda = 1;
-		sh_missing_entries_finish (frame, this);
-		return 0;
-	}
-
-	if (!type) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"no source found for %s. all nodes down?. returning.",
-			local->loc.path);
-		/* subvolumes down and/or file does not exist */
-		sh_missing_entries_finish (frame, this);
-		return 0;
+			if (ntoh32 (pending_int[idx]))
+				return _gf_true;
+		}
 	}
 
-	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;
-	}
+	for (i = 0; i < priv->child_count; i++) {
+		if (dict_get_ptr (xdata, priv->pending_key[i],
+				  &pending_raw))
+			continue;
+		if (!pending_raw)
+			continue;
+		pending_int = pending_raw;
 
-	switch (type) {
-	case S_IFSOCK:
-	case S_IFREG:
-	case S_IFBLK:
-	case S_IFCHR:
-	case S_IFIFO:
-		sh_missing_entries_mknod (frame, this);
-		break;
-	case S_IFLNK:
-		sh_missing_entries_readlink (frame, this);
-		break;
-	case S_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 (ntoh32 (pending_int[idx]))
+			return _gf_true;
 	}
 
-	return 0;
+	return _gf_false;
 }
 
 
-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 stat *buf, dict_t *xattr,
-                               struct stat *postparent)
+gf_boolean_t
+afr_is_data_set (xlator_t *this, dict_t *xdata)
 {
-	int              child_index = 0;
-	afr_local_t     *local = NULL;
-	int              call_count = 0;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	child_index = (long) cookie;
+	return afr_is_pending_set (this, xdata, AFR_DATA_TRANSACTION);
+}
 
-	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->st_mode);
+gf_boolean_t
+afr_is_metadata_set (xlator_t *this, dict_t *xdata)
+{
+	return afr_is_pending_set (this, xdata, AFR_METADATA_TRANSACTION);
+}
 
-			local->self_heal.buf[child_index] = *buf;
+gf_boolean_t
+afr_is_entry_set (xlator_t *this, dict_t *xdata)
+{
+	return afr_is_pending_set (this, xdata, AFR_ENTRY_TRANSACTION);
+}
 
-		} else {
-			gf_log (this->name, GF_LOG_TRACE,
-				"path %s on subvolume %s => -1 (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
 
-			local->self_heal.child_errno[child_index] = op_errno;
-		}
+void
+afr_inode_link (inode_t *inode, struct iatt *iatt)
+{
+	inode_t *linked_inode = NULL;
 
-	}
-	UNLOCK (&frame->lock);
+	linked_inode = inode_link (inode, NULL, NULL, iatt);
 
-	call_count = afr_frame_return (frame);
+	uuid_copy (inode->gfid, iatt->ia_gfid);
+	inode->ia_type = iatt->ia_type;
 
-	if (call_count == 0) {
-		sh_missing_entries_create (frame, this);
+	if (linked_inode) {
+		inode_lookup (linked_inode);
+		inode_unref (linked_inode);
 	}
-
-	return 0;
 }
 
 
-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;
+/*
+ * 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.
+ */
 
-	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));
-                }
-        }
+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;
 
-	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;
-		}
-	}
-	
-	if (xattr_req)
-		dict_unref (xattr_req);
+	priv = this->private;
 
-	return 0;
-}
+	replies = alloca0 (sizeof (*replies) * priv->child_count);
 
+	ret = afr_selfheal_unlocked_discover (frame, inode, gfid, replies);
+	if (ret)
+		return ret;
 
-static int
-sh_missing_entries_lk_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;
-	int              call_count = 0;
-	int              child_index = (long) cookie;
+	for (i = 0; i < priv->child_count; i++) {
+		if (!replies[i].valid)
+			continue;
+		if (replies[i].op_ret == -1)
+			continue;
 
+		if (afr_is_data_set (this, replies[i].xdata))
+			*data_selfheal = _gf_true;
 
-	local = frame->local;
-	sh    = &local->self_heal;
+		if (afr_is_metadata_set (this, replies[i].xdata))
+			*metadata_selfheal = _gf_true;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			sh->op_failed = 1;
+		if (afr_is_entry_set (this, replies[i].xdata))
+			*entry_selfheal = _gf_true;
 
-                        sh->locked_nodes[child_index] = 0;
-			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 {
-                        sh->locked_nodes[child_index] = 1;
-			gf_log (this->name, GF_LOG_TRACE,
-				"inode of %s on child %d locked",
-				local->loc.path, child_index);
+		valid_cnt ++;
+		if (valid_cnt == 1) {
+			first = replies[i].poststat;
+			continue;
 		}
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
 
-	if (call_count == 0) {
-		if (sh->op_failed == 1) {
-			sh_missing_entries_finish (frame, this);
-			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;
 		}
 
-		sh_missing_entries_lookup (frame, this);
-	}
-
-	return 0;
-}
+		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));
 
+			*metadata_selfheal = _gf_true;
+		}
 
-static int
-afr_self_heal_missing_entries (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;
-	int              call_count = 0;
+		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));
 
+			*metadata_selfheal = _gf_true;
+		}
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+		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));
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"attempting to recreate missing entries for path=%s",
-		local->loc.path);
+			*metadata_selfheal = _gf_true;
+		}
 
-	afr_build_parent_loc (&sh->parent_loc, &local->loc);
+		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));
 
-	call_count = afr_up_children_count (priv->child_count,
-                                            local->child_up);
+			*data_selfheal = _gf_true;
+		}
+	}
 
-	local->call_count = call_count;
+	if (valid_cnt > 0)
+		afr_inode_link (inode, &first);
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, sh_missing_entries_lk_cbk,
-                                           (void *) (long) i,
-                                           priv->children[i],
-                                           priv->children[i]->fops->entrylk,
-                                           this->name,
-                                           &sh->parent_loc, local->loc.name,
-                                           ENTRYLK_LOCK_NB, ENTRYLK_WRLCK);
-			if (!--call_count)
-				break;
-		}
- 	}
+	if (valid_cnt < 2)
+		return -ENOTCONN;
 
 	return 0;
 }
 
 
-afr_local_t *afr_local_copy (afr_local_t *l, xlator_t *this)
+inode_t *
+afr_inode_find (xlator_t *this, uuid_t gfid)
 {
-        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 = CALLOC (1, sizeof (afr_local_t));
-        shc = &lc->self_heal;
+	inode_table_t *table = NULL;
+	inode_t *inode = NULL;
 
-        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)
-                shc->healing_fd = fd_ref (sh->healing_fd);
-        shc->background = sh->background;
-        shc->mode = sh->mode;
+	table = this->itable;
+	if (!table)
+		return NULL;
 
-        if (l->loc.path)
-                loc_copy (&lc->loc, &l->loc);
+	inode = inode_find (table, gfid);
+	if (inode)
+		return inode;
 
-        lc->child_up  = memdup (l->child_up, priv->child_count);
-        if (l->xattr_req)
-                lc->xattr_req = dict_copy_with_ref (l->xattr_req, NULL);
+	inode = inode_new (table);
+	if (!inode)
+		return NULL;
 
-        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_copy_with_ref (l->cont.lookup.xattr, NULL);
+	uuid_copy (inode->gfid, gfid);
 
-        return lc;
+	return inode;
 }
 
 
-int
-afr_self_heal_completion_cbk (call_frame_t *bgsh_frame, xlator_t *this)
+call_frame_t *
+afr_frame_create (xlator_t *this)
 {
-        afr_private_t *   priv  = NULL;
-	afr_local_t *     local = NULL;
-        afr_self_heal_t * sh    = NULL;
+	call_frame_t *frame = NULL;
+	afr_local_t *local = NULL;
+	int op_errno = 0;
+	pid_t pid = -1;
 
-        priv  = this->private;
-	local = bgsh_frame->local;
-        sh    = &local->self_heal;
+	frame = create_frame (this, this->ctx->pool);
+	if (!frame)
+		return NULL;
 
-	if (local->govinda_gOvinda) {
-                afr_set_split_brain (this, local->cont.lookup.inode);
+	local = AFR_FRAME_INIT (frame, op_errno);
+	if (!local) {
+		STACK_DESTROY (frame->root);
+		return NULL;
 	}
 
-        gf_log (this->name, GF_LOG_TRACE,
-                "background self-heal completed");
-
-        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;
 }
 
 
+/*
+ * This is the entry point for healing a given GFID
+ */
+
 int
-afr_self_heal (call_frame_t *frame, xlator_t *this)
+afr_selfheal (xlator_t *this, uuid_t gfid)
 {
-	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;
-
-	local = frame->local;
-	priv  = this->private;
-
-        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);
+	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_frame        = copy_frame (frame);
-        sh_local        = afr_local_copy (local, this);
-        sh_frame->local = sh_local;
-        sh              = &sh_local->self_heal;
+	inode = afr_inode_find (this, gfid);
+	if (!inode)
+		goto out;
 
-        sh->orig_frame  = frame;
+	frame = afr_frame_create (this);
+	if (!frame)
+		goto out;
 
-        sh->completion_cbk = afr_self_heal_completion_cbk;
+	ret = afr_selfheal_unlocked_inspect (frame, this, inode, gfid,
+					     &data_selfheal,
+					     &metadata_selfheal,
+					     &entry_selfheal);
+	if (ret)
+		goto out;
 
+	if (data_selfheal)
+		afr_selfheal_data (frame, this, inode);
 
-	sh->buf          = CALLOC (priv->child_count, sizeof (struct stat));
-	sh->child_errno  = CALLOC (priv->child_count, sizeof (int));
-	sh->success      = CALLOC (priv->child_count, sizeof (int));
-	sh->xattr        = CALLOC (priv->child_count, sizeof (dict_t *));
-	sh->sources      = CALLOC (priv->child_count, sizeof (*sh->sources));
-	sh->locked_nodes = CALLOC (priv->child_count, sizeof (*sh->locked_nodes));
+	if (metadata_selfheal)
+		afr_selfheal_metadata (frame, this, inode);
 
-	sh->pending_matrix = CALLOC (sizeof (int32_t *), priv->child_count);
-	for (i = 0; i < priv->child_count; i++) {
-		sh->pending_matrix[i] = CALLOC (sizeof (int32_t),
-						priv->child_count);
-	}
+	if (entry_selfheal)
+		afr_selfheal_entry (frame, this, inode);
 
-	sh->delta_matrix = CALLOC (sizeof (int32_t *), priv->child_count);
-	for (i = 0; i < priv->child_count; i++) {
-		sh->delta_matrix[i] = CALLOC (sizeof (int32_t),
-					      priv->child_count);
-	}
-
-	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);
-
-		afr_sh_missing_entries_done (sh_frame, this);
-	}
+	inode_forget (inode, 1);
+out:
+	if (inode)
+		inode_unref (inode);
+	if (frame)
+		AFR_STACK_DESTROY (frame);
 
-	return 0;
+	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 be556a3c7..000000000
--- a/xlators/cluster/afr/src/afr-self-heal-common.h
+++ /dev/null
@@ -1,70 +0,0 @@
-/*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#ifndef __AFR_SELF_HEAL_COMMON_H__
-#define __AFR_SELF_HEAL_COMMON_H__
-
-#define FILE_HAS_HOLES(buf) (((buf)->st_size) > ((buf)->st_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);
-
-
-#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 75090ad79..c0548d995 100644
--- a/xlators/cluster/afr/src/afr-self-heal-data.c
+++ b/xlators/cluster/afr/src/afr-self-heal-data.c
@@ -1,1206 +1,635 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 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;
-        int              i = 0;
+	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;
-	afr_self_heal_t *sh = NULL;
-	int              call_count = 0;
-
-        int child_index = (long) cookie;
+	int i = (long) cookie;
+	afr_local_t *local = NULL;
 
 	local = frame->local;
-	sh = &local->self_heal;
-	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 stat *statpre, struct stat *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;
-        int  active_sinks = 0;
-        int32_t valid     = 0;
-
-        struct stat stbuf = {0,};
-        
-        local = frame->local;
-        sh    = &local->self_heal;
-        priv  = this->private;
-
-        source       = sh->source;
-        active_sinks = sh->active_sinks;
-
-        valid |= (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME);
-
-#ifdef HAVE_STRUCT_STAT_ST_ATIM_TV_NSEC
-	stbuf.st_atim = sh->buf[source].st_atim;
-	stbuf.st_mtim = sh->buf[source].st_mtim;
-        
-#elif HAVE_STRUCT_STAT_ST_ATIMESPEC_TV_NSEC
-	stbuf.st_atimespec = sh->buf[source].st_atimespec;
-	stbuf.st_mtimespec = sh->buf[source].st_mtimespec;
-#else
-	stbuf.st_atime = sh->buf[source].st_atime;
-	stbuf.st_mtime = sh->buf[source].st_mtime;
-#endif
-
-        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;
-
-	/* closed source */
-	gf_log (this->name, GF_LOG_TRACE,
-		"closing fd of %s on %s",
-		local->loc.path, priv->children[sh->source]->name);
+	AFR_ONLIST (wind_subvols, frame, __checksum_cbk, rchecksum, fd,
+		    offset, size, NULL);
 
-	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;
+		if (memcmp (local->replies[source].checksum,
+			    local->replies[i].checksum,
+			    MD5_DIGEST_LENGTH))
+			return _gf_false;
 	}
 
-	return 0;
+	return _gf_true;
 }
 
 
-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);
-		}
-	}
-	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)
+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)
 {
-	struct flock flock;			
-	int i = 0;				
-	int call_count = 0;		     
-
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	afr_self_heal_t * sh  = 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;
 
-        if (sh->data_lock_held) {
-                /* not our job to unlock, proceed to close */
-
-                afr_sh_data_close (frame, this);
-                return 0;
-        }
-
-        for (i = 0; i < priv->child_count; i++) {
-                if (sh->locked_nodes[i])
-                        call_count++;
-        }
-
-        if (call_count == 0) {
-                afr_sh_data_close (frame, this);
-                return 0;
-        }
+	ret = syncop_readv (priv->children[source], fd, size, offset, 0,
+			    &iovec, &count, &iobref);
+	if (ret <= 0)
+		return ret;
 
-	local->call_count = call_count;		
+	for (i = 0; i < priv->child_count; i++) {
+		if (!healed_sinks[i])
+			continue;
 
-	flock.l_start = 0;
-	flock.l_len   = 0;
-	flock.l_type  = F_UNLCK;
+		/*
+		 * 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 (sh->locked_nodes[i]) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"unlocking %s on subvolume %s",
-				local->loc.path, priv->children[i]->name);
-
-			STACK_WIND_COOKIE (frame, afr_sh_data_unlck_cbk,
-					   (void *) (long) i,
-					   priv->children[i], 
-					   priv->children[i]->fops->inodelk,
-                                           this->name,
-					   &local->loc, F_SETLK, &flock); 
-			if (!--call_count)
-				break;
+		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;
 		}
 	}
+	if (iobref)
+		iobref_unref (iobref);
 
-	return 0;
+	return ret;
 }
 
 
-int
-afr_sh_data_finish (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t   *local = NULL;
-
-	local = frame->local;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"finishing data selfheal of %s", local->loc.path);
-
-	afr_sh_data_unlock (frame, this);
-
-	return 0;
-}
-
-
-int
-afr_sh_data_erase_pending_cbk (call_frame_t *frame, void *cookie,
-			       xlator_t *this, int32_t op_ret,
-			       int32_t op_errno, dict_t *xattr)
+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_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int             call_count = 0;
+	int ret = -1;
+	int sink_count = 0;
+	afr_private_t *priv = NULL;
+	unsigned char *data_lock = NULL;
 
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
+	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);
 	{
-	}
-	UNLOCK (&frame->lock);
-
-	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)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              call_count = 0;
-	int              i = 0;
-	dict_t          **erase_xattr = NULL;
-
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix, sh->success,
-                                 priv->child_count, AFR_DATA_TRANSACTION);
-
-	erase_xattr = CALLOC (sizeof (*erase_xattr), 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 (ret < sink_count) {
+			ret = -ENOTCONN;
+			goto unlock;
 		}
-	}
-
-	afr_sh_delta_to_xattr (priv, sh->delta_matrix, erase_xattr,
-			       priv->child_count, AFR_DATA_TRANSACTION);
-
-	local->call_count = call_count;
-	for (i = 0; i < priv->child_count; i++) {
-		if (!erase_xattr[i])
-			continue;
 
-		gf_log (this->name, GF_LOG_TRACE,
-			"erasing pending flags from %s on %s",
-			local->loc.path, priv->children[i]->name);
-
-		STACK_WIND_COOKIE (frame, afr_sh_data_erase_pending_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->fxattrop,
-				   sh->healing_fd,
-				   GF_XATTROP_ADD_ARRAY, erase_xattr[i]);
-		if (!--call_count)
-			break;
-	}
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (erase_xattr[i]) {
-			dict_unref (erase_xattr[i]);
+		if (type == AFR_SELFHEAL_DATA_DIFF &&
+		    __afr_selfheal_data_checksums_match (frame, this, fd, source,
+							 healed_sinks, offset, size)) {
+			ret = 0;
+			goto unlock;
 		}
-	}
-	FREE (erase_xattr);
 
-	return 0;
+		ret = __afr_selfheal_data_read_write (frame, this, fd, source,
+						      healed_sinks, offset, size,
+						      replies);
+	}
+unlock:
+	afr_selfheal_uninodelk (frame, this, fd->inode, this->name,
+				offset, size, data_lock);
+	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 stat *prebuf,
-                      struct stat *postbuf)
+
+static int
+afr_selfheal_data_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
+			 unsigned char *healed_sinks)
 {
-	afr_private_t * priv = NULL;
-	afr_local_t * local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              call_count = 0;
-	int              child_index = 0;
+	afr_local_t *local = NULL;
+	afr_private_t *priv = NULL;
+	int i = 0;
 
-	priv = this->private;
 	local = frame->local;
-	sh = &local->self_heal;
-
-	child_index = (long) cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1)
-			gf_log (this->name, GF_LOG_DEBUG,
-				"ftruncate of %s on subvolume %s failed (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-		else
-			gf_log (this->name, GF_LOG_TRACE,
-				"ftruncate of %s on subvolume %s completed",
-				local->loc.path,
-				priv->children[child_index]->name);
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
+	priv = this->private;
 
-	if (call_count == 0) {
-		afr_sh_data_erase_pending (frame, this);
-	}
+	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;
 }
 
 
-int
-afr_sh_data_trim_sinks (call_frame_t *frame, xlator_t *this)
+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)
 {
-	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;
+	loc_t loc = {0, };
 
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
+	AFR_ONLIST (healed_sinks, frame, attr_cbk, setattr, &loc,
+		    &replies[source].poststat,
+		    (GF_SET_ATTR_ATIME|GF_SET_ATTR_MTIME), NULL);
 
-	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); 
-
-		if (!--call_count)
-			break;
-	}
+	loc_wipe (&loc);
 
 	return 0;
 }
 
-
-static struct afr_sh_algorithm *
-sh_algo_from_name (xlator_t *this, char *name)
+static int
+afr_data_self_heal_type_get (afr_private_t *priv, unsigned char *healed_sinks,
+                             int source, struct afr_reply *replies)
 {
+        int type = AFR_SELFHEAL_DATA_FULL;
         int i = 0;
 
-        while (afr_self_heal_algorithms[i].name) {
-                if (!strcmp (name, afr_self_heal_algorithms[i].name)) {
-                        return &afr_self_heal_algorithms[i];
+        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;
+                        }
                 }
-
-                i++;
+        } 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 NULL;
+        return type;
 }
 
-
 static int
-sh_zero_byte_files_exist (afr_self_heal_t *sh, int child_count)
-{
-        int i;
-        int ret = 0;
-
-        for (i = 0; i < child_count; i++) {
-                if (sh->buf[i].st_size == 0) {
-                        ret = 1;
-                        break;
-                }
-        }
-
-        return ret;
-}
-
-
-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");
-                }
-        }
-
-        return algo;
-}
-
-
-int
-afr_sh_data_sync_prepare (call_frame_t *frame, xlator_t *this)
+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_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;
-}
-
-
-int
-afr_sh_data_fix (call_frame_t *frame, xlator_t *this)
-{
-	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].st_size;
-
-	if (FILE_HAS_HOLES (&sh->buf[source]))
-		sh->file_has_holes = 1;
-
-        orig_local = sh->orig_frame->local;
-        orig_local->cont.lookup.buf.st_size = sh->buf[source].st_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);
-
-        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);
 
+        type = afr_data_self_heal_type_get (priv, healed_sinks, source,
+                                            replies);
 
-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;
+	iter_frame = afr_copy_frame (frame);
+	if (!iter_frame)
+		return -ENOMEM;
 
-	int              nsources = 0;
-	int              source = 0;
-	int              i = 0;
+	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   = &local->self_heal;
-	priv = this->private;
-
-	sh->pending_matrix = CALLOC (sizeof (int32_t *), priv->child_count);
-	for (i = 0; i < priv->child_count; i++) {
-		sh->pending_matrix[i] = CALLOC (sizeof (int32_t),
-						priv->child_count);
+		AFR_STACK_RESET (iter_frame);
 	}
 
-	sh->sources = CALLOC (priv->child_count, sizeof (*sh->sources));
+	afr_selfheal_data_restore_time (frame, this, fd->inode, source,
+					healed_sinks, replies);
 
-	afr_sh_build_pending_matrix (priv, sh->pending_matrix, xattr,
-				     priv->child_count, AFR_DATA_TRANSACTION);
+	ret = afr_selfheal_data_fsync (frame, this, fd, healed_sinks);
 
-	nsources = afr_sh_mark_sources (sh, priv->child_count,
-                                        AFR_SELF_HEAL_DATA);
-
-	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 stat *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);
-
-			sh->xattr[child_index] = dict_ref (xattr);
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
+	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);
 
-	if (call_count == 0) {
-		afr_sh_data_fstat (frame, this);
+	if (locked_count == healed_sinks_count || !sources_count) {
+		/* split brain */
+		return -EIO;
 	}
 
-	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));
-                }
-        }
-
 	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 (size <= replies[i].poststat.ia_size) {
+			size = replies[i].poststat.ia_size;
+			source = i;
 		}
 	}
-	
-	if (xattr_req)
-		dict_unref (xattr_req);
-
-	return 0;
-}
-
 
-int
-afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this, int child_index);
-
-int
-afr_sh_data_lock_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;
-	int              child_index = (long) cookie;
-
-	/* TODO: what if lock fails? */
-	
-	local = frame->local;
-	sh = &local->self_heal;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-                        sh->locked_nodes[child_index] = 0;
-
-			gf_log (this->name, GF_LOG_DEBUG,
-				"locking of %s on child %d failed: %s",
-				local->loc.path, child_index,
-				strerror (op_errno));
-		} else {
-                        sh->locked_nodes[child_index] = 1;
-                        sh->lock_count++;
-
-			gf_log (this->name, GF_LOG_TRACE,
-				"inode of %s on child %d locked",
-				local->loc.path, child_index);
+	for (i = 0; i < priv->child_count; i++) {
+		if (!sources[i])
+			continue;
+		if (replies[i].poststat.ia_size < size) {
+			sources[i] = 0;
+			sinks[i] = 1;
 		}
 	}
-	UNLOCK (&frame->lock);
-
-        afr_sh_data_lock_rec (frame, this, child_index + 1);
 
-	return 0;
+	return source;
 }
 
-
-int
-afr_sh_data_lock_rec (call_frame_t *frame, xlator_t *this, int child_index)
+/*
+ * __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)
 {
-	struct flock flock;			
-	int i = 0;				
-
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	afr_self_heal_t * sh  = NULL;
+	int ret = -1;
+	int source = -1;
+	afr_private_t *priv = NULL;
+	int i = 0;
 
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-	flock.l_start = 0;
-	flock.l_len   = 0;
-	flock.l_type  = F_WRLCK;			
+	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];
 
-	/* 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) &&
-	    sh->lock_count == 0) {
-
-		gf_log (this->name, GF_LOG_DEBUG,
-			"unable to lock on even one child");
-
-                afr_sh_data_done (frame, this);
-		return 0;
-	}
-
-        if ((child_index == priv->child_count)
-            || (sh->lock_count == afr_lock_server_count (priv, AFR_DATA_TRANSACTION))) {
-                afr_sh_data_fxattrop (frame, this);
-                return 0;
-        }
-
-        gf_log (this->name, GF_LOG_TRACE,
-                "locking %s on subvolume %s",
-                local->loc.path, priv->children[i]->name);
-
-        STACK_WIND_COOKIE (frame, afr_sh_data_lock_cbk,
-                           (void *) (long) child_index,
-                           priv->children[i],
-                           priv->children[i]->fops->inodelk,
-                           this->name,
-                           &local->loc, F_SETLKW, &flock);
-
-	return 0;
+	return source;
 }
 
 
-int
-afr_sh_data_lock (call_frame_t *frame, xlator_t *this)
+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_private_t * priv  = NULL;
-	afr_self_heal_t * sh  = NULL;
-
-        int i = 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;
+	priv = this->private;
 
-        if (sh->data_lock_held) {
-                /* caller has held the lock already,
-                   so skip locking */
+	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);
 
-                afr_sh_data_fxattrop (frame, this);
-                return 0;
-        }
+	locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
 
-        for (i = 0; i < priv->child_count; i++)
-                sh->locked_nodes[i] = 0;
+	ret = afr_selfheal_inodelk (frame, this, fd->inode, this->name, 0, 0,
+				    data_lock);
+	{
+		if (ret < 2) {
+			ret = -ENOTCONN;
+			goto unlock;
+		}
 
-        return afr_sh_data_lock_rec (frame, this, 0);
+		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:
+	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;
 }
 
 
-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 fd_t *
+afr_selfheal_data_open (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;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	child_index = (long) cookie;
+	loc_t loc = {0,};
+	int ret = 0;
+	fd_t *fd = NULL;
 
-	/* 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 */
+	fd = fd_create (inode, 0);
+	if (!fd)
+		return NULL;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"open of %s failed on child %s (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-			sh->op_failed = 1;
-		}
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-                gf_log (this->name, GF_LOG_TRACE,
-                        "open of %s succeeded on child %s",
-                        local->loc.path,
-                        priv->children[child_index]->name);
+	ret = syncop_open (this, &loc, O_RDWR|O_LARGEFILE, fd);
+	if (ret) {
+		fd_unref (fd);
+		fd = NULL;
+	} else {
+		fd_bind (fd);
 	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-		if (sh->op_failed) {
-			afr_sh_data_finish (frame, this);
-			return 0;
-		}
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"fd for %s opened, commencing sync",
-			local->loc.path);
 
-                afr_sh_data_lock (frame, this);
-	}
+	loc_wipe (&loc);
 
-	return 0;
+	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 */
-
-                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;
+	fd = afr_selfheal_data_open (this, inode);
+	if (!fd)
+		return -EIO;
 
-	/* 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 420b7155c..9605d69f4 100644
--- a/xlators/cluster/afr/src/afr-self-heal-entry.c
+++ b/xlators/cluster/afr/src/afr-self-heal-entry.c
@@ -1,2562 +1,631 @@
 /*
-  Copyright (c) 2008-2009 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 General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
-#include <libgen.h>
-#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_entry_done (call_frame_t *frame, xlator_t *this)
+static int
+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_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-        int              i = 0;
+	afr_private_t *priv = NULL;
+	xlator_t *subvol = NULL;
+	int ret = 0;
+	loc_t loc = {0, };
+	char g[64];
 
-	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;
+	subvol = priv->children[child];
 
-        for (i = 0; i < priv->child_count; i++) {
-                sh->locked_nodes[i] = 0;
-        }
+	loc.parent = inode_ref (dir);
+	uuid_copy (loc.pargfid, dir->gfid);
+	loc.name = name;
+	loc.inode = inode_ref (inode);
 
-	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_unlck_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;
-	int           call_count = 0;
-	int           child_index = (long) cookie;
-
-	/* TODO: what if lock fails? */
-	
-	local = frame->local;
-	sh = &local->self_heal;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"unlocking inode of %s on child %d failed: %s",
-				local->loc.path, child_index,
-				strerror (op_errno));
-		} else {
-			gf_log (this->name, GF_LOG_TRACE,
-				"unlocked inode of %s on child %d",
-				local->loc.path, child_index);
+	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;
+		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;
 		}
 	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
 
-	if (call_count == 0) {
-		afr_sh_entry_done (frame, this);
-	}
+	loc_wipe (&loc);
 
-	return 0;
+	return ret;
 }
 
 
 int
-afr_sh_entry_unlock (call_frame_t *frame, xlator_t *this)
+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 i = 0;				
-	int call_count = 0;		     
-
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	afr_self_heal_t * sh  = NULL;
-
+	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,};
 
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-        for (i = 0; i < priv->child_count; i++) {
-                if (sh->locked_nodes[i])
-                        call_count++;
-        }
-
-        if (call_count == 0) {
-                afr_sh_entry_done (frame, this);
-                return 0;
-        }
-
-	local->call_count = call_count;		
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->locked_nodes[i]) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"unlocking %s on subvolume %s",
-				local->loc.path, priv->children[i]->name);
-
-			STACK_WIND_COOKIE (frame, afr_sh_entry_unlck_cbk,
-					   (void *) (long) i,
-					   priv->children[i], 
-					   priv->children[i]->fops->entrylk,
-                                           this->name,
-					   &local->loc, NULL,
-					   ENTRYLK_UNLOCK, ENTRYLK_WRLCK);
-			if (!--call_count)
-				break;
-		}
-	}
-
-	return 0;
-}
+	xdata = dict_new();
+	if (!xdata)
+		return -ENOMEM;
 
+	loc.parent = inode_ref (dir);
+	uuid_copy (loc.pargfid, dir->gfid);
+	loc.name = name;
+	loc.inode = inode_ref (inode);
 
-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);
+	ret = afr_selfheal_entry_delete (frame, this, dir, name, inode, dst,
+					 replies);
+	if (ret)
+		goto out;
 
-	return 0;
-}
+	ret = dict_set_static_bin (xdata, "gfid-req",
+				   replies[source].poststat.ia_gfid, 16);
+	if (ret)
+		goto out;
 
+	iatt = &replies[source].poststat;
 
-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)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int             call_count = 0;
+	srcloc.inode = inode_ref (inode);
+	uuid_copy (srcloc.gfid, iatt->ia_gfid);
 
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+	mode = st_mode_from_ia (iatt->ia_prot, iatt->ia_type);
 
-	LOCK (&frame->lock);
-	{
+	switch (iatt->ia_type) {
+	case IA_IFDIR:
+		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:
+		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;
 	}
-	UNLOCK (&frame->lock);
 
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		afr_sh_entry_finish (frame, this);
-
-	return 0;
+out:
+	if (xdata)
+		dict_unref (xdata);
+	loc_wipe (&loc);
+	loc_wipe (&srcloc);
+	return ret;
 }
 
 
-int
-afr_sh_entry_erase_pending (call_frame_t *frame, xlator_t *this)
+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_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;
-
+	int ret = 0;
+	int i = 0;
+	afr_private_t *priv = NULL;
+	dict_t *xattr = NULL;
+	int **changelog = NULL;
+	int idx = 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 = CALLOC (sizeof (*erase_xattr), priv->child_count);
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->xattr[i]) {
-			call_count++;
+	idx = afr_index_for_transaction_type (AFR_DATA_TRANSACTION);
 
-			erase_xattr[i] = get_new_dict();
-			dict_ref (erase_xattr[i]);
-		}
-	}
+	uuid_copy (inode->gfid, replies[source].poststat.ia_gfid);
 
-        if (call_count == 0)
-                need_unwind = 1;
+	changelog = afr_matrix_create (priv->child_count, AFR_NUM_CHANGE_LOGS);
 
-	afr_sh_delta_to_xattr (priv, sh->delta_matrix, erase_xattr,
-			       priv->child_count, AFR_ENTRY_TRANSACTION);
+	xattr = dict_new();
+	if (!xattr)
+		return -ENOMEM;
 
-	local->call_count = call_count;
 	for (i = 0; i < priv->child_count; i++) {
-		if (!erase_xattr[i])
+		if (!newentry[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;
+		changelog[i][idx] = hton32(1);
 	}
 
+	afr_set_pending_dict (priv, xattr, changelog);
+
 	for (i = 0; i < priv->child_count; i++) {
-		if (erase_xattr[i]) {
-			dict_unref (erase_xattr[i]);
-		}
+		if (!sources[i])
+			continue;
+		afr_selfheal_post_op (frame, this, inode, i, xattr);
 	}
-	FREE (erase_xattr);
-
-        if (need_unwind)
-                afr_sh_entry_finish (frame, this);
 
-	return 0;
+	dict_unref (xattr);
+	return ret;
 }
 
 
-
 static int
-next_active_source (call_frame_t *frame, xlator_t *this,
-		    int current_active_source)
+__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              source = -1;
-	int              next_active_source = -1;
-	int              i = 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;
+	newentry = alloca0 (priv->child_count);
 
-	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
-	*/
+	if (!replies[source].valid)
+		return -EIO;
 
 	for (i = 0; i < priv->child_count; i++) {
-		if ((sh->sources[i] == 0)
-		    && (local->child_up[i] == 1)
-		    && (i > current_active_source)) {
-
-			next_active_source = i;
-			break;
+		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 {
+			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;
+			}
 		}
+		if (ret < 0)
+			break;
 	}
-out:
-	return next_active_source;
-}
 
+	if (AFR_COUNT (newentry, priv->child_count))
+		afr_selfheal_newentry_mark (frame, this, inode, source, replies,
+					    sources, newentry);
+	return ret;
+}
 
 
 static int
-next_active_sink (call_frame_t *frame, xlator_t *this,
-		  int current_active_sink)
+__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     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              next_active_sink = -1;
-	int              i = 0;
+	int ret = 0;
+	afr_private_t *priv = NULL;
+	int i = 0;
+	int source = -1;
 
 	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;
+		if (replies[i].valid && replies[i].op_ret == 0) {
+			source = i;
 			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 = asprintf ((char **)&child->path, "/%s", name);
-	else
-		ret = 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;
+	if (source == -1) {
+		/* entry got deleted in the mean time? */
+		return 0;
 	}
 
-	child->name = strrchr (child->path, '/');
-	if (child->name)
-		child->name++;
+	for (i = 0; i < priv->child_count; i++) {
+		if (i == source || !healed_sinks[i])
+			continue;
 
-	child->parent = inode_ref (parent->inode);
-	child->inode = inode_new (parent->inode->table);
+		if (replies[i].op_errno != ENOENT)
+			continue;
 
-	if (!child->inode) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
+		ret = afr_selfheal_recreate_entry (frame, this, i, source,
+						   fd->inode, name, inode,
+						   replies);
 	}
 
-	ret = 0;
-out:
-	if (ret == -1)
-		loc_wipe (child);
-
 	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)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              call_count = 0;
-
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	LOCK (&frame->lock);
-	{
-	}
-	UNLOCK (&frame->lock);
-
-	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 stat *preop, struct stat *postop)
-{
-	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_rename_cbk (call_frame_t *expunge_frame, void *cookie,
-				 xlator_t *this,
-				 int32_t op_ret, int32_t op_errno,
-                                 struct stat *buf,
-                                 struct stat *preoldparent,
-                                 struct stat *postoldparent,
-                                 struct stat *prenewparent,
-                                 struct stat *postnewparent)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
-	afr_self_heal_t *expunge_sh = NULL;
-	int              active_src = 0;
-	call_frame_t    *frame = NULL;
-
-        int32_t valid = 0;
-
-	priv = this->private;
-	expunge_local = expunge_frame->local;
-	expunge_sh = &expunge_local->self_heal;
-	frame = expunge_sh->sh_frame;
-
-	active_src = (long) cookie;
-
-	if (op_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"removed %s on %s",
-			expunge_local->loc.path,
-			priv->children[active_src]->name);
-	} else {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"removing %s on %s failed (%s)",
-			expunge_local->loc.path,
-			priv->children[active_src]->name,
-			strerror (op_errno));
-	}
-
-        valid = GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
-        afr_build_parent_loc (&expunge_sh->parent_loc, &expunge_local->loc);
-
-        STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_parent_setattr_cbk,
-                           (void *) (long) active_src,
-                           priv->children[active_src],
-                           priv->children[active_src]->fops->setattr,
-                           &expunge_sh->parent_loc,
-                           &expunge_sh->parentbuf,
-                           valid);
-
-        return 0;
-}
-
-
-static void
-init_trash_loc (loc_t *trash_loc, inode_table_t *table)
-{
-        trash_loc->path   = strdup ("/" GF_REPLICATE_TRASH_DIR);
-        trash_loc->name   = GF_REPLICATE_TRASH_DIR;
-        trash_loc->parent = table->root;
-        trash_loc->inode  = inode_new (table);
-}
-
-
-char *
-make_trash_path (const char *path)
-{
-        char *c  = NULL;
-        char *tp = NULL;
-
-        tp = CALLOC (strlen ("/" GF_REPLICATE_TRASH_DIR) + strlen (path) + 1, sizeof (char));
-
-        strcpy (tp, GF_REPLICATE_TRASH_DIR);
-        strcat (tp, path);
-
-        c = strchr (tp, '/') + 1;
-        while (*c++)
-                if (*c == '/')
-                        *c = '-';
-
-        return tp;
-}
-
-
-int
-afr_sh_entry_expunge_rename (call_frame_t *expunge_frame, xlator_t *this,
-                             int active_src, inode_t *trash_inode)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
-
-        loc_t rename_loc;
-
-	priv          = this->private;
-	expunge_local = expunge_frame->local;
-
-        rename_loc.inode = inode_ref (expunge_local->loc.inode);
-        rename_loc.path  = make_trash_path (expunge_local->loc.path);
-        rename_loc.name  = strrchr (rename_loc.path, '/') + 1;
-        rename_loc.parent = trash_inode;
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"moving file/directory %s on %s to %s",
-		expunge_local->loc.path, priv->children[active_src]->name,
-                rename_loc.path);
-
-	STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_rename_cbk,
-			   (void *) (long) active_src,
-			   priv->children[active_src],
-			   priv->children[active_src]->fops->rename,
-			   &expunge_local->loc, &rename_loc);
-
-        loc_wipe (&rename_loc);
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_expunge_mkdir_cbk (call_frame_t *expunge_frame, void *cookie, xlator_t *this,
-                                int32_t op_ret, int32_t op_errno, inode_t *inode,
-                                struct stat *buf, struct stat *preparent,
-                                struct stat *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 = (long) cookie;
-
-        inode_t *trash_inode = NULL;
-
-        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,
-                        "mkdir of /" GF_REPLICATE_TRASH_DIR " failed on %s",
-                        priv->children[active_src]->name);
-
-                goto out;
-        }
-
-        /* mkdir successful */
-
-        trash_inode = inode_link (inode, expunge_local->loc.inode->table->root,
-                                  GF_REPLICATE_TRASH_DIR, buf);
-
-        afr_sh_entry_expunge_rename (expunge_frame, this, active_src,
-                                     trash_inode);
-        return 0;
-out:
-        AFR_STACK_DESTROY (expunge_frame);
-        afr_sh_entry_expunge_entry_done (frame, this, active_src);
-        return 0;
-}
-
-
-int
-afr_sh_entry_expunge_lookup_trash_cbk (call_frame_t *expunge_frame, void *cookie,
-                                       xlator_t *this,
-                                       int32_t op_ret, int32_t op_errno,
-                                       inode_t *inode, struct stat *buf,
-                                       dict_t *xattr, struct stat *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 = (long) cookie;
-
-        inode_t *trash_inode;
-        loc_t    trash_loc;
-
-        priv          = this->private;
-        expunge_local = expunge_frame->local;
-	expunge_sh    = &expunge_local->self_heal;
-	frame         = expunge_sh->sh_frame;
-
-        if ((op_ret != 0) && (op_errno == ENOENT)) {
-                init_trash_loc (&trash_loc, expunge_local->loc.inode->table);
-
-                gf_log (this->name, GF_LOG_TRACE,
-                        "creating directory " GF_REPLICATE_TRASH_DIR " on subvolume %s",
-                        priv->children[active_src]->name);
-
-                STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_mkdir_cbk,
-                                   (void *) (long) active_src,
-                                   priv->children[active_src],
-                                   priv->children[active_src]->fops->mkdir,
-                                   &trash_loc, 0777);
-
-                loc_wipe (&trash_loc);
-                return 0;
-        }
-
-        if (op_ret != 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "lookup of /" GF_REPLICATE_TRASH_DIR " failed on %s",
-                        priv->children[active_src]->name);
-                goto out;
-        }
-
-        /* lookup successful */
-
-        trash_inode = inode_link (inode, expunge_local->loc.inode->table->root,
-                                  GF_REPLICATE_TRASH_DIR, buf);
-
-        afr_sh_entry_expunge_rename (expunge_frame, this, active_src,
-                                     trash_inode);
-        return 0;
-out:
-        AFR_STACK_DESTROY (expunge_frame);
-        afr_sh_entry_expunge_entry_done (frame, this, active_src);
-        return 0;
-}
-
-
-int
-afr_sh_entry_expunge_lookup_trash (call_frame_t *expunge_frame, xlator_t *this,
-                                   int active_src)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *expunge_local = NULL;
-
-        inode_t *root  = NULL;
-        inode_t *trash = NULL;
-        loc_t trash_loc;
-
-	priv          = this->private;
-	expunge_local = expunge_frame->local;
-
-        root = expunge_local->loc.inode->table->root;
-
-        trash = inode_grep (root->table, root, GF_REPLICATE_TRASH_DIR);
-
-        if (trash) {
-                /* inode is in cache, so no need to mkdir */
-
-                afr_sh_entry_expunge_rename (expunge_frame, this, active_src,
-                                             trash);
-                return 0;
-        }
-
-        /* Not in cache, so look it up */
-
-        init_trash_loc (&trash_loc, expunge_local->loc.inode->table);
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"looking up /" GF_REPLICATE_TRASH_DIR " on %s",
-                priv->children[active_src]->name);
-
-	STACK_WIND_COOKIE (expunge_frame, afr_sh_entry_expunge_lookup_trash_cbk,
-			   (void *) (long) active_src,
-			   priv->children[active_src],
-			   priv->children[active_src]->fops->lookup,
-			   &trash_loc, NULL);
-
-        loc_wipe (&trash_loc);
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_expunge_remove (call_frame_t *expunge_frame, xlator_t *this,
-			     int active_src, struct stat *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;
-
-	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->st_mode & S_IFMT);
-
-	switch (type) {
-	case S_IFSOCK:
-	case S_IFREG:
-	case S_IFBLK:
-	case S_IFCHR:
-	case S_IFIFO:
-	case S_IFLNK:
-	case S_IFDIR:
-		afr_sh_entry_expunge_lookup_trash (expunge_frame, this, active_src);
-		break;
-	default:
-		gf_log (this->name, GF_LOG_ERROR,
-			"%s has unknown file type on %s: 0%o",
-			expunge_local->loc.path,
-			priv->children[source]->name, type);
-		goto out;
-		break;
-	}
-
-	return 0;
-out:
-	AFR_STACK_DESTROY (expunge_frame);
-	afr_sh_entry_expunge_entry_done (frame, this, active_src);
-
-	return 0;
-}
-
-
-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 stat *buf, dict_t *x,
-                                struct stat *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;
-}
-
-
-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 stat *buf, dict_t *x,
-                                struct stat *postparent)
+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     *expunge_local = NULL;
-	afr_self_heal_t *expunge_sh = NULL;
-	int              source = 0;
-	call_frame_t    *frame = NULL;
-	int              active_src = 0;
-
-
-	priv = this->private;
-	expunge_local = expunge_frame->local;
-	expunge_sh = &expunge_local->self_heal;
-	frame = expunge_sh->sh_frame;
-	active_src = expunge_sh->active_source;
-	source = (long) cookie;
-
-	if (op_ret == -1 && op_errno == ENOENT) {
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"missing entry %s on %s",
-			expunge_local->loc.path,
-			priv->children[source]->name);
+	int ret = -1;
 
-                expunge_sh->parentbuf = *postparent;
-
-		afr_sh_entry_expunge_purge (expunge_frame, this, active_src);
-
-		return 0;
-	}
-
-	if (op_ret == 0) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"%s exists under %s",
-			expunge_local->loc.path,
-			priv->children[source]->name);
-	} else {
-		gf_log (this->name, GF_LOG_TRACE,
-			"looking up %s under %s failed (%s)",
-			expunge_local->loc.path,
-			priv->children[source]->name,
-			strerror (op_errno));
-	}
-
-	AFR_STACK_DESTROY (expunge_frame);
-	afr_sh_entry_expunge_entry_done (frame, this, active_src);
-
-	return 0;
+	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_expunge_entry (call_frame_t *frame, xlator_t *this,
-			    char *name)
+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     *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;
+	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;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	active_src = sh->active_source;
-	source = sh->source;
-
-	if ((strcmp (name, ".") == 0)
-	    || (strcmp (name, "..") == 0)
-            || ((strcmp (local->loc.path, "/") == 0)
-                && (strcmp (name, GF_REPLICATE_TRASH_DIR) == 0))) {
-
-                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);
+	locked_on = alloca0 (priv->child_count);
 
-	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);
+	replies = alloca0 (priv->child_count * sizeof(*replies));
 
-	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));
-		} else {
-			gf_log (this->name, GF_LOG_TRACE,
-				"readdir of %s on subvolume %s complete",
-				local->loc.path,
-				priv->children[active_src]->name);
+	ret = afr_selfheal_entrylk (frame, this, fd->inode, this->name,
+				    name, locked_on);
+	{
+		if (ret < 2) {
+			ret = -ENOTCONN;
+			goto unlock;
 		}
 
-		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;
+		inode = afr_selfheal_unlocked_lookup_on (frame, fd->inode, name,
+							 replies, locked_on);
+		if (!inode) {
+			ret = -ENOMEM;
+			goto unlock;
+		}
 
-	list_for_each_entry (entry, &entries->list, list) {
-                afr_sh_entry_expunge_entry (frame, this, entry->d_name);
+		ret = __afr_selfheal_entry_dirent (frame, this, fd, name, inode,
+						   source, sources, healed_sinks,
+						   locked_on, replies);
 	}
-
-	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;
+unlock:
+	afr_selfheal_unentrylk (frame, this, fd->inode, this->name, name,
+				locked_on);
+	if (inode)
+		inode_unref (inode);
+	return ret;
 }
 
 
-int
-afr_sh_entry_expunge_all (call_frame_t *frame, xlator_t *this)
+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     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              active_src = -1;
+	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;
-	local = frame->local;
-	sh = &local->self_heal;
+	subvol = priv->children[child];
 
-	sh->offset = 0;
+	INIT_LIST_HEAD (&entries.list);
 
-	if (sh->source == -1) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"no active sources for %s to expunge entries",
-			local->loc.path);
-		goto out;
-	}
+	iter_frame = afr_copy_frame (frame);
+	if (!iter_frame)
+		return -ENOMEM;
 
-	active_src = next_active_sink (frame, this, sh->active_source);
-	sh->active_source = active_src;
+	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 (sh->op_failed) {
-		goto out;
-	}
+			if (!strcmp (entry->d_name, ".") ||
+			    !strcmp (entry->d_name, ".."))
+				continue;
 
-	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;
+			if (__is_root_gfid (fd->inode->gfid) &&
+			    !strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR))
+				continue;
 
-}
-
-
-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);
+			ret = afr_selfheal_entry_dirent (iter_frame, this, fd,
+							 source, sources,
+							 healed_sinks,
+							 entry->d_name);
+			AFR_STACK_RESET (iter_frame);
 
-int
-afr_sh_entry_impunge_entry_done (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;
-	int              call_count = 0;
-
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
+			if (ret)
+				break;
+		}
 
-	LOCK (&frame->lock);
-	{
+		gf_dirent_free (&entries);
+		if (ret)
+			break;
 	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		afr_sh_entry_impunge_subvol (frame, this, active_src);
 
-	return 0;
+	AFR_STACK_DESTROY (iter_frame);
+	return ret;
 }
 
-
-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 stat *preop, struct stat *postop)
+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)
 {
-	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;
+	int i = 0;
+	afr_private_t *priv = NULL;
+	int ret = 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);
+	gf_log (this->name, GF_LOG_INFO, "performing entry selfheal on %s",
+		uuid_utoa (fd->inode->gfid));
 
-	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 (i != source && !healed_sinks[i])
+			continue;
+		ret = afr_selfheal_entry_do_subvol (frame, this, fd, i, source,
+						    sources, healed_sinks);
+		if (ret)
+			break;
 	}
-
-	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;
-	call_frame_t    *frame = NULL;
-	int              child_index = 0;
-
-        struct stat stbuf;
-        int32_t     valid = 0;
-
-	priv          = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh    = &impunge_local->self_heal;
-	frame         = impunge_sh->sh_frame;
-
-	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.st_uid,
-		impunge_local->cont.lookup.buf.st_gid);
-
-#ifdef HAVE_STRUCT_STAT_ST_ATIM_TV_NSEC
-	stbuf.st_atim = impunge_local->cont.lookup.buf.st_atim;
-	stbuf.st_mtim = impunge_local->cont.lookup.buf.st_mtim;
-
-#elif HAVE_STRUCT_STAT_ST_ATIMESPEC_TV_NSEC
-	stbuf.st_atimespec = impunge_local->cont.lookup.buf.st_atimespec;
-	stbuf.st_mtimespec = impunge_local->cont.lookup.buf.st_mtimespec;
-#else
-	stbuf.st_atime = impunge_local->cont.lookup.buf.st_atime;
-	stbuf.st_mtime = impunge_local->cont.lookup.buf.st_mtime;
-#endif
-
-        stbuf.st_uid = impunge_local->cont.lookup.buf.st_uid;
-        stbuf.st_gid = impunge_local->cont.lookup.buf.st_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 stat *preop, struct stat *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);
-
-        FREE (parent_loc);
-
-        AFR_STACK_DESTROY (setattr_frame);
-        return 0;
+	return ret;
 }
 
 
-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 stat *stbuf,
-                                  struct stat *preparent,
-                                  struct stat *postparent)
+static int
+__afr_selfheal_entry_finalize_source (xlator_t *this, unsigned char *sources,
+				      unsigned char *sinks,
+				      unsigned char *locked_on,
+				      struct afr_reply *replies)
 {
-	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 stat parentbuf;
+	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;
-	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->st_mode = stbuf->st_mode;
+	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);
 
-        xattr = get_new_dict ();
-        dict_ref (xattr);
-
-        idx = afr_index_for_transaction_type (AFR_METADATA_TRANSACTION);
-        pending_array[idx] = hton32 (1);
-        if (S_ISDIR (stbuf->st_mode))
-                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));
-
-        valid         = GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
-        parentbuf     = impunge_sh->parentbuf;
-        setattr_frame = copy_frame (impunge_frame);
-
-        parent_loc = CALLOC (1, sizeof (*parent_loc));
-        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;
+	if (locked_count == sinks_count || !sources_count) {
+		return -1;
 	}
-	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 (sources[i]) {
+			source = i;
+			break;
+		}
 	}
 
-	return 0;
-}
-
-
-int
-afr_sh_entry_impunge_mknod (call_frame_t *impunge_frame, xlator_t *this,
-			    int child_index, struct stat *stbuf)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-
-	gf_log (this->name, GF_LOG_DEBUG,
-		"creating missing file %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->mknod,
-			   &impunge_local->loc,
-			   stbuf->st_mode, stbuf->st_rdev);
-
-	return 0;
-}
-
-
-
-int
-afr_sh_entry_impunge_mkdir (call_frame_t *impunge_frame, xlator_t *this,
-			    int child_index, struct stat *stbuf)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-
-	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, stbuf->st_mode);
-
-	return 0;
+	return source;
 }
 
 
-int
-afr_sh_entry_impunge_symlink (call_frame_t *impunge_frame, xlator_t *this,
-			      int child_index, const char *linkname)
+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     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-
+	int ret = -1;
+	int source = -1;
+	afr_private_t *priv = NULL;
+	int i = 0;
 
 	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-
-	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);
-
-	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 stat *preparent,
-                                         struct stat *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);
-	}
-
-	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;
-	afr_self_heal_t *impunge_sh    = NULL;
-
-	priv          = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh    = &impunge_local->self_heal;
-
-	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 stat *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;
-	}
-
-        /* 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;
-        }
+	ret = afr_selfheal_unlocked_discover (frame, fd->inode, fd->inode->gfid,
+					      replies);
+	if (ret)
+		return ret;
 
-out:
-	LOCK (&impunge_frame->lock);
-	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
+	ret = afr_selfheal_find_direction (frame, this, replies,
+					   AFR_ENTRY_TRANSACTION,
+					   locked_on, sources, sinks);
+	if (ret)
+		return ret;
 
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
+	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;
 
-	return 0;
-}
+	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];
 
-int
-afr_sh_entry_impunge_readlink_sink (call_frame_t *impunge_frame, xlator_t *this,
-                                    int child_index)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-
-	gf_log (this->name, GF_LOG_DEBUG,
-		"checking symlink target of %s on %s",
-		impunge_local->loc.path, priv->children[child_index]->name);
-
-	STACK_WIND_COOKIE (impunge_frame, afr_sh_entry_impunge_readlink_sink_cbk,
-			   (void *) (long) child_index,
-			   priv->children[child_index],
-			   priv->children[child_index]->fops->readlink,
-			   &impunge_local->loc, 4096);
-
-	return 0;
+	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 stat *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;
+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;
-	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 = strdup (linkname);
+	sources = alloca0 (priv->child_count);
+	sinks = alloca0 (priv->child_count);
+	healed_sinks = alloca0 (priv->child_count);
+	data_lock = alloca0 (priv->child_count);
 
-	afr_sh_entry_impunge_readlink_sink (impunge_frame, this, child_index);
+	locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
 
-	return 0;
-
-out:
-	LOCK (&impunge_frame->lock);
+	ret = afr_selfheal_entrylk (frame, this, fd->inode, this->name, NULL,
+				    data_lock);
 	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
+		if (ret < 2) {
+			ret = -ENOTCONN;
+			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_prepare (frame, this, fd, data_lock,
+						    sources, sinks, healed_sinks,
+						    locked_replies, &source);
 	}
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_impunge_readlink (call_frame_t *impunge_frame, xlator_t *this,
-			       int child_index, struct stat *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;
-
-	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 stat *buf,
-					  dict_t *xattr,struct stat *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));
+unlock:
+	afr_selfheal_unentrylk (frame, this, fd->inode, this->name, NULL,
+				data_lock);
+	if (ret < 0)
 		goto out;
-	}
 
-        impunge_sh->parentbuf = *postparent;
-
-	impunge_local->cont.lookup.buf = *buf;
-	type = (buf->st_mode & S_IFMT);
-
-	switch (type) {
-	case S_IFSOCK:
-	case S_IFREG:
-	case S_IFBLK:
-	case S_IFCHR:
-	case S_IFIFO:
-		afr_sh_entry_impunge_mknod (impunge_frame, this,
-					    child_index, buf);
-		break;
-	case S_IFLNK:
-		afr_sh_entry_impunge_readlink (impunge_frame, this,
-					       child_index, buf);
-		break;
-	case S_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);
+	ret = afr_selfheal_entry_do (frame, this, fd, source, sources,
+				     healed_sinks, locked_replies);
+	if (ret)
 		goto out;
-		break;
-	}
-
-	return 0;
 
+	ret = afr_selfheal_undo_pending (frame, this, fd->inode, sources, sinks,
+					 healed_sinks, AFR_ENTRY_TRANSACTION,
+					 locked_replies, data_lock);
 out:
-	LOCK (&impunge_frame->lock);
-	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
-
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
-	}
-
-	return 0;
-}
-
-
-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;
+	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 stat *buf, dict_t *x,
-                                struct stat *postparent)
+static fd_t *
+afr_selfheal_data_opendir (xlator_t *this, inode_t *inode)
 {
-	afr_private_t   *priv = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              call_count = 0;
-	int              child_index = 0;
-	call_frame_t    *frame = NULL;
-	int              active_src = 0;
-
-	priv = this->private;
-	impunge_local = impunge_frame->local;
-	impunge_sh = &impunge_local->self_heal;
-	frame = impunge_sh->sh_frame;
-	child_index = (long) cookie;
-	active_src = impunge_sh->active_source;
-
-	if ((op_ret == -1 && op_errno == ENOENT)
-            || (S_ISLNK (impunge_sh->impunging_entry_mode))) {
-
-                /*
-                 * A symlink's target might have changed, so
-                 * always go down the recreate path for them.
-                 */
+	loc_t loc = {0,};
+	int ret = 0;
+	fd_t *fd = NULL;
 
-		/* decrease call_count in recreate-callback */
+	fd = fd_create (inode, 0);
+	if (!fd)
+		return NULL;
 
-		gf_log (this->name, GF_LOG_TRACE,
-			"missing entry %s on %s",
-			impunge_local->loc.path,
-			priv->children[child_index]->name);
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-		afr_sh_entry_impunge_recreate (impunge_frame, this,
-					       child_index);
-		return 0;
-	}
-
-	if (op_ret == 0) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"%s exists under %s",
-			impunge_local->loc.path,
-			priv->children[child_index]->name);
-
-                impunge_sh->parentbuf = *postparent;
+	ret = syncop_opendir (this, &loc, fd);
+	if (ret) {
+		fd_unref (fd);
+		fd = NULL;
 	} else {
-		gf_log (this->name, GF_LOG_TRACE,
-			"looking up %s under %s failed (%s)",
-			impunge_local->loc.path,
-			priv->children[child_index]->name,
-			strerror (op_errno));
-	}
-
-	LOCK (&impunge_frame->lock);
-	{
-		call_count = --impunge_local->call_count;
-	}
-	UNLOCK (&impunge_frame->lock);
-
-	if (call_count == 0) {
-		AFR_STACK_DESTROY (impunge_frame);
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
-	}
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_impunge_entry (call_frame_t *frame, xlator_t *this,
-			    gf_dirent_t *entry)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	int              ret = -1;
-	call_frame_t    *impunge_frame = NULL;
-	afr_local_t     *impunge_local = NULL;
-	afr_self_heal_t *impunge_sh = NULL;
-	int              active_src = 0;
-	int              i = 0;
-	int              call_count = 0;
-	int              op_errno = 0;
-
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	active_src = sh->active_source;
-
-	if ((strcmp (entry->d_name, ".") == 0)
-	    || (strcmp (entry->d_name, "..") == 0)
-            || ((strcmp (local->loc.path, "/") == 0)
-                && (strcmp (entry->d_name, GF_REPLICATE_TRASH_DIR) == 0))) {
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"skipping inspection of %s under %s",
-			entry->d_name, local->loc.path);
-		goto out;
-	}
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"inspecting existance of %s under %s",
-		entry->d_name, local->loc.path);
-
-	impunge_frame = copy_frame (frame);
-	if (!impunge_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-	ALLOC_OR_GOTO (impunge_local, afr_local_t, out);
-
-	impunge_frame->local = impunge_local;
-	impunge_sh = &impunge_local->self_heal;
-	impunge_sh->sh_frame = frame;
-	impunge_sh->active_source = active_src;
-
-        impunge_sh->impunging_entry_mode = entry->d_stat.st_mode;
-
-	ret = build_child_loc (this, &impunge_local->loc, &local->loc, entry->d_name);
-	if (ret != 0) {
-		goto out;
-	}
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (i == active_src)
-			continue;
-		if (local->child_up[i] == 0)
-			continue;
-		if (sh->sources[i] == 1)
-			continue;
-		call_count++;
-	}
-
-	impunge_local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (i == active_src)
-			continue;
-		if (local->child_up[i] == 0)
-			continue;
-		if (sh->sources[i] == 1)
-			continue;
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"looking up %s on %s", impunge_local->loc.path,
-			priv->children[i]->name);
-
-		STACK_WIND_COOKIE (impunge_frame,
-				   afr_sh_entry_impunge_entry_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->lookup,
-				   &impunge_local->loc, 0);
-
-		if (!--call_count)
-			break;
-	}
-
-	ret = 0;
-out:
-	if (ret == -1)
-		afr_sh_entry_impunge_entry_done (frame, this, active_src);
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_impunge_readdir_cbk (call_frame_t *frame, void *cookie,
-				  xlator_t *this,
-				  int32_t op_ret, int32_t op_errno,
-				  gf_dirent_t *entries)
-{
-	afr_private_t   *priv = NULL;
-	afr_local_t     *local  = NULL;
-	afr_self_heal_t *sh  = NULL;
-	gf_dirent_t     *entry = NULL;
-	off_t            last_offset = 0;
-	int              active_src = 0;
-	int              entry_count = 0;
-
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	active_src = sh->active_source;
-
-	if (op_ret <= 0) {
-		if (op_ret < 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"readdir of %s on subvolume %s failed (%s)",
-				local->loc.path,
-				priv->children[active_src]->name,
-				strerror (op_errno));
-		} else {
-			gf_log (this->name, GF_LOG_TRACE,
-				"readdir of %s on subvolume %s complete",
-				local->loc.path,
-				priv->children[active_src]->name);
-		}
-
-		afr_sh_entry_impunge_all (frame, this);
-		return 0;
-	}
-
-	list_for_each_entry (entry, &entries->list, list) {
-		last_offset = entry->d_off;
-                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_impunge_entry (frame, this, entry);
-	}
-
-	return 0;
-}
-				  
-
-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)
-{
-	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;
-
-	active_src = next_active_source (frame, this, sh->active_source);
-	sh->active_source = active_src;
-
-	if (sh->op_failed) {
-		afr_sh_entry_finish (frame, this);
-		return 0;
+		fd_bind (fd);
 	}
 
-	if (active_src == -1) {
-		/* completed creating missing files on all subvolumes */
-		afr_sh_entry_expunge_all (frame, this);
-		return 0;
-	}
-
-	gf_log (this->name, GF_LOG_TRACE,
-		"impunging entries of %s on %s to other sinks",
-		local->loc.path, priv->children[active_src]->name);
+	loc_wipe (&loc);
 
-	afr_sh_entry_impunge_subvol (frame, this, active_src);
-
-	return 0;
+	return fd;
 }
 
 
-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;
-	priv = this->private;
-
-	child_index = (long) cookie;
-
-	/* TODO: some of the open's might fail.
-	   In that case, modify cleanup fn to send flush on those 
-	   fd's which are already open */
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"opendir of %s failed on child %s (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-			sh->op_failed = 1;
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-		if (sh->op_failed) {
-			afr_sh_entry_finish (frame, this);
-			return 0;
-		}
-		gf_log (this->name, GF_LOG_TRACE,
-			"fd for %s opened, commencing sync",
-			local->loc.path);
-
-		sh->active_source = -1;
-		afr_sh_entry_impunge_all (frame, this);
-	}
-
-	return 0;
-}
-
 
 int
-afr_sh_entry_open (call_frame_t *frame, xlator_t *this)
+afr_selfheal_entry (call_frame_t *frame, xlator_t *this, inode_t *inode)
 {
-	int i = 0;				
-	int call_count = 0;		     
-
-	int source = -1;
-	int *sources = NULL;
-
+	afr_private_t *priv = NULL;
+	unsigned char *locked_on = NULL;
 	fd_t *fd = NULL;
-
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	afr_self_heal_t *sh = NULL;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	source  = local->self_heal.source;
-	sources = local->self_heal.sources;
-
-	sh->block_size = 131072;
-	sh->offset = 0;
-
-	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--;
-	}
-
-	/* 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;
-	}
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_sync_prepare (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              active_sinks = 0;
-	int              source = 0;
-	int              i = 0;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	source = sh->source;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->sources[i] == 0 && local->child_up[i] == 1) {
-			active_sinks++;
-			sh->success[i] = 1;
-		}
-	}
-	if (source != -1)
-		sh->success[source] = 1;
-
-	if (active_sinks == 0) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"no active sinks for self-heal on dir %s",
-			local->loc.path);
-		afr_sh_entry_finish (frame, this);
-		return 0;
-	}
-	if (source == -1 && active_sinks < 2) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"cannot sync with 0 sources and 1 sink on dir %s",
-			local->loc.path);
-		afr_sh_entry_finish (frame, this);
-		return 0;
-	}
-	sh->active_sinks = active_sinks;
-
-	if (source != -1)
-		gf_log (this->name, GF_LOG_DEBUG,
-			"self-healing directory %s from subvolume %s to "
-                        "%d other",
-			local->loc.path, priv->children[source]->name,
-			active_sinks);
-	else
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no active sources for %s found. "
-			"merging all entries as a conservative decision",
-			local->loc.path);
-
-	afr_sh_entry_open (frame, this);
-
-	return 0;
-}
-
-
-int
-afr_sh_entry_fix (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              source = 0;
-
-        int nsources = 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;
-        }
-
-	afr_sh_supress_errenous_children (sh->sources, sh->child_errno,
-					  priv->child_count);
-
-	source = afr_sh_select_source (sh->sources, priv->child_count);
-
-        sh->source = source;
-
-        if (sh->background) {
-                sh->unwind (sh->orig_frame, this);
-                sh->unwound = _gf_true;
-        }
-
-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 stat *buf, dict_t *xattr,
-                         struct stat *postparent)
-{
-	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) {
-			sh->xattr[child_index] = dict_ref (xattr);
-			sh->buf[child_index] = *buf;
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-		afr_sh_entry_fix (frame, this);
-	}
-
-	return 0;
-}
-
-
-
-int
-afr_sh_entry_lookup (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;
 	int ret = 0;
-	int call_count = 0;
-	int i = 0;
-
-	priv  = this->private;
-	local = frame->local;
-	sh    = &local->self_heal;
-
-	call_count = afr_up_children_count (priv->child_count,
-                                            local->child_up);
-
-	local->call_count = call_count;
-	
-	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));
-                }
-        }
-
-	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;
-}
 
+	priv = this->private;
 
+	fd = afr_selfheal_data_opendir (this, inode);
+	if (!fd)
+		return -EIO;
 
-int
-afr_sh_entry_lock_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;
-	int              call_count = 0;
-	int              child_index = (long) cookie;
+	locked_on = alloca0 (priv->child_count);
 
-	/* TODO: what if lock fails? */
-	
-	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->op_failed = 1;
-
-                        sh->locked_nodes[child_index] = 0;
-			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 {
-                        sh->locked_nodes[child_index] = 1;
-			gf_log (this->name, GF_LOG_TRACE,
-				"inode of %s on child %d locked",
-				local->loc.path, child_index);
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-		if (sh->op_failed == 1) {
-			afr_sh_entry_finish (frame, this);
-			return 0;
+		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;
 		}
 
-		afr_sh_entry_lookup (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_lock (call_frame_t *frame, xlator_t *this)
-{
-	int i = 0;				
-	int call_count = 0;		     
-
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	afr_self_heal_t * sh  = NULL;
+	if (fd)
+		fd_unref (fd);
 
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	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]) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"locking %s on subvolume %s",
-				local->loc.path, priv->children[i]->name);
-
-			STACK_WIND_COOKIE (frame, afr_sh_entry_lock_cbk,
-					   (void *) (long) i,
-					   priv->children[i], 
-					   priv->children[i]->fops->entrylk,
-                                           this->name,
-					   &local->loc, NULL,
-					   ENTRYLK_LOCK_NB, ENTRYLK_WRLCK);
-			if (!--call_count)
-				break;
-		}
-	}
-
-	return 0;
-}
-
-
-int
-afr_self_heal_entry (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t   *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-
-
-	priv = this->private;
-	local = frame->local;
-	sh = &local->self_heal;
-
-	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 bddccf682..b31a33237 100644
--- a/xlators/cluster/afr/src/afr-self-heal-metadata.c
+++ b/xlators/cluster/afr/src/afr-self-heal-metadata.c
@@ -1,822 +1,281 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 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 "byte-order.h"
-
-#include "afr-transaction.h"
 #include "afr-self-heal.h"
-#include "afr-self-heal-common.h"
-
-
-int
-afr_sh_metadata_done (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              i = 0;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-//	memset (sh->child_errno, 0, sizeof (int) * priv->child_count);
-	memset (sh->buf, 0, sizeof (struct stat) * 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 (S_ISREG (sh->mode)) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"proceeding to data check on %s",
-				local->loc.path);
-			afr_self_heal_data (frame, this);
-			return 0;
-		}
-
-		if (S_ISDIR (sh->mode)) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"proceeding to entry check on %s",
-				local->loc.path);
-			afr_self_heal_entry (frame, this);
-			return 0;
-		}
-		gf_log (this->name, GF_LOG_DEBUG,
-			"completed self heal of %s",
-			local->loc.path);
-
-		sh->completion_cbk (frame, this);
-	}
-
-	return 0;
-}
-
-
-int
-afr_sh_metadata_unlck_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			   int32_t op_ret, int32_t op_errno)
-{
-	afr_local_t      *local = NULL;
-	int               call_count = 0;
-
-
-	local = frame->local;
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		afr_sh_metadata_done (frame, this);
-
-	return 0;
-}
-
-
-int
-afr_sh_metadata_finish (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              i = 0;
-	int              call_count = 0;
-	struct flock     flock = {0, };
-
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-        for (i = 0; i < priv->child_count; i++) {
-                if (sh->locked_nodes[i])
-                        call_count++;
-        }
-
-        if (call_count == 0) {
-                afr_sh_metadata_done (frame, this);
-                return 0;
-        }
-
-	local->call_count = call_count;
-
-	for (i = 0; i < priv->child_count; i++) {
-		flock.l_start   = 0;
-		flock.l_len     = 0;
-		flock.l_type    = F_UNLCK;
-
-		if (sh->locked_nodes[i]) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"unlocking %s on subvolume %s",
-				local->loc.path, priv->children[i]->name);
-
-			STACK_WIND (frame, afr_sh_metadata_unlck_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->inodelk,
-                                    this->name,
-				    &local->loc, F_SETLK, &flock);
-
-			if (!--call_count)
-				break;
-		}
-	}
-
-	return 0;
-}
-
-
-int
-afr_sh_metadata_erase_pending_cbk (call_frame_t *frame, void *cookie,
-				   xlator_t *this, int32_t op_ret,
-				   int32_t op_errno, dict_t *xattr)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int             call_count = 0;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	LOCK (&frame->lock);
-	{
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
+#include "byte-order.h"
 
-	if (call_count == 0)
-		afr_sh_metadata_finish (frame, this);
-
-	return 0;
-}
 
+#define AFR_HEAL_ATTR (GF_SET_ATTR_UID|GF_SET_ATTR_GID|GF_SET_ATTR_MODE)
 
 int
-afr_sh_metadata_erase_pending (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              call_count = 0;
-	int              i = 0;
-	dict_t          **erase_xattr = NULL;
-
+	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;
 
-	afr_sh_pending_to_delta (priv, sh->xattr, sh->delta_matrix, 
-                                 sh->success, priv->child_count,
-                                 AFR_METADATA_TRANSACTION);
+	loc.inode = inode_ref (inode);
+	uuid_copy (loc.gfid, inode->gfid);
 
-	erase_xattr = CALLOC (sizeof (*erase_xattr), priv->child_count);
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (sh->xattr[i]) {
-			call_count++;
+	gf_log (this->name, GF_LOG_INFO, "performing metadata selfheal on %s",
+		uuid_utoa (inode->gfid));
 
-			erase_xattr[i] = get_new_dict();
-			dict_ref (erase_xattr[i]);
-		}
+	ret = syncop_getxattr (priv->children[source], &loc, &xattr, NULL);
+	if (ret < 0) {
+		loc_wipe (&loc);
+		return -EIO;
 	}
 
-	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);
-	}
+	afr_filter_xattrs (xattr);
+	dict_del (xattr, GF_SELINUX_XATTR_KEY);
 
 	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_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]);
+		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);
 		}
-	}
-	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;
-		}
+		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_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 stat *preop, struct stat *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 stat      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
-	 */
+	loc_wipe (&loc);
 	if (xattr)
-		call_count = active_sinks * 2;
-	else
-		call_count = active_sinks;
-
-	local->call_count = call_count;
-
-#ifdef HAVE_STRUCT_STAT_ST_ATIM_TV_NSEC
-	stbuf.st_atim = sh->buf[source].st_atim;
-	stbuf.st_mtim = sh->buf[source].st_mtim;
-        
-#elif HAVE_STRUCT_STAT_ST_ATIMESPEC_TV_NSEC
-	stbuf.st_atimespec = sh->buf[source].st_atimespec;
-	stbuf.st_mtimespec = sh->buf[source].st_mtimespec;
-#else
-	stbuf.st_atime = sh->buf[source].st_atime;
-	stbuf.st_mtime = sh->buf[source].st_mtime;
-#endif
-
-        stbuf.st_uid = sh->buf[source].st_uid;
-        stbuf.st_gid = sh->buf[source].st_gid;
-
-        stbuf.st_mode = sh->buf[source].st_mode;
-        
-        valid = GF_SET_ATTR_MODE  | 
-                GF_SET_ATTR_UID   | GF_SET_ATTR_GID |
-                GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (call_count == 0) {
-			break;
-		}
-		if (sh->sources[i] || !local->child_up[i])
-			continue;
-
-		gf_log (this->name, GF_LOG_DEBUG,
-			"self-healing metadata of %s from %s to %s",
-			local->loc.path, priv->children[source]->name,
-			priv->children[i]->name);
-
-		STACK_WIND_COOKIE (frame, afr_sh_metadata_setattr_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->setattr,
-				   &local->loc, &stbuf, valid);
-
-		call_count--;
-
-		if (!xattr)
-			continue;
-
-		STACK_WIND_COOKIE (frame, afr_sh_metadata_xattr_cbk,
-				   (void *) (long) i,
-				   priv->children[i],
-				   priv->children[i]->fops->setxattr,
-				   &local->loc, xattr, 0);
-		call_count--;
-	}
+		dict_unref (xattr);
 
 	return 0;
 }
 
 
-int
-afr_sh_metadata_getxattr_cbk (call_frame_t *frame, void *cookie,
-			      xlator_t *this,
-			      int32_t op_ret, int32_t op_errno, dict_t *xattr)
+/*
+ * 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              source = 0;
-
-        int i;
+	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;
 
-	source = sh->source;
+	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 (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);
+	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;
+			}
+		}
 	}
 
-	return 0;
-}
-
-
-int
-afr_sh_metadata_sync_prepare (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t     *local = NULL;
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              active_sinks = 0;
-	int              source = 0;
-	int              i = 0;
-
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
-
-	source = sh->source;
-
 	for (i = 0; i < priv->child_count; i++) {
-		if (sh->sources[i] == 0 && local->child_up[i] == 1) {
-			active_sinks++;
-			sh->success[i] = 1;
+		if (!sources[i])
+			continue;
+		if (source == -1) {
+			source = i;
+			first = replies[i].poststat;
 		}
 	}
-	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;
-
-		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.");
-
-                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])
+		if (!sources[i])
 			continue;
-
-		if (PERMISSION_DIFFERS (&sh->buf[i], &sh->buf[source]))
-			sh->sources[i] = 0;
-
-		if (OWNERSHIP_DIFFERS (&sh->buf[i], &sh->buf[source]))
-			sh->sources[i] = 0;
+		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;
+		}
 	}
 
-	afr_sh_metadata_sync_prepare (frame, this);
-
-	return 0;
+	return source;
 }
 
 
-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 stat *buf, dict_t *xattr,
-                            struct stat *postparent)
+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              call_count = 0;
-	int              child_index = 0;
-
+	int ret = -1;
+	int source = -1;
+	afr_private_t *priv = NULL;
+	int i = 0;
 
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-	child_index = (long) cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"path %s on subvolume %s is of mode 0%o",
-				local->loc.path,
-				priv->children[child_index]->name,
-				buf->st_mode);
-
-			sh->buf[child_index] = *buf;
-			if (xattr)
-				sh->xattr[child_index] = dict_ref (xattr);
-		} else {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"path %s on subvolume %s => -1 (%s)",
-				local->loc.path,
-				priv->children[child_index]->name,
-				strerror (op_errno));
-
-			sh->child_errno[child_index] = op_errno;
-		}
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		afr_sh_metadata_fix (frame, this);
-
-	return 0;
+	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_lookup (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              i = 0;
-	int              call_count = 0;
-	dict_t          *xattr_req = NULL;
-	int              ret = 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;
 
-	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));
-                }
-        }
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"looking up %s on %s",
-				local->loc.path, priv->children[i]->name);
-
-			STACK_WIND_COOKIE (frame, afr_sh_metadata_lookup_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->lookup,
-					   &local->loc, xattr_req);
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	if (xattr_req)
-		dict_unref (xattr_req);
-
-	return 0;
-}
+	priv = this->private;
 
+	sources = alloca0 (priv->child_count);
+	sinks = alloca0 (priv->child_count);
+	healed_sinks = alloca0 (priv->child_count);
+	data_lock = alloca0 (priv->child_count);
 
-int
-afr_sh_metadata_lk_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 = (long) cookie;
-
-	/* TODO: what if lock fails? */
-	
-	local = frame->local;
-	sh = &local->self_heal;
-	priv = this->private;
+	locked_replies = alloca0 (sizeof (*locked_replies) * priv->child_count);
 
-	LOCK (&frame->lock);
+	ret = afr_selfheal_inodelk (frame, this, inode, this->name,
+				    LLONG_MAX - 1, 0, data_lock);
 	{
-		if (op_ret == -1) {
-			sh->op_failed = 1;
-
-                        sh->locked_nodes[child_index] = 0;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"locking of %s on child %d failed: %s",
-				local->loc.path, child_index,
-				strerror (op_errno));
-		} else {
-                        sh->locked_nodes[child_index] = 1;
-			gf_log (this->name, GF_LOG_TRACE,
-				"inode of %s on child %d locked",
-				local->loc.path, child_index);
+		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_metadata_finish (frame, this);
-			return 0;
-		}
+		ret = __afr_selfheal_metadata_prepare (frame, this, inode, data_lock,
+						       sources, sinks, healed_sinks,
+						       locked_replies);
+		if (ret < 0)
+			goto unlock;
 
-		afr_sh_metadata_lookup (frame, this);
+		source = ret;
+		ret = 0;
 	}
-
-	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_lock (call_frame_t *frame, xlator_t *this)
+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              i = 0;
-	int              call_count = 0;
-	struct flock     flock = {0, };
-
+	afr_private_t *priv = NULL;
+	unsigned char *locked_on = NULL;
+	int ret = 0;
 
-	local = frame->local;
-	sh = &local->self_heal;
 	priv = this->private;
 
-	call_count = afr_up_children_count (priv->child_count,
-                                            local->child_up);
-	local->call_count = call_count;
+	locked_on = alloca0 (priv->child_count);
 
-	for (i = 0; i < priv->child_count; i++) {
-		flock.l_start   = 0;
-		flock.l_len     = 0;
-		flock.l_type    = F_WRLCK;
-
-		if (local->child_up[i]) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"locking %s on subvolume %s",
-				local->loc.path, priv->children[i]->name);
-
-			STACK_WIND_COOKIE (frame, afr_sh_metadata_lk_cbk,
-					   (void *) (long) i,
-					   priv->children[i],
-					   priv->children[i]->fops->inodelk,
-                                           this->name,
-					   &local->loc, F_SETLK, &flock);
-
-			if (!--call_count)
-				break;
+	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;
 		}
-	}
-
-	return 0;
-}
-
 
-int
-afr_self_heal_metadata (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 (local->self_heal.need_metadata_self_heal && priv->metadata_self_heal) {
-		afr_sh_metadata_lock (frame, this);
-	} else {
-		afr_sh_metadata_done (frame, this);
+		ret = __afr_selfheal_metadata (frame, this, inode, locked_on);
 	}
+unlock:
+	afr_selfheal_uninodelk (frame, this, inode, priv->sh_domain, 0, 0, locked_on);
 
-	return 0;
+	return ret;
 }
-
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 1c2743a48..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) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 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) ((S_IFMT & ((struct stat *)buf1)->st_mode) != (S_IFMT & ((struct stat *)buf2)->st_mode))
-#define PERMISSION_DIFFERS(buf1,buf2) ((((struct stat *)buf1)->st_mode) != (((struct stat *)buf2)->st_mode))
-#define OWNERSHIP_DIFFERS(buf1,buf2) ((((struct stat *)buf1)->st_uid) != (((struct stat *)buf2)->st_uid) || (((struct stat *)buf1)->st_gid != (((struct stat *)buf2)->st_gid)))
-#define SIZE_DIFFERS(buf1,buf2) ((((struct stat *)buf1)->st_size) != (((struct stat *)buf2)->st_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) ((((struct stat *)buf1)->st_size > (((struct stat *)buf2)->st_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 06e8931f2..205ff759e 100644
--- a/xlators/cluster/afr/src/afr-transaction.c
+++ b/xlators/cluster/afr/src/afr-transaction.c
@@ -1,1407 +1,1579 @@
 /*
-  Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include "dict.h"
 #include "byte-order.h"
 #include "common-utils.h"
+#include "timer.h"
 
 #include "afr.h"
 #include "afr-transaction.h"
 
 #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 */
-
+gf_boolean_t
+afr_changelog_pre_op_update (call_frame_t *frame, xlator_t *this);
 
-static void
-afr_pid_save (call_frame_t *frame)
-{
-        afr_local_t * local = NULL;
-
-        local = frame->local;
-
-        local->saved_pid = frame->root->pid;
-}
+int
+afr_changelog_do (call_frame_t *frame, xlator_t *this, dict_t *xattr,
+		  afr_changelog_resume_t changelog_resume);
 
 
-static void
-afr_pid_restore (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;
 
-        frame->root->pid = local->saved_pid;
-}
-
-
-static void
-__mark_all_pending (int32_t *pending[], int child_count,
-                    afr_transaction_type type)
-{	
-	int i;
-        int j;
+        call_count = AFR_COUNT (local->transaction.pre_op, priv->child_count);
 
-        for (i = 0; i < child_count; i++) {
-                j = afr_index_for_transaction_type (type);
-		pending[i][j] = hton32 (1);
+        if (call_count == 0) {
+                local->transaction.resume (frame, this);
+                return 0;
         }
-}
 
+        local->call_count = call_count;
 
-static void
-__mark_child_dead (int32_t *pending[], int child_count, int child,
-                   afr_transaction_type type)
-{
-        int j;
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.pre_op[i]) {
+			local->transaction.wind (frame, this, i);
 
-        j = afr_index_for_transaction_type (type);
-        
-	pending[child][j] = 0;
+                        if (!--call_count)
+                                break;
+                }
+        }
+
+        return 0;
 }
 
 
-static void
-__mark_fop_failed_on_fd (fd_t *fd, xlator_t *this,
-                         int child_index)
+int
+__afr_txn_write_done (call_frame_t *frame, xlator_t *this)
 {
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
-
-        int ret = 0;
+        afr_local_t *local = NULL;
 
-        ret = fd_ctx_get (fd, this, &ctx);
+        local = frame->local;
 
-        if (ret < 0)
-                goto out;
+        local->transaction.unwind (frame, this);
 
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+        AFR_STACK_DESTROY (frame);
 
-        fd_ctx->child_failed[child_index] = 1;
-out:
-        return;
+        return 0;
 }
 
 
-static void
-__mark_failed_children (int32_t *pending[], int child_count, 
-                        xlator_t *this, fd_t *fd, afr_transaction_type type)
+call_frame_t*
+afr_transaction_detach_fop_frame (call_frame_t *frame)
 {
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
-
-        int ret = 0;
-        int i   = 0;
-        int j   = 0;
+        afr_local_t *   local = NULL;
+        call_frame_t   *fop_frame = NULL;
 
-        ret = fd_ctx_get (fd, this, &ctx);
-
-        if (ret < 0)
-                goto out;
-
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-        for (i = 0; i < child_count; i++) {
-                j = afr_index_for_transaction_type (type);
+        local = frame->local;
 
-                if (fd_ctx->child_failed[i])
-                        pending[i][j] = 0;
+        LOCK (&frame->lock);
+        {
+                fop_frame = local->transaction.main_frame;
+                local->transaction.main_frame = NULL;
         }
-        
-out:
-        return;
+        UNLOCK (&frame->lock);
+
+        return fop_frame;
 }
 
 
 static void
-__mark_pre_op_done_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;
-
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
-        int ret = 0;
+        afr_local_t * local = NULL;
 
         local = frame->local;
 
-        ret = fd_ctx_get (local->fd, this, &ctx);
+        local->saved_lk_owner = frame->root->lk_owner;
+}
 
-        if (ret < 0)
-                goto out;
 
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+static void
+afr_restore_lk_owner (call_frame_t *frame)
+{
+        afr_local_t * local = NULL;
 
-        if ((local->op == GF_FOP_WRITE)
-            || (local->op == GF_FOP_FTRUNCATE)) {
-                fd_ctx->pre_op_done[child_index] = 1;
-        }
+        local = frame->local;
 
-out:
-        return;
+        frame->root->lk_owner = local->saved_lk_owner;
 }
 
-
-static void
-__mark_down_children (int32_t *pending[], int child_count, 
-                      unsigned char *child_up, 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);
+	local = frame->local;
+	priv = this->private;
 
-		if (!child_up[i])
-			pending[i][j] = 0;
-        }
+	for (i = 0; i < priv->child_count; i++) {
+		local->transaction.failed_subvols[i] = 0;
+	}
 }
 
 
-static void
-__mark_all_success (int32_t *pending[], int child_count,
-                    afr_transaction_type type)
+int
+afr_transaction_perform_fop (call_frame_t *frame, xlator_t *this)
 {
-	int i;
-        int j;
+        afr_local_t     *local = NULL;
+        fd_t            *fd   = NULL;
 
-	for (i = 0; i < child_count; i++) {
-                j = afr_index_for_transaction_type (type);
-		pending[i][j] = hton32 (-1);
-        }
+        local = frame->local;
+        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 0;
 }
 
 
 static int
-__is_first_write_on_fd (xlator_t *this, fd_t *fd)
+__changelog_enabled (afr_private_t *priv, afr_transaction_type type)
 {
-        int op_ret     = 0;
-        int _ret       = -1;
-        int i          = 0;
-
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
+        int ret = 0;
 
-        afr_private_t *priv = NULL;
+        switch (type) {
+        case AFR_DATA_TRANSACTION:
+                if (priv->data_change_log)
+                        ret = 1;
 
-        priv = this->private;
+                break;
 
-        LOCK (&fd->lock);
-        {
-                _ret = __fd_ctx_get (fd, this, &ctx);
-                
-                if (_ret < 0) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "could not get fd ctx on fd=%p",
-                                fd);
-                        goto out;
-                }
+        case AFR_METADATA_TRANSACTION:
+                if (priv->metadata_change_log)
+                        ret = 1;
 
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+                break;
 
-                op_ret = 1;
-                for (i = 0; i < priv->child_count; i++) {
-                        if (fd_ctx->pre_op_done[i] == 0)
-                                continue;
+        case AFR_ENTRY_TRANSACTION:
+        case AFR_ENTRY_RENAME_TRANSACTION:
+                if (priv->entry_change_log)
+                        ret = 1;
 
-                        op_ret = 0;
-                }
+                break;
         }
-out:
-        UNLOCK (&fd->lock);
 
-        return op_ret;
+        return ret;
 }
 
 
 static int
-__if_fd_pre_op_done (xlator_t *this, fd_t *fd, int child_index)
+__fop_changelog_needed (call_frame_t *frame, xlator_t *this)
 {
+        afr_private_t * priv  = NULL;
+        afr_local_t   * local = NULL;
         int op_ret = 0;
-        int _ret   = -1;
+        afr_transaction_type type = -1;
 
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
+        priv  = this->private;
+        local = frame->local;
+        type  = local->transaction.type;
 
-        LOCK (&fd->lock);
-        {
-                _ret = __fd_ctx_get (fd, this, &ctx);
+        if (__changelog_enabled (priv, type)) {
+                switch (local->op) {
 
-                if (_ret < 0) {
-                        goto out;
-                }
+                case GF_FOP_WRITE:
+                case GF_FOP_FTRUNCATE:
+                        op_ret = 1;
+                        break;
 
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+                case GF_FOP_FLUSH:
+                        op_ret = 0;
+                        break;
 
-                if (fd_ctx->pre_op_done[child_index]) {
+                default:
                         op_ret = 1;
                 }
-                fd_ctx->pre_op_done[child_index] = 0;
         }
-out:
-        UNLOCK (&fd->lock);
 
         return op_ret;
 }
 
 
-static int
-afr_pre_op_done_count (xlator_t *this, fd_t *fd, unsigned char *child_up)
+int
+afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int **pending)
 {
         int i = 0;
-        int count = 0;
+        int ret = 0;
+	int pending_zero[AFR_NUM_CHANGE_LOGS] = {0, };
 
-        int _ret = 0;
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
+        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;
 
-        afr_private_t *priv = NULL;
+                ret = dict_set_static_bin (xattr, priv->pending_key[i],
+					   pending[i],
+					   AFR_NUM_CHANGE_LOGS * sizeof (int));
+                /* 3 = data+metadata+entry */
 
-        priv = this->private;
+                if (ret)
+                        break;
+        }
 
-        LOCK (&fd->lock);
-        {
-                _ret = __fd_ctx_get (fd, this, &ctx);
+        return ret;
+}
 
-                if (_ret < 0) {
-                        goto out;
-                }
+int
+afr_lock_server_count (afr_private_t *priv, afr_transaction_type type)
+{
+        int ret = 0;
 
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+        switch (type) {
+        case AFR_DATA_TRANSACTION:
+                ret = priv->child_count;
+                break;
 
-                for (i = 0; i < priv->child_count; i++) {
-                        if (fd_ctx->pre_op_done[i] && child_up[i]) {
-                                count++;
-                        }
-                }
+        case AFR_METADATA_TRANSACTION:
+                ret = priv->child_count;
+                break;
+
+        case AFR_ENTRY_TRANSACTION:
+        case AFR_ENTRY_RENAME_TRANSACTION:
+                ret = priv->child_count;
+                break;
         }
-out:
-        UNLOCK (&fd->lock);
 
-        return count;
+        return ret;
 }
 
+/* {{{ pending */
 
-static int
-__changelog_enabled (afr_private_t *priv, afr_transaction_type type)
+
+int
+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:
-		if (priv->data_change_log)
-			ret = 1;
-		
-		break;
+	local = frame->local;
+	priv = this->private;
+        int_lock = &local->internal_lock;
 
-	case AFR_METADATA_TRANSACTION:
-		if (priv->metadata_change_log)
-			ret = 1;
+	if (local->transaction.resume_stub) {
+		call_resume (local->transaction.resume_stub);
+		local->transaction.resume_stub = NULL;
+	}
 
-		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);
+	}
 
-	case AFR_ENTRY_TRANSACTION:
-	case AFR_ENTRY_RENAME_TRANSACTION:
-		if (priv->entry_change_log)
-			ret = 1;
+	return 0;
+}
 
-		break;
-		
-	case AFR_FLUSH_TRANSACTION:
-		ret = 1;
-	}
 
-	return ret;
+afr_inodelk_t*
+afr_get_inodelk (afr_internal_lock_t *int_lock, char *dom)
+{
+        afr_inodelk_t *inodelk = NULL;
+        int           i = 0;
+
+        for (i = 0; int_lock->inodelk[i].domain; i++) {
+                inodelk = &int_lock->inodelk[i];
+                if (strcmp (dom, inodelk->domain) == 0)
+                        return inodelk;
+        }
+        return NULL;
 }
 
+unsigned char*
+afr_locked_nodes_get (afr_transaction_type type, afr_internal_lock_t *int_lock)
+{
+        unsigned char *locked_nodes = NULL;
+        afr_inodelk_t *inodelk = NULL;
+        switch (type) {
+        case AFR_DATA_TRANSACTION:
+        case AFR_METADATA_TRANSACTION:
+                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;
+}
 
-static int
-__changelog_needed_pre_op (call_frame_t *frame, xlator_t *this)
+
+int
+afr_changelog_call_count (afr_transaction_type type,
+			  unsigned char *pre_op_subvols,
+			  unsigned int child_count)
+{
+        int call_count = 0;
+
+	call_count = AFR_COUNT(pre_op_subvols, child_count);
+
+        if (type == AFR_ENTRY_RENAME_TRANSACTION)
+                call_count *= 2;
+
+        return call_count;
+}
+
+
+gf_boolean_t
+afr_txn_nothing_failed (call_frame_t *frame, xlator_t *this)
 {
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-	fd_t *          fd    = NULL;
+        afr_private_t *priv = NULL;
+        afr_local_t *local = NULL;
+        int i = 0;
+
+        local = frame->local;
+	priv = this->private;
+
+        for (i = 0; i < priv->child_count; i++) {
+                if (local->transaction.failed_subvols[i])
+                        return _gf_false;
+        }
+
+        return _gf_true;
+}
 
-	int op_ret   = 0;
 
-	priv  = this->private;
+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;
+
+	priv = this->private;
 	local = frame->local;
-	
-	if (__changelog_enabled (priv, local->transaction.type)) {
-		switch (local->op) {
-
-		case GF_FOP_WRITE:
-		case GF_FOP_FTRUNCATE:
-			/* 
-			   if it's a data transaction, we write the changelog
-			   only on the first write on an fd 
-			*/
-			
-			fd = local->fd;
-			if (!fd || __is_first_write_on_fd (this, fd))
-				op_ret = 1;
 
+	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;
 
-		case GF_FOP_FLUSH:
-			/* only do post-op on flush() */
-
-			op_ret = 0;
+		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;
+		}
 
-		default:
-			op_ret = 1;
+		if (!op_errno) {
+			op_errno = i_errno;
+		} else if (op_errno != i_errno) {
+			/* Mismatching op_errno's */
+			matching_errors = _gf_false;
+			break;
 		}
 	}
 
-	return op_ret;
+	if (matching_errors)
+		__mark_all_success (frame, this);
 }
 
 
-static int
-__changelog_needed_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  = NULL;
-	afr_local_t   * local = NULL;
+        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;
 
-	int op_ret = 0;
-	afr_transaction_type type = -1;
+        local = frame->local;
+	idx = afr_index_for_transaction_type (local->transaction.type);
 
-	priv  = this->private;
-	local = frame->local;
-	type  = local->transaction.type;
+        nothing_failed = afr_txn_nothing_failed (frame, this);
 
-	if (__changelog_enabled (priv, type)) {
-                switch (local->op) {
+	if (afr_changelog_pre_op_uninherit (frame, this))
+		need_undirty = _gf_false;
+	else
+		need_undirty = _gf_true;
 
-                case GF_FOP_WRITE:
-                case GF_FOP_FTRUNCATE:
-                        op_ret = 0;
-                        break;
+	if (nothing_failed && !need_undirty) {
+		afr_changelog_post_op_done (frame, this);
+                goto out;
+	}
 
-                case GF_FOP_FLUSH:
-                        op_ret = 1;
-                        break;
+	xattr = dict_new ();
+	if (!xattr) {
+		local->op_ret = -1;
+		local->op_errno = ENOMEM;
+		afr_changelog_post_op_done (frame, this);
+		goto out;
+	}
 
-                default:
-                        op_ret = 1;
-                }
-        }
+	if (need_undirty) {
+		local->dirty[idx] = hton32(-1);
 
-	return op_ret;
-}
+		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;
+		}
 
+	}
 
-static int
-afr_set_pending_dict (afr_private_t *priv, dict_t *xattr, int32_t **pending)
-{
-        int i;
-        int ret = 0;
+	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++) {
-                ret = dict_set_static_bin (xattr, priv->pending_key[i],
-                                           pending[i], 3 * sizeof (int32_t));
-                /* 3 = data+metadata+entry */
-                
-                if (ret < 0)
-                        goto out;
-        }
+	}
 
+	afr_changelog_do (frame, this, xattr, afr_changelog_post_op_done);
 out:
-        return ret;
+	if (xattr)
+                dict_unref (xattr);
+
+        return 0;
 }
 
 
-int
-afr_lock_server_count (afr_private_t *priv, afr_transaction_type type)
+gf_boolean_t
+afr_changelog_pre_op_uninherit (call_frame_t *frame, xlator_t *this)
 {
-	int ret = 0;
+	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;
+	priv = this->private;
+	fd = local->fd;
+
+	type = afr_index_for_transaction_type (local->transaction.type);
+	if (type != AFR_DATA_TRANSACTION)
+		return !local->transaction.dirtied;
+
+	if (!fd)
+		return !local->transaction.dirtied;
 
-	switch (type) {
-	case AFR_FLUSH_TRANSACTION:
-	case AFR_DATA_TRANSACTION:
-		ret = priv->data_lock_server_count;
-		break;
+	fd_ctx = afr_fd_ctx_get (fd, this);
+	if (!fd_ctx)
+		return _gf_false;
 
-	case AFR_METADATA_TRANSACTION:
-		ret = priv->metadata_lock_server_count;
-		break;
+	if (local->transaction.no_uninherit)
+		return _gf_false;
 
-	case AFR_ENTRY_TRANSACTION:
-	case AFR_ENTRY_RENAME_TRANSACTION:
-		ret = priv->entry_lock_server_count;
-		break;
+	/* This function must be idempotent. So check if we
+	   were called before and return the same answer again.
+
+	   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;
+
+	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;
+			}
+		}
+
+		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 (!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);
+
+	local->transaction.uninherit_done = _gf_true;
+	local->transaction.uninherit_value = ret;
 
 	return ret;
 }
 
 
-/* {{{ unlock */
-
-static int
-afr_transaction_locked_nodes_count (afr_local_t *local, int child_count)
+gf_boolean_t
+afr_changelog_pre_op_inherit (call_frame_t *frame, xlator_t *this)
 {
-        int i;
-        int call_count = 0;
+	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;
 
-        for (i = 0; i < child_count; i++) {
-                if (local->transaction.locked_nodes[i] & LOCKED_YES)
-                        call_count++;
+	local = frame->local;
+	priv = this->private;
+	fd = local->fd;
 
-                if ((local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION)
-                    && (local->transaction.locked_nodes[i] & LOCKED_LOWER)) {
-                        call_count++;
-                }
-        }
+	if (local->transaction.type != AFR_DATA_TRANSACTION)
+		return _gf_false;
 
-        return call_count;
-}
+	type = afr_index_for_transaction_type (local->transaction.type);
 
+	if (!fd)
+		return _gf_false;
 
-static loc_t *
-lower_path (loc_t *l1, const char *b1, loc_t *l2, const char *b2)
-{
-	int ret = 0;
+	fd_ctx = afr_fd_ctx_get (fd, this);
+	if (!fd_ctx)
+		return _gf_false;
 
-	ret = strcmp (l1->path, l2->path);
-	
-	if (ret == 0) 
-		ret = strcmp (b1, b2);
+	LOCK(&fd->lock);
+	{
+		if (!fd_ctx->on_disk[type]) {
+			/* nothing to inherit yet */
+			ret = _gf_false;
+			goto unlock;
+		}
 
-	if (ret <= 0)
-		return l1;
-	else
-		return l2;
+		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;
+			}
+		}
+
+		fd_ctx->inherited[type]++;
+
+		ret = _gf_true;
+
+		local->transaction.inherited = _gf_true;
+	}
+unlock:
+	UNLOCK(&fd->lock);
+
+	return ret;
 }
 
 
-int32_t
-afr_unlock_common_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int32_t op_ret, int32_t op_errno)
+gf_boolean_t
+afr_changelog_pre_op_update (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *local;
-	int call_count = 0;
+	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;
 
-	LOCK (&frame->lock);
+	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);
 	{
-		call_count = --local->call_count;
-	}
-	UNLOCK (&frame->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]++;
 
-	if (call_count == 0) {
-		local->transaction.done (frame, this);
+		ret = _gf_true;
 	}
-	
-	return 0;
+unlock:
+	UNLOCK(&fd->lock);
+
+	return ret;
 }
 
 
 int
-afr_unlock (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)
 {
-	struct flock flock;			
+        afr_local_t *local = NULL;
+        int call_count = -1;
 
-	int i = 0;				
-	int call_count = 0;		     
+        local = frame->local;
 
-	afr_local_t *local = NULL;
-	afr_private_t * priv = this->private;
+	if (op_ret == -1)
+		afr_transaction_fop_failed (frame, this, (long) cookie);
 
-        loc_t * lower  = NULL;
-        loc_t * higher = NULL;
+	call_count = afr_frame_return (frame);
 
-        const char *lower_name  = NULL;
-        const char *higher_name = NULL;
+        if (call_count == 0)
+		local->transaction.changelog_resume (frame, this);
 
-	local = frame->local;
+        return 0;
+}
 
-        /*
-          pid has been restored to saved_pid in the fop,
-          so set it back to frame->root 
-        */
 
-        frame->root->pid = (long) frame->root;
+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 = frame->local;
+	priv = this->private;
 
-	call_count = afr_transaction_locked_nodes_count (local,
-                                                         priv->child_count);
+        call_count = afr_changelog_call_count (local->transaction.type,
+					       local->transaction.pre_op,
+					       priv->child_count);
 
 	if (call_count == 0) {
-		local->transaction.done (frame, this);
+		changelog_resume (frame, this);
 		return 0;
 	}
 
-	local->call_count = call_count;		
+	local->call_count = call_count;
+
+	local->transaction.changelog_resume = changelog_resume;
 
-	for (i = 0; i < priv->child_count; i++) {				
-		flock.l_start = local->transaction.start;			
-		flock.l_len   = local->transaction.len;
-		flock.l_type  = F_UNLCK;			
+        for (i = 0; i < priv->child_count; i++) {
+                if (!local->transaction.pre_op[i])
+                        continue;
 
                 switch (local->transaction.type) {
                 case AFR_DATA_TRANSACTION:
                 case AFR_METADATA_TRANSACTION:
-                case AFR_FLUSH_TRANSACTION:
-
-                        if (local->transaction.locked_nodes[i] & LOCKED_YES) {
-                                if (local->fd) {
-                                        STACK_WIND (frame, afr_unlock_common_cbk,	
-                                                    priv->children[i], 
-                                                    priv->children[i]->fops->finodelk, 
-                                                    this->name, local->fd, 
-                                                    F_SETLK, &flock); 
-                                } else {
-                                        STACK_WIND (frame, afr_unlock_common_cbk,	
-                                                    priv->children[i], 
-                                                    priv->children[i]->fops->inodelk, 
-                                                    this->name, &local->loc,
-                                                    F_SETLK, &flock); 
-                                }
-
-                                call_count--;
+                        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;
-
+			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);
-
-                        if (local->transaction.locked_nodes[i] & LOCKED_LOWER) {
-                                STACK_WIND (frame, afr_unlock_common_cbk,	
-                                            priv->children[i], 
-                                            priv->children[i]->fops->entrylk, 
-                                            this->name,
-                                            lower, lower_name,
-                                            ENTRYLK_UNLOCK, ENTRYLK_WRLCK);
-
-                                call_count--;
-                        }
 
-                        if (local->transaction.locked_nodes[i] & LOCKED_YES) {
-                                STACK_WIND (frame, afr_unlock_common_cbk,	
-                                            priv->children[i], 
-                                            priv->children[i]->fops->entrylk, 
-                                            this->name,
-                                            higher, higher_name,
-                                            ENTRYLK_UNLOCK, ENTRYLK_WRLCK);
+			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--;
 
-                                call_count--;
-                        }
-
-                        break;
+                /* fall through */
 
                 case AFR_ENTRY_TRANSACTION:
-                        if (local->transaction.locked_nodes[i] & LOCKED_YES) {
-                                if (local->fd) {
-                                        STACK_WIND (frame, afr_unlock_common_cbk,	
-                                                    priv->children[i], 
-                                                    priv->children[i]->fops->fentrylk, 
-                                                    this->name, local->fd, 
-                                                    local->transaction.basename,
-                                                    ENTRYLK_UNLOCK, ENTRYLK_WRLCK);
-                                } else {
-                                        STACK_WIND (frame, afr_unlock_common_cbk,	
-                                                    priv->children[i], 
-                                                    priv->children[i]->fops->entrylk, 
-                                                    this->name,
-                                                    &local->transaction.parent_loc, 
-                                                    local->transaction.basename,
-                                                    ENTRYLK_UNLOCK, ENTRYLK_WRLCK);
-
-                                }
-
-                                call_count--;
-                        }
-
-                        break;
-                }
+                        if (local->fd)
+                                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_cbk,
+						   (void *) (long) i,
+						   priv->children[i],
+						   priv->children[i]->fops->xattrop,
+						   &local->transaction.parent_loc,
+						   GF_XATTROP_ADD_ARRAY, xattr,
+						   NULL);
+			break;
+		}
 
-                if (!call_count)
+                if (!--call_count)
                         break;
-	}
+        }
 
 	return 0;
 }
 
-/* }}} */
 
+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;
+
+        local = frame->local;
+        int_lock = &local->internal_lock;
+	idx = afr_index_for_transaction_type (local->transaction.type);
 
-/* {{{ pending */
+        locked_nodes = afr_locked_nodes_get (local->transaction.type, int_lock);
 
-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_private_t * priv  = NULL;
-	afr_local_t *   local = NULL;
+	pending_subvols = alloca0 (priv->child_count);
 
-	int call_count = -1;
+	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;
+		}
+	}
 
-        int (*post_post_op) (call_frame_t *, xlator_t *);
+	/* TBD: quorum check w/ call_count */
 
-	priv  = this->private;
-	local = frame->local;
+        if (call_count == 0) {
+		op_errno = ENOTCONN;
+		goto err;
+	}
 
-	LOCK (&frame->lock);
-	{
-		call_count = --local->call_count;
+	xdata_req = dict_new();
+	if (!xdata_req) {
+		op_errno = ENOMEM;
+		goto err;
 	}
-	UNLOCK (&frame->lock);
 
-	if (call_count == 0) {
-                if (local->transaction.post_post_op) {
-                        post_post_op = local->transaction.post_post_op;
+	pre_nop = _gf_true;
 
-                        if (afr_lock_server_count (priv, local->transaction.type) == 0) {
-                                local->transaction.post_post_op = local->transaction.done;
-                        } else {
-                                local->transaction.post_post_op = afr_unlock;
-                        }
+	if (afr_changelog_pre_op_inherit (frame, this))
+		goto next;
 
-                        post_post_op (frame, this);
-                } else {
-                        if (afr_lock_server_count (priv, local->transaction.type) == 0) {
-                                local->transaction.done (frame, this);
-                        } else {
-                                afr_unlock (frame, this);
-                        }
-                }
+	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;
 	}
 
-	return 0;	
+	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;
 }
 
 
-int 
-afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
+int
+afr_post_blocking_inodelk_cbk (call_frame_t *frame, xlator_t *this)
 {
-	afr_private_t * priv = this->private;
+        afr_internal_lock_t *int_lock = NULL;
+        afr_local_t         *local    = NULL;
 
-	int ret        = 0;
-	int i          = 0;				
-	int call_count = 0;
-	
-	afr_local_t *  local = NULL;	
-	dict_t        **xattr = NULL;
+        local    = frame->local;
+        int_lock = &local->internal_lock;
 
-	local = frame->local;
+        if (int_lock->lock_op_ret < 0) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "Blocking inodelks failed.");
+                local->transaction.done (frame, this);
+        } else {
 
-	__mark_down_children (local->pending, priv->child_count, 
-                              local->child_up, local->transaction.type);
-        
-        if (local->op == GF_FOP_FLUSH) {
-                __mark_failed_children (local->pending, priv->child_count,
-                                        this, local->fd,
-                                        local->transaction.type);
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Blocking inodelks done. Proceeding to FOP");
+                afr_internal_lock_finish (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]);
-        }
+        return 0;
+}
+
 
-        if (local->op == GF_FOP_FLUSH) {
-                call_count = afr_pre_op_done_count (this, local->fd, local->child_up);
+int
+afr_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;
+
+        /* Initiate blocking locks if non-blocking has failed */
+        if (int_lock->lock_op_ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Non blocking inodelks failed. Proceeding to blocking");
+                int_lock->lock_cbk = afr_post_blocking_inodelk_cbk;
+                afr_blocking_lock (frame, this);
         } else {
-                call_count = afr_up_children_count (priv->child_count, local->child_up); 
 
-                if (local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
-                        call_count *= 2;
-                }
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Non blocking inodelks done. Proceeding to FOP");
+                afr_internal_lock_finish (frame, this);
         }
 
-	local->call_count = call_count;		
+        return 0;
+}
 
-	if (call_count == 0) {
-		/* no child is up */
-                for (i = 0; i < priv->child_count; i++) {
-                        dict_unref (xattr[i]);
-                }
-                
-		afr_unlock (frame, this);
-		return 0;
-	}
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-                        ret = afr_set_pending_dict (priv, xattr[i], 
-                                                    local->pending);
-
-			if (ret < 0)
-				gf_log (this->name, GF_LOG_DEBUG,
-					"failed to set pending entry");
-
-
-			switch (local->transaction.type) {
-			case AFR_DATA_TRANSACTION:
-			case AFR_METADATA_TRANSACTION:
-			{
-				if (local->fd)
-					STACK_WIND (frame, afr_changelog_post_op_cbk,
-						    priv->children[i],
-						    priv->children[i]->fops->fxattrop,
-						    local->fd,
-						    GF_XATTROP_ADD_ARRAY, xattr[i]);
-				else
-					STACK_WIND (frame, afr_changelog_post_op_cbk,
-						    priv->children[i],
-						    priv->children[i]->fops->xattrop,
-						    &local->loc,
-						    GF_XATTROP_ADD_ARRAY, xattr[i]);
-                                call_count--;
-			}
-			break;
+int
+afr_post_blocking_entrylk_cbk (call_frame_t *frame, xlator_t *this)
+{
+        afr_internal_lock_t *int_lock = NULL;
+        afr_local_t         *local    = NULL;
 
-			case AFR_FLUSH_TRANSACTION:
-			{
-				if (__if_fd_pre_op_done (this, local->fd, i)) {
-					STACK_WIND (frame, afr_changelog_post_op_cbk,
-						    priv->children[i], 
-						    priv->children[i]->fops->fxattrop,
-						    local->fd,
-						    GF_XATTROP_ADD_ARRAY, xattr[i]);
-                                        call_count--;
-                                }
-			}
-			break;
+        local    = frame->local;
+        int_lock = &local->internal_lock;
 
-			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--;
-			}
+        if (int_lock->lock_op_ret < 0) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "Blocking entrylks failed.");
+                local->transaction.done (frame, this);
+        } else {
 
-			/* 
-			   set it again because previous stack_wind
-			   might have already returned (think of case
-			   where subvolume is posix) and would have
-			   used the dict as placeholder for return
-			   value
-			*/
-                        
-			ret = afr_set_pending_dict (priv, xattr[i], 
-                                                    local->pending);
-
-			if (ret < 0)
-				gf_log (this->name, GF_LOG_DEBUG,
-					"failed to set pending entry");
-
-			/* fall through */
-
-			case AFR_ENTRY_TRANSACTION:
-			{
-				if (local->fd)
-					STACK_WIND (frame, afr_changelog_post_op_cbk,
-						    priv->children[i], 
-						    priv->children[i]->fops->fxattrop,
-						    local->fd, 
-						    GF_XATTROP_ADD_ARRAY, xattr[i]);
-				else 
-					STACK_WIND (frame, afr_changelog_post_op_cbk,
-						    priv->children[i], 
-						    priv->children[i]->fops->xattrop,
-						    &local->transaction.parent_loc, 
-						    GF_XATTROP_ADD_ARRAY, xattr[i]);
-                                call_count--;
-			}
-			break;
-			}
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Blocking entrylks done. Proceeding to FOP");
+                afr_internal_lock_finish (frame, this);
+        }
 
-			if (!call_count)
-				break;
-		}
-	}
+        return 0;
+}
 
-        for (i = 0; i < priv->child_count; i++) {
-                dict_unref (xattr[i]);
+
+int
+afr_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;
+
+        /* Initiate blocking locks if non-blocking has failed */
+        if (int_lock->lock_op_ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Non blocking entrylks failed. Proceeding to blocking");
+                int_lock->lock_cbk = afr_post_blocking_entrylk_cbk;
+                afr_blocking_lock (frame, this);
+        } else {
+
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Non blocking entrylks done. Proceeding to FOP");
+                afr_internal_lock_finish (frame, this);
         }
 
-	return 0;
+        return 0;
 }
 
 
-int32_t
-afr_changelog_pre_op_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			      int32_t op_ret, int32_t op_errno, dict_t *xattr)
+int
+afr_post_blocking_rename_cbk (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = this->private;
-	loc_t       *   loc   = NULL;
+        afr_internal_lock_t *int_lock = NULL;
+        afr_local_t         *local    = NULL;
 
-	int call_count  = -1;
-	int child_index = (long) cookie;
+        local    = frame->local;
+        int_lock = &local->internal_lock;
 
-	local = frame->local;
-	loc   = &local->loc;
+        if (int_lock->lock_op_ret < 0) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "Blocking entrylks failed.");
+                local->transaction.done (frame, this);
+        } else {
 
-	LOCK (&frame->lock);
-	{
-                if (op_ret == 0) {
-                        __mark_pre_op_done_on_fd (frame, this, child_index);
-                }
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Blocking entrylks done. Proceeding to FOP");
+                afr_internal_lock_finish (frame, this);
+        }
+        return 0;
+}
 
-		if (op_ret == -1) {
-			local->child_up[child_index] = 0;
-			
-			if (op_errno == ENOTSUP) {
-				gf_log (this->name, GF_LOG_ERROR,
-					"xattrop not supported by %s",
-					priv->children[child_index]->name);
-				local->op_ret = -1;
-                                
-			} else if (!child_went_down (op_ret, op_errno)) {
-				gf_log (this->name, GF_LOG_ERROR,
-					"xattrop failed on child %s: %s",
-					priv->children[child_index]->name, 
-					strerror (op_errno));
-			}
-			local->op_errno = op_errno;
-		}
 
-		call_count = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
+int
+afr_post_lower_unlock_cbk (call_frame_t *frame, xlator_t *this)
+{
+        afr_internal_lock_t *int_lock = NULL;
+        afr_local_t         *local    = NULL;
 
-	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);
+        local    = frame->local;
+        int_lock = &local->internal_lock;
 
-                        afr_pid_restore (frame);
+        GF_ASSERT (!int_lock->higher_locked);
 
-			local->transaction.fop (frame, this);
-		}
-	}
+        int_lock->lock_cbk = afr_post_blocking_rename_cbk;
+        afr_blocking_lock (frame, this);
 
-	return 0;	
+        return 0;
 }
 
 
-int 
-afr_changelog_pre_op (call_frame_t *frame, xlator_t *this)
+int
+afr_set_transaction_flock (afr_local_t *local)
 {
-	afr_private_t * priv = this->private;
+        afr_internal_lock_t *int_lock = NULL;
+        afr_inodelk_t       *inodelk  = NULL;
 
-	int i = 0;				
-	int ret = 0;
-	int call_count = 0;		     
-	dict_t **xattr = NULL;
+        int_lock = &local->internal_lock;
+        inodelk = afr_get_inodelk (int_lock, int_lock->domain);
 
-	afr_local_t *local = NULL;
+        inodelk->flock.l_len   = local->transaction.len;
+        inodelk->flock.l_start = local->transaction.start;
+        inodelk->flock.l_type  = F_WRLCK;
 
-	local = frame->local;
-        
-        xattr = alloca (priv->child_count * sizeof (*xattr));
-        memset (xattr, 0, (priv->child_count * sizeof (*xattr)));
+        return 0;
+}
 
-	for (i = 0; i < priv->child_count; i++) {
-                xattr[i] = get_new_dict ();
-                dict_ref (xattr[i]);
-        }
+int
+afr_lock_rec (call_frame_t *frame, xlator_t *this)
+{
+        afr_internal_lock_t *int_lock = NULL;
+        afr_local_t         *local    = NULL;
 
-	call_count = afr_up_children_count (priv->child_count, 
-					    local->child_up); 
+        local    = frame->local;
+        int_lock = &local->internal_lock;
 
-	if (local->transaction.type == AFR_ENTRY_RENAME_TRANSACTION) {
-		call_count *= 2;
-	}
+        int_lock->transaction_lk_type = AFR_TRANSACTION_LK;
+        int_lock->domain = this->name;
 
-	if (call_count == 0) {
-		/* no child is up */
-                for (i = 0; i < priv->child_count; i++) {
-                        dict_unref (xattr[i]);
-                }
-                
-		afr_unlock (frame, this);
-		return 0;
-	}
+        switch (local->transaction.type) {
+        case AFR_DATA_TRANSACTION:
+        case AFR_METADATA_TRANSACTION:
+                afr_set_transaction_flock (local);
 
-	local->call_count = call_count;		
+                int_lock->lock_cbk = afr_post_nonblocking_inodelk_cbk;
 
-	__mark_all_pending (local->pending, priv->child_count,
-                            local->transaction.type);
+                afr_nonblocking_inodelk (frame, this);
+                break;
 
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			ret = afr_set_pending_dict (priv, xattr[i], 
-                                                    local->pending);
-
-			if (ret < 0)
-				gf_log (this->name, GF_LOG_DEBUG,
-					"failed to set pending entry");
-
-
-			switch (local->transaction.type) {
-			case AFR_DATA_TRANSACTION:
-			case AFR_METADATA_TRANSACTION:
-			case AFR_FLUSH_TRANSACTION:
-			{
-				if (local->fd)
-					STACK_WIND_COOKIE (frame, 
-							   afr_changelog_pre_op_cbk,
-							   (void *) (long) i,
-							   priv->children[i], 
-							   priv->children[i]->fops->fxattrop,
-							   local->fd,
-							   GF_XATTROP_ADD_ARRAY, xattr[i]);
-				else
-					STACK_WIND_COOKIE (frame, 
-							   afr_changelog_pre_op_cbk,
-							   (void *) (long) i,
-							   priv->children[i], 
-							   priv->children[i]->fops->xattrop,
-							   &(local->loc), 
-							   GF_XATTROP_ADD_ARRAY, xattr[i]);
-			}
-			break;
-				
-			case AFR_ENTRY_RENAME_TRANSACTION: 
-			{
-				STACK_WIND_COOKIE (frame, 
-						   afr_changelog_pre_op_cbk,
-						   (void *) (long) i,
-						   priv->children[i], 
-						   priv->children[i]->fops->xattrop,
-						   &local->transaction.new_parent_loc, 
-						   GF_XATTROP_ADD_ARRAY, xattr[i]);
+        case AFR_ENTRY_RENAME_TRANSACTION:
 
-				call_count--;
-			}
+                int_lock->lock_cbk = afr_post_nonblocking_entrylk_cbk;
+                afr_nonblocking_entrylk (frame, this);
+                break;
 
+        case AFR_ENTRY_TRANSACTION:
+                int_lock->lk_basename = local->transaction.basename;
+                if (&local->transaction.parent_loc)
+                        int_lock->lk_loc = &local->transaction.parent_loc;
+                else
+                        GF_ASSERT (local->fd);
 
-			/* 
-			   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_lock->lock_cbk = afr_post_nonblocking_entrylk_cbk;
+                afr_nonblocking_entrylk (frame, this);
+                break;
+        }
 
-			ret = afr_set_pending_dict (priv, xattr[i], 
-                                                    local->pending);
-
-			if (ret < 0)
-				gf_log (this->name, GF_LOG_DEBUG,
-					"failed to set pending entry");
-
-			/* fall through */
-				
-			case AFR_ENTRY_TRANSACTION:
-			{
-				if (local->fd)
-					STACK_WIND_COOKIE (frame, 
-							   afr_changelog_pre_op_cbk,
-							   (void *) (long) i,
-							   priv->children[i], 
-							   priv->children[i]->fops->fxattrop,
-							   local->fd, 
-							   GF_XATTROP_ADD_ARRAY, xattr[i]);
-				else
-					STACK_WIND_COOKIE (frame, 
-							   afr_changelog_pre_op_cbk,
-							   (void *) (long) i,
-							   priv->children[i], 
-							   priv->children[i]->fops->xattrop,
-							   &local->transaction.parent_loc, 
-							   GF_XATTROP_ADD_ARRAY, xattr[i]);
-			}
+        return 0;
+}
 
-			break;
-			}
 
-			if (!--call_count)
-				break;
-		}
-	}
-        
-        for (i = 0; i < priv->child_count; i++) {
-                dict_unref (xattr[i]);
-        }
-        
-	return 0;
+int
+afr_lock (call_frame_t *frame, xlator_t *this)
+{
+        afr_set_lock_number (frame, this);
+
+        return afr_lock_rec (frame, this);
 }
 
+
 /* }}} */
 
-/* {{{ lock */
+int
+afr_internal_lock_finish (call_frame_t *frame, xlator_t *this)
+{
+        if (__fop_changelog_needed (frame, this)) {
+                afr_changelog_pre_op (frame, this);
+        } else {
+                afr_transaction_perform_fop (frame, this);
+        }
+
+        return 0;
+}
 
-static
-int afr_lock_rec (call_frame_t *frame, xlator_t *this, int child_index);
 
-int32_t
-afr_lock_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	      int32_t op_ret, int32_t op_errno)
+void
+afr_set_delayed_post_op (call_frame_t *frame, xlator_t *this)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv = NULL;
-	int done = 0;
-	int child_index = (long) cookie;
+        afr_local_t    *local = NULL;
+        afr_private_t  *priv = NULL;
 
-	local = frame->local;
-	priv  = this->private;
+        /* call this function from any of the related optimizations
+           which benefit from delaying post op are enabled, namely:
 
-	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;
-				done = 1;
-			}
+           - changelog piggybacking
+           - eager locking
+        */
 
-			local->child_up[child_index] = 0;
-			local->op_errno = op_errno;
-		}
-	}
-	UNLOCK (&frame->lock);
-	
-        if ((op_ret == -1) &&
-            (op_errno == ENOSYS)) {
-                afr_unlock (frame, this);
-        } else {
-                if (op_ret == 0) {
-                        local->transaction.locked_nodes[child_index]
-                                |= LOCKED_YES;
-                        local->transaction.lock_count++;
-                }
-                afr_lock_rec (frame, this, child_index + 1);
-        }
+        priv = this->private;
+        if (!priv)
+                return;
 
-	return 0;
-}
+        if (!priv->post_op_delay_secs)
+                return;
 
+        local = frame->local;
+        if (!local->transaction.eager_lock_on)
+                return;
+
+        if (!local)
+                return;
+
+        if (!local->fd)
+                return;
+
+        if (local->op == GF_FOP_WRITE)
+                local->delayed_post_op = _gf_true;
+}
 
-int32_t
-afr_lock_lower_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno)
+gf_boolean_t
+afr_are_multiple_fds_opened (fd_t *fd, xlator_t *this)
 {
-        afr_private_t *priv  = NULL;
-        afr_local_t   *local = NULL;
+        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
+         */
 
-        int child_index = (long) cookie;
+        fd_ctx = afr_fd_ctx_get (fd, this);
+        if (!fd_ctx)
+                return _gf_true;
 
-	loc_t * lower  = NULL;
-	loc_t * higher = NULL;
+        if (fd_ctx->open_fd_count > 1)
+                return _gf_true;
 
-	const char *lower_name  = NULL;
-	const char *higher_name = NULL;
+        return _gf_false;
+}
 
-        priv  = this->private;
-        local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			if (op_errno == ENOSYS) {
-				/* return ENOTSUP */
+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;
 
-				gf_log (this->name, GF_LOG_ERROR,
-					"subvolume does not support locking. "
-					"please load features/posix-locks xlator on server");
+        local = frame->local;
+        if (!local)
+                goto out;
 
-				local->op_ret   = op_ret;
-			}
+        if (!local->delayed_post_op)
+                goto out;
 
-			local->child_up[child_index] = 0;
-			local->op_errno = op_errno;
-		}
-	}
-	UNLOCK (&frame->lock);
+        //Mark pending changelog ASAP
+        if (!afr_txn_nothing_failed (frame, this))
+                goto out;
 
-        if (op_ret != 0) {
-                afr_unlock (frame, this);
+        if (local->fd && afr_are_multiple_fds_opened (local->fd, this))
                 goto out;
-        } else {
-                local->transaction.locked_nodes[child_index] |= LOCKED_LOWER;
-        }
 
-        /* The lower path has been locked. Now lock the higher path */
+        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;
 
-        lower = lower_path (&local->transaction.parent_loc,
-                            local->transaction.basename,
-                            &local->transaction.new_parent_loc,
-                            local->transaction.new_basename);
+        fd = data;
 
-        lower_name = (lower == &local->transaction.parent_loc ?
-                      local->transaction.basename :
-                      local->transaction.new_basename);
+        afr_delayed_changelog_wake_up (THIS, fd);
+}
 
-        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);
+/* SET operation */
+int
+afr_fd_report_unstable_write (xlator_t *this, fd_t *fd)
+{
+        afr_fd_ctx_t *fdctx = NULL;
 
-        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);
+        fdctx = afr_fd_ctx_get (fd, this);
+
+        LOCK(&fd->lock);
+        {
+                fdctx->witnessed_unstable_write = _gf_true;
+        }
+        UNLOCK(&fd->lock);
 
-out:
         return 0;
 }
 
-
-static
-int afr_lock_rec (call_frame_t *frame, xlator_t *this, int child_index)
+/* TEST and CLEAR operation */
+gf_boolean_t
+afr_fd_has_witnessed_unstable_write (xlator_t *this, fd_t *fd)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
+        afr_fd_ctx_t *fdctx = NULL;
+        gf_boolean_t witness = _gf_false;
 
-        uint64_t      ctx;
-        afr_fd_ctx_t *fd_ctx;
+        fdctx = afr_fd_ctx_get (fd, this);
+        if (!fdctx)
+                return _gf_true;
 
-	struct flock flock;
+        LOCK(&fd->lock);
+        {
+                if (fdctx->witnessed_unstable_write) {
+                        witness = _gf_true;
+                        fdctx->witnessed_unstable_write = _gf_false;
+                }
+        }
+        UNLOCK (&fd->lock);
 
-        int ret = 0;
+        return witness;
+}
 
-	loc_t * lower  = NULL;
-	loc_t * higher = NULL;
 
-	const char *lower_name  = NULL;
-	const char *higher_name = NULL;
+int
+afr_changelog_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int op_ret, int op_errno, struct iatt *pre,
+                         struct iatt *post, dict_t *xdata)
+{
+        afr_private_t *priv = NULL;
+        int child_index = (long) cookie;
+        int call_count = -1;
+        afr_local_t *local = NULL;
 
-	local = frame->local;
-	priv  = this->private;
+        priv = this->private;
+        local = frame->local;
 
-	flock.l_start = local->transaction.start;
-	flock.l_len   = local->transaction.len;
-	flock.l_type  = F_WRLCK;
+        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);
+	}
 
-        if (local->fd) {
-                ret = fd_ctx_get (local->fd, this, &ctx);
+	call_count = afr_frame_return (frame);
 
-                if (ret < 0) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "unable to get fd ctx for fd=%p",
-                                local->fd);
+	if (call_count == 0)
+		afr_changelog_post_op_now (frame, this);
 
-                        local->op_ret   = -1;
-                        local->op_errno = EINVAL;
+	return 0;
+}
 
-                        afr_unlock (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;
 
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
+        local = frame->local;
+        priv = this->private;
 
-                /* skip over children that or down
-                   or don't have the fd open */
+        call_count = AFR_COUNT (local->transaction.pre_op, priv->child_count);
 
-                while ((child_index < priv->child_count)
-                       && (!local->child_up[child_index]
-                           || !fd_ctx->opened_on[child_index]))
+        if (!call_count) {
+                /* will go straight to unlock */
+                afr_changelog_post_op_now (frame, this);
+                return 0;
+        }
 
-                        child_index++;
-        } else {
-                /* skip over children that are down */
-                while ((child_index < priv->child_count)
-                       && !local->child_up[child_index])
-                        child_index++;
+        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 ((child_index == priv->child_count) &&
-	    local->transaction.lock_count == 0) {
+        if (xdata)
+                dict_unref (xdata);
 
-		gf_log (this->name, GF_LOG_DEBUG,
-			"unable to lock on even one child");
+        return 0;
+}
 
-		local->op_ret   = -1;
-		local->op_errno = EAGAIN;
 
-                afr_unlock (frame, this);
-		
-		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 ((child_index == priv->child_count) 
-	    || (local->transaction.lock_count == 
-		afr_lock_server_count (priv, local->transaction.type))) {
+        if (!local->fd || local->transaction.type != AFR_DATA_TRANSACTION) {
+                afr_changelog_post_op_now (frame, this);
+                return 0;
+        }
 
-		/* we're done locking */
+        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;
+        }
 
-		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);
+        /* 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;
+        }
 
-                        afr_pid_restore (frame);
+        /* 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);
+        }
 
-			local->transaction.fop (frame, this);
-		}
+        return 0;
+}
 
-		return 0;
-	}
 
-	switch (local->transaction.type) {
-	case AFR_DATA_TRANSACTION:		
-	case AFR_METADATA_TRANSACTION:
-	case AFR_FLUSH_TRANSACTION:
-
-		if (local->fd) {
-			STACK_WIND_COOKIE (frame, afr_lock_cbk,
-					   (void *) (long) child_index,
-					   priv->children[child_index], 
-					   priv->children[child_index]->fops->finodelk,
-					   this->name, local->fd, 
-                                           F_SETLKW, &flock);
-			
-		} else {
-			STACK_WIND_COOKIE (frame, afr_lock_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:
+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);
 	{
-		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);
-
-		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;
+		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;
 	}
-		
-	case AFR_ENTRY_TRANSACTION:
-		if (local->fd) {
-			STACK_WIND_COOKIE (frame, afr_lock_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 {
-			STACK_WIND_COOKIE (frame, afr_lock_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);
-		}
+unlock:
+	pthread_mutex_unlock (&fd_ctx->delay_lock);
 
-		break;
+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);
 	}
-
-	return 0;
 }
 
 
-int32_t afr_lock (call_frame_t *frame, xlator_t *this)
+void
+afr_changelog_post_op (call_frame_t *frame, xlator_t *this)
 {
-        afr_pid_save (frame);
+        afr_local_t  *local = NULL;
 
-        frame->root->pid = (long) frame->root;
+        local = frame->local;
 
-	return afr_lock_rec (frame, this, 0);
+        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);
 }
 
 
-/* }}} */
 
-int32_t
+/* 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;
-	afr_private_t * priv  = NULL;
+        afr_local_t         *local    = NULL;
 
-	local = frame->local;
-	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 {
-			afr_unlock (frame, this);
-		}
-	}
+        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);
+        }
 
-	return 0;
+        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;
 }
 
 
@@ -1410,54 +1582,141 @@ afr_transaction_resume (call_frame_t *frame, xlator_t *this)
  */
 
 void
-afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this, int child_index)
+afr_transaction_fop_failed (call_frame_t *frame, xlator_t *this,
+                            int child_index)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
+        afr_local_t *   local = NULL;
 
-	local = frame->local;
-	priv  = this->private;
+        local = frame->local;
 
-        switch (local->op) {
-        case GF_FOP_WRITE:
-                __mark_fop_failed_on_fd (local->fd, this, child_index);
-                break;
-        default:
-                __mark_child_dead (local->pending, priv->child_count,
-                                   child_index, local->transaction.type);
-                break;
+	local->transaction.failed_subvols[child_index] = 1;
+}
+
+
+
+        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);
 }
 
 
-int32_t
+int
 afr_transaction (call_frame_t *frame, xlator_t *this, afr_transaction_type type)
 {
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
+        afr_local_t *   local = NULL;
+        afr_private_t * priv  = NULL;
+        fd_t            *fd   = NULL;
+        int             ret   = -1;
 
-	local = frame->local;
-	priv  = this->private;
+        local = frame->local;
+        priv  = this->private;
 
-	afr_transaction_local_init (local, priv);
+        local->transaction.resume = afr_transaction_resume;
+        local->transaction.type   = type;
 
-	local->transaction.resume = afr_transaction_resume;
-	local->transaction.type   = type;
+        ret = afr_transaction_local_init (local, this);
+        if (ret < 0)
+            goto out;
 
-	if (afr_lock_server_count (priv, local->transaction.type) == 0) {
-		if (__changelog_needed_pre_op (frame, this)) {
-			afr_changelog_pre_op (frame, this);
-		} else {
-                        __mark_all_success (local->pending, priv->child_count,
-                                            local->transaction.type);
+        afr_transaction_eager_lock_init (local, this);
 
-                        afr_pid_restore (frame);
+        if (local->fd && local->transaction.eager_lock_on)
+                afr_set_lk_owner (frame, this, local->fd);
+        else
+                afr_set_lk_owner (frame, this, frame->root);
 
-			local->transaction.fop (frame, this);
-		}
-	} else {
-		afr_lock (frame, this);
-	}
+        if (!local->transaction.eager_lock_on && local->loc.inode) {
+                fd = fd_lookup (local->loc.inode, frame->root->pid);
+                if (fd == NULL)
+                        fd = fd_lookup_anonymous (local->loc.inode);
 
-	return 0;
+                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 0d3d4443e..77cc8eed0 100644
--- a/xlators/cluster/afr/src/afr-transaction.h
+++ b/xlators/cluster/afr/src/afr-transaction.h
@@ -1,25 +1,18 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef __TRANSACTION_H__
 #define __TRANSACTION_H__
 
+#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 f81b8005d..5e12910b7 100644
--- a/xlators/cluster/afr/src/afr.c
+++ b/xlators/cluster/afr/src/afr.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #include <libgen.h>
@@ -28,2988 +19,484 @@
 #define _CONFIG_H
 #include "config.h"
 #endif
+#include "afr-common.c"
 
-#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 "byte-order.h"
-#include "statedump.h"
-
-#include "fd.h"
-
-#include "afr-inode-read.h"
-#include "afr-inode-write.h"
-#include "afr-dir-read.h"
-#include "afr-dir-write.h"
-#include "afr-transaction.h"
-
-#include "afr-self-heal.h"
-
-#define AFR_ICTX_OPENDIR_DONE_MASK     0x0000000200000000ULL
-#define AFR_ICTX_SPLIT_BRAIN_MASK      0x0000000100000000ULL
-#define AFR_ICTX_READ_CHILD_MASK       0x00000000FFFFFFFFULL
-
-
-uint64_t
-afr_is_split_brain (xlator_t *this, inode_t *inode)
-{
-        int ret = 0;
-
-        uint64_t ctx         = 0;
-        uint64_t split_brain = 0;
-
-        VALIDATE_OR_GOTO (inode, out);
+struct volume_options options[];
 
-        LOCK (&inode->lock);
-        {
-                ret = __inode_ctx_get (inode, this, &ctx);
-
-                if (ret < 0)
-                        goto unlock;
-                
-                split_brain = ctx & AFR_ICTX_SPLIT_BRAIN_MASK;
-        }
-unlock:
-        UNLOCK (&inode->lock);
-
-out:
-        return split_brain;
-}
-
-
-void
-afr_set_split_brain (xlator_t *this, inode_t *inode)
+int32_t
+notify (xlator_t *this, int32_t event,
+        void *data, ...)
 {
-        uint64_t ctx = 0;
-        int      ret = 0;
-
-        VALIDATE_OR_GOTO (inode, out);
+        int ret = -1;
+        va_list         ap;
+        void *data2 = NULL;
 
-        LOCK (&inode->lock);
-        {
-                ret = __inode_ctx_get (inode, this, &ctx);
+        va_start (ap, data);
+        data2 = va_arg (ap, dict_t*);
+        va_end (ap);
+        ret = afr_notify (this, event, data, data2);
 
-                if (ret < 0) {
-                        ctx = 0;
-                }
-
-                ctx = (~AFR_ICTX_SPLIT_BRAIN_MASK & ctx)
-                        | (0xFFFFFFFFFFFFFFFFULL & AFR_ICTX_SPLIT_BRAIN_MASK);
-
-                __inode_ctx_put (inode, this, ctx);
-        }
-        UNLOCK (&inode->lock);
-out:
-        return;
+        return ret;
 }
 
-
-uint64_t
-afr_is_opendir_done (xlator_t *this, inode_t *inode)
+int32_t
+mem_acct_init (xlator_t *this)
 {
-        int ret = 0;
-
-        uint64_t ctx          = 0;
-        uint64_t opendir_done = 0;
+        int     ret = -1;
 
-        VALIDATE_OR_GOTO (inode, out);
+        if (!this)
+                return ret;
 
-        LOCK (&inode->lock);
-        {
-                ret = __inode_ctx_get (inode, this, &ctx);
+        ret = xlator_mem_acct_init (this, gf_afr_mt_end + 1);
 
-                if (ret < 0)
-                        goto unlock;
-                
-                opendir_done = ctx & AFR_ICTX_OPENDIR_DONE_MASK;
+        if (ret != 0) {
+                gf_log(this->name, GF_LOG_ERROR, "Memory accounting init"
+                       "failed");
+                return ret;
         }
-unlock:
-        UNLOCK (&inode->lock);
 
-out:
-        return opendir_done;
+        return ret;
 }
 
 
-void
-afr_set_opendir_done (xlator_t *this, inode_t *inode)
+int
+xlator_subvolume_index (xlator_t *this, xlator_t *subvol)
 {
-        uint64_t ctx = 0;
-        int      ret = 0;
-
-        VALIDATE_OR_GOTO (inode, out);
+        int index = -1;
+        int i = 0;
+        xlator_list_t *list = NULL;
 
-        LOCK (&inode->lock);
-        {
-                ret = __inode_ctx_get (inode, this, &ctx);
+        list = this->children;
 
-                if (ret < 0) {
-                        ctx = 0;
+        while (list) {
+                if (subvol == list->xlator ||
+                    strcmp (subvol->name, list->xlator->name) == 0) {
+                        index = i;
+                        break;
                 }
-
-                ctx = (~AFR_ICTX_OPENDIR_DONE_MASK & ctx)
-                        | (0xFFFFFFFFFFFFFFFFULL & AFR_ICTX_OPENDIR_DONE_MASK);
-
-                __inode_ctx_put (inode, this, ctx);
+                list = list->next;
+                i++;
         }
-        UNLOCK (&inode->lock);
-out:
-        return;
-}
-
-
-uint64_t
-afr_read_child (xlator_t *this, inode_t *inode)
-{
-        int ret = 0;
-
-        uint64_t ctx         = 0;
-        uint64_t read_child  = 0;
 
-        VALIDATE_OR_GOTO (inode, out);
-
-        LOCK (&inode->lock);
-        {
-                ret = __inode_ctx_get (inode, this, &ctx);
-
-                if (ret < 0)
-                        goto unlock;
-                
-                read_child = ctx & AFR_ICTX_READ_CHILD_MASK;
-        }
-unlock:
-        UNLOCK (&inode->lock);
-
-out:
-        return read_child;
+        return index;
 }
 
-
 void
-afr_set_read_child (xlator_t *this, inode_t *inode, int32_t read_child)
+fix_quorum_options (xlator_t *this, afr_private_t *priv, char *qtype)
 {
-        uint64_t ctx = 0;
-        int      ret = 0;
-
-        VALIDATE_OR_GOTO (inode, out);
-
-        LOCK (&inode->lock);
-        {
-                ret = __inode_ctx_get (inode, this, &ctx);
-
-                if (ret < 0) {
-                        ctx = 0;
-                }
-
-                ctx = (~AFR_ICTX_READ_CHILD_MASK & ctx) 
-                        | (AFR_ICTX_READ_CHILD_MASK & read_child);
-
-                __inode_ctx_put (inode, this, ctx);
+        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);
         }
-        UNLOCK (&inode->lock);
-        
-out:
-        return;
-}
-
-
-/**
- * afr_local_cleanup - cleanup everything in frame->local
- */
-
-void
-afr_local_sh_cleanup (afr_local_t *local, xlator_t *this)
-{
-	afr_self_heal_t *sh = NULL;
-	afr_private_t   *priv = NULL;
-	int              i = 0;
-
-
-	sh = &local->self_heal;
-	priv = this->private;
-
-	if (sh->buf)
-		FREE (sh->buf);
-
-	if (sh->xattr) {
-		for (i = 0; i < priv->child_count; i++) {
-			if (sh->xattr[i]) {
-				dict_unref (sh->xattr[i]);
-				sh->xattr[i] = NULL;
-			}
-		}
-		FREE (sh->xattr);
-	}
-
-	if (sh->child_errno)
-		FREE (sh->child_errno);
-
-	if (sh->pending_matrix) {
-		for (i = 0; i < priv->child_count; i++) {
-			FREE (sh->pending_matrix[i]);
-		}
-		FREE (sh->pending_matrix);
-	}
-
-	if (sh->delta_matrix) {
-		for (i = 0; i < priv->child_count; i++) {
-			FREE (sh->delta_matrix[i]);
-		}
-		FREE (sh->delta_matrix);
-	}
-
-	if (sh->sources)
-		FREE (sh->sources);
-
-	if (sh->success)
-		FREE (sh->success);
-
-	if (sh->locked_nodes)
-		FREE (sh->locked_nodes);
-
-	if (sh->healing_fd && !sh->healing_fd_opened) {
-		fd_unref (sh->healing_fd);
-		sh->healing_fd = NULL;
-	}
-
-        if (sh->linkname)
-                FREE (sh->linkname);
-
-	loc_wipe (&sh->parent_loc);
-}
-
-
-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);
-
-	FREE (local->child_errno);
-
-        priv = this->private;
-
-        for (i = 0; i < priv->child_count; i++) {
-                if (local->pending && local->pending[i])
-                        FREE (local->pending[i]);
+        if (!strcmp(qtype,"none")) {
+                priv->quorum_count = 0;
         }
-
-        FREE (local->pending);
-
-	loc_wipe (&local->loc);
-	loc_wipe (&local->newloc);
-
-	FREE (local->transaction.locked_nodes);
-	FREE (local->transaction.child_errno);
-
-	FREE (local->transaction.basename);
-	FREE (local->transaction.new_basename);
-
-	loc_wipe (&local->transaction.parent_loc);	
-	loc_wipe (&local->transaction.new_parent_loc);
-
-	if (local->fd)
-		fd_unref (local->fd);
-	
-	if (local->xattr_req)
-		dict_unref (local->xattr_req);
-
-	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;
-                                }
-                        }
-                        FREE (local->cont.lookup.xattrs);
-                        local->cont.lookup.xattrs = NULL;
-                }
-
-		if (local->cont.lookup.xattr) {
-                        dict_unref (local->cont.lookup.xattr);
-                }
-	}
-
-	{ /* getxattr */
-		if (local->cont.getxattr.name)
-			FREE (local->cont.getxattr.name);
-	}
-
-	{ /* lk */
-		if (local->cont.lk.locked_nodes)
-			FREE (local->cont.lk.locked_nodes);
-	}
-
-	{ /* checksum */
-		if (local->cont.checksum.file_checksum)
-			FREE (local->cont.checksum.file_checksum);
-		if (local->cont.checksum.dir_checksum)
-			FREE (local->cont.checksum.dir_checksum);
-	}
-
-	{ /* create */
-		if (local->cont.create.fd)
-			fd_unref (local->cont.create.fd);
-	}
-
-	{ /* writev */
-		FREE (local->cont.writev.vector);
-	}
-
-	{ /* setxattr */
-		if (local->cont.setxattr.dict)
-			dict_unref (local->cont.setxattr.dict);
-	}
-
-	{ /* removexattr */
-		FREE (local->cont.removexattr.name);
-	}
-
-	{ /* symlink */
-		FREE (local->cont.symlink.linkpath);
-	}
-
-        { /* opendir */
-                if (local->cont.opendir.checksum)
-                        FREE (local->cont.opendir.checksum);
+        else if (!strcmp(qtype,"auto")) {
+                priv->quorum_count = AFR_QUORUM_AUTO;
         }
 }
 
-
-int
-afr_frame_return (call_frame_t *frame)
-{
-	afr_local_t *local = NULL;
-	int          call_count = 0;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		call_count = --local->call_count;
-	}
-	UNLOCK (&frame->lock);
-
-	return call_count;
-}
-
-
-/**
- * up_children_count - return the number of children that are up
- */
-
-int
-afr_up_children_count (int child_count, unsigned char *child_up)
-{
-	int i   = 0;
-	int ret = 0;
-
-	for (i = 0; i < child_count; i++)
-		if (child_up[i])
-			ret++;
-	return ret;
-}
-
-
-int
-afr_locked_nodes_count (unsigned char *locked_nodes, int child_count)
-{
-	int ret = 0;
-	int i;
-
-	for (i = 0; i < child_count; i++)
-		if (locked_nodes[i])
-			ret++;
-
-	return ret;
-}
-
-
-ino64_t
-afr_itransform (ino64_t ino, int child_count, int child_index)
-{
-	ino64_t scaled_ino = -1;
-
-	if (ino == ((uint64_t) -1)) {
-		scaled_ino = ((uint64_t) -1);
-		goto out;
-	}
-
-	scaled_ino = (ino * child_count) + child_index;
-
-out:
-	return scaled_ino;
-}
-
-
 int
-afr_deitransform_orig (ino64_t ino, int child_count)
+reconfigure (xlator_t *this, dict_t *options)
 {
-	int index = -1;
-
-	index = ino % child_count;
-
-	return index;
-}
-
-
-int
-afr_deitransform (ino64_t ino, int child_count)
-{
-	return 0;
-}
-
-
-int
-afr_self_heal_lookup_unwind (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t *local = NULL;
-
-	local = frame->local;
-
-	if (local->govinda_gOvinda) {
-                afr_set_split_brain (this, local->cont.lookup.inode);
-	}
-
-	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;
-}
-
-
-static void
-afr_lookup_collect_xattr (afr_local_t *local, xlator_t *this,
-                          int child_index, dict_t *xattr)
-{
-	uint32_t        open_fd_count = 0;
-        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;
-
-        if (afr_sh_has_entry_pending (xattr, child_index, this))
-                local->self_heal.need_entry_self_heal = _gf_true;
-
-        if (afr_sh_has_data_pending (xattr, child_index, this))
-                local->self_heal.need_data_self_heal = _gf_true;
-
-        ret = dict_get_uint32 (xattr, GLUSTERFS_OPEN_FD_COUNT,
-                               &open_fd_count);
-        if (ret == 0)
-                local->open_fd_count += open_fd_count;
-
-        ret = dict_get_uint32 (xattr, GLUSTERFS_INODELK_COUNT,
-                               &inodelk_count);
-        if (ret == 0)
-                local->inodelk_count += inodelk_count;
+        afr_private_t *priv        = NULL;
+        xlator_t      *read_subvol = NULL;
+        int            read_subvol_index = -1;
+        int            ret         = -1;
+        int            index       = -1;
+        char          *qtype       = NULL;
 
-        ret = dict_get_uint32 (xattr, GLUSTERFS_ENTRYLK_COUNT,
-                               &entrylk_count);
-        if (ret == 0)
-                local->entrylk_count += entrylk_count;
-}
+        priv = this->private;
 
+	GF_OPTION_RECONF ("afr-dirty-xattr",
+			  priv->afr_dirty, options, str,
+			  out);
 
-static void
-afr_lookup_self_heal_check (afr_local_t *local, struct stat *buf,
-                            struct stat *lookup_buf)
-{
-        if (FILETYPE_DIFFERS (buf, lookup_buf)) {
-                /* mismatching filetypes with same name
-                   -- Govinda !! GOvinda !!!
-                */
+	GF_OPTION_RECONF ("metadata-splitbrain-forced-heal",
+			  priv->metadata_splitbrain_forced_heal, options, bool,
+			  out);
 
-                gf_log ("afr", GF_LOG_TRACE,
-                        "file %s is govinda!", local->loc.path);
+        GF_OPTION_RECONF ("background-self-heal-count",
+                          priv->background_self_heal_count, options, uint32,
+                          out);
 
-                local->govinda_gOvinda = 1;
-        }
+        GF_OPTION_RECONF ("metadata-self-heal",
+                          priv->metadata_self_heal, options, bool, out);
 
-        if (PERMISSION_DIFFERS (buf, lookup_buf)) {
-                /* mismatching permissions */
-                local->self_heal.need_metadata_self_heal = _gf_true;
-        }
+        GF_OPTION_RECONF ("data-self-heal", priv->data_self_heal, options, str,
+                          out);
 
-        if (OWNERSHIP_DIFFERS (buf, lookup_buf)) {
-                /* mismatching permissions */
-                local->self_heal.need_metadata_self_heal = _gf_true;
-        }
+        GF_OPTION_RECONF ("entry-self-heal", priv->entry_self_heal, options,
+                          bool, out);
 
-        if (SIZE_DIFFERS (buf, lookup_buf)
-            && S_ISREG (buf->st_mode)) {
-                local->self_heal.need_data_self_heal = _gf_true;
-        }
+        GF_OPTION_RECONF ("data-self-heal-window-size",
+                          priv->data_self_heal_window_size, options,
+                          uint32, out);
 
-}
+        GF_OPTION_RECONF ("data-change-log", priv->data_change_log, options,
+                          bool, out);
 
+        GF_OPTION_RECONF ("metadata-change-log",
+                          priv->metadata_change_log, options, bool, out);
 
-static void
-afr_lookup_done (call_frame_t *frame, xlator_t *this, struct stat *lookup_buf)
-{
-        int unwind = 1;
-        int source = -1;
-
-        afr_local_t *local = NULL;
+        GF_OPTION_RECONF ("entry-change-log", priv->entry_change_log, options,
+                          bool, out);
 
-        local = frame->local;
+        GF_OPTION_RECONF ("data-self-heal-algorithm",
+                          priv->data_self_heal_algorithm, options, str, out);
 
-        local->cont.lookup.postparent.st_ino  = local->cont.lookup.parent_ino;
+        GF_OPTION_RECONF ("read-subvolume", read_subvol, options, xlator, out);
 
-        if (local->cont.lookup.ino) {
-                local->cont.lookup.buf.st_ino = local->cont.lookup.ino;
-                local->cont.lookup.buf.st_dev = local->cont.lookup.gen;
-        }
+        GF_OPTION_RECONF ("read-hash-mode", priv->hash_mode,
+                          options, uint32, out);
 
-        if (local->op_ret == 0) {
-                /* KLUDGE: assuming DHT will not itransform in 
-                   revalidate */
-                if (local->cont.lookup.inode->ino) {
-                        local->cont.lookup.buf.st_ino = 
-                                local->cont.lookup.inode->ino;
-                        local->cont.lookup.buf.st_dev =
-                                local->cont.lookup.inode->generation;
+        if (read_subvol) {
+                index = xlator_subvolume_index (this, read_subvol);
+                if (index == -1) {
+                        gf_log (this->name, GF_LOG_ERROR, "%s not a subvolume",
+                                read_subvol->name);
+                        goto out;
                 }
+                priv->read_child = index;
         }
 
-        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;
-        }
-
-        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_OPTION_RECONF ("read-subvolume-index",read_subvol_index, options,int32,out);
 
-        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->open_fd_count
-                    || 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 */
-
-                        if (S_ISREG (local->cont.lookup.inode->st_mode)) {
-                                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->st_mode) {
-                                /* fix for RT #602 */
-                                local->cont.lookup.inode->st_mode =
-                                        lookup_buf->st_mode;
-                        }
-
-                        local->self_heal.background = _gf_true;
-                        local->self_heal.mode       = local->cont.lookup.buf.st_mode;
-                        local->self_heal.unwind     = afr_self_heal_lookup_unwind;
-
-                        unwind = 0;
-
-                        afr_self_heal (frame, this);
+        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;
         }
 
-        if (unwind) {
-                AFR_STACK_UNWIND (lookup, frame, local->op_ret,
-                                  local->op_errno,
-                                  local->cont.lookup.inode, 
-                                  &local->cont.lookup.buf,
-                                  local->cont.lookup.xattr,
-                                  &local->cont.lookup.postparent);
-        }
-}
-
-
-/*
- * During a lookup, some errors are more "important" than
- * others in that they must be given higher priority while
- * returning to the user.
- *
- * The hierarchy is ESTALE > ENOENT > others
- *
- */
-
-static gf_boolean_t
-__error_more_important (int32_t old_errno, int32_t new_errno)
-{
-        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;
+        GF_OPTION_RECONF ("pre-op-compat", priv->pre_op_compat, options, bool, out);
 
-        return ret;
-}
+        GF_OPTION_RECONF ("eager-lock", priv->eager_lock, options, bool, out);
+        GF_OPTION_RECONF ("quorum-type", qtype, options, str, out);
+        GF_OPTION_RECONF ("quorum-count", priv->quorum_count, options,
+                          uint32, out);
+        fix_quorum_options(this,priv,qtype);
 
+	GF_OPTION_RECONF ("post-op-delay-secs", priv->post_op_delay_secs, options,
+			  uint32, out);
 
-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 stat *buf, dict_t *xattr,
-                      struct stat *postparent)
-{
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	struct stat *   lookup_buf = NULL;
-
-	int             call_count      = -1;
-	int             child_index     = -1;
-        int             first_up_child  = -1;
+        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);
 
-	child_index = (long) cookie;
-	priv = this->private;
+	GF_OPTION_RECONF ("self-heal-daemon", priv->shd.enabled, options,
+			  bool, out);
 
-	LOCK (&frame->lock);
-	{
-		local = frame->local;
+	GF_OPTION_RECONF ("iam-self-heal-daemon", priv->shd.iamshd, options,
+			  bool, out);
 
-                lookup_buf = &local->cont.lookup.buf;
+        priv->did_discovery = _gf_false;
 
-		if (op_ret == -1) {
-			if (op_errno == ENOENT)
-				local->enoent_count++;
-
-                        if (__error_more_important (local->op_errno, op_errno))
-                                local->op_errno = op_errno;
-
-                        if (local->op_errno == ESTALE) {
-                                local->op_ret = -1;
-                        }
-
-                        goto unlock;
-		}
-
-                afr_lookup_collect_xattr (local, this, child_index, xattr);
-
-                first_up_child = afr_first_up_child (priv);
-
-                if (child_index == first_up_child) {
-                        local->cont.lookup.ino = 
-                                afr_itransform (buf->st_ino,
-                                                priv->child_count,
-                                                first_up_child);
-                        local->cont.lookup.gen = buf->st_dev;
-                }
-
-		if (local->success_count == 0) {
-                        if (local->op_errno != ESTALE)
-                                local->op_ret = op_ret;
-
-			local->cont.lookup.inode               = inode;
-			local->cont.lookup.xattr               = dict_ref (xattr);
-			local->cont.lookup.xattrs[child_index] = dict_ref (xattr);
-                        local->cont.lookup.postparent          = *postparent;
-
-                        *lookup_buf = *buf;
-
-                        lookup_buf->st_ino = afr_itransform (buf->st_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);
-                        }
-
-		} else {
-                        afr_lookup_self_heal_check (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.inode               = inode;
-                                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);
-                                }
-                        }
-
-		}
-
-		local->success_count++;
-	}
-unlock:
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-                afr_lookup_done (frame, this, lookup_buf);
-	}
+        ret = 0;
+out:
+        return ret;
 
-	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 stat *buf, dict_t *xattr,
-                           struct stat *postparent)
-{
-	afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-	struct stat *   lookup_buf = NULL;
-        
-	int             call_count      = -1;
-	int             child_index     = -1;
-        int             first_up_child  = -1;
-
-	child_index = (long) cookie;
-	priv = this->private;
-
-	LOCK (&frame->lock);
-	{
-		local = frame->local;
-
-		lookup_buf = &local->cont.lookup.buf;
-
-		if (op_ret == -1) {
-			if (op_errno == ENOENT)
-				local->enoent_count++;
-
-                        if (__error_more_important (local->op_errno, op_errno))
-                            local->op_errno = op_errno;
-
-                            if (local->op_errno == ESTALE) {
-                                    local->op_ret = -1;
-                            }
-
-                            goto unlock;
-                }
-
-                afr_lookup_collect_xattr (local, this, child_index, xattr);
-
-                first_up_child = afr_first_up_child (priv);
-
-                if (child_index == first_up_child) {
-                        local->cont.lookup.ino = 
-                                afr_itransform (buf->st_ino,
-                                                priv->child_count,
-                                                first_up_child);
-                        local->cont.lookup.gen = buf->st_dev;
-                }
-
-		/* 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 */
-
-		/* inode number should be preserved across revalidates */
-
-		if (local->success_count == 0) {
-                        if (local->op_errno != ESTALE)
-                                local->op_ret = op_ret;
-
-			local->cont.lookup.inode               = inode;
-			local->cont.lookup.xattr               = dict_ref (xattr);
-			local->cont.lookup.xattrs[child_index] = dict_ref (xattr);
-                        local->cont.lookup.postparent          = *postparent;
-
-			*lookup_buf = *buf;
-
-                        lookup_buf->st_ino = afr_itransform (buf->st_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);
-                        }
-
-		} else {
-                        afr_lookup_self_heal_check (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.inode               = inode;
-                                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);
-                                }
-                        }
-
-		}
-
-		local->success_count++;
-	}
-unlock:
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-                afr_lookup_done (frame, this, lookup_buf);
-	}
-
-	return 0;
-}
+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.";
 
 
-int
-afr_lookup (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, dict_t *xattr_req)
+int32_t
+init (xlator_t *this)
 {
-	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;
-
-	priv = this->private;
-
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-
-	local->op_ret = -1;
-
-	frame->local = local;
-
-        if (!strcmp (loc->path, "/" GF_REPLICATE_TRASH_DIR)) {
-                op_errno = ENOENT;
-                goto out;
+        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;
         }
 
-	loc_copy (&local->loc, loc);
-
-        ret = inode_ctx_get (loc->inode, this, &ctx);
-        if (ret == 0) {
-                /* lookup is a revalidate */
-
-                callback = afr_revalidate_lookup_cbk;
-
-                local->cont.lookup.is_revalidate = _gf_true;
-                local->read_child_index          = afr_read_child (this,
-                                                                   loc->inode);
-        } else {
-                callback = afr_fresh_lookup_cbk;
-
-                LOCK (&priv->read_child_lock);
-                {
-                        local->read_child_index = (++priv->read_child_rr)
-                                % (priv->child_count);
-                }
-                UNLOCK (&priv->read_child_lock);
+        if (!this->parents) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Volume is dangling.");
         }
 
-	local->child_up = memdup (priv->child_up, priv->child_count);
-
-        local->cont.lookup.xattrs = CALLOC (priv->child_count,
-                                            sizeof (*local->cont.lookup.xattr));
-
-	local->call_count = afr_up_children_count (priv->child_count,
-                                                   local->child_up);
-        call_count = local->call_count;
-
-        if (local->call_count == 0) {
-                ret      = -1;
-                op_errno = ENOTCONN;
+	this->private = GF_CALLOC (1, sizeof (afr_private_t),
+                                   gf_afr_mt_afr_private_t);
+        if (!this->private)
                 goto out;
-        }
-
-	/* By default assume ENOTCONN. On success it will be set to 0. */
-	local->op_errno = ENOTCONN;
-	
-	if (xattr_req == NULL)
-		local->xattr_req = dict_new ();
-	else
-		local->xattr_req = dict_ref (xattr_req);
-
-        for (i = 0; i < priv->child_count; i++) {
-		ret = dict_set_uint64 (local->xattr_req, priv->pending_key[i],
-				       3 * sizeof(int32_t));
-                
-                /* 3 = data+metadata+entry */
-        }
-
-	ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_OPEN_FD_COUNT, 0);
-        ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_INODELK_COUNT, 0);
-        ret = dict_set_uint64 (local->xattr_req, GLUSTERFS_ENTRYLK_COUNT, 0);
-
-	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;
-                }
-	}
-
-	ret = 0;
-out:
-	if (ret == -1)
-		AFR_STACK_UNWIND (lookup, frame, -1, op_errno,
-                                  NULL, NULL, NULL, NULL);
-
-	return 0;
-}
-
-
-/* {{{ open */
-
-int
-afr_fd_ctx_set (xlator_t *this, fd_t *fd)
-{
-        afr_private_t * priv = NULL;
-
-        int op_ret = 0;
-        int ret    = 0;
-
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
-
-        VALIDATE_OR_GOTO (this->private, out);
-        VALIDATE_OR_GOTO (fd, out);
 
         priv = this->private;
+        LOCK_INIT (&priv->lock);
 
-        LOCK (&fd->lock);
-        {
-                ret = __fd_ctx_get (fd, this, &ctx);
-                
-                if (ret == 0)
-                        goto unlock;
-
-                fd_ctx = CALLOC (1, sizeof (afr_fd_ctx_t));
-                if (!fd_ctx) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        
-                        op_ret = -ENOMEM;
-                        goto unlock;
-                }
-
-                fd_ctx->pre_op_done = CALLOC (sizeof (*fd_ctx->pre_op_done),
-                                              priv->child_count);
-                if (!fd_ctx->pre_op_done) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        op_ret = -ENOMEM;
-                        goto unlock;
-                }
+        child_count = xlator_subvolume_count (this);
 
-                fd_ctx->opened_on = CALLOC (sizeof (*fd_ctx->opened_on),
-                                            priv->child_count);
-                if (!fd_ctx->opened_on) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        op_ret = -ENOMEM;
-                        goto unlock;
-                }
+        priv->child_count = child_count;
 
-                fd_ctx->child_failed = CALLOC (sizeof (*fd_ctx->child_failed),
-                                               priv->child_count);
-                
-                if (!fd_ctx->child_failed) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
+        priv->read_child = -1;
 
-                        op_ret = -ENOMEM;
-                        goto unlock;
-                }
+	GF_OPTION_INIT ("afr-dirty-xattr", priv->afr_dirty, str, out);
 
-                fd_ctx->up_count   = priv->up_count;
-                fd_ctx->down_count = priv->down_count;
+	GF_OPTION_INIT ("metadata-splitbrain-forced-heal",
+			priv->metadata_splitbrain_forced_heal, bool, out);
 
-                ret = __fd_ctx_set (fd, this, (uint64_t)(long) fd_ctx);
-                if (ret < 0) {
-                        op_ret = ret;
+        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;
                 }
         }
-unlock:
-        UNLOCK (&fd->lock);
-out:
-        return ret;
-}
-
-/* {{{ flush */
-
-int
-afr_flush_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 (flush, main_frame,
-                                  local->op_ret, local->op_errno);
-	}
-        
-	return 0;
-}
-
-
-int
-afr_flush_wind_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		      int32_t op_ret, int32_t op_errno)
-{
-        afr_local_t *   local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int call_count  = -1;
-	int child_index = (long) cookie;
-	int need_unwind = 0;
-
-	local = frame->local;
-	priv  = this->private;
-
-	LOCK (&frame->lock);
-	{
-		if (afr_fop_failed (op_ret, op_errno))
-			afr_transaction_fop_failed (frame, this, child_index);
-
-		if (op_ret != -1) {
-			if (local->success_count == 0) {
-				local->op_ret = op_ret;
-			}
-			local->success_count++;
-
-			if (local->success_count == priv->wait_count) {
-				need_unwind = 1;
-			}
-		}
-
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	if (need_unwind)
-		afr_flush_unwind (frame, this);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0) {
-		local->transaction.resume (frame, this);
-	}
-	
-	return 0;
-}
-
-
-int
-afr_flush_wind (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t *local = NULL;
-	afr_private_t *priv = NULL;
-	
-	int i = 0;
-	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;
-
-	for (i = 0; i < priv->child_count; i++) {				
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_flush_wind_cbk, 
-					   (void *) (long) i,	
-					   priv->children[i], 
-					   priv->children[i]->fops->flush,
-					   local->fd);
-		
-			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_plain_flush_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 = -1;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0)
-			local->op_ret = 0;
-
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		AFR_STACK_UNWIND (flush, frame, local->op_ret, local->op_errno);
-
-	return 0;
-}
-
-
-static int
-__no_pre_op_done (xlator_t *this, fd_t *fd)
-{
-        int i      = 0;
-        int op_ret = 1;
-
-        int _ret = 0;
-        uint64_t       ctx;
-        afr_fd_ctx_t * fd_ctx = NULL;
-
-        afr_private_t *priv = NULL;
-
-        priv = this->private;
-
-        LOCK (&fd->lock);
-        {
-                _ret = __fd_ctx_get (fd, this, &ctx);
-
-                if (_ret < 0) {
+        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;
                 }
-
-                fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-                for (i = 0; i < priv->child_count; i++) {
-                        if (fd_ctx->pre_op_done[i]) {
-                                op_ret = 0;
-                                break;
-                        }
-                }
+                priv->read_child = read_subvol_index;
         }
-out:
-        UNLOCK (&fd->lock);
+        GF_OPTION_INIT ("choose-local", priv->choose_local, bool, out);
 
-        return op_ret;
-}
-
-
-int
-afr_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
-{
-	afr_private_t * priv  = NULL;
-	afr_local_t   * local = NULL;
-
-        call_frame_t  * transaction_frame = NULL;
-
-	int ret        = -1;
+        GF_OPTION_INIT ("read-hash-mode", priv->hash_mode, uint32, out);
 
-	int op_ret   = -1;
-	int op_errno = 0;
-
-        int i          = 0;
-        int call_count = 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;
-	}
-
-        call_count = afr_up_children_count (priv->child_count, local->child_up);
-
-        if (__no_pre_op_done (this, fd)) {
-                frame->local = local;
-
-                for (i = 0; i < priv->child_count; i++) {
-                        if (local->child_up[i]) {
-                                STACK_WIND_COOKIE (frame, afr_plain_flush_cbk,
-                                                   (void *) (long) i,
-                                                   priv->children[i],
-                                                   priv->children[i]->fops->flush,
-                                                   fd);
-                                if (!--call_count)
-                                        break;
-                        }
-                }
-        } else {
-                transaction_frame = copy_frame (frame);
-                if (!transaction_frame) {
-                        op_errno = ENOMEM;
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory.");
+        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;
                 }
-
-                transaction_frame->local = local;
-
-                local->op = GF_FOP_FLUSH;
-        
-                local->transaction.fop    = afr_flush_wind;
-                local->transaction.done   = afr_flush_done;
-                local->transaction.unwind = afr_flush_unwind;
-
-                local->fd                 = fd_ref (fd);
-
-                local->transaction.main_frame = frame;
-                local->transaction.start  = LLONG_MAX - 1;
-                local->transaction.len    = 0;
-
-                afr_transaction (transaction_frame, this, AFR_FLUSH_TRANSACTION);
+                gf_log (this->name, GF_LOG_WARNING,
+                        favorite_child_warning_str, fav_child->name,
+                        fav_child->name, fav_child->name);
         }
 
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-                if (transaction_frame)
-                        AFR_STACK_DESTROY (transaction_frame);
-
-		AFR_STACK_UNWIND (flush, frame, op_ret, op_errno);
-	}
-
-	return 0;
-}
-
-/* }}} */
-
 
-int
-afr_release (xlator_t *this, fd_t *fd)
-{
-        uint64_t        ctx;
-        afr_fd_ctx_t *  fd_ctx;
+        GF_OPTION_INIT ("background-self-heal-count",
+                        priv->background_self_heal_count, uint32, out);
 
-        int ret = 0;
+        GF_OPTION_INIT ("data-self-heal", priv->data_self_heal, str, out);
 
-        ret = fd_ctx_get (fd, this, &ctx);
-
-        if (ret < 0)
-                goto out;
+        GF_OPTION_INIT ("data-self-heal-algorithm",
+                        priv->data_self_heal_algorithm, str, out);
 
-        fd_ctx = (afr_fd_ctx_t *)(long) ctx;
-
-        if (fd_ctx) {
-                if (fd_ctx->child_failed)
-                        FREE (fd_ctx->child_failed);
-
-                FREE (fd_ctx);
-        }
-        
-out:
-        return 0;
-}
+        GF_OPTION_INIT ("data-self-heal-window-size",
+                        priv->data_self_heal_window_size, uint32, out);
 
+        GF_OPTION_INIT ("metadata-self-heal", priv->metadata_self_heal, bool,
+                        out);
 
-/* {{{ fsync */
+        GF_OPTION_INIT ("entry-self-heal", priv->entry_self_heal, bool, out);
 
-int
-afr_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-               int32_t op_ret, int32_t op_errno, struct stat *prebuf,
-               struct stat *postbuf)
-{
-	afr_local_t *local = NULL;
-	
-	int call_count = -1;
+        GF_OPTION_INIT ("data-change-log", priv->data_change_log, bool, out);
 
-        int child_index = (long) cookie;
-        int read_child  = 0;
+        GF_OPTION_INIT ("metadata-change-log", priv->metadata_change_log, bool,
+                        out);
 
-	local = frame->local;
+        GF_OPTION_INIT ("entry-change-log", priv->entry_change_log, bool, out);
 
-        read_child = afr_read_child (this, local->fd->inode);
+        GF_OPTION_INIT ("optimistic-change-log", priv->optimistic_change_log,
+                        bool, out);
 
-	LOCK (&frame->lock);
-	{
-                if (child_index == read_child) {
-                        local->read_child_returned = _gf_true;
-                }
+        GF_OPTION_INIT ("inodelk-trace", priv->inodelk_trace, bool, out);
 
-		if (op_ret == 0) {
-			local->op_ret = 0;
+        GF_OPTION_INIT ("entrylk-trace", priv->entrylk_trace, bool, out);
 
-			if (local->success_count == 0) {
-				local->cont.fsync.prebuf  = *prebuf;
-				local->cont.fsync.postbuf = *postbuf;
-			}
+        GF_OPTION_INIT ("pre-op-compat", priv->pre_op_compat, bool, out);
 
-                        if (child_index == read_child) {
-                                local->cont.fsync.prebuf  = *prebuf;
-                                local->cont.fsync.postbuf = *postbuf;
-                        }
+        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);
 
-			local->success_count++;
-                }
+	GF_OPTION_INIT ("post-op-delay-secs", priv->post_op_delay_secs, uint32, out);
+        GF_OPTION_INIT ("ensure-durability", priv->ensure_durability, bool,
+                        out);
 
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
+	GF_OPTION_INIT ("self-heal-daemon", priv->shd.enabled, bool, out);
 
-	call_count = afr_frame_return (frame);
+	GF_OPTION_INIT ("iam-self-heal-daemon", priv->shd.iamshd, bool, out);
 
-	if (call_count == 0) {
-                local->cont.fsync.prebuf.st_ino  = local->cont.fsync.ino;
-                local->cont.fsync.postbuf.st_ino = local->cont.fsync.ino;
+        priv->wait_count = 1;
 
-		AFR_STACK_UNWIND (fsync, frame, local->op_ret, local->op_errno,
-                                  &local->cont.fsync.prebuf,
-                                  &local->cont.fsync.postbuf);
+        priv->child_up = GF_CALLOC (sizeof (unsigned char), child_count,
+                                    gf_afr_mt_char);
+        if (!priv->child_up) {
+                ret = -ENOMEM;
+                goto out;
         }
 
-	return 0;
-}
-
-
-int
-afr_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd,
-	   int32_t datasync)
-{
-	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;
-	}
-
-	call_count = local->call_count;
-	frame->local = local;
-
-        local->fd             = fd_ref (fd);
-        local->cont.fsync.ino = fd->inode->ino;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND_COOKIE (frame, afr_fsync_cbk,
-                                           (void *) (long) i,
-                                           priv->children[i],
-                                           priv->children[i]->fops->fsync,
-                                           fd, datasync);
-			if (!--call_count)
-				break;
-		}
-	}
-
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (fsync, frame, op_ret, op_errno, NULL, NULL);
-	}
-	return 0;
-}
-
-/* }}} */
-
-/* {{{ fsync */
-
-int32_t
-afr_fsyncdir_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 = -1;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0)
-			local->op_ret = 0;
-
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		AFR_STACK_UNWIND (fsyncdir, frame, local->op_ret,
-                                  local->op_errno);
-
-	return 0;
-}
-
-
-int32_t
-afr_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
-	      int32_t datasync)
-{
-	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;
-	}
-
-	call_count = local->call_count;
-	frame->local = local;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_fsyncdir_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->fsyncdir,
-				    fd, datasync);
-			if (!--call_count)
-				break;
-		}
-	}
-
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (fsyncdir, frame, op_ret, op_errno);
-	}
-	return 0;
-}
-
-/* }}} */
-
-/* {{{ xattrop */
-
-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)
-{
-	afr_local_t *local = NULL;
-	
-	int call_count = -1;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0)
-			local->op_ret = 0;
-
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		AFR_STACK_UNWIND (xattrop, frame, local->op_ret, local->op_errno,
-                                  xattr);
-
-	return 0;
-}
-
-
-int32_t
-afr_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
-	     gf_xattrop_flags_t optype, dict_t *xattr)
-{
-	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;
-	}
-
-	call_count = local->call_count;
-	frame->local = local;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_xattrop_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->xattrop,
-				    loc, optype, xattr);
-			if (!--call_count)
-				break;
-		}
-	}
-
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (xattrop, frame, op_ret, op_errno, NULL);
-	}
-	return 0;
-}
-
-/* }}} */
-
-/* {{{ fxattrop */
-
-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)
-{
-	afr_local_t *local = NULL;
-	
-	int call_count = -1;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0)
-			local->op_ret = 0;
-
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		AFR_STACK_UNWIND (fxattrop, frame, local->op_ret, local->op_errno,
-                                  xattr);
-
-	return 0;
-}
-
-
-int32_t
-afr_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
-	      gf_xattrop_flags_t optype, dict_t *xattr)
-{
-	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;
-	}
-
-	call_count = local->call_count;
-	frame->local = local;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_fxattrop_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->fxattrop,
-				    fd, optype, xattr);
-			if (!--call_count)
-				break;
-		}
-	}
-
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, NULL);
-	}
-	return 0;
-}
-
-/* }}} */
-
-
-int32_t
-afr_inodelk_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 = -1;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0)
-			local->op_ret = 0;
-
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		AFR_STACK_UNWIND (inodelk, frame, local->op_ret,
-                                  local->op_errno);
-
-	return 0;
-}
-
-
-int32_t
-afr_inodelk (call_frame_t *frame, xlator_t *this, 
-             const char *volume, loc_t *loc, int32_t cmd, struct flock *flock)
-{
-	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;
-	}
-
-	call_count = local->call_count;
-	frame->local = local;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_inodelk_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->inodelk,
-				    volume, loc, cmd, flock);
-
-			if (!--call_count)
-				break;
-		}
-	}
-
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (inodelk, frame, op_ret, op_errno);
-	}
-	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_local_t *local = NULL;
-	
-	int call_count = -1;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0)
-			local->op_ret = 0;
-
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		AFR_STACK_UNWIND (finodelk, frame, local->op_ret,
-                                  local->op_errno);
-
-	return 0;
-}
-
-
-int32_t
-afr_finodelk (call_frame_t *frame, xlator_t *this, 
-              const char *volume, fd_t *fd, int32_t cmd, struct flock *flock)
-{
-	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;
-	}
-
-	call_count = local->call_count;
-	frame->local = local;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_finodelk_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->finodelk,
-				    volume, fd, cmd, flock);
-
-			if (!--call_count)
-				break;
-		}
-	}
-
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (finodelk, frame, op_ret, op_errno);
-	}
-	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_local_t *local = NULL;
-	
-	int call_count = -1;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0)
-			local->op_ret = 0;
-
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		AFR_STACK_UNWIND (entrylk, frame, local->op_ret,
-                                  local->op_errno);
-
-	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)
-{
-	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;
-	}
-
-	call_count = local->call_count;
-	frame->local = local;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_entrylk_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->entrylk,
-				    volume, loc, basename, cmd, type);
-
-			if (!--call_count)
-				break;
-		}
-	}
-
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (entrylk, frame, op_ret, op_errno);
-	}
-	return 0;
-}
-
-
-
-int32_t
-afr_fentrylk_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 = -1;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0)
-			local->op_ret = 0;
-
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		AFR_STACK_UNWIND (fentrylk, frame, local->op_ret,
-                                  local->op_errno);
-
-	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)
-{
-	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;
-	}
-
-	call_count = local->call_count;
-	frame->local = local;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_fentrylk_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->fentrylk,
-				    volume, fd, basename, cmd, type);
-
-			if (!--call_count)
-				break;
-		}
-	}
-
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (fentrylk, frame, op_ret, op_errno);
-	}
-	return 0;
-}
-
-
-int32_t
-afr_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)
-		
-{
-	afr_local_t *local = NULL;
-	
-	int call_count = -1;
-
-	local = frame->local;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == 0 && (local->op_ret != 0)) {
-			local->op_ret = 0;
-
-			local->cont.checksum.file_checksum = MALLOC (NAME_MAX);
-			memcpy (local->cont.checksum.file_checksum, file_checksum, 
-				NAME_MAX);
-
-			local->cont.checksum.dir_checksum = MALLOC (NAME_MAX);
-			memcpy (local->cont.checksum.dir_checksum, dir_checksum, 
-				NAME_MAX);
-
-		}
-
-		local->op_errno = op_errno;
-	}
-	UNLOCK (&frame->lock);
-
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		AFR_STACK_UNWIND (checksum, frame, local->op_ret, local->op_errno,
-				  local->cont.checksum.file_checksum, 
-				  local->cont.checksum.dir_checksum);
-
-	return 0;
-}
-
-
-int32_t
-afr_checksum (call_frame_t *frame, xlator_t *this, loc_t *loc,
-	      int32_t flag)
-{
-	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;
-	}
-
-	call_count = local->call_count;
-	frame->local = local;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_checksum_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->checksum,
-				    loc, flag);
-
-			if (!--call_count)
-				break;
-		}
-	}
-
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (checksum, frame, op_ret, op_errno,
-                                  NULL, 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)
-{
-	afr_local_t *local = NULL;
-
-	int call_count = 0;
-
-	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)
-			local->op_errno = op_errno;
-
-	}
-	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);
-
-	return 0;
-}
-
-
-int32_t
-afr_statfs (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc)
-{
-	afr_private_t *  priv        = NULL;
-	int              child_count = 0;
-	afr_local_t   *  local       = NULL;
-	int              i           = 0;
-
-	int ret = -1;
-	int              call_count = 0;
-	int32_t          op_ret      = -1;
-	int32_t          op_errno    = 0;
-
-	VALIDATE_OR_GOTO (this, out);
-	VALIDATE_OR_GOTO (this->private, out);
-	VALIDATE_OR_GOTO (loc, out);
-
-	priv = this->private;
-	child_count = priv->child_count;
-
-	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;
-
-	for (i = 0; i < child_count; i++) {
-		if (local->child_up[i]) {
-			STACK_WIND (frame, afr_statfs_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->statfs, 
-				    loc);
-			if (!--call_count)
-				break;
-		}
-	}
-	
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (statfs, frame, op_ret, op_errno, NULL);
-	}
-	return 0;
-}
-
-
-int32_t
-afr_lk_unlock_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-		   int32_t op_ret, int32_t op_errno, struct flock *lock)
-{
-	afr_local_t * local = NULL;
-
-	int call_count = -1;
-
-	local = frame->local;
-	call_count = afr_frame_return (frame);
-
-	if (call_count == 0)
-		AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
-				  lock);
-
-	return 0;
-}
-
-
-int32_t 
-afr_lk_unlock (call_frame_t *frame, xlator_t *this)
-{
-	afr_local_t   * local = NULL;
-	afr_private_t * priv  = NULL;
-
-	int i;
-	int call_count = 0;
-
-	local = frame->local;
-	priv  = this->private;
-
-	call_count = afr_locked_nodes_count (local->cont.lk.locked_nodes, 
-					     priv->child_count);
-
-	if (call_count == 0) {
-		AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
-				  &local->cont.lk.flock);
-		return 0;
-	}
-
-	local->call_count = call_count;
-
-	local->cont.lk.flock.l_type = F_UNLCK;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if (local->cont.lk.locked_nodes[i]) {
-			STACK_WIND (frame, afr_lk_unlock_cbk,
-				    priv->children[i],
-				    priv->children[i]->fops->lk,
-				    local->fd, F_SETLK, 
-				    &local->cont.lk.flock);
-
-			if (!--call_count)
-				break;
-		}
-	}
-
-	return 0;
-}
-
-
-int32_t
-afr_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
-	    int32_t op_ret, int32_t op_errno, struct flock *lock)
-{
-	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;
-
-	call_count = --local->call_count;
-
-	if (!child_went_down (op_ret, op_errno) && (op_ret == -1)) {
-		local->op_ret   = -1;
-		local->op_errno = op_errno;
-
-		afr_lk_unlock (frame, this);
-		return 0;
-	}
-
-	if (op_ret == 0) {
-		local->op_ret        = 0;
-		local->op_errno      = 0;
-		local->cont.lk.locked_nodes[child_index] = 1;
-	}
-
-	child_index++;
-
-	if (child_index < priv->child_count) {
-		STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) child_index,
-				   priv->children[child_index],
-				   priv->children[child_index]->fops->lk,
-				   local->fd, local->cont.lk.cmd, 
-				   &local->cont.lk.flock);
-	} else if (local->op_ret == -1) {
-		/* all nodes have gone down */
-		
-		AFR_STACK_UNWIND (lk, frame, -1, ENOTCONN, &local->cont.lk.flock);
-	} else {
-		/* locking has succeeded on all nodes that are up */
-		
-		AFR_STACK_UNWIND (lk, frame, local->op_ret, local->op_errno,
-			      &local->cont.lk.flock);
-	}
-
-	return 0;
-}
-
-
-int
-afr_lk (call_frame_t *frame, xlator_t *this,
-	fd_t *fd, int32_t cmd,
-	struct flock *flock)
-{
-	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);
-
-	priv = this->private;
-
-	ALLOC_OR_GOTO (local, afr_local_t, out);
-	AFR_LOCAL_INIT (local, priv);
-
-	frame->local  = local;
-
-	local->cont.lk.locked_nodes = CALLOC (priv->child_count, 
-					      sizeof (*local->cont.lk.locked_nodes));
-	
-	if (!local->cont.lk.locked_nodes) {
-		gf_log (this->name, GF_LOG_ERROR, "Out of memory");
-		op_errno = ENOMEM;
-		goto out;
-	}
-
-	local->fd            = fd_ref (fd);
-	local->cont.lk.cmd   = cmd;
-	local->cont.lk.flock = *flock;
-
-	STACK_WIND_COOKIE (frame, afr_lk_cbk, (void *) (long) 0,
-			   priv->children[i],
-			   priv->children[i]->fops->lk,
-			   fd, cmd, flock);
-
-	op_ret = 0;
-out:
-	if (op_ret == -1) {
-		AFR_STACK_UNWIND (lk, frame, op_ret, op_errno, NULL);
-	}
-	return 0;
-}
-
-int
-afr_priv_dump (xlator_t *this)
-{
-	afr_private_t *priv = NULL;
-        char  key_prefix[GF_DUMP_MAX_BUF_LEN];
-        char  key[GF_DUMP_MAX_BUF_LEN];
-        int   i = 0;
-
-
-        assert(this);
-	priv = this->private;
-
-        assert(priv);
-        snprintf(key_prefix, GF_DUMP_MAX_BUF_LEN, "%s.%s", this->type, this->name);
-        gf_proc_dump_add_section(key_prefix);
-        gf_proc_dump_build_key(key, key_prefix, "child_count");
-        gf_proc_dump_write(key, "%u", priv->child_count);
-        gf_proc_dump_build_key(key, key_prefix, "read_child_rr");
-        gf_proc_dump_write(key, "%u", priv->read_child_rr);
-	for (i = 0; i < priv->child_count; i++) {
-                gf_proc_dump_build_key(key, key_prefix, "child_up[%d]", i);
-                gf_proc_dump_write(key, "%d", priv->child_up[i]);
-                gf_proc_dump_build_key(key, key_prefix, 
-                                        "pending_key[%d]", i);
-                gf_proc_dump_write(key, "%s", priv->pending_key[i]);
+        for (i = 0; i < child_count; i++)
+                priv->child_up[i] = -1; /* start with unknown state.
+                                           this initialization needed
+                                           for afr_notify() to work
+                                           reliably
+                                        */
+
+        priv->children = GF_CALLOC (sizeof (xlator_t *), child_count,
+                                    gf_afr_mt_xlator_t);
+        if (!priv->children) {
+                ret = -ENOMEM;
+                goto out;
         }
-        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);
-
-        return 0;
-}
-
-
-/**
- * find_child_index - find the child's index in the array of subvolumes
- * @this: AFR
- * @child: child
- */
 
-static int
-find_child_index (xlator_t *this, xlator_t *child)
-{
-	afr_private_t *priv = NULL;
-
-	int i = -1;
-
-	priv = this->private;
-
-	for (i = 0; i < priv->child_count; i++) {
-		if ((xlator_t *) child == priv->children[i])
-			break;
-	}
-
-	return i;
-}
-
-
-int32_t
-notify (xlator_t *this, int32_t event,
-	void *data, ...)
-{
-	afr_private_t *     priv     = NULL;
-	unsigned char *     child_up = NULL;
-
-	int i           = -1;
-	int up_children = 0;
-
-	priv = this->private;
-
-	if (!priv)
-		return 0;
-
-	child_up = priv->child_up;
-
-	switch (event) {
-	case GF_EVENT_CHILD_UP:
-		i = find_child_index (this, data);
-
-		child_up[i] = 1;
-
-                LOCK (&priv->lock);
-                {
-                        priv->up_count++;
-                }
-                UNLOCK (&priv->lock);
-
-		/* 
-		   if all the children were down, and one child came up, 
-		   send notify to parent
-		*/
-
-		for (i = 0; i < priv->child_count; i++)
-			if (child_up[i])
-				up_children++;
-
-		if (up_children == 1) {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "Subvolume '%s' came back up; "
-                                "going online.", ((xlator_t *)data)->name);
-                        
-			default_notify (this, event, data);
-                }
-
-		break;
-
-	case GF_EVENT_CHILD_DOWN:
-		i = find_child_index (this, data);
+        priv->pending_key = GF_CALLOC (sizeof (*priv->pending_key),
+                                       child_count,
+                                       gf_afr_mt_char);
+        if (!priv->pending_key) {
+                ret = -ENOMEM;
+                goto out;
+        }
 
-		child_up[i] = 0;
+        trav = this->children;
+        i = 0;
+        while (i < child_count) {
+                priv->children[i] = trav->xlator;
 
-                LOCK (&priv->lock);
-                {
-                        priv->down_count++;
-                }
-                UNLOCK (&priv->lock);
-		
-		/* 
-		   if all children are down, and this was the last to go down,
-		   send notify to parent
-		*/
-
-		for (i = 0; i < priv->child_count; i++)
-			if (child_up[i])
-				up_children++;
-
-		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.");
-
-			default_notify (this, event, data);
+                ret = gf_asprintf (&priv->pending_key[i], "%s.%s",
+                                   AFR_XATTR_PREFIX,
+                                   trav->xlator->name);
+                if (-1 == ret) {
+                        ret = -ENOMEM;
+                        goto out;
                 }
 
-		break;
-
-	default:
-		default_notify (this, event, data);
-	}
-
-	return 0;
-}
-
-
-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 
-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;
-
-        int32_t background_count  = 0;
-	int32_t lock_server_count = 1;
-        int32_t window_size;
-
-	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 "
-                        "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.",
-			window_size);
-
-		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 = strdup (algo);
+                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->data_self_heal_window_size = 16;
-
-	dict_ret = dict_get_int32 (this->options, "data-self-heal-window-size",
-				   &window_size);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Setting data self-heal window size to %d.",
-			window_size);
-
-		priv->data_self_heal_window_size = window_size;
-	}
-
-	dict_ret = dict_get_str (this->options, "metadata-self-heal",
-				 &self_heal);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (self_heal, &priv->metadata_self_heal);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option metadata-self-heal %s'. "
-				"Defaulting to metadata-self-heal as 'on'.", 
-				self_heal);
-			priv->metadata_self_heal = 1;
-		} 
-	}
-
-	dict_ret = dict_get_str (this->options, "entry-self-heal", &self_heal);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (self_heal, &priv->entry_self_heal);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option entry-self-heal %s'. "
-				"Defaulting to entry-self-heal as 'on'.", 
-				self_heal);
-			priv->entry_self_heal = 1;
-		} 
-	}
-
-	/* Change log options */
-
-	priv->data_change_log     = 1;
-	priv->metadata_change_log = 0;
-	priv->entry_change_log    = 1;
-
-	dict_ret = dict_get_str (this->options, "data-change-log",
-				 &change_log);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (change_log, &priv->data_change_log);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option data-change-log %s'. "
-				"Defaulting to data-change-log as 'on'.", 
-				change_log);
-			priv->data_change_log = 1;
-		} 
-	}
-
-	dict_ret = dict_get_str (this->options, "metadata-change-log",
-				 &change_log);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (change_log,
-					 &priv->metadata_change_log);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option metadata-change-log %s'. "
-				"Defaulting to metadata-change-log as 'off'.",
-				change_log);
-			priv->metadata_change_log = 0;
-		} 
-	}
-
-	dict_ret = dict_get_str (this->options, "entry-change-log",
-				 &change_log);
-	if (dict_ret == 0) {
-		ret = gf_string2boolean (change_log, &priv->entry_change_log);
-		if (ret < 0) {
-			gf_log (this->name, GF_LOG_WARNING,
-				"Invalid 'option entry-change-log %s'. "
-				"Defaulting to entry-change-log as 'on'.", 
-				change_log);
-			priv->entry_change_log = 1;
-		} 
-	}
-
-	/* Locking options */
-
-	priv->data_lock_server_count = 1;
-	priv->metadata_lock_server_count = 0;
-	priv->entry_lock_server_count = 1;
-
-	dict_ret = dict_get_int32 (this->options, "data-lock-server-count", 
-				   &lock_server_count);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Setting data lock server count to %d.",
-			lock_server_count);
-
-		if (lock_server_count == 0) 
-			gf_log (this->name, GF_LOG_WARNING, "%s",
-                                no_lock_servers_warning_str);
-
-		priv->data_lock_server_count = lock_server_count;
-	}
-
-
-	dict_ret = dict_get_int32 (this->options,
-				   "metadata-lock-server-count", 
-				   &lock_server_count);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Setting metadata lock server count to %d.",
-			lock_server_count);
-		priv->metadata_lock_server_count = lock_server_count;
-	}
-
-
-	dict_ret = dict_get_int32 (this->options, "entry-lock-server-count", 
-				   &lock_server_count);
-	if (dict_ret == 0) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"Setting entry lock server count to %d.",
-			lock_server_count);
-
-		priv->entry_lock_server_count = lock_server_count;
-	}
-
-	trav = this->children;
-	while (trav) {
-		if (!read_ret && !strcmp (read_subvol, trav->xlator->name)) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"Subvolume '%s' specified as read child.",
-				trav->xlator->name);
-
-			priv->read_child = child_count;
-		}
-
-		if (fav_ret == 0 && !strcmp (fav_child, trav->xlator->name)) {
-			gf_log (this->name, GF_LOG_WARNING,
-				favorite_child_warning_str, trav->xlator->name,
-				trav->xlator->name, trav->xlator->name);
-			priv->favorite_child = child_count;
-		}
-
-		child_count++;
-		trav = trav->next;
-	}
-
-	priv->wait_count = 1;
-
-	priv->child_count = child_count;
-
-	LOCK_INIT (&priv->lock);
-        LOCK_INIT (&priv->read_child_lock);
-
-	priv->child_up = CALLOC (sizeof (unsigned char), child_count);
-	if (!priv->child_up) {
-		gf_log (this->name, GF_LOG_ERROR,	
-			"Out of memory.");		
-		op_errno = 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;
+        }
 
-	priv->children = CALLOC (sizeof (xlator_t *), child_count);
-	if (!priv->children) {
-		gf_log (this->name, GF_LOG_ERROR,	
-			"Out of memory.");		
-		op_errno = ENOMEM;			
+	ret = afr_selfheal_daemon_init (this);
+	if (ret) {
+		ret = -ENOMEM;
 		goto out;
 	}
 
-        priv->pending_key = CALLOC (sizeof (*priv->pending_key), child_count);
-        if (!priv->pending_key) {
+        /* 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,
-                        "Out of memory.");
-                op_errno = ENOMEM;
+                        "failed to create local_t's memory pool");
                 goto out;
         }
 
-	trav = this->children;
-	i = 0;
-	while (i < child_count) {
-		priv->children[i] = trav->xlator;
+        priv->root_inode = NULL;
 
-                ret = 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");
-                        op_errno = ENOMEM;
-                        goto out;
-                }
-
-		trav = trav->next;
-		i++;
-	}
-
-	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,
-	.checksum    = afr_checksum,
-
-	/* 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,
-	.getdents    = afr_getdents,
-
-	/* 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,
-	.setdents    = afr_setdents,
+        .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,
 };
 
 
-struct xlator_mops mops = {
-};
-
 struct xlator_dumpops dumpops = {
         .priv       = afr_priv_dump,
 };
@@ -3017,57 +504,246 @@ struct xlator_dumpops dumpops = {
 
 struct xlator_cbks cbks = {
         .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-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  = {"metadata-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  = {"entry-lock-server-count"},  
-	  .type = GF_OPTION_TYPE_INT,
-	  .min  = 0
+	{ .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 c5db12a44..36042f7b2 100644
--- a/xlators/cluster/afr/src/afr.h
+++ b/xlators/cluster/afr/src/afr.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -26,143 +17,147 @@
 #include "config.h"
 #endif
 
-#include "scheduler.h"
 #include "call-stub.h"
 #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)
+
+#define AFR_LOCKEE_COUNT_MAX    3
+#define AFR_DOM_COUNT_MAX    3
+#define AFR_NUM_CHANGE_LOGS            3 /*data + metadata + entry*/
+
+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);
+
+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   */
+        gf_lock_t lock;               /* to guard access to child_count, etc */
+        unsigned int child_count;     /* total number of children   */
 
-        unsigned int read_child_rr;   /* round-robin index of the read_child */
-        gf_lock_t read_child_lock;    /* lock to protect above */
-        
-	xlator_t **children;
+        xlator_t **children;
 
-	unsigned char *child_up;
+        inode_t *root_inode;
+
+        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;
 
-	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 */
-} 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 */
-
-        mode_t mode;                      /* 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 stat *buf;
-        struct stat 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;
-	int *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;
+        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;
 
-	loc_t parent_loc;
+        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;
 
-        call_frame_t *orig_frame;
-        gf_boolean_t unwound;
+	afr_self_heald_t       shd;
 
-        /* private data for the particular self-heal algorithm */
-        void *private;
+	/* pump dependencies */
+	void                   *pump_private;
+	gf_boolean_t           use_afr_in_pump;
+} afr_private_t;
 
-        int (*flush_self_heal_cbk) (call_frame_t *frame, xlator_t *this);
 
-	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);
+typedef enum {
+        AFR_DATA_TRANSACTION,          /* truncate, write, ... */
+        AFR_METADATA_TRANSACTION,      /* chmod, chown, ... */
+        AFR_ENTRY_TRANSACTION,         /* create, rmdir, ... */
+        AFR_ENTRY_RENAME_TRANSACTION,  /* rename */
+} afr_transaction_type;
 
-	call_frame_t *sh_frame;
-} afr_self_heal_t;
+typedef enum {
+        AFR_TRANSACTION_LK,
+        AFR_SELFHEAL_LK,
+} transaction_lk_type_t;
 
+typedef enum {
+        AFR_LOCK_OP,
+        AFR_UNLOCK_OP,
+} afr_lock_op_type_t;
 
 typedef enum {
-	AFR_DATA_TRANSACTION,          /* truncate, write, ... */
-	AFR_METADATA_TRANSACTION,      /* chmod, chown, ... */
-	AFR_ENTRY_TRANSACTION,         /* create, rmdir, ... */
-	AFR_ENTRY_RENAME_TRANSACTION,  /* rename */
-	AFR_FLUSH_TRANSACTION,         /* flush */
-} afr_transaction_type;
+        AFR_DATA_SELF_HEAL_LK,
+        AFR_METADATA_SELF_HEAL_LK,
+        AFR_ENTRY_SELF_HEAL_LK,
+}selfheal_lk_type_t;
 
+typedef enum {
+        AFR_INODELK_TRANSACTION,
+        AFR_INODELK_NB_TRANSACTION,
+        AFR_ENTRYLK_TRANSACTION,
+        AFR_ENTRYLK_NB_TRANSACTION,
+        AFR_INODELK_SELFHEAL,
+        AFR_INODELK_NB_SELFHEAL,
+        AFR_ENTRYLK_SELFHEAL,
+        AFR_ENTRYLK_NB_SELFHEAL,
+} afr_lock_call_type_t;
 
 /*
   xattr format: trusted.afr.volume = [x y z]
@@ -175,9 +170,8 @@ static inline int
 afr_index_for_transaction_type (afr_transaction_type type)
 {
         switch (type) {
-                
+
         case AFR_DATA_TRANSACTION:
-        case AFR_FLUSH_TRANSACTION:
                 return 0;
 
         case AFR_METADATA_TRANSACTION:
@@ -191,560 +185,792 @@ 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;
+
+        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;
+        char lower_locked;
+        char higher_locked;
+
+        unsigned char *locked_nodes;
+        unsigned char *lower_locked_nodes;
+
+        selfheal_lk_type_t selfheal_lk_type;
+        transaction_lk_type_t transaction_lk_type;
+
+        int32_t 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;
+
+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];
+};
 
 typedef enum {
-        AFR_CHILD_UP_FLUSH,
-        AFR_CHILD_DOWN_FLUSH,
-} afr_flush_type;
+        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_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;
-	
-	dict_t  *xattr_req;
-	int      open_fd_count;
+	   pointer to function which will perform the read operation on a given
+	   subvolume. Used in read transactions.
+	*/
 
-	int32_t  inodelk_count;
-	int32_t  entrylk_count;
+	afr_read_txn_wind_t readfn;
 
-        int (*up_down_flush_cbk) (call_frame_t *, xlator_t *);
+	/* @refreshed:
 
-	/* 
-	   This struct contains the arguments for the "continuation"
-	   (scheme-like) of fops
+	   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;
 
-	int   op;
-	struct {
-		struct {
-			unsigned char buf_set;
-			struct statvfs buf;
-		} statfs;
+	/* @inode:
 
-		struct {
-			inode_t *inode;
-			struct stat buf;
-                        struct stat read_child_buf;
-                        struct stat postparent;
-                        ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			dict_t *xattr;
-			dict_t **xattrs;
-                        gf_boolean_t is_revalidate;
-		} lookup;
+	   the inode on which the read txn is performed on. ref'ed and copied
+	   from either fd->inode or loc.inode
+	*/
 
-		struct {
-			int32_t flags;
-                        int32_t wbflags;
-		} open;
+	inode_t *inode;
 
-		struct {
-			int32_t cmd;
-			struct flock flock;
-			unsigned char *locked_nodes;
-		} lk;
+	/* @parent[2]:
 
-		struct {
-			uint8_t *file_checksum;
-			uint8_t *dir_checksum;
-		} checksum;
+	   parent inode[s] on which directory transactions are performed.
+	*/
 
-		/* inode read */
+	inode_t *parent;
+	inode_t *parent2;
 
-		struct {
-			int32_t mask;
-			int last_tried;  /* index of the child we tried previously */
-		} access;
+	/* @readable:
 
-		struct {
-			int last_tried;
-			ino_t ino;
-		} stat;
+	   array of flags representing servers from which a read can be
+	   performed. This is the output of afr_inode_refresh()
+	*/
+	unsigned char *readable;
 
-		struct {
-			int last_tried;
-			ino_t ino;
-		} fstat;
+	afr_inode_refresh_cbk_t refreshfn;
 
-		struct {
-			size_t size;
-			int last_tried;
-                        ino_t ino;
-		} readlink;
+	/* @refreshinode:
 
-		struct {
-			const char *name;
-			int last_tried;
-		} getxattr;
+	   Inode currently getting refreshed.
+	*/
+	inode_t *refreshinode;
 
-		struct {
-                        ino_t ino;
-			size_t size;
-			off_t offset;
-			int last_tried;
-		} readv;
+	/*
+	  @pre_op_compat:
 
-		/* dir read */
+	  compatibility mode of pre-op. send a separate pre-op and
+	  op operations as part of transaction, rather than combining
+	*/
 
-		struct {
-			int success_count;
-			int32_t op_ret;
-			int32_t op_errno;
+	gf_boolean_t pre_op_compat;
 
-                        uint32_t *checksum;
-		} opendir;
+        dict_t  *xattr_req;
 
-		struct {
-			int32_t op_ret;
-			int32_t op_errno;
-			size_t size;
-			off_t offset;
+        afr_internal_lock_t internal_lock;
 
-			int last_tried;
-		} readdir;
+        dict_t  *dict;
 
-		struct {
-			int32_t op_ret;
-			int32_t op_errno;
+        int      optimistic_change_log;
+	gf_boolean_t      delayed_post_op;
 
-			size_t size;
-			off_t offset;
-			int32_t flag;
+	/* Is the current writev() going to perform a stable write?
+	   i.e, is fd->flags or @flags writev param have O_SYNC or
+	   O_DSYNC?
+	*/
+	gf_boolean_t      stable_write;
+
+	/* This write appended to the file. Nnot necessarily O_APPEND,
+	   just means the offset of write was at the end of file.
+	*/
+	gf_boolean_t      append_write;
 
-			int last_tried;
-		} getdents;
+        /*
+          This struct contains the arguments for the "continuation"
+          (scheme-like) of fops
+        */
 
-		/* inode write */
+        struct {
+                struct {
+                        unsigned char buf_set;
+                        struct statvfs buf;
+                } statfs;
 
-		struct {
-			ino_t ino;
-			struct stat prebuf;
-			struct stat postbuf;
+                struct {
+                        int32_t flags;
+                } open;
 
-			int32_t op_ret;
+                struct {
+                        int32_t cmd;
+                        struct gf_flock user_flock;
+                        struct gf_flock ret_flock;
+                        unsigned char *locked_nodes;
+                } lk;
 
-			struct iovec *vector;
-			struct iobref *iobref;
-			int32_t count;
-			off_t offset;
-		} writev;
+                /* inode read */
 
                 struct {
-                        ino_t ino;
-                        struct stat prebuf;
-                        struct stat postbuf;
-                } fsync;
+                        int32_t mask;
+                        int last_index;  /* index of the child we tried previously */
+                } access;
 
-		struct {
-			ino_t ino;
-			off_t offset;
-			struct stat prebuf;
-                        struct stat postbuf;
-		} truncate;
+                struct {
+                        int last_index;
+                } stat;
 
-		struct {
-			ino_t ino;
-			off_t offset;
-			struct stat prebuf;
-                        struct stat postbuf;
-		} ftruncate;
+                struct {
+                        int last_index;
+                } fstat;
 
-		struct {
-			ino_t ino;
-			struct stat in_buf;
+                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 {
+                        struct iatt prebuf;
+                        struct iatt postbuf;
+                } inode_wfop; //common structure for all inode-write-fops
+
+                struct {
+                        int32_t op_ret;
+
+                        struct iovec *vector;
+                        struct iobref *iobref;
+                        int32_t count;
+                        off_t offset;
+                        uint32_t flags;
+                } writev;
+
+                struct {
+                        off_t offset;
+                } truncate;
+
+                struct {
+                        off_t offset;
+                } ftruncate;
+
+                struct {
+                        struct iatt in_buf;
                         int32_t valid;
-                        struct stat preop_buf;
-                        struct stat postop_buf;
-		} setattr;
+                } setattr;
 
-		struct {
-			ino_t ino;
-			struct stat in_buf;
+                struct {
+                        struct iatt in_buf;
                         int32_t valid;
-                        struct stat preop_buf;
-                        struct stat postop_buf;
-		} fsetattr;
+                } fsetattr;
 
-		struct {
-			dict_t *dict;
-			int32_t flags;
-		} setxattr;
+                struct {
+                        dict_t *dict;
+                        int32_t flags;
+                } setxattr;
 
-		struct {
-			const char *name;
-		} removexattr;
+                struct {
+                        dict_t *dict;
+                        int32_t flags;
+                } fsetxattr;
 
-		/* dir write */
-		
-		struct {
-			ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			fd_t *fd;
-			int32_t flags;
-			mode_t mode;
-			inode_t *inode;
-			struct stat buf;
-                        struct stat preparent;
-                        struct stat postparent;
-                        struct stat read_child_buf;
-		} create;
+                struct {
+                        char *name;
+                } removexattr;
 
-		struct {
-			ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			dev_t dev;
-			mode_t mode;
-			inode_t *inode;
-			struct stat buf;
-                        struct stat preparent;
-                        struct stat postparent;
-                        struct stat read_child_buf;
-		} mknod;
+                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 prenewparent;
+                        struct iatt postnewparent;
+                } dir_fop; //common structure for all dir fops
+
+                struct {
+                        fd_t *fd;
+                        dict_t *params;
+                        int32_t flags;
+                        mode_t mode;
+                } create;
+
+                struct {
+                        dev_t dev;
+                        mode_t mode;
+                        dict_t *params;
+                } mknod;
+
+                struct {
+                        int32_t mode;
+                        dict_t *params;
+                } mkdir;
+
+                struct {
+                        int flags;
+                } rmdir;
+
+                struct {
+                        dict_t *params;
+                        char *linkpath;
+                } symlink;
 
 		struct {
-			ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
 			int32_t mode;
-			inode_t *inode;
-			struct stat buf;
-                        struct stat read_child_buf;
-                        struct stat preparent;
-                        struct stat postparent;
-		} mkdir;
+			off_t offset;
+			size_t len;
+		} fallocate;
 
 		struct {
-                        ino_t parent_ino;
-			int32_t op_ret;
-			int32_t op_errno;
-                        struct stat preparent;
-                        struct stat postparent;
-		} unlink;
+			off_t offset;
+			size_t len;
+		} discard;
 
-		struct {
-                        ino_t parent_ino;
-			int32_t op_ret;
-			int32_t op_errno;
-                        struct stat preparent;
-                        struct stat postparent;
-		} rmdir;
+                struct {
+                        off_t offset;
+                        off_t len;
+                        struct iatt prebuf;
+                        struct iatt postbuf;
+                } zerofill;
 
-		struct {
-                        ino_t oldparent_ino;
-                        ino_t newparent_ino;
-			ino_t ino;
-			struct stat buf;
-                        struct stat read_child_buf;
-                        struct stat preoldparent;
-                        struct stat prenewparent;
-                        struct stat postoldparent;
-                        struct stat postnewparent;
-		} rename;
 
-		struct {
-			ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			inode_t *inode;
-			struct stat buf;
-                        struct stat read_child_buf;
-                        struct stat preparent;
-                        struct stat postparent;
-		} link;
+        } cont;
 
-		struct {
-			ino_t ino;
-                        uint64_t gen;
-                        ino_t parent_ino;
-			inode_t *inode;
-			struct stat buf;
-                        struct stat read_child_buf;
-			char *linkpath;
-                        struct stat preparent;
-                        struct stat postparent;
-		} symlink;
+        struct {
+                off_t start, len;
 
-		struct {
-			int32_t flags;
-			dir_entry_t *entries;
-			int32_t count;
-		} setdents;
-	} cont;
-	
-	struct {
-		off_t start, len;
+                gf_boolean_t    eager_lock_on;
+                int *eager_lock;
+
+                char *basename;
+                char *new_basename;
+
+                loc_t parent_loc;
+                loc_t new_parent_loc;
+
+                afr_transaction_type type;
+
+		/* stub to resume on destruction
+		   of the transaction frame */
+		call_stub_t      *resume_stub;
 
-		unsigned char *locked_nodes;
-		int lock_count;
+		struct list_head  eager_locked;
 
-		const char *basename;
-		const char *new_basename;
+                unsigned char   *pre_op;
 
-		loc_t parent_loc;
-		loc_t new_parent_loc;
+		/* @fop_subvols: subvolumes on which FOP will be attempted */
+                unsigned char   *fop_subvols;
 
-		afr_transaction_type type;
+		/* @failed_subvols: subvolumes on which FOP failed. Always
+		   a subset of @fop_subvols */
+                unsigned char   *failed_subvols;
 
-		int success_count;
-		int erase_pending;
-		int failure_count;
+		/* @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
 
-		int last_tried;
-		int32_t *child_errno;
+		   A 'true' value set in @dirtied flag means an 'undirtying'
+		   has to be done in POST-OP phase.
+		*/
+		gf_boolean_t  dirtied;
 
-		call_frame_t *main_frame;
+		/* @inherited: flag which indicates that the dirty flags
+		   of the previous transaction were inherited
+		*/
+		gf_boolean_t  inherited;
 
-		int (*fop) (call_frame_t *frame, xlator_t *this);
+		/*
+		  @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;
 
-		int (*done) (call_frame_t *frame, xlator_t *this);
+		/* @uninherit_done:
+		   @uninherit_value:
 
-		int (*resume) (call_frame_t *frame, xlator_t *this);
+		   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;
 
-		int (*unwind) (call_frame_t *frame, xlator_t *this);
+		/* @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 */
-                int (*post_post_op) (call_frame_t *frame, xlator_t *this);
-	} transaction;
+        } transaction;
+
+	syncbarrier_t barrier;
 
-	afr_self_heal_t self_heal;
+        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;
 
 
-typedef struct {
-        unsigned char *pre_op_done;
-        unsigned char *opened_on;     /* which subvolumes the fd is open on */
-        unsigned char *child_failed;
-        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 */
-} afr_fd_ctx_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)))
+
+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);
 
-/* try alloc and if it fails, goto label */
-#define ALLOC_OR_GOTO(var, type, label) do {			\
-		var = CALLOC (sizeof (type), 1);		\
-		if (!var) {					\
-			gf_log (this->name, GF_LOG_ERROR,	\
-				"out of memory :(");		\
-			op_errno = ENOMEM;			\
-			goto label;				\
-		}						\
-	} while (0);
+int
+afr_inode_read_subvol_reset (inode_t *inode, xlator_t *this);
 
+int
+afr_read_subvol_select_by_policy (inode_t *inode, xlator_t *this,
+				  unsigned char *readable);
 
-/* 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)))
+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_fop_failed(op_ret, op_errno) ((op_ret) == -1)
+#define afr_data_subvol_get(i, t, s, e) \
+	afr_read_subvol_get(i, t, s, e, AFR_DATA_TRANSACTION)
 
-/* have we tried all children? */
-#define all_tried(i, count)  ((i) == (count) - 1)
+#define afr_metadata_subvol_get(i, t, s, e) \
+	afr_read_subvol_get(i, t, s, e, AFR_METADATA_TRANSACTION)
 
 int
-afr_fd_ctx_set (xlator_t *this, fd_t *fd);
+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, void *data2);
 
-uint64_t
-afr_read_child (xlator_t *this, inode_t *inode);
+int
+afr_init_entry_lockee (afr_entry_lockee_t *lockee, afr_local_t *local,
+                       loc_t *loc, char *basename, int child_count);
 
 void
-afr_set_read_child (xlator_t *this, inode_t *inode, int32_t read_child);
+afr_entry_lockee_cleanup (afr_internal_lock_t *int_lock);
+
+int
+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_build_parent_loc (loc_t *parent, loc_t *child);
+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);
+
+int32_t
+afr_unlock (call_frame_t *frame, xlator_t *this);
 
 int
-afr_up_children_count (int child_count, unsigned char *child_up);
+afr_nonblocking_entrylk (call_frame_t *frame, xlator_t *this);
 
 int
-afr_locked_nodes_count (unsigned char *locked_nodes, int child_count);
+afr_nonblocking_inodelk (call_frame_t *frame, xlator_t *this);
 
-ino64_t
-afr_itransform (ino64_t ino, int child_count, int child_index);
+int
+afr_blocking_lock (call_frame_t *frame, xlator_t *this);
 
 int
-afr_deitransform (ino64_t ino, int child_count);
+afr_internal_lock_finish (call_frame_t *frame, xlator_t *this);
 
-void
-afr_local_cleanup (afr_local_t *local, xlator_t *this);
+int
+afr_lk_transfer_datalock (call_frame_t *dst, call_frame_t *src, char *dom,
+                          unsigned int child_count);
 
 int
-afr_frame_return (call_frame_t *frame);
+__afr_fd_ctx_set (xlator_t *this, fd_t *fd);
 
-uint64_t
-afr_is_split_brain (xlator_t *this, inode_t *inode);
+int
+afr_fd_ctx_set (xlator_t *this, fd_t *fd);
+
+afr_fd_ctx_t *
+afr_fd_ctx_get (fd_t *fd, xlator_t *this);
+
+int
+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);
+
+int
+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_set_split_brain (xlator_t *this, inode_t *inode);
+afr_local_cleanup (afr_local_t *local, xlator_t *this);
 
 int
-afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
-          fd_t *fd, int32_t wbflags);
+afr_frame_return (call_frame_t *frame);
 
 int
-afr_up_down_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, afr_flush_type type);
+afr_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+          fd_t *fd, dict_t *xdata);
 
 void
-afr_set_opendir_done (xlator_t *this, inode_t *inode);
-
-uint64_t
-afr_is_opendir_done (xlator_t *this, inode_t *inode);
-
-#define AFR_STACK_UNWIND(fop, frame, params ...)        \
-	do {						\
-		afr_local_t *__local = NULL;		\
-		xlator_t    *__this = NULL;		\
-		__local = frame->local;			\
-		__this = frame->this;			\
-		frame->local = NULL;                    \
-		STACK_UNWIND_STRICT (fop, frame, params);       \
-		afr_local_cleanup (__local, __this);	\
-		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);	\
-		free (__local);				\
-} while (0);					
+afr_local_transaction_cleanup (afr_local_t *local, xlator_t *this);
+
+int
+afr_cleanup_fd_ctx (xlator_t *this, fd_t *fd);
+
+#define AFR_STACK_UNWIND(fop, frame, params ...)                \
+        do {                                                    \
+                afr_local_t *__local = NULL;                    \
+                xlator_t    *__this = NULL;                     \
+                if (frame) {                                    \
+                        __local = frame->local;                 \
+                        __this = frame->this;                   \
+                        frame->local = NULL;                    \
+                }                                               \
+                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);                      \
+                }                                               \
+        } 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)						
+static inline char *
+AFR_BASENAME (const char *str)
 {
-	char *__tmp_str = NULL;				
-	char *__basename_str = NULL;			
-	__tmp_str = strdup (str);			
-	__basename_str = strdup (basename (__tmp_str));	
-	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 = CALLOC (sizeof (*local->child_up),
-				  priv->child_count);
-	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);
 
-	return 0;
-}
+int
+afr_final_errno (afr_local_t *local, afr_private_t *priv);
+
+int
+afr_xattr_req_prepare (xlator_t *this, dict_t *xattr_req);
+
+void
+afr_fix_open (fd_t *fd, xlator_t *this);
+
+afr_fd_ctx_t *
+afr_fd_ctx_get (fd_t *fd, xlator_t *this);
+
+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);
+
+gf_boolean_t
+afr_have_quorum (char *logname, afr_private_t *priv);
+
+void
+afr_matrix_cleanup (int32_t **pending, unsigned int m);
 
+int32_t**
+afr_matrix_create (unsigned int m, unsigned int n);
 
-/**
- * first_up_child - return the index of the first child that is up
+void
+afr_filter_xattrs (dict_t *xattr);
+
+/*
+ * 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
 
-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;
-}
+/*
+ * 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);
 
+int
+afr_fd_report_unstable_write (xlator_t *this, fd_t *fd);
 
-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 = CALLOC (sizeof (*local->child_errno),
-				     priv->child_count);
-	if (!local->child_errno) {
-		return -ENOMEM;
-	}
-
-	local->pending = CALLOC (sizeof (*local->pending),
-                                 priv->child_count);
-        
-	if (!local->pending) {
-		return -ENOMEM;
-	}
-
-        for (i = 0; i < priv->child_count; i++) {
-                local->pending[i] = CALLOC (sizeof (*local->pending[i]),
-                                            3); /* data + metadata + entry */
-                if (!local->pending[i])
-                        return -ENOMEM;
-        }
-        
-	local->transaction.locked_nodes = CALLOC (sizeof (*local->transaction.locked_nodes),
-						  priv->child_count);
+gf_boolean_t
+afr_fd_has_witnessed_unstable_write (xlator_t *this, fd_t *fd);
 
-	local->transaction.child_errno = CALLOC (sizeof (*local->transaction.child_errno),
-						  priv->child_count);
+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
new file mode 100644
index 000000000..eed509956
--- /dev/null
+++ b/xlators/cluster/afr/src/pump.c
@@ -0,0 +1,2473 @@
+/*
+  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
+#include "config.h"
+#endif
+
+#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)
+{
+        afr_private_t  *priv      = NULL;
+        pump_private_t *pump_priv = NULL;
+
+        priv      = this->private;
+        pump_priv = priv->pump_private;
+
+        pump_priv->pump_start_pending = 1;
+
+        return 0;
+}
+
+static int
+is_pump_start_pending (xlator_t *this)
+{
+        afr_private_t  *priv      = NULL;
+        pump_private_t *pump_priv = NULL;
+
+        priv      = this->private;
+        pump_priv = priv->pump_private;
+
+        return (pump_priv->pump_start_pending);
+}
+
+static int
+pump_remove_start_pending (xlator_t *this)
+{
+        afr_private_t  *priv      = NULL;
+        pump_private_t *pump_priv = NULL;
+
+        priv      = this->private;
+        pump_priv = priv->pump_private;
+
+        pump_priv->pump_start_pending = 0;
+
+        return 0;
+}
+
+static pump_state_t
+pump_get_state ()
+{
+        xlator_t *this = NULL;
+        afr_private_t *priv = NULL;
+        pump_private_t *pump_priv = NULL;
+
+        pump_state_t ret;
+
+        this = THIS;
+        priv = this->private;
+        pump_priv = priv->pump_private;
+
+        LOCK (&pump_priv->pump_state_lock);
+        {
+                ret = pump_priv->pump_state;
+        }
+        UNLOCK (&pump_priv->pump_state_lock);
+
+        return ret;
+}
+
+int
+pump_change_state (xlator_t *this, pump_state_t state)
+{
+        afr_private_t *priv = NULL;
+        pump_private_t *pump_priv = NULL;
+
+        pump_state_t state_old;
+        pump_state_t state_new;
+
+
+        priv = this->private;
+        pump_priv = priv->pump_private;
+
+        GF_ASSERT (pump_priv);
+
+        LOCK (&pump_priv->pump_state_lock);
+        {
+                state_old = pump_priv->pump_state;
+                state_new = state;
+
+                pump_priv->pump_state = state;
+
+        }
+        UNLOCK (&pump_priv->pump_state_lock);
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Pump changing state from %d to %d",
+                state_old,
+                state_new);
+
+        return  0;
+}
+
+static int
+pump_set_resume_path (xlator_t *this, const char *path)
+{
+        int ret = 0;
+
+        afr_private_t *priv = NULL;
+        pump_private_t *pump_priv = NULL;
+
+        priv = this->private;
+        pump_priv = priv->pump_private;
+
+        GF_ASSERT (pump_priv);
+
+        LOCK (&pump_priv->resume_path_lock);
+        {
+                strncpy (pump_priv->resume_path, path, strlen (path) + 1);
+        }
+        UNLOCK (&pump_priv->resume_path_lock);
+
+        return ret;
+}
+
+static int
+pump_save_path (xlator_t *this, const char *path)
+{
+        afr_private_t *priv = NULL;
+        pump_state_t state;
+        dict_t *dict = NULL;
+        loc_t  loc = {0};
+        int dict_ret = 0;
+        int ret = -1;
+
+        state = pump_get_state ();
+        if (state == PUMP_STATE_RESUME)
+                return 0;
+
+        priv = this->private;
+
+        GF_ASSERT (priv->root_inode);
+
+        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_INFO,
+                        "setxattr failed - could not save path=%s", path);
+        } else {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "setxattr succeeded - saved path=%s", path);
+        }
+
+        dict_unref (dict);
+
+        loc_wipe (&loc);
+        return 0;
+}
+
+static int
+pump_check_and_update_status (xlator_t *this)
+{
+        pump_state_t state;
+        int ret = -1;
+
+        state = pump_get_state ();
+
+        switch (state) {
+
+        case PUMP_STATE_RESUME:
+        case PUMP_STATE_RUNNING:
+        {
+                ret = 0;
+                break;
+        }
+        case PUMP_STATE_PAUSE:
+        {
+                ret = -1;
+                break;
+        }
+        case PUMP_STATE_ABORT:
+        {
+                pump_save_path (this, "/");
+                ret = -1;
+                break;
+        }
+        default:
+        {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Unknown pump state");
+                ret = -1;
+                break;
+        }
+
+        }
+
+        return ret;
+}
+
+static const char *
+pump_get_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_priv->resume_path;
+
+        return resume_path;
+}
+
+static int
+pump_update_resume_state (xlator_t *this, const char *path)
+{
+        pump_state_t state;
+        const char *resume_path = NULL;
+
+        state = pump_get_state ();
+
+        if (state == PUMP_STATE_RESUME) {
+                resume_path = pump_get_resume_path (this);
+                if (strcmp (resume_path, "/") == 0) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "Reached the resume path (/). Proceeding to change state"
+                                " to running");
+                        pump_change_state (this, PUMP_STATE_RUNNING);
+                } else if (strcmp (resume_path, path) == 0) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "Reached the resume path. Proceeding to change state"
+                                " to running");
+                        pump_change_state (this, PUMP_STATE_RUNNING);
+                } else {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "Not yet hit the resume path:res-path=%s,path=%s",
+                                resume_path, path);
+                }
+        }
+
+        return 0;
+}
+
+static gf_boolean_t
+is_pump_traversal_allowed (xlator_t *this, const char *path)
+{
+        pump_state_t state;
+        const char *resume_path = NULL;
+        gf_boolean_t ret = _gf_true;
+
+        state = pump_get_state ();
+
+        if (state == PUMP_STATE_RESUME) {
+                resume_path = pump_get_resume_path (this);
+                if (strstr (resume_path, path)) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "On the right path to resumption path");
+                        ret = _gf_true;
+                } else {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "Not the right path to resuming=> ignoring traverse");
+                        ret = _gf_false;
+                }
+        }
+
+        return ret;
+}
+
+static int
+pump_save_file_stats (xlator_t *this, const char *path)
+{
+        afr_private_t  *priv        = NULL;
+        pump_private_t *pump_priv   = NULL;
+
+        priv      = this->private;
+        pump_priv = priv->pump_private;
+
+        LOCK (&pump_priv->resume_path_lock);
+        {
+                pump_priv->number_files_pumped++;
+
+                strncpy (pump_priv->current_file, path,
+                         PATH_MAX);
+        }
+        UNLOCK (&pump_priv->resume_path_lock);
+
+        return 0;
+}
+
+static int
+gf_pump_traverse_directory (loc_t *loc)
+{
+        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;
+
+        GF_ASSERT (loc->inode);
+
+	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 (this, loc, fd);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "opendir failed on %s", loc->path);
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_TRACE,
+                "pump opendir on %s returned=%d",
+                loc->path, ret);
+
+        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,
+                                "no more entries in directory");
+                        goto out;
+                }
+
+                list_for_each_entry_safe (entry, tmp, &entries.list, list) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "found readdir entry=%s", entry->d_name);
+
+                        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;
+
+			if ((strcmp (entry->d_name, ".") == 0) ||
+			    (strcmp (entry->d_name, "..") == 0))
+				continue;
+
+			is_directory_empty = _gf_false;
+			gf_log (this->name, GF_LOG_DEBUG,
+				"lookup %s => %"PRId64,
+				entry_loc.path,
+				iatt.ia_ino);
+
+			ret = syncop_lookup (this, &entry_loc, NULL, &iatt,
+					     &xattr_rsp, &parent);
+
+			if (ret) {
+				gf_log (this->name, GF_LOG_ERROR,
+					"%s: lookup failed", entry_loc.path);
+				continue;
+			}
+
+			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;
+			}
+
+			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_fill_loc_info (&entry_loc, &iatt, &parent);
+
+			pump_update_resume_state (this, entry_loc.path);
+
+			pump_save_path (this, entry_loc.path);
+			pump_save_file_stats (this, entry_loc.path);
+
+			ret = pump_check_and_update_status (this);
+			if (ret < 0) {
+				gf_log (this->name, GF_LOG_DEBUG,
+					"Pump beginning to exit out");
+				goto out;
+			}
+
+			if (IA_ISDIR (iatt.ia_type)) {
+				if (is_pump_traversal_allowed (this, entry_loc.path)) {
+					gf_log (this->name, GF_LOG_TRACE,
+						"entering dir=%s", entry->d_name);
+					gf_pump_traverse_directory (&entry_loc);
+				}
+                        }
+                }
+
+                gf_dirent_free (&entries);
+                free_entries = _gf_false;
+                gf_log (this->name, GF_LOG_TRACE, "offset incremented to %d",
+                        (int32_t ) offset);
+
+        }
+
+        ret = syncop_close (fd);
+        if (ret < 0)
+                gf_log (this->name, GF_LOG_DEBUG, "closing the fd failed");
+
+        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)
+{
+        const char *resume_path = NULL;
+
+        resume_path = pump_get_resume_path (this);
+
+        if (resume_path) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Found a path to resume from: %s",
+                        resume_path);
+
+        }else {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Did not find a path=> setting to '/'");
+                pump_set_resume_path (this, "/");
+        }
+
+        pump_change_state (this, PUMP_STATE_RESUME);
+
+        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)
+{
+        afr_private_t *priv = NULL;
+        pump_private_t *pump_priv = NULL;
+        dict_t *dict = NULL;
+        pump_state_t state;
+        loc_t  loc = {0};
+        int dict_ret = 0;
+        int ret = -1;
+
+        priv = this->private;
+        pump_priv = priv->pump_private;
+
+        GF_ASSERT (priv->root_inode);
+
+        afr_build_root_loc (this, &loc);
+
+        dict = dict_new ();
+
+        state = pump_get_state ();
+        if (state == PUMP_STATE_RUNNING) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Pump finished pumping");
+
+                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) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "setxattr failed - while  notifying source complete");
+                }
+                dict_ret = dict_set_str (dict, PUMP_SINK_COMPLETE, "jargon");
+                if (dict_ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to set the key %s",
+                                loc.path, PUMP_SINK_COMPLETE);
+
+                ret = syncop_setxattr (PUMP_SINK_CHILD (this), &loc, dict, 0);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "setxattr failed - while notifying sink complete");
+                }
+
+                pump_save_path (this, "/");
+
+        } else if (state == PUMP_STATE_ABORT) {
+                gf_log (this->name, GF_LOG_DEBUG, "Starting cleanup "
+                        "of pump internal xattrs");
+                call_resume (pump_priv->cleaner);
+        }
+
+        loc_wipe (&loc);
+        return 0;
+}
+
+static int
+pump_lookup_sink (loc_t *loc)
+{
+        xlator_t *this = NULL;
+	struct iatt iatt, parent;
+	dict_t *xattr_rsp;
+        dict_t *xattr_req = NULL;
+        int ret = 0;
+
+        this = THIS;
+
+        xattr_req = dict_new ();
+
+        ret = afr_set_root_gfid (xattr_req);
+        if (ret)
+                goto out;
+
+        ret = syncop_lookup (PUMP_SINK_CHILD (this), loc,
+                             xattr_req, &iatt, &xattr_rsp, &parent);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Lookup on sink child failed");
+                ret = -1;
+                goto out;
+        }
+
+out:
+        if (xattr_req)
+                dict_unref (xattr_req);
+
+        return ret;
+}
+
+static int
+pump_task (void *data)
+{
+	xlator_t *this = NULL;
+        afr_private_t *priv = NULL;
+
+
+        loc_t loc = {0};
+	struct iatt iatt, parent;
+	dict_t *xattr_rsp = NULL;
+        dict_t *xattr_req = NULL;
+
+        int ret = -1;
+
+        this = THIS;
+        priv = this->private;
+
+        GF_ASSERT (priv->root_inode);
+
+        afr_build_root_loc (this, &loc);
+        xattr_req = dict_new ();
+        if (!xattr_req) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Out of memory");
+                ret = -1;
+                goto out;
+        }
+
+        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: path=%s gfid=%s",
+                loc.path, uuid_utoa (loc.inode->gfid));
+
+        ret = pump_check_and_update_status (this);
+        if (ret < 0) {
+                goto out;
+        }
+
+        pump_update_resume_path (this);
+
+        afr_set_root_gfid (xattr_req);
+        ret = pump_lookup_sink (&loc);
+        if (ret) {
+                pump_update_resume_path (this);
+                goto out;
+        }
+
+        gf_pump_traverse_directory (&loc);
+
+        pump_complete_migration (this);
+out:
+        if (xattr_req)
+                dict_unref (xattr_req);
+
+        loc_wipe (&loc);
+	return 0;
+}
+
+
+static int
+pump_task_completion (int ret, call_frame_t *sync_frame, void *data)
+{
+        xlator_t *this = NULL;
+        afr_private_t *priv = NULL;
+
+        this = THIS;
+
+        priv = this->private;
+
+        inode_unref (priv->root_inode);
+        STACK_DESTROY (sync_frame->root);
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Pump xlator exiting");
+	return 0;
+}
+
+int
+pump_start (call_frame_t *pump_frame, xlator_t *this)
+{
+	afr_private_t *priv = NULL;
+	pump_private_t *pump_priv = NULL;
+
+	int ret = -1;
+
+	priv = this->private;
+        pump_priv = priv->pump_private;
+
+        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, NULL);
+        if (ret == -1) {
+                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_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:
+	return ret;
+}
+
+static int
+pump_start_synctask (xlator_t *this)
+{
+        call_frame_t *frame = NULL;
+        int ret = 0;
+
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Out of memory");
+                ret = -1;
+                goto out;
+        }
+
+        pump_change_state (this, PUMP_STATE_RUNNING);
+
+        ret = pump_start (frame, this);
+
+out:
+        return ret;
+}
+
+int32_t
+pump_cmd_start_setxattr_cbk (call_frame_t *frame,
+                             void *cookie,
+                             xlator_t *this,
+                             int32_t op_ret,
+                             int32_t op_errno, dict_t *xdata)
+
+{
+        call_frame_t *prev = NULL;
+        afr_local_t *local = NULL;
+        int ret = 0;
+
+        local = frame->local;
+
+        if (op_ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Could not initiate destination "
+                        "brick connect");
+                ret = op_ret;
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Successfully initiated destination "
+                "brick connect");
+
+        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);
+
+        return 0;
+}
+
+static int
+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;
+        data_t        *data      = NULL;
+        char          *clnt_cmd  = NULL;
+        loc_t loc = {0};
+
+        int ret = 0;
+
+        priv  = this->private;
+        local = frame->local;
+
+        GF_ASSERT (priv->root_inode);
+
+        afr_build_root_loc (this, &loc);
+
+        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;
+        }
+
+        dict = dict_new ();
+        if (!dict) {
+                ret = -1;
+                goto out;
+        }
+
+        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_dynstr (dict, CLIENT_CMD_CONNECT, clnt_cmd);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Could not inititiate destination brick "
+                        "connect");
+                goto out;
+        }
+
+	STACK_WIND (frame,
+		    pump_cmd_start_setxattr_cbk,
+		    PUMP_SINK_CHILD(this),
+		    PUMP_SINK_CHILD(this)->fops->setxattr,
+		    &loc,
+		    dict,
+		    0, NULL);
+
+        ret = 0;
+
+out:
+        if (dict)
+                dict_unref (dict);
+
+        if (data)
+                data_unref (data);
+
+        if (ret && clnt_cmd)
+                GF_FREE (clnt_cmd);
+
+        loc_wipe (&loc);
+        return ret;
+}
+
+static int
+is_pump_aborted (xlator_t *this)
+{
+        pump_state_t state;
+
+        state = pump_get_state ();
+
+        return ((state == PUMP_STATE_ABORT));
+}
+
+int32_t
+pump_cmd_start_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;
+        char *path = NULL;
+
+        pump_state_t state;
+        int ret = 0;
+        int need_unwind = 0;
+        int dict_ret = -1;
+
+        local = frame->local;
+
+        if (op_ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "getxattr failed - changing pump "
+                        "state to RUNNING with '/'");
+                path = "/";
+                ret = op_ret;
+        } else {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "getxattr succeeded");
+
+                dict_ret =  dict_get_str (dict, PUMP_PATH, &path);
+                if (dict_ret < 0)
+                        path = "/";
+        }
+
+        state = pump_get_state ();
+        if ((state == PUMP_STATE_RUNNING) ||
+            (state == PUMP_STATE_RESUME)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Pump is already started");
+                ret = -1;
+                goto out;
+        }
+
+        pump_set_resume_path (this, path);
+
+        if (is_pump_aborted (this))
+                /* We're re-starting pump afresh */
+                ret = pump_initiate_sink_connect (frame, this);
+        else {
+                /* We're re-starting pump from a previous
+                   pause */
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "about to start synctask");
+                ret = pump_start_synctask (this);
+                need_unwind = 1;
+        }
+
+out:
+        if ((ret < 0) || (need_unwind == 1)) {
+                local->op_ret = ret;
+                pump_command_reply (frame, this);
+        }
+	return 0;
+}
+
+int
+pump_execute_status (call_frame_t *frame, xlator_t *this)
+{
+        afr_private_t *priv = NULL;
+        pump_private_t *pump_priv = NULL;
+
+        uint64_t number_files = 0;
+
+        char filename[PATH_MAX];
+        char summary[PATH_MAX+256];
+        char *dict_str = NULL;
+
+        int32_t op_ret = 0;
+        int32_t op_errno = 0;
+
+        dict_t *dict = NULL;
+        int ret = -1;
+
+        priv = this->private;
+        pump_priv = priv->pump_private;
+
+        LOCK (&pump_priv->resume_path_lock);
+        {
+                number_files  = pump_priv->number_files_pumped;
+                strncpy (filename, pump_priv->current_file, PATH_MAX);
+        }
+        UNLOCK (&pump_priv->resume_path_lock);
+
+        dict_str     = GF_CALLOC (1, PATH_MAX + 256, gf_afr_mt_char);
+        if (!dict_str) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Out of memory");
+                op_ret = -1;
+                op_errno = ENOMEM;
+                goto out;
+        }
+
+        if (pump_priv->pump_finished) {
+                snprintf (summary, PATH_MAX+256,
+                          "no_of_files=%"PRIu64, number_files);
+        } else {
+                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_dynstr (dict, RB_PUMP_CMD_STATUS, dict_str);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "dict_set_dynstr returned negative value");
+        } else {
+                dict_str = NULL;
+        }
+
+        op_ret = 0;
+
+out:
+
+        AFR_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, NULL);
+
+        if (dict)
+                dict_unref (dict);
+
+        GF_FREE (dict_str);
+
+        return 0;
+}
+
+int
+pump_execute_pause (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *local = NULL;
+
+        local = frame->local;
+
+        pump_change_state (this, PUMP_STATE_PAUSE);
+
+        local->op_ret = 0;
+        pump_command_reply (frame, this);
+
+        return 0;
+}
+
+int
+pump_execute_start (call_frame_t *frame, xlator_t *this)
+{
+        afr_private_t *priv = NULL;
+        afr_local_t   *local = NULL;
+
+        int ret = 0;
+        loc_t loc = {0};
+
+        priv = this->private;
+        local = frame->local;
+
+        if (!priv->root_inode) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Pump xlator cannot be started without an initial "
+                        "lookup");
+                ret = -1;
+                goto out;
+        }
+
+        GF_ASSERT (priv->root_inode);
+
+        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, NULL);
+
+        ret = 0;
+
+out:
+        if (ret < 0) {
+                local->op_ret = ret;
+                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;
+        call_frame_t   *sync_frame = NULL;
+        int             ret        = 0;
+
+        priv      = this->private;
+        pump_priv = priv->pump_private;
+        local     = frame->local;
+
+        pump_change_state (this, PUMP_STATE_ABORT);
+
+        LOCK (&pump_priv->resume_path_lock);
+        {
+                pump_priv->number_files_pumped = 0;
+                pump_priv->current_file[0] = '\0';
+        }
+        UNLOCK (&pump_priv->resume_path_lock);
+
+        local->op_ret = 0;
+        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;
+}
+
+gf_boolean_t
+pump_command_status (xlator_t *this, dict_t *dict)
+{
+        char *cmd = NULL;
+        int dict_ret = -1;
+        int ret = _gf_true;
+
+        dict_ret = dict_get_str (dict, RB_PUMP_CMD_STATUS, &cmd);
+        if (dict_ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Not a pump status command");
+                ret = _gf_false;
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Hit a pump command - status");
+        ret = _gf_true;
+
+out:
+        return ret;
+
+}
+
+gf_boolean_t
+pump_command_pause (xlator_t *this, dict_t *dict)
+{
+        char *cmd = NULL;
+        int dict_ret = -1;
+        int ret = _gf_true;
+
+        dict_ret = dict_get_str (dict, RB_PUMP_CMD_PAUSE, &cmd);
+        if (dict_ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Not a pump pause command");
+                ret = _gf_false;
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Hit a pump command - pause");
+        ret = _gf_true;
+
+out:
+        return ret;
+
+}
+
+gf_boolean_t
+pump_command_commit (xlator_t *this, dict_t *dict)
+{
+        char *cmd = NULL;
+        int dict_ret = -1;
+        int ret = _gf_true;
+
+        dict_ret = dict_get_str (dict, RB_PUMP_CMD_COMMIT, &cmd);
+        if (dict_ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Not a pump commit command");
+                ret = _gf_false;
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Hit a pump command - commit");
+        ret = _gf_true;
+
+out:
+        return ret;
+
+}
+
+gf_boolean_t
+pump_command_abort (xlator_t *this, dict_t *dict)
+{
+        char *cmd = NULL;
+        int dict_ret = -1;
+        int ret = _gf_true;
+
+        dict_ret = dict_get_str (dict, RB_PUMP_CMD_ABORT, &cmd);
+        if (dict_ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Not a pump abort command");
+                ret = _gf_false;
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Hit a pump command - abort");
+        ret = _gf_true;
+
+out:
+        return ret;
+
+}
+
+gf_boolean_t
+pump_command_start (xlator_t *this, dict_t *dict)
+{
+        char *cmd = NULL;
+        int dict_ret = -1;
+        int ret = _gf_true;
+
+        dict_ret = dict_get_str (dict, RB_PUMP_CMD_START, &cmd);
+        if (dict_ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Not a pump start command");
+                ret = _gf_false;
+                goto out;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Hit a pump command - start");
+        ret = _gf_true;
+
+out:
+        return ret;
+
+}
+
+int
+pump_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+	       const char *name, dict_t *xdata)
+{
+	afr_private_t *priv = NULL;
+	int op_errno = 0;
+	int ret = 0;
+
+	priv = this->private;
+
+        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,
+                              strlen (AFR_XATTR_PREFIX))) {
+
+                        op_errno = ENODATA;
+                        goto out;
+                }
+
+                if (!strcmp (name, RB_PUMP_CMD_STATUS)) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "Hit pump command - status");
+                        pump_execute_status (frame, this);
+                        ret = 0;
+                        goto out;
+                }
+        }
+
+	afr_getxattr (frame, this, loc, name, xdata);
+
+	ret = 0;
+out:
+	if (ret < 0)
+		AFR_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
+	return 0;
+}
+
+int
+pump_command_reply (call_frame_t *frame, xlator_t *this)
+{
+        afr_local_t *local = NULL;
+
+        local = frame->local;
+
+        if (local->op_ret < 0)
+                gf_log (this->name, GF_LOG_INFO,
+                        "Command failed");
+        else
+                gf_log (this->name, GF_LOG_INFO,
+                        "Command succeeded");
+
+        AFR_STACK_UNWIND (setxattr,
+                          frame,
+                          local->op_ret,
+                          local->op_errno, NULL);
+
+        return 0;
+}
+
+int
+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)) {
+                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)) {
+		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)) {
+		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, dict_t *xdata)
+{
+	afr_private_t *priv = NULL;
+	int ret = -1;
+	int op_errno = 0;
+
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.glusterfs.pump*", dict, op_errno, out);
+
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame, default_setxattr_cbk,
+                            FIRST_CHILD (this),
+                            (FIRST_CHILD (this))->fops->setxattr,
+                            loc, dict, flags, xdata);
+                return 0;
+        }
+
+        ret = pump_parse_command (frame, this, dict, &op_errno);
+        if (ret >= 0)
+                goto out;
+
+	afr_setxattr (frame, this, loc, dict, flags, xdata);
+
+	ret = 0;
+out:
+	if (ret < 0) {
+		AFR_STACK_UNWIND (setxattr, frame, -1, op_errno, NULL);
+	}
+
+	return 0;
+}
+
+/* Defaults */
+static int32_t
+pump_lookup (call_frame_t *frame,
+             xlator_t *this,
+             loc_t *loc,
+             dict_t *xattr_req)
+{
+	afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_lookup_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->lookup,
+                            loc,
+                            xattr_req);
+                return 0;
+        }
+
+        afr_lookup (frame, this, loc, xattr_req);
+        return 0;
+}
+
+
+static int32_t
+pump_truncate (call_frame_t *frame,
+               xlator_t *this,
+               loc_t *loc,
+               off_t offset, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_truncate_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->truncate,
+                            loc,
+                            offset, xdata);
+                return 0;
+        }
+
+        afr_truncate (frame, this, loc, offset, xdata);
+        return 0;
+}
+
+
+static int32_t
+pump_ftruncate (call_frame_t *frame,
+                xlator_t *this,
+                fd_t *fd,
+                off_t offset, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_ftruncate_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->ftruncate,
+                            fd,
+                            offset, xdata);
+                return 0;
+        }
+
+        afr_ftruncate (frame, this, fd, offset, xdata);
+        return 0;
+}
+
+
+
+
+int
+pump_mknod (call_frame_t *frame, xlator_t *this,
+            loc_t *loc, mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame, default_mknod_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->mknod,
+                            loc, mode, rdev, umask, xdata);
+                return 0;
+        }
+        afr_mknod (frame, this, loc, mode, rdev, umask, xdata);
+        return 0;
+
+}
+
+
+
+int
+pump_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;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame, default_mkdir_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->mkdir,
+                            loc, mode, umask, xdata);
+                return 0;
+        }
+        afr_mkdir (frame, this, loc, mode, umask, xdata);
+        return 0;
+
+}
+
+
+static int32_t
+pump_unlink (call_frame_t *frame,
+             xlator_t *this,
+             loc_t *loc, int xflag, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_unlink_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->unlink,
+                            loc, xflag, xdata);
+                return 0;
+        }
+        afr_unlink (frame, this, loc, xflag, xdata);
+        return 0;
+
+}
+
+
+static int
+pump_rmdir (call_frame_t *frame, xlator_t *this,
+            loc_t *loc, int flags, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+
+	priv = this->private;
+
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame, default_rmdir_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->rmdir,
+                            loc, flags, xdata);
+                return 0;
+        }
+
+        afr_rmdir (frame, this, loc, flags, xdata);
+        return 0;
+
+}
+
+
+
+int
+pump_symlink (call_frame_t *frame, xlator_t *this,
+              const char *linkpath, loc_t *loc, mode_t umask, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame, default_symlink_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->symlink,
+                            linkpath, loc, umask, xdata);
+                return 0;
+        }
+        afr_symlink (frame, this, linkpath, loc, umask, xdata);
+        return 0;
+
+}
+
+
+static int32_t
+pump_rename (call_frame_t *frame,
+             xlator_t *this,
+             loc_t *oldloc,
+             loc_t *newloc, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_rename_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->rename,
+                            oldloc, newloc, xdata);
+                return 0;
+        }
+        afr_rename (frame, this, oldloc, newloc, xdata);
+        return 0;
+
+}
+
+
+static int32_t
+pump_link (call_frame_t *frame,
+           xlator_t *this,
+           loc_t *oldloc,
+           loc_t *newloc, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_link_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->link,
+                            oldloc, newloc, xdata);
+                return 0;
+        }
+        afr_link (frame, this, oldloc, newloc, xdata);
+        return 0;
+
+}
+
+
+static int32_t
+pump_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;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame, default_create_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->create,
+                            loc, flags, mode, umask, fd, xdata);
+                return 0;
+        }
+        afr_create (frame, this, loc, flags, mode, umask, fd, xdata);
+        return 0;
+
+}
+
+
+static int32_t
+pump_open (call_frame_t *frame,
+           xlator_t *this,
+           loc_t *loc,
+           int32_t flags, fd_t *fd, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_open_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->open,
+                            loc, flags, fd, xdata);
+                return 0;
+        }
+        afr_open (frame, this, loc, flags, fd, xdata);
+        return 0;
+
+}
+
+
+static int32_t
+pump_writev (call_frame_t *frame,
+             xlator_t *this,
+             fd_t *fd,
+             struct iovec *vector,
+             int32_t count,
+             off_t off, uint32_t flags,
+             struct iobref *iobref, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_writev_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->writev,
+                            fd,
+                            vector,
+                            count,
+                            off, flags,
+                            iobref, xdata);
+                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, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_flush_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->flush,
+                            fd, xdata);
+                return 0;
+        }
+        afr_flush (frame, this, fd, xdata);
+        return 0;
+
+}
+
+
+static int32_t
+pump_fsync (call_frame_t *frame,
+            xlator_t *this,
+            fd_t *fd,
+            int32_t flags, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_fsync_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->fsync,
+                            fd,
+                            flags, xdata);
+                return 0;
+        }
+        afr_fsync (frame, this, fd, flags, xdata);
+        return 0;
+
+}
+
+
+static int32_t
+pump_opendir (call_frame_t *frame,
+              xlator_t *this,
+              loc_t *loc, fd_t *fd, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_opendir_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->opendir,
+                            loc, fd, xdata);
+                return 0;
+        }
+        afr_opendir (frame, this, loc, fd, xdata);
+        return 0;
+
+}
+
+
+static int32_t
+pump_fsyncdir (call_frame_t *frame,
+               xlator_t *this,
+               fd_t *fd,
+               int32_t flags, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_fsyncdir_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->fsyncdir,
+                            fd,
+                            flags, xdata);
+                return 0;
+        }
+        afr_fsyncdir (frame, this, fd, flags, xdata);
+        return 0;
+
+}
+
+
+static int32_t
+pump_xattrop (call_frame_t *frame,
+              xlator_t *this,
+              loc_t *loc,
+              gf_xattrop_flags_t flags,
+              dict_t *dict, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_xattrop_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->xattrop,
+                            loc,
+                            flags,
+                            dict, xdata);
+                return 0;
+        }
+        afr_xattrop (frame, this, loc, flags, dict, xdata);
+        return 0;
+
+}
+
+static int32_t
+pump_fxattrop (call_frame_t *frame,
+               xlator_t *this,
+               fd_t *fd,
+               gf_xattrop_flags_t flags,
+               dict_t *dict, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_fxattrop_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->fxattrop,
+                            fd,
+                            flags,
+                            dict, xdata);
+                return 0;
+        }
+        afr_fxattrop (frame, this, fd, flags, dict, xdata);
+        return 0;
+
+}
+
+
+static int32_t
+pump_removexattr (call_frame_t *frame,
+                  xlator_t *this,
+                  loc_t *loc,
+                  const char *name, dict_t *xdata)
+{
+        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,
+                            default_removexattr_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->removexattr,
+                            loc,
+                            name, xdata);
+                return 0;
+        }
+        afr_removexattr (frame, this, loc, name, xdata);
+
+ out:
+        if (op_errno)
+                AFR_STACK_UNWIND (removexattr, frame, -1, op_errno, NULL);
+        return 0;
+
+}
+
+
+
+static int32_t
+pump_readdir (call_frame_t *frame,
+              xlator_t *this,
+              fd_t *fd,
+              size_t size,
+              off_t off, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_readdir_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->readdir,
+                            fd, size, off, xdata);
+                return 0;
+        }
+        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, dict_t *dict)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_readdirp_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->readdirp,
+                            fd, size, off, dict);
+                return 0;
+        }
+        afr_readdirp (frame, this, fd, size, off, dict);
+        return 0;
+
+}
+
+
+
+static int32_t
+pump_releasedir (xlator_t *this,
+                 fd_t *fd)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (priv->use_afr_in_pump)
+                afr_releasedir (this, fd);
+	return 0;
+
+}
+
+static int32_t
+pump_release (xlator_t *this,
+              fd_t *fd)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (priv->use_afr_in_pump)
+                afr_release (this, fd);
+	return 0;
+
+}
+
+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, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_setattr_cbk,
+                            FIRST_CHILD (this),
+                            FIRST_CHILD (this)->fops->setattr,
+                            loc, stbuf, valid, xdata);
+                return 0;
+        }
+        afr_setattr (frame, this, loc, stbuf, valid, xdata);
+        return 0;
+
+}
+
+
+static int32_t
+pump_fsetattr (call_frame_t *frame,
+               xlator_t *this,
+               fd_t *fd,
+               struct iatt *stbuf,
+               int32_t valid, dict_t *xdata)
+{
+        afr_private_t *priv  = NULL;
+	priv = this->private;
+        if (!priv->use_afr_in_pump) {
+                STACK_WIND (frame,
+                            default_fsetattr_cbk,
+                            FIRST_CHILD (this),
+                            FIRST_CHILD (this)->fops->fsetattr,
+                            fd, stbuf, valid, xdata);
+                return 0;
+        }
+        afr_fsetattr (frame, this, fd, stbuf, valid, xdata);
+        return 0;
+
+}
+
+
+/* End of defaults */
+
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_afr_mt_end + 1);
+
+        if (ret != 0) {
+                gf_log(this->name, GF_LOG_ERROR, "Memory accounting init"
+                                "failed");
+                return ret;
+        }
+
+        return ret;
+}
+
+static int
+is_xlator_pump_sink (xlator_t *child)
+{
+        return (child == PUMP_SINK_CHILD(THIS));
+}
+
+static int
+is_xlator_pump_source (xlator_t *child)
+{
+        return (child == PUMP_SOURCE_CHILD(THIS));
+}
+
+int32_t
+notify (xlator_t *this, int32_t event,
+	void *data, ...)
+{
+        int ret = -1;
+        xlator_t *child_xl = NULL;
+
+        child_xl = (xlator_t *) data;
+
+        ret = afr_notify (this, event, data, NULL);
+
+	switch (event) {
+	case GF_EVENT_CHILD_DOWN:
+                if (is_xlator_pump_source (child_xl))
+                        pump_change_state (this, PUMP_STATE_ABORT);
+                break;
+
+        case GF_EVENT_CHILD_UP:
+                if (is_xlator_pump_sink (child_xl))
+                        if (is_pump_start_pending (this)) {
+                                gf_log (this->name, GF_LOG_DEBUG,
+                                        "about to start synctask");
+                                ret = pump_start_synctask (this);
+                                if (ret < 0)
+                                        gf_log (this->name, GF_LOG_DEBUG,
+                                                "Could not start pump "
+                                                "synctask");
+                                else
+                                        pump_remove_start_pending (this);
+                        }
+        }
+
+        return ret;
+}
+
+int32_t
+init (xlator_t *this)
+{
+	afr_private_t * priv        = NULL;
+        pump_private_t *pump_priv   = NULL;
+	int             child_count = 0;
+	xlator_list_t * trav        = NULL;
+	int             i           = 0;
+	int             ret         = -1;
+	GF_UNUSED int   op_errno    = 0;
+
+        int source_child = 0;
+
+	if (!this->children) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"pump translator needs a source and sink"
+                        "subvolumes defined.");
+		return -1;
+	}
+
+	if (!this->parents) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"Volume is dangling.");
+	}
+
+	priv = GF_CALLOC (1, sizeof (afr_private_t), gf_afr_mt_afr_private_t);
+        if (!priv)
+                goto out;
+
+        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     = "on";
+	priv->metadata_self_heal = 1;
+	priv->entry_self_heal    = 1;
+
+        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 */
+
+        /* Lock server count infact does not matter. Locks are held
+           on all subvolumes, in this case being the source
+           and the sink.
+        */
+
+	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.");
+		op_errno = ENOMEM;
+		goto out;
+	}
+
+	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.");
+		op_errno = ENOMEM;
+		goto out;
+	}
+
+        priv->pending_key = GF_CALLOC (sizeof (*priv->pending_key),
+                                       child_count,
+                                       gf_afr_mt_char);
+        if (!priv->pending_key) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Out of memory.");
+                op_errno = ENOMEM;
+                goto out;
+        }
+
+	trav = this->children;
+	i = 0;
+	while (i < child_count) {
+		priv->children[i] = trav->xlator;
+
+                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");
+                        op_errno = ENOMEM;
+                        goto out;
+                }
+
+		trav = trav->next;
+		i++;
+	}
+
+        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) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"Out of memory");
+                op_errno = ENOMEM;
+		goto out;
+	}
+
+        LOCK_INIT (&pump_priv->resume_path_lock);
+        LOCK_INIT (&pump_priv->pump_state_lock);
+
+        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");
+                ret = -1;
+                goto out;
+        }
+
+	pump_priv->env = this->ctx->env;
+        if (!pump_priv->env) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Could not create new sync-environment");
+                ret = -1;
+                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;
+
+        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;
+}
+
+
+struct xlator_fops fops = {
+	.lookup      = pump_lookup,
+	.open        = pump_open,
+	.flush       = pump_flush,
+	.fsync       = pump_fsync,
+	.fsyncdir    = pump_fsyncdir,
+	.xattrop     = pump_xattrop,
+	.fxattrop    = pump_fxattrop,
+        .getxattr    = pump_getxattr,
+
+	/* inode write */
+	.writev      = pump_writev,
+	.truncate    = pump_truncate,
+	.ftruncate   = pump_ftruncate,
+	.setxattr    = pump_setxattr,
+        .setattr     = pump_setattr,
+	.fsetattr    = pump_fsetattr,
+	.removexattr = pump_removexattr,
+
+	/* dir read */
+	.opendir     = pump_opendir,
+	.readdir     = pump_readdir,
+	.readdirp    = pump_readdirp,
+
+	/* dir write */
+	.create      = pump_create,
+	.mknod       = pump_mknod,
+	.mkdir       = pump_mkdir,
+	.unlink      = pump_unlink,
+	.rmdir       = pump_rmdir,
+	.link        = pump_link,
+	.symlink     = pump_symlink,
+	.rename      = pump_rename,
+};
+
+struct xlator_dumpops dumpops = {
+        .priv       = afr_priv_dump,
+};
+
+
+struct xlator_cbks cbks = {
+	.release     = pump_release,
+	.releasedir  = pump_releasedir,
+        .forget      = pump_forget,
+};
+
+struct volume_options options[] = {
+	{ .key  = {NULL} },
+};
diff --git a/xlators/cluster/afr/src/pump.h b/xlators/cluster/afr/src/pump.h
new file mode 100644
index 000000000..9d0b6db6a
--- /dev/null
+++ b/xlators/cluster/afr/src/pump.h
@@ -0,0 +1,81 @@
+/*
+  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__
+#define __PUMP_H__
+
+#include "syncop.h"
+
+/* FIXME: Needs to be defined in a common file */
+#define CLIENT_CMD_CONNECT "trusted.glusterfs.client-connect"
+#define CLIENT_CMD_DISCONNECT "trusted.glusterfs.client-disconnect"
+
+#define PUMP_SOURCE_COMPLETE "trusted.glusterfs.pump-source-complete"
+#define PUMP_SINK_COMPLETE "trusted.glusterfs.pump-sink-complete"
+
+#define PUMP_PATH "trusted.glusterfs.pump-path"
+
+#define PUMP_SOURCE_CHILD(xl) (xl->children->xlator)
+#define PUMP_SINK_CHILD(xl) (xl->children->next->xlator)
+
+typedef enum {
+        PUMP_STATE_RUNNING,             /* Pump is running and migrating files */
+        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 */
+        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 */
+        char current_file[PATH_MAX];    /* Current file being pumped */
+        uint64_t number_files_pumped;   /* Number of files pumped */
+        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
+build_root_loc (inode_t *inode, loc_t *loc);
+int pump_start (call_frame_t *frame, xlator_t *this);
+
+gf_boolean_t
+pump_command_start (xlator_t *this, dict_t *dict);
+
+int
+pump_execute_start (call_frame_t *frame, xlator_t *this);
+
+gf_boolean_t
+pump_command_pause (xlator_t *this, dict_t *dict);
+
+int
+pump_execute_pause (call_frame_t *frame, xlator_t *this);
+
+gf_boolean_t
+pump_command_abort (xlator_t *this, dict_t *dict);
+
+int
+pump_execute_abort (call_frame_t *frame, xlator_t *this);
+
+gf_boolean_t
+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 70b89521e..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
+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 3a38b67ff..3868fc38f 100644
--- a/xlators/cluster/dht/src/dht-common.c
+++ b/xlators/cluster/dht/src/dht-common.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2009-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -27,12 +18,88 @@
 
 #include "glusterfs.h"
 #include "xlator.h"
+#include "libxlator.h"
 #include "dht-common.h"
 #include "defaults.h"
+#include "byte-order.h"
+#include "glusterfs-acl.h"
 
 #include <sys/time.h>
 #include <libgen.h>
 
+int
+dht_aggregate (dict_t *this, char *key, data_t *value, void *data)
+{
+        dict_t  *dst  = NULL;
+        int64_t *ptr  = 0, *size = NULL;
+        int32_t  ret  = -1;
+        data_t  *dict_data = NULL;
+
+        dst = data;
+
+        if (strcmp (key, GF_XATTR_QUOTA_SIZE_KEY) == 0) {
+                ret = dict_get_bin (dst, key, (void **)&size);
+                if (ret < 0) {
+                        size = GF_CALLOC (1, sizeof (int64_t),
+                                          gf_common_mt_char);
+                        if (size == NULL) {
+                                gf_log ("dht", GF_LOG_WARNING,
+                                        "memory allocation failed");
+                                return -1;
+                        }
+                        ret = dict_set_bin (dst, key, size, sizeof (int64_t));
+                        if (ret < 0) {
+                                gf_log ("dht", GF_LOG_WARNING,
+                                        "dht aggregate dict set failed");
+                                GF_FREE (size);
+                                return -1;
+                        }
+                }
+
+                ptr = data_to_bin (value);
+                if (ptr == NULL) {
+                        gf_log ("dht", GF_LOG_WARNING, "data to bin failed");
+                        return -1;
+                }
+
+                *size = hton64 (ntoh64 (*size) + ntoh64 (*ptr));
+
+        } else if (fnmatch (GF_XATTR_STIME_PATTERN, key, FNM_NOESCAPE) == 0) {
+                ret = gf_get_min_stime (THIS, dst, key, value);
+                if (ret < 0)
+                        return ret;
+        } else {
+                /* compare user xattrs only */
+                if (!strncmp (key, "user.", strlen ("user."))) {
+                        ret = dict_lookup (dst, key, &dict_data);
+                        if (!ret && dict_data && value) {
+                                ret = is_data_equal (dict_data, value);
+                                if (!ret)
+                                        gf_log ("dht", GF_LOG_DEBUG,
+                                                "xattr mismatch for %s", key);
+                        }
+                }
+                ret = dict_set (dst, key, value);
+                if (ret)
+                        gf_log ("dht", GF_LOG_WARNING, "xattr dict set failed");
+        }
+
+        return 0;
+}
+
+
+void
+dht_aggregate_xattr (dict_t *dst, dict_t *src)
+{
+        if ((dst == NULL) || (src == NULL)) {
+                goto out;
+        }
+
+        dict_foreach (src, dht_aggregate, dst);
+out:
+        return;
+}
+
 /* TODO:
    - use volumename in xattr instead of "dht"
    - use NS locks
@@ -43,102 +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;
-
-	if (ret == 0) {
-		layout = local->selfheal.layout;
-		ret = dht_layout_set (this, local->inode, layout);
-
-		if (local->st_ino) {
-			local->stbuf.st_ino = local->st_ino;
-		} else {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"could not find hashed subvolume for %s",
-				local->loc.path);
-		}
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
 
-                if (local->loc.parent)
-                        local->postparent.st_ino = local->loc.parent->ino;
-	}
+        local = frame->local;
+        ret = op_ret;
 
-	DHT_STACK_UNWIND (lookup, frame, ret, local->op_errno, local->inode,
-			  &local->stbuf, local->xattr, &local->postparent);
+        FRAME_SU_UNDO (frame, dht_local_t);
 
-	return 0;
+        if (ret == 0) {
+                layout = local->selfheal.layout;
+                ret = dht_layout_set (this, local->inode, layout);
+        }
+
+        if (local->loc.parent) {
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           &local->postparent, 1);
+        }
+
+        DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+
+        DHT_STACK_UNWIND (lookup, frame, ret, local->op_errno, local->inode,
+                          &local->stbuf, local->xattr, &local->postparent);
+
+out:
+        return ret;
+}
+
+
+int
+dht_discover_complete (xlator_t *this, call_frame_t *discover_frame)
+{
+        dht_local_t     *local = NULL;
+        call_frame_t    *main_frame = NULL;
+        int              op_errno = 0;
+        int              ret = -1;
+        dht_layout_t    *layout = NULL;
+        dht_conf_t      *conf = NULL;
+
+        local = discover_frame->local;
+        layout = local->layout;
+        conf = this->private;
+
+        LOCK(&discover_frame->lock);
+        {
+                main_frame = local->main_frame;
+                local->main_frame = NULL;
+        }
+        UNLOCK(&discover_frame->lock);
+
+        if (!main_frame)
+                return 0;
+
+        if (local->file_count && local->dir_count) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "path %s exists as a file on one subvolume "
+                        "and directory on another. "
+                        "Please fix it manually",
+                        local->loc.path);
+                op_errno = EIO;
+                goto out;
+        }
+
+        if (local->cached_subvol) {
+                ret = dht_layout_preset (this, local->cached_subvol,
+                                         local->inode);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "failed to set layout for subvolume %s",
+                                local->cached_subvol ? local->cached_subvol->name : "<nil>");
+                        op_errno = EINVAL;
+                        goto out;
+                }
+        } else {
+                ret = dht_layout_normalize (this, &local->loc, layout);
+                if ((ret < 0) || ((ret > 0) && (local->op_ret != 0))) {
+                        /* either the layout is incorrect or the directory is
+                         * not found even in one subvolume.
+                         */
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "normalizing failed on %s "
+                                "(overlaps/holes present: %s, "
+                                "ENOENT errors: %d)", local->loc.path,
+                                (ret < 0) ? "yes" : "no", (ret > 0) ? ret : 0);
+                        if ((ret > 0) && (ret == conf->subvolume_cnt)) {
+                                op_errno = ESTALE;
+                                goto out;
+                        }
+                }
+
+                if (local->inode)
+                        dht_layout_set (this, local->inode, layout);
+        }
+
+        DHT_STACK_UNWIND (lookup, main_frame, local->op_ret, local->op_errno,
+                          local->inode, &local->stbuf, local->xattr,
+                          &local->postparent);
+        return 0;
+out:
+        DHT_STACK_UNWIND (lookup, main_frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL);
+
+        return ret;
+}
+
+
+int
+dht_discover_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int op_ret, int op_errno,
+                  inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+                  struct iatt *postparent)
+{
+        dht_local_t  *local                   = NULL;
+        int           this_call_cnt           = 0;
+        call_frame_t *prev                    = NULL;
+        dht_layout_t *layout                  = NULL;
+        int           ret                     = -1;
+        int           is_dir                  = 0;
+        int           is_linkfile             = 0;
+        int           attempt_unwind          = 0;
+        dht_conf_t   *conf                    = 0;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+        local = frame->local;
+        prev  = cookie;
+        conf = this->private;
+
+        layout = local->layout;
+
+        /* Check if the gfid is different for file from other node */
+        if (!op_ret && uuid_compare (local->gfid, stbuf->ia_gfid)) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: gfid different on %s",
+                        local->loc.path, prev->this->name);
+        }
+
+
+        LOCK (&frame->lock);
+        {
+                /* TODO: assert equal mode on stbuf->st_mode and
+                   local->stbuf->st_mode
+
+                   else mkdir/chmod/chown and fix
+                */
+                ret = dht_layout_merge (this, layout, prev->this,
+                                        op_ret, op_errno, xattr);
+                if (ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to merge layouts", local->loc.path);
+
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "lookup of %s on %s returned error (%s)",
+                                local->loc.path, prev->this->name,
+                                strerror (op_errno));
+
+                        goto unlock;
+                }
+
+                is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+                                                 conf->link_xattr_name);
+                is_dir = check_is_dir (inode, stbuf, xattr);
+
+                if (is_dir) {
+                        local->dir_count ++;
+                } else {
+                        local->file_count ++;
+
+                        if (!is_linkfile) {
+                                /* real file */
+                                local->cached_subvol = prev->this;
+                                attempt_unwind = 1;
+                        } else {
+                                goto unlock;
+                        }
+                }
+
+                local->op_ret = 0;
+
+                if (local->xattr == NULL) {
+                        local->xattr = dict_ref (xattr);
+                } else {
+                        dht_aggregate_xattr (local->xattr, xattr);
+                }
+
+                if (local->inode == NULL)
+                        local->inode = inode_ref (inode);
+
+                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+                dht_iatt_merge (this, &local->postparent, postparent,
+                                prev->this);
+        }
+unlock:
+        UNLOCK (&frame->lock);
+out:
+        this_call_cnt = dht_frame_return (frame);
+
+        if (is_last_call (this_call_cnt) || attempt_unwind) {
+                dht_discover_complete (this, frame);
+        }
+
+        if (is_last_call (this_call_cnt))
+                DHT_STACK_DESTROY (frame);
+
+        return 0;
+}
+
+
+int
+dht_discover (call_frame_t *frame, xlator_t *this, loc_t *loc)
+{
+        int          ret;
+        dht_local_t *local = NULL;
+        dht_conf_t  *conf = NULL;
+        int          call_cnt = 0;
+        int          op_errno = EINVAL;
+        int          i = 0;
+        call_frame_t *discover_frame = NULL;
+
+        conf = this->private;
+        local = frame->local;
+
+        ret = dict_set_uint32 (local->xattr_req, conf->xattr_name, 4 * 4);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to set '%s' key",
+                        loc->path, conf->xattr_name);
+
+        ret = dict_set_uint32 (local->xattr_req, conf->link_xattr_name, 256);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to set '%s' key",
+                        loc->path, conf->link_xattr_name);
+
+        call_cnt        = conf->subvolume_cnt;
+        local->call_cnt = call_cnt;
+
+        local->layout = dht_layout_new (this, conf->subvolume_cnt);
+
+        if (!local->layout) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        uuid_copy (local->gfid, loc->gfid);
+
+        discover_frame = copy_frame (frame);
+        if (!discover_frame) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        discover_frame->local = local;
+        frame->local = NULL;
+        local->main_frame = frame;
+
+        for (i = 0; i < call_cnt; i++) {
+                STACK_WIND (discover_frame, dht_discover_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->lookup,
+                            &local->loc, local->xattr_req);
+        }
+
+        return 0;
+
+err:
+        DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL);
+
+        return 0;
 }
 
 
 int
 dht_lookup_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int op_ret, int op_errno,
-                    inode_t *inode, struct stat *stbuf, dict_t *xattr,
-                    struct stat *postparent)
+                    inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+                    struct iatt *postparent)
 {
-	dht_conf_t   *conf                    = NULL;
         dht_local_t  *local                   = NULL;
         int           this_call_cnt           = 0;
         call_frame_t *prev                    = NULL;
-	dht_layout_t *layout                  = NULL;
-	int           ret                     = 0;
-	int           is_dir                  = 0;
+        dht_layout_t *layout                  = NULL;
+        int           ret                     = -1;
+        int           is_dir                  = 0;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
 
-	conf  = this->private;
         local = frame->local;
         prev  = cookie;
 
-	layout = local->layout;
+        layout = local->layout;
+
+        if (!op_ret && uuid_is_null (local->gfid))
+                memcpy (local->gfid, stbuf->ia_gfid, 16);
+
+        /* Check if the gfid is different for file from other node */
+        if (!op_ret && uuid_compare (local->gfid, stbuf->ia_gfid)) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: gfid different on %s",
+                        local->loc.path, prev->this->name);
+        }
 
         LOCK (&frame->lock);
         {
                 /* TODO: assert equal mode on stbuf->st_mode and
-		   local->stbuf->st_mode
+                   local->stbuf->st_mode
 
-		   else mkdir/chmod/chown and fix
-		*/
-		ret = dht_layout_merge (this, layout, prev->this,
-					op_ret, op_errno, xattr);
+                   else mkdir/chmod/chown and fix
+                */
+                ret = dht_layout_merge (this, layout, prev->this,
+                                        op_ret, op_errno, xattr);
 
-		if (op_ret == -1) {
-			local->op_errno = ENOENT;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"lookup of %s on %s returned error (%s)",
-				local->loc.path, prev->this->name,
-				strerror (op_errno));
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "lookup of %s on %s returned error (%s)",
+                                local->loc.path, prev->this->name,
+                                strerror (op_errno));
 
-			goto unlock;
-		}
+                        goto unlock;
+                }
 
- 		is_dir = check_is_dir (inode, stbuf, xattr);
- 		if (!is_dir) {
+                is_dir = check_is_dir (inode, stbuf, xattr);
+                if (!is_dir) {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "lookup of %s on %s returned non dir 0%o",
                                 local->loc.path, prev->this->name,
-                                stbuf->st_mode);
+                                stbuf->ia_type);
                         local->need_selfheal = 1;
- 			goto unlock;
+                        goto unlock;
                 }
 
- 		local->op_ret = 0;
- 		if (local->xattr == NULL)
- 			local->xattr = dict_ref (xattr);
- 		if (local->inode == NULL)
- 			local->inode = inode_ref (inode);
-
-		dht_stat_merge (this, &local->stbuf, stbuf, prev->this);
-                dht_stat_merge (this, &local->postparent, postparent,
-                                prev->this);
+                local->op_ret = 0;
+                if (local->xattr == NULL) {
+                        local->xattr = dict_ref (xattr);
+                } else {
+                        dht_aggregate_xattr (local->xattr, xattr);
+                }
 
-		if (prev->this == local->hashed_subvol)
-			local->st_ino = local->stbuf.st_ino;
+                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);
@@ -153,185 +487,270 @@ unlock:
                         return 0;
                 }
 
-		if (local->op_ret == 0) {
-			ret = dht_layout_normalize (this, &local->loc, layout);
+                if (local->op_ret == 0) {
+                        ret = dht_layout_normalize (this, &local->loc, layout);
 
-			if (ret != 0) {
-				gf_log (this->name, GF_LOG_DEBUG,
-					"fixing assignment on %s",
-					local->loc.path);
-				goto selfheal;
-			}
-			
-			dht_layout_set (this, local->inode, layout);
-			
-			if (local->st_ino) {
-				local->stbuf.st_ino = local->st_ino;
-			} else {
-				gf_log (this->name, GF_LOG_DEBUG,
-					"could not find hashed subvol for %s",
-					local->loc.path);
-			}
+                        if (ret != 0) {
+                                gf_log (this->name, GF_LOG_DEBUG,
+                                        "fixing assignment on %s",
+                                        local->loc.path);
+                                goto selfheal;
+                        }
 
-                        if (local->loc.parent)
-                                local->postparent.st_ino =
-                                        local->loc.parent->ino;
-		}
+                        dht_layout_set (this, local->inode, layout);
+                }
+
+                if (local->loc.parent) {
+                        dht_inode_ctx_time_update (local->loc.parent, this,
+                                                   &local->postparent, 1);
+                }
 
-		DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
-				  local->inode, &local->stbuf, local->xattr,
+                DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+                DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
+                                  local->inode, &local->stbuf, local->xattr,
                                   &local->postparent);
         }
 
-	return 0;
+        return 0;
 
 selfheal:
-	ret = dht_selfheal_directory (frame, dht_lookup_selfheal_cbk,
-				      &local->loc, layout);
-
-	return 0;
+        FRAME_SU_DO (frame, dht_local_t);
+        uuid_copy (local->loc.gfid, local->gfid);
+        ret = dht_selfheal_directory (frame, dht_lookup_selfheal_cbk,
+                                      &local->loc, layout);
+out:
+        return ret;
 }
 
 int
 dht_revalidate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int op_ret, int op_errno,
-                    inode_t *inode, struct stat *stbuf, dict_t *xattr,
-                    struct stat *postparent)
+                    inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+                    struct iatt *postparent)
 {
         dht_local_t  *local         = NULL;
         int           this_call_cnt = 0;
         call_frame_t *prev          = NULL;
-	dht_layout_t *layout        = NULL;
-	dht_conf_t   *conf          = NULL;
-	int           ret  = -1;
-	int           is_dir = 0;
-	int           is_linkfile = 0;
+        dht_layout_t *layout        = NULL;
+        dht_conf_t   *conf          = NULL;
+        int           ret  = -1;
+        int           is_dir = 0;
+        int           is_linkfile = 0;
+        call_frame_t *copy          = NULL;
+        dht_local_t  *copy_local    = NULL;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, err);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, err);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, err);
 
         local = frame->local;
         prev  = cookie;
-	conf = this->private;
+        conf = this->private;
+        if (!conf)
+                goto out;
 
         LOCK (&frame->lock);
         {
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
 
-			if ((op_errno != ENOTCONN) 
+                        if ((op_errno != ENOTCONN)
                             && (op_errno != ENOENT)
                             && (op_errno != ESTALE)) {
-				gf_log (this->name, GF_LOG_DEBUG,
-					"subvolume %s returned -1 (%s)",
-					prev->this->name, strerror (op_errno));
+				gf_log (this->name, GF_LOG_INFO,
+					"subvolume %s for %s returned -1 (%s)",
+					prev->this->name, local->loc.path,
+                                        strerror (op_errno));
 			}
-                        
                         if (op_errno == ESTALE) {
-                                /* propogate the ESTALE to parent. 
-                                 * setting local->layout_mismatch would send
+                                /* propagate the ESTALE to parent.
+                                 * setting local->return_estale would send
                                  * ESTALE to parent. */
-                                local->layout_mismatch = 1;
+                                local->return_estale = 1;
                         }
 
-			goto unlock;
-		}
+                        /* if it is ENOENT, we may have to do a
+                         * 'lookup_everywhere()' to make sure
+                         * the file is not migrated */
+                        if (op_errno == ENOENT) {
+                                if (IA_ISREG (local->loc.inode->ia_type)) {
+                                        local->need_lookup_everywhere = 1;
+                                }
+                        }
+                        goto unlock;
+                }
 
-		if (S_IFMT & (stbuf->st_mode ^ local->inode->st_mode)) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"mismatching filetypes 0%o v/s 0%o for %s",
-				(stbuf->st_mode & S_IFMT),
-				(local->inode->st_mode & S_IFMT),
-				local->loc.path);
+                if (stbuf->ia_type != local->inode->ia_type) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "mismatching filetypes 0%o v/s 0%o for %s",
+                                (stbuf->ia_type), (local->inode->ia_type),
+                                local->loc.path);
 
-			local->op_ret = -1;
-			local->op_errno = EINVAL;
+                        local->op_ret = -1;
+                        local->op_errno = EINVAL;
 
-			goto unlock;
-		}
+                        goto unlock;
+                }
 
-		layout = local->layout;
-		
-		is_dir = check_is_dir (inode, stbuf, xattr);
-		is_linkfile = check_is_linkfile (inode, stbuf, xattr);
-		
-		if (is_linkfile) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"linkfile found in revalidate for %s",
-				local->loc.path);
-			local->layout_mismatch = 1;
-
-			goto unlock;
-		}
+                layout = local->layout;
 
-		if (is_dir) {
-			ret = dht_layout_dir_mismatch (this, layout,
-						       prev->this, &local->loc,
-						       xattr);
-			if (ret != 0) {
-				gf_log (this->name, GF_LOG_DEBUG,
-					"mismatching layouts for %s", 
-					local->loc.path);
-			
-				local->layout_mismatch = 1;
-
-				goto unlock;
-			}
-		} 
-		
-		dht_stat_merge (this, &local->stbuf, stbuf, prev->this);
-                dht_stat_merge (this, &local->postparent, postparent,
+                is_dir = check_is_dir (inode, stbuf, xattr);
+                is_linkfile = check_is_linkfile (inode, stbuf, xattr,
+                                                 conf->link_xattr_name);
+
+                if (is_linkfile) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "linkfile found in revalidate for %s",
+                                local->loc.path);
+                        local->return_estale = 1;
+
+                        goto unlock;
+                }
+
+                if (is_dir) {
+                        ret = dht_dir_has_layout (xattr, conf->xattr_name);
+                        if (ret >= 0) {
+                                if (is_greater_time(local->stbuf.ia_ctime,
+                                                    local->stbuf.ia_ctime_nsec,
+                                                    stbuf->ia_ctime,
+                                                    stbuf->ia_ctime_nsec)) {
+                                        local->prebuf.ia_gid = stbuf->ia_gid;
+                                        local->prebuf.ia_uid = stbuf->ia_uid;
+                                }
+                        }
+                        if (local->stbuf.ia_type != IA_INVAL)
+                        {
+                                if ((local->stbuf.ia_gid != stbuf->ia_gid) ||
+                                    (local->stbuf.ia_uid != stbuf->ia_uid)) {
+                                        local->need_selfheal = 1;
+                                }
+                        }
+                        ret = dht_layout_dir_mismatch (this, layout,
+                                                       prev->this, &local->loc,
+                                                       xattr);
+                        if (ret != 0) {
+                                gf_log (this->name, GF_LOG_INFO,
+                                        "mismatching layouts for %s",
+                                        local->loc.path);
+
+                                local->layout_mismatch = 1;
+
+                                goto unlock;
+                        }
+                }
+
+                dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+                dht_iatt_merge (this, &local->postparent, postparent,
                                 prev->this);
-		
-		local->op_ret = 0;
-		local->stbuf.st_ino = local->st_ino;
-                if (local->loc.parent)
-                        local->postparent.st_ino = local->loc.parent->ino;
-
-		if (!local->xattr)
-			local->xattr = dict_ref (xattr);
-	}
-unlock:
-	UNLOCK (&frame->lock);
 
+                local->op_ret = 0;
+
+                if (!local->xattr) {
+                        local->xattr = dict_ref (xattr);
+                } else if (is_dir) {
+                        dht_aggregate_xattr (local->xattr, xattr);
+                }
+        }
+unlock:
+        UNLOCK (&frame->lock);
+out:
         this_call_cnt = dht_frame_return (frame);
 
         if (is_last_call (this_call_cnt)) {
-		if (!S_ISDIR (local->stbuf.st_mode)
-		    && (local->hashed_subvol != local->cached_subvol)
-		    && (local->stbuf.st_nlink == 1)
-		    && (conf->unhashed_sticky_bit)) {
-			local->stbuf.st_mode |= S_ISVTX;
-		}
-
+                if (!IA_ISDIR (local->stbuf.ia_type)
+                    && (local->hashed_subvol != local->cached_subvol)
+                    && (local->stbuf.ia_nlink == 1)
+                    && (conf && conf->unhashed_sticky_bit)) {
+                        local->stbuf.ia_prot.sticky = 1;
+                }
+                if (local->need_selfheal) {
+                        local->need_selfheal = 0;
+                        uuid_copy (local->gfid, local->stbuf.ia_gfid);
+                        local->stbuf.ia_gid = local->prebuf.ia_gid;
+                        local->stbuf.ia_uid = local->prebuf.ia_uid;
+                        copy = create_frame (this, this->ctx->pool);
+                        if (copy) {
+                                copy_local = dht_local_init (copy, &local->loc,
+                                                             NULL, 0);
+                                if (!copy_local)
+                                        goto cont;
+                                copy_local->stbuf = local->stbuf;
+                                copy->local = copy_local;
+                                FRAME_SU_DO (copy, dht_local_t);
+                                ret = synctask_new (this->ctx->env,
+                                                    dht_dir_attr_heal,
+                                                    dht_dir_attr_heal_done,
+                                                    copy, copy);
+                        }
+                }
+cont:
 		if (local->layout_mismatch) {
-			local->op_ret = -1;
-			local->op_errno = ESTALE;
-		}
-			
-		DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
-				  local->inode, &local->stbuf, local->xattr,
+                        /* 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;
+                }
+
+                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->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,
-                                inode_t *inode, struct stat *stbuf,
-                                struct stat *preparent, struct stat *postparent)
-{
-	dht_local_t  *local = NULL;
-	xlator_t     *cached_subvol = NULL;
-	dht_conf_t   *conf = NULL;
+                                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)
+{
+        dht_local_t  *local = NULL;
+        xlator_t     *cached_subvol = NULL;
+        dht_conf_t   *conf = NULL;
         int           ret = -1;
 
-	local = frame->local;
-	cached_subvol = local->cached_subvol;
-	conf = this->private;
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+
+        local = frame->local;
+        cached_subvol = local->cached_subvol;
+        conf = this->private;
 
-        ret = dht_layout_preset (this, local->cached_subvol, inode);
+        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",
@@ -341,70 +760,232 @@ dht_lookup_linkfile_create_cbk (call_frame_t *frame, void *cookie,
                 goto unwind;
         }
 
-	local->op_ret = 0;
-	if ((local->stbuf.st_nlink == 1)
-	    && (conf->unhashed_sticky_bit)) {
-		local->stbuf.st_mode |= S_ISVTX;
-	}
+        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.st_ino = local->loc.parent->ino;
+        if (local->loc.parent) {
+                dht_inode_ctx_time_update (local->loc.parent, this,
+                                           postparent, 1);
+        }
 
 unwind:
-	DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
-			  local->inode, &local->stbuf, local->xattr,
+        if (local->linked == _gf_true)
+                dht_linkfile_attr_heal (frame, this);
+
+        DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+        DHT_STACK_UNWIND (lookup, frame, local->op_ret, local->op_errno,
+                          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,
-                           inode_t *inode, struct stat *buf, dict_t *xattr,
-                           struct stat *postparent)
+                           int32_t op_ret, int32_t op_errno,
+                           inode_t *inode, struct iatt *buf, dict_t *xattr,
+                           struct iatt *postparent)
 {
-	dht_conf_t   *conf          = NULL;
         dht_local_t  *local         = NULL;
         int           this_call_cnt = 0;
         call_frame_t *prev          = NULL;
-	int           is_linkfile   = 0;
-	int           is_dir        = 0;
-	xlator_t     *subvol        = NULL;
-	loc_t        *loc           = NULL;
-	xlator_t     *link_subvol   = NULL;
-	xlator_t     *hashed_subvol = NULL;
-	xlator_t     *cached_subvol = NULL;
-        int           ret = -1;
+        int           is_linkfile   = 0;
+        int           is_dir        = 0;
+        xlator_t     *subvol        = NULL;
+        loc_t        *loc           = NULL;
+        xlator_t     *link_subvol   = NULL;
+        int           ret           = -1;
+        int32_t       fd_count      = 0;
+        dht_conf_t   *conf          = NULL;
 
-	conf   = this->private;
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", cookie, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
 
-	local  = frame->local;
-	loc    = &local->loc;
+        local  = frame->local;
+        loc    = &local->loc;
+        conf   = this->private;
 
-	prev   = cookie;
-	subvol = prev->this;
+        prev   = cookie;
+        subvol = prev->this;
 
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			if (op_errno != ENOENT)
-				local->op_errno = op_errno;
-			goto unlock;
-		}
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        if (op_errno != ENOENT)
+                                local->op_errno = op_errno;
+                        goto unlock;
+                }
 
-		is_linkfile = check_is_linkfile (inode, buf, xattr);
-		is_dir = check_is_dir (inode, buf, xattr);
-
-		if (is_linkfile) {
-			link_subvol = dht_linkfile_subvol (this, inode, buf,
-							   xattr);
-			gf_log (this->name, GF_LOG_DEBUG,
-				"found on %s linkfile %s (-> %s)",
-				subvol->name, loc->path,
-				link_subvol ? link_subvol->name : "''");
-			goto unlock;
-		}
+                if (uuid_is_null (local->gfid))
+                        uuid_copy (local->gfid, buf->ia_gfid);
+
+                if (uuid_compare (local->gfid, buf->ia_gfid)) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: gfid differs on subvolume %s",
+                                loc->path, prev->this->name);
+                }
+
+                is_linkfile = check_is_linkfile (inode, buf, xattr,
+                                                 conf->link_xattr_name);
+                is_dir = check_is_dir (inode, buf, xattr);
+
+                if (is_linkfile) {
+                        link_subvol = dht_linkfile_subvol (this, inode, buf,
+                                                           xattr);
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "found on %s linkfile %s (-> %s)",
+                                subvol->name, loc->path,
+                                link_subvol ? link_subvol->name : "''");
+                        goto unlock;
+                }
+
+                /* non linkfile GFID takes precedence */
+                uuid_copy (local->gfid, buf->ia_gfid);
 
                 if (is_dir) {
                         local->dir_count++;
@@ -417,193 +998,175 @@ dht_lookup_everywhere_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
                         if (!local->cached_subvol) {
                                 /* found one file */
-                                dht_stat_merge (this, &local->stbuf, buf,
+                                dht_iatt_merge (this, &local->stbuf, buf,
                                                 subvol);
                                 local->xattr = dict_ref (xattr);
                                 local->cached_subvol = subvol;
                                 gf_log (this->name, GF_LOG_DEBUG,
                                         "found on %s file %s",
                                         subvol->name, loc->path);
-                                
-                                dht_stat_merge (this, &local->postparent,
+
+                                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.st_ino =
-                                        local->loc.parent->ino;
+        UNLOCK (&frame->lock);
 
-                        DHT_STACK_UNWIND (lookup, frame, local->op_ret,
-                                          local->op_errno, local->inode,
-                                          &local->stbuf, local->xattr,
-                                          &local->postparent);
+        if (is_linkfile) {
+                ret = dict_get_int32 (xattr, GLUSTERFS_OPEN_FD_COUNT, &fd_count);
+                /* Delete the linkfile only if there are no open fds on it.
+                   if there is a open-fd, it may be in migration */
+                if (!ret && (fd_count == 0)) {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "deleting stale linkfile %s on %s",
+                                loc->path, subvol->name);
+                        STACK_WIND (frame, dht_lookup_unlink_cbk,
+                                    subvol, subvol->fops->unlink, loc, 0, NULL);
                         return 0;
                 }
+        }
 
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "linking file %s existing on %s to %s (hash)",
-                        loc->path, cached_subvol->name,
-                        hashed_subvol->name);
-                        
-                dht_linkfile_create (frame, 
-                                     dht_lookup_linkfile_create_cbk,
-                                     cached_subvol, hashed_subvol, loc);
-	}
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                dht_lookup_everywhere_done (frame, this);
+        }
 
-	return 0;
+out:
+        return ret;
 }
 
 
 int
 dht_lookup_everywhere (call_frame_t *frame, xlator_t *this, loc_t *loc)
 {
-	dht_conf_t     *conf = NULL;
-	dht_local_t    *local = NULL;
-	int             i = 0;
-	int             call_cnt = 0;
+        dht_conf_t     *conf = NULL;
+        dht_local_t    *local = NULL;
+        int             i = 0;
+        int             call_cnt = 0;
 
-	conf = this->private;
-	local = frame->local;
+        GF_VALIDATE_OR_GOTO ("dht", frame, err);
+        GF_VALIDATE_OR_GOTO ("dht", this, out);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, out);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, out);
+        GF_VALIDATE_OR_GOTO ("dht", loc, out);
 
-	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);
 
-	return 0;
+        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);
+err:
+        return -1;
 }
 
 
 int
 dht_lookup_linkfile_cbk (call_frame_t *frame, void *cookie,
                          xlator_t *this, int op_ret, int op_errno,
-                         inode_t *inode, struct stat *stbuf, dict_t *xattr,
-                         struct stat *postparent)
+                         inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+                         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->st_nlink == 1)
-	    && (conf->unhashed_sticky_bit)) {
-		stbuf->st_mode |= S_ISVTX;
-	}
-        dht_itransform (this, prev->this, stbuf->st_ino, &stbuf->st_ino);
-        if (local->loc.parent)
-                postparent->st_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:
+        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);
 
@@ -611,7 +1174,7 @@ out:
 
 err:
         dht_lookup_everywhere (frame, this, loc);
-
+out:
         return 0;
 }
 
@@ -623,21 +1186,38 @@ dht_lookup_directory (call_frame_t *frame, xlator_t *this, loc_t *loc)
         int           i = 0;
         dht_conf_t   *conf = NULL;
         dht_local_t  *local = NULL;
+        int           ret = 0;
+
+        GF_VALIDATE_OR_GOTO ("dht", frame, out);
+        GF_VALIDATE_OR_GOTO ("dht", this, unwind);
+        GF_VALIDATE_OR_GOTO ("dht", frame->local, unwind);
+        GF_VALIDATE_OR_GOTO ("dht", this->private, unwind);
+        GF_VALIDATE_OR_GOTO ("dht", loc, unwind);
 
         conf = this->private;
         local = frame->local;
 
         call_cnt        = conf->subvolume_cnt;
         local->call_cnt = call_cnt;
-		
+
         local->layout = dht_layout_new (this, conf->subvolume_cnt);
         if (!local->layout) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                DHT_STACK_UNWIND (lookup, frame, -1, ENOMEM, NULL, NULL, NULL, NULL);
-                return 0;
+                goto unwind;
+        }
+
+        if (local->xattr != NULL) {
+                dict_unref (local->xattr);
+                local->xattr = NULL;
         }
-		
+
+        if (!uuid_is_null (local->gfid)) {
+                ret = dict_set_static_bin (local->xattr_req, "gfid-req",
+                                           local->gfid, 16);
+                if (ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to set gfid", local->loc.path);
+        }
+
         for (i = 0; i < call_cnt; i++) {
                 STACK_WIND (frame, dht_lookup_dir_cbk,
                             conf->subvolumes[i],
@@ -645,14 +1225,19 @@ dht_lookup_directory (call_frame_t *frame, xlator_t *this, loc_t *loc)
                             &local->loc, local->xattr_req);
         }
         return 0;
+unwind:
+        DHT_STACK_UNWIND (lookup, frame, -1, ENOMEM, NULL, NULL, NULL, NULL);
+out:
+        return 0;
+
 }
 
 
 int
 dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 int op_ret, int op_errno,
-                inode_t *inode, struct stat *stbuf, dict_t *xattr,
-                struct stat *postparent)
+                inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+                struct iatt *postparent)
 {
         char          is_linkfile   = 0;
         char          is_dir        = 0;
@@ -662,6 +1247,13 @@ dht_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         loc_t        *loc           = NULL;
         call_frame_t *prev          = NULL;
         int           ret           = 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;
 
@@ -669,83 +1261,140 @@ dht_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;
-		}
-	}
+        /* This is required for handling stale linkfile deletion,
+         * or any more call which happens from this 'loc'.
+         */
+        if (!op_ret && uuid_is_null (local->gfid))
+                memcpy (local->gfid, stbuf->ia_gfid, 16);
+
+        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;
+                }
+                if ((conf->search_unhashed == GF_DHT_LOOKUP_UNHASHED_AUTO) &&
+                    (loc->parent)) {
+                        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->st_ino,
-				&stbuf->st_ino);
-                if (loc->parent)
-                        postparent->st_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->st_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_stat_merge for reference.
-         */ 
+         * from each of the subvolume. See dht_iatt_merge for reference.
+         */
+
+        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,
@@ -753,56 +1402,68 @@ 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;
-
-        local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        conf = this->private;
+        if (!conf)
+                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 ();
-	}
 
-	hashed_subvol = dht_subvol_get_hashed (this, loc);
-	cached_subvol = dht_subvol_get_cached (this, loc->inode);
+        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);
 
-	local->cached_subvol = cached_subvol;
-	local->hashed_subvol = hashed_subvol;
+                /* we no more need 'new_loc' entries */
+                loc_wipe (&new_loc);
 
-        if (is_revalidate (loc)) {
-		local->layout = layout = dht_layout_get (this, loc->inode);
+                /* check if loc_dup() is successful */
+                if (ret == -1) {
+                        op_errno = errno;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "copying location failed for path=%s",
+                                loc->path);
+                        goto err;
+                }
+        }
 
+        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);
+        local->hashed_subvol = hashed_subvol;
+
+        if (is_revalidate (loc)) {
+                layout = local->layout;
                 if (!layout) {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "revalidate without cache. path=%s",
@@ -811,71 +1472,93 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
                         goto err;
                 }
 
-		if (layout->gen && (layout->gen < conf->gen)) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"incomplete layout failure for path=%s",
-				loc->path);
+                if (layout->gen && (layout->gen < conf->gen)) {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "incomplete layout failure for path=%s",
+                                loc->path);
 
                         dht_layout_unref (this, local->layout);
                         local->layout = NULL;
-			goto do_fresh_lookup;
-		}
+                        local->cached_subvol = NULL;
+                        goto do_fresh_lookup;
+                }
+
+                local->inode = inode_ref (loc->inode);
+
+                /* NOTE: we don't require 'trusted.glusterfs.dht.linkto' attribute,
+                 *       revalidates directly go to the cached-subvolume.
+                 */
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->xattr_name, 4 * 4);
+
+                if (IA_ISDIR (local->inode->ia_type)) {
+                        local->call_cnt = call_cnt = conf->subvolume_cnt;
+                        for (i = 0; i < call_cnt; i++) {
+                                STACK_WIND (frame, dht_revalidate_cbk,
+                                            conf->subvolumes[i],
+                                            conf->subvolumes[i]->fops->lookup,
+                                            loc, local->xattr_req);
+                        }
+                        return 0;
+                }
+
+                call_cnt = local->call_cnt = layout->cnt;
 
-		local->inode    = inode_ref (loc->inode);
-		local->st_ino   = loc->inode->ino;
-		
-		local->call_cnt = layout->cnt;
-		call_cnt = local->call_cnt;
-		
-		/* NOTE: we don't require 'trusted.glusterfs.dht.linkto' attribute,
-		 *       revalidates directly go to the cached-subvolume.
-		 */
-		ret = dict_set_uint32 (local->xattr_req, 
-				       "trusted.glusterfs.dht", 4 * 4);
-
-		for (i = 0; i < layout->cnt; i++) {
+                /* need it for 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,
-				    loc, local->xattr_req);
+				    &local->loc, local->xattr_req);
 
-			if (!--call_cnt)
-				break;
 		}
         } else {
         do_fresh_lookup:
-		/* TODO: remove the hard-coding */
-		ret = dict_set_uint32 (local->xattr_req, 
-				       "trusted.glusterfs.dht", 4 * 4);
+                /* TODO: remove the hard-coding */
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->xattr_name, 4 * 4);
+
+                ret = dict_set_uint32 (local->xattr_req,
+                                       conf->link_xattr_name, 256);
+
+                /* need it for self-healing linkfiles which is
+                   'in-migration' state */
+                ret = dict_set_uint32 (local->xattr_req,
+                                       GLUSTERFS_OPEN_FD_COUNT, 4);
 
-		ret = dict_set_uint32 (local->xattr_req, 
-				       "trusted.glusterfs.dht.linkto", 256);
+                /* need it for dir self-heal */
+                dht_check_and_set_acl_xattr_req (loc->inode, local->xattr_req);
 
                 if (!hashed_subvol) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"no subvolume in layout for path=%s, "
-				"checking on all the subvols to see if "
-				"it is a directory", loc->path);
- 			call_cnt        = conf->subvolume_cnt;
- 			local->call_cnt = call_cnt;
- 			
- 			local->layout = dht_layout_new (this,
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "no subvolume in layout for path=%s, "
+                                "checking on all the subvols to see if "
+                                "it is a directory", loc->path);
+                        call_cnt        = conf->subvolume_cnt;
+                        local->call_cnt = call_cnt;
+
+                        local->layout = dht_layout_new (this,
                                                         conf->subvolume_cnt);
- 			if (!local->layout) {
- 				op_errno = ENOMEM;
- 				gf_log (this->name, GF_LOG_ERROR,
- 					"Out of memory");
- 				goto err;
- 			}
-
-			for (i = 0; i < call_cnt; i++) {
- 				STACK_WIND (frame, dht_lookup_dir_cbk,
- 					    conf->subvolumes[i],
- 					    conf->subvolumes[i]->fops->lookup,
- 					    &local->loc, local->xattr_req);
- 			}
- 			return 0;
+                        if (!local->layout) {
+                                op_errno = ENOMEM;
+                                goto err;
+                        }
+
+                        for (i = 0; i < call_cnt; i++) {
+                                STACK_WIND (frame, dht_lookup_dir_cbk,
+                                            conf->subvolumes[i],
+                                            conf->subvolumes[i]->fops->lookup,
+                                            &local->loc, local->xattr_req);
+                        }
+                        return 0;
                 }
 
                 STACK_WIND (frame, dht_lookup_cbk,
@@ -886,1200 +1569,1585 @@ dht_lookup (call_frame_t *frame, xlator_t *this,
         return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-        DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL,
+                          NULL);
         return 0;
 }
 
 
 int
-dht_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	      int op_ret, int op_errno, struct stat *prebuf,
-              struct stat *postbuf)
+dht_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)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	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) {
-			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_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_stat_merge (this, &local->prebuf, prebuf, prev->this);
-		dht_stat_merge (this, &local->stbuf, postbuf, prev->this);
+                local->op_ret = 0;
 
-		if (local->inode) {
-			local->stbuf.st_ino = local->inode->ino;
-                        local->prebuf.st_ino = local->inode->ino;
-                }
+                local->postparent = *postparent;
+                local->preparent = *preparent;
 
-		local->op_ret = 0;
-	}
+                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);
 
-	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);
+        DHT_STACK_UNWIND (unlink, frame, local->op_ret, local->op_errno,
+                          &local->preparent, &local->postparent, NULL);
 
         return 0;
 }
 
 
-
 int
-dht_attr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	      int op_ret, int op_errno, struct stat *stbuf)
+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, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	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;
 
-		dht_stat_merge (this, &local->stbuf, stbuf, prev->this);
-		
-		if (local->inode)
-			local->stbuf.st_ino = local->inode->ino;
-		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);
 
-	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);
+        if (local->op_ret == -1)
+                goto err;
+
+        cached_subvol = dht_subvol_get_cached (this, local->loc.inode);
+        if (!cached_subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s",
+                        local->loc.path);
+                local->op_errno = EINVAL;
+                goto err;
+        }
+
+        STACK_WIND (frame, dht_unlink_cbk,
+                    cached_subvol, cached_subvol->fops->unlink,
+                    &local->loc, local->flags, NULL);
 
         return 0;
-}
 
+err:
+        DHT_STACK_UNWIND (unlink, frame, -1, local->op_errno,
+                          NULL, NULL, NULL);
+        return 0;
+}
 
 int
-dht_stat (call_frame_t *frame, xlator_t *this,
-	  loc_t *loc)
+dht_err_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+             int op_ret, int op_errno, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-	dht_layout_t *layout = NULL;
-	int           i = 0;
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
 
+        local = frame->local;
+        prev = cookie;
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
+        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);
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        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);
+        }
 
-	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;
-	}
+        return 0;
+}
 
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = layout->cnt;
+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);
+        }
+}
 
-	for (i = 0; i < layout->cnt; i++) {
-		subvol = layout->list[i].xlator;
+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);
+}
 
-		STACK_WIND (frame, dht_attr_cbk,
-			    subvol, subvol->fops->stat,
-			    loc);
-	}
+int
+dht_vgetxattr_alloc_and_fill (dht_local_t *local, dict_t *xattr, xlator_t *this,
+                              int op_errno)
+{
+        int      ret       = -1;
+        char    *value     = NULL;
+        int32_t  plen      = 0;
 
-	return 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;
+        }
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (stat, frame, -1, op_errno, NULL);
+        local->alloc_len += strlen(value);
 
-	return 0;
-}
+        if (!local->xattr_val) {
+                local->alloc_len += (strlen (DHT_PATHINFO_HEADER) + 10);
+                local->xattr_val = GF_CALLOC (local->alloc_len, sizeof (char),
+                                              gf_common_mt_char);
+                if (!local->xattr_val) {
+                        ret = -1;
+                        goto out;
+                }
+        }
 
+        if (local->xattr_val) {
+                plen = strlen (local->xattr_val);
+                if (plen) {
+                        /* extra byte(s) for \0 to be safe */
+                        local->alloc_len += (plen + 2);
+                        local->xattr_val = GF_REALLOC (local->xattr_val,
+                                                       local->alloc_len);
+                        if (!local->xattr_val) {
+                                ret = -1;
+                                goto out;
+                        }
+                }
 
-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;
+                (void) strcat (local->xattr_val, value);
+                (void) strcat (local->xattr_val, " ");
+                local->op_ret = 0;
+        }
 
+        ret = 0;
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+ out:
+        return ret;
+}
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+int
+dht_vgetxattr_fill_and_set (dht_local_t *local, dict_t **dict, xlator_t *this,
+                            gf_boolean_t flag)
+{
+        int   ret             = -1;
+        char *xattr_buf       = NULL;
+        char layout_buf[8192] = {0,};
 
-	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 (flag)
+                fill_layout_info (local->layout, layout_buf);
 
-	local->inode    = inode_ref (fd->inode);
-	local->call_cnt = layout->cnt;;
+        *dict = dict_new ();
+        if (!*dict)
+                goto out;
 
-	for (i = 0; i < layout->cnt; i++) {
-		subvol = layout->list[i].xlator;
-		STACK_WIND (frame, dht_attr_cbk,
-			    subvol, subvol->fops->fstat,
-			    fd);
-	}
+        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;
 
-	return 0;
+        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 {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Unknown local->xsel (%s)", local->xsel);
+                GF_FREE (xattr_buf);
+                goto out;
+        }
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (fstat, frame, -1, op_errno, NULL);
+        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_truncate (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, off_t offset)
+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, out);
+        VALIDATE_OR_GOTO (frame->local, 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);
+        local = frame->local;
 
-	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;
-	}
+        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 = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+                        goto unlock;
+                }
 
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = 1;
+                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);
 
-	STACK_WIND (frame, dht_truncate_cbk,
-		    subvol, subvol->fops->truncate,
-		    loc, offset);
+        if (!is_last_call (this_call_cnt))
+                goto out;
 
-	return 0;
+        /* -- last call: do patch ups -- */
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL);
+        if (local->op_ret == -1) {
+                goto unwind;
+        }
 
-	return 0;
-}
+        ret = dht_vgetxattr_fill_and_set (local, &dict, this, _gf_true);
+        if (ret)
+                goto unwind;
 
+        DHT_STACK_UNWIND (getxattr, frame, 0, 0, dict, xdata);
+        goto cleanup;
+
+ unwind:
+        DHT_STACK_UNWIND (getxattr, frame, -1, local->op_errno, NULL, NULL);
+ cleanup:
+        if (dict)
+                dict_unref (dict);
+ out:
+        return 0;
+}
 
 int
-dht_ftruncate (call_frame_t *frame, xlator_t *this,
-	       fd_t *fd, off_t offset)
+dht_vgetxattr_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;
-
+        dht_local_t  *local         = NULL;
+        int           ret           = 0;
+        dict_t       *dict          = NULL;
+        call_frame_t *prev          = NULL;
+        gf_boolean_t  flag          = _gf_true;
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        local = frame->local;
+        prev = cookie;
 
-	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;
-	}
+        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;
+        }
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        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;
+        }
 
-	local->inode = inode_ref (fd->inode);
-	local->call_cnt = 1;
+        flag = (local->layout->cnt > 1) ? _gf_true : _gf_false;
 
-	STACK_WIND (frame, dht_truncate_cbk,
-		    subvol, subvol->fops->ftruncate,
-		    fd, offset);
+        ret = dht_vgetxattr_fill_and_set (local, &dict, this, flag);
+        if (ret)
+                goto unwind;
 
-	return 0;
+        DHT_STACK_UNWIND (getxattr, frame, 0, 0, dict, xdata);
+        goto cleanup;
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL);
+ unwind:
+        DHT_STACK_UNWIND (getxattr, frame, -1, local->op_errno,
+                          NULL, NULL);
+ cleanup:
+        if (dict)
+                dict_unref (dict);
 
-	return 0;
+        return 0;
 }
 
-
 int
-dht_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                int op_ret, int op_errno, struct stat *preparent,
-                struct stat *postparent)
+dht_linkinfo_getxattr_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;
-
-
-	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   ret   = 0;
+        char *value = NULL;
+
+        if (op_ret != -1) {
+                ret = dict_get_str (xattr, GF_XATTR_PATHINFO_KEY, &value);
+                if (!ret) {
+                        ret = dict_set_str (xattr, GF_XATTR_LINKINFO_KEY, value);
+                        if (!ret)
+                                gf_log (this->name, GF_LOG_TRACE,
+                                        "failed to set linkinfo");
+                }
+        }
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (unlink, frame, local->op_ret, local->op_errno,
-                                  NULL, NULL);
+        DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr, xdata);
 
         return 0;
 }
 
-
 int
-dht_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this, int op_ret,
-               int op_errno, struct stat *prebuf, struct stat *postbuf)
+dht_getxattr_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             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);
 
-	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;
-		}
+        conf = this->private;
+        local = frame->local;
 
-		local->op_ret = 0;
-	}
-unlock:
-	UNLOCK (&frame->lock);
+        this_call_cnt = dht_frame_return (frame);
+
+        if (!xattr || (op_ret == -1))
+                goto out;
 
-        if (local && (op_ret == 0)) {
-                prebuf->st_ino = local->st_ino;
-                postbuf->st_ino = local->st_ino;
+        if (dict_get (xattr, conf->xattr_name)) {
+                dict_del (xattr, conf->xattr_name);
         }
 
-	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);
+        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 {
+                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, 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_err_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	     int op_ret, int op_errno)
+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)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
+        int             this_call_cnt = 0;
+        dht_local_t     *local = 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;
+
+	if (op_ret != -1) {
+		if (local->xattr)
+			dict_unref (local->xattr);
+		local->xattr = dict_ref (xattr);
 
-		local->op_ret = 0;
+		if (local->xattr_req)
+			dict_unref (local->xattr_req);
+		local->xattr_req = dict_ref (xdata);
 	}
-unlock:
-	UNLOCK (&frame->lock);
 
 	this_call_cnt = dht_frame_return (frame);
 	if (is_last_call (this_call_cnt)) {
-		DHT_STACK_UNWIND (setxattr, frame, local->op_ret, local->op_errno);
-        }
+		DHT_STACK_UNWIND (getxattr, frame, local->op_ret, op_errno,
+				  local->xattr, local->xattr_req);
+	}
 
-        return 0;
+	return 0;
 }
 
 
 int
-dht_access (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, int32_t mask)
+dht_getxattr_get_real_filename (call_frame_t *frame, xlator_t *this,
+				loc_t *loc, const char *key, 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;
-	}
+	dht_local_t     *local = NULL;
+	int              i = 0;
+	dht_layout_t    *layout = NULL;
+	int              cnt = 0;
+	xlator_t        *subvol = NULL;
 
-	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;
+	local = frame->local;
+	layout = local->layout;
 
-	STACK_WIND (frame, dht_err_cbk,
-		    subvol, subvol->fops->access,
-		    loc, mask);
+	cnt = local->call_cnt = layout->cnt;
 
-	return 0;
+	local->op_ret = -1;
+	local->op_errno = ENODATA;
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (access, frame, -1, op_errno);
+	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_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		  int op_ret, int op_errno, const char *path, struct stat *sbuf)
+dht_getxattr (call_frame_t *frame, xlator_t *this,
+              loc_t *loc, const char *key, dict_t *xdata)
+#define DHT_IS_DIR(layout)  (layout->cnt > 1)
 {
-        dht_local_t *local = NULL;
 
-        local = frame->local;
-        if (local) {
-                sbuf->st_ino = local->st_ino;
-        } else {
-                op_ret = -1;
-                op_errno = EINVAL;
-        }
-
-        DHT_STACK_UNWIND (readlink, frame, op_ret, op_errno, path, sbuf);
-
-        return 0;
-}
-
-
-int
-dht_readlink (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, size_t size)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-        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           i             = 0;
+        int           cnt           = 0;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (loc, err);
         VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	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, loc, NULL, GF_FOP_GETXATTR);
+        if (!local) {
+                op_errno = ENOMEM;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
+                goto err;
+        }
+
+        layout = local->layout;
+        if (!layout) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "layout is NULL");
+                op_errno = ENOENT;
+                goto err;
+        }
+
+        if (key) {
+                local->key = gf_strdup (key);
+                if (!local->key) {
+                        op_errno = ENOMEM;
+                        goto err;
+                }
+        }
+
+	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;
 	}
 
-        local->st_ino = loc->inode->ino;
- 
-	STACK_WIND (frame, dht_readlink_cbk,
-		    subvol, subvol->fops->readlink,
-		    loc, size);
+        /* 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;
+        }
 
-	return 0;
+        /* 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);
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (readlink, frame, -1, op_errno, NULL, NULL);
+                local->call_cnt = 1;
+                STACK_WIND (frame, dht_vgetxattr_cbk, cached_subvol,
+                            cached_subvol->fops->getxattr, loc, key, NULL);
 
-	return 0;
-}
+                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;
+                }
 
-int
-dht_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		  int op_ret, int op_errno, dict_t *xattr)
-{
-        DHT_STACK_UNWIND (getxattr, frame, op_ret, op_errno, xattr);
+                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;
+                }
 
-        return 0;
-}
+                if (hashed_subvol == cached_subvol) {
+                        op_errno = ENODATA;
+                        goto err;
+                }
+                if (hashed_subvol) {
+                        STACK_WIND (frame, dht_linkinfo_getxattr_cbk, hashed_subvol,
+                                    hashed_subvol->fops->getxattr, loc,
+                                    GF_XATTR_PATHINFO_KEY, NULL);
+                        return 0;
+                }
+                op_errno = ENODATA;
+                goto err;
+        }
 
+        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;
+                }
 
-int
-dht_getxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, const char *key)
-{
-	xlator_t     *subvol = NULL;
-        int           op_errno = -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_UUID_TYPE, marker_uuid_default_gauge,
+                                           conf->vol_uuid)) {
+                        op_errno = EINVAL;
+                        goto err;
+                }
 
-        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);
+                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;
-	}
+        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;
+        }
 
-	STACK_WIND (frame, dht_getxattr_cbk,
-		    subvol, subvol->fops->getxattr,
-		    loc, key);
+        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;
+                        }
 
-	return 0;
+                        return 0;
+                }
+        }
+
+        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);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (getxattr, frame, -1, op_errno, NULL, NULL);
 
-	return 0;
+        return 0;
 }
-
+#undef DHT_IS_DIR
 
 int
-dht_setxattr (call_frame_t *frame, xlator_t *this,
-	      loc_t *loc, dict_t *xattr, int flags)
+dht_fgetxattr (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, const char *key, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
+        xlator_t     *subvol        = NULL;
+        dht_local_t  *local         = NULL;
+        dht_layout_t *layout        = NULL;
+        int           op_errno      = -1;
+        int           i             = 0;
+        int           cnt           = 0;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (fd->inode, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FGETXATTR);
+        if (!local) {
+                op_errno = ENOMEM;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+                goto err;
+        }
+
+        layout = local->layout;
+        if (!layout) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "layout is NULL");
+                op_errno = ENOENT;
+                goto err;
+        }
 
-	local->call_cnt = 1;
+        if (key) {
+                local->key = gf_strdup (key);
+                if (!local->key) {
+                        op_errno = ENOMEM;
+                        goto err;
+                }
+        }
 
-	STACK_WIND (frame, dht_err_cbk,
-		    subvol, subvol->fops->setxattr,
-		    loc, xattr, flags);
+        if ((fd->inode->ia_type == IA_IFDIR)
+	    && 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;
+        }
 
-	return 0;
+        for (i = 0; i < cnt; i++) {
+                subvol = layout->list[i].xlator;
+                STACK_WIND (frame, dht_getxattr_cbk,
+                            subvol, subvol->fops->fgetxattr,
+                            fd, key, NULL);
+        }
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (setxattr, frame, -1, op_errno);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fgetxattr, frame, -1, op_errno, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
-
 int
-dht_removexattr (call_frame_t *frame, xlator_t *this,
-		 loc_t *loc, const char *key)
+dht_fsetxattr (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, dict_t *xattr, int flags, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
+        xlator_t     *subvol   = NULL;
+        dht_local_t  *local    = NULL;
+        int           op_errno = EINVAL;
+        dht_conf_t   *conf     = NULL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (fd->inode, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        conf = this->private;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        GF_IF_INTERNAL_XATTR_GOTO (conf->wild_xattr_name, xattr,
+                                   op_errno, err);
 
-	local->call_cnt = 1;
+        local = dht_local_init (frame, NULL, fd, GF_FOP_FSETXATTR);
+        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;
+        }
 
-	STACK_WIND (frame, dht_err_cbk,
-		    subvol, subvol->fops->removexattr,
-		    loc, key);
+        local->call_cnt = 1;
 
-	return 0;
+        STACK_WIND (frame, dht_err_cbk, subvol, subvol->fops->fsetxattr,
+                    fd, xattr, flags, NULL);
+
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (removexattr, frame, -1, op_errno);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fsetxattr, frame, -1, op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
 
+static int
+dht_common_setxattr_cbk (call_frame_t *frame, void *cookie,
+                         xlator_t *this, int32_t op_ret, int32_t op_errno,
+                         dict_t *xdata)
+{
+        DHT_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
+
+        return 0;
+}
+
 int
-dht_fd_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	      int op_ret, int op_errno, fd_t *fd)
+dht_checking_pathinfo_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                           int op_ret, int op_errno, dict_t *xattr,
+                           dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
+        int           i     = -1;
+        int           ret   = -1;
+        char         *value = NULL;
+        dht_local_t  *local = NULL;
+        dht_conf_t   *conf  = NULL;
+        call_frame_t *prev  = NULL;
+        int           this_call_cnt = 0;
 
+        local = frame->local;
+        prev = cookie;
+        conf = this->private;
 
-	local = frame->local;
-	prev = cookie;
-
-	LOCK (&frame->lock);
-	{
-		if (op_ret == -1) {
-			local->op_errno = op_errno;
-			gf_log (this->name, GF_LOG_DEBUG,
-				"subvolume %s returned -1 (%s)",
-				prev->this->name, strerror (op_errno));
-			goto unlock;
-		}
+        if (op_ret == -1)
+                goto out;
 
-		local->op_ret = 0;
-	}
-unlock:
-	UNLOCK (&frame->lock);
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (open, frame, local->op_ret, local->op_errno,
-				  local->fd);
+        ret = dict_get_str (xattr, GF_XATTR_PATHINFO_KEY, &value);
+        if (ret)
+                goto out;
+
+        if (!strcmp (value, local->key)) {
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (conf->subvolumes[i] == prev->this)
+                                conf->decommissioned_bricks[i] = prev->this;
+                }
+        }
 
+out:
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                DHT_STACK_UNWIND (setxattr, frame, local->op_ret, ENOTSUP, NULL);
+        }
         return 0;
-}
 
+}
 
 int
-dht_open (call_frame_t *frame, xlator_t *this,
-	  loc_t *loc, int flags, fd_t *fd, int wbflags)
+dht_setxattr (call_frame_t *frame, xlator_t *this,
+              loc_t *loc, dict_t *xattr, int flags, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
-	int           ret = -1;
-        int           op_errno = -1;
-	dht_local_t  *local = NULL;
-
+        xlator_t     *subvol   = NULL;
+        dht_local_t  *local    = NULL;
+        dht_conf_t   *conf     = NULL;
+        dht_layout_t *layout   = NULL;
+        int           i        = 0;
+        int           op_errno = EINVAL;
+        int           ret      = -1;
+        data_t       *tmp      = NULL;
+        uint32_t      dir_spread = 0;
+        char          value[4096] = {0,};
+        gf_dht_migrate_data_type_t forced_rebalance = GF_DHT_MIGRATE_DATA;
+        int           call_cnt = 0;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
-
-	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;
-	}
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        conf   = this->private;
 
-	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;
-	}
+        GF_IF_INTERNAL_XATTR_GOTO (conf->wild_xattr_name, xattr,
+                                   op_errno, err);
 
-	local->call_cnt = 1;
+        local = dht_local_init (frame, loc, NULL, GF_FOP_SETXATTR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	STACK_WIND (frame, dht_fd_cbk,
-		    subvol, subvol->fops->open,
-		    loc, flags, fd, wbflags);
+        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;
+        }
 
-	return 0;
+        layout = local->layout;
+        if (!layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no layout for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (open, frame, -1, op_errno, NULL);
+        local->call_cnt = call_cnt = layout->cnt;
 
-	return 0;
-}
+        tmp = dict_get (xattr, "distribute.migrate-data");
+        if (tmp) {
+                if (IA_ISDIR (loc->inode->ia_type)) {
+                        op_errno = ENOTSUP;
+                        goto err;
+                }
 
+                /* 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;
 
-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 stat *stbuf,
-               struct iobref *iobref)
-{
-        DHT_STACK_UNWIND (readv, frame, op_ret, op_errno, vector, count, stbuf,
-                          iobref);
+                if (conf->decommission_in_progress)
+                        forced_rebalance = GF_DHT_MIGRATE_HARDLINK;
 
-        return 0;
-}
+                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;
+                }
 
+                local->rebalance.from_subvol = local->cached_subvol;
 
-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;
+                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;
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "%s: failed to create a new synctask",
+                                loc->path);
+                }
+                op_errno = EINVAL;
+                goto err;
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+        }
 
-	STACK_WIND (frame, dht_readv_cbk,
-		    subvol, subvol->fops->readv,
-		    fd, size, off);
+        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;
+                }
 
-	return 0;
+                memcpy (value, tmp->data, ((tmp->len < 4095) ? tmp->len : 4095));
+                local->key = gf_strdup (value);
+                local->call_cnt = conf->subvolume_cnt;
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, 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;
+        }
 
-	return 0;
-}
+        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);
 
+                ret = dht_fix_directory_layout (frame, dht_common_setxattr_cbk,
+                                                layout);
+                if (ret) {
+                        op_errno = ENOTCONN;
+                        goto err;
+                }
+                return ret;
+        }
 
-int
-dht_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		int op_ret, int op_errno, struct stat *prebuf,
-                struct stat *postbuf)
-{
-        dht_local_t *local = NULL;
+        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;
+        }
 
-        if (op_ret == -1) {
-                goto out;
+        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);
         }
 
-        local = frame->local;
-        if (!local) {
-                op_ret = -1;
-                op_errno = EINVAL;
-                goto out;
-        } 
-        
-        prebuf->st_ino = local->st_ino;
-        postbuf->st_ino = local->st_ino;
+        return 0;
 
-out:
-        DHT_STACK_UNWIND (writev, frame, op_ret, op_errno, prebuf, postbuf);
+err:
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (setxattr, frame, -1, 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_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int op_ret, int op_errno, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
         dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        local = frame->local;
+        prev = cookie;
 
-	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;
-	}
+        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 = dht_local_init (frame);
-        if (!local) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                op_errno = ENOMEM;
-                goto err;
+                local->op_ret = 0;
         }
+unlock:
+        UNLOCK (&frame->lock);
 
-        local->st_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);
+        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;
+        return 0;
 }
 
 
 int
-dht_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
+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_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;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        GF_IF_NATIVE_XATTR_GOTO (conf->wild_xattr_name, key, op_errno, err);
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
 
-	local->fd = fd_ref (fd);
-	local->call_cnt = 1;
+        local = dht_local_init (frame, loc, NULL, GF_FOP_REMOVEXATTR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	STACK_WIND (frame, dht_err_cbk,
-		    subvol, subvol->fops->flush, fd);
+        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;
+        }
 
-	return 0;
+        layout = local->layout;
+        if (!local->layout) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no layout for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        local->call_cnt = call_cnt = layout->cnt;
+        local->key = gf_strdup (key);
+
+        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;
 
 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->st_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);
+        }
 
-int
-dht_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	    int op_ret, int op_errno, struct flock *flock)
-{
-        DHT_STACK_UNWIND (lk, frame, op_ret, op_errno, flock);
+        return 0;
+
+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 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;
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
 
+        local = frame->local;
+        prev = cookie;
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        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;
+                }
 
-	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->op_ret = 0;
+        }
+unlock:
+        UNLOCK (&frame->lock);
 
-	STACK_WIND (frame, dht_lk_cbk,
-		    subvol, subvol->fops->lk,
-		    fd, cmd, flock);
+        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;
+}
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (lk, frame, -1, op_errno, NULL);
+/*
+ * dht_normalize_stats -
+ */
+static void
+dht_normalize_stats (struct statvfs *buf, unsigned long bsize,
+                     unsigned long frsize)
+{
+        double factor = 0;
 
-	return 0;
-}
+        if (buf->f_bsize != bsize) {
+                buf->f_bsize = bsize;
+        }
 
+        if (buf->f_frsize != frsize) {
+                factor = ((double) buf->f_frsize) / frsize;
+                buf->f_frsize = frsize;
+                buf->f_blocks = (fsblkcnt_t) (factor * buf->f_blocks);
+                buf->f_bfree  = (fsblkcnt_t) (factor * buf->f_bfree);
+                buf->f_bavail = (fsblkcnt_t) (factor * buf->f_bavail);
+
+        }
+}
 
 int
 dht_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		int op_ret, int op_errno, struct statvfs *statvfs)
+                int op_ret, int op_errno, struct statvfs *statvfs, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
+        dht_local_t *local              = NULL;
+        int          this_call_cnt      = 0;
+        int          bsize              = 0;
+        int          frsize             = 0;
+        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;
-
-		/* TODO: normalize sizes */
-		local->statvfs.f_bsize    = statvfs->f_bsize;
-		local->statvfs.f_frsize   = statvfs->f_frsize;
-
-		local->statvfs.f_blocks  += statvfs->f_blocks;
-		local->statvfs.f_bfree   += statvfs->f_bfree;
-		local->statvfs.f_bavail  += statvfs->f_bavail;
-		local->statvfs.f_files   += statvfs->f_files;
-		local->statvfs.f_ffree   += statvfs->f_ffree;
-		local->statvfs.f_favail  += statvfs->f_favail;
-		local->statvfs.f_fsid     = statvfs->f_fsid;
-		local->statvfs.f_flag     = statvfs->f_flag;
-		local->statvfs.f_namemax  = statvfs->f_namemax;
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        goto unlock;
+                }
+                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);
+                        frsize = max(local->statvfs.f_frsize, statvfs->f_frsize);
+                        dht_normalize_stats(&local->statvfs, bsize, frsize);
+                        dht_normalize_stats(statvfs, bsize, frsize);
+                } else {
+                        local->statvfs.f_bsize    = statvfs->f_bsize;
+                        local->statvfs.f_frsize   = statvfs->f_frsize;
+                }
+
+                local->statvfs.f_blocks  += statvfs->f_blocks;
+                local->statvfs.f_bfree   += statvfs->f_bfree;
+                local->statvfs.f_bavail  += statvfs->f_bavail;
+                local->statvfs.f_files   += statvfs->f_files;
+                local->statvfs.f_ffree   += statvfs->f_ffree;
+                local->statvfs.f_favail  += statvfs->f_favail;
+                local->statvfs.f_fsid     = statvfs->f_fsid;
+                local->statvfs.f_flag     = statvfs->f_flag;
+                local->statvfs.f_namemax  = statvfs->f_namemax;
+
+
+        }
 unlock:
-	UNLOCK (&frame->lock);
+        UNLOCK (&frame->lock);
 
 
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt))
-		DHT_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno,
-				  &local->statvfs);
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt))
+                DHT_STACK_UNWIND (statfs, frame, local->op_ret, local->op_errno,
+                                  &local->statvfs, xdata);
 
         return 0;
 }
 
 
 int
-dht_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc)
+dht_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
-	dht_local_t  *local  = NULL;
-	dht_conf_t   *conf = NULL;
+        xlator_t     *subvol = NULL;
+        dht_local_t  *local  = NULL;
+        dht_conf_t   *conf = NULL;
         int           op_errno = -1;
-	int           i = -1;
-
+        int           i = -1;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (loc, err);
         VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (this->private, err);
+
+        conf = this->private;
+
+        local = dht_local_init (frame, NULL, NULL, GF_FOP_STATFS);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	conf = this->private;
+        if (IA_ISDIR (loc->inode->ia_type)) {
+                local->call_cnt = conf->subvolume_cnt;
 
-	local = dht_local_init (frame);
-	local->call_cnt = conf->subvolume_cnt;
+                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;
+        }
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_statfs_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->statfs, loc);
-	}
+        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;
+        }
 
-	return 0;
+        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)
-{
-	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;
+                  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;
         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;
+        if (op_ret < 0)
+                goto done;
 
-	if (op_ret < 0)
-		goto done;
+        if (!local->layout)
+                local->layout = dht_layout_get (this, local->fd->inode);
 
-	list_for_each_entry (orig_entry, (&orig_entries->list), list) {
-                next_offset = orig_entry->d_off;
+        layout = local->layout;
 
-                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)))) {
+        list_for_each_entry (orig_entry, (&orig_entries->list), list) {
+                next_offset = orig_entry->d_off;
+                if (check_is_dir (NULL, (&orig_entry->d_stat), NULL) &&
+                    (prev->this != 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;
                 }
 
-                entry->d_stat = orig_entry->d_stat;
+                /* Do this if conf->search_unhashed is set to "auto" */
+                if (conf->search_unhashed == GF_DHT_LOOKUP_UNHASHED_AUTO) {
+                        subvol = dht_layout_search (this, layout,
+                                                    orig_entry->d_name);
+                        if (!subvol || (subvol != prev->this)) {
+                                /* TODO: Count the number of entries which need
+                                   linkfile to prove its existence in fs */
+                                layout->search_unhashed++;
+                        }
+                }
 
-                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.st_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
+         * posix for each subvolume sends a ENOENT on end-of-directory but in
+         * distribute we're not concerned only with a posix's view of the
+         * directory but the aggregated namespace' view of the directory.
+         */
+        if (prev->this != dht_last_up_subvol (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
                 */
@@ -2089,23 +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;
 }
@@ -2114,56 +3196,69 @@ 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)
-{
-	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;
+                 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;
         off_t         next_offset = 0;
-	int           count = 0;
+        int           count = 0;
+        dht_layout_t *layout = 0;
+        xlator_t     *subvol = 0;
 
+        INIT_LIST_HEAD (&entries.list);
+        prev = cookie;
+        local = frame->local;
 
-	INIT_LIST_HEAD (&entries.list);
-	prev = cookie;
-	local = frame->local;
+        if (op_ret < 0)
+                goto done;
+
+        if (!local->layout)
+                local->layout = dht_layout_get (this, local->fd->inode);
 
-	if (op_ret < 0)
-		goto done;
+        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)))) {
-                        continue;
-                }
+                subvol = dht_layout_search (this, layout, orig_entry->d_name);
 
-                entry = gf_dirent_for_name (orig_entry->d_name);
-                if (!entry) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "Out of memory");
-                        goto unwind;
-                }
+                if (!subvol || (subvol == prev->this)) {
+                        entry = gf_dirent_for_name (orig_entry->d_name);
+                        if (!entry) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "memory allocation failed :(");
+                                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);
+                        dht_itransform (this, prev->this, orig_entry->d_off,
+                                        &entry->d_off);
 
-                entry->d_type = orig_entry->d_type;
-                entry->d_len  = orig_entry->d_len;
+                        entry->d_ino  = orig_entry->d_ino;
+                        entry->d_type = orig_entry->d_type;
+                        entry->d_len  = orig_entry->d_len;
 
-                list_add_tail (&entry->list, &entries.list);
-                count++;
-	}
-	op_ret = count;
+                        list_add_tail (&entry->list, &entries.list);
+                        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
+         * posix for each subvolume sends a ENOENT on end-of-directory but in
+         * distribute we're not concerned only with a posix's view of the
+         * directory but the aggregated namespace' view of the directory.
+         */
+        if (prev->this != dht_last_up_subvol (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
                 */
@@ -2173,23 +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;
 }
@@ -2197,65 +3292,113 @@ 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;
-
-	local = dht_local_init (frame);
-	if (!local) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
+        conf = this->private;
 
-	local->fd = fd_ref (fd);
-	local->size = size;
+        local = dht_local_init (frame, NULL, NULL, whichop);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	dht_deitransform (this, yoff, &xvol, (uint64_t *)&xoff);
+        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);
+
+        /* 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)
 {
-        dht_do_readdir (frame, this, fd, size, yoff, GF_FOP_READDIR);
+        int          op = GF_FOP_READDIR;
+        dht_conf_t  *conf = NULL;
+        int          i = 0;
+
+        conf = this->private;
+        if (!conf)
+                goto out;
+
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (!conf->subvolume_status[i]) {
+                        op = GF_FOP_READDIRP;
+                        break;
+                }
+        }
+
+        if (conf->use_readdirp)
+                op = GF_FOP_READDIRP;
+
+out:
+        dht_do_readdir (frame, this, fd, size, yoff, op, 0);
         return 0;
 }
 
 int
 dht_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-              off_t yoff)
+              off_t yoff, dict_t *dict)
 {
-        dht_do_readdir (frame, this, fd, size, yoff, GF_FOP_READDIRP);
+        dht_do_readdir (frame, this, fd, size, yoff, GF_FOP_READDIRP, dict);
         return 0;
 }
 
@@ -2263,88 +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,
-                 inode_t *inode, struct stat *stbuf, struct stat *preparent,
-                 struct stat *postparent)
+                 int op_ret, int op_errno,
+                 inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
+                 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) {
@@ -2353,421 +3496,416 @@ dht_newfile_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 goto out;
         }
 
-	prev = cookie;
+        prev = cookie;
 
-	dht_itransform (this, prev->this, stbuf->st_ino, &stbuf->st_ino);
         if (local->loc.parent) {
-                preparent->st_ino = local->loc.parent->ino;
-                postparent->st_ino = local->loc.parent->ino;
+
+                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: st_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_stat_merge for reference.
-         */ 
-
-	DHT_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode, stbuf, preparent,
-                          postparent);
-	return 0;
+         * See dht_iatt_merge for reference.
+         */
+        DHT_STRIP_PHASE1_FLAGS (stbuf);
+        DHT_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode, stbuf,
+                          preparent, postparent, xdata);
+        return 0;
 }
 
 int
 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 stat *stbuf,
-                               struct stat *preparent, struct stat *postparent)
+                               inode_t *inode, struct iatt *stbuf,
+                               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;
+        }
 
-        STACK_WIND (frame, dht_newfile_cbk,
-                    cached_subvol, cached_subvol->fops->mknod,
-                    &local->loc, local->mode, local->rdev);
+        cached_subvol = local->cached_subvol;
 
+        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,
+                          NULL);
         return 0;
- err:
- 	DHT_STACK_UNWIND (mknod, frame, -1, op_errno, 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)
+           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);
+
+                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);
+
+                        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)
+             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;
 
+        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);
-
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        local = dht_local_init (frame, loc, NULL, GF_FOP_SYMLINK);
+        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;
-	}
+        subvol = dht_subvol_get_hashed (this, loc);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        loc->path);
+                op_errno = ENOENT;
+                goto err;
+        }
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"creating %s on %s", loc->path, subvol->name);
+        gf_log (this->name, GF_LOG_TRACE,
+                "creating %s on %s", loc->path, subvol->name);
 
-	STACK_WIND (frame, dht_newfile_cbk,
-		    subvol, subvol->fops->symlink,
-		    linkname, loc);
+        STACK_WIND_COOKIE (frame, dht_newfile_cbk, (void *)subvol, subvol,
+                           subvol->fops->symlink, linkname, loc, umask,
+                           params);
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (link, frame, -1, op_errno,
-                          NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (link, frame, -1, op_errno,
+                          NULL, NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
-dht_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
+dht_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+            dict_t *xdata)
 {
-	xlator_t    *cached_subvol = NULL;
-	xlator_t    *hashed_subvol = NULL;
-        int          ret = -1;
-	int          op_errno = -1;
-	dht_local_t *local = NULL;
-
-
-	VALIDATE_OR_GOTO (frame, err);
-	VALIDATE_OR_GOTO (this, err);
-	VALIDATE_OR_GOTO (loc, err);
-
-	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;
-	}
-
-	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;
-	}
+        xlator_t    *cached_subvol = NULL;
+        xlator_t    *hashed_subvol = NULL;
+        int          op_errno = -1;
+        dht_local_t *local = NULL;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		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) {
-		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,
+                                       &cached_subvol)) {
+                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, xflag, xdata);
+                goto done;
+        }
 
-	local->call_cnt = 1;
-	if (hashed_subvol != cached_subvol)
-		local->call_cnt++;
+        local = dht_local_init (frame, loc, NULL, GF_FOP_UNLINK);
+        if (!local) {
+                op_errno = ENOMEM;
 
-	STACK_WIND (frame, dht_unlink_cbk,
-		    cached_subvol, cached_subvol->fops->unlink, loc);
+                goto err;
+        }
 
-	if (hashed_subvol != cached_subvol)
-		STACK_WIND (frame, dht_unlink_cbk,
-			    hashed_subvol, hashed_subvol->fops->unlink, loc);
+        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);
+        }
 
-	return 0;
+        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;
+        }
 
+        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,
+                            xflag, xdata);
+        }
+done:
+        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 (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,
-              inode_t *inode, struct stat *stbuf, struct stat *preparent,
-              struct stat *postparent)
+              int op_ret, int op_errno,
+              inode_t *inode, struct iatt *stbuf, struct iatt *preparent,
+              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->st_ino = local->loc.inode->ino;
-
-        preparent->st_ino = local->loc2.parent->ino;
-        postparent->st_ino = local->loc2.parent->ino;
+        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,
-                       inode_t *inode, struct stat *stbuf,
-                       struct stat *preparent, struct stat *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;
-	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)
-{
-	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;
-	}
+          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;
 
-	hashed_subvol = dht_subvol_get_hashed (this, newloc);
-	if (!hashed_subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no subvolume in layout for path=%s",
-			newloc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (oldloc, err);
+        VALIDATE_OR_GOTO (newloc, err);
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        local = dht_local_init (frame, oldloc, NULL, GF_FOP_LINK);
+        if (!local) {
+                op_errno = ENOMEM;
 
-	ret = loc_copy (&local->loc, oldloc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+                goto err;
+        }
 
-	ret = loc_copy (&local->loc2, newloc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        cached_subvol = local->cached_subvol;
+        if (!cached_subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", oldloc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	if (hashed_subvol != cached_subvol) {
-		dht_linkfile_create (frame, dht_link_linkfile_cbk,
-				     cached_subvol, hashed_subvol, newloc);
-	} else {
-		STACK_WIND (frame, dht_link_cbk,
-			    cached_subvol, cached_subvol->fops->link,
-			    oldloc, newloc);
-	}
+        hashed_subvol = dht_subvol_get_hashed (this, newloc);
+        if (!hashed_subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no subvolume in layout for path=%s",
+                        newloc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	return 0;
+        ret = loc_copy (&local->loc2, newloc);
+        if (ret == -1) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        if (hashed_subvol != cached_subvol) {
+                uuid_copy (local->gfid, oldloc->inode->gfid);
+                dht_linkfile_create (frame, dht_link_linkfile_cbk, this,
+                                     cached_subvol, hashed_subvol, newloc);
+        } else {
+                STACK_WIND (frame, dht_link_cbk,
+                            cached_subvol, cached_subvol->fops->link,
+                            oldloc, newloc, xdata);
+        }
+
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		 int op_ret, int op_errno,
-		 fd_t *fd, inode_t *inode, struct stat *stbuf,
-                 struct stat *preparent, struct stat *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) {
@@ -2776,98 +3914,105 @@ dht_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 goto out;
         }
 
-	prev = cookie;
+        prev = cookie;
 
-	dht_itransform (this, prev->this, stbuf->st_ino, &stbuf->st_ino);
         if (local->loc.parent) {
-                preparent->st_ino = local->loc.parent->ino;
-                postparent->st_ino = local->loc.parent->ino;
+                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 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,
-                                inode_t *inode, struct stat *stbuf,
-                                struct stat *preparent, struct stat *postparent)
+                                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)
 {
-	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->loc, local->flags, local->mode,
+                    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)
+            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;
+        }
 
-	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_filter_loc_subvol_key (this, loc, &local->loc,
+                                       &subvol)) {
+                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, 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");
+        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;
         }
 
@@ -2876,143 +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);
-        } else {
-                /* Choose the minimum filled volume, and create the 
-                   files there */
-                /* TODO */
-                avail_subvol = dht_free_disk_available_subvol (this, subvol);
-                if (avail_subvol != subvol) {
-                        local->fd = fd_ref (fd);
-                        local->flags = flags;
-                        local->mode = mode;
-
-                        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);
-                } else {
-                        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);
-                        
-                }
+                            loc, flags, mode, umask, fd, params);
+                goto done;
         }
-
-	return 0;
+        /* Choose the minimum filled volume, and create the
+           files there */
+        avail_subvol = dht_free_disk_available_subvol (this, subvol, local);
+        if (avail_subvol != subvol) {
+                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,
+                                     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, umask, fd, params);
+done:
+        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.st_ino = local->st_ino;
                 if (local->loc.parent) {
-                        local->preparent.st_ino = local->loc.parent->ino;
-                        local->postparent.st_ino = local->loc.parent->ino;
+                        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 (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 stat *stbuf,
-               struct stat *preparent, struct stat *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;
-	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_stat_merge (this, &local->stbuf, stbuf, prev->this);
-                dht_stat_merge (this, &local->preparent, preparent, prev->this);
-                dht_stat_merge (this, &local->postparent, postparent,
+                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);
-	}
+        }
 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,
-                      inode_t *inode, struct stat *stbuf,
-                      struct stat *preparent, struct stat *postparent)
+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,
+                      dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           ret = -1;
-	call_frame_t *prev = NULL;
-	dht_layout_t *layout = NULL;
-	dht_conf_t   *conf = NULL;
-	int           i = 0;
-	xlator_t     *hashed_subvol = NULL;
+        dht_local_t  *local = NULL;
+        int           ret = -1;
+        call_frame_t *prev = NULL;
+        dht_layout_t *layout = NULL;
+        dht_conf_t   *conf = NULL;
+        int           i = 0;
+        xlator_t     *hashed_subvol = NULL;
 
-	local = frame->local;
-	prev  = cookie;
-	layout = local->layout;
-	conf = this->private;
-	hashed_subvol = local->hashed_subvol;
+        VALIDATE_OR_GOTO (this->private, err);
+
+        local = frame->local;
+        prev  = cookie;
+        layout = local->layout;
+        conf = this->private;
+        hashed_subvol = local->hashed_subvol;
+
+        if (uuid_is_null (local->loc.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,
@@ -3020,48 +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_stat_merge (this, &local->stbuf, stbuf, prev->this);
-        dht_stat_merge (this, &local->preparent, preparent, prev->this);
-        dht_stat_merge (this, &local->postparent, postparent, 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);
 
-	local->st_ino = local->stbuf.st_ino;
+        if (op_ret == -1) {
+                local->op_errno = op_errno;
+                goto err;
+        }
+        local->op_ret = 0;
 
-	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);
-	}
-	return 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->call_cnt = conf->subvolume_cnt - 1;
+
+        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)
+           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);
@@ -3069,144 +4236,234 @@ dht_mkdir (call_frame_t *frame, xlator_t *this,
         VALIDATE_OR_GOTO (loc, err);
         VALIDATE_OR_GOTO (loc->inode, err);
         VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (this->private, err);
 
-	conf = this->private;
+        conf = this->private;
 
         dht_get_du_info (frame, this, loc);
 
-	local = dht_local_init (frame);
-	if (!local) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
-
-	hashed_subvol = dht_subvol_get_hashed (this, loc);
+        local = dht_local_init (frame, loc, NULL, GF_FOP_MKDIR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	if (hashed_subvol == NULL) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"hashed subvol not found for %s",
+        hashed_subvol = dht_subvol_get_hashed (this, loc);
+        if (hashed_subvol == NULL) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "hashed subvol not found for %s",
                         loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
-
-	local->hashed_subvol = hashed_subvol;
-	local->inode = inode_ref (loc->inode);
-	ret = loc_copy (&local->loc, loc);
-	local->mode = mode;
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		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);
+        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.st_ino = local->loc.parent->ino;
-                local->postparent.st_ino = local->loc.parent->ino;
         }
+unlock:
+        UNLOCK (&frame->lock);
 
-	DHT_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
-                          &local->preparent, &local->postparent);
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+               if (local->need_selfheal) {
+                        local->layout =
+                                dht_layout_get (this, local->loc.inode);
 
-	return 0;
+                        /* TODO: neater interface needed below */
+                        local->stbuf.ia_type = local->loc.inode->ia_type;
+
+                        uuid_copy (local->gfid, local->loc.inode->gfid);
+                        dht_selfheal_restore (frame, dht_rmdir_selfheal_cbk,
+                                              &local->loc, local->layout);
+               } else {
+
+                        if (local->loc.parent) {
+                                dht_inode_ctx_time_update (local->loc.parent,
+                                                           this,
+                                                           &local->preparent,
+                                                           0);
+
+                                dht_inode_ctx_time_update (local->loc.parent,
+                                                           this,
+                                                           &local->postparent,
+                                                           1);
+                        }
+
+                        DHT_STACK_UNWIND (rmdir, frame, local->op_ret,
+                                          local->op_errno, &local->preparent,
+                                          &local->postparent, NULL);
+               }
+        }
+
+        return 0;
 }
 
 
 int
 dht_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-	       int op_ret, int op_errno, struct stat *preparent,
-               struct stat *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;
+                }
 
-                dht_stat_merge (this, &local->preparent, preparent, prev->this);
-                dht_stat_merge (this, &local->postparent, postparent,
+                /* 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.st_mode = local->loc.inode->st_mode;
+                        /* 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.st_ino =
-                                        local->loc.parent->ino;
-                                local->postparent.st_ino =
-                                        local->loc.parent->ino;
+
+                                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;
 }
@@ -3215,753 +4472,869 @@ unlock:
 int
 dht_rmdir_do (call_frame_t *frame, xlator_t *this)
 {
-	dht_local_t  *local = NULL;
-	dht_conf_t   *conf = NULL;
-	int           i = 0;
+        dht_local_t  *local = NULL;
+        dht_conf_t   *conf = NULL;
+        int           i = 0;
+        xlator_t     *hashed_subvol = NULL;
 
-	conf = this->private;
-	local = frame->local;
+        VALIDATE_OR_GOTO (this->private, err);
 
-	if (local->op_ret == -1)
-		goto err;
+        conf = this->private;
+        local = frame->local;
 
-	local->call_cnt = conf->subvolume_cnt;
+        if (local->op_ret == -1)
+                goto err;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_rmdir_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->rmdir,
-			    &local->loc);
-	}
+        local->call_cnt = conf->subvolume_cnt;
 
-	return 0;
+        /* first remove from non-hashed_subvol */
+        hashed_subvol = dht_subvol_get_hashed (this, &local->loc);
+
+        if (!hashed_subvol) {
+                gf_log (this->name, GF_LOG_WARNING, "failed to get hashed "
+                        "subvol for %s",local->loc.path);
+        } else {
+                local->hashed_subvol = hashed_subvol;
+        }
+
+        /* When DHT has only 1 child */
+        if (conf->subvolume_cnt == 1) {
+                STACK_WIND (frame, dht_rmdir_hashed_subvol_cbk,
+                            conf->subvolumes[0],
+                            conf->subvolumes[0]->fops->rmdir,
+                            &local->loc, local->flags, NULL);
+                return 0;
+        }
+
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (hashed_subvol &&
+                    (hashed_subvol == conf->subvolumes[i]))
+                        continue;
+
+                STACK_WIND (frame, dht_rmdir_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->rmdir,
+                            &local->loc, local->flags, NULL);
+        }
+
+        return 0;
 
 err:
-	DHT_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
-                          &local->preparent, &local->postparent);
-	return 0;
+        DHT_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
+                          &local->preparent, &local->postparent, NULL);
+        return 0;
 }
 
 
 int
-dht_rmdir_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		       int op_ret, int op_errno, gf_dirent_t *entries)
+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, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = -1;
-	call_frame_t *prev = NULL;
+        dht_local_t    *local = NULL;
+        call_frame_t   *prev = NULL;
+        xlator_t       *src = NULL;
+        call_frame_t   *main_frame = NULL;
+        dht_local_t    *main_local = NULL;
+        int             this_call_cnt = 0;
+
+        local  = frame->local;
+        prev   = cookie;
+        src    = prev->this;
 
-	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;
-	}
+        main_frame = local->main_frame;
+        main_local = main_frame->local;
 
-	this_call_cnt = dht_frame_return (frame);
+        if (op_ret == 0) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "unlinked linkfile %s on %s",
+                        local->loc.path, src->name);
+        } else {
+                main_local->op_ret   = -1;
+                main_local->op_errno = op_errno;
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "unlink of %s on %s failed (%s)",
+                        local->loc.path, src->name, strerror (op_errno));
+        }
 
-	if (is_last_call (this_call_cnt)) {
-		dht_rmdir_do (frame, this);
-	}
+        this_call_cnt = dht_frame_return (main_frame);
+        if (is_last_call (this_call_cnt))
+                dht_rmdir_do (main_frame, this);
 
-	return 0;
+        DHT_STACK_DESTROY (frame);
+        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_rmdir_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;
-	int           this_call_cnt = -1;
-	call_frame_t *prev = NULL;
+        dht_local_t    *local = NULL;
+        call_frame_t   *prev = 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;
 
+        local = frame->local;
+        prev  = cookie;
+        src   = prev->this;
 
-	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;
-	}
+        main_frame = local->main_frame;
+        main_local = main_frame->local;
 
-	STACK_WIND (frame, dht_rmdir_readdir_cbk,
-		    prev->this, prev->this->fops->readdir,
-		    local->fd, 4096, 0);
+        if (op_ret != 0)
+                goto err;
 
-	return 0;
+        if (!check_is_linkfile (inode, stbuf, xattr, conf->link_xattr_name)) {
+                main_local->op_ret  = -1;
+                main_local->op_errno = ENOTEMPTY;
 
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s on %s found to be not a linkfile (type=0%o)",
+                        local->loc.path, src->name, stbuf->ia_type);
+                goto err;
+        }
+
+        STACK_WIND (frame, dht_rmdir_linkfile_unlink_cbk,
+                    src, src->fops->unlink, &local->loc, 0, NULL);
+        return 0;
 err:
-	this_call_cnt = dht_frame_return (frame);
 
-	if (is_last_call (this_call_cnt)) {
-		dht_rmdir_do (frame, this);
-	}
+        this_call_cnt = dht_frame_return (main_frame);
+        if (is_last_call (this_call_cnt))
+                dht_rmdir_do (main_frame, this);
 
-	return 0;
+        DHT_STACK_DESTROY (frame);
+        return 0;
 }
 
 
 int
-dht_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc)
+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;
-	dht_conf_t   *conf = NULL;
-        int           op_errno = -1;
-	int           i = -1;
-	int           ret = -1;
+        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;
 
-        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;
+        main_frame = local->main_frame;
+        main_local = main_frame->local;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto err;
-	}
+        if (op_ret == 0) {
+                main_local->op_ret  = -1;
+                main_local->op_errno = ENOTEMPTY;
 
-	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;
-	}
+                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;
+        }
 
-	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;
-	}
+        xattrs = dict_new ();
+        if (!xattrs) {
+                gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
+                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);
-	}
+        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;
+        }
 
-	return 0;
+        STACK_WIND (frame, dht_rmdir_lookup_cbk,
+                    src, src->fops->lookup, &local->loc, xattrs);
+        if (xattrs)
+                dict_unref (xattrs);
 
+        return 0;
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (rmdir, frame, -1, op_errno,
-                          NULL, NULL);
-
-	return 0;
-}
 
+        this_call_cnt = dht_frame_return (main_frame);
+        if (is_last_call (this_call_cnt))
+                dht_rmdir_do (main_frame, this);
 
-int
-dht_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		 int32_t op_ret, int32_t op_errno, dict_t *dict)
-{
-	DHT_STACK_UNWIND (xattrop, frame, op_ret, op_errno, dict);
-	return 0;
+        DHT_STACK_DESTROY (frame);
+        return 0;
 }
 
 
 int
-dht_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
-	     gf_xattrop_flags_t flags, dict_t *dict)
+dht_rmdir_is_subvol_empty (call_frame_t *frame, xlator_t *this,
+                           gf_dirent_t *entries, xlator_t *src)
 {
-	xlator_t     *subvol = NULL;
-        int           op_errno = -1;
-	dht_local_t  *local = 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;
+        dict_t             *xattrs       = NULL;
+        dht_conf_t         *conf         = this->private;
+        xlator_t           *subvol       = 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 = frame->local;
 
-	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;
-	}
+        list_for_each_entry (trav, &entries->list, list) {
+                if (strcmp (trav->d_name, ".") == 0)
+                        continue;
+                if (strcmp (trav->d_name, "..") == 0)
+                        continue;
+                if (check_is_linkfile (NULL, (&trav->d_stat), trav->dict,
+                                              conf->link_xattr_name)) {
+                        ret++;
+                        continue;
+                }
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+                /* this entry is either a directory which is neither "." nor "..",
+                   or a non directory which is not a linkfile. the directory is to
+                   be treated as non-empty
+                */
+                return 0;
+        }
 
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = 1;
+        xattrs = dict_new ();
+        if (!xattrs) {
+                gf_log (this->name, GF_LOG_ERROR, "dict_new failed");
+                return -1;
+        }
 
-	STACK_WIND (frame,
-		    dht_xattrop_cbk,
-		    subvol, subvol->fops->xattrop,
-		    loc, flags, dict);
+        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;
+        }
 
-	return 0;
+        list_for_each_entry (trav, &entries->list, list) {
+                if (strcmp (trav->d_name, ".") == 0)
+                        continue;
+                if (strcmp (trav->d_name, "..") == 0)
+                        continue;
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (xattrop, frame, -1, op_errno, NULL);
+                lookup_frame = NULL;
+                lookup_local = NULL;
 
-	return 0;
-}
+                lookup_frame = copy_frame (frame);
+                if (!lookup_frame) {
+                        /* out of memory, let the rmdir fail
+                           (as non-empty, unfortunately) */
+                        goto err;
+                }
 
+                lookup_local = mem_get0 (this->local_pool);
+                if (!lookup_local) {
+                        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;
-}
+                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;
 
-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;
+                uuid_copy (lookup_local->loc.gfid, trav->d_stat.ia_gfid);
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+                gf_log (this->name, GF_LOG_TRACE,
+                        "looking up %s on %s",
+                        lookup_local->loc.path, src->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;
-	}
+                LOCK (&frame->lock);
+                {
+                        local->call_cnt++;
+                }
+                UNLOCK (&frame->lock);
 
-	STACK_WIND (frame,
-		    dht_fxattrop_cbk,
-		    subvol, subvol->fops->fxattrop,
-		    fd, flags, dict);
+                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++;
+        }
 
-	return 0;
+        if (xattrs)
+                dict_unref (xattrs);
 
+        return ret;
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (fxattrop, frame, -1, op_errno, NULL);
+        if (xattrs)
+                dict_unref (xattrs);
 
-	return 0;
+        DHT_STACK_DESTROY (lookup_frame);
+        return 0;
 }
 
 
 int
-dht_inodelk_cbk (call_frame_t *frame, void *cookie,
-		 xlator_t *this, int32_t op_ret, int32_t op_errno)
-
+dht_rmdir_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int op_ret, int op_errno, gf_dirent_t *entries,
+                        dict_t *xdata)
 {
-	DHT_STACK_UNWIND (inodelk, frame, op_ret, op_errno);
-	return 0;
+        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;
+        src   = prev->this;
+
+        if (op_ret > 2) {
+                ret = dht_rmdir_is_subvol_empty (frame, this, entries, src);
+
+                switch (ret) {
+                case 0: /* non linkfiles exist */
+                        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;
+                        break;
+                default:
+                        /* @ret number of linkfiles are getting unlinked */
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "readdir on %s for %s found %d linkfiles",
+                                prev->this->name, local->loc.path, ret);
+                        break;
+                }
+        }
+
+        this_call_cnt = dht_frame_return (frame);
+
+        if (is_last_call (this_call_cnt)) {
+                dht_rmdir_do (frame, this);
+        }
+
+        return 0;
 }
 
 
-int32_t
-dht_inodelk (call_frame_t *frame, xlator_t *this,
-	     const char *volume, loc_t *loc, int32_t cmd, struct flock *lock)
+int
+dht_rmdir_opendir_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;
-	dht_local_t  *local = 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;
 
-        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 = frame->local;
+        prev  = cookie;
 
-	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;
-	}
+        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;
+        }
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        if (!is_last_call (this_call_cnt))
+                return 0;
 
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = 1;
+        if (local->op_ret == -1)
+                goto err;
 
-	STACK_WIND (frame,
-		    dht_inodelk_cbk,
-		    subvol, subvol->fops->inodelk,
-		    volume, loc, cmd, lock);
+        dict = dict_new ();
+        if (!dict) {
+                local->op_ret = -1;
+                local->op_errno = ENOMEM;
+                goto err;
+        }
 
-	return 0;
+        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);
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (inodelk, frame, -1, op_errno);
+        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);
+        }
 
-	return 0;
-}
+        if (dict)
+                dict_unref (dict);
 
+        return 0;
 
-int
-dht_finodelk_cbk (call_frame_t *frame, void *cookie,
-		  xlator_t *this, int32_t op_ret, int32_t op_errno)
+err:
+        if (is_last_call (this_call_cnt)) {
+                dht_rmdir_do (frame, this);
+        }
 
-{
-	DHT_STACK_UNWIND (finodelk, frame, op_ret, op_errno);
-	return 0;
+        return 0;
 }
 
 
 int
-dht_finodelk (call_frame_t *frame, xlator_t *this,
-	      const char *volume, fd_t *fd, int32_t cmd, struct flock *lock)
+dht_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+           dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
+        dht_local_t  *local  = NULL;
+        dht_conf_t   *conf = NULL;
         int           op_errno = -1;
+        int           i = -1;
 
         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);
+        VALIDATE_OR_GOTO (loc->path, 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, loc, NULL, GF_FOP_RMDIR);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
+        local->call_cnt = conf->subvolume_cnt;
+        local->op_ret   = 0;
+        local->fop_succeeded = 0;
 
-	STACK_WIND (frame,
-		    dht_finodelk_cbk,
-		    subvol, subvol->fops->finodelk,
-		    volume, fd, cmd, lock);
+        local->flags = flags;
 
-	return 0;
+        local->fd = fd_create (local->loc.inode, frame->root->pid);
+        if (!local->fd) {
+
+                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, NULL);
+        }
+
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (finodelk, frame, -1, op_errno);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (rmdir, frame, -1, op_errno,
+                          NULL, NULL, NULL);
 
-	return 0;
+        return 0;
 }
 
-
 int
 dht_entrylk_cbk (call_frame_t *frame, void *cookie,
-		 xlator_t *this, int32_t op_ret, int32_t op_errno)
+                 xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
 
 {
-	DHT_STACK_UNWIND (entrylk, frame, op_ret, op_errno);
-	return 0;
+        DHT_STACK_UNWIND (entrylk, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
 
 int
 dht_entrylk (call_frame_t *frame, xlator_t *this,
-	     const char *volume, loc_t *loc, const char *basename,
-	     entrylk_cmd cmd, entrylk_type type)
+             const char *volume, loc_t *loc, const char *basename,
+             entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
+        xlator_t     *subvol = NULL;
         int           op_errno = -1;
-	dht_local_t  *local = NULL;
+        dht_local_t  *local = NULL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (loc, err);
         VALIDATE_OR_GOTO (loc->inode, err);
-        VALIDATE_OR_GOTO (loc->path, err);
 
-	subvol = dht_subvol_get_cached (this, loc->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for path=%s", loc->path);
-		op_errno = EINVAL;
-		goto err;
-	}
+        local = dht_local_init (frame, loc, NULL, GF_FOP_ENTRYLK);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        subvol = local->cached_subvol;
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for path=%s", loc->path);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = 1;
+        local->call_cnt = 1;
 
-	STACK_WIND (frame, dht_entrylk_cbk,
-		    subvol, subvol->fops->entrylk,
-		    volume, loc, basename, cmd, type);
+        STACK_WIND (frame, dht_entrylk_cbk,
+                    subvol, subvol->fops->entrylk,
+                    volume, loc, basename, cmd, type, xdata);
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (entrylk, frame, -1, op_errno);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (entrylk, frame, -1, op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_fentrylk_cbk (call_frame_t *frame, void *cookie,
-		  xlator_t *this, int32_t op_ret, int32_t op_errno)
+                  xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
 
 {
-	DHT_STACK_UNWIND (fentrylk, frame, op_ret, op_errno);
-	return 0;
+        DHT_STACK_UNWIND (fentrylk, frame, op_ret, op_errno, NULL);
+        return 0;
 }
 
 
 int
 dht_fentrylk (call_frame_t *frame, xlator_t *this,
-	      const char *volume, fd_t *fd, const char *basename,
-	      entrylk_cmd cmd, entrylk_type type)
+              const char *volume, fd_t *fd, const char *basename,
+              entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
-	xlator_t     *subvol = NULL;
+        xlator_t     *subvol = NULL;
         int           op_errno = -1;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
         VALIDATE_OR_GOTO (fd, err);
 
-	subvol = dht_subvol_get_cached (this, fd->inode);
-	if (!subvol) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no cached subvolume for fd=%p", fd);
-		op_errno = EINVAL;
-		goto err;
-	}
+        subvol = dht_subvol_get_cached (this, fd->inode);
+        if (!subvol) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "no cached subvolume for fd=%p", fd);
+                op_errno = EINVAL;
+                goto err;
+        }
 
-	STACK_WIND (frame, dht_fentrylk_cbk,
-		    subvol, subvol->fops->fentrylk,
-		    volume, fd, basename, cmd, type);
+        STACK_WIND (frame, dht_fentrylk_cbk,
+                    subvol, subvol->fops->fentrylk,
+                    volume, fd, basename, cmd, type, xdata);
 
-	return 0;
+        return 0;
 
 err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (fentrylk, frame, -1, op_errno);
+        op_errno = (op_errno == -1) ? errno : op_errno;
+        DHT_STACK_UNWIND (fentrylk, frame, -1, op_errno, NULL);
 
-	return 0;
+        return 0;
 }
 
 
 int
-dht_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                 int op_ret, int op_errno, struct stat *statpre,
-                 struct stat *statpost)
+dht_forget (xlator_t *this, inode_t *inode)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
+        uint64_t        ctx_int = 0;
+        dht_inode_ctx_t *ctx    = NULL;
+        dht_layout_t *layout = NULL;
 
+        inode_ctx_del (inode, this, &ctx_int);
 
-	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 (!ctx_int)
+                return 0;
 
-		dht_stat_merge (this, &local->prebuf, statpre, prev->this);
-                dht_stat_merge (this, &local->stbuf, statpost, prev->this);
-		
-		if (local->inode) {
-			local->prebuf.st_ino = local->inode->ino;
-                        local->stbuf.st_ino = local->inode->ino;
-                }
+        ctx = (dht_inode_ctx_t *) (long) ctx_int;
 
-		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);
+        layout = ctx->layout;
+        ctx->layout = NULL;
+        dht_layout_unref (this, layout);
+        GF_FREE (ctx);
 
         return 0;
 }
 
 
 int
-dht_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-             struct stat *stbuf, int32_t valid)
+dht_notify (xlator_t *this, int event, void *data, ...)
 {
-	dht_layout_t *layout = NULL;
-	dht_local_t  *local  = NULL;
-        int           op_errno = -1;
-	int           i = -1;
-
+        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;
 
-        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;
-	}
+        conf = this->private;
+        if (!conf)
+                return ret;
+
+        /* had all subvolumes reported status once till now? */
+        had_heard_from_all = 1;
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (!conf->last_event[i]) {
+                        had_heard_from_all = 0;
+                }
+        }
 
-	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;
-	}
+        switch (event) {
+        case GF_EVENT_CHILD_UP:
+                subvol = data;
 
-	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;
-	}
+                conf->gen++;
 
-	local->inode = inode_ref (loc->inode);
-	local->call_cnt = layout->cnt;
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (subvol == conf->subvolumes[i]) {
+                                cnt = i;
+                                break;
+                        }
+                }
 
-	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);
-	}
+                if (cnt == -1) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "got GF_EVENT_CHILD_UP bad subvolume %s",
+                                subvol->name);
+                        break;
+                }
 
-	return 0;
+                gettimeofday (&time, NULL);
+                LOCK (&conf->subvolume_lock);
+                {
+                        conf->subvolume_status[cnt] = 1;
+                        conf->last_event[cnt] = event;
+                        conf->subvol_up_time[cnt] = time.tv_sec;
+                }
+                UNLOCK (&conf->subvolume_lock);
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL);
+                /* one of the node came back up, do a stat update */
+                dht_get_du_info_for_subvol (this, cnt);
 
-	return 0;
-}
+                break;
 
+        case GF_EVENT_CHILD_MODIFIED:
+                subvol = data;
 
-int
-dht_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd, struct stat *stbuf,
-              int32_t valid)
-{
-	dht_layout_t *layout = NULL;
-	dht_local_t  *local  = NULL;
-        int           op_errno = -1;
-	int           i = -1;
+                conf->gen++;
+                propagate = 1;
 
+                break;
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
+        case GF_EVENT_CHILD_DOWN:
+                subvol = data;
 
-	local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+                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);
+                        }
+                }
 
-	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;
-	}
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (subvol == conf->subvolumes[i]) {
+                                cnt = i;
+                                break;
+                        }
+                }
 
-	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 (cnt == -1) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "got GF_EVENT_CHILD_DOWN bad subvolume %s",
+                                subvol->name);
+                        break;
+                }
 
-	local->inode = inode_ref (fd->inode);
-	local->call_cnt = layout->cnt;
+                LOCK (&conf->subvolume_lock);
+                {
+                        conf->subvolume_status[cnt] = 0;
+                        conf->last_event[cnt] = event;
+                        conf->subvol_up_time[cnt] = 0;
+                }
+                UNLOCK (&conf->subvolume_lock);
 
-	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);
-	}
+                break;
 
-	return 0;
+        case GF_EVENT_CHILD_CONNECTING:
+                subvol = data;
 
-err:
-	op_errno = (op_errno == -1) ? errno : op_errno;
-	DHT_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL);
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (subvol == conf->subvolumes[i]) {
+                                cnt = i;
+                                break;
+                        }
+                }
 
-	return 0;
-}
+                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);
+                {
+                        conf->last_event[cnt] = event;
+                }
+                UNLOCK (&conf->subvolume_lock);
 
-int
-dht_forget (xlator_t *this, inode_t *inode)
-{
-	uint64_t      tmp_layout = 0;
-	dht_layout_t *layout = NULL;
+                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;
+        }
 
-	inode_ctx_get (inode, this, &tmp_layout);
+        default:
+                propagate = 1;
+                break;
+        }
 
-	if (!tmp_layout)
-		return 0;
 
-	layout = (dht_layout_t *)(long)tmp_layout;
-        dht_layout_unref (this, layout);
+        /* have all subvolumes reported status once by now? */
+        have_heard_from_all = 1;
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (!conf->last_event[i])
+                        have_heard_from_all = 0;
+        }
 
-	return 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;
 
+        }
 
 
-int
-dht_init_subvolumes (xlator_t *this, dht_conf_t *conf)
-{
-        xlator_list_t *subvols = NULL;
-        int            cnt = 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;
 
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (conf->last_event[i] == GF_EVENT_CHILD_UP) {
+                                event = GF_EVENT_CHILD_UP;
+                                break;
+                        }
 
-        for (subvols = this->children; subvols; subvols = subvols->next)
-                cnt++;
+                        if (conf->last_event[i] == GF_EVENT_CHILD_CONNECTING) {
+                                event = GF_EVENT_CHILD_CONNECTING;
+                                /* continue to check other events for CHILD_UP */
+                        }
+                }
 
-        conf->subvolumes = CALLOC (cnt, sizeof (xlator_t *));
-        if (!conf->subvolumes) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                return -1;
+                /* 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);
+                        }
+                }
         }
-        conf->subvolume_cnt = cnt;
 
-        cnt = 0;
-        for (subvols = this->children; subvols; subvols = subvols->next)
-                conf->subvolumes[cnt++] = subvols->xlator;
+        ret = 0;
+        if (propagate)
+                ret = default_notify (this, event, data);
 
-	conf->subvolume_status = CALLOC (cnt, sizeof (char));
-	if (!conf->subvolume_status) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		return -1;
-	}
-
-        return 0;
+        return ret;
 }
 
-
 int
-dht_notify (xlator_t *this, int event, void *data, ...)
+dht_inode_ctx_layout_get (inode_t *inode, xlator_t *this, dht_layout_t **layout)
 {
-	xlator_t   *subvol = NULL;
-	int         cnt    = -1;
-	int         i      = -1;
-	dht_conf_t *conf   = NULL;
-	int         ret    = -1;
+        dht_inode_ctx_t         *ctx            = NULL;
+        int                      ret            = -1;
 
+        ret = dht_inode_ctx_get (inode, this, &ctx);
 
-	conf = this->private;
-
-	switch (event) {
-	case GF_EVENT_CHILD_UP:
-		subvol = data;
-
-		conf->gen++;
-
-		for (i = 0; i < conf->subvolume_cnt; i++) {
-			if (subvol == conf->subvolumes[i]) {
-				cnt = i;
-				break;
-			}
-		}
-
-		if (cnt == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"got GF_EVENT_CHILD_UP bad subvolume %s",
-				subvol->name);
-			break;
-		}
-
-		LOCK (&conf->subvolume_lock);
-		{
-			conf->subvolume_status[cnt] = 1;
-		}
-		UNLOCK (&conf->subvolume_lock);
-
-                /* one of the node came back up, do a stat update */
-                dht_get_du_info_for_subvol (this, cnt);
-
-		break;
-
-	case GF_EVENT_CHILD_DOWN:
-		subvol = data;
-
-		for (i = 0; i < conf->subvolume_cnt; i++) {
-			if (subvol == conf->subvolumes[i]) {
-				cnt = i;
-				break;
-			}
-		}
-
-		if (cnt == -1) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"got GF_EVENT_CHILD_DOWN bad subvolume %s",
-				subvol->name);
-			break;
-		}
-
-		LOCK (&conf->subvolume_lock);
-		{
-			conf->subvolume_status[cnt] = 0;
-		}
-		UNLOCK (&conf->subvolume_lock);
-
-		break;
-	}
-
-	ret = default_notify (this, event, data);
+        if (!ret && ctx) {
+                if (ctx->layout) {
+                        if (layout)
+                                *layout = ctx->layout;
+                        ret = 0;
+                } else {
+                        ret = -1;
+                }
+        }
 
-	return ret;
+        return ret;
 }
-
diff --git a/xlators/cluster/dht/src/dht-common.h b/xlators/cluster/dht/src/dht-common.h
index c417a8596..d391b87d5 100644
--- a/xlators/cluster/dht/src/dht-common.h
+++ b/xlators/cluster/dht/src/dht-common.h
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -22,240 +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     "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                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_layout dht_layout_t;
 
+typedef struct dht_stat_time dht_stat_time_t;
+
+struct dht_inode_ctx {
+        dht_layout_t    *layout;
+        dht_stat_time_t  time;
+};
+
+typedef struct dht_inode_ctx dht_inode_ctx_t;
+
+
+typedef enum {
+        DHT_HASH_TYPE_DM,
+        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 stat              stbuf;
-        struct stat              prebuf;
-        struct stat              preoldparent;
-        struct stat              postoldparent; 
-        struct stat              preparent;
-        struct stat              postparent;
-	struct statvfs           statvfs;
-	fd_t                    *fd;
-	inode_t                 *inode;
-	dict_t                  *xattr;
-	dict_t                  *xattr_req;
-	dht_layout_t            *layout;
-	size_t                   size;
-	ino_t                    st_ino;
-	xlator_t                *src_hashed, *src_cached;
-	xlator_t                *dst_hashed, *dst_cached;
-	xlator_t                *cached_subvol;
-	xlator_t                *hashed_subvol;
-	char                     need_selfheal;
+        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;
+        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;
         int                      file_count;
         int                      dir_count;
-	struct {
-		fop_mknod_cbk_t  linkfile_cbk;
-		struct stat      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;
-
-	/* needed by nufa */
-	int32_t flags;
-	mode_t  mode;
-	dev_t   rdev;
+        call_frame_t            *main_frame;
+        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;
+        mode_t  umask;
+
+        /* need for file-info */
+        char   *xattr_val;
+        char   *key;
+
+        /* which xattr request? */
+        char xsel[256];
+        int32_t alloc_len;
+
+        char   *newpath;
+
+        /* gfid related */
+        uuid_t  gfid;
+
+        /*Marker Related*/
+        struct marker_str    marker;
+
+        /* flag used to make sure we need to return estale in
+           {lookup,revalidate}_cbk */
+        char return_estale;
+        char need_lookup_everywhere;
+
+        glusterfs_fop_t      fop;
+
+        gf_boolean_t     linked;
+        xlator_t        *link_subvol;
+
+        struct dht_rebalance_ rebalance;
+        xlator_t        *first_up_subvol;
+
 };
 typedef struct dht_local dht_local_t;
 
 /* du - disk-usage */
 struct dht_du {
         double   avail_percent;
+	double   avail_inodes;
         uint64_t avail_space;
         uint32_t log;
 };
 typedef struct dht_du dht_du_t;
 
+enum gf_defrag_type {
+        GF_DEFRAG_CMD_START = 1,
+        GF_DEFRAG_CMD_STOP = 1 + 1,
+        GF_DEFRAG_CMD_STATUS = 1 + 2,
+        GF_DEFRAG_CMD_START_LAYOUT_FIX = 1 + 3,
+        GF_DEFRAG_CMD_START_FORCE = 1 + 4,
+};
+typedef enum gf_defrag_type gf_defrag_type;
+
+enum gf_defrag_status_t {
+        GF_DEFRAG_STATUS_NOT_STARTED,
+        GF_DEFRAG_STATUS_STARTED,
+        GF_DEFRAG_STATUS_STOPPED,
+        GF_DEFRAG_STATUS_COMPLETE,
+        GF_DEFRAG_STATUS_FAILED,
+        GF_DEFRAG_STATUS_LAYOUT_FIX_STARTED,
+        GF_DEFRAG_STATUS_LAYOUT_FIX_STOPPED,
+        GF_DEFRAG_STATUS_LAYOUT_FIX_COMPLETE,
+        GF_DEFRAG_STATUS_LAYOUT_FIX_FAILED,
+};
+typedef enum gf_defrag_status_t gf_defrag_status_t;
+
+typedef struct gf_defrag_pattern_list gf_defrag_pattern_list_t;
+
+struct gf_defrag_pattern_list {
+        char                       path_pattern[256];
+        uint64_t                   size;
+        gf_defrag_pattern_list_t  *next;
+};
+
+struct gf_defrag_info_ {
+        uint64_t                     total_files;
+        uint64_t                     total_data;
+        uint64_t                     num_files_lookedup;
+        uint64_t                     total_failures;
+        uint64_t                     skipped;
+        gf_lock_t                    lock;
+        int                          cmd;
+        pthread_t                    th;
+        gf_defrag_status_t           defrag_status;
+        struct rpc_clnt             *rpc;
+        uint32_t                     connected;
+        uint32_t                     is_exiting;
+        pid_t                        pid;
+        inode_t                     *root_inode;
+        uuid_t                       node_uuid;
+        struct timeval               start_time;
+        gf_boolean_t                 stats;
+        gf_defrag_pattern_list_t    *defrag_pattern;
+};
+
+typedef struct gf_defrag_info_ gf_defrag_info_t;
+
 struct dht_conf {
-	gf_lock_t      subvolume_lock;
+        gf_lock_t      subvolume_lock;
         int            subvolume_cnt;
         xlator_t     **subvolumes;
-	char          *subvolume_status;
-	dht_layout_t **file_layouts;
-	dht_layout_t **dir_layouts;
-	dht_layout_t  *default_dir_layout;
-	gf_boolean_t   search_unhashed;
-	int            gen;
+        char          *subvolume_status;
+        int           *last_event;
+        dht_layout_t **file_layouts;
+        dht_layout_t **dir_layouts;
+        gf_boolean_t   search_unhashed;
+        int            gen;
         dht_du_t      *du_stats;
-        uint64_t       min_free_disk;
+        double         min_free_disk;
+	double         min_free_inodes;
         char           disk_unit;
         int32_t        refresh_interval;
         gf_boolean_t   unhashed_sticky_bit;
-	struct timeval last_stat_fetch;
+        struct timeval last_stat_fetch;
         gf_lock_t      layout_lock;
         void          *private;     /* Can be used by wrapper xlators over
                                        dht */
+        gf_boolean_t   use_readdirp;
+        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 ENTRY_MISSING(op_ret, op_errno) (op_ret == -1 && op_errno == ENOENT)
 
-#define is_fs_root(loc) (strcmp (loc->path, "/") == 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_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) ((s->st_mode & ~S_IFMT) == DHT_LINKFILE_MODE)
+#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 check_is_dir(i,s,x) (S_ISDIR(s->st_mode))
+#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;                          \
-                __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);
+                dht_local_t *__local = NULL;                    \
+                xlator_t    *__xl    = NULL;                    \
+                if (frame) {                                    \
+                        __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)
+
+#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 stat *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, int32_t *disk_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 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_stat_merge (xlator_t *this, struct stat *to, struct stat *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_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,
+                               xlator_t **subvol);
+
+int                                     dht_rename_cleanup (call_frame_t *frame);
+int dht_rename_links_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                          int32_t           op_ret, int32_t op_errno,
+                          inode_t          *inode, struct iatt *stbuf,
+                          struct iatt      *preparent, struct iatt *postparent,
+                          dict_t *xdata);
+
+int dht_fix_directory_layout (call_frame_t *frame,
+                              dht_selfheal_dir_cbk_t  dir_cbk,
+                              dht_layout_t           *layout);
+
+int dht_init_subvolumes (xlator_t *this, dht_conf_t *conf);
+
+/* migration/rebalance */
+int dht_start_rebalance_task (xlator_t *this, call_frame_t *frame);
+
+int dht_rebalance_in_progress_check (xlator_t *this, call_frame_t *frame);
+int dht_rebalance_complete_check (xlator_t *this, call_frame_t *frame);
+
+
+/* FOPS */
+int32_t dht_lookup (call_frame_t *frame,
+                    xlator_t *this,
+                    loc_t    *loc,
+                    dict_t   *xattr_req);
+
+int32_t dht_stat (call_frame_t *frame,
+                  xlator_t *this,
+                  loc_t    *loc, dict_t *xdata);
+
+int32_t dht_fstat (call_frame_t *frame,
+                   xlator_t *this,
+                   fd_t     *fd, dict_t *xdata);
+
+int32_t dht_truncate (call_frame_t *frame,
+                      xlator_t *this,
+                      loc_t    *loc,
+                      off_t     offset, dict_t *xdata);
+
+int32_t dht_ftruncate (call_frame_t *frame,
+                       xlator_t *this,
+                       fd_t     *fd,
+                       off_t     offset, dict_t *xdata);
+
+int32_t dht_access (call_frame_t *frame,
+                    xlator_t *this,
+                    loc_t    *loc,
+                    int32_t   mask, dict_t *xdata);
+
+int32_t dht_readlink (call_frame_t *frame,
+                      xlator_t *this,
+                      loc_t    *loc,
+                      size_t    size, dict_t *xdata);
+
+int32_t dht_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                   mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata);
+
+int32_t dht_mkdir (call_frame_t *frame, xlator_t *this,
+                   loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata);
+
+int32_t dht_unlink (call_frame_t *frame,
+                    xlator_t *this,
+                    loc_t    *loc, int xflag, dict_t *xdata);
+
+int32_t dht_rmdir (call_frame_t *frame, xlator_t *this,
+                   loc_t    *loc, int flags, dict_t *xdata);
+
+int32_t dht_symlink (call_frame_t   *frame, xlator_t *this,
+                     const char *linkpath, loc_t *loc, mode_t umask,
+                     dict_t *xdata);
+
+int32_t dht_rename (call_frame_t *frame,
+                    xlator_t *this,
+                    loc_t    *oldloc,
+                    loc_t    *newloc, dict_t *xdata);
+
+int32_t dht_link (call_frame_t *frame,
+                  xlator_t *this,
+                  loc_t    *oldloc,
+                  loc_t    *newloc, dict_t *xdata);
+
+int32_t dht_create (call_frame_t *frame, xlator_t *this,
+                    loc_t    *loc, int32_t flags, mode_t mode,
+                    mode_t umask, fd_t     *fd, dict_t *params);
+
+int32_t dht_open (call_frame_t *frame,
+                  xlator_t *this,
+                  loc_t    *loc,
+                  int32_t   flags, fd_t *fd, dict_t *xdata);
+
+int32_t dht_readv (call_frame_t *frame,
+                   xlator_t *this,
+                   fd_t     *fd,
+                   size_t    size,
+                   off_t     offset, uint32_t flags, dict_t *xdata);
+
+int32_t dht_writev (call_frame_t      *frame,
+                    xlator_t      *this,
+                    fd_t          *fd,
+                    struct iovec  *vector,
+                    int32_t        count,
+                    off_t          offset,
+                    uint32_t       flags,
+                    struct iobref *iobref, dict_t *xdata);
+
+int32_t dht_flush (call_frame_t *frame,
+                   xlator_t *this,
+                   fd_t     *fd, dict_t *xdata);
+
+int32_t dht_fsync (call_frame_t *frame,
+                   xlator_t *this,
+                   fd_t     *fd,
+                   int32_t   datasync, dict_t *xdata);
+
+int32_t dht_opendir (call_frame_t *frame,
+                     xlator_t *this,
+                     loc_t    *loc, fd_t *fd, dict_t *xdata);
+
+int32_t dht_fsyncdir (call_frame_t *frame,
+                      xlator_t *this,
+                      fd_t     *fd,
+                      int32_t   datasync, dict_t *xdata);
+
+int32_t dht_statfs (call_frame_t *frame,
+                    xlator_t *this,
+                    loc_t    *loc, dict_t *xdata);
+
+int32_t dht_setxattr (call_frame_t *frame,
+                      xlator_t *this,
+                      loc_t    *loc,
+                      dict_t   *dict,
+                      int32_t   flags, dict_t *xdata);
+
+int32_t dht_getxattr (call_frame_t   *frame,
+                      xlator_t   *this,
+                      loc_t      *loc,
+                      const char *name, dict_t *xdata);
+
+int32_t dht_fsetxattr (call_frame_t *frame,
+                       xlator_t *this,
+                       fd_t     *fd,
+                       dict_t   *dict,
+                       int32_t   flags, dict_t *xdata);
+
+int32_t dht_fgetxattr (call_frame_t   *frame,
+                       xlator_t   *this,
+                       fd_t       *fd,
+                       const char *name, dict_t *xdata);
+
+int32_t dht_removexattr (call_frame_t   *frame,
+                         xlator_t   *this,
+                         loc_t      *loc,
+                         const char *name, dict_t *xdata);
+int32_t dht_fremovexattr (call_frame_t   *frame,
+                          xlator_t   *this,
+                          fd_t      *fd,
+                          const char *name, dict_t *xdata);
+
+int32_t dht_lk (call_frame_t        *frame,
+                xlator_t        *this,
+                fd_t            *fd,
+                int32_t          cmd,
+                struct gf_flock *flock, dict_t *xdata);
+
+int32_t dht_inodelk (call_frame_t *frame, xlator_t *this,
+                     const char      *volume, loc_t *loc, int32_t cmd,
+                     struct gf_flock *flock, dict_t *xdata);
+
+int32_t dht_finodelk (call_frame_t        *frame, xlator_t *this,
+                      const char      *volume, fd_t *fd, int32_t cmd,
+                      struct gf_flock *flock, dict_t *xdata);
+
+int32_t dht_entrylk (call_frame_t    *frame, xlator_t *this,
+                     const char  *volume, loc_t *loc, const char *basename,
+                     entrylk_cmd  cmd, entrylk_type type, dict_t *xdata);
+
+int32_t dht_fentrylk (call_frame_t    *frame, xlator_t *this,
+                      const char  *volume, fd_t *fd, const char *basename,
+                      entrylk_cmd  cmd, entrylk_type type, dict_t *xdata);
+
+int32_t dht_readdir (call_frame_t  *frame,
+                     xlator_t *this,
+                     fd_t     *fd,
+                     size_t    size, off_t off, dict_t *xdata);
+
+int32_t dht_readdirp (call_frame_t *frame,
+                      xlator_t *this,
+                      fd_t     *fd,
+                      size_t    size, off_t off, dict_t *dict);
+
+int32_t dht_xattrop (call_frame_t           *frame,
+                     xlator_t           *this,
+                     loc_t              *loc,
+                     gf_xattrop_flags_t  flags,
+                     dict_t             *dict, dict_t *xdata);
+
+int32_t dht_fxattrop (call_frame_t *frame,
+                      xlator_t           *this,
+                      fd_t               *fd,
+                      gf_xattrop_flags_t  flags,
+                      dict_t             *dict, dict_t *xdata);
+
+int32_t dht_forget (xlator_t *this, inode_t *inode);
+int32_t dht_setattr (call_frame_t  *frame, xlator_t *this, loc_t *loc,
+                     struct iatt   *stbuf, int32_t valid, dict_t *xdata);
+int32_t dht_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                      struct iatt  *stbuf, int32_t valid, dict_t *xdata);
+int32_t dht_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd,
+		      int32_t mode, off_t offset, size_t len, dict_t *xdata);
+int32_t dht_discard(call_frame_t *frame, xlator_t *this, fd_t *fd,
+		    off_t offset, size_t len, dict_t *xdata);
+int32_t dht_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd,
+                    off_t offset, 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 */
+#endif/* _DHT_H */
diff --git a/xlators/cluster/dht/src/dht-diskusage.c b/xlators/cluster/dht/src/dht-diskusage.c
index bcc385a2f..fe3955ecb 100644
--- a/xlators/cluster/dht/src/dht-diskusage.c
+++ b/xlators/cluster/dht/src/dht-diskusage.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -33,52 +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;
-	dht_local_t   *local = 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;
-
-	local = frame->local;
-        conf = this->private;
-        prev = cookie;
-
-        if (op_ret == -1) 
-                goto out;
-
-        if (statvfs && statvfs->f_blocks) {
-                percent = (statvfs->f_bfree * 100) / statvfs->f_blocks;
-                bytes = (statvfs->f_bfree * statvfs->f_bsize);
-        }
-        
-        LOCK (&conf->subvolume_lock);
-        {
-                for (i = 0; i < conf->subvolume_cnt; i++)
-                        if (prev->this == conf->subvolumes[i]) {
-                                conf->du_stats[i].avail_percent = percent;
-                                conf->du_stats[i].avail_space   = bytes;
-                                gf_log (this->name, GF_LOG_DEBUG,
-                                        "on subvolume '%s': avail_percent is: "
-                                        "%.2f and avail_space is: %"PRIu64"",
-                                        prev->this->name, 
-                                        conf->du_stats[i].avail_percent,
-                                        conf->du_stats[i].avail_space);
-                        }
-        }
-        UNLOCK (&conf->subvolume_lock);
+	int            i = 0;
+	double         percent = 0;
+	double         percent_inodes = 0;
+	uint64_t       bytes = 0;
 
- 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
@@ -87,177 +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;
-                }
-
-                statfs_local = dht_local_init (statfs_frame);
-                if (!statfs_local) {
-                        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;
 
-                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_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);
+        }
 
-        /* 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;
-                                        }
-                                }
-                        }
+	{
+                avail_subvol = dht_subvol_with_free_space_inodes(this, subvol,
+                                                                 layout);
+                if(!avail_subvol)
+                {
+                        avail_subvol = dht_subvol_maxspace_nonzeroinode(this,
+                                                                        subvol,
+                                                                        layout);
+                }
+
+	}
+	UNLOCK (&conf->subvolume_lock);
+out:
+	if (!avail_subvol) {
+		gf_log (this->name,
+                        GF_LOG_DEBUG,
+			"no subvolume has enough free space and/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;
                 }
         }
-        UNLOCK (&conf->subvolume_lock);
-
-        if (subvol_filled) {
-                if (!(conf->du_stats[i].log++ % GF_UNIVERSAL_ANSWER)) {
-                        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));
+        ret = 0;
+out:
+        return ret;
+}
+
+/*Get subvolume which has both space and inodes more than the min criteria*/
+xlator_t *
+dht_subvol_with_free_space_inodes(xlator_t *this, xlator_t *subvol,
+                                  dht_layout_t *layout)
+{
+        int i = 0;
+        double max = 0;
+        double max_inodes = 0;
+        int    ignore_subvol = 0;
+
+        xlator_t *avail_subvol = NULL;
+        dht_conf_t *conf = NULL;
+
+        conf = this->private;
+
+        for(i=0; i < conf->subvolume_cnt; i++) {
+                /* check if subvol has layout errors, before selecting it */
+                ignore_subvol = dht_subvol_has_err (conf->subvolumes[i],
+                                                    layout);
+                if (ignore_subvol)
+                        continue;
+
+                if ((conf->disk_unit == 'p') &&
+                    (conf->du_stats[i].avail_percent > conf->min_free_disk) &&
+                    (conf->du_stats[i].avail_inodes  > conf->min_free_inodes)) {
+                        if ((conf->du_stats[i].avail_inodes > max_inodes) ||
+                            (conf->du_stats[i].avail_percent > max)) {
+                                max = conf->du_stats[i].avail_percent;
+                                max_inodes = conf->du_stats[i].avail_inodes;
+                                avail_subvol = conf->subvolumes[i];
+                        }
+                }
+
+                if ((conf->disk_unit != 'p') &&
+                    (conf->du_stats[i].avail_space > conf->min_free_disk) &&
+                    (conf->du_stats[i].avail_inodes  > conf->min_free_inodes)) {
+                        if ((conf->du_stats[i].avail_inodes > max_inodes) ||
+                            (conf->du_stats[i].avail_space > max)) {
+                                max = conf->du_stats[i].avail_space;
+                                max_inodes = conf->du_stats[i].avail_inodes;
+                                avail_subvol = conf->subvolumes[i];
+                        }
                 }
         }
 
-        return subvol_filled;
+        return avail_subvol;
 }
 
+
+/* Get subvol which has atleast one inode and maximum space */
 xlator_t *
-dht_free_disk_available_subvol (xlator_t *this, xlator_t *subvol) 
+dht_subvol_maxspace_nonzeroinode (xlator_t *this, xlator_t *subvol,
+                                  dht_layout_t *layout)
 {
         int         i = 0;
-        double      max= 0;
+        double      max = 0;
+        int         ignore_subvol = 0;
+
         xlator_t   *avail_subvol = NULL;
-	dht_conf_t *conf = NULL;
+        dht_conf_t *conf = NULL;
 
         conf = this->private;
-        avail_subvol = subvol;
 
-        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') {
+                        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];
+                         }
+               }
         }
-        UNLOCK (&conf->subvolume_lock);
-
-        if (max < conf->min_free_disk)
-                avail_subvol = subvol;
 
-        if (avail_subvol == subvol) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "no subvolume has enough free space to create");
-        }
-                
         return avail_subvol;
 }
diff --git a/xlators/cluster/dht/src/dht-hashfn.c b/xlators/cluster/dht/src/dht-hashfn.c
index 3d1b7c152..656cf23a0 100644
--- a/xlators/cluster/dht/src/dht-hashfn.c
+++ b/xlators/cluster/dht/src/dht-hashfn.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -29,58 +20,92 @@
 #include "hashfn.h"
 
 
-typedef enum {
-	DHT_HASH_TYPE_DM,
-} dht_hashfn_type_t;
-
-
 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 1b49046ce..18a501f04 100644
--- a/xlators/cluster/dht/src/dht-helper.c
+++ b/xlators/cluster/dht/src/dht-helper.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -27,184 +18,406 @@
 #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;
 
-	if (!frame)
-		return -1;
+        LOCK (&frame->lock);
+        {
+                this_call_cnt = --local->call_cnt;
+        }
+        UNLOCK (&frame->lock);
+
+        return this_call_cnt;
+}
 
-	local = frame->local;
 
-	LOCK (&frame->lock);
-	{
-		this_call_cnt = --local->call_cnt;
+static uint64_t
+dht_bits_for (uint64_t num)
+{
+	uint64_t bits = 0, ctrl = 1;
+
+	while (ctrl < num) {
+		ctrl *= 2;
+		bits ++;
 	}
-	UNLOCK (&frame->lock);
 
-	return this_call_cnt;
+	return bits;
 }
 
+/*
+ * A slightly "updated" version of the algorithm described in the commit log
+ * is used here.
+ *
+ * The only enhancement is that:
+ *
+ * - The number of bits used by the backend filesystem for HUGE d_off which
+ *   is described as 63, and
+ * - The number of bits used by the d_off presented by the transformation
+ *   upwards which is described as 64, are both made "configurable."
+ */
+
+
+#define BACKEND_D_OFF_BITS 63
+#define PRESENT_D_OFF_BITS 63
+
+#define ONE 1ULL
+#define MASK (~0ULL)
+#define PRESENT_MASK (MASK >> (64 - PRESENT_D_OFF_BITS))
+#define BACKEND_MASK (MASK >> (64 - BACKEND_D_OFF_BITS))
+
+#define TOP_BIT (ONE << (PRESENT_D_OFF_BITS - 1))
+#define SHIFT_BITS (max (0, (BACKEND_D_OFF_BITS - PRESENT_D_OFF_BITS + 1)))
 
 int
 dht_itransform (xlator_t *this, xlator_t *subvol, uint64_t x, uint64_t *y_p)
 {
-	dht_conf_t *conf = NULL;
-	int         cnt = 0;
-	int         max = 0;
-	uint64_t    y = 0;
+        dht_conf_t *conf = NULL;
+        int         cnt = 0;
+        int         max = 0;
+        uint64_t    y = 0;
+        uint64_t    hi_mask = 0;
+        uint64_t    off_mask = 0;
+        int         max_bits = 0;
+
+        if (x == ((uint64_t) -1)) {
+                y = (uint64_t) -1;
+                goto out;
+        }
 
+        conf = this->private;
+        if (!conf)
+                goto out;
+
+        max = conf->subvolume_cnt;
+        cnt = dht_subvol_cnt (this, subvol);
 
-	if (x == ((uint64_t) -1)) {
-		y = (uint64_t) -1;
+	if (max == 1) {
+		y = x;
 		goto out;
 	}
 
-	conf = this->private;
+        max_bits = dht_bits_for (max);
 
-	max = conf->subvolume_cnt;
-	cnt = dht_subvol_cnt (this, subvol);
+        hi_mask = ~(PRESENT_MASK >> (max_bits + 1));
 
-	y = ((x * max) + cnt);
+        if (x & hi_mask) {
+                /* HUGE d_off */
+                off_mask = MASK << max_bits;
+                y = TOP_BIT | ((x >> SHIFT_BITS) & off_mask) | cnt;
+        } else {
+                /* small d_off */
+                y = ((x * max) + cnt);
+        }
 
 out:
-	if (y_p)
-		*y_p = y;
+        if (y_p)
+                *y_p = y;
 
-	return 0;
+        return 0;
 }
 
+int
+dht_filter_loc_subvol_key (xlator_t *this, loc_t *loc, loc_t *new_loc,
+                           xlator_t **subvol)
+{
+        char          *new_name  = NULL;
+        char          *new_path  = NULL;
+        xlator_list_t *trav      = NULL;
+        char           key[1024] = {0,};
+        int            ret       = 0; /* not found */
+
+        /* Why do other tasks if first required 'char' itself is not there */
+        if (!new_loc || !loc || !loc->name || !strchr (loc->name, '@'))
+                goto out;
+
+        trav = this->children;
+        while (trav) {
+                snprintf (key, 1024, "*@%s:%s", this->name, trav->xlator->name);
+                if (fnmatch (key, loc->name, FNM_NOESCAPE) == 0) {
+                        new_name = GF_CALLOC(strlen (loc->name),
+                                             sizeof (char),
+                                             gf_common_mt_char);
+                        if (!new_name)
+                                goto out;
+                        if (fnmatch (key, loc->path, FNM_NOESCAPE) == 0) {
+                                new_path = GF_CALLOC(strlen (loc->path),
+                                                     sizeof (char),
+                                                     gf_common_mt_char);
+                                if (!new_path)
+                                        goto out;
+                                strncpy (new_path, loc->path, (strlen (loc->path) -
+                                                               strlen (key) + 1));
+                        }
+                        strncpy (new_name, loc->name, (strlen (loc->name) -
+                                                       strlen (key) + 1));
+
+                        if (new_loc) {
+                                new_loc->path   = ((new_path) ? new_path:
+                                                   gf_strdup (loc->path));
+                                new_loc->name   = new_name;
+                                new_loc->inode  = inode_ref (loc->inode);
+                                new_loc->parent = inode_ref (loc->parent);
+                        }
+                        *subvol         = trav->xlator;
+                        ret = 1;  /* success */
+                        goto out;
+                }
+                trav = trav->next;
+        }
+out:
+        if (!ret) {
+                /* !success */
+                GF_FREE (new_path);
+                GF_FREE (new_name);
+        }
+        return ret;
+}
 
 int
 dht_deitransform (xlator_t *this, uint64_t y, xlator_t **subvol_p,
-		  uint64_t *x_p)
+                  uint64_t *x_p)
 {
-	dht_conf_t *conf = NULL;
-	int         cnt = 0;
-	int         max = 0;
-	uint64_t    x = 0;
-	xlator_t   *subvol = 0;
+        dht_conf_t *conf = NULL;
+        int         cnt = 0;
+        int         max = 0;
+        uint64_t    x = 0;
+        xlator_t   *subvol = 0;
+        int         max_bits = 0;
+        uint64_t    off_mask = 0;
+        uint64_t    host_mask = 0;
+
+        if (!this->private)
+                return -1;
+
+        conf = this->private;
+        max = conf->subvolume_cnt;
+
+	if (max == 1) {
+		x = y;
+		cnt = 0;
+		goto out;
+	}
 
+        if (y & TOP_BIT) {
+                /* HUGE d_off */
+                max_bits = dht_bits_for (max);
+                off_mask = (MASK << max_bits);
+                host_mask = ~(off_mask);
 
-	conf = this->private;
-	max = conf->subvolume_cnt;
+                x = ((y & ~TOP_BIT) & off_mask) << SHIFT_BITS;
 
-	cnt = y % max;
-	x   = y / max;
+                cnt = y & host_mask;
+	} else {
+                /* small d_off */
+                cnt = y % max;
+                x = y / max;
+        }
 
-	subvol = conf->subvolumes[cnt];
+out:
+        subvol = conf->subvolumes[cnt];
 
-	if (subvol_p)
-		*subvol_p = subvol;
+        if (subvol_p)
+                *subvol_p = subvol;
 
-	if (x_p)
-		*x_p = x;
+        if (x_p)
+                *x_p = x;
 
-	return 0;
+        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->xattr_req)
-		dict_unref (local->xattr_req);
+        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->selfheal.layout) {
                 dht_layout_unref (this, local->selfheal.layout);
                 local->selfheal.layout = NULL;
         }
 
-	FREE (local);
+        GF_FREE (local->newpath);
+
+        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;
+        dht_local_t *local = NULL;
+        inode_t     *inode = NULL;
+        int          ret   = 0;
 
-	/* TODO: use mem-pool */
-	local = CALLOC (1, sizeof (*local));
+        local = mem_get0 (THIS->local_pool);
+        if (!local)
+                goto out;
 
-	if (!local)
-		return NULL;
+        if (loc) {
+                ret = loc_copy (&local->loc, loc);
+                if (ret)
+                        goto out;
 
-	local->op_ret = -1;
-	local->op_errno = EUCLEAN;
+                inode = loc->inode;
+        }
 
-	frame->local = local;
+        if (fd) {
+                local->fd = fd_ref (fd);
+                if (!inode)
+                        inode = fd->inode;
+        }
 
-	return local;
-}
+        local->op_ret   = -1;
+        local->op_errno = EUCLEAN;
+        local->fop      = fop;
 
+        if (inode) {
+                local->layout   = dht_layout_get (frame->this, inode);
+                local->cached_subvol = dht_subvol_get_cached (frame->this,
+                                                              inode);
+        }
+
+        frame->local = local;
+
+out:
+        if (ret) {
+                if (local)
+                        mem_put (local);
+                local = NULL;
+        }
+        return local;
+}
 
-char *
-basestr (const char *str)
+xlator_t *
+dht_first_up_subvol (xlator_t *this)
 {
-        char *basestr = NULL;
+        dht_conf_t *conf = NULL;
+        xlator_t   *child = NULL;
+        int         i = 0;
+        time_t      time = 0;
 
-        basestr = strrchr (str, '/');
-        if (basestr)
-                basestr ++;
+        conf = this->private;
+        if (!conf)
+                goto out;
+
+        LOCK (&conf->subvolume_lock);
+        {
+                for (i = 0; i < conf->subvolume_cnt; i++) {
+                        if (conf->subvol_up_time[i]) {
+                                if (!time) {
+                                        time = conf->subvol_up_time[i];
+                                        child = conf->subvolumes[i];
+                                } else if (time > conf->subvol_up_time[i]) {
+                                        time  = conf->subvol_up_time[i];
+                                        child = conf->subvolumes[i];
+                                }
+                        }
+                }
+        }
+        UNLOCK (&conf->subvolume_lock);
 
-        return basestr;
+out:
+        return child;
 }
 
 xlator_t *
-dht_first_up_subvol (xlator_t *this)
+dht_last_up_subvol (xlator_t *this)
 {
-	dht_conf_t *conf = NULL;
-	xlator_t   *child = NULL;
-	int         i = 0;
-
-	conf = this->private;
-	
-	LOCK (&conf->subvolume_lock);
-	{
-		for (i = 0; i < conf->subvolume_cnt; i++) {
-			if (conf->subvolume_status[i]) {
-				child = conf->subvolumes[i];
-				break;
-			}
-		}
-	}
-	UNLOCK (&conf->subvolume_lock);
-	
-	return child;
+        dht_conf_t *conf = NULL;
+        xlator_t   *child = NULL;
+        int         i = 0;
+
+        conf = this->private;
+        if (!conf)
+                goto out;
+
+        LOCK (&conf->subvolume_lock);
+        {
+                for (i = conf->subvolume_cnt-1; i >= 0; i--) {
+                        if (conf->subvolume_status[i]) {
+                                child = conf->subvolumes[i];
+                                break;
+                        }
+                }
+        }
+        UNLOCK (&conf->subvolume_lock);
+
+out:
+        return child;
 }
 
 xlator_t *
@@ -213,17 +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;
         }
 
@@ -251,6 +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);
 
@@ -258,7 +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) {
@@ -272,71 +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;
-
-	conf = this->private;
-
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		if (conf->subvolumes[i] == prev) {
-			if ((i + 1) < conf->subvolume_cnt)
-				next = conf->subvolumes[i + 1];
-			break;
-		}
-	}
+        dht_conf_t *conf = NULL;
+        int         i = 0;
+        xlator_t   *next = NULL;
+
+        conf = this->private;
+        if (!conf)
+                goto out;
 
-	return next;
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (conf->subvolumes[i] == prev) {
+                        if ((i + 1) < conf->subvolume_cnt)
+                                next = conf->subvolumes[i + 1];
+                        break;
+                }
+        }
+
+out:
+        return next;
 }
 
+/* This func wraps around, if prev is actually the last subvol.
+ */
+xlator_t *
+dht_subvol_next_available (xlator_t *this, xlator_t *prev)
+{
+        dht_conf_t *conf = NULL;
+        int         i = 0;
+        xlator_t   *next = NULL;
+
+        conf = this->private;
+        if (!conf)
+                goto out;
+
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (conf->subvolumes[i] == prev) {
+                        /* if prev is last in conf->subvolumes, then wrap
+                         * around.
+                         */
+                        if ((i + 1) < conf->subvolume_cnt) {
+                                next = conf->subvolumes[i + 1];
+                        } else {
+                                next = conf->subvolumes[0];
+                        }
+                        break;
+                }
+        }
 
+out:
+        return next;
+}
 int
 dht_subvol_cnt (xlator_t *this, xlator_t *subvol)
 {
-	int i = 0;
-	int ret = -1;
-	dht_conf_t *conf = NULL;
+        int i = 0;
+        int ret = -1;
+        dht_conf_t *conf = NULL;
+
+        conf = this->private;
+        if (!conf)
+                goto out;
+
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (subvol == conf->subvolumes[i]) {
+                        ret = i;
+                        break;
+                }
+        }
+
+out:
+        return ret;
+}
 
 
-	conf = this->private;
+#define set_if_greater(a, b) do {               \
+                if ((a) < (b))                  \
+                        (a) = (b);              \
+        } while (0)
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		if (subvol == conf->subvolumes[i]) {
-			ret = i;
-			break;
-		}
-	}
 
-	return ret;
+#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)
+{
+        if (!from || !to)
+                return 0;
+
+        to->ia_dev      = from->ia_dev;
+
+        uuid_copy (to->ia_gfid, from->ia_gfid);
+
+        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;
+        }
+
+        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) {
+                goto err;
+        }
+
+        return 0;
+err:
+        loc_wipe (child);
+        return -1;
 }
 
 
-#define set_if_greater(a, b) do {		\
-		if ((a) < (b))			\
-			(a) = (b);		\
-	} while (0)
 
 int
-dht_stat_merge (xlator_t *this, struct stat *to,
-		struct stat *from, xlator_t *subvol)
+dht_init_subvolumes (xlator_t *this, dht_conf_t *conf)
 {
-	to->st_dev      = from->st_dev;
+        xlator_list_t *subvols = NULL;
+        int            cnt = 0;
+
+        if (!conf)
+                return -1;
+
+        for (subvols = this->children; subvols; subvols = subvols->next)
+                cnt++;
 
-	dht_itransform (this, subvol, from->st_ino, &to->st_ino);
+        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;
+        }
 
-	to->st_mode     = from->st_mode;
-	to->st_nlink    = from->st_nlink;
-	to->st_rdev     = from->st_rdev;
-	to->st_size    += from->st_size;
-	to->st_blksize  = from->st_blksize;
-	to->st_blocks  += from->st_blocks;
+        conf->last_event = GF_CALLOC (cnt, sizeof (int),
+                                      gf_dht_mt_char);
+        if (!conf->last_event) {
+                return -1;
+        }
 
-	set_if_greater (to->st_uid, from->st_uid);
-	set_if_greater (to->st_gid, from->st_gid);
+        conf->subvol_up_time = GF_CALLOC (cnt, sizeof (time_t),
+                                          gf_dht_mt_subvol_time);
+        if (!conf->subvol_up_time) {
+                return -1;
+        }
 
-	set_if_greater (to->st_atime, from->st_atime);
-	set_if_greater (to->st_mtime, from->st_mtime);
-	set_if_greater (to->st_ctime, from->st_ctime);
+        conf->du_stats = GF_CALLOC (conf->subvolume_cnt, sizeof (dht_du_t),
+                                    gf_dht_mt_dht_du_t);
+        if (!conf->du_stats) {
+                return -1;
+        }
 
-	return 0;
+        conf->decommissioned_bricks = GF_CALLOC (cnt, sizeof (xlator_t *),
+                                                 gf_dht_mt_xlator_t);
+        if (!conf->decommissioned_bricks) {
+                return -1;
+        }
+
+        return 0;
+}
+
+
+
+
+static int
+dht_migration_complete_check_done (int op_ret, call_frame_t *frame, void *data)
+{
+        dht_local_t *local = NULL;
+
+        local = frame->local;
+
+        local->rebalance.target_op_fn (THIS, frame, op_ret);
+
+        return 0;
+}
+
+
+int
+dht_migration_complete_check_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;
+        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;
+        }
+
+        inode = (!local->fd) ? local->loc.inode : local->fd->inode;
+
+        /* getxattr on cached_subvol for 'linkto' value. Do path based getxattr
+         * as root:root. If a fd is already open, access check wont be done*/
+
+        if (!local->loc.inode) {
+                ret = syncop_fgetxattr (src_node, local->fd, &dict,
+                                        conf->link_xattr_name);
+        } else {
+                SYNCTASK_SETID (0, 0);
+                ret = syncop_getxattr (src_node, &local->loc, &dict,
+                                       conf->link_xattr_name);
+                SYNCTASK_SETID (frame->root->uid, frame->root->gid);
+        }
+
+        if (!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,
+                        "%s: failed to get the destination node",
+                        local->loc.path);
+                ret = -1;
+                local->op_errno = EINVAL;
+                goto out;
+        }
+
+        /* 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;
+                }
+
+                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;
+                }
+        }
+
+        /* 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,
+                        "%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;
+}
+
+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 4b7b44fc4..e1a37b77c 100644
--- a/xlators/cluster/dht/src/dht-layout.c
+++ b/xlators/cluster/dht/src/dht-layout.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -34,30 +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 = CALLOC (1, layout_size (cnt));
-	if (!layout) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto out;
-	}
+        layout = GF_CALLOC (1, layout_size (cnt),
+                            gf_dht_mt_dht_layout_t);
+        if (!layout) {
+                goto out;
+        }
+
+        layout->type = DHT_HASH_TYPE_DM;
+        layout->cnt = cnt;
 
-	layout->cnt = cnt;
-        if (conf)
+        if (conf) {
+                layout->spread_cnt = conf->dir_spread_cnt;
                 layout->gen = conf->gen;
+        }
 
         layout->ref = 1;
+
 out:
-	return layout;
+        return layout;
 }
 
 
@@ -65,21 +58,22 @@ dht_layout_t *
 dht_layout_get (xlator_t *this, inode_t *inode)
 {
         dht_conf_t   *conf = NULL;
-        uint64_t      layout_int = 0;
         dht_layout_t *layout = NULL;
-        int           ret    = -1;
 
         conf = this->private;
+        if (!conf)
+                goto out;
+
         LOCK (&conf->layout_lock);
         {
-                ret = inode_ctx_get (inode, this, &layout_int);
-                if (ret == 0) {
-                        layout = (dht_layout_t *) (unsigned long) layout_int;
+                dht_inode_ctx_layout_get (inode, this, &layout);
+                if (layout) {
                         layout->ref++;
                 }
         }
         UNLOCK (&conf->layout_lock);
 
+out:
         return layout;
 }
 
@@ -91,24 +85,24 @@ dht_layout_set (xlator_t *this, inode_t *inode, dht_layout_t *layout)
         int           oldret = -1;
         int           ret = 0;
         dht_layout_t *old_layout;
-        uint64_t      old_layout_int;
 
         conf = this->private;
+        if (!conf)
+                goto out;
+
         LOCK (&conf->layout_lock);
         {
-                oldret = inode_ctx_get (inode, this, &old_layout_int);
-
+                oldret = dht_inode_ctx_layout_get (inode, this, &old_layout);
                 layout->ref++;
-                ret = inode_ctx_put (inode, this, (uint64_t) (unsigned long)
-                                     layout);
+                dht_inode_ctx_layout_set (inode, this, layout);
         }
         UNLOCK (&conf->layout_lock);
 
-        if (oldret == 0) {
-                old_layout = (dht_layout_t *) (unsigned long) old_layout_int;
+        if (!oldret) {
                 dht_layout_unref (this, old_layout);
         }
 
+out:
         return ret;
 }
 
@@ -119,10 +113,11 @@ dht_layout_unref (xlator_t *this, dht_layout_t *layout)
         dht_conf_t  *conf = NULL;
         int          ref = 0;
 
-        if (layout->preset)
+        if (!layout || layout->preset || !this->private)
                 return;
 
         conf = this->private;
+
         LOCK (&conf->layout_lock);
         {
                 ref = --layout->ref;
@@ -130,7 +125,7 @@ dht_layout_unref (xlator_t *this, dht_layout_t *layout)
         UNLOCK (&conf->layout_lock);
 
         if (!ref)
-                FREE (layout);
+                GF_FREE (layout);
 }
 
 
@@ -139,7 +134,7 @@ dht_layout_ref (xlator_t *this, dht_layout_t *layout)
 {
         dht_conf_t  *conf = NULL;
 
-        if (layout->preset)
+        if (layout->preset || !this->private)
                 return layout;
 
         conf = this->private;
@@ -156,527 +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;
+                }
+        }
 
-	return layout;
+out:
+        return layout;
 }
 
 
 int
 dht_layouts_init (xlator_t *this, dht_conf_t *conf)
 {
-	dht_layout_t *layout = NULL;
-	int           i = 0;
-	int           ret = -1;
-	
-
-	conf->file_layouts = CALLOC (conf->subvolume_cnt,
-				     sizeof (dht_layout_t *));
-	if (!conf->file_layouts) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto out;
-	}
+        dht_layout_t *layout = NULL;
+        int           i = 0;
+        int           ret = -1;
+
+        if (!conf)
+                goto out;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		layout = dht_layout_new (this, 1);
+        conf->file_layouts = GF_CALLOC (conf->subvolume_cnt,
+                                        sizeof (dht_layout_t *),
+                                        gf_dht_mt_dht_layout_t);
+        if (!conf->file_layouts) {
+                goto out;
+        }
 
-		if (!layout) {
-			goto out;
-		}
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                layout = dht_layout_new (this, 1);
 
-		layout->preset = 1;
+                if (!layout) {
+                        goto out;
+                }
 
-		layout->list[0].xlator = conf->subvolumes[i];
+                layout->preset = 1;
 
-		conf->file_layouts[i] = layout;
-	}
+                layout->list[0].xlator = conf->subvolumes[i];
+
+                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;
-
-	disk_layout = CALLOC (5, sizeof (int));
-	if (!disk_layout) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto out;
-	}
+        int      ret = -1;
+        int32_t *disk_layout = NULL;
+
+        disk_layout = GF_CALLOC (5, sizeof (int),
+                                 gf_dht_mt_int32_t);
+        if (!disk_layout) {
+                goto out;
+        }
+
+        disk_layout[0] = hton32 (1);
+        disk_layout[1] = hton32 (layout->type);
+        disk_layout[2] = hton32 (layout->list[pos].start);
+        disk_layout[3] = hton32 (layout->list[pos].stop);
 
-	disk_layout[0] = hton32 (1);
-	disk_layout[1] = hton32 (layout->type);
-	disk_layout[2] = hton32 (layout->list[pos].start);
-	disk_layout[3] = hton32 (layout->list[pos].stop);
+        if (disk_layout_p)
+                *disk_layout_p = disk_layout;
+        else
+                GF_FREE (disk_layout);
 
-	if (disk_layout_p)
-		*disk_layout_p = disk_layout;
-	ret = 0;
+        ret = 0;
 
 out:
-	return ret;
+        return ret;
 }
 
 
 int
 dht_disk_layout_merge (xlator_t *this, dht_layout_t *layout,
-		       int pos, int32_t *disk_layout)
+		       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];
+
+	if (!disk_layout_raw) {
+		gf_log (this->name, GF_LOG_CRITICAL,
+                        "error no layout on disk for merge");
+		return -1;
+	}
+
+	GF_ASSERT (disk_layout_len == sizeof (disk_layout));
 
-	/* TODO: assert disk_layout_ptr is of required length */
+        memcpy (disk_layout, disk_layout_raw, disk_layout_len);
+
+        cnt  = ntoh32 (disk_layout[0]);
+        if (cnt != 1) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "disk layout has invalid count %d", cnt);
+                return -1;
+        }
 
-	cnt  = ntoh32 (disk_layout[0]);
-	if (cnt != 1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"disk layout has invalid count %d", cnt);
+        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;
-	int32_t *disk_layout = 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",
-				    VOID(&disk_layout));
-	}
+        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);
-	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;
+
+        /* 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;
+        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)
 {
-	dht_conf_t *conf = NULL;
-	uint32_t    holes    = 0;
-	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;
-
-
-	conf = this->private;
-
-	/* 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++;
-			holes += (layout->list[i].start - (prev_stop + 1));
-		}
-
-		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++;
-	holes += (last_stop - prev_stop);
+        if (holes_p)
+                *holes_p = hole_cnt;
 
-	if (holes_p)
-		*holes_p = hole_cnt;
+        if (overlaps_p)
+                *overlaps_p = overlap_cnt;
 
-	if (overlaps_p)
-		*overlaps_p = overlap_cnt;
+        if (missing_p)
+                *missing_p = missing;
 
-	if (missing_p)
-		*missing_p = missing;
+        if (down_p)
+                *down_p = down;
 
-	if (down_p)
-		*down_p = down;
+        if (misc_p)
+                *misc_p = misc;
 
-	if (misc_p)
-		*misc_p = misc;
+        if (no_space_p)
+                *no_space_p = no_space;
 
-	return ret;
+        return ret;
 }
 
 
 int
 dht_layout_normalize (xlator_t *this, loc_t *loc, dht_layout_t *layout)
 {
-	int          ret   = 0;
-	int          i = 0;
-	uint32_t     holes = 0;
-	uint32_t     overlaps = 0;
-	uint32_t     missing = 0;
-	uint32_t     down = 0;
-	uint32_t     misc = 0;
-
-
-	ret = dht_layout_sort (layout);
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"sort failed?! how the ....");
-		goto out;
-	}
+        int          ret   = 0;
+        int          i = 0;
+        uint32_t     holes = 0;
+        uint32_t     overlaps = 0;
+        uint32_t     missing = 0;
+        uint32_t     down = 0;
+        uint32_t     misc = 0;
+
+        ret = dht_layout_sort (layout);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "sort failed?! how the ....");
+                goto out;
+        }
 
-	ret = dht_layout_anomalies (this, loc, layout,
-				    &holes, &overlaps,
-				    &missing, &down, &misc);
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"error while finding anomalies in %s -- not good news",
-			loc->path);
-		goto out;
-	}
+        ret = dht_layout_anomalies (this, loc, layout,
+                                    &holes, &overlaps,
+                                    &missing, &down, &misc, NULL);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "error while finding anomalies in %s -- not good news",
+                        loc->path);
+                goto out;
+        }
 
-	if (holes || overlaps) {
-		if (missing == layout->cnt) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"directory %s looked up first time",
-				loc->path);
-		} else {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"found anomalies in %s. holes=%d overlaps=%d",
-				loc->path, holes, overlaps);
-		}
-		ret = 1;
-	}
+        if (holes || overlaps) {
+                if (missing == layout->cnt) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "directory %s looked up first time",
+                                loc->path);
+                } else {
+                        gf_log (this->name, GF_LOG_INFO,
+                                "found anomalies in %s. holes=%d overlaps=%d",
+                                loc->path, holes, overlaps);
+                }
+                ret = -1;
+        }
+
+        for (i = 0; i < layout->cnt; i++) {
+                /* TODO During DHT selfheal rewrite (almost) find a better place
+                 * to detect this - probably in dht_layout_anomalies()
+                 */
+                if (layout->list[i].err > 0) {
+                        gf_log_callingfn (this->name, GF_LOG_DEBUG,
+                                          "path=%s err=%s on subvol=%s",
+                                          loc->path,
+                                          strerror (layout->list[i].err),
+                                          (layout->list[i].xlator ?
+                                           layout->list[i].xlator->name
+                                           : "<>"));
+                        if ((layout->list[i].err == ENOENT) && (ret >= 0)) {
+                                ret++;
+                        }
+                }
+        }
 
-	for (i = 0; i < layout->cnt; i++) {
-	/* TODO During DHT selfheal rewrite (almost) find a better place to 
-	 * detect this - probably in dht_layout_anomalies() 
-	 */
-		if (layout->list[i].err > 0) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"path=%s err=%s on subvol=%s",
-                                loc->path, strerror (layout->list[i].err),
-                                (layout->list[i].xlator ?
-                                 layout->list[i].xlator->name : "<>"));
-                        if (layout->list[i].err == ENOENT)
-                                ret = 1;
-		}
-	}
 
 out:
-	return ret;
+        return ret;
 }
 
+int
+dht_dir_has_layout (dict_t *xattr, char *name)
+{
+
+        void     *disk_layout_raw = NULL;
+
+        return dict_get_ptr (xattr, name, &disk_layout_raw);
+}
 
 int
 dht_layout_dir_mismatch (xlator_t *this, dht_layout_t *layout, xlator_t *subvol,
-			 loc_t *loc, dict_t *xattr)
+                         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 = NULL;
-	int32_t   count = -1;
-	uint32_t  start_off = -1;
-	uint32_t  stop_off = -1;
-
-
-	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         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",
-                                 VOID(&disk_layout));
-	
-	if (dict_ret < 0) {
-                if (err == 0) {
-                        gf_log (this->name, GF_LOG_DEBUG,
+        dict_ret = dict_get_ptr (xattr, conf->xattr_name,
+                                 &disk_layout_raw);
+
+        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;
-	} 
-
-	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;
-	}
+                goto out;
+        }
 
-	start_off = ntoh32 (disk_layout[2]);
-	stop_off  = ntoh32 (disk_layout[3]);
-	
-	if ((layout->list[pos].start != start_off)
-	    || (layout->list[pos].stop != stop_off)) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"subvol: %s; inode layout - %"PRId32" - %"PRId32"; "
-			"disk layout - %"PRId32" - %"PRId32,
-			layout->list[pos].xlator->name,
-			layout->list[pos].start, layout->list[pos].stop,
-			start_off, stop_off);
-		ret = 1;
-	} else {
-		ret = 0;
-	}
+        memcpy (disk_layout, disk_layout_raw, sizeof (disk_layout));
+
+        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;
+        }
 out:
-	return ret;
+        return ret;
 }
 
 
@@ -688,19 +782,21 @@ dht_layout_preset (xlator_t *this, xlator_t *subvol, inode_t *inode)
         dht_conf_t   *conf = NULL;
 
         conf = this->private;
-
-	layout = dht_layout_for_subvol (this, subvol);
-	if (!layout) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"no pre-set layout for subvolume %s",
-			subvol ? subvol->name : "<nil>");
-		ret = -1;
-		goto out;
-	}
+        if (!conf)
+                goto out;
+
+        layout = dht_layout_for_subvol (this, subvol);
+        if (!layout) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "no pre-set layout for subvolume %s",
+                        subvol ? subvol->name : "<nil>");
+                ret = -1;
+                goto out;
+        }
 
         LOCK (&conf->layout_lock);
         {
-                inode_ctx_put (inode, this, (uint64_t)(long)layout);
+                dht_inode_ctx_layout_set (inode, this, layout);
         }
         UNLOCK (&conf->layout_lock);
 
diff --git a/xlators/cluster/dht/src/dht-linkfile.c b/xlators/cluster/dht/src/dht-linkfile.c
index 77e7818ef..dbc9d0b3c 100644
--- a/xlators/cluster/dht/src/dht-linkfile.c
+++ b/xlators/cluster/dht/src/dht-linkfile.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -28,199 +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 stat *stbuf, struct stat *preparent,
-                         struct stat *postparent)
+                         struct iatt *stbuf, struct iatt *preparent,
+                         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");
-		op_errno = EINVAL;
-	}
-	str_data = NULL;
-
-	local->linkfile.stbuf = *stbuf;
-
-	STACK_WIND (frame, dht_linkfile_xattr_cbk,
-		    prev->this, prev->this->fops->setxattr,
-		    &local->linkfile.loc, local->linkfile.xattr, 0);
-
-	return 0;
-
-err:
-	if (str_data) {
-		data_destroy (str_data);
-		str_data = NULL;
-	}
-
+        dht_local_t  *local = NULL;
+        xlator_t     *subvol = NULL;
+        call_frame_t *prev = NULL;
+        dict_t       *xattrs = NULL;
+        dht_conf_t   *conf = NULL;
+        int           ret = -1;
+
+        local = frame->local;
+
+        if (!op_ret)
+                local->linked = _gf_true;
+
+        FRAME_SU_UNDO (frame, dht_local_t);
+
+        if (op_ret && (op_errno == EEXIST)) {
+                conf = this->private;
+                prev = cookie;
+                subvol = prev->this;
+                if (!subvol)
+                        goto out;
+                xattrs = dict_new ();
+                if (!xattrs)
+                        goto out;
+                ret = dict_set_uint32 (xattrs, conf->link_xattr_name, 256);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to set linkto key");
+                        goto out;
+                }
+
+                STACK_WIND (frame, dht_linkfile_lookup_cbk, subvol,
+                            subvol->fops->lookup, &local->loc, xattrs);
+                if (xattrs)
+                        dict_unref (xattrs);
+                return 0;
+        }
+out:
         local->linkfile.linkfile_cbk (frame, cookie, this, op_ret, op_errno,
-                                      inode, stbuf, preparent, postparent);
-	return 0;
+                                      inode, stbuf, preparent, postparent,
+                                      xdata);
+        if (xattrs)
+                dict_unref (xattrs);
+        return 0;
 }
 
 
 int
 dht_linkfile_create (call_frame_t *frame, fop_mknod_cbk_t linkfile_cbk,
-		     xlator_t *tovol, xlator_t *fromvol, loc_t *loc)
+                     xlator_t *this,
+                     xlator_t *tovol, xlator_t *fromvol, loc_t *loc)
 {
-	dht_local_t *local = NULL;
-
-
-	local = frame->local;
-	local->linkfile.linkfile_cbk = linkfile_cbk;
-	local->linkfile.srcvol = tovol;
-	loc_copy (&local->linkfile.loc, loc);
+        dht_local_t *local = NULL;
+        dict_t      *dict = NULL;
+        int          need_unref = 0;
+        int          ret = 0;
+        dht_conf_t  *conf = this->private;
+
+        local = frame->local;
+        local->linkfile.linkfile_cbk = linkfile_cbk;
+        local->linkfile.srcvol = tovol;
+
+        local->linked = _gf_false;
+
+        dict = local->params;
+        if (!dict) {
+                dict = dict_new ();
+                if (!dict)
+                        goto out;
+                need_unref = 1;
+        }
+
+        if (!uuid_is_null (local->gfid)) {
+                ret = dict_set_static_bin (dict, "gfid-req", local->gfid, 16);
+                if (ret)
+                        gf_log ("dht-linkfile", GF_LOG_INFO,
+                                "%s: gfid set failed", loc->path);
+        }
+
+        ret = dict_set_str (dict, GLUSTERFS_INTERNAL_FOP_KEY, "yes");
+        if (ret)
+                gf_log ("dht-linkfile", GF_LOG_INFO,
+                        "%s: internal-fop set failed", loc->path);
+
+        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, NULL);
 
-	STACK_WIND (frame, dht_linkfile_create_cbk,
-		    fromvol, fromvol->fops->mknod, loc,
-		    S_IFREG | DHT_LINKFILE_MODE, 0);
+        if (need_unref && dict)
+                dict_unref (dict);
 
-	return 0;
+        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 stat *preparent, struct stat *postparent)
+                         int32_t op_ret, int32_t op_errno,
+                         struct iatt *preparent, struct iatt *postparent,
+                         dict_t *xdata)
 {
-	dht_local_t   *local = NULL;
-	call_frame_t  *prev = NULL;
-	xlator_t      *subvol = NULL;
+        dht_local_t   *local = NULL;
+        call_frame_t  *prev = NULL;
+        xlator_t      *subvol = NULL;
 
-	local = frame->local;
-	prev = cookie;
-	subvol = prev->this;
+        local = frame->local;
+        prev = cookie;
+        subvol = prev->this;
 
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"unlinking linkfile %s on %s failed (%s)",
-			local->loc.path, subvol->name, strerror (op_errno));
-	}
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "unlinking linkfile %s on %s failed (%s)",
+                        local->loc.path, subvol->name, strerror (op_errno));
+        }
 
-	DHT_STACK_DESTROY (frame);
+        DHT_STACK_DESTROY (frame);
 
-	return 0;
+        return 0;
 }
 
 
 int
 dht_linkfile_unlink (call_frame_t *frame, xlator_t *this,
-		     xlator_t *subvol, loc_t *loc)
+                     xlator_t *subvol, loc_t *loc)
 {
-	call_frame_t *unlink_frame = NULL;
-	dht_local_t  *unlink_local = NULL;
-
-	unlink_frame = copy_frame (frame);
-	if (!unlink_frame) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	unlink_local = dht_local_init (unlink_frame);
-	if (!unlink_local) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
-
-	loc_copy (&unlink_local->loc, loc);
-
-	STACK_WIND (unlink_frame, dht_linkfile_unlink_cbk,
-		    subvol, subvol->fops->unlink,
-		    &unlink_local->loc);
-
-	return 0;
+        call_frame_t *unlink_frame = NULL;
+        dht_local_t  *unlink_local = NULL;
+
+        unlink_frame = copy_frame (frame);
+        if (!unlink_frame) {
+                goto err;
+        }
+
+        /* Using non-fop value here, as anyways, 'local->fop' is not used in
+           this particular case */
+        unlink_local = dht_local_init (unlink_frame, loc, NULL,
+                                       GF_FOP_MAXVALUE);
+        if (!unlink_local) {
+                goto err;
+        }
+
+        STACK_WIND (unlink_frame, dht_linkfile_unlink_cbk,
+                    subvol, subvol->fops->unlink,
+                    &unlink_local->loc, 0, NULL);
+
+        return 0;
 err:
-	if (unlink_frame)
-		DHT_STACK_DESTROY (unlink_frame);
+        if (unlink_frame)
+                DHT_STACK_DESTROY (unlink_frame);
 
-	return -1;
+        return -1;
 }
 
 
 xlator_t *
-dht_linkfile_subvol (xlator_t *this, inode_t *inode, struct stat *stbuf,
-		     dict_t *xattr)
+dht_linkfile_subvol (xlator_t *this, inode_t *inode, struct iatt *stbuf,
+                     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
new file mode 100644
index 000000000..e893eb48f
--- /dev/null
+++ b/xlators/cluster/dht/src/dht-mem-types.h
@@ -0,0 +1,35 @@
+/*
+  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 __DHT_MEM_TYPES_H__
+#define __DHT_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_dht_mem_types_ {
+        gf_dht_mt_dht_du_t = gf_common_mt_end + 1,
+        gf_dht_mt_dht_conf_t,
+        gf_dht_mt_char,
+        gf_dht_mt_int32_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 92e305613..925538cc8 100644
--- a/xlators/cluster/dht/src/dht-rename.c
+++ b/xlators/cluster/dht/src/dht-rename.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 /* TODO: link(oldpath, newpath) fails if newpath already exists. DHT should
@@ -33,422 +24,801 @@
 
 int
 dht_rename_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno, struct stat *stbuf,
-                    struct stat *preoldparent, struct stat *postoldparent,
-                    struct stat *prenewparent, struct stat *postnewparent)
+                    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_local_t  *local = NULL;
-	int           this_call_cnt = 0;
-	call_frame_t *prev = NULL;
-
-
-	local = frame->local;
-	prev = cookie;
-
-	if (op_ret == -1) {
-		/* TODO: undo the damage */
+        dht_local_t  *local = NULL;
+        int           this_call_cnt = 0;
+        call_frame_t *prev = NULL;
+
+        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);
 
-		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_stat_merge (this, &local->stbuf, stbuf, prev->this);
-                dht_stat_merge (this, &local->preoldparent, preoldparent,
-                                prev->this);
-                dht_stat_merge (this, &local->postoldparent, postoldparent,
-                                prev->this);
-                dht_stat_merge (this, &local->preparent, prenewparent,
-                                prev->this);
-                dht_stat_merge (this, &local->postparent, postnewparent,
-                                prev->this);
-	}
-
-	this_call_cnt = dht_frame_return (frame);
-	if (is_last_call (this_call_cnt)) {
-                local->stbuf.st_ino = local->loc.inode->ino;
-
-                local->preoldparent.st_ino = local->loc.parent->ino;
-                local->postoldparent.st_ino = local->loc.parent->ino;
-
-                local->preparent.st_ino = local->loc2.parent->ino;
-                local->postparent.st_ino = local->loc2.parent->ino;
-
-		DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
-				  &local->stbuf, &local->preoldparent,
+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_STRIP_PHASE1_FLAGS (&local->stbuf);
+                DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
+                                  &local->stbuf, &local->preoldparent,
                                   &local->postoldparent,
-                                  &local->preparent, &local->postparent);
-	}
+                                  &local->preparent, &local->postparent, xdata);
+        }
 
-	return 0;
+        return 0;
 }
 
 
+int
+dht_rename_hashed_dir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                           int32_t op_ret, int32_t op_errno, struct iatt *stbuf,
+                           struct iatt *preoldparent,
+                           struct iatt *postoldparent,
+                           struct iatt *prenewparent,
+                           struct iatt *postnewparent, dict_t *xdata)
+{
+        dht_conf_t   *conf = NULL;
+        dht_local_t  *local = NULL;
+        int           call_cnt = 0;
+        call_frame_t *prev = NULL;
+        int           i = 0;
+
+        conf = this->private;
+        local = frame->local;
+        prev = cookie;
+
+        if (op_ret == -1) {
+                /* TODO: undo the damage */
+
+                gf_log (this->name, GF_LOG_INFO,
+                        "rename %s -> %s on %s failed (%s)",
+                        local->loc.path, local->loc2.path,
+                        prev->this->name, strerror (op_errno));
+
+                local->op_ret   = op_ret;
+                local->op_errno = op_errno;
+                goto unwind;
+        }
+        /* TODO: construct proper stbuf for dir */
+        /*
+         * FIXME: is this the correct way to build stbuf and
+         * parent bufs?
+         */
+        dht_iatt_merge (this, &local->stbuf, stbuf, prev->this);
+        dht_iatt_merge (this, &local->preoldparent, preoldparent,
+                        prev->this);
+        dht_iatt_merge (this, &local->postoldparent, postoldparent,
+                        prev->this);
+        dht_iatt_merge (this, &local->preparent, prenewparent,
+                        prev->this);
+        dht_iatt_merge (this, &local->postparent, postnewparent,
+                        prev->this);
+
+        call_cnt = local->call_cnt = conf->subvolume_cnt - 1;
+
+        if (!local->call_cnt)
+                goto unwind;
+
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (conf->subvolumes[i] == local->dst_hashed)
+                        continue;
+                STACK_WIND (frame, dht_rename_dir_cbk,
+                            conf->subvolumes[i],
+                            conf->subvolumes[i]->fops->rename,
+                            &local->loc, &local->loc2, NULL);
+                if (!--call_cnt)
+                        break;
+        }
+
+
+        return 0;
+unwind:
+        WIPE (&local->preoldparent);
+        WIPE (&local->postoldparent);
+        WIPE (&local->preparent);
+        WIPE (&local->postparent);
+
+        DHT_STRIP_PHASE1_FLAGS (&local->stbuf);
+        DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
+                          &local->stbuf, &local->preoldparent,
+                          &local->postoldparent,
+                          &local->preparent, &local->postparent, NULL);
+
+        return 0;
+}
+
 
 int
 dht_rename_dir_do (call_frame_t *frame, xlator_t *this)
 {
-	dht_local_t  *local = NULL;
-	dht_conf_t   *conf = NULL;
-	int           i = 0;
-
-	conf = this->private;
-	local = frame->local;
+        dht_local_t  *local = NULL;
 
-	if (local->op_ret == -1)
-		goto err;
+        local = frame->local;
 
-	local->call_cnt = conf->subvolume_cnt;
-	local->op_ret = 0;
+        if (local->op_ret == -1)
+                goto err;
 
-	for (i = 0; i < conf->subvolume_cnt; i++) {
-		STACK_WIND (frame, dht_rename_dir_cbk,
-			    conf->subvolumes[i],
-			    conf->subvolumes[i]->fops->rename,
-			    &local->loc, &local->loc2);
-	}
+        local->op_ret = 0;
 
-	return 0;
+        STACK_WIND (frame, dht_rename_hashed_dir_cbk,
+                    local->dst_hashed,
+                    local->dst_hashed->fops->rename,
+                    &local->loc, &local->loc2, NULL);
+        return 0;
 
 err:
-	DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno, NULL, NULL,
-                          NULL, NULL, NULL);
-	return 0;
+        DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno, NULL, NULL,
+                          NULL, NULL, NULL, NULL);
+        return 0;
 }
 
 
 int
 dht_rename_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			int op_ret, int op_errno, gf_dirent_t *entries)
+                        int op_ret, int op_errno, gf_dirent_t *entries,
+                        dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	int           this_call_cnt = -1;
-	call_frame_t *prev = NULL;
+        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 > 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;
-	}
+        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;
 
-	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++) {
+                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->op_ret = 0;
+        local->fd = fd_create (local->loc.inode, frame->root->pid);
+        if (!local->fd) {
+                op_errno = ENOMEM;
+                goto err;
+        }
 
-	if (!local->dst_cached) {
-		dht_rename_dir_do (frame, this);
-		return 0;
-	}
+        local->op_ret = 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);
-	}
+        if (!local->dst_cached) {
+                dht_rename_dir_do (frame, this);
+                return 0;
+        }
 
-	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, NULL);
+        }
+
+        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 stat *preparent,
-                       struct stat *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;
+
+        if (!local) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "!local, should not happen");
+                goto out;
+        }
+
+        this_call_cnt = dht_frame_return (frame);
+
+        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_rename_done (frame, this);
+        }
+
+out:
+        return 0;
+}
 
-	local = frame->local;
-	prev  = cookie;
 
-	this_call_cnt = dht_frame_return (frame);
+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;
+        dict_t      *xattr      = NULL;
 
-	if (op_ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"unlink on %s failed (%s)",
-			prev->this->name, strerror (op_errno));
-	}
+        local = frame->local;
+        this  = frame->this;
 
-	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);
+        src_hashed = local->src_hashed;
+        src_cached = local->src_cached;
+        dst_hashed = local->dst_hashed;
+        dst_cached = local->dst_cached;
 
-	return 0;
+        if (src_cached == dst_cached)
+                goto nolinks;
+
+        if (dst_hashed != src_hashed && dst_hashed != src_cached)
+                call_cnt++;
+
+        if (src_cached != dst_hashed)
+                call_cnt++;
+
+        local->call_cnt = call_cnt;
+
+        if (!call_cnt)
+                goto nolinks;
+
+        DHT_MARK_FOP_INTERNAL (xattr);
+
+        if (dst_hashed != src_hashed && dst_hashed != src_cached) {
+                dict_t *xattr_new = NULL;
+
+                gf_log (this->name, GF_LOG_TRACE,
+                        "unlinking linkfile %s @ %s => %s",
+                        local->loc.path, dst_hashed->name, src_cached->name);
+
+                xattr_new = dict_copy_with_ref (xattr, NULL);
+
+
+                DHT_MARKER_DONT_ACCOUNT(xattr_new);
+
+                STACK_WIND (frame, dht_rename_unlink_cbk,
+                            dst_hashed, dst_hashed->fops->unlink,
+                            &local->loc, 0, xattr_new);
+
+                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;
+
+nolinks:
+        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_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;
 }
 
 
 int
 dht_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		int32_t op_ret, int32_t op_errno, struct stat *stbuf,
-                struct stat *preoldparent, struct stat *postoldparent,
-                struct stat *prenewparent, struct stat *postnewparent)
+                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_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;
+        }
+
+        if ((src_cached == dst_cached) && (dst_hashed != dst_cached)) {
+                link_frame = copy_frame (frame);
+                if (!link_frame) {
+                        goto err;
+                }
+
+                /* 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 (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_linkfile_create (link_frame, dht_rename_links_create_cbk,
+                                     this, src_cached, dst_hashed,
+                                     &link_local->loc);
+        }
 
-	local = frame->local;
-	prev = cookie;
+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);
+        }
 
-	src_hashed = local->src_hashed;
-	src_cached = local->src_cached;
-	dst_hashed = local->dst_hashed;
-	dst_cached = local->dst_cached;
 
-	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 unwind;
-	}
+        /* 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;
 
-        dht_stat_merge (this, &local->stbuf, stbuf, prev->this);
-        dht_stat_merge (this, &local->preoldparent, preoldparent, prev->this);
-        dht_stat_merge (this, &local->postoldparent, postoldparent, prev->this);
-        dht_stat_merge (this, &local->preparent, prenewparent, prev->this);
-        dht_stat_merge (this, &local->postparent, postnewparent, prev->this);
+        /* TODO: delete files in background */
 
-        local->stbuf.st_ino = local->loc.inode->ino;
+        if (src_cached != dst_hashed && src_cached != dst_cached)
+                local->call_cnt++;
 
-        local->preoldparent.st_ino = local->loc.parent->ino;
-        local->postoldparent.st_ino = local->loc.parent->ino;
+        if (src_hashed != rename_subvol && src_hashed != src_cached)
+                local->call_cnt++;
 
-        local->preparent.st_ino = local->loc2.parent->ino;
-        local->postparent.st_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;
+        if (dst_cached && dst_cached != dst_hashed && dst_cached != src_cached)
+                local->call_cnt++;
 
-	/* TODO: delete files in background */
+        if (local->call_cnt == 0)
+                goto unwind;
 
-	if (src_cached != dst_hashed && src_cached != dst_cached)
-		local->call_cnt++;
+        DHT_MARK_FOP_INTERNAL (xattr);
 
-	if (src_hashed != rename_subvol && src_hashed != src_cached)
-		local->call_cnt++;
+        if (src_cached != dst_hashed && src_cached != dst_cached) {
+                dict_t *xattr_new = NULL;
 
-	if (dst_cached && dst_cached != dst_hashed && dst_cached != src_cached)
-		local->call_cnt++;
+                xattr_new = dict_copy_with_ref (xattr, NULL);
 
-	if (local->call_cnt == 0)
-		goto unwind;
+                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) {
-		gf_log (this->name, GF_LOG_TRACE,
-			"deleting old src datafile %s @ %s",
-			local->loc.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_unlink_cbk,
-			    src_cached, src_cached->fops->unlink,
-			    &local->loc);
-	}
+                STACK_WIND (frame, dht_rename_unlink_cbk,
+                            src_cached, src_cached->fops->unlink,
+                            &local->loc, 0, xattr_new);
 
-	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);
+                dict_unref (xattr_new);
+                xattr_new = NULL;
+        }
 
-		STACK_WIND (frame, dht_rename_unlink_cbk,
-			    src_hashed, src_hashed->fops->unlink,
-			    &local->loc);
-	}
+        if (src_hashed != rename_subvol && src_hashed != src_cached) {
+                dict_t *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);
+                xattr_new = dict_copy_with_ref (xattr, NULL);
 
-		STACK_WIND (frame, dht_rename_unlink_cbk,
-			    dst_cached, dst_cached->fops->unlink,
-			    &local->loc2);
-	}
-	return 0;
+                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:
-	DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
-			  &local->stbuf, &local->preoldparent,
-                          &local->postoldparent, &local->preparent,
-                          &local->postparent);
+        WIPE (&local->preoldparent);
+        WIPE (&local->postoldparent);
+        WIPE (&local->preparent);
+        WIPE (&local->postparent);
+        if (xattr)
+                dict_unref (xattr);
 
-	return 0;
+        dht_rename_done (frame, this);
+
+        return 0;
+
+cleanup:
+        if (xattr)
+                dict_unref (xattr);
+        dht_rename_cleanup (frame);
+
+        return 0;
 }
 
 
 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;
+
+        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_hashed) && (rename_subvol == dst_hashed)) {
+                DHT_MARKER_DONT_ACCOUNT(dict);
+        }
+
+        gf_log (this->name, GF_LOG_TRACE,
+                "renaming %s => %s (%s)",
+                local->loc.path, local->loc2.path, rename_subvol->name);
+
+        if (local->linked == _gf_true)
+                FRAME_SU_DO (frame, dht_local_t);
+        STACK_WIND (frame, dht_rename_cbk,
+                    rename_subvol, rename_subvol->fops->rename,
+                    &local->loc, &local->loc2, dict);
+
+        return 0;
+}
 
 
-	local = frame->local;
-	this  = frame->this;
+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)
+{
+        dht_local_t  *local = NULL;
+        call_frame_t *prev = NULL;
+        int           this_call_cnt  = 0;
 
-	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;
+        local = frame->local;
+        prev = cookie;
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"renaming %s => %s (%s)",
-		local->loc.path, local->loc2.path, rename_subvol->name);
+        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);
+        }
 
-	STACK_WIND (frame, dht_rename_cbk,
-		    rename_subvol, rename_subvol->fops->rename,
-		    &local->loc, &local->loc2);
+        this_call_cnt = dht_frame_return (frame);
+        if (is_last_call (this_call_cnt)) {
+                if (local->op_ret == -1)
+                        goto cleanup;
 
-	return 0;
+                dht_do_rename (frame);
+        }
+
+        return 0;
+
+cleanup:
+        dht_rename_cleanup (frame);
+
+        return 0;
 }
 
 
 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 stat *stbuf,
-                      struct stat *preparent, struct stat *postparent)
+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 unwind;
-		
-		dht_do_rename (frame);
-	}
+        if (local->op_ret == -1)
+                goto cleanup;
+
+        dht_do_rename (frame);
 
 	return 0;
 
-unwind:
-	DHT_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
-			  &local->stbuf, &local->preoldparent,
-                          &local->postoldparent, &local->preparent,
-                          &local->postparent);
+cleanup:
+        dht_rename_cleanup (frame);
 
 	return 0;
 }
@@ -457,155 +827,183 @@ unwind:
 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);
-		dht_linkfile_create (frame, dht_rename_links_cbk,
+			"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, 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);
 
-nolinks:
-	if (!call_cnt) {
-		/* skip to next step */
-		dht_do_rename (frame);
+                dict_unref (xattr_new);
 	}
 
-	return 0;
+nolinks:
+        if (!call_cnt) {
+                /* skip to next step */
+                dht_do_rename (frame);
+        }
+        if (xattr)
+                dict_unref (xattr);
+
+        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;
-	}
-
-	if (newloc->inode)
-		dst_cached = dht_subvol_get_cached (this, newloc->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_copy (&local->loc, oldloc);
-	if (ret == -1) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		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;
-	}
-
-	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 (S_ISDIR (oldloc->inode->st_mode)) {
-		dht_rename_dir (frame, this);
-	} else {
-		local->op_ret = 0;
-		dht_rename_create_links (frame);
-	}
-
-	return 0;
+        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;
+        }
+
+        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;
+        }
+
+        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;
+        }
+
+        if (newloc->inode)
+                dst_cached = dht_subvol_get_cached (this, newloc->inode);
+
+        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);
+        }
+
+        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 a254b0ddc..0e6527544 100644
--- a/xlators/cluster/dht/src/dht-selfheal.c
+++ b/xlators/cluster/dht/src/dht-selfheal.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 #ifndef _CONFIG_H
@@ -26,256 +17,530 @@
 #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;
 
-	subvol = layout->list[i].xlator;
-	this = frame->this;
+        xattr = get_new_dict ();
+        if (!xattr) {
+                goto err;
+        }
 
-	xattr = get_new_dict ();
-	if (!xattr) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        ret = dht_disk_layout_extract (this, layout, i, &disk_layout);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: (subvol %s) failed to extract disk layout",
+                        loc->path, subvol->name);
+                goto err;
+        }
 
-	ret = dht_disk_layout_extract (this, layout, i, &disk_layout);
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"failed to extract disk layout");
-		goto err;
-	}
+        ret = dict_set_bin (xattr, conf->xattr_name, disk_layout, 4 * 4);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: (subvol %s) failed to set xattr dictionary",
+                        loc->path, subvol->name);
+                goto err;
+        }
+        disk_layout = NULL;
+
+        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);
 
-	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;
+        STACK_WIND (frame, dht_selfheal_dir_xattr_cbk,
+                    subvol, subvol->fops->setxattr,
+                    loc, xattr, 0, 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_unref (xattr);
 
-	dict_ref (xattr);
+        return 0;
 
-	STACK_WIND (frame, dht_selfheal_dir_xattr_cbk,
-		    subvol, subvol->fops->setxattr,
-		    loc, xattr, 0);
+err:
+        if (xattr)
+                dict_destroy (xattr);
 
-	dict_unref (xattr);
+        GF_FREE (disk_layout);
 
-	return 0;
+        dht_selfheal_dir_xattr_cbk (frame, subvol, frame->this,
+                                    -1, ENOMEM, NULL);
+        return 0;
+}
 
-err:
-	if (xattr)
-		dict_destroy (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;
+
+        local = frame->local;
+        this = frame->this;
+        conf = this->private;
 
-	if (disk_layout)
-		FREE (disk_layout);
+        gf_log (this->name, GF_LOG_DEBUG,
+                "writing the new range for all subvolumes");
 
-	dht_selfheal_dir_xattr_cbk (frame, subvol, frame->this,
-				    -1, ENOMEM);
-	return 0;
-}
+        local->call_cnt = count = conf->subvolume_cnt;
+
+        for (i = 0; i < layout->cnt; i++) {
+                dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i, NULL);
 
+                if (--count == 0)
+                        goto out;
+        }
+        /* if we are here, subvolcount > layout_count. subvols-per-directory
+         * option might be set here. We need to clear out layout from the
+         * non-participating subvolumes, else it will result in overlaps */
+        dummy = dht_layout_new (this, 1);
+        if (!dummy)
+                goto out;
+        for (i = 0; i < conf->subvolume_cnt; i++) {
+                if (_gf_false ==
+                    dht_is_subvol_in_layout (layout, conf->subvolumes[i])) {
+                        dht_selfheal_dir_xattr_persubvol (frame, loc, dummy, 0,
+                                                          conf->subvolumes[i]);
+                        if (--count == 0)
+                                break;
+                }
+        }
+
+        dht_layout_unref (this, dummy);
+out:
+        return 0;
+}
 
 int
 dht_selfheal_dir_xattr (call_frame_t *frame, loc_t *loc, dht_layout_t *layout)
 {
-	dht_local_t *local = NULL;
-	int          missing_xattr = 0;
-	int          i = 0;
-	int          ret = 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;
 
-	gf_log (this->name, GF_LOG_TRACE,
-		"%d subvolumes missing xattr for %s",
-		missing_xattr, loc->path);
+        local = frame->local;
+        this = frame->this;
+        conf = this->private;
 
-	if (missing_xattr == 0) {
-		dht_selfheal_dir_finish (frame, this, 0);
-		return 0;
-	}
+        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);
 
-	local->call_cnt = missing_xattr;
+        if (missing_xattr == 0) {
+                dht_selfheal_dir_finish (frame, this, 0);
+                return 0;
+        }
 
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].err != -1 || !layout->list[i].stop)
-			continue;
+        local->call_cnt = missing_xattr;
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err != -1 || !layout->list[i].stop)
+                        continue;
 
-		ret = dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i);
+                dht_selfheal_dir_xattr_persubvol (frame, loc, layout, i, NULL);
 
-		if (--missing_xattr == 0)
-			break;
-	}
-	return 0;
+                if (--missing_xattr == 0)
+                        break;
+        }
+        dummy = dht_layout_new (this, 1);
+        if (!dummy)
+                goto out;
+        for (i = 0; i < conf->subvolume_cnt && 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;
+}
+
+int
+dht_selfheal_dir_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                              int op_ret, int op_errno, struct iatt *statpre,
+                              struct iatt *statpost, dict_t *xdata)
+{
+        dht_local_t   *local = NULL;
+        dht_layout_t  *layout = NULL;
+        int            this_call_cnt = 0;
+
+        local  = frame->local;
+        layout = local->selfheal.layout;
+
+        this_call_cnt = dht_frame_return (frame);
+
+        if (is_last_call (this_call_cnt)) {
+                dht_selfheal_dir_xattr (frame, &local->loc, layout);
+        }
+
+        return 0;
 }
 
 
 int
+dht_selfheal_dir_setattr (call_frame_t *frame, loc_t *loc, struct iatt *stbuf,
+                          int32_t valid, dht_layout_t *layout)
+{
+        int           missing_attr = 0;
+        int           i     = 0;
+        dht_local_t  *local = NULL;
+        xlator_t     *this = NULL;
+
+        local = frame->local;
+        this = frame->this;
+
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err == -1)
+                        missing_attr++;
+        }
+
+        if (missing_attr == 0) {
+                dht_selfheal_dir_xattr (frame, loc, layout);
+                return 0;
+        }
+
+        if (!uuid_is_null (local->gfid))
+                uuid_copy (loc->gfid, local->gfid);
+
+        local->call_cnt = missing_attr;
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err == -1) {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "setattr for %s on subvol %s",
+                                loc->path, layout->list[i].xlator->name);
+
+                        STACK_WIND (frame, dht_selfheal_dir_setattr_cbk,
+                                    layout->list[i].xlator,
+                                    layout->list[i].xlator->fops->setattr,
+                                    loc, stbuf, valid, NULL);
+                }
+        }
+
+        return 0;
+}
+
+int
 dht_selfheal_dir_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-			    int op_ret, int op_errno,
-                            inode_t *inode, struct stat *stbuf,
-                            struct stat *preparent, struct stat *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;
-	dht_layout_t  *layout = NULL;
-	call_frame_t  *prev = NULL;
-	xlator_t      *subvol = NULL;
-	int            i = 0;
-	int            this_call_cnt = 0;
+        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;
+        local  = frame->local;
+        layout = local->selfheal.layout;
+        prev   = cookie;
+        subvol = prev->this;
 
-        dht_stat_merge (this, &local->stbuf, stbuf, prev->this);
-        if (prev->this == local->hashed_subvol)
-                local->st_ino = local->stbuf.st_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;
+                        }
+                }
+        }
 
-        dht_stat_merge (this, &local->preparent, preparent, prev->this);
-        dht_stat_merge (this, &local->postparent, postparent, prev->this);
+        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;
+        }
 
-	if ((op_ret == 0) || (op_errno == EEXIST)) {
-		for (i = 0; i < layout->cnt; i++) {
-			if (layout->list[i].xlator == subvol) {
-				layout->list[i].err = -1;
-				break;
-			}
-		}
-	}
+        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);
 
-	this_call_cnt = dht_frame_return (frame);
+out:
+        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;
-	dht_local_t  *local = NULL;
-	xlator_t     *this = NULL;
-
-
-	local = frame->local;
-	this = frame->this;
+        int           missing_dirs = 0;
+        int           i     = 0;
+        int           ret   = -1;
+        dht_local_t  *local = NULL;
+        xlator_t     *this = NULL;
+        dict_t       *dict = NULL;
+
+        local = frame->local;
+        this = frame->this;
+
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err == ENOENT || force)
+                        missing_dirs++;
+        }
 
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].err == ENOENT || force)
-			missing_dirs++;
-	}
+        if (missing_dirs == 0) {
+                dht_selfheal_dir_setattr (frame, loc, &local->stbuf, 0xffffffff, layout);
+                return 0;
+        }
 
-	if (missing_dirs == 0) {
-		dht_selfheal_dir_xattr (frame, loc, layout);
-		return 0;
-	}
+        local->call_cnt = missing_dirs;
+        if (!uuid_is_null (local->gfid)) {
+                dict = dict_new ();
+                if (!dict)
+                        return -1;
+
+                ret = dict_set_static_bin (dict, "gfid-req", local->gfid, 16);
+                if (ret)
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s: failed to set gfid in dict", loc->path);
+        } else if (local->params) {
+                /* Send the dictionary from higher layers directly */
+                dict = dict_ref (local->params);
+        }
+        /* Set acls */
+        if (local->xattr && dict)
+                dht_selfheal_dir_mkdir_setacl (local->xattr, dict);
+
+        if (!dict)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "dict is NULL, need to make sure gfids are same");
+
+        for (i = 0; i < layout->cnt; i++) {
+                if (layout->list[i].err == ENOENT || force) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "creating directory %s on subvol %s",
+                                loc->path, layout->list[i].xlator->name);
+
+                        STACK_WIND (frame, dht_selfheal_dir_mkdir_cbk,
+                                    layout->list[i].xlator,
+                                    layout->list[i].xlator->fops->mkdir,
+                                    loc,
+                                    st_mode_from_ia (local->stbuf.ia_prot,
+                                                     local->stbuf.ia_type),
+                                    0, dict);
+                }
+        }
 
-	local->call_cnt = missing_dirs;
-	for (i = 0; i < layout->cnt; i++) {
-		if (layout->list[i].err == ENOENT || force) {
-			gf_log (this->name, GF_LOG_TRACE,
-				"creating directory %s on subvol %s",
-				loc->path, layout->list[i].xlator->name);
-
-			STACK_WIND (frame, dht_selfheal_dir_mkdir_cbk,
-				    layout->list[i].xlator,
-				    layout->list[i].xlator->fops->mkdir,
-				    loc, local->stbuf.st_mode);
-		}
-	}
+        if (dict)
+                dict_unref (dict);
 
-	return 0;
+        return 0;
 }
 
 
@@ -284,13 +549,10 @@ dht_selfheal_layout_alloc_start (xlator_t *this, loc_t *loc,
                                  dht_layout_t *layout)
 {
         int           start = 0;
-        dht_conf_t   *conf = NULL;
         uint32_t      hashval = 0;
         int           ret = 0;
 
-        conf = this->private;
-
-        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);
         }
@@ -298,237 +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)
 {
-	dht_conf_t  *conf = NULL;
-	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;
+        int i = 0;
+        int j = 0;
+        int err = 0;
+        int count = 0;
+        dht_conf_t *conf = NULL;
 
-	this = frame->this;
-	conf = this->private;
+        /* 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;
+        /* 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);
 
-	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;
-			}
-		}
+void dht_layout_entry_swap (dht_layout_t *layout, int i, int j);
+void dht_layout_range_swap (dht_layout_t *layout, int i, int j);
+
+/*
+ * It's a bit icky using local variables in a macro, but it makes the rest
+ * of the code a lot clearer.
+ */
+#define OV_ENTRY(x,y)      table[x*new->cnt+y]
+
+void
+dht_selfheal_layout_maximize_overlap (call_frame_t *frame, loc_t *loc,
+				      dht_layout_t *new, dht_layout_t *old)
+{
+	int           i            = 0;
+	int           j            = 0;
+	uint32_t      curr_overlap = 0;
+	uint32_t      max_overlap  = 0;
+	int           max_overlap_idx = -1;
+	uint32_t      overlap      = 0;
+        uint32_t     *table = NULL;
+
+	dht_layout_sort_volname (old);
+	/* Now both old_layout->list[] and new_layout->list[]
+	   are match the same xlators/subvolumes. i.e,
+	   old_layout->[i] and new_layout->[i] are referring
+	   to the same subvolumes
+	*/
+
+        /* Build a table of overlaps between new[i] and old[j]. */
+        table = alloca(sizeof(overlap)*old->cnt*new->cnt);
+        if (!table) {
+                return;
+        }
+        memset(table,0,sizeof(overlap)*old->cnt*new->cnt);
+        for (i = 0; i < new->cnt; ++i) {
+                for (j = 0; j < old->cnt; ++j) {
+                        OV_ENTRY(i,j) = dht_overlap_calc(old,j,new,i);
+                }
+        }
+
+	for (i = 0; i < new->cnt; i++) {
+		if (new->list[i].err > 0) {
+			/* Subvol might be marked for decommission
+			   with EINVAL, or some other serious error
+			   marked with positive errno.
+			*/
+			continue;
+                }
+
+                max_overlap = 0;
+                max_overlap_idx = i;
+                for (j = (i + 1); j < new->cnt; ++j) {
+                        if (new->list[j].err > 0) {
+			        /* Subvol might be marked for decommission
+			        with EINVAL, or some other serious error
+			        marked with positive errno.
+			        */
+			        continue;
+                        }
+                        /* Calculate the overlap now. */
+                        curr_overlap = OV_ENTRY(i,i) + OV_ENTRY(j,j);
+                        /* Calculate the overlap after the proposed swap. */
+                        overlap = OV_ENTRY(i,j) + OV_ENTRY(j,i);
+                        /* Are we better than status quo? */
+                        if (overlap > curr_overlap) {
+                                overlap -= curr_overlap;
+                                /* Are we better than the previous choice? */
+                                if (overlap > max_overlap) {
+                                        max_overlap = overlap;
+                                        max_overlap_idx = j;
+                                }
+                        }
+                }
+
+		if (max_overlap_idx != i) {
+ 			dht_layout_range_swap (new, i, max_overlap_idx);
+                        /* Need to swap the table values too. */
+                        for (j = 0; j < old->cnt; ++j) {
+                                overlap = OV_ENTRY(i,j);
+                                OV_ENTRY(i,j) = OV_ENTRY(max_overlap_idx,j);
+                                OV_ENTRY(max_overlap_idx,j) = overlap;
+                        }
+                }
 	}
 }
 
 
+dht_layout_t *
+dht_fix_layout_of_directory (call_frame_t *frame, loc_t *loc,
+                             dht_layout_t *layout)
+{
+        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 || missing || 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;
+        }
 
-	/* TODO: give a fix to these non-virgins */
+        for (i = 0; i < layout->cnt; i++) {
+                /* directory not present */
+                if (layout->list[i].err == ENOENT) {
+                        ret = 0;
+                        break;
+                }
+        }
+
+        /* TODO: give a fix to these non-virgins */
+
+        return ret;
+}
+
+int
+dht_selfheal_new_directory (call_frame_t *frame,
+                            dht_selfheal_dir_cbk_t dir_cbk,
+                            dht_layout_t *layout)
+{
+        dht_local_t *local = NULL;
 
-	return ret;
+        local = frame->local;
+
+        local->selfheal.dir_cbk = dir_cbk;
+        local->selfheal.layout = dht_layout_ref (frame->this, layout);
+
+        dht_layout_sort_volname (layout);
+        dht_selfheal_layout_new_directory (frame, &local->loc, layout);
+        dht_selfheal_dir_xattr (frame, &local->loc, layout);
+        return 0;
 }
 
 int
-dht_selfheal_new_directory (call_frame_t *frame, 
-			    dht_selfheal_dir_cbk_t dir_cbk,
-			    dht_layout_t *layout)
+dht_fix_directory_layout (call_frame_t *frame,
+                          dht_selfheal_dir_cbk_t dir_cbk,
+                          dht_layout_t *layout)
 {
-	dht_local_t *local = NULL;
+        dht_local_t  *local = NULL;
+        dht_layout_t *tmp_layout = NULL;
 
-	local = frame->local;
+        local = frame->local;
 
-	local->selfheal.dir_cbk = dir_cbk;
-	local->selfheal.layout = dht_layout_ref (frame->this, layout);
+        local->selfheal.dir_cbk = dir_cbk;
+        local->selfheal.layout = dht_layout_ref (frame->this, layout);
+
+        /* No layout sorting required here */
+        tmp_layout = dht_fix_layout_of_directory (frame, &local->loc, layout);
+        if (!tmp_layout) {
+                return -1;
+        }
+        dht_fix_dir_xattr (frame, &local->loc, tmp_layout);
 
-	dht_layout_sort_volname (layout);
-	dht_selfheal_layout_new_directory (frame, &local->loc, layout);	
-	dht_selfheal_dir_xattr (frame, &local->loc, layout);
-	return 0;
+        return 0;
 }
 
 
 int
 dht_selfheal_directory (call_frame_t *frame, dht_selfheal_dir_cbk_t dir_cbk,
-			loc_t *loc, dht_layout_t *layout)
+                        loc_t *loc, dht_layout_t *layout)
 {
-	dht_local_t *local    = NULL;
-	uint32_t     holes    = 0;
-	uint32_t     overlaps = 0;
-	uint32_t     missing  = 0;
-	uint32_t     down     = 0;
-	uint32_t     misc     = 0;
-	int          ret      = 0;
-	xlator_t    *this     = NULL;
-
-	local = frame->local;
-	this = frame->this;
-
-	ret = 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;
-	overlaps = local->selfheal.overlaps_cnt;
-	missing  = local->selfheal.missing;
-	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 = 0;
-		goto sorry_no_fix;
-	}
+        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;
+
+        dht_layout_anomalies (this, loc, layout,
+                              &local->selfheal.hole_cnt,
+                              &local->selfheal.overlaps_cnt,
+                              NULL, &local->selfheal.down,
+                              &local->selfheal.misc, NULL);
+
+        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_WARNING,
+                        "%d subvolumes down -- not fixing", down);
+                ret = 0;
+                goto sorry_no_fix;
+        }
 
-	if (misc) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"%d subvolumes have unrecoverable errors", misc);
-		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);
+        dht_layout_sort_volname (layout);
+        ret = dht_selfheal_dir_getafix (frame, loc, layout);
 
-	if (ret == -1) {
-		gf_log (this->name, GF_LOG_DEBUG,
-			"not able to form layout for the directory");
-		goto sorry_no_fix;
-	}
+        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;
+}
 
-	return 0;
+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;
+}
+
+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 d947ae814..fc0ca2f77 100644
--- a/xlators/cluster/dht/src/dht.c
+++ b/xlators/cluster/dht/src/dht.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -23,376 +14,65 @@
 #include "config.h"
 #endif
 
-/* TODO: add NS locking */
-
 #include "statedump.h"
-#include "dht-common.c"
-
-/* TODO:
-   - use volumename in xattr instead of "dht"
-   - use NS locks
-   - handle all cases in self heal layout reconstruction
-   - complete linkfile selfheal
-*/
-
-
-void
-dht_layout_dump (dht_layout_t  *layout, const char *prefix)
-{
-
-        char    key[GF_DUMP_MAX_BUF_LEN];
-        int     i = 0;
-
-        if (!layout)
-                return;
-
-        gf_proc_dump_build_key(key, prefix, "cnt");
-        gf_proc_dump_write(key, "%d", layout->cnt);
-        gf_proc_dump_build_key(key, prefix, "preset");
-        gf_proc_dump_write(key, "%d", layout->preset);
-        gf_proc_dump_build_key(key, prefix, "gen");
-        gf_proc_dump_write(key, "%d", layout->gen);
-        gf_proc_dump_build_key(key, prefix, "type");
-        gf_proc_dump_write(key, "%d", layout->type);
-
-        for (i = 0; i < layout->cnt; i++) {
-                gf_proc_dump_build_key(key, prefix,"list[%d].err", i);
-                gf_proc_dump_write(key, "%d", layout->list[i].err);
-                gf_proc_dump_build_key(key, prefix,"list[%d].start", i);
-                gf_proc_dump_write(key, "%u", layout->list[i].start);
-                gf_proc_dump_build_key(key, prefix,"list[%d].stop", i);
-                gf_proc_dump_write(key, "%u", layout->list[i].stop);
-                if (layout->list[i].xlator) {
-                        gf_proc_dump_build_key(key, prefix,
-                                               "list[%d].xlator.type", i);
-                        gf_proc_dump_write(key, "%s",
-                                           layout->list[i].xlator->type);
-                        gf_proc_dump_build_key(key, prefix,
-                                               "list[%d].xlator.name", i);
-                        gf_proc_dump_write(key, "%s",
-                                           layout->list[i].xlator->name);
-                }
-        }
-}
-
-
-int32_t
-dht_priv_dump (xlator_t *this)
-{
-	char            key_prefix[GF_DUMP_MAX_BUF_LEN];
-	char            key[GF_DUMP_MAX_BUF_LEN];
-        int             i = 0;
-        dht_conf_t      *conf = NULL;
-        int             ret = 0;
-
-        if (!this)
-                return -1;
-
-        conf = this->private;
-
-        if (!conf)
-                return -1;
-
-        ret = TRY_LOCK(&conf->subvolume_lock);
-
-        if (ret != 0) {
-                gf_log("", GF_LOG_WARNING, "Unable to lock dht subvolume %s",
-                                this->name);
-                return ret;
-        }
-
-        gf_proc_dump_add_section("xlator.cluster.dht.%s.priv", this->name);
-        gf_proc_dump_build_key(key_prefix,"xlator.cluster.dht","%s.priv",
-                               this->name);
-        gf_proc_dump_build_key(key, key_prefix, "subvolume_cnt");
-        gf_proc_dump_write(key,"%d", conf->subvolume_cnt);
-        for (i = 0; i < conf->subvolume_cnt; i++) {
-                gf_proc_dump_build_key(key, key_prefix, "subvolumes[%d]", i);
-                gf_proc_dump_write(key, "%s.%s", conf->subvolumes[i]->type,
-                                   conf->subvolumes[i]->name);
-                if (conf->file_layouts && conf->file_layouts[i]){
-                        gf_proc_dump_build_key(key, key_prefix,
-                                               "file_layouts[%d]",i);
-                        dht_layout_dump(conf->file_layouts[i], key);
-                }
-                if (conf->dir_layouts && conf->dir_layouts[i]) {
-                        gf_proc_dump_build_key(key, key_prefix,
-                                              "dir_layouts[%d]",i);
-                        dht_layout_dump(conf->dir_layouts[i], key);
-                }
-                if (conf->subvolume_status) {
-                        gf_proc_dump_build_key(key, key_prefix,
-                                                "subvolume_status[%d]", i);
-                        gf_proc_dump_write(key, "%d",
-                                               (int)conf->subvolume_status[i]);
-                }
-
-        }
-
-        gf_proc_dump_build_key(key, key_prefix,"default_dir_layout");
-        dht_layout_dump(conf->default_dir_layout, key);
-
-        gf_proc_dump_build_key(key, key_prefix, "search_unhashed");
-        gf_proc_dump_write(key, "%d", conf->search_unhashed);
-        gf_proc_dump_build_key(key, key_prefix, "gen");
-        gf_proc_dump_write(key, "%d", conf->gen);
-        gf_proc_dump_build_key(key, key_prefix, "min_free_disk");
-        gf_proc_dump_write(key, "%lu", conf->min_free_disk);
-        gf_proc_dump_build_key(key, key_prefix, "disk_unit");
-        gf_proc_dump_write(key, "%c", conf->disk_unit);
-        gf_proc_dump_build_key(key, key_prefix, "refresh_interval");
-        gf_proc_dump_write(key, "%d", conf->refresh_interval);
-        gf_proc_dump_build_key(key, key_prefix, "unhashed_sticky_bit");
-        gf_proc_dump_write(key, "%d", conf->unhashed_sticky_bit);
-        if (conf ->du_stats) {
-                gf_proc_dump_build_key(key, key_prefix,
-                                "du_stats.avail_percent");
-                gf_proc_dump_write(key, "%lf", conf->du_stats->avail_percent);
-                gf_proc_dump_build_key(key, key_prefix,
-                                "du_stats.avail_space");
-                gf_proc_dump_write(key, "%lu", conf->du_stats->avail_space);
-                gf_proc_dump_build_key(key, key_prefix,
-                                "du_stats.log");
-                gf_proc_dump_write(key, "%lu", conf->du_stats->log);
-        }
-        gf_proc_dump_build_key(key, key_prefix, "last_stat_fetch");
-        gf_proc_dump_write(key, "%s", ctime(&conf->last_stat_fetch.tv_sec));
-
-        UNLOCK(&conf->subvolume_lock);
-
-        return 0;
-}
-
-int32_t
-dht_inodectx_dump (xlator_t *this, inode_t *inode)
-{
-        int             ret = -1;
-        char            key_prefix[GF_DUMP_MAX_BUF_LEN];
-        dht_layout_t    *layout = NULL;
-	uint64_t        tmp_layout = 0;
-
-        if (!inode)
-                return -1;
-
-	ret = inode_ctx_get (inode, this, &tmp_layout);
-
-        if (ret != 0)
-                return ret;
-
-        layout = (dht_layout_t *)(long)tmp_layout;
-
-        if (!layout)
-                return -1;
-
-        gf_proc_dump_build_key(key_prefix, "xlator.cluster.dht",
-                               "%s.inode.%ld", this->name, inode->ino);
-        dht_layout_dump(layout, key_prefix);
-
-        return 0;
-}
-
-int
-notify (xlator_t *this, int event, void *data, ...)
-{
-	int ret = -1;
-
-	ret = dht_notify (this, event, data);
-
-	return ret;
-}
-
-void
-fini (xlator_t *this)
-{
-        int         i = 0;
-        dht_conf_t *conf = NULL;
-
-	conf = this->private;
-
-        if (conf) {
-                if (conf->file_layouts) {
-                        for (i = 0; i < conf->subvolume_cnt; i++) {
-                                FREE (conf->file_layouts[i]);
-                        }
-                        FREE (conf->file_layouts);
-                }
-
-                if (conf->default_dir_layout)
-                        FREE (conf->default_dir_layout);
-
-                if (conf->subvolumes)
-                        FREE (conf->subvolumes);
-
-		if (conf->subvolume_status)
-			FREE (conf->subvolume_status);
-
-                FREE (conf);
-        }
-
-	return;
-}
-
-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 = CALLOC (1, sizeof (*conf));
-        if (!conf) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
-        }
-
-	conf->search_unhashed = 0;
-
-	if (dict_get_str (this->options, "lookup-unhashed", &temp_str) == 0) {
-		gf_string2boolean (temp_str, &conf->search_unhashed);
-	}
-
-	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->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 = CALLOC (conf->subvolume_cnt, sizeof (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++) {
-                                FREE (conf->file_layouts[i]);
-                        }
-                        FREE (conf->file_layouts);
-                }
-
-                if (conf->default_dir_layout)
-                        FREE (conf->default_dir_layout);
-
-                if (conf->subvolumes)
-                        FREE (conf->subvolumes);
-
-		if (conf->subvolume_status)
-			FREE (conf->subvolume_status);
-
-                if (conf->du_stats)
-                        FREE (conf->du_stats);
-
-                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,
-#if 0
-	.setdents    = dht_setdents,
-	.getdents    = dht_getdents,
-	.checksum    = dht_checksum,
-#endif
-};
-
-
-struct xlator_mops mops = {
+	.fallocate   = dht_fallocate,
+	.discard     = dht_discard,
+        .zerofill    = dht_zerofill,
 };
 
 struct xlator_dumpops dumpops = {
@@ -402,21 +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"}, 
-	  .type = GF_OPTION_TYPE_BOOL 
-	},
-        { .key  = {"min-free-disk"},
-          .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
-        },
-        { .key = {"unhashed-sticky-bit"},
-          .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 5fe233dfe..e934acdf0 100644
--- a/xlators/cluster/dht/src/nufa.c
+++ b/xlators/cluster/dht/src/nufa.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -23,15 +14,17 @@
 #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,
-                       inode_t *inode, struct stat *stbuf, dict_t *xattr,
-                       struct stat *postparent)
+                       int op_ret, int op_errno,
+                       inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+                       struct iatt *postparent)
 {
         xlator_t     *subvol      = NULL;
         char          is_linkfile = 0;
@@ -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->st_ino,
-				&stbuf->st_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,
@@ -118,52 +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);
-		op_errno = EINVAL;
-		goto err;
-	}
+        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);
@@ -172,40 +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",
@@ -214,150 +187,155 @@ 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->st_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.");
+                        op_errno = -1;
+                        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.");
+                        op_errno = -1;
+                        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.");
+                        op_errno = -1;
+                        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,
-                                 inode_t *inode, struct stat *stbuf,
-                                 struct stat *preparent,
-                                 struct stat *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;
-	call_frame_t *prev = NULL;
-	dht_conf_t   *conf  = NULL;
+        dht_local_t  *local = NULL;
 
-	local = frame->local;
-	prev  = cookie;
-	conf  = this->private;
+        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);
+        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)
+             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;
         }
 
@@ -366,382 +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);
+                    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 stat *stbuf, struct stat *preparent,
-                         struct stat *postparent)
+                         struct iatt *stbuf, struct iatt *preparent,
+                         struct iatt *postparent, dict_t *xdata)
 {
-	dht_local_t  *local = NULL;
-	call_frame_t *prev = NULL;
-	dht_conf_t   *conf  = NULL;
-
-	local = frame->local;
-	prev  = cookie;
-	conf  = this->private;
-
-	if (op_ret >= 0) {
-		STACK_WIND (frame, dht_newfile_cbk,
-			    local->cached_subvol,
-			    local->cached_subvol->fops->mknod,
-			    &local->loc, local->mode, local->rdev);
-
-		return 0;
-	}
-
-	DHT_STACK_UNWIND (link, frame, op_ret, op_errno,
-                          inode, stbuf, preparent, postparent);
-	return 0;
+        dht_local_t  *local = NULL;
+
+        local = frame->local;
+        if (!local || !local->cached_subvol) {
+                op_errno = EINVAL;
+                op_ret = -1;
+                goto err;
+        }
+
+        if (op_ret >= 0) {
+                STACK_WIND_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, xdata);
+        return 0;
 }
 
 
 int
 nufa_mknod (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, mode_t mode, dev_t rdev)
+            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;
-		}
-
-		local->mode = mode;
-		local->rdev = rdev;
-		local->cached_subvol = avail_subvol;
-
-		dht_linkfile_create (frame, nufa_mknod_linkfile_cbk,
+        if (avail_subvol != subvol) {
+                /* Create linkfile first */
+
+                local->params = dict_ref (params);
+                local->mode = mode;
+                local->umask = umask;
+                local->rdev = rdev;
+                local->cached_subvol = avail_subvol;
+
+                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);
+        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;
-}
-
-void
-fini (xlator_t *this)
-{
-        int         i = 0;
-        dht_conf_t *conf = NULL;
-
-	conf = this->private;
-
-        if (conf) {
-                if (conf->file_layouts) {
-                        for (i = 0; i < conf->subvolume_cnt; i++) {
-                                FREE (conf->file_layouts[i]);
-                        }
-                        FREE (conf->file_layouts);
+        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;
+        }
+}
 
-                if (conf->default_dir_layout)
-                        FREE (conf->default_dir_layout);
-
-                if (conf->subvolumes)
-                        FREE (conf->subvolumes);
+typedef struct nufa_args {
+        xlator_t    *this;
+        char        *volname;
+        gf_boolean_t addr_match;
+} nufa_args_t;
 
-		if (conf->subvolume_status)
-			FREE (conf->subvolume_status);
+static void
+nufa_find_local_brick (xlator_t *xl, void *data)
+{
+        nufa_args_t     *args = data;
+        xlator_t        *this = args->this;
+        char            *local_volname = args->volname;
+        gf_boolean_t    addr_match = args->addr_match;
+        char            *brick_host = NULL;
+        dht_conf_t      *conf = this->private;
+        int             ret = -1;
+
+        /*This means a local subvol was already found. We pick the first brick
+         * that is local*/
+        if (conf->private)
+                return;
+
+        if (strcmp (xl->name, local_volname) == 0) {
+                conf->private = xl;
+                gf_log (this->name, GF_LOG_INFO, "Using specified subvol %s",
+                        local_volname);
+                return;
+        }
 
-                FREE (conf);
+        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;
         }
 
-	return;
 }
 
-int
-init (xlator_t *this)
+static void
+nufa_to_dht (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;
-        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 = CALLOC (1, sizeof (*conf));
-        if (!conf) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
-        }
-
-	conf->search_unhashed = 0;
+        GF_ASSERT (this);
+        GF_ASSERT (this->fops);
 
-	if (dict_get_str (this->options, "lookup-unhashed",
-			  &temp_str) == 0) {
-		gf_string2boolean (temp_str,
-				   &conf->search_unhashed);
-	}
-
-        ret = dht_init_subvolumes (this, conf);
-        if (ret == -1) {
-                goto err;
-        }
+        this->fops->lookup = dht_lookup;
+        this->fops->create = dht_create;
+        this->fops->mknod  = dht_mknod;
+}
 
-        ret = dht_layouts_init (this, conf);
-        if (ret == -1) {
-                goto err;
+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;
         }
 
-	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';
+        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;
                 }
-        }
 
-        conf->du_stats = CALLOC (conf->subvolume_cnt, sizeof (dht_du_t));
-        if (!conf->du_stats) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
+                candidate = parent;
+                trav = parent->parents;
         }
 
-        this->private = conf;
-
-        return 0;
+        return ret;
+}
 
-err:
-        if (conf) {
-                if (conf->file_layouts) {
-                        for (i = 0; i < conf->subvolume_cnt; i++) {
-                                FREE (conf->file_layouts[i]);
-                        }
-                        FREE (conf->file_layouts);
-                }
+int
+nufa_init (xlator_t *this)
+{
+        data_t        *data = NULL;
+        char          *local_volname = NULL;
+        int            ret = -1;
+        char           my_hostname[256];
+        gf_boolean_t   addr_match = _gf_false;
+        nufa_args_t    args = {0, };
 
-                if (conf->default_dir_layout)
-                        FREE (conf->default_dir_layout);
+        ret = dht_init(this);
+        if (ret) {
+                return ret;
+        }
 
-                if (conf->subvolumes)
-                        FREE (conf->subvolumes);
+        if ((data = dict_get (this->options, "local-volume-name"))) {
+                local_volname = data->data;
 
-		if (conf->subvolume_status)
-			FREE (conf->subvolume_status);
+        } else {
+                addr_match = _gf_true;
+                local_volname = "localhost";
+                ret = gethostname (my_hostname, 256);
+                if (ret == 0)
+                        local_volname = my_hostname;
 
-                if (conf->du_stats)
-                        FREE (conf->du_stats);
+                else
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "could not find hostname (%s)",
+                                strerror (errno));
 
-                FREE (conf);
         }
 
-        return -1;
+        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);
+        }
+        return 0;
 }
 
 
-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,
-        .setattr     = dht_setattr,
-#if 0
-	.setdents    = dht_setdents,
-	.getdents    = dht_getdents,
-	.checksum    = dht_checksum,
-#endif
+class_methods_t class_methods = {
+        .init           = nufa_init,
+        .fini           = dht_fini,
+        .reconfigure    = dht_reconfigure,
+        .notify         = dht_notify
 };
 
 
-struct xlator_mops mops = {
+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,
+        .setattr     = dht_setattr,
 };
 
 
 struct xlator_cbks cbks = {
-	.forget     = dht_forget
-};
-
-
-struct volume_options options[] = {
-	{ .key  = {"local-volume-name"},
-	  .type = GF_OPTION_TYPE_XLATOR
-	},
-        { .key  = {"lookup-unhashed"},
-	  .type = GF_OPTION_TYPE_BOOL
-	},
-        { .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 6670f0e3b..2717ce975 100644
--- a/xlators/cluster/dht/src/switch.c
+++ b/xlators/cluster/dht/src/switch.c
@@ -1,20 +1,11 @@
 /*
-   Copyright (c) 2009 Gluster, Inc. <http://www.gluster.com>
-   This file is part of GlusterFS.
-
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -23,28 +14,31 @@
 #include "config.h"
 #endif
 
-#include "dht-common.c"
+#include "dht-common.h"
+#include "dht-mem-types.h"
 
 #include <sys/time.h>
 #include <stdlib.h>
 #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 */
@@ -73,39 +67,47 @@ 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;
+
+        pathname = gf_strdup (path);
+        if (!pathname)
+                goto out;
 
         trav = cond;
-        pathname = strdup (path);
-	while (trav) {
-		if (fnmatch (trav->path_pattern,
-			     pathname, FNM_NOESCAPE) == 0) {
+        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;
-	}
-	free (pathname);
-	return hashed_subvol;
+                        subvol = trav->array[idx].xl;
+                        goto out;
+                }
+                trav = trav->next;
+        }
+out:
+        GF_FREE (pathname);
+
+        return subvol;
 }
 
 
 int
 switch_local_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                          int op_ret, int op_errno,
-                         inode_t *inode, struct stat *stbuf, dict_t *xattr,
-                         struct stat *postparent)
+                         inode_t *inode, struct iatt *stbuf, dict_t *xattr,
+                         struct iatt *postparent)
 {
         xlator_t     *subvol      = NULL;
         char          is_linkfile = 0;
@@ -115,7 +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;
@@ -124,57 +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->st_ino,
-				&stbuf->st_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,
@@ -191,33 +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);
-		op_errno = EINVAL;
-		goto err;
-	}
+        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;
@@ -225,18 +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);
@@ -245,40 +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;
-
-        local = dht_local_init (frame);
-	if (!local) {
-		op_errno = ENOMEM;
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		goto err;
-	}
+        conf = this->private;
 
-        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",
@@ -287,71 +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->st_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);
-
-		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);
+                 */
+                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 %s",
+                                conf->xattr_name);
+
+                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);
 
-			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,
-				       "trusted.glusterfs.dht.linkto", 256);
+                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 %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 %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) {
@@ -370,101 +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 stat *stbuf,
-                                 struct stat *preparent,
-                                 struct stat *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;
-	call_frame_t *prev = NULL;
-	dht_conf_t   *conf  = NULL;
+        dht_local_t  *local = NULL;
 
-	local = frame->local;
-	prev  = cookie;
-	conf  = this->private;
+        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);
+        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)
+               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 = 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;
         }
 
@@ -473,178 +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);
+                    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 stat *stbuf, struct stat *preparent,
-                         struct stat *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;
-	call_frame_t *prev = NULL;
-	dht_conf_t   *conf  = NULL;
-
-	local = frame->local;
-	prev  = cookie;
-	conf  = this->private;
-
-	if (op_ret >= 0) {
-		STACK_WIND (frame, dht_newfile_cbk,
-			    local->cached_subvol,
-			    local->cached_subvol->fops->mknod,
-			    &local->loc, local->mode, local->rdev);
-
-		return 0;
-	}
-
-	DHT_STACK_UNWIND (link, frame, op_ret, op_errno,
-                          inode, stbuf, preparent, postparent);
-	return 0;
+        dht_local_t  *local = NULL;
+
+        local = frame->local;
+        if (!local || !local->cached_subvol) {
+                op_errno = EINVAL;
+                op_ret = -1;
+                goto err;
+        }
+
+        if (op_ret >= 0) {
+                STACK_WIND_COOKIE (frame, dht_newfile_cbk,
+                            (void *)local->cached_subvol, local->cached_subvol,
+                            local->cached_subvol->fops->mknod,
+                            &local->loc, local->mode, local->rdev,
+                            local->umask, local->params);
+
+                return 0;
+        }
+err:
+        DHT_STACK_UNWIND (link, frame, op_ret, op_errno,
+                          inode, stbuf, preparent, postparent, xdata);
+        return 0;
 }
 
 
 int
-switch_mknod (call_frame_t *frame, xlator_t *this,
-	    loc_t *loc, mode_t mode, dev_t rdev)
+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->mode = mode;
-		local->rdev = rdev;
-		local->cached_subvol = avail_subvol;
+                local->params = dict_ref (params);
+                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);
+        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)
-                                FREE (trav->array);
+                        GF_FREE (trav->array);
                         prev = trav;
                         trav = trav->next;
-                        FREE (prev);
+                        GF_FREE (prev);
                 }
-
-                if (conf->file_layouts) {
-                        for (i = 0; i < conf->subvolume_cnt; i++) {
-                                FREE (conf->file_layouts[i]);
-                        }
-                        FREE (conf->file_layouts);
-                }
-
-                if (conf->default_dir_layout)
-                        FREE (conf->default_dir_layout);
-
-                if (conf->subvolumes)
-                        FREE (conf->subvolumes);
-
-		if (conf->subvolume_status)
-			FREE (conf->subvolume_status);
-
-                FREE (conf);
         }
 
-	return;
+        dht_fini(this);
 }
 
 int
@@ -669,56 +618,62 @@ 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 = CALLOC ((index + 1),
-                                   sizeof (struct switch_sched_array));
+        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 */
-        option_string = strdup (pattern_str);
+        option_string = gf_strdup (pattern_str);
         switch_str = strtok_r (option_string, ";", &tmp_str);
         while (switch_str) {
-                dup_str = strdup (switch_str);
-                switch_opt = CALLOC (1, sizeof (struct switch_struct));
-                if (!switch_opt)
+                dup_str = gf_strdup (switch_str);
+                switch_opt = GF_CALLOC (1, sizeof (struct switch_struct),
+                                        gf_switch_mt_switch_struct);
+                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);
-                        free (dup_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 = strdup (childs);
+                        dup_childs = gf_strdup (childs);
                         child = strtok_r (dup_childs, ",", &tmp);
                         while (child) {
                                 if (gf_switch_valid_child (this, child)) {
@@ -733,11 +688,12 @@ set_switch_pattern (xlator_t *this, dht_conf_t *conf,
                                         goto err;
                                 }
                         }
-                        free (dup_childs);
+                        GF_FREE (dup_childs);
                         child = strtok_r (childs, ",", &tmp1);
                         switch_opt->num_child = idx;
-                        switch_opt->array = CALLOC (1, (idx *
-                                                        sizeof (struct switch_sched_array)));
+                        switch_opt->array = GF_CALLOC (1, (idx *
+                                                           sizeof (struct switch_sched_array)),
+                                                       gf_switch_mt_switch_sched_array);
                         if (!switch_opt->array)
                                 goto err;
                         idx = 0;
@@ -771,8 +727,7 @@ set_switch_pattern (xlator_t *this, dht_conf_t *conf,
                                 "option in unify volume. Exiting");
                         goto err;
                 }
-                free (dup_str);
-                
+
                 /* Link it to the main structure */
                 if (switch_buf) {
                         /* there are already few entries */
@@ -784,269 +739,166 @@ 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 = 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 =
-			CALLOC (1, flag * sizeof (struct switch_sched_array));
+                /* 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)
-                        FREE (switch_buf_array);
                 trav = switch_buf;
                 while (trav) {
-                        if (trav->array)
-                                FREE (trav->array);
+                        GF_FREE (trav->array);
                         switch_opt = trav;
                         trav = trav->next;
-                        FREE (switch_opt);
+                        GF_FREE (switch_opt);
                 }
         }
         return -1;
 }
 
 
-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 = CALLOC (1, sizeof (*conf));
-        if (!conf) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory");
-                goto err;
-        }
 
-	conf->search_unhashed = 1;
-	if (dict_get_str (this->options, "lookup-unhashed", &temp_str) == 0) {
-		gf_string2boolean (temp_str, &conf->search_unhashed);
-	}
-
-	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 = CALLOC (conf->subvolume_cnt, sizeof (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++) {
-                                FREE (conf->file_layouts[i]);
-                        }
-                        FREE (conf->file_layouts);
-                }
-
-                if (conf->default_dir_layout)
-                        FREE (conf->default_dir_layout);
-
-                if (conf->subvolumes)
-                        FREE (conf->subvolumes);
-
-		if (conf->subvolume_status)
-			FREE (conf->subvolume_status);
-
-                if (conf->du_stats)
-                        FREE (conf->du_stats);
-
-                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,
-#if 0
-	.setdents    = dht_setdents,
-	.getdents    = dht_getdents,
-	.checksum    = dht_checksum,
-#endif
+class_methods_t class_methods = {
+        .init           = switch_init,
+        .fini           = switch_fini,
+        .reconfigure    = dht_reconfigure,
+        .notify         = dht_notify
 };
 
 
-struct xlator_mops mops = {
+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 xlator_cbks cbks = {
-	.forget     = dht_forget
-};
-
-
-struct volume_options options[] = {
-	{ .key  = {"pattern.switch.case"},
-	  .type = GF_OPTION_TYPE_ANY
-	},
-        { .key  = {"lookup-unhashed"},
-	  .type = GF_OPTION_TYPE_BOOL
-	},
-        { .key  = {"min-free-disk"},
-          .type = GF_OPTION_TYPE_PERCENT_OR_SIZET,
-        },
-	{ .key  = {NULL} },
+        .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 c23c5676c..19be1ed27 100644
--- a/xlators/cluster/ha/src/ha-helpers.c
+++ b/xlators/cluster/ha/src/ha-helpers.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #include "xlator.h"
 #include "call-stub.h"
 #include "defaults.h"
@@ -49,12 +39,14 @@ int ha_alloc_init_fd (call_frame_t *frame, fd_t *fd)
 			goto out;
 		}
 		hafdp = (hafd_t *)(long)tmp_hafdp;
-		local = frame->local = CALLOC (1, sizeof (*local));
+                local = frame->local = GF_CALLOC (1, sizeof (*local), 
+                                                  gf_ha_mt_ha_local_t);
 		if (local == NULL) {
 			ret = -ENOMEM;
 			goto out;
 		}
-		local->state = CALLOC (1, child_count);
+                local->state = GF_CALLOC (1, child_count, 
+                                          gf_ha_mt_child_count);
 		if (local->state == NULL) {
 			ret = -ENOMEM;
 			goto out;
@@ -147,7 +139,7 @@ int ha_handle_cbk (call_frame_t *frame, void *cookie, int op_ret, int op_errno)
         }
 
 	if (local->fd) {
-		FREE (local->state);
+		GF_FREE (local->state);
                 local->state = NULL;
 
 		fd_unref (local->fd);
@@ -170,7 +162,8 @@ int ha_alloc_init_inode (call_frame_t *frame, inode_t *inode)
 	local = frame->local;
 
 	if (local == NULL) {
-		local = frame->local = CALLOC (1, sizeof (*local));
+                local = frame->local = GF_CALLOC (1, sizeof (*local), 
+                                                  gf_ha_mt_ha_local_t);
 		if (local == NULL) {
 			ret = -ENOMEM;
 			goto out;
diff --git a/xlators/cluster/ha/src/ha-mem-types.h b/xlators/cluster/ha/src/ha-mem-types.h
new file mode 100644
index 000000000..e5e97d237
--- /dev/null
+++ b/xlators/cluster/ha/src/ha-mem-types.h
@@ -0,0 +1,26 @@
+/*
+   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_MEM_TYPES_H__
+#define __HA_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_ha_mem_types_ {
+        gf_ha_mt_ha_local_t = gf_common_mt_end + 1,
+        gf_ha_mt_hafd_t,
+        gf_ha_mt_char,
+        gf_ha_mt_child_count,
+        gf_ha_mt_xlator_t,
+        gf_ha_mt_ha_private_t,
+        gf_ha_mt_end
+};
+#endif
+
diff --git a/xlators/cluster/ha/src/ha.c b/xlators/cluster/ha/src/ha.c
index 442de21df..3eccb516b 100644
--- a/xlators/cluster/ha/src/ha.c
+++ b/xlators/cluster/ha/src/ha.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 /* generate errors randomly, code is simple now, better alogorithm
  * can be written to decide what error to be returned and when
  */
@@ -50,7 +40,7 @@ ha_local_wipe (ha_local_t *local)
         }
 
         if (local->state) {
-                FREE (local->state);
+                GF_FREE (local->state);
                 local->state = NULL;
         }
 
@@ -71,7 +61,7 @@ ha_local_wipe (ha_local_t *local)
                 local->inode = NULL;
         }
 
-        FREE (local);
+        GF_FREE (local);
         return;
 }
 
@@ -84,7 +74,7 @@ ha_forget (xlator_t *this,
 	char *state = NULL;
 	if (!inode_ctx_del (inode, this, &stateino)) {
 		state =  ((char *)(long)stateino);
-		FREE (state);
+		GF_FREE (state);
 	}
 
 	return 0;
@@ -98,9 +88,9 @@ ha_lookup_cbk (call_frame_t *frame,
 	       int32_t op_ret,
 	       int32_t op_errno,
 	       inode_t *inode,
-	       struct stat *buf,
+	       struct iatt *buf,
                dict_t *dict,
-               struct stat *postparent)
+               struct iatt *postparent)
 {
 	ha_local_t *local = NULL;
 	ha_private_t *pvt = NULL;
@@ -193,7 +183,8 @@ ha_lookup (call_frame_t *frame,
 	child_count = pvt->child_count;
 	children = pvt->children;
 
-	frame->local = local = CALLOC (1, sizeof (*local));
+	frame->local = local = GF_CALLOC (1, sizeof (*local), 
+                                          gf_ha_mt_ha_local_t);
         if (!local) {
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
                 op_errno = ENOMEM;
@@ -205,7 +196,7 @@ ha_lookup (call_frame_t *frame,
 
 	ret = inode_ctx_get (loc->inode, this, NULL);
 	if (ret) {
-		state = CALLOC (1, child_count);
+		state = GF_CALLOC (1, child_count, gf_ha_mt_child_count);
                 if (state == NULL) {
                         gf_log (this->name, GF_LOG_ERROR, "out of memory");
                         op_errno = ENOMEM;
@@ -245,7 +236,7 @@ ha_stat_cbk (call_frame_t *frame,
 	     xlator_t *this,
 	     int32_t op_ret,
 	     int32_t op_errno,
-	     struct stat *buf)
+	     struct iatt *buf)
 {
 	int ret = -1;
 
@@ -290,8 +281,8 @@ err:
 
 int32_t
 ha_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                int32_t op_ret, int32_t op_errno, struct stat *statpre,
-                struct stat *statpost)
+                int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+                struct iatt *statpost)
 {
 	int ret = -1;
 
@@ -305,7 +296,7 @@ ha_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 
 int32_t
-ha_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc, struct stat *stbuf,
+ha_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *stbuf,
             int32_t valid)
 {
 	ha_local_t *local = NULL;
@@ -333,7 +324,7 @@ err:
 
 
 int32_t
-ha_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd, struct stat *stbuf,
+ha_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *stbuf,
              int32_t valid)
 {
 	ha_local_t *local = NULL;
@@ -366,8 +357,8 @@ ha_truncate_cbk (call_frame_t *frame,
 		 xlator_t *this,
 		 int32_t op_ret,
 		 int32_t op_errno,
-		 struct stat *prebuf,
-                 struct stat *postbuf)
+		 struct iatt *prebuf,
+                 struct iatt *postbuf)
 {
 	int ret = -1;
 
@@ -423,8 +414,8 @@ ha_ftruncate_cbk (call_frame_t *frame,
 		  xlator_t *this,
 		  int32_t op_ret,
 		  int32_t op_errno,
-		  struct stat *prebuf,
-                  struct stat *postbuf)
+		  struct iatt *prebuf,
+                  struct iatt *postbuf)
 {
 	int ret = -1;
 
@@ -546,7 +537,7 @@ ha_readlink_cbk (call_frame_t *frame,
 		 int32_t op_ret,
 		 int32_t op_errno,
 		 const char *path,
-                 struct stat *sbuf)
+                 struct iatt *sbuf)
 {
 	int ret = -1;
 
@@ -599,9 +590,9 @@ ha_mknod_lookup_cbk (call_frame_t *frame,
 		     int32_t op_ret,
 		     int32_t op_errno,
 		     inode_t *inode,
-		     struct stat *buf,
+		     struct iatt *buf,
                      dict_t *dict,
-                     struct stat *postparent)
+                     struct iatt *postparent)
 {
 	ha_local_t *local = NULL;
 	ha_private_t *pvt = NULL;
@@ -645,7 +636,7 @@ ha_mknod_lookup_cbk (call_frame_t *frame,
 
 	if (cnt == 0) {
 		call_stub_t *stub = local->stub;
-		FREE (local->state);
+		GF_FREE (local->state);
 		STACK_UNWIND (frame,
 			      local->op_ret,
 			      local->op_errno,
@@ -664,9 +655,9 @@ ha_mknod_cbk (call_frame_t *frame,
 	      int32_t op_ret,
 	      int32_t op_errno,
 	      inode_t *inode,
-              struct stat *buf,
-              struct stat *preparent,
-              struct stat *postparent)
+              struct iatt *buf,
+              struct iatt *preparent,
+              struct iatt *postparent)
 {
 	ha_local_t *local = NULL;
 	ha_private_t *pvt = NULL;
@@ -715,7 +706,7 @@ ha_mknod_cbk (call_frame_t *frame,
 
 	if (cnt == 0 || i == child_count) {
 		call_stub_t *stub = local->stub;
-		FREE (local->state);
+		GF_FREE (local->state);
 		stub = local->stub;
 		STACK_UNWIND (frame, local->op_ret, local->op_errno,
                               local->stub->args.mknod.loc.inode, &local->buf,
@@ -770,7 +761,8 @@ ha_mknod (call_frame_t *frame,
 	pvt = this->private;
 	child_count = pvt->child_count;
 
-	frame->local = local = CALLOC (1, sizeof (*local));
+	frame->local = local = GF_CALLOC (1, sizeof (*local), 
+                                          gf_ha_mt_ha_local_t);
         if (!local) {
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
                 op_errno = ENOMEM;
@@ -786,7 +778,7 @@ ha_mknod (call_frame_t *frame,
 
 	local->op_ret = -1;
 	local->op_errno = ENOTCONN;
-	local->state = CALLOC (1, child_count);
+	local->state = GF_CALLOC (1, child_count, gf_ha_mt_char);
         if (!local->state) {
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
                 op_errno = ENOMEM;
@@ -796,7 +788,7 @@ ha_mknod (call_frame_t *frame,
 	memcpy (local->state, pvt->state, child_count);
 	local->active = -1;
 
-	stateino = CALLOC (1, child_count);
+	stateino = GF_CALLOC (1, child_count, gf_ha_mt_char);
         if (!stateino) {
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
                 op_errno = ENOMEM;
@@ -837,9 +829,9 @@ ha_mkdir_lookup_cbk (call_frame_t *frame,
 		     int32_t op_ret,
 		     int32_t op_errno,
 		     inode_t *inode,
-		     struct stat *buf,
+		     struct iatt *buf,
                      dict_t *dict,
-                     struct stat *postparent)
+                     struct iatt *postparent)
 {
 	ha_local_t *local = NULL;
 	ha_private_t *pvt = NULL;
@@ -875,7 +867,7 @@ ha_mkdir_lookup_cbk (call_frame_t *frame,
 
 	if (cnt == 0) {
 		call_stub_t *stub = local->stub;
-		FREE (local->state);
+		GF_FREE (local->state);
 		STACK_UNWIND (frame,
 			      local->op_ret,
 			      local->op_errno,
@@ -893,9 +885,9 @@ ha_mkdir_cbk (call_frame_t *frame,
 	      int32_t op_ret,
 	      int32_t op_errno,
 	      inode_t *inode,
-              struct stat *buf,
-              struct stat *preparent,
-              struct stat *postparent)
+              struct iatt *buf,
+              struct iatt *preparent,
+              struct iatt *postparent)
 {
 	ha_local_t *local = NULL;
 	ha_private_t *pvt = NULL;
@@ -940,7 +932,7 @@ ha_mkdir_cbk (call_frame_t *frame,
 
 	if (cnt == 0 || i == child_count) {
 		call_stub_t *stub = local->stub;
-		FREE (local->state);
+		GF_FREE (local->state);
 		stub = local->stub;
 		STACK_UNWIND (frame, local->op_ret, local->op_errno,
                               local->stub->args.mkdir.loc.inode, &local->buf,
@@ -993,7 +985,8 @@ ha_mkdir (call_frame_t *frame,
 	pvt = this->private;
 	child_count = pvt->child_count;
 
-	frame->local = local = CALLOC (1, sizeof (*local));
+	frame->local = local = GF_CALLOC (1, sizeof (*local), 
+                                          gf_ha_mt_ha_local_t);
         if (!frame->local) {
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
                 op_errno = ENOMEM;
@@ -1009,7 +1002,7 @@ ha_mkdir (call_frame_t *frame,
 
 	local->op_ret = -1;
 	local->op_errno = ENOTCONN;
-	local->state = CALLOC (1, child_count);
+	local->state = GF_CALLOC (1, child_count, gf_ha_mt_char);
         if (!local->state) {
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
                 op_errno = ENOMEM;
@@ -1019,7 +1012,7 @@ ha_mkdir (call_frame_t *frame,
 	memcpy (local->state, pvt->state, child_count);
 	local->active = -1;
 
-	stateino = CALLOC (1, child_count);
+	stateino = GF_CALLOC (1, child_count, gf_ha_mt_char);
         if (!stateino) {
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
                 op_errno = ENOMEM;
@@ -1056,8 +1049,8 @@ ha_unlink_cbk (call_frame_t *frame,
 	       xlator_t *this,
 	       int32_t op_ret,
 	       int32_t op_errno,
-               struct stat *preparent,
-               struct stat *postparent)
+               struct iatt *preparent,
+               struct iatt *postparent)
 {
 	int ret = -1;
 
@@ -1108,8 +1101,8 @@ ha_rmdir_cbk (call_frame_t *frame,
 	      xlator_t *this,
 	      int32_t op_ret,
 	      int32_t op_errno,
-              struct stat *preparent,
-              struct stat *postparent)
+              struct iatt *preparent,
+              struct iatt *postparent)
 {
 	int ret = -1;
 
@@ -1166,9 +1159,9 @@ ha_symlink_lookup_cbk (call_frame_t *frame,
 		       int32_t op_ret,
 		       int32_t op_errno,
 		       inode_t *inode,
-		       struct stat *buf,
+		       struct iatt *buf,
                        dict_t *dict,
-                       struct stat *postparent)
+                       struct iatt *postparent)
 {
 	ha_local_t *local = NULL;
 	ha_private_t *pvt = NULL;
@@ -1204,7 +1197,7 @@ ha_symlink_lookup_cbk (call_frame_t *frame,
 
 	if (cnt == 0) {
 		call_stub_t *stub = local->stub;
-		FREE (local->state);
+		GF_FREE (local->state);
 		STACK_UNWIND (frame,
 			      local->op_ret,
 			      local->op_errno,
@@ -1222,9 +1215,9 @@ ha_symlink_cbk (call_frame_t *frame,
 		int32_t op_ret,
 		int32_t op_errno,
 		inode_t *inode,
-                struct stat *buf,
-                struct stat *preparent,
-                struct stat *postparent)
+                struct iatt *buf,
+                struct iatt *preparent,
+                struct iatt *postparent)
 {
 	ha_local_t *local = NULL;
 	ha_private_t *pvt = NULL;
@@ -1268,7 +1261,7 @@ ha_symlink_cbk (call_frame_t *frame,
 
 	if (cnt == 0 || i == child_count) {
 		call_stub_t *stub = local->stub;
-		FREE (local->state);
+		GF_FREE (local->state);
 		stub = local->stub;
 		STACK_UNWIND (frame, local->op_ret, local->op_errno, 
 			      local->stub->args.symlink.loc.inode, &local->buf,
@@ -1321,7 +1314,8 @@ ha_symlink (call_frame_t *frame,
 	pvt = this->private;
 	child_count = pvt->child_count;
 
-	frame->local = local = CALLOC (1, sizeof (*local));
+	frame->local = local = GF_CALLOC (1, sizeof (*local), 
+                                          gf_ha_mt_ha_local_t);
         if (!local) {
                 op_errno = ENOMEM;
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
@@ -1337,7 +1331,7 @@ ha_symlink (call_frame_t *frame,
 
 	local->op_ret = -1;
 	local->op_errno = ENOTCONN;
-	local->state = CALLOC (1, child_count);
+	local->state = GF_CALLOC (1, child_count, gf_ha_mt_char);
         if (!local->state) {
                 op_errno = ENOMEM;
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
@@ -1347,7 +1341,7 @@ ha_symlink (call_frame_t *frame,
 	memcpy (local->state, pvt->state, child_count);
 	local->active = -1;
 
-	stateino = CALLOC (1, child_count);
+	stateino = GF_CALLOC (1, child_count, gf_ha_mt_char);
         if (!stateino) {
                 op_errno = ENOMEM;
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
@@ -1385,11 +1379,11 @@ ha_rename_cbk (call_frame_t *frame,
 	       xlator_t *this,
 	       int32_t op_ret,
 	       int32_t op_errno,
-	       struct stat *buf,
-               struct stat *preoldparent,
-               struct stat *postoldparent,
-               struct stat *prenewparent,
-               struct stat *postnewparent)
+	       struct iatt *buf,
+               struct iatt *preoldparent,
+               struct iatt *postoldparent,
+               struct iatt *prenewparent,
+               struct iatt *postnewparent)
 {
 	int ret = -1;
 
@@ -1443,9 +1437,9 @@ ha_link_lookup_cbk (call_frame_t *frame,
 		    int32_t op_ret,
 		    int32_t op_errno,
 		    inode_t *inode,
-		    struct stat *buf,
+		    struct iatt *buf,
                     dict_t *dict,
-                    struct stat *postparent)
+                    struct iatt *postparent)
 {
 	ha_local_t *local = NULL;
 	ha_private_t *pvt = NULL;
@@ -1481,7 +1475,7 @@ ha_link_lookup_cbk (call_frame_t *frame,
 
 	if (cnt == 0) {
 		call_stub_t *stub = local->stub;
-		FREE (local->state);
+		GF_FREE (local->state);
 		STACK_UNWIND (frame,
 			      local->op_ret,
 			      local->op_errno,
@@ -1499,9 +1493,9 @@ ha_link_cbk (call_frame_t *frame,
 	     int32_t op_ret,
 	     int32_t op_errno,
 	     inode_t *inode,
-             struct stat *buf,
-             struct stat *preparent,
-             struct stat *postparent)
+             struct iatt *buf,
+             struct iatt *preparent,
+             struct iatt *postparent)
 {
 	ha_local_t *local = NULL;
 	ha_private_t *pvt = NULL;
@@ -1545,7 +1539,7 @@ ha_link_cbk (call_frame_t *frame,
 
 	if (cnt == 0 || i == child_count) {
 		call_stub_t *stub = local->stub;
-		FREE (local->state);
+		GF_FREE (local->state);
 		stub = local->stub;
 		STACK_UNWIND (frame, local->op_ret, local->op_errno,
                               local->stub->args.link.oldloc.inode, &local->buf,
@@ -1613,7 +1607,8 @@ ha_link (call_frame_t *frame,
 	pvt = this->private;
 	child_count = pvt->child_count;
 
-	frame->local = local = CALLOC (1, sizeof (*local));
+	frame->local = local = GF_CALLOC (1, sizeof (*local), 
+                                          gf_ha_mt_ha_local_t);
         if (!frame->local) {
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
                 op_errno = ENOMEM;
@@ -1629,7 +1624,7 @@ ha_link (call_frame_t *frame,
 
 	local->op_ret = -1;
 	local->op_errno = ENOTCONN;
-	local->state = CALLOC (1, child_count);
+	local->state = GF_CALLOC (1, child_count, gf_ha_mt_char);
         if (!local->state) {
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
                 op_errno = ENOMEM;
@@ -1669,9 +1664,9 @@ ha_create_cbk (call_frame_t *frame,
 	       int32_t op_errno,
 	       fd_t *fd,
 	       inode_t *inode,
-	       struct stat *buf,
-               struct stat *preparent,
-               struct stat *postparent)
+	       struct iatt *buf,
+               struct iatt *preparent,
+               struct iatt *postparent)
 {
 	ha_local_t *local = NULL;
 	ha_private_t *pvt = NULL;
@@ -1741,7 +1736,7 @@ ha_create_cbk (call_frame_t *frame,
 			      stub->args.create.fd,
 			      stub->args.create.loc.inode, &local->buf,
                               &local->preparent, &local->postparent);
-		FREE (state);
+		GF_FREE (state);
 		call_stub_destroy (stub);
 		return 0;
 	}
@@ -1785,7 +1780,8 @@ ha_create (call_frame_t *frame,
 	children = pvt->children;
 
 	if (local == NULL) {
-		local = frame->local = CALLOC (1, sizeof (*local));
+                frame->local = local = GF_CALLOC (1, sizeof (*local), 
+                                                  gf_ha_mt_ha_local_t);
                 if (!local) {
                         op_errno = ENOMEM;
                         gf_log (this->name, GF_LOG_ERROR, "out of memory");
@@ -1799,7 +1795,7 @@ ha_create (call_frame_t *frame,
                         goto err;
                 }
 
-		local->state = CALLOC (1, child_count);
+                local->state = GF_CALLOC (1, child_count, gf_ha_mt_char);
                 if (!local->state) {
                         op_errno = ENOMEM;
                         gf_log (this->name, GF_LOG_ERROR, "out of memory");
@@ -1819,28 +1815,28 @@ ha_create (call_frame_t *frame,
 			}
 		}
 		/* FIXME handle active -1 */
-		stateino = CALLOC (1, child_count);
+		stateino = GF_CALLOC (1, child_count, gf_ha_mt_char);
                 if (!stateino) {
                         op_errno = ENOMEM;
                         gf_log (this->name, GF_LOG_ERROR, "out of memory");
                         goto err;
                 }
 
-		hafdp = CALLOC (1, sizeof (*hafdp));
+		hafdp = GF_CALLOC (1, sizeof (*hafdp), gf_ha_mt_hafd_t);
                 if (!hafdp) {
                         op_errno = ENOMEM;
                         gf_log (this->name, GF_LOG_ERROR, "out of memory");
                         goto err;
                 }
 
-		hafdp->fdstate = CALLOC (1, child_count);
+		hafdp->fdstate = GF_CALLOC (1, child_count, gf_ha_mt_char);
                 if (!hafdp->fdstate) {
                         op_errno = ENOMEM;
                         gf_log (this->name, GF_LOG_ERROR, "out of memory");
                         goto err;
                 }
 
-		hafdp->path = strdup(loc->path);
+		hafdp->path = gf_strdup(loc->path);
                 if (!hafdp->path) {
                         op_errno = ENOMEM;
                         gf_log (this->name, GF_LOG_ERROR, "out of memory");
@@ -1865,20 +1861,16 @@ err:
         ha_local_wipe (local);
 
         if (stateino) {
-                FREE (stateino);
+                GF_FREE (stateino);
                 stateino = NULL;
         }
 
         if (hafdp) {
-                if (hafdp->fdstate) {
-                        FREE (hafdp->fdstate);
-                }
+                GF_FREE (hafdp->fdstate);
 
-                if (hafdp->path) {
-                        FREE (hafdp->path);
-                }
+                GF_FREE (hafdp->path);
 
-                FREE (hafdp);
+                GF_FREE (hafdp);
         }
 
         return 0;
@@ -1955,7 +1947,8 @@ ha_open (call_frame_t *frame,
 	child_count = pvt->child_count;
 
 
-	local = frame->local = CALLOC (1, sizeof (*local));
+	frame->local = local = GF_CALLOC (1, sizeof (*local), 
+                                          gf_ha_mt_ha_local_t);
         if (!local) {
                 op_errno = ENOMEM;
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
@@ -1966,21 +1959,21 @@ ha_open (call_frame_t *frame,
 	local->op_errno = ENOTCONN;
 	local->fd = fd;
 
-	hafdp = CALLOC (1, sizeof (*hafdp));
+	hafdp = GF_CALLOC (1, sizeof (*hafdp), gf_ha_mt_hafd_t);
         if (!hafdp) {
                 op_errno = ENOMEM;
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
                 goto err;
         }
 
-	hafdp->fdstate = CALLOC (1, child_count);
+	hafdp->fdstate = GF_CALLOC (1, child_count, gf_ha_mt_char);
         if (!hafdp->fdstate) {
                 op_errno = ENOMEM;
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
                 goto err;
         }
 
-	hafdp->path = strdup (loc->path);
+	hafdp->path = gf_strdup (loc->path);
         if (!hafdp->path) {
                 op_errno = ENOMEM;
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
@@ -2020,16 +2013,16 @@ err:
 	STACK_UNWIND (frame, -1, op_errno, fd);
         if (hafdp) {
                 if (hafdp->fdstate) {
-                        FREE (hafdp->fdstate);
+                        GF_FREE (hafdp->fdstate);
                         hafdp->fdstate = NULL;
                 }
 
                 if (hafdp->path) {
-                        FREE (hafdp->path);
+                        GF_FREE (hafdp->path);
                         hafdp->path = NULL;
                 }
 
-                FREE (hafdp);
+                GF_FREE (hafdp);
         }
 
         ha_local_wipe (local);
@@ -2044,7 +2037,7 @@ ha_readv_cbk (call_frame_t *frame,
 	      int32_t op_errno,
 	      struct iovec *vector,
 	      int32_t count,
-	      struct stat *stbuf,
+	      struct iatt *stbuf,
               struct iobref *iobref)
 {
 	int ret = 0;
@@ -2111,8 +2104,8 @@ ha_writev_cbk (call_frame_t *frame,
 	       xlator_t *this,
 	       int32_t op_ret,
 	       int32_t op_errno,
-               struct stat *prebuf,
-	       struct stat *postbuf)
+               struct iatt *prebuf,
+	       struct iatt *postbuf)
 {
 	int ret = 0;
 	ret = ha_handle_cbk (frame, cookie, op_ret, op_errno);
@@ -2237,8 +2230,8 @@ ha_fsync_cbk (call_frame_t *frame,
 	      xlator_t *this,
 	      int32_t op_ret,
 	      int32_t op_errno,
-              struct stat *prebuf,
-              struct stat *postbuf)
+              struct iatt *prebuf,
+              struct iatt *postbuf)
 {
 	int ret = 0;
 	ret = ha_handle_cbk (frame, cookie, op_ret, op_errno);
@@ -2297,7 +2290,7 @@ ha_fstat_cbk (call_frame_t *frame,
 	      xlator_t *this,
 	      int32_t op_ret,
 	      int32_t op_errno,
-	      struct stat *buf)
+	      struct iatt *buf)
 {
 	int ret = 0;
 
@@ -2420,7 +2413,8 @@ ha_opendir (call_frame_t *frame,
 	children = pvt->children;
 	child_count = pvt->child_count;
 
-	local = frame->local = CALLOC (1, sizeof (*local));
+	frame->local = local = GF_CALLOC (1, sizeof (*local), 
+                                          gf_ha_mt_ha_local_t);
         if (!local) {
                 op_errno = ENOMEM;
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
@@ -2431,21 +2425,21 @@ ha_opendir (call_frame_t *frame,
 	local->op_errno = ENOTCONN;
 	local->fd = fd;
 
-	hafdp = CALLOC (1, sizeof (*hafdp));
+	hafdp = GF_CALLOC (1, sizeof (*hafdp), gf_ha_mt_hafd_t);
         if (!hafdp) {
                 op_errno = ENOMEM;
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
                 goto err;
         }
 
-	hafdp->fdstate = CALLOC (1, child_count);
+	hafdp->fdstate = GF_CALLOC (1, child_count, gf_ha_mt_char);
         if (!hafdp->fdstate) {
                 op_errno = ENOMEM;
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
                 goto err;
         }
 
-	hafdp->path = strdup (loc->path);
+	hafdp->path = gf_strdup (loc->path);
         if (!hafdp->path) {
                 op_errno = ENOMEM;
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
@@ -2484,16 +2478,16 @@ err:
         ha_local_wipe (local);
         if (hafdp) {
                 if (hafdp->fdstate) {
-                        FREE (hafdp->fdstate);
+                        GF_FREE (hafdp->fdstate);
                         hafdp->fdstate = NULL;
                 }
 
                 if (hafdp->path) {
-                        FREE (hafdp->path);
+                        GF_FREE (hafdp->path);
                         hafdp->path = NULL;
                 }
 
-                FREE (hafdp);
+                GF_FREE (hafdp);
         }
         return 0;
 }
@@ -2733,7 +2727,8 @@ ha_statfs (call_frame_t *frame,
         /* The normal way of handling failover doesn't work here
          * as loc->inode may be null in this case.
          */
-        local = CALLOC (1, sizeof (*local));
+        local = GF_CALLOC (1, sizeof (*local), 
+                           gf_ha_mt_ha_local_t);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -3057,7 +3052,7 @@ ha_lk_setlk_unlck_cbk (call_frame_t *frame,
 		       xlator_t *this,
 		       int32_t op_ret,
 		       int32_t op_errno,
-		       struct flock *lock)
+		       struct gf_flock *lock)
 {
 	ha_local_t *local = NULL;
 	int cnt = 0;
@@ -3073,7 +3068,7 @@ ha_lk_setlk_unlck_cbk (call_frame_t *frame,
 
 	if (cnt == 0) {
 		stub = local->stub;
-		FREE (local->state);
+		GF_FREE (local->state);
 		if (stub->args.lk.lock.l_type == F_UNLCK) {
 			STACK_UNWIND (frame, local->op_ret, local->op_errno, &stub->args.lk.lock);
 		} else {
@@ -3090,7 +3085,7 @@ ha_lk_setlk_cbk (call_frame_t *frame,
 		 xlator_t *this,
 		 int32_t op_ret,
 		 int32_t op_errno,
-		 struct flock *lock)
+		 struct gf_flock *lock)
 {
 	ha_local_t *local = NULL;
 	ha_private_t *pvt = NULL;
@@ -3122,7 +3117,7 @@ ha_lk_setlk_cbk (call_frame_t *frame,
 		}
 		if (i == child_count) {
 			call_stub_t *stub = local->stub;
-			FREE (local->state);
+			GF_FREE (local->state);
 			STACK_UNWIND (frame, 0, op_errno, &stub->args.lk.lock);
 			call_stub_destroy (stub);
 			return 0;
@@ -3146,7 +3141,7 @@ ha_lk_setlk_cbk (call_frame_t *frame,
 				cnt++;
 		}
 		if (cnt) {
-			struct flock lock;
+			struct gf_flock lock;
 			lock = local->stub->args.lk.lock;
 			for (i = 0; i < child_count; i++) {
 				if (state[i]) {
@@ -3163,7 +3158,7 @@ ha_lk_setlk_cbk (call_frame_t *frame,
 			}
 			return 0;
 		} else {
-			FREE (local->state);
+			GF_FREE (local->state);
 			call_stub_destroy (local->stub);
 			STACK_UNWIND (frame,
 				      op_ret,
@@ -3180,7 +3175,7 @@ ha_lk_getlk_cbk (call_frame_t *frame,
 		 xlator_t *this,
 		 int32_t op_ret,
 		 int32_t op_errno,
-		 struct flock *lock)
+		 struct gf_flock *lock)
 {
 	ha_local_t *local = NULL;
 	ha_private_t *pvt = NULL;
@@ -3197,7 +3192,7 @@ ha_lk_getlk_cbk (call_frame_t *frame,
 	prev_frame = cookie;
 
 	if (op_ret == 0) {
-		FREE (local->state);
+		GF_FREE (local->state);
 		call_stub_destroy (local->stub);
 		STACK_UNWIND (frame, 0, 0, lock);
 		return 0;
@@ -3214,7 +3209,7 @@ ha_lk_getlk_cbk (call_frame_t *frame,
 	}
 
 	if (i == child_count) {
-		FREE (local->state);
+		GF_FREE (local->state);
 		call_stub_destroy (local->stub);
 		STACK_UNWIND (frame, op_ret, op_errno, lock);
 		return 0;
@@ -3235,7 +3230,7 @@ ha_lk (call_frame_t *frame,
        xlator_t *this,
        fd_t *fd,
        int32_t cmd,
-       struct flock *lock)
+       struct gf_flock *lock)
 {
 	ha_local_t *local = NULL;
 	ha_private_t *pvt = NULL;
@@ -3255,7 +3250,8 @@ ha_lk (call_frame_t *frame,
 		gf_log (this->name, GF_LOG_ERROR, "fd_ctx_get failed");
 
 	if (local == NULL) {
-		local = frame->local = CALLOC (1, sizeof (*local));
+                local = frame->local = GF_CALLOC (1, sizeof (*local),
+                                                  gf_ha_mt_ha_local_t);
                 if (!local) {
                         op_errno = ENOMEM;
                         gf_log (this->name, GF_LOG_ERROR, "out of memory");
@@ -3280,7 +3276,7 @@ ha_lk (call_frame_t *frame,
                 goto err;
         }
 
-	local->state = CALLOC (1, child_count);
+	local->state = GF_CALLOC (1, child_count, gf_ha_mt_char);
         if (!local->state) {
                 op_errno = ENOMEM;
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
@@ -3368,7 +3364,7 @@ ha_inodelk (call_frame_t *frame,
             const char *volume,
 	    loc_t *loc,
 	    int32_t cmd,
-	    struct flock *lock)
+	    struct gf_flock *lock)
 {
 	ha_local_t *local = NULL;
 	int op_errno = 0;
@@ -3636,7 +3632,8 @@ ha_stats (call_frame_t *frame,
 	int i = 0;
         int32_t op_errno = EINVAL;
 
-	local = frame->local = CALLOC (1, sizeof (*local));
+	local = frame->local = GF_CALLOC (1, sizeof (*local),
+                                          gf_ha_mt_ha_local_t);
         if (!local) {
                 op_errno = ENOMEM;
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
@@ -3737,7 +3734,8 @@ ha_getspec (call_frame_t *frame,
 	int i = 0;
         int32_t op_errno = EINVAL;
 
-	local = frame->local = CALLOC (1, sizeof (*local));
+	local = frame->local = GF_CALLOC (1, sizeof (*local), 
+                                          gf_ha_mt_ha_local_t);
         if (!local) {
                 op_errno = ENOMEM;
                 gf_log (this->name, GF_LOG_ERROR, "out of memory");
@@ -3791,8 +3789,8 @@ ha_closedir (xlator_t *this,
 	}
 	hafdp = (hafd_t *)(long)tmp_hafdp;
 
-	FREE (hafdp->fdstate);
-	FREE (hafdp->path);
+	GF_FREE (hafdp->fdstate);
+	GF_FREE (hafdp->path);
 	LOCK_DESTROY (&hafdp->lock);
 	return 0;
 }
@@ -3812,8 +3810,8 @@ ha_close (xlator_t *this,
 	}
 	hafdp = (hafd_t *)(long)tmp_hafdp;
 
-	FREE (hafdp->fdstate);
-	FREE (hafdp->path);
+	GF_FREE (hafdp->fdstate);
+	GF_FREE (hafdp->path);
 	LOCK_DESTROY (&hafdp->lock);
 	return 0;
 }
@@ -3884,6 +3882,25 @@ notify (xlator_t *this,
 	return 0;
 }
 
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_ha_mt_end + 1);
+        
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
+                                "failed");
+                return ret;
+        }
+
+        return ret;
+}
+
 int
 init (xlator_t *this)
 {
@@ -3891,6 +3908,7 @@ init (xlator_t *this)
 	xlator_list_t *trav = NULL;
 	int count = 0, ret = 0;
 
+
 	if (!this->children) {
 		gf_log (this->name,GF_LOG_ERROR, 
 			"FATAL: ha should have one or more child defined");
@@ -3903,7 +3921,7 @@ init (xlator_t *this)
 	}
   
 	trav = this->children;
-	pvt = CALLOC (1, sizeof (ha_private_t));
+	pvt = GF_CALLOC (1, sizeof (ha_private_t), gf_ha_mt_ha_private_t);
 
 	ret = dict_get_int32 (this->options, "preferred-subvolume", 
 			      &pvt->pref_subvol);
@@ -3918,7 +3936,8 @@ init (xlator_t *this)
 	}
 
 	pvt->child_count = count;
-	pvt->children = CALLOC (count, sizeof (xlator_t*));
+	pvt->children = GF_CALLOC (count, sizeof (xlator_t*), 
+                                   gf_ha_mt_xlator_t);
 
 	trav = this->children;
 	count = 0;
@@ -3928,7 +3947,7 @@ init (xlator_t *this)
 		trav = trav->next;
 	}
 
-	pvt->state = CALLOC (1, count);
+	pvt->state = GF_CALLOC (1, count, gf_ha_mt_char);
 	this->private = pvt;
 	return 0;
 }
@@ -3938,7 +3957,7 @@ fini (xlator_t *this)
 {
 	ha_private_t *priv = NULL;
 	priv = this->private;
-	FREE (priv);
+	GF_FREE (priv);
 	return;
 }
 
@@ -3983,11 +4002,6 @@ struct xlator_fops fops = {
         .fsetattr    = ha_fsetattr,
 };
 
-struct xlator_mops mops = {
-	.stats   = ha_stats,
-	.getspec = ha_getspec,
-};
-
 struct xlator_cbks cbks = {
 	.release    = ha_close,
 	.releasedir = ha_closedir,
diff --git a/xlators/cluster/ha/src/ha.h b/xlators/cluster/ha/src/ha.h
index 0da31850c..e2ed7eaa6 100644
--- a/xlators/cluster/ha/src/ha.h
+++ b/xlators/cluster/ha/src/ha.h
@@ -1,25 +1,17 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef __HA_H_
 #define __HA_H_
 
+#include "ha-mem-types.h"
+
 typedef struct {
 	call_stub_t *stub;
 	int32_t op_ret, op_errno;
@@ -28,9 +20,9 @@ typedef struct {
 	char *state, *pattern;
 	dict_t *dict;
 	loc_t loc;
-	struct stat buf;
-        struct stat postparent;
-        struct stat preparent;
+	struct iatt buf;
+        struct iatt postparent;
+        struct iatt preparent;
 	fd_t *fd;
 	inode_t *inode;
 	int32_t flags;
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 b4c8ad525..851397b68 100644
--- a/xlators/cluster/map/src/map-helper.c
+++ b/xlators/cluster/map/src/map-helper.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2009-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2009-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -256,14 +246,15 @@ verify_dir_and_assign_subvol (xlator_t *this,
 				goto out;
 			}
 
-			tmp_map = CALLOC (1, sizeof (struct map_pattern));
+			tmp_map = GF_CALLOC (1, sizeof (struct map_pattern),
+                                             gf_map_mt_map_pattern);
 			tmp_map->xl = trav->xlator;
 			tmp_map->dir_len = strlen (directory);
 
 			/* make sure that the top level directory starts 
 			 * with '/' and ends without '/'
 			 */
-			tmp_map->directory = strdup (directory);
+			tmp_map->directory = gf_strdup (directory);
 			if (directory[tmp_map->dir_len - 1] == '/') {
 				tmp_map->dir_len--;
 			}
diff --git a/xlators/cluster/map/src/map-mem-types.h b/xlators/cluster/map/src/map-mem-types.h
new file mode 100644
index 000000000..3e89f4736
--- /dev/null
+++ b/xlators/cluster/map/src/map-mem-types.h
@@ -0,0 +1,24 @@
+/*
+   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 __MAP_MEM_TYPES_H__
+#define __MAP_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_map_mem_types_ {
+        gf_map_mt_map_private_t = gf_common_mt_end + 1,
+        gf_map_mt_map_local_t,
+        gf_map_mt_map_xlator_array,
+        gf_map_mt_map_pattern,
+        gf_map_mt_end
+};
+#endif
+
diff --git a/xlators/cluster/map/src/map.c b/xlators/cluster/map/src/map.c
index 72464fc82..6150a33ce 100644
--- a/xlators/cluster/map/src/map.c
+++ b/xlators/cluster/map/src/map.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -36,13 +26,13 @@ map_stat_cbk (call_frame_t *frame,
 		  xlator_t *this,
 		  int32_t op_ret,
 		  int32_t op_errno,
-		  struct stat *buf)
+		  struct iatt *buf)
 
 {
         call_frame_t *prev = NULL;
         prev  = cookie;
 	
-	map_itransform (this, prev->this, buf->st_ino, &buf->st_ino);
+	map_itransform (this, prev->this, buf->ia_ino, &buf->ia_ino);
 
 	STACK_UNWIND (frame, op_ret, op_errno, buf);
 	return 0;
@@ -54,14 +44,14 @@ map_setattr_cbk (call_frame_t *frame,
                  xlator_t *this,
                  int32_t op_ret,
                  int32_t op_errno,
-                 struct stat *statpre,
-                 struct stat *statpost)
+                 struct iatt *statpre,
+                 struct iatt *statpost)
 {
         call_frame_t *prev = NULL;
         prev  = cookie;
 
-	map_itransform (this, prev->this, statpre->st_ino, &statpre->st_ino);
-	map_itransform (this, prev->this, statpost->st_ino, &statpost->st_ino);
+	map_itransform (this, prev->this, statpre->ia_ino, &statpre->ia_ino);
+	map_itransform (this, prev->this, statpost->ia_ino, &statpost->ia_ino);
 
 	STACK_UNWIND (frame, op_ret, op_errno, statpre, statpost);
 	return 0;
@@ -73,14 +63,14 @@ map_fsetattr_cbk (call_frame_t *frame,
                   xlator_t *this,
                   int32_t op_ret,
                   int32_t op_errno,
-                  struct stat *statpre,
-                  struct stat *statpost)
+                  struct iatt *statpre,
+                  struct iatt *statpost)
 {
         call_frame_t *prev = NULL;
         prev  = cookie;
 
-	map_itransform (this, prev->this, statpre->st_ino, &statpre->st_ino);
-	map_itransform (this, prev->this, statpost->st_ino, &statpost->st_ino);
+	map_itransform (this, prev->this, statpre->ia_ino, &statpre->ia_ino);
+	map_itransform (this, prev->this, statpost->ia_ino, &statpost->ia_ino);
 
 	STACK_UNWIND (frame, op_ret, op_errno, statpre, statpost);
 	return 0;
@@ -92,13 +82,13 @@ map_truncate_cbk (call_frame_t *frame,
 		      xlator_t *this,
 		      int32_t op_ret,
 		      int32_t op_errno,
-		      struct stat *prebuf,
-                      struct stat *postbuf)
+		      struct iatt *prebuf,
+                      struct iatt *postbuf)
 {
         call_frame_t *prev = NULL;
         prev  = cookie;
 	
-	map_itransform (this, prev->this, postbuf->st_ino, &postbuf->st_ino);
+	map_itransform (this, prev->this, postbuf->ia_ino, &postbuf->ia_ino);
 
 	STACK_UNWIND (frame, op_ret, op_errno, prebuf, postbuf);
 	return 0;
@@ -110,13 +100,13 @@ map_ftruncate_cbk (call_frame_t *frame,
 		       xlator_t *this,
 		       int32_t op_ret,
 		       int32_t op_errno,
-		       struct stat *prebuf,
-                       struct stat *postbuf)
+		       struct iatt *prebuf,
+                       struct iatt *postbuf)
 {
         call_frame_t *prev = NULL;
         prev  = cookie;
 	
-	map_itransform (this, prev->this, postbuf->st_ino, &postbuf->st_ino);
+	map_itransform (this, prev->this, postbuf->ia_ino, &postbuf->ia_ino);
 
 	STACK_UNWIND (frame, op_ret, op_errno, prebuf, postbuf);
 	return 0;
@@ -141,7 +131,7 @@ map_readlink_cbk (call_frame_t *frame,
 		      int32_t op_ret,
 		      int32_t op_errno,
 		      const char *path,
-                      struct stat *sbuf)
+                      struct iatt *sbuf)
 {
 	STACK_UNWIND (frame, op_ret, op_errno, path, sbuf);
 	return 0;
@@ -153,8 +143,8 @@ map_unlink_cbk (call_frame_t *frame,
 		    xlator_t *this,
 		    int32_t op_ret,
 		    int32_t op_errno,
-                    struct stat *preparent,
-                    struct stat *postparent)
+                    struct iatt *preparent,
+                    struct iatt *postparent)
 {
 	STACK_UNWIND (frame, op_ret, op_errno, preparent, postparent);
 	return 0;
@@ -166,8 +156,8 @@ map_rmdir_cbk (call_frame_t *frame,
 		   xlator_t *this,
 		   int32_t op_ret,
 		   int32_t op_errno,
-                   struct stat *preparent,
-                   struct stat *postparent)
+                   struct iatt *preparent,
+                   struct iatt *postparent)
 {
 	STACK_UNWIND (frame, op_ret, op_errno, preparent, postparent);
 	return 0;
@@ -180,16 +170,16 @@ map_rename_cbk (call_frame_t *frame,
 		    xlator_t *this,
 		    int32_t op_ret,
 		    int32_t op_errno,
-		    struct stat *buf,
-                    struct stat *preoldparent,
-                    struct stat *postoldparent,
-                    struct stat *prenewparent,
-                    struct stat *postnewparent)
+		    struct iatt *buf,
+                    struct iatt *preoldparent,
+                    struct iatt *postoldparent,
+                    struct iatt *prenewparent,
+                    struct iatt *postnewparent)
 {
         call_frame_t *prev = NULL;
         prev  = cookie;
 	
-	map_itransform (this, prev->this, buf->st_ino, &buf->st_ino);
+	map_itransform (this, prev->this, buf->ia_ino, &buf->ia_ino);
 
 	STACK_UNWIND (frame, op_ret, op_errno, buf);
 	return 0;
@@ -202,14 +192,14 @@ map_link_cbk (call_frame_t *frame,
 		  int32_t op_ret,
 		  int32_t op_errno,
 		  inode_t *inode,
-                  struct stat *buf,
-                  struct stat *preparent,
-                  struct stat *postparent)
+                  struct iatt *buf,
+                  struct iatt *preparent,
+                  struct iatt *postparent)
 {
         call_frame_t *prev = NULL;
         prev  = cookie;
 	
-	map_itransform (this, prev->this, buf->st_ino, &buf->st_ino);
+	map_itransform (this, prev->this, buf->ia_ino, &buf->ia_ino);
 
 	STACK_UNWIND (frame, op_ret, op_errno, inode,	buf);
 	return 0;
@@ -235,13 +225,13 @@ map_readv_cbk (call_frame_t *frame,
                int32_t op_errno,
                struct iovec *vector,
                int32_t count,
-               struct stat *stbuf,
+               struct iatt *stbuf,
                struct iobref *iobref)
 {
         call_frame_t *prev = NULL;
         prev  = cookie;
 	
-	map_itransform (this, prev->this, stbuf->st_ino, &stbuf->st_ino);
+	map_itransform (this, prev->this, stbuf->ia_ino, &stbuf->ia_ino);
 
 	STACK_UNWIND (frame, op_ret, op_errno, vector, count, stbuf, iobref);
 	return 0;
@@ -253,13 +243,13 @@ map_writev_cbk (call_frame_t *frame,
 		    xlator_t *this,
 		    int32_t op_ret,
 		    int32_t op_errno,
-                    struct stat *prebuf,
-		    struct stat *postbuf)
+                    struct iatt *prebuf,
+		    struct iatt *postbuf)
 {
         call_frame_t *prev = NULL;
         prev  = cookie;
 	
-	map_itransform (this, prev->this, postbuf->st_ino, &postbuf->st_ino);
+	map_itransform (this, prev->this, postbuf->ia_ino, &postbuf->ia_ino);
 
 	STACK_UNWIND (frame, op_ret, op_errno, prebuf, postbuf);
 	return 0;
@@ -283,8 +273,8 @@ map_fsync_cbk (call_frame_t *frame,
 		   xlator_t *this,
 		   int32_t op_ret,
 		   int32_t op_errno,
-                   struct stat *prebuf,
-                   struct stat *postbuf)
+                   struct iatt *prebuf,
+                   struct iatt *postbuf)
 {
 	STACK_UNWIND (frame, op_ret, op_errno, prebuf, postbuf);
 	return 0;
@@ -297,12 +287,12 @@ map_fstat_cbk (call_frame_t *frame,
 		   xlator_t *this,
 		   int32_t op_ret,
 		   int32_t op_errno,
-		   struct stat *buf)
+		   struct iatt *buf)
 {
         call_frame_t *prev = NULL;
         prev  = cookie;
 	
-	map_itransform (this, prev->this, buf->st_ino, &buf->st_ino);
+	map_itransform (this, prev->this, buf->ia_ino, &buf->ia_ino);
 
 	STACK_UNWIND (frame, op_ret, op_errno, buf);
 	return 0;
@@ -438,7 +428,7 @@ map_lk_cbk (call_frame_t *frame,
 		xlator_t *this,
 		int32_t op_ret,
 		int32_t op_errno,
-		struct flock *lock)
+		struct gf_flock *lock)
 {
 	STACK_UNWIND (frame, op_ret, op_errno, lock);
 	return 0;
@@ -492,14 +482,14 @@ map_newentry_cbk (call_frame_t *frame,
 		  int32_t op_ret,
 		  int32_t op_errno,
 		  inode_t *inode,
-                  struct stat *buf,
-                  struct stat *preparent,
-                  struct stat *postparent)
+                  struct iatt *buf,
+                  struct iatt *preparent,
+                  struct iatt *postparent)
 {
         call_frame_t *prev = NULL;
         prev  = cookie;
 	
-	map_itransform (this, prev->this, buf->st_ino, &buf->st_ino);
+	map_itransform (this, prev->this, buf->ia_ino, &buf->ia_ino);
 
 	STACK_UNWIND (frame, op_ret, op_errno, inode, buf);
 	return 0;
@@ -515,14 +505,14 @@ map_create_cbk (call_frame_t *frame,
 		int32_t op_errno,
 		fd_t *fd,
 		inode_t *inode,
-		struct stat *buf,
-                struct stat *preparent,
-                struct stat *postparent)
+		struct iatt *buf,
+                struct iatt *preparent,
+                struct iatt *postparent)
 {
         call_frame_t *prev = NULL;
         prev  = cookie;
 	
-	map_itransform (this, prev->this, buf->st_ino, &buf->st_ino);
+	map_itransform (this, prev->this, buf->ia_ino, &buf->ia_ino);
 
 	STACK_UNWIND (frame, op_ret, op_errno, fd, inode, buf);
 	return 0;
@@ -624,14 +614,14 @@ map_single_lookup_cbk (call_frame_t *frame,
 		       int32_t op_ret,
 		       int32_t op_errno,
 		       inode_t *inode,
-		       struct stat *buf,
+		       struct iatt *buf,
                        dict_t *dict,
-                       struct stat *postparent)
+                       struct iatt *postparent)
 {
         call_frame_t *prev = NULL;
         prev  = cookie;
 	
-	map_itransform (this, prev->this, buf->st_ino, &buf->st_ino);
+	map_itransform (this, prev->this, buf->ia_ino, &buf->ia_ino);
 
 	STACK_UNWIND (frame, op_ret, op_errno, inode, buf, dict);
 
@@ -645,9 +635,9 @@ map_root_lookup_cbk (call_frame_t *frame,
                      int32_t op_ret,
                      int32_t op_errno,
                      inode_t *inode,
-                     struct stat *buf,
+                     struct iatt *buf,
                      dict_t *dict,
-                     struct stat *postparent)
+                     struct iatt *postparent)
 {
 	int          callcnt = 0;
 	map_local_t *local = NULL;
@@ -762,7 +752,7 @@ map_single_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 	list_for_each_entry (orig_entry, &entries->list, list) {
 		map_itransform (this, prev->this, orig_entry->d_ino,
 				&orig_entry->d_ino);
-                orig_entry->d_stat.st_ino = orig_entry->d_ino;
+                orig_entry->d_stat.ia_ino = orig_entry->d_ino;
 	}
 	STACK_UNWIND (frame, op_ret, op_errno, entries);
 	return 0;
@@ -815,7 +805,7 @@ map_generic_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 				&entry->d_off);
 		
                 if (whichop == GF_FOP_READDIRP)
-                        entry->d_stat.st_ino = entry->d_ino;
+                        entry->d_stat.ia_ino = entry->d_ino;
 		entry->d_type = orig_entry->d_type;
 		entry->d_len  = orig_entry->d_len;
 		
@@ -905,23 +895,6 @@ map_checksum_cbk (call_frame_t *frame,
 }
 
 
-int32_t
-map_lock_notify_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;
-}
-
-
-int32_t
-map_lock_fnotify_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;
-}
-
 /* Fops starts here */
 
 int32_t
@@ -957,7 +930,7 @@ int32_t
 map_setattr (call_frame_t *frame,
              xlator_t *this,
              loc_t *loc,
-             struct stat *stbuf,
+             struct iatt *stbuf,
              int32_t valid)
 {
 	int32_t op_errno = 1;
@@ -989,7 +962,7 @@ int32_t
 map_fsetattr (call_frame_t *frame,
               xlator_t *this,
               fd_t *fd,
-              struct stat *stbuf,
+              struct iatt *stbuf,
               int32_t valid)
 {
 	int32_t op_errno = 1;
@@ -1787,7 +1760,7 @@ map_lk (call_frame_t *frame,
 	xlator_t *this,
 	fd_t *fd,
 	int32_t cmd,
-	struct flock *lock)
+	struct gf_flock *lock)
 {
 	int32_t op_errno = 1;
 	xlator_t *subvol   = NULL;
@@ -1816,7 +1789,7 @@ map_lk (call_frame_t *frame,
 
 int32_t
 map_inodelk (call_frame_t *frame, xlator_t *this,
-	     const char *volume, loc_t *loc, int32_t cmd, struct flock *lock)
+	     const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *lock)
 {
 	int32_t op_errno = 1;
 	xlator_t *subvol   = NULL;
@@ -1846,7 +1819,7 @@ map_inodelk (call_frame_t *frame, xlator_t *this,
 
 int32_t
 map_finodelk (call_frame_t *frame, xlator_t *this,
-	      const char *volume, fd_t *fd, int32_t cmd, struct flock *lock)
+	      const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *lock)
 {
 	int32_t op_errno = 1;
 	xlator_t *subvol   = NULL;
@@ -2164,7 +2137,8 @@ map_lookup (call_frame_t *frame,
 	return 0;
 
  root_inode:
-	local = CALLOC (1, sizeof (map_local_t));
+	local = GF_CALLOC (1, sizeof (map_local_t),
+                           gf_map_mt_map_local_t);
 
 	frame->local = local;
 	local->call_count = priv->child_count;
@@ -2216,7 +2190,8 @@ map_statfs (call_frame_t *frame,
 	return 0;
 
  root_inode:
-	local = CALLOC (1, sizeof (map_local_t));
+	local = GF_CALLOC (1, sizeof (map_local_t),
+                           gf_map_mt_map_local_t);
 
 	priv = this->private;
 	frame->local = local;
@@ -2268,7 +2243,8 @@ map_opendir (call_frame_t *frame,
 	return 0;
 
  root_inode:
-	local = CALLOC (1, sizeof (map_local_t));
+	local = GF_CALLOC (1, sizeof (map_local_t),
+                           gf_map_mt_map_local_t);
 
 	priv = this->private;
 	frame->local = local;
@@ -2327,7 +2303,8 @@ map_do_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
 
  root_inode:
 	/* readdir on '/' */
-	local = CALLOC (1, sizeof (map_local_t));
+	local = GF_CALLOC (1, sizeof (map_local_t),
+                           gf_map_mt_map_local_t);
 	if (!local) {
 		gf_log (this->name, GF_LOG_ERROR,
 			"memory allocation failed :(");
@@ -2388,22 +2365,40 @@ fini (xlator_t *this)
 	priv = this->private;
 
 	if (priv) {
-		if (priv->xlarray)
-			FREE (priv->xlarray);
+		GF_FREE (priv->xlarray);
 
 		trav_map = priv->map;
 		while (trav_map) {
 			tmp_map = trav_map;
 			trav_map = trav_map->next;
-			FREE (tmp_map);
+			GF_FREE (tmp_map);
 		}
 
-		FREE(priv);
+		GF_FREE(priv);
 	}
 
 	return;
 }
 
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_map_mt_end + 1);
+        
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
+                                "failed");
+                return ret;
+        }
+
+        return ret;
+}
+
 int
 init (xlator_t *this)
 {
@@ -2420,6 +2415,7 @@ init (xlator_t *this)
 	char *subvol_str = NULL;
 	char *map_xl = NULL;
 
+
 	if (!this->children) {
 		gf_log (this->name,GF_LOG_ERROR,
 			"FATAL: map should have one or more child defined");
@@ -2431,7 +2427,8 @@ init (xlator_t *this)
 			"dangling volume. check volfile ");
 	}
 
-	priv = CALLOC (1, sizeof (map_private_t));
+	priv = GF_CALLOC (1, sizeof (map_private_t),
+                          gf_map_mt_map_private_t);
 	this->private = priv;
 
 	/* allocate xlator array */
@@ -2440,7 +2437,8 @@ init (xlator_t *this)
 		count++;
 		trav = trav->next;
 	}
-	priv->xlarray = CALLOC (1, sizeof (struct map_xlator_array) * count);
+	priv->xlarray = GF_CALLOC (1, sizeof (struct map_xlator_array) * count,
+                                   gf_map_mt_map_xlator_array);
 	priv->child_count = count;
 
 	/* build xlator array */
@@ -2460,7 +2458,7 @@ init (xlator_t *this)
 	}
 	map_pair_str = strtok_r (pattern_string, ";", &tmp_str);
 	while (map_pair_str) {
-		dup_map_pair = strdup (map_pair_str);
+		dup_map_pair = gf_strdup (map_pair_str);
 		dir_str = strtok_r (dup_map_pair, ":", &tmp_str1);
 		if (!dir_str) {
 			gf_log (this->name, GF_LOG_ERROR, 
@@ -2482,7 +2480,7 @@ init (xlator_t *this)
 			goto err;
 		}
 		
-		FREE (dup_map_pair);
+		GF_FREE (dup_map_pair);
 
 		map_pair_str = strtok_r (NULL, ";", &tmp_str);
 	}
@@ -2553,9 +2551,6 @@ struct xlator_fops fops = {
         .fsetattr    = map_fsetattr,
 };
 
-struct xlator_mops mops = {
-};
-
 struct xlator_cbks cbks = {
 };
 
diff --git a/xlators/cluster/map/src/map.h b/xlators/cluster/map/src/map.h
index b5f57518b..7703a543e 100644
--- a/xlators/cluster/map/src/map.h
+++ b/xlators/cluster/map/src/map.h
@@ -1,26 +1,17 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef __MAP_H__
 #define __MAP_H__
 
 #include "xlator.h"
+#include "map-mem-types.h"
 
 struct map_pattern {
 	struct map_pattern *next;
@@ -46,7 +37,7 @@ typedef struct {
 	int32_t        op_errno;
 	int            call_count;
 	struct statvfs statvfs;
-	struct stat    stbuf;
+	struct iatt    stbuf;
 	inode_t       *inode;
 	dict_t        *dict;
 	fd_t          *fd;
diff --git a/xlators/cluster/stripe/src/Makefile.am b/xlators/cluster/stripe/src/Makefile.am
index 180d0da1f..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
+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
new file mode 100644
index 000000000..e9ac9cf46
--- /dev/null
+++ b/xlators/cluster/stripe/src/stripe-mem-types.h
@@ -0,0 +1,31 @@
+/*
+  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 __STRIPE_MEM_TYPES_H__
+#define __STRIPE_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_stripe_mem_types_ {
+        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 ff2f84036..79e80b513 100644
--- a/xlators/cluster/stripe/src/stripe.c
+++ b/xlators/cluster/stripe/src/stripe.c
@@ -1,418 +1,184 @@
 /*
-  Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 /**
  * xlators/cluster/stripe:
- *    Stripe translator, stripes the data accross its child nodes, 
- *    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 
- *    show file size bigger than the actual size. But when one does 
+ *    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
+ *    show file size bigger than the actual size. But when one does
  *    'df' or 'du <file>', real size of the file on the server is shown.
  *
  * WARNING:
  *  Stripe translator can't regenerate data if a child node gets disconnected.
- *  So, no 'self-heal' for stripe. Hence the advice, use stripe only when its 
- *  very much necessary, or else, use it in combination with AFR, to have a 
- *  backup copy. 
- */
-
-/* TODO:
- * 1. Implement basic self-heal ability to manage the basic backend 
- *    layout missmatch.
- * 
+ *  So, no 'self-heal' for stripe. Hence the advice, use stripe only when its
+ *  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"
 
-/**
- * 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;
-        pathname   = strdup (path);
-        trav       = opts;
-
-        while (trav) {
-                if (!fnmatch (trav->path_pattern, pathname, FNM_NOESCAPE)) {
-                        block_size = trav->block_size;
-                        break;
-                }
-                trav = trav->next;
-        }
-        free (pathname);
-        
-        return block_size;
-}
-
-
-/*
- * stripe_common_remove_cbk -
- */
-int32_t
-stripe_common_remove_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                          int32_t op_ret, int32_t op_errno,
-                          struct stat *preparent, struct stat *postparent)
-{
-        STACK_UNWIND (frame, op_ret, op_errno, preparent, postparent);
-        return 0;
-}
-
+struct volume_options options[];
 
-/*
- * stripe_common_cbk -
- */
 int32_t
-stripe_common_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno)
+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, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno);
-        return 0;
-}
-
-/**
- * stripe_stack_unwind_cbk -  This function is used for all the _cbk without 
- *     any extra arguments (other than the minimum given)
- * This is called from functions like fsync,unlink,rmdir etc.
- *
- */
-int32_t 
-stripe_stack_unwind_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                         int32_t op_ret, int32_t op_errno)
-{
-        int32_t         callcnt = 0;
+        int             callcnt = -1;
         stripe_local_t *local   = NULL;
 
+        if (!this || !frame || !frame->local) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+
         local = frame->local;
 
         LOCK (&frame->lock);
         {
                 callcnt = --local->call_count;
-
-                if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG, 
-                                "%s returned %s",
-                                ((call_frame_t *)cookie)->this->name, 
-                                strerror (op_errno));                   
-                        local->op_errno = op_errno;
-                        if (op_errno == ENOTCONN) 
-                                local->failed = 1;
-                }
-                if (op_ret >= 0) 
-                        local->op_ret = op_ret;
         }
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                if (local->failed)
-                        local->op_ret = -1;
-
-                if (local->loc.path)
-                        loc_wipe (&local->loc);
-                if (local->loc2.path)
-                        loc_wipe (&local->loc2);
-
-                STACK_UNWIND (frame, local->op_ret, local->op_errno);
+                STRIPE_STACK_DESTROY (frame);
         }
+out:
         return 0;
 }
 
-
-
 int32_t
-stripe_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                  int32_t op_ret, int32_t op_errno, struct stat *prebuf,
-                  struct stat *postbuf)
+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, dict_t *xdata)
 {
-        int32_t         callcnt = 0;
-        stripe_local_t *local   = NULL;
-
-        local = frame->local;
-
-        LOCK (&frame->lock);
-        {
-                callcnt = --local->call_count;
+        stripe_local_t *local = NULL;
+        call_frame_t    *prev = NULL;
 
-                if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG, 
-                                "%s returned %s",
-                                ((call_frame_t *)cookie)->this->name,
-                                strerror (op_errno));
-                        local->op_errno = op_errno;
-                        if (op_errno == ENOTCONN)
-                                local->failed = 1;
-                }
-                if (op_ret >= 0) {
-                        local->op_ret = op_ret;
-                        if (!local->pre_buf.st_blksize) {
-                                local->pre_buf = *prebuf;
-                                local->pre_buf.st_blocks = 0;
-                        }
-                        if (!local->post_buf.st_blksize) {
-                                local->post_buf = *postbuf;
-                                local->post_buf.st_blocks = 0;
-                        }
-                        if (FIRST_CHILD(this) ==
-                           ((call_frame_t *)cookie)->this) {
-                                local->pre_buf.st_ino    = prebuf->st_ino;
-                                local->pre_buf.st_mtime  = prebuf->st_mtime;
-                                local->pre_buf.st_ctime  = prebuf->st_ctime;
-                                local->post_buf.st_ino   = postbuf->st_ino;
-                                local->post_buf.st_mtime = postbuf->st_mtime;
-                                local->post_buf.st_ctime = postbuf->st_ctime;
-                        }
-                        local->pre_buf.st_blocks += prebuf->st_blocks;
-                        if (local->pre_buf.st_size < prebuf->st_size)
-                                local->pre_buf.st_size = prebuf->st_size;
-                        if (local->pre_buf.st_blksize != prebuf->st_blksize) {
-                                /* TODO: add to blocks in terms of
-                                   original block size */
-                        }
-                        local->post_buf.st_blocks += postbuf->st_blocks;
-                        if (local->post_buf.st_size < postbuf->st_size)
-                                local->post_buf.st_size = postbuf->st_size;
-                        if (local->post_buf.st_blksize != postbuf->st_blksize) {
-                                /* TODO: add to blocks in terms of
-                                   original block size */
-                        }
-                }
+        if (!frame || !frame->local || !cookie || !this) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
         }
-        UNLOCK (&frame->lock);
 
-        if (!callcnt) {
-                if (local->failed)
-                        local->op_ret = -1;
+        prev  = cookie;
+        local = frame->local;
 
-                if (local->loc.path)
-                        loc_wipe (&local->loc);
-                if (local->loc2.path)
-                        loc_wipe (&local->loc2);
+        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), NULL);
 
-                STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-                              &local->pre_buf, &local->post_buf);
-        }
+out:
         return 0;
 }
 
-
-/**
- * stripe_stack_unwind_unlink_cbk -  
- * This is called from functions like unlink,rmdir.
- *
- */
 int32_t
-stripe_stack_unwind_unlink_cbk (call_frame_t *frame, void *cookie,
-                                xlator_t *this, int32_t op_ret,
-                                int32_t op_errno, struct stat *preparent,
-                                struct stat *postparent)
+stripe_entry_self_heal (call_frame_t *frame, xlator_t *this,
+                        stripe_local_t *local)
 {
-        int32_t         callcnt = 0;
-        stripe_local_t *local   = NULL;
-
-        local = frame->local;
-
-        LOCK (&frame->lock);
-        {
-                callcnt = --local->call_count;
+        xlator_list_t    *trav   = NULL;
+        call_frame_t     *rframe = NULL;
+        stripe_local_t   *rlocal = NULL;
+        stripe_private_t *priv   = NULL;
+        dict_t           *xdata  = NULL;
+        int               ret    = 0;
 
-                if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "%s returned %s",
-                                ((call_frame_t *)cookie)->this->name,
-                                strerror (op_errno));
-                        local->op_errno = op_errno;
-                        if (op_errno == ENOTCONN)
-                                local->failed = 1;
-                }
-                if (op_ret >= 0) {
-                        local->op_ret = op_ret;
-                        if (!local->pre_parent_buf.st_blksize) {
-                                local->pre_parent_buf = *preparent;
-                                local->pre_parent_buf.st_blocks = 0;
-                        }
-                        if (!local->post_parent_buf.st_blksize) {
-                                local->post_parent_buf = *postparent;
-                                local->post_parent_buf.st_blocks = 0;
-                        }
-                        if (FIRST_CHILD(this) ==
-                           ((call_frame_t *)cookie)->this) {
-                                local->pre_parent_buf.st_ino    = 
-                                        preparent->st_ino;
-                                local->pre_parent_buf.st_mtime  = 
-                                        preparent->st_mtime;
-                                local->pre_parent_buf.st_ctime  =
-                                        preparent->st_ctime;
-                                local->post_parent_buf.st_ino   = 
-                                        postparent->st_ino;
-                                local->post_parent_buf.st_mtime = 
-                                        postparent->st_mtime;
-                                local->post_parent_buf.st_ctime =
-                                        postparent->st_ctime;
-                        }
-                        local->pre_parent_buf.st_blocks += preparent->st_blocks;
-                        if (local->pre_parent_buf.st_size < preparent->st_size)
-                                local->pre_parent_buf.st_size = 
-                                        preparent->st_size;
-                        if (local->pre_parent_buf.st_blksize != 
-                            preparent->st_blksize) {
-                                /* TODO: add to blocks in terms of
-                                   original block size */
-                        }
-                        local->post_parent_buf.st_blocks += 
-                                postparent->st_blocks;
-                        if (local->post_parent_buf.st_size < 
-                            postparent->st_size)
-                                local->post_parent_buf.st_size = 
-                                        postparent->st_size;
-                        if (local->post_parent_buf.st_blksize != 
-                            postparent->st_blksize) {
-                                /* TODO: add to blocks in terms of
-                                   original block size */
-                        }
-                }
+        if (!local || !this || !frame) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
         }
-        UNLOCK (&frame->lock);
 
-        if (!callcnt) {
-                if (local->failed)
-                        local->op_ret = -1;
-
-                if (local->loc.path)
-                        loc_wipe (&local->loc);
-                if (local->loc2.path)
-                        loc_wipe (&local->loc2);
+        if (!(IA_ISREG (local->stbuf.ia_type) ||
+              IA_ISDIR (local->stbuf.ia_type)))
+                return 0;
 
-                STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-                              &local->pre_parent_buf, &local->post_parent_buf);
+        priv = this->private;
+        trav = this->children;
+        rframe = copy_frame (frame);
+        if (!rframe) {
+                goto out;
         }
-        return 0;
-}
-
-int32_t
-stripe_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                     int32_t op_ret, int32_t op_errno, struct stat *prebuf,
-                     struct stat *postbuf)
-{
-        int32_t         callcnt = 0;
-        stripe_local_t *local = NULL;
+        rlocal = mem_get0 (this->local_pool);
+        if (!rlocal) {
+                goto out;
+        }
+        rframe->local = rlocal;
+        rlocal->call_count = priv->child_count;
+        loc_copy (&rlocal->loc, &local->loc);
+        memcpy (&rlocal->stbuf, &local->stbuf, sizeof (struct iatt));
 
-        local = frame->local;
+        xdata = dict_new ();
+        if (!xdata)
+                goto out;
 
-        LOCK (&frame->lock);
-        {
-                callcnt = --local->call_count;
+        ret = dict_set_static_bin (xdata, "gfid-req", local->stbuf.ia_gfid, 16);
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "%s: failed to set gfid-req", local->loc.path);
 
-                if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "%s returned error %s",
-                                ((call_frame_t *)cookie)->this->name,
-                                strerror (op_errno));
-                        local->op_errno = op_errno;
-                        if (op_errno == ENOTCONN)
-                                local->failed = 1;
+        while (trav) {
+                if (IA_ISREG (local->stbuf.ia_type)) {
+                        STACK_WIND (rframe, stripe_sh_make_entry_cbk,
+                                    trav->xlator, trav->xlator->fops->mknod,
+                                    &local->loc,
+                                    st_mode_from_ia (local->stbuf.ia_prot,
+                                                     local->stbuf.ia_type),
+                                    0, 0, xdata);
                 }
-
-                if (op_ret == 0) {
-                        local->op_ret = 0;
-                        if (!local->pre_buf.st_blksize) {
-                                local->pre_buf = *prebuf;
-                                local->pre_buf.st_blocks = 0;
-                        }
-                        if (!local->post_buf.st_blksize) {
-                                local->post_buf = *postbuf;
-                                local->post_buf.st_blocks = 0;
-                        }
-                        if (FIRST_CHILD(this) ==
-                           ((call_frame_t *)cookie)->this) {
-                                local->pre_buf.st_ino    = prebuf->st_ino;
-                                local->pre_buf.st_mtime  = prebuf->st_mtime;
-                                local->pre_buf.st_ctime  = prebuf->st_ctime;
-                                local->post_buf.st_ino   = postbuf->st_ino;
-                                local->post_buf.st_mtime = postbuf->st_mtime;
-                                local->post_buf.st_ctime = postbuf->st_ctime;
-                        }
-                        local->pre_buf.st_blocks += prebuf->st_blocks;
-                        if (local->pre_buf.st_size < prebuf->st_size)
-                                local->pre_buf.st_size = prebuf->st_size;
-                        if (local->pre_buf.st_blksize != prebuf->st_blksize) {
-                                /* TODO: add to blocks in terms of
-                                   original block size */
-                        }
-                        local->post_buf.st_blocks += postbuf->st_blocks;
-                        if (local->post_buf.st_size < postbuf->st_size)
-                                local->post_buf.st_size = postbuf->st_size;
-                        if (local->post_buf.st_blksize != postbuf->st_blksize) {
-                                /* TODO: add to blocks in terms of
-                                   original block size */
-                        }
+                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),
+                                    0, xdata);
                 }
+                trav = trav->next;
         }
-        UNLOCK (&frame->lock);
-
-        if (!callcnt) {
-                if (local->failed)
-                        local->op_ret = -1;
 
-                if (local->loc.path)
-                        loc_wipe (&local->loc);
-                if (local->loc2.path)
-                        loc_wipe (&local->loc2);
+        if (xdata)
+                dict_unref (xdata);
+        return 0;
 
-                STACK_UNWIND (frame, local->op_ret, local->op_errno,
-                              &local->pre_buf, &local->post_buf);
-        }
+out:
+        if (rframe)
+                STRIPE_STACK_DESTROY (rframe);
+        if (xdata)
+                dict_unref (xdata);
 
         return 0;
 }
 
 
-int32_t 
-stripe_common_buf_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                       int32_t op_ret, int32_t op_errno, struct stat *buf)
+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 *xdata, struct iatt *postparent)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, buf);
-        return 0;
-}
+        int32_t         callcnt     = 0;
+        stripe_local_t *local       = NULL;
+        call_frame_t   *prev        = NULL;
+        int             ret         = 0;
 
-/**
- * stripe_stack_unwind_buf_cbk -  This function is used for all the _cbk with 
- *    'struct stat *buf' as extra argument (other than minimum)
- * This is called from functions like, chmod, fchmod, chown, fchown,
- * truncate, ftruncate, utimens etc.
- *
- * @cookie - this argument should be always 'xlator_t *' of child node 
- */
-int32_t 
-stripe_stack_unwind_buf_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                             int32_t op_ret, int32_t op_errno,
-                             struct stat *buf)
-{
-        int32_t         callcnt = 0;
-        stripe_local_t *local = 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);
@@ -420,240 +186,226 @@ stripe_stack_unwind_buf_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 callcnt = --local->call_count;
 
                 if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG, 
-                                "%s returned error %s",
-                                ((call_frame_t *)cookie)->this->name, 
-                                strerror (op_errno));
-                        local->op_errno = op_errno;
-                        if (op_errno == ENOTCONN)
+                        if (op_errno != ENOENT)
+                                gf_log (this->name, GF_LOG_DEBUG,
+                                        "%s returned error %s",
+                                        prev->this->name,
+                                        strerror (op_errno));
+                        if (local->op_errno != ESTALE)
+                                local->op_errno = op_errno;
+                        if (((op_errno != ENOENT) && (op_errno != ENOTCONN)) ||
+                            (prev->this == FIRST_CHILD (this)))
                                 local->failed = 1;
+                        if (op_errno == ENOENT)
+                                local->entry_self_heal_needed = 1;
                 }
-    
-                if (op_ret == 0) {
+
+                if (op_ret >= 0) {
                         local->op_ret = 0;
-                        if (local->post_buf.st_blksize == 0) {
-                                local->post_buf = *buf;
-                                /* Because st_blocks gets added again */
-                                local->post_buf.st_blocks = 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) == 
-                            ((call_frame_t *)cookie)->this) {
-                                /* Always, pass the inode number of 
-                                   first child to the above layer */
-                                local->post_buf.st_ino = buf->st_ino;
-                                local->post_buf.st_ctime = buf->st_ctime;
-                                local->post_buf.st_mtime = buf->st_mtime;
+                        if (FIRST_CHILD(this) == prev->this) {
+                                local->stbuf      = *buf;
+                                local->postparent = *postparent;
+                                local->inode = inode_ref (inode);
+                                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->post_buf.st_blocks += buf->st_blocks;
-                        if (local->post_buf.st_size < buf->st_size)
-                                local->post_buf.st_size = buf->st_size;
-                        if (local->post_buf.st_blksize != buf->st_blksize) {
-                                /* TODO: add to blocks in terms of 
-                                   original block size */
+                        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 &&
+                    !uuid_is_null (local->loc.inode->gfid))
+                        stripe_entry_self_heal (frame, this, local);
+
                 if (local->failed)
                         local->op_ret = -1;
 
-                if (local->loc.path)
-                        loc_wipe (&local->loc);
-                if (local->loc2.path)
-                        loc_wipe (&local->loc2);
+                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);
+                }
 
-                STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-                              &local->post_buf);
+                STRIPE_STACK_UNWIND (lookup, frame, local->op_ret,
+                                     local->op_errno, local->inode,
+                                     &local->stbuf, local->xdata,
+                                     &local->postparent);
         }
-
+out:
         return 0;
 }
 
-/* In case of symlink, mknod, the file is created on just first node */
-int32_t 
-stripe_common_inode_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                         int32_t op_ret, int32_t op_errno, inode_t *inode,
-                         struct stat *buf, struct stat *preparent,
-                         struct stat *postparent)
-{
-        STACK_UNWIND (frame, op_ret, op_errno, inode, buf, preparent,
-                      postparent);
-        return 0;
-}
-
-/**
- * stripe_stack_unwind_inode_cbk - This is called by the function like, 
- *                   link (), symlink (), mkdir (), mknod () 
- *           This creates a inode for new inode. It keeps a list of all 
- *           the inodes received from the child nodes. It is used while 
- *           forwarding any fops to child nodes.
- *
- */
-int32_t 
-stripe_stack_unwind_inode_cbk (call_frame_t *frame, void *cookie,
-                               xlator_t *this, int32_t op_ret, 
-                               int32_t op_errno, inode_t *inode,
-                               struct stat *buf, struct stat *preparent,
-                               struct stat *postparent)
+int32_t
+stripe_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
+               dict_t *xdata)
 {
-        int32_t         callcnt = 0;
-        stripe_local_t  *local   = NULL;
-        inode_t         *local_inode = 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;
 
-        local = frame->local;
+        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);
 
-        LOCK (&frame->lock);
-        {
-                callcnt = --local->call_count;
-    
-                if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG, 
-                                "%s returned error %s",
-                                ((call_frame_t *)cookie)->this->name, 
-                                strerror (op_errno));
-                        local->op_errno = op_errno;
-                        if (op_errno == ENOTCONN)
-                                local->failed = 1;
-                }
- 
-                if (op_ret >= 0) {
-                        local->op_ret = 0;
-                        if (!local->inode)
-                                local->inode = inode_ref (inode);
+        priv = this->private;
+        trav = this->children;
 
-                        if (!local->post_buf.st_blksize) {
-                                local->post_buf = *buf;
-                                local->post_buf.st_blocks = 0;
-                        }
-                        if (!local->pre_parent_buf.st_blksize) {
-                                local->pre_parent_buf = *preparent;
-                                local->pre_parent_buf.st_blocks = 0;
-                        }
-                        if (!local->post_parent_buf.st_blksize) {
-                                local->post_parent_buf = *postparent;
-                                local->post_parent_buf.st_blocks = 0;
-                        }
-                        if (FIRST_CHILD(this) ==
-                            ((call_frame_t *)cookie)->this) {
-                                local->post_buf.st_ino   = buf->st_ino;
-                                local->post_buf.st_mtime = buf->st_mtime;
-                                local->post_buf.st_ctime = buf->st_ctime;
-
-                                local->pre_parent_buf.st_ino = preparent->st_ino;
-                                local->pre_parent_buf.st_mtime = preparent->st_mtime;
-                                local->pre_parent_buf.st_ctime = preparent->st_ctime;
-                                local->post_parent_buf.st_ino = postparent->st_ino;
-                                local->post_parent_buf.st_mtime = postparent->st_mtime;
-                                local->post_parent_buf.st_ctime = postparent->st_ctime;
-                        }
-                        local->post_buf.st_blocks += buf->st_blocks;
-                        if (local->post_buf.st_size < buf->st_size)
-                                local->post_buf.st_size =  buf->st_size;
-
-                        local->pre_parent_buf.st_blocks += preparent->st_blocks;
-                        if (local->pre_parent_buf.st_size < 
-                            preparent->st_size)
-                                local->pre_parent_buf.st_size = 
-                                        preparent->st_size;
-                        local->post_parent_buf.st_blocks += 
-                                postparent->st_blocks;
-                        if (local->post_parent_buf.st_size < 
-                            postparent->st_size)
-                                local->post_parent_buf.st_size = 
-                                        postparent->st_size;
-                }
+        /* Initialization */
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                op_errno = ENOMEM;
+                goto err;
         }
-        UNLOCK (&frame->lock);
+        local->op_ret = -1;
+        frame->local = local;
+        loc_copy (&local->loc, loc);
 
-        if (!callcnt) {
-                if (local->failed)
-                        local->op_ret = -1;
+        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);
 
-                local_inode = local->inode;
-                STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-                              local->inode, &local->post_buf, 
-                              &local->pre_parent_buf, 
-                              &local->post_parent_buf);
-                if (local_inode)
-                        inode_unref (local_inode);
         }
 
+        /* 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, xdata);
+                trav = trav->next;
+        }
+
+	dict_unref(xdata);
+
+        return 0;
+err:
+        STRIPE_STACK_UNWIND (lookup, frame, -1, op_errno, NULL, NULL, NULL, NULL);
         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 stat *buf, dict_t *dict, struct stat *postparent)
+
+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, dict_t *xdata)
 {
         int32_t         callcnt = 0;
-        dict_t         *tmp_dict = NULL;
-        inode_t        *tmp_inode = NULL;
         stripe_local_t *local = NULL;
         call_frame_t   *prev = NULL;
 
-        prev = cookie;
+        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) {
-                        if (op_errno != ENOENT)
-                                gf_log (this->name, GF_LOG_DEBUG, 
-                                        "%s returned error %s",
-                                        prev->this->name,
-                                        strerror (op_errno));
-                        if (local->op_errno != ESTALE)
-                                local->op_errno = op_errno;
-                        if ((op_errno == ENOTCONN) || (op_errno == ESTALE))
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "%s returned error %s",
+                                prev->this->name, strerror (op_errno));
+                        local->op_errno = op_errno;
+                        if ((op_errno != ENOENT) ||
+                            (prev->this == FIRST_CHILD (this)))
                                 local->failed = 1;
-                        /* TODO: bring in self-heal ability */
-                        /* 
-                         * if (local->op_ret == 0) {
-                         *       if (S_ISREG (local->stbuf.st_mode) || 
-                         *           S_ISDIR (local->stbuf.st_mode))
-                         *               local->entry_self_heal_needed = 1;
-                         * }
-                         */
-                }
- 
-                if (op_ret >= 0) {
+                }
+
+                if (op_ret == 0) {
                         local->op_ret = 0;
 
-                        if (local->post_buf.st_blksize == 0) {
-                                local->inode = inode_ref (inode);
-                                local->post_buf = *buf;
-                                local->post_parent_buf = *postparent;
-                                /* Because st_blocks gets added again */
-                                local->post_buf.st_blocks = 0;
-                                local->post_parent_buf.st_blocks = 0;
-                        }
                         if (FIRST_CHILD(this) == prev->this) {
-                                local->post_buf.st_ino = buf->st_ino;
-                                local->post_buf.st_mtime = buf->st_mtime;
-                                local->post_buf.st_ctime = buf->st_ctime;
-                                local->post_parent_buf.st_ino = postparent->st_ino;
-                                local->post_parent_buf.st_mtime = postparent->st_mtime;
-                                local->post_parent_buf.st_ctime = postparent->st_mtime;
-                                if (local->dict)
-                                        dict_unref (local->dict);
-                                local->dict = dict_ref (dict);
-                        } else {
-                                if (!local->dict)
-                                        local->dict = dict_ref (dict);
+                                local->stbuf = *buf;
                         }
-                        local->post_buf.st_blocks += buf->st_blocks;
-                        if (local->post_buf.st_size < buf->st_size)
-                                local->post_buf.st_size = buf->st_size;
-                        local->post_parent_buf.st_blocks += postparent->st_blocks;
-                        if (local->post_parent_buf.st_size < postparent->st_size)
-                                local->post_parent_buf.st_size = postparent->st_size;
+
+                        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;
                 }
         }
         UNLOCK (&frame->lock);
@@ -662,34 +414,26 @@ stripe_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (local->failed)
                         local->op_ret = -1;
 
-                tmp_dict = local->dict;
-                tmp_inode = local->inode;
-
-                STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-                              local->inode, &local->post_buf, local->dict, 
-                              &local->post_parent_buf);
+                if (local->op_ret != -1) {
+                        local->stbuf.ia_size   = local->stbuf_size;
+                        local->stbuf.ia_blocks = local->stbuf_blocks;
+                }
 
-                if (tmp_inode)
-                        inode_unref (tmp_inode);
-                if (tmp_dict)
-                        dict_unref (tmp_dict);
+                STRIPE_STACK_UNWIND (stat, frame, local->op_ret,
+                                     local->op_errno, &local->stbuf, NULL);
         }
-
+out:
         return 0;
 }
 
-/**
- * stripe_lookup -
- */
-int32_t 
-stripe_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc, 
-               dict_t *xattr_req)
+int32_t
+stripe_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
-        stripe_local_t   *local = NULL;
         xlator_list_t    *trav = NULL;
+        stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
-        char              send_lookup_to_all = 0;
-        int32_t           op_errno = 1;
+	stripe_fd_ctx_t  *fctx = NULL;
+        int32_t           op_errno = EINVAL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -700,113 +444,53 @@ stripe_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
         priv = this->private;
         trav = this->children;
 
+        if (priv->first_child_down) {
+                op_errno = ENOTCONN;
+                goto err;
+        }
+
         /* Initialization */
-        local = CALLOC (1, sizeof (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 ((!loc->inode->st_mode) || S_ISDIR (loc->inode->st_mode) || 
-            S_ISREG (loc->inode->st_mode)) {
-                send_lookup_to_all = 1;
-        }
-
-        if (send_lookup_to_all) {
-                /* 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 = trav->next;
-                }
-        } else {
-                local->call_count = 1;
-                
-                STACK_WIND (frame, stripe_lookup_cbk, FIRST_CHILD(this), 
-                            FIRST_CHILD(this)->fops->lookup,
-                            loc, xattr_req);
-        }
-  
-        return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno, NULL, NULL, NULL);
-        return 0;
-}
-
-/**
- * stripe_stat -
- */
-int32_t
-stripe_stat (call_frame_t *frame, xlator_t *this, loc_t *loc)
-{ 
-        int               send_lookup_to_all = 0;
-        xlator_list_t    *trav = NULL;
-        stripe_local_t   *local = NULL;
-        stripe_private_t *priv = NULL;
-        int32_t           op_errno = 1;
-
-        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;
-
-        if (priv->first_child_down) {
-                op_errno = ENOTCONN;
-                goto err;
-        }
-
-        if (S_ISDIR (loc->inode->st_mode) || S_ISREG (loc->inode->st_mode))
-                send_lookup_to_all = 1;
+	if (IA_ISREG(loc->inode->ia_type)) {
+		inode_ctx_get(loc->inode, this, (uint64_t *) &fctx);
+		if (!fctx)
+			goto err;
+		local->fctx = fctx;
+	}
 
-        if (!send_lookup_to_all) {
-                STACK_WIND (frame, stripe_common_buf_cbk, FIRST_CHILD(this), 
-                            FIRST_CHILD(this)->fops->stat, loc);
-        } else {
-                /* Initialization */
-                local = CALLOC (1, sizeof (stripe_local_t));
-                if (!local) {
-                        op_errno = ENOMEM;
-                        goto err;
-                }
-                local->op_ret = -1;
-                frame->local = local;
-                local->inode = loc->inode;
-                local->call_count = priv->child_count;
-    
-                while (trav) {
-                        STACK_WIND (frame, stripe_stack_unwind_buf_cbk,
-                                    trav->xlator, trav->xlator->fops->stat, 
-                                    loc);
-                        trav = trav->next;
-                }
+        while (trav) {
+                STACK_WIND (frame, stripe_stat_cbk, trav->xlator,
+                            trav->xlator->fops->stat, loc, NULL);
+                trav = trav->next;
         }
 
         return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno, NULL);
+
+err:
+        STRIPE_STACK_UNWIND (stat, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
 
-/**
- * stripe_statfs_cbk - 
- */
 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;
 
+        if (!this || !frame || !frame->local) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
         local = frame->local;
 
         LOCK(&frame->lock);
@@ -833,32 +517,32 @@ stripe_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 }
         }
         UNLOCK (&frame->lock);
-  
+
         if (!callcnt) {
-                STACK_UNWIND (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;
 }
 
-
-/**
- * stripe_statfs - 
- */
 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;
         stripe_private_t *priv = NULL;
-        int32_t           op_errno = 1;
+        int32_t           op_errno = EINVAL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (loc, err);
 
         trav = this->children;
         priv = this->private;
 
         /* Initialization */
-        local = CALLOC (1, sizeof (stripe_local_t));
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -870,28 +554,99 @@ 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 (frame, -1, op_errno, NULL);
+err:
+        STRIPE_STACK_UNWIND (statfs, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
 
-/**
- * stripe_truncate - 
- */
+
 int32_t
-stripe_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset)
+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, 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 error %s",
+                                prev->this->name, strerror (op_errno));
+                        local->op_errno = op_errno;
+                        if ((op_errno != ENOENT) ||
+                            (prev->this == FIRST_CHILD (this)))
+                                local->failed = 1;
+                }
+
+                if (op_ret == 0) {
+                        local->op_ret = 0;
+                        if (FIRST_CHILD(this) == prev->this) {
+                                local->pre_buf  = *prebuf;
+                                local->post_buf = *postbuf;
+                        }
+
+                        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;
+
+                        if (local->postbuf_size < postbuf->ia_size)
+                                local->postbuf_size = postbuf->ia_size;
+                }
+        }
+        UNLOCK (&frame->lock);
+
+        if (!callcnt) {
+                if (local->failed)
+                        local->op_ret = -1;
+
+                if (local->op_ret != -1) {
+                        local->pre_buf.ia_blocks  = local->prebuf_blocks;
+                        local->pre_buf.ia_size    = local->prebuf_size;
+                        local->post_buf.ia_blocks = local->postbuf_blocks;
+                        local->post_buf.ia_size   = local->postbuf_size;
+                }
+
+                STRIPE_STACK_UNWIND (truncate, frame, local->op_ret,
+                                     local->op_errno, &local->pre_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, dict_t *xdata)
 {
-        int               send_fop_to_all = 0;
-        xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
-        int32_t           op_errno = 1;
+	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);
@@ -900,7 +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;
@@ -908,48 +662,83 @@ stripe_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset)
         }
 
         /* Initialization */
-        local = CALLOC (1, sizeof (stripe_local_t));
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         local->op_ret = -1;
         frame->local = local;
-        local->inode = loc->inode;
-        local->call_count = 1;
+        local->call_count = priv->child_count;
 
-        if (S_ISDIR (loc->inode->st_mode) || S_ISREG (loc->inode->st_mode))
-                send_fop_to_all = 1;
-
-        if (!send_fop_to_all) {
-                STACK_WIND (frame, stripe_truncate_cbk, trav->xlator,
-                            trav->xlator->fops->truncate, loc, offset);
-        } else {
-                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 (frame, -1, op_errno, NULL, NULL);
+err:
+        STRIPE_STACK_UNWIND (truncate, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 
-int32_t 
+int32_t
 stripe_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                     int32_t op_ret, int32_t op_errno,
-                    struct stat *preop, struct stat *postop)
+                    struct iatt *preop, struct iatt *postop, 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);
@@ -959,53 +748,31 @@ stripe_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (op_ret == -1) {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "%s returned error %s",
-                                ((call_frame_t *)cookie)->this->name,
-                                strerror (op_errno));
+                                prev->this->name, strerror (op_errno));
                         local->op_errno = op_errno;
-                        if (op_errno == ENOTCONN)
+                        if ((op_errno != ENOENT) ||
+                            (prev->this == FIRST_CHILD (this)))
                                 local->failed = 1;
                 }
 
                 if (op_ret == 0) {
                         local->op_ret = 0;
 
-                        if (local->stbuf.st_blksize == 0) {
-                                local->pre_buf = *preop;
-                                local->stbuf       = *postop;
-
-                                /* Because st_blocks gets added again */
-
-                                local->pre_buf.st_blocks = 0;
-                                local->stbuf.st_blocks     = 0;
-                        }
-
-                        if (FIRST_CHILD(this) ==
-                            ((call_frame_t *)cookie)->this) {
-                                /* Always, pass the inode number of
-                                   first child to the above layer */
-
-                                local->pre_buf.st_ino   = preop->st_ino;
-                                local->pre_buf.st_mtime = preop->st_mtime;
-                                local->pre_buf.st_ctime = preop->st_ctime;
-
-                                local->stbuf.st_ino   = postop->st_ino;
-                                local->stbuf.st_mtime = postop->st_mtime;
-                                local->stbuf.st_ctime = postop->st_ctime;
+                        if (FIRST_CHILD(this) == prev->this) {
+                                local->pre_buf  = *preop;
+                                local->post_buf = *postop;
                         }
 
-                        local->pre_buf.st_blocks += preop->st_blocks;
-                        local->stbuf.st_blocks     += postop->st_blocks;
-
+                        local->prebuf_blocks  += preop->ia_blocks;
+                        local->postbuf_blocks += postop->ia_blocks;
 
-                        if (local->pre_buf.st_size < preop->st_size)
-                                local->pre_buf.st_size = preop->st_size;
-                        if (local->stbuf.st_size < postop->st_size)
-                                local->stbuf.st_size = postop->st_size;
+			correct_file_size(preop, local->fctx, prev);
+			correct_file_size(postop, local->fctx, prev);
 
-                        if (local->stbuf.st_blksize != postop->st_blksize) {
-                                /* TODO: add to blocks in terms of
-                                   original block size */
-                        }
+                        if (local->prebuf_size < preop->ia_size)
+                                local->prebuf_size = preop->ia_size;
+                        if (local->postbuf_size < postop->ia_size)
+                                local->postbuf_size = postop->ia_size;
                 }
         }
         UNLOCK (&frame->lock);
@@ -1014,28 +781,31 @@ stripe_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (local->failed)
                         local->op_ret = -1;
 
-                if (local->loc.path)
-                        loc_wipe (&local->loc);
-                if (local->loc2.path)
-                        loc_wipe (&local->loc2);
+                if (local->op_ret != -1) {
+                        local->pre_buf.ia_blocks  = local->prebuf_blocks;
+                        local->pre_buf.ia_size    = local->prebuf_size;
+                        local->post_buf.ia_blocks = local->postbuf_blocks;
+                        local->post_buf.ia_size   = local->postbuf_size;
+                }
 
-                STACK_UNWIND (frame, local->op_ret, local->op_errno,
-                              &local->pre_buf, &local->stbuf);
+                STRIPE_STACK_UNWIND (setattr, frame, local->op_ret,
+                                     local->op_errno, &local->pre_buf,
+                                     &local->post_buf, NULL);
         }
-
+out:
         return 0;
 }
 
 
-int32_t 
+int32_t
 stripe_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                struct stat *stbuf, int32_t valid)
+                struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
-        int               send_fop_to_all = 0;
         xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
-        int32_t           op_errno = 1;
+	stripe_fd_ctx_t	 *fctx = NULL;
+        int32_t           op_errno = EINVAL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -1052,48 +822,52 @@ stripe_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
         }
 
         /* Initialization */
-        local = CALLOC (1, sizeof (stripe_local_t));
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         local->op_ret = -1;
         frame->local = local;
-        local->inode = loc->inode;
-        local->call_count = 1;
+        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 (S_ISDIR (loc->inode->st_mode) || S_ISREG (loc->inode->st_mode))
-                send_fop_to_all = 1;
+	if (IA_ISREG(loc->inode->ia_type)) {
+		inode_ctx_get(loc->inode, this, (uint64_t *) &fctx);
+		if (!fctx)
+			goto err;
+		local->fctx = fctx;
+	}
 
-        if (!send_fop_to_all) {
-                STACK_WIND (frame, stripe_setattr_cbk, trav->xlator,
-                            trav->xlator->fops->setattr, loc, stbuf, valid);
-        } else {
-                local->call_count = priv->child_count;
-    
-                while (trav) {
-                        STACK_WIND (frame, stripe_setattr_cbk,
-                                    trav->xlator, trav->xlator->fops->setattr,
-                                    loc, stbuf, valid);
-                        trav = trav->next;
-                }
+        local->call_count = priv->child_count;
+        while (trav) {
+                STACK_WIND (frame, stripe_setattr_cbk,
+                            trav->xlator, trav->xlator->fops->setattr,
+                            loc, stbuf, valid, NULL);
+                trav = trav->next;
         }
 
         return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno, NULL, NULL);
+err:
+        STRIPE_STACK_UNWIND (setattr, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 
-int32_t 
+int32_t
 stripe_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
-                 struct stat *stbuf, int32_t valid)
+                 struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
         xlator_list_t    *trav = NULL;
-        int32_t           op_errno = 1;
+        int32_t           op_errno = EINVAL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -1104,37 +878,44 @@ stripe_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
         trav = this->children;
 
         /* Initialization */
-        local = CALLOC (1, sizeof (stripe_local_t));
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         local->op_ret = -1;
         frame->local = local;
-        local->inode = fd->inode;
         local->call_count = priv->child_count;
-        
+
         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 (frame, -1, op_errno, NULL, NULL);
+err:
+        STRIPE_STACK_UNWIND (fsetattr, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 int32_t
 stripe_stack_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                         int32_t op_ret, int32_t op_errno, struct stat *buf,
-                         struct stat *preoldparent, struct stat *postoldparent,
-                         struct stat *prenewparent, struct stat *postnewparent)
+                         int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                         struct iatt *preoldparent, struct iatt *postoldparent,
+                         struct iatt *prenewparent, struct iatt *postnewparent,
+                         dict_t *xdata)
 {
         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);
@@ -1144,45 +925,38 @@ stripe_stack_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (op_ret == -1) {
                         gf_log (this->name, GF_LOG_DEBUG,
                                 "%s returned error %s",
-                                ((call_frame_t *)cookie)->this->name,
-                                strerror (op_errno));
+                                prev->this->name, strerror (op_errno));
                         local->op_errno = op_errno;
-                        if (op_errno == ENOTCONN)
+                        if ((op_errno != ENOENT) ||
+                            (prev->this == FIRST_CHILD (this)))
                                 local->failed = 1;
                 }
 
                 if (op_ret == 0) {
                         local->op_ret = 0;
- 
-                        local->post_buf.st_blocks += buf->st_blocks;
-                        if (local->post_buf.st_size < buf->st_size)
-                                local->post_buf.st_size =  buf->st_size;
-
-                        local->pre_parent_buf.st_blocks += 
-                                preoldparent->st_blocks;
-                        if (local->pre_parent_buf.st_size < 
-                            preoldparent->st_size)
-                                local->pre_parent_buf.st_size = 
-                                        preoldparent->st_size;
-                        local->post_parent_buf.st_blocks += 
-                                postoldparent->st_blocks;
-                        if (local->post_parent_buf.st_size < 
-                            postoldparent->st_size)
-                                local->post_parent_buf.st_size = 
-                                        postoldparent->st_size;
-                        
-                        local->pre_parent_buf.st_blocks += 
-                                prenewparent->st_blocks;
-                        if (local->pre_parent_buf.st_size < 
-                            prenewparent->st_size)
-                                local->pre_parent_buf.st_size = 
-                                        prenewparent->st_size;
-                        local->post_parent_buf.st_blocks += 
-                                postnewparent->st_blocks;
-                        if (local->post_parent_buf.st_size < 
-                            postnewparent->st_size)
-                                local->post_parent_buf.st_size = 
-                                        postnewparent->st_size;
+
+                        local->stbuf.ia_blocks      += buf->ia_blocks;
+                        local->preparent.ia_blocks  += preoldparent->ia_blocks;
+                        local->postparent.ia_blocks += postoldparent->ia_blocks;
+                        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;
+
+                        if (local->preparent.ia_size < preoldparent->ia_size)
+                                local->preparent.ia_size = preoldparent->ia_size;
+
+                        if (local->postparent.ia_size < postoldparent->ia_size)
+                                local->postparent.ia_size = postoldparent->ia_size;
+
+                        if (local->pre_buf.ia_size < prenewparent->ia_size)
+                                local->pre_buf.ia_size = prenewparent->ia_size;
+
+                        if (local->post_buf.ia_size < postnewparent->ia_size)
+                                local->post_buf.ia_size = postnewparent->ia_size;
                 }
         }
         UNLOCK (&frame->lock);
@@ -1191,42 +965,43 @@ stripe_stack_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 if (local->failed)
                         local->op_ret = -1;
 
-                if (local->loc.path)
-                        loc_wipe (&local->loc);
-                if (local->loc2.path)
-                        loc_wipe (&local->loc2);
-
-                STACK_UNWIND (frame, local->op_ret, local->op_errno,
-                              &local->post_buf, &local->pre_parent_buf,
-                              &local->post_parent_buf, 
-                              &local->pre_newparent_buf, 
-                              &local->post_newparent_buf);
+                STRIPE_STACK_UNWIND (rename, frame, local->op_ret, local->op_errno,
+                                     &local->stbuf, &local->preparent,
+                                     &local->postparent,  &local->pre_buf,
+                                     &local->post_buf, NULL);
         }
-
+out:
         return 0;
 }
 
-int32_t 
+int32_t
 stripe_first_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                         int32_t op_ret, int32_t op_errno, struct stat *buf,
-                         struct stat *preoldparent, struct stat *postoldparent,
-                         struct stat *prenewparent, struct stat *postnewparent)
+                         int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                         struct iatt *preoldparent, struct iatt *postoldparent,
+                         struct iatt *prenewparent, struct iatt *postnewparent,
+                         dict_t *xdata)
 {
         stripe_local_t *local = NULL;
         xlator_list_t  *trav = NULL;
 
-        local = frame->local;
-        trav = this->children;
+        if (!this || !frame || !frame->local) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                op_errno = EINVAL;
+                goto unwind;
+        }
 
         if (op_ret == -1) {
                 goto unwind;
         }
 
-        local->post_buf = *buf;
-        local->pre_parent_buf = *preoldparent;
-        local->post_parent_buf = *postoldparent;
-        local->pre_newparent_buf = *prenewparent;
-        local->post_newparent_buf = *postnewparent;
+        local = frame->local;
+        trav = this->children;
+
+        local->stbuf      = *buf;
+        local->preparent  = *preoldparent;
+        local->postparent = *postoldparent;
+        local->pre_buf    = *prenewparent;
+        local->post_buf   = *postnewparent;
 
         local->op_ret = 0;
         local->call_count--;
@@ -1235,32 +1010,26 @@ 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:
-        if (local->loc.path)
-                loc_wipe (&local->loc);
-        if (local->loc2.path)
-                loc_wipe (&local->loc2);
-
-       STACK_UNWIND (frame, op_ret, op_errno, buf, preoldparent,
-                      postoldparent, prenewparent, postnewparent);
-       return 0;
+unwind:
+        STRIPE_STACK_UNWIND (rename, frame, -1, op_errno, buf, preoldparent,
+                             postoldparent, prenewparent, postnewparent, NULL);
+        return 0;
 }
-/**
- * stripe_rename - 
- */
+
 int32_t
-stripe_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, 
-               loc_t *newloc)
+stripe_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+               loc_t *newloc, dict_t *xdata)
 {
         stripe_private_t *priv = NULL;
         stripe_local_t   *local = NULL;
         xlator_list_t    *trav = NULL;
-        int32_t           op_errno = 1;
+	stripe_fd_ctx_t	 *fctx = NULL;
+        int32_t           op_errno = EINVAL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -1279,100 +1048,132 @@ stripe_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
         }
 
         /* Initialization */
-        local = CALLOC (1, sizeof (stripe_local_t));
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
+
+        frame->local = local;
+
         local->op_ret = -1;
-        local->inode = oldloc->inode;
         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 (frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+err:
+        STRIPE_STACK_UNWIND (rename, frame, -1, op_errno, NULL, NULL, NULL,
+                             NULL, NULL, NULL);
         return 0;
 }
-
-/**
- * stripe_access - 
- */
 int32_t
-stripe_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask)
+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)
 {
-        int32_t op_errno = 1;
+        stripe_local_t *local   = NULL;
+        call_frame_t   *prev = NULL;
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->path, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
 
-        STACK_WIND (frame, stripe_common_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->access, loc, mask);
+        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;
- err:
-        STACK_UNWIND (frame, -1, op_errno);
+out:
+        STRIPE_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
+
         return 0;
 }
 
 
-/**
- * stripe_readlink_cbk - 
- */
-int32_t 
-stripe_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                     int32_t op_ret, int32_t op_errno, const char *path,
-                     struct stat *sbuf)
-{
-        STACK_UNWIND (frame, op_ret, op_errno, path, sbuf);
-        return 0;
-}
 
 
-/**
- * stripe_readlink - 
- */
 int32_t
-stripe_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size)
+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, dict_t *xdata)
 {
-        int32_t op_errno = 1;
+        int32_t         callcnt = 0;
+        stripe_local_t *local   = NULL;
+        call_frame_t   *prev = NULL;
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (loc, err);
-        VALIDATE_OR_GOTO (loc->path, err);
-        VALIDATE_OR_GOTO (loc->inode, err);
+        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;
 
-        STACK_WIND (frame, stripe_readlink_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->readlink, loc, size);
+                if (op_ret == -1) {
+                        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) {
+                                local->failed = 1;
+                                local->op_ret = op_ret;
+                        }
+                }
+        }
+        UNLOCK (&frame->lock);
 
+        if (callcnt == 1) {
+                if (local->failed) {
+                        op_errno = local->op_errno;
+                        goto out;
+                }
+                STACK_WIND(frame, stripe_first_unlink_cbk, FIRST_CHILD (this),
+                           FIRST_CHILD (this)->fops->unlink, &local->loc,
+                           local->xflag, local->xdata);
+        }
         return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno, NULL, NULL);
+out:
+        STRIPE_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
+
         return 0;
 }
 
-
-/**
- * stripe_unlink - 
- */
 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;
         stripe_private_t *priv = NULL;
-        int32_t           op_errno = 1;
+        int32_t           op_errno = EINVAL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -1387,7 +1188,7 @@ stripe_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
                 op_errno = ENOTCONN;
                 goto err;
         }
- 
+
         /* Don't unlink a file if a node is down */
         if (priv->nodes_down) {
                 op_errno = ENOTCONN;
@@ -1395,71 +1196,124 @@ stripe_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc)
         }
 
         /* Initialization */
-        local = CALLOC (1, sizeof (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_stack_unwind_unlink_cbk,
+                STACK_WIND (frame, stripe_unlink_cbk,
                             trav->xlator, trav->xlator->fops->unlink,
-                            loc);
+                            loc, xflag, xdata);
                 trav = trav->next;
         }
 
         return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno, NULL, NULL);
+err:
+        STRIPE_STACK_UNWIND (unlink, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 
-int32_t 
+int32_t
 stripe_first_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int32_t op_ret, int32_t op_errno,struct stat *preparent,
-                        struct stat *postparent)
-
+                        int32_t op_ret, int32_t op_errno,struct iatt *preparent,
+                        struct iatt *postparent, dict_t *xdata)
 {
-        xlator_list_t *trav = NULL;
         stripe_local_t *local = NULL;
 
+        if (!this || !frame || !frame->local) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                op_errno = EINVAL;
+                goto err;
+        }
+
         if (op_ret == -1) {
-                STACK_UNWIND (frame, op_ret, op_errno, NULL, NULL);
-                return 0;
+                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->pre_parent_buf  = *preparent;
-        local->post_parent_buf = *postparent;
-        
-        while (trav) {
-                STACK_WIND (frame, stripe_stack_unwind_unlink_cbk, trav->xlator,
-                            trav->xlator->fops->rmdir, &local->loc);
-                trav = trav->next;
+        local->preparent  = *preparent;
+        local->postparent = *postparent;
+        local->preparent_size  = preparent->ia_size;
+        local->postparent_size = postparent->ia_size;
+        local->preparent_blocks  += preparent->ia_blocks;
+        local->postparent_blocks += postparent->ia_blocks;
+
+        STRIPE_STACK_UNWIND (rmdir, frame, local->op_ret, local->op_errno,
+                             &local->preparent, &local->postparent, xdata);
+        return 0;
+err:
+        STRIPE_STACK_UNWIND (rmdir, frame, op_ret, op_errno, NULL, NULL, NULL);
+        return 0;
+
+}
+
+int32_t
+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;
 }
 
-/**
- * stripe_rmdir - 
- */
 int32_t
-stripe_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc)
+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;
         stripe_private_t *priv = NULL;
-        int32_t           op_errno = 1;
+        int32_t           op_errno = EINVAL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -1477,69 +1331,45 @@ stripe_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc)
         }
 
         /* Initialization */
-        local = CALLOC (1, sizeof (stripe_local_t));
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         local->op_ret = -1;
         frame->local = local;
-        local->inode = loc->inode;
         loc_copy (&local->loc, loc);
+        local->flags = flags;
         local->call_count = priv->child_count;
-  
-        STACK_WIND (frame, stripe_first_rmdir_cbk,  trav->xlator,
-                    trav->xlator->fops->rmdir, loc);
-
-        return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno, NULL, NULL);
-        return 0;
-}
-
-
-/**
- * stripe_setxattr - 
- */
-int32_t
-stripe_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                 dict_t *dict, int32_t flags)
-{
-        stripe_private_t *priv = NULL;
-        int32_t           op_errno = 1;
+        trav = trav->next; /* skip the first child */
 
-        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;
-
-        if (priv->first_child_down) {
-                op_errno = ENOTCONN;
-                goto err;
+        while (trav) {
+                STACK_WIND (frame, stripe_rmdir_cbk,  trav->xlator,
+                            trav->xlator->fops->rmdir, loc, flags, NULL);
+                trav = trav->next;
         }
 
-        STACK_WIND (frame, stripe_common_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->setxattr, loc, dict, flags);
-
         return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno);
+err:
+        STRIPE_STACK_UNWIND (rmdir, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 
-int32_t 
+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 stat *preparent,
-                                    struct stat *postparent)
+                                    int32_t op_errno, struct iatt *preparent,
+                                    struct iatt *postparent, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
 
+        if (!this || !frame || !frame->local) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+
         local = frame->local;
 
         LOCK (&frame->lock);
@@ -1549,12 +1379,11 @@ stripe_mknod_ifreg_fail_unlink_cbk (call_frame_t *frame, void *cookie,
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                loc_wipe (&local->loc);
-                STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-                              local->inode, &local->post_buf,
-                              &local->pre_parent_buf, &local->post_parent_buf);
+                STRIPE_STACK_UNWIND (mknod, frame, local->op_ret, local->op_errno,
+                                     local->inode, &local->stbuf,
+                                     &local->preparent, &local->postparent, NULL);
         }
-
+out:
         return 0;
 }
 
@@ -1564,25 +1393,31 @@ stripe_mknod_ifreg_fail_unlink_cbk (call_frame_t *frame, void *cookie,
 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;
         stripe_private_t *priv = NULL;
         xlator_list_t    *trav = NULL;
+        call_frame_t     *prev = NULL;
 
-        priv = this->private;
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+
+        prev  = 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",
-                                ((call_frame_t *)cookie)->this->name, 
-                                strerror (op_errno));
+                                prev->this->name, strerror (op_errno));
                         local->op_ret = -1;
                         local->op_errno = op_errno;
                 }
@@ -1597,157 +1432,239 @@ 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;
                 }
 
-                loc_wipe (&local->loc);
-                STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-                              local->inode, &local->post_buf,
-                              &local->pre_parent_buf, &local->post_parent_buf);
+                STRIPE_STACK_UNWIND (mknod, frame, local->op_ret, local->op_errno,
+                                     local->inode, &local->stbuf,
+                                     &local->preparent, &local->postparent, NULL);
         }
+out:
         return 0;
 }
 
-/**
- */
 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 stat *buf, struct stat *preparent,
-                        struct stat *postparent)
+                        struct iatt *buf, struct iatt *preparent,
+                        struct iatt *postparent, dict_t *xdata)
 {
-        int               ret = 0;
         int32_t           callcnt = 0;
         stripe_local_t   *local = NULL;
-        xlator_list_t    *trav = NULL;
         stripe_private_t *priv = NULL;
+        call_frame_t     *prev = NULL;
+        xlator_list_t    *trav = NULL;
 
-	priv = this->private;
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+
+        prev  = 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, 
+                        gf_log (this->name, GF_LOG_DEBUG,
                                 "%s returned error %s",
-                                ((call_frame_t *)cookie)->this->name, 
-                                strerror (op_errno));
-                        local->failed = 1;
+                                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 (local->post_buf.st_blksize == 0) {
-                                local->post_buf = *buf;
-                                local->pre_parent_buf = *preparent;
-                                local->post_parent_buf = *postparent;
-                                /* Because st_blocks gets added again */
-                                local->post_buf.st_blocks = 0;
-                                local->pre_buf.st_blocks = 0;
-                                local->post_parent_buf.st_blocks = 0;
-                        }
 
-                        /* Always, pass the inode number of first child 
-                           to the above layer */
-                        if (FIRST_CHILD(this) == 
-                            ((call_frame_t *)cookie)->this) {
-                                local->post_buf.st_ino = buf->st_ino;
-                                local->pre_parent_buf.st_ino = preparent->st_ino;
-                                local->post_parent_buf.st_ino = postparent->st_ino;
-                        }
-      
-                        local->post_buf.st_blocks += buf->st_blocks;
-                        if (local->post_buf.st_size < buf->st_size)
-                                local->post_buf.st_size = buf->st_size;
+                        /* 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);
 
-                        local->pre_parent_buf.st_blocks += preparent->st_blocks;
-                        if (local->pre_parent_buf.st_size < preparent->st_size)
-                                local->pre_parent_buf.st_size = preparent->st_size;
+			if (stripe_ctx_handle(this, prev, local, xdata))
+				gf_log(this->name, GF_LOG_ERROR,
+					"Error getting fctx info from dict");
 
-                        local->post_parent_buf.st_blocks += postparent->st_blocks;
-                        if (local->post_parent_buf.st_size < postparent->st_size)
-                                local->post_parent_buf.st_size = postparent->st_size;
+                        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) 
+                if (local->failed)
                         local->op_ret = -1;
 
-                if ((local->op_ret != -1) && priv->xattr_supported) {
-                        /* Send a setxattr request to nodes where the
-                           files are created */
-                        int32_t index = 0;
-                        char size_key[256] = {0,};
-                        char index_key[256] = {0,};
-                        char count_key[256] = {0,};
-                        dict_t *dict = NULL;
-
-                        trav = this->children;
-                        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);
-
+                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) {
-                                dict = get_new_dict ();
-                                dict_ref (dict);
-                                /* TODO: check return value */
-                                ret = dict_set_int64 (dict, size_key, 
-                                                      local->stripe_size);
-                                ret = dict_set_int32 (dict, count_key, 
-                                                      priv->child_count);
-                                ret = dict_set_int32 (dict, index_key, index);
-
                                 STACK_WIND (frame,
-                                            stripe_mknod_ifreg_setxattr_cbk,
+                                            stripe_mknod_ifreg_fail_unlink_cbk,
                                             trav->xlator,
-                                            trav->xlator->fops->setxattr,
-                                            &local->loc, dict, 0);
-        
-                                dict_unref (dict);
-                                index++;
+                                            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;
+                        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;
+                        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;
+}
+
+
+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;
+
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+
+        prev  = cookie;
+        priv  = this->private;
+        local = frame->local;
+        trav = this->children;
+
+        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->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);
+                        }
+                        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 {
-                        /* Create itself has failed.. so return 
-                           without setxattring */
-                        loc_wipe (&local->loc);
-                        STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-                                      local->inode, &local->post_buf, 
-                                      &local->pre_parent_buf, 
-                                      &local->post_parent_buf);
+                        dict = local->xattr;
                 }
+
+                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);
         }
-  
+
         return 0;
+
+out:
+
+       STRIPE_STACK_UNWIND (mknod, frame, op_ret, op_errno, NULL, NULL, NULL, NULL, NULL);
+       return 0;
 }
 
 
-/**
- * stripe_mknod - 
- */
 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, dict_t *xdata)
+{
+        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)
+              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 = 1;
+        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);
@@ -1756,71 +1673,216 @@ 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;
                 goto err;
         }
 
         if (S_ISREG(mode)) {
-                /* NOTE: on older kernels (older than 2.6.9), 
-                   creat() fops is sent as mknod() + open(). Hence handling 
+                /* NOTE: on older kernels (older than 2.6.9),
+                   creat() fops is sent as mknod() + open(). Hence handling
                    S_IFREG files is necessary */
                 if (priv->nodes_down) {
-                        gf_log (this->name, GF_LOG_WARNING, 
+                        gf_log (this->name, GF_LOG_WARNING,
                                 "Some node down, returning EIO");
                         op_errno = EIO;
                         goto err;
                 }
-                
+
                 /* Initialization */
-                local = CALLOC (1, sizeof (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);
-                        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. */
-                return 0; 
-        }
+                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_common_inode_cbk,
+        STACK_WIND (frame, stripe_single_mknod_cbk,
                     FIRST_CHILD(this), FIRST_CHILD(this)->fops->mknod,
-                    loc, mode, rdev);
+                    loc, mode, rdev, umask, xdata);
 
         return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno, NULL, NULL, NULL, NULL);
+err:
+        STRIPE_STACK_UNWIND (mknod, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
+        return 0;
+}
+
+
+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, 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 error %s",
+                                prev->this->name, strerror (op_errno));
+                        local->op_errno = op_errno;
+                        if ((op_errno != ENOENT) ||
+                            (prev->this == FIRST_CHILD (this)))
+                                local->failed = 1;
+                }
+
+                if (op_ret >= 0) {
+                        local->op_ret = 0;
+
+                        local->stbuf_blocks      += buf->ia_blocks;
+                        local->preparent_blocks  += preparent->ia_blocks;
+                        local->postparent_blocks += postparent->ia_blocks;
+
+                        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 != -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;
+                }
+                STRIPE_STACK_UNWIND (mkdir, frame, local->op_ret,
+                                     local->op_errno, local->inode,
+                                     &local->stbuf, &local->preparent,
+                                     &local->postparent, NULL);
+        }
+out:
         return 0;
 }
 
 
-/**
- * stripe_mkdir - 
- */
 int32_t
-stripe_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode)
+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,
+              mode_t umask, dict_t *xdata)
 {
         stripe_private_t *priv = NULL;
         stripe_local_t   *local = NULL;
@@ -1835,72 +1897,130 @@ stripe_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode)
 
         priv = this->private;
         trav = this->children;
-  
+
         if (priv->first_child_down) {
                 op_errno = ENOTCONN;
                 goto err;
         }
 
         /* Initialization */
-        local = CALLOC (1, sizeof (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_stack_unwind_inode_cbk,
-                            trav->xlator, trav->xlator->fops->mkdir,
-                            loc, mode);
-                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 (frame, -1, op_errno, NULL, NULL, NULL, NULL);
+err:
+        STRIPE_STACK_UNWIND (mkdir, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
         return 0;
 }
 
 
-/**
- * stripe_symlink - 
- */
 int32_t
-stripe_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
-                loc_t *loc)
+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, dict_t *xdata)
 {
-        int32_t           op_errno = 1;
-        stripe_private_t *priv = NULL;
-  
-        priv = this->private;
+        int32_t         callcnt = 0;
+        stripe_local_t  *local   = NULL;
+        call_frame_t    *prev = NULL;
+	stripe_fd_ctx_t *fctx = NULL;
 
-        if (priv->first_child_down) {
-                op_errno = ENOTCONN;
-                goto err;
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
         }
 
-        /* send symlink to only first node */
-        STACK_WIND (frame, stripe_common_inode_cbk, FIRST_CHILD(this), 
-                    FIRST_CHILD(this)->fops->symlink, linkpath, loc);
+        prev  = cookie;
+        local = frame->local;
 
-        return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno, NULL, NULL, NULL, NULL);
+        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_errno = op_errno;
+                        if ((op_errno != ENOENT) ||
+                            (prev->this == FIRST_CHILD (this)))
+                                local->failed = 1;
+                }
+
+                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;
+                                local->postparent = *postparent;
+                                local->preparent  = *preparent;
+                        }
+                        local->stbuf_blocks      += buf->ia_blocks;
+                        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)
+                                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;
+                        local->stbuf.ia_size        = local->stbuf_size;
+                        local->stbuf.ia_blocks      = local->stbuf_blocks;
+                }
+                STRIPE_STACK_UNWIND (link, frame, local->op_ret,
+                                     local->op_errno, local->inode,
+                                     &local->stbuf, &local->preparent,
+                                     &local->postparent, NULL);
+        }
+out:
         return 0;
 }
 
-/**
- * stripe_link -
- */
 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)
 {
-        int               send_fop_to_all = 0;
         xlator_list_t    *trav = NULL;
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
@@ -1921,50 +2041,45 @@ stripe_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc)
                 goto err;
         }
 
-        if (S_ISREG (oldloc->inode->st_mode))
-                send_fop_to_all = 1;
+        /* Initialization */
+        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 (!send_fop_to_all) {
-                STACK_WIND (frame, stripe_common_inode_cbk,
+        /* Everytime in stripe lookup, all child
+           nodes should be looked up */
+        while (trav) {
+                STACK_WIND (frame, stripe_link_cbk,
                             trav->xlator, trav->xlator->fops->link,
-                            oldloc, newloc);
-        } else {
-                /* Initialization */
-                local = CALLOC (1, sizeof (stripe_local_t));
-                if (!local) {
-                        op_errno = ENOMEM;
-                        goto err;
-                }
-                local->op_ret = -1;
-                frame->local = local;
-                local->call_count = priv->child_count;
-
-                /* Everytime in stripe lookup, all child
-                   nodes should be looked up */
-                while (trav) {
-                        STACK_WIND (frame, stripe_stack_unwind_inode_cbk,
-                                    trav->xlator, trav->xlator->fops->link,
-                                    oldloc, newloc);
-                        trav = trav->next;
-                }
+                            oldloc, newloc, NULL);
+                trav = trav->next;
         }
 
         return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno, NULL, NULL, NULL, NULL);
+err:
+        STRIPE_STACK_UNWIND (link, frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
         return 0;
 }
 
-int32_t 
+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 stat *preparent,
-                               struct stat *postparent)
+                               int32_t op_errno, struct iatt *preparent,
+                               struct iatt *postparent, dict_t *xdata)
 {
         int32_t         callcnt = 0;
-        fd_t           *lfd = NULL;
         stripe_local_t *local = NULL;
 
+        if (!this || !frame || !frame->local) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+
         local = frame->local;
 
         LOCK (&frame->lock);
@@ -1974,49 +2089,78 @@ stripe_create_fail_unlink_cbk (call_frame_t *frame, void *cookie,
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                lfd = local->fd;
-                loc_wipe (&local->loc);
-                STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-                              local->fd, local->inode, &local->post_buf,
-                              &local->pre_parent_buf, &local->post_parent_buf);
-                fd_unref (lfd);
+                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;
 }
 
 
-/**
- * stripe_create_setxattr_cbk - 
- */
 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)
 {
-        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;
 
-	priv = this->private;        
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+
+        prev  = 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",
-                                ((call_frame_t *)cookie)->this->name, 
-                                strerror (op_errno));
-                        local->op_ret = -1;
+                                prev->this->name, strerror (op_errno));
+                        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;
@@ -2025,190 +2169,180 @@ 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;
-                loc_wipe (&local->loc);
-                STACK_UNWIND (frame, local->op_ret, local->op_errno,
-                              local->fd, local->inode, &local->post_buf,
-                              &local->pre_parent_buf, &local->post_parent_buf);
-                fd_unref (lfd);
+                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_copy_xl_array(local->fctx->xl_array,
+				             priv->xl_array,
+					     local->fctx->stripe_count);
+			inode_ctx_put(local->inode, this,
+					(uint64_t) local->fctx);
+                }
+
+                /* 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;
 }
 
-/**
- * stripe_create_cbk - 
- */
+
+
 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 stat *buf, struct stat *preparent,
-                   struct stat *postparent)
+                   inode_t *inode, struct iatt *buf, struct iatt *preparent,
+                   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;
+        call_frame_t     *prev = NULL;
+        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");
+                goto out;
+        }
 
+        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",
-                                ((call_frame_t *)cookie)->this->name, 
-                                strerror (op_errno));
-                        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 (!local->post_buf.st_blksize) {
-                                local->post_buf = *buf;
-                                local->post_buf.st_blocks = 0;
-                        }
-                        if (!local->pre_parent_buf.st_blksize) {
-                                local->pre_parent_buf = *preparent;
-                                local->pre_parent_buf.st_blocks = 0;
-                        }
-                        if (!local->post_parent_buf.st_blksize) {
-                                local->post_parent_buf = *postparent;
-                                local->post_parent_buf.st_blocks = 0;
-                        }
-                        if (FIRST_CHILD(this) ==
-                            ((call_frame_t *)cookie)->this) {
-                                local->post_buf.st_ino   = buf->st_ino;
-                                local->post_buf.st_mtime = buf->st_mtime;
-                                local->post_buf.st_ctime = buf->st_ctime;
-
-                                local->pre_parent_buf.st_ino = preparent->st_ino;
-                                local->pre_parent_buf.st_mtime = preparent->st_mtime;
-                                local->pre_parent_buf.st_ctime = preparent->st_ctime;
-                                local->post_parent_buf.st_ino = postparent->st_ino;
-                                local->post_parent_buf.st_mtime = postparent->st_mtime;
-                                local->post_parent_buf.st_ctime = postparent->st_ctime;
-                        }
-                        local->post_buf.st_blocks += buf->st_blocks;
-                        if (local->post_buf.st_size < buf->st_size)
-                                local->post_buf.st_size =  buf->st_size;
-
-                        local->pre_parent_buf.st_blocks += preparent->st_blocks;
-                        if (local->pre_parent_buf.st_size < 
-                            preparent->st_size)
-                                local->pre_parent_buf.st_size = 
-                                        preparent->st_size;
-                        local->post_parent_buf.st_blocks += 
-                                postparent->st_blocks;
-                        if (local->post_parent_buf.st_size < 
-                            postparent->st_size)
-                                local->post_parent_buf.st_size = 
-                                        postparent->st_size;
-                }
+        --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->failed = 1;
+                 local->op_errno = op_errno;
         }
-        UNLOCK (&frame->lock);
 
-        if (!callcnt) {
-                if (local->failed)
-                        local->op_ret = -1;
+        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->stbuf_blocks += buf->ia_blocks;
+        local->preparent_blocks  += preparent->ia_blocks;
+        local->postparent_blocks += postparent->ia_blocks;
+
+        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 (local->failed)
+                local->op_ret = -1;
 
-                /* */
-                if (local->op_ret >= 0) {
-                        fctx = CALLOC (1, sizeof (stripe_fd_ctx_t));
-                        if (fctx) {
-                                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->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;
+        }
 
-                if ((local->op_ret != -1) && 
-                    local->stripe_size && priv->xattr_supported) {
-                        /* Send a setxattr request to nodes where
-                           the files are created */
-                        int            ret = 0;
-                        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);
+        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;
+        }
 
-                        local->call_count = priv->child_count;
-        
-                        for (i = 0; i < priv->child_count; i++) {
-                                dict = get_new_dict ();
-                                dict_ref (dict);
-
-                                /* TODO: check return values */
-                                ret = dict_set_int64 (dict, size_key, 
-                                                      local->stripe_size);
-                                ret = dict_set_int32 (dict, count_key, 
-                                                      priv->child_count);
-                                ret = dict_set_int32 (dict, index_key, i);
-        
-                                STACK_WIND (frame, stripe_create_setxattr_cbk,
-                                            priv->xl_array[i],
-                                            priv->xl_array[i]->fops->setxattr,
-                                            &local->loc, dict, 0);
-        
-                                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);
                         }
+                        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 {
-                        /* Create itself has failed.. so return
-                           without setxattring */
-                        lfd = local->fd;
-                        loc_wipe (&local->loc);
-                        STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-                                      local->fd, local->inode, &local->post_buf,
-                                      &local->pre_parent_buf, 
-                                      &local->post_parent_buf);
-      
-                        fd_unref (lfd);
+                        dict = local->xattr;
                 }
+
+                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:
         return 0;
 }
 
 
+
 /**
- * stripe_create - If a block-size is specified for the 'name', create the 
+ * 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.
  *
  * @name- complete path of the file to be created.
  */
 int32_t
 stripe_create (call_frame_t *frame, xlator_t *this, loc_t *loc,
-               int32_t flags, mode_t mode, fd_t *fd)
+               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 = 1;
+        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);
+        VALIDATE_OR_GOTO (loc, err);
+        VALIDATE_OR_GOTO (loc->path, err);
+        VALIDATE_OR_GOTO (loc->inode, err);
 
         priv = this->private;
 
@@ -2216,121 +2350,86 @@ stripe_create (call_frame_t *frame, xlator_t *this, loc_t *loc,
         flags &= ~O_APPEND;
 
         if (priv->first_child_down || priv->nodes_down) {
-                gf_log (this->name, GF_LOG_DEBUG, 
+                gf_log (this->name, GF_LOG_DEBUG,
                         "First node down, returning EIO");
                 op_errno = EIO;
                 goto err;
         }
 
         /* Initialization */
-        local = CALLOC (1, sizeof (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->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;
-        
-        trav = this->children;
-        while (trav) {
-                STACK_WIND (frame, stripe_create_cbk, trav->xlator,
-                            trav->xlator->fops->create, loc, flags, mode, fd);
-                trav = trav->next;
-        }
-       
-        return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno, NULL, NULL, NULL, NULL, NULL);
-        return 0;
-}
+        /* Send a setxattr request to nodes where the
+           files are created */
 
-/**
- * stripe_open_cbk - 
- */
-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         callcnt = 0;
-        stripe_local_t *local = NULL;
-        fd_t           *lfd = NULL;
-
-        local = frame->local;
+        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;
 
-        LOCK (&frame->lock);
-        {
-                callcnt = --local->call_count;
+                dict_copy (xdata, dict);
 
-                if (op_ret == -1) {
-                        local->failed = 1;
-                        gf_log (this->name, GF_LOG_DEBUG, 
-                                "%s returned error %s",
-                                ((call_frame_t *)cookie)->this->name, 
-                                strerror (op_errno));
-                        local->op_ret = -1;
-                        local->op_errno = op_errno;
-                }
-    
-                if (op_ret >= 0)
-                        local->op_ret = op_ret;
+                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;
         }
-        UNLOCK (&frame->lock);
-  
-        if (!callcnt) {
-                if (local->failed)
-                        local->op_ret = -1;
 
-                if (local->op_ret == -1) {
-                        if (local->fctx) {
-                                if (!local->fctx->static_array)
-                                        FREE (local->fctx->xl_array);
-                                FREE (local->fctx);
-                        }
-                } else {
-                        fd_ctx_set (local->fd, this, 
-                                    (uint64_t)(long)local->fctx);
-                }
 
-                lfd = local->fd;
-                loc_wipe (&local->loc);
-                STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-                              local->fd);
-                fd_unref (lfd);
+        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:
+        STRIPE_STACK_UNWIND (create, frame, -1, op_errno, NULL, NULL, NULL,
+                             NULL, NULL, xdata);
+        return 0;
 }
 
-
-/**
- * stripe_getxattr_cbk - 
- */
 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)
+stripe_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
-        int32_t           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;
+        int32_t         callcnt = 0;
+        stripe_local_t *local = NULL;
+        call_frame_t   *prev = NULL;
 
-        prev  = (call_frame_t *)cookie;
-        priv  = this->private;
+        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);
@@ -2338,145 +2437,39 @@ stripe_open_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
                 callcnt = --local->call_count;
 
                 if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG, 
+
+                        gf_log (this->name, GF_LOG_DEBUG,
                                 "%s returned error %s",
-                                ((call_frame_t *)cookie)->this->name, 
-                                strerror (op_errno));
-                        local->op_ret = -1;
-                        if (local->op_errno != EIO)
-                                local->op_errno = op_errno;
-                        if (op_errno == ENOTCONN)
+                                prev->this->name, strerror (op_errno));
+                        if ((op_errno != ENOENT) ||
+                            (prev->this == FIRST_CHILD (this)))
                                 local->failed = 1;
-                        goto unlock;
-                }
-
-                if (!local->fctx) {
-                        local->fctx =  CALLOC (1, sizeof (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_DEBUG,
-                                        "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 = 
-                                CALLOC (local->fctx->stripe_count, 
-                                        sizeof (xlator_t *));
-                }
-                if (local->fctx->stripe_count != data_to_int32 (data)) {
-                        gf_log (this->name, GF_LOG_ERROR, 
-                                "error with stripe-count xattr");
-                        local->op_ret   = -1;
-                        local->op_errno = EIO;
-                        goto unlock;
+                        local->op_errno = op_errno;
                 }
 
-                /* 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");
-                        local->op_ret   = -1;
-                        local->op_errno = EIO;
-                        goto unlock;
-                }
-                if (local->fctx->xl_array)
-                        local->fctx->xl_array[index] = prev->this;
-                local->entry_count++;
-                local->op_ret = 0;
+                if (op_ret >= 0)
+                        local->op_ret = op_ret;
         }
- 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->op_ret)
-                        goto err;
 
-                if (local->entry_count != local->fctx->stripe_count) {
-                        local->op_ret = -1;
-                        local->op_errno = EIO;
-                        goto err;
-                }
-                if (!local->fctx->stripe_size) {
+        if (!callcnt) {
+                if (local->failed)
                         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;
-                }
+                STRIPE_STACK_UNWIND (open, frame, local->op_ret,
+                                     local->op_errno, local->fd, xdata);
         }
-
-        return 0;
- err:
-        lfd = local->fd;
-        loc_wipe (&local->loc);
-        STACK_UNWIND (frame, local->op_ret, local->op_errno, local->fd);
-        fd_unref (lfd);
-        
+out:
         return 0;
 }
 
+
 /**
- * stripe_open - 
+ * 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;
@@ -2498,7 +2491,7 @@ stripe_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
         }
 
         /* Initialization */
-        local = CALLOC (1, sizeof (stripe_local_t));
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -2509,98 +2502,76 @@ stripe_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
 
         local->fd = fd_ref (fd);
         frame->local = local;
-        local->inode = loc->inode;
         loc_copy (&local->loc, 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;
-                }
-        } else {
-                local->fctx =  CALLOC (1, sizeof (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;
-                
-                while (trav) {
-                        STACK_WIND (frame, stripe_open_cbk, trav->xlator,
-                                    trav->xlator->fops->open,
-                                    &local->loc, local->flags, local->fd,
-                                    wbflags);
-                        trav = trav->next;
-                }
-        }
+        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,
+                            xdata);
+                trav = trav->next;
+        }
         return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno, NULL);
+err:
+        STRIPE_STACK_UNWIND (open, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
-/**
- * stripe_opendir_cbk - 
- */
+
 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 = frame->local;
+        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, 
+                        gf_log (this->name, GF_LOG_DEBUG,
                                 "%s returned error %s",
-                                ((call_frame_t *)cookie)->this->name, 
-                                strerror (op_errno));
+                                prev->this->name, strerror (op_errno));
                         local->op_ret = -1;
-                        local->failed = 1;
                         local->op_errno = op_errno;
                 }
-    
-                if (op_ret >= 0) 
+
+                if (op_ret >= 0)
                         local->op_ret = op_ret;
         }
         UNLOCK (&frame->lock);
 
         if (!callcnt) {
-                STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-                              local->fd);
+                STRIPE_STACK_UNWIND (opendir, frame, local->op_ret,
+                                     local->op_errno, local->fd, NULL);
         }
-
+out:
         return 0;
 }
 
 
-/**
- * stripe_opendir - 
- */
 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;
         stripe_private_t *priv = NULL;
-        int32_t           op_errno = 1;
+        int32_t           op_errno = EINVAL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -2617,118 +2588,61 @@ stripe_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd)
         }
 
         /* Initialization */
-        local = CALLOC (1, sizeof (stripe_local_t));
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         frame->local = local;
-        local->inode = loc->inode;
-        local->fd = fd;
         local->call_count = priv->child_count;
+        local->fd = fd_ref (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 (frame, -1, op_errno, NULL);
-        return 0;
-}
-
-
-/**
- * stripe_getxattr_cbk - 
- */
-int32_t
-stripe_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                     int32_t op_ret, int32_t op_errno, dict_t *value)
-{
-        STACK_UNWIND (frame, op_ret, op_errno, value);
-        return 0;
-}
-
-
-/**
- * stripe_getxattr - 
- */
-int32_t
-stripe_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                 const char *name)
-{
-        int32_t op_errno = 1;
-
-        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);
-
-        STACK_WIND (frame, stripe_getxattr_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->getxattr, loc, name);
-
-        return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno, NULL);
-        return 0;
-}
-
-/**
- * stripe_removexattr - 
- */
-int32_t
-stripe_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
-                    const char *name)
-{
-        int32_t op_errno = 1;
-
-        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);
-
-        STACK_WIND (frame, stripe_common_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->removexattr, loc, name);
 
         return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno);
+err:
+        STRIPE_STACK_UNWIND (opendir, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
-
-/**
- * stripe_lk_cbk - 
- */
 int32_t
 stripe_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-               int32_t op_ret, int32_t op_errno, struct flock *lock)
+               int32_t op_ret, int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
 {
         int32_t         callcnt = 0;
         stripe_local_t *local = NULL;
+        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, 
+                        gf_log (this->name, GF_LOG_DEBUG,
                                 "%s returned error %s",
-                                ((call_frame_t *)cookie)->this->name, 
-                                strerror (op_errno));
+                                prev->this->name, strerror (op_errno));
                         local->op_errno = op_errno;
-                        if (op_errno == ENOTCONN)
+                        if ((op_errno != ENOENT) ||
+                            (prev->this == FIRST_CHILD (this)))
                                 local->failed = 1;
                 }
-                if (op_ret == 0 && local->op_ret == -1) {
-                        /* First successful call, copy the *lock */
-                        local->op_ret = 0;
-                        local->lock = *lock;
+                if (op_ret >= 0) {
+                        if (FIRST_CHILD(this) == prev->this) {
+                                /* First successful call, copy the *lock */
+                                local->op_ret = op_ret;
+                                local->lock = *lock;
+                        }
                 }
         }
         UNLOCK (&frame->lock);
@@ -2736,24 +2650,21 @@ stripe_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         if (!callcnt) {
                 if (local->failed)
                         local->op_ret = -1;
-                STACK_UNWIND (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;
 }
 
-
-/**
- * stripe_lk - 
- */
 int32_t
 stripe_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
-           struct flock *lock)
+           struct gf_flock *lock, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         xlator_list_t    *trav = NULL;
         stripe_private_t *priv = NULL;
-        int32_t           op_errno = 1;
+        int32_t           op_errno = EINVAL;
 
         VALIDATE_OR_GOTO (frame, err);
         VALIDATE_OR_GOTO (this, err);
@@ -2764,77 +2675,75 @@ stripe_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
         priv = this->private;
 
         /* Initialization */
-        local = CALLOC (1, sizeof (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;
-        
+
         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 (frame, -1, op_errno, NULL);
+err:
+        STRIPE_STACK_UNWIND (lk, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
-/**
- * stripe_writedir - 
- */
+
 int32_t
-stripe_setdents (call_frame_t *frame, xlator_t *this, fd_t *fd,
-                 int32_t flags, dir_entry_t *entries, int32_t count)
+stripe_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        stripe_local_t   *local = NULL;
-        stripe_private_t *priv = NULL;
-        xlator_list_t    *trav = NULL;
-        int32_t           op_errno = 1;
+        int32_t         callcnt = 0;
+        stripe_local_t *local   = NULL;
+        call_frame_t   *prev = NULL;
 
-        VALIDATE_OR_GOTO (frame, err);
-        VALIDATE_OR_GOTO (this, err);
-        VALIDATE_OR_GOTO (fd, err);
-        VALIDATE_OR_GOTO (fd->inode, err);
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
 
-        priv = this->private;
-        trav = this->children;
+        prev  = cookie;
+        local = frame->local;
 
-        /* Initialization */
-        local = CALLOC (1, sizeof (stripe_local_t));
-        if (!local) {
-                op_errno = ENOMEM;
-                goto err;
-        }
-        local->op_ret = -1;
-        frame->local = local;
-        local->call_count = priv->child_count;
+        LOCK (&frame->lock);
+        {
+                callcnt = --local->call_count;
 
-        while (trav) {
-                STACK_WIND (frame, stripe_stack_unwind_cbk, trav->xlator,
-                            trav->xlator->fops->setdents, fd, flags, entries, 
-                            count);
-                trav = trav->next;
+                if (op_ret == -1) {
+                        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)))
+                                local->failed = 1;
+                }
+                if (op_ret >= 0)
+                        local->op_ret = op_ret;
         }
+        UNLOCK (&frame->lock);
 
-        return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno);
+        if (!callcnt) {
+                if (local->failed)
+                        local->op_ret = -1;
+
+                STRIPE_STACK_UNWIND (flush, frame, local->op_ret,
+                                     local->op_errno, NULL);
+        }
+out:
         return 0;
 }
 
-
-/**
- * stripe_flush - 
- */
 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;
@@ -2854,7 +2763,7 @@ stripe_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
                 goto err;
         }
         /* Initialization */
-        local = CALLOC (1, sizeof (stripe_local_t));
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -2862,29 +2771,98 @@ stripe_flush (call_frame_t *frame, xlator_t *this, fd_t *fd)
         local->op_ret = -1;
         frame->local = local;
         local->call_count = priv->child_count;
-        
+
         while (trav) {
-                STACK_WIND (frame, stripe_stack_unwind_cbk,  trav->xlator,
-                            trav->xlator->fops->flush, fd);
+                STACK_WIND (frame, stripe_flush_cbk,  trav->xlator,
+                            trav->xlator->fops->flush, fd, NULL);
                 trav = trav->next;
         }
 
         return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno);
+err:
+        STRIPE_STACK_UNWIND (flush, frame, -1, op_errno, NULL);
         return 0;
 }
 
 
-/**
- * stripe_fsync - 
- */
+
 int32_t
-stripe_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
+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, 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));
+                        local->op_errno = op_errno;
+                        if ((op_errno != ENOENT) ||
+                            (prev->this == FIRST_CHILD (this)))
+                                local->failed = 1;
+                }
+                if (op_ret >= 0) {
+                        local->op_ret = op_ret;
+                        if (FIRST_CHILD(this) == prev->this) {
+                                local->pre_buf  = *prebuf;
+                                local->post_buf = *postbuf;
+                        }
+                        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;
+
+                        if (local->postbuf_size < postbuf->ia_size)
+                                local->postbuf_size = postbuf->ia_size;
+                }
+        }
+        UNLOCK (&frame->lock);
+
+        if (!callcnt) {
+                if (local->failed)
+                        local->op_ret = -1;
+
+                if (local->op_ret != -1) {
+                        local->pre_buf.ia_blocks  = local->prebuf_blocks;
+                        local->pre_buf.ia_size    = local->prebuf_size;
+                        local->post_buf.ia_blocks = local->postbuf_blocks;
+                        local->post_buf.ia_size   = local->postbuf_size;
+                }
+
+                STRIPE_STACK_UNWIND (fsync, frame, local->op_ret,
+                                     local->op_errno, &local->pre_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, 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);
@@ -2896,39 +2874,107 @@ stripe_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
         trav = this->children;
 
         /* Initialization */
-        local = CALLOC (1, sizeof (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 (frame, -1, op_errno, NULL, NULL);
+err:
+        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, 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 error %s",
+                                prev->this->name, strerror (op_errno));
+                        local->op_errno = op_errno;
+                        if ((op_errno != ENOENT) ||
+                            (prev->this == FIRST_CHILD (this)))
+                                local->failed = 1;
+                }
+
+                if (op_ret == 0) {
+                        local->op_ret = 0;
+
+                        if (FIRST_CHILD(this) == prev->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;
+                }
+        }
+        UNLOCK (&frame->lock);
+
+        if (!callcnt) {
+                if (local->failed)
+                        local->op_ret = -1;
+
+                if (local->op_ret != -1) {
+                        local->stbuf.ia_size   = local->stbuf_size;
+                        local->stbuf.ia_blocks = local->stbuf_blocks;
+                }
+
+                STRIPE_STACK_UNWIND (fstat, frame, local->op_ret,
+                                     local->op_errno, &local->stbuf, NULL);
+        }
+
+out:
+        return 0;
+}
 
-/**
- * stripe_fstat - 
- */
 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);
@@ -2940,39 +2986,44 @@ stripe_fstat (call_frame_t *frame,
         trav = this->children;
 
         /* Initialization */
-        local = CALLOC (1, sizeof (stripe_local_t));
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         local->op_ret = -1;
         frame->local = local;
-        local->inode = fd->inode;
         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_stack_unwind_buf_cbk, trav->xlator,
-                            trav->xlator->fops->fstat, fd);
+                STACK_WIND (frame, stripe_fstat_cbk, trav->xlator,
+                            trav->xlator->fops->fstat, fd, NULL);
                 trav = trav->next;
         }
 
         return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno, NULL);
+err:
+        STRIPE_STACK_UNWIND (fstat, frame, -1, op_errno, NULL, NULL);
         return 0;
 }
 
 
-/**
- * stripe_ftruncate - 
- */
 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);
@@ -2980,37 +3031,115 @@ 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 = CALLOC (1, sizeof (stripe_local_t));
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
         local->op_ret = -1;
         frame->local = local;
-        local->inode = fd->inode;
         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 (frame, -1, op_errno);
+err:
+        STRIPE_STACK_UNWIND (ftruncate, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 
-/**
- * stripe_fsyncdir - 
- */
 int32_t
-stripe_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
+stripe_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, 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));
+                        local->op_errno = op_errno;
+                        if ((op_errno != ENOENT) ||
+                            (prev->this == FIRST_CHILD (this)))
+                                local->failed = 1;
+                }
+                if (op_ret >= 0)
+                        local->op_ret = op_ret;
+        }
+        UNLOCK (&frame->lock);
+
+        if (!callcnt) {
+                if (local->failed)
+                        local->op_ret = -1;
+
+                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, dict_t *xdata)
 {
         stripe_local_t   *local = NULL;
         stripe_private_t *priv = NULL;
@@ -3026,7 +3155,7 @@ stripe_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
         trav = this->children;
 
         /* Initialization */
-        local = CALLOC (1, sizeof (stripe_local_t));
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
@@ -3036,29 +3165,114 @@ stripe_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags)
         local->call_count = priv->child_count;
 
         while (trav) {
-                STACK_WIND (frame, stripe_stack_unwind_cbk, trav->xlator,
-                            trav->xlator->fops->fsyncdir, fd, flags);
+                STACK_WIND (frame, stripe_fsyncdir_cbk, trav->xlator,
+                            trav->xlator->fops->fsyncdir, fd, flags, NULL);
                 trav = trav->next;
         }
 
         return 0;
- err:
-        STACK_UNWIND (frame, -1, op_errno, NULL, NULL);
+err:
+        STRIPE_STACK_UNWIND (fsyncdir, frame, -1, op_errno, NULL);
         return 0;
 }
 
 
-/**
- * stripe_single_readv_cbk - This function is used as return fn, when the 
- *     file name doesn't match the pattern specified for striping.
- */
 int32_t
-stripe_single_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 stat *stbuf, struct iobref *iobref)
+stripe_readv_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)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, vector, count, stbuf, iobref);
+        int32_t         i = 0;
+        int32_t         callcnt = 0;
+        int32_t         count = 0;
+        stripe_local_t *local = NULL;
+        struct iovec   *vec = NULL;
+        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");
+                goto out;
+        }
+
+        local = frame->local;
+	prev = cookie;
+
+        LOCK (&frame->lock);
+        {
+                callcnt = --local->call_count;
+                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);
+
+        if (!callcnt) {
+                op_ret = 0;
+
+                /* Keep extra space for filling in '\0's */
+                vec = GF_CALLOC ((local->count * 2), sizeof (struct iovec),
+                                 gf_stripe_mt_iovec);
+                if (!vec) {
+                        op_ret = -1;
+                        goto done;
+                }
+
+                for (i = 0; i < local->wind_count; i++) {
+                        if (local->replies[i].op_ret) {
+                                memcpy ((vec + count), local->replies[i].vector,
+                                        (local->replies[i].count * sizeof (struct iovec)));
+                                count +=  local->replies[i].count;
+                                op_ret += local->replies[i].op_ret;
+                        }
+                        if ((local->replies[i].op_ret <
+                             local->replies[i].requested_size) &&
+                            (local->stbuf_size > (local->offset + op_ret))) {
+                                /* Fill in 0s here */
+                                vec[count].iov_len  =
+                                        (local->replies[i].requested_size -
+                                         local->replies[i].op_ret);
+                                iobuf = iobuf_get2 (this->ctx->iobuf_pool,
+                                                    vec[count].iov_len);
+                                if (!iobuf) {
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "Out of memory.");
+                                        op_ret = -1;
+                                        op_errno = ENOMEM;
+                                        goto done;
+                                }
+                                memset (iobuf->ptr, 0, vec[count].iov_len);
+                                vec[count].iov_base = iobuf->ptr;
+
+                                iobref_add (local->iobref, iobuf);
+                                iobuf_unref(iobuf);
+
+                                op_ret += vec[count].iov_len;
+                                count++;
+                        }
+                        GF_FREE (local->replies[i].vector);
+                }
+
+                /* FIXME: notice that st_ino, and st_dev (gen) will be
+                 * different than what inode will have. Make sure this doesn't
+                 * cause any bugs at higher levels */
+                memcpy (&tmp_stbuf, &local->replies[0].stbuf,
+                        sizeof (struct iatt));
+                tmp_stbuf.ia_size = local->stbuf_size;
+
+        done:
+                GF_FREE (local->replies);
+                tmp_iobref = local->iobref;
+                STRIPE_STACK_UNWIND (readv, frame, op_ret, op_errno, vec,
+                                     count, &tmp_stbuf, tmp_iobref, NULL);
+
+                iobref_unref (tmp_iobref);
+                GF_FREE (vec);
+        }
+out:
         return 0;
 }
 
@@ -3067,116 +3281,155 @@ stripe_single_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
  *        to above layer after putting it in a single vector.
  */
 int32_t
-stripe_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this, 
+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 stat *stbuf, struct iobref *iobref)
+                  int32_t count, struct iatt *stbuf, struct iobref *iobref, dict_t *xdata)
 {
         int32_t         index = 0;
         int32_t         callcnt = 0;
-        call_frame_t   *main_frame = NULL;
-        stripe_local_t *main_local = NULL;
-        stripe_local_t *local = frame->local;
+        int32_t         final_count = 0;
+        int32_t         need_to_check_proper_size = 0;
+        call_frame_t   *mframe = NULL;
+        stripe_local_t *mlocal = NULL;
+        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;
 
-        index = local->node_index;
-        main_frame = local->orig_frame;
-        main_local = main_frame->local;
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto end;
+        }
+
+        local  = frame->local;
+        index  = local->node_index;
+	prev = cookie;
+        mframe = local->orig_frame;
+        if (!mframe)
+                goto out;
+
+        mlocal = mframe->local;
+        if (!mlocal)
+                goto out;
 
-        LOCK (&main_frame->lock);
+        fctx = mlocal->fctx;
+
+        LOCK (&mframe->lock);
         {
-                main_local->replies[index].op_ret = op_ret;
-                main_local->replies[index].op_errno = op_errno;
+                mlocal->replies[index].op_ret = op_ret;
+                mlocal->replies[index].op_errno = op_errno;
+                mlocal->replies[index].requested_size = local->readv_size;
                 if (op_ret >= 0) {
-                        main_local->replies[index].stbuf  = *stbuf;
-                        main_local->replies[index].count  = count;
-                        main_local->replies[index].vector = 
-                                iov_dup (vector, count);
+                        mlocal->replies[index].stbuf  = *stbuf;
+                        mlocal->replies[index].count  = count;
+                        mlocal->replies[index].vector = iov_dup (vector, count);
+
+			correct_file_size(stbuf, fctx, prev);
 
-                        if (!main_local->iobref)
-                                main_local->iobref = iobref_new ();
-                        iobref_merge (main_local->iobref, iobref);
+                        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);
                 }
-                callcnt = ++main_local->call_count;
+                callcnt = ++mlocal->call_count;
         }
-        UNLOCK(&main_frame->lock);
-
-        if (callcnt == main_local->wind_count) {
-                int32_t        final_count = 0;
-                struct iovec  *final_vec = NULL;
-                struct stat    tmp_stbuf = {0,};
-                struct iobref *iobref = NULL;
+        UNLOCK(&mframe->lock);
 
+        if (callcnt == mlocal->wind_count) {
                 op_ret = 0;
-                memcpy (&tmp_stbuf, &main_local->replies[0].stbuf, 
-                        sizeof (struct stat));
-                for (index=0; index < main_local->wind_count; index++) {
-                        /* TODO: check whether each stripe returned 'expected'
-                         * number of bytes 
-                         */
-                        if (main_local->replies[index].op_ret == -1) {
+
+                for (index=0; index < mlocal->wind_count; index++) {
+                        /* check whether each stripe returned
+                         * 'expected' number of bytes */
+                        if (mlocal->replies[index].op_ret == -1) {
                                 op_ret = -1;
-                                op_errno = main_local->replies[index].op_errno;
+                                op_errno = mlocal->replies[index].op_errno;
                                 break;
                         }
-                        op_ret += main_local->replies[index].op_ret;
-                        final_count += main_local->replies[index].count;
-                        /* TODO: Do I need to send anything more in stbuf? */
-                        if (tmp_stbuf.st_size < 
-                            main_local->replies[index].stbuf.st_size) {
-                                tmp_stbuf.st_size = 
-                                        main_local->replies[index].stbuf.st_size;
+                        /* TODO: handle the 'holes' within the read range
+                           properly */
+                        if (mlocal->replies[index].op_ret <
+                            mlocal->replies[index].requested_size) {
+                                need_to_check_proper_size = 1;
                         }
+
+                        op_ret       += mlocal->replies[index].op_ret;
+                        mlocal->count += mlocal->replies[index].count;
                 }
-                if (op_ret != -1) {
-                        final_vec = CALLOC (final_count, 
-                                            sizeof (struct iovec));
-                        if (!final_vec) {
-                                op_ret = -1;
-                                final_count = 0;
-                                goto done;
-                        }
+                if (op_ret == -1)
+                        goto done;
+                if (need_to_check_proper_size)
+                        goto check_size;
 
-                        final_count = 0;
+                final_vec = GF_CALLOC (mlocal->count, sizeof (struct iovec),
+                                       gf_stripe_mt_iovec);
 
-                        for (index=0; 
-                             index < main_local->wind_count; index++) {
-                                memcpy (final_vec + final_count,
-                                        main_local->replies[index].vector,
-                                        (main_local->replies[index].count * 
-                                         sizeof (struct iovec)));
-                                final_count += 
-                                        main_local->replies[index].count;
+                if (!final_vec) {
+                        op_ret = -1;
+                        goto done;
+                }
 
-                                free (main_local->replies[index].vector);
-                        }
-                } else {
-                        final_vec = NULL;
-                        final_count = 0;
+                for (index = 0; index < mlocal->wind_count; index++) {
+                        memcpy ((final_vec + final_count),
+                                mlocal->replies[index].vector,
+                                (mlocal->replies[index].count *
+                                 sizeof (struct iovec)));
+                        final_count +=  mlocal->replies[index].count;
+                        GF_FREE (mlocal->replies[index].vector);
                 }
 
+                /* FIXME: notice that st_ino, and st_dev (gen) will be
+                 * different than what inode will have. Make sure this doesn't
+                 * 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:
                 /* */
-                FREE (main_local->replies);
-                iobref = main_local->iobref;
-                STACK_UNWIND (main_frame, op_ret, op_errno, 
-                              final_vec, final_count, &tmp_stbuf, iobref);
+                GF_FREE (mlocal->replies);
+                tmp_iobref = mlocal->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);
+                GF_FREE (final_vec);
+        }
+
+        goto out;
 
-                iobref_unref (iobref);
-                if (final_vec)
-                        FREE (final_vec);
+check_size:
+        mlocal->call_count = fctx->stripe_count;
+
+        for (index = 0; index < fctx->stripe_count; index++) {
+                STACK_WIND (mframe, stripe_readv_fstat_cbk,
+                            (fctx->xl_array[index]),
+                            (fctx->xl_array[index])->fops->fstat,
+                            mlocal->fd, NULL);
         }
 
-        STACK_DESTROY (frame->root);
+out:
+        STRIPE_STACK_DESTROY (frame);
+end:
         return 0;
 }
 
-/**
- * stripe_readv - 
- */
+
 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 = 1;
+        int32_t           op_errno = EINVAL;
         int32_t           idx = 0;
         int32_t           index = 0;
         int32_t           num_stripe = 0;
@@ -3187,17 +3440,18 @@ 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;
-        xlator_list_t    *trav = NULL;
-        stripe_private_t *priv = NULL;
         stripe_fd_ctx_t  *fctx = NULL;
 
-        trav = this->children;
-        priv = this->private;
+        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 = EBADFD;
                 goto err;
@@ -3205,124 +3459,183 @@ 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;
 
-        /* The file is stripe across the child nodes. Send the read request 
-         * to the child nodes appropriately after checking which region of 
+        STRIPE_VALIDATE_FCTX (fctx, err);
+
+        if (!stripe_size) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Wrong stripe size for the file");
+                goto err;
+        }
+        /* The file is stripe across the child nodes. Send the read request
+         * to the child nodes appropriately after checking which region of
          * the file is in which child node. Always '0-<stripe_size>' part of
          * the file resides in the first child.
          */
         rounded_start = floor (offset, stripe_size);
         rounded_end = roof (offset+size, stripe_size);
-        num_stripe = rounded_end/stripe_size - rounded_start/stripe_size;
-        
-        local = CALLOC (1, sizeof (stripe_local_t));
+        num_stripe = (rounded_end- rounded_start)/stripe_size;
+
+        local = mem_get0 (this->local_pool);
         if (!local) {
                 op_errno = ENOMEM;
                 goto err;
         }
-        local->wind_count = num_stripe;
         frame->local = local;
-        
+
         /* This is where all the vectors should be copied. */
-        local->replies = CALLOC (num_stripe, sizeof (struct 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;
         }
-        
+
         off_index = (offset / stripe_size) % fctx->stripe_count;
-    
+        local->wind_count = num_stripe;
+        local->readv_size = size;
+        local->offset     = offset;
+        local->fd         = fd_ref (fd);
+        local->fctx       = fctx;
+
         for (index = off_index; index < (num_stripe + off_index); index++) {
                 rframe = copy_frame (frame);
-                rlocal = CALLOC (1, sizeof (stripe_local_t));
+                rlocal = mem_get0 (this->local_pool);
                 if (!rlocal) {
                         op_errno = ENOMEM;
                         goto err;
                 }
-                
+
                 frame_size = min (roof (frame_offset+1, stripe_size),
                                   (offset + size)) - frame_offset;
-                
+
                 rlocal->node_index = index - off_index;
                 rlocal->orig_frame = frame;
+                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:
-        STACK_UNWIND (frame, -1, op_errno, NULL);
+err:
+        if (rframe)
+                STRIPE_STACK_DESTROY (rframe);
+
+        STRIPE_STACK_UNWIND (readv, frame, -1, op_errno, NULL, 0, NULL, NULL, NULL);
         return 0;
 }
 
 
-/**
- * stripe_writev_cbk - 
- */
 int32_t
 stripe_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct stat *prebuf,
-                   struct stat *postbuf)
+                   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;
+	struct stripe_replies *reply = NULL;
+	int32_t		i = 0;
 
+        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 = ++local->call_count;
-    
-                if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG, 
-                                "%s returned error %s",
-                                ((call_frame_t *)cookie)->this->name, 
-                                strerror (op_errno));
-                        local->op_errno = op_errno;
-                        local->op_ret = -1;
-                }
+                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 >= 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 (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;
 }
 
-
-/**
- * stripe_single_writev_cbk - 
- */
-int32_t
-stripe_single_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                          int32_t op_ret, int32_t op_errno, struct stat *stbuf)
-{
-        STACK_UNWIND (frame, op_ret, op_errno, stbuf);
-        return 0;
-}
-/**
- * stripe_writev - 
- */
 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 = vector;
-        stripe_private_t *priv = NULL;
+        struct iovec     *tmp_vec = NULL;
         stripe_local_t   *local = NULL;
-        xlator_list_t    *trav = NULL;
         stripe_fd_ctx_t  *fctx = NULL;
         int32_t           op_errno = 1;
         int32_t           idx = 0;
@@ -3333,10 +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;
 
-        priv = this->private;
+        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;
@@ -3344,29 +3666,57 @@ 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 += tmp_vec[idx].iov_len;
+                total_size += vector[idx].iov_len;
         }
         remaining_size = total_size;
 
-        local = CALLOC (1, sizeof (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) {
-                /* Send striped chunk of the vector to child 
+		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. */
-                trav = this->children;
-                
-                idx = (((offset + offset_offset) / 
+                idx = (((offset + offset_offset) /
                         local->stripe_size) % fctx->stripe_count);
 
-                fill_size = (local->stripe_size - 
+                fill_size = (local->stripe_size -
                              ((offset + offset_offset) % local->stripe_size));
                 if (fill_size > remaining_size)
                         fill_size = remaining_size;
@@ -3375,158 +3725,597 @@ stripe_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
 
                 tmp_count = iov_subset (vector, count, offset_offset,
                                         offset_offset + fill_size, NULL);
-                tmp_vec = CALLOC (tmp_count, sizeof (struct iovec));
+                tmp_vec = GF_CALLOC (tmp_count, sizeof (struct iovec),
+                                     gf_stripe_mt_iovec);
                 if (!tmp_vec) {
                         op_errno = ENOMEM;
                         goto err;
                 }
                 tmp_count = iov_subset (vector, count, offset_offset,
                                         offset_offset + fill_size, tmp_vec);
-                
+
                 local->wind_count++;
                 if (remaining_size == 0)
                         local->unwind = 1;
 
-                STACK_WIND(frame, stripe_writev_cbk, fctx->xl_array[idx],
-                           fctx->xl_array[idx]->fops->writev, fd, tmp_vec, 
-                           tmp_count, offset + offset_offset, iobref);
-                FREE (tmp_vec);
+		/*
+		 * 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, 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 (frame, -1, op_errno, NULL);
+err:
+	if (wframe)
+		STRIPE_STACK_DESTROY(wframe);
+
+        STRIPE_STACK_UNWIND (writev, frame, -1, op_errno, NULL, NULL, NULL);
         return 0;
 }
 
 
+int32_t
+stripe_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                     struct iatt *postbuf, dict_t *xdata)
+{
+        int32_t         callcnt = 0;
+        stripe_local_t *local = NULL;
+	stripe_local_t *mlocal = NULL;
+        call_frame_t   *prev = NULL;
+	call_frame_t   *mframe = NULL;
 
-/* Management operations */
+        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;
+}
 
-/**
- * stripe_stats_cbk - Add all the fields received from different clients. 
- *    Once all the clients return, send stats to above layer.
- * 
- */
 int32_t
-stripe_stats_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                  int32_t op_ret, int32_t op_errno, struct xlator_stats *stats)
+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 = --local->call_count;
-    
-                if (op_ret == -1) {
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                "%s returned error %s",
-                                ((call_frame_t *)cookie)->this->name, 
-                                strerror (op_errno));
-                        local->op_ret = -1;
-                        local->op_errno = op_errno;
-                }
+                callcnt = ++mlocal->call_count;
+
                 if (op_ret == 0) {
-                        if (local->op_ret == -2) {
-                                /* This is to make sure this is the 
-                                   first time */
-                                local->stats = *stats;
-                                local->op_ret = 0;
-                        } else {
-                                local->stats.nr_files += stats->nr_files;
-                                local->stats.free_disk += stats->free_disk;
-                                local->stats.disk_usage += stats->disk_usage;
-                                local->stats.nr_clients += stats->nr_clients;
-                        }
+                        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) {
-                STACK_UNWIND (frame, local->op_ret, local->op_errno,
-                              &local->stats);
-        }
+        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;
 }
 
-/**
- * stripe_stats - 
- */
 int32_t
-stripe_stats (call_frame_t *frame, xlator_t *this, int32_t flags)
+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;
-        xlator_list_t    *trav = NULL;
-        stripe_private_t *priv = 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;
 
-        priv = this->private;
-        trav = this->children;
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+        VALIDATE_OR_GOTO (fd->inode, err);
 
-        local = CALLOC (1, sizeof (stripe_local_t));
+        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->op_ret = -2; /* to be used as a flag in _cbk */
-        local->call_count = priv->child_count;
+        local->stripe_size = stripe_size;
+	local->fctx = fctx;
 
-        while (trav) {
-                STACK_WIND (frame, stripe_stats_cbk, trav->xlator,
-                            trav->xlator->mops->stats, flags);
-                trav = trav->next;
+        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:
-        STACK_UNWIND (frame, -1, op_errno, NULL);
+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;
 }
 
 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;
 
-        fd_ctx_del (fd, this, &tmp_fctx);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (inode, err);
+
+        (void) inode_ctx_del (inode, this, &tmp_fctx);
         if (!tmp_fctx) {
-                goto out;
+                goto err;
         }
-        
+
         fctx = (stripe_fd_ctx_t *)(long)tmp_fctx;
 
         if (!fctx->static_array)
-                FREE (fctx->xl_array);
-        
-        FREE (fctx);
-                
- out:
-	return 0;
+                GF_FREE (fctx->xl_array);
+
+        GF_FREE (fctx);
+err:
+        return 0;
 }
 
-/**
- * notify
- */
 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;
 
         priv = this->private;
         if (!priv)
                 return 0;
 
-        switch (event) 
+        switch (event)
         {
         case GF_EVENT_CHILD_UP:
         {
@@ -3535,24 +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)) {
+                        if (data == FIRST_CHILD (this))
                                 priv->first_child_down = 0;
-                                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 */
@@ -3560,20 +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)) {
+                        if (data == FIRST_CHILD (this))
                                 priv->first_child_down = 1;
-                                default_notify (this, event, data);
-                        }
+                        priv->last_event[i] = event;
                 }
                 UNLOCK (&priv->lock);
         }
@@ -3583,84 +4375,751 @@ 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;    
-        
-        /* Get the pattern for striping. 
-           "option block-size *avi:10MB" etc */
-        stripe_str = strtok_r (data, ",", &tmp_str);
-        while (stripe_str) {
-                dup_str = strdup (stripe_str);
-                stripe_opt = CALLOC (1, sizeof (struct stripe_options));
-                if (!stripe_opt) {
-                        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;
+        }
+
+        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;
                 }
+        }
+
+ 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;
 
-                pattern = strtok_r (dup_str, ":", &tmp_str1);
-                num = strtok_r (NULL, ":", &tmp_str1);
-                if (!num) {
-                        num = pattern;
-                        pattern = "*";
+        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);
                 }
-                if (gf_string2bytesize (num, &stripe_opt->block_size) != 0) {
-                        gf_log (this->name, GF_LOG_ERROR, 
-                                "invalid number format \"%s\"", num);
+        } else {
+                local->wind_count = 1;
+                STACK_WIND (frame, stripe_setxattr_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->setxattr,
+                            loc, dict, flags, xdata);
+        }
+
+        return 0;
+err:
+        STRIPE_STACK_UNWIND (setxattr, frame, -1,  op_errno, NULL);
+        return 0;
+}
+
+
+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;
+                }
+        }
+        UNLOCK (&frame->lock);
+
+        if (call_count == 0) {
+                STRIPE_STACK_UNWIND (fsetxattr, frame, local->op_ret,
+                                     local->op_errno, NULL);
+        }
+        return 0;
+}
+
+int
+stripe_fsetxattr_to_everyone (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                              dict_t *dict, int flags, dict_t *xdata)
+{
+        xlator_list_t    *trav  = NULL;
+        stripe_private_t *priv  = NULL;
+        int               ret   = -1;
+        stripe_local_t   *local = NULL;
+
+        priv = this->private;
+
+        local = mem_get0 (this->local_pool);
+        if (local == NULL) {
+                goto out;
+        }
+
+        frame->local = local;
+
+        local->wind_count = priv->child_count;
+
+        trav = this->children;
+
+        while (trav) {
+                STACK_WIND (frame, stripe_fsetxattr_everyone_cbk,
+                            trav->xlator, trav->xlator->fops->fsetxattr,
+                            fd, dict, flags, xdata);
+                trav = trav->next;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+inline gf_boolean_t
+stripe_fsetxattr_is_special (dict_t *dict)
+{
+        gf_boolean_t is_spl = _gf_false;
+
+        if (dict == NULL) {
+                goto out;
+        }
+
+        dict_foreach (dict, stripe_is_special_key, &is_spl);
+
+out:
+        return is_spl;
+}
+
+int
+stripe_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                  dict_t *dict, int flags, dict_t *xdata)
+{
+        int32_t      op_ret = -1, ret = -1, op_errno = EINVAL;
+        gf_boolean_t is_spl = _gf_false;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        GF_IF_INTERNAL_XATTR_GOTO ("trusted.*stripe*", dict,
+                                   op_errno, err);
+
+        is_spl = stripe_fsetxattr_is_special (dict);
+        if (is_spl) {
+                ret = stripe_fsetxattr_to_everyone (frame, this, fd, dict,
+                                                    flags, xdata);
+                if (ret < 0) {
+                        op_errno = ENOMEM;
+                        goto err;
+                }
+
+                goto out;
+        }
+
+        STACK_WIND (frame, stripe_fsetxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsetxattr,
+                    fd, dict, flags, xdata);
+out:
+        return 0;
+err:
+        STRIPE_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, NULL);
+        return 0;
+}
+
+int
+stripe_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        STRIPE_STACK_UNWIND (removexattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int
+stripe_removexattr (call_frame_t *frame, xlator_t *this,
+                    loc_t *loc, const char *name, dict_t *xdata)
+{
+        int32_t         op_errno = EINVAL;
+
+        VALIDATE_OR_GOTO (this, err);
+
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.*stripe*",
+                                 name, op_errno, err);
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (loc, err);
+
+        STACK_WIND (frame, stripe_removexattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->removexattr,
+                    loc, name, xdata);
+        return 0;
+err:
+        STRIPE_STACK_UNWIND (removexattr, frame, -1,  op_errno, NULL);
+        return 0;
+}
+
+
+int
+stripe_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        STRIPE_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int
+stripe_fremovexattr (call_frame_t *frame, xlator_t *this,
+                     fd_t *fd, const char *name, dict_t *xdata)
+{
+        int32_t         op_ret   = -1;
+        int32_t         op_errno = EINVAL;
+
+        VALIDATE_OR_GOTO (frame, err);
+        VALIDATE_OR_GOTO (this, err);
+        VALIDATE_OR_GOTO (fd, err);
+
+        GF_IF_NATIVE_XATTR_GOTO ("trusted.*stripe*",
+                                 name, op_errno, err);
+
+        STACK_WIND (frame, stripe_fremovexattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fremovexattr,
+                    fd, name, xdata);
+        return 0;
+ err:
+        STRIPE_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int32_t
+stripe_readdirp_lookup_cbk (call_frame_t *frame, void *cookie,
+                            xlator_t *this, int op_ret, int op_errno,
+                            inode_t *inode, struct iatt *stbuf,
+                            dict_t *xattr, struct iatt *parent)
+{
+        stripe_local_t          *local          = NULL;
+        call_frame_t            *main_frame     = NULL;
+        stripe_local_t          *main_local     = NULL;
+        gf_dirent_t             *entry          = NULL;
+        call_frame_t            *prev           = NULL;
+        int                      done           = 0;
+
+        local = frame->local;
+        prev = cookie;
+
+        entry = local->dirent;
+
+        main_frame = local->orig_frame;
+        main_local = main_frame->local;
+        LOCK (&frame->lock);
+        {
+
+                local->call_count--;
+                if (!local->call_count)
+                        done = 1;
+                if (op_ret == -1) {
+                        local->op_errno = op_errno;
+                        local->op_ret = op_ret;
+                        goto unlock;
+                }
+
+		if (stripe_ctx_handle(this, prev, local, xattr))
+			gf_log(this->name, GF_LOG_ERROR,
+				"Error getting fctx info from dict.");
+
+		correct_file_size(stbuf, local->fctx, prev);
+
+                stripe_iatt_merge (stbuf, &entry->d_stat);
+                local->stbuf_blocks += stbuf->ia_blocks;
+        }
+unlock:
+        UNLOCK(&frame->lock);
+
+        if (done) {
+                inode_ctx_put (entry->inode, this,
+                               (uint64_t) (long)local->fctx);
+
+                done = 0;
+                LOCK (&main_frame->lock);
+                {
+                        main_local->wind_count--;
+                        if (!main_local->wind_count)
+                                done = 1;
+                        if (local->op_ret == -1) {
+                                main_local->op_errno = local->op_errno;
+                                main_local->op_ret = local->op_ret;
+                        }
+                        entry->d_stat.ia_blocks = local->stbuf_blocks;
+                }
+                UNLOCK (&main_frame->lock);
+                if (done) {
+                        main_frame->local = NULL;
+                        STRIPE_STACK_UNWIND (readdir, main_frame,
+                                             main_local->op_ret,
+                                             main_local->op_errno,
+                                             &main_local->entries, NULL);
+                        gf_dirent_free (&main_local->entries);
+                        stripe_local_wipe (main_local);
+                        mem_put (main_local);
+                }
+                frame->local = NULL;
+                stripe_local_wipe (local);
+                mem_put (local);
+                STRIPE_STACK_DESTROY (frame);
+        }
+
+        return 0;
+}
+
+int32_t
+stripe_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno,
+                     gf_dirent_t *orig_entries, dict_t *xdata)
+{
+        stripe_local_t *local = NULL;
+        call_frame_t   *prev = NULL;
+        gf_dirent_t    *local_entry = NULL;
+        gf_dirent_t    *tmp_entry = NULL;
+        xlator_list_t  *trav = NULL;
+        loc_t          loc = {0, };
+        int32_t        count = 0;
+        stripe_private_t *priv = NULL;
+        int32_t        subvols = 0;
+        dict_t         *xattrs = NULL;
+        call_frame_t   *local_frame = NULL;
+        stripe_local_t *local_ent = NULL;
+
+        if (!this || !frame || !frame->local || !cookie) {
+                gf_log ("stripe", GF_LOG_DEBUG, "possible NULL deref");
+                goto out;
+        }
+        prev  = cookie;
+        local = frame->local;
+        trav = this->children;
+        priv = this->private;
+
+        subvols = priv->child_count;
+
+        LOCK (&frame->lock);
+        {
+                if (op_ret == -1) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "%s returned error %s",
+                                prev->this->name, strerror (op_errno));
+                        local->op_errno = op_errno;
+                        local->op_ret = op_ret;
+                        goto unlock;
+                } else {
+                        local->op_ret = op_ret;
+                        list_splice_init (&orig_entries->list,
+                                          &local->entries.list);
+                        local->wind_count = op_ret;
+                }
+
+        }
+unlock:
+        UNLOCK (&frame->lock);
+
+        if (op_ret == -1)
+                goto out;
+
+        xattrs = dict_new ();
+        if (xattrs)
+                (void) stripe_xattr_request_build (this, xattrs, 0, 0, 0, 0);
+        count = op_ret;
+        list_for_each_entry_safe (local_entry, tmp_entry,
+                                  (&local->entries.list), list) {
+
+                if (!local_entry)
+                        break;
+                if (!IA_ISREG (local_entry->d_stat.ia_type) || !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;
+                }
+
+                local_ent = mem_get0 (this->local_pool);
+                if (!local_ent) {
+                        op_errno = ENOMEM;
+                        op_ret = -1;
                         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) {
-                        priv->pattern = stripe_opt;
+                }
+
+                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:
+        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 = -1;
+        volume_option_t  *opt = NULL;
+
+        GF_ASSERT (this);
+        GF_ASSERT (this->private);
+
+        priv = this->private;
+
+
+        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 {
-                        temp_stripeopt = priv->pattern;
-                        while (temp_stripeopt->next)
-                                temp_stripeopt = temp_stripeopt->next;
-                        temp_stripeopt->next = stripe_opt;
+                        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;
+                        }
                 }
-                stripe_str = strtok_r (NULL, ",", &tmp_str);
+
+		GF_OPTION_RECONF("coalesce", priv->coalesce, options, bool,
+				unlock);
         }
+ unlock:
+        UNLOCK (&priv->lock);
+        if (ret)
+                goto out;
 
         ret = 0;
  out:
         return ret;
+
 }
 
 /**
- * init - This function is called when xlator-graph gets initialized. 
+ * init - This function is called when xlator-graph gets initialized.
  *     The option given in volfiles are parsed here.
- * @this - 
+ * @this -
  */
 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;
         int               ret = -1;
 
+        if (!this)
+                goto out;
+
         trav = this->children;
         while (trav) {
                 count++;
@@ -3685,15 +5144,20 @@ init (xlator_t *this)
                         " please check the volume. exiting");
                 goto out;
         }
-        priv = CALLOC (1, sizeof (stripe_private_t));
+
+        priv = GF_CALLOC (1, sizeof (stripe_private_t),
+                          gf_stripe_mt_stripe_private_t);
+
         if (!priv)
                 goto out;
-        priv->xl_array = CALLOC (count, sizeof (xlator_t *));
+        priv->xl_array = GF_CALLOC (count, sizeof (xlator_t *),
+                                    gf_stripe_mt_xlator_t);
         if (!priv->xl_array)
                 goto out;
 
-        priv->state = CALLOC (count, sizeof (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;
@@ -3713,125 +5177,614 @@ 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:
+out:
         if (ret) {
                 if (priv) {
-                        if (priv->xl_array)
-                                FREE (priv->xl_array);
-                        FREE (priv);
+                        GF_FREE (priv->xl_array);
+                        GF_FREE (priv);
                 }
         }
         return ret;
-} 
+}
 
-/** 
+/**
  * fini -   Free all the private variables
- * @this - 
+ * @this -
  */
-void 
+void
 fini (xlator_t *this)
 {
         stripe_private_t      *priv = NULL;
         struct stripe_options *prev = NULL;
         struct stripe_options *trav = NULL;
 
+        if (!this)
+                goto out;
+
         priv = this->private;
         if (priv) {
-                if (priv->xl_array)
-                        FREE (priv->xl_array);
+                this->private = NULL;
+                GF_FREE (priv->xl_array);
 
                 trav = priv->pattern;
                 while (trav) {
                         prev = trav;
                         trav = trav->next;
-                        FREE (prev);
+                        GF_FREE (prev);
                 }
+                GF_FREE (priv->last_event);
                 LOCK_DESTROY (&priv->lock);
-                FREE (priv);
+                GF_FREE (priv);
         }
 
+out:
         return;
 }
 
+int32_t
+stripe_getxattr_unwind (call_frame_t *frame,
+                        int op_ret, int op_errno, dict_t *dict, dict_t *xdata)
 
-struct xlator_fops fops = {
-        .stat        = stripe_stat,
-        .unlink      = stripe_unlink,
-        .symlink     = stripe_symlink,
-        .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,
-        .setxattr    = stripe_setxattr,
-        .getxattr    = stripe_getxattr,
-        .removexattr = stripe_removexattr,
-        .access      = stripe_access,
-        .ftruncate   = stripe_ftruncate,
-        .fstat       = stripe_fstat,
-        .readlink    = stripe_readlink,
-        .mkdir       = stripe_mkdir,
-        .rmdir       = stripe_rmdir,
-        .lk          = stripe_lk,
-        .opendir     = stripe_opendir,
-        .fsyncdir    = stripe_fsyncdir,
-        .setattr     = stripe_setattr,
-        .fsetattr    = stripe_fsetattr,
-        .lookup      = stripe_lookup,
-        .setdents    = stripe_setdents,
-        .mknod       = stripe_mknod,
-};
+{
+        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;
 
-struct xlator_mops mops = {
-        .stats  = stripe_stats,
+                        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,
+        .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 
+        { .key  = {"block-size"},
+          .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
+        { .key  = {"use-xattr"},
+          .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 ec4782686..5673d18f3 100644
--- a/xlators/cluster/stripe/src/stripe.h
+++ b/xlators/cluster/stripe/src/stripe.h
@@ -1,20 +1,11 @@
 /*
-  Copyright (c) 2009 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
 
@@ -32,12 +23,63 @@
 #include "common-utils.h"
 #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 
+ * struct stripe_options : This keeps the pattern and the block-size
  *     information, which is used for striping on a file.
  */
 struct stripe_options {
@@ -47,7 +89,7 @@ struct stripe_options {
 };
 
 /**
- * Private structure for stripe translator 
+ * Private structure for stripe translator
  */
 struct stripe_private {
         struct stripe_options  *pattern;
@@ -56,27 +98,31 @@ 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
         int32_t       op_errno;
-        struct stat   stbuf;    /* 'stbuf' is also a part of reply */
+        int32_t       requested_size;
+        struct iatt   stbuf;    /* 'stbuf' is also a part of reply */
 };
 
 typedef struct _stripe_fd_ctx {
         off_t      stripe_size;
         int        stripe_count;
+	int	   stripe_coalesce;
         int        static_array;
-        xlator_t **xl_array;        
+        xlator_t **xl_array;
 } stripe_fd_ctx_t;
 
 
@@ -87,36 +133,45 @@ struct stripe_local; /* this itself is used inside the structure; */
 
 struct stripe_local {
         struct stripe_local *next;
-        call_frame_t        *orig_frame; 
-        
+        call_frame_t        *orig_frame;
+
         stripe_fd_ctx_t     *fctx;
 
         /* Used by _cbk functions */
-        struct stat          stbuf;
-        struct stat          pre_buf;
-        struct stat          post_buf;
-        struct stat          pre_parent_buf;
-        struct stat          post_parent_buf;
+        struct iatt          stbuf;
+        struct iatt          pre_buf;
+        struct iatt          post_buf;
+        struct iatt          preparent;
+        struct iatt          postparent;
 
-        struct stat          pre_newparent_buf;
-        struct stat          post_newparent_buf;
+        off_t                stbuf_size;
+        off_t                prebuf_size;
+        off_t                postbuf_size;
+        off_t                preparent_size;
+        off_t                postparent_size;
 
-        struct readv_replies *replies;
-        struct statvfs       statvfs_buf;
-        dir_entry_t         *entry;
-        struct xlator_stats  stats;
+        blkcnt_t             stbuf_blocks;
+        blkcnt_t             prebuf_blocks;
+        blkcnt_t             postbuf_blocks;
+        blkcnt_t             preparent_blocks;
+        blkcnt_t             postparent_blocks;
+
+        struct stripe_replies *replies;
+        struct statvfs        statvfs_buf;
+        dir_entry_t          *entry;
 
         int8_t               revalidate;
         int8_t               failed;
         int8_t               unwind;
 
+        size_t               readv_size;
         int32_t              entry_count;
         int32_t              node_index;
         int32_t              call_count;
-        int32_t              wind_count; /* used instead of child_cound 
+        int32_t              wind_count; /* used instead of child_cound
                                             in case of read and write */
         int32_t              op_ret;
-        int32_t              op_errno; 
+        int32_t              op_errno;
         int32_t              count;
         int32_t              flags;
         char                *name;
@@ -125,8 +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;
@@ -136,13 +200,89 @@ struct stripe_local {
         int entry_self_heal_needed;
 
         int8_t              *list;
-        struct flock         lock;
+        struct gf_flock         lock;
         fd_t                *fd;
         void                *value;
         struct iobref       *iobref;
+        gf_dirent_t          entries;
+        gf_dirent_t         *dirent;
+        dict_t              *xattr;
+        uuid_t               ia_gfid;
+
+        int                  xflag;
+        mode_t               umask;
 };
 
 typedef struct stripe_local   stripe_local_t;
 typedef struct stripe_private stripe_private_t;
 
+/*
+ * Determine the stripe index of a particular frame based on the translator.
+ */
+static inline int32_t stripe_get_frame_index(stripe_fd_ctx_t *fctx,
+					     call_frame_t *prev)
+{
+	int32_t i, idx = -1;
+
+	for (i = 0; i < fctx->stripe_count; i++) {
+		if (fctx->xl_array[i] == prev->this) {
+			idx = i;
+			break;
+		}
+	}
+
+	return idx;
+}
+
+static inline void stripe_copy_xl_array(xlator_t **dst, xlator_t **src,
+					int count)
+{
+	int i;
+
+	for (i = 0; i < count; i++)
+		dst[i] = src[i];
+}
+
+void stripe_local_wipe (stripe_local_t *local);
+int32_t stripe_ctx_handle (xlator_t *this, call_frame_t *prev,
+                           stripe_local_t *local, dict_t *dict);
+void stripe_aggregate_xattr (dict_t *dst, dict_t *src);
+int32_t stripe_xattr_request_build (xlator_t *this, dict_t *dict,
+                                    uint64_t stripe_size, uint32_t stripe_count,
+                                    uint32_t stripe_index,
+				    uint32_t stripe_coalesce);
+int32_t stripe_get_matching_bs (const char *path, stripe_private_t *priv);
+int set_stripe_block_size (xlator_t *this, stripe_private_t *priv, char *data);
+int32_t stripe_iatt_merge (struct iatt *from, struct iatt *to);
+int32_t stripe_fill_pathinfo_xattr (xlator_t *this, stripe_local_t *local,
+                                    char **xattr_serz);
+int32_t stripe_free_xattr_str (stripe_local_t *local);
+int32_t stripe_xattr_aggregate (char *buffer, stripe_local_t *local,
+                                int32_t *total);
+off_t coalesced_offset(off_t offset, uint64_t stripe_size, int stripe_count);
+off_t uncoalesced_size(off_t size, uint64_t stripe_size, int stripe_count,
+			int stripe_index);
+int32_t
+stripe_fill_lockinfo_xattr (xlator_t *this, stripe_local_t *local,
+                            void **xattr_serz);
+
+/*
+ * Adjust the size attribute for files if coalesce is enabled.
+ */
+static inline void correct_file_size(struct iatt *buf, stripe_fd_ctx_t *fctx,
+	call_frame_t *prev)
+{
+	int index;
+
+	if (!IA_ISREG(buf->ia_type))
+		return;
+
+	if (!fctx || !fctx->stripe_coalesce)
+		return;
+
+	index = stripe_get_frame_index(fctx, prev);
+	buf->ia_size = uncoalesced_size(buf->ia_size, fctx->stripe_size,
+		fctx->stripe_count, index);
+}
+
 #endif /* _STRIPE_H_ */
diff --git a/xlators/cluster/unify/src/Makefile.am b/xlators/cluster/unify/src/Makefile.am
deleted file mode 100644
index 2a1fe8372..000000000
--- a/xlators/cluster/unify/src/Makefile.am
+++ /dev/null
@@ -1,16 +0,0 @@
-
-xlator_LTLIBRARIES = unify.la
-xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/legacy/cluster
-
-unify_la_LDFLAGS = -module -avoidversion
-
-unify_la_SOURCES = unify.c unify-self-heal.c
-unify_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
-
-noinst_HEADERS = unify.h
-
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
-
-CLEANFILES = 
-
diff --git a/xlators/cluster/unify/src/unify-self-heal.c b/xlators/cluster/unify/src/unify-self-heal.c
deleted file mode 100644
index 3e4affe8c..000000000
--- a/xlators/cluster/unify/src/unify-self-heal.c
+++ /dev/null
@@ -1,1227 +0,0 @@
-/*
-  Copyright (c) 2007-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-/**
- * unify-self-heal.c : 
- *   This file implements few functions which enables 'unify' translator 
- *  to be consistent in its behaviour when 
- *     > a node fails, 
- *     > a node gets added, 
- *     > a failed node comes back
- *     > a new namespace server is added (ie, an fresh namespace server).
- * 
- *  This functionality of 'unify' will enable glusterfs to support storage
- *  system failure, and maintain consistancy. This works both ways, ie, when
- *  an entry (either file or directory) is found on namespace server, and not
- *  on storage nodes, its created in storage nodes and vica-versa.
- * 
- *  The two fops, where it can be implemented are 'getdents ()' and 'lookup ()'
- *
- */
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#include "glusterfs.h"
-#include "unify.h"
-#include "dict.h"
-#include "xlator.h"
-#include "hashfn.h"
-#include "logging.h"
-#include "stack.h"
-#include "common-utils.h"
-
-int32_t
-unify_sh_getdents_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno,
-		       dir_entry_t *entry,
-		       int32_t count);
-
-int32_t
-unify_sh_ns_getdents_cbk (call_frame_t *frame,
-			  void *cookie,
-			  xlator_t *this,
-			  int32_t op_ret,
-			  int32_t op_errno,
-			  dir_entry_t *entry,
-			  int32_t count);
-
-int32_t
-unify_bgsh_getdents_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno,
-			 dir_entry_t *entry,
-			 int32_t count);
-
-int32_t
-unify_bgsh_ns_getdents_cbk (call_frame_t *frame,
-			    void *cookie,
-			    xlator_t *this,
-			    int32_t op_ret,
-			    int32_t op_errno,
-			    dir_entry_t *entry,
-			    int32_t count);
-
-/**
- * unify_local_wipe - free all the extra allocation of local->* here.
- */
-static void 
-unify_local_wipe (unify_local_t *local)
-{
-	/* Free the strdup'd variables in the local structure */
-	if (local->name) {
-		FREE (local->name);
-	}
-
-	if (local->sh_struct) {
-		if (local->sh_struct->offset_list)
-			FREE (local->sh_struct->offset_list);
-
-		if (local->sh_struct->entry_list)
-			FREE (local->sh_struct->entry_list);
-
-		if (local->sh_struct->count_list)
-			FREE (local->sh_struct->count_list);
-
-		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;
-				FREE (trav->name);
-				if (S_ISLNK (trav->buf.st_mode))
-					FREE (trav->link);
-				FREE (trav);
-				trav = prev->next;
-			}
-			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 = CALLOC (1, sizeof (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;
-				FREE (trav->name);
-				if (S_ISLNK (trav->buf.st_mode))
-					FREE (trav->link);
-				FREE (trav);
-				trav = prev->next;
-			}
-			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 = CALLOC (1, sizeof (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 = 
-					calloc (priv->child_count, 
-						sizeof (off_t));
-				ERR_ABORT (local->sh_struct->offset_list);
-	
-				local->sh_struct->entry_list = 
-					calloc (priv->child_count, 
-						sizeof (dir_entry_t *));
-				ERR_ABORT (local->sh_struct->entry_list);
-
-				local->sh_struct->count_list = 
-					calloc (priv->child_count, 
-						sizeof (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;
-				FREE (trav->name);
-				if (S_ISLNK (trav->buf.st_mode))
-					FREE (trav->link);
-				FREE (trav);
-				trav = prev->next;
-			}
-			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 = CALLOC (1, sizeof (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;
-			FREE (trav->name);
-			if (S_ISLNK (trav->buf.st_mode))
-				FREE (trav->link);
-			FREE (trav);
-			trav = prev->next;
-		}
-		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 = CALLOC (1, sizeof (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 = 
-					calloc (priv->child_count, 
-						sizeof (off_t));
-				ERR_ABORT (local->sh_struct->offset_list);
-	
-				local->sh_struct->entry_list = 
-					calloc (priv->child_count, 
-						sizeof (dir_entry_t *));
-				ERR_ABORT (local->sh_struct->entry_list);
-
-				local->sh_struct->count_list = 
-					calloc (priv->child_count, 
-						sizeof (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 = 
-				calloc (1, sizeof (struct 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 = 
-			calloc (1, sizeof (struct 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 88ae4ca84..000000000
--- a/xlators/cluster/unify/src/unify.c
+++ /dev/null
@@ -1,4563 +0,0 @@
-/*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-/**
- * xlators/cluster/unify:
- *     - This xlator is one of the main translator in GlusterFS, which
- *   actually does the clustering work of the file system. One need to 
- *   understand that, unify assumes file to be existing in only one of 
- *   the child node, and directories to be present on all the nodes. 
- *
- * NOTE:
- *   Now, unify has support for global namespace, which is used to keep a 
- * global view of fs's namespace tree. The stat for directories are taken
- * just from the namespace, where as for files, just 'st_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) {
-		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 (!S_ISDIR (loc->inode->st_mode)) {
-		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 stat *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->st_ino = buf->st_ino;
-				/* If the entry is directory, get the stat
-				   from NS node */
-				if (S_ISDIR (buf->st_mode) || 
-				    !local->stbuf.st_blksize) {
-					local->stbuf = *buf;
-				}
-			}
-
-			if ((!S_ISDIR (buf->st_mode)) && 
-			    (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->st_ino)
-			local->op_ret = -1;
-
-		local->stbuf.st_ino = local->st_ino;
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno, 
-			      &local->stbuf);
-	}
-
-	return 0;
-}
-
-#define check_if_dht_linkfile(s) ((s->st_mode & ~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 stat *buf,
-		  dict_t *dict,
-                  struct stat *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.st_mode && local->stbuf.st_blksize) {
-				/* make sure we already have a stbuf
-				   stored in local->stbuf */
-				if (S_ISDIR (local->stbuf.st_mode) && 
-				    !S_ISDIR (buf->st_mode)) {
-					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 (!S_ISDIR (local->stbuf.st_mode) && 
-				    S_ISDIR (buf->st_mode)) {
-					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 && !S_ISDIR (buf->st_mode)) {
-				/* This is the first time lookup on file*/
-				if (!local->list) {
-					/* list is not allocated, allocate 
-					   the max possible range */
-					local->list = CALLOC (1, 2 * (priv->child_count + 2));
-					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 && S_ISDIR (buf->st_mode)) {
-                                /* 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->st_ino = buf->st_ino;
-				if (S_ISDIR (buf->st_mode) || 
-				    !(local->stbuf.st_blksize)) {
-					local->stbuf = *buf;
-                                        local->oldpostparent = *postparent;
-				}
-			} else if (!S_ISDIR (buf->st_mode)) {
-				/* If file, then get the stat from 
-				   storage node */
-				local->stbuf = *buf;
-			}
-
-			if (local->st_nlink < buf->st_nlink) {
-				local->st_nlink = buf->st_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.st_blksize) {
-			/* Inode not present */
-			local->op_ret = -1;
-		} else {
-			if (!local->revalidate && 
-			    !S_ISDIR (local->stbuf.st_mode)) { 
-				/* If its a file, big array is useless, 
-				   allocate the smaller one */
-				int16_t *list = NULL;
-				list = CALLOC (1, 2 * (local->index + 1));
-				ERR_ABORT (list);
-				memcpy (list, local->list, 2 * local->index);
-				/* Make the end of the list as -1 */
-				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 (S_ISDIR(local->loc1.inode->st_mode)) {
-				/* 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.st_ino = local->st_ino;
-			}
-	  
-			local->stbuf.st_nlink = local->st_nlink;
-		}
-		if (local->op_ret == -1) {
-			if (!local->revalidate && local->list)
-				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) && 
-		     S_ISDIR(local->stbuf.st_mode))) {
-			/* 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)
-            && S_ISDIR (loc->inode->st_mode)) {
-                local->revalidate = 1;
-        }
-
-	if (!inode_ctx_get (loc->inode, this, NULL) && 
-	    loc->inode->st_mode && 
-	    !S_ISDIR (loc->inode->st_mode)) {
-		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->st_mode) {
-			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->st_ino = loc->inode->ino;
-	if (S_ISDIR (loc->inode->st_mode)) {
-		/* 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 stat *buf,
-                 struct stat *preparent,
-                 struct stat *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 stat *buf,
-                    struct stat *preparent,
-                    struct stat *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 stat *preparent,
-                 struct stat *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 stat *preparent,
-                    struct stat *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 stat *buf,
-		       dict_t *dict)
-{
-	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 (S_ISDIR (buf->st_mode))
-				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)
-{
-	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 = strdup (path);
-		ERR_ABORT (local->name);
-	} else {
-		char *tmp_str = strdup (local->loc1.path);
-		char *tmp_base = dirname (tmp_str);
-		local->name = CALLOC (1, ZR_PATH_MAX);
-		strcpy (local->name, tmp_base);
-		strncat (local->name, "/", 1);
-		strcat (local->name, path);
-		FREE (tmp_str);
-	}
-  
-	local->list = CALLOC (1, sizeof (int16_t) * 3);
-	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 (S_ISLNK (loc->inode->st_mode)) {
-		/* 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 stat *preparent,
-                         struct stat *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 stat *buf,
-			 dict_t *dict,
-                         struct stat *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->st_ino = buf->st_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.st_ino = local->st_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 stat *buf,
-                  struct stat *preparent,
-                  struct stat *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.st_ino = local->st_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 stat *buf,
-                     struct stat *preparent,
-                     struct stat *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->st_ino = buf->st_ino;
-  
-                local->oldpreparent = *preparent;
-                local->oldpostparent = *postparent;
-
-		local->op_ret = -1;
-
-		/* Start the mapping list */
-		list = CALLOC (1, sizeof (int16_t) * 3);
-		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 = CALLOC (1, sizeof (int16_t) * 3);
-		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 stat *statpre,
-                   struct stat *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->st_ino = statpost->st_ino;
-				/* If the entry is directory, get the stat
-				   from NS node */
-				if (S_ISDIR (statpost->st_mode) || 
-				    !local->stpost.st_blksize) {
-					local->stpre = *statpre;
-                                        local->stpost = *statpost;
-				}
-			}
-
-			if ((!S_ISDIR (statpost->st_mode)) && 
-			    (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->st_ino)
-			local->op_ret = -1;
-
-                local->stpre.st_ino = local->st_ino;
-		local->stpost.st_ino = local->st_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 stat *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 (S_ISDIR (loc->inode->st_mode)) {
-		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 stat *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 stat *prebuf,
-                    struct stat *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->st_ino = postbuf->st_ino;
-				/* If the entry is directory, get the 
-				   stat from NS node */
-				if (S_ISDIR (postbuf->st_mode) ||
-				    !local->stbuf.st_blksize) {
-					local->stbuf = *prebuf;
-                                        local->poststbuf = *postbuf;
-				}
-			}
-
-			if ((!S_ISDIR (postbuf->st_mode)) &&
-			    (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->st_ino) {
-			local->stbuf.st_ino = local->st_ino;
-                        local->poststbuf.st_ino = local->st_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->st_ino = loc->inode->ino;
-
-	if (S_ISDIR (loc->inode->st_mode)) {
-		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 stat *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 stat *preparent,
-                  struct stat *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 stat *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 stat *prebuf,
-		  struct stat *postbuf)
-{
-	unify_local_t *local = NULL;
-
-        local = frame->local;
-
-        local->stbuf = *prebuf;
-        local->stbuf.st_ino = local->st_ino;
-
-        local->poststbuf = *postbuf;
-        local->poststbuf.st_ino = local->st_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->st_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 stat *prebuf,
-                 struct stat *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->st_ino = fd->inode->ino;
-
-	if (!fd_ctx_get (fd, this, &tmp_child)) {
-		/* If its set, then its file */
-		child = (xlator_t *)(long)tmp_child;		     
-		local->call_count = 2;
-
-		STACK_WIND (frame, unify_buf_cbk, child,
-			    child->fops->fstat, fd);
-
-		STACK_WIND (frame, unify_buf_cbk, NS(this),
-			    NS(this)->fops->fstat, fd);
-
-	} else {
-		/* this is an directory */
-		local->call_count = 1;
-		STACK_WIND (frame, unify_buf_cbk, NS(this),
-			    NS(this)->fops->fstat, fd);
-	}
-
-	return 0;
-}
-
-/**
- * unify_getdents_cbk - 
- */
-int32_t
-unify_getdents_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    dir_entry_t *entry,
-		    int32_t count)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, entry, count);
-	return 0;
-}
-
-/**
- * unify_getdents - send the FOP request to all the nodes.
- */
-int32_t
-unify_getdents (call_frame_t *frame,
-		xlator_t *this,
-		fd_t *fd,
-		size_t size,
-		off_t offset,
-		int32_t flag)
-{
-	UNIFY_CHECK_FD_AND_UNWIND_ON_ERR (fd);
-
-	STACK_WIND (frame, unify_getdents_cbk, NS(this),
-		    NS(this)->fops->getdents, fd, size, offset, flag);
-
-	return 0;
-}
-
-
-/**
- * unify_readdir_cbk - 
- */
-int32_t
-unify_readdir_cbk (call_frame_t *frame,
-		   void *cookie,
-		   xlator_t *this,
-		   int32_t op_ret,
-		   int32_t op_errno,
-		   gf_dirent_t *buf)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, buf);
-
-	return 0;
-}
-
-/**
- * unify_readdir - send the FOP request to all the nodes.
- */
-int32_t
-unify_readdir (call_frame_t *frame,
-	       xlator_t *this,
-	       fd_t *fd,
-	       size_t size,
-	       off_t offset)
-{
-	UNIFY_CHECK_FD_AND_UNWIND_ON_ERR (fd);
-
-	STACK_WIND (frame, unify_readdir_cbk, NS(this),
-		    NS(this)->fops->readdir, fd, size, offset);
-
-	return 0;
-}
-
-
-int32_t
-unify_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, gf_dirent_t *buf)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, buf);
-
-	return 0;
-}
-
-
-int32_t
-unify_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-                off_t offset)
-{
-	UNIFY_CHECK_FD_AND_UNWIND_ON_ERR (fd);
-
-	STACK_WIND (frame, unify_readdirp_cbk, NS(this),
-		    NS(this)->fops->readdirp, fd, size, offset);
-
-	return 0;
-}
-
-
-/**
- * unify_fsyncdir_cbk - 
- */
-int32_t
-unify_fsyncdir_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-
-	return 0;
-}
-
-/**
- * unify_fsyncdir -
- */
-int32_t
-unify_fsyncdir (call_frame_t *frame,
-		xlator_t *this,
-		fd_t *fd,
-		int32_t flags)
-{
-	UNIFY_CHECK_FD_AND_UNWIND_ON_ERR (fd);
-
-	STACK_WIND (frame, unify_fsyncdir_cbk,
-		    NS(this), NS(this)->fops->fsyncdir, fd, flags);
-
-	return 0;
-}
-
-/**
- * unify_lk_cbk - UNWIND frame with the proper return arguments.
- */
-int32_t
-unify_lk_cbk (call_frame_t *frame,
-	      void *cookie,
-	      xlator_t *this,
-	      int32_t op_ret,
-	      int32_t op_errno,
-	      struct flock *lock)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, lock);
-	return 0;
-}
-
-/**
- * unify_lk - Send it to all the storage nodes, (should be 1) which has file.
- */
-int32_t
-unify_lk (call_frame_t *frame,
-	  xlator_t *this,
-	  fd_t *fd,
-	  int32_t cmd,
-	  struct flock *lock)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame, unify_lk_cbk, child,
-		    child->fops->lk, fd, cmd, lock);
-
-	return 0;
-}
-
-
-int32_t
-unify_setxattr_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno);
-
-static int32_t
-unify_setxattr_file_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno)
-{
-	unify_private_t *private = this->private;
-	unify_local_t *local = frame->local;
-	xlator_t *sched_xl = NULL;
-	struct sched_ops *sched_ops = NULL;
-
-	if (op_ret == -1) {
-		if (!ENOTSUP)
-			gf_log (this->name, GF_LOG_ERROR,
-				"setxattr with XATTR_CREATE on ns: "
-				"path(%s) key(%s): %s",
-				local->loc1.path, local->name, 
-				strerror (op_errno));
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, op_ret, op_errno);
-		return 0;
-	} 
-
-	LOCK (&frame->lock);
-	{
-		local->failed = 0;
-		local->op_ret = 0;
-		local->op_errno = 0;
-		local->call_count = 1;
-	}
-	UNLOCK (&frame->lock);
-
-	/* schedule XATTR_CREATE on one of the child node */
-	sched_ops = private->sched_ops;
-    
-	/* Send create request to the scheduled node now */
-	sched_xl = sched_ops->schedule (this, local->name); 
-	if (!sched_xl) {
-		STACK_UNWIND (frame, -1, ENOTCONN);
-		return 0;
-	}
-
-	STACK_WIND (frame,
-		    unify_setxattr_cbk,
-		    sched_xl,
-		    sched_xl->fops->setxattr,
-		    &local->loc1,
-		    local->dict,
-		    local->flags);
-	return 0;
-}
-
-/**
- * unify_setxattr_cbk - When all the child nodes return, UNWIND frame.
- */
-int32_t
-unify_setxattr_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno)
-{
-	int32_t callcnt = 0;
-	unify_local_t *local = frame->local;
-	call_frame_t *prev_frame = cookie;
-	dict_t *dict = NULL;
-
-	LOCK (&frame->lock);
-	{
-		callcnt = --local->call_count;
-    
-		if (op_ret == -1) {
-			gf_log (this->name, (((op_errno == ENOENT) || 
-					      (op_errno == ENOTSUP))? 
-					     GF_LOG_DEBUG : GF_LOG_ERROR), 
-				"child(%s): path(%s): %s", 
-				prev_frame->this->name, 
-				(local->loc1.path)?local->loc1.path:"", 
-				strerror (op_errno));
-			if (local->failed == -1) {
-				local->failed = 1;
-			}
-			local->op_errno = op_errno;
-		} else {
-			local->failed = 0;
-			local->op_ret = op_ret;
-		}
-	}
-	UNLOCK (&frame->lock);
-  
-	if (!callcnt) {
-		if (local->failed && local->name && 
-		    ZR_FILE_CONTENT_REQUEST(local->name)) {      
-			dict = get_new_dict ();
-			dict_set (dict, local->dict->members_list->key, 
-				  data_from_dynptr(NULL, 0));
-			dict_ref (dict);
-
-			local->call_count = 1;
-
-			STACK_WIND (frame,
-				    unify_setxattr_file_cbk,
-				    NS(this),
-				    NS(this)->fops->setxattr,
-				    &local->loc1,
-				    dict,
-				    XATTR_CREATE);
-
-			dict_unref (dict);
-			return 0;
-		}
-    
-		unify_local_wipe (local);
-		STACK_UNWIND (frame, local->op_ret, local->op_errno);
-	}
-
-	return 0;
-}
-
-/**
- * unify_sexattr - This function should be sent to all the storage nodes, 
- *       which contains the file, (excluding namespace).
- */
-int32_t
-unify_setxattr (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		dict_t *dict,
-		int32_t flags)
-{
-	unify_private_t *priv = this->private;
-	unify_local_t *local = NULL;
-	int16_t *list = NULL;
-	int16_t index = 0;
-	int32_t call_count = 0;
-  	uint64_t tmp_list = 0;
-	data_pair_t *trav = dict->members_list;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (loc);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-	local->failed = -1;
-	loc_copy (&local->loc1, loc);
-
-	if (S_ISDIR (loc->inode->st_mode)) {
-
-		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 = 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 (S_ISDIR (loc->inode->st_mode)) {
-		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 (S_ISDIR (loc->inode->st_mode)) {
-		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 stat *preparent,
-                        struct stat *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 stat *buf,
-                 struct stat *preparent,
-                 struct stat *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.st_ino = local->st_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 stat *buf,
-                    struct stat *preparent,
-                    struct stat *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->st_ino = buf->st_ino;
-
-        local->oldpreparent = *preparent;
-        local->oldpostparent = *postparent;
-
-	list = CALLOC (1, sizeof (int16_t) * 3);
-	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 stat *preparent,
-                          struct stat *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 stat *buf,
-                   struct stat *preparent,
-                   struct stat *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.st_ino = local->st_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 stat *buf,
-                      struct stat *preparent,
-                      struct stat *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->st_ino = buf->st_ino;
-  
-        local->oldpreparent = *preparent;
-        local->oldpostparent = *postparent;
-
-	/* Start the mapping list */
-
-	list = CALLOC (1, sizeof (int16_t) * 3);
-	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 = 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 stat *preparent,
-                         struct stat *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.st_ino = local->st_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 stat *buf,
-                          struct stat *preoldparent,
-                          struct stat *postoldparent,
-                          struct stat *prenewparent,
-                          struct stat *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.st_ino = local->st_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 stat *buf,
-                  struct stat *preoldparent,
-                  struct stat *postoldparent,
-                  struct stat *prenewparent,
-                  struct stat *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 (!S_ISDIR (buf->st_mode))
-				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.st_ino = local->st_ino;
-		if (S_ISDIR (local->loc1.inode->st_mode)) {
-			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 = CALLOC (1, index * 2);
-				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;
-						}
-					}
-
-					FREE (tmp_list);
-					return 0;
-				}
-				if (tmp_list)
-					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 stat *buf,
-                     struct stat *preoldparent,
-                     struct stat *postoldparent,
-                     struct stat *prenewparent,
-                     struct stat *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->st_ino = buf->st_ino;
-
-        local->oldpreparent = *preoldparent;
-        local->oldpostparent = *postoldparent;
-        local->newpreparent = *prenewparent;
-        local->newpostparent = *postnewparent;
-
-	/* Everything is fine. */
-	if (S_ISDIR (buf->st_mode)) {
-		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 stat *buf,
-                struct stat *preparent,
-                struct stat *postparent)
-{
-	unify_local_t *local = frame->local;
-
-	if (op_ret >= 0) 
-		local->stbuf = *buf;
-	local->stbuf.st_ino = local->st_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 stat *buf,
-                   struct stat *preparent,
-                   struct stat *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->st_ino = buf->st_ino;
-
-        local->oldpreparent = *preparent;
-        local->oldpostparent = *postparent;
-
-	/* Send link request to the node now */
-	for (index = 0; list[index] != -1; index++) {
-		char need_break = (list[index+1] == -1);
-		if (priv->xl_array[list[index]] != NS (this)) {
-			STACK_WIND (frame,
-				    unify_link_cbk,
-				    priv->xl_array[list[index]],
-				    priv->xl_array[list[index]]->fops->link,
-				    &local->loc1,
-				    &local->loc2);
-                        break;
-		}
-		if (need_break)
-			break;
-	}
-
-	return 0;
-}
-
-/**
- * unify_link - 
- */
-int32_t
-unify_link (call_frame_t *frame,
-	    xlator_t *this,
-	    loc_t *oldloc,
-	    loc_t *newloc)
-{
-	unify_local_t *local = NULL;
-  	uint64_t tmp_list = 0;
-
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (oldloc);
-	UNIFY_CHECK_INODE_CTX_AND_UNWIND_ON_ERR (newloc);
-
-	/* Initialization */
-	INIT_LOCAL (frame, local);
-
-	loc_copy (&local->loc1, oldloc);
-	loc_copy (&local->loc2, newloc);
-
-	inode_ctx_get (oldloc->inode, this, &tmp_list);
-	local->list = (int16_t *)(long)tmp_list;
-
-	STACK_WIND (frame,
-		    unify_ns_link_cbk,
-		    NS(this),
-		    NS(this)->fops->link,
-		    oldloc,
-		    newloc);
-
-	return 0;
-}
-
-
-/**
- * unify_checksum_cbk - 
- */
-int32_t
-unify_checksum_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    uint8_t *fchecksum,
-		    uint8_t *dchecksum)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, fchecksum, dchecksum);
-
-	return 0;
-}
-
-/**
- * unify_checksum - 
- */
-int32_t
-unify_checksum (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		int32_t flag)
-{
-	STACK_WIND (frame,
-		    unify_checksum_cbk,
-		    NS(this),
-		    NS(this)->fops->checksum,
-		    loc,
-		    flag);
-
-	return 0;
-}
-
-
-/**
- * unify_finodelk_cbk - 
- */
-int
-unify_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-/**
- * unify_finodelk
- */
-int
-unify_finodelk (call_frame_t *frame, xlator_t *this,
-		const char *volume, fd_t *fd, int cmd, struct flock *flock)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame, unify_finodelk_cbk,
-		    child, child->fops->finodelk,
-		    volume, fd, cmd, flock);
-
-	return 0;
-}
-
-
-
-/**
- * unify_fentrylk_cbk - 
- */
-int
-unify_fentrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-/**
- * unify_fentrylk
- */
-int
-unify_fentrylk (call_frame_t *frame, xlator_t *this,
-		const char *volume, fd_t *fd, const char *basename,
-		entrylk_cmd cmd, entrylk_type type)
-		
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame, unify_fentrylk_cbk,
-		    child, child->fops->fentrylk,
-		    volume, fd, basename, cmd, type);
-
-	return 0;
-}
-
-
-
-/**
- * unify_fxattrop_cbk - 
- */
-int
-unify_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		    int32_t op_ret, int32_t op_errno, dict_t *xattr)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, xattr);
-	return 0;
-}
-
-/**
- * unify_fxattrop
- */
-int
-unify_fxattrop (call_frame_t *frame, xlator_t *this,
-		fd_t *fd, gf_xattrop_flags_t optype, dict_t *xattr)
-{
-	UNIFY_CHECK_FD_CTX_AND_UNWIND_ON_ERR (fd);
-	xlator_t *child = NULL;
-	uint64_t tmp_child = 0;
-
-	fd_ctx_get (fd, this, &tmp_child);
-	child = (xlator_t *)(long)tmp_child;		     
-
-	STACK_WIND (frame, unify_fxattrop_cbk,
-		    child, child->fops->fxattrop,
-		    fd, optype, xattr);
-
-	return 0;
-}
-
-
-/**
- * unify_inodelk_cbk - 
- */
-int
-unify_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		   int32_t op_ret, int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-
-/**
- * unify_inodelk
- */
-int
-unify_inodelk (call_frame_t *frame, xlator_t *this,
-	       const char *volume, loc_t *loc, int cmd, struct flock *flock)
-{
-	xlator_t *child = NULL;
-
-	child = unify_loc_subvol (loc, this);
-
-	STACK_WIND (frame, unify_inodelk_cbk,
-		    child, child->fops->inodelk,
-		    volume, loc, cmd, flock);
-
-	return 0;
-}
-
-
-
-/**
- * unify_entrylk_cbk - 
- */
-int
-unify_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		   int32_t op_ret, int32_t op_errno)
-{
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-/**
- * unify_entrylk
- */
-int
-unify_entrylk (call_frame_t *frame, xlator_t *this,
-	       const char *volume, loc_t *loc, const char *basename,
-	       entrylk_cmd cmd, entrylk_type type)
-		
-{
-	xlator_t *child = NULL;
-
-	child = unify_loc_subvol (loc, this);
-
-	STACK_WIND (frame, unify_entrylk_cbk,
-		    child, child->fops->entrylk,
-		    volume, loc, basename, cmd, type);
-
-	return 0;
-}
-
-
-
-/**
- * unify_xattrop_cbk - 
- */
-int
-unify_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-		   int32_t op_ret, int32_t op_errno, dict_t *xattr)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, xattr);
-	return 0;
-}
-
-/**
- * unify_xattrop
- */
-int
-unify_xattrop (call_frame_t *frame, xlator_t *this,
-		loc_t *loc, gf_xattrop_flags_t optype, dict_t *xattr)
-{
-	xlator_t *child = NULL;
-
-	child = unify_loc_subvol (loc, this);
-
-	STACK_WIND (frame, unify_xattrop_cbk,
-		    child, child->fops->xattrop,
-		    loc, optype, xattr);
-
-	return 0;
-}
-
-int
-unify_forget (xlator_t *this,
-              inode_t *inode)
-{
-        int16_t *list = NULL;
-        uint64_t tmp_list = 0;
-
-        if (inode->st_mode && (!S_ISDIR(inode->st_mode))) {
-                inode_ctx_get (inode, this, &tmp_list);
-                if (tmp_list) {
-                        list = (int16_t *)(long)tmp_list;
-                        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;
-}
-
-/** 
- * 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 = CALLOC (1, sizeof (*_private));
-	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");
-		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");
-		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 = CALLOC (1, 
-					     sizeof (xlator_t) * (count + 1));
-		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;
-	{
-		/* 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");
-			FREE (_private);
-			return -1;
-		}
-
-		ret = 0;
-
-		/* This section is required because some fops may look 
-		 * for 'xl->parent' variable 
-		 */
-		xlparent = CALLOC (1, sizeof (*xlparent));
-		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;
-		}
-		/* Initialize the namespace volume */
-		if (!ns_xl->ready) {
-			ret = xlator_tree_init (ns_xl);
-			if (ret) {
-				gf_log (this->name, GF_LOG_ERROR, 
-					"initializing namespace node failed, "
-					"Exiting");
-			FREE (_private);
-			return -1;
-			}
-		}
-	}
-
-	/* 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);
-	FREE (priv->xl_array);
-	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_mops mops = {
-};
-
-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 b81946697..000000000
--- a/xlators/cluster/unify/src/unify.h
+++ /dev/null
@@ -1,145 +0,0 @@
-/*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
-
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
-
-#ifndef _UNIFY_H
-#define _UNIFY_H
-
-#include "scheduler.h"
-#include "list.h"
-
-#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 = CALLOC (1, sizeof (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 stat stbuf;
-        struct stat stpre;
-        struct stat stpost;
-	struct statvfs statvfs_buf;
-	struct timespec tv[2];
-	char *name;
-	int32_t revalidate;
-
-	ino_t st_ino;
-	nlink_t st_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 stat poststbuf;
-        /* When not used for rename, old*
-         * are used as the attrs for the current
-         * parent directory.
-         */
-        struct stat oldpreparent;
-        struct stat oldpostparent;
-        struct stat newpreparent;
-        struct stat newpostparent;
-        int32_t wbflags;
-};
-typedef struct _unify_local_t unify_local_t;
-
-int32_t zr_unify_self_heal (call_frame_t *frame,
-			    xlator_t *this,
-			    unify_local_t *local);
-
-#endif /* _UNIFY_H */
diff --git a/xlators/debug/error-gen/src/Makefile.am b/xlators/debug/error-gen/src/Makefile.am
index f353b61e6..5075c59a8 100644
--- a/xlators/debug/error-gen/src/Makefile.am
+++ b/xlators/debug/error-gen/src/Makefile.am
@@ -2,15 +2,16 @@
 xlator_LTLIBRARIES = error-gen.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/debug
 
-error_gen_la_LDFLAGS = -module -avoidversion
+error_gen_la_LDFLAGS = -module -avoid-version
 
 error_gen_la_SOURCES = error-gen.c
 error_gen_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-noinst_HEADERS = error-gen.h
+noinst_HEADERS = error-gen.h error-gen-mem-types.h
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/debug/error-gen/src/error-gen-mem-types.h b/xlators/debug/error-gen/src/error-gen-mem-types.h
new file mode 100644
index 000000000..f02280535
--- /dev/null
+++ b/xlators/debug/error-gen/src/error-gen-mem-types.h
@@ -0,0 +1,20 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef __ERROR_GEN_MEM_TYPES_H__
+#define __ERROR_GEN_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_error_gen_mem_types_ {
+        gf_error_gen_mt_eg_t = gf_common_mt_end + 1,
+        gf_error_gen_mt_end
+};
+#endif
diff --git a/xlators/debug/error-gen/src/error-gen.c b/xlators/debug/error-gen/src/error-gen.c
index 8d97a46f1..ec0874b35 100644
--- a/xlators/debug/error-gen/src/error-gen.c
+++ b/xlators/debug/error-gen/src/error-gen.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -24,166 +14,165 @@
 
 #include "xlator.h"
 #include "error-gen.h"
+#include "statedump.h"
 
 sys_error_t error_no_list[] = {
-        [ERR_LOOKUP]            = { .error_no_count = 4,
+        [GF_FOP_LOOKUP]            = { .error_no_count = 4,
                                     .error_no = {ENOENT,ENOTDIR,
                                                  ENAMETOOLONG,EAGAIN}},
-        [ERR_STAT]              = { .error_no_count = 7,
+        [GF_FOP_STAT]              = { .error_no_count = 7,
                                     .error_no = {EACCES,EBADF,EFAULT,
                                                  ENAMETOOLONG,ENOENT,
                                                  ENOMEM,ENOTDIR}},
-        [ERR_READLINK]          = { .error_no_count = 8,
+        [GF_FOP_READLINK]          = { .error_no_count = 8,
                                     .error_no = {EACCES,EFAULT,EINVAL,EIO,
                                                  ENAMETOOLONG,ENOENT,ENOMEM,
                                                  ENOTDIR}},
-        [ERR_MKNOD]             = { .error_no_count = 11,
+        [GF_FOP_MKNOD]             = { .error_no_count = 11,
                                     .error_no = {EACCES,EEXIST,EFAULT,
                                                  EINVAL,ENAMETOOLONG,
                                                  ENOENT,ENOMEM,ENOSPC,
                                                  ENOTDIR,EPERM,EROFS}},
-        [ERR_MKDIR]             = { .error_no_count = 10,
+        [GF_FOP_MKDIR]             = { .error_no_count = 10,
                                     .error_no = {EACCES,EEXIST,EFAULT,
                                                  ENAMETOOLONG,ENOENT,
                                                  ENOMEM,ENOSPC,ENOTDIR,
 						 EPERM,EROFS}},
-        [ERR_UNLINK]            = { .error_no_count = 10,
+        [GF_FOP_UNLINK]            = { .error_no_count = 10,
                                     .error_no = {EACCES,EBUSY,EFAULT,EIO,
                                                  EISDIR,ENAMETOOLONG,
                                                  ENOENT,ENOMEM,ENOTDIR,
                                                  EPERM,EROFS}},
-        [ERR_RMDIR]             = { .error_no_count = 8,
+        [GF_FOP_RMDIR]             = { .error_no_count = 8,
                                     .error_no = {EACCES,EBUSY,EFAULT,
                                                  ENOMEM,ENOTDIR,ENOTEMPTY,
                                                  EPERM,EROFS}},
-        [ERR_SYMLINK]           = { .error_no_count = 11,
+        [GF_FOP_SYMLINK]           = { .error_no_count = 11,
                                     .error_no = {EACCES,EEXIST,EFAULT,EIO,
                                                  ENAMETOOLONG,ENOENT,ENOMEM,
                                                  ENOSPC,ENOTDIR,EPERM,
                                                  EROFS}},
-        [ERR_RENAME]            = { .error_no_count = 13,
+        [GF_FOP_RENAME]            = { .error_no_count = 13,
                                     .error_no = {EACCES,EBUSY,EFAULT,
                                                  EINVAL,EISDIR,EMLINK,
                                                  ENAMETOOLONG,ENOENT,ENOMEM,
                                                  ENOSPC,ENOTDIR,EEXIST,
                                                  EXDEV}},
-        [ERR_LINK]              = { .error_no_count = 13,
+        [GF_FOP_LINK]              = { .error_no_count = 13,
                                     .error_no = {EACCES,EFAULT,EEXIST,EIO,
                                                  EMLINK,ENAMETOOLONG,
                                                  ENOENT,ENOMEM,ENOSPC,
                                                  ENOTDIR,EPERM,EROFS,
                                                  EXDEV}},
-        [ERR_TRUNCATE]          = { .error_no_count = 10,
+        [GF_FOP_TRUNCATE]          = { .error_no_count = 10,
                                     .error_no = {EACCES,EFAULT,EFBIG,
                                                  EINTR,EINVAL,EIO,EISDIR,
                                                  ENAMETOOLONG,ENOENT,
                                                  EISDIR}},
-        [ERR_CREATE]            = {.error_no_count = 10,
+        [GF_FOP_CREATE]            = {.error_no_count = 10,
                                    .error_no = {EACCES,EEXIST,EFAULT,
                                                 EISDIR,EMFILE,ENAMETOOLONG,
                                                 ENFILE,ENODEV,ENOENT,
                                                 ENODEV}},
-        [ERR_OPEN]              = { .error_no_count = 10,
+        [GF_FOP_OPEN]              = { .error_no_count = 10,
                                     .error_no = {EACCES,EEXIST,EFAULT,
                                                  EISDIR,EMFILE,
                                                  ENAMETOOLONG,ENFILE,
                                                  ENODEV,ENOENT,ENOMEM}},
-        [ERR_READV]             = { .error_no_count = 5,
+        [GF_FOP_READ]              = { .error_no_count = 5,
                                     .error_no = {EINVAL,EBADF,EFAULT,EISDIR,
                                                  ENAMETOOLONG}},
-        [ERR_WRITEV]            = { .error_no_count = 5,
+        [GF_FOP_WRITE]             = { .error_no_count = 7,
                                     .error_no = {EINVAL,EBADF,EFAULT,EISDIR,
-                                                 ENAMETOOLONG}},
-        [ERR_STATFS]            = {.error_no_count = 10,
+                                                 ENAMETOOLONG,ENOSPC,
+						 GF_ERROR_SHORT_WRITE}},
+        [GF_FOP_STATFS]            = {.error_no_count = 10,
                                    .error_no = {EACCES,EBADF,EFAULT,EINTR,
                                                 EIO,ENAMETOOLONG,ENOENT,
                                                 ENOMEM,ENOSYS,ENOTDIR}},
-        [ERR_FLUSH]             = { .error_no_count = 5,
+        [GF_FOP_FLUSH]             = { .error_no_count = 5,
                                     .error_no = {EACCES,EFAULT,
                                                  ENAMETOOLONG,ENOSYS,
                                                  ENOENT}},
-        [ERR_FSYNC]             = { .error_no_count = 4,
+        [GF_FOP_FSYNC]             = { .error_no_count = 4,
                                     .error_no = {EBADF,EIO,EROFS,EINVAL}},
-        [ERR_SETXATTR]          = { .error_no_count = 4,
+        [GF_FOP_SETXATTR]          = { .error_no_count = 4,
                                     .error_no = {EACCES,EBADF,EINTR,
                                                  ENAMETOOLONG}},
-        [ERR_GETXATTR]          = { .error_no_count = 4,
+        [GF_FOP_GETXATTR]          = { .error_no_count = 4,
                                     .error_no = {EACCES,EBADF,ENAMETOOLONG,
                                                  EINTR}},
-        [ERR_REMOVEXATTR]       = { .error_no_count = 4,
+        [GF_FOP_REMOVEXATTR]       = { .error_no_count = 4,
                                     .error_no = {EACCES,EBADF,ENAMETOOLONG,
                                                  EINTR}},
-        [ERR_OPENDIR]           = { .error_no_count = 8,
+        [GF_FOP_FSETXATTR]          = { .error_no_count = 4,
+                                        .error_no = {EACCES,EBADF,EINTR,
+                                                     ENAMETOOLONG}},
+        [GF_FOP_FGETXATTR]          = { .error_no_count = 4,
+                                        .error_no = {EACCES,EBADF,ENAMETOOLONG,
+                                                     EINTR}},
+        [GF_FOP_FREMOVEXATTR]       = { .error_no_count = 4,
+                                        .error_no = {EACCES,EBADF,ENAMETOOLONG,
+                                                     EINTR}},
+        [GF_FOP_OPENDIR]           = { .error_no_count = 8,
                                     .error_no = {EACCES,EEXIST,EFAULT,
                                                  EISDIR,EMFILE,
                                                  ENAMETOOLONG,ENFILE,
                                                  ENODEV}},
-        [ERR_READDIR]           = { .error_no_count = 5,
+        [GF_FOP_READDIR]           = { .error_no_count = 5,
                                     .error_no = {EINVAL,EACCES,EBADF,
                                                  EMFILE,ENOENT}},
-        [ERR_READDIRP]          = { .error_no_count = 5,
+        [GF_FOP_READDIRP]          = { .error_no_count = 5,
                                     .error_no = {EINVAL,EACCES,EBADF,
                                                  EMFILE,ENOENT}},
-        [ERR_GETDENTS]          = { .error_no_count = 5,
-                                    .error_no = {EBADF,EFAULT,EINVAL,
-                                                 ENOENT,ENOTDIR}},
-        [ERR_FSYNCDIR]          = { .error_no_count = 4,
+        [GF_FOP_FSYNCDIR]          = { .error_no_count = 4,
                                     .error_no = {EBADF,EIO,EROFS,EINVAL}},
-        [ERR_ACCESS]            = { .error_no_count = 8,
+        [GF_FOP_ACCESS]            = { .error_no_count = 8,
                                     .error_no = {EACCES,ENAMETOOLONG,
                                                  ENOENT,ENOTDIR,EROFS,
                                                  EFAULT,EINVAL,EIO}},
-        [ERR_FTRUNCATE]         = { .error_no_count = 9,
+        [GF_FOP_FTRUNCATE]         = { .error_no_count = 9,
                                     .error_no = {EACCES,EFAULT,EFBIG,
                                                  EINTR,EINVAL,EIO,EISDIR,
                                                  ENAMETOOLONG,ENOENT}},
-        [ERR_FSTAT]             = { .error_no_count = 7,
+        [GF_FOP_FSTAT]             = { .error_no_count = 7,
                                     .error_no = {EACCES,EBADF,EFAULT,
                                                  ENAMETOOLONG,ENOENT,
                                                  ENOMEM,ENOTDIR}},
-        [ERR_LK]                = { .error_no_count = 4,
+        [GF_FOP_LK]                = { .error_no_count = 4,
                                     .error_no = {EACCES,EFAULT,ENOENT,
                                                  EINTR}},
-        [ERR_SETDENTS]          = { .error_no_count = 4,
-                                    .error_no = {EACCES,EBADF,EINTR,
-                                                 ENAMETOOLONG}},
-        [ERR_CHECKSUM]          = { .error_no_count = 4,
-                                    .error_no = {EACCES,EBADF,
-                                                 ENAMETOOLONG,EINTR}},
-        [ERR_XATTROP]           = { .error_no_count = 5,
+        [GF_FOP_XATTROP]           = { .error_no_count = 5,
                                     .error_no = {EACCES,EFAULT,
                                                  ENAMETOOLONG,ENOSYS,
                                                  ENOENT}},
-        [ERR_FXATTROP]          = { .error_no_count = 4,
+        [GF_FOP_FXATTROP]          = { .error_no_count = 4,
                                     .error_no = {EBADF,EIO,EROFS,EINVAL}},
-        [ERR_INODELK]           = { .error_no_count = 4,
+        [GF_FOP_INODELK]           = { .error_no_count = 4,
                                     .error_no = {EACCES,EBADF,EINTR,
                                                  ENAMETOOLONG}},
-        [ERR_FINODELK]          = { .error_no_count = 4,
+        [GF_FOP_FINODELK]          = { .error_no_count = 4,
                                     .error_no = {EACCES,EBADF,EINTR,
                                                  ENAMETOOLONG}},
-        [ERR_ENTRYLK]           = { .error_no_count = 4,
+        [GF_FOP_ENTRYLK]           = { .error_no_count = 4,
                                     .error_no = {EACCES,EBADF,
                                                  ENAMETOOLONG,EINTR}},
-        [ERR_FENTRYLK]          = { .error_no_count = 10,
+        [GF_FOP_FENTRYLK]          = { .error_no_count = 10,
                                     .error_no = {EACCES,EEXIST,EFAULT,
                                                  EISDIR,EMFILE,
                                                  ENAMETOOLONG,ENFILE,
                                                  ENODEV,ENOENT,ENOMEM}},
-        [ERR_SETATTR]           = {.error_no_count = 11,
+        [GF_FOP_SETATTR]           = {.error_no_count = 11,
                                     .error_no = {EACCES,EFAULT,EIO,
                                                  ENAMETOOLONG,ENOENT,
                                                  ENOMEM,ENOTDIR,EPERM,
                                                  EROFS,EBADF,EIO}},
-        [ERR_FSETATTR]          = { .error_no_count = 11,
+        [GF_FOP_FSETATTR]          = { .error_no_count = 11,
                                     .error_no = {EACCES,EFAULT,EIO,
                                                  ENAMETOOLONG,ENOENT,
                                                  ENOMEM,ENOTDIR,EPERM,
                                                  EROFS,EBADF,EIO}},
-        [ERR_STATS]             = { .error_no_count = 4,
-                                    .error_no = {EACCES,EBADF,ENAMETOOLONG,
-                                                 EINTR}},
-        [ERR_GETSPEC]           = { .error_no_count = 4,
+        [GF_FOP_GETSPEC]           = { .error_no_count = 4,
                                     .error_no = {EACCES,EBADF,ENAMETOOLONG,
                                                  EINTR}}
 };
@@ -193,7 +182,7 @@ generate_rand_no (int op_no)
 {
         int             rand_no = 0;
 
-        if (op_no < NO_OF_FOPS)
+        if (op_no < GF_FOP_MAXVALUE)
                 rand_no = rand () % error_no_list[op_no].error_no_count;
         return rand_no;
 }
@@ -249,6 +238,8 @@ conv_errno_to_int (char **error_no)
                 return EINTR;
         else if (!strcmp ((*error_no), "EFBIG"))
                 return EFBIG;
+	else if (!strcmp((*error_no), "GF_ERROR_SHORT_WRITE"))
+		return GF_ERROR_SHORT_WRITE;
         else
                 return EAGAIN;
 }
@@ -257,89 +248,87 @@ int
 get_fop_int (char **op_no_str)
 {
         if (!strcmp ((*op_no_str), "lookup"))
-                return ERR_LOOKUP;
+                return GF_FOP_LOOKUP;
         else if (!strcmp ((*op_no_str), "stat"))
-                return ERR_STAT;
+                return GF_FOP_STAT;
         else if (!strcmp ((*op_no_str), "readlink"))
-                return ERR_READLINK;
+                return GF_FOP_READLINK;
         else if (!strcmp ((*op_no_str), "mknod"))
-                return ERR_MKNOD;
+                return GF_FOP_MKNOD;
         else if (!strcmp ((*op_no_str), "mkdir"))
-                return ERR_MKDIR;
+                return GF_FOP_MKDIR;
         else if (!strcmp ((*op_no_str), "unlink"))
-                return ERR_UNLINK;
+                return GF_FOP_UNLINK;
         else if (!strcmp ((*op_no_str), "rmdir"))
-                return ERR_RMDIR;
+                return GF_FOP_RMDIR;
         else if (!strcmp ((*op_no_str), "symlink"))
-                return ERR_SYMLINK;
+                return GF_FOP_SYMLINK;
         else if (!strcmp ((*op_no_str), "rename"))
-                return ERR_RENAME;
+                return GF_FOP_RENAME;
         else if (!strcmp ((*op_no_str), "link"))
-                return ERR_LINK;
+                return GF_FOP_LINK;
         else if (!strcmp ((*op_no_str), "truncate"))
-                return ERR_TRUNCATE;
+                return GF_FOP_TRUNCATE;
         else if (!strcmp ((*op_no_str), "create"))
-                return ERR_CREATE;
+                return GF_FOP_CREATE;
         else if (!strcmp ((*op_no_str), "open"))
-                return ERR_OPEN;
+                return GF_FOP_OPEN;
         else if (!strcmp ((*op_no_str), "readv"))
-                return ERR_READV;
+                return GF_FOP_READ;
         else if (!strcmp ((*op_no_str), "writev"))
-                return ERR_WRITEV;
+                return GF_FOP_WRITE;
         else if (!strcmp ((*op_no_str), "statfs"))
-                return ERR_STATFS;
+                return GF_FOP_STATFS;
         else if (!strcmp ((*op_no_str), "flush"))
-                return ERR_FLUSH;
+                return GF_FOP_FLUSH;
         else if (!strcmp ((*op_no_str), "fsync"))
-                return ERR_FSYNC;
+                return GF_FOP_FSYNC;
         else if (!strcmp ((*op_no_str), "setxattr"))
-                return ERR_SETXATTR;
+                return GF_FOP_SETXATTR;
         else if (!strcmp ((*op_no_str), "getxattr"))
-                return ERR_GETXATTR;
+                return GF_FOP_GETXATTR;
         else if (!strcmp ((*op_no_str), "removexattr"))
-                return ERR_REMOVEXATTR;
+                return GF_FOP_REMOVEXATTR;
+        else if (!strcmp ((*op_no_str), "fsetxattr"))
+                return GF_FOP_FSETXATTR;
+        else if (!strcmp ((*op_no_str), "fgetxattr"))
+                return GF_FOP_FGETXATTR;
+        else if (!strcmp ((*op_no_str), "fremovexattr"))
+                return GF_FOP_FREMOVEXATTR;
         else if (!strcmp ((*op_no_str), "opendir"))
-                return ERR_OPENDIR;
+                return GF_FOP_OPENDIR;
         else if (!strcmp ((*op_no_str), "readdir"))
-                return ERR_READDIR;
+                return GF_FOP_READDIR;
         else if (!strcmp ((*op_no_str), "readdirp"))
-                return ERR_READDIRP;
-	else if (!strcmp ((*op_no_str), "getdents"))
-                return ERR_GETDENTS;
+                return GF_FOP_READDIRP;
         else if (!strcmp ((*op_no_str), "fsyncdir"))
-                return ERR_FSYNCDIR;
+                return GF_FOP_FSYNCDIR;
         else if (!strcmp ((*op_no_str), "access"))
-                return ERR_ACCESS;
+                return GF_FOP_ACCESS;
         else if (!strcmp ((*op_no_str), "ftruncate"))
-                return ERR_FTRUNCATE;
+                return GF_FOP_FTRUNCATE;
         else if (!strcmp ((*op_no_str), "fstat"))
-                return ERR_FSTAT;
+                return GF_FOP_FSTAT;
         else if (!strcmp ((*op_no_str), "lk"))
-                return ERR_LK;
-        else if (!strcmp ((*op_no_str), "setdents"))
-                return ERR_SETDENTS;
-        else if (!strcmp ((*op_no_str), "checksum"))
-                return ERR_CHECKSUM;
+                return GF_FOP_LK;
         else if (!strcmp ((*op_no_str), "xattrop"))
-                return ERR_XATTROP;
+                return GF_FOP_XATTROP;
         else if (!strcmp ((*op_no_str), "fxattrop"))
-                return ERR_FXATTROP;
+                return GF_FOP_FXATTROP;
         else if (!strcmp ((*op_no_str), "inodelk"))
-                return ERR_INODELK;
+                return GF_FOP_INODELK;
         else if (!strcmp ((*op_no_str), "finodelk"))
-                return ERR_FINODELK;
+                return GF_FOP_FINODELK;
         else if (!strcmp ((*op_no_str), "etrylk"))
-                return ERR_ENTRYLK;
+                return GF_FOP_ENTRYLK;
         else if (!strcmp ((*op_no_str), "fentrylk"))
-                return ERR_FENTRYLK;
+                return GF_FOP_FENTRYLK;
         else if (!strcmp ((*op_no_str), "setattr"))
-                return ERR_SETATTR;
+                return GF_FOP_SETATTR;
         else if (!strcmp ((*op_no_str), "fsetattr"))
-                return ERR_FSETATTR;
-        else if (!strcmp ((*op_no_str), "stats"))
-                return ERR_STATS;
+                return GF_FOP_FSETATTR;
         else if (!strcmp ((*op_no_str), "getspec"))
-                return ERR_GETSPEC;
+                return GF_FOP_GETSPEC;
 	else
                 return -1;
 }
@@ -376,1937 +365,1732 @@ error_gen (xlator_t *this, int op_no)
                 else {
 
                         rand_no = generate_rand_no (op_no);
-                        if (op_no >= NO_OF_FOPS)
+                        if (op_no >= GF_FOP_MAXVALUE)
                                 op_no = 0;
                         if (rand_no >= error_no_list[op_no].error_no_count)
                                 rand_no = 0;
                         ret = error_no_list[op_no].error_no[rand_no];
                 }
+                if (egp->random_failure == _gf_true)
+                        egp->failure_iter_no = 3 + (rand () % GF_UNIVERSAL_ANSWER);
         }
         return ret;
 }
 
-static int32_t
-error_gen_lookup_cbk (call_frame_t *frame,
-		      void *cookie,
-		      xlator_t *this,
-		      int32_t op_ret,
-		      int32_t op_errno,
-		      inode_t *inode,
-		      struct stat *buf,
-		      dict_t *dict,
-                      struct stat *postparent)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      inode,
-		      buf,
-		      dict, postparent);
-	return 0;
+
+int
+error_gen_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		      int32_t op_ret, int32_t op_errno, inode_t *inode,
+		      struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+{
+	STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode,
+                             buf, xdata, postparent);
+        return 0;
 }
 
-int32_t
-error_gen_lookup (call_frame_t *frame,
-		  xlator_t *this,
-		  loc_t *loc,
-		  dict_t *xattr_req)
+
+int
+error_gen_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                  dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_LOOKUP];
+        enable = egp->enable[GF_FOP_LOOKUP];
 
         if (enable)
-                op_errno = error_gen (this, ERR_LOOKUP);
+                op_errno = error_gen (this, GF_FOP_LOOKUP);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, NULL, NULL,
-                              NULL);
-		return 0;
+		STACK_UNWIND_STRICT (lookup, frame, -1, op_errno, NULL, NULL, NULL,
+                                     NULL);
+        return 0;
 	}
-	STACK_WIND (frame,
-		    error_gen_lookup_cbk,
+
+	STACK_WIND (frame, error_gen_lookup_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->lookup,
-		    loc,
-		    xattr_req);
-	return 0;
+		    loc, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_forget (xlator_t *this,
-		  inode_t *inode)
-{
-	return 0;
-}
 
-int32_t
-error_gen_stat_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    struct stat *buf)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      buf);
-	return 0;
+int
+error_gen_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		    int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (stat, frame, op_ret, op_errno, buf, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_stat (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc)
+int
+error_gen_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_STAT];
+        enable = egp->enable[GF_FOP_STAT];
 
         if (enable)
-                op_errno = error_gen (this, ERR_STAT);
+                op_errno = error_gen (this, GF_FOP_STAT);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (stat, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
-	STACK_WIND (frame,
-		    error_gen_stat_cbk,
+
+	STACK_WIND (frame, error_gen_stat_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->stat,
-		    loc);
-	return 0;
+		    loc, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_setattr_cbk (call_frame_t *frame,
-                       void *cookie,
-                       xlator_t *this,
-                       int32_t op_ret,
-                       int32_t op_errno,
-                       struct stat *preop,
-                       struct stat *postop)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      preop, postop);
-	return 0;
+
+int
+error_gen_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno,
+                       struct iatt *preop, struct iatt *postop, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (setattr, frame, op_ret, op_errno, preop, postop, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_setattr (call_frame_t *frame,
-                   xlator_t *this,
-                   loc_t *loc,
-                   struct stat *stbuf,
-                   int32_t valid)
+
+int
+error_gen_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                   struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_SETATTR];
+        enable = egp->enable[GF_FOP_SETATTR];
 
         if (enable)
-                op_errno = error_gen (this, ERR_SETATTR);
+                op_errno = error_gen (this, GF_FOP_SETATTR);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (setattr, frame, -1, op_errno, NULL, NULL, xdata);
+        return 0;
 	}
-	STACK_WIND (frame,
-		    error_gen_setattr_cbk,
+
+	STACK_WIND (frame, error_gen_setattr_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->setattr,
-		    loc,
-		    stbuf, valid);
-	return 0;
+		    loc, stbuf, valid, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_fsetattr (call_frame_t *frame,
-                    xlator_t *this,
-                    fd_t *fd,
-                    struct stat *stbuf,
-                    int32_t valid)
+
+int
+error_gen_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                    struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
 	int             op_errno = 0;
         eg_t            *egp = NULL;
         int             enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_FSETATTR];
+        enable = egp->enable[GF_FOP_FSETATTR];
 
         if (enable)
-                op_errno = error_gen (this, ERR_FSETATTR);
+                op_errno = error_gen (this, GF_FOP_FSETATTR);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (fsetattr, frame, -1, op_errno, NULL, NULL, xdata);
+        return 0;
 	}
-	STACK_WIND (frame,
-		    error_gen_setattr_cbk,
+
+	STACK_WIND (frame, error_gen_setattr_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->fsetattr,
-		    fd,
-		    stbuf, valid);
-	return 0;
+		    fd, stbuf, valid, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_truncate_cbk (call_frame_t *frame,
-			void *cookie,
-			xlator_t *this,
-			int32_t op_ret,
-			int32_t op_errno,
-			struct stat *prebuf,
-                        struct stat *postbuf)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      prebuf,
-                      postbuf);
-	return 0;
+
+int
+error_gen_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+			int32_t op_ret, int32_t op_errno,
+			struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno,
+                             prebuf, postbuf, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_truncate (call_frame_t *frame,
-		    xlator_t *this,
-		    loc_t *loc,
-		    off_t offset)
+
+int
+error_gen_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                    off_t offset, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_TRUNCATE];
+        enable = egp->enable[GF_FOP_TRUNCATE];
 
         if (enable)
-                op_errno = error_gen (this, ERR_TRUNCATE);
+                op_errno = error_gen (this, GF_FOP_TRUNCATE);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL,
-                              NULL, NULL,  /* pre & post old parent attr */
-                              NULL, NULL); /* pre & post new parent attr */
-		return 0;
+		STACK_UNWIND_STRICT (truncate, frame, -1, op_errno,
+                                     NULL, NULL, xdata);
+        return 0;
 	}
-	STACK_WIND (frame,
-		    error_gen_truncate_cbk,
+
+	STACK_WIND (frame, error_gen_truncate_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->truncate,
-		    loc,
-		    offset);
-	return 0;
+		    loc, offset, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_ftruncate_cbk (call_frame_t *frame,
-			 void *cookie,
-			 xlator_t *this,
-			 int32_t op_ret,
-			 int32_t op_errno,
-			 struct stat *prebuf,
-                         struct stat *postbuf)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      prebuf,
-                      postbuf);
-	return 0;
+
+int
+error_gen_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+			 int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                         struct iatt *postbuf, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno,
+                             prebuf, postbuf, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_ftruncate (call_frame_t *frame,
-		     xlator_t *this,
-		     fd_t *fd,
-		     off_t offset)
+
+int
+error_gen_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd,
+		     off_t offset, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp =NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_FTRUNCATE];
+        enable = egp->enable[GF_FOP_FTRUNCATE];
 
         if (enable)
-                op_errno = error_gen (this, ERR_FTRUNCATE);
+                op_errno = error_gen (this, GF_FOP_FTRUNCATE);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (ftruncate, frame, -1, op_errno,
+                                     NULL, NULL, xdata);
+        return 0;
 	}
-	STACK_WIND (frame,
-		    error_gen_ftruncate_cbk,
+
+	STACK_WIND (frame, error_gen_ftruncate_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->ftruncate,
-		    fd,
-		    offset);
-	return 0;
+		    fd, offset, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_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;
+
+int
+error_gen_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		      int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (access, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_access (call_frame_t *frame,
-		  xlator_t *this,
-		  loc_t *loc,
-		  int32_t mask)
+
+int
+error_gen_access (call_frame_t *frame, xlator_t *this, loc_t *loc,
+		  int32_t mask, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_ACCESS];
+        enable = egp->enable[GF_FOP_ACCESS];
 
         if (enable)
-                op_errno = error_gen (this, ERR_ACCESS);
+                op_errno = error_gen (this, GF_FOP_ACCESS);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (access, frame, -1, op_errno, xdata);
+        return 0;
 	}
-	STACK_WIND (frame,
-		    error_gen_access_cbk,
+
+	STACK_WIND (frame, error_gen_access_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->access,
-		    loc,
-		    mask);
-	return 0;
+		    loc, mask, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_readlink_cbk (call_frame_t *frame,
-			void *cookie,
-			xlator_t *this,
-			int32_t op_ret,
-			int32_t op_errno,
-			const char *path,
-                        struct stat *sbuf)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      path,
-                      sbuf);
-	return 0;
+
+int
+error_gen_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+			int32_t op_ret, int32_t op_errno,
+			const char *path, struct iatt *sbuf, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, path, sbuf, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_readlink (call_frame_t *frame,
-		    xlator_t *this,
-		    loc_t *loc,
-		    size_t size)
+
+int
+error_gen_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                    size_t size, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_READLINK];
+        enable = egp->enable[GF_FOP_READLINK];
 
         if (enable)
-                op_errno = error_gen (this, ERR_READLINK);
+                op_errno = error_gen (this, GF_FOP_READLINK);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (readlink, frame, -1, op_errno, NULL, NULL, xdata);
+        return 0;
 	}
-	STACK_WIND (frame,
-		    error_gen_readlink_cbk,
+
+	STACK_WIND (frame, error_gen_readlink_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->readlink,
-		    loc,
-		    size);
-	return 0;
+		    loc, size, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_mknod_cbk (call_frame_t *frame,
-		     void *cookie,
-		     xlator_t *this,
-		     int32_t op_ret,
-		     int32_t op_errno,
-		     inode_t *inode,
-                     struct stat *buf,
-                     struct stat *preparent,
-                     struct stat *postparent)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      inode,
-		      buf,
-                      preparent, postparent);
-	return 0;
+
+int
+error_gen_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		     int32_t op_ret, int32_t op_errno, inode_t *inode,
+                     struct iatt *buf, struct iatt *preparent,
+                     struct iatt *postparent, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (mknod, frame, op_ret, op_errno,
+                             inode, buf,
+                             preparent, postparent, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_mknod (call_frame_t *frame,
-		 xlator_t *this,
-		 loc_t *loc,
-		 mode_t mode,
-		 dev_t rdev)
+
+int
+error_gen_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
+		 mode_t mode, dev_t rdev, mode_t umask, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_MKNOD];
+        enable = egp->enable[GF_FOP_MKNOD];
 
         if (enable)
-                op_errno = error_gen (this, ERR_MKNOD);
+                op_errno = error_gen (this, GF_FOP_MKNOD);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (mknod, frame, -1, op_errno, NULL, NULL,
+                                     NULL, NULL, xdata);
+        return 0;
 	}
-	STACK_WIND (frame,
-		    error_gen_mknod_cbk,
+
+	STACK_WIND (frame, error_gen_mknod_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->mknod,
-		    loc, mode, rdev);
-	return 0;
+		    loc, mode, rdev, umask, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_mkdir_cbk (call_frame_t *frame,
-		     void *cookie,
-		     xlator_t *this,
-		     int32_t op_ret,
-		     int32_t op_errno,
-		     inode_t *inode,
-                     struct stat *buf,
-                     struct stat *preparent,
-                     struct stat *postparent)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      inode,
-		      buf,
-                      preparent, postparent);
-	return 0;
+
+int
+error_gen_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		     int32_t op_ret, int32_t op_errno, inode_t *inode,
+                     struct iatt *buf, struct iatt *preparent,
+                     struct iatt *postparent, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno,
+                             inode, buf,
+                             preparent, postparent, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_mkdir (call_frame_t *frame,
-		 xlator_t *this,
-		 loc_t *loc,
-		 mode_t mode)
+int
+error_gen_mkdir (call_frame_t *frame, xlator_t *this,
+		 loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_MKDIR];
+        enable = egp->enable[GF_FOP_MKDIR];
 
         if (enable)
-                op_errno = error_gen (this, ERR_MKDIR);
+                op_errno = error_gen (this, GF_FOP_MKDIR);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (mkdir, frame, -1, op_errno, NULL, NULL,
+                                     NULL, NULL, xdata);
+                return 0;
 	}
-	STACK_WIND (frame,
-		    error_gen_mkdir_cbk,
+
+	STACK_WIND (frame, error_gen_mkdir_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->mkdir,
-		    loc, mode);
-	return 0;
+		    loc, mode, umask, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_unlink_cbk (call_frame_t *frame,
-		      void *cookie,
-		      xlator_t *this,
-		      int32_t op_ret,
-		      int32_t op_errno,
-                      struct stat *preparent,
-                      struct stat *postparent)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, preparent, postparent);
-	return 0;
+
+int
+error_gen_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		      int32_t op_ret, int32_t op_errno,
+                      struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno,
+                             preparent, postparent, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_unlink (call_frame_t *frame,
-		  xlator_t *this,
-		  loc_t *loc)
+
+int
+error_gen_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+                  dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_UNLINK];
+        enable = egp->enable[GF_FOP_UNLINK];
 
         if (enable)
-                op_errno = error_gen (this, ERR_UNLINK);
+                op_errno = error_gen (this, GF_FOP_UNLINK);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (unlink, frame, -1, op_errno, NULL, NULL,
+                                     xdata);
+                return 0;
 	}
-	STACK_WIND (frame,
-		    error_gen_unlink_cbk,
+
+	STACK_WIND (frame, error_gen_unlink_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->unlink,
-		    loc);
-	return 0;
+		    loc, xflag, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_rmdir_cbk (call_frame_t *frame,
-		     void *cookie,
-		     xlator_t *this,
-		     int32_t op_ret,
-		     int32_t op_errno,
-                     struct stat *preparent,
-                     struct stat *postparent)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-                      preparent, postparent);
-	return 0;
+
+int
+error_gen_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		     int32_t op_ret, int32_t op_errno,
+                     struct iatt *preparent, struct iatt *postparent,
+                     dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno,
+                             preparent, postparent, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_rmdir (call_frame_t *frame,
-		 xlator_t *this,
-		 loc_t *loc)
+
+int
+error_gen_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+                 dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_RMDIR];
+        enable = egp->enable[GF_FOP_RMDIR];
 
         if (enable)
-                op_errno = error_gen (this, ERR_RMDIR);
+                op_errno = error_gen (this, GF_FOP_RMDIR);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, NULL,
-                              NULL, NULL); /* pre & post parent attr */
-		return 0;
+		STACK_UNWIND_STRICT (rmdir, frame, -1, op_errno, NULL, NULL, xdata);
+        return 0;
 	}
-	STACK_WIND (frame,
-		    error_gen_rmdir_cbk,
+
+	STACK_WIND (frame, error_gen_rmdir_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->rmdir,
-		    loc);
-	return 0;
+		    loc, flags, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_symlink_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno,
-		       inode_t *inode,
-                       struct stat *buf,
-                       struct stat *preparent,
-                       struct stat *postparent)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, inode, buf,
-                      preparent, postparent);
-	return 0;
+
+int
+error_gen_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		       int32_t op_ret, int32_t op_errno, inode_t *inode,
+                       struct iatt *buf, struct iatt *preparent,
+                       struct iatt *postparent, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_symlink (call_frame_t *frame,
-		   xlator_t *this,
-		   const char *linkpath,
-		   loc_t *loc)
+
+int
+error_gen_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
+		   loc_t *loc, mode_t umask, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_SYMLINK];
+        enable = egp->enable[GF_FOP_SYMLINK];
 
         if (enable)
-                op_errno = error_gen (this, ERR_SYMLINK);
+                op_errno = error_gen (this, GF_FOP_SYMLINK);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, NULL,
-                              NULL, NULL); /* pre & post parent attr */
+		STACK_UNWIND_STRICT (symlink, frame, -1, op_errno, NULL, NULL,
+                                     NULL, NULL, NULL); /* pre & post parent attr */
 		return 0;
 	}
-	STACK_WIND (frame,
-		    error_gen_symlink_cbk,
+
+	STACK_WIND (frame, error_gen_symlink_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->symlink,
-		    linkpath, loc);
-	return 0;
+		    linkpath, loc, umask, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_rename_cbk (call_frame_t *frame,
-		      void *cookie,
-		      xlator_t *this,
-		      int32_t op_ret,
-		      int32_t op_errno,
-		      struct stat *buf,
-                      struct stat *preoldparent,
-                      struct stat *postoldparent,
-                      struct stat *prenewparent,
-                      struct stat *postnewparent)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, buf,
-                      preoldparent, postoldparent,
-                      prenewparent, postnewparent);
-	return 0;
+
+int
+error_gen_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		      int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                      struct iatt *preoldparent, struct iatt *postoldparent,
+                      struct iatt *prenewparent, struct iatt *postnewparent,
+                      dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (rename, frame, op_ret, op_errno, buf,
+                             preoldparent, postoldparent,
+                             prenewparent, postnewparent, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_rename (call_frame_t *frame,
-		  xlator_t *this,
-		  loc_t *oldloc,
-		  loc_t *newloc)
+
+int
+error_gen_rename (call_frame_t *frame, xlator_t *this,
+		  loc_t *oldloc, loc_t *newloc, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_RENAME];
+        enable = egp->enable[GF_FOP_RENAME];
 
         if (enable)
-                op_errno = error_gen (this, ERR_RENAME);
+                op_errno = error_gen (this, GF_FOP_RENAME);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, NULL,
-                              NULL, NULL); /* pre & post parent attr */
+		STACK_UNWIND_STRICT (rename, frame, -1, op_errno, NULL,
+                                     NULL, NULL, NULL, NULL, NULL);
 		return 0;
 	}
-	STACK_WIND (frame,
-		    error_gen_rename_cbk,
+
+	STACK_WIND (frame, error_gen_rename_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->rename,
-		    oldloc, newloc);
-	return 0;
+		    oldloc, newloc, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_link_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    inode_t *inode,
-                    struct stat *buf,
-                    struct stat *preparent,
-                    struct stat *postparent)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, inode, buf,
-                      preparent, postparent);
-	return 0;
+
+int
+error_gen_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		    int32_t op_ret, int32_t op_errno, inode_t *inode,
+                    struct iatt *buf, struct iatt *preparent,
+                    struct iatt *postparent, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_link (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *oldloc,
-		loc_t *newloc)
+
+int
+error_gen_link (call_frame_t *frame, xlator_t *this,
+		loc_t *oldloc, loc_t *newloc, dict_t *xdata)
 {
 	int             op_errno = 0;
         eg_t            *egp = NULL;
         int             enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_LINK];
+        enable = egp->enable[GF_FOP_LINK];
 
         if (enable)
-                op_errno = error_gen (this, ERR_LINK);
+                op_errno = error_gen (this, GF_FOP_LINK);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, NULL,
-                              NULL, NULL); /* pre & post parent attr */
+		STACK_UNWIND_STRICT (link, frame, -1, op_errno, NULL, NULL,
+                                     NULL, NULL, NULL);
 		return 0;
 	}
-	STACK_WIND (frame,
-		    error_gen_link_cbk,
+
+	STACK_WIND (frame, error_gen_link_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->link,
-		    oldloc, newloc);
-	return 0;
+		    oldloc, newloc, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_create_cbk (call_frame_t *frame,
-		      void *cookie,
-		      xlator_t *this,
-		      int32_t op_ret,
-		      int32_t op_errno,
-		      fd_t *fd,
-		      inode_t *inode,
-		      struct stat *buf,
-                      struct stat *preparent,
-                      struct stat *postparent)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, fd, inode, buf,
-                      preparent, postparent);
-	return 0;
+
+int
+error_gen_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		      int32_t op_ret, int32_t op_errno,
+		      fd_t *fd, inode_t *inode, struct iatt *buf,
+                      struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
+                             preparent, postparent, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_create (call_frame_t *frame,
-		  xlator_t *this,
-		  loc_t *loc,
-		  int32_t flags,
-		  mode_t mode, fd_t *fd)
+
+int
+error_gen_create (call_frame_t *frame, xlator_t *this, loc_t *loc,
+		  int32_t flags, mode_t mode, mode_t umask, fd_t *fd,
+                  dict_t *xdata)
 {
 	int             op_errno = 0;
         eg_t            *egp = NULL;
         int             enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_CREATE];
+        enable = egp->enable[GF_FOP_CREATE];
 
         if (enable)
-                op_errno = error_gen (this, ERR_CREATE);
+                op_errno = error_gen (this, GF_FOP_CREATE);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, NULL, NULL,
-                              NULL, NULL); /* pre & post attr */
+		STACK_UNWIND_STRICT (create, frame, -1, op_errno, NULL, NULL,
+                                     NULL, NULL, NULL, NULL);
 		return 0;
 	}
 
 	STACK_WIND (frame, error_gen_create_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->create,
-		    loc, flags, mode, fd);
-	return 0;
+		    loc, flags, mode, umask, fd, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_open_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;
+
+int
+error_gen_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		    int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_open (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		int32_t flags, fd_t *fd, int32_t wbflags)
+
+int
+error_gen_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
+		int32_t flags, fd_t *fd, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_OPEN];
+        enable = egp->enable[GF_FOP_OPEN];
 
         if (enable)
-                op_errno = error_gen (this, ERR_OPEN);
+                op_errno = error_gen (this, GF_FOP_OPEN);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (open, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_open_cbk,
+	STACK_WIND (frame, error_gen_open_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->open,
-		    loc, flags, fd, wbflags);
-	return 0;
+		    loc, flags, fd, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_readv_cbk (call_frame_t *frame,
-		     void *cookie,
-		     xlator_t *this,
-		     int32_t op_ret,
-		     int32_t op_errno,
-		     struct iovec *vector,
-		     int32_t count,
-		     struct stat *stbuf,
-                     struct iobref *iobref)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      vector,
-		      count,
-		      stbuf,
-                      iobref);
-	return 0;
+
+int
+error_gen_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		     int32_t op_ret, int32_t op_errno,
+		     struct iovec *vector, int32_t count,
+		     struct iatt *stbuf, struct iobref *iobref, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno,
+                             vector, count, stbuf, iobref, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_readv (call_frame_t *frame,
-		 xlator_t *this,
-		 fd_t *fd,
-		 size_t size,
-		 off_t offset)
+
+int
+error_gen_readv (call_frame_t *frame, xlator_t *this,
+		 fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_READV];
+        enable = egp->enable[GF_FOP_READ];
 
         if (enable)
-                op_errno = error_gen (this, ERR_READV);
+                op_errno = error_gen (this, GF_FOP_READ);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, 0, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (readv, frame, -1, op_errno, NULL, 0,
+                                     NULL, NULL, xdata);
+        return 0;
 	}
 
 
-	STACK_WIND (frame,
-		    error_gen_readv_cbk,
+	STACK_WIND (frame, error_gen_readv_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->readv,
-		    fd,
-		    size,
-		    offset);
-	return 0;
+		    fd, size, offset, flags, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_writev_cbk (call_frame_t *frame,
-		      void *cookie,
-		      xlator_t *this,
-		      int32_t op_ret,
-		      int32_t op_errno,
-                      struct stat *prebuf,
-		      struct stat *postbuf)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-                      prebuf,
-		      postbuf);
-	return 0;
+
+int
+error_gen_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		      int32_t op_ret, int32_t op_errno,
+                      struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_writev (call_frame_t *frame,
-		  xlator_t *this,
-		  fd_t *fd,
-		  struct iovec *vector,
-		  int32_t count,
-		  off_t off,
-                  struct iobref *iobref)
+
+int
+error_gen_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
+		  struct iovec *vector, int32_t count,
+		  off_t off, uint32_t flags, struct iobref *iobref, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_WRITEV];
+        enable = egp->enable[GF_FOP_WRITE];
 
         if (enable)
-                op_errno = error_gen (this, ERR_WRITEV);
-
-	if (op_errno) {
-		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL);
+                op_errno = error_gen (this, GF_FOP_WRITE);
+
+	if (op_errno == GF_ERROR_SHORT_WRITE) {
+		struct iovec *shortvec;
+
+		/*
+		 * A short write error returns some value less than what was
+		 * requested from a write. To simulate this, replace the vector
+		 * with one half the size;
+		 */
+		shortvec = iov_dup(vector, 1);
+		shortvec->iov_len /= 2;
+
+		STACK_WIND(frame, error_gen_writev_cbk, FIRST_CHILD(this),
+			   FIRST_CHILD(this)->fops->writev, fd, shortvec, count,
+			   off, flags, iobref, xdata);
+		GF_FREE(shortvec);
 		return 0;
+	} else if (op_errno) {
+		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
+		STACK_UNWIND_STRICT (writev, frame, -1, op_errno, NULL, NULL, xdata);
+        	return 0;
 	}
 
-
-	STACK_WIND (frame,
-		    error_gen_writev_cbk,
+	STACK_WIND (frame, error_gen_writev_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->writev,
-		    fd,
-		    vector,
-		    count,
-		    off,
-                    iobref);
-	return 0;
+		    fd, vector, count, off, flags, iobref, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_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;
+
+int
+error_gen_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		     int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_flush (call_frame_t *frame,
-		 xlator_t *this,
-		 fd_t *fd)
+
+int
+error_gen_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_FLUSH];
+        enable = egp->enable[GF_FOP_FLUSH];
 
         if (enable)
-                op_errno = error_gen (this, ERR_FLUSH);
+                op_errno = error_gen (this, GF_FOP_FLUSH);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (flush, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_flush_cbk,
-		    FIRST_CHILD(this),
+	STACK_WIND (frame, error_gen_flush_cbk,
+                    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->flush,
-		    fd);
-	return 0;
+		    fd, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_fsync_cbk (call_frame_t *frame,
-		     void *cookie,
-		     xlator_t *this,
-		     int32_t op_ret,
-		     int32_t op_errno,
-                     struct stat *prebuf,
-                     struct stat *postbuf)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno);
-	return 0;
+
+int
+error_gen_fsync_cbk (call_frame_t *frame, void *cookie,
+		     xlator_t *this, int32_t op_ret,
+		     int32_t op_errno, struct iatt *prebuf,
+                     struct iatt *postbuf, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_fsync (call_frame_t *frame,
-		 xlator_t *this,
-		 fd_t *fd,
-		 int32_t flags)
+
+int
+error_gen_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_FSYNC];
+        enable = egp->enable[GF_FOP_FSYNC];
 
         if (enable)
-                op_errno = error_gen (this, ERR_FSYNC);
+                op_errno = error_gen (this, GF_FOP_FSYNC);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (fsync, frame, -1, op_errno, NULL, NULL, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_fsync_cbk,
+	STACK_WIND (frame, error_gen_fsync_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->fsync,
-		    fd,
-		    flags);
-	return 0;
+		    fd, flags, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_fstat_cbk (call_frame_t *frame,
-		     void *cookie,
-		     xlator_t *this,
-		     int32_t op_ret,
-		     int32_t op_errno,
-		     struct stat *buf)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      buf);
-	return 0;
+
+int
+error_gen_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, buf, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_fstat (call_frame_t *frame,
-		 xlator_t *this,
-		 fd_t *fd)
+
+int
+error_gen_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_FSTAT];
+        enable = egp->enable[GF_FOP_FSTAT];
 
         if (enable)
-                op_errno = error_gen (this, ERR_FSTAT);
+                op_errno = error_gen (this, GF_FOP_FSTAT);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (fstat, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_fstat_cbk,
+	STACK_WIND (frame, error_gen_fstat_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->fstat,
-		    fd);
-	return 0;
+		    fd, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_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;
+
+int
+error_gen_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		       int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_opendir (call_frame_t *frame,
-		   xlator_t *this,
-		   loc_t *loc, fd_t *fd)
+
+int
+error_gen_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_OPENDIR];
+        enable = egp->enable[GF_FOP_OPENDIR];
 
         if (enable)
-                op_errno = error_gen (this, ERR_OPENDIR);
+                op_errno = error_gen (this, GF_FOP_OPENDIR);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (opendir, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_opendir_cbk,
+	STACK_WIND (frame, error_gen_opendir_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->opendir,
-		    loc, fd);
-	return 0;
+		    loc, fd, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_getdents_cbk (call_frame_t *frame,
-			void *cookie,
-			xlator_t *this,
-			int32_t op_ret,
-			int32_t op_errno,
-			dir_entry_t *entries,
-			int32_t count)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      entries,
-		      count);
-	return 0;
+int
+error_gen_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+			int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (fsyncdir, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_getdents (call_frame_t *frame,
-		    xlator_t *this,
-		    fd_t *fd,
-		    size_t size,
-		    off_t offset,
-		    int32_t flag)
+
+int
+error_gen_fsyncdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                    int32_t flags, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_GETDENTS];
+        enable = egp->enable[GF_FOP_FSYNCDIR];
 
         if (enable)
-                op_errno = error_gen (this, ERR_GETDENTS);
+                op_errno = error_gen (this, GF_FOP_FSYNCDIR);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, 0);
-		return 0;
+		STACK_UNWIND_STRICT (fsyncdir, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_getdents_cbk,
+	STACK_WIND (frame, error_gen_fsyncdir_cbk,
 		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->getdents,
-		    fd,
-		    size,
-		    offset,
-		    flag);
-	return 0;
+		    FIRST_CHILD(this)->fops->fsyncdir,
+		    fd, flags, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_setdents_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;
+
+int
+error_gen_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		      int32_t op_ret, int32_t op_errno, struct statvfs *buf, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, buf, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_setdents (call_frame_t *frame,
-		    xlator_t *this,
-		    fd_t *fd,
-		    int32_t flags,
-		    dir_entry_t *entries,
-		    int32_t count)
+
+int
+error_gen_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_SETDENTS];
+        enable = egp->enable[GF_FOP_STATFS];
 
         if (enable)
-                op_errno = error_gen (this, ERR_SETDENTS);
+                op_errno = error_gen (this, GF_FOP_STATFS);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, 0);
-		return 0;
+		STACK_UNWIND_STRICT (statfs, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_setdents_cbk,
+	STACK_WIND (frame, error_gen_statfs_cbk,
 		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->setdents,
-		    fd,
-		    flags,
-		    entries,
-		    count);
-	return 0;
+		    FIRST_CHILD(this)->fops->statfs,
+		    loc, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_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;
+
+int
+error_gen_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+			int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_fsyncdir (call_frame_t *frame,
-		    xlator_t *this,
-		    fd_t *fd,
-		    int32_t flags)
+
+int
+error_gen_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+		    dict_t *dict, int32_t flags, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_FSYNCDIR];
+        enable = egp->enable[GF_FOP_SETXATTR];
 
         if (enable)
-                op_errno = error_gen (this, ERR_FSYNCDIR);
+                op_errno = error_gen (this, GF_FOP_SETXATTR);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (setxattr, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_fsyncdir_cbk,
+	STACK_WIND (frame, error_gen_setxattr_cbk,
 		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->fsyncdir,
-		    fd,
-		    flags);
-	return 0;
+		    FIRST_CHILD(this)->fops->setxattr,
+		    loc, dict, flags, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_statfs_cbk (call_frame_t *frame,
-		      void *cookie,
-		      xlator_t *this,
-		      int32_t op_ret,
-		      int32_t op_errno,
-		      struct statvfs *buf)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      buf);
-	return 0;
+
+int
+error_gen_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+			int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_statfs (call_frame_t *frame,
-		  xlator_t *this,
-		  loc_t *loc)
+
+int
+error_gen_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+		    const char *name, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_STATFS];
+        enable = egp->enable[GF_FOP_GETXATTR];
 
         if (enable)
-                op_errno = error_gen (this, ERR_STATFS);
+                op_errno = error_gen (this, GF_FOP_GETXATTR);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (getxattr, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_statfs_cbk,
+	STACK_WIND (frame, error_gen_getxattr_cbk,
 		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->statfs,
-		    loc);
-	return 0;
+		    FIRST_CHILD(this)->fops->getxattr,
+		    loc, name, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_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);
-	return 0;
+int
+error_gen_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_setxattr (call_frame_t *frame,
-		    xlator_t *this,
-		    loc_t *loc,
-		    dict_t *dict,
-		    int32_t flags)
+
+int
+error_gen_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                     dict_t *dict, int32_t flags, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_SETXATTR];
+        enable = egp->enable[GF_FOP_FSETXATTR];
 
         if (enable)
-                op_errno = error_gen (this, ERR_SETXATTR);
+                op_errno = error_gen (this, GF_FOP_FSETXATTR);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (fsetxattr, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_setxattr_cbk,
+	STACK_WIND (frame, error_gen_fsetxattr_cbk,
 		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->setxattr,
-		    loc,
-		    dict,
-		    flags);
-	return 0;
+		    FIRST_CHILD(this)->fops->fsetxattr,
+		    fd, dict, flags, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_getxattr_cbk (call_frame_t *frame,
-			void *cookie,
-			xlator_t *this,
-			int32_t op_ret,
-			int32_t op_errno,
-			dict_t *dict)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      dict);
-	return 0;
+
+int
+error_gen_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_getxattr (call_frame_t *frame,
-		    xlator_t *this,
-		    loc_t *loc,
-		    const char *name)
+
+int
+error_gen_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                     const char *name, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_GETXATTR];
+        enable = egp->enable[GF_FOP_FGETXATTR];
 
         if (enable)
-                op_errno = error_gen (this, ERR_GETXATTR);
+                op_errno = error_gen (this, GF_FOP_FGETXATTR);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (fgetxattr, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_getxattr_cbk,
+	STACK_WIND (frame, error_gen_fgetxattr_cbk,
 		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->getxattr,
-		    loc,
-		    name);
-	return 0;
+		    FIRST_CHILD(this)->fops->fgetxattr,
+		    fd, name, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_xattrop_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno,
-		       dict_t *dict)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, dict);
-	return 0;
+
+int
+error_gen_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (xattrop, frame, op_ret, op_errno, dict, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_xattrop (call_frame_t *frame,
-		   xlator_t *this,
-		   loc_t *loc,
-		   gf_xattrop_flags_t flags,
-		   dict_t *dict)
+
+int
+error_gen_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
+		   gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_XATTROP];
+        enable = egp->enable[GF_FOP_XATTROP];
 
         if (enable)
-                op_errno = error_gen (this, ERR_XATTROP);
+                op_errno = error_gen (this, GF_FOP_XATTROP);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (xattrop, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_xattrop_cbk,
+	STACK_WIND (frame, error_gen_xattrop_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->xattrop,
-		    loc, flags, dict);
-	return 0;
+		    loc, flags, dict, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_fxattrop_cbk (call_frame_t *frame,
-			void *cookie,
-			xlator_t *this,
-			int32_t op_ret,
-			int32_t op_errno,
-			dict_t *dict)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, dict);
-	return 0;
+
+int
+error_gen_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+			int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (fxattrop, frame, op_ret, op_errno, dict, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_fxattrop (call_frame_t *frame,
-		    xlator_t *this,
-		    fd_t *fd,
-		    gf_xattrop_flags_t flags,
-		    dict_t *dict)
+
+int
+error_gen_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
+		    gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_FXATTROP];
+        enable = egp->enable[GF_FOP_FXATTROP];
 
         if (enable)
-                op_errno = error_gen (this, ERR_FXATTROP);
+                op_errno = error_gen (this, GF_FOP_FXATTROP);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (fxattrop, frame, -1, op_errno, NULL, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_fxattrop_cbk,
+	STACK_WIND (frame, error_gen_fxattrop_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->fxattrop,
-		    fd, flags, dict);
-	return 0;
+		    fd, flags, dict, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_removexattr_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;
+
+int
+error_gen_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+			   int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_removexattr (call_frame_t *frame,
-		       xlator_t *this,
-		       loc_t *loc,
-		       const char *name)
+
+int
+error_gen_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+		       const char *name, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_REMOVEXATTR];
+        enable = egp->enable[GF_FOP_REMOVEXATTR];
 
         if (enable)
-                op_errno = error_gen (this, ERR_REMOVEXATTR);
+                op_errno = error_gen (this, GF_FOP_REMOVEXATTR);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (removexattr, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_removexattr_cbk,
+	STACK_WIND (frame, error_gen_removexattr_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->removexattr,
-		    loc,
-		    name);
-	return 0;
+		    loc, name, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_lk_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-		  struct flock *lock)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      lock);
-	return 0;
+int
+error_gen_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+			   int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (fremovexattr, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_lk (call_frame_t *frame,
-	      xlator_t *this,
-	      fd_t *fd,
-	      int32_t cmd,
-	      struct flock *lock)
+
+int
+error_gen_fremovexattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                        const char *name, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_LK];
+        enable = egp->enable[GF_FOP_FREMOVEXATTR];
 
         if (enable)
-                op_errno = error_gen (this, ERR_LK);
+                op_errno = error_gen (this, GF_FOP_FREMOVEXATTR);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL);
-		return 0;
+		STACK_UNWIND_STRICT (fremovexattr, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_lk_cbk,
+	STACK_WIND (frame, error_gen_fremovexattr_cbk,
+		    FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->fremovexattr,
+		    fd, name, xdata);
+        return 0;
+}
+
+
+int
+error_gen_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		  int32_t op_ret, int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, lock, xdata);
+        return 0;
+}
+
+
+int
+error_gen_lk (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t cmd,
+	      struct gf_flock *lock, dict_t *xdata)
+{
+	int              op_errno = 0;
+        eg_t            *egp = NULL;
+        int              enable = 1;
+
+        egp = this->private;
+        enable = egp->enable[GF_FOP_LK];
+
+        if (enable)
+                op_errno = error_gen (this, GF_FOP_LK);
+
+	if (op_errno) {
+		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
+		STACK_UNWIND_STRICT (lk, frame, -1, op_errno, NULL, xdata);
+        return 0;
+	}
+
+	STACK_WIND (frame, error_gen_lk_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->lk,
-		    fd,
-		    cmd,
-		    lock);
-	return 0;
+		    fd, cmd, lock, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_inodelk_cbk (call_frame_t *frame, void *cookie,
-		       xlator_t *this, int32_t op_ret, int32_t op_errno)
 
+int
+error_gen_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
+	STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
-int32_t
+
+int
 error_gen_inodelk (call_frame_t *frame, xlator_t *this,
 		   const char *volume, loc_t *loc, int32_t cmd,
-                   struct flock *lock)
+                   struct gf_flock *lock, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_INODELK];
+        enable = egp->enable[GF_FOP_INODELK];
 
         if (enable)
-                op_errno = error_gen (this, ERR_INODELK);
+                op_errno = error_gen (this, GF_FOP_INODELK);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (inodelk, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_inodelk_cbk,
+	STACK_WIND (frame, error_gen_inodelk_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->inodelk,
-		    volume, loc, cmd, lock);
-	return 0;
+		    volume, loc, cmd, lock, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_finodelk_cbk (call_frame_t *frame, void *cookie,
-			xlator_t *this, int32_t op_ret, int32_t op_errno)
 
+int
+error_gen_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
+	STACK_UNWIND_STRICT (finodelk, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
-int32_t
+
+int
 error_gen_finodelk (call_frame_t *frame, xlator_t *this,
 		    const char *volume, fd_t *fd, int32_t cmd,
-                    struct flock *lock)
+                    struct gf_flock *lock, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_FINODELK];
+        enable = egp->enable[GF_FOP_FINODELK];
 
         if (enable)
-                op_errno = error_gen (this, ERR_FINODELK);
+                op_errno = error_gen (this, GF_FOP_FINODELK);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (finodelk, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_finodelk_cbk,
+	STACK_WIND (frame, error_gen_finodelk_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->finodelk,
-		    volume, fd, cmd, lock);
-	return 0;
+		    volume, fd, cmd, lock, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_entrylk_cbk (call_frame_t *frame, void *cookie,
-		       xlator_t *this, int32_t op_ret, int32_t op_errno)
 
+int
+error_gen_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
+	STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
-int32_t
+
+int
 error_gen_entrylk (call_frame_t *frame, xlator_t *this,
 		   const char *volume, loc_t *loc, const char *basename,
-		   entrylk_cmd cmd, entrylk_type type)
+		   entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_ENTRYLK];
+        enable = egp->enable[GF_FOP_ENTRYLK];
 
         if (enable)
-                op_errno = error_gen (this, ERR_ENTRYLK);
+                op_errno = error_gen (this, GF_FOP_ENTRYLK);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (entrylk, frame, -1, op_errno, xdata);
+        return 0;
 	}
 
 	STACK_WIND (frame, error_gen_entrylk_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->entrylk,
-		    volume, loc, basename, cmd, type);
-	return 0;
+		    volume, loc, basename, cmd, type, xdata);
+        return 0;
 }
 
-int32_t
-error_gen_fentrylk_cbk (call_frame_t *frame, void *cookie,
-			xlator_t *this, int32_t op_ret, int32_t op_errno)
 
+int
+error_gen_fentrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
+	STACK_UNWIND_STRICT (fentrylk, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
-int32_t
+
+int
 error_gen_fentrylk (call_frame_t *frame, xlator_t *this,
 		    const char *volume, fd_t *fd, const char *basename,
-		    entrylk_cmd cmd, entrylk_type type)
+		    entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
 	int             op_errno = 0;
         eg_t            *egp = NULL;
         int             enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_FENTRYLK];
+        enable = egp->enable[GF_FOP_FENTRYLK];
 
         if (enable)
-                op_errno = error_gen (this, ERR_FENTRYLK);
+                op_errno = error_gen (this, GF_FOP_FENTRYLK);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno);
-		return 0;
+		STACK_UNWIND_STRICT (fentrylk, frame, -1, op_errno, xdata);
+                return 0;
 	}
 
 	STACK_WIND (frame, error_gen_fentrylk_cbk,
 		    FIRST_CHILD(this),
 		    FIRST_CHILD(this)->fops->fentrylk,
-		    volume, fd, basename, cmd, type);
-	return 0;
+		    volume, fd, basename, cmd, type, xdata);
+        return 0;
 }
 
 
 /* Management operations */
 
-int32_t
-error_gen_stats_cbk (call_frame_t *frame,
-		     void *cookie,
-		     xlator_t *this,
-		     int32_t op_ret,
-		     int32_t op_errno,
-		     struct xlator_stats *stats)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      stats);
-	return 0;
+
+int
+error_gen_getspec_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		       int32_t op_ret, int32_t op_errno, char *spec_data)
+{
+	STACK_UNWIND_STRICT (getspec, frame, op_ret, op_errno, spec_data);
+        return 0;
 }
 
-int32_t
-error_gen_stats (call_frame_t *frame,
-		 xlator_t *this,
-		 int32_t flags)
+
+int
+error_gen_getspec (call_frame_t *frame, xlator_t *this, const char *key,
+		   int32_t flags)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_STATS];
+        enable = egp->enable[GF_FOP_GETSPEC];
 
         if (enable)
-                op_errno = error_gen (this, ERR_STATS);
+                op_errno = error_gen (this, GF_FOP_GETSPEC);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL);
+		STACK_UNWIND_STRICT (getspec, frame, -1, op_errno, NULL);
 		return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_stats_cbk,
+	STACK_WIND (frame, error_gen_getspec_cbk,
 		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->mops->stats,
-		    flags);
+		    FIRST_CHILD(this)->fops->getspec,
+		    key, flags);
 	return 0;
 }
 
-int32_t
-error_gen_getspec_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno,
-		       char *spec_data)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      spec_data);
+
+int
+error_gen_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		       int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+                       dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, entries, xdata);
 	return 0;
 }
 
-int32_t
-error_gen_getspec (call_frame_t *frame,
-		   xlator_t *this,
-		   const char *key,
-		   int32_t flags)
+
+int
+error_gen_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
+		   size_t size, off_t off, dict_t *xdata)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_GETSPEC];
+        enable = egp->enable[GF_FOP_READDIR];
 
         if (enable)
-                op_errno = error_gen (this, ERR_GETSPEC);
+                op_errno = error_gen (this, GF_FOP_READDIR);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL);
+		STACK_UNWIND_STRICT (readdir, frame, -1, op_errno, NULL, xdata);
 		return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_getspec_cbk,
+	STACK_WIND (frame, error_gen_readdir_cbk,
 		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->mops->getspec,
-		    key, flags);
+		    FIRST_CHILD(this)->fops->readdir,
+		    fd, size, off, xdata);
 	return 0;
 }
 
-int32_t
-error_gen_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)
-{
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      file_checksum,
-		      dir_checksum);
+
+int
+error_gen_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+                        dict_t *xdata)
+{
+	STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries, xdata);
 	return 0;
 }
 
-int32_t
-error_gen_checksum (call_frame_t *frame,
-		    xlator_t *this,
-		    loc_t *loc,
-		    int32_t flag)
+
+int
+error_gen_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+                    off_t off, dict_t *dict)
 {
-	int             op_errno = 0;
+	int              op_errno = 0;
         eg_t            *egp = NULL;
-        int             enable = 1;
+        int              enable = 1;
 
         egp = this->private;
-        enable = egp->enable[ERR_CHECKSUM];
+        enable = egp->enable[GF_FOP_READDIRP];
 
         if (enable)
-                op_errno = error_gen (this, ERR_CHECKSUM);
+                op_errno = error_gen (this, GF_FOP_READDIRP);
 
 	if (op_errno) {
 		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL, NULL);
+		STACK_UNWIND_STRICT (readdirp, frame, -1, op_errno, NULL, NULL);
 		return 0;
 	}
 
-	STACK_WIND (frame,
-		    error_gen_checksum_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->checksum,
-		    loc,
-		    flag);
+	STACK_WIND (frame, error_gen_readdirp_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readdirp,
+                    fd, size, off, dict);
 	return 0;
 }
 
-int32_t
-error_gen_readdir_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno,
-		       gf_dirent_t *entries)
-{
-	STACK_UNWIND (frame, op_ret, op_errno, entries);
-	return 0;
+static void
+error_gen_set_failure (eg_t *pvt, int percent)
+{
+        GF_ASSERT (pvt);
+
+        if (percent)
+                pvt->failure_iter_no = 100/percent;
+        else
+                pvt->failure_iter_no = 100/GF_FAILURE_DEFAULT;
 }
 
-int32_t
-error_gen_readdir (call_frame_t *frame,
-		   xlator_t *this,
-		   fd_t *fd,
-		   size_t size,
-		   off_t off)
+static void
+error_gen_parse_fill_fops (eg_t *pvt, char *enable_fops)
 {
-	int             op_errno = 0;
-        eg_t            *egp = NULL;
-        int             enable = 1;
+        char            *op_no_str = NULL;
+        int              op_no = -1;
+        int              i = 0;
+        xlator_t        *this = THIS;
+        char            *saveptr = NULL;
 
-        egp = this->private;
-        enable = egp->enable[ERR_READDIR];
+        GF_ASSERT (pvt);
+        GF_ASSERT (this);
 
-        if (enable)
-                op_errno = error_gen (this, ERR_READDIR);
-
-	if (op_errno) {
-		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL);
-		return 0;
-	}
+        for (i = 0; i < GF_FOP_MAXVALUE; i++)
+                pvt->enable[i] = 0;
 
-	STACK_WIND (frame,
-		    error_gen_readdir_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->readdir,
-		    fd, size, off);
-	return 0;
+        if (!enable_fops) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "All fops are enabled.");
+                for (i = 0; i < GF_FOP_MAXVALUE; i++)
+                        pvt->enable[i] = 1;
+        } else {
+                op_no_str = strtok_r (enable_fops, ",", &saveptr);
+                while (op_no_str) {
+                        op_no = get_fop_int (&op_no_str);
+                        if (op_no == -1) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "Wrong option value %s", op_no_str);
+                        } else
+                                pvt->enable[op_no] = 1;
+
+                        op_no_str = strtok_r (NULL, ",", &saveptr);
+                }
+        }
 }
 
 int32_t
-error_gen_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int32_t op_ret, int32_t op_errno, gf_dirent_t *entries)
+error_gen_priv_dump (xlator_t *this)
 {
-	STACK_UNWIND (frame, op_ret, op_errno, entries);
-	return 0;
-}
+        char            key_prefix[GF_DUMP_MAX_BUF_LEN];
+        int             ret = -1;
+        eg_t            *conf = NULL;
 
+        if (!this)
+            goto out;
 
-int32_t
-error_gen_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-                    off_t off)
-{
-	int             op_errno = 0;
-        eg_t            *egp = NULL;
-        int             enable = 1;
+        conf = this->private;
+        if (!conf)
+            goto out;
 
-        egp = this->private;
-        enable = egp->enable[ERR_READDIRP];
+        ret = TRY_LOCK(&conf->lock);
+        if (ret != 0) {
+                return ret;
+        }
 
-        if (enable)
-                op_errno = error_gen (this, ERR_READDIRP);
+        gf_proc_dump_add_section("xlator.debug.error-gen.%s.priv", this->name);
+        gf_proc_dump_build_key(key_prefix,"xlator.debug.error-gen","%s.priv",
+                               this->name);
 
-	if (op_errno) {
-		GF_ERROR(this, "unwind(-1, %s)", strerror (op_errno));
-		STACK_UNWIND (frame, -1, op_errno, NULL);
-		return 0;
-	}
+        gf_proc_dump_write("op_count", "%d", conf->op_count);
+        gf_proc_dump_write("failure_iter_no", "%d", conf->failure_iter_no);
+        gf_proc_dump_write("error_no", "%s", conf->error_no);
+        gf_proc_dump_write("random_failure", "%d", conf->random_failure);
 
-	STACK_WIND (frame, error_gen_readdirp_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->readdirp, fd, size, off);
-	return 0;
+        UNLOCK(&conf->lock);
+out:
+        return ret;
 }
 
 int32_t
-error_gen_closedir (xlator_t *this,
-		    fd_t *fd)
+mem_acct_init (xlator_t *this)
 {
-	return 0;
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_error_gen_mt_end + 1);
+
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
+                        " failed");
+                return ret;
+        }
+
+        return ret;
 }
 
-int32_t
-error_gen_close (xlator_t *this,
-		 fd_t *fd)
+int
+reconfigure (xlator_t *this, dict_t *options)
 {
-	return 0;
+        eg_t            *pvt = NULL;
+        int32_t          ret = 0;
+        char            *error_enable_fops = NULL;
+        int32_t          failure_percent_int = 0;
+
+        if (!this || !this->private)
+                goto out;
+
+        pvt = this->private;
+
+        GF_OPTION_RECONF ("error-no", pvt->error_no, options, str, out);
+
+        GF_OPTION_RECONF ("failure", failure_percent_int, options, int32,
+                          out);
+
+        GF_OPTION_RECONF ("enable", error_enable_fops, options, str, out);
+
+        GF_OPTION_RECONF ("random-failure", pvt->random_failure, options,
+                          bool, out);
+
+        error_gen_parse_fill_fops (pvt, error_enable_fops);
+        error_gen_set_failure (pvt, failure_percent_int);
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "reconfigure returning %d", ret);
+        return ret;
 }
 
 int
 init (xlator_t *this)
 {
         eg_t            *pvt = NULL;
-        data_t          *error_no = NULL;
-        data_t          *failure_percent = NULL;
-        data_t          *enable = NULL;
-        int32_t         ret = 0;
+        int32_t          ret = 0;
         char            *error_enable_fops = NULL;
-        char            *op_no_str = NULL;
-        int             op_no = -1;
-        int             i = 0;
-        int32_t         failure_percent_int = 0;
+        int32_t          failure_percent_int = 0;
 
         if (!this->children || this->children->next) {
                 gf_log (this->name, GF_LOG_ERROR,
@@ -2320,74 +2104,38 @@ init (xlator_t *this)
                         "dangling volume. check volfile ");
         }
 
-        error_no = dict_get (this->options, "error-no");
-        failure_percent = dict_get (this->options, "failure");
-        enable = dict_get (this->options, "enable");
-
-        pvt = CALLOC (1, sizeof (eg_t));
+        pvt = GF_CALLOC (1, sizeof (eg_t), gf_error_gen_mt_eg_t);
 
         if (!pvt) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "out of memory.");
                 ret = -1;
                 goto out;
         }
 
         LOCK_INIT (&pvt->lock);
 
-        for (i = 0; i < NO_OF_FOPS; i++)
-                pvt->enable[i] = 0;
-        if (!error_no) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "error-no not specified.");
-        } else {
-                pvt->error_no = data_to_str (error_no);
-        }
+        GF_OPTION_INIT ("error-no", pvt->error_no, str, out);
 
-        if (!failure_percent) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "failure percent not specified.");
-                pvt->failure_iter_no = 100/GF_FAILURE_DEFAULT;
-        } else {
-                failure_percent_int = data_to_int32 (failure_percent);
-                if (failure_percent_int)
-                        pvt->failure_iter_no = 100/failure_percent_int;
-                else
-                        pvt->failure_iter_no = 100/GF_FAILURE_DEFAULT;
-        }
+        GF_OPTION_INIT ("failure", failure_percent_int, int32, out);
+
+        GF_OPTION_INIT ("enable", error_enable_fops, str, out);
+
+        GF_OPTION_INIT ("random-failure", pvt->random_failure, bool, out);
+
+
+        error_gen_parse_fill_fops (pvt, error_enable_fops);
+        error_gen_set_failure (pvt, failure_percent_int);
 
-        if (!enable) {
-                gf_log (this->name, GF_LOG_WARNING,
-                        "All fops are enabled.");
-                for (i = 0; i < NO_OF_FOPS; i++)
-                        pvt->enable[i] = 1;
-        } else {
-                error_enable_fops = data_to_str (enable);
-                op_no_str = error_enable_fops;
-                while ((*error_enable_fops) != '\0') {
-                        error_enable_fops++;
-                        if (((*error_enable_fops) == ',') ||
-                            ((*error_enable_fops) == '\0')) {
-                                if ((*error_enable_fops) != '\0') {
-                                        (*error_enable_fops) = '\0';
-                                        error_enable_fops++;
-                                }
-                                op_no = get_fop_int (&op_no_str);
-                                if (op_no == -1) {
-                                        gf_log (this->name, GF_LOG_WARNING,
-                                                "Wrong option value %s",
-                                                op_no_str);
-                                } else
-                                        pvt->enable[op_no] = 1;
-                                op_no_str = error_enable_fops;
-                        }
-                }
-        }
         this->private = pvt;
+
+        /* Give some seed value here */
+        srand (time(NULL));
 out:
+        if (ret)
+                GF_FREE (pvt);
         return ret;
 }
 
+
 void
 fini (xlator_t *this)
 {
@@ -2399,12 +2147,18 @@ fini (xlator_t *this)
 
         if (pvt) {
                 LOCK_DESTROY (&pvt->lock);
-                FREE (pvt);
+                GF_FREE (pvt);
                 gf_log (this->name, GF_LOG_DEBUG, "fini called");
         }
         return;
 }
 
+struct xlator_dumpops dumpops = {
+        .priv = error_gen_priv_dump,
+};
+
+struct xlator_fops cbks;
+
 struct xlator_fops fops = {
 	.lookup      = error_gen_lookup,
 	.stat        = error_gen_stat,
@@ -2427,18 +2181,18 @@ struct xlator_fops fops = {
 	.setxattr    = error_gen_setxattr,
 	.getxattr    = error_gen_getxattr,
 	.removexattr = error_gen_removexattr,
+	.fsetxattr    = error_gen_fsetxattr,
+	.fgetxattr    = error_gen_fgetxattr,
+	.fremovexattr = error_gen_fremovexattr,
 	.opendir     = error_gen_opendir,
 	.readdir     = error_gen_readdir,
 	.readdirp    = error_gen_readdirp,
-	.getdents    = error_gen_getdents,
 	.fsyncdir    = error_gen_fsyncdir,
 	.access      = error_gen_access,
 	.ftruncate   = error_gen_ftruncate,
 	.fstat       = error_gen_fstat,
 	.lk          = error_gen_lk,
-	.setdents    = error_gen_setdents,
 	.lookup_cbk  = error_gen_lookup_cbk,
-	.checksum    = error_gen_checksum,
 	.xattrop     = error_gen_xattrop,
 	.fxattrop    = error_gen_fxattrop,
 	.inodelk     = error_gen_inodelk,
@@ -2447,29 +2201,32 @@ struct xlator_fops fops = {
 	.fentrylk    = error_gen_fentrylk,
         .setattr     = error_gen_setattr,
         .fsetattr    = error_gen_fsetattr,
-};
-
-struct xlator_mops mops = {
-	.stats = error_gen_stats,
-	.getspec = error_gen_getspec,
-};
-
-struct xlator_cbks cbks = {
-	.release = error_gen_close,
-	.releasedir = error_gen_closedir,
+	.getspec     = error_gen_getspec,
 };
 
 struct volume_options options[] = {
         { .key  = {"failure"},
-          .type = GF_OPTION_TYPE_INT },
+          .type = GF_OPTION_TYPE_INT,
+          .description = "Percentage failure of operations when enabled.",
+        },
+
         { .key  = {"error-no"},
           .value = {"ENOENT","ENOTDIR","ENAMETOOLONG","EACCES","EBADF",
                     "EFAULT","ENOMEM","EINVAL","EIO","EEXIST","ENOSPC",
                     "EPERM","EROFS","EBUSY","EISDIR","ENOTEMPTY","EMLINK"
                     "ENODEV","EXDEV","EMFILE","ENFILE","ENOSYS","EINTR",
-                    "EFBIG","EAGAIN"},
-          .type = GF_OPTION_TYPE_STR },
+                    "EFBIG","EAGAIN","GF_ERROR_SHORT_WRITE"},
+          .type = GF_OPTION_TYPE_STR,
+        },
+
+        { .key = {"random-failure"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "off",
+        },
+
         { .key  = {"enable"},
-          .type = GF_OPTION_TYPE_STR },
+          .type = GF_OPTION_TYPE_STR,
+        },
+
         { .key  = {NULL} }
 };
diff --git a/xlators/debug/error-gen/src/error-gen.h b/xlators/debug/error-gen/src/error-gen.h
index 7fb5fdfb5..d92c23062 100644
--- a/xlators/debug/error-gen/src/error-gen.h
+++ b/xlators/debug/error-gen/src/error-gen.h
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef _ERROR_GEN_H
 #define _ERROR_GEN_H
 
@@ -25,59 +15,28 @@
 #include "config.h"
 #endif
 
+#include "error-gen-mem-types.h"
+
 #define GF_FAILURE_DEFAULT 10
-#define NO_OF_FOPS 42
 
-enum {
-	ERR_LOOKUP,
-	ERR_STAT,
-        ERR_READLINK,
-	ERR_MKNOD,
-	ERR_MKDIR,
-	ERR_UNLINK,
-	ERR_RMDIR,
-	ERR_SYMLINK,
-	ERR_RENAME,
-	ERR_LINK,
-	ERR_TRUNCATE,
-	ERR_CREATE,
-	ERR_OPEN,
-	ERR_READV,
-	ERR_WRITEV,
-	ERR_STATFS,
-	ERR_FLUSH,
-	ERR_FSYNC,
-	ERR_SETXATTR,
-	ERR_GETXATTR,
-	ERR_REMOVEXATTR,
-	ERR_OPENDIR,
-	ERR_READDIR,
-	ERR_READDIRP,
-	ERR_GETDENTS,
-	ERR_FSYNCDIR,
-	ERR_ACCESS,
-	ERR_FTRUNCATE,
-	ERR_FSTAT,
-	ERR_LK,
-	ERR_SETDENTS,
-	ERR_CHECKSUM,
-	ERR_XATTROP,
-	ERR_FXATTROP,
-	ERR_INODELK,
-	ERR_FINODELK,
-	ERR_ENTRYLK,
-	ERR_FENTRYLK,
-        ERR_SETATTR,
-	ERR_FSETATTR,
-	ERR_STATS,
-        ERR_GETSPEC
+/*
+ * Pseudo-errors refer to errors beyond the scope of traditional <-1, op_errno>
+ * returns. This facilitates the ability to return unexpected, but not -1 values
+ * and/or to inject operations that lead to implicit error conditions. The range
+ * for pseudo errors resides at a high value to avoid conflicts with the errno
+ * range.
+ */
+enum GF_PSEUDO_ERRORS {
+	GF_ERROR_SHORT_WRITE = 1000,	/* short writev return value */
+	GF_ERROR_MAX
 };
 
 typedef struct {
-        int enable[NO_OF_FOPS];
+        int enable[GF_FOP_MAXVALUE];
         int op_count;
         int failure_iter_no;
         char *error_no;
+        gf_boolean_t random_failure;
         gf_lock_t lock;
 } eg_t;
 
diff --git a/xlators/debug/io-stats/src/Makefile.am b/xlators/debug/io-stats/src/Makefile.am
index 91aea6ebb..dff294cd8 100644
--- a/xlators/debug/io-stats/src/Makefile.am
+++ b/xlators/debug/io-stats/src/Makefile.am
@@ -2,12 +2,17 @@
 xlator_LTLIBRARIES = io-stats.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/debug
 
-io_stats_la_LDFLAGS = -module -avoidversion
+io_stats_la_LDFLAGS = -module -avoid-version
 
 io_stats_la_SOURCES = io-stats.c
 io_stats_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+noinst_HEADERS = io-stats-mem-types.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
+	-I$(top_srcdir)/rpc/xdr/src \
+	-I$(top_srcdir)/rpc/rpc-lib/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES =
diff --git a/xlators/debug/io-stats/src/io-stats-mem-types.h b/xlators/debug/io-stats/src/io-stats-mem-types.h
new file mode 100644
index 000000000..c30dfb17e
--- /dev/null
+++ b/xlators/debug/io-stats/src/io-stats-mem-types.h
@@ -0,0 +1,24 @@
+/*
+   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 __IO_STATS_MEM_TYPES_H__
+#define __IO_STATS_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_io_stats_mem_types_ {
+        gf_io_stats_mt_ios_conf = gf_common_mt_end + 1,
+        gf_io_stats_mt_ios_fd,
+        gf_io_stats_mt_ios_stat,
+        gf_io_stats_mt_ios_stat_list,
+        gf_io_stats_mt_end
+};
+#endif
+
diff --git a/xlators/debug/io-stats/src/io-stats.c b/xlators/debug/io-stats/src/io-stats.c
index eb950a33d..9e48a7c6e 100644
--- a/xlators/debug/io-stats/src/io-stats.c
+++ b/xlators/debug/io-stats/src/io-stats.c
@@ -1,1475 +1,2937 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
+#include "xlator.h"
 #endif
 
 /**
  * xlators/debug/io_stats :
- *    This translator maintains a counter for each fop (that is,
- *    a counter which stores the number of invocations for the certain
- *    fop), and makes these counters accessible as extended attributes.
+ *    This translator maintains statistics of all filesystem activity
+ *    happening through it. The kind of statistics include:
+ *
+ *  a) total read data - since process start, last interval and per fd
+ *  b) total write data - since process start, last interval and per fd
+ *  c) counts of read IO block size - since process start, last interval and per fd
+ *  d) counts of write IO block size - since process start, last interval and per fd
+ *  e) counts of all FOP types passing through it
+ *
+ *  Usage: setfattr -n io-stats-dump /tmp/filename /mnt/gluster
+ *
  */
 
+#include <fnmatch.h>
 #include <errno.h>
 #include "glusterfs.h"
 #include "xlator.h"
+#include "io-stats-mem-types.h"
+#include <stdarg.h>
+#include "defaults.h"
+#include "logging.h"
+#include "cli1-xdr.h"
+
+#define MAX_LIST_MEMBERS 100
+
+typedef enum {
+        IOS_STATS_TYPE_NONE,
+        IOS_STATS_TYPE_OPEN,
+        IOS_STATS_TYPE_READ,
+        IOS_STATS_TYPE_WRITE,
+        IOS_STATS_TYPE_OPENDIR,
+        IOS_STATS_TYPE_READDIRP,
+        IOS_STATS_TYPE_READ_THROUGHPUT,
+        IOS_STATS_TYPE_WRITE_THROUGHPUT,
+        IOS_STATS_TYPE_MAX
+}ios_stats_type_t;
 
-#define BUMP_HIT(op)                                            \
-do {                                                            \
-        LOCK (&((io_stats_private_t *)this->private)->lock);    \
-        {                                                       \
-                ((io_stats_private_t *)this->private)           \
-                  ->fop_records[GF_FOP_##op].hits++;            \
-        }                                                       \
-        UNLOCK (&((io_stats_private_t *)this->private)->lock);  \
- } while (0)
-
-struct io_stats_io_count {
-        size_t  size;
-        int64_t hits;        
+typedef enum {
+        IOS_STATS_THRU_READ,
+        IOS_STATS_THRU_WRITE,
+        IOS_STATS_THRU_MAX,
+}ios_stats_thru_t;
+
+struct ios_stat_lat {
+        struct timeval  time;
+        double          throughput;
 };
+
+struct ios_stat {
+        gf_lock_t       lock;
+        uuid_t          gfid;
+        char           *filename;
+        uint64_t        counters [IOS_STATS_TYPE_MAX];
+        struct ios_stat_lat thru_counters [IOS_STATS_THRU_MAX];
+        int             refcnt;
+};
+
+struct ios_stat_list {
+        struct list_head  list;
+        struct ios_stat  *iosstat;
+        double            value;
+};
+
+struct ios_stat_head {
+       gf_lock_t                lock;
+       double                   min_cnt;
+       uint64_t                 members;
+       struct ios_stat_list    *iosstats;
+};
+
+struct ios_lat {
+        double  min;
+        double  max;
+        double  avg;
+};
+
+struct ios_global_stats {
+        uint64_t        data_written;
+        uint64_t        data_read;
+        uint64_t        block_count_write[32];
+        uint64_t        block_count_read[32];
+        uint64_t        fop_hits[GF_FOP_MAXVALUE];
+        struct timeval  started_at;
+        struct ios_lat  latency[GF_FOP_MAXVALUE];
+        uint64_t        nr_opens;
+        uint64_t        max_nr_opens;
+        struct timeval  max_openfd_time;
+};
+
+
+struct ios_conf {
+        gf_lock_t                 lock;
+        struct ios_global_stats   cumulative;
+        uint64_t                  increment;
+        struct ios_global_stats   incremental;
+        gf_boolean_t              dump_fd_stats;
+        gf_boolean_t              count_fop_hits;
+        gf_boolean_t              measure_latency;
+        struct ios_stat_head      list[IOS_STATS_TYPE_MAX];
+        struct ios_stat_head      thru_list[IOS_STATS_THRU_MAX];
+};
+
+
+struct ios_fd {
+        char           *filename;
+        uint64_t        data_written;
+        uint64_t        data_read;
+        uint64_t        block_count_write[32];
+        uint64_t        block_count_read[32];
+        struct timeval  opened_at;
+};
+
 typedef enum {
-        GF_IO_STAT_BLK_SIZE_1K,
-        GF_IO_STAT_BLK_SIZE_2K,
-        GF_IO_STAT_BLK_SIZE_4K,
-        GF_IO_STAT_BLK_SIZE_8K,
-        GF_IO_STAT_BLK_SIZE_16K,
-        GF_IO_STAT_BLK_SIZE_32K,
-        GF_IO_STAT_BLK_SIZE_64K,
-        GF_IO_STAT_BLK_SIZE_128K,
-        GF_IO_STAT_BLK_SIZE_MAX,
-} gf_io_stat_blk_t;
-
-struct io_stats_private {
-        gf_lock_t lock;
-        struct {
-                char *name;
-                int enabled;
-                uint32_t hits;
-        } fop_records[GF_FOP_MAXVALUE];
-        struct io_stats_io_count read[GF_IO_STAT_BLK_SIZE_MAX + 1];
-        struct io_stats_io_count write[GF_IO_STAT_BLK_SIZE_MAX + 1];        
+        IOS_DUMP_TYPE_NONE = 0,
+        IOS_DUMP_TYPE_FILE = 1,
+        IOS_DUMP_TYPE_DICT = 2,
+        IOS_DUMP_TYPE_MAX  = 3
+} ios_dump_type_t;
+
+struct ios_dump_args {
+        ios_dump_type_t type;
+        union {
+                FILE *logfp;
+                dict_t *dict;
+        } u;
 };
-typedef struct io_stats_private io_stats_private_t;
 
-int32_t
-io_stats_create_cbk (call_frame_t *frame,
-                     void *cookie,
-                     xlator_t *this,
-                     int32_t op_ret,
-                     int32_t op_errno,
-                     fd_t *fd,
-                     inode_t *inode,
-                     struct stat *buf,
-                     struct stat *preparent,
-                     struct stat *postparent)
+typedef int (*block_dump_func) (xlator_t *, struct ios_dump_args*,
+                                    int , int , uint64_t ) ;
+
+struct ios_local {
+        struct timeval  wind_at;
+        struct timeval  unwind_at;
+};
+
+struct volume_options options[];
+
+inline static int
+is_fop_latency_started (call_frame_t *frame)
+{
+        GF_ASSERT (frame);
+        struct timeval epoch = {0,};
+        return memcmp (&frame->begin, &epoch, sizeof (epoch));
+}
+
+#define END_FOP_LATENCY(frame, op)                                      \
+        do {                                                            \
+                struct ios_conf  *conf = NULL;                          \
+                                                                        \
+                conf = this->private;                                   \
+                if (conf && conf->measure_latency) {                    \
+                        gettimeofday (&frame->end, NULL);               \
+                        update_ios_latency (conf, frame, GF_FOP_##op);  \
+                }                                                       \
+        } while (0)
+
+#define START_FOP_LATENCY(frame)                                         \
+        do {                                                             \
+                struct ios_conf  *conf = NULL;                           \
+                                                                         \
+                conf = this->private;                                    \
+                if (conf && conf->measure_latency) {                     \
+                        gettimeofday (&frame->begin, NULL);              \
+                } else {                                                 \
+                        memset (&frame->begin, 0, sizeof (frame->begin));\
+                }                                                        \
+        } while (0)
+
+
+#define BUMP_FOP(op)                                                    \
+        do {                                                            \
+                struct ios_conf  *conf = NULL;                          \
+                                                                        \
+                conf = this->private;                                   \
+                if (!conf)                                              \
+                        break;                                          \
+                conf->cumulative.fop_hits[GF_FOP_##op]++;               \
+                conf->incremental.fop_hits[GF_FOP_##op]++;              \
+        } while (0)
+
+#define UPDATE_PROFILE_STATS(frame, op)                                       \
+        do {                                                                  \
+                struct ios_conf  *conf = NULL;                                \
+                                                                              \
+                if (!is_fop_latency_started (frame))                          \
+                        break;                                                \
+                conf = this->private;                                         \
+                LOCK (&conf->lock);                                           \
+                {                                                             \
+                        if (conf && conf->measure_latency &&                  \
+                            conf->count_fop_hits) {                           \
+                                BUMP_FOP(op);                                 \
+                                gettimeofday (&frame->end, NULL);             \
+                                update_ios_latency (conf, frame, GF_FOP_##op);\
+                        }                                                     \
+                }                                                             \
+                UNLOCK (&conf->lock);                                         \
+        } while (0)
+
+#define BUMP_READ(fd, len)                                              \
+        do {                                                            \
+                struct ios_conf  *conf = NULL;                          \
+                struct ios_fd    *iosfd = NULL;                         \
+                int               lb2 = 0;                              \
+                                                                        \
+                conf = this->private;                                   \
+                lb2 = log_base2 (len);                                  \
+                ios_fd_ctx_get (fd, this, &iosfd);                      \
+                if (!conf)                                              \
+                        break;                                          \
+                                                                        \
+                LOCK (&conf->lock);                                     \
+                {                                                       \
+                        conf->cumulative.data_read += len;              \
+                        conf->incremental.data_read += len;             \
+                        conf->cumulative.block_count_read[lb2]++;       \
+                        conf->incremental.block_count_read[lb2]++;      \
+                                                                        \
+                        if (iosfd) {                                    \
+                                iosfd->data_read += len;                \
+                                iosfd->block_count_read[lb2]++;         \
+                        }                                               \
+                }                                                       \
+                UNLOCK (&conf->lock);                                   \
+        } while (0)
+
+
+#define BUMP_WRITE(fd, len)                                             \
+        do {                                                            \
+                struct ios_conf  *conf = NULL;                          \
+                struct ios_fd    *iosfd = NULL;                         \
+                int               lb2 = 0;                              \
+                                                                        \
+                conf = this->private;                                   \
+                lb2 = log_base2 (len);                                  \
+                ios_fd_ctx_get (fd, this, &iosfd);                      \
+                if (!conf)                                              \
+                        break;                                          \
+                                                                        \
+                LOCK (&conf->lock);                                     \
+                {                                                       \
+                        conf->cumulative.data_written += len;           \
+                        conf->incremental.data_written += len;          \
+                        conf->cumulative.block_count_write[lb2]++;      \
+                        conf->incremental.block_count_write[lb2]++;     \
+                                                                        \
+                        if (iosfd) {                                    \
+                                iosfd->data_written += len;             \
+                                iosfd->block_count_write[lb2]++;        \
+                        }                                               \
+                }                                                       \
+                UNLOCK (&conf->lock);                                   \
+        } while (0)
+
+
+#define BUMP_STATS(iosstat, type)                                               \
+        do {                                                                    \
+                struct ios_conf         *conf = NULL;                           \
+                uint64_t                 value = 0;                             \
+                                                                                \
+                conf = this->private;                                           \
+                                                                                \
+                LOCK(&iosstat->lock);                                           \
+                {                                                               \
+                        iosstat->counters[type]++;                              \
+                        value = iosstat->counters[type];                        \
+                }                                                               \
+                UNLOCK (&iosstat->lock);                                        \
+                ios_stat_add_to_list (&conf->list[type],                        \
+                                     value, iosstat);                           \
+                                                                                \
+        } while (0)
+
+
+#define BUMP_THROUGHPUT(iosstat, type)						\
+        do {									\
+                struct ios_conf         *conf = NULL;				\
+                double                   elapsed;				\
+                struct timeval          *begin, *end;				\
+                double                   throughput;				\
+               int                      flag = 0;                              \
+                                                                                \
+                begin = &frame->begin;						\
+                end   = &frame->end;						\
+                                                                                \
+                elapsed = (end->tv_sec - begin->tv_sec) * 1e6			\
+                        + (end->tv_usec - begin->tv_usec);			\
+                throughput = op_ret / elapsed;					\
+                                                                                \
+                conf = this->private;						\
+                LOCK(&iosstat->lock);						\
+                {								\
+                        if (iosstat->thru_counters[type].throughput             \
+                                <= throughput) {                                \
+                                iosstat->thru_counters[type].throughput =       \
+                                                                throughput;     \
+                                gettimeofday (&iosstat->                        \
+                                             thru_counters[type].time, NULL);   \
+                               flag = 1;                                       \
+                        }							\
+                }								\
+                UNLOCK (&iosstat->lock);					\
+               if (flag)                                                       \
+                       ios_stat_add_to_list (&conf->thru_list[type],           \
+                                               throughput, iosstat);           \
+        } while (0)
+
+int
+ios_fd_ctx_get (fd_t *fd, xlator_t *this, struct ios_fd **iosfd)
+{
+        uint64_t      iosfd64 = 0;
+        unsigned long iosfdlong = 0;
+        int           ret = 0;
+
+        ret = fd_ctx_get (fd, this, &iosfd64);
+        iosfdlong = iosfd64;
+        if (ret != -1)
+                *iosfd = (void *) iosfdlong;
+
+        return ret;
+}
+
+
+
+int
+ios_fd_ctx_set (fd_t *fd, xlator_t *this, struct ios_fd *iosfd)
+{
+        uint64_t   iosfd64 = 0;
+        int        ret = 0;
+
+        iosfd64 = (unsigned long) iosfd;
+        ret = fd_ctx_set (fd, this, iosfd64);
+
+        return ret;
+}
+
+int
+ios_stat_ref (struct ios_stat *iosstat)
+{
+        LOCK (&iosstat->lock);
+        {
+                iosstat->refcnt++;
+        }
+        UNLOCK (&iosstat->lock);
+
+        return iosstat->refcnt;
+}
+
+int
+ios_stat_unref (struct ios_stat *iosstat)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, fd, inode, buf,
-                      preparent, postparent);
+        int cleanup = 0;
+        LOCK (&iosstat->lock);
+        {
+                iosstat->refcnt--;
+                if (iosstat->refcnt == 0) {
+                        if (iosstat->filename) {
+                                GF_FREE (iosstat->filename);
+                                iosstat->filename = NULL;
+                        }
+                        cleanup = 1;
+                }
+        }
+        UNLOCK (&iosstat->lock);
+
+        if (cleanup) {
+                GF_FREE (iosstat);
+                iosstat = NULL;
+        }
+
         return 0;
 }
 
-int32_t
-io_stats_open_cbk (call_frame_t *frame,
-                   void *cookie,
-                   xlator_t *this,
-                   int32_t op_ret,
-                   int32_t op_errno,
-                   fd_t *fd)
+int
+ios_inode_ctx_set (inode_t *inode, xlator_t *this, struct ios_stat *iosstat)
+{
+        uint64_t   iosstat64 = 0;
+        int        ret     = 0;
+
+        ios_stat_ref (iosstat);
+        iosstat64 = (unsigned long )iosstat;
+        ret = inode_ctx_put (inode, this, iosstat64);
+        return ret;
+}
+
+int
+ios_inode_ctx_get (inode_t *inode, xlator_t *this, struct ios_stat **iosstat)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, fd);
+        uint64_t      iosstat64 = 0;
+        unsigned long iosstatlong = 0;
+        int           ret = 0;
+
+        ret = inode_ctx_get (inode, this, &iosstat64);
+        iosstatlong = iosstat64;
+        if (ret != -1)
+                *iosstat = (void *) iosstatlong;
+
+        return ret;
+
+}
+
+int
+ios_stat_add_to_list (struct ios_stat_head *list_head, uint64_t value,
+                            struct ios_stat *iosstat)
+{
+        struct ios_stat_list *new = NULL;
+        struct ios_stat_list *entry = NULL;
+        struct ios_stat_list *t = NULL;
+        struct ios_stat_list *list_entry = NULL;
+        struct ios_stat_list *tmp = NULL;
+        struct ios_stat_list *last = NULL;
+        struct ios_stat      *stat = NULL;
+        int                   cnt = 0;
+        int                   found = 0;
+        int                   reposition = 0;
+        double                min_count = 0;
+
+        LOCK (&list_head->lock);
+        {
+
+                if (list_head->min_cnt == 0)
+                        list_head->min_cnt = value;
+                if ((list_head->members == MAX_LIST_MEMBERS) &&
+                     (list_head->min_cnt > value))
+                        goto out;
+
+                list_for_each_entry_safe (entry, t,
+                              &list_head->iosstats->list, list) {
+                        cnt++;
+                        if (cnt == list_head->members)
+                                last = entry;
+
+                        if (!uuid_compare (iosstat->gfid,
+                            entry->iosstat->gfid)) {
+                                list_entry = entry;
+                                found = cnt;
+                                entry->value = value;
+                                if (!reposition) {
+                                        if (cnt == list_head->members)
+                                                list_head->min_cnt = value;
+                                        goto out;
+                                }
+                                break;
+                        } else if (entry->value <= value && !reposition) {
+                                reposition = cnt;
+                                tmp = entry;
+                                if (cnt == list_head->members - 1)
+                                        min_count = entry->value;
+                        }
+                }
+                if (found) {
+                        list_del (&list_entry->list);
+                        list_add_tail (&list_entry->list, &tmp->list);
+                        if (min_count)
+                                list_head->min_cnt = min_count;
+                        goto out;
+                } else if (list_head->members == MAX_LIST_MEMBERS && reposition) {
+                        new = GF_CALLOC (1, sizeof (*new),
+                                        gf_io_stats_mt_ios_stat_list);
+                        new->iosstat = iosstat;
+                        new->value = value;
+                        ios_stat_ref (iosstat);
+                        list_add_tail (&new->list, &tmp->list);
+                        stat = last->iosstat;
+                        last->iosstat = NULL;
+                        ios_stat_unref (stat);
+                        list_del (&last->list);
+                        GF_FREE (last);
+                        if (reposition == MAX_LIST_MEMBERS)
+                                list_head->min_cnt = value;
+                        else if (min_count) {
+                                list_head->min_cnt = min_count;
+                        }
+                } else if (list_head->members < MAX_LIST_MEMBERS) {
+                        new  = GF_CALLOC (1, sizeof (*new),
+                                          gf_io_stats_mt_ios_stat_list);
+                        new->iosstat = iosstat;
+                        new->value = value;
+                        ios_stat_ref (iosstat);
+                        if (reposition) {
+                                list_add_tail (&new->list, &tmp->list);
+                        } else {
+                                list_add_tail (&new->list, &entry->list);
+                        }
+                        list_head->members++;
+                        if (list_head->min_cnt > value)
+                                list_head->min_cnt = value;
+                }
+        }
+out:
+        UNLOCK (&list_head->lock);
         return 0;
 }
 
-int32_t
-io_stats_stat_cbk (call_frame_t *frame,
-                   void *cookie,
-                   xlator_t *this,
-                   int32_t op_ret,
-                   int32_t op_errno,
-                   struct stat *buf)
+inline int
+ios_stats_cleanup (xlator_t *this, inode_t *inode)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, buf);
+
+        struct ios_stat *iosstat = NULL;
+        uint64_t         iosstat64 = 0;
+
+        inode_ctx_del (inode, this, &iosstat64);
+        if (!iosstat64) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "could not get inode ctx");
+                return 0;
+        }
+        iosstat = (void *) (long)iosstat64;
+        if (iosstat) {
+                ios_stat_unref (iosstat);
+        }
         return 0;
 }
 
-int32_t
-io_stats_readv_cbk (call_frame_t *frame,
-                    void *cookie,
-                    xlator_t *this,
-                    int32_t op_ret,
-                    int32_t op_errno,
-                    struct iovec *vector,
-                    int32_t count,
-                    struct stat *buf,
-                    struct iobref *iobref)
+#define ios_log(this, logfp, fmt ...)                           \
+        do {                                                    \
+                if (logfp) {                                    \
+                        fprintf (logfp, fmt);                   \
+                        fprintf (logfp, "\n");                  \
+                }                                               \
+                gf_log (this->name, GF_LOG_INFO, fmt);        \
+        } while (0)
+
+int
+ios_dump_file_stats (struct ios_stat_head *list_head, xlator_t *this, FILE* logfp)
 {
-        int i = 0;
-        io_stats_private_t *priv = NULL;
+        struct ios_stat_list *entry = NULL;
 
-        priv = this->private;
+        LOCK (&list_head->lock);
+        {
+                list_for_each_entry (entry, &list_head->iosstats->list, list) {
+                        ios_log (this, logfp, "%-12.0f %s",
+                                entry->value, entry->iosstat->filename);
+                }
+        }
+        UNLOCK (&list_head->lock);
+        return 0;
+}
 
-        if (op_ret > 0) {
-                for (i=0; i < GF_IO_STAT_BLK_SIZE_MAX; i++) {
-                        if (priv->read[i].size > iov_length (vector, count)) {
-                                break;
+int
+ios_dump_throughput_stats (struct ios_stat_head *list_head, xlator_t *this,
+                            FILE* logfp, ios_stats_type_t type)
+{
+        struct ios_stat_list *entry = NULL;
+        struct timeval        time  = {0, };
+        char                  timestr[256] = {0, };
+
+        LOCK (&list_head->lock);
+        {
+                list_for_each_entry (entry, &list_head->iosstats->list, list) {
+                        gf_time_fmt (timestr, sizeof timestr,
+                                     entry->iosstat->thru_counters[type].time.tv_sec,
+                                     gf_timefmt_FT);
+                        snprintf (timestr + strlen (timestr), sizeof timestr - strlen (timestr),
+                          ".%"GF_PRI_SUSECONDS, time.tv_usec);
+
+                        ios_log (this, logfp, "%s \t %-10.2f  \t  %s",
+                                 timestr, entry->value, entry->iosstat->filename);
+                }
+        }
+        UNLOCK (&list_head->lock);
+        return 0;
+}
+
+int
+io_stats_dump_global_to_logfp (xlator_t *this, struct ios_global_stats *stats,
+                               struct timeval *now, int interval, FILE* logfp)
+{
+        int                   i = 0;
+        int                   per_line = 0;
+        int                   index = 0;
+        struct ios_stat_head *list_head = NULL;
+        struct ios_conf      *conf = NULL;
+        char                  timestr[256] = {0, };
+        char                  str_header[128] = {0};
+        char                  str_read[128] = {0};
+        char                  str_write[128] = {0};
+
+        conf = this->private;
+
+        if (interval == -1)
+                ios_log (this, logfp, "\n=== Cumulative stats ===");
+        else
+                ios_log (this, logfp, "\n=== Interval %d stats ===",
+                         interval);
+        ios_log (this, logfp, "      Duration : %"PRId64" secs",
+                 (uint64_t) (now->tv_sec - stats->started_at.tv_sec));
+        ios_log (this, logfp, "     BytesRead : %"PRId64,
+                 stats->data_read);
+        ios_log (this, logfp, "  BytesWritten : %"PRId64"\n",
+                 stats->data_written);
+
+        snprintf (str_header, sizeof (str_header), "%-12s %c", "Block Size", ':');
+        snprintf (str_read, sizeof (str_read), "%-12s %c", "Read Count", ':');
+        snprintf (str_write, sizeof (str_write), "%-12s %c", "Write Count", ':');
+        index = 14;
+        for (i = 0; i < 32; i++) {
+                if ((stats->block_count_read[i] == 0) &&
+                    (stats->block_count_write[i] == 0))
+                        continue;
+                per_line++;
+
+                snprintf (str_header+index, sizeof (str_header)-index,
+                          "%16dB+", (1<<i));
+                if (stats->block_count_read[i])
+                        snprintf (str_read+index, sizeof (str_read)-index,
+                                  "%18"PRId64, stats->block_count_read[i]);
+                else    snprintf (str_read+index, sizeof (str_read)-index,
+                                  "%18s", "0");
+                if (stats->block_count_write[i])
+                        snprintf (str_write+index, sizeof (str_write)-index,
+                                  "%18"PRId64, stats->block_count_write[i]);
+                else    snprintf (str_write+index, sizeof (str_write)-index,
+                                  "%18s", "0");
+
+                index += 18;
+                if (per_line == 3) {
+                        ios_log (this, logfp, "%s", str_header);
+                        ios_log (this, logfp, "%s", str_read);
+                        ios_log (this, logfp, "%s\n", str_write);
+
+                        memset (str_header, 0, sizeof (str_header));
+                        memset (str_read, 0, sizeof (str_read));
+                        memset (str_write, 0, sizeof (str_write));
+
+                        snprintf (str_header, sizeof (str_header), "%-12s %c",
+                                  "Block Size", ':');
+                        snprintf (str_read, sizeof (str_read), "%-12s %c",
+                                  "Read Count", ':');
+                        snprintf (str_write, sizeof (str_write), "%-12s %c",
+                                  "Write Count", ':');
+
+                        index = 14;
+                        per_line = 0;
+                }
+        }
+
+        if (per_line != 0) {
+                ios_log (this, logfp, "%s", str_header);
+                ios_log (this, logfp, "%s", str_read);
+                ios_log (this, logfp, "%s\n", str_write);
+        }
+
+        ios_log (this, logfp, "%-13s %10s %14s %14s %14s", "Fop",
+                 "Call Count", "Avg-Latency", "Min-Latency",
+                 "Max-Latency");
+        ios_log (this, logfp, "%-13s %10s %14s %14s %14s", "---", "----------",
+                 "-----------", "-----------", "-----------");
+
+        for (i = 0; i < GF_FOP_MAXVALUE; i++) {
+                if (stats->fop_hits[i] && !stats->latency[i].avg)
+                        ios_log (this, logfp, "%-13s %10"PRId64" %11s "
+                                 "us %11s us %11s us", gf_fop_list[i],
+                                 stats->fop_hits[i], "0", "0", "0");
+                else if (stats->fop_hits[i] && stats->latency[i].avg)
+                        ios_log (this, logfp, "%-13s %10"PRId64" %11.2lf us "
+                                 "%11.2lf us %11.2lf us", gf_fop_list[i],
+                                 stats->fop_hits[i], stats->latency[i].avg,
+                                 stats->latency[i].min, stats->latency[i].max);
+        }
+        ios_log (this, logfp, "------ ----- ----- ----- ----- ----- ----- ----- "
+                 " ----- ----- ----- -----\n");
+
+        if (interval == -1) {
+                LOCK (&conf->lock);
+                {
+                        gf_time_fmt (timestr, sizeof timestr,
+                                     conf->cumulative.max_openfd_time.tv_sec,
+                                     gf_timefmt_FT);
+                        snprintf (timestr + strlen (timestr), sizeof timestr - strlen (timestr),
+                                  ".%"GF_PRI_SUSECONDS,
+                                  conf->cumulative.max_openfd_time.tv_usec);
+                        ios_log (this, logfp, "Current open fd's: %"PRId64
+                                 " Max open fd's: %"PRId64" time %s",
+                                 conf->cumulative.nr_opens,
+                                 conf->cumulative.max_nr_opens, timestr);
+                }
+                UNLOCK (&conf->lock);
+                ios_log (this, logfp, "\n==========Open File Stats========");
+                ios_log (this, logfp, "\nCOUNT:  \t  FILE NAME");
+                list_head = &conf->list[IOS_STATS_TYPE_OPEN];
+                ios_dump_file_stats (list_head, this, logfp);
+
+
+                ios_log (this, logfp, "\n==========Read File Stats========");
+                ios_log (this, logfp, "\nCOUNT:  \t  FILE NAME");
+                list_head = &conf->list[IOS_STATS_TYPE_READ];
+                ios_dump_file_stats (list_head, this, logfp);
+
+                ios_log (this, logfp, "\n==========Write File Stats========");
+                ios_log (this, logfp, "\nCOUNT:  \t  FILE NAME");
+                list_head = &conf->list[IOS_STATS_TYPE_WRITE];
+                ios_dump_file_stats (list_head, this, logfp);
+
+                ios_log (this, logfp, "\n==========Directory open stats========");
+                ios_log (this, logfp, "\nCOUNT:  \t  DIRECTORY NAME");
+                list_head = &conf->list[IOS_STATS_TYPE_OPENDIR];
+                ios_dump_file_stats (list_head, this, logfp);
+
+                ios_log (this, logfp, "\n========Directory readdirp Stats=======");
+                ios_log (this, logfp, "\nCOUNT:  \t  DIRECTORY NAME");
+                list_head = &conf->list[IOS_STATS_TYPE_READDIRP];
+                ios_dump_file_stats (list_head, this, logfp);
+
+                ios_log (this, logfp, "\n========Read Throughput File Stats=====");
+                ios_log (this, logfp, "\nTIMESTAMP \t\t\t THROUGHPUT(KBPS)"
+                         "\tFILE NAME");
+                list_head = &conf->thru_list[IOS_STATS_THRU_READ];
+                ios_dump_throughput_stats(list_head, this, logfp, IOS_STATS_THRU_READ);
+
+                ios_log (this, logfp, "\n======Write Throughput File Stats======");
+                ios_log (this, logfp, "\nTIMESTAMP \t\t\t THROUGHPUT(KBPS)"
+                         "\tFILE NAME");
+                list_head = &conf->thru_list[IOS_STATS_THRU_WRITE];
+                ios_dump_throughput_stats (list_head, this, logfp, IOS_STATS_THRU_WRITE);
+        }
+        return 0;
+}
+
+int
+io_stats_dump_global_to_dict (xlator_t *this, struct ios_global_stats *stats,
+                              struct timeval *now, int interval, dict_t *dict)
+{
+        int             ret = 0;
+        char            key[256] = {0};
+        uint64_t        sec = 0;
+        int             i = 0;
+        uint64_t        count = 0;
+
+        GF_ASSERT (stats);
+        GF_ASSERT (now);
+        GF_ASSERT (dict);
+        GF_ASSERT (this);
+
+        if (interval == -1)
+                snprintf (key, sizeof (key), "cumulative");
+        else
+                snprintf (key, sizeof (key), "interval");
+        ret = dict_set_int32 (dict, key, interval);
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR, "failed to set "
+                        "interval %d", interval);
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "%d-duration", interval);
+        sec = (uint64_t) (now->tv_sec - stats->started_at.tv_sec);
+        ret = dict_set_uint64 (dict, key, sec);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to set "
+                        "duration(%d) - %"PRId64, interval, sec);
+                goto out;
+        }
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "%d-total-read", interval);
+        ret = dict_set_uint64 (dict, key, stats->data_read);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to set total "
+                       "read(%d) - %"PRId64, interval, stats->data_read);
+                goto out;
+        }
+
+        memset (key, 0, sizeof (key));
+        snprintf (key, sizeof (key), "%d-total-write", interval);
+        ret = dict_set_uint64 (dict, key, stats->data_written);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to set total "
+                        "write(%d) - %"PRId64, interval, stats->data_written);
+                goto out;
+        }
+        for (i = 0; i < 32; i++) {
+                if (stats->block_count_read[i]) {
+                        memset (key, 0, sizeof (key));
+                        snprintf (key, sizeof (key), "%d-read-%d", interval,
+                                  (1 << i));
+                        count = stats->block_count_read[i];
+                        ret = dict_set_uint64 (dict, key, count);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                        "set read-%db+, with: %"PRId64,
+                                        (1<<i), count);
+                                goto out;
+                        }
+                }
+        }
+
+        for (i = 0; i < 32; i++) {
+                if (stats->block_count_write[i]) {
+                        snprintf (key, sizeof (key), "%d-write-%d", interval,
+                                  (1<<i));
+                        count = stats->block_count_write[i];
+                        ret = dict_set_uint64 (dict, key, count);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                        "set write-%db+, with: %"PRId64,
+                                        (1<<i), count);
+                                goto out;
                         }
                 }
-                priv->read[i].hits++;
         }
 
-        STACK_UNWIND (frame, op_ret, op_errno, vector, count, buf, iobref);
+        for (i = 0; i < GF_FOP_MAXVALUE; i++) {
+                if (stats->fop_hits[i] == 0)
+                        continue;
+                snprintf (key, sizeof (key), "%d-%d-hits", interval, i);
+                ret = dict_set_uint64 (dict, key, stats->fop_hits[i]);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to "
+                                "set %s-fop-hits: %"PRIu64, gf_fop_list[i],
+                                stats->fop_hits[i]);
+                        goto out;
+                }
+
+                if (stats->latency[i].avg == 0)
+                        continue;
+                snprintf (key, sizeof (key), "%d-%d-avglatency", interval, i);
+                ret = dict_set_double (dict, key, stats->latency[i].avg);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to set %s "
+                                "avglatency(%d) with %f", gf_fop_list[i],
+                                interval, stats->latency[i].avg);
+                        goto out;
+                }
+                snprintf (key, sizeof (key), "%d-%d-minlatency", interval, i);
+                ret = dict_set_double (dict, key, stats->latency[i].min);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to set %s "
+                                "minlatency(%d) with %f", gf_fop_list[i],
+                                interval, stats->latency[i].min);
+                        goto out;
+                }
+                snprintf (key, sizeof (key), "%d-%d-maxlatency", interval, i);
+                ret = dict_set_double (dict, key, stats->latency[i].max);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to set %s "
+                                "maxlatency(%d) with %f", gf_fop_list[i],
+                                interval, stats->latency[i].max);
+                        goto out;
+                }
+        }
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "returning %d", ret);
+        return ret;
+}
+
+int
+io_stats_dump_global (xlator_t *this, struct ios_global_stats *stats,
+                      struct timeval *now, int interval,
+                      struct ios_dump_args *args)
+{
+        int     ret = -1;
+
+        GF_ASSERT (args);
+        GF_ASSERT (now);
+        GF_ASSERT (stats);
+        GF_ASSERT (this);
+
+
+
+        switch (args->type) {
+        case IOS_DUMP_TYPE_FILE:
+                ret = io_stats_dump_global_to_logfp (this, stats, now,
+                                                     interval, args->u.logfp);
+        break;
+        case IOS_DUMP_TYPE_DICT:
+                ret = io_stats_dump_global_to_dict (this, stats, now,
+                                                    interval, args->u.dict);
+        break;
+        default:
+                GF_ASSERT (0);
+                ret = -1;
+        break;
+        }
+        return ret;
+}
+
+int
+ios_dump_args_init (struct ios_dump_args *args, ios_dump_type_t type,
+                    void *output)
+{
+        int             ret = 0;
+
+        GF_ASSERT (args);
+        GF_ASSERT (type > IOS_DUMP_TYPE_NONE && type < IOS_DUMP_TYPE_MAX);
+        GF_ASSERT (output);
+
+        args->type = type;
+        switch (args->type) {
+        case IOS_DUMP_TYPE_FILE:
+                args->u.logfp = output;
+                break;
+        case IOS_DUMP_TYPE_DICT:
+                args->u.dict = output;
+                break;
+        default:
+                GF_ASSERT (0);
+                ret = -1;
+        }
+
+        return ret;
+}
+
+static void
+ios_global_stats_clear (struct ios_global_stats *stats, struct timeval *now)
+{
+        GF_ASSERT (stats);
+        GF_ASSERT (now);
+
+        memset (stats, 0, sizeof (*stats));
+        stats->started_at = *now;
+}
+
+int
+io_stats_dump (xlator_t *this, struct ios_dump_args *args,
+               gf1_cli_info_op op, gf_boolean_t is_peek)
+{
+        struct ios_conf         *conf = NULL;
+        struct ios_global_stats  cumulative = {0, };
+        struct ios_global_stats  incremental = {0, };
+        int                      increment = 0;
+        struct timeval           now;
+
+        GF_ASSERT (this);
+        GF_ASSERT (args);
+        GF_ASSERT (args->type > IOS_DUMP_TYPE_NONE);
+        GF_ASSERT (args->type < IOS_DUMP_TYPE_MAX);
+
+        conf = this->private;
+
+        gettimeofday (&now, NULL);
+        LOCK (&conf->lock);
+        {
+                if (op == GF_CLI_INFO_ALL ||
+                    op == GF_CLI_INFO_CUMULATIVE)
+                        cumulative  = conf->cumulative;
+
+                if (op == GF_CLI_INFO_ALL ||
+                    op == GF_CLI_INFO_INCREMENTAL) {
+                        incremental = conf->incremental;
+                        increment = conf->increment;
+
+                        if (!is_peek) {
+                                increment = conf->increment++;
+
+                                ios_global_stats_clear (&conf->incremental,
+                                                        &now);
+                        }
+                }
+        }
+        UNLOCK (&conf->lock);
+
+        if (op == GF_CLI_INFO_ALL ||
+            op == GF_CLI_INFO_CUMULATIVE)
+                io_stats_dump_global (this, &cumulative, &now, -1, args);
+
+        if (op == GF_CLI_INFO_ALL ||
+            op == GF_CLI_INFO_INCREMENTAL)
+                io_stats_dump_global (this, &incremental, &now, increment, args);
+
         return 0;
 }
 
-int32_t
-io_stats_writev_cbk (call_frame_t *frame,
-                     void *cookie,
-                     xlator_t *this,
-                     int32_t op_ret,
-                     int32_t op_errno,
-                     struct stat *prebuf,
-                     struct stat *postbuf)
+
+int
+io_stats_dump_fd (xlator_t *this, struct ios_fd *iosfd)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, prebuf, postbuf);
+        struct ios_conf         *conf = NULL;
+        struct timeval           now;
+        uint64_t                 sec = 0;
+        uint64_t                 usec = 0;
+        int                      i = 0;
+
+        conf = this->private;
+
+        if (!conf->dump_fd_stats)
+                return 0;
+
+        if (!iosfd)
+                return 0;
+
+        gettimeofday (&now, NULL);
+
+        if (iosfd->opened_at.tv_usec > now.tv_usec) {
+                now.tv_usec += 1000000;
+                now.tv_usec--;
+        }
+
+        sec = now.tv_sec - iosfd->opened_at.tv_sec;
+        usec = now.tv_usec - iosfd->opened_at.tv_usec;
+
+        gf_log (this->name, GF_LOG_INFO,
+                "--- fd stats ---");
+
+        if (iosfd->filename)
+                gf_log (this->name, GF_LOG_INFO,
+                        "      Filename : %s",
+                        iosfd->filename);
+
+        if (sec)
+                gf_log (this->name, GF_LOG_INFO,
+                        "      Lifetime : %"PRId64"secs, %"PRId64"usecs",
+                        sec, usec);
+
+        if (iosfd->data_read)
+                gf_log (this->name, GF_LOG_INFO,
+                        "     BytesRead : %"PRId64" bytes",
+                        iosfd->data_read);
+
+        if (iosfd->data_written)
+                gf_log (this->name, GF_LOG_INFO,
+                        "  BytesWritten : %"PRId64" bytes",
+                        iosfd->data_written);
+
+        for (i = 0; i < 32; i++) {
+                if (iosfd->block_count_read[i])
+                        gf_log (this->name, GF_LOG_INFO,
+                                " Read %06db+ : %"PRId64,
+                                (1 << i), iosfd->block_count_read[i]);
+        }
+        for (i = 0; i < 32; i++) {
+                if (iosfd->block_count_write[i])
+                        gf_log (this->name, GF_LOG_INFO,
+                                "Write %06db+ : %"PRId64,
+                                (1 << i), iosfd->block_count_write[i]);
+        }
         return 0;
 }
 
-int32_t
-io_stats_getdents_cbk (call_frame_t *frame,
-                       void *cookie,
-                       xlator_t *this,
-                       int32_t op_ret,
-                       int32_t op_errno,
-                       dir_entry_t *entries,
-                       int32_t count)
+static void
+update_ios_latency_stats (struct ios_global_stats   *stats, double elapsed,
+                          glusterfs_fop_t op)
+{
+        double avg;
+
+        GF_ASSERT (stats);
+
+        if (!stats->latency[op].min)
+                stats->latency[op].min = elapsed;
+        if (stats->latency[op].min > elapsed)
+                stats->latency[op].min = elapsed;
+        if (stats->latency[op].max < elapsed)
+                stats->latency[op].max = elapsed;
+
+        avg = stats->latency[op].avg;
+
+        stats->latency[op].avg = avg + (elapsed - avg) / stats->fop_hits[op];
+}
+
+int
+update_ios_latency (struct ios_conf *conf, call_frame_t *frame,
+                    glusterfs_fop_t op)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, entries, count);
+        double elapsed;
+        struct timeval *begin, *end;
+
+        begin = &frame->begin;
+        end   = &frame->end;
+
+        elapsed = (end->tv_sec - begin->tv_sec) * 1e6
+                + (end->tv_usec - begin->tv_usec);
+
+        update_ios_latency_stats (&conf->cumulative, elapsed, op);
+        update_ios_latency_stats (&conf->incremental, elapsed, op);
+
         return 0;
 }
 
 int32_t
-io_stats_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                       int32_t op_ret, int32_t op_errno, gf_dirent_t *buf)
+io_stats_dump_stats_to_dict (xlator_t *this, dict_t *resp,
+                             ios_stats_type_t flags, int32_t list_cnt)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, buf);
+        struct ios_conf         *conf = NULL;
+        int                      cnt  = 0;
+        char                     key[256];
+        struct ios_stat_head    *list_head = NULL;
+        struct ios_stat_list    *entry = NULL;
+        int                      ret = -1;
+        ios_stats_thru_t         index = IOS_STATS_THRU_MAX;
+        char                     timestr[256] = {0, };
+        char                    *dict_timestr = NULL;
+
+        conf = this->private;
+
+        switch (flags) {
+                case IOS_STATS_TYPE_OPEN:
+                        list_head = &conf->list[IOS_STATS_TYPE_OPEN];
+                        LOCK (&conf->lock);
+                        {
+                                ret = dict_set_uint64 (resp, "current-open",
+                                                     conf->cumulative.nr_opens);
+                                if (ret)
+                                        goto unlock;
+                                ret = dict_set_uint64 (resp, "max-open",
+                                                       conf->cumulative.max_nr_opens);
+
+                                gf_time_fmt (timestr, sizeof timestr,
+                                             conf->cumulative.max_openfd_time.tv_sec,
+                                             gf_timefmt_FT);
+                                if (conf->cumulative.max_openfd_time.tv_sec)
+                                        snprintf (timestr + strlen (timestr), sizeof timestr - strlen (timestr),
+                                                  ".%"GF_PRI_SUSECONDS,
+                                                  conf->cumulative.max_openfd_time.tv_usec);
+
+                                dict_timestr = gf_strdup (timestr);
+                                if (!dict_timestr)
+                                        goto unlock;
+                                ret = dict_set_dynstr (resp, "max-openfd-time",
+                                                       dict_timestr);
+                                if (ret)
+                                        goto unlock;
+                        }
+        unlock:
+                        UNLOCK (&conf->lock);
+                        /* Do not proceed if we came here because of some error
+                         * during the dict operation */
+                        if (ret)
+                                goto out;
+                        break;
+                case IOS_STATS_TYPE_READ:
+                        list_head = &conf->list[IOS_STATS_TYPE_READ];
+                        break;
+                case IOS_STATS_TYPE_WRITE:
+                        list_head = &conf->list[IOS_STATS_TYPE_WRITE];
+                        break;
+                case IOS_STATS_TYPE_OPENDIR:
+                        list_head = &conf->list[IOS_STATS_TYPE_OPENDIR];
+                        break;
+                case IOS_STATS_TYPE_READDIRP:
+                        list_head = &conf->list[IOS_STATS_TYPE_READDIRP];
+                        break;
+                case IOS_STATS_TYPE_READ_THROUGHPUT:
+                        list_head = &conf->thru_list[IOS_STATS_THRU_READ];
+                        index = IOS_STATS_THRU_READ;
+                        break;
+                case IOS_STATS_TYPE_WRITE_THROUGHPUT:
+                        list_head = &conf->thru_list[IOS_STATS_THRU_WRITE];
+                        index = IOS_STATS_THRU_WRITE;
+                        break;
+
+                default:
+                       goto out;
+        }
+        ret = dict_set_int32 (resp, "top-op", flags);
+        if (!list_cnt)
+                goto out;
+        LOCK (&list_head->lock);
+        {
+                list_for_each_entry (entry, &list_head->iosstats->list, list) {
+
+                        cnt++;
+                        snprintf (key, 256, "%s-%d", "filename", cnt);
+                        ret = dict_set_str (resp, key, entry->iosstat->filename);
+                        if (ret)
+                                goto unlock_list_head;
+                         snprintf (key, 256, "%s-%d", "value",cnt);
+                         ret = dict_set_uint64 (resp, key, entry->value);
+                         if (ret)
+                                 goto unlock_list_head;
+                         if (index != IOS_STATS_THRU_MAX) {
+                                 snprintf (key, 256, "%s-%d", "time-sec", cnt);
+                                 ret = dict_set_int32 (resp, key,
+                                         entry->iosstat->thru_counters[index].time.tv_sec);
+                                 if (ret)
+                                         goto unlock_list_head;
+                                 snprintf (key, 256, "%s-%d", "time-usec", cnt);
+                                 ret = dict_set_int32 (resp, key,
+                                         entry->iosstat->thru_counters[index].time.tv_usec);
+                                 if (ret)
+                                         goto unlock_list_head;
+                         }
+                         if (cnt == list_cnt)
+                                 break;
+
+                }
+        }
+unlock_list_head:
+        UNLOCK (&list_head->lock);
+        /* ret is !=0 if some dict operation in the above critical region
+         * failed. */
+        if (ret)
+                goto out;
+        ret = dict_set_int32 (resp, "members", cnt);
+ out:
+        return ret;
+}
+
+int
+io_stats_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, fd_t *fd,
+                     inode_t *inode, struct iatt *buf,
+                     struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
+{
+        struct ios_fd *iosfd = NULL;
+        char          *path = NULL;
+        struct ios_stat *iosstat = NULL;
+        struct ios_conf   *conf = NULL;
+
+        conf = this->private;
+
+        path = frame->local;
+        frame->local = NULL;
+
+        if (!path)
+                goto unwind;
+
+        if (op_ret < 0) {
+                GF_FREE (path);
+                goto unwind;
+        }
+
+        iosfd = GF_CALLOC (1, sizeof (*iosfd), gf_io_stats_mt_ios_fd);
+        if (!iosfd) {
+                GF_FREE (path);
+                goto unwind;
+        }
+
+        iosfd->filename = path;
+        gettimeofday (&iosfd->opened_at, NULL);
+
+        ios_fd_ctx_set (fd, this, iosfd);
+        LOCK (&conf->lock);
+        {
+                conf->cumulative.nr_opens++;
+                if (conf->cumulative.nr_opens > conf->cumulative.max_nr_opens) {
+                        conf->cumulative.max_nr_opens = conf->cumulative.nr_opens;
+                        conf->cumulative.max_openfd_time = iosfd->opened_at;
+                }
+        }
+        UNLOCK (&conf->lock);
+
+        iosstat = GF_CALLOC (1, sizeof (*iosstat), gf_io_stats_mt_ios_stat);
+        if (!iosstat) {
+                GF_FREE (path);
+                goto unwind;
+        }
+        iosstat->filename = gf_strdup (path);
+        uuid_copy (iosstat->gfid, buf->ia_gfid);
+        LOCK_INIT (&iosstat->lock);
+        ios_inode_ctx_set (fd->inode, this, iosstat);
+
+unwind:
+        UPDATE_PROFILE_STATS (frame, CREATE);
+        STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
+                             preparent, postparent, xdata);
         return 0;
 }
 
 
-int32_t
-io_stats_readdir_cbk (call_frame_t *frame,
-                      void *cookie,
-                      xlator_t *this,
-                      int32_t op_ret,
-                      int32_t op_errno,
-                      gf_dirent_t *buf)
+int
+io_stats_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, buf);
+        struct ios_fd *iosfd = NULL;
+        char          *path = NULL;
+        struct   ios_stat *iosstat = NULL;
+        struct ios_conf   *conf = NULL;
+
+        conf = this->private;
+        path = frame->local;
+        frame->local = NULL;
+
+        if (!path)
+                goto unwind;
+
+        if (op_ret < 0) {
+                GF_FREE (path);
+                goto unwind;
+        }
+
+        iosfd = GF_CALLOC (1, sizeof (*iosfd), gf_io_stats_mt_ios_fd);
+        if (!iosfd) {
+                GF_FREE (path);
+                goto unwind;
+        }
+
+        iosfd->filename = path;
+        gettimeofday (&iosfd->opened_at, NULL);
+
+        ios_fd_ctx_set (fd, this, iosfd);
+
+        ios_inode_ctx_get (fd->inode, this, &iosstat);
+        if (!iosstat) {
+                iosstat = GF_CALLOC (1, sizeof (*iosstat),
+                                     gf_io_stats_mt_ios_stat);
+                if (iosstat) {
+                        iosstat->filename = gf_strdup (path);
+                        uuid_copy (iosstat->gfid, fd->inode->gfid);
+                        LOCK_INIT (&iosstat->lock);
+                        ios_inode_ctx_set (fd->inode, this, iosstat);
+                }
+        }
+
+        LOCK (&conf->lock);
+        {
+                conf->cumulative.nr_opens++;
+                if (conf->cumulative.nr_opens > conf->cumulative.max_nr_opens) {
+                        conf->cumulative.max_nr_opens = conf->cumulative.nr_opens;
+                        conf->cumulative.max_openfd_time = iosfd->opened_at;
+                }
+        }
+        UNLOCK (&conf->lock);
+        if (iosstat) {
+              BUMP_STATS (iosstat, IOS_STATS_TYPE_OPEN);
+              iosstat = NULL;
+        }
+unwind:
+        UPDATE_PROFILE_STATS (frame, OPEN);
+
+        STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
         return 0;
+
 }
 
-int32_t
-io_stats_fsync_cbk (call_frame_t *frame,
-                    void *cookie,
-                    xlator_t *this,
-                    int32_t op_ret,
-                    int32_t op_errno,
-                    struct stat *prebuf,
-                    struct stat *postbuf)
+
+int
+io_stats_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, prebuf, postbuf);
+        UPDATE_PROFILE_STATS (frame, STAT);
+        STACK_UNWIND_STRICT (stat, frame, op_ret, op_errno, buf, xdata);
         return 0;
 }
 
-int32_t
-io_stats_setattr_cbk (call_frame_t *frame,
-                      void *cookie,
-                      xlator_t *this,
-                      int32_t op_ret,
-                      int32_t op_errno,
-                      struct stat *preop,
-                      struct stat *postop)
+
+int
+io_stats_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno,
+                    struct iovec *vector, int32_t count,
+                    struct iatt *buf, struct iobref *iobref, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, preop, postop);
+        int              len = 0;
+        fd_t            *fd = NULL;
+        struct ios_stat *iosstat = NULL;
+
+        fd = frame->local;
+        frame->local = NULL;
+
+        if (op_ret > 0) {
+                len = iov_length (vector, count);
+                BUMP_READ (fd, len);
+        }
+
+        UPDATE_PROFILE_STATS (frame, READ);
+        ios_inode_ctx_get (fd->inode, this, &iosstat);
+
+        if (iosstat) {
+              BUMP_STATS (iosstat, IOS_STATS_TYPE_READ);
+              BUMP_THROUGHPUT (iosstat, IOS_STATS_THRU_READ);
+              iosstat = NULL;
+        }
+
+        STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno,
+                             vector, count, buf, iobref, xdata);
         return 0;
+
 }
 
-int32_t
-io_stats_unlink_cbk (call_frame_t *frame,
-                     void *cookie,
-                     xlator_t *this,
-                     int32_t op_ret,
-                     int32_t op_errno,
-                     struct stat *preparent,
-                     struct stat *postparent)
+
+int
+io_stats_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno,
+                     struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, preparent, postparent);
+        struct ios_stat *iosstat = NULL;
+        inode_t         *inode   = NULL;
+
+        UPDATE_PROFILE_STATS (frame, WRITE);
+        if (frame->local){
+                inode = frame->local;
+                frame->local = NULL;
+                ios_inode_ctx_get (inode, this, &iosstat);
+                if (iosstat) {
+                        BUMP_STATS (iosstat, IOS_STATS_TYPE_WRITE);
+                        BUMP_THROUGHPUT (iosstat, IOS_STATS_THRU_WRITE);
+                        inode = NULL;
+                        iosstat = NULL;
+                }
+        }
+
+        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf, xdata);
         return 0;
+
 }
 
-int32_t
-io_stats_rename_cbk (call_frame_t *frame,
-                     void *cookie,
-                     xlator_t *this,
-                     int32_t op_ret,
-                     int32_t op_errno,
-                     struct stat *buf,
-                     struct stat *preoldparent,
-                     struct stat *postoldparent,
-                     struct stat *prenewparent,
-                     struct stat *postnewparent)
+
+
+
+int
+io_stats_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, gf_dirent_t *buf, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, buf,
-                      preoldparent, postoldparent,
-                      prenewparent, postnewparent);
+        struct ios_stat *iosstat = NULL;
+        inode_t         *inode   = frame->local;
+
+        frame->local = NULL;
+
+        UPDATE_PROFILE_STATS (frame, READDIRP);
+
+        ios_inode_ctx_get (inode, this, &iosstat);
+
+        if (iosstat) {
+              BUMP_STATS (iosstat, IOS_STATS_TYPE_READDIRP);
+              iosstat = NULL;
+        }
+
+        STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, buf, xdata);
         return 0;
 }
 
-int32_t
-io_stats_readlink_cbk (call_frame_t *frame,
-                       void *cookie,
-                       xlator_t *this,
-                       int32_t op_ret,
-                       int32_t op_errno,
-                       const char *buf,
-                       struct stat *sbuf)
+
+int
+io_stats_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, gf_dirent_t *buf, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, buf, sbuf);
+        UPDATE_PROFILE_STATS (frame, READDIR);
+        STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, buf, xdata);
         return 0;
 }
 
-int32_t
-io_stats_lookup_cbk (call_frame_t *frame,
-                     void *cookie,
-                     xlator_t *this,
-                     int32_t op_ret,
-                     int32_t op_errno,
-                     inode_t *inode,
-                     struct stat *buf,
-                     dict_t *xattr,
-                     struct stat *postparent)
+
+int
+io_stats_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno,
+                    struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, inode, buf, xattr,
-                      postparent);
+        UPDATE_PROFILE_STATS (frame, FSYNC);
+        STACK_UNWIND_STRICT (fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
         return 0;
 }
 
-int32_t
-io_stats_symlink_cbk (call_frame_t *frame,
-                      void *cookie,
-                      xlator_t *this,
-                      int32_t op_ret,
-                      int32_t op_errno,
-                      inode_t *inode,
-                      struct stat *buf,
-                      struct stat *preparent,
-                      struct stat *postparent)
+
+int
+io_stats_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno,
+                      struct iatt *preop, struct iatt *postop, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, inode, buf,
-                      preparent, postparent);
+        UPDATE_PROFILE_STATS (frame, SETATTR);
+        STACK_UNWIND_STRICT (setattr, frame, op_ret, op_errno, preop, postop, xdata);
         return 0;
 }
 
-int32_t
-io_stats_mknod_cbk (call_frame_t *frame,
-                    void *cookie,
-                    xlator_t *this,
-                    int32_t op_ret,
-                    int32_t op_errno,
-                    inode_t *inode,
-                    struct stat *buf,
-                    struct stat *preparent,
-                    struct stat *postparent)
+
+int
+io_stats_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno,
+                     struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, inode, buf,
-                      preparent, postparent);
+        UPDATE_PROFILE_STATS (frame, UNLINK);
+        STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno,
+                             preparent, postparent, xdata);
         return 0;
+
 }
 
 
-int32_t
-io_stats_mkdir_cbk (call_frame_t *frame,
-                    void *cookie,
-                    xlator_t *this,
-                    int32_t op_ret,
-                    int32_t op_errno,
-                    inode_t *inode,
-                    struct stat *buf,
-                    struct stat *preparent,
-                    struct stat *postparent)
+int
+io_stats_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                     struct iatt *preoldparent, struct iatt *postoldparent,
+                     struct iatt *prenewparent, struct iatt *postnewparent, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, inode, buf,
-                      preparent, postparent);
+        UPDATE_PROFILE_STATS (frame, RENAME);
+        STACK_UNWIND_STRICT (rename, frame, op_ret, op_errno, buf,
+                             preoldparent, postoldparent,
+                             prenewparent, postnewparent, xdata);
         return 0;
 }
 
-int32_t
-io_stats_link_cbk (call_frame_t *frame,
-                   void *cookie,
-                   xlator_t *this,
-                   int32_t op_ret,
-                   int32_t op_errno,
-                   inode_t *inode,
-                   struct stat *buf,
-                   struct stat *preparent,
-                   struct stat *postparent)
+
+int
+io_stats_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, const char *buf,
+                       struct iatt *sbuf, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, inode, buf,
-                      preparent, postparent);
+        UPDATE_PROFILE_STATS (frame, READLINK);
+        STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, buf, sbuf, xdata);
         return 0;
 }
 
-int32_t
-io_stats_flush_cbk (call_frame_t *frame,
-                    void *cookie,
-                    xlator_t *this,
-                    int32_t op_ret,
-                    int32_t op_errno)
+
+int
+io_stats_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno,
+                     inode_t *inode, struct iatt *buf,
+                     dict_t *xdata, struct iatt *postparent)
 {
-        STACK_UNWIND (frame, op_ret, op_errno);
+        UPDATE_PROFILE_STATS (frame, LOOKUP);
+        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf, xdata,
+                             postparent);
         return 0;
 }
 
 
-int32_t
-io_stats_opendir_cbk (call_frame_t *frame,
-                      void *cookie,
-                      xlator_t *this,
-                      int32_t op_ret,
-                      int32_t op_errno,
-                      fd_t *fd)
+int
+io_stats_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno,
+                      inode_t *inode, struct iatt *buf,
+                      struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, fd);
+        UPDATE_PROFILE_STATS (frame, SYMLINK);
+        STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
         return 0;
 }
 
-int32_t
-io_stats_rmdir_cbk (call_frame_t *frame,
-                    void *cookie,
-                    xlator_t *this,
-                    int32_t op_ret,
-                    int32_t op_errno,
-                    struct stat *preparent,
-                    struct stat *postparent)
+
+int
+io_stats_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno,
+                    inode_t *inode, struct iatt *buf,
+                    struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, preparent, postparent);
+        UPDATE_PROFILE_STATS (frame, MKNOD);
+        STACK_UNWIND_STRICT (mknod, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
         return 0;
 }
 
-int32_t
-io_stats_truncate_cbk (call_frame_t *frame,
-                       void *cookie,
-                       xlator_t *this,
-                       int32_t op_ret,
-                       int32_t op_errno,
-                       struct stat *prebuf,
-                       struct stat *postbuf)
+
+int
+io_stats_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno,
+                    inode_t *inode, struct iatt *buf,
+                    struct iatt *preparent, struct iatt *postparent,
+                    dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, prebuf, postbuf);
+        struct ios_stat *iosstat = NULL;
+        char   *path = frame->local;
+
+        UPDATE_PROFILE_STATS (frame, MKDIR);
+        if (op_ret < 0)
+                goto unwind;
+
+        iosstat = GF_CALLOC (1, sizeof (*iosstat), gf_io_stats_mt_ios_stat);
+        if (iosstat) {
+                LOCK_INIT (&iosstat->lock);
+                iosstat->filename = gf_strdup(path);
+                uuid_copy (iosstat->gfid, buf->ia_gfid);
+                ios_inode_ctx_set (inode, this, iosstat);
+        }
+
+unwind:
+        /* local is assigned with path */
+        GF_FREE (frame->local);
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
         return 0;
 }
 
-int32_t
-io_stats_utimens_cbk (call_frame_t *frame,
-                      void *cookie,
-                      xlator_t *this,
-                      int32_t op_ret,
-                      int32_t op_errno,
-                      struct stat *buf)
+
+int
+io_stats_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno,
+                   inode_t *inode, struct iatt *buf,
+                   struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, buf);
+        UPDATE_PROFILE_STATS (frame, LINK);
+        STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
         return 0;
 }
 
-int32_t
-io_stats_statfs_cbk (call_frame_t *frame,
-                     void *cookie,
-                     xlator_t *this,
-                     int32_t op_ret,
-                     int32_t op_errno,
-                     struct statvfs *buf)
+
+int
+io_stats_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, buf);
+        UPDATE_PROFILE_STATS (frame, FLUSH);
+        STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
-int32_t
-io_stats_setxattr_cbk (call_frame_t *frame,
-                       void *cookie,
-                       xlator_t *this,
-                       int32_t op_ret,
-                       int32_t op_errno)
+
+int
+io_stats_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno);
+        struct ios_stat *iosstat = NULL;
+        int              ret     = -1;
+
+        UPDATE_PROFILE_STATS (frame, OPENDIR);
+        if (op_ret < 0)
+                goto unwind;
+
+        ios_fd_ctx_set (fd, this, 0);
+
+        ret = ios_inode_ctx_get (fd->inode, this, &iosstat);
+        if (!ret)
+                BUMP_STATS (iosstat, IOS_STATS_TYPE_OPENDIR);
+
+unwind:
+        STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd, xdata);
         return 0;
 }
 
-int32_t
-io_stats_getxattr_cbk (call_frame_t *frame,
-                       void *cookie,
-                       xlator_t *this,
-                       int32_t op_ret,
-                       int32_t op_errno,
-                       dict_t *dict)
-{
-        io_stats_private_t *priv = this->private;
-        int i = 0;
-        char keycont[] = "trusted.glusterfs.hits." /* 23 chars */
-                         "0123456789"
-                         "0123456789"
-                         "0123456789"
-                         "0123456789";
-        int ret = -1;
-
-        memset (keycont + 23, '\0', 40);
 
-        for (i = 0; i < GF_FOP_MAXVALUE; i++) {
-                if (!(priv->fop_records[i].enabled &&
-                      priv->fop_records[i].name))
-                        continue;
+int
+io_stats_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno,
+                    struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
+{
 
-                strncpy (keycont + 23, priv->fop_records[i].name, 40);
-                LOCK(&priv->lock);
-                {
-                        ret = dict_set_uint32 (dict, keycont,
-                                               priv->fop_records[i].hits);
-                }
-                UNLOCK(&priv->lock);
-                if (ret < 0) {
-                        gf_log (this->name, GF_LOG_ERROR,
-                                "dict set failed for xattr %s",
-                                keycont);
-                        break;
-                }
-        }
+        UPDATE_PROFILE_STATS (frame, RMDIR);
 
-        STACK_UNWIND (frame, op_ret, op_errno, dict);
+        STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno,
+                             preparent, postparent, xdata);
         return 0;
 }
 
-int32_t
-io_stats_removexattr_cbk (call_frame_t *frame,
-                          void *cookie,
-                          xlator_t *this,
-                          int32_t op_ret,
-                          int32_t op_errno)
+
+int
+io_stats_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno,
+                       struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno);
+        UPDATE_PROFILE_STATS (frame, TRUNCATE);
+        STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno,
+                             prebuf, postbuf, xdata);
         return 0;
 }
 
 
-int32_t
-io_stats_fsyncdir_cbk (call_frame_t *frame,
-                       void *cookie,
-                       xlator_t *this,
-                       int32_t op_ret,
-                       int32_t op_errno)
+int
+io_stats_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct statvfs *buf, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno);
+        UPDATE_PROFILE_STATS (frame, STATFS);
+        STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, buf, xdata);
         return 0;
 }
 
-int32_t
-io_stats_access_cbk (call_frame_t *frame,
-                     void *cookie,
-                     xlator_t *this,
-                     int32_t op_ret,
-                     int32_t op_errno)
+
+int
+io_stats_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno);
+        UPDATE_PROFILE_STATS (frame, SETXATTR);
+        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
-int32_t
-io_stats_ftruncate_cbk (call_frame_t *frame,
-                        void *cookie,
-                        xlator_t *this,
-                        int32_t op_ret,
-                        int32_t op_errno,
-                        struct stat *prebuf,
-                        struct stat *postbuf)
+
+int
+io_stats_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, prebuf, postbuf);
+        UPDATE_PROFILE_STATS (frame, GETXATTR);
+        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
         return 0;
 }
 
-int32_t
-io_stats_fstat_cbk (call_frame_t *frame,
-                    void *cookie,
-                    xlator_t *this,
-                    int32_t op_ret,
-                    int32_t op_errno,
-                    struct stat *buf)
+
+int
+io_stats_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                          int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, buf);
+        UPDATE_PROFILE_STATS (frame, REMOVEXATTR);
+        STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
-int32_t
-io_stats_lk_cbk (call_frame_t *frame,
-                 void *cookie,
-                 xlator_t *this,
-                 int32_t op_ret,
-                 int32_t op_errno,
-                 struct flock *lock)
+int
+io_stats_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, lock);
+        UPDATE_PROFILE_STATS (frame, FSETXATTR);
+        STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
 
-int32_t
-io_stats_setdents_cbk (call_frame_t *frame,
-                       void *cookie,
-                       xlator_t *this,
-                       int32_t op_ret,
-                       int32_t op_errno)
+int
+io_stats_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno);
+        UPDATE_PROFILE_STATS (frame, FGETXATTR);
+        STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict, xdata);
         return 0;
 }
 
-int32_t
-io_stats_entrylk_cbk (call_frame_t *frame,
-                      void *cookie,
-                      xlator_t *this,
-                      int32_t op_ret,
-                      int32_t op_errno)
+
+int
+io_stats_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                           int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno);
+        UPDATE_PROFILE_STATS (frame, FREMOVEXATTR);
+        STACK_UNWIND_STRICT (fremovexattr, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
 
-int32_t
-io_stats_xattrop_cbk (call_frame_t *frame,
-                      void *cookie,
-                      xlator_t *this,
-                      int32_t op_ret,
-                      int32_t op_errno,
-                      dict_t *dict)
+int
+io_stats_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, dict);
+        UPDATE_PROFILE_STATS (frame, FSYNCDIR);
+        STACK_UNWIND_STRICT (fsyncdir, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
-int32_t
-io_stats_fxattrop_cbk (call_frame_t *frame,
-                       void *cookie,
-                       xlator_t *this,
-                       int32_t op_ret,
-                       int32_t op_errno,
-                       dict_t *dict)
+
+int
+io_stats_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, dict);
+        UPDATE_PROFILE_STATS (frame, ACCESS);
+        STACK_UNWIND_STRICT (access, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
-int32_t
-io_stats_inodelk_cbk (call_frame_t *frame,
-                      void *cookie,
-                      xlator_t *this,
-                      int32_t op_ret,
-                      int32_t op_errno)
+
+int
+io_stats_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno,
+                        struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
 {
-        STACK_UNWIND (frame, op_ret, op_errno);
+        UPDATE_PROFILE_STATS (frame, FTRUNCATE);
+        STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno,
+                             prebuf, postbuf, xdata);
         return 0;
 }
 
 
-int32_t
+int
+io_stats_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
+{
+        UPDATE_PROFILE_STATS (frame, FSTAT);
+        STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, buf, xdata);
+        return 0;
+}
+
+
+int
+io_stats_fallocate_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+		       int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+		       struct iatt *postbuf, dict_t *xdata)
+{
+	UPDATE_PROFILE_STATS(frame, FALLOCATE);
+	STACK_UNWIND_STRICT(fallocate, frame, op_ret, op_errno, prebuf, postbuf,
+			    xdata);
+	return 0;
+}
+
+
+int
+io_stats_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+		     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+		     struct iatt *postbuf, dict_t *xdata)
+{
+	UPDATE_PROFILE_STATS(frame, DISCARD);
+	STACK_UNWIND_STRICT(discard, frame, op_ret, op_errno, prebuf, postbuf,
+			    xdata);
+	return 0;
+}
+
+int
+io_stats_zerofill_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                     struct iatt *postbuf, dict_t *xdata)
+{
+        UPDATE_PROFILE_STATS(frame, ZEROFILL);
+        STACK_UNWIND_STRICT(zerofill, frame, op_ret, op_errno, prebuf, postbuf,
+                            xdata);
+        return 0;
+}
+
+int
+io_stats_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
+{
+        UPDATE_PROFILE_STATS (frame, LK);
+        STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, lock, xdata);
+        return 0;
+}
+
+
+int
+io_stats_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        UPDATE_PROFILE_STATS (frame, ENTRYLK);
+        STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+
+int
+io_stats_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        UPDATE_PROFILE_STATS (frame, XATTROP);
+        STACK_UNWIND_STRICT (xattrop, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+}
+
+
+int
+io_stats_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        UPDATE_PROFILE_STATS (frame, FXATTROP);
+        STACK_UNWIND_STRICT (fxattrop, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+}
+
+
+int
+io_stats_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        UPDATE_PROFILE_STATS (frame, INODELK);
+        STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int
 io_stats_entrylk (call_frame_t *frame, xlator_t *this,
                   const char *volume, loc_t *loc, const char *basename,
-                  entrylk_cmd cmd, entrylk_type type)
+                  entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
-        BUMP_HIT(ENTRYLK);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_entrylk_cbk,
+        STACK_WIND (frame, io_stats_entrylk_cbk,
                     FIRST_CHILD (this),
                     FIRST_CHILD (this)->fops->entrylk,
-                    volume, loc, basename, cmd, type);
+                    volume, loc, basename, cmd, type, xdata);
         return 0;
 }
 
-int32_t
-io_stats_inodelk (call_frame_t *frame,
-                  xlator_t *this,
-                  const char *volume, loc_t *loc, int32_t cmd, struct flock *flock)
+
+int
+io_stats_inodelk (call_frame_t *frame, xlator_t *this,
+                  const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
 {
 
-        BUMP_HIT(INODELK);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_inodelk_cbk,
+        STACK_WIND (frame, io_stats_inodelk_cbk,
                     FIRST_CHILD (this),
                     FIRST_CHILD (this)->fops->inodelk,
-                    volume, loc, cmd, flock);
+                    volume, loc, cmd, flock, xdata);
         return 0;
 }
 
 
-int32_t
-io_stats_finodelk_cbk (call_frame_t *frame,
-                       void *cookie,
-                       xlator_t *this,
-                       int32_t op_ret,
-                       int32_t op_errno)
+int
+io_stats_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
 
-        STACK_UNWIND (frame, op_ret, op_errno);
+        UPDATE_PROFILE_STATS (frame, FINODELK);
+        STACK_UNWIND_STRICT (finodelk, frame, op_ret, op_errno, xdata);
         return 0;
 }
 
-int32_t
-io_stats_finodelk (call_frame_t *frame,
-                   xlator_t *this,
-                   const char *volume, fd_t *fd, int32_t cmd, struct flock *flock)
+
+int
+io_stats_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
+                   fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
 {
-        BUMP_HIT(FINODELK);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_finodelk_cbk,
+        STACK_WIND (frame, io_stats_finodelk_cbk,
                     FIRST_CHILD (this),
                     FIRST_CHILD (this)->fops->finodelk,
-                    volume, fd, cmd, flock);
+                    volume, fd, cmd, flock, xdata);
         return 0;
 }
 
 
-int32_t
-io_stats_xattrop (call_frame_t *frame,
-                  xlator_t *this,
-                  loc_t *loc,
-                  gf_xattrop_flags_t flags,
-                  dict_t *dict)
+int
+io_stats_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                  gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
 {
-        BUMP_HIT(XATTROP);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_xattrop_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->xattrop,
-                    loc, flags, dict);
-
+                    loc, flags, dict, xdata);
         return 0;
 }
 
-int32_t
-io_stats_fxattrop (call_frame_t *frame,
-                   xlator_t *this,
-                   fd_t *fd,
-                   gf_xattrop_flags_t flags,
-                   dict_t *dict)
+
+int
+io_stats_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                   gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
 {
-        BUMP_HIT(FXATTROP);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_fxattrop_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fxattrop,
-                    fd, flags, dict);
-
+                    fd, flags, dict, xdata);
         return 0;
 }
 
-int32_t
-io_stats_lookup (call_frame_t *frame,
-                 xlator_t *this,
-                 loc_t *loc,
-                 dict_t *xattr_req)
+
+int
+io_stats_lookup (call_frame_t *frame, xlator_t *this,
+                 loc_t *loc, dict_t *xdata)
 {
-        BUMP_HIT(LOOKUP);
+        START_FOP_LATENCY (frame);
 
         STACK_WIND (frame, io_stats_lookup_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->lookup,
-                    loc, xattr_req);
-
+                    loc, xdata);
         return 0;
 }
 
-int32_t
-io_stats_stat (call_frame_t *frame,
-               xlator_t *this,
-               loc_t *loc)
+
+int
+io_stats_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
-        BUMP_HIT(STAT);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_stat_cbk,
+        STACK_WIND (frame, io_stats_stat_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->stat,
-                    loc);
-
+                    loc, xdata);
         return 0;
 }
 
-int32_t
-io_stats_readlink (call_frame_t *frame,
-                   xlator_t *this,
-                   loc_t *loc,
-                   size_t size)
+
+int
+io_stats_readlink (call_frame_t *frame, xlator_t *this,
+                   loc_t *loc, size_t size, dict_t *xdata)
 {
-        BUMP_HIT(READLINK);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_readlink_cbk,
+        STACK_WIND (frame, io_stats_readlink_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->readlink,
-                    loc,
-                    size);
-
+                    loc, size, xdata);
         return 0;
 }
 
-int32_t
-io_stats_mknod (call_frame_t *frame,
-                xlator_t *this,
-                loc_t *loc,
-                mode_t mode,
-                dev_t dev)
+
+int
+io_stats_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                mode_t mode, dev_t dev, mode_t umask, dict_t *xdata)
 {
-        BUMP_HIT(MKNOD);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_mknod_cbk,
+        STACK_WIND (frame, io_stats_mknod_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->mknod,
-                    loc,
-                    mode,
-                    dev);
-
+                    loc, mode, dev, umask, xdata);
         return 0;
 }
 
-int32_t
-io_stats_mkdir (call_frame_t *frame,
-                xlator_t *this,
-                loc_t *loc,
-                mode_t mode)
+
+int
+io_stats_mkdir (call_frame_t *frame, xlator_t *this,
+                loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata)
 {
-        BUMP_HIT(MKDIR);
+        frame->local = gf_strdup (loc->path);
 
-        STACK_WIND (frame,
-                    io_stats_mkdir_cbk,
+        START_FOP_LATENCY (frame);
+
+        STACK_WIND (frame, io_stats_mkdir_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->mkdir,
-                    loc,
-                    mode);
+                    loc, mode, umask, xdata);
         return 0;
 }
 
-int32_t
-io_stats_unlink (call_frame_t *frame,
-                 xlator_t *this,
-                 loc_t *loc)
+
+int
+io_stats_unlink (call_frame_t *frame, xlator_t *this,
+                 loc_t *loc, int xflag, dict_t *xdata)
 {
-        BUMP_HIT(UNLINK);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_unlink_cbk,
+        STACK_WIND (frame, io_stats_unlink_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->unlink,
-                    loc);
+                    loc, xflag, xdata);
         return 0;
 }
 
-int32_t
-io_stats_rmdir (call_frame_t *frame,
-                xlator_t *this,
-                loc_t *loc)
+
+int
+io_stats_rmdir (call_frame_t *frame, xlator_t *this,
+                loc_t *loc, int flags, dict_t *xdata)
 {
-        BUMP_HIT(RMDIR);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_rmdir_cbk,
+        STACK_WIND (frame, io_stats_rmdir_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->rmdir,
-                    loc);
-
+                    loc, flags, xdata);
         return 0;
 }
 
-int32_t
-io_stats_symlink (call_frame_t *frame,
-                  xlator_t *this,
-                  const char *linkpath,
-                  loc_t *loc)
+
+int
+io_stats_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
+                  loc_t *loc, mode_t umask, dict_t *xdata)
 {
-        BUMP_HIT(SYMLINK);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_symlink_cbk,
+        STACK_WIND (frame, io_stats_symlink_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->symlink,
-                    linkpath,
-                    loc);
-
+                    linkpath, loc, umask, xdata);
         return 0;
 }
 
-int32_t
-io_stats_rename (call_frame_t *frame,
-                 xlator_t *this,
-                 loc_t *oldloc,
-                 loc_t *newloc)
+
+int
+io_stats_rename (call_frame_t *frame, xlator_t *this,
+                 loc_t *oldloc, loc_t *newloc, dict_t *xdata)
 {
-        BUMP_HIT(RENAME);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_rename_cbk,
+        STACK_WIND (frame, io_stats_rename_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->rename,
-                    oldloc,
-                    newloc);
-
+                    oldloc, newloc, xdata);
         return 0;
 }
 
-int32_t
-io_stats_link (call_frame_t *frame,
-               xlator_t *this,
-               loc_t *oldloc,
-               loc_t *newloc)
+
+int
+io_stats_link (call_frame_t *frame, xlator_t *this,
+               loc_t *oldloc, loc_t *newloc, dict_t *xdata)
 {
-        BUMP_HIT(LINK);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_link_cbk,
+        STACK_WIND (frame, io_stats_link_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->link,
-                    oldloc,
-                    newloc);
+                    oldloc, newloc, xdata);
         return 0;
 }
 
-int32_t
-io_stats_setattr (call_frame_t *frame,
-                  xlator_t *this,
-                  loc_t *loc,
-                  struct stat *stbuf,
-                  int32_t valid)
+
+int
+io_stats_setattr (call_frame_t *frame, xlator_t *this,
+                  loc_t *loc, struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
-        BUMP_HIT(SETATTR);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_setattr_cbk,
+        STACK_WIND (frame, io_stats_setattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->setattr,
-                    loc,
-                    stbuf, valid);
-
+                    loc, stbuf, valid, xdata);
         return 0;
 }
 
-int32_t
-io_stats_truncate (call_frame_t *frame,
-                   xlator_t *this,
-                   loc_t *loc,
-                   off_t offset)
+
+int
+io_stats_truncate (call_frame_t *frame, xlator_t *this,
+                   loc_t *loc, off_t offset, dict_t *xdata)
 {
-        BUMP_HIT(TRUNCATE);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_truncate_cbk,
+        STACK_WIND (frame, io_stats_truncate_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->truncate,
-                    loc,
-                    offset);
-
+                    loc, offset, xdata);
         return 0;
 }
 
-int32_t
-io_stats_open (call_frame_t *frame,
-               xlator_t *this,
-               loc_t *loc,
-               int32_t flags,
-               fd_t *fd, int32_t wbflags)
+
+int
+io_stats_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
+               int32_t flags, fd_t *fd, dict_t *xdata)
 {
-        BUMP_HIT(OPEN);
+        frame->local = gf_strdup (loc->path);
+
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_open_cbk,
+        STACK_WIND (frame, io_stats_open_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->open,
-                    loc,
-                    flags,
-                    fd, wbflags);
+                    loc, flags, fd, xdata);
         return 0;
 }
 
-int32_t
-io_stats_create (call_frame_t *frame,
-                 xlator_t *this,
-                 loc_t *loc,
-                 int32_t flags,
-                 mode_t mode,
-                 fd_t *fd)
+
+int
+io_stats_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)
 {
-        BUMP_HIT(CREATE);
+        frame->local = gf_strdup (loc->path);
 
-        STACK_WIND (frame,
-                    io_stats_create_cbk,
+        START_FOP_LATENCY (frame);
+
+        STACK_WIND (frame, io_stats_create_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->create,
-                    loc,
-                    flags,
-                    mode,
-                    fd);
+                    loc, flags, mode, umask, fd, xdata);
         return 0;
 }
 
-int32_t
-io_stats_readv (call_frame_t *frame,
-                xlator_t *this,
-                fd_t *fd,
-                size_t size,
-                off_t offset)
+
+int
+io_stats_readv (call_frame_t *frame, xlator_t *this,
+                fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata)
 {
-        BUMP_HIT(READ);
+        frame->local = fd;
 
-        STACK_WIND (frame,
-                    io_stats_readv_cbk,
+        START_FOP_LATENCY (frame);
+
+        STACK_WIND (frame, io_stats_readv_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->readv,
-                    fd,
-                    size,
-                    offset);
+                    fd, size, offset, flags, xdata);
         return 0;
 }
 
-int32_t
-io_stats_writev (call_frame_t *frame,
-                 xlator_t *this,
-                 fd_t *fd,
-                 struct iovec *vector,
-                 int32_t count,
-                 off_t offset,
-                 struct iobref *iobref)
-{
-        int i = 0;
-        io_stats_private_t *priv = NULL;
 
-        priv = this->private;
+int
+io_stats_writev (call_frame_t *frame, xlator_t *this,
+                 fd_t *fd, struct iovec *vector,
+                 int32_t count, off_t offset,
+                 uint32_t flags, struct iobref *iobref, dict_t *xdata)
+{
+        int                 len = 0;
 
-        BUMP_HIT(WRITE);
+        if (fd->inode)
+                frame->local = fd->inode;
+        len = iov_length (vector, count);
 
-        for (i=0; i < GF_IO_STAT_BLK_SIZE_MAX; i++) {
-                if (priv->write[i].size > iov_length (vector, count)) {
-                        break;
-                }
-        }
-        priv->write[i].hits++;
+        BUMP_WRITE (fd, len);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_writev_cbk,
+        STACK_WIND (frame, io_stats_writev_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->writev,
-                    fd,
-                    vector,
-                    count,
-                    offset,
-                    iobref);
+                    fd, vector, count, offset, flags, iobref, xdata);
         return 0;
+
 }
 
-int32_t
-io_stats_statfs (call_frame_t *frame,
-                 xlator_t *this,
-                 loc_t *loc)
+
+int
+io_stats_statfs (call_frame_t *frame, xlator_t *this,
+                 loc_t *loc, dict_t *xdata)
 {
-        BUMP_HIT(STATFS);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_statfs_cbk,
-                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->statfs,
-                    loc);
+        STACK_WIND (frame, io_stats_statfs_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->statfs,
+                    loc, xdata);
         return 0;
 }
 
-int32_t
-io_stats_flush (call_frame_t *frame,
-                xlator_t *this,
-                fd_t *fd)
+
+int
+io_stats_flush (call_frame_t *frame, xlator_t *this,
+                fd_t *fd, dict_t *xdata)
 {
-        BUMP_HIT(FLUSH);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_flush_cbk,
+        STACK_WIND (frame, io_stats_flush_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->flush,
-                    fd);
+                    fd, xdata);
         return 0;
 }
 
 
-int32_t
-io_stats_fsync (call_frame_t *frame,
-                xlator_t *this,
-                fd_t *fd,
-                int32_t flags)
+int
+io_stats_fsync (call_frame_t *frame, xlator_t *this,
+                fd_t *fd, int32_t flags, dict_t *xdata)
 {
-        BUMP_HIT(FSYNC);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_fsync_cbk,
+        STACK_WIND (frame, io_stats_fsync_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fsync,
-                    fd,
-                    flags);
+                    fd, flags, xdata);
         return 0;
 }
 
-int32_t
-io_stats_setxattr (call_frame_t *frame,
-                   xlator_t *this,
-                   loc_t *loc,
-                   dict_t *dict,
-                   int32_t flags)
+
+int
+conditional_dump (dict_t *dict, char *key, data_t *value, void *data)
+{
+        struct {
+                xlator_t       *this;
+                inode_t        *inode;
+                const char     *path;
+        } *stub;
+        xlator_t             *this = NULL;
+        char                 *filename = NULL;
+        FILE                 *logfp = NULL;
+        struct ios_dump_args args = {0};
+
+        stub  = data;
+        this  = stub->this;
+
+        filename = alloca (value->len + 1);
+        memset (filename, 0, value->len + 1);
+        memcpy (filename, data_to_str (value), value->len);
+
+        if (fnmatch ("*io*stat*dump", key, 0) == 0) {
+
+                if (!strncmp (filename, "", 1)) {
+                        gf_log (this->name, GF_LOG_ERROR, "No filename given");
+                        return -1;
+                }
+                logfp = fopen (filename, "w+");
+                if (!logfp) {
+                        gf_log (this->name, GF_LOG_ERROR, "failed to open %s "
+                                "for writing", filename);
+                        return -1;
+                    }
+                (void) ios_dump_args_init (&args, IOS_DUMP_TYPE_FILE,
+                                           logfp);
+                io_stats_dump (this, &args, GF_CLI_INFO_ALL, _gf_false);
+                fclose (logfp);
+        }
+        return 0;
+}
+
+
+int
+io_stats_setxattr (call_frame_t *frame, xlator_t *this,
+                   loc_t *loc, dict_t *dict,
+                   int32_t flags, dict_t *xdata)
 {
-        BUMP_HIT(SETXATTR);
+        struct {
+                xlator_t     *this;
+                inode_t      *inode;
+                const char   *path;
+        } stub;
+
+        stub.this  = this;
+        stub.inode = loc->inode;
+        stub.path  = loc->path;
+
+        dict_foreach (dict, conditional_dump, &stub);
 
-        STACK_WIND (frame,
-                    io_stats_setxattr_cbk,
+        START_FOP_LATENCY (frame);
+
+        STACK_WIND (frame, io_stats_setxattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->setxattr,
-                    loc,
-                    dict,
-                    flags);
+                    loc, dict, flags, xdata);
         return 0;
 }
 
-int32_t
-io_stats_getxattr (call_frame_t *frame,
-                   xlator_t *this,
-                   loc_t *loc,
-                   const char *name)
+
+int
+io_stats_getxattr (call_frame_t *frame, xlator_t *this,
+                   loc_t *loc, const char *name, dict_t *xdata)
 {
-        BUMP_HIT(GETXATTR);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_getxattr_cbk,
+        STACK_WIND (frame, io_stats_getxattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->getxattr,
-                    loc,
-                    name);
+                    loc, name, xdata);
         return 0;
 }
 
-int32_t
-io_stats_removexattr (call_frame_t *frame,
-                      xlator_t *this,
-                      loc_t *loc,
-                      const char *name)
+
+int
+io_stats_removexattr (call_frame_t *frame, xlator_t *this,
+                      loc_t *loc, const char *name, dict_t *xdata)
 {
-        BUMP_HIT(REMOVEXATTR);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_removexattr_cbk,
+        STACK_WIND (frame, io_stats_removexattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->removexattr,
-                    loc,
-                    name);
+                    loc, name, xdata);
+        return 0;
+}
+
 
+int
+io_stats_fsetxattr (call_frame_t *frame, xlator_t *this,
+                    fd_t *fd, dict_t *dict,
+                    int32_t flags, dict_t *xdata)
+{
+        START_FOP_LATENCY (frame);
+
+        STACK_WIND (frame, io_stats_fsetxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsetxattr,
+                    fd, dict, flags, xdata);
         return 0;
 }
 
-int32_t
-io_stats_opendir (call_frame_t *frame,
-                  xlator_t *this,
-                  loc_t *loc,
-                  fd_t *fd)
+
+int
+io_stats_fgetxattr (call_frame_t *frame, xlator_t *this,
+                    fd_t *fd, const char *name, dict_t *xdata)
 {
-        BUMP_HIT(OPENDIR);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_opendir_cbk,
+        STACK_WIND (frame, io_stats_fgetxattr_cbk,
                     FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->opendir,
-                    loc,
-                    fd);
+                    FIRST_CHILD(this)->fops->fgetxattr,
+                    fd, name, xdata);
         return 0;
 }
 
-int32_t
-io_stats_getdents (call_frame_t *frame,
-                   xlator_t *this,
-                   fd_t *fd,
-                   size_t size,
-                   off_t offset,
-                   int32_t flag)
+
+int
+io_stats_fremovexattr (call_frame_t *frame, xlator_t *this,
+                       fd_t *fd, const char *name, dict_t *xdata)
 {
-        BUMP_HIT(GETDENTS);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_getdents_cbk,
+        STACK_WIND (frame, io_stats_fremovexattr_cbk,
                     FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->getdents,
-                    fd,
-                    size,
-                    offset,
-                    flag);
+                    FIRST_CHILD(this)->fops->fremovexattr,
+                    fd, name, xdata);
         return 0;
 }
 
 
-int32_t
-io_stats_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-                   off_t offset)
+int
+io_stats_opendir (call_frame_t *frame, xlator_t *this,
+                  loc_t *loc, fd_t *fd, dict_t *xdata)
 {
-        BUMP_HIT(READDIRP);
 
-        STACK_WIND (frame, io_stats_readdirp_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->readdir, fd, size, offset);
+        START_FOP_LATENCY (frame);
 
+        STACK_WIND (frame, io_stats_opendir_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->opendir,
+                    loc, fd, xdata);
         return 0;
 }
 
+int
+io_stats_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+                   off_t offset, dict_t *dict)
+{
+        frame->local = fd->inode;
+        START_FOP_LATENCY (frame);
 
+        STACK_WIND (frame, io_stats_readdirp_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readdirp,
+                    fd, size, offset, dict);
+        return 0;
+}
 
-int32_t
-io_stats_readdir (call_frame_t *frame,
-                  xlator_t *this,
-                  fd_t *fd,
-                  size_t size,
-                  off_t offset)
+
+int
+io_stats_readdir (call_frame_t *frame, xlator_t *this,
+                  fd_t *fd, size_t size, off_t offset, dict_t *xdata)
 {
-        BUMP_HIT(READDIR);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_readdir_cbk,
+        STACK_WIND (frame, io_stats_readdir_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->readdir,
-                    fd,
-                    size,
-                    offset);
-
+                    fd, size, offset, xdata);
         return 0;
 }
 
 
-int32_t
-io_stats_fsyncdir (call_frame_t *frame,
-                   xlator_t *this,
-                   fd_t *fd,
-                   int32_t datasync)
+int
+io_stats_fsyncdir (call_frame_t *frame, xlator_t *this,
+                   fd_t *fd, int32_t datasync, dict_t *xdata)
 {
-        BUMP_HIT(FSYNCDIR);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_fsyncdir_cbk,
+        STACK_WIND (frame, io_stats_fsyncdir_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fsyncdir,
-                    fd,
-                    datasync);
+                    fd, datasync, xdata);
         return 0;
 }
 
-int32_t
-io_stats_access (call_frame_t *frame,
-                 xlator_t *this,
-                 loc_t *loc,
-                 int32_t mask)
+
+int
+io_stats_access (call_frame_t *frame, xlator_t *this,
+                 loc_t *loc, int32_t mask, dict_t *xdata)
 {
-        BUMP_HIT(ACCESS);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_access_cbk,
+        STACK_WIND (frame, io_stats_access_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->access,
-                    loc,
-                    mask);
+                    loc, mask, xdata);
         return 0;
 }
 
-int32_t
-io_stats_ftruncate (call_frame_t *frame,
-                    xlator_t *this,
-                    fd_t *fd,
-                    off_t offset)
+
+int
+io_stats_ftruncate (call_frame_t *frame, xlator_t *this,
+                    fd_t *fd, off_t offset, dict_t *xdata)
 {
-        BUMP_HIT(FTRUNCATE);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_ftruncate_cbk,
+        STACK_WIND (frame, io_stats_ftruncate_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->ftruncate,
-                    fd,
-                    offset);
-
+                    fd, offset, xdata);
         return 0;
 }
 
-int32_t
-io_stats_fsetattr (call_frame_t *frame,
-                   xlator_t *this,
-                   fd_t *fd,
-                   struct stat *stbuf,
-                   int32_t valid)
+
+int
+io_stats_fsetattr (call_frame_t *frame, xlator_t *this,
+                   fd_t *fd, struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
-        BUMP_HIT(FSETATTR);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_setattr_cbk,
+        STACK_WIND (frame, io_stats_setattr_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fsetattr,
-                    fd,
-                    stbuf, valid);
+                    fd, stbuf, valid, xdata);
         return 0;
 }
 
-int32_t
-io_stats_fstat (call_frame_t *frame,
-                xlator_t *this,
-                fd_t *fd)
+
+int
+io_stats_fstat (call_frame_t *frame, xlator_t *this,
+                fd_t *fd, dict_t *xdata)
 {
-        BUMP_HIT(FSTAT);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_fstat_cbk,
+        STACK_WIND (frame, io_stats_fstat_cbk,
                     FIRST_CHILD(this),
                     FIRST_CHILD(this)->fops->fstat,
-                    fd);
+                    fd, xdata);
         return 0;
 }
 
-int32_t
-io_stats_lk (call_frame_t *frame,
-             xlator_t *this,
-             fd_t *fd,
-             int32_t cmd,
-             struct flock *lock)
+
+int
+io_stats_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+		   off_t offset, size_t len, dict_t *xdata)
 {
-        BUMP_HIT(LK);
+	START_FOP_LATENCY(frame);
+
+	STACK_WIND(frame, io_stats_fallocate_cbk, FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fallocate, fd, mode, offset, len,
+		   xdata);
+
+	return 0;
+}
+
+
+int
+io_stats_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+		 size_t len, dict_t *xdata)
+{
+	START_FOP_LATENCY(frame);
+
+	STACK_WIND(frame, io_stats_discard_cbk, FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->discard, fd, offset, len, xdata);
+
+	return 0;
+}
+
+int
+io_stats_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+                 off_t len, dict_t *xdata)
+{
+        START_FOP_LATENCY(frame);
+
+        STACK_WIND(frame, io_stats_zerofill_cbk, FIRST_CHILD(this),
+                   FIRST_CHILD(this)->fops->zerofill, fd, offset, len, xdata);
 
-        STACK_WIND (frame,
-                    io_stats_lk_cbk,
-                    FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->lk,
-                    fd,
-                    cmd,
-                    lock);
         return 0;
 }
 
-int32_t
-io_stats_setdents (call_frame_t *frame,
-                   xlator_t *this,
-                   fd_t *fd,
-                   int32_t flags,
-                   dir_entry_t *entries,
-                   int32_t count)
+
+int
+io_stats_lk (call_frame_t *frame, xlator_t *this,
+             fd_t *fd, int32_t cmd, struct gf_flock *lock, dict_t *xdata)
 {
-        BUMP_HIT(SETDENTS);
+        START_FOP_LATENCY (frame);
 
-        STACK_WIND (frame,
-                    io_stats_setdents_cbk,
+        STACK_WIND (frame, io_stats_lk_cbk,
                     FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->setdents,
-                    fd,
-                    flags,
-                    entries,
-                    count);
+                    FIRST_CHILD(this)->fops->lk,
+                    fd, cmd, lock, xdata);
         return 0;
 }
 
 
-int32_t
-io_stats_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)
+int
+io_stats_release (xlator_t *this, fd_t *fd)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, fchecksum, dchecksum);
+        struct ios_fd  *iosfd = NULL;
+        struct ios_conf *conf = NULL;
+
+        BUMP_FOP (RELEASE);
+
+        conf = this->private;
+
+        LOCK (&conf->lock);
+        {
+                conf->cumulative.nr_opens--;
+        }
+        UNLOCK (&conf->lock);
+
+        ios_fd_ctx_get (fd, this, &iosfd);
+        if (iosfd) {
+                io_stats_dump_fd (this, iosfd);
+
+                GF_FREE (iosfd->filename);
+                GF_FREE (iosfd);
+        }
 
         return 0;
 }
 
-int32_t
-io_stats_checksum (call_frame_t *frame,
-                   xlator_t *this,
-                   loc_t *loc,
-                   int32_t flag)
+
+int
+io_stats_releasedir (xlator_t *this, fd_t *fd)
 {
-        STACK_WIND (frame,
-                    io_stats_checksum_cbk,
-                    FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->checksum,
-                    loc,
-                    flag);
+        BUMP_FOP (RELEASEDIR);
 
         return 0;
 }
 
 
-int32_t
-io_stats_stats_cbk (call_frame_t *frame,
-                    void *cookie,
-                    xlator_t *this,
-                    int32_t op_ret,
-                    int32_t op_errno,
-                    struct xlator_stats *stats)
+int
+io_stats_forget (xlator_t *this, inode_t *inode)
 {
-        STACK_UNWIND (frame, op_ret, op_errno, stats);
+        BUMP_FOP (FORGET);
+        ios_stats_cleanup (this, inode);
         return 0;
 }
 
-int32_t
-io_stats_stats (call_frame_t *frame,
-                xlator_t *this,
-                int32_t flags)
+static int
+ios_init_top_stats (struct ios_conf *conf)
 {
-        STACK_WIND (frame,
-                    io_stats_stats_cbk,
-                    FIRST_CHILD(this),
-                    FIRST_CHILD(this)->mops->stats,
-                    flags);
+        int     i = 0;
+
+        GF_ASSERT (conf);
+
+        for (i = 0; i <IOS_STATS_TYPE_MAX; i++) {
+                conf->list[i].iosstats = GF_CALLOC (1,
+                                         sizeof(*conf->list[i].iosstats),
+                                         gf_io_stats_mt_ios_stat);
+
+                if (!conf->list[i].iosstats)
+                        return -1;
+
+                INIT_LIST_HEAD(&conf->list[i].iosstats->list);
+                LOCK_INIT (&conf->list[i].lock);
+        }
+
+        for (i = 0; i < IOS_STATS_THRU_MAX; i ++) {
+                conf->thru_list[i].iosstats = GF_CALLOC (1,
+                                 sizeof (*conf->thru_list[i].iosstats),
+                                 gf_io_stats_mt_ios_stat);
+
+                if (!conf->thru_list[i].iosstats)
+                        return -1;
+
+                INIT_LIST_HEAD(&conf->thru_list[i].iosstats->list);
+                LOCK_INIT (&conf->thru_list[i].lock);
+        }
 
         return 0;
 }
 
-void
-enable_all_calls (io_stats_private_t *priv, int enabled)
+static void
+ios_destroy_top_stats (struct ios_conf *conf)
 {
-        int i;
-        for (i = 0; i < GF_FOP_MAXVALUE; i++)
-                priv->fop_records[i].enabled = enabled;
-}
+        int                     i = 0;
+        struct ios_stat_head    *list_head = NULL;
+        struct ios_stat_list    *entry     = NULL;
+        struct ios_stat_list    *tmp       = NULL;
+        struct ios_stat_list    *list      = NULL;
+        struct ios_stat         *stat      = NULL;
 
-void
-enable_call (io_stats_private_t *priv, const char *name, int enabled)
-{
-        int i;
-        for (i = 0; i < GF_FOP_MAXVALUE; i++) {
-                if (!priv->fop_records[i].name)
+        GF_ASSERT (conf);
+
+        LOCK (&conf->lock);
+
+        conf->cumulative.nr_opens = 0;
+        conf->cumulative.max_nr_opens = 0;
+        conf->cumulative.max_openfd_time.tv_sec = 0;
+        conf->cumulative.max_openfd_time.tv_usec = 0;
+
+        for (i = 0; i < IOS_STATS_TYPE_MAX; i++) {
+                list_head = &conf->list[i];
+                if (!list_head)
                         continue;
-                if (!strcasecmp(priv->fop_records[i].name, name))
-                        priv->fop_records[i].enabled = enabled;
+                list_for_each_entry_safe (entry, tmp,
+                                          &list_head->iosstats->list, list) {
+                        list = entry;
+                        stat = list->iosstat;
+                        ios_stat_unref (stat);
+                        list_del (&list->list);
+                        GF_FREE (list);
+                        list_head->members--;
+                }
+        }
+
+        for (i = 0; i < IOS_STATS_THRU_MAX; i++) {
+                list_head = &conf->thru_list[i];
+                if (!list_head)
+                        continue;
+                list_for_each_entry_safe (entry, tmp,
+                                          &list_head->iosstats->list, list) {
+                        list = entry;
+                        stat = list->iosstat;
+                        ios_stat_unref (stat);
+                        list_del (&list->list);
+                        GF_FREE (list);
+                        list_head->members--;
+                }
         }
+
+        UNLOCK (&conf->lock);
+
+        return;
 }
 
+static int
+io_stats_clear (struct ios_conf *conf)
+{
+        struct timeval      now;
+        int                 ret = -1;
+
+        GF_ASSERT (conf);
 
-/*
-   include = 1 for "include-ops"
-           = 0 for "exclude-ops"
-*/
-void
-process_call_list (io_stats_private_t *priv, const char *list, int include)
+        if (!gettimeofday (&now, NULL))
+        {
+            LOCK (&conf->lock);
+            {
+                    ios_global_stats_clear (&conf->cumulative, &now);
+                    ios_global_stats_clear (&conf->incremental, &now);
+                    conf->increment = 0;
+            }
+            UNLOCK (&conf->lock);
+            ret = 0;
+        }
+
+        return ret;
+}
+
+int
+reconfigure (xlator_t *this, dict_t *options)
 {
-        enable_all_calls (priv, include ? 0 : 1);
+        struct ios_conf    *conf = NULL;
+        int                 ret = -1;
+        char               *sys_log_str = NULL;
+        int                 sys_log_level = -1;
+        char               *log_str = NULL;
+        int                 log_level = -1;
+
+        if (!this || !this->private)
+                goto out;
+
+        conf = this->private;
 
-        char *call = strsep ((char **)&list, ",");
-        while (call) {
-                enable_call (priv, call, include);
-                call = strsep ((char **)&list, ",");
+        GF_OPTION_RECONF ("dump-fd-stats", conf->dump_fd_stats, options, bool,
+                          out);
+
+        GF_OPTION_RECONF ("count-fop-hits", conf->count_fop_hits, options, bool,
+                          out);
+
+        GF_OPTION_RECONF ("latency-measurement", conf->measure_latency,
+                          options, bool, out);
+
+        GF_OPTION_RECONF ("sys-log-level", sys_log_str, options, str, out);
+        if (sys_log_str) {
+                sys_log_level = glusterd_check_log_level (sys_log_str);
+                set_sys_log_level (sys_log_level);
+        }
+
+        GF_OPTION_RECONF ("log-level", log_str, options, str, out);
+        if (log_str) {
+                log_level = glusterd_check_log_level (log_str);
+                gf_log_set_loglevel (log_level);
         }
+
+        ret = 0;
+out:
+        gf_log (this->name, GF_LOG_DEBUG, "reconfigure returning %d", ret);
+        return ret;
 }
 
 
 int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_io_stats_mt_end + 1);
+
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
+                        " failed");
+                return ret;
+        }
+
+        return ret;
+}
+
+int
 init (xlator_t *this)
 {
-        dict_t *options = NULL;
-        char *includes = NULL, *excludes = NULL;
-        io_stats_private_t *priv = NULL;
-        size_t size = 0;
-        int i = 0;
+        struct ios_conf    *conf = NULL;
+        char               *sys_log_str = NULL;
+        int                 sys_log_level = -1;
+        char               *log_str = NULL;
+        int                 log_level = -1;
+        int                 ret = -1;
 
         if (!this)
                 return -1;
 
-        if (!this->children || this->children->next) {
+        if (!this->children) {
                 gf_log (this->name, GF_LOG_ERROR,
-                        "io_stats translator requires one subvolume");
+                        "io_stats translator requires atleast one subvolume");
                 return -1;
         }
+
         if (!this->parents) {
-                gf_log (this->name, GF_LOG_WARNING,
+                /* This is very much valid as io-stats currently is loaded
+                 * on top of volumes on both client and server, hence this is
+                 * not an warning message */
+                gf_log (this->name, GF_LOG_DEBUG,
                         "dangling volume. check volfile ");
         }
 
-        priv = CALLOC (1, sizeof(*priv));
-        ERR_ABORT (priv);
-
-        options = this->options;
-        includes = data_to_str (dict_get (options, "include-ops"));
-        excludes = data_to_str (dict_get (options, "exclude-ops"));
+        conf = GF_CALLOC (1, sizeof(*conf), gf_io_stats_mt_ios_conf);
 
-        for (i = 0; i < GF_FOP_MAXVALUE; i++) {
-                priv->fop_records[i].name = gf_fop_list[i];
-                priv->fop_records[i].enabled = 1;
-        }
-
-        if (includes && excludes) {
+        if (!conf) {
                 gf_log (this->name, GF_LOG_ERROR,
-                        "must specify only one of 'include-ops' and "
-                        "'exclude-ops'");
+                        "Out of memory.");
                 return -1;
         }
-        if (includes)
-                process_call_list (priv, includes, 1);
-        if (excludes)
-                process_call_list (priv, excludes, 0);
-
-        LOCK_INIT (&priv->lock);
-
-        /* Set this translator's inode table pointer to child node's pointer. */
-        size = GF_UNIT_KB;
-        for (i=0; i < GF_IO_STAT_BLK_SIZE_MAX; i++) {
-                priv->read[i].size = size;
-                priv->write[i].size = size;
-                size *= 2;
+
+        LOCK_INIT (&conf->lock);
+
+        gettimeofday (&conf->cumulative.started_at, NULL);
+        gettimeofday (&conf->incremental.started_at, NULL);
+
+        ret = ios_init_top_stats (conf);
+        if (ret)
+                return -1;
+
+        GF_OPTION_INIT ("dump-fd-stats", conf->dump_fd_stats, bool, out);
+
+        GF_OPTION_INIT ("count-fop-hits", conf->count_fop_hits, bool, out);
+
+        GF_OPTION_INIT ("latency-measurement", conf->measure_latency,
+                          bool, out);
+
+        GF_OPTION_INIT ("sys-log-level", sys_log_str, str, out);
+        if (sys_log_str) {
+                sys_log_level = glusterd_check_log_level (sys_log_str);
+                set_sys_log_level (sys_log_level);
         }
 
-        this->itable = FIRST_CHILD (this)->itable;
-        this->private = priv;
+        GF_OPTION_INIT ("log-level", log_str, str, out);
+        if (log_str) {
+                log_level = glusterd_check_log_level (log_str);
+                gf_log_set_loglevel (log_level);
+        }
 
-        return 0;
+        this->private = conf;
+        ret = 0;
+out:
+        return ret;
 }
 
+
 void
 fini (xlator_t *this)
 {
-        io_stats_private_t *priv = NULL;
+        struct ios_conf *conf = NULL;
 
         if (!this)
                 return;
 
-        priv = this->private;
-        if (priv) {
-                LOCK_DESTROY (&priv->lock);
-                FREE (priv);
-        }
+        conf = this->private;
+
+        if (!conf)
+                return;
+        this->private = NULL;
+
+        ios_destroy_top_stats (conf);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        GF_FREE(conf);
+
+        gf_log (this->name, GF_LOG_INFO,
                 "io-stats translator unloaded");
         return;
 }
 
+int
+notify (xlator_t *this, int32_t event, void *data, ...)
+{
+        int          ret = 0;
+        struct ios_dump_args args = {0};
+        dict_t       *output = NULL;
+        dict_t       *dict = NULL;
+        int32_t       op = 0;
+        int32_t       list_cnt = 0;
+        double        throughput = 0;
+        double        time = 0;
+        gf_boolean_t  is_peek = _gf_false;
+        va_list ap;
+
+        dict = data;
+        va_start (ap, data);
+        output = va_arg (ap, dict_t*);
+        va_end (ap);
+        switch (event) {
+        case GF_EVENT_TRANSLATOR_INFO:
+                ret = dict_get_str_boolean (dict, "clear-stats", _gf_false);
+                if (ret) {
+                        ret = dict_set_int32 (output, "top-op", op);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to set top-op in dict");
+                                goto out;
+                        }
+                        ios_destroy_top_stats (this->private);
+                        ret = ios_init_top_stats (this->private);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to reset top stats");
+                        ret = dict_set_int32 (output, "stats-cleared",
+                                              ret ? 0 : 1);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to set stats-cleared"
+                                        " in dict");
+                        goto out;
+                }
+
+                ret = dict_get_int32 (dict, "top-op", &op);
+                if (!ret) {
+                        ret = dict_get_int32 (dict, "list-cnt", &list_cnt);
+                        if (op > IOS_STATS_TYPE_NONE &&
+                            op < IOS_STATS_TYPE_MAX)
+                                ret = io_stats_dump_stats_to_dict (this, output,
+                                                             op, list_cnt);
+                        if (op == IOS_STATS_TYPE_READ_THROUGHPUT ||
+                                op == IOS_STATS_TYPE_WRITE_THROUGHPUT) {
+                                ret = dict_get_double (dict, "throughput",
+                                                        &throughput);
+                                if (!ret) {
+                                        ret = dict_get_double (dict, "time",
+                                                                &time);
+                                        if (ret)
+                                                goto out;
+                                        ret = dict_set_double (output,
+                                                "throughput", throughput);
+                                        if (ret)
+                                                goto out;
+                                        ret = dict_set_double (output, "time",
+                                                        time);
+                                        if (ret)
+                                                goto out;
+                                }
+                                ret = 0;
+
+                        }
+                } else {
+                        ret = dict_get_int32 (dict, "info-op", &op);
+                        if (ret || op < GF_CLI_INFO_ALL ||
+                            GF_CLI_INFO_CLEAR < op)
+                            op = GF_CLI_INFO_ALL;
+
+                        ret = dict_set_int32 (output, "info-op", op);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Failed to set info-op in dict");
+                                goto out;
+                        }
+
+                        if (GF_CLI_INFO_CLEAR == op) {
+                                ret = io_stats_clear (this->private);
+                                if (ret)
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "Failed to clear info stats");
+
+                                ret = dict_set_int32 (output, "stats-cleared",
+                                              ret ? 0 : 1);
+                                if (ret)
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "Failed to set stats-cleared"
+                                                " in dict");
+                        }
+                        else {
+                                ret = dict_get_str_boolean (dict, "peek",
+                                                            _gf_false);
+                                if (-1 != ret)
+                                        is_peek = ret;
+
+                                (void) ios_dump_args_init (&args,
+                                                IOS_DUMP_TYPE_DICT, output);
+                                ret = io_stats_dump (this, &args, op, is_peek);
+                        }
+                }
+                break;
+        default:
+                default_notify (this, event, data);
+                break;
+
+        }
+out:
+        return ret;
+}
+
 struct xlator_fops fops = {
         .stat        = io_stats_stat,
         .readlink    = io_stats_readlink,
@@ -1490,6 +2952,9 @@ struct xlator_fops fops = {
         .setxattr    = io_stats_setxattr,
         .getxattr    = io_stats_getxattr,
         .removexattr = io_stats_removexattr,
+        .fsetxattr    = io_stats_fsetxattr,
+        .fgetxattr    = io_stats_fgetxattr,
+        .fremovexattr = io_stats_fremovexattr,
         .opendir     = io_stats_opendir,
         .readdir     = io_stats_readdir,
         .readdirp    = io_stats_readdirp,
@@ -1503,30 +2968,65 @@ struct xlator_fops fops = {
         .finodelk    = io_stats_finodelk,
         .entrylk     = io_stats_entrylk,
         .lookup      = io_stats_lookup,
-        .setdents    = io_stats_setdents,
-        .getdents    = io_stats_getdents,
-        .checksum    = io_stats_checksum,
         .xattrop     = io_stats_xattrop,
         .fxattrop    = io_stats_fxattrop,
         .setattr     = io_stats_setattr,
         .fsetattr    = io_stats_fsetattr,
-};
-
-struct xlator_mops mops = {
-        .stats    = io_stats_stats,
+	.fallocate   = io_stats_fallocate,
+	.discard     = io_stats_discard,
+        .zerofill    = io_stats_zerofill,
 };
 
 struct xlator_cbks cbks = {
+        .release     = io_stats_release,
+        .releasedir  = io_stats_releasedir,
+        .forget      = io_stats_forget,
 };
 
 struct volume_options options[] = {
-        { .key  = {"include-ops", "include"},
+        { .key  = {"dump-fd-stats"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "off",
+          .description = "If on stats related to file-operations would be "
+                         "tracked inside GlusterFS data-structures."
+        },
+        { .key  = { "latency-measurement" },
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "off",
+          .description = "If on stats related to the latency of each operation "
+                         "would be tracked inside GlusterFS data-structures. "
+        },
+        { .key  = {"count-fop-hits"},
+          .type = GF_OPTION_TYPE_BOOL,
+        },
+        { .key = {"log-level"},
           .type = GF_OPTION_TYPE_STR,
-          /*.value = { ""} */
+          .value = { "DEBUG", "WARNING", "ERROR", "INFO",
+                     "CRITICAL", "NONE", "TRACE"}
         },
-        { .key  = {"exclude-ops", "exclude"},
-          .type = GF_OPTION_TYPE_STR
-          /*.value = { ""} */
+
+        /* These are synthetic entries to assist validation of CLI's  *
+         *  volume set  command                                       */
+        { .key = {"client-log-level"},
+          .type = GF_OPTION_TYPE_STR,
+          .default_value = "INFO",
+          .description = "Changes the log-level of the clients",
+          .value = { "DEBUG", "WARNING", "ERROR", "INFO",
+                     "CRITICAL", "NONE", "TRACE"}
+        },
+        { .key = {"sys-log-level"},
+          .type = GF_OPTION_TYPE_STR,
+          .default_value = "CRITICAL",
+          .description = "Gluster's syslog log-level",
+          .value = { "WARNING", "ERROR", "INFO", "CRITICAL"}
+        },
+        { .key = {"brick-log-level"},
+          .type = GF_OPTION_TYPE_STR,
+          .default_value = "INFO",
+          .description = "Changes the log-level of the bricks",
+          .value = { "DEBUG", "WARNING", "ERROR", "INFO",
+                     "CRITICAL", "NONE", "TRACE"}
         },
         { .key  = {NULL} },
+
 };
diff --git a/xlators/debug/trace/src/Makefile.am b/xlators/debug/trace/src/Makefile.am
index 0f1679a04..7b2597b4d 100644
--- a/xlators/debug/trace/src/Makefile.am
+++ b/xlators/debug/trace/src/Makefile.am
@@ -2,13 +2,15 @@
 xlator_LTLIBRARIES = trace.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/debug
 
-trace_la_LDFLAGS = -module -avoidversion
+trace_la_LDFLAGS = -module -avoid-version
 
 trace_la_SOURCES = trace.c
 trace_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+noinst_HEADERS = trace.h trace-mem-types.h
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/debug/trace/src/trace-mem-types.h b/xlators/debug/trace/src/trace-mem-types.h
new file mode 100644
index 000000000..9fa7d97c2
--- /dev/null
+++ b/xlators/debug/trace/src/trace-mem-types.h
@@ -0,0 +1,21 @@
+/*
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef __TRACE_MEM_TYPES_H__
+#define __TRACE_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_trace_mem_types_ {
+        gf_trace_mt_trace_conf_t = gf_common_mt_end + 1,
+        gf_trace_mt_end
+};
+#endif
diff --git a/xlators/debug/trace/src/trace.c b/xlators/debug/trace/src/trace.c
index b4fcf29e1..c9d839356 100644
--- a/xlators/debug/trace/src/trace.c
+++ b/xlators/debug/trace/src/trace.c
@@ -1,2546 +1,3274 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
+  Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
 */
 
-#ifndef _CONFIG_H
-#define _CONFIG_H
-#include "config.h"
-#endif
+#include "trace.h"
+#include "trace-mem-types.h"
 
 /**
  * xlators/debug/trace :
- *    This translator logs all the arguments to the fops/mops and also 
- *    their _cbk functions, which later passes the call to next layer. 
+ *    This translator logs all the arguments to the fops/mops and also
+ *    their _cbk functions, which later passes the call to next layer.
  *    Very helpful translator for debugging.
  */
 
-#include <time.h>
-#include <errno.h>
-#include "glusterfs.h"
-#include "xlator.h"
-#include "common-utils.h"
-
-#define ERR_EINVAL_NORETURN(cond)                \
-do                                               \
-  {						 \
-    if ((cond))					 \
-      {						 \
-	gf_log ("ERROR", GF_LOG_ERROR,   	 \
-		"%s: %s: (%s) is true", 	 \
-		__FILE__, __FUNCTION__, #cond);	 \
-      }                                          \
-  } while (0)
-
-typedef struct trace_private {
-	int32_t debug_flag;
-} trace_private_t;
-
-struct {
-	char *name;
-	int enabled;
-} trace_fop_names[GF_FOP_MAXVALUE];
-
-static char *
-trace_stat_to_str (struct stat *stbuf)
-{
-        char *statstr = NULL;
-	char atime_buf[256] = {0,};
-        char mtime_buf[256] = {0,};
-        char ctime_buf[256] = {0,};
-        int  asprint_ret_value = 0;
-
-        strftime (atime_buf, 256, "[%b %d %H:%M:%S]",
-                  localtime (&stbuf->st_atime));
-        strftime (mtime_buf, 256, "[%b %d %H:%M:%S]",
-                  localtime (&stbuf->st_mtime));
-        strftime (ctime_buf, 256, "[%b %d %H:%M:%S]",
-                  localtime (&stbuf->st_ctime));
-
-        asprint_ret_value = asprintf (&statstr,
-                                      "st_ino=%"PRIu64", st_mode=%o, st_nlink=%"GF_PRI_NLINK", "
-                                      "st_uid=%d, st_gid=%d, st_size=%"PRId64", st_blocks=%"PRId64
-                                      ", st_atime=%s, st_mtime=%s, st_ctime=%s",
-                                      stbuf->st_ino, stbuf->st_mode, stbuf->st_nlink, stbuf->st_uid,
-                                      stbuf->st_gid, stbuf->st_size, stbuf->st_blocks, atime_buf,
-                                      mtime_buf, ctime_buf);
-
-        if (asprint_ret_value < 0)
-                statstr = NULL;
-
-        return statstr;
-}
-
-int32_t 
-trace_create_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-		  fd_t *fd,
-		  inode_t *inode,
-		  struct stat *buf,
-                  struct stat *preparent,
-                  struct stat *postparent)
-{
-        char *statstr = NULL, *preparentstr = NULL, *postparentstr = NULL;
-
-	ERR_EINVAL_NORETURN (!this);
-
-	if (trace_fop_names[GF_FOP_CREATE].enabled) {
-		if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        preparentstr = trace_stat_to_str (preparent);
-                        postparentstr = trace_stat_to_str (postparent);
-
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, fd=%p, ino=%"PRIu64" "
-				"*stbuf {%s}, *preparent {%s}, *postparent = "
-                                "{%s})",
-				frame->root->unique, op_ret, fd, inode->ino,
-                                statstr, preparentstr, postparentstr);
-
-                        if (statstr)
-                                FREE (statstr);
-                        if (preparentstr)
-                                FREE (preparentstr);
-                        if (postparentstr)
-                                FREE (postparentstr);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}    
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, fd, inode, buf,
-                      preparent, postparent);
-	return 0;
-}
-
-int32_t 
-trace_open_cbk (call_frame_t *frame,
-		void *cookie,
-		xlator_t *this,
-		int32_t op_ret,
-		int32_t op_errno,
-		fd_t *fd)
-{
-	ERR_EINVAL_NORETURN (!this);
-
-	if (trace_fop_names[GF_FOP_OPEN].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (op_ret=%d, op_errno=%d, *fd=%p)",
-			frame->root->unique, op_ret, op_errno, fd);
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, fd);
-	return 0;
-}
-
-int32_t 
-trace_stat_cbk (call_frame_t *frame,
-		void *cookie,
-		xlator_t *this,
-		int32_t op_ret,
-		int32_t op_errno,
-		struct stat *buf)
-{
-	char atime_buf[256], mtime_buf[256], ctime_buf[256];
-	ERR_EINVAL_NORETURN (!this);
-  
-	if (trace_fop_names[GF_FOP_STAT].enabled) {
-
-		if (op_ret >= 0) {
-			strftime (atime_buf, 256, "[%b %d %H:%M:%S]", localtime (&buf->st_atime));
-			strftime (mtime_buf, 256, "[%b %d %H:%M:%S]", localtime (&buf->st_mtime));
-			strftime (ctime_buf, 256, "[%b %d %H:%M:%S]", localtime (&buf->st_ctime));
-
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": (op_ret=%d, buf {st_dev=%"GF_PRI_DEV", "
-				"st_ino=%"PRIu64", st_mode=%o, st_nlink=%"GF_PRI_NLINK", "
-				"st_uid=%d, st_gid=%d, st_rdev=%"GF_PRI_DEV", st_size=%"PRId64
-				", st_blksize=%"GF_PRI_BLKSIZE", st_blocks=%"PRId64", "
-				"st_atime=%s, st_mtime=%s, st_ctime=%s})",
-				frame->root->unique, op_ret, buf->st_dev, buf->st_ino, 
-				buf->st_mode, buf->st_nlink, buf->st_uid, buf->st_gid, 
-				buf->st_rdev, buf->st_size, buf->st_blksize, 
-				buf->st_blocks, atime_buf, mtime_buf, ctime_buf);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}    
-	}    
-
-	STACK_UNWIND (frame, op_ret, op_errno, buf);
-	return 0;
-}
-
-int32_t 
-trace_readv_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-		 struct iovec *vector,
-		 int32_t count,
-		 struct stat *buf,
-                 struct iobref *iobref)
-{
-	char atime_buf[256], mtime_buf[256], ctime_buf[256];
-	ERR_EINVAL_NORETURN (!this);
-
-	if (trace_fop_names[GF_FOP_READ].enabled) {
-
-		if (op_ret >= 0) {
-			strftime (atime_buf, 256, "[%b %d %H:%M:%S]", localtime (&buf->st_atime));
-			strftime (mtime_buf, 256, "[%b %d %H:%M:%S]", localtime (&buf->st_mtime));
-			strftime (ctime_buf, 256, "[%b %d %H:%M:%S]", localtime (&buf->st_ctime));
-
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": (op_ret=%d, *buf {st_dev=%"GF_PRI_DEV", "
-				"st_ino=%"PRIu64", st_mode=%o, st_nlink=%"GF_PRI_NLINK", "
-				"st_uid=%d, st_gid=%d, st_rdev=%"GF_PRI_DEV", "
-				"st_size=%"PRId64", st_blksize=%"GF_PRI_BLKSIZE", "
-				"st_blocks=%"PRId64", st_atime=%s, st_mtime=%s, st_ctime=%s})",
-				frame->root->unique, op_ret, buf->st_dev, buf->st_ino, 
-				buf->st_mode, buf->st_nlink, buf->st_uid, buf->st_gid, 
-				buf->st_rdev, buf->st_size, buf->st_blksize, buf->st_blocks, 
-				atime_buf, mtime_buf, ctime_buf);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}    
-	}
-  
-	STACK_UNWIND (frame, op_ret, op_errno, vector, count, buf, iobref);
-	return 0;
-}
-
-int32_t 
-trace_writev_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-                  struct stat *prebuf,
-		  struct stat *postbuf)
-{
-        char *preopstr = NULL, *postopstr = NULL;
-	ERR_EINVAL_NORETURN (!this);
-
-	if (trace_fop_names[GF_FOP_WRITE].enabled) {
-		if (op_ret >= 0) {
-                        preopstr = trace_stat_to_str (prebuf);
-                        preopstr = trace_stat_to_str (postbuf);
-
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, ino = %"PRIu64
-                                ", *prebuf = {%s}, *postbuf = {%s})",
-				frame->root->unique, op_ret, postbuf->st_ino,
-                                preopstr, postopstr);
-
-                        if (preopstr)
-                                FREE (preopstr);
-
-                        if (postopstr)
-                                FREE (postopstr);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}    
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, prebuf, postbuf);
-	return 0;
-}
-
-int32_t 
-trace_getdents_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    dir_entry_t *entries,
-		    int32_t count)
-{
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_GETDENTS].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (op_ret=%d, op_errno=%d, count=%d)",
-			frame->root->unique, op_ret, op_errno, count);
-	}
-  
-	STACK_UNWIND (frame, op_ret, op_errno, entries, count);
-	return 0;
-}
-
-int32_t 
-trace_readdir_cbk (call_frame_t *frame,
-		   void *cookie,
-		   xlator_t *this,
-		   int32_t op_ret,
-		   int32_t op_errno,
-		   gf_dirent_t *buf)
-{
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_READDIR].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64" :(op_ret=%d, op_errno=%d)",
-			frame->root->unique, op_ret, op_errno);
-	}
-  
-	STACK_UNWIND (frame, op_ret, op_errno, buf);
-
-	return 0;
+int
+dump_history_trace (circular_buffer_t *cb, void *data)
+{
+        char    *string      = NULL;
+        struct  tm      *tm  = NULL;
+        char    timestr[256] = {0,};
+
+        string = (char *)cb->data;
+        tm    = localtime (&cb->tv.tv_sec);
+
+        /* Since we are continuing with adding entries to the buffer even when
+           gettimeofday () fails, it's safe to check tm and then dump the time
+           at which the entry was added to the buffer */
+        if (tm) {
+                strftime (timestr, 256, "%Y-%m-%d %H:%M:%S", tm);
+                snprintf (timestr + strlen (timestr), 256 - strlen (timestr),
+                          ".%"GF_PRI_SUSECONDS, cb->tv.tv_usec);
+                gf_proc_dump_write ("TIME", "%s", timestr);
+        }
+
+        gf_proc_dump_write ("FOP", "%s\n", string);
+
+        return 0;
 }
 
-int32_t
+int
+trace_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, fd_t *fd,
+                  inode_t *inode, struct iatt *buf,
+                  struct iatt *preparent, struct iatt *postparent,
+                  dict_t *xdata)
+{
+        char          statstr[4096]       = {0, };
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_CREATE].enabled) {
+                char  string[4096] = {0,};
+                if (op_ret >= 0) {
+                        trace_stat_to_str (buf, statstr);
+                        trace_stat_to_str (preparent, preparentstr);
+                        trace_stat_to_str (postparent, postparentstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s (op_ret=%d, fd=%p"
+                                  "*stbuf {%s}, *preparent {%s}, "
+                                  "*postparent = {%s})",
+                                  frame->root->unique,
+                                  uuid_utoa (inode->gfid), op_ret, fd,
+                                  statstr, preparentstr, postparentstr);
+
+                        /* for 'release' log */
+                        fd_ctx_set (fd, this, 0);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, op_errno=%d)",
+                                  frame->root->unique, op_ret,
+                                  op_errno);
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (create, frame, op_ret, op_errno, fd, inode, buf,
+                            preparent, postparent, xdata);
+        return 0;
+}
+
+int
+trace_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+        trace_conf_t      *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_OPEN].enabled) {
+                char     string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d, "
+                          "*fd=%p", frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno,
+                          fd);
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        /* for 'release' log */
+        if (op_ret >= 0)
+                fd_ctx_set (fd, this, 0);
+
+        TRACE_STACK_UNWIND (open, frame, op_ret, op_errno, fd, xdata);
+        return 0;
+}
+
+int
+trace_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                dict_t *xdata)
+{
+        char          statstr[4096] = {0, };
+        trace_conf_t  *conf         = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_STAT].enabled) {
+                char string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (buf, statstr);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d buf=%s",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  statstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d)",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (stat, frame, op_ret, op_errno, buf, xdata);
+        return 0;
+}
+
+int
+trace_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, struct iovec *vector,
+                 int32_t count, struct iatt *buf, struct iobref *iobref,
+                 dict_t *xdata)
+{
+        char          statstr[4096] = {0, };
+        trace_conf_t  *conf         = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_READ].enabled) {
+                char  string[4096] = {0,};
+                if (op_ret >= 0) {
+                        trace_stat_to_str (buf, statstr);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d buf=%s",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  statstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d)",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (readv, frame, op_ret, op_errno, vector, count,
+                            buf, iobref, xdata);
+        return 0;
+}
+
+int
+trace_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno,
+                  struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+{
+        char         preopstr[4096]  = {0, };
+        char         postopstr[4096] = {0, };
+        trace_conf_t *conf           = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_WRITE].enabled) {
+                char  string[4096] = {0,};
+                if (op_ret >= 0) {
+                        trace_stat_to_str (prebuf, preopstr);
+                        trace_stat_to_str (postbuf, postopstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, "
+                                  "*prebuf = {%s}, *postbuf = {%s})",
+                                  frame->root->unique, op_ret,
+                                  preopstr, postopstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d", frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (writev, frame, op_ret, op_errno, prebuf, postbuf,
+                            xdata);
+        return 0;
+}
+
+int
+trace_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, gf_dirent_t *buf,
+                   dict_t *xdata)
+{
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_READDIR].enabled) {
+                char    string[4096]  = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64" : gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique, uuid_utoa (frame->local),
+                          op_ret, op_errno);
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (readdir, frame, op_ret, op_errno, buf, xdata);
+
+        return 0;
+}
+
+int
 trace_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                    int32_t op_ret, int32_t op_errno, gf_dirent_t *buf)
+                    int32_t op_ret, int32_t op_errno, gf_dirent_t *buf,
+                    dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_READDIRP].enabled) {
+                char        string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64" : gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique, uuid_utoa (frame->local),
+                          op_ret, op_errno);
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        TRACE_STACK_UNWIND (readdirp, frame, op_ret, op_errno, buf, xdata);
+        return 0;
+}
+
+int
+trace_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno,
+                 struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+{
+        char          preopstr[4096]  = {0, };
+        char          postopstr[4096] = {0, };
+        trace_conf_t  *conf           = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FSYNC].enabled) {
+                char  string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (prebuf, preopstr);
+                        trace_stat_to_str (postbuf, postopstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, "
+                                  "*prebuf = {%s}, *postbuf = {%s}",
+                                  frame->root->unique, op_ret,
+                                  preopstr, postopstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d", frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
+
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (fsync, frame, op_ret, op_errno, prebuf, postbuf,
+                            xdata);
+
+        return 0;
+}
+
+int
+trace_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno,
+                   struct iatt *statpre, struct iatt *statpost, dict_t *xdata)
+{
+        char          preopstr[4096]  = {0, };
+        char          postopstr[4096] = {0, };
+        trace_conf_t  *conf           = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_SETATTR].enabled) {
+                char  string[4096]  = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (statpre, preopstr);
+                        trace_stat_to_str (statpost, postopstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, "
+                                  "*prebuf = {%s}, *postbuf = {%s})",
+                                  frame->root->unique, op_ret,
+                                  preopstr, postopstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d)", frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (setattr, frame, op_ret, op_errno, statpre,
+                            statpost, xdata);
+        return 0;
+}
+
+int
+trace_fsetattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno,
+                    struct iatt *statpre, struct iatt *statpost, dict_t *xdata)
+{
+        char          preopstr[4096]  = {0, };
+        char          postopstr[4096] = {0, };
+        trace_conf_t  *conf           = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FSETATTR].enabled) {
+                char  string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (statpre, preopstr);
+                        trace_stat_to_str (statpost, postopstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, "
+                                  "*prebuf = {%s}, *postbuf = {%s})",
+                                  frame->root->unique, op_ret,
+                                  preopstr, postopstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, op_errno=%d)",
+                                  frame->root->unique, uuid_utoa (frame->local),
+                                  op_ret, op_errno);
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (fsetattr, frame, op_ret, op_errno,
+                            statpre, statpost, xdata);
+        return 0;
+}
+
+int
+trace_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno,
+                  struct iatt *preparent, struct iatt *postparent,
+                  dict_t *xdata)
+{
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_UNLINK].enabled) {
+                char string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (preparent, preparentstr);
+                        trace_stat_to_str (postparent, postparentstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  " *preparent = {%s}, "
+                                  "*postparent = {%s})",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local),
+                                  op_ret, preparentstr,
+                                  postparentstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d)",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (unlink, frame, op_ret, op_errno,
+                            preparent, postparent, xdata);
+        return 0;
+}
+
+int
+trace_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                  struct iatt *preoldparent, struct iatt *postoldparent,
+                  struct iatt *prenewparent, struct iatt *postnewparent,
+                  dict_t *xdata)
+{
+        char           statstr[4096]          = {0, };
+        char           preoldparentstr[4096]  = {0, };
+        char           postoldparentstr[4096] = {0, };
+        char           prenewparentstr[4096]  = {0, };
+        char           postnewparentstr[4096] = {0, };
+        trace_conf_t   *conf                  = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_RENAME].enabled) {
+                char  string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (buf, statstr);
+                        trace_stat_to_str (preoldparent, preoldparentstr);
+                        trace_stat_to_str (postoldparent, postoldparentstr);
+                        trace_stat_to_str (prenewparent, prenewparentstr);
+                        trace_stat_to_str (postnewparent, postnewparentstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, "
+                                  "*stbuf = {%s}, *preoldparent = {%s},"
+                                  " *postoldparent = {%s}"
+                                  " *prenewparent = {%s}, "
+                                  "*postnewparent = {%s})",
+                                  frame->root->unique, op_ret, statstr,
+                                  preoldparentstr, postoldparentstr,
+                                  prenewparentstr, postnewparentstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d", frame->root->unique,
+                                  uuid_utoa (frame->local),
+                                  op_ret, op_errno);
+
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (rename, frame, op_ret, op_errno, buf,
+                            preoldparent, postoldparent,
+                            prenewparent, postnewparent, xdata);
+        return 0;
+}
+
+int
+trace_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno,
+                    const char *buf, struct iatt *stbuf, dict_t *xdata)
+{
+        char          statstr[4096] = {0, };
+        trace_conf_t  *conf         = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_READLINK].enabled) {
+                char string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (stbuf, statstr);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, op_errno=%d,"
+                                  "buf=%s, stbuf = { %s })",
+                                  frame->root->unique, op_ret, op_errno,
+                                  buf, statstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
+                }
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (readlink, frame, op_ret, op_errno, buf, stbuf,
+                            xdata);
+        return 0;
+}
+
+int
+trace_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno,
+                  inode_t *inode, struct iatt *buf,
+                  dict_t *xdata, struct iatt *postparent)
+{
+        char          statstr[4096]       = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_LOOKUP].enabled) {
+                char  string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (buf, statstr);
+                        trace_stat_to_str (postparent, postparentstr);
+                        /* print buf->ia_gfid instead of inode->gfid,
+                         * since if the inode is not yet linked to the
+                         * inode table (fresh lookup) then null gfid
+                         * will be printed.
+                         */
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s (op_ret=%d "
+                                  "*buf {%s}, *postparent {%s}",
+                                  frame->root->unique,
+                                  uuid_utoa (buf->ia_gfid),
+                                  op_ret, statstr, postparentstr);
+
+                        /* For 'forget' */
+                        inode_ctx_put (inode, this, 0);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d)",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, buf,
+                            xdata, postparent);
+        return 0;
+}
+
+int
+trace_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno,
+                   inode_t *inode, struct iatt *buf,
+                   struct iatt *preparent, struct iatt *postparent,
+                   dict_t *xdata)
+{
+        char          statstr[4096]       = {0, };
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_SYMLINK].enabled) {
+                char  string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (buf, statstr);
+                        trace_stat_to_str (preparent, preparentstr);
+                        trace_stat_to_str (postparent, postparentstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s (op_ret=%d "
+                                  "*stbuf = {%s}, *preparent = {%s}, "
+                                  "*postparent = {%s})",
+                                  frame->root->unique,
+                                  uuid_utoa (inode->gfid),
+                                  op_ret, statstr, preparentstr,
+                                  postparentstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": op_ret=%d, op_errno=%d",
+                                  frame->root->unique, op_ret,
+                                  op_errno);
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (symlink, frame, op_ret, op_errno, inode, buf,
+                            preparent, postparent, xdata);
+        return 0;
+}
+
+int
+trace_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno,
+                 inode_t *inode, struct iatt *buf,
+                 struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
+{
+        char          statstr[4096]       = {0, };
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        char string[4096]  = {0,};
+        if (trace_fop_names[GF_FOP_MKNOD].enabled) {
+                if (op_ret == 0) {
+                        trace_stat_to_str (buf, statstr);
+                        trace_stat_to_str (preparent, preparentstr);
+                        trace_stat_to_str (postparent, postparentstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s (op_ret=%d "
+                                  "*stbuf = {%s}, *preparent = {%s}, "
+                                  "*postparent = {%s})",
+                                  frame->root->unique,
+                                  uuid_utoa (inode->gfid),
+                                  op_ret, statstr, preparentstr,
+                                  postparentstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, op_errno=%d)",
+                                  frame->root->unique, op_ret,
+                                  op_errno);
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (mknod, frame, op_ret, op_errno, inode, buf,
+                            preparent, postparent, xdata);
+        return 0;
+}
+
+int
+trace_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno,
+                 inode_t *inode, struct iatt *buf,
+                 struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
+{
+        char          statstr[4096]       = {0, };
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_MKDIR].enabled) {
+                char  string[4096]  = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (buf, statstr);
+                        trace_stat_to_str (preparent, preparentstr);
+                        trace_stat_to_str (postparent, postparentstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s (op_ret=%d "
+                                  ", *stbuf = {%s}, *prebuf = {%s}, "
+                                  "*postbuf = {%s} )",
+                                  frame->root->unique,
+                                  uuid_utoa (inode->gfid),
+                                  op_ret, statstr, preparentstr,
+                                  postparentstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, op_errno=%d)",
+                                  frame->root->unique, op_ret,
+                                  op_errno);
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (mkdir, frame, op_ret, op_errno, inode, buf,
+                            preparent, postparent, xdata);
+        return 0;
+}
+
+int
+trace_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno,
+                inode_t *inode, struct iatt *buf,
+                struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
+{
+        char          statstr[4096]       = {0, };
+        char          preparentstr[4096]  = {0, };
+        char          postparentstr[4096] = {0, };
+        trace_conf_t  *conf               = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        char  string[4096]  = {0,};
+        if (trace_fop_names[GF_FOP_LINK].enabled) {
+                if (op_ret == 0) {
+                        trace_stat_to_str (buf, statstr);
+                        trace_stat_to_str (preparent, preparentstr);
+                        trace_stat_to_str (postparent, postparentstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, "
+                                  "*stbuf = {%s},  *prebuf = {%s},"
+                                  " *postbuf = {%s})",
+                                  frame->root->unique, op_ret,
+                                  statstr, preparentstr, postparentstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local),
+                                  op_ret, op_errno);
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (link, frame, op_ret, op_errno, inode, buf,
+                            preparent, postparent, xdata);
+        return 0;
+}
+
+int
+trace_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        char      string[4096] = {0,};
+        if (trace_fop_names[GF_FOP_FLUSH].enabled) {
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique, uuid_utoa (frame->local),
+                          op_ret, op_errno);
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (flush, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int
+trace_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        char    string[4096] = {0,};
+        if (trace_fop_names[GF_FOP_OPENDIR].enabled) {
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d,"
+                          " fd=%p",
+                          frame->root->unique, uuid_utoa (frame->local),
+                          op_ret, op_errno, fd);
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        /* for 'releasedir' log */
+        if (op_ret >= 0)
+                fd_ctx_set (fd, this, 0);
+
+        TRACE_STACK_UNWIND (opendir, frame, op_ret, op_errno, fd, xdata);
+        return 0;
+}
+
+int
+trace_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno,
+                 struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
+{
+        char           preparentstr[4096]  = {0, };
+        char           postparentstr[4096] = {0, };
+        trace_conf_t   *conf               = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_RMDIR].enabled) {
+                char  string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (preparent, preparentstr);
+                        trace_stat_to_str (postparent, postparentstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "*prebuf={%s},  *postbuf={%s}",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local),
+                                  op_ret, preparentstr, postparentstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d", frame->root->unique,
+                                  uuid_utoa (frame->local),
+                                  op_ret, op_errno);
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (rmdir, frame, op_ret, op_errno,
+                            preparent, postparent, xdata);
+        return 0;
+}
+
+int
+trace_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno,
+                    struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+{
+        char           preopstr[4096]  = {0, };
+        char           postopstr[4096] = {0, };
+        trace_conf_t   *conf           = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_TRUNCATE].enabled) {
+                char   string[4096] = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (prebuf, preopstr);
+                        trace_stat_to_str (postbuf, postopstr);
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, "
+                                  "*prebuf = {%s}, *postbuf = {%s} )",
+                                  frame->root->unique, op_ret,
+                                  preopstr, postopstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d", frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (truncate, frame, op_ret, op_errno, prebuf,
+                            postbuf, xdata);
+        return 0;
+}
+
+int
+trace_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, struct statvfs *buf,
+                  dict_t *xdata)
+{
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_STATFS].enabled) {
+                char   string[4096] = {0,};
+                if (op_ret == 0) {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": ({f_bsize=%lu, "
+                                  "f_frsize=%lu, "
+                                  "f_blocks=%"GF_PRI_FSBLK
+                                  ", f_bfree=%"GF_PRI_FSBLK", "
+                                  "f_bavail=%"GF_PRI_FSBLK", "
+                                  "f_files=%"GF_PRI_FSBLK", "
+                                  "f_ffree=%"GF_PRI_FSBLK", "
+                                  "f_favail=%"GF_PRI_FSBLK", "
+                                  "f_fsid=%lu, f_flag=%lu, "
+                                  "f_namemax=%lu}) => ret=%d",
+                                  frame->root->unique, buf->f_bsize,
+                                  buf->f_frsize, buf->f_blocks,
+                                  buf->f_bfree, buf->f_bavail,
+                                  buf->f_files, buf->f_ffree,
+                                  buf->f_favail, buf->f_fsid,
+                                  buf->f_flag, buf->f_namemax, op_ret);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": (op_ret=%d, "
+                                  "op_errno=%d)",
+                                  frame->root->unique, op_ret,
+                                  op_errno);
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (statfs, frame, op_ret, op_errno, buf, xdata);
+        return 0;
+}
+
+int
+trace_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-	ERR_EINVAL_NORETURN (!this );
+        trace_conf_t     *conf = NULL;
 
-	if (trace_fop_names[GF_FOP_READDIRP].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL,
-			"%"PRId64" :(op_ret=%d, op_errno=%d)",
-			frame->root->unique, op_ret, op_errno);
-	}
+        conf = this->private;
 
-	STACK_UNWIND (frame, op_ret, op_errno, buf);
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_SETXATTR].enabled) {
+                char      string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret,
+                          op_errno);
 
-	return 0;
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
+int
+trace_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *dict,
+                    dict_t *xdata)
+{
+        trace_conf_t    *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_GETXATTR].enabled) {
+                char       string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d,"
+                          " dict=%p", frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno,
+                          dict);
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (getxattr, frame, op_ret, op_errno, dict, xdata);
+
+        return 0;
+}
 
-int32_t 
-trace_fsync_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-                 struct stat *prebuf,
-                 struct stat *postbuf)
+int
+trace_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *xdata)
 {
-        char *preopstr = NULL, *postopstr = NULL;
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
 
-	ERR_EINVAL_NORETURN (!this );
+        if (!conf->log_file && !conf->log_history)
+                goto out;
+        if (trace_fop_names[GF_FOP_FSETXATTR].enabled) {
+                char    string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
 
-	if (trace_fop_names[GF_FOP_FSYNC].enabled) {
-		if (op_ret >= 0) {
-                        preopstr = trace_stat_to_str (prebuf);
-                        preopstr = trace_stat_to_str (postbuf);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
 
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, ino = %"PRIu64
-                                ", *prebuf = {%s}, *postbuf = {%s}",
-				frame->root->unique, op_ret, postbuf->st_ino,
-                                preopstr, postopstr);
+int
+trace_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *dict,
+                     dict_t *xdata)
+{
+        trace_conf_t    *conf = NULL;
 
-                        if (preopstr)
-                                FREE (preopstr);
+        conf = this->private;
 
-                        if (postopstr)
-                                FREE (postopstr);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}
-	}
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FGETXATTR].enabled) {
+                char      string[4096]  = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d,"
+                          " dict=%p", frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno,
+                          dict);
 
-	STACK_UNWIND (frame, op_ret, op_errno, prebuf, postbuf);
-	return 0;
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (fgetxattr, frame, op_ret, op_errno, dict, xdata);
+
+        return 0;
 }
 
-int32_t
-trace_setattr_cbk (call_frame_t *frame,
-                   void *cookie,
-                   xlator_t *this,
-                   int32_t op_ret,
-                   int32_t op_errno,
-                   struct stat *statpre,
-                   struct stat *statpost)
-{
-	char atime_pre[256] = {0,};
-        char mtime_pre[256] = {0,};
-        char ctime_pre[256] = {0,};
-	char atime_post[256] = {0,};
-        char mtime_post[256] = {0,};
-        char ctime_post[256] = {0,};
-
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_SETATTR].enabled) {
-		if (op_ret >= 0) {
-			strftime (atime_pre, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&statpre->st_atime));
-			strftime (mtime_pre, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&statpre->st_mtime));
-			strftime (ctime_pre, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&statpre->st_ctime));
-
-			strftime (atime_post, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&statpost->st_atime));
-			strftime (mtime_post, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&statpost->st_mtime));
-			strftime (ctime_post, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&statpost->st_ctime));
-    
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, *statpre "
-                                "{st_ino=%"PRIu64", st_mode=%o, st_uid=%d, "
-                                "st_gid=%d, st_atime=%s, st_mtime=%s, "
-                                "st_ctime=%s}, *statpost {st_ino=%"PRIu64", "
-                                "st_mode=%o, st_uid=%d, st_gid=%d, st_atime=%s,"
-                                " st_mtime=%s, st_ctime=%s})",
-				frame->root->unique, op_ret, statpre->st_ino,
-                                statpre->st_mode, statpre->st_uid,
-                                statpre->st_gid, atime_pre, mtime_pre,
-                                ctime_pre, statpost->st_ino, statpost->st_mode,
-                                statpost->st_uid, statpost->st_gid, atime_post,
-                                mtime_post, ctime_post);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}    
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, statpre, statpost);
-	return 0;
+int
+trace_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_REMOVEXATTR].enabled) {
+                char       string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (removexattr, frame, op_ret, op_errno, xdata);
+
+        return 0;
 }
 
-int32_t
-trace_fsetattr_cbk (call_frame_t *frame,
-                    void *cookie,
-                    xlator_t *this,
-                    int32_t op_ret,
-                    int32_t op_errno,
-                    struct stat *statpre,
-                    struct stat *statpost)
-{
-	char atime_pre[256] = {0,};
-        char mtime_pre[256] = {0,};
-        char ctime_pre[256] = {0,};
-	char atime_post[256] = {0,};
-        char mtime_post[256] = {0,};
-        char ctime_post[256] = {0,};
-
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_FSETATTR].enabled) {
-		if (op_ret >= 0) {
-			strftime (atime_pre, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&statpre->st_atime));
-			strftime (mtime_pre, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&statpre->st_mtime));
-			strftime (ctime_pre, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&statpre->st_ctime));
-
-			strftime (atime_post, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&statpost->st_atime));
-			strftime (mtime_post, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&statpost->st_mtime));
-			strftime (ctime_post, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&statpost->st_ctime));
-    
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, *statpre "
-                                "{st_ino=%"PRIu64", st_mode=%o, st_uid=%d, "
-                                "st_gid=%d, st_atime=%s, st_mtime=%s, "
-                                "st_ctime=%s}, *statpost {st_ino=%"PRIu64", "
-                                "st_mode=%o, st_uid=%d, st_gid=%d, st_atime=%s,"
-                                " st_mtime=%s, st_ctime=%s})",
-				frame->root->unique, op_ret, statpre->st_ino,
-                                statpre->st_mode, statpre->st_uid,
-                                statpre->st_gid, atime_pre, mtime_pre,
-                                ctime_pre, statpost->st_ino, statpost->st_mode,
-                                statpost->st_uid, statpost->st_gid, atime_post,
-                                mtime_post, ctime_post);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}    
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, statpre, statpost);
-	return 0;
-}
-
-
-int32_t 
-trace_unlink_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-                  struct stat *preparent,
-                  struct stat *postparent)
-{
-        char *preparentstr = NULL, *postparentstr = NULL;
-
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_UNLINK].enabled) {
-		if (op_ret >= 0) {
-                        preparentstr = trace_stat_to_str (preparent);
-                        preparentstr = trace_stat_to_str (postparent);
-
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, *preparent = {%s}, "
-                                "*postparent = {%s})",
-				frame->root->unique, op_ret, preparentstr,
-                                postparentstr);
-
-                        if (preparentstr)
-                                FREE (preparentstr);
-
-                        if (postparentstr)
-                                FREE (postparentstr);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, preparent, postparent);
-	return 0;
-}
-
-int32_t 
-trace_rename_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-		  struct stat *buf,
-                  struct stat *preoldparent,
-                  struct stat *postoldparent,
-                  struct stat *prenewparent,
-                  struct stat *postnewparent)
-{
-        char *statstr = NULL;
-        char *preoldparentstr = NULL, *postoldparentstr = NULL;
-        char *prenewparentstr = NULL, *postnewparentstr = NULL;
-
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_RENAME].enabled) {
-		if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        preoldparentstr = trace_stat_to_str (preoldparent);
-                        postoldparentstr = trace_stat_to_str (postoldparent);
-
-                        prenewparentstr = trace_stat_to_str (prenewparent);
-                        postnewparentstr = trace_stat_to_str (postnewparent);
-
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, *stbuf = {%s}, "
-                                "*preoldparent = {%s}, *postoldparent = {%s}"
-                                " *prenewparent = {%s}, *postnewparent = {%s})",
-				frame->root->unique, op_ret, statstr,
-                                preoldparentstr, postoldparentstr,
-                                prenewparentstr, postnewparentstr);
-
-                        if (preoldparentstr)
-                                FREE (preoldparentstr);
-
-                        if (postoldparentstr)
-                                FREE (postoldparentstr);
-
-                        if (prenewparentstr)
-                                FREE (prenewparentstr);
-
-                        if (postnewparentstr)
-                                FREE (postnewparentstr);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (op_ret=%d, op_errno=%d, buf {st_ino=%"PRIu64"})",
-			frame->root->unique, op_ret, op_errno, 
-			(buf? buf->st_ino : 0));
-	}
-  
-	STACK_UNWIND (frame, op_ret, op_errno, buf,
-                      preoldparent, postoldparent,
-                      prenewparent, postnewparent);
-	return 0;
-}
-
-int32_t 
-trace_readlink_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    const char *buf,
-                    struct stat *stbuf)
-{
-        char *statstr = NULL;
-
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_READLINK].enabled) {  
-		statstr = trace_stat_to_str (stbuf);
-
-                gf_log (this->name, GF_LOG_NORMAL,
-			"%"PRId64": (op_ret=%d, op_errno=%d, buf=%s, "
-                        "stbuf = { %s })",
-			frame->root->unique, op_ret, op_errno, buf, statstr);
-
-                if (statstr)
-                        FREE (statstr);
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, buf);
-	return 0;
-}
-
-int32_t 
-trace_lookup_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-		  inode_t *inode,
-		  struct stat *buf,
-		  dict_t *xattr,
-                  struct stat *postparent)
-{
-        char *statstr = NULL, *postparentstr = NULL;
-
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_LOOKUP].enabled) {
-		if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        postparentstr = trace_stat_to_str (buf);
-
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, ino=%"PRIu64", "
-				"*buf {%s}, *postparent {%s}",
-				frame->root->unique, op_ret, inode->ino,
-                                statstr, postparentstr);
-
-                        if (statstr)
-                                FREE (statstr);
-                        if (postparentstr)
-                                FREE (postparentstr);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}    
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, inode, buf, xattr, postparent);
-	return 0;
-}
-
-int32_t 
-trace_symlink_cbk (call_frame_t *frame,
-		   void *cookie,
-		   xlator_t *this,
-		   int32_t op_ret,
-		   int32_t op_errno,
-		   inode_t *inode,
-                   struct stat *buf,
-                   struct stat *preparent,
-                   struct stat *postparent)
-{
-        char *statstr = NULL;
-        char *preparentstr = NULL, *postparentstr = NULL;
-
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_SYMLINK].enabled) {
-		if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        preparentstr = trace_stat_to_str (preparent);
-                        postparentstr = trace_stat_to_str (postparent);
-
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, ino=%"PRIu64", "
-                                "*stbuf = {%s}, *preparent = {%s}, "
-                                "*postparent = {%s})",
-				frame->root->unique, op_ret, inode->ino,
-                                statstr, preparentstr, postparentstr);
-
-                        if (statstr)
-                                FREE (statstr);
-
-                        if (preparentstr)
-                                FREE (preparentstr);
-
-                        if (postparentstr)
-                                FREE (postparentstr);
-
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}    
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, inode, buf,
-                      preparent, postparent);
-	return 0;
-}
-
-int32_t 
-trace_mknod_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-		 inode_t *inode,
-                 struct stat *buf,
-                 struct stat *preparent,
-                 struct stat *postparent)
-{
-        char *statstr = NULL;
-        char *preparentstr = NULL, *postparentstr = NULL;
-
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_MKNOD].enabled) {  
-		if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        preparentstr = trace_stat_to_str (preparent);
-                        postparentstr = trace_stat_to_str (postparent);
-
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, ino=%"PRIu64", "
-                                "*stbuf = {%s}, *preparent = {%s}, "
-                                "*postparent = {%s})",
-				frame->root->unique, op_ret, inode->ino,
-                                statstr, preparentstr, postparentstr);
-
-                        if (statstr)
-                                FREE (statstr);
-
-                        if (preparentstr)
-                                FREE (preparentstr);
-
-                        if (postparentstr)
-                                FREE (postparentstr);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}    
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, inode, buf,
-                      preparent, postparent);
-	return 0;
-}
-  
-
-int32_t 
-trace_mkdir_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-		 inode_t *inode,
-                 struct stat *buf,
-                 struct stat *preparent,
-                 struct stat *postparent)
-{
-        char *statstr = NULL, *preparentstr = NULL, *postparentstr = NULL;
-
-        ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_MKDIR].enabled) {  
-                if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        preparentstr = trace_stat_to_str (preparent);
-                        preparentstr = trace_stat_to_str (postparent);
+int
+trace_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        trace_conf_t     *conf = NULL;
 
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, ino = %"PRIu64
-                                ", *stbuf = {%s}, *prebuf = {%s}, "
-                                "*postbuf = {%s} )",
-                                frame->root->unique, op_ret, buf->st_ino,
-                                statstr, preparentstr, postparentstr);
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FSYNCDIR].enabled) {
+                char       string[4096]   =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
 
-                        if (statstr)
-                                FREE (statstr);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (fsyncdir, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int
+trace_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_ACCESS].enabled) {
+                char      string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, "
+                          "op_errno=%d)", frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
 
-                        if (preparentstr)
-                                FREE (preparentstr);
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (access, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int
+trace_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno,
+                     struct iatt *prebuf, struct iatt *postbuf, dict_t *xdata)
+{
+        char          prebufstr[4096]  = {0, };
+        char          postbufstr[4096] = {0, };
+        trace_conf_t  *conf            = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FTRUNCATE].enabled) {
+                char  string[4096]  = {0,};
+                if (op_ret == 0) {
+                        trace_stat_to_str (prebuf, prebufstr);
+                        trace_stat_to_str (postbuf, postbufstr);
 
-                        if (postparentstr)
-                                FREE (postparentstr);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": op_ret=%d, "
+                                  "*prebuf = {%s}, *postbuf = {%s} )",
+                                  frame->root->unique, op_ret,
+                                  prebufstr, postbufstr);
                 } else {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, op_errno=%d)",
-                                frame->root->unique, op_ret, op_errno);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d", frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
                 }
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, inode, buf,
-                      preparent, postparent);
-	return 0;
-}
-  
-int32_t 
-trace_link_cbk (call_frame_t *frame,
-		void *cookie,
-		xlator_t *this,
-		int32_t op_ret,
-		int32_t op_errno,
-		inode_t *inode,
-                struct stat *buf,
-                struct stat *preparent,
-                struct stat *postparent)
-{
-        char *statstr = NULL, *preparentstr = NULL, *postparentstr = NULL;
-
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_LINK].enabled) {
-		if (op_ret >= 0) {
-                        statstr = trace_stat_to_str (buf);
-                        preparentstr = trace_stat_to_str (preparent);
-                        preparentstr = trace_stat_to_str (postparent);
-
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, ino = %"PRIu64
-                                ", *stbuf = {%s}, *prebuf = {%s}, "
-                                "*postbuf = {%s})",
-				frame->root->unique, op_ret, buf->st_ino,
-                                statstr, preparentstr, postparentstr);
-
-                        if (statstr)
-                                FREE (statstr);
-
-                        if (preparentstr)
-                                FREE (preparentstr);
-
-                        if (postparentstr)
-                                FREE (postparentstr);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}    
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, inode, buf,
-                      preparent, postparent);
-	return 0;
-}
-
-int32_t 
-trace_flush_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno)
-{
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_FLUSH].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (op_ret=%d, op_errno=%d)",
-			frame->root->unique, op_ret, op_errno);
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-
-int32_t 
-trace_opendir_cbk (call_frame_t *frame,
-		   void *cookie,
-		   xlator_t *this,
-		   int32_t op_ret,
-		   int32_t op_errno,
-		   fd_t *fd)
-{
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_OPENDIR].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (op_ret=%d, op_errno=%d, fd=%p)",
-			frame->root->unique, op_ret, op_errno, fd);
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, fd);
-	return 0;
-}
-
-int32_t 
-trace_rmdir_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-                 struct stat *preparent,
-                 struct stat *postparent)
-{
-        char *preparentstr = NULL, *postparentstr = NULL;
-
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_RMDIR].enabled) {
-		if (op_ret >= 0) {
-                        preparentstr = trace_stat_to_str (preparent);
-                        preparentstr = trace_stat_to_str (postparent);
-
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, *prebuf = {%s}, "
-                                "*postbuf = {%s}",
-				frame->root->unique, op_ret, preparentstr,
-                                postparentstr);
-
-                        if (preparentstr)
-                                FREE (preparentstr);
-
-                        if (postparentstr)
-                                FREE (postparentstr);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno,
-                      preparent, postparent);
-	return 0;
-}
-
-int32_t 
-trace_truncate_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    struct stat *prebuf,
-                    struct stat *postbuf)
-{
-        char *preopstr = NULL, *postopstr = NULL;
-
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_TRUNCATE].enabled) {  
-		if (op_ret >= 0) {
-                        preopstr = trace_stat_to_str (prebuf);
-                        postopstr = trace_stat_to_str (prebuf);
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, *prebuf = {%s}, "
-                                "*postbuf = {%s} )",
-                                frame->root->unique, op_ret, preopstr,
-                                postopstr);
-
-                        if (preopstr)
-                                FREE (preopstr);
-
-                        if (postopstr)
-                                FREE (postopstr);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}    
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, prebuf, postbuf);
-	return 0;
-}
-
-int32_t 
-trace_statfs_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno,
-		  struct statvfs *buf)
-{
-	ERR_EINVAL_NORETURN (!this);
-
-	if (trace_fop_names[GF_FOP_STATFS].enabled) {
-		if (op_ret >= 0) {
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": ({f_bsize=%lu, f_frsize=%lu, f_blocks=%"GF_PRI_FSBLK
-				", f_bfree=%"GF_PRI_FSBLK", f_bavail=%"GF_PRI_FSBLK", "
-				"f_files=%"GF_PRI_FSBLK", f_ffree=%"GF_PRI_FSBLK", f_favail=%"
-				GF_PRI_FSBLK", f_fsid=%lu, f_flag=%lu, f_namemax=%lu}) => ret=%d",
-				frame->root->unique, buf->f_bsize, buf->f_frsize, buf->f_blocks, 
-				buf->f_bfree, buf->f_bavail, buf->f_files, buf->f_ffree, 
-				buf->f_favail, buf->f_fsid, buf->f_flag, buf->f_namemax, op_ret);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}    
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, buf);
-	return 0;
-}
-
-int32_t 
-trace_setxattr_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno)
-{
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_SETXATTR].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (op_ret=%d, op_errno=%d)",
-			frame->root->unique, op_ret, op_errno);
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-int32_t 
-trace_getxattr_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    dict_t *dict)
-{
-	ERR_EINVAL_NORETURN (!this || !dict);
-
-	if (trace_fop_names[GF_FOP_GETXATTR].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (op_ret=%d, op_errno=%d, dict=%p)",
-			frame->root->unique, op_ret, op_errno, dict);
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, dict);
-	return 0;
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (ftruncate, frame, op_ret, op_errno, prebuf, postbuf,
+                            xdata);
+        return 0;
+}
+
+int
+trace_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
+{
+        char          statstr[4096] = {0, };
+        trace_conf_t  *conf         = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FSTAT].enabled) {
+                char string[4096]  = {0.};
+                if (op_ret == 0) {
+                        trace_stat_to_str (buf, statstr);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d "
+                                  "buf=%s", frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  statstr);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d", frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
+                }
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (fstat, frame, op_ret, op_errno, buf, xdata);
+        return 0;
+}
+
+int
+trace_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+              int32_t op_ret, int32_t op_errno, struct gf_flock *lock,
+              dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_LK].enabled) {
+                char      string[4096] = {0,};
+                if (op_ret == 0) {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "{l_type=%d, l_whence=%d, "
+                                  "l_start=%"PRId64", "
+                                  "l_len=%"PRId64", l_pid=%u})",
+                                  frame->root->unique,
+                                  uuid_utoa (frame->local),
+                                  op_ret, lock->l_type, lock->l_whence,
+                                  lock->l_start, lock->l_len,
+                                  lock->l_pid);
+                } else {
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s op_ret=%d, "
+                                  "op_errno=%d)", frame->root->unique,
+                                  uuid_utoa (frame->local), op_ret,
+                                  op_errno);
+                }
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (lk, frame, op_ret, op_errno, lock, xdata);
+        return 0;
+}
+
+int
+trace_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_ENTRYLK].enabled) {
+                char   string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (entrylk, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
-int32_t 
-trace_removexattr_cbk (call_frame_t *frame,
-		       void *cookie,
-		       xlator_t *this,
-		       int32_t op_ret,
-		       int32_t op_errno)
-{
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_REMOVEXATTR].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (op_ret=%d, op_errno=%d)",
-			frame->root->unique, op_ret, op_errno);
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-
-int32_t 
-trace_fsyncdir_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno)
-{
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_FSYNCDIR].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (op_ret=%d, op_errno=%d)",
-			frame->root->unique, op_ret, op_errno);
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-int32_t 
-trace_access_cbk (call_frame_t *frame,
-		  void *cookie,
-		  xlator_t *this,
-		  int32_t op_ret,
-		  int32_t op_errno)
-{
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_ACCESS].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (op_ret=%d, op_errno=%d)",
-			frame->root->unique, op_ret, op_errno);
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-int32_t 
-trace_ftruncate_cbk (call_frame_t *frame,
-		     void *cookie,
-		     xlator_t *this,
-		     int32_t op_ret,
-		     int32_t op_errno,
-		     struct stat *prebuf,
-                     struct stat *postbuf)
-{
-        char *prebufstr = NULL, *postbufstr = NULL;
-
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_FTRUNCATE].enabled) {
-		if (op_ret >= 0) {
-                        prebufstr = trace_stat_to_str (prebuf);
-                        postbufstr = trace_stat_to_str (postbuf);
-
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (op_ret=%d, *prebuf = {%s}, "
-                                "*postbuf = {%s} )",
-                                frame->root->unique, op_ret,
-                                prebufstr, postbufstr);
-
-                        if (prebufstr)
-                                FREE (prebufstr);
-
-                        if (postbufstr)
-                                FREE (postbufstr);
-
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}    
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, prebuf, postbuf);
-	return 0;
-}
-
-int32_t 
-trace_fstat_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-		 struct stat *buf)
-{
-	char atime_buf[256], mtime_buf[256], ctime_buf[256];
-	ERR_EINVAL_NORETURN (!this );
-  
-	if (trace_fop_names[GF_FOP_FSTAT].enabled) {
-		if (op_ret >= 0) {
-			strftime (atime_buf, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&buf->st_atime));
-			strftime (mtime_buf, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&buf->st_mtime));
-			strftime (ctime_buf, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&buf->st_ctime));
-    
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": (op_ret=%d, *buf {st_dev=%"GF_PRI_DEV", "
-				"st_ino=%"PRIu64", st_mode=%o, st_nlink=%"GF_PRI_NLINK", "
-				"st_uid=%d, st_gid=%d, st_rdev=%"GF_PRI_DEV", st_size=%"PRId64", "
-				"st_blksize=%"GF_PRI_BLKSIZE", st_blocks=%"PRId64", st_atime=%s, "
-				"st_mtime=%s, st_ctime=%s})",
-				frame->root->unique, op_ret, buf->st_dev, buf->st_ino, 
-				buf->st_mode, buf->st_nlink, buf->st_uid, buf->st_gid, 
-				buf->st_rdev, buf->st_size, buf->st_blksize, 
-				buf->st_blocks, atime_buf, mtime_buf, ctime_buf);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}    
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, buf);
-	return 0;
-}
-
-int32_t 
-trace_lk_cbk (call_frame_t *frame,
-	      void *cookie,
-	      xlator_t *this,
-	      int32_t op_ret,
-	      int32_t op_errno,
-	      struct flock *lock)
-{
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_LK].enabled) {
-		if (op_ret >= 0) {
-			gf_log (this->name, GF_LOG_NORMAL,
-				"%"PRId64": (op_ret=%d, {l_type=%d, l_whence=%d, "
-				"l_start=%"PRId64", l_len=%"PRId64", l_pid=%u})",
-				frame->root->unique, op_ret, lock->l_type, lock->l_whence, 
-				lock->l_start, lock->l_len, lock->l_pid);
-		} else {
-			gf_log (this->name, GF_LOG_NORMAL, 
-				"%"PRId64": (op_ret=%d, op_errno=%d)",
-				frame->root->unique, op_ret, op_errno);
-		}    
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, lock);
-	return 0;
-}
-
-
-int32_t 
-trace_setdents_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno)
-{
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_SETDENTS].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL,
-			"%"PRId64": op_ret=%d, op_errno=%d",
-			frame->root->unique, op_ret, op_errno);
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-int32_t 
-trace_entrylk_cbk (call_frame_t *frame,
-		     void *cookie,
-		     xlator_t *this,
-		     int32_t op_ret,
-		     int32_t op_errno)
-{
-	ERR_EINVAL_NORETURN (!this );
-
-	if (trace_fop_names[GF_FOP_ENTRYLK].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL,
-			"%"PRId64": op_ret=%d, op_errno=%d",
-			frame->root->unique, op_ret, op_errno);
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
-}
-
-
-int32_t 
-trace_xattrop_cbk (call_frame_t *frame,
-		   void *cookie,
-		   xlator_t *this,
-		   int32_t op_ret,
-		   int32_t op_errno,
-		   dict_t *dict)
-{
-	ERR_EINVAL_NORETURN (!this || !dict);
-
-	if (trace_fop_names[GF_FOP_XATTROP].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (op_ret=%d, op_errno=%d)",
-			frame->root->unique, op_ret, op_errno);
-	}
-
-	STACK_UNWIND (frame, op_ret, op_errno, dict);
-	return 0;
+int
+trace_fentrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        trace_conf_t    *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FENTRYLK].enabled) {
+                char      string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (fentrylk, frame, op_ret, op_errno, xdata);
+        return 0;
 }
 
-int32_t 
-trace_fxattrop_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    dict_t *dict)
+int
+trace_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, dict_t *dict,
+                   dict_t *xdata)
 {
-	ERR_EINVAL_NORETURN (!this || !dict);
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
 
-	if (trace_fop_names[GF_FOP_FXATTROP].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (op_ret=%d, op_errno=%d)",
-			frame->root->unique, op_ret, op_errno);
-	}
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_XATTROP].enabled) {
+                char    string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
 
-	STACK_UNWIND (frame, op_ret, op_errno, dict);
-	return 0;
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (xattrop, frame, op_ret, op_errno, dict, xdata);
+        return 0;
 }
 
-int32_t 
-trace_inodelk_cbk (call_frame_t *frame,
-		   void *cookie,
-		   xlator_t *this,
-		   int32_t op_ret,
-		   int32_t op_errno)
+int
+trace_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *dict,
+                    dict_t *xdata)
 {
-	ERR_EINVAL_NORETURN (!this );
+        trace_conf_t    *conf = NULL;
 
-	if (trace_fop_names[GF_FOP_INODELK].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL,
-			"%"PRId64": op_ret=%d, op_errno=%d",
-			frame->root->unique, op_ret, op_errno);
-	}
+        conf = this->private;
 
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FXATTROP].enabled) {
+                char      string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, dict, xdata);
+        return 0;
 }
 
+int
+trace_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
 
-int32_t
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_INODELK].enabled) {
+                char       string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local),op_ret, op_errno);
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (inodelk, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int
+trace_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FINODELK].enabled) {
+                char      string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d, op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        TRACE_STACK_UNWIND (finodelk, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int
+trace_rchecksum_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno,
+                     uint32_t weak_checksum, uint8_t *strong_checksum,
+                     dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+                goto out;
+        if (trace_fop_names[GF_FOP_RCHECKSUM].enabled) {
+                char      string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s op_ret=%d op_errno=%d",
+                          frame->root->unique,
+                          uuid_utoa (frame->local), op_ret, op_errno);
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        TRACE_STACK_UNWIND (rchecksum, frame, op_ret, op_errno, weak_checksum,
+                            strong_checksum, xdata);
+
+        return 0;
+}
+
+/* *_cbk section over <----------> fop section start */
+
+int
 trace_entrylk (call_frame_t *frame, xlator_t *this,
-	       const char *volume, loc_t *loc, const char *basename,
-	       entrylk_cmd cmd, entrylk_type type)
+               const char *volume, loc_t *loc, const char *basename,
+               entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_ENTRYLK].enabled) {
+                char     string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s volume=%s, (path=%s "
+                          "basename=%s, cmd=%s, type=%s)",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid),
+                          volume, loc->path, basename,
+                          ((cmd == ENTRYLK_LOCK) ? "ENTRYLK_LOCK" :
+                           "ENTRYLK_UNLOCK"),
+                          ((type == ENTRYLK_RDLCK) ? "ENTRYLK_RDLCK" :
+                           "ENTRYLK_WRLCK"));
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_entrylk_cbk,
+                    FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->entrylk,
+                    volume, loc, basename, cmd, type, xdata);
+        return 0;
+}
+
+int
+trace_inodelk (call_frame_t *frame, xlator_t *this, const char *volume,
+               loc_t *loc, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
+{
+        char         *cmd_str  = NULL;
+        char         *type_str = NULL;
+        trace_conf_t *conf     = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_INODELK].enabled) {
+                char string[4096]  = {0,};
+                switch (cmd) {
+#if F_GETLK != F_GETLK64
+                case F_GETLK64:
+#endif
+                case F_GETLK:
+                        cmd_str = "GETLK";
+                        break;
+
+#if F_SETLK != F_SETLK64
+                case F_SETLK64:
+#endif
+                case F_SETLK:
+                        cmd_str = "SETLK";
+                        break;
+
+#if F_SETLKW != F_SETLKW64
+                case F_SETLKW64:
+#endif
+                case F_SETLKW:
+                        cmd_str = "SETLKW";
+                        break;
+
+                default:
+                        cmd_str = "UNKNOWN";
+                        break;
+                }
+
+                switch (flock->l_type) {
+                case F_RDLCK:
+                        type_str = "READ";
+                        break;
+                case F_WRLCK:
+                        type_str = "WRITE";
+                        break;
+                case F_UNLCK:
+                        type_str = "UNLOCK";
+                        break;
+                default:
+                        type_str = "UNKNOWN";
+                        break;
+                }
+
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s volume=%s, (path=%s "
+                          "cmd=%s, type=%s, start=%llu, len=%llu, "
+                          "pid=%llu)", frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), volume,
+                          loc->path, cmd_str, type_str,
+                          (unsigned long long)flock->l_start,
+                          (unsigned long long) flock->l_len,
+                          (unsigned long long) flock->l_pid);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_inodelk_cbk,
+                    FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->inodelk,
+                    volume, loc, cmd, flock, xdata);
+        return 0;
+}
+
+int
+trace_finodelk (call_frame_t *frame, xlator_t *this, const char *volume,
+                fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
+{
+        char          *cmd_str  = NULL;
+        char          *type_str = NULL;
+        trace_conf_t  *conf     = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FINODELK].enabled) {
+                char  string[4096] = {0,};
+                switch (cmd) {
+#if F_GETLK != F_GETLK64
+                case F_GETLK64:
+#endif
+                case F_GETLK:
+                        cmd_str = "GETLK";
+                        break;
+
+#if F_SETLK != F_SETLK64
+                case F_SETLK64:
+#endif
+                case F_SETLK:
+                        cmd_str = "SETLK";
+                        break;
+
+#if F_SETLKW != F_SETLKW64
+                case F_SETLKW64:
+#endif
+                case F_SETLKW:
+                        cmd_str = "SETLKW";
+                        break;
+
+                default:
+                        cmd_str = "UNKNOWN";
+                        break;
+                }
+
+                switch (flock->l_type) {
+                case F_RDLCK:
+                        type_str = "READ";
+                        break;
+                case F_WRLCK:
+                        type_str = "WRITE";
+                        break;
+                case F_UNLCK:
+                        type_str = "UNLOCK";
+                        break;
+                default:
+                        type_str = "UNKNOWN";
+                        break;
+                }
+
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s volume=%s, (fd =%p "
+                          "cmd=%s, type=%s, start=%llu, len=%llu, "
+                          "pid=%llu)", frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), volume, fd,
+                          cmd_str, type_str,
+                          (unsigned long long) flock->l_start,
+                          (unsigned long long) flock->l_len,
+                          (unsigned long long) flock->l_pid);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        STACK_WIND (frame, trace_finodelk_cbk,
+                    FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->finodelk,
+                    volume, fd, cmd, flock, xdata);
+        return 0;
+}
+
+int
+trace_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
+               gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
 {
-	ERR_EINVAL_NORETURN (!this || !loc || !basename);
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_XATTROP].enabled) {
+                char    string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s (path=%s flags=%d)",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          flags);
+
+                frame->local = loc->inode->gfid;
 
-	if (trace_fop_names[GF_FOP_ENTRYLK].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": volume=%s, (loc= {path=%s, ino=%"PRIu64"} basename=%s, cmd=%s, type=%s)",
-			frame->root->unique, volume, loc->path, loc->inode->ino, basename, 
-			((cmd == ENTRYLK_LOCK) ? "ENTRYLK_LOCK" : "ENTRYLK_UNLOCK"), 
-			((type == ENTRYLK_RDLCK) ? "ENTRYLK_RDLCK" : "ENTRYLK_WRLCK"));
-	}
+                LOG_ELEMENT (conf, string);
+        }
 
-	STACK_WIND (frame, 
-		    trace_entrylk_cbk,
-		    FIRST_CHILD (this),
-		    FIRST_CHILD (this)->fops->entrylk,
-		    volume, loc, basename, cmd, type);
-	return 0;
+out:
+        STACK_WIND (frame, trace_xattrop_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->xattrop,
+                    loc, flags, dict, xdata);
+
+        return 0;
 }
 
-int32_t
-trace_inodelk (call_frame_t *frame,
-	       xlator_t *this,
-	       const char *volume, loc_t *loc, int32_t cmd, struct flock *flock)
+int
+trace_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+{
+        trace_conf_t    *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FXATTROP].enabled) {
+                char    string[4096]  = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p, flags=%d",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd, flags);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_fxattrop_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fxattrop,
+                    fd, flags, dict, xdata);
+
+        return 0;
+}
+
+int
+trace_lookup (call_frame_t *frame, xlator_t *this,
+              loc_t *loc, dict_t *xdata)
+{
+        trace_conf_t *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_LOOKUP].enabled) {
+                char string[4096] = {0,};
+                /* TODO: print all the keys mentioned in xattr_req */
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_lookup_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup,
+                    loc, xdata);
+
+        return 0;
+}
+
+int
+trace_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
 {
-	ERR_EINVAL_NORETURN (!this || !loc);
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+                goto out;
+        if (trace_fop_names[GF_FOP_STAT].enabled) {
+                char  string[4096]  = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_stat_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->stat,
+                    loc, xdata);
+
+        return 0;
+}
+
+int
+trace_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc, size_t size,
+                dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_READLINK].enabled) {
+                char      string[4096] = {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s, "
+                          "size=%"GF_PRI_SIZET")", frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          size);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
 
-	if (trace_fop_names[GF_FOP_INODELK].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": volume=%s, (loc {path=%s, ino=%"PRIu64"}, cmd=%s)",
-			frame->root->unique, volume, loc->path, loc->inode->ino, 
-			((cmd == F_SETLK)? "F_SETLK" : "unknown"));
-	}
+out:
+        STACK_WIND (frame, trace_readlink_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readlink,
+                    loc, size, xdata);
 
-	STACK_WIND (frame, 
-		    trace_inodelk_cbk,
-		    FIRST_CHILD (this),
-		    FIRST_CHILD (this)->fops->inodelk,
-		    volume, loc, cmd, flock);
-	return 0;
+        return 0;
 }
 
+int
+trace_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc,
+             mode_t mode, dev_t dev, mode_t umask, dict_t *xdata)
+{
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+                goto out;
+        if (trace_fop_names[GF_FOP_MKNOD].enabled) {
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s mode=%d "
+                          "umask=0%o, dev=%"GF_PRI_DEV")",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          mode, umask, dev);
+
+                LOG_ELEMENT (conf, string);
+        }
 
-int32_t 
-trace_finodelk_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno)
+out:
+        STACK_WIND (frame, trace_mknod_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->mknod,
+                    loc, mode, dev, umask, xdata);
+
+        return 0;
+}
+
+int
+trace_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+             mode_t umask, dict_t *xdata)
 {
-	ERR_EINVAL_NORETURN (!this );
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
 
-	if (trace_fop_names[GF_FOP_FINODELK].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL,
-			"%"PRId64": op_ret=%d, op_errno=%d",
-			frame->root->unique, op_ret, op_errno);
-	}
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_MKDIR].enabled) {
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s mode=%d"
+                          " umask=0%o", frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          mode, umask);
 
-	STACK_UNWIND (frame, op_ret, op_errno);
-	return 0;
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_mkdir_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->mkdir,
+                    loc, mode, umask, xdata);
+        return 0;
 }
 
-int32_t
-trace_finodelk (call_frame_t *frame,
-		xlator_t *this,
-		const char *volume, fd_t *fd, int32_t cmd, struct flock *flock)
+int
+trace_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+              dict_t *xdata)
 {
-	ERR_EINVAL_NORETURN (!this || !fd);
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_UNLINK].enabled) {
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s flag=%d",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          xflag);
 
-	if (trace_fop_names[GF_FOP_FINODELK].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": volume=%s, (fd=%p, cmd=%s)",
-			frame->root->unique, volume, fd, 
-			((cmd == F_SETLK) ? "F_SETLK" : "unknown"));
-	}
+                frame->local = loc->inode->gfid;
 
-	STACK_WIND (frame, 
-		    trace_finodelk_cbk,
-		    FIRST_CHILD (this),
-		    FIRST_CHILD (this)->fops->finodelk,
-		    volume, fd, cmd, flock);
-	return 0;
+                LOG_ELEMENT (conf, string);
+        }
+out:
+        STACK_WIND (frame, trace_unlink_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->unlink,
+                    loc, xflag, xdata);
+        return 0;
 }
 
+int
+trace_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+             dict_t *xdata)
+{
+        trace_conf_t    *conf = NULL;
 
-int32_t
-trace_xattrop (call_frame_t *frame,
-	       xlator_t *this,
-	       loc_t *loc,
-	       gf_xattrop_flags_t flags,
-	       dict_t *dict)
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_RMDIR].enabled) {
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s flags=%d",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          flags);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_rmdir_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->rmdir,
+                    loc, flags, xdata);
+
+        return 0;
+}
+
+int
+trace_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
+               loc_t *loc, mode_t umask, dict_t *xdata)
 {
-	ERR_EINVAL_NORETURN (!this || !loc);
+        trace_conf_t     *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_SYMLINK].enabled) {
+                char     string[4096]   =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s linkpath=%s, path=%s"
+                          " umask=0%o", frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), linkpath,
+                          loc->path, umask);
 
-	if (trace_fop_names[GF_FOP_XATTROP].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (path=%s, ino=%"PRIu64" flags=%d)",
-			frame->root->unique, loc->path, loc->inode->ino, flags);
-			
-	}
-  
-	STACK_WIND (frame, trace_xattrop_cbk,
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->xattrop, 
-		    loc, flags, dict);
+                LOG_ELEMENT (conf, string);
+        }
 
-	return 0;
+out:
+        STACK_WIND (frame, trace_symlink_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->symlink,
+                    linkpath, loc, umask, xdata);
+
+        return 0;
 }
 
-int32_t
-trace_fxattrop (call_frame_t *frame,
-		xlator_t *this,
-		fd_t *fd,
-		gf_xattrop_flags_t flags,
-		dict_t *dict)
-{
-	ERR_EINVAL_NORETURN (!this || !fd);
-
-	if (trace_fop_names[GF_FOP_FXATTROP].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (fd=%p, flags=%d)",
-			frame->root->unique, fd, flags);
-			
-	}
-  
-	STACK_WIND (frame, trace_fxattrop_cbk,
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->fxattrop, 
-		    fd, flags, dict);
-
-	return 0;
-}
-
-int32_t 
-trace_lookup (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *loc,
-	      dict_t *xattr_req)
-{
-	ERR_EINVAL_NORETURN (!this || !loc);
-
-	if (trace_fop_names[GF_FOP_LOOKUP].enabled) {  
-		/* TODO: print all the keys mentioned in xattr_req */
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (loc {path=%s, ino=%"PRIu64"})",
-			frame->root->unique, loc->path,
-			loc->inode->ino);
-	}
-  
-	STACK_WIND (frame, trace_lookup_cbk,
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->lookup, 
-		    loc, xattr_req);
-
-	return 0;
-}
-
-int32_t 
-trace_stat (call_frame_t *frame,
-	    xlator_t *this,
-	    loc_t *loc)
-{
-	ERR_EINVAL_NORETURN (!this || !loc );
-
-
-	if (trace_fop_names[GF_FOP_STAT].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (loc {path=%s, ino=%"PRIu64"})",
-			frame->root->unique, loc->path, loc->inode->ino);
-	}
-
-	STACK_WIND (frame, 
-		    trace_stat_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->stat, 
-		    loc);
-  
-	return 0;
-}
-
-int32_t 
-trace_readlink (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		size_t size)
-{
-	ERR_EINVAL_NORETURN (!this || !loc || (size < 1));
-
-	if (trace_fop_names[GF_FOP_READLINK].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (loc {path=%s, ino=%"PRIu64"}, size=%"GF_PRI_SIZET")",
-			frame->root->unique, loc->path, loc->inode->ino, size);
-	}
-
-	STACK_WIND (frame, 
-		    trace_readlink_cbk,
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->readlink, 
-		    loc, 
-		    size);
-  
-	return 0;
-}
-
-int32_t 
-trace_mknod (call_frame_t *frame,
-	     xlator_t *this,
-	     loc_t *loc,
-	     mode_t mode,
-	     dev_t dev)
-{
-	ERR_EINVAL_NORETURN (!this || !loc->path);
-
-	if (trace_fop_names[GF_FOP_MKNOD].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (loc {path=%s, ino=%"PRIu64"}, mode=%d, dev=%"GF_PRI_DEV")",
-			frame->root->unique, loc->path, loc->inode->ino, mode, dev);
-	}
-
-	STACK_WIND (frame, 
-		    trace_mknod_cbk,
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->mknod, 
-		    loc,
-		    mode, 
-		    dev);
-  
-	return 0;
-}
-
-int32_t 
-trace_mkdir (call_frame_t *frame,
-	     xlator_t *this,
-	     loc_t *loc,
-	     mode_t mode)
-{
-	ERR_EINVAL_NORETURN (!this || !loc || !loc->path);
-
-	if (trace_fop_names[GF_FOP_MKDIR].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (path=%s, ino=%"PRIu64", mode=%d)",
-			frame->root->unique, loc->path, 
-			((loc->inode)? loc->inode->ino : 0), mode);
-	}
-  
-	STACK_WIND (frame, 
-		    trace_mkdir_cbk,
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->mkdir, 
-		    loc,
-		    mode);
-	return 0;
-}
-
-int32_t 
-trace_unlink (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *loc)
-{
-	ERR_EINVAL_NORETURN (!this || !loc);
-
-	if (trace_fop_names[GF_FOP_UNLINK].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (loc {path=%s, ino=%"PRIu64"})",
-			frame->root->unique, loc->path, loc->inode->ino);
-	}
-
-	STACK_WIND (frame, 
-		    trace_unlink_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->unlink, 
-		    loc);
-	return 0;
-}
-
-int32_t 
-trace_rmdir (call_frame_t *frame,
-	     xlator_t *this,
-	     loc_t *loc)
-{
-	ERR_EINVAL_NORETURN (!this || !loc);
-
-	if (trace_fop_names[GF_FOP_RMDIR].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (loc {path=%s, ino=%"PRIu64"})",
-			frame->root->unique, loc->path, loc->inode->ino);
-	}
-
-	STACK_WIND (frame, 
-		    trace_rmdir_cbk,
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->rmdir, 
-		    loc);
-  
-	return 0;
-}
-
-int32_t 
-trace_symlink (call_frame_t *frame,
-	       xlator_t *this,
-	       const char *linkpath,
-	       loc_t *loc)
-{
-	ERR_EINVAL_NORETURN (!this || !linkpath || !loc || !loc->path);
-
-	if (trace_fop_names[GF_FOP_SYMLINK].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (linkpath=%s, loc {path=%s, ino=%"PRIu64"})",
-			frame->root->unique, linkpath, loc->path, 
-			((loc->inode)? loc->inode->ino : 0));
-	}
-
-	STACK_WIND (frame, 
-		    trace_symlink_cbk,
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->symlink, 
-		    linkpath,
-		    loc);
-  
-	return 0;
-}
-
-int32_t 
-trace_rename (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *oldloc,
-	      loc_t *newloc)
-{  
-	ERR_EINVAL_NORETURN (!this || !oldloc || !newloc);
-
-	if (trace_fop_names[GF_FOP_RENAME].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (oldloc {path=%s, ino=%"PRIu64"}, "
-			"newloc{path=%s, ino=%"PRIu64"})",
-			frame->root->unique, oldloc->path, oldloc->ino, 
-			newloc->path, newloc->ino);
-	}
-
-	STACK_WIND (frame, 
-		    trace_rename_cbk,
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->rename, 
-		    oldloc,
-		    newloc);
-  
-	return 0;
-}
-
-int32_t 
-trace_link (call_frame_t *frame,
-	    xlator_t *this,
-	    loc_t *oldloc,
-	    loc_t *newloc)
-{
-	ERR_EINVAL_NORETURN (!this || !oldloc || !newloc);
-
-	if (trace_fop_names[GF_FOP_LINK].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (oldloc {path=%s, ino=%"PRIu64"}, "
-			"newloc {path=%s, ino=%"PRIu64"})",
-			frame->root->unique, oldloc->path, oldloc->inode->ino, 
-			newloc->path, newloc->inode->ino);
-	}
-
-	STACK_WIND (frame, 
-		    trace_link_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->link, 
-		    oldloc,
-		    newloc);
-	return 0;
+int
+trace_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+              dict_t *xdata)
+{
+        char         oldgfid[50] = {0,};
+        char         newgfid[50] = {0,};
+        trace_conf_t *conf       = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_RENAME].enabled) {
+                char string[4096] = {0,};
+                if (newloc->inode)
+                        uuid_utoa_r (newloc->inode->gfid, newgfid);
+                else
+                        strcpy (newgfid, "0");
+
+                uuid_utoa_r (oldloc->inode->gfid, oldgfid);
+
+                snprintf (string, sizeof (string),
+                          "%"PRId64": oldgfid=%s oldpath=%s --> "
+                          "newgfid=%s newpath=%s",
+                          frame->root->unique, oldgfid,
+                          oldloc->path, newgfid, newloc->path);
+
+                frame->local = oldloc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_rename_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->rename,
+                    oldloc, newloc, xdata);
+
+        return 0;
 }
 
-int32_t
-trace_setattr (call_frame_t *frame,
-               xlator_t *this,
-               loc_t *loc,
-               struct stat *stbuf,
-               int32_t valid)
+int
+trace_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+            dict_t *xdata)
+{
+        char         oldgfid[50] = {0,};
+        char         newgfid[50] = {0,};
+        trace_conf_t *conf       = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_LINK].enabled) {
+                char string[4096]  = {0,};
+                if (newloc->inode)
+                        uuid_utoa_r (newloc->inode->gfid, newgfid);
+                else
+                        strcpy (newgfid, "0");
+
+                uuid_utoa_r (oldloc->inode->gfid, oldgfid);
+
+                snprintf (string, sizeof (string),
+                          "%"PRId64": oldgfid=%s oldpath=%s --> "
+                          "newgfid=%s newpath=%s", frame->root->unique,
+                          oldgfid, oldloc->path, newgfid,
+                          newloc->path);
+
+                frame->local = oldloc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_link_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->link,
+                    oldloc, newloc, xdata);
+        return 0;
+}
+
+int
+trace_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+               struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
-	char actime_str[256] = {0,};
-	char modtime_str[256] = {0,};
+        uint64_t     ia_time          = 0;
+        char         actime_str[256]  = {0,};
+        char         modtime_str[256] = {0,};
+        trace_conf_t *conf            = NULL;
 
-	ERR_EINVAL_NORETURN (!this || !loc || !stbuf);
+        conf = this->private;
 
-	if (trace_fop_names[GF_FOP_SETATTR].enabled) {
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_SETATTR].enabled) {
+                char     string[4096]  =  {0,};
                 if (valid & GF_SET_ATTR_MODE) {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (loc {path=%s, ino=%"PRIu64"},"
-                                " mode=%o)", frame->root->unique, loc->path,
-                                loc->inode->ino, stbuf->st_mode);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s path=%s mode=%o)",
+                                  frame->root->unique,
+                                  uuid_utoa (loc->inode->gfid),
+                                  loc->path,
+                                  st_mode_from_ia (stbuf->ia_prot,
+                                                   stbuf->ia_type));
+
+                        LOG_ELEMENT (conf, string);
+                        memset (string, 0 , sizeof (string));
                 }
 
                 if (valid & (GF_SET_ATTR_UID | GF_SET_ATTR_GID)) {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (loc {path=%s, ino=%"PRIu64"},"
-                                " uid=%o, gid=%o)",
-                                frame->root->unique, loc->path, loc->inode->ino,
-                                stbuf->st_uid, stbuf->st_gid);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s path=%s uid=%o,"
+                                  " gid=%o", frame->root->unique,
+                                  uuid_utoa (loc->inode->gfid),
+                                  loc->path, stbuf->ia_uid,
+                                  stbuf->ia_gid);
+
+                        LOG_ELEMENT (conf, string);
+                        memset (string, 0 , sizeof (string));
                 }
 
                 if (valid & (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME)) {
+                        ia_time = stbuf->ia_atime;
                         strftime (actime_str, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&stbuf->st_atime));
+                                  localtime ((time_t *)&ia_time));
+
+                        ia_time = stbuf->ia_mtime;
                         strftime (modtime_str, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&stbuf->st_mtime));
+                                  localtime ((time_t *)&ia_time));
+
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s path=%s "
+                                  "ia_atime=%s, ia_mtime=%s",
+                                  frame->root->unique,
+                                  uuid_utoa (loc->inode->gfid),
+                                  loc->path, actime_str, modtime_str);
 
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (loc {path=%s, ino=%"PRIu64"}, "
-                                "*stbuf=%p {st_atime=%s, st_mtime=%s})",
-                                frame->root->unique, loc->path, loc->inode->ino,
-                                stbuf, actime_str, modtime_str);
+                        LOG_ELEMENT (conf, string);
+                        memset (string, 0 , sizeof (string));
                 }
-	}
+                frame->local = loc->inode->gfid;
+        }
 
-	STACK_WIND (frame,
-		    trace_setattr_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->setattr,
-		    loc,
-		    stbuf, valid);
+out:
+        STACK_WIND (frame, trace_setattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setattr,
+                    loc, stbuf, valid, xdata);
 
-	return 0;
+        return 0;
 }
 
-int32_t
-trace_fsetattr (call_frame_t *frame,
-                xlator_t *this,
-                fd_t *fd,
-                struct stat *stbuf,
-                int32_t valid)
+int
+trace_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                struct iatt *stbuf, int32_t valid, dict_t *xdata)
 {
-	char actime_str[256] = {0,};
-	char modtime_str[256] = {0,};
+        uint64_t       ia_time          = 0;
+        char           actime_str[256]  = {0,};
+        char           modtime_str[256] = {0,};
+        trace_conf_t   *conf            = NULL;
 
-	ERR_EINVAL_NORETURN (!this || !fd || !stbuf);
+        conf = this->private;
 
-	if (trace_fop_names[GF_FOP_FSETATTR].enabled) {
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FSETATTR].enabled) {
+                char     string[4096]  =  {0,};
                 if (valid & GF_SET_ATTR_MODE) {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (*fd=%p, mode=%o)",
-                                frame->root->unique, fd,
-                                stbuf->st_mode);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s fd=%p, mode=%o",
+                                  frame->root->unique,
+                                  uuid_utoa (fd->inode->gfid), fd,
+                                  st_mode_from_ia (stbuf->ia_prot,
+                                                   stbuf->ia_type));
+
+                        LOG_ELEMENT (conf, string);
+                        memset (string, 0, sizeof (string));
                 }
 
                 if (valid & (GF_SET_ATTR_UID | GF_SET_ATTR_GID)) {
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (*fd=%p, uid=%o, gid=%o)",
-                                frame->root->unique, fd,
-                                stbuf->st_uid, stbuf->st_gid);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s fd=%p, uid=%o, "
+                                  "gid=%o", frame->root->unique,
+                                  uuid_utoa (fd->inode->gfid),
+                                  fd, stbuf->ia_uid, stbuf->ia_gid);
+
+                        LOG_ELEMENT (conf, string);
+                        memset (string, 0, sizeof (string));
                 }
 
                 if (valid & (GF_SET_ATTR_ATIME | GF_SET_ATTR_MTIME)) {
+                        ia_time = stbuf->ia_atime;
                         strftime (actime_str, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&stbuf->st_atime));
+                                  localtime ((time_t *)&ia_time));
+
+                        ia_time = stbuf->ia_mtime;
                         strftime (modtime_str, 256, "[%b %d %H:%M:%S]",
-                                  localtime (&stbuf->st_mtime));
+                                  localtime ((time_t *)&ia_time));
 
-                        gf_log (this->name, GF_LOG_NORMAL,
-                                "%"PRId64": (*fd=%p"
-                                "*stbuf=%p {st_atime=%s, st_mtime=%s})",
-                                frame->root->unique, fd, stbuf, actime_str,
-                                modtime_str);
+                        snprintf (string, sizeof (string),
+                                  "%"PRId64": gfid=%s fd=%p "
+                                  "ia_atime=%s, ia_mtime=%s",
+                                  frame->root->unique,
+                                  uuid_utoa (fd->inode->gfid),
+                                  fd, actime_str, modtime_str);
+
+                        LOG_ELEMENT (conf, string);
+                        memset (string, 0, sizeof (string));
                 }
-	}
-
-	STACK_WIND (frame,
-		    trace_fsetattr_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->fsetattr,
-		    fd,
-		    stbuf, valid);
-
-	return 0;
-}
-
-int32_t 
-trace_truncate (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		off_t offset)
-{
-	ERR_EINVAL_NORETURN (!this || !loc);
-
-	if (trace_fop_names[GF_FOP_TRUNCATE].enabled) { 
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (loc {path=%s, ino=%"PRIu64"}, offset=%"PRId64")",
-			frame->root->unique, loc->path, loc->inode->ino, offset);
-	}
-
-	STACK_WIND (frame, 
-		    trace_truncate_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->truncate, 
-		    loc,
-		    offset);
-  
-	return 0;
-}
-
-int32_t 
-trace_open (call_frame_t *frame,
-	    xlator_t *this,
-	    loc_t *loc,
-	    int32_t flags,
-	    fd_t *fd, int32_t wbflags)
-{
-	ERR_EINVAL_NORETURN (!this || !loc);
-
-	if (trace_fop_names[GF_FOP_OPEN].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (loc {path=%s, ino=%"PRIu64"}, flags=%d, "
-                        "fd=%p, wbflags=%d)",
-			frame->root->unique, loc->path, loc->inode->ino, flags,
-                        fd, wbflags);
-	}
-
-	STACK_WIND (frame, 
-		    trace_open_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->open, 
-		    loc,
-		    flags,
-		    fd, wbflags);
-	return 0;
-}
-
-int32_t 
-trace_create (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *loc,
-	      int32_t flags,
-	      mode_t mode,
-	      fd_t *fd)
-{
-	ERR_EINVAL_NORETURN (!this || !loc->path);
-
-	if (trace_fop_names[GF_FOP_CREATE].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (loc {path=%s, ino=%"PRIu64"}, flags=0%o mode=0%o)",
-			frame->root->unique, loc->path, loc->inode->ino, flags, mode);
-	}
-
-	STACK_WIND (frame, 
-		    trace_create_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->create, 
-		    loc,
-		    flags,
-		    mode,
-		    fd);
-	return 0;
-}
-
-int32_t 
-trace_readv (call_frame_t *frame,
-	     xlator_t *this,
-	     fd_t *fd,
-	     size_t size,
-	     off_t offset)
-{
-	ERR_EINVAL_NORETURN (!this || !fd || (size < 1));
-
-	if (trace_fop_names[GF_FOP_READ].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (*fd=%p, size=%"GF_PRI_SIZET", offset=%"PRId64")",
-			frame->root->unique, fd, size, offset);
-	}
-
-	STACK_WIND (frame, 
-		    trace_readv_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->readv,
-		    fd,
-		    size,
-		    offset);
-	return 0;
-}
-
-int32_t 
-trace_writev (call_frame_t *frame,
-	      xlator_t *this,
-	      fd_t *fd,
-	      struct iovec *vector,
-	      int32_t count,
-	      off_t offset,
-              struct iobref *iobref)
-{
-	ERR_EINVAL_NORETURN (!this || !fd || !vector || (count < 1));
-
-	if (trace_fop_names[GF_FOP_WRITE].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (*fd=%p, *vector=%p, count=%d, offset=%"PRId64")",
-			frame->root->unique, fd, vector, count, offset);
-	}
-
-	STACK_WIND (frame, 
-		    trace_writev_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->writev, 
-		    fd,
-		    vector,
-		    count,
-		    offset,
-                    iobref);
-	return 0;
-}
-
-int32_t 
-trace_statfs (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *loc)
-{
-	ERR_EINVAL_NORETURN (!this || !loc);
-
-	if (trace_fop_names[GF_FOP_STATFS].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (loc {path=%s, ino=%"PRIu64"})",
-			frame->root->unique, loc->path, 
-			((loc->inode)? loc->inode->ino : 0));
-	}
-
-	STACK_WIND (frame, 
-		    trace_statfs_cbk, 
-		    FIRST_CHILD(this), FIRST_CHILD(this)->fops->statfs, 
-		    loc);
-	return 0; 
-}
-
-int32_t 
-trace_flush (call_frame_t *frame,
-	     xlator_t *this,
-	     fd_t *fd)
-{
-	ERR_EINVAL_NORETURN (!this || !fd);
-
-	if (trace_fop_names[GF_FOP_FLUSH].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (*fd=%p)", 
-			frame->root->unique, fd);
-	}
-
-	STACK_WIND (frame, 
-		    trace_flush_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->flush, 
-		    fd);
-	return 0;
-}
-
-
-int32_t 
-trace_fsync (call_frame_t *frame,
-	     xlator_t *this,
-	     fd_t *fd,
-	     int32_t flags)
-{
-	ERR_EINVAL_NORETURN (!this || !fd);
-
-	if (trace_fop_names[GF_FOP_FSYNC].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (flags=%d, *fd=%p)", 
-			frame->root->unique, flags, fd);
-	}
-
-	STACK_WIND (frame, 
-		    trace_fsync_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->fsync, 
-		    fd,
-		    flags);
-	return 0;
-}
-
-int32_t 
-trace_setxattr (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		dict_t *dict,
-		int32_t flags)
-{
-	ERR_EINVAL_NORETURN (!this || !loc || !dict);
-
-	if (trace_fop_names[GF_FOP_SETXATTR].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (loc {path=%s, ino=%"PRIu64"}, dict=%p, flags=%d)",
-			frame->root->unique, loc->path, 
-			((loc->inode)? loc->inode->ino : 0), dict, flags);
-	}
-
-	STACK_WIND (frame, 
-		    trace_setxattr_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->setxattr, 
-		    loc,
-		    dict,
-		    flags);
-	return 0;
-}
-
-int32_t 
-trace_getxattr (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		const char *name)
-{
-	ERR_EINVAL_NORETURN (!this || !loc);
-
-	if (trace_fop_names[GF_FOP_GETXATTR].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (loc {path=%s, ino=%"PRIu64"}), name=%s",
-			frame->root->unique, loc->path, 
-			((loc->inode)? loc->inode->ino : 0), name);
-	}
-
-	STACK_WIND (frame, 
-		    trace_getxattr_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->getxattr,
-		    loc,
-		    name);
-	return 0;
-}
-
-int32_t 
-trace_removexattr (call_frame_t *frame,
-		   xlator_t *this,
-		   loc_t *loc,
-		   const char *name)
-{
-	ERR_EINVAL_NORETURN (!this || !loc || !name);
-
-	if (trace_fop_names[GF_FOP_REMOVEXATTR].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (loc {path=%s, ino=%"PRIu64"}, name=%s)",
-			frame->root->unique, loc->path, 
-			((loc->inode)? loc->inode->ino : 0), name);
-	}
-
-	STACK_WIND (frame, 
-		    trace_removexattr_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->removexattr, 
-		    loc,
-		    name);
-
-	return 0;
-}
-
-int32_t 
-trace_opendir (call_frame_t *frame,
-	       xlator_t *this,
-	       loc_t *loc,
-	       fd_t *fd)
-{
-	ERR_EINVAL_NORETURN (!this || !loc );
-
-	if (trace_fop_names[GF_FOP_OPENDIR].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64":( loc {path=%s, ino=%"PRIu64"}, fd=%p)",
-			frame->root->unique, loc->path, loc->inode->ino, fd);
-	}
-
-	STACK_WIND (frame, 
-		    trace_opendir_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->opendir, 
-		    loc,
-		    fd);
-	return 0;
-}
-
-int32_t 
-trace_getdents (call_frame_t *frame,
-		xlator_t *this,
-		fd_t *fd,
-		size_t size,
-		off_t offset,
-		int32_t flag)
-{
-	ERR_EINVAL_NORETURN (!this || !fd);  
-
-	if (trace_fop_names[GF_FOP_GETDENTS].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (fd=%p, size=%"GF_PRI_SIZET", offset=%"PRId64", flag=0x%x)",
-			frame->root->unique, fd, size, offset, flag);
-	}
-
-	STACK_WIND (frame, 
-		    trace_getdents_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->getdents, 
-		    fd,
-		    size, 
-		    offset, 
-		    flag);
-	return 0;
+                frame->local = fd->inode->gfid;
+        }
+
+out:
+        STACK_WIND (frame, trace_fsetattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsetattr,
+                    fd, stbuf, valid, xdata);
+
+        return 0;
 }
 
-int32_t
+int
+trace_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                off_t offset, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_TRUNCATE].enabled) {
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s, "
+                          "offset=%"PRId64"", frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          offset);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_truncate_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->truncate,
+                    loc, offset, xdata);
+
+        return 0;
+}
+
+int
+trace_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
+            int32_t flags, fd_t *fd, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_OPEN].enabled) {
+                char      string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s flags=%d fd=%p",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          flags, fd);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_open_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->open,
+                    loc, flags, fd, xdata);
+        return 0;
+}
+
+int
+trace_create (call_frame_t *frame, xlator_t *this, loc_t *loc,
+              int32_t flags, mode_t mode, mode_t umask, fd_t *fd,
+              dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_CREATE].enabled) {
+                char    string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s, fd=%p, "
+                          "flags=0%o mode=0%o umask=0%o",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          fd, flags, mode, umask);
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_create_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->create,
+                    loc, flags, mode, umask, fd, xdata);
+        return 0;
+}
+
+int
+trace_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
+             size_t size, off_t offset, uint32_t flags, dict_t *xdata)
+{
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_READ].enabled) {
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p, size=%"
+                          GF_PRI_SIZET"offset=%"PRId64" flags=0%x)",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd, size,
+                          offset, flags);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_readv_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readv,
+                    fd, size, offset, flags, xdata);
+        return 0;
+}
+
+int
+trace_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
+              struct iovec *vector, int32_t count,
+              off_t offset, uint32_t flags, struct iobref *iobref, dict_t *xdata)
+{
+        trace_conf_t    *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_WRITE].enabled) {
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p, count=%d, "
+                          " offset=%"PRId64" flags=0%x)",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd, count,
+                          offset, flags);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_writev_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->writev,
+                    fd, vector, count, offset, flags, iobref, xdata);
+        return 0;
+}
+
+int
+trace_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_STATFS].enabled) {
+                char  string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s",
+                          frame->root->unique, (loc->inode)?
+                          uuid_utoa (loc->inode->gfid):"0", loc->path);
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_statfs_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->statfs,
+                    loc, xdata);
+        return 0;
+}
+
+int
+trace_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+{
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FLUSH].enabled) {
+                char    string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_flush_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->flush,
+                    fd, xdata);
+        return 0;
+}
+
+int
+trace_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t flags,
+             dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FSYNC].enabled) {
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s flags=%d fd=%p",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), flags, fd);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_fsync_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsync,
+                    fd, flags, xdata);
+        return 0;
+}
+
+int
+trace_setxattr (call_frame_t *frame, xlator_t *this,
+                loc_t *loc, dict_t *dict, int32_t flags, dict_t *xdata)
+{
+        trace_conf_t    *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_SETXATTR].enabled) {
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s flags=%d",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          flags);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_setxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr,
+                    loc, dict, flags, xdata);
+        return 0;
+}
+
+int
+trace_getxattr (call_frame_t *frame, xlator_t *this,
+                loc_t *loc, const char *name, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_GETXATTR].enabled) {
+                char    string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s name=%s",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          name);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_getxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->getxattr,
+                    loc, name, xdata);
+        return 0;
+}
+
+int
+trace_removexattr (call_frame_t *frame, xlator_t *this,
+                   loc_t *loc, const char *name, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_REMOVEXATTR].enabled) {
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s name=%s",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path,
+                          name);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_removexattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->removexattr,
+                    loc, name, xdata);
+
+        return 0;
+}
+
+int
+trace_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc, fd_t *fd,
+               dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_OPENDIR].enabled) {
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s fd=%p",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid), loc->path, fd);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_opendir_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->opendir,
+                    loc, fd, xdata);
+        return 0;
+}
+
+int
 trace_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
-               off_t offset)
+                off_t offset, dict_t *dict)
+{
+        trace_conf_t *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_READDIRP].enabled) {
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p, size=%"GF_PRI_SIZET
+                          ", offset=%"PRId64" dict=%p",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd, size,
+                          offset, dict);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_readdirp_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readdirp,
+                    fd, size, offset, dict);
+
+        return 0;
+}
+
+int
+trace_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
+               size_t size, off_t offset, dict_t *xdata)
+{
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_READDIR].enabled) {
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p, size=%"GF_PRI_SIZET
+                          ", offset=%"PRId64,
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd, size,
+                          offset);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_readdir_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readdir,
+                    fd, size, offset, xdata);
+
+        return 0;
+}
+
+int
+trace_fsyncdir (call_frame_t *frame, xlator_t *this,
+                fd_t *fd, int32_t datasync, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FSYNCDIR].enabled) {
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s datasync=%d fd=%p",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), datasync, fd);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_fsyncdir_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsyncdir,
+                    fd, datasync, xdata);
+        return 0;
+}
+
+int
+trace_access (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t mask,
+              dict_t *xdata)
+{
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_ACCESS].enabled) {
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s path=%s mask=0%o",
+                          frame->root->unique,
+                          uuid_utoa (loc->inode->gfid),
+                          loc->path, mask);
+
+                frame->local = loc->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_access_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->access,
+                    loc, mask, xdata);
+        return 0;
+}
+
+int32_t
+trace_rchecksum (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+                 int32_t len, dict_t *xdata)
+{
+
+        trace_conf_t *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_RCHECKSUM].enabled) {
+                char    string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s offset=%"PRId64
+                          "len=%u fd=%p", frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), offset, len, fd);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_rchecksum_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->rchecksum,
+                    fd, offset, len, xdata);
+
+        return 0;
+
+}
+
+int32_t
+trace_fentrylk (call_frame_t *frame, xlator_t *this, const char *volume,
+                fd_t *fd, const char *basename, entrylk_cmd cmd,
+                entrylk_type type, dict_t *xdata)
 {
-	ERR_EINVAL_NORETURN (!this || !fd);
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FENTRYLK].enabled) {
+                char      string[4096]   =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s volume=%s, (fd=%p "
+                          "basename=%s, cmd=%s, type=%s)",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), volume, fd,
+                          basename,
+                          ((cmd == ENTRYLK_LOCK) ? "ENTRYLK_LOCK" :
+                           "ENTRYLK_UNLOCK"),
+                          ((type == ENTRYLK_RDLCK) ? "ENTRYLK_RDLCK" :
+                           "ENTRYLK_WRLCK"));
+
+                frame->local = fd->inode->gfid;
 
-	if (trace_fop_names[GF_FOP_READDIRP].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL,
-			"%"PRId64": (fd=%p, size=%"GF_PRI_SIZET", offset=%"PRId64")",
-			frame->root->unique, fd, size, offset);
-	}
+                LOG_ELEMENT (conf, string);
+        }
 
-	STACK_WIND (frame, trace_readdirp_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->readdirp, fd, size, offset);
+out:
+        STACK_WIND (frame, trace_fentrylk_cbk,
+                    FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->fentrylk,
+                    volume, fd, basename, cmd, type, xdata);
+        return 0;
 
-	return 0;
 }
 
+int32_t
+trace_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                 const char *name, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+                goto out;
+        if (trace_fop_names[GF_FOP_FGETXATTR].enabled) {
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p name=%s",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd, name);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_fgetxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fgetxattr,
+                    fd, name, xdata);
+        return 0;
+}
+
+int32_t
+trace_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                 dict_t *dict, int32_t flags, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FSETXATTR].enabled) {
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p flags=%d",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd, flags);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
 
-int32_t 
-trace_readdir (call_frame_t *frame,
-	       xlator_t *this,
-	       fd_t *fd,
-	       size_t size,
-	       off_t offset)
+out:
+        STACK_WIND (frame, trace_fsetxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsetxattr,
+                    fd, dict, flags, xdata);
+        return 0;
+}
+
+int
+trace_ftruncate (call_frame_t *frame, xlator_t *this,
+                 fd_t *fd, off_t offset, dict_t *xdata)
 {
-	ERR_EINVAL_NORETURN (!this || !fd);  
+        trace_conf_t  *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FTRUNCATE].enabled) {
+                char    string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s offset=%"PRId64" fd=%p",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), offset, fd);
 
-	if (trace_fop_names[GF_FOP_READDIR].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (fd=%p, size=%"GF_PRI_SIZET", offset=%"PRId64")",
-			frame->root->unique, fd, size, offset);
-	}
+                frame->local = fd->inode->gfid;
 
-	STACK_WIND (frame, 
-		    trace_readdir_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->readdir,
-		    fd,
-		    size, 
-		    offset);
+                LOG_ELEMENT (conf, string);
+        }
 
-	return 0;
+out:
+        STACK_WIND (frame, trace_ftruncate_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->ftruncate,
+                    fd, offset, xdata);
+
+        return 0;
 }
 
+int
+trace_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_FSTAT].enabled) {
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_fstat_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fstat,
+                    fd, xdata);
+        return 0;
+}
 
-int32_t 
-trace_fsyncdir (call_frame_t *frame,
-		xlator_t *this,
-		fd_t *fd,
-		int32_t datasync)
+int
+trace_lk (call_frame_t *frame, xlator_t *this, fd_t *fd,
+          int32_t cmd, struct gf_flock *lock, dict_t *xdata)
 {
-	ERR_EINVAL_NORETURN (!this || !fd);
-
-	if (trace_fop_names[GF_FOP_FSYNCDIR].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (datasync=%d, *fd=%p)", 
-			frame->root->unique, datasync, fd);
-	}
-
-	STACK_WIND (frame, 
-		    trace_fsyncdir_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->fsyncdir, 
-		    fd,
-		    datasync);
-	return 0;
-}
-
-int32_t 
-trace_access (call_frame_t *frame,
-	      xlator_t *this,
-	      loc_t *loc,
-	      int32_t mask)
-{
-	ERR_EINVAL_NORETURN (!this || !loc);
-
-	if (trace_fop_names[GF_FOP_ACCESS].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (*loc {path=%s, ino=%"PRIu64"}, mask=0%o)",
-			frame->root->unique, loc->path, 
-			((loc->inode)? loc->inode->ino : 0), mask);
-	}
-
-	STACK_WIND (frame, 
-		    trace_access_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->access, 
-		    loc,
-		    mask);
-	return 0;
-}
-
-int32_t 
-trace_ftruncate (call_frame_t *frame,
-		 xlator_t *this,
-		 fd_t *fd,
-		 off_t offset)
-{
-	ERR_EINVAL_NORETURN (!this || !fd);
-
-	if (trace_fop_names[GF_FOP_FTRUNCATE].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (offset=%"PRId64", *fd=%p)", 
-			frame->root->unique, offset, fd);
-	}
-
-	STACK_WIND (frame, 
-		    trace_ftruncate_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->ftruncate, 
-		    fd,
-		    offset);
-
-	return 0;
-}
-
-int32_t 
-trace_fstat (call_frame_t *frame,
-	     xlator_t *this,
-	     fd_t *fd)
-{
-	ERR_EINVAL_NORETURN (!this || !fd);
-
-	if (trace_fop_names[GF_FOP_FSTAT].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (*fd=%p)", 
-			frame->root->unique, fd);
-	}
-
-	STACK_WIND (frame, 
-		    trace_fstat_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->fstat, 
-		    fd);
-	return 0;
-}
-
-int32_t 
-trace_lk (call_frame_t *frame,
-	  xlator_t *this,
-	  fd_t *fd,
-	  int32_t cmd,
-	  struct flock *lock)
-{
-	ERR_EINVAL_NORETURN (!this || !fd);
-
-	if (trace_fop_names[GF_FOP_LK].enabled) {  
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (*fd=%p, cmd=%d, lock {l_type=%d, l_whence=%d, "
-			"l_start=%"PRId64", l_len=%"PRId64", l_pid=%u})",
-			frame->root->unique, fd, cmd, lock->l_type, lock->l_whence, 
-			lock->l_start, lock->l_len, lock->l_pid);
-	}
-
-	STACK_WIND (frame, 
-		    trace_lk_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->lk, 
-		    fd,
-		    cmd,
-		    lock);
-	return 0;
-}
-
-int32_t 
-trace_setdents (call_frame_t *frame,
-		xlator_t *this,
-		fd_t *fd,
-		int32_t flags,
-		dir_entry_t *entries,
-		int32_t count)
-{
-	if (trace_fop_names[GF_FOP_SETDENTS].enabled) {
-		gf_log (this->name, GF_LOG_NORMAL, 
-			"%"PRId64": (*fd=%p, flags=%d, count=%d", 
-			frame->root->unique, fd, flags, count);
-	}
-
-	STACK_WIND (frame, 
-		    trace_setdents_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->setdents, 
-		    fd,
-		    flags,
-		    entries,
-		    count);
-	return 0;
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_LK].enabled) {
+                char     string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "%"PRId64": gfid=%s fd=%p, cmd=%d, "
+                          "lock {l_type=%d, "
+                          "l_whence=%d, l_start=%"PRId64", "
+                          "l_len=%"PRId64", l_pid=%u})",
+                          frame->root->unique,
+                          uuid_utoa (fd->inode->gfid), fd, cmd,
+                          lock->l_type, lock->l_whence,
+                          lock->l_start, lock->l_len, lock->l_pid);
+
+                frame->local = fd->inode->gfid;
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        STACK_WIND (frame, trace_lk_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lk,
+                    fd, cmd, lock, xdata);
+        return 0;
 }
 
+int32_t
+trace_forget (xlator_t *this, inode_t *inode)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+        /* If user want to understand when a lookup happens,
+           he should know about 'forget' too */
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_LOOKUP].enabled) {
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "gfid=%s", uuid_utoa (inode->gfid));
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        return 0;
+}
 
 int32_t
-trace_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)
+trace_releasedir (xlator_t *this, fd_t *fd)
 {
-	gf_log (this->name, GF_LOG_NORMAL, 
-		"%"PRId64": op_ret (%d), op_errno(%d)", 
-		frame->root->unique, op_ret, op_errno);
+        trace_conf_t  *conf = NULL;
 
-	STACK_UNWIND (frame, op_ret, op_errno, fchecksum, dchecksum);
+        conf = this->private;
 
-	return 0;
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_OPENDIR].enabled) {
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "gfid=%s fd=%p",
+                          uuid_utoa (fd->inode->gfid), fd);
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        return 0;
 }
 
 int32_t
-trace_checksum (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		int32_t flag)
-{
-	gf_log (this->name, GF_LOG_NORMAL, 
-		"%"PRId64": loc->path (%s) flag (%d)", 
-		frame->root->unique, loc->path, flag);
-  
-	STACK_WIND (frame,
-		    trace_checksum_cbk,
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->fops->checksum, 
-		    loc,
-		    flag);
-
-	return 0;
-}
-
-
-int32_t 
-trace_stats_cbk (call_frame_t *frame,
-		 void *cookie,
-		 xlator_t *this,
-		 int32_t op_ret,
-		 int32_t op_errno,
-		 struct xlator_stats *stats)
-{
-	gf_log (this->name, GF_LOG_NORMAL, 
-		"%"PRId64": op_ret (%d), op_errno(%d)", 
-		frame->root->unique, op_ret, op_errno);
-
-	STACK_UNWIND (frame, op_ret, op_errno, stats);
-	return 0;
-}
-
-int32_t 
-trace_stats (call_frame_t *frame,
-	     xlator_t *this, 
-	     int32_t flags)
-{
-	ERR_EINVAL_NORETURN (!this);
-  
-	gf_log (this->name, GF_LOG_NORMAL, 
-		"%"PRId64": (flags=%d)", 
-		frame->root->unique, flags);
-	
-	STACK_WIND (frame, 
-		    trace_stats_cbk, 
-		    FIRST_CHILD(this), 
-		    FIRST_CHILD(this)->mops->stats, 
-		    flags);
-
-	return 0;
+trace_release (xlator_t *this, fd_t *fd)
+{
+        trace_conf_t   *conf = NULL;
+
+        conf = this->private;
+
+        if (!conf->log_file && !conf->log_history)
+		goto out;
+        if (trace_fop_names[GF_FOP_OPEN].enabled ||
+            trace_fop_names[GF_FOP_CREATE].enabled) {
+                char   string[4096]  =  {0,};
+                snprintf (string, sizeof (string),
+                          "gfid=%s fd=%p",
+                          uuid_utoa (fd->inode->gfid), fd);
+
+                LOG_ELEMENT (conf, string);
+        }
+
+out:
+        return 0;
 }
 
+
 void
 enable_all_calls (int enabled)
 {
-	int i;
-	for (i = 0; i < GF_FOP_MAXVALUE; i++)
-		trace_fop_names[i].enabled = enabled;
+        int i;
+
+        for (i = 0; i < GF_FOP_MAXVALUE; i++)
+                trace_fop_names[i].enabled = enabled;
 }
 
-void 
+void
 enable_call (const char *name, int enabled)
 {
-	int i;
-	for (i = 0; i < GF_FOP_MAXVALUE; i++)
-		if (!strcasecmp(trace_fop_names[i].name, name))
-			trace_fop_names[i].enabled = enabled;
+        int i;
+        for (i = 0; i < GF_FOP_MAXVALUE; i++)
+                if (!strcasecmp(trace_fop_names[i].name, name))
+                        trace_fop_names[i].enabled = enabled;
 }
 
 
-/* 
-   include = 1 for "include-ops"
-           = 0 for "exclude-ops" 
+/*
+  include = 1 for "include-ops"
+  = 0 for "exclude-ops"
 */
 void
 process_call_list (const char *list, int include)
 {
-	enable_all_calls (include ? 0 : 1);
-	
-	char *call = strsep ((char **)&list, ",");
-	while (call) {
-		enable_call (call, include);
-		call = strsep ((char **)&list, ",");
-	}
+        enable_all_calls (include ? 0 : 1);
+
+        char *call = strsep ((char **)&list, ",");
+
+        while (call) {
+                enable_call (call, include);
+                call = strsep ((char **)&list, ",");
+        }
+}
+
+int32_t
+trace_dump_history (xlator_t *this)
+{
+        int ret = -1;
+        char key_prefix[GF_DUMP_MAX_BUF_LEN] = {0,};
+        trace_conf_t *conf = NULL;
+
+        GF_VALIDATE_OR_GOTO ("trace", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, this->history, out);
+
+        conf = this->private;
+        // Is it ok to return silently if log-history option his off?
+        if (conf && conf->log_history == _gf_true) {
+                gf_proc_dump_build_key (key_prefix, "xlator.debug.trace",
+                                        "history");
+                gf_proc_dump_add_section (key_prefix);
+                eh_dump (this->history, NULL, dump_history_trace);
+        }
+        ret = 0;
+
+out:
+        return ret;
+}
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_trace_mt_end + 1);
+
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
+                        " failed");
+                return ret;
+        }
+
+        return ret;
 }
 
+int
+reconfigure (xlator_t *this, dict_t *options)
+{
+        int32_t       ret            = -1;
+        trace_conf_t    *conf           = NULL;
+        char            *includes = NULL, *excludes = NULL;
+
+        GF_VALIDATE_OR_GOTO ("quick-read", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, this->private, out);
+        GF_VALIDATE_OR_GOTO (this->name, options, out);
+
+        conf = this->private;
+
+        includes = data_to_str (dict_get (options, "include-ops"));
+        excludes = data_to_str (dict_get (options, "exclude-ops"));
+
+        {
+                int i;
+                for (i = 0; i < GF_FOP_MAXVALUE; i++) {
+                        if (gf_fop_list[i])
+                                strncpy (trace_fop_names[i].name,
+                                         gf_fop_list[i],
+                                         strlen (gf_fop_list[i]));
+                        else
+                                strncpy (trace_fop_names[i].name, ":O",
+                                         strlen (":O"));
+                        trace_fop_names[i].enabled = 1;
+                }
+        }
+
+        if (includes && excludes) {
+                gf_log (this->name,
+                        GF_LOG_ERROR,
+                        "must specify only one of 'include-ops' and "
+                        "'exclude-ops'");
+                goto out;
+        }
+
+        if (includes)
+                process_call_list (includes, 1);
+        if (excludes)
+                process_call_list (excludes, 0);
+
+        /* Should resizing of the event-history be allowed in reconfigure?
+         * for which a new event_history might have to be allocated and the
+         * older history has to be freed.
+         */
+        GF_OPTION_RECONF ("log-file", conf->log_file, options, bool, out);
+
+        GF_OPTION_RECONF ("log-history", conf->log_history, options, bool, out);
+
+        ret = 0;
+
+out:
+        return ret;
+}
 
-int32_t 
+int32_t
 init (xlator_t *this)
 {
-  dict_t *options = NULL;
-  char *includes = NULL, *excludes = NULL;
-  
-  if (!this)
-	  return -1;
-  
-  if (!this->children || this->children->next) {
-	  gf_log (this->name, GF_LOG_ERROR, 
-		  "trace translator requires one subvolume");
-	  return -1;
-  }
-  if (!this->parents) {
-	  gf_log (this->name, GF_LOG_WARNING,
-		  "dangling volume. check volfile ");
-  }
-  
-  
-  options = this->options;
-  includes = data_to_str (dict_get (options, "include-ops"));
-  excludes = data_to_str (dict_get (options, "exclude-ops"));
-  
-  {
-	  int i;
-	  for (i = 0; i < GF_FOP_MAXVALUE; i++) {
-		  trace_fop_names[i].name = (gf_fop_list[i] ?
-					     gf_fop_list[i] : ":O");
-		  trace_fop_names[i].enabled = 1;
-	  }
-  }
-  
-  if (includes && excludes) {
-	  gf_log (this->name, 
-		  GF_LOG_ERROR,
-		  "must specify only one of 'include-ops' and 'exclude-ops'");
-	  return -1;
-  }
-  if (includes)
-	  process_call_list (includes, 1);
-  if (excludes)
-	  process_call_list (excludes, 0);
-  
-  gf_log_set_loglevel (GF_LOG_NORMAL);
-  
-  /* Set this translator's inode table pointer to child node's pointer. */
-  this->itable = FIRST_CHILD (this)->itable;
-  
-  return 0;
+        dict_t *options = NULL;
+        char *includes = NULL, *excludes = NULL;
+        char *forced_loglevel = NULL;
+        eh_t *history = NULL;
+        int  ret = -1;
+        size_t  history_size = TRACE_DEFAULT_HISTORY_SIZE;
+        trace_conf_t    *conf = NULL;
+
+        if (!this)
+                return -1;
+
+        if (!this->children || this->children->next) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "trace translator requires one subvolume");
+                return -1;
+        }
+        if (!this->parents) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "dangling volume. check volfile ");
+        }
+
+        conf = GF_CALLOC (1, sizeof (trace_conf_t), gf_trace_mt_trace_conf_t);
+        if (!conf) {
+                gf_log (this->name, GF_LOG_ERROR, "cannot allocate "
+                        "xl->private");
+                return -1;
+        }
+
+        options = this->options;
+        includes = data_to_str (dict_get (options, "include-ops"));
+        excludes = data_to_str (dict_get (options, "exclude-ops"));
+
+        {
+                int i;
+                for (i = 0; i < GF_FOP_MAXVALUE; i++) {
+                        if (gf_fop_list[i])
+                                strncpy (trace_fop_names[i].name,
+                                         gf_fop_list[i],
+                                         strlen (gf_fop_list[i]));
+                        else
+                                strncpy (trace_fop_names[i].name, ":O",
+                                         strlen (":O"));
+                        trace_fop_names[i].enabled = 1;
+                }
+        }
+
+        if (includes && excludes) {
+                gf_log (this->name,
+                        GF_LOG_ERROR,
+                        "must specify only one of 'include-ops' and "
+                        "'exclude-ops'");
+                return -1;
+        }
+
+        if (includes)
+                process_call_list (includes, 1);
+        if (excludes)
+                process_call_list (excludes, 0);
+
+
+        GF_OPTION_INIT ("history-size", conf->history_size, size, out);
+
+        gf_log (this->name, GF_LOG_INFO, "history size %"GF_PRI_SIZET,
+                history_size);
+
+        GF_OPTION_INIT ("log-file", conf->log_file, bool, out);
+
+        gf_log (this->name, GF_LOG_INFO, "logging to file %s",
+                (conf->log_file == _gf_true)?"enabled":"disabled");
+
+        GF_OPTION_INIT ("log-history", conf->log_history, bool, out);
+
+        gf_log (this->name, GF_LOG_DEBUG, "logging to history %s",
+                (conf->log_history == _gf_true)?"enabled":"disabled");
+
+        history = eh_new (history_size, _gf_false, NULL);
+        if (!history) {
+                gf_log (this->name, GF_LOG_ERROR, "event history cannot be "
+                        "initialized");
+                return -1;
+        }
+
+        this->history = history;
+
+        conf->trace_log_level = GF_LOG_INFO;
+
+        if (dict_get (options, "force-log-level")) {
+                forced_loglevel = data_to_str (dict_get (options,
+                                                         "force-log-level"));
+                if (!forced_loglevel)
+                        goto setloglevel;
+
+                if (strcmp (forced_loglevel, "INFO") == 0)
+                        conf->trace_log_level = GF_LOG_INFO;
+                else if (strcmp (forced_loglevel, "TRACE") == 0)
+                        conf->trace_log_level = GF_LOG_TRACE;
+                else if (strcmp (forced_loglevel, "ERROR") == 0)
+                        conf->trace_log_level = GF_LOG_ERROR;
+                else if (strcmp (forced_loglevel, "DEBUG") == 0)
+                        conf->trace_log_level = GF_LOG_DEBUG;
+                else if (strcmp (forced_loglevel, "WARNING") == 0)
+                        conf->trace_log_level = GF_LOG_WARNING;
+                else if (strcmp (forced_loglevel, "CRITICAL") == 0)
+                        conf->trace_log_level = GF_LOG_CRITICAL;
+                else if (strcmp (forced_loglevel, "NONE") == 0)
+                        conf->trace_log_level = GF_LOG_NONE;
+        }
+
+setloglevel:
+        gf_log_set_loglevel (conf->trace_log_level);
+        this->private = conf;
+        ret = 0;
+out:
+        if (ret == -1) {
+                if (history)
+                        GF_FREE (history);
+                if (conf)
+                        GF_FREE (conf);
+        }
+
+        return ret;
 }
 
 void
 fini (xlator_t *this)
 {
-	if (!this)
-		return;
-	
-	gf_log (this->name, GF_LOG_NORMAL, 
-		"trace translator unloaded");
-	return;
+        if (!this)
+                return;
+
+        if (this->history)
+                eh_destroy (this->history);
+
+        gf_log (this->name, GF_LOG_INFO,
+                "trace translator unloaded");
+        return;
 }
 
 struct xlator_fops fops = {
-  .stat        = trace_stat,
-  .readlink    = trace_readlink,
-  .mknod       = trace_mknod,
-  .mkdir       = trace_mkdir,
-  .unlink      = trace_unlink,
-  .rmdir       = trace_rmdir,
-  .symlink     = trace_symlink,
-  .rename      = trace_rename,
-  .link        = trace_link,
-  .truncate    = trace_truncate,
-  .open        = trace_open,
-  .readv       = trace_readv,
-  .writev      = trace_writev,
-  .statfs      = trace_statfs,
-  .flush       = trace_flush,
-  .fsync       = trace_fsync,
-  .setxattr    = trace_setxattr,
-  .getxattr    = trace_getxattr,
-  .removexattr = trace_removexattr,
-  .opendir     = trace_opendir,
-  .readdir     = trace_readdir, 
-  .readdirp    = trace_readdirp,
-  .fsyncdir    = trace_fsyncdir,
-  .access      = trace_access,
-  .ftruncate   = trace_ftruncate,
-  .fstat       = trace_fstat,
-  .create      = trace_create,
-  .lk          = trace_lk,
-  .inodelk     = trace_inodelk,
-  .finodelk    = trace_finodelk,
-  .entrylk     = trace_entrylk,
-  .lookup      = trace_lookup,
-  .setdents    = trace_setdents,
-  .getdents    = trace_getdents,
-  .checksum    = trace_checksum,
-  .xattrop     = trace_xattrop,
-  .fxattrop    = trace_fxattrop,
-  .setattr     = trace_setattr,
-  .fsetattr    = trace_fsetattr,
-};
-
-struct xlator_mops mops = {
-	.stats    = trace_stats,
+        .stat        = trace_stat,
+        .readlink    = trace_readlink,
+        .mknod       = trace_mknod,
+        .mkdir       = trace_mkdir,
+        .unlink      = trace_unlink,
+        .rmdir       = trace_rmdir,
+        .symlink     = trace_symlink,
+        .rename      = trace_rename,
+        .link        = trace_link,
+        .truncate    = trace_truncate,
+        .open        = trace_open,
+        .readv       = trace_readv,
+        .writev      = trace_writev,
+        .statfs      = trace_statfs,
+        .flush       = trace_flush,
+        .fsync       = trace_fsync,
+        .setxattr    = trace_setxattr,
+        .getxattr    = trace_getxattr,
+        .fsetxattr   = trace_fsetxattr,
+        .fgetxattr   = trace_fgetxattr,
+        .removexattr = trace_removexattr,
+        .opendir     = trace_opendir,
+        .readdir     = trace_readdir,
+        .readdirp    = trace_readdirp,
+        .fsyncdir    = trace_fsyncdir,
+        .access      = trace_access,
+        .ftruncate   = trace_ftruncate,
+        .fstat       = trace_fstat,
+        .create      = trace_create,
+        .lk          = trace_lk,
+        .inodelk     = trace_inodelk,
+        .finodelk    = trace_finodelk,
+        .entrylk     = trace_entrylk,
+        .fentrylk    = trace_fentrylk,
+        .lookup      = trace_lookup,
+        .rchecksum   = trace_rchecksum,
+        .xattrop     = trace_xattrop,
+        .fxattrop    = trace_fxattrop,
+        .setattr     = trace_setattr,
+        .fsetattr    = trace_fsetattr,
 };
 
 struct xlator_cbks cbks = {
+        .release     = trace_release,
+        .releasedir  = trace_releasedir,
+        .forget      = trace_forget,
 };
 
 struct volume_options options[] = {
-	{ .key  = {"include-ops", "include"}, 
-	  .type = GF_OPTION_TYPE_STR,
-	  /*.value = { ""} */
-	},
-	{ .key  = {"exclude-ops", "exclude"}, 
-	  .type = GF_OPTION_TYPE_STR 
-	  /*.value = { ""} */	  
-	},
-	{ .key  = {NULL} },
+        { .key  = {"include-ops", "include"},
+          .type = GF_OPTION_TYPE_STR,
+          /*.value = { ""} */
+        },
+        { .key  = {"exclude-ops", "exclude"},
+          .type = GF_OPTION_TYPE_STR
+          /*.value = { ""} */
+        },
+        { .key  = {"history-size"},
+          .type = GF_OPTION_TYPE_SIZET,
+          .default_value = "1024",
+        },
+        { .key  = {"log-file"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "no",
+        },
+        { .key  = {"log-history"},
+          .type = GF_OPTION_TYPE_BOOL,
+          .default_value = "no",
+        },
+        { .key  = {NULL} },
 };
 
+struct xlator_dumpops dumpops = {
+        .history = trace_dump_history
+};
diff --git a/xlators/debug/trace/src/trace.h b/xlators/debug/trace/src/trace.h
new file mode 100644
index 000000000..045eefb36
--- /dev/null
+++ b/xlators/debug/trace/src/trace.h
@@ -0,0 +1,98 @@
+/*
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include <time.h>
+#include <errno.h>
+#include "glusterfs.h"
+#include "xlator.h"
+#include "common-utils.h"
+#include "event-history.h"
+#include "logging.h"
+#include "circ-buff.h"
+#include "statedump.h"
+#include "options.h"
+
+#define TRACE_DEFAULT_HISTORY_SIZE 1024
+
+typedef struct {
+        /* Since the longest fop name is fremovexattr i.e 12 characters, array size
+         * is kept 24, i.e double of the maximum.
+         */
+        char name[24];
+        int enabled;
+} trace_fop_name_t;
+
+trace_fop_name_t trace_fop_names[GF_FOP_MAXVALUE];
+
+typedef struct {
+        gf_boolean_t log_file;
+        gf_boolean_t log_history;
+        size_t       history_size;
+        int trace_log_level;
+} trace_conf_t;
+
+#define TRACE_STACK_UNWIND(op, frame, params ...)                       \
+        do {                                                            \
+                frame->local = NULL;                                    \
+                STACK_UNWIND_STRICT (op, frame, params);                \
+        } while (0);
+
+#define LOG_ELEMENT(_conf, _string)                                     \
+        do {                                                            \
+                if (_conf) {                                            \
+                        if ((_conf->log_history) == _gf_true)           \
+                                gf_log_eh ("%s", _string);              \
+                        if ((_conf->log_file) == _gf_true)              \
+                                gf_log (THIS->name, _conf->trace_log_level, \
+                                        "%s", _string);                 \
+                }                                                       \
+        } while (0);
+
+#define trace_stat_to_str(buf, statstr)                                 \
+        do {                                                            \
+                char     atime_buf[256]    = {0,};                      \
+                char     mtime_buf[256]    = {0,};                      \
+                char     ctime_buf[256]    = {0,};                      \
+                uint64_t ia_time           = 0;                         \
+                                                                        \
+                if (!buf)                                               \
+                        break;                                          \
+                                                                        \
+                ia_time = buf->ia_atime;                                \
+                strftime (atime_buf, 256, "[%b %d %H:%M:%S]",           \
+                          localtime ((time_t *)&ia_time));              \
+                                                                        \
+                ia_time = buf->ia_mtime;                                \
+                strftime (mtime_buf, 256, "[%b %d %H:%M:%S]",           \
+                          localtime ((time_t *)&ia_time));              \
+                                                                        \
+                ia_time = buf->ia_ctime;                                \
+                strftime (ctime_buf, 256, "[%b %d %H:%M:%S]",           \
+                          localtime ((time_t *)&ia_time));              \
+                                                                        \
+                snprintf (statstr, sizeof (statstr),                    \
+                          "gfid=%s ino=%"PRIu64", mode=%o, "            \
+                          "nlink=%"GF_PRI_NLINK", uid=%u, "             \
+                          "gid=%u, size=%"PRIu64", "                    \
+                          "blocks=%"PRIu64", atime=%s, "                \
+                          "mtime=%s, ctime=%s",                         \
+                          uuid_utoa (buf->ia_gfid), buf->ia_ino,        \
+                          st_mode_from_ia (buf->ia_prot,                \
+                                           buf->ia_type),               \
+                          buf->ia_nlink, buf->ia_uid,                   \
+                          buf->ia_gid, buf->ia_size,                    \
+                          buf->ia_blocks, atime_buf,                    \
+                          mtime_buf, ctime_buf);                        \
+        } while (0);
diff --git a/xlators/encryption/Makefile.am b/xlators/encryption/Makefile.am
index 2cbde680f..36efc6698 100644
--- a/xlators/encryption/Makefile.am
+++ b/xlators/encryption/Makefile.am
@@ -1,3 +1,3 @@
-SUBDIRS = rot-13
+SUBDIRS = rot-13 crypt
 
 CLEANFILES = 
diff --git a/xlators/cluster/unify/Makefile.am b/xlators/encryption/crypt/Makefile.am
index d471a3f92..d471a3f92 100644
--- a/xlators/cluster/unify/Makefile.am
+++ b/xlators/encryption/crypt/Makefile.am
diff --git a/xlators/encryption/crypt/src/Makefile.am b/xlators/encryption/crypt/src/Makefile.am
new file mode 100644
index 000000000..b13f65043
--- /dev/null
+++ b/xlators/encryption/crypt/src/Makefile.am
@@ -0,0 +1,24 @@
+if ENABLE_CRYPT_XLATOR
+
+xlator_LTLIBRARIES = crypt.la
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/encryption
+
+crypt_la_LDFLAGS = -module -avoid-version -lssl -lcrypto
+
+crypt_la_SOURCES = keys.c data.c metadata.c atom.c crypt.c
+crypt_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+noinst_HEADERS = crypt-common.h crypt-mem-types.h crypt.h metadata.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+CLEANFILES =
+
+else
+
+noinst_DIST = keys.c data.c metadata.c atom.c crypt.c
+noinst_HEADERS = crypt-common.h crypt-mem-types.h crypt.h metadata.h
+
+endif
+\ No newline at end of file
diff --git a/xlators/encryption/crypt/src/atom.c b/xlators/encryption/crypt/src/atom.c
new file mode 100644
index 000000000..1ec41495c
--- /dev/null
+++ b/xlators/encryption/crypt/src/atom.c
@@ -0,0 +1,962 @@
+/*
+  Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "defaults.h"
+#include "crypt-common.h"
+#include "crypt.h"
+
+/*
+ *                              Glossary
+ *
+ *
+ * cblock            (or cipher block). A logical unit in a file.
+ *                   cblock size is defined as the number of bits
+ *                   in an input (or output) block of the block
+ *                   cipher (*). Cipher block size is a property of
+ *                   cipher algorithm. E.g. cblock size is 64 bits
+ *                   for DES, 128 bits for AES, etc.
+ *
+ * atomic cipher     A cipher algorithm, which requires some chunks of
+ * algorithm         text to be padded at left and(or) right sides before
+ *                   cipher transaform.
+ *
+ *
+ * block (atom)      Minimal chunk of file's data, which doesn't require
+ *                   padding. We'll consider logical units in a file of
+ *                   block size (atom size).
+ *
+ * cipher algorithm  Atomic cipher algorithm, which requires the last
+ * with EOF issue    incomplete cblock in a file to be padded with some
+ *                   data (usually zeros).
+ *
+ *
+ * operation, which  reading/writing from offset, which is not aligned to
+ * forms a gap at    to atom size
+ * the beginning
+ *
+ *
+ * operation, which  reading/writing count bytes starting from offset off,
+ * forms a gap at    so that off+count is not aligned to atom_size
+ * the end
+ *
+ * head block        the first atom affected by an operation, which forms
+ *                   a gap at the beginning, or(and) at the end.
+ *                   Сomment. Head block has at least one gap (either at
+ *                   the beginning, or at the end)
+ *
+ *
+ * tail block        the last atom different from head, affected by an
+ *                   operation, which forms a gap at the end.
+ *                   Сomment: Tail block has exactly one gap (at the end).
+ *
+ *
+ * partial block     head or tail block
+ *
+ *
+ * full block        block without gaps.
+ *
+ *
+ * (*) Recommendation for Block Cipher Modes of Operation
+ *     Methods and Techniques
+ *     NIST Special Publication 800-38A Edition 2001
+ */
+
+/*
+ * atom->offset_at()
+ */
+static off_t offset_at_head(struct avec_config *conf)
+{
+	return conf->aligned_offset;
+}
+
+static off_t offset_at_hole_head(call_frame_t *frame,
+					 struct object_cipher_info *object)
+{
+	return offset_at_head(get_hole_conf(frame));
+}
+
+static off_t offset_at_data_head(call_frame_t *frame,
+					 struct object_cipher_info *object)
+{
+	return offset_at_head(get_data_conf(frame));
+}
+
+
+static off_t offset_at_tail(struct avec_config *conf,
+				    struct object_cipher_info *object)
+{
+	return conf->aligned_offset +
+		(conf->off_in_head ? get_atom_size(object) : 0) +
+		(conf->nr_full_blocks << get_atom_bits(object));
+}
+
+static off_t offset_at_hole_tail(call_frame_t *frame,
+					 struct object_cipher_info *object)
+{
+	return offset_at_tail(get_hole_conf(frame), object);
+}
+
+
+static off_t offset_at_data_tail(call_frame_t *frame,
+					 struct object_cipher_info *object)
+{
+	return offset_at_tail(get_data_conf(frame), object);
+}
+
+static off_t offset_at_full(struct avec_config *conf,
+				    struct object_cipher_info *object)
+{
+	return conf->aligned_offset +
+		(conf->off_in_head ? get_atom_size(object) : 0);
+}
+
+static off_t offset_at_data_full(call_frame_t *frame,
+					 struct object_cipher_info *object)
+{
+	return offset_at_full(get_data_conf(frame), object);
+}
+
+static off_t offset_at_hole_full(call_frame_t *frame,
+					 struct object_cipher_info *object)
+{
+	return offset_at_full(get_hole_conf(frame), object);
+}
+
+/*
+ * atom->io_size_nopad()
+ */
+
+static uint32_t io_size_nopad_head(struct avec_config *conf,
+				   struct object_cipher_info *object)
+{
+	uint32_t gap_at_beg;
+	uint32_t gap_at_end;
+
+	check_head_block(conf);
+
+	gap_at_beg = conf->off_in_head;
+
+	if (has_tail_block(conf) || has_full_blocks(conf) || conf->off_in_tail == 0 )
+		gap_at_end = 0;
+	else
+		gap_at_end = get_atom_size(object) - conf->off_in_tail;
+
+	return get_atom_size(object) - (gap_at_beg + gap_at_end);
+}
+
+static uint32_t io_size_nopad_tail(struct avec_config *conf,
+				   struct object_cipher_info *object)
+{
+	check_tail_block(conf);
+	return conf->off_in_tail;
+}
+
+static uint32_t io_size_nopad_full(struct avec_config *conf,
+				   struct object_cipher_info *object)
+{	
+	check_full_block(conf);
+	return get_atom_size(object);
+}
+
+static uint32_t io_size_nopad_data_head(call_frame_t *frame,
+					struct object_cipher_info *object)
+{
+	return io_size_nopad_head(get_data_conf(frame), object);
+}
+
+static uint32_t io_size_nopad_hole_head(call_frame_t *frame,
+					struct object_cipher_info *object)
+{
+	return io_size_nopad_head(get_hole_conf(frame), object);
+}
+
+static uint32_t io_size_nopad_data_tail(call_frame_t *frame,
+					struct object_cipher_info *object)
+{
+	return io_size_nopad_tail(get_data_conf(frame), object);
+}
+
+static uint32_t io_size_nopad_hole_tail(call_frame_t *frame,
+					struct object_cipher_info *object)
+{
+	return io_size_nopad_tail(get_hole_conf(frame), object);
+}
+
+static uint32_t io_size_nopad_data_full(call_frame_t *frame,
+					struct object_cipher_info *object)
+{
+	return io_size_nopad_full(get_data_conf(frame), object);
+}
+
+static uint32_t io_size_nopad_hole_full(call_frame_t *frame,
+					struct object_cipher_info *object)
+{
+	return io_size_nopad_full(get_hole_conf(frame), object);
+}
+
+static uint32_t offset_in_head(struct avec_config *conf)
+{
+	check_cursor_head(conf);
+
+	return conf->off_in_head;
+}
+
+static uint32_t offset_in_tail(call_frame_t *frame,
+			       struct object_cipher_info *object)
+{
+	return 0;
+}
+
+static uint32_t offset_in_full(struct avec_config *conf,
+			       struct object_cipher_info *object)
+{
+	check_cursor_full(conf);
+
+	if (has_head_block(conf))
+		return (conf->cursor - 1) << get_atom_bits(object);
+	else
+		return conf->cursor << get_atom_bits(object);
+}
+
+static uint32_t offset_in_data_head(call_frame_t *frame,
+				    struct object_cipher_info *object)
+{
+	return offset_in_head(get_data_conf(frame));
+}
+
+static uint32_t offset_in_hole_head(call_frame_t *frame,
+				    struct object_cipher_info *object)
+{
+	return offset_in_head(get_hole_conf(frame));
+}
+
+static uint32_t offset_in_data_full(call_frame_t *frame,
+				    struct object_cipher_info *object)
+{
+	return offset_in_full(get_data_conf(frame), object);
+}
+
+static uint32_t offset_in_hole_full(call_frame_t *frame,
+				    struct object_cipher_info *object)
+{
+	return offset_in_full(get_hole_conf(frame), object);
+}
+
+/*
+ * atom->rmw()
+ */
+/*
+ * Pre-conditions:
+ * @vec contains plain text of the latest
+ * version.
+ *
+ * Uptodate gaps of the @partial block with
+ * this plain text, encrypt the whole block
+ * and write the result to disk.
+ */
+static int32_t rmw_partial_block(call_frame_t *frame,
+				 void *cookie,
+				 xlator_t *this,
+				 int32_t op_ret,
+				 int32_t op_errno,
+				 struct iovec *vec,
+				 int32_t count,
+				 struct iatt *stbuf,
+				 struct iobref *iobref,
+				 struct rmw_atom *atom)
+{
+	size_t was_read = 0;
+	uint64_t file_size;
+	crypt_local_t *local = frame->local;
+	struct object_cipher_info *object = &local->info->cinfo;
+
+	struct iovec *partial = atom->get_iovec(frame, 0);
+	struct avec_config *conf = atom->get_config(frame);
+	end_writeback_handler_t end_writeback_partial_block;
+#if DEBUG_CRYPT
+	gf_boolean_t check_last_cblock = _gf_false;
+#endif
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret < 0)
+		goto exit;
+
+	file_size = local->cur_file_size;
+	was_read = op_ret;
+
+	if (atom->locality == HEAD_ATOM && conf->off_in_head) {
+		/*
+		 * head atom with a non-uptodate gap
+		 * at the beginning
+		 *
+		 * fill the gap with plain text of the
+		 * latest version. Convert a part of hole
+		 * (if any) to zeros.
+		 */
+		int32_t i;
+		int32_t copied = 0;
+		int32_t to_gap; /* amount of data needed to uptodate
+				   the gap at the beginning */
+#if 0
+		int32_t hole = 0; /* The part of the hole which
+				   * got in the head block */
+#endif /* 0 */
+		to_gap = conf->off_in_head;
+
+		if (was_read < to_gap) {
+			if (file_size >
+			    offset_at_head(conf) + was_read) {
+				/*
+				 * It is impossible to uptodate
+				 * head block: too few bytes have
+				 * been read from disk, so that
+				 * partial write is impossible.
+				 *
+				 * It could happen because of many
+				 * reasons: IO errors, (meta)data
+				 * corruption in the local file system,
+				 * etc.
+				 */
+				gf_log(this->name, GF_LOG_WARNING,
+				     "Can not uptodate a gap at the beginning");
+				local->op_ret = -1;
+				local->op_errno = EIO;
+				goto exit;
+			}
+#if 0
+			hole = to_gap - was_read;
+#endif /* 0 */
+			to_gap = was_read;
+		}
+		/*
+		 * uptodate the gap at the beginning
+		 */
+		for (i = 0; i < count && copied < to_gap; i++) {
+			int32_t to_copy;
+
+			to_copy = vec[i].iov_len;
+			if (to_copy > to_gap - copied)
+				to_copy = to_gap - copied;
+
+			memcpy(partial->iov_base, vec[i].iov_base, to_copy);
+			copied += to_copy;
+		}
+#if 0
+		/*
+		 * If possible, convert part of the
+		 * hole, which got in the head block
+		 */
+		ret = TRY_LOCK(&local->hole_lock);
+		if (!ret) {
+			if (local->hole_handled)
+				/*
+				 * already converted by
+				 * crypt_writev_cbk()
+				 */
+				UNLOCK(&local->hole_lock);
+			else {
+				/*
+				 * convert the part of the hole
+				 * which got in the head block
+				 * to zeros.
+				 *
+				 * Update the orig_offset to make
+				 * sure writev_cbk() won't care
+				 * about this part of the hole.
+				 *
+				 */
+				memset(partial->iov_base + to_gap, 0, hole);
+
+				conf->orig_offset -= hole;
+				conf->orig_size += hole;
+				UNLOCK(&local->hole_lock);
+			}
+		}
+		else /*
+		      * conversion is being performed
+		      * by crypt_writev_cbk()
+		      */
+			;
+#endif /* 0 */
+	}
+	if (atom->locality == TAIL_ATOM ||
+	    (!has_tail_block(conf) && conf->off_in_tail)) {
+		/*
+		 * tail atom, or head atom with a non-uptodate
+		 * gap at the end.
+		 *
+		 * fill the gap at the end of the block
+		 * with plain text of the latest version.
+		 * Pad the result, (if needed)
+		 */
+		int32_t i;
+		int32_t to_gap;
+		int copied;
+		off_t off_in_tail;
+		int32_t to_copy;
+
+		off_in_tail = conf->off_in_tail;
+		to_gap = conf->gap_in_tail;
+
+		if (to_gap && was_read < off_in_tail + to_gap) {
+			/*
+			 * It is impossible to uptodate
+			 * the gap at the end: too few bytes
+			 * have been read from disk, so that
+			 * partial write is impossible.
+			 *
+			 * It could happen because of many
+			 * reasons: IO errors, (meta)data
+			 * corruption in the local file system,
+			 * etc.
+			 */
+			gf_log(this->name, GF_LOG_WARNING,
+			       "Can not uptodate a gap at the end");
+			local->op_ret = -1;
+			local->op_errno = EIO;
+			goto exit;
+		}
+		/*
+		 * uptodate the gap at the end
+		 */
+		copied = 0;
+		to_copy = to_gap;
+		for(i = count - 1; i >= 0 && to_copy > 0; i--) {
+			uint32_t from_vec, off_in_vec;
+
+			off_in_vec = 0;
+			from_vec = vec[i].iov_len;
+			if (from_vec > to_copy) {
+				off_in_vec = from_vec - to_copy;
+				from_vec = to_copy;
+			}
+			memcpy(partial->iov_base +
+			       off_in_tail + to_gap - copied - from_vec,
+			       vec[i].iov_base + off_in_vec,
+			       from_vec);
+
+			gf_log(this->name, GF_LOG_DEBUG,
+			       "uptodate %d bytes at tail. Offset at target(source): %d(%d)",
+			       (int)from_vec,
+			       (int)off_in_tail + to_gap - copied - from_vec,
+			       (int)off_in_vec);
+
+			copied += from_vec;
+			to_copy -= from_vec;
+		}
+		partial->iov_len = off_in_tail + to_gap;
+
+		if (object_alg_should_pad(object)) {
+			int32_t resid = 0;
+			resid = partial->iov_len & (object_alg_blksize(object) - 1);
+			if (resid) {
+				/*
+				 * append a new EOF padding
+				 */
+				local->eof_padding_size =
+					object_alg_blksize(object) - resid;
+
+				gf_log(this->name, GF_LOG_DEBUG,
+				       "set padding size %d",
+				       local->eof_padding_size);
+
+				memset(partial->iov_base + partial->iov_len,
+				       1,
+				       local->eof_padding_size);
+				partial->iov_len += local->eof_padding_size;
+#if DEBUG_CRYPT
+				gf_log(this->name, GF_LOG_DEBUG,
+				       "pad cblock with %d zeros:",
+				       local->eof_padding_size);
+				dump_cblock(this,
+					    (unsigned char *)partial->iov_base +
+					    partial->iov_len - object_alg_blksize(object));
+				check_last_cblock = _gf_true;
+#endif
+			}
+		}
+	}
+	/*
+	 * encrypt the whole block
+	 */
+	encrypt_aligned_iov(object,
+			    partial,
+			    1,
+			    atom->offset_at(frame, object));
+#if DEBUG_CRYPT
+	if (check_last_cblock == _gf_true) {
+		gf_log(this->name, GF_LOG_DEBUG,
+		       "encrypt last cblock with offset %llu",
+		       (unsigned long long)atom->offset_at(frame, object));
+		dump_cblock(this, (unsigned char *)partial->iov_base +
+			    partial->iov_len - object_alg_blksize(object));
+	}
+#endif
+	set_local_io_params_writev(frame, object, atom,
+				   atom->offset_at(frame, object),
+				   iovec_get_size(partial, 1));
+	/*
+	 * write the whole block to disk
+	 */
+	end_writeback_partial_block = dispatch_end_writeback(local->fop);
+	conf->cursor ++;
+	STACK_WIND(frame,
+		   end_writeback_partial_block,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->writev,
+		   local->fd,
+		   partial,
+		   1,
+		   atom->offset_at(frame, object),
+		   local->flags,
+		   local->iobref_data,
+		   local->xdata);
+
+	gf_log("crypt", GF_LOG_DEBUG,
+	       "submit partial block: %d bytes from %d offset",
+	       (int)iovec_get_size(partial, 1),
+	       (int)atom->offset_at(frame, object));
+ exit:
+	return 0;
+}
+
+/*
+ * Perform a (read-)modify-write sequence.
+ * This should be performed only after approval
+ * of upper server-side manager, i.e. the caller
+ * needs to make sure this is his turn to rmw.
+ */
+void submit_partial(call_frame_t *frame,
+		    xlator_t *this,
+		    fd_t *fd,
+		    atom_locality_type ltype)
+{
+	int32_t ret;
+	dict_t *dict;
+	struct rmw_atom *atom;
+	crypt_local_t *local = frame->local;
+	struct object_cipher_info *object = &local->info->cinfo;
+
+	atom = atom_by_types(local->active_setup, ltype);
+	/*
+	 * To perform the "read" component of the read-modify-write
+	 * sequence the crypt translator does stack_wind to itself.
+	 * 
+	 * Pass current file size to crypt_readv()
+	 */
+	dict = dict_new();
+	if (!dict) {
+		/*
+		 * FIXME: Handle the error
+		 */
+		gf_log("crypt", GF_LOG_WARNING, "Can not alloc dict");
+		return;
+	}
+	ret = dict_set(dict,
+		       FSIZE_XATTR_PREFIX,
+		       data_from_uint64(local->cur_file_size));
+	if (ret) {
+		/*
+		 * FIXME: Handle the error
+		 */
+		dict_unref(dict);
+		gf_log("crypt", GF_LOG_WARNING, "Can not set dict");
+		goto exit;
+	}
+	STACK_WIND(frame,
+		   atom->rmw,
+		   this,
+		   this->fops->readv, /* crypt_readv */
+		   fd,
+		   atom->count_to_uptodate(frame, object), /* count */
+		   atom->offset_at(frame, object), /* offset to read from */
+		   0,
+		   dict);
+ exit:
+	dict_unref(dict);
+}
+
+/*
+ * submit blocks of FULL_ATOM type
+ */
+void submit_full(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+	struct object_cipher_info *object = &local->info->cinfo;
+	struct rmw_atom *atom = atom_by_types(local->active_setup, FULL_ATOM);
+	uint32_t count; /* total number of full blocks to submit */
+	uint32_t granularity; /* number of blocks to submit in one iteration */
+
+	uint64_t off_in_file; /* start offset in the file, bytes */
+	uint32_t off_in_atom; /* start offset in the atom, blocks */
+	uint32_t blocks_written = 0; /* blocks written for this submit */
+
+	struct avec_config *conf = atom->get_config(frame);
+	end_writeback_handler_t end_writeback_full_block;
+	/*
+	 * Write full blocks by groups of granularity size.
+	 */
+	end_writeback_full_block = dispatch_end_writeback(local->fop);
+
+	if (is_ordered_mode(frame)) {
+		uint32_t skip = has_head_block(conf) ? 1 : 0;
+		count = 1;
+		granularity = 1;
+		/*
+		 * calculate start offset using cursor value;
+		 * here we should take into accout head block,
+		 * which corresponds to cursor value 0.
+		 */
+		off_in_file = atom->offset_at(frame, object) +
+			((conf->cursor - skip) << get_atom_bits(object));
+		off_in_atom = conf->cursor - skip;
+	}
+	else {
+		/*
+		 * in parallel mode
+		 */
+		count = conf->nr_full_blocks;
+		granularity = MAX_IOVEC;
+		off_in_file = atom->offset_at(frame, object);
+		off_in_atom = 0;
+	}
+	while (count) {
+		uint32_t blocks_to_write = count;
+
+		if (blocks_to_write > granularity)
+			blocks_to_write = granularity;
+		if (conf->type == HOLE_ATOM)
+			/*
+			 * reset iovec before encryption
+			 */
+			memset(atom->get_iovec(frame, 0)->iov_base,
+			       0,
+			       get_atom_size(object));
+		/*
+		 * encrypt the group
+		 */
+		encrypt_aligned_iov(object,
+				    atom->get_iovec(frame,
+						    off_in_atom +
+						    blocks_written),
+				    blocks_to_write,
+				    off_in_file + (blocks_written <<
+						   get_atom_bits(object)));
+		
+		set_local_io_params_writev(frame, object, atom,
+		        off_in_file + (blocks_written << get_atom_bits(object)),
+			blocks_to_write <<  get_atom_bits(object));
+
+		conf->cursor += blocks_to_write;
+
+		STACK_WIND(frame,
+			   end_writeback_full_block,
+			   FIRST_CHILD(this),
+			   FIRST_CHILD(this)->fops->writev,
+			   local->fd,
+			   atom->get_iovec(frame, off_in_atom + blocks_written),
+			   blocks_to_write,
+			   off_in_file + (blocks_written << get_atom_bits(object)),
+			   local->flags,
+			   local->iobref_data ? local->iobref_data : local->iobref,
+			   local->xdata);
+
+		gf_log("crypt", GF_LOG_DEBUG, "submit %d full blocks from %d offset",
+		       blocks_to_write,
+		       (int)(off_in_file + (blocks_written << get_atom_bits(object))));
+
+		count -= blocks_to_write;
+		blocks_written += blocks_to_write;
+	}
+	return;
+}
+
+static int32_t rmw_data_head(call_frame_t *frame,
+			     void *cookie,
+			     xlator_t *this,
+			     int32_t op_ret,
+			     int32_t op_errno,
+			     struct iovec *vec,
+			     int32_t count,
+			     struct iatt *stbuf,
+			     struct iobref *iobref,
+			     dict_t *xdata)
+{
+	return rmw_partial_block(frame,
+				 cookie,
+				 this,
+				 op_ret,
+				 op_errno,
+				 vec,
+				 count,
+				 stbuf,
+				 iobref,
+				 atom_by_types(DATA_ATOM, HEAD_ATOM));
+}
+
+static int32_t rmw_data_tail(call_frame_t *frame,
+			     void *cookie,
+			     xlator_t *this,
+			     int32_t op_ret,
+			     int32_t op_errno,
+			     struct iovec *vec,
+			     int32_t count,
+			     struct iatt *stbuf,
+			     struct iobref *iobref,
+			     dict_t *xdata)
+{
+	return rmw_partial_block(frame,
+				 cookie,
+				 this,
+				 op_ret,
+				 op_errno,
+				 vec,
+				 count,
+				 stbuf,
+				 iobref,
+				 atom_by_types(DATA_ATOM, TAIL_ATOM));
+}
+
+static int32_t rmw_hole_head(call_frame_t *frame,
+			     void *cookie,
+			     xlator_t *this,
+			     int32_t op_ret,
+			     int32_t op_errno,
+			     struct iovec *vec,
+			     int32_t count,
+			     struct iatt *stbuf,
+			     struct iobref *iobref,
+			     dict_t *xdata)
+{
+	return rmw_partial_block(frame,
+				 cookie,
+				 this,
+				 op_ret,
+				 op_errno,
+				 vec,
+				 count,
+				 stbuf,
+				 iobref,
+				 atom_by_types(HOLE_ATOM, HEAD_ATOM));
+}
+
+static int32_t rmw_hole_tail(call_frame_t *frame,
+			     void *cookie,
+			     xlator_t *this,
+			     int32_t op_ret,
+			     int32_t op_errno,
+			     struct iovec *vec,
+			     int32_t count,
+			     struct iatt *stbuf,
+			     struct iobref *iobref,
+			     dict_t *xdata)
+{
+	return rmw_partial_block(frame,
+				 cookie,
+				 this,
+				 op_ret,
+				 op_errno,
+				 vec,
+				 count,
+				 stbuf,
+				 iobref,
+				 atom_by_types(HOLE_ATOM, TAIL_ATOM));
+}
+
+/*
+ * atom->count_to_uptodate()
+ */
+static uint32_t count_to_uptodate_head(struct avec_config *conf,
+				       struct object_cipher_info *object)
+{
+	if (conf->acount == 1 && conf->off_in_tail)
+		return get_atom_size(object);
+	else
+		/* there is no need to read the whole head block */
+		return conf->off_in_head;
+}
+
+static uint32_t count_to_uptodate_tail(struct avec_config *conf,
+				       struct object_cipher_info *object)
+{
+	/* we need to read the whole tail block */
+	return get_atom_size(object);
+}
+
+static uint32_t count_to_uptodate_data_head(call_frame_t *frame,
+					    struct object_cipher_info *object)
+{
+	return count_to_uptodate_head(get_data_conf(frame), object);
+}
+
+static uint32_t count_to_uptodate_data_tail(call_frame_t *frame,
+					    struct object_cipher_info *object)
+{
+	return count_to_uptodate_tail(get_data_conf(frame), object);
+}
+
+static uint32_t count_to_uptodate_hole_head(call_frame_t *frame,
+					    struct object_cipher_info *object)
+{
+	return count_to_uptodate_head(get_hole_conf(frame), object);
+}
+
+static uint32_t count_to_uptodate_hole_tail(call_frame_t *frame,
+					    struct object_cipher_info *object)
+{
+	return count_to_uptodate_tail(get_hole_conf(frame), object);
+}
+
+/* atom->get_config() */
+
+static struct avec_config *get_config_data(call_frame_t *frame)
+{
+	return &((crypt_local_t *)frame->local)->data_conf;
+}
+
+static struct avec_config *get_config_hole(call_frame_t *frame)
+{
+	return &((crypt_local_t *)frame->local)->hole_conf;
+}
+
+/*
+ * atom->get_iovec()
+ */
+static struct iovec *get_iovec_hole_head(call_frame_t *frame,
+					 uint32_t count)
+{
+	struct avec_config *conf = get_hole_conf(frame);
+
+	return conf->avec;
+}
+
+static struct iovec *get_iovec_hole_full(call_frame_t *frame,
+					 uint32_t count)
+{
+	struct avec_config *conf = get_hole_conf(frame);
+
+	return conf->avec + (conf->off_in_head ? 1 : 0);
+}
+
+static inline struct iovec *get_iovec_hole_tail(call_frame_t *frame,
+						uint32_t count)
+{
+	struct avec_config *conf = get_hole_conf(frame);
+
+	return conf->avec + (conf->blocks_in_pool - 1);
+}
+
+static inline struct iovec *get_iovec_data_head(call_frame_t *frame,
+						uint32_t count)
+{
+	struct avec_config *conf = get_data_conf(frame);
+
+	return conf->avec;
+}
+
+static inline struct iovec *get_iovec_data_full(call_frame_t *frame,
+						uint32_t count)
+{
+	struct avec_config *conf = get_data_conf(frame);
+
+	return conf->avec + (conf->off_in_head ? 1 : 0) + count;
+}
+
+static inline struct iovec *get_iovec_data_tail(call_frame_t *frame,
+						uint32_t count)
+{
+	struct avec_config *conf = get_data_conf(frame);
+
+	return conf->avec +
+		(conf->off_in_head ? 1 : 0) +
+		conf->nr_full_blocks;
+}
+
+static struct rmw_atom atoms[LAST_DATA_TYPE][LAST_LOCALITY_TYPE] = {
+	[DATA_ATOM][HEAD_ATOM] =
+	{ .locality = HEAD_ATOM,
+	  .rmw = rmw_data_head,
+	  .offset_at = offset_at_data_head,
+	  .offset_in = offset_in_data_head,
+	  .get_iovec = get_iovec_data_head,
+	  .io_size_nopad = io_size_nopad_data_head,
+	  .count_to_uptodate = count_to_uptodate_data_head,
+	  .get_config = get_config_data
+	},
+	[DATA_ATOM][TAIL_ATOM] =
+	{ .locality = TAIL_ATOM,
+	  .rmw = rmw_data_tail,
+	  .offset_at = offset_at_data_tail,
+	  .offset_in = offset_in_tail,
+	  .get_iovec = get_iovec_data_tail,
+	  .io_size_nopad = io_size_nopad_data_tail,
+	  .count_to_uptodate = count_to_uptodate_data_tail,
+	  .get_config = get_config_data
+	},
+	[DATA_ATOM][FULL_ATOM] =
+	{ .locality = FULL_ATOM,
+	  .offset_at = offset_at_data_full,
+	  .offset_in = offset_in_data_full,
+	  .get_iovec = get_iovec_data_full,
+	  .io_size_nopad = io_size_nopad_data_full,
+	  .get_config = get_config_data
+	},
+	[HOLE_ATOM][HEAD_ATOM] =
+	{ .locality = HEAD_ATOM,
+	  .rmw = rmw_hole_head,
+	  .offset_at = offset_at_hole_head,
+	  .offset_in = offset_in_hole_head,
+	  .get_iovec = get_iovec_hole_head,
+	  .io_size_nopad = io_size_nopad_hole_head,
+	  .count_to_uptodate = count_to_uptodate_hole_head,
+	  .get_config = get_config_hole
+	},
+	[HOLE_ATOM][TAIL_ATOM] =
+	{ .locality = TAIL_ATOM,
+	  .rmw = rmw_hole_tail,
+	  .offset_at = offset_at_hole_tail,
+	  .offset_in = offset_in_tail,
+	  .get_iovec = get_iovec_hole_tail,
+	  .io_size_nopad = io_size_nopad_hole_tail,
+	  .count_to_uptodate = count_to_uptodate_hole_tail,
+	  .get_config = get_config_hole
+	},
+	[HOLE_ATOM][FULL_ATOM] =
+	{ .locality = FULL_ATOM,
+	  .offset_at = offset_at_hole_full,
+	  .offset_in = offset_in_hole_full,
+	  .get_iovec = get_iovec_hole_full,
+	  .io_size_nopad = io_size_nopad_hole_full,
+	  .get_config = get_config_hole
+	}
+};
+
+struct rmw_atom *atom_by_types(atom_data_type data,
+			       atom_locality_type locality)
+{
+	return &atoms[data][locality];
+}
+
+/*
+  Local variables:
+  c-indentation-style: "K&R"
+  mode-name: "LC"
+  c-basic-offset: 8
+  tab-width: 8
+  fill-column: 80
+  scroll-step: 1
+  End:
+*/
diff --git a/xlators/encryption/crypt/src/crypt-common.h b/xlators/encryption/crypt/src/crypt-common.h
new file mode 100644
index 000000000..7c212ad5d
--- /dev/null
+++ b/xlators/encryption/crypt/src/crypt-common.h
@@ -0,0 +1,141 @@
+/*
+  Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef __CRYPT_COMMON_H__
+#define __CRYPT_COMMON_H__
+
+#define INVAL_SUBVERSION_NUMBER (0xff)
+#define CRYPT_INVAL_OP   (GF_FOP_NULL)
+
+#define CRYPTO_FORMAT_PREFIX  "trusted.glusterfs.crypt.att.cfmt"
+#define FSIZE_XATTR_PREFIX    "trusted.glusterfs.crypt.att.size"
+#define SUBREQ_PREFIX         "trusted.glusterfs.crypt.msg.sreq"
+#define FSIZE_MSG_PREFIX      "trusted.glusterfs.crypt.msg.size"
+#define DE_MSG_PREFIX         "trusted.glusterfs.crypt.msg.dent"
+#define REQUEST_ID_PREFIX     "trusted.glusterfs.crypt.msg.rqid"
+#define MSGFLAGS_PREFIX       "trusted.glusterfs.crypt.msg.xfgs"
+
+
+/* messages for crypt_open() */
+#define MSGFLAGS_REQUEST_MTD_RLOCK   1 /* take read lock and don't unlock */
+#define MSGFLAGS_REQUEST_MTD_WLOCK   2 /* take write lock and don't unlock */
+
+#define AES_BLOCK_BITS (4) /* AES_BLOCK_SIZE == 1 << AES_BLOCK_BITS */
+
+#define noop   do {; } while (0)
+#define cassert(cond) ({ switch (-1) { case (cond): case 0: break; } })
+#define __round_mask(x, y) ((__typeof__(x))((y)-1))
+#define round_up(x, y) ((((x)-1) | __round_mask(x, y))+1)
+
+/*
+ * Format of file's metadata
+ */
+struct crypt_format {
+	uint8_t loader_id;  /* version of metadata loader */
+	uint8_t versioned[0]; /* file's metadata of specific version */
+} __attribute__((packed));
+
+typedef enum {
+	AES_CIPHER_ALG,
+	LAST_CIPHER_ALG
+} cipher_alg_t;
+
+typedef enum {
+	XTS_CIPHER_MODE,
+	LAST_CIPHER_MODE
+} cipher_mode_t;
+
+typedef enum {
+	MTD_LOADER_V1,
+	LAST_MTD_LOADER
+} mtd_loader_id;
+
+static inline void msgflags_set_mtd_rlock(uint32_t *flags)
+{
+	*flags |= MSGFLAGS_REQUEST_MTD_RLOCK;
+}
+
+static inline void msgflags_set_mtd_wlock(uint32_t *flags)
+{
+	*flags |= MSGFLAGS_REQUEST_MTD_WLOCK;
+}
+
+static inline gf_boolean_t msgflags_check_mtd_rlock(uint32_t *flags)
+{
+	return *flags & MSGFLAGS_REQUEST_MTD_RLOCK;
+}
+
+static inline gf_boolean_t msgflags_check_mtd_wlock(uint32_t *flags)
+{
+	return *flags & MSGFLAGS_REQUEST_MTD_WLOCK;
+}
+
+static inline gf_boolean_t msgflags_check_mtd_lock(uint32_t *flags)
+{
+	return  msgflags_check_mtd_rlock(flags) ||
+		msgflags_check_mtd_wlock(flags);
+}
+
+/*
+ * returns number of logical blocks occupied
+ * (maybe partially) by @count bytes
+ * at offset @start.
+ */
+static inline off_t logical_blocks_occupied(uint64_t start, off_t count,
+					    int blkbits)
+{
+	return ((start + count - 1) >> blkbits) - (start >> blkbits) + 1;
+}
+
+/*
+ * are two bytes (represented by offsets @off1
+ * and @off2 respectively) in the same logical
+ * block.
+ */
+static inline int in_same_lblock(uint64_t off1, uint64_t off2,
+				 int blkbits)
+{
+	return off1 >> blkbits == off2 >> blkbits;
+}
+
+static inline void dump_cblock(xlator_t *this, unsigned char *buf)
+{
+	gf_log(this->name, GF_LOG_DEBUG,
+	       "dump cblock: %x %x %x %x %x %x %x %x %x %x %x %x %x %x %x %x",
+	       (buf)[0],
+	       (buf)[1],
+	       (buf)[2],
+	       (buf)[3],
+	       (buf)[4],
+	       (buf)[5],
+	       (buf)[6],
+	       (buf)[7],
+	       (buf)[8],
+	       (buf)[9],
+	       (buf)[10],
+	       (buf)[11],
+	       (buf)[12],
+	       (buf)[13],
+	       (buf)[14],
+	       (buf)[15]);
+}
+
+#endif /* __CRYPT_COMMON_H__ */
+
+/*
+  Local variables:
+  c-indentation-style: "K&R"
+  mode-name: "LC"
+  c-basic-offset: 8
+  tab-width: 8
+  fill-column: 80
+  scroll-step: 1
+  End:
+*/
diff --git a/xlators/encryption/crypt/src/crypt-mem-types.h b/xlators/encryption/crypt/src/crypt-mem-types.h
new file mode 100644
index 000000000..2eab921fc
--- /dev/null
+++ b/xlators/encryption/crypt/src/crypt-mem-types.h
@@ -0,0 +1,44 @@
+/*
+  Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef __CRYPT_MEM_TYPES_H__
+#define __CRYPT_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_crypt_mem_types_ {
+	gf_crypt_mt_priv = gf_common_mt_end + 1,
+	gf_crypt_mt_inode,
+	gf_crypt_mt_data,
+	gf_crypt_mt_mtd,
+	gf_crypt_mt_loc,
+	gf_crypt_mt_iatt,
+	gf_crypt_mt_key,
+	gf_crypt_mt_iovec,
+	gf_crypt_mt_char,
+	gf_crypt_mt_end,
+};
+
+#endif /* __CRYPT_MEM_TYPES_H__ */
+
+/*
+  Local variables:
+  c-indentation-style: "K&R"
+  mode-name: "LC"
+  c-basic-offset: 8
+  tab-width: 8
+  fill-column: 80
+  scroll-step: 1
+  End:
+*/
+
+
+
diff --git a/xlators/encryption/crypt/src/crypt.c b/xlators/encryption/crypt/src/crypt.c
new file mode 100644
index 000000000..1abdad31d
--- /dev/null
+++ b/xlators/encryption/crypt/src/crypt.c
@@ -0,0 +1,4522 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+#include <ctype.h>
+#include <sys/uio.h>
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "xlator.h"
+#include "logging.h"
+#include "defaults.h"
+
+#include "crypt-common.h"
+#include "crypt.h"
+
+static void init_inode_info_head(struct crypt_inode_info *info, fd_t *fd);
+static int32_t init_inode_info_tail(struct crypt_inode_info *info,
+				    struct master_cipher_info *master);
+static int32_t prepare_for_submit_hole(call_frame_t *frame, xlator_t *this,
+				       uint64_t from, off_t size);
+static int32_t load_file_size(call_frame_t *frame, void *cookie,
+			      xlator_t *this, int32_t op_ret, int32_t op_errno,
+			      dict_t *dict, dict_t *xdata);
+static void do_ordered_submit(call_frame_t *frame, xlator_t *this,
+			      atom_data_type dtype);
+static void do_parallel_submit(call_frame_t *frame, xlator_t *this,
+			       atom_data_type dtype);
+static void put_one_call_open(call_frame_t *frame);
+static void put_one_call_readv(call_frame_t *frame, xlator_t *this);
+static void put_one_call_writev(call_frame_t *frame, xlator_t *this);
+static void put_one_call_ftruncate(call_frame_t *frame, xlator_t *this);
+static void free_avec(struct iovec *avec, char **pool, int blocks_in_pool);
+static void free_avec_data(crypt_local_t *local);
+static void free_avec_hole(crypt_local_t *local);
+
+static crypt_local_t *crypt_alloc_local(call_frame_t *frame, xlator_t *this,
+					glusterfs_fop_t fop)
+{
+	crypt_local_t *local = NULL;
+
+	local = mem_get0(this->local_pool);
+	if (!local) {
+		gf_log(this->name, GF_LOG_ERROR, "out of memory");
+		return NULL;
+	}
+	local->fop = fop;
+	LOCK_INIT(&local->hole_lock);
+	LOCK_INIT(&local->call_lock);
+	LOCK_INIT(&local->rw_count_lock);
+
+	frame->local = local;
+	return local;
+}
+
+struct crypt_inode_info *get_crypt_inode_info(inode_t *inode, xlator_t *this)
+{
+	int ret;
+	uint64_t value = 0;
+	struct crypt_inode_info *info;
+
+	ret = inode_ctx_get(inode, this, &value);
+	if (ret == -1) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"Can not get inode info");
+		return NULL;
+	}
+	info = (struct crypt_inode_info *)(long)value;
+	if (info == NULL) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"Can not obtain inode info");
+		return NULL;
+	}
+	return info;
+}
+
+static struct crypt_inode_info *local_get_inode_info(crypt_local_t *local,
+						      xlator_t *this)
+{
+	if (local->info)
+		return local->info;
+	local->info = get_crypt_inode_info(local->fd->inode, this);
+	return local->info;
+}
+
+static struct crypt_inode_info *alloc_inode_info(crypt_local_t *local,
+						 loc_t *loc)
+{
+	struct crypt_inode_info *info;
+
+	info = GF_CALLOC(1, sizeof(*info), gf_crypt_mt_inode);
+	if (!info) {
+		local->op_ret = -1;
+		local->op_errno = ENOMEM;
+		gf_log ("crypt", GF_LOG_WARNING,
+			"Can not allocate inode info");
+		return NULL;
+	}
+	memset(info, 0, sizeof(*info));
+#if DEBUG_CRYPT
+	info->loc = GF_CALLOC(1, sizeof(*loc), gf_crypt_mt_loc);
+	if (!info->loc) {
+		gf_log("crypt", GF_LOG_WARNING, "Can not allocate loc");
+		GF_FREE(info);
+		return NULL;
+	}
+	if (loc_copy(info->loc, loc)){
+		GF_FREE(info->loc);
+		GF_FREE(info);
+		return NULL;
+	}
+#endif /* DEBUG_CRYPT */
+
+	local->info = info;
+	return info;
+}
+
+static void free_inode_info(struct crypt_inode_info *info)
+{
+#if DEBUG_CRYPT
+	loc_wipe(info->loc);
+	GF_FREE(info->loc);
+#endif
+	memset(info, 0, sizeof(*info));
+	GF_FREE(info);
+}
+
+int crypt_forget (xlator_t *this, inode_t *inode)
+{
+        uint64_t ctx_addr = 0;
+        if (!inode_ctx_del (inode, this, &ctx_addr))
+                free_inode_info((struct crypt_inode_info *)(long)ctx_addr);
+        return 0;
+}
+
+#if DEBUG_CRYPT
+static void check_read(call_frame_t *frame, xlator_t *this, int32_t read,
+		       struct iovec *vec, int32_t count, struct iatt *stbuf)
+{
+	crypt_local_t *local = frame->local;
+	struct object_cipher_info *object = get_object_cinfo(local->info);
+	struct avec_config *conf = &local->data_conf;
+	uint32_t resid = stbuf->ia_size & (object_alg_blksize(object) - 1);
+
+	if (read <= 0)
+		return;
+	if (read != iovec_get_size(vec, count))
+		gf_log ("crypt", GF_LOG_DEBUG,
+			"op_ret differs from amount of read bytes");
+
+	if (object_alg_should_pad(object) && (read & (object_alg_blksize(object) - 1)))
+		gf_log ("crypt", GF_LOG_DEBUG,
+			"bad amount of read bytes (!= 0 mod(cblock size))");
+
+	if (conf->aligned_offset + read >
+	    stbuf->ia_size + (resid ? object_alg_blksize(object) - resid : 0))
+		gf_log ("crypt", GF_LOG_DEBUG,
+			"bad amount of read bytes (too large))");
+
+}
+
+#define PT_BYTES_TO_DUMP (32)
+static void dump_plain_text(crypt_local_t *local, struct iovec *avec)
+{
+	int32_t to_dump;
+	char str[PT_BYTES_TO_DUMP + 1];
+
+	if (!avec)
+		return;
+	to_dump = avec->iov_len;
+	if (to_dump > PT_BYTES_TO_DUMP)
+		to_dump = PT_BYTES_TO_DUMP;
+	memcpy(str, avec->iov_base, to_dump);
+	memset(str + to_dump, '0', 1);
+	gf_log("crypt", GF_LOG_DEBUG, "Read file: %s", str);
+}
+
+static int32_t data_conf_invariant(struct avec_config *conf)
+{
+	return conf->acount ==
+		!!has_head_block(conf) +
+		!!has_tail_block(conf)+
+		conf->nr_full_blocks;
+}
+
+static int32_t hole_conf_invariant(struct avec_config *conf)
+{
+	return conf->blocks_in_pool ==
+		!!has_head_block(conf) +
+		!!has_tail_block(conf)+
+		!!has_full_blocks(conf);
+}
+
+static void crypt_check_conf(struct avec_config *conf)
+{
+	int32_t ret = 0;
+	const char *msg;
+
+	switch (conf->type) {
+	case DATA_ATOM:
+		msg = "data";
+		ret = data_conf_invariant(conf);
+		break;
+	case HOLE_ATOM:
+		msg = "hole";
+		ret = hole_conf_invariant(conf);
+		break;
+	default:
+		msg = "unknown";
+	}
+	if (!ret)
+		gf_log("crypt", GF_LOG_DEBUG, "bad %s conf", msg);
+}
+
+static void check_buf(call_frame_t *frame, xlator_t *this, struct iatt *buf)
+{
+	crypt_local_t *local = frame->local;
+	struct object_cipher_info *object = &local->info->cinfo;
+	uint64_t local_file_size;
+
+	switch(local->fop) {
+	case GF_FOP_FTRUNCATE:
+		return;
+	case GF_FOP_WRITE:
+		local_file_size = local->new_file_size;
+		break;
+	case GF_FOP_READ:
+		if (parent_is_crypt_xlator(frame, this))
+			return;
+		local_file_size = local->cur_file_size;
+		break;
+	default:
+		gf_log("crypt", GF_LOG_DEBUG, "bad file operation");
+		return;
+	}
+	if (buf->ia_size != round_up(local_file_size,
+					 object_alg_blksize(object)))
+		gf_log("crypt", GF_LOG_DEBUG,
+		       "bad ia_size in buf (%llu), should be %llu",
+		       (unsigned long long)buf->ia_size,
+		       (unsigned long long)round_up(local_file_size,
+						    object_alg_blksize(object)));
+}
+
+#else
+#define check_read(frame, this, op_ret, vec, count, stbuf) noop
+#define dump_plain_text(local, avec) noop
+#define crypt_check_conf(conf) noop
+#define check_buf(frame, this, buf) noop
+#endif /* DEBUG_CRYPT */
+
+/*
+ * Pre-conditions:
+ * @vec represents a ciphertext of expanded size and
+ * aligned offset.
+ *
+ * Compound a temporal vector @avec with block-aligned
+ * components, decrypt and fix it up to represent a chunk
+ * of data corresponding to the original size and offset.
+ * Pass the result to the next translator.
+ */
+int32_t crypt_readv_cbk(call_frame_t *frame,
+			void *cookie,
+			xlator_t *this,
+			int32_t op_ret,
+			int32_t op_errno,
+			struct iovec *vec,
+			int32_t count,
+			struct iatt *stbuf,
+			struct iobref *iobref,
+			dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	struct avec_config *conf = &local->data_conf;
+	struct object_cipher_info *object = &local->info->cinfo;
+
+	struct iovec *avec;
+	uint32_t i;
+	uint32_t to_vec;
+	uint32_t to_user;
+
+	check_buf(frame, this, stbuf);
+	check_read(frame, this, op_ret, vec, count, stbuf);
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+	local->iobref = iobref_ref(iobref);
+
+	local->buf = *stbuf;
+	local->buf.ia_size = local->cur_file_size;
+
+	if (op_ret <= 0 || count == 0 || vec[0].iov_len == 0)
+		goto put_one_call;
+
+	if (conf->orig_offset >= local->cur_file_size) {
+		local->op_ret = 0;
+		goto put_one_call;
+	}
+	/* 
+	 * correct config params with real file size
+	 * and actual amount of bytes read
+	 */
+	set_config_offsets(frame, this,
+			   conf->orig_offset, op_ret, DATA_ATOM, 0);
+
+	if (conf->orig_offset + conf->orig_size > local->cur_file_size)
+		conf->orig_size = local->cur_file_size - conf->orig_offset;
+	/*
+	 * calculate amount of data to be returned
+	 * to user.
+	 */
+	to_user = op_ret;
+	if (conf->aligned_offset + to_user <= conf->orig_offset) {
+		gf_log(this->name, GF_LOG_WARNING, "Incomplete read");
+		local->op_ret = -1;
+		local->op_errno = EIO;
+		goto put_one_call;
+	}
+	to_user -= (conf->aligned_offset - conf->orig_offset);
+
+	if (to_user > conf->orig_size)
+		to_user = conf->orig_size;
+	local->rw_count = to_user;
+
+	op_errno = set_config_avec_data(this, local,
+					conf, object, vec, count);
+	if (op_errno) {
+		local->op_ret = -1;
+		local->op_errno = op_errno;
+		goto put_one_call;
+	}
+	avec = conf->avec;
+#if DEBUG_CRYPT
+	if (conf->off_in_tail != 0 &&
+	    conf->off_in_tail < object_alg_blksize(object) &&
+	    object_alg_should_pad(object))
+		gf_log(this->name, GF_LOG_DEBUG, "Bad offset in tail %d",
+		       conf->off_in_tail);
+	if (iovec_get_size(vec, count) != 0 &&
+	    in_same_lblock(conf->orig_offset + iovec_get_size(vec, count) - 1,
+			   local->cur_file_size - 1,
+			   object_alg_blkbits(object))) {
+		gf_log(this->name, GF_LOG_DEBUG, "Compound last cblock");
+		dump_cblock(this,
+			    (unsigned char *)(avec[conf->acount - 1].iov_base) +
+			    avec[conf->acount - 1].iov_len - object_alg_blksize(object));
+		dump_cblock(this,
+			    (unsigned char *)(vec[count - 1].iov_base) +
+			    vec[count - 1].iov_len - object_alg_blksize(object));
+	}
+#endif
+	decrypt_aligned_iov(object, avec,
+			    conf->acount, conf->aligned_offset);
+	/*
+	 * pass proper plain data to user
+	 */
+	avec[0].iov_base += (conf->aligned_offset - conf->orig_offset);
+	avec[0].iov_len  -= (conf->aligned_offset - conf->orig_offset);
+
+	to_vec = to_user;
+	for (i = 0; i < conf->acount; i++) {
+		if (avec[i].iov_len > to_vec)
+			avec[i].iov_len = to_vec;
+		to_vec -= avec[i].iov_len;
+	}
+ put_one_call:
+	put_one_call_readv(frame, this);
+	return 0;
+}
+
+static int32_t do_readv(call_frame_t *frame,
+			void *cookie,
+			xlator_t *this,
+			int32_t op_ret,
+			int32_t op_errno,
+			dict_t *dict,
+			dict_t *xdata)
+{
+	data_t *data;
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0)
+		goto error;	
+	/*
+	 * extract regular file size
+	 */
+	data = dict_get(dict, FSIZE_XATTR_PREFIX);
+	if (!data) {
+		gf_log("crypt", GF_LOG_WARNING, "Regular file size not found");
+		op_errno = EIO;
+		goto error;
+	}
+	local->cur_file_size = data_to_uint64(data);
+	
+	get_one_call(frame);
+	STACK_WIND(frame,
+		   crypt_readv_cbk,
+		   FIRST_CHILD (this),
+		   FIRST_CHILD (this)->fops->readv,
+		   local->fd,
+		   /*
+		    * FIXME: read amount can be reduced
+		    */
+		   local->data_conf.expanded_size,
+		   local->data_conf.aligned_offset,
+		   local->flags,
+		   local->xdata);
+	return 0;
+ error:
+	local->op_ret = -1;
+	local->op_errno = op_errno;
+
+	get_one_call(frame);
+	put_one_call_readv(frame, this);
+	return 0;
+}
+
+static int32_t crypt_readv_finodelk_cbk(call_frame_t *frame,
+					void *cookie,
+					xlator_t *this,
+					int32_t op_ret,
+					int32_t op_errno,
+					dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0)
+		goto error;
+	/*
+	 * An access has been granted,
+	 * retrieve file size
+	 */
+	STACK_WIND(frame,
+		   do_readv,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fgetxattr,
+		   local->fd,
+		   FSIZE_XATTR_PREFIX,
+		   NULL);
+	return  0;
+ error:
+	fd_unref(local->fd);
+	if (local->xdata)
+		dict_unref(local->xdata);
+	STACK_UNWIND_STRICT(readv,
+			    frame,
+			    -1,
+			    op_errno,
+			    NULL,
+			    0,
+			    NULL,
+			    NULL,
+			    NULL);
+	return 0;
+}
+
+static int32_t readv_trivial_completion(call_frame_t *frame,
+					void *cookie,
+					xlator_t *this,
+					int32_t op_ret,
+					int32_t op_errno,
+					struct iatt *buf,
+					dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret < 0) {
+		gf_log(this->name, GF_LOG_WARNING,
+		       "stat failed (%d)", op_errno);
+		goto error;
+	}
+	local->buf = *buf;
+	STACK_WIND(frame,
+		   load_file_size,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->getxattr,
+		   local->loc,
+		   FSIZE_XATTR_PREFIX,
+		   NULL);
+	return 0;	
+ error:
+	STACK_UNWIND_STRICT(readv, frame, op_ret, op_errno,
+			    NULL, 0, NULL, NULL, NULL);
+	return 0;
+}
+
+int32_t crypt_readv(call_frame_t *frame,
+		    xlator_t *this,
+		    fd_t *fd,
+		    size_t size,
+		    off_t offset,
+		    uint32_t flags, dict_t *xdata)
+{
+	int32_t ret;
+	crypt_local_t *local;
+	struct crypt_inode_info *info;
+	struct gf_flock lock = {0, };
+
+#if DEBUG_CRYPT
+	gf_log("crypt", GF_LOG_DEBUG, "reading %d bytes from offset %llu",
+	       (int)size, (long long)offset);
+	if (parent_is_crypt_xlator(frame, this))
+		gf_log("crypt", GF_LOG_DEBUG, "parent is crypt");
+#endif	
+	local = crypt_alloc_local(frame, this, GF_FOP_READ);
+	if (!local) {
+		ret = ENOMEM;
+		goto error;
+	}
+	if (size == 0)
+		goto trivial;
+
+	local->fd = fd_ref(fd);
+	local->flags = flags;
+
+	info = local_get_inode_info(local, this);
+	if (info == NULL) {
+		ret = EINVAL;
+		fd_unref(fd);
+		goto error;
+	}
+	if (!object_alg_atomic(&info->cinfo)) {
+		ret = EINVAL;
+		fd_unref(fd);
+		goto error;
+	}
+	set_config_offsets(frame, this, offset, size,
+			   DATA_ATOM, 0);
+	if (parent_is_crypt_xlator(frame, this)) {
+		data_t *data;
+		/*
+		 * We are called by crypt_writev (or cypt_ftruncate)
+		 * to perform the "read" component of the read-modify-write
+		 * (or read-prune-write) sequence for some atom;
+		 *
+		 * don't ask for access:
+		 * it has already been acquired
+		 *
+		 * Retrieve current file size
+		 */
+		if (!xdata) {
+			gf_log("crypt", GF_LOG_WARNING,
+			       "Regular file size hasn't been passed");
+			ret = EIO;
+			goto error;
+		}
+		data = dict_get(xdata, FSIZE_XATTR_PREFIX);
+		if (!data) {
+			gf_log("crypt", GF_LOG_WARNING,
+			       "Regular file size not found");
+			ret = EIO;
+			goto error;
+		}
+		local->old_file_size =
+			local->cur_file_size = data_to_uint64(data);
+
+		get_one_call(frame);
+		STACK_WIND(frame,
+			   crypt_readv_cbk,
+			   FIRST_CHILD(this),
+			   FIRST_CHILD(this)->fops->readv,
+			   local->fd,
+			   /*
+			    * FIXME: read amount can be reduced
+			    */
+			   local->data_conf.expanded_size,
+			   local->data_conf.aligned_offset,
+			   flags,
+			   NULL);
+		return 0;
+	}
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+
+	lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = F_RDLCK;
+        lock.l_whence = SEEK_SET;
+
+	STACK_WIND(frame,
+		   crypt_readv_finodelk_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);
+	return 0;
+ trivial:
+	STACK_WIND(frame,
+		   readv_trivial_completion,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fstat,
+		   fd,
+		   NULL);
+	return 0;
+ error:
+	STACK_UNWIND_STRICT(readv,
+			    frame,
+			    -1,
+			    ret,
+			    NULL,
+			    0,
+			    NULL,
+			    NULL,
+			    NULL);
+	return 0;
+}
+
+void set_local_io_params_writev(call_frame_t *frame,
+				struct object_cipher_info *object,
+				struct rmw_atom *atom,
+				off_t io_offset,
+				uint32_t io_size)
+{
+	crypt_local_t *local = frame->local;
+
+	local->io_offset = io_offset;
+	local->io_size = io_size;
+
+	local->io_offset_nopad =
+		atom->offset_at(frame, object) + atom->offset_in(frame, object);
+
+	gf_log("crypt", GF_LOG_DEBUG,
+	       "set nopad offset to %llu",
+	       (unsigned long long)local->io_offset_nopad);
+
+	local->io_size_nopad = atom->io_size_nopad(frame, object);
+
+	gf_log("crypt", GF_LOG_DEBUG,
+	       "set nopad size to %llu",
+	       (unsigned long long)local->io_size_nopad);
+
+	local->update_disk_file_size = 0;
+	/*
+	 * NOTE: eof_padding_size is 0 for all full atoms;
+	 * For head and tail atoms it will be set up at rmw_partial block()
+	 */
+	local->new_file_size = local->cur_file_size;
+
+	if (local->io_offset_nopad + local->io_size_nopad > local->cur_file_size) {
+
+		local->new_file_size = local->io_offset_nopad + local->io_size_nopad;
+
+		gf_log("crypt", GF_LOG_DEBUG,
+		       "set new file size to %llu",
+		       (unsigned long long)local->new_file_size);
+
+		local->update_disk_file_size = 1;
+	}
+}
+
+void set_local_io_params_ftruncate(call_frame_t *frame,
+				   struct object_cipher_info *object)
+{
+	uint32_t resid;
+	crypt_local_t *local = frame->local;
+	struct avec_config *conf = &local->data_conf;
+
+	resid = conf->orig_offset & (object_alg_blksize(object) - 1);
+	if (resid) {
+		local->eof_padding_size =
+			object_alg_blksize(object) - resid;
+		local->new_file_size = conf->aligned_offset;
+		local->update_disk_file_size = 0;
+		/*
+		 * file size will be updated
+		 * in the ->writev() stack,
+		 * when submitting file tail
+		 */
+	}
+	else {	
+		local->eof_padding_size = 0;
+		local->new_file_size = conf->orig_offset;
+		local->update_disk_file_size = 1;
+		/*
+		 * file size will be updated
+		 * in this ->ftruncate stack
+		 */
+	}
+}
+
+static inline void submit_head(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+	submit_partial(frame, this, local->fd, HEAD_ATOM);
+}
+
+static inline void submit_tail(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+	submit_partial(frame, this, local->fd, TAIL_ATOM);
+}
+
+static void submit_hole(call_frame_t *frame, xlator_t *this)
+{
+	/*
+	 * hole conversion always means
+	 * appended write and goes in ordered fashion
+	 */
+	do_ordered_submit(frame, this, HOLE_ATOM);
+}
+
+static void submit_data(call_frame_t *frame, xlator_t *this)
+{
+	if (is_ordered_mode(frame)) {
+		do_ordered_submit(frame, this, DATA_ATOM);
+		return;
+	}
+	gf_log("crypt", GF_LOG_WARNING, "Bad submit mode");
+	get_nr_calls(frame, nr_calls_data(frame));
+	do_parallel_submit(frame, this, DATA_ATOM);
+	return;
+}
+
+/*
+ * heplers called by writev_cbk, fruncate_cbk in ordered mode
+ */
+
+static inline int32_t should_submit_hole(crypt_local_t *local)
+{
+	struct avec_config *conf = &local->hole_conf;
+
+	return conf->avec != NULL;
+}
+
+static inline int32_t should_resume_submit_hole(crypt_local_t *local)
+{
+	struct avec_config *conf = &local->hole_conf;
+
+	if (local->fop == GF_FOP_WRITE && has_tail_block(conf))
+		/*
+		 * Don't submit a part of hole, which
+		 * fits into a data block:
+ 		 * this part of hole will be converted
+		 * as a gap filled by zeros in data head
+		 * block.
+		 */
+		return conf->cursor < conf->acount - 1;
+	else
+		return conf->cursor < conf->acount;
+}
+
+static inline int32_t should_resume_submit_data(call_frame_t *frame)
+{
+	crypt_local_t *local = frame->local;
+	struct avec_config *conf = &local->data_conf;
+
+	if (is_ordered_mode(frame))
+		return conf->cursor < conf->acount;
+	/*
+	 * parallel writes
+	 */
+	return 0;
+}
+
+static inline int32_t should_submit_data_after_hole(crypt_local_t *local)
+{
+	return local->data_conf.avec != NULL;
+}
+
+static void update_local_file_params(call_frame_t *frame,
+				     xlator_t *this,
+				     struct iatt *prebuf,
+				     struct iatt *postbuf)
+{
+	crypt_local_t *local = frame->local;
+
+	check_buf(frame, this, postbuf);
+
+	local->prebuf = *prebuf;
+	local->postbuf = *postbuf;
+
+	local->prebuf.ia_size  = local->cur_file_size;
+	local->postbuf.ia_size = local->new_file_size;
+
+	local->cur_file_size = local->new_file_size;	
+}
+
+static int32_t end_writeback_writev(call_frame_t *frame,
+				    void *cookie,
+				    xlator_t *this,
+				    int32_t op_ret,
+				    int32_t op_errno,
+				    struct iatt *prebuf,
+				    struct iatt *postbuf,
+				    dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret <= 0) {
+		gf_log(this->name, GF_LOG_WARNING,
+		       "writev iteration failed");
+		goto put_one_call;
+	}
+	/*
+	 * op_ret includes paddings (atom's head, atom's tail and EOF)
+	 */
+	if (op_ret < local->io_size) {
+		gf_log(this->name, GF_LOG_WARNING,
+		       "Incomplete writev iteration");
+		goto put_one_call;
+	}
+	op_ret -= local->eof_padding_size;
+	local->op_ret = op_ret;
+
+	update_local_file_params(frame, this, prebuf, postbuf);
+
+	if (data_write_in_progress(local)) {
+
+		LOCK(&local->rw_count_lock);
+		local->rw_count += op_ret;
+		UNLOCK(&local->rw_count_lock);
+
+		if (should_resume_submit_data(frame))
+			submit_data(frame, this);
+	}
+	else {
+		/*
+		 * hole conversion is going on;
+		 * don't take into account written zeros
+		 */
+		if (should_resume_submit_hole(local))
+			submit_hole(frame, this);
+
+		else if (should_submit_data_after_hole(local))
+			submit_data(frame, this);
+	}
+ put_one_call:
+	put_one_call_writev(frame, this);
+	return 0;
+}
+
+#define crypt_writev_cbk end_writeback_writev
+
+#define HOLE_WRITE_CHUNK_BITS 12
+#define HOLE_WRITE_CHUNK_SIZE (1 << HOLE_WRITE_CHUNK_BITS)
+
+/*
+ * Convert hole of size @size at offset @off to
+ * zeros and prepare respective iovecs for submit.
+ * The hole lock should be held.
+ *
+ * Pre-conditions:
+ * @local->file_size is set and valid.
+ */
+int32_t prepare_for_submit_hole(call_frame_t *frame, xlator_t *this,
+				uint64_t off, off_t size)
+{
+	int32_t ret;
+	crypt_local_t *local = frame->local;
+	struct object_cipher_info *object = &local->info->cinfo;
+
+	set_config_offsets(frame, this, off, size, HOLE_ATOM, 1);
+
+	ret = set_config_avec_hole(this, local,
+				   &local->hole_conf, object, local->fop);
+	crypt_check_conf(&local->hole_conf);
+
+	return ret;
+}
+
+/*
+ * prepare for submit @count bytes at offset @from
+ */
+int32_t prepare_for_submit_data(call_frame_t *frame, xlator_t *this,
+				off_t from, int32_t size, struct iovec *vec,
+				int32_t vec_count, int32_t setup_gap)
+{
+	uint32_t ret;
+	crypt_local_t *local = frame->local;
+	struct object_cipher_info *object = &local->info->cinfo;
+
+	set_config_offsets(frame, this, from, size,
+			   DATA_ATOM, setup_gap);
+
+	ret = set_config_avec_data(this, local,
+				   &local->data_conf, object, vec, vec_count);
+	crypt_check_conf(&local->data_conf);
+
+	return ret;
+}
+
+static void free_avec(struct iovec *avec,
+		      char **pool, int blocks_in_pool)
+{
+	if (!avec)
+		return;
+	GF_FREE(pool);
+	GF_FREE(avec);
+}
+
+static void free_avec_data(crypt_local_t *local)
+{
+	return free_avec(local->data_conf.avec,
+			 local->data_conf.pool,
+			 local->data_conf.blocks_in_pool);
+}
+
+static void free_avec_hole(crypt_local_t *local)
+{
+	return free_avec(local->hole_conf.avec,
+			 local->hole_conf.pool,
+			 local->hole_conf.blocks_in_pool);
+}
+
+
+static void do_parallel_submit(call_frame_t *frame, xlator_t *this,
+			       atom_data_type dtype)
+{
+	crypt_local_t *local = frame->local;
+	struct avec_config *conf;
+
+	local->active_setup = dtype;
+	conf = conf_by_type(frame, dtype);
+
+	if (has_head_block(conf))
+		submit_head(frame, this);
+
+	if (has_full_blocks(conf))
+		submit_full(frame, this);
+
+	if (has_tail_block(conf))
+		submit_tail(frame, this);
+	return;
+}
+
+static void do_ordered_submit(call_frame_t *frame, xlator_t *this,
+			      atom_data_type dtype)
+{
+	crypt_local_t *local = frame->local;
+	struct avec_config *conf;
+
+	local->active_setup = dtype;
+	conf = conf_by_type(frame, dtype);
+
+	if (should_submit_head_block(conf)) {
+		get_one_call_nolock(frame);
+		submit_head(frame, this);
+	}
+	else if (should_submit_full_block(conf)) {
+		get_one_call_nolock(frame);
+		submit_full(frame, this);
+	}
+	else if (should_submit_tail_block(conf)) {
+		get_one_call_nolock(frame);
+		submit_tail(frame, this);
+	}
+	else
+		gf_log("crypt", GF_LOG_DEBUG,
+		       "nothing has been submitted in ordered mode");
+	return;
+}
+
+static int32_t do_writev(call_frame_t *frame,
+			 void *cookie,
+			 xlator_t *this,
+			 int32_t op_ret,
+			 int32_t op_errno,
+			 dict_t *dict,
+			 dict_t *xdata)
+{
+	data_t *data;
+	crypt_local_t *local = frame->local;
+	struct object_cipher_info *object = &local->info->cinfo;
+	/*
+	 * extract regular file size
+	 */
+	data = dict_get(dict, FSIZE_XATTR_PREFIX);
+	if (!data) {
+		gf_log("crypt", GF_LOG_WARNING, "Regular file size not found");
+		op_ret = -1;
+		op_errno = EIO;
+		goto error;
+	}
+	local->old_file_size = local->cur_file_size = data_to_uint64(data);
+
+	set_gap_at_end(frame, object, &local->data_conf, DATA_ATOM);
+
+	if (local->cur_file_size < local->data_conf.orig_offset) {
+		/*
+		 * Set up hole config
+		 */
+		op_errno = prepare_for_submit_hole(frame,
+			       this,
+			       local->cur_file_size,
+			       local->data_conf.orig_offset - local->cur_file_size);
+		if (op_errno) {
+			local->op_ret = -1;
+			local->op_errno = op_errno;
+			goto error;
+		}
+	}
+	if (should_submit_hole(local))
+		submit_hole(frame, this);
+	else 
+		submit_data(frame, this);
+	return 0;
+ error:
+	get_one_call_nolock(frame);
+	put_one_call_writev(frame, this);
+	return 0;
+}
+
+static int32_t crypt_writev_finodelk_cbk(call_frame_t *frame,
+					 void *cookie,
+					 xlator_t *this,
+					 int32_t op_ret,
+					 int32_t op_errno,
+					 dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret < 0)
+		goto error;
+	/*
+	 * An access has been granted,
+	 * retrieve file size first
+	 */
+	STACK_WIND(frame,
+		   do_writev,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fgetxattr,
+		   local->fd,
+		   FSIZE_XATTR_PREFIX,
+		   NULL);
+	return 0;
+ error:
+	get_one_call_nolock(frame);
+	put_one_call_writev(frame, this);
+	return 0;	
+}
+
+static int32_t writev_trivial_completion(call_frame_t *frame,
+					 void *cookie,
+					 xlator_t *this,
+					 int32_t op_ret,
+					 int32_t op_errno,
+					 struct iatt *buf,
+					 dict_t *dict)
+{
+	crypt_local_t *local = frame->local;
+	
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+	local->prebuf = *buf;
+	local->postbuf = *buf;
+
+	local->prebuf.ia_size = local->cur_file_size;
+	local->postbuf.ia_size = local->cur_file_size;
+
+	get_one_call(frame);
+	put_one_call_writev(frame, this);
+	return 0;
+}
+
+int crypt_writev(call_frame_t *frame,
+		 xlator_t *this,
+		 fd_t *fd,
+		 struct iovec *vec,
+		 int32_t count,
+		 off_t offset,
+		 uint32_t flags,
+		 struct iobref *iobref,
+		 dict_t *xdata)
+{
+	int32_t ret;
+	crypt_local_t *local;
+	struct crypt_inode_info *info;
+	struct gf_flock lock = {0, };
+#if DEBUG_CRYPT
+	gf_log ("crypt", GF_LOG_DEBUG, "writing %d bytes from offset %llu",
+		(int)iovec_get_size(vec, count), (long long)offset);
+#endif
+	local = crypt_alloc_local(frame, this, GF_FOP_WRITE);
+	if (!local) {
+		ret = ENOMEM;
+		goto error;
+	}
+	local->fd = fd_ref(fd);
+
+	if (iobref)
+		local->iobref = iobref_ref(iobref);
+	/*
+	 * to update real file size on the server
+	 */
+	local->xattr = dict_new();
+	if (!local->xattr) {
+		ret = ENOMEM;
+		goto error;
+	}
+	local->flags = flags;
+
+	info = local_get_inode_info(local, this);
+	if (info == NULL) {
+		ret = EINVAL;
+		goto error;
+	}
+	if (!object_alg_atomic(&info->cinfo)) {
+		ret = EINVAL;
+		goto error;
+	}
+	if (iovec_get_size(vec, count) == 0)
+		goto trivial;
+
+	ret = prepare_for_submit_data(frame, this, offset,
+				      iovec_get_size(vec, count),
+				      vec, count, 0 /* don't setup gup
+						       in tail: we don't
+						       know file size yet */);
+	if (ret)
+		goto error;
+
+	if (parent_is_crypt_xlator(frame, this)) {
+		data_t *data;
+		/*
+		 * we are called by shinking crypt_ftruncate(),
+		 * which doesn't perform hole conversion;
+		 *
+		 * don't ask for access:
+		 * it has already been acquired
+		 */
+
+		/*
+		 * extract file size
+		 */
+		if (!xdata) {
+			gf_log("crypt", GF_LOG_WARNING,
+			       "Regular file size hasn't been passed");
+			ret = EIO;
+			goto error;
+		}
+		data = dict_get(xdata, FSIZE_XATTR_PREFIX);
+		if (!data) {
+			gf_log("crypt", GF_LOG_WARNING,
+			       "Regular file size not found");
+			ret = EIO;
+			goto error;
+		}
+		local->old_file_size =
+			local->cur_file_size = data_to_uint64(data);
+
+		submit_data(frame, this);
+		return 0;
+	}
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+	/*
+	 * lock the file and retrieve its size
+	 */
+	lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = F_WRLCK;
+        lock.l_whence = SEEK_SET;
+
+	STACK_WIND(frame,
+		   crypt_writev_finodelk_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);
+	return 0;
+ trivial:
+	STACK_WIND(frame,
+		   writev_trivial_completion,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fstat,
+		   fd,
+		   NULL);
+	return 0;
+ error:
+	if (local && local->fd)
+		fd_unref(fd);
+	if (local && local->iobref)
+		iobref_unref(iobref);
+	if (local && local->xdata)
+		dict_unref(xdata);
+	if (local && local->xattr)
+		dict_unref(local->xattr);
+	if (local && local->info)
+		free_inode_info(local->info);
+
+	STACK_UNWIND_STRICT(writev, frame, -1, ret, NULL, NULL, NULL);
+	return 0;
+}
+
+int32_t prepare_for_prune(call_frame_t *frame, xlator_t *this, uint64_t offset)
+{
+	set_config_offsets(frame, this,
+			   offset,
+			   0, /* count */
+			   DATA_ATOM,
+			   0 /* since we prune, there is no
+				gap in tail to uptodate */);
+	return 0;
+}
+
+/*
+ * Finish the read-prune-modify sequence
+ *
+ * Can be invoked as
+ * 1) ->ftruncate_cbk() for cblock-aligned, or trivial prune
+ * 2) ->writev_cbk() for non-cblock-aligned prune
+ */
+
+static int32_t prune_complete(call_frame_t *frame,
+			      void *cookie,
+			      xlator_t *this,
+			      int32_t op_ret,
+			      int32_t op_errno,
+			      struct iatt *prebuf,
+			      struct iatt *postbuf,
+			      dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	update_local_file_params(frame, this, prebuf, postbuf);
+
+	put_one_call_ftruncate(frame, this);
+	return 0;
+}
+
+/*
+ * This is called as ->ftruncate_cbk()
+ *
+ * Perform the "write" component of the
+ * read-prune-write sequence.
+ *
+ * submuit the rest of the file
+ */
+static int32_t prune_submit_file_tail(call_frame_t *frame,
+				      void *cookie,
+				      xlator_t *this,
+				      int32_t op_ret,
+				      int32_t op_errno,
+				      struct iatt *prebuf,
+				      struct iatt *postbuf,
+				      dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	struct avec_config *conf = &local->data_conf;
+	dict_t *dict;
+
+	if (op_ret < 0)
+		goto put_one_call;
+
+	if (local->xdata) {
+		dict_unref(local->xdata);
+		local->xdata = NULL;
+	}
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+
+	dict = dict_new();
+	if (!dict) {
+		op_errno = ENOMEM;
+		goto error;
+	}
+
+	update_local_file_params(frame, this, prebuf, postbuf);
+	local->new_file_size = conf->orig_offset;
+
+	/*
+	 * The rest of the file is a partial block and, hence,
+	 * should be written via RMW sequence, so the crypt xlator
+	 * does STACK_WIND to itself.
+	 *
+	 * Pass current file size to crypt_writev()
+	 */
+	op_errno = dict_set(dict,
+			    FSIZE_XATTR_PREFIX,
+			    data_from_uint64(local->cur_file_size));
+	if (op_errno) {
+		gf_log("crypt", GF_LOG_WARNING,
+		       "can not set key to update file size");
+		dict_unref(dict);
+		goto error;
+	}
+	gf_log("crypt", GF_LOG_DEBUG,
+	       "passing current file size (%llu) to crypt_writev",
+	       (unsigned long long)local->cur_file_size);
+	/*
+	 * Padding will be filled with
+	 * zeros by rmw_partial_block()
+	 */
+	STACK_WIND(frame,
+		   prune_complete,
+		   this,
+		   this->fops->writev, /* crypt_writev */
+		   local->fd,
+		   &local->vec,
+		   1,
+		   conf->aligned_offset, /* offset to write from */
+		   0,
+		   local->iobref,
+		   dict);
+
+	dict_unref(dict);
+	return 0;
+ error:
+	local->op_ret = -1;
+	local->op_errno = op_errno;
+ put_one_call:
+	put_one_call_ftruncate(frame, this);
+	return 0;
+}
+
+/*
+ * This is called as a callback of ->writev() invoked in behalf
+ * of ftruncate(): it can be
+ * 1) ordered writes issued by hole conversion in the case of
+ *    expanded truncate, or
+ * 2) an rmw partial data block issued by non-cblock-aligned
+ * prune.
+ */
+int32_t end_writeback_ftruncate(call_frame_t *frame,
+				void *cookie,
+				xlator_t *this,
+				int32_t op_ret,
+				int32_t op_errno,
+				struct iatt *prebuf,
+				struct iatt *postbuf,
+				dict_t *xdata)
+{
+	crypt_local_t  *local = frame->local;
+	/*
+	 * if nothing has been written,
+	 * then it must be an error
+	 */
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret < 0)
+		goto put_one_call;
+
+	update_local_file_params(frame, this, prebuf, postbuf);
+
+	if (data_write_in_progress(local))
+		/* case (2) */
+		goto put_one_call;
+	/* case (1) */
+	if (should_resume_submit_hole(local))
+		submit_hole(frame, this);
+	/*
+	 * case of hole, when we should't resume
+	 */
+ put_one_call:
+	put_one_call_ftruncate(frame, this);
+	return 0;
+}
+
+/*
+ * Perform prune and write components of the
+ * read-prune-write sequence.
+ *
+ * Called as ->readv_cbk()
+ *
+ * Pre-conditions:
+ * @vec contains the latest atom of the file
+ * (plain text)
+ */
+static int32_t prune_write(call_frame_t *frame,
+			   void *cookie,
+			   xlator_t *this,
+			   int32_t op_ret,
+			   int32_t op_errno,
+			   struct iovec *vec,
+			   int32_t count,
+			   struct iatt *stbuf,
+			   struct iobref *iobref,
+			   dict_t *xdata)
+{
+	int32_t i;
+	size_t to_copy;
+	size_t copied = 0;
+	crypt_local_t *local = frame->local;
+	struct avec_config *conf = &local->data_conf;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+	if (op_ret == -1)
+		goto put_one_call;
+
+	/*
+	 * At first, uptodate head block
+	 */
+	if (iovec_get_size(vec, count) < conf->off_in_head) {
+		gf_log(this->name, GF_LOG_WARNING,
+		       "Failed to uptodate head block for prune");
+		local->op_ret = -1;
+		local->op_errno = EIO;
+		goto put_one_call;
+	}
+	local->vec.iov_len = conf->off_in_head;
+	local->vec.iov_base = GF_CALLOC(1, local->vec.iov_len,
+					gf_crypt_mt_data);
+
+	if (local->vec.iov_base == NULL) {
+	        gf_log(this->name, GF_LOG_WARNING,
+                       "Failed to calloc head block for prune");
+		local->op_ret = -1;
+		local->op_errno = ENOMEM;
+	        goto put_one_call;
+	}
+	for (i = 0; i < count; i++) {
+		to_copy = vec[i].iov_len;
+		if (to_copy > local->vec.iov_len - copied)
+			to_copy = local->vec.iov_len - copied;
+
+		memcpy((char *)local->vec.iov_base + copied,
+		       vec[i].iov_base,
+		       to_copy);
+		copied += to_copy;
+		if (copied == local->vec.iov_len)
+			break;
+	}
+	/*
+	 * perform prune with aligned offset
+	 * (i.e. at this step we prune a bit
+	 * more then it is needed
+	 */
+	STACK_WIND(frame,
+		   prune_submit_file_tail,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->ftruncate,
+		   local->fd,
+		   conf->aligned_offset,
+		   local->xdata);
+	return 0;
+ put_one_call:
+	put_one_call_ftruncate(frame, this);
+	return 0;
+}
+
+/*
+ * Perform a read-prune-write sequence
+ */
+int32_t read_prune_write(call_frame_t *frame, xlator_t *this)
+{
+	int32_t ret = 0;
+	dict_t *dict = NULL;
+	crypt_local_t  *local = frame->local;
+	struct avec_config *conf = &local->data_conf;
+	struct object_cipher_info *object = &local->info->cinfo;
+
+	set_local_io_params_ftruncate(frame, object);
+	get_one_call_nolock(frame);
+
+	if ((conf->orig_offset & (object_alg_blksize(object) - 1)) == 0) {
+		/*
+		 * cblock-aligned prune:
+		 * we don't need read and write components,
+		 * just cut file body
+		 */
+		gf_log("crypt", GF_LOG_DEBUG,
+		       "prune without RMW (at offset %llu",
+		       (unsigned long long)conf->orig_offset);
+
+		STACK_WIND(frame,
+			   prune_complete,
+			   FIRST_CHILD(this),
+			   FIRST_CHILD(this)->fops->ftruncate,
+			   local->fd,
+			   conf->orig_offset,
+			   local->xdata);
+		return 0;
+	}
+	gf_log("crypt", GF_LOG_DEBUG,
+	       "prune with RMW (at offset %llu",
+	       (unsigned long long)conf->orig_offset);	
+	/*
+	 * We are about to perform the "read" component of the
+	 * read-prune-write sequence. It means that we need to
+	 * read encrypted data from disk and decrypt it.
+	 * So, the crypt translator does STACK_WIND to itself.
+	 * 
+	 * Pass current file size to crypt_readv()
+	 
+	 */
+	dict = dict_new();
+	if (!dict) {
+		gf_log("crypt", GF_LOG_WARNING, "Can not alloc dict");
+		ret = ENOMEM;
+		goto exit;
+	}
+	ret = dict_set(dict,
+		       FSIZE_XATTR_PREFIX,
+		       data_from_uint64(local->cur_file_size));
+	if (ret) {
+		gf_log("crypt", GF_LOG_WARNING, "Can not set dict");
+		goto exit;
+	}
+	STACK_WIND(frame,
+		   prune_write,
+		   this,
+		   this->fops->readv, /* crypt_readv */
+		   local->fd,
+		   get_atom_size(object), /* bytes to read */
+		   conf->aligned_offset, /* offset to read from */
+		   0,
+		   dict);
+ exit:
+	if (dict)
+		dict_unref(dict);
+	return ret;
+}
+
+/*
+ * File prune is more complicated than expand.
+ * First we need to read the latest atom to not lose info
+ * needed for proper update. Also we need to make sure that
+ * every component of read-prune-write sequence leaves data
+ * consistent
+ *
+ * Non-cblock aligned prune is performed as read-prune-write
+ * sequence:
+ *
+ * 1) read the latest atom;
+ * 2) perform cblock-aligned prune
+ * 3) issue a write request for the end-of-file
+ */
+int32_t prune_file(call_frame_t *frame, xlator_t *this, uint64_t offset)
+{
+	int32_t ret;
+
+	ret = prepare_for_prune(frame, this, offset);
+	if (ret)
+		return ret;
+	return read_prune_write(frame, this);
+}
+
+int32_t expand_file(call_frame_t *frame, xlator_t *this,
+		    uint64_t offset)
+{
+	int32_t ret;
+	crypt_local_t *local = frame->local;
+
+	ret = prepare_for_submit_hole(frame, this,
+				      local->old_file_size,
+				      offset - local->old_file_size);
+	if (ret)
+		return ret;
+	submit_hole(frame, this);
+	return 0;
+}
+
+static int32_t ftruncate_trivial_completion(call_frame_t *frame,
+					    void *cookie,
+					    xlator_t *this,
+					    int32_t op_ret,
+					    int32_t op_errno,
+					    struct iatt *buf,
+					    dict_t *dict)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+	local->prebuf = *buf;
+	local->postbuf = *buf;
+
+	local->prebuf.ia_size = local->cur_file_size;
+	local->postbuf.ia_size = local->cur_file_size;
+
+	get_one_call(frame);
+	put_one_call_ftruncate(frame, this);
+	return 0;
+}
+
+static int32_t do_ftruncate(call_frame_t *frame,
+			    void *cookie,
+			    xlator_t *this,
+			    int32_t op_ret,
+			    int32_t op_errno,
+			    dict_t *dict,
+			    dict_t *xdata)
+{
+	data_t *data;
+	crypt_local_t *local = frame->local;
+
+	if (op_ret)
+		goto error;	
+	/*
+	 * extract regular file size
+	 */
+	data = dict_get(dict, FSIZE_XATTR_PREFIX);
+	if (!data) {
+		gf_log("crypt", GF_LOG_WARNING, "Regular file size not found");
+		op_errno = EIO;
+		goto error;
+	}
+	local->old_file_size = local->cur_file_size = data_to_uint64(data);
+
+	if (local->data_conf.orig_offset == local->cur_file_size) {
+#if DEBUG_CRYPT
+		gf_log("crypt", GF_LOG_DEBUG,
+		       "trivial ftruncate (current file size %llu)",
+		       (unsigned long long)local->cur_file_size);
+#endif
+		goto trivial;
+	}
+	else if (local->data_conf.orig_offset < local->cur_file_size) {
+#if DEBUG_CRYPT
+		gf_log("crypt", GF_LOG_DEBUG, "prune from %llu to %llu",
+		       (unsigned long long)local->cur_file_size,
+		       (unsigned long long)local->data_conf.orig_offset);
+#endif
+		op_errno = prune_file(frame,
+				      this,
+				      local->data_conf.orig_offset);
+	}
+	else {
+#if DEBUG_CRYPT
+		gf_log("crypt", GF_LOG_DEBUG, "expand from %llu to %llu",
+		       (unsigned long long)local->cur_file_size,
+		       (unsigned long long)local->data_conf.orig_offset);
+#endif
+		op_errno = expand_file(frame,
+				       this,
+				       local->data_conf.orig_offset);
+	}
+	if (op_errno)
+		goto error;
+	return 0;
+ trivial:
+	STACK_WIND(frame,
+		   ftruncate_trivial_completion,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fstat,
+		   local->fd,
+		   NULL);
+	return 0;
+ error:
+	/*
+	 * finish with ftruncate
+	 */
+	local->op_ret = -1;
+	local->op_errno = op_errno;
+
+	get_one_call_nolock(frame);
+	put_one_call_ftruncate(frame, this);
+	return 0;
+}
+
+static int32_t crypt_ftruncate_finodelk_cbk(call_frame_t *frame,
+					    void *cookie,
+					    xlator_t *this,
+					    int32_t op_ret,
+					    int32_t op_errno,
+					    dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret < 0)
+		goto error;
+	/*
+	 * An access has been granted,
+	 * retrieve file size first
+	 */
+	STACK_WIND(frame,
+		   do_ftruncate,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fgetxattr,
+		   local->fd,
+		   FSIZE_XATTR_PREFIX,
+		   NULL);
+	return 0;
+ error:
+	get_one_call_nolock(frame);
+	put_one_call_ftruncate(frame, this);
+	return 0;
+}
+
+/*
+ * ftruncate is performed in 2 steps:
+ * . recieve file size;
+ * . expand or prune file.
+ */
+static int32_t crypt_ftruncate(call_frame_t *frame,
+			       xlator_t *this,
+			       fd_t *fd,
+			       off_t offset,
+			       dict_t *xdata)
+{
+	int32_t ret;
+	crypt_local_t *local;
+	struct crypt_inode_info *info;
+	struct gf_flock lock = {0, };
+
+	local = crypt_alloc_local(frame, this, GF_FOP_FTRUNCATE);
+	if (!local) {
+		ret = ENOMEM;
+		goto error;
+	}
+	local->xattr = dict_new();
+	if (!local->xattr) {
+		ret = ENOMEM;
+		goto error;
+	}
+	local->fd = fd_ref(fd);
+	info = local_get_inode_info(local, this);
+	if (info == NULL) {
+		ret = EINVAL;
+		goto error;
+	}
+	if (!object_alg_atomic(&info->cinfo)) {
+		ret = EINVAL;
+		goto error;
+	}
+	local->data_conf.orig_offset = offset;
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+
+	lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = F_WRLCK;
+        lock.l_whence = SEEK_SET;
+
+	STACK_WIND(frame,
+		   crypt_ftruncate_finodelk_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);
+	return 0;
+ error:
+	if (local && local->fd)
+		fd_unref(fd);
+	if (local && local->xdata)
+		dict_unref(xdata);
+	if (local && local->xattr)
+		dict_unref(local->xattr);
+	if (local && local->info)
+		free_inode_info(local->info);
+
+	STACK_UNWIND_STRICT(ftruncate, frame, -1, ret, NULL, NULL, NULL);
+	return 0;
+}
+
+/* ->flush_cbk() */
+int32_t truncate_end(call_frame_t *frame,
+		     void *cookie,
+		     xlator_t *this,
+		     int32_t op_ret,
+		     int32_t op_errno,
+		     dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	STACK_UNWIND_STRICT(truncate,
+			    frame,
+			    op_ret,
+			    op_errno,
+			    &local->prebuf,
+			    &local->postbuf,
+			    local->xdata);
+	return 0;
+}
+
+/* ftruncate_cbk() */
+int32_t truncate_flush(call_frame_t *frame,
+		       void *cookie,
+		       xlator_t *this,
+		       int32_t op_ret,
+		       int32_t op_errno,
+		       struct iatt *prebuf,
+		       struct iatt *postbuf,
+		       dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	fd_t *fd = local->fd;
+	local->prebuf = *prebuf;
+	local->postbuf = *postbuf;
+
+	STACK_WIND(frame,
+		   truncate_end,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->flush,
+		   fd,
+		   NULL);
+	fd_unref(fd);
+	return 0;
+}
+
+/*
+ * is called as ->open_cbk()
+ */
+static int32_t truncate_begin(call_frame_t *frame,
+			      void *cookie,
+			      xlator_t *this,
+			      int32_t op_ret,
+			      int32_t op_errno,
+			      fd_t *fd,
+			      dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0) {
+		fd_unref(fd);
+		STACK_UNWIND_STRICT(truncate,
+				    frame,
+				    op_ret,
+				    op_errno, NULL, NULL, NULL);
+		return 0;
+	}
+	/*
+	 * crypt_truncate() is implemented via crypt_ftruncate(),
+	 * so the crypt xlator does STACK_WIND to itself here
+	 */
+	STACK_WIND(frame,
+		   truncate_flush,
+		   this,
+		   this->fops->ftruncate, /* crypt_ftruncate */
+		   fd,
+		   local->offset,
+		   NULL);
+	return 0;
+}
+
+/*
+ * crypt_truncate() is implemented via crypt_ftruncate() as a
+ * sequence crypt_open() - crypt_ftruncate() - truncate_flush()
+ */
+int32_t crypt_truncate(call_frame_t *frame,
+		       xlator_t *this,
+		       loc_t *loc,
+		       off_t offset,
+		       dict_t *xdata)
+{
+	fd_t *fd;
+	crypt_local_t *local;
+
+#if DEBUG_CRYPT
+	gf_log(this->name, GF_LOG_DEBUG,
+	       "truncate file %s at offset %llu",
+	       loc->path, (unsigned long long)offset);
+#endif
+	local = crypt_alloc_local(frame, this, GF_FOP_TRUNCATE);
+	if (!local)
+		goto error;
+
+	fd = fd_create(loc->inode, frame->root->pid);
+	if (!fd) {
+		gf_log(this->name, GF_LOG_ERROR, "Can not create fd");
+		goto error;
+	}
+	local->fd = fd;
+	local->offset = offset;
+	local->xdata = xdata;
+	STACK_WIND(frame,
+		   truncate_begin,
+		   this,
+		   this->fops->open, /* crypt_open() */
+		   loc,
+		   O_RDWR,
+		   fd,
+		   NULL);
+	return 0;
+ error:
+	STACK_UNWIND_STRICT(truncate, frame, -1, EINVAL, NULL, NULL, NULL);
+	return 0;
+}
+
+end_writeback_handler_t dispatch_end_writeback(glusterfs_fop_t fop)
+{
+	switch (fop) {
+	case GF_FOP_WRITE:
+		return end_writeback_writev;
+	case GF_FOP_FTRUNCATE:
+		return end_writeback_ftruncate;
+	default:
+		gf_log("crypt", GF_LOG_WARNING, "Bad wb operation %d", fop);
+		return NULL;
+	}
+}
+
+/*
+ * true, if the caller needs metadata string
+ */
+static int32_t is_custom_mtd(dict_t *xdata)
+{
+	data_t *data;
+	uint32_t flags;
+
+	if (!xdata)
+		return 0;
+
+	data = dict_get(xdata, MSGFLAGS_PREFIX);
+	if (!data)
+		return 0;
+	if (data->len != sizeof(uint32_t)) {
+		gf_log("crypt", GF_LOG_WARNING,
+		       "Bad msgflags size (%d)", data->len);
+		return -1;
+	}
+	flags = *((uint32_t *)data->data);
+	return msgflags_check_mtd_lock(&flags);
+}
+
+static int32_t crypt_open_done(call_frame_t *frame,
+			       void *cookie,
+			       xlator_t *this,
+			       int32_t op_ret,
+			       int32_t op_errno, dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+	if (op_ret < 0)
+		gf_log(this->name, GF_LOG_WARNING, "mtd unlock failed (%d)",
+		       op_errno);
+	put_one_call_open(frame);
+	return 0;
+}
+
+static void crypt_open_tail(call_frame_t *frame, xlator_t *this)
+{
+	struct gf_flock  lock  = {0, };
+	crypt_local_t *local = frame->local;
+
+        lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+	STACK_WIND(frame,
+		   crypt_open_done,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   local->fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);
+}
+
+/*
+ * load private inode info at open time
+ * called as ->fgetxattr_cbk()
+ */
+static int load_mtd_open(call_frame_t *frame,
+			 void *cookie,
+			 xlator_t *this,
+			 int32_t op_ret,
+			 int32_t op_errno,
+			 dict_t *dict,
+			 dict_t *xdata)
+{
+	int32_t ret;
+	gf_boolean_t upload_info;
+	data_t *mtd;
+	uint64_t value = 0;
+	struct crypt_inode_info *info;
+	crypt_local_t *local = frame->local;
+	crypt_private_t *priv = this->private;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (local->fd->inode->ia_type == IA_IFLNK)
+		goto exit;
+	if (op_ret < 0)
+		goto exit;
+	/*
+	 * first, check for cached info
+	 */
+	ret = inode_ctx_get(local->fd->inode, this, &value);
+	if (ret != -1) {
+		info = (struct crypt_inode_info *)(long)value;
+		if (info == NULL) {
+			gf_log(this->name, GF_LOG_WARNING,
+			       "Inode info expected, but not found");
+			local->op_ret = -1;
+			local->op_errno = EIO;
+			goto exit;
+		}
+		/*
+		 * info has been found in the cache
+		 */
+		upload_info = _gf_false;
+	}
+	else {
+		/*
+		 * info hasn't been found in the cache.
+		 */
+		info = alloc_inode_info(local, local->loc);
+		if (!info) {
+			local->op_ret = -1;
+			local->op_errno = ENOMEM;
+			goto exit;
+		}
+		init_inode_info_head(info, local->fd);
+		upload_info = _gf_true;
+	}
+	/*
+	 * extract metadata
+	 */
+	mtd = dict_get(dict, CRYPTO_FORMAT_PREFIX);
+	if (!mtd) {
+		local->op_ret = -1;
+		local->op_errno = ENOENT;
+		gf_log (this->name, GF_LOG_WARNING,
+			"Format string wasn't found");
+		goto exit;
+	}
+	/*
+	 * authenticate metadata against the path
+	 */	
+	ret = open_format((unsigned char *)mtd->data,
+			  mtd->len,
+			  local->loc,
+			  info,
+			  get_master_cinfo(priv),
+			  local,
+			  upload_info);
+	if (ret) {
+		local->op_ret = -1;
+		local->op_errno = ret;
+		goto exit;
+	}
+	if (upload_info) {
+		ret = init_inode_info_tail(info, get_master_cinfo(priv));
+		if (ret) {
+			local->op_ret = -1;
+			local->op_errno = ret;
+			goto exit;	
+		}
+		ret = inode_ctx_put(local->fd->inode,
+				    this, (uint64_t)(long)info);
+		if (ret == -1) {
+			local->op_ret = -1;
+			local->op_errno = EIO;
+			goto exit;
+		}
+	}
+	if (local->custom_mtd) {
+		/*
+		 * pass the metadata string to the customer
+		 */
+		ret = dict_set_static_bin(local->xdata,
+					  CRYPTO_FORMAT_PREFIX,
+					  mtd->data,
+					  mtd->len);
+		if (ret) {
+			local->op_ret = -1;
+			local->op_errno = ret;
+			goto exit;
+		}
+	}
+ exit:
+	if (!local->custom_mtd)
+		crypt_open_tail(frame, this);
+	else
+		put_one_call_open(frame);
+	return 0;
+}
+
+static int32_t crypt_open_finodelk_cbk(call_frame_t *frame,
+				       void *cookie,
+				       xlator_t *this,
+				       int32_t op_ret,
+				       int32_t op_errno,
+				       dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret < 0) {
+		gf_log(this->name, GF_LOG_WARNING, "finodelk (LOCK) failed");
+		goto exit;
+	}
+	STACK_WIND(frame,
+		   load_mtd_open,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fgetxattr,
+		   local->fd,
+		   CRYPTO_FORMAT_PREFIX,
+		   NULL);
+	return 0;
+ exit:
+	put_one_call_open(frame);
+	return 0;
+}
+
+/*
+ * verify metadata against the specified pathname
+ */
+static int32_t crypt_open_cbk(call_frame_t *frame,
+			      void *cookie,
+			      xlator_t *this,
+			      int32_t op_ret,
+			      int32_t op_errno,
+			      fd_t *fd,
+			      dict_t *xdata)
+{
+	struct gf_flock lock = {0, };
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (local->fd->inode->ia_type == IA_IFLNK)
+		goto exit;
+	if (op_ret < 0)
+		goto exit;
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+	else if (local->custom_mtd){
+		local->xdata = dict_new();
+		if (!local->xdata) {
+			local->op_ret = -1;
+			local->op_errno = ENOMEM;
+			gf_log ("crypt", GF_LOG_ERROR,
+				"Can not get new dict for mtd string");
+			goto exit;
+		}
+	}
+	lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = local->custom_mtd ? F_WRLCK : F_RDLCK;
+        lock.l_whence = SEEK_SET;
+
+	STACK_WIND(frame,
+		   crypt_open_finodelk_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);
+	return 0;
+ exit:
+	put_one_call_open(frame);
+	return 0;
+}
+
+static int32_t crypt_open(call_frame_t *frame,
+			  xlator_t *this,
+			  loc_t *loc,
+			  int32_t flags,
+			  fd_t *fd,
+			  dict_t *xdata)
+{
+	int32_t ret = ENOMEM;
+	crypt_local_t *local;
+
+	local = crypt_alloc_local(frame, this, GF_FOP_OPEN);
+	if (!local)
+		goto error;
+	local->loc = GF_CALLOC(1, sizeof(*loc), gf_crypt_mt_loc);
+	if (!local->loc) {
+		ret = ENOMEM;
+		goto error;
+	}
+	memset(local->loc, 0, sizeof(*local->loc));
+	ret = loc_copy(local->loc, loc);
+	if (ret) {
+		GF_FREE(local->loc);
+		goto error;
+	}
+	local->fd = fd_ref(fd);
+
+	ret = is_custom_mtd(xdata);
+	if (ret < 0) {
+		loc_wipe(local->loc);
+		GF_FREE(local->loc);
+		ret = EINVAL;
+		goto error;
+	}
+	local->custom_mtd = ret;
+
+	if ((flags & O_ACCMODE) == O_WRONLY)
+		/*
+		 * we can't open O_WRONLY, because
+		 * we need to do read-modify-write
+		 */
+		flags = (flags & ~O_ACCMODE) | O_RDWR;
+	/*
+	 * Make sure that out translated offsets
+	 * and counts won't be ignored
+	 */
+	flags &= ~O_APPEND;
+	get_one_call_nolock(frame);
+	STACK_WIND(frame,
+		   crypt_open_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->open,
+		   loc,
+		   flags,
+		   fd,
+		   xdata);
+	return 0;
+ error:
+	STACK_UNWIND_STRICT(open,
+			    frame,
+			    -1,
+			    ret,
+			    NULL,
+			    NULL);
+	return 0;
+}
+
+static int32_t init_inode_info_tail(struct crypt_inode_info *info,
+				    struct master_cipher_info *master)
+{
+	int32_t ret;
+	struct object_cipher_info *object = &info->cinfo;
+
+#if DEBUG_CRYPT
+	gf_log("crypt", GF_LOG_DEBUG, "Init inode info for object %s",
+	       uuid_utoa(info->oid));
+#endif
+	ret = data_cipher_algs[object->o_alg][object->o_mode].set_private(info,
+									master);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, "Set private info failed");
+		return ret;
+	}
+	return 0;
+}
+
+/*
+ * Init inode info at ->create() time
+ */
+static void init_inode_info_create(struct crypt_inode_info *info,
+				   struct master_cipher_info *master,
+				   data_t *data)
+{
+	struct object_cipher_info *object;
+
+	info->nr_minor = CRYPT_XLATOR_ID;
+	memcpy(info->oid, data->data, data->len);
+
+	object = &info->cinfo;
+
+	object->o_alg        = master->m_alg;
+	object->o_mode       = master->m_mode;
+	object->o_block_bits = master->m_block_bits;
+	object->o_dkey_size  = master->m_dkey_size;
+}
+
+static void init_inode_info_head(struct crypt_inode_info *info, fd_t *fd)
+{
+	memcpy(info->oid, fd->inode->gfid, sizeof(uuid_t));
+}
+
+static int32_t crypt_create_done(call_frame_t *frame,
+				 void *cookie,
+				 xlator_t *this,
+				 int32_t op_ret,
+				 int32_t op_errno, dict_t *xdata)
+{
+	crypt_private_t *priv = this->private;
+	crypt_local_t *local = frame->local;
+	struct crypt_inode_info *info = local->info;
+	fd_t *local_fd = local->fd;
+	dict_t *local_xdata = local->xdata;
+	inode_t *local_inode = local->inode;
+
+	if (op_ret < 0) {
+		free_inode_info(info);
+		goto unwind;
+	}
+	op_errno = init_inode_info_tail(info, get_master_cinfo(priv));
+	if (op_errno) {
+		op_ret = -1;
+		free_inode_info(info);
+		goto unwind;
+	}
+	/*
+	 * FIXME: drop major subversion number
+	 */
+	op_ret = inode_ctx_put(local->fd->inode, this, (uint64_t)(long)info);
+	if (op_ret == -1) {
+		op_errno = EIO;
+		free_inode_info(info);
+		goto unwind;
+	}
+ unwind:
+	free_format(local);
+	STACK_UNWIND_STRICT(create,
+			    frame,
+			    op_ret,
+			    op_errno,
+			    local_fd,
+			    local_inode,
+			    &local->buf,
+			    &local->prebuf,
+			    &local->postbuf,
+			    local_xdata);
+	fd_unref(local_fd);
+	inode_unref(local_inode);
+	if (local_xdata)
+		dict_unref(local_xdata);
+	return 0;
+}
+
+static int crypt_create_tail(call_frame_t *frame,
+			     void *cookie,
+			     xlator_t *this,
+			     int32_t op_ret,
+			     int32_t op_errno,
+			     dict_t *xdata)
+{
+	struct gf_flock  lock  = {0, };
+	crypt_local_t *local = frame->local;
+	fd_t *local_fd = local->fd;
+	dict_t *local_xdata = local->xdata;
+	inode_t *local_inode = local->inode;
+
+	dict_unref(local->xattr);
+
+	if (op_ret < 0)
+		goto error;
+
+        lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+	STACK_WIND(frame,
+		   crypt_create_done,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   local->fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);
+	return 0;
+ error:
+	free_inode_info(local->info);
+	free_format(local);
+
+	STACK_UNWIND_STRICT(create,
+			    frame,
+			    op_ret,
+			    op_errno,
+			    local_fd,
+			    local_inode,
+			    &local->buf,
+			    &local->prebuf,
+			    &local->postbuf,
+			    local_xdata);
+
+	fd_unref(local_fd);
+	inode_unref(local_inode);
+	if (local_xdata)
+		dict_unref(local_xdata);
+	return 0;
+}
+
+static int32_t crypt_create_finodelk_cbk(call_frame_t *frame,
+					 void *cookie,
+					 xlator_t *this,
+					 int32_t op_ret,
+					 int32_t op_errno,
+					 dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	struct crypt_inode_info *info = local->info;
+
+	if (op_ret < 0)
+		goto error;
+
+	STACK_WIND(frame,
+		   crypt_create_tail,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fsetxattr,
+		   local->fd,
+		   local->xattr, /* CRYPTO_FORMAT_PREFIX */
+		   0,
+		   NULL);
+	return 0;
+ error: 
+	free_inode_info(info);
+	free_format(local);
+	fd_unref(local->fd);
+	dict_unref(local->xattr);
+	if (local->xdata)
+		dict_unref(local->xdata);
+
+	STACK_UNWIND_STRICT(create,
+			    frame,
+			    op_ret,
+			    op_errno,
+			    NULL,
+			    NULL,
+			    NULL,
+			    NULL,
+			    NULL,
+			    NULL);
+	return 0;
+}
+
+/*
+ * Create and store crypt-specific format on disk;
+ * Populate cache with private inode info
+ */
+static int32_t crypt_create_cbk(call_frame_t *frame,
+				void *cookie,
+				xlator_t *this,
+				int32_t op_ret,
+				int32_t op_errno,
+				fd_t *fd,
+				inode_t *inode,
+				struct iatt *buf,
+				struct iatt *preparent,
+				struct iatt *postparent,
+				dict_t *xdata)
+{
+	struct gf_flock lock = {0, };
+	crypt_local_t *local = frame->local;
+	struct crypt_inode_info *info = local->info;
+
+	if (op_ret < 0)
+		goto error;
+	if (xdata)
+		local->xdata = dict_ref(xdata);	
+	local->inode = inode_ref(inode);
+	local->buf = *buf;
+	local->prebuf = *preparent;
+	local->postbuf = *postparent;
+
+        lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = F_WRLCK;
+        lock.l_whence = SEEK_SET;
+
+        STACK_WIND(frame,
+		   crypt_create_finodelk_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   local->fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);	
+	return 0;
+ error:
+	free_inode_info(info);
+	free_format(local);
+	fd_unref(local->fd);
+	dict_unref(local->xattr);
+
+	STACK_UNWIND_STRICT(create,
+			    frame,
+			    op_ret,
+			    op_errno,
+			    NULL, NULL, NULL,
+			    NULL, NULL, NULL);
+	return 0;
+}
+
+static int32_t crypt_create(call_frame_t *frame,
+			    xlator_t *this,
+			    loc_t *loc,
+			    int32_t flags,
+			    mode_t mode,
+			    mode_t umask,
+			    fd_t *fd,
+			    dict_t *xdata)
+{
+	int ret;
+	data_t *data;
+	crypt_local_t *local;
+	crypt_private_t *priv;
+	struct master_cipher_info *master;
+	struct crypt_inode_info *info;
+
+	priv = this->private;
+	master = get_master_cinfo(priv);
+
+	if (master_alg_atomic(master)) {
+		/*
+		 * We can't open O_WRONLY, because we
+		 * need to do read-modify-write.
+		 */
+		if ((flags & O_ACCMODE) == O_WRONLY)
+			flags = (flags & ~O_ACCMODE) | O_RDWR;
+		/*
+		 * Make sure that out translated offsets
+		 * and counts won't be ignored
+		 */
+		flags &= ~O_APPEND;
+	}
+	local = crypt_alloc_local(frame, this, GF_FOP_CREATE);
+	if (!local) {
+		ret = ENOMEM;
+		goto error;
+	}
+	data = dict_get(xdata, "gfid-req");
+	if (!data) {
+		ret = EINVAL;
+		gf_log("crypt", GF_LOG_WARNING, "gfid not found");
+		goto error;
+	}
+	if (data->len != sizeof(uuid_t)) {
+		ret = EINVAL;
+		gf_log("crypt", GF_LOG_WARNING,
+		       "bad gfid size (%d), should be %d",
+		       (int)data->len, (int)sizeof(uuid_t));
+		goto error;
+	}
+	info = alloc_inode_info(local, loc);
+	if (!info){
+		ret = ENOMEM;
+		goto error;
+	}
+	/*
+	 * NOTE:
+	 * format has to be created BEFORE
+	 * proceeding to the untrusted server
+	 */
+	ret = alloc_format_create(local);
+	if (ret) {
+		free_inode_info(info);
+		goto error;
+	}
+	init_inode_info_create(info, master, data);
+
+	ret = create_format(local->format,
+			    loc,
+			    info,
+			    master);
+	if (ret) {
+		free_inode_info(info);
+		goto error;		
+	}
+	local->xattr = dict_new();
+	if (!local->xattr) {
+		free_inode_info(info);
+		free_format(local);	
+		goto error;
+	}
+	ret = dict_set_static_bin(local->xattr,
+				  CRYPTO_FORMAT_PREFIX,
+				  local->format,
+				  new_format_size());
+	if (ret) {
+		dict_unref(local->xattr);
+		free_inode_info(info);
+		free_format(local);
+		goto error;
+	}
+	ret = dict_set(local->xattr, FSIZE_XATTR_PREFIX, data_from_uint64(0));
+	if (ret) {
+		dict_unref(local->xattr);
+		free_inode_info(info);
+		free_format(local);
+		goto error;
+	}
+	local->fd = fd_ref(fd);
+
+	STACK_WIND(frame,
+		   crypt_create_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->create,
+		   loc,
+		   flags,
+		   mode,
+		   umask,
+		   fd,
+		   xdata);
+	return 0;
+ error:
+	gf_log("crypt", GF_LOG_WARNING, "can not create file");
+	STACK_UNWIND_STRICT(create,
+			    frame,
+			    -1,
+			    ret,
+			    NULL, NULL, NULL,
+			    NULL, NULL, NULL);
+	return 0;
+}
+
+/*
+ * FIXME: this should depends on the version of format string
+ */
+static int32_t filter_crypt_xattr(dict_t *dict,
+				  char *key, data_t *value, void *data)
+{
+        dict_del(dict, key);
+        return 0;
+}
+
+static int32_t crypt_fsetxattr(call_frame_t *frame,
+			       xlator_t *this,
+			       fd_t *fd,
+			       dict_t *dict,
+			       int32_t flags, dict_t *xdata)
+{
+	dict_foreach_fnmatch(dict, "trusted.glusterfs.crypt*",
+			     filter_crypt_xattr, NULL);
+	STACK_WIND(frame,
+		   default_fsetxattr_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fsetxattr,
+		   fd,
+		   dict,
+		   flags,
+		   xdata);
+	return 0;
+}
+
+/*
+ * TBD: verify file metadata before wind
+ */
+static int32_t crypt_setxattr(call_frame_t *frame,
+			      xlator_t *this,
+			      loc_t *loc,
+			      dict_t *dict,
+			      int32_t flags, dict_t *xdata)
+{
+	dict_foreach_fnmatch(dict, "trusted.glusterfs.crypt*",
+			     filter_crypt_xattr, NULL);
+	STACK_WIND(frame,
+		   default_setxattr_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->setxattr,
+		   loc,
+		   dict,
+		   flags,
+		   xdata);
+	return 0;
+}
+
+/*
+ * called as flush_cbk()
+ */
+static int32_t linkop_end(call_frame_t *frame,
+			  void *cookie,
+			  xlator_t *this,
+			  int32_t op_ret,
+			  int32_t op_errno,
+			  dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	linkop_unwind_handler_t unwind_fn;
+	unwind_fn = linkop_unwind_dispatch(local->fop);
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret < 0 &&
+	    op_errno == ENOENT &&
+	    local->loc->inode->ia_type == IA_IFLNK) {
+		local->op_ret = 0;
+		local->op_errno = 0;
+	}
+	unwind_fn(frame);
+	return 0;
+}
+
+/*
+ * unpin inode on the server
+ */
+static int32_t link_flush(call_frame_t *frame,
+			  void *cookie,
+			  xlator_t *this,
+			  int32_t op_ret,
+			  int32_t op_errno,
+			  inode_t *inode,
+			  struct iatt *buf,
+			  struct iatt *preparent,
+			  struct iatt *postparent, dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0)
+		goto error;
+	if (local->xdata) {
+		dict_unref(local->xdata);
+		local->xdata = NULL;
+	}
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+	local->inode = inode_ref(inode);
+	local->buf = *buf;
+	local->prebuf = *preparent;
+	local->postbuf = *postparent;
+
+	STACK_WIND(frame,
+		   linkop_end,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->flush,
+		   local->fd,
+		   NULL);
+	return 0;
+ error:
+	local->op_ret = -1;
+	local->op_errno = op_errno;
+	link_unwind(frame);
+	return 0;
+}
+
+void link_unwind(call_frame_t *frame)
+{
+	crypt_local_t *local = frame->local;
+	dict_t *xdata;
+	dict_t *xattr;
+	inode_t *inode;
+
+	if (!local) {
+		STACK_UNWIND_STRICT(link,
+				    frame,
+				    -1,
+				    ENOMEM,
+				    NULL,
+				    NULL,
+				    NULL,
+				    NULL,
+				    NULL);
+		return;
+	}
+	xdata = local->xdata;
+	xattr = local->xattr;
+	inode = local->inode;
+
+	if (local->loc){
+		loc_wipe(local->loc);
+		GF_FREE(local->loc);
+	}
+	if (local->newloc) {
+		loc_wipe(local->newloc);
+		GF_FREE(local->newloc);
+	}
+	if (local->fd)
+		fd_unref(local->fd);
+	if (local->format)
+		GF_FREE(local->format);
+
+	STACK_UNWIND_STRICT(link,
+			    frame,
+			    local->op_ret,
+			    local->op_errno,
+			    inode,
+			    &local->buf,
+			    &local->prebuf,
+			    &local->postbuf,
+			    xdata);
+	if (xdata)
+		dict_unref(xdata);
+	if (xattr)
+		dict_unref(xattr);
+	if (inode)
+		inode_unref(inode);
+}
+
+void link_wind(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+
+	STACK_WIND(frame,
+		   link_flush,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->link,
+		   local->loc,
+		   local->newloc,
+		   local->xdata);
+}
+
+/*
+ * unlink()
+ */
+static int32_t unlink_flush(call_frame_t *frame,
+			    void *cookie,
+			    xlator_t *this,
+			    int32_t op_ret,
+			    int32_t op_errno,
+			    struct iatt *preparent,
+			    struct iatt *postparent, dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0)
+		goto error;
+	local->prebuf = *preparent;
+	local->postbuf = *postparent;
+	if (local->xdata) {
+		dict_unref(local->xdata);
+		local->xdata = NULL;
+	}
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+
+	STACK_WIND(frame,
+		   linkop_end,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->flush,
+		   local->fd,
+		   NULL);
+	return 0;
+ error:
+	local->op_ret = -1;
+	local->op_errno = op_errno;
+	unlink_unwind(frame);
+	return 0;
+}
+
+void unlink_unwind(call_frame_t *frame)
+{
+	crypt_local_t *local = frame->local;
+	dict_t *xdata;
+	dict_t *xattr;
+
+	if (!local) {
+		STACK_UNWIND_STRICT(unlink,
+				    frame,
+				    -1,
+				    ENOMEM,
+				    NULL,
+				    NULL,
+				    NULL);
+		return;
+	}
+	xdata = local->xdata;
+	xattr = local->xattr;
+	if (local->loc){
+		loc_wipe(local->loc);
+		GF_FREE(local->loc);
+	}
+	if (local->fd)
+		fd_unref(local->fd);
+	if (local->format)
+		GF_FREE(local->format);
+
+	STACK_UNWIND_STRICT(unlink,
+			    frame,
+			    local->op_ret,
+			    local->op_errno,
+			    &local->prebuf,
+			    &local->postbuf,
+			    xdata);
+	if (xdata)
+		dict_unref(xdata);
+	if (xattr)
+		dict_unref(xattr);
+}
+
+void unlink_wind(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+
+	STACK_WIND(frame,
+		   unlink_flush,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->unlink,
+		   local->loc,
+		   local->flags,
+		   local->xdata);
+}
+
+void rename_unwind(call_frame_t *frame)
+{
+	crypt_local_t *local = frame->local;
+	dict_t *xdata;
+	dict_t *xattr;
+	struct iatt *prenewparent;
+	struct iatt *postnewparent;
+
+	if (!local) {
+		STACK_UNWIND_STRICT(rename,
+				    frame,
+				    -1,
+				    ENOMEM,
+				    NULL,
+				    NULL,
+				    NULL,
+				    NULL,
+				    NULL,
+				    NULL);	
+		return;
+	}
+	xdata = local->xdata;
+	xattr = local->xattr;
+	prenewparent = local->prenewparent;
+	postnewparent = local->postnewparent;	
+
+	if (local->loc){
+		loc_wipe(local->loc);
+		GF_FREE(local->loc);
+	}
+	if (local->newloc){
+		loc_wipe(local->newloc);
+		GF_FREE(local->newloc);
+	}
+	if (local->fd)
+		fd_unref(local->fd);
+	if (local->format)
+		GF_FREE(local->format);
+
+	STACK_UNWIND_STRICT(rename,
+			    frame,
+			    local->op_ret,
+			    local->op_errno,
+			    &local->buf,
+			    &local->prebuf,
+			    &local->postbuf,
+			    prenewparent,
+			    postnewparent,
+			    xdata);
+	if (xdata)
+		dict_unref(xdata);
+	if (xattr)
+		dict_unref(xattr);
+	if (prenewparent)
+		GF_FREE(prenewparent);
+	if (postnewparent)
+		GF_FREE(postnewparent);
+}
+
+/*
+ * called as flush_cbk()
+ */
+static int32_t rename_end(call_frame_t *frame,
+			  void *cookie,
+			  xlator_t *this,
+			  int32_t op_ret,
+			  int32_t op_errno,
+			  dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	rename_unwind(frame);
+	return 0;
+}
+
+static int32_t rename_flush(call_frame_t *frame,
+			    void *cookie,
+			    xlator_t *this,
+			    int32_t op_ret,
+			    int32_t op_errno,
+			    struct iatt *buf,
+			    struct iatt *preoldparent,
+			    struct iatt *postoldparent,
+			    struct iatt *prenewparent,
+			    struct iatt *postnewparent,
+			    dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0)
+		goto error;
+	dict_unref(local->xdata);
+	local->xdata = NULL;
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+
+	local->buf = *buf;
+	local->prebuf = *preoldparent;
+	local->postbuf = *postoldparent;
+	if (prenewparent) {
+		local->prenewparent = GF_CALLOC(1, sizeof(*prenewparent),
+						gf_crypt_mt_iatt);
+		if (!local->prenewparent) {
+			op_errno = ENOMEM;
+			goto error;
+		}
+		*local->prenewparent = *prenewparent;
+	}
+	if (postnewparent) {
+		local->postnewparent = GF_CALLOC(1, sizeof(*postnewparent),
+						 gf_crypt_mt_iatt);
+		if (!local->postnewparent) {
+			op_errno = ENOMEM;
+			goto error;
+		}
+		*local->postnewparent = *postnewparent;
+	}
+	STACK_WIND(frame,
+		   rename_end,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->flush,
+		   local->fd,
+		   NULL);
+	return 0;
+ error:
+	local->op_ret = -1;
+	local->op_errno = op_errno;
+	rename_unwind(frame);
+	return 0;
+}
+
+void rename_wind(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+
+	STACK_WIND(frame,
+		   rename_flush,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->rename,
+		   local->loc,
+		   local->newloc,
+		   local->xdata);
+}
+
+static int32_t __do_linkop(call_frame_t *frame,
+			   void *cookie,
+			   xlator_t *this,
+			   int32_t op_ret,
+			   int32_t op_errno, dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	linkop_wind_handler_t wind_fn;
+	linkop_unwind_handler_t unwind_fn;
+
+	wind_fn = linkop_wind_dispatch(local->fop);
+	unwind_fn = linkop_unwind_dispatch(local->fop);
+
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if (op_ret >= 0)
+		wind_fn(frame, this);
+	else {
+		gf_log(this->name, GF_LOG_WARNING, "mtd unlock failed (%d)",
+		       op_errno);
+		unwind_fn(frame);
+	}
+	return 0;
+}
+
+static int32_t do_linkop(call_frame_t *frame,
+			 void *cookie,
+			 xlator_t *this,
+			 int32_t op_ret,
+			 int32_t op_errno,
+			 dict_t *xdata)
+{
+	struct gf_flock  lock  = {0, };
+	crypt_local_t *local = frame->local;
+	linkop_unwind_handler_t unwind_fn;
+
+	unwind_fn = linkop_unwind_dispatch(local->fop);
+	local->op_ret = op_ret;
+	local->op_errno = op_errno;
+
+	if(op_ret < 0)
+		goto error;
+
+	lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+	STACK_WIND(frame,
+		   __do_linkop,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   local->fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);
+	return 0;
+ error:		
+	unwind_fn(frame);
+	return 0;
+}
+
+/*
+ * Update the metadata string (against the new pathname);
+ * submit the result
+ */
+static int32_t linkop_begin(call_frame_t *frame,
+			    void *cookie,
+			    xlator_t *this,
+			    int32_t op_ret,
+			    int32_t op_errno,
+			    fd_t *fd,
+			    dict_t *xdata)
+{
+	gf_boolean_t upload_info;
+	crypt_local_t *local = frame->local;
+	crypt_private_t *priv = this->private;
+	struct crypt_inode_info *info;
+	data_t *old_mtd;
+	uint32_t new_mtd_size;
+	uint64_t value = 0;
+	void (*unwind_fn)(call_frame_t *frame);
+	void (*wind_fn)(call_frame_t *frame, xlator_t *this);
+	mtd_op_t mop;
+
+	wind_fn = linkop_wind_dispatch(local->fop);
+	unwind_fn = linkop_unwind_dispatch(local->fop);
+	mop = linkop_mtdop_dispatch(local->fop);
+
+	if (local->fd->inode->ia_type == IA_IFLNK)
+		goto wind;
+	if (op_ret < 0)
+		/*
+		 * verification failed
+		 */
+		goto error;
+
+	old_mtd = dict_get(xdata, CRYPTO_FORMAT_PREFIX);
+	if (!old_mtd) {
+		op_errno = EIO;
+		gf_log (this->name, GF_LOG_DEBUG,
+			"Metadata string wasn't found");
+		goto error;
+	}
+	new_mtd_size = format_size(mop, old_mtd->len);
+	op_errno = alloc_format(local, new_mtd_size);
+	if (op_errno)
+		goto error;
+	/*
+	 * check for cached info 
+	 */
+	op_ret = inode_ctx_get(fd->inode, this, &value);
+	if (op_ret != -1) {
+		info = (struct crypt_inode_info *)(long)value;
+		if (info == NULL) {
+			gf_log (this->name, GF_LOG_WARNING,
+				"Inode info was not found");
+			op_errno = EINVAL;
+			goto error;
+		}
+		/*
+		 * info was found in the cache
+		 */
+		local->info = info;
+		upload_info = _gf_false;
+	}
+	else {
+		/*
+		 * info wasn't found in the cache;
+		 */
+		info = alloc_inode_info(local, local->loc);
+		if (!info)
+			goto error;
+		init_inode_info_head(info, fd);
+		local->info = info;
+		upload_info = _gf_true;
+	}
+	op_errno = open_format((unsigned char *)old_mtd->data,
+			       old_mtd->len,
+			       local->loc,
+			       info,
+			       get_master_cinfo(priv),
+			       local,
+			       upload_info);
+	if (op_errno)
+		goto error;
+	if (upload_info == _gf_true) {
+		op_errno = init_inode_info_tail(info,
+						get_master_cinfo(priv));
+		if (op_errno)
+			goto error;
+		op_errno = inode_ctx_put(fd->inode, this,
+					 (uint64_t)(long)(info));
+		if (op_errno == -1) {
+			op_errno = EIO;
+			goto error;
+		}
+	}
+	/*
+	 * update the format string (append/update/cup a MAC)
+	 */
+	op_errno = update_format(local->format,
+				 (unsigned char *)old_mtd->data,
+				 old_mtd->len,
+				 local->mac_idx,
+				 mop,
+				 local->newloc,
+				 info,
+				 get_master_cinfo(priv),
+				 local);
+	if (op_errno)
+		goto error;
+	/*
+	 * store the new format string on the server
+	 */
+	if (new_mtd_size) {
+		op_errno = dict_set_static_bin(local->xattr,
+					       CRYPTO_FORMAT_PREFIX,
+					       local->format,
+					       new_mtd_size);
+		if (op_errno)
+			goto error;
+	}
+	STACK_WIND(frame,
+		   do_linkop,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->setxattr,
+		   local->loc,
+		   local->xattr,
+		   0,
+		   NULL);
+	return 0;
+ wind:
+	wind_fn(frame, this);
+	return 0;
+ error:
+	local->op_ret = -1;
+	local->op_errno = op_errno;
+	unwind_fn(frame);
+	return 0;
+}
+
+static int32_t linkop_grab_local(call_frame_t *frame,
+				 xlator_t *this,
+				 loc_t *oldloc,
+				 loc_t *newloc,
+				 int flags, dict_t *xdata,
+				 glusterfs_fop_t op)
+{
+	int32_t ret = ENOMEM;
+	fd_t *fd;
+	crypt_local_t *local;
+
+	local = crypt_alloc_local(frame, this, op);
+	if (!local)
+		goto error;
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+
+	fd = fd_create(oldloc->inode, frame->root->pid);
+	if (!fd) {
+		gf_log(this->name, GF_LOG_ERROR, "Can not create fd");
+		goto error;
+	}
+	local->fd = fd;
+	local->flags = flags;
+	local->loc = GF_CALLOC(1, sizeof(*oldloc), gf_crypt_mt_loc);
+	if (!local->loc)
+		goto error;
+	memset(local->loc, 0, sizeof(*local->loc));
+	ret = loc_copy(local->loc, oldloc);
+	if (ret) {
+		GF_FREE(local->loc);
+		local->loc = NULL;
+		goto error;
+	}
+	if (newloc) {
+		local->newloc = GF_CALLOC(1, sizeof(*newloc), gf_crypt_mt_loc);
+		if (!local->newloc) {
+			loc_wipe(local->loc);
+			GF_FREE(local->loc);
+			goto error;
+		}
+		memset(local->newloc, 0, sizeof(*local->newloc));
+		ret = loc_copy(local->newloc, newloc);
+		if (ret) {
+			loc_wipe(local->loc);
+			GF_FREE(local->loc);
+			GF_FREE(local->newloc);
+			goto error;
+		}
+	}
+	local->xattr = dict_new();
+	if (!local->xattr) {
+		gf_log(this->name, GF_LOG_ERROR, "Can not create dict");
+		ret = ENOMEM;
+		goto error;
+	}
+	return 0;
+
+error:
+        if (local) {
+                if (local->xdata)
+                        dict_unref(local->xdata);
+                if (local->fd)
+                        fd_unref(local->fd);
+                local->fd = 0;
+                local->loc = NULL;
+                local->newloc = NULL;
+                local->op_ret = -1;
+                local->op_errno = ret;
+        }
+
+        return ret;
+}
+
+/*
+ * read and verify locked metadata against the old pathname (via open);
+ * update the metadata string in accordance with the new pathname;
+ * submit modified metadata;
+ * wind;
+ */
+static int32_t linkop(call_frame_t *frame,
+		      xlator_t *this,
+		      loc_t *oldloc,
+		      loc_t *newloc,
+		      int flags,
+		      dict_t *xdata,
+		      glusterfs_fop_t op)
+{
+	int32_t ret;
+	dict_t *dict;
+	crypt_local_t *local;
+	void (*unwind_fn)(call_frame_t *frame);
+
+	unwind_fn = linkop_unwind_dispatch(op);
+
+	ret = linkop_grab_local(frame, this, oldloc, newloc, flags, xdata, op);
+	local = frame->local;
+	if (ret)
+		goto error;
+	dict = dict_new();
+	if (!dict) {
+		gf_log(this->name, GF_LOG_ERROR, "Can not create dict");
+		ret = ENOMEM;
+		goto error;
+	}
+	/*
+	 * Set a message to crypt_open() that we need
+	 * locked metadata string.
+	 * All link operations (link, unlink, rename)
+	 * need write lock
+	 */
+	msgflags_set_mtd_wlock(&local->msgflags);
+	ret = dict_set_static_bin(dict,
+				  MSGFLAGS_PREFIX,
+				  &local->msgflags,
+				  sizeof(local->msgflags));
+	if (ret) {
+		gf_log(this->name, GF_LOG_ERROR, "Can not set dict");
+		dict_unref(dict);
+		goto error;
+	}
+	/*
+	 * verify metadata against the old pathname
+	 * and retrieve locked metadata string
+	 */
+	STACK_WIND(frame,
+		   linkop_begin,
+		   this,
+		   this->fops->open, /* crypt_open() */
+		   oldloc,
+		   O_RDWR,
+		   local->fd,
+		   dict);
+	dict_unref(dict);
+	return 0;
+ error:
+	local->op_ret = -1;
+	local->op_errno = ret;
+	unwind_fn(frame);
+	return 0;
+}
+
+static int32_t crypt_link(call_frame_t *frame, xlator_t *this,
+			  loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+{
+	return linkop(frame, this, oldloc, newloc, 0, xdata, GF_FOP_LINK);
+}
+
+static int32_t crypt_unlink(call_frame_t *frame, xlator_t *this,
+			    loc_t *loc, int flags, dict_t *xdata)
+{
+	return linkop(frame, this, loc, NULL, flags, xdata, GF_FOP_UNLINK);
+}
+
+static int32_t crypt_rename(call_frame_t *frame, xlator_t *this,
+			    loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+{
+	return linkop(frame, this, oldloc, newloc, 0, xdata, GF_FOP_RENAME);
+}
+
+static void put_one_call_open(call_frame_t *frame)
+{
+	crypt_local_t *local = frame->local;
+	if (put_one_call(local)) {
+		fd_t *fd = local->fd;
+		loc_t *loc = local->loc;
+		dict_t *xdata = local->xdata;
+
+		STACK_UNWIND_STRICT(open,
+				    frame,
+				    local->op_ret,
+				    local->op_errno,
+				    fd,
+				    xdata);
+		fd_unref(fd);
+		if (xdata)
+			dict_unref(xdata);
+		loc_wipe(loc);
+		GF_FREE(loc);
+	}
+}
+
+static int32_t __crypt_readv_done(call_frame_t *frame,
+				  void *cookie,
+				  xlator_t *this,
+				  int32_t op_ret,
+				  int32_t op_errno, dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	fd_t *local_fd = local->fd;
+	dict_t *local_xdata = local->xdata;
+	/* read deals with data configs only */
+	struct iovec *avec = local->data_conf.avec;
+	char **pool = local->data_conf.pool;
+	int blocks_in_pool = local->data_conf.blocks_in_pool;
+	struct iobref *iobref = local->iobref;
+	struct iobref *iobref_data = local->iobref_data;
+
+	if (op_ret < 0) {
+		gf_log(this->name, GF_LOG_WARNING,
+		       "readv unlock failed (%d)", op_errno);
+		if (local->op_ret >= 0) {
+			local->op_ret = op_ret;
+			local->op_errno = op_errno;
+		}
+	}
+	dump_plain_text(local, avec);
+
+	gf_log("crypt", GF_LOG_DEBUG,
+	       "readv: ret_to_user: %d, iovec len: %d, ia_size: %llu",
+	       (int)(local->rw_count > 0 ? local->rw_count : local->op_ret),
+	       (int)(local->rw_count > 0 ? iovec_get_size(avec, local->data_conf.acount) : 0),
+	       (unsigned long long)local->buf.ia_size);
+
+	STACK_UNWIND_STRICT(readv,
+			    frame,
+			    local->rw_count > 0 ? local->rw_count : local->op_ret,
+			    local->op_errno,
+			    avec,
+			    avec ? local->data_conf.acount : 0,
+			    &local->buf,
+			    local->iobref,
+			    local_xdata);
+
+	free_avec(avec, pool, blocks_in_pool);
+	fd_unref(local_fd);
+	if (local_xdata)
+		dict_unref(local_xdata);
+	if (iobref)
+		iobref_unref(iobref);
+	if (iobref_data)
+		iobref_unref(iobref_data);
+	return 0;
+}
+
+static void crypt_readv_done(call_frame_t *frame, xlator_t *this)
+{
+	if (parent_is_crypt_xlator(frame, this))
+		/*
+		 *  don't unlock (it will be done by the parent)
+		 */
+		__crypt_readv_done(frame, NULL, this, 0, 0, NULL);
+	else {
+		crypt_local_t *local = frame->local;
+		struct gf_flock  lock  = {0, };
+
+		lock.l_type   = F_UNLCK;
+		lock.l_whence = SEEK_SET;
+		lock.l_start  = 0;
+		lock.l_len    = 0;
+		lock.l_pid    = 0;
+
+		STACK_WIND(frame,
+			   __crypt_readv_done,
+			   FIRST_CHILD(this),
+			   FIRST_CHILD(this)->fops->finodelk,
+			   this->name,
+			   local->fd,
+			   F_SETLKW,
+			   &lock,
+			   NULL);
+	}
+}
+
+static void put_one_call_readv(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+	if (put_one_call(local))
+		crypt_readv_done(frame, this);
+}
+
+static int32_t __crypt_writev_done(call_frame_t *frame,
+				   void *cookie,
+				   xlator_t *this,
+				   int32_t op_ret,
+				   int32_t op_errno, dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	fd_t *local_fd = local->fd;
+	dict_t *local_xdata = local->xdata;
+	int32_t ret_to_user;
+
+	if (local->xattr)
+		dict_unref(local->xattr);
+	/*
+	 * Calculate amout of butes to be returned
+	 * to user. We need to subtract paddings that
+	 * have been written as a part of atom.
+	 */
+	/*
+	 * subtract head padding
+	 */
+	if (local->rw_count == 0)
+		/*
+		 * Nothing has been written, it must be an error
+		 */
+		ret_to_user = local->op_ret;
+	else if (local->rw_count <= local->data_conf.off_in_head) {
+		gf_log("crypt", GF_LOG_WARNING, "Incomplete write");
+		ret_to_user = 0;
+	}
+	else
+		ret_to_user = local->rw_count -
+			local->data_conf.off_in_head;
+	/*
+	 * subtract tail padding
+	 */
+	if (ret_to_user > local->data_conf.orig_size)
+		ret_to_user = local->data_conf.orig_size;
+
+	if (local->iobref)
+		iobref_unref(local->iobref);
+	if (local->iobref_data)
+		iobref_unref(local->iobref_data);
+	free_avec_data(local);
+	free_avec_hole(local);
+
+	gf_log("crypt", GF_LOG_DEBUG,
+	       "writev: ret_to_user: %d", ret_to_user);
+
+	STACK_UNWIND_STRICT(writev,
+			    frame,
+			    ret_to_user,
+			    local->op_errno,
+			    &local->prebuf,
+			    &local->postbuf,
+			    local_xdata);
+	fd_unref(local_fd);
+	if (local_xdata)
+		dict_unref(local_xdata);
+	return 0;
+}
+
+static int32_t crypt_writev_done(call_frame_t *frame,
+				 void *cookie,
+				 xlator_t *this,
+				 int32_t op_ret,
+				 int32_t op_errno,
+				 dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0)
+		gf_log("crypt", GF_LOG_WARNING, "can not update file size");
+
+	if (parent_is_crypt_xlator(frame, this))
+		/*
+		 * don't unlock (it will be done by the parent)
+		 */
+		__crypt_writev_done(frame, NULL, this, 0, 0, NULL);
+	else {
+		struct gf_flock  lock  = {0, };
+
+		lock.l_type   = F_UNLCK;
+		lock.l_whence = SEEK_SET;
+		lock.l_start  = 0;
+		lock.l_len    = 0;
+		lock.l_pid    = 0;
+
+		STACK_WIND(frame,
+			   __crypt_writev_done,
+			   FIRST_CHILD(this),
+			   FIRST_CHILD(this)->fops->finodelk,
+			   this->name,
+			   local->fd,
+			   F_SETLKW,
+			   &lock,
+			   NULL);
+	}
+	return 0;
+}
+
+static void put_one_call_writev(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+	if (put_one_call(local)) {
+		if (local->update_disk_file_size) {
+			int32_t ret;
+			/*
+			 * update file size, unlock the file and unwind
+			 */
+			ret = dict_set(local->xattr,
+				       FSIZE_XATTR_PREFIX,
+				       data_from_uint64(local->cur_file_size));
+			if (ret) {
+				gf_log("crypt", GF_LOG_WARNING,
+				       "can not set key to update file size");
+				crypt_writev_done(frame, NULL,
+						  this, 0, 0, NULL);
+				return;
+			}
+			gf_log("crypt", GF_LOG_DEBUG,
+			       "Updating disk file size to %llu",
+			       (unsigned long long)local->cur_file_size);
+			STACK_WIND(frame,
+				   crypt_writev_done,
+				   FIRST_CHILD(this),
+				   FIRST_CHILD(this)->fops->fsetxattr,
+				   local->fd,
+				   local->xattr, /* CRYPTO_FORMAT_PREFIX */
+				   0,
+				   NULL);
+		}
+		else
+			crypt_writev_done(frame, NULL, this, 0, 0, NULL);
+	}
+}
+
+static int32_t __crypt_ftruncate_done(call_frame_t *frame,
+				      void *cookie,
+				      xlator_t *this,
+				      int32_t op_ret,
+				      int32_t op_errno, dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	fd_t *local_fd = local->fd;
+	dict_t *local_xdata = local->xdata;
+	char *iobase = local->vec.iov_base;
+
+	if (op_ret < 0) {
+		gf_log(this->name, GF_LOG_WARNING,
+		       "ftruncate unlock failed (%d)", op_errno);
+		if (local->op_ret >= 0) {
+			local->op_ret = op_ret;
+			local->op_errno = op_errno;
+		}
+	}
+	if (local->iobref_data)
+		iobref_unref(local->iobref_data);
+	free_avec_data(local);
+	free_avec_hole(local);
+
+	gf_log("crypt", GF_LOG_DEBUG,
+	       "ftruncate, return to user: presize=%llu, postsize=%llu",
+	       (unsigned long long)local->prebuf.ia_size,
+	       (unsigned long long)local->postbuf.ia_size);
+
+	STACK_UNWIND_STRICT(ftruncate,
+			    frame,
+			    local->op_ret < 0 ? -1 : 0,
+			    local->op_errno,
+			    &local->prebuf,
+			    &local->postbuf,
+			    local_xdata);
+	fd_unref(local_fd);
+	if (local_xdata)
+		dict_unref(local_xdata);
+	if (iobase)
+		GF_FREE(iobase);
+	return 0;
+}
+
+static int32_t crypt_ftruncate_done(call_frame_t *frame,
+				    void *cookie,
+				    xlator_t *this,
+				    int32_t op_ret,
+				    int32_t op_errno,
+				    dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+	struct gf_flock  lock  = {0, };
+
+	dict_unref(local->xattr);
+	if (op_ret < 0)
+		gf_log("crypt", GF_LOG_WARNING, "can not update file size");
+
+	lock.l_type   = F_UNLCK;
+	lock.l_whence = SEEK_SET;
+	lock.l_start  = 0;
+	lock.l_len    = 0;
+	lock.l_pid    = 0;
+
+	STACK_WIND(frame,
+		   __crypt_ftruncate_done,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->finodelk,
+		   this->name,
+		   local->fd,
+		   F_SETLKW,
+		   &lock,
+		   NULL);
+	return 0;
+}
+
+static void put_one_call_ftruncate(call_frame_t *frame, xlator_t *this)
+{
+	crypt_local_t *local = frame->local;
+	if (put_one_call(local)) {
+		if (local->update_disk_file_size) {
+			int32_t ret;
+			/*
+			 * update file size, unlock the file and unwind
+			 */
+			ret = dict_set(local->xattr,
+				       FSIZE_XATTR_PREFIX,
+				       data_from_uint64(local->cur_file_size));
+			if (ret) {
+				gf_log("crypt", GF_LOG_WARNING,
+				       "can not set key to update file size");
+				crypt_ftruncate_done(frame, NULL,
+						     this, 0, 0, NULL);
+				return;
+			}
+			gf_log("crypt", GF_LOG_DEBUG,
+			       "Updating disk file size to %llu",
+			       (unsigned long long)local->cur_file_size);
+			STACK_WIND(frame,
+				   crypt_ftruncate_done,
+				   FIRST_CHILD(this),
+				   FIRST_CHILD(this)->fops->fsetxattr,
+				   local->fd,
+				   local->xattr, /* CRYPTO_FORMAT_PREFIX */
+				   0,
+				   NULL);
+		}
+		else
+			crypt_ftruncate_done(frame, NULL, this, 0, 0, NULL);
+	}
+}
+
+/*
+ * load regular file size for some FOPs
+ */
+static int32_t load_file_size(call_frame_t *frame,
+			      void *cookie,
+			      xlator_t *this,
+			      int32_t op_ret,
+			      int32_t op_errno,
+			      dict_t *dict,
+			      dict_t *xdata)
+{
+	data_t *data;
+	crypt_local_t *local = frame->local;
+
+	dict_t *local_xdata = local->xdata;
+	inode_t *local_inode = local->inode;
+
+	if (op_ret < 0)
+		goto unwind;
+	/*
+	 * load regular file size
+	 */
+	data = dict_get(dict, FSIZE_XATTR_PREFIX);
+	if (!data) {
+		if (local->xdata)
+			dict_unref(local->xdata);
+		gf_log("crypt", GF_LOG_WARNING, "Regular file size not found");
+		op_ret = -1;
+		op_errno = EIO;
+		goto unwind;
+	}
+	local->buf.ia_size = data_to_uint64(data);
+
+	gf_log(this->name, GF_LOG_DEBUG,
+	       "FOP %d: Translate regular file to %llu",
+	       local->fop,
+	       (unsigned long long)local->buf.ia_size);
+ unwind:	
+	if (local->fd)
+		fd_unref(local->fd);
+	if (local->loc) {
+		loc_wipe(local->loc);
+		GF_FREE(local->loc);
+	}
+	switch (local->fop) {
+	case GF_FOP_FSTAT:
+		STACK_UNWIND_STRICT(fstat,
+				    frame,
+				    op_ret,
+				    op_errno,
+				    op_ret >= 0 ? &local->buf : NULL,
+				    local->xdata);
+		break;
+	case GF_FOP_STAT:
+		STACK_UNWIND_STRICT(stat,
+				    frame,
+				    op_ret,
+				    op_errno,
+				    op_ret >= 0 ? &local->buf : NULL,
+				    local->xdata);
+		break;
+	case GF_FOP_LOOKUP:
+		STACK_UNWIND_STRICT(lookup,
+				    frame,
+				    op_ret,
+				    op_errno,
+				    op_ret >= 0 ? local->inode : NULL,
+				    op_ret >= 0 ? &local->buf : NULL,				    
+				    local->xdata,
+				    op_ret >= 0 ? &local->postbuf : NULL);
+		break;
+	case GF_FOP_READ:
+		STACK_UNWIND_STRICT(readv,
+				    frame,
+				    op_ret,
+				    op_errno,
+				    NULL,
+				    0,
+				    op_ret >= 0 ? &local->buf : NULL,
+				    NULL,
+				    NULL);
+		break;
+	default:
+		gf_log(this->name, GF_LOG_WARNING,
+		       "Improper file operation %d", local->fop);
+	}
+	if (local_xdata)
+		dict_unref(local_xdata);
+	if (local_inode)
+		inode_unref(local_inode);
+	return 0;
+}
+
+static int32_t crypt_stat_common_cbk(call_frame_t *frame,
+				     void *cookie,
+				     xlator_t *this,
+				     int32_t op_ret,
+				     int32_t op_errno,
+				     struct iatt *buf, dict_t *xdata)
+{
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0)
+		goto unwind;
+	if (!IA_ISREG(buf->ia_type))
+		goto unwind;
+
+	local->buf = *buf;
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+
+	switch (local->fop) {
+	case GF_FOP_FSTAT:
+		STACK_WIND(frame,
+			   load_file_size,
+			   FIRST_CHILD(this),
+			   FIRST_CHILD(this)->fops->fgetxattr,
+			   local->fd,
+			   FSIZE_XATTR_PREFIX,
+			   NULL);
+		break;
+	case GF_FOP_STAT:
+		STACK_WIND(frame,
+			   load_file_size,
+			   FIRST_CHILD(this),
+			   FIRST_CHILD(this)->fops->getxattr,
+			   local->loc,
+			   FSIZE_XATTR_PREFIX,
+			   NULL);
+		break;
+	default:
+		gf_log (this->name, GF_LOG_WARNING,
+			"Improper file operation %d", local->fop);
+	}
+	return 0;
+ unwind:
+	if (local->fd)
+		fd_unref(local->fd);
+	if (local->loc) {
+		loc_wipe(local->loc);
+		GF_FREE(local->loc);
+	}
+	switch (local->fop) {
+	case GF_FOP_FSTAT:
+		STACK_UNWIND_STRICT(fstat,
+				    frame,
+				    op_ret,
+				    op_errno,
+				    op_ret >= 0 ? buf : NULL,
+				    op_ret >= 0 ? xdata : NULL);
+		break;
+	case GF_FOP_STAT:
+		STACK_UNWIND_STRICT(stat,
+				    frame,
+				    op_ret,
+				    op_errno,
+				    op_ret >= 0 ? buf : NULL,
+				    op_ret >= 0 ? xdata : NULL);
+		break;
+	default:
+		gf_log (this->name, GF_LOG_WARNING,
+			"Improper file operation %d", local->fop);
+	}
+	return 0;
+}
+
+static int32_t crypt_fstat(call_frame_t *frame,
+			   xlator_t *this,
+			   fd_t *fd, dict_t *xdata)
+{
+	crypt_local_t *local;
+
+	local = crypt_alloc_local(frame, this, GF_FOP_FSTAT);
+	if (!local)
+		goto error;
+	local->fd = fd_ref(fd);
+	STACK_WIND(frame,
+		   crypt_stat_common_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->fstat,
+		   fd,
+		   xdata);	
+	return 0;
+ error:
+	STACK_UNWIND_STRICT(fstat,
+			    frame,
+			    -1,
+			    ENOMEM,
+			    NULL,
+			    NULL);
+	return 0;
+}
+
+static int32_t crypt_stat(call_frame_t *frame,
+			  xlator_t *this,
+			  loc_t *loc, dict_t *xdata)
+{
+	int32_t ret;
+	crypt_local_t *local;
+
+	local = crypt_alloc_local(frame, this, GF_FOP_STAT);
+	if (!local)
+		goto error;
+	local->loc = GF_CALLOC(1, sizeof(*loc), gf_crypt_mt_loc);
+	if (!local->loc)
+		goto error;
+	memset(local->loc, 0, sizeof(*local->loc));
+	ret = loc_copy(local->loc, loc);
+	if (ret) {
+		GF_FREE(local->loc);
+		goto error;
+	}
+	STACK_WIND(frame,
+		   crypt_stat_common_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->stat,
+		   loc,
+		   xdata);	
+	return 0;
+ error:
+	STACK_UNWIND_STRICT(stat,
+			    frame,
+			    -1,
+			    ENOMEM,
+			    NULL,
+			    NULL);
+	return 0;
+}
+
+static int32_t crypt_lookup_cbk(call_frame_t *frame,
+				void *cookie,
+				xlator_t *this,
+				int32_t op_ret,
+				int32_t op_errno,
+				inode_t *inode,
+				struct iatt *buf, dict_t *xdata,
+				struct iatt *postparent)
+{
+	crypt_local_t *local = frame->local;
+
+	if (op_ret < 0)
+		goto unwind;
+	if (!IA_ISREG(buf->ia_type))
+		goto unwind;
+
+	local->inode = inode_ref(inode);
+	local->buf = *buf;
+	local->postbuf = *postparent;
+	if (xdata)
+		local->xdata = dict_ref(xdata);
+	uuid_copy(local->loc->gfid, buf->ia_gfid);
+
+	STACK_WIND(frame,
+		   load_file_size,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->getxattr,
+		   local->loc,
+		   FSIZE_XATTR_PREFIX,
+		   NULL);
+	return 0;
+ unwind:
+	loc_wipe(local->loc);
+	GF_FREE(local->loc);
+	STACK_UNWIND_STRICT(lookup,
+			    frame,
+			    op_ret,
+			    op_errno,
+			    inode,
+			    buf,
+			    xdata,
+			    postparent);
+	return 0;
+}
+
+static int32_t crypt_lookup(call_frame_t *frame,
+			    xlator_t *this,
+			    loc_t *loc, dict_t *xdata)
+{
+	int32_t ret;
+	crypt_local_t *local;
+
+	local = crypt_alloc_local(frame, this, GF_FOP_LOOKUP);
+	if (!local)
+		goto error;
+	local->loc = GF_CALLOC(1, sizeof(*loc), gf_crypt_mt_loc);
+	if (!local->loc)
+		goto error;
+	memset(local->loc, 0, sizeof(*local->loc));
+	ret = loc_copy(local->loc, loc);
+	if (ret) {
+		GF_FREE(local->loc);
+		goto error;
+	}
+	gf_log(this->name, GF_LOG_DEBUG, "Lookup %s", loc->path);
+	STACK_WIND(frame,
+		   crypt_lookup_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->lookup,
+		   loc,
+		   xdata);
+	return 0;
+ error:
+	STACK_UNWIND_STRICT(lookup,
+			    frame,
+			    -1,
+			    ENOMEM,
+			    NULL,
+			    NULL,
+			    NULL,
+			    NULL);
+	return 0;
+}
+
+/*
+ * for every regular directory entry find its real file size
+ * and update stat's buf properly
+ */
+static int32_t crypt_readdirp_cbk(call_frame_t *frame,
+				  void *cookie,
+				  xlator_t *this,
+				  int32_t op_ret,
+				  int32_t op_errno,
+				  gf_dirent_t *entries, dict_t *xdata)
+{
+	gf_dirent_t *entry = NULL;
+
+	if (op_ret < 0)
+		goto unwind;
+
+	list_for_each_entry (entry, (&entries->list), list) {
+		data_t *data;
+
+		if (!IA_ISREG(entry->d_stat.ia_type))
+			continue;
+		data = dict_get(entry->dict, FSIZE_XATTR_PREFIX);
+		if (!data){
+			gf_log("crypt", GF_LOG_WARNING,
+			       "Regular file size of direntry not found");
+			op_errno = EIO;
+			op_ret = -1;
+			break;
+		}
+		entry->d_stat.ia_size = data_to_uint64(data);
+	}
+ unwind:
+	STACK_UNWIND_STRICT(readdirp, frame, op_ret, op_errno, entries, xdata);
+	return 0;
+}
+
+/*
+ * ->readdirp() fills in-core inodes, so we need to set proper
+ * file sizes for all directory entries of the parent @fd.
+ * Actual updates take place in ->crypt_readdirp_cbk()
+ */
+static int32_t crypt_readdirp(call_frame_t *frame, xlator_t *this,
+			      fd_t *fd, size_t size, off_t offset,
+			      dict_t *xdata)
+{
+	int32_t ret = ENOMEM;
+
+	if (!xdata) {
+		xdata = dict_new();
+		if (!xdata)
+			goto error;
+	}
+	else
+		dict_ref(xdata);
+	/*
+	 * make sure that we'll have real file sizes at ->readdirp_cbk()
+	 */
+	ret = dict_set(xdata, FSIZE_XATTR_PREFIX, data_from_uint64(0));
+	if (ret) {
+		dict_unref(xdata);
+		goto error;
+	}
+	STACK_WIND(frame,
+		   crypt_readdirp_cbk,
+		   FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->readdirp,
+		   fd,
+		   size,
+		   offset,
+		   xdata);
+	dict_unref(xdata);
+	return 0;
+ error:
+	STACK_UNWIND_STRICT(readdirp, frame, -1, ret, NULL, NULL);
+	return 0;
+}
+
+static int32_t crypt_access(call_frame_t *frame,
+			    xlator_t *this,
+			    loc_t *loc,
+			    int32_t mask, dict_t *xdata)
+{
+	gf_log(this->name, GF_LOG_WARNING,
+	       "NFS mounts of encrypted volumes are unsupported");
+	STACK_UNWIND_STRICT(access, frame, -1, EPERM, NULL);
+	return 0;
+}
+
+int32_t master_set_block_size (xlator_t *this, crypt_private_t *priv,
+			       dict_t *options)
+{
+	uint64_t block_size = 0;
+	struct master_cipher_info *master = get_master_cinfo(priv);
+
+	if (options != NULL)
+		GF_OPTION_RECONF("block-size", block_size, options,
+				 size, error);
+	else
+		GF_OPTION_INIT("block-size", block_size, size, error);
+
+	switch (block_size) {
+	case 512:
+		master->m_block_bits = 9;
+		break;
+	case 1024:
+		master->m_block_bits = 10;
+		break;
+	case 2048:
+		master->m_block_bits = 11;
+		break;
+	case 4096:
+		master->m_block_bits = 12;
+		break;
+	default:
+		gf_log("crypt", GF_LOG_ERROR,
+		       "FATAL: unsupported block size %llu",
+		       (unsigned long long)block_size);
+		goto error;
+	}
+	return 0;
+ error:
+	return -1;
+}
+
+int32_t master_set_alg(xlator_t *this, crypt_private_t *priv)
+{
+	struct master_cipher_info *master = get_master_cinfo(priv);
+	master->m_alg = AES_CIPHER_ALG;
+	return 0;
+}
+
+int32_t master_set_mode(xlator_t *this, crypt_private_t *priv)
+{
+	struct master_cipher_info *master = get_master_cinfo(priv);
+	master->m_mode = XTS_CIPHER_MODE;
+	return 0;
+}
+
+/*
+ * set key size in bits to the master info
+ * Pre-conditions: cipher mode in the master info is uptodate.
+ */
+static int master_set_data_key_size (xlator_t *this, crypt_private_t *priv,
+				     dict_t *options)
+{
+	int32_t ret;
+	uint64_t key_size = 0;
+	struct master_cipher_info *master = get_master_cinfo(priv);
+
+	if (options != NULL)
+		GF_OPTION_RECONF("data-key-size", key_size, options,
+				 size, error);
+	else
+		GF_OPTION_INIT("data-key-size", key_size, size, error);
+
+	ret = data_cipher_algs[master->m_alg][master->m_mode].check_key(key_size);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR,
+		       "FATAL: wrong bin key size %llu for alg %d mode %d",
+		       (unsigned long long)key_size,
+		       (int)master->m_alg,
+		       (int)master->m_mode);
+		goto error;
+	}
+	master->m_dkey_size = key_size;
+	return 0;
+ error:
+	return -1;
+}
+
+static int is_hex(char *s) {
+	return  ('0' <= *s && *s <= '9') || ('a' <= *s && *s <= 'f');
+}
+
+static int parse_hex_buf(xlator_t *this, char *src, unsigned char *dst,
+			 int hex_size)
+{
+	int i;
+	int hex_byte = 0;
+
+	for (i = 0; i < (hex_size / 2); i++) {
+		if (!is_hex(src + i*2) || !is_hex(src + i*2 + 1)) {
+			gf_log("crypt", GF_LOG_ERROR,
+			       "FATAL: not hex symbol in key");
+			return -1;
+		}
+		if (sscanf(src + i*2, "%2x", &hex_byte) != 1) {
+			gf_log("crypt", GF_LOG_ERROR,
+			       "FATAL: can not parse hex key");
+			return -1;
+		}
+		dst[i] = hex_byte & 0xff;
+	}
+	return 0;
+}
+
+/*
+ * Parse options;
+ * install master volume key
+ */
+int32_t master_set_master_vol_key(xlator_t *this, crypt_private_t *priv)
+{
+	int32_t ret;
+	FILE *file = NULL;
+
+	int32_t key_size;
+	char *opt_key_file_pathname = NULL;
+
+	unsigned char bin_buf[MASTER_VOL_KEY_SIZE];
+	char hex_buf[2 * MASTER_VOL_KEY_SIZE];
+
+	struct master_cipher_info *master = get_master_cinfo(priv);
+	/*
+	 * extract master key passed via option
+	 */
+	GF_OPTION_INIT("master-key", opt_key_file_pathname, path, bad_key);
+
+	if (!opt_key_file_pathname) {
+		gf_log(this->name, GF_LOG_ERROR, "FATAL: missing master key");
+		return -1;
+	}
+	gf_log(this->name, GF_LOG_DEBUG, "handling file key %s",
+	       opt_key_file_pathname);
+
+	file = fopen(opt_key_file_pathname, "r");
+	if (file == NULL) {
+		gf_log(this->name, GF_LOG_ERROR,
+		       "FATAL: can not open file with master key");
+		return -1;
+	}
+	/*
+	 * extract hex key
+	 */
+	key_size = fread(hex_buf, 1, sizeof(hex_buf), file);
+	if (key_size < sizeof(hex_buf)) {
+		gf_log(this->name, GF_LOG_ERROR,
+		       "FATAL: master key is too short");
+		goto bad_key;
+	}
+	ret = parse_hex_buf(this, hex_buf, bin_buf, key_size);
+	if (ret)
+		goto bad_key;
+	memcpy(master->m_key, bin_buf, MASTER_VOL_KEY_SIZE);
+	memset(hex_buf, 0, sizeof(hex_buf));
+	fclose(file);
+
+ 	memset(bin_buf, 0, sizeof(bin_buf));
+	return 0;
+ bad_key:
+	gf_log(this->name, GF_LOG_ERROR, "FATAL: bad master key");
+	if (file)
+		fclose(file);
+	memset(bin_buf, 0, sizeof(bin_buf));
+	return -1;
+}
+
+/*
+ * Derive volume key for object-id authentication
+ */
+int32_t master_set_nmtd_vol_key(xlator_t *this, crypt_private_t *priv)
+{
+	return get_nmtd_vol_key(get_master_cinfo(priv));
+}
+
+int32_t crypt_init_xlator(xlator_t *this)
+{
+	int32_t ret;
+	crypt_private_t *priv = this->private; 
+
+	ret = master_set_alg(this, priv);
+	if (ret)
+		return ret;
+	ret = master_set_mode(this, priv);
+	if (ret)
+		return ret;
+	ret = master_set_block_size(this, priv, NULL);
+	if (ret)
+		return ret;
+	ret = master_set_data_key_size(this, priv, NULL);
+	if (ret)
+		return ret;
+	ret = master_set_master_vol_key(this, priv);
+	if (ret)
+		return ret;
+	return master_set_nmtd_vol_key(this, priv);
+}
+
+static int32_t crypt_alloc_private(xlator_t *this)
+{
+	this->private = GF_CALLOC(1, sizeof(crypt_private_t), gf_crypt_mt_priv);
+	if (!this->private) {
+		gf_log("crypt", GF_LOG_ERROR,
+		       "Can not allocate memory for private data");
+		return ENOMEM;
+	}
+	return 0;
+}
+
+static void crypt_free_private(xlator_t *this)
+{
+	crypt_private_t *priv = this->private;
+	if (priv) {
+		memset(priv, 0, sizeof(*priv));
+		GF_FREE(priv);
+	}
+}
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_crypt_mt_end);
+
+        if (ret != 0) {
+                gf_log(this->name, GF_LOG_ERROR, "Memory accounting init"
+                       "failed");
+                return ret;
+        }
+
+        return ret;
+}
+
+int32_t reconfigure (xlator_t *this, dict_t *options)
+{
+	int32_t ret = -1;
+	crypt_private_t *priv = NULL;
+
+	GF_VALIDATE_OR_GOTO ("crypt", this, error);
+	GF_VALIDATE_OR_GOTO (this->name, this->private, error);
+	GF_VALIDATE_OR_GOTO (this->name, options, error);
+
+	priv = this->private;
+
+	ret = master_set_block_size(this, priv, options);
+	if (ret) {
+		gf_log("this->name", GF_LOG_ERROR,
+		       "Failed to reconfure block size");
+		goto error;
+	}
+	ret = master_set_data_key_size(this, priv, options);
+	if (ret) {
+		gf_log("this->name", GF_LOG_ERROR,
+		       "Failed to reconfure data key size");
+		goto error;
+	}
+	return 0;
+ error:
+	return ret;
+}
+
+int32_t init(xlator_t *this)
+{
+	int32_t ret;
+
+	if (!this->children || this->children->next) {
+		gf_log ("crypt", GF_LOG_ERROR,
+			"FATAL: crypt should have exactly one child");
+		return EINVAL;
+	}
+	if (!this->parents) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"dangling volume. check volfile ");
+	}
+	ret = crypt_alloc_private(this);
+	if (ret)
+		return ret;
+	ret = crypt_init_xlator(this);
+	if (ret)
+		goto error;
+	this->local_pool = mem_pool_new(crypt_local_t, 64);
+        if (!this->local_pool) {
+		gf_log(this->name, GF_LOG_ERROR,
+		       "failed to create local_t's memory pool");
+		ret = ENOMEM;
+                goto error;
+        }
+	gf_log ("crypt", GF_LOG_INFO, "crypt xlator loaded");
+	return 0;
+ error:
+	crypt_free_private(this);
+	return ret;
+}
+
+void fini (xlator_t *this)
+{
+	crypt_free_private(this);
+}
+
+struct xlator_fops fops = {
+	.readv        = crypt_readv,
+	.writev       = crypt_writev,
+	.truncate     = crypt_truncate,
+	.ftruncate    = crypt_ftruncate,
+	.setxattr     = crypt_setxattr,
+	.fsetxattr    = crypt_fsetxattr,
+	.link         = crypt_link,
+	.unlink       = crypt_unlink,
+	.rename       = crypt_rename,
+	.open         = crypt_open,
+	.create       = crypt_create,
+	.stat         = crypt_stat,
+	.fstat        = crypt_fstat,
+	.lookup       = crypt_lookup,
+	.readdirp     = crypt_readdirp,
+	.access       = crypt_access   
+};
+
+struct xlator_cbks cbks = {
+	.forget       = crypt_forget
+};
+
+struct volume_options options[] = {
+	{ .key = {"master-key"},
+	  .type = GF_OPTION_TYPE_PATH,
+	  .description = "Pathname of regular file which contains master volume key"
+	},
+	{ .key = {"data-key-size"},
+	  .type = GF_OPTION_TYPE_SIZET,
+	  .description = "Data key size (bits)",
+	  .min = 256,
+	  .max = 512,
+	  .default_value = "256",
+	},
+	{ .key = {"block-size"},
+	  .type = GF_OPTION_TYPE_SIZET,
+	  .description = "Atom size (bits)",
+	  .min = 512,
+	  .max = 4096,
+	  .default_value = "4096"
+	},
+	{ .key  = {NULL} },
+};
+
+/*
+  Local variables:
+  c-indentation-style: "K&R"
+  mode-name: "LC"
+  c-basic-offset: 8
+  tab-width: 8
+  fill-column: 80
+  scroll-step: 1
+  End:
+*/
diff --git a/xlators/encryption/crypt/src/crypt.h b/xlators/encryption/crypt/src/crypt.h
new file mode 100644
index 000000000..ff8eb571b
--- /dev/null
+++ b/xlators/encryption/crypt/src/crypt.h
@@ -0,0 +1,903 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef __CRYPT_H__
+#define __CRYPT_H__
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+#include <openssl/aes.h>
+#include <openssl/evp.h>
+#include <openssl/sha.h>
+#include <openssl/hmac.h>
+#include <openssl/cmac.h>
+#include <openssl/modes.h>
+#include "crypt-mem-types.h"
+
+#define CRYPT_XLATOR_ID  (0)
+
+#define MAX_IOVEC_BITS (3)
+#define MAX_IOVEC (1 << MAX_IOVEC_BITS)
+#define KEY_FACTOR_BITS  (6)
+
+#define DEBUG_CRYPT (0)
+#define TRIVIAL_TFM (0)
+
+#define CRYPT_MIN_BLOCK_BITS (9)
+#define CRYPT_MAX_BLOCK_BITS (12)
+
+#define MASTER_VOL_KEY_SIZE (32)
+#define NMTD_VOL_KEY_SIZE (16)
+
+#ifdef __NetBSD__
+typedef off_t loff_t;
+#endif
+
+struct crypt_key {
+	uint32_t len;
+	const char *label;	
+};
+
+/*
+ * Add new key types to the end of this
+ * enumeration but before LAST_KEY_TYPE
+ */
+typedef enum {
+	MASTER_VOL_KEY,
+	NMTD_VOL_KEY,
+	NMTD_LINK_KEY,
+	EMTD_FILE_KEY,
+	DATA_FILE_KEY_256,
+	DATA_FILE_KEY_512,
+	LAST_KEY_TYPE
+}crypt_key_type;
+
+struct kderive_context {
+	const unsigned char *pkey;/* parent key */
+	uint32_t pkey_len;        /* parent key size, bits */
+	uint32_t ckey_len;        /* child key size, bits */
+	unsigned char *fid;       /* fixed input data, NIST 800-108, 5.1 */
+	uint32_t fid_len;         /* fid len, bytes */
+	unsigned char *out;       /* contains child keying material */
+	uint32_t out_len;         /* out len, bytes */
+};
+
+typedef enum {
+	DATA_ATOM,
+	HOLE_ATOM,
+	LAST_DATA_TYPE
+}atom_data_type;
+
+typedef enum {
+	HEAD_ATOM,
+	TAIL_ATOM,
+	FULL_ATOM,
+	LAST_LOCALITY_TYPE
+}atom_locality_type;
+
+typedef enum {
+	MTD_CREATE,
+	MTD_APPEND,
+	MTD_OVERWRITE,
+	MTD_CUT,
+	MTD_LAST_OP
+} mtd_op_t;
+
+struct xts128_context {
+	void      *key1, *key2;
+	block128_f block1,block2;
+};
+
+struct object_cipher_info {
+	cipher_alg_t  o_alg;
+        cipher_mode_t o_mode;
+	uint32_t      o_block_bits;
+        uint32_t      o_dkey_size; /* raw data key size in bits */
+	union {
+		struct {
+			unsigned char ivec[16];
+			AES_KEY dkey[2];
+			AES_KEY tkey;    /* key used for tweaking */
+			XTS128_CONTEXT xts;
+		} aes_xts;
+	} u;
+};
+
+struct master_cipher_info {
+	/*
+	 * attributes inherited by newly created regular files
+	 */
+	cipher_alg_t  m_alg;
+        cipher_mode_t m_mode;
+	uint32_t      m_block_bits;
+        uint32_t      m_dkey_size; /* raw key size in bits */
+	/*
+	 * master key
+	 */
+	unsigned char m_key[MASTER_VOL_KEY_SIZE];
+	/* 
+	 * volume key for oid authentication
+	 */
+	unsigned char m_nmtd_key[NMTD_VOL_KEY_SIZE];
+};
+
+/*
+* This info is not changed during file's life
+ */
+struct crypt_inode_info {
+#if DEBUG_CRYPT
+	loc_t *loc; /* pathname that the file has been
+		       opened, or created with */
+#endif
+	uint16_t nr_minor;
+	uuid_t oid;
+	struct object_cipher_info cinfo;
+};
+
+/*
+ * this should locate in secure memory
+ */
+typedef struct {
+	struct master_cipher_info master;
+} crypt_private_t;
+
+static inline struct master_cipher_info *get_master_cinfo(crypt_private_t *priv)
+{
+	return &priv->master;
+}
+
+static inline struct object_cipher_info *get_object_cinfo(struct crypt_inode_info
+							 *info)
+{
+	return &info->cinfo;
+}
+
+/*
+ * this describes layouts and properties
+ * of atoms in an aligned vector
+ */
+struct avec_config {
+	uint32_t atom_size;
+	atom_data_type type;
+	size_t orig_size;
+	off_t orig_offset;
+	size_t expanded_size;
+	off_t aligned_offset;
+
+	uint32_t off_in_head;
+	uint32_t off_in_tail;
+	uint32_t gap_in_tail;
+	uint32_t nr_full_blocks;
+
+	struct iovec *avec; /* aligned vector */
+	uint32_t acount; /* number of avec components. The same
+			  * as number of occupied logical blocks */
+	char **pool;
+	uint32_t blocks_in_pool;
+	uint32_t cursor; /* makes sense only for ordered writes,
+			  * so there is no races on this counter.
+			  *
+			  * Cursor is per-config object, we don't
+			  * reset cursor for atoms of different
+			  * localities (head, tail, full)
+			  */
+};
+
+
+typedef struct {
+	glusterfs_fop_t fop; /* code of FOP this local info built for */
+	fd_t *fd;
+	inode_t *inode;
+	loc_t *loc;
+	int32_t mac_idx;
+	loc_t *newloc;
+	int32_t flags;
+	int32_t wbflags;
+	struct crypt_inode_info *info;
+	struct iobref *iobref;
+	struct iobref *iobref_data;
+	off_t offset;
+
+	uint64_t old_file_size; /* per FOP, retrieved under lock held */
+	uint64_t cur_file_size; /* per iteration, before issuing IOs */
+	uint64_t new_file_size; /* per iteration, after issuing IOs */
+
+	uint64_t io_offset; /* offset of IOs issued per iteration */
+	uint64_t io_offset_nopad; /* offset of user's data in the atom */
+	uint32_t io_size; /* size of IOs issued per iteration */
+	uint32_t io_size_nopad; /* size of user's data in the IOs */
+	uint32_t eof_padding_size; /* size od EOF padding in the IOs */
+
+	gf_lock_t call_lock; /* protect nr_calls from many cbks */
+	int32_t nr_calls;
+
+	atom_data_type active_setup; /* which setup (hole or date)
+					is currently active */
+	/* data setup */
+	struct avec_config data_conf;
+
+	/* hole setup */
+	int hole_conv_in_proggress;
+	gf_lock_t hole_lock; /* protect hole config from many cbks */
+	int hole_handled;
+	struct avec_config hole_conf;
+	struct iatt buf;
+	struct iatt prebuf;
+	struct iatt postbuf;
+	struct iatt *prenewparent;
+	struct iatt *postnewparent;
+	int32_t op_ret;
+	int32_t op_errno;
+	int32_t rw_count; /* total read or written */
+	gf_lock_t rw_count_lock; /* protect the counter above */
+	unsigned char *format; /* for create, update format string */
+	uint32_t format_size;
+	uint32_t msgflags; /* messages for crypt_open() */
+	dict_t *xdata;
+	dict_t *xattr;
+	struct iovec vec; /* contains last file's atom for
+			     read-prune-write sequence */
+	gf_boolean_t custom_mtd;
+	/*
+	 * the next 3 fields are used by readdir and friends
+	 */
+	gf_dirent_t *de; /* directory entry */
+	char *de_path; /* pathname of directory entry */
+	uint32_t de_prefix_len; /* lenght of the parent's pathname */
+	gf_dirent_t *entries;
+
+	uint32_t update_disk_file_size:1;
+} crypt_local_t;
+
+/* This represents a (read)modify-write atom */
+struct rmw_atom {
+	atom_locality_type locality;
+	/*
+	 * read-modify-write sequence of the atom
+	 */
+	int32_t (*rmw)(call_frame_t *frame,
+		       void *cookie,
+		       xlator_t *this,
+		       int32_t op_ret,
+		       int32_t op_errno,
+		       struct iovec *vec,
+		       int32_t count,
+		       struct iatt *stbuf,
+		       struct iobref *iobref,
+		       dict_t *xdata);
+	/*
+	 * offset of the logical block in a file
+	 */
+	loff_t (*offset_at)(call_frame_t *frame,
+			    struct object_cipher_info *object);
+	/*
+	 * IO offset in an atom
+	 */
+	uint32_t (*offset_in)(call_frame_t *frame,
+			      struct object_cipher_info *object);
+	/*
+	 * number of bytes of plain text of this atom that user
+	 * wants to read/write.
+	 * It can be smaller than atom_size in the case of head
+	 * or tail atoms.
+	 */
+	uint32_t (*io_size_nopad)(call_frame_t *frame,
+				  struct object_cipher_info *object);
+	/*
+	 * which iovec represents the atom
+	 */
+	struct iovec *(*get_iovec)(call_frame_t *frame, uint32_t count);
+	/*
+	 * how many bytes of partial block should be uptodated by
+	 * reading from disk.
+	 * This is used to perform a read component of RMW (read-modify-write).
+	 */
+	uint32_t (*count_to_uptodate)(call_frame_t *frame, struct object_cipher_info *object);
+	struct avec_config *(*get_config)(call_frame_t *frame);
+};
+
+struct data_cipher_alg {
+	gf_boolean_t atomic; /* true means that algorithm requires
+				to pad data before cipher transform */
+	gf_boolean_t should_pad; /* true means that algorithm requires
+				    to pad the end of file with extra-data */
+	uint32_t blkbits; /* blksize = 1 << blkbits */
+	/*
+	 * any preliminary sanity checks goes here
+	 */
+	int32_t (*init)(void);
+	/*
+	 * set alg-mode specific inode info
+	 */
+	int32_t (*set_private)(struct crypt_inode_info *info,
+			       struct master_cipher_info *master);
+	/*
+	 * check alg-mode specific data key
+	 */
+	int32_t (*check_key)(uint32_t key_size);
+	void (*set_iv)(off_t offset, struct object_cipher_info *object);
+	int32_t (*encrypt)(const unsigned char *from, unsigned char *to,
+			   size_t length, off_t offset, const int enc,
+			   struct object_cipher_info *object);
+};
+
+/*
+ * version-dependent metadata loader
+ */
+struct crypt_mtd_loader {
+	/*
+	 * return core format size
+	 */
+	size_t (*format_size)(mtd_op_t op, size_t old_size);
+	/*
+	 * pack version-specific metadata of an object
+	 * at ->create()
+	 */
+	int32_t (*create_format)(unsigned char *wire,
+				 loc_t *loc,
+				 struct crypt_inode_info *info,
+				 struct master_cipher_info *master);
+	/*
+	 * extract version-specific metadata of an object
+	 * at ->open() time
+	 */
+	int32_t (*open_format)(unsigned char *wire,
+			       int32_t len,
+			       loc_t *loc,
+			       struct crypt_inode_info *info,
+			       struct master_cipher_info *master,
+			       crypt_local_t *local,
+			       gf_boolean_t load_info);
+	int32_t (*update_format)(unsigned char *new,
+				 unsigned char *old,
+				 size_t old_len,
+				 int32_t mac_idx,
+				 mtd_op_t op,
+				 loc_t *loc,
+				 struct crypt_inode_info *info,
+				 struct master_cipher_info *master,
+				 crypt_local_t *local);
+};
+
+typedef int32_t (*end_writeback_handler_t)(call_frame_t *frame,
+					   void *cookie,
+					   xlator_t *this,
+					   int32_t op_ret,
+					   int32_t op_errno,
+					   struct iatt *prebuf,
+					   struct iatt *postbuf,
+					   dict_t *xdata);
+typedef void (*linkop_wind_handler_t)(call_frame_t *frame, xlator_t *this);
+typedef void (*linkop_unwind_handler_t)(call_frame_t *frame);
+
+
+/* Declarations */
+
+/* keys.c */
+extern struct crypt_key crypt_keys[LAST_KEY_TYPE];
+int32_t get_nmtd_vol_key(struct master_cipher_info *master);
+int32_t get_nmtd_link_key(loc_t *loc,
+			  struct master_cipher_info *master,
+			  unsigned char *result);
+int32_t get_emtd_file_key(struct crypt_inode_info *info,
+			  struct master_cipher_info *master,
+			  unsigned char *result);
+int32_t get_data_file_key(struct crypt_inode_info *info,
+			  struct master_cipher_info *master,
+			  uint32_t keysize,
+			  unsigned char *key);
+/* data.c */
+extern struct data_cipher_alg data_cipher_algs[LAST_CIPHER_ALG][LAST_CIPHER_MODE];
+void encrypt_aligned_iov(struct object_cipher_info *object,
+			 struct iovec *vec,
+			 int count,
+			 off_t off);
+void decrypt_aligned_iov(struct object_cipher_info *object,
+			 struct iovec *vec,
+			 int count,
+			 off_t off);
+int32_t align_iov_by_atoms(xlator_t *this,
+			   crypt_local_t *local,
+			   struct object_cipher_info *object,
+			   struct iovec *vec /* input vector */,
+			   int32_t count /* number of vec components */,
+			   struct iovec *avec /* aligned vector */,
+			   char **blocks /* pool of blocks */,
+			   uint32_t *blocks_allocated,
+			   struct avec_config *conf);
+int32_t set_config_avec_data(xlator_t *this,
+			     crypt_local_t *local,
+			     struct avec_config *conf,
+			     struct object_cipher_info *object,
+			     struct iovec *vec,
+			     int32_t vec_count);
+int32_t set_config_avec_hole(xlator_t *this,
+			     crypt_local_t *local,
+			     struct avec_config *conf,
+			     struct object_cipher_info *object,
+			     glusterfs_fop_t fop);
+void set_gap_at_end(call_frame_t *frame, struct object_cipher_info *object,
+		    struct avec_config *conf, atom_data_type dtype);
+void set_config_offsets(call_frame_t *frame,
+			xlator_t *this,
+			uint64_t offset,
+			uint64_t count,
+			atom_data_type dtype,
+			int32_t setup_gap_in_tail);
+
+/* metadata.c */
+extern struct crypt_mtd_loader mtd_loaders [LAST_MTD_LOADER];
+
+int32_t alloc_format(crypt_local_t *local, size_t size);
+int32_t alloc_format_create(crypt_local_t *local);
+void free_format(crypt_local_t *local);
+size_t format_size(mtd_op_t op, size_t old_size);
+size_t new_format_size(void);
+int32_t open_format(unsigned char *str, int32_t len, loc_t *loc,
+		    struct crypt_inode_info *info,
+		    struct master_cipher_info *master, crypt_local_t *local,
+		    gf_boolean_t load_info);
+int32_t update_format(unsigned char *new, unsigned char *old,
+		      size_t old_len, int32_t mac_idx, mtd_op_t op, loc_t *loc,
+		      struct crypt_inode_info *info,
+		      struct master_cipher_info *master,
+		      crypt_local_t *local);
+int32_t create_format(unsigned char *wire,
+		      loc_t *loc,
+		      struct crypt_inode_info *info,
+		      struct master_cipher_info *master);
+
+/* atom.c */
+struct rmw_atom *atom_by_types(atom_data_type data,
+			       atom_locality_type locality);
+void submit_partial(call_frame_t *frame,
+		    xlator_t *this,
+		    fd_t *fd,
+		    atom_locality_type ltype);
+void submit_full(call_frame_t *frame, xlator_t *this);
+
+/* crypt.c */
+
+end_writeback_handler_t dispatch_end_writeback(glusterfs_fop_t fop);
+static size_t iovec_get_size(struct iovec *vec, uint32_t count);
+void set_local_io_params_writev(call_frame_t *frame,
+				struct object_cipher_info *object,
+				struct rmw_atom *atom, off_t io_offset,
+				uint32_t io_size);
+void link_wind(call_frame_t *frame, xlator_t *this);
+void unlink_wind(call_frame_t *frame, xlator_t *this);
+void link_unwind(call_frame_t *frame);
+void unlink_unwind(call_frame_t *frame);
+void rename_wind(call_frame_t *frame, xlator_t *this);
+void rename_unwind(call_frame_t *frame);
+
+/* Inline functions */
+
+static inline size_t iovec_get_size(struct iovec *vec, uint32_t count)
+{
+	int i;
+	size_t size = 0;
+	for (i = 0; i < count; i++)
+		size += vec[i].iov_len;
+	return size;
+}
+
+static inline int32_t crypt_xlator_id(void)
+{
+	return CRYPT_XLATOR_ID;
+}
+
+static inline mtd_loader_id current_mtd_loader(void)
+{
+	return MTD_LOADER_V1;
+}
+
+static inline uint32_t master_key_size (void)
+{
+	return crypt_keys[MASTER_VOL_KEY].len >> 3;
+}
+
+static inline uint32_t nmtd_vol_key_size (void)
+{
+	return crypt_keys[NMTD_VOL_KEY].len >> 3;
+}
+
+static inline uint32_t alg_mode_blkbits(cipher_alg_t alg,
+					cipher_mode_t mode)
+{
+	return data_cipher_algs[alg][mode].blkbits;
+}
+
+static inline uint32_t alg_mode_blksize(cipher_alg_t alg,
+					cipher_mode_t mode)
+{
+	return 1 << alg_mode_blkbits(alg, mode);
+}
+
+static inline gf_boolean_t alg_mode_atomic(cipher_alg_t alg,
+					   cipher_mode_t mode)
+{
+	return data_cipher_algs[alg][mode].atomic;
+}
+
+static inline gf_boolean_t alg_mode_should_pad(cipher_alg_t alg,
+					       cipher_mode_t mode)
+{
+	return data_cipher_algs[alg][mode].should_pad;
+}
+
+static inline uint32_t master_alg_blksize(struct master_cipher_info *mr)
+{
+	return alg_mode_blksize(mr->m_alg, mr->m_mode);
+}
+
+static inline uint32_t master_alg_blkbits(struct master_cipher_info *mr)
+{
+	return alg_mode_blkbits(mr->m_alg, mr->m_mode);
+}
+
+static inline gf_boolean_t master_alg_atomic(struct master_cipher_info *mr)
+{
+	return alg_mode_atomic(mr->m_alg, mr->m_mode);
+}
+
+static inline gf_boolean_t master_alg_should_pad(struct master_cipher_info *mr)
+{
+	return alg_mode_should_pad(mr->m_alg, mr->m_mode);
+}
+
+static inline uint32_t object_alg_blksize(struct object_cipher_info *ob)
+{
+	return alg_mode_blksize(ob->o_alg, ob->o_mode);
+}
+
+static inline uint32_t object_alg_blkbits(struct object_cipher_info *ob)
+{
+	return alg_mode_blkbits(ob->o_alg, ob->o_mode);
+}
+
+static inline gf_boolean_t object_alg_atomic(struct object_cipher_info *ob)
+{
+	return alg_mode_atomic(ob->o_alg, ob->o_mode);
+}
+
+static inline gf_boolean_t object_alg_should_pad(struct object_cipher_info *ob)
+{
+	return alg_mode_should_pad(ob->o_alg, ob->o_mode);
+}
+
+static inline uint32_t aes_raw_key_size(struct master_cipher_info *master)
+{
+	return master->m_dkey_size >> 3;
+}
+
+static inline struct avec_config *get_hole_conf(call_frame_t *frame)
+{
+	return &(((crypt_local_t *)frame->local)->hole_conf);
+}
+
+static inline struct avec_config *get_data_conf(call_frame_t *frame)
+{
+	return &(((crypt_local_t *)frame->local)->data_conf);
+}
+
+static inline int32_t get_atom_bits (struct object_cipher_info *object)
+{
+	return object->o_block_bits;
+}
+
+static inline int32_t get_atom_size (struct object_cipher_info *object)
+{
+	return 1 << get_atom_bits(object);
+}
+
+static inline int32_t has_head_block(struct avec_config *conf)
+{
+	return conf->off_in_head ||
+		(conf->acount == 1 && conf->off_in_tail);
+}
+
+static inline int32_t has_tail_block(struct avec_config *conf)
+{
+	return conf->off_in_tail && conf->acount > 1;
+}
+
+static inline int32_t has_full_blocks(struct avec_config *conf)
+{
+	return conf->nr_full_blocks;
+}
+
+static inline int32_t should_submit_head_block(struct avec_config *conf)
+{
+	return has_head_block(conf) && (conf->cursor == 0);
+}
+
+static inline int32_t should_submit_tail_block(struct avec_config *conf)
+{
+	return has_tail_block(conf) && (conf->cursor == conf->acount - 1);
+}
+
+static inline int32_t should_submit_full_block(struct avec_config *conf)
+{
+	uint32_t start = has_head_block(conf) ? 1 : 0;
+
+	return has_full_blocks(conf) &&
+		conf->cursor >= start &&
+		conf->cursor < start + conf->nr_full_blocks;
+}
+
+#if DEBUG_CRYPT
+static inline void crypt_check_input_len(size_t len,
+					 struct object_cipher_info *object)
+{
+	if (object_alg_should_pad(object) && (len & (object_alg_blksize(object) - 1)))
+		gf_log ("crypt", GF_LOG_DEBUG, "bad input len: %d", (int)len);
+}
+
+static inline void check_head_block(struct avec_config *conf)
+{
+	if (!has_head_block(conf))
+		gf_log("crypt", GF_LOG_DEBUG, "not a head atom");
+}
+
+static inline void check_tail_block(struct avec_config *conf)
+{
+	if (!has_tail_block(conf))
+		gf_log("crypt", GF_LOG_DEBUG, "not a tail atom");
+}
+
+static inline void check_full_block(struct avec_config *conf)
+{
+	if (!has_full_blocks(conf))
+		gf_log("crypt", GF_LOG_DEBUG, "not a full atom");
+}
+
+static inline void check_cursor_head(struct avec_config *conf)
+{
+	if (!has_head_block(conf))
+		gf_log("crypt",
+		       GF_LOG_DEBUG, "Illegal call of head atom method");
+	else if (conf->cursor != 0)
+		gf_log("crypt",
+		       GF_LOG_DEBUG, "Cursor (%d) is not at head atom",
+		       conf->cursor);
+}
+
+static inline void check_cursor_full(struct avec_config *conf)
+{
+	if (!has_full_blocks(conf))
+		gf_log("crypt",
+		       GF_LOG_DEBUG, "Illegal call of full atom method");
+	if (has_head_block(conf) && (conf->cursor == 0))
+		gf_log("crypt",
+		       GF_LOG_DEBUG, "Cursor is not at full atom");
+}
+
+/*
+ * FIXME: use avec->iov_len to check setup
+ */
+static inline int data_local_invariant(crypt_local_t *local)
+{
+	return 0;
+}
+
+#else
+#define crypt_check_input_len(len, object) noop
+#define check_head_block(conf) noop
+#define check_tail_block(conf) noop
+#define check_full_block(conf) noop
+#define check_cursor_head(conf) noop
+#define check_cursor_full(conf) noop
+
+#endif /* DEBUG_CRYPT */
+
+static inline struct avec_config *conf_by_type(call_frame_t *frame,
+					       atom_data_type dtype)
+{
+	struct avec_config *conf = NULL;
+
+	switch (dtype) {
+	case HOLE_ATOM:
+		conf = get_hole_conf(frame);
+		break;
+	case DATA_ATOM:
+		conf = get_data_conf(frame);
+		break;
+	default:
+		gf_log("crypt", GF_LOG_DEBUG, "bad atom type");
+	}
+	return conf;
+}
+
+static inline uint32_t nr_calls_head(struct avec_config *conf)
+{
+	return has_head_block(conf) ? 1 : 0;
+}
+
+static inline uint32_t nr_calls_tail(struct avec_config *conf)
+{
+	return has_tail_block(conf) ? 1 : 0;
+}
+
+static inline uint32_t nr_calls_full(struct avec_config *conf)
+{
+	switch(conf->type) {
+	case HOLE_ATOM:
+		return has_full_blocks(conf);
+	case DATA_ATOM:
+		return has_full_blocks(conf) ?
+			logical_blocks_occupied(0,
+						conf->nr_full_blocks,
+						MAX_IOVEC_BITS)	: 0;
+	default:
+		gf_log("crypt", GF_LOG_DEBUG, "bad atom data type");
+		return 0;
+	}
+}
+
+static inline uint32_t nr_calls(struct avec_config *conf)
+{
+	return nr_calls_head(conf) + nr_calls_tail(conf) + nr_calls_full(conf);
+}
+
+static inline uint32_t nr_calls_data(call_frame_t *frame)
+{
+	return nr_calls(get_data_conf(frame));
+}
+
+static inline uint32_t nr_calls_hole(call_frame_t *frame)
+{
+	return nr_calls(get_hole_conf(frame));
+}
+
+static inline void get_one_call_nolock(call_frame_t *frame)
+{
+	crypt_local_t *local = frame->local;
+
+	++local->nr_calls;
+
+	//gf_log("crypt", GF_LOG_DEBUG, "get %d calls", 1);
+}
+
+static inline void get_one_call(call_frame_t *frame)
+{
+	crypt_local_t *local = frame->local;
+
+	LOCK(&local->call_lock);
+	get_one_call_nolock(frame);
+	UNLOCK(&local->call_lock);
+}
+
+static inline void get_nr_calls_nolock(call_frame_t *frame, int32_t nr)
+{
+	crypt_local_t *local = frame->local;
+
+	local->nr_calls += nr;
+
+	//gf_log("crypt", GF_LOG_DEBUG, "get %d calls", nr);
+}
+
+static inline void get_nr_calls(call_frame_t *frame, int32_t nr)
+{
+	crypt_local_t *local = frame->local;
+
+	LOCK(&local->call_lock);
+	get_nr_calls_nolock(frame, nr);
+	UNLOCK(&local->call_lock);
+}
+
+static inline int put_one_call(crypt_local_t *local)
+{
+	uint32_t last = 0;
+
+	LOCK(&local->call_lock);
+	if (--local->nr_calls == 0)
+		last = 1;
+
+	//gf_log("crypt", GF_LOG_DEBUG, "put %d calls", 1);
+
+	UNLOCK(&local->call_lock);
+	return last;
+}
+
+static inline int is_appended_write(call_frame_t *frame)
+{
+	crypt_local_t *local = frame->local;
+	struct avec_config *conf = get_data_conf(frame);
+
+	return conf->orig_offset + conf->orig_size > local->old_file_size;
+}
+
+static inline int is_ordered_mode(call_frame_t *frame)
+{
+#if 0
+	crypt_local_t *local = frame->local;
+	return local->fop == GF_FOP_FTRUNCATE ||
+		(local->fop == GF_FOP_WRITE && is_appended_write(frame));
+#endif
+	return 1;
+}
+
+static inline int32_t hole_conv_completed(crypt_local_t *local)
+{
+	struct avec_config *conf = &local->hole_conf;
+	return conf->cursor == conf->acount;
+}
+
+static inline int32_t data_write_in_progress(crypt_local_t *local)
+{
+	return local->active_setup == DATA_ATOM;
+}
+
+static inline int32_t parent_is_crypt_xlator(call_frame_t *frame,
+					     xlator_t *this)
+{
+	return frame->parent->this == this;
+}
+
+static inline linkop_wind_handler_t linkop_wind_dispatch(glusterfs_fop_t fop)
+{
+	switch(fop){
+	case GF_FOP_LINK:
+		return link_wind;
+	case GF_FOP_UNLINK:
+		return unlink_wind;
+	case GF_FOP_RENAME:
+		return rename_wind;
+	default:
+		gf_log("crypt", GF_LOG_ERROR, "Bad link operation %d", fop);
+		return NULL;
+	}
+}
+
+static inline linkop_unwind_handler_t linkop_unwind_dispatch(glusterfs_fop_t fop)
+{
+	switch(fop){
+	case GF_FOP_LINK:
+		return link_unwind;
+	case GF_FOP_UNLINK:
+		return unlink_unwind;
+	case GF_FOP_RENAME:
+		return rename_unwind;
+	default:
+		gf_log("crypt", GF_LOG_ERROR, "Bad link operation %d", fop);
+		return NULL;		
+	}	
+}
+
+static inline mtd_op_t linkop_mtdop_dispatch(glusterfs_fop_t fop)
+{
+	switch (fop) {
+	case GF_FOP_LINK:
+		return MTD_APPEND;
+	case GF_FOP_UNLINK:
+		return MTD_CUT;
+	case GF_FOP_RENAME:
+		return MTD_OVERWRITE;
+	default:
+		gf_log("crypt", GF_LOG_WARNING, "Bad link operation %d", fop);
+		return MTD_LAST_OP;
+	}
+}
+
+#endif /* __CRYPT_H__ */
+
+/*
+  Local variables:
+  c-indentation-style: "K&R"
+  mode-name: "LC"
+  c-basic-offset: 8
+  tab-width: 8
+  fill-column: 80
+  scroll-step: 1
+  End:
+*/
diff --git a/xlators/encryption/crypt/src/data.c b/xlators/encryption/crypt/src/data.c
new file mode 100644
index 000000000..762fa554a
--- /dev/null
+++ b/xlators/encryption/crypt/src/data.c
@@ -0,0 +1,769 @@
+/*
+  Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "defaults.h"
+#include "crypt-common.h"
+#include "crypt.h"
+
+static void set_iv_aes_xts(off_t offset, struct object_cipher_info *object)
+{
+	unsigned char *ivec;
+
+	ivec = object->u.aes_xts.ivec;
+
+	/* convert the tweak into a little-endian byte
+	 * array (IEEE P1619/D16, May 2007, section 5.1)
+	 */
+
+	*((uint64_t *)ivec) = htole64(offset);
+
+	/* ivec is padded with zeroes */
+}
+
+static int32_t aes_set_keys_common(unsigned char *raw_key, uint32_t key_size,
+				   AES_KEY *keys)
+{
+	int32_t ret;
+
+	ret = AES_set_encrypt_key(raw_key,
+				  key_size,
+				  &keys[AES_ENCRYPT]);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, "Set encrypt key failed");
+		return ret;
+	}
+	ret = AES_set_decrypt_key(raw_key,
+				  key_size,
+				  &keys[AES_DECRYPT]);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, "Set decrypt key failed");
+		return ret;
+	}
+	return 0;
+}
+
+/*
+ * set private cipher info for xts mode
+ */
+static int32_t set_private_aes_xts(struct crypt_inode_info *info,
+				   struct master_cipher_info *master)
+{
+	int ret;
+	struct object_cipher_info *object = get_object_cinfo(info);
+	unsigned char *data_key;
+	uint32_t subkey_size;
+
+	/* init tweak value */
+	memset(object->u.aes_xts.ivec, 0, 16);
+
+	data_key = GF_CALLOC(1, object->o_dkey_size, gf_crypt_mt_key);
+	if (!data_key)
+		return ENOMEM;
+
+	/*
+	 * retrieve data keying meterial
+	 */
+	ret = get_data_file_key(info, master, object->o_dkey_size, data_key);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, "Failed to retrieve data key");
+		GF_FREE(data_key);
+		return ret;
+	}
+	/*
+	 * parse compound xts key
+	 */
+	subkey_size = object->o_dkey_size >> 4; /* (xts-key-size-in-bytes / 2) */
+	/*
+	 * install key for data encryption
+	 */
+	ret = aes_set_keys_common(data_key,
+				  subkey_size << 3, object->u.aes_xts.dkey);
+	if (ret) {
+		GF_FREE(data_key);
+		return ret;
+	}
+	/*
+	 * set up key used to encrypt tweaks
+	 */
+	ret = AES_set_encrypt_key(data_key + subkey_size,
+				  object->o_dkey_size / 2,
+				  &object->u.aes_xts.tkey);
+	if (ret < 0)
+		gf_log("crypt", GF_LOG_ERROR, "Set tweak key failed");
+
+	GF_FREE(data_key);
+	return ret;
+}
+
+static int32_t aes_xts_init(void)
+{
+	cassert(AES_BLOCK_SIZE == (1 << AES_BLOCK_BITS));
+	return 0;
+}
+
+static int32_t check_key_aes_xts(uint32_t keysize)
+{
+	switch(keysize) {
+	case 256:
+	case 512:
+		return 0;
+	default:
+		break;
+	}
+	return -1;
+}
+
+static int32_t encrypt_aes_xts(const unsigned char *from,
+			       unsigned char *to, size_t length,
+			       off_t offset, const int enc,
+			       struct object_cipher_info *object)
+{
+	XTS128_CONTEXT ctx;
+	if (enc) {
+		ctx.key1 = &object->u.aes_xts.dkey[AES_ENCRYPT];
+		ctx.block1 = (block128_f)AES_encrypt;
+	}
+	else {
+		ctx.key1 = &object->u.aes_xts.dkey[AES_DECRYPT];
+		ctx.block1 = (block128_f)AES_decrypt;
+	}
+	ctx.key2 = &object->u.aes_xts.tkey;
+	ctx.block2 = (block128_f)AES_encrypt;
+
+	return CRYPTO_xts128_encrypt(&ctx,
+				     object->u.aes_xts.ivec,
+				     from,
+				     to,
+				     length, enc);
+}
+
+/*
+ * Cipher input chunk @from of length @len;
+ * @to: result of cipher transform;
+ * @off: offset in a file (must be cblock-aligned);
+ */
+static void cipher_data(struct object_cipher_info *object,
+			char *from,
+			char *to,
+			off_t off,
+			size_t len,
+			const int enc)
+{
+	crypt_check_input_len(len, object);
+
+#if TRIVIAL_TFM && DEBUG_CRYPT
+	return;
+#endif
+	data_cipher_algs[object->o_alg][object->o_mode].set_iv(off, object);
+	data_cipher_algs[object->o_alg][object->o_mode].encrypt
+		((const unsigned char *)from,
+		 (unsigned char *)to,
+		 len,
+		 off,
+		 enc,
+		 object);
+}
+
+#define MAX_CIPHER_CHUNK (1 << 30)
+
+/*
+ * Do cipher (encryption/decryption) transform of a
+ * continuous region of memory.
+ *
+ * @len: a number of bytes to transform;
+ * @buf: data to transform;
+ * @off: offset in a file, should be block-aligned
+ *       for atomic cipher modes and ksize-aligned
+ *       for other modes).
+ * @dir: direction of transform (encrypt/decrypt).
+ */
+static void cipher_region(struct object_cipher_info *object,
+			  char *from,
+			  char *to,
+			  off_t off,
+			  size_t len,
+			  int dir)
+{
+	while (len > 0) {
+		size_t to_cipher;
+
+		to_cipher = len;
+		if (to_cipher > MAX_CIPHER_CHUNK)
+			to_cipher = MAX_CIPHER_CHUNK;
+
+		/* this will reset IV */
+		cipher_data(object,
+			    from,
+			    to,
+			    off,
+			    to_cipher,
+			    dir);
+		from += to_cipher;
+		to   += to_cipher;
+		off  += to_cipher;
+		len  -= to_cipher;
+	}
+}
+
+/*
+ * Do cipher transform (encryption/decryption) of
+ * plaintext/ciphertext represented by @vec.
+ *
+ * Pre-conditions: @vec represents a continuous piece
+ * of data in a file at offset @off to be ciphered
+ * (encrypted/decrypted).
+ * @count is the number of vec's components. All the
+ * components must be block-aligned, the caller is
+ * responsible for this. @dir is "direction" of
+ * transform (encrypt/decrypt).
+ */
+static void cipher_aligned_iov(struct object_cipher_info *object,
+			       struct iovec *vec,
+			       int count,
+			       off_t off,
+			       int32_t dir)
+{
+	int i;
+	int len = 0;
+
+	for (i = 0; i < count; i++) {
+		cipher_region(object,
+			      vec[i].iov_base,
+			      vec[i].iov_base,
+			      off + len,
+			      vec[i].iov_len,
+			      dir);
+		len += vec[i].iov_len;
+	}
+}
+
+void encrypt_aligned_iov(struct object_cipher_info *object,
+			 struct iovec *vec,
+			 int count,
+			 off_t off)
+{
+	cipher_aligned_iov(object, vec, count, off, 1);
+}
+
+void decrypt_aligned_iov(struct object_cipher_info *object,
+			 struct iovec *vec,
+			 int count,
+			 off_t off)
+{
+	cipher_aligned_iov(object, vec, count, off, 0);
+}
+
+#if DEBUG_CRYPT
+static void compound_stream(struct iovec *vec, int count, char *buf, off_t skip)
+{
+	int i;
+	int off = 0;
+	for (i = 0; i < count; i++) {
+		memcpy(buf + off,
+		       vec[i].iov_base + skip,
+		       vec[i].iov_len - skip);
+
+		off += (vec[i].iov_len - skip);
+		skip = 0;
+	}
+}
+
+static void check_iovecs(struct iovec *vec, int cnt,
+			 struct iovec *avec, int acnt, uint32_t off_in_head)
+{
+	char *s1, *s2;
+	uint32_t size, asize;
+
+	size = iovec_get_size(vec, cnt);
+	asize = iovec_get_size(avec, acnt) - off_in_head;
+	if (size != asize) {
+		gf_log("crypt", GF_LOG_DEBUG, "size %d is not eq asize %d",
+		       size, asize);
+		return;
+	}
+	s1 = GF_CALLOC(1, size, gf_crypt_mt_data);
+	if (!s1) {
+		gf_log("crypt", GF_LOG_DEBUG, "Can not allocate stream ");
+		return;
+	}
+	s2 = GF_CALLOC(1, asize, gf_crypt_mt_data);
+	if (!s2) {
+		GF_FREE(s1);
+		gf_log("crypt", GF_LOG_DEBUG, "Can not allocate stream ");
+		return;
+	}
+	compound_stream(vec, cnt, s1, 0);
+	compound_stream(avec, acnt, s2, off_in_head);
+	if (memcmp(s1, s2, size))
+		gf_log("crypt", GF_LOG_DEBUG, "chunks of different data");
+	GF_FREE(s1);
+	GF_FREE(s2);
+}
+
+#else
+#define check_iovecs(vec, count, avec, avecn, off) noop
+#endif /* DEBUG_CRYPT */
+
+static char *data_alloc_block(xlator_t *this, crypt_local_t *local,
+			      int32_t block_size)
+{
+	struct iobuf *iobuf = NULL;
+
+	iobuf = iobuf_get2(this->ctx->iobuf_pool, block_size);
+	if (!iobuf) {
+		gf_log("crypt", GF_LOG_ERROR,
+		       "Failed to get iobuf");
+		return NULL;
+	}
+	if (!local->iobref_data) {
+		local->iobref_data = iobref_new();
+		if (!local->iobref_data) {
+			gf_log("crypt", GF_LOG_ERROR,
+			       "Failed to get iobref");
+			iobuf_unref(iobuf);
+			return NULL;
+		}
+	}
+	iobref_add(local->iobref_data, iobuf);
+	return iobuf->ptr;
+}
+
+/*
+ * Compound @avec, which represent the same data
+ * chunk as @vec, but has aligned components of
+ * specified block size. Alloc blocks, if needed.
+ * In particular, incomplete head and tail blocks
+ * must be allocated.
+ * Put number of allocated blocks to @num_blocks.
+ *
+ * Example:
+ *
+ * input: data chunk represented by 4 components
+ * [AB],[BC],[CD],[DE];
+ * output: 5 logical blocks (0, 1, 2, 3, 4).
+ *
+ *   A     B       C     D               E
+ *   *-----*+------*-+---*----+--------+-*
+ *   |     ||      | |   |    |        | |
+ * *-+-----+*------+-*---+----*--------*-+------*
+ * 0        1        2        3        4
+ *
+ * 0 - incomplete compound (head);
+ * 1, 2 - full compound;
+ * 3 - full non-compound (the case of reuse);
+ * 4 - incomplete non-compound (tail).
+ */
+int32_t align_iov_by_atoms(xlator_t *this,
+			   crypt_local_t *local,
+			   struct object_cipher_info *object,
+			   struct iovec *vec /* input vector */,
+			   int32_t count /* number of vec components */,
+			   struct iovec *avec /* aligned vector */,
+			   char **blocks /* pool of blocks */,
+			   uint32_t *blocks_allocated,
+			   struct avec_config *conf)
+{
+	int vecn = 0;      /* number of the current component in vec */
+	int avecn = 0;     /* number of the current component in avec */
+	off_t vec_off = 0; /* offset in the current vec component,
+			    * i.e. the number of bytes have already
+			    * been copied */
+	int32_t block_size = get_atom_size(object);
+	size_t to_process; /* number of vec's bytes to copy and(or) re-use */
+	int32_t off_in_head = conf->off_in_head;
+
+	to_process = iovec_get_size(vec, count);
+
+	while (to_process > 0) {
+		if (off_in_head ||
+		    vec[vecn].iov_len - vec_off < block_size) {
+			/*
+			 * less than block_size:
+			 * the case of incomplete (head or tail),
+			 * or compound block
+			 */
+			size_t copied = 0;
+			/*
+			 * populate the pool with a new block
+			 */
+			blocks[*blocks_allocated] = data_alloc_block(this,
+								    local,
+								    block_size);
+			if (!blocks[*blocks_allocated])
+				return -ENOMEM;
+			memset(blocks[*blocks_allocated], 0, off_in_head);
+			/*
+			 * fill the block with vec components
+			 */
+			do {
+				size_t to_copy;
+
+				to_copy = vec[vecn].iov_len - vec_off;
+				if (to_copy > block_size - off_in_head)
+					to_copy = block_size - off_in_head;
+
+				memcpy(blocks[*blocks_allocated] + off_in_head + copied,
+				       vec[vecn].iov_base + vec_off,
+				       to_copy);
+
+				copied += to_copy;
+				to_process -= to_copy;
+
+				vec_off += to_copy;
+				if (vec_off == vec[vecn].iov_len) {
+					/* finished with this vecn */
+					vec_off = 0;
+					vecn++;
+				}
+			} while (copied < (block_size - off_in_head) && to_process > 0);
+			/*
+			 * update avec
+			 */
+			avec[avecn].iov_len = off_in_head + copied;
+			avec[avecn].iov_base = blocks[*blocks_allocated];
+
+			(*blocks_allocated)++;
+			off_in_head = 0;
+		} else {
+			/*
+			 * the rest of the current vec component
+			 * is not less than block_size, so reuse
+			 * the memory buffer of the component.
+			 */
+			size_t to_reuse;
+			to_reuse = (to_process > block_size ?
+				    block_size :
+				    to_process);
+			avec[avecn].iov_len = to_reuse;
+			avec[avecn].iov_base = vec[vecn].iov_base + vec_off;
+
+			vec_off += to_reuse;
+			if (vec_off == vec[vecn].iov_len) {
+				/* finished with this vecn */
+				vec_off = 0;
+				vecn++;
+			}
+			to_process -= to_reuse;
+		}
+		avecn++;
+	}
+	check_iovecs(vec, count, avec, avecn, conf->off_in_head);
+	return 0;
+}
+
+/*
+ * allocate and setup aligned vector for data submission
+ * Pre-condition: @conf is set.
+ */
+int32_t set_config_avec_data(xlator_t *this,
+			     crypt_local_t *local,
+			     struct avec_config *conf,
+			     struct object_cipher_info *object,
+			     struct iovec *vec,
+			     int32_t vec_count)
+{
+	int32_t ret = ENOMEM;
+	struct iovec *avec;
+	char **pool;
+	uint32_t blocks_in_pool = 0;
+
+	conf->type = DATA_ATOM;
+
+	avec = GF_CALLOC(conf->acount, sizeof(*avec), gf_crypt_mt_iovec);
+	if (!avec)
+		return ret;
+	pool = GF_CALLOC(conf->acount, sizeof(pool), gf_crypt_mt_char);
+	if (!pool) {
+		GF_FREE(avec);
+		return ret;
+	}
+	if (!vec) {
+		/*
+		 * degenerated case: no data
+		 */
+		pool[0] = data_alloc_block(this, local, get_atom_size(object));
+		if (!pool[0])
+			goto free;
+		blocks_in_pool = 1;
+		avec->iov_base = pool[0];
+		avec->iov_len = conf->off_in_tail;
+	}
+	else {
+		ret = align_iov_by_atoms(this, local, object, vec, vec_count,
+					 avec, pool, &blocks_in_pool, conf);
+		if (ret)
+			goto free;
+	}
+	conf->avec = avec;
+	conf->pool = pool;
+	conf->blocks_in_pool = blocks_in_pool;
+	return 0;
+ free:
+	GF_FREE(avec);
+	GF_FREE(pool);
+	return ret;
+}
+
+/*
+ * allocate and setup aligned vector for hole submission
+ */
+int32_t set_config_avec_hole(xlator_t *this,
+			     crypt_local_t *local,
+			     struct avec_config *conf,
+			     struct object_cipher_info *object,
+			     glusterfs_fop_t fop)
+{
+	uint32_t i, idx;
+	struct iovec *avec;
+	char **pool;
+	uint32_t num_blocks;
+	uint32_t blocks_in_pool = 0;
+
+	conf->type = HOLE_ATOM;
+
+	num_blocks = conf->acount -
+		(conf->nr_full_blocks ? conf->nr_full_blocks - 1 : 0);
+
+	switch (fop) {
+	case GF_FOP_WRITE:
+		/*
+		 * hole goes before data
+		 */
+		if (num_blocks == 1 && conf->off_in_tail != 0)
+		/*
+		 * we won't submit a hole which fits into
+		 * a data atom: this part of hole will be
+		 * submitted with data write
+		 */
+			return 0;
+		break;
+	case GF_FOP_FTRUNCATE:
+		/*
+		 * expanding truncate, hole goes after data,
+		 * and will be submited in any case.
+ 		 */
+		break;
+	default:
+		gf_log("crypt", GF_LOG_WARNING,
+		       "bad file operation %d", fop);
+		return 0;
+	}
+	avec = GF_CALLOC(num_blocks, sizeof(*avec), gf_crypt_mt_iovec);
+	if (!avec)
+		return ENOMEM;
+	pool = GF_CALLOC(num_blocks, sizeof(pool), gf_crypt_mt_char);
+	if (!pool) {
+		GF_FREE(avec);
+		return ENOMEM;
+	}
+	for (i = 0; i < num_blocks; i++) {
+		pool[i] = data_alloc_block(this, local, get_atom_size(object));
+		if (pool[i] == NULL)
+			goto free;
+		blocks_in_pool++;
+	}
+	if (has_head_block(conf)) {
+		/* set head block */
+		idx = 0;
+		avec[idx].iov_base = pool[idx];
+		avec[idx].iov_len = get_atom_size(object);
+		memset(avec[idx].iov_base + conf->off_in_head,
+		       0,
+		       get_atom_size(object) - conf->off_in_head);
+	}
+	if (has_tail_block(conf)) {
+		/* set tail block */
+		idx = num_blocks - 1;
+		avec[idx].iov_base = pool[idx];
+		avec[idx].iov_len = get_atom_size(object);
+		memset(avec[idx].iov_base, 0, conf->off_in_tail);
+	}
+	if (has_full_blocks(conf)) {
+		/* set full block */
+		idx = conf->off_in_head ? 1 : 0;
+		avec[idx].iov_base = pool[idx];
+		avec[idx].iov_len = get_atom_size(object);
+		/*
+		 * since we re-use the buffer,
+		 * zeroes will be set every time
+		 * before encryption, see submit_full()
+		 */
+	}
+	conf->avec = avec;
+	conf->pool = pool;
+	conf->blocks_in_pool = blocks_in_pool;
+	return 0;
+ free:
+	GF_FREE(avec);
+	GF_FREE(pool);
+	return ENOMEM;
+}
+
+/* A helper for setting up config of partial atoms (which
+ * participate in read-modify-write sequence).
+ *
+ * Calculate and setup precise amount of "extra-bytes"
+ * that should be uptodated at the end of partial (not
+ * necessarily tail!) block.
+ *
+ * Pre-condition: local->old_file_size is valid!
+ * @conf contains setup, which is enough for correct calculation
+ * of has_tail_block(), ->get_offset().
+ */
+void set_gap_at_end(call_frame_t *frame, struct object_cipher_info *object,
+		    struct avec_config *conf, atom_data_type dtype)
+{
+	uint32_t to_block;
+	crypt_local_t *local = frame->local;
+	uint64_t old_file_size = local->old_file_size;
+	struct rmw_atom *partial = atom_by_types(dtype,
+						 has_tail_block(conf) ?
+						 TAIL_ATOM : HEAD_ATOM);
+
+	if (old_file_size <= partial->offset_at(frame, object))
+		to_block = 0;
+	else {
+		to_block = old_file_size - partial->offset_at(frame, object);
+		if (to_block > get_atom_size(object))
+			to_block = get_atom_size(object);
+	}
+	if (to_block > conf->off_in_tail)
+		conf->gap_in_tail = to_block - conf->off_in_tail;
+	else
+		/*
+		 * nothing to uptodate
+		 */
+		conf->gap_in_tail = 0;
+}
+
+/*
+ * fill struct avec_config with offsets layouts
+ */
+void set_config_offsets(call_frame_t *frame,
+			xlator_t *this,
+			uint64_t offset,
+			uint64_t count,
+			atom_data_type dtype,
+			int32_t set_gap)
+{
+	crypt_local_t *local;
+	struct object_cipher_info *object;
+	struct avec_config *conf;
+	uint32_t resid;
+
+	uint32_t atom_size;
+	uint32_t atom_bits;
+
+	size_t orig_size;
+	off_t orig_offset;
+	size_t expanded_size;
+	off_t aligned_offset;
+
+	uint32_t off_in_head = 0;
+	uint32_t off_in_tail = 0;
+	uint32_t nr_full_blocks;
+	int32_t size_full_blocks;
+
+	uint32_t acount; /* number of alifned components to write.
+			  * The same as number of occupied logical
+			  * blocks (atoms)
+			  */
+	local = frame->local;
+	object = &local->info->cinfo;
+	conf = (dtype == DATA_ATOM ?
+		get_data_conf(frame) : get_hole_conf(frame));
+
+	orig_offset = offset;
+	orig_size = count;
+
+	atom_size = get_atom_size(object);
+	atom_bits = get_atom_bits(object);
+
+	/*
+	 * Round-down the start,
+	 * round-up the end.
+	 */
+	resid = offset & (uint64_t)(atom_size - 1);
+
+	if (resid)
+		off_in_head = resid;
+	aligned_offset = offset - off_in_head;
+	expanded_size = orig_size + off_in_head;
+
+	/* calculate tail,
+	   expand size forward  */
+	resid = (offset + orig_size) & (uint64_t)(atom_size - 1);
+
+	if (resid) {
+		off_in_tail = resid;
+		expanded_size += (atom_size - off_in_tail);
+	}
+	/*
+	 * calculate number of occupied blocks
+	 */
+	acount = expanded_size >> atom_bits;
+	/*
+	 * calculate number of full blocks
+	 */
+	size_full_blocks = expanded_size;
+	if (off_in_head)
+		size_full_blocks -= atom_size;
+	if (off_in_tail && size_full_blocks > 0)
+		size_full_blocks -= atom_size;
+	nr_full_blocks = size_full_blocks >> atom_bits;
+
+	conf->atom_size = atom_size;
+	conf->orig_size = orig_size;
+	conf->orig_offset = orig_offset;
+	conf->expanded_size = expanded_size;
+	conf->aligned_offset = aligned_offset;
+
+	conf->off_in_head = off_in_head;
+	conf->off_in_tail = off_in_tail;
+	conf->nr_full_blocks = nr_full_blocks;
+	conf->acount = acount;
+	/*
+	 * Finally, calculate precise amount of
+	 * "extra-bytes" that should be uptodated
+	 * at the end.
+	 * Only if RMW is expected.
+	 */
+	if (off_in_tail && set_gap)
+		set_gap_at_end(frame, object, conf, dtype);
+}
+
+struct data_cipher_alg data_cipher_algs[LAST_CIPHER_ALG][LAST_CIPHER_MODE] = {
+	[AES_CIPHER_ALG][XTS_CIPHER_MODE] =
+	{ .atomic = _gf_true,
+	  .should_pad = _gf_true,
+	  .blkbits = AES_BLOCK_BITS,
+	  .init = aes_xts_init,
+	  .set_private = set_private_aes_xts,
+	  .check_key = check_key_aes_xts,
+	  .set_iv = set_iv_aes_xts,
+	  .encrypt = encrypt_aes_xts
+	}
+};
+
+/*
+  Local variables:
+  c-indentation-style: "K&R"
+  mode-name: "LC"
+  c-basic-offset: 8
+  tab-width: 8
+  fill-column: 80
+  scroll-step: 1
+  End:
+*/
diff --git a/xlators/encryption/crypt/src/keys.c b/xlators/encryption/crypt/src/keys.c
new file mode 100644
index 000000000..4a1d3bb5a
--- /dev/null
+++ b/xlators/encryption/crypt/src/keys.c
@@ -0,0 +1,302 @@
+/*
+  Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "defaults.h"
+#include "crypt-common.h"
+#include "crypt.h"
+
+/* Key hierarchy
+                  
+                   +----------------+
+                   | MASTER_VOL_KEY |
+                   +-------+--------+
+                           |
+                           |
+          +----------------+----------------+
+          |                |                |
+          |                |                |
+  +-------+------+ +-------+-------+ +------+--------+
+  | NMTD_VOL_KEY | | EMTD_FILE_KEY | | DATA_FILE_KEY |
+  +-------+------+ +---------------+ +---------------+
+          |
+          |
+  +-------+-------+
+  | NMTD_LINK_KEY |
+  +---------------+
+
+ */
+
+#if DEBUG_CRYPT
+static void check_prf_iters(uint32_t num_iters)
+{
+	if (num_iters == 0)
+		gf_log ("crypt", GF_LOG_DEBUG,
+			"bad number of prf iterations : %d", num_iters);
+}
+#else
+#define check_prf_iters(num_iters) noop
+#endif /* DEBUG_CRYPT */
+
+unsigned char crypt_fake_oid[16] =
+	{0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0};
+
+/*
+ * derive key in the counter mode using
+ * sha256-based HMAC as PRF, see
+ * NIST Special Publication 800-108, 5.1)
+ */
+
+#define PRF_OUTPUT_SIZE  SHA256_DIGEST_LENGTH
+
+static int32_t kderive_init(struct kderive_context *ctx,
+			    const unsigned char *pkey, /* parent key */
+			    uint32_t pkey_size,  /* parent key size */
+			    const unsigned char *idctx, /* id-context */
+			    uint32_t idctx_size,
+			    crypt_key_type type  /* type of child key */)
+{
+	unsigned char *pos;
+	uint32_t llen = strlen(crypt_keys[type].label);	
+	/*
+	 * Compoud the fixed input data for KDF:
+	 * [i]_2 || Label || 0x00 || Id-Context || [L]_2),
+	 * NIST SP 800-108, 5.1
+	 */
+	ctx->fid_len =
+		sizeof(uint32_t) +
+		llen +
+		1 +
+		idctx_size +
+		sizeof(uint32_t);
+
+	ctx->fid = GF_CALLOC(ctx->fid_len, 1, gf_crypt_mt_key);
+	if (!ctx->fid)
+		return ENOMEM;
+	ctx->out_len = round_up(crypt_keys[type].len >> 3,
+				PRF_OUTPUT_SIZE);
+	ctx->out = GF_CALLOC(ctx->out_len, 1, gf_crypt_mt_key);
+	if (!ctx->out) {
+		GF_FREE(ctx->fid);
+		return ENOMEM;
+	}
+	ctx->pkey = pkey;
+	ctx->pkey_len = pkey_size;
+	ctx->ckey_len = crypt_keys[type].len;
+
+	pos = ctx->fid;
+
+	/* counter will be set up in kderive_rfn() */
+	pos += sizeof(uint32_t);
+
+	memcpy(pos, crypt_keys[type].label, llen);
+	pos += llen;
+
+	/* set up zero octet */
+	*pos = 0;
+	pos += 1;
+
+	memcpy(pos, idctx, idctx_size);
+	pos += idctx_size;
+
+	*((uint32_t *)pos) = htobe32(ctx->ckey_len);
+
+	return 0;
+}
+
+static void kderive_update(struct kderive_context *ctx)
+{	
+	uint32_t i;
+	HMAC_CTX hctx;
+	unsigned char *pos = ctx->out;
+	uint32_t *p_iter = (uint32_t *)ctx->fid;
+	uint32_t num_iters = ctx->out_len / PRF_OUTPUT_SIZE;
+
+	check_prf_iters(num_iters);
+
+	HMAC_CTX_init(&hctx);
+	for (i = 0; i < num_iters; i++) {
+		/*
+		 * update the iteration number in the fid
+		 */
+		*p_iter = htobe32(i);
+		HMAC_Init_ex(&hctx,
+			     ctx->pkey, ctx->pkey_len >> 3,
+			     EVP_sha256(),
+			     NULL);
+		HMAC_Update(&hctx, ctx->fid, ctx->fid_len);
+		HMAC_Final(&hctx, pos, NULL);
+
+		pos += PRF_OUTPUT_SIZE;
+	}
+	HMAC_CTX_cleanup(&hctx);
+}
+
+static void kderive_final(struct kderive_context *ctx, unsigned char *child)
+{
+	memcpy(child, ctx->out, ctx->ckey_len >> 3);
+	GF_FREE(ctx->fid);
+	GF_FREE(ctx->out);
+	memset(ctx, 0, sizeof(*ctx));
+}
+
+/*
+ * derive per-volume key for object ids aithentication
+ */
+int32_t get_nmtd_vol_key(struct master_cipher_info *master)
+{
+	int32_t ret;
+	struct kderive_context ctx;
+
+	ret = kderive_init(&ctx,
+			   master->m_key,
+			   master_key_size(),
+			   crypt_fake_oid, sizeof(uuid_t), NMTD_VOL_KEY);
+	if (ret)
+		return ret;
+	kderive_update(&ctx);
+	kderive_final(&ctx, master->m_nmtd_key);
+	return 0;
+}
+
+/*
+ * derive per-link key for aithentication of non-encrypted
+ * meta-data (nmtd)
+ */
+int32_t get_nmtd_link_key(loc_t *loc,
+			  struct master_cipher_info *master,
+			  unsigned char *result)
+{
+	int32_t ret;
+	struct kderive_context ctx;
+
+	ret = kderive_init(&ctx,
+			   master->m_nmtd_key,
+			   nmtd_vol_key_size(),
+			   (const unsigned char *)loc->path,
+			   strlen(loc->path), NMTD_LINK_KEY);
+	if (ret)
+		return ret;
+	kderive_update(&ctx);
+	kderive_final(&ctx, result);
+	return 0;
+}
+
+/*
+ * derive per-file key for encryption and authentication
+ * of encrypted part of metadata (emtd)
+ */
+int32_t get_emtd_file_key(struct crypt_inode_info *info,
+			  struct master_cipher_info *master,
+			  unsigned char *result)
+{
+	int32_t ret;
+	struct kderive_context ctx;
+
+	ret = kderive_init(&ctx,
+			   master->m_key,
+			   master_key_size(),
+			   info->oid, sizeof(uuid_t), EMTD_FILE_KEY);
+	if (ret)
+		return ret;
+	kderive_update(&ctx);
+	kderive_final(&ctx, result);
+	return 0;
+}
+
+static int32_t data_key_type_by_size(uint32_t keysize, crypt_key_type *type)
+{
+	int32_t ret = 0;
+	switch (keysize) {
+	case 256:
+		*type = DATA_FILE_KEY_256;
+		break;
+	case 512:
+		*type = DATA_FILE_KEY_512;
+		break;
+	default:
+		gf_log("crypt", GF_LOG_ERROR, "Unsupported data key size %d",
+		       keysize);
+		ret = ENOTSUP;
+		break;
+	}
+	return ret;
+}
+
+/*
+ * derive per-file key for data encryption
+ */
+int32_t get_data_file_key(struct crypt_inode_info *info,
+			  struct master_cipher_info *master,
+			  uint32_t keysize,
+			  unsigned char *key)
+{
+	int32_t ret;
+	struct kderive_context ctx;
+	crypt_key_type type;
+
+	ret = data_key_type_by_size(keysize, &type);
+	if (ret)
+		return ret;
+	ret = kderive_init(&ctx,
+			   master->m_key,
+			   master_key_size(),
+			   info->oid, sizeof(uuid_t), type);
+	if (ret)
+		return ret;
+	kderive_update(&ctx);
+	kderive_final(&ctx, key);
+	return 0;
+}
+
+/*
+ * NOTE: Don't change existing keys: it will break compatibility;
+ */
+struct crypt_key crypt_keys[LAST_KEY_TYPE] = {
+	[MASTER_VOL_KEY] =
+	{ .len = MASTER_VOL_KEY_SIZE << 3,
+	  .label = "volume-master",
+	},
+	[NMTD_VOL_KEY] =
+	{ .len = NMTD_VOL_KEY_SIZE << 3,
+	  .label = "volume-nmtd-key-generation"
+	},
+	[NMTD_LINK_KEY] =
+	{ .len = 128,
+	  .label = "link-nmtd-authentication"
+	},
+	[EMTD_FILE_KEY] =
+	{ .len = 128,
+	  .label = "file-emtd-encryption-and-auth"
+	},
+	[DATA_FILE_KEY_256] =
+	{ .len = 256,
+	  .label = "file-data-encryption-256"
+	},
+	[DATA_FILE_KEY_512] =
+	{ .len = 512,
+	  .label = "file-data-encryption-512"
+	}
+};
+
+/*
+  Local variables:
+  c-indentation-style: "K&R"
+  mode-name: "LC"
+  c-basic-offset: 8
+  tab-width: 8
+  fill-column: 80
+  scroll-step: 1
+  End:
+*/
diff --git a/xlators/encryption/crypt/src/metadata.c b/xlators/encryption/crypt/src/metadata.c
new file mode 100644
index 000000000..36b14c055
--- /dev/null
+++ b/xlators/encryption/crypt/src/metadata.c
@@ -0,0 +1,605 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "defaults.h"
+#include "crypt-common.h"
+#include "crypt.h"
+#include "metadata.h"
+
+int32_t alloc_format(crypt_local_t *local, size_t size)
+{
+	if (size > 0) {
+		local->format = GF_CALLOC(1, size, gf_crypt_mt_mtd);
+		if (!local->format)
+			return ENOMEM;
+	}
+	local->format_size = size;
+	return 0;
+}
+
+int32_t alloc_format_create(crypt_local_t *local)
+{
+	return alloc_format(local, new_format_size());
+}
+
+void free_format(crypt_local_t *local)
+{
+	GF_FREE(local->format);
+}
+
+/*
+ * Check compatibility with extracted metadata
+ */
+static int32_t check_file_metadata(struct crypt_inode_info *info)
+{
+	struct object_cipher_info *object = &info->cinfo;
+
+	if (info->nr_minor != CRYPT_XLATOR_ID) {
+		gf_log("crypt", GF_LOG_WARNING,
+		       "unsupported minor subversion %d", info->nr_minor);
+		return EINVAL;
+	}
+	if (object->o_alg > LAST_CIPHER_ALG) {
+		gf_log("crypt", GF_LOG_WARNING,
+		       "unsupported cipher algorithm %d",
+		       object->o_alg);
+		return EINVAL;
+	}
+	if (object->o_mode > LAST_CIPHER_MODE) {
+		gf_log("crypt", GF_LOG_WARNING,
+		       "unsupported cipher mode %d",
+		       object->o_mode);
+		return EINVAL;
+	}
+	if (object->o_block_bits < CRYPT_MIN_BLOCK_BITS ||
+	    object->o_block_bits > CRYPT_MAX_BLOCK_BITS) {
+		gf_log("crypt", GF_LOG_WARNING, "unsupported block bits %d",
+		       object->o_block_bits);
+		return EINVAL;
+	}
+	/* TBD: check data key size */
+	return 0;
+}
+
+static size_t format_size_v1(mtd_op_t op, size_t old_size)
+{
+	
+	switch (op) {
+	case MTD_CREATE:
+		return sizeof(struct mtd_format_v1);
+	case MTD_OVERWRITE:
+		return old_size;
+	case MTD_APPEND:
+		return old_size + NMTD_8_MAC_SIZE;
+	case MTD_CUT:
+		if (old_size > sizeof(struct mtd_format_v1))
+			return old_size - NMTD_8_MAC_SIZE;
+		else
+			return 0;
+	default:
+		gf_log("crypt", GF_LOG_WARNING, "Bad mtd operation");
+		return 0;
+	}
+}
+
+/*
+ * Calculate size of the updated format string.
+ * Returned zero means that we don't need to update the format string.
+ */
+size_t format_size(mtd_op_t op, size_t old_size)
+{
+	size_t versioned;
+
+	versioned = mtd_loaders[current_mtd_loader()].format_size(op,
+				 old_size - sizeof(struct crypt_format));
+	if (versioned != 0)
+		return versioned + sizeof(struct crypt_format);
+	return 0;
+}
+
+/*
+ * size of the format string of newly created file (nr_links = 1)
+ */
+size_t new_format_size(void)
+{
+	return format_size(MTD_CREATE, 0);
+}
+
+/*
+ * Calculate per-link MAC by pathname
+ */
+static int32_t calc_link_mac_v1(struct mtd_format_v1 *fmt,
+				loc_t *loc,
+				unsigned char *result,
+				struct crypt_inode_info *info,
+				struct master_cipher_info *master)
+{
+	int32_t ret;	
+	unsigned char nmtd_link_key[16];
+	CMAC_CTX *cctx;
+	size_t len;
+
+	ret = get_nmtd_link_key(loc, master, nmtd_link_key);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, "Can not get nmtd link key");
+		return -1;
+	}
+	cctx = CMAC_CTX_new();
+	if (!cctx) {
+		gf_log("crypt", GF_LOG_ERROR, "CMAC_CTX_new failed");
+		return -1; 
+	}
+	ret = CMAC_Init(cctx, nmtd_link_key, sizeof(nmtd_link_key),
+			EVP_aes_128_cbc(), 0);
+	if (!ret) {
+		gf_log("crypt", GF_LOG_ERROR, "CMAC_Init failed");
+		CMAC_CTX_free(cctx);
+		return -1;
+	}
+	ret = CMAC_Update(cctx, get_NMTD_V1(info), SIZE_OF_NMTD_V1);
+	if (!ret) {
+		gf_log("crypt", GF_LOG_ERROR, "CMAC_Update failed");
+		CMAC_CTX_free(cctx);
+		return -1;
+	}
+	ret = CMAC_Final(cctx, result, &len);
+	CMAC_CTX_free(cctx);
+	if (!ret) {
+		gf_log("crypt", GF_LOG_ERROR, "CMAC_Final failed");
+		return -1;
+	}
+	return 0;
+}
+
+/*
+ * Create per-link MAC of index @idx by pathname
+ */
+static int32_t create_link_mac_v1(struct mtd_format_v1 *fmt,
+				  uint32_t idx,
+				  loc_t *loc,
+				  struct crypt_inode_info *info,
+				  struct master_cipher_info *master)
+{
+	int32_t ret;
+	unsigned char *mac;
+	unsigned char cmac[16];
+
+	mac = get_NMTD_V1_MAC(fmt) + idx * SIZE_OF_NMTD_V1_MAC;
+
+	ret = calc_link_mac_v1(fmt, loc, cmac, info, master);
+	if (ret)
+		return -1;
+	memcpy(mac, cmac, SIZE_OF_NMTD_V1_MAC);
+	return 0;
+}
+
+static int32_t create_format_v1(unsigned char *wire,
+				loc_t *loc,
+				struct crypt_inode_info *info,
+				struct master_cipher_info *master)
+{
+	int32_t ret;
+	struct mtd_format_v1 *fmt;
+	unsigned char mtd_key[16];
+	AES_KEY EMTD_KEY;
+	unsigned char nmtd_link_key[16];
+	uint32_t ad;
+	GCM128_CONTEXT *gctx;
+
+	fmt = (struct mtd_format_v1 *)wire;
+
+	fmt->minor_id = info->nr_minor;
+	fmt->alg_id = AES_CIPHER_ALG;
+	fmt->dkey_factor = master->m_dkey_size >> KEY_FACTOR_BITS;
+	fmt->block_bits = master->m_block_bits;
+	fmt->mode_id = master->m_mode;
+	/*
+	 * retrieve keys for the parts of metadata
+	 */
+	ret = get_emtd_file_key(info, master, mtd_key);
+	if (ret)
+		return ret;
+	ret = get_nmtd_link_key(loc, master, nmtd_link_key);
+	if (ret)
+		return ret;
+	
+	AES_set_encrypt_key(mtd_key, sizeof(mtd_key)*8, &EMTD_KEY);
+
+	gctx = CRYPTO_gcm128_new(&EMTD_KEY, (block128_f)AES_encrypt);
+
+	/* TBD: Check return values */
+
+	CRYPTO_gcm128_setiv(gctx, info->oid, sizeof(uuid_t));
+
+	ad = htole32(MTD_LOADER_V1);
+	ret = CRYPTO_gcm128_aad(gctx, (const unsigned char *)&ad, sizeof(ad));
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, " CRYPTO_gcm128_aad failed");
+		CRYPTO_gcm128_release(gctx);
+		return ret;
+	}
+	ret = CRYPTO_gcm128_encrypt(gctx,
+				    get_EMTD_V1(fmt),
+				    get_EMTD_V1(fmt),
+				    SIZE_OF_EMTD_V1);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, " CRYPTO_gcm128_encrypt failed");
+		CRYPTO_gcm128_release(gctx);
+		return ret;
+	}
+	/*
+	 * set MAC of encrypted part of metadata
+	 */
+	CRYPTO_gcm128_tag(gctx, get_EMTD_V1_MAC(fmt), SIZE_OF_EMTD_V1_MAC);
+	CRYPTO_gcm128_release(gctx);
+	/*
+	 * set the first MAC of non-encrypted part of metadata 
+	 */
+	return create_link_mac_v1(fmt, 0, loc, info, master);
+}
+
+/*
+ * Called by fops:
+ * ->create();
+ * ->link();
+ *
+ * Pack common and version-specific parts of file's metadata
+ * Pre-conditions: @info contains valid object-id.
+ */
+int32_t create_format(unsigned char *wire,
+		      loc_t *loc,
+		      struct crypt_inode_info *info,
+		      struct master_cipher_info *master)
+{
+	struct crypt_format *fmt = (struct crypt_format *)wire;
+
+	fmt->loader_id = current_mtd_loader();
+
+	wire += sizeof(struct crypt_format);
+	return mtd_loaders[current_mtd_loader()].create_format(wire, loc,
+							       info, master);
+}
+
+/*
+ * Append or overwrite per-link mac of @mac_idx index
+ * in accordance with the new pathname
+ */
+int32_t appov_link_mac_v1(unsigned char *new,
+			  unsigned char *old,
+			  uint32_t old_size,
+			  int32_t mac_idx,
+			  loc_t *loc,
+			  struct crypt_inode_info *info,
+			  struct master_cipher_info *master,
+			  crypt_local_t *local)
+{
+	memcpy(new, old, old_size);
+	return create_link_mac_v1((struct mtd_format_v1 *)new, mac_idx,
+				  loc, info, master);
+}
+
+/*
+ * Cut per-link mac of @mac_idx index
+ */
+static int32_t cut_link_mac_v1(unsigned char *new,
+			       unsigned char *old,			       
+			       uint32_t old_size,
+			       int32_t mac_idx,
+			       loc_t *loc,
+			       struct crypt_inode_info *info,
+			       struct master_cipher_info *master,
+			       crypt_local_t *local)
+{
+	memcpy(new,
+	       old,
+	       sizeof(struct mtd_format_v1) + NMTD_8_MAC_SIZE * (mac_idx - 1));
+
+	memcpy(new + sizeof(struct mtd_format_v1) + NMTD_8_MAC_SIZE * (mac_idx - 1),
+	       old + sizeof(struct mtd_format_v1) + NMTD_8_MAC_SIZE * mac_idx,
+	       old_size - (sizeof(struct mtd_format_v1) + NMTD_8_MAC_SIZE * mac_idx));
+	return 0;
+}
+
+int32_t update_format_v1(unsigned char *new,
+			 unsigned char *old,
+				size_t old_len,
+				int32_t mac_idx, /* of old name */
+				mtd_op_t op,
+				loc_t *loc,
+				struct crypt_inode_info *info,
+				struct master_cipher_info *master,
+				crypt_local_t *local)
+{
+	switch (op) {
+	case MTD_APPEND:
+		mac_idx = 1 + (old_len - sizeof(struct mtd_format_v1))/8;
+	case MTD_OVERWRITE:
+		return appov_link_mac_v1(new, old, old_len, mac_idx,
+					 loc, info, master, local);
+	case MTD_CUT:
+		return cut_link_mac_v1(new, old, old_len, mac_idx,
+				       loc, info, master, local);
+	default:
+		gf_log("crypt", GF_LOG_ERROR, "Bad  mtd operation %d", op);
+		return -1;
+	}
+}
+
+/*
+ * Called by fops:
+ *
+ * ->link()
+ * ->unlink()
+ * ->rename()
+ *
+ */
+int32_t update_format(unsigned char *new,
+		      unsigned char *old,
+		      size_t old_len,
+		      int32_t mac_idx,
+		      mtd_op_t op,
+		      loc_t *loc,
+		      struct crypt_inode_info *info,
+		      struct master_cipher_info *master,
+		      crypt_local_t *local)
+{
+	if (!new)
+		return 0;
+	memcpy(new, old, sizeof(struct crypt_format));
+
+	old += sizeof(struct crypt_format);
+	new += sizeof(struct crypt_format);
+	old_len -= sizeof(struct crypt_format);
+
+	return mtd_loaders[current_mtd_loader()].update_format(new, old,
+							       old_len,
+							       mac_idx, op,
+							       loc, info,
+							       master, local);
+}
+
+/*
+ * Perform preliminary checks of found metadata
+ * Return < 0 on errors;
+ * Return number of object-id MACs (>= 1) on success
+ */
+int32_t check_format_v1(uint32_t len, unsigned char *wire)
+{
+	uint32_t nr_links;
+
+	if (len < sizeof(struct mtd_format_v1)) {
+		gf_log("crypt", GF_LOG_ERROR,
+		       "v1-loader: bad metadata size %d", len);
+		goto error;
+	}
+	len -= sizeof(struct mtd_format_v1);
+	if (len % sizeof(nmtd_8_mac_t)) {
+		gf_log("crypt", GF_LOG_ERROR,
+		       "v1-loader: bad metadata format");
+		goto error;
+	}
+	nr_links = 1 + len / sizeof(nmtd_8_mac_t);
+	if (nr_links > _POSIX_LINK_MAX)
+		goto error;
+	return nr_links;
+ error:
+	return EIO;
+}
+
+/*
+ * Verify per-link MAC specified by index @idx
+ *
+ * return:
+ * -1 on errors;
+ *  0 on failed verification;
+ *  1 on sucessful verification
+ */
+static int32_t verify_link_mac_v1(struct mtd_format_v1 *fmt,
+				  uint32_t idx /* index of the mac to verify */,
+				  loc_t *loc,
+				  struct crypt_inode_info *info,
+				  struct master_cipher_info *master)
+{
+	int32_t ret;
+	unsigned char *mac;
+	unsigned char cmac[16];
+
+	mac = get_NMTD_V1_MAC(fmt) + idx * SIZE_OF_NMTD_V1_MAC;
+
+	ret = calc_link_mac_v1(fmt, loc, cmac, info, master);
+	if (ret)
+		return -1;
+	if (memcmp(cmac, mac, SIZE_OF_NMTD_V1_MAC))
+		return 0;
+	return 1;
+}
+
+/*
+ * Lookup per-link MAC by pathname.
+ *
+ * return index of the MAC, if it was found;
+ * return < 0 on errors, or if the MAC wasn't found
+ */
+static int32_t lookup_link_mac_v1(struct mtd_format_v1 *fmt,
+				  uint32_t nr_macs,
+				  loc_t *loc,
+				  struct crypt_inode_info *info,
+				  struct master_cipher_info *master)
+{
+	int32_t ret;
+	uint32_t idx;
+
+	for (idx = 0; idx < nr_macs; idx++) {
+		ret = verify_link_mac_v1(fmt, idx, loc, info, master);
+		if (ret < 0)
+			return ret;
+		if (ret > 0)
+			return idx;
+	}
+	return -ENOENT;
+}
+
+/*
+ * Extract version-specific part of metadata
+ */
+static int32_t open_format_v1(unsigned char *wire,
+			      int32_t len,
+			      loc_t *loc,
+			      struct crypt_inode_info *info,
+			      struct master_cipher_info *master,
+			      crypt_local_t *local,
+			      gf_boolean_t load_info)
+{
+	int32_t ret;
+	int32_t num_nmtd_macs;
+	struct mtd_format_v1 *fmt;
+	unsigned char mtd_key[16];
+	AES_KEY EMTD_KEY;
+	GCM128_CONTEXT *gctx;
+	uint32_t ad;
+	emtd_8_mac_t gmac;
+	struct object_cipher_info *object;
+
+	num_nmtd_macs = check_format_v1(len, wire);
+	if (num_nmtd_macs <= 0)
+		return EIO;
+	fmt = (struct mtd_format_v1 *)wire;
+
+	ret = lookup_link_mac_v1(fmt, num_nmtd_macs, loc, info, master);
+	if (ret < 0) {
+		gf_log("crypt", GF_LOG_ERROR, "NMTD verification failed");
+		return EINVAL;
+	}
+	local->mac_idx = ret;
+	if (load_info == _gf_false)
+		/* the case of partial open */
+		return 0;
+
+	object = &info->cinfo;
+
+	ret = get_emtd_file_key(info, master, mtd_key);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, "Can not retrieve metadata key");
+		return ret;
+	}
+	/*
+	 * decrypt encrypted meta-data
+	 */
+	ret = AES_set_encrypt_key(mtd_key, sizeof(mtd_key)*8, &EMTD_KEY);
+	if (ret < 0) {
+		gf_log("crypt", GF_LOG_ERROR, "Can not set encrypt key");
+		return ret;
+	}
+	gctx = CRYPTO_gcm128_new(&EMTD_KEY, (block128_f)AES_encrypt);
+	if (!gctx) {
+		gf_log("crypt", GF_LOG_ERROR, "Can not alloc gcm context");
+		return ENOMEM;
+	}
+	CRYPTO_gcm128_setiv(gctx, info->oid, sizeof(uuid_t));
+
+	ad = htole32(MTD_LOADER_V1);
+	ret = CRYPTO_gcm128_aad(gctx, (const unsigned char *)&ad, sizeof(ad));
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, " CRYPTO_gcm128_aad failed");
+		CRYPTO_gcm128_release(gctx);
+		return ret;	
+	}
+	ret = CRYPTO_gcm128_decrypt(gctx,
+				    get_EMTD_V1(fmt),
+				    get_EMTD_V1(fmt),
+				    SIZE_OF_EMTD_V1);
+	if (ret) {
+		gf_log("crypt", GF_LOG_ERROR, " CRYPTO_gcm128_decrypt failed");
+		CRYPTO_gcm128_release(gctx);
+		return ret;
+	}
+	/*
+	 * verify metadata
+	 */
+	CRYPTO_gcm128_tag(gctx, gmac, sizeof(gmac));
+	CRYPTO_gcm128_release(gctx);
+	if (memcmp(gmac, get_EMTD_V1_MAC(fmt), SIZE_OF_EMTD_V1_MAC)) {
+		gf_log("crypt", GF_LOG_ERROR, "EMTD verification failed");
+		return EINVAL;
+	}
+	/*
+	 * load verified metadata to the private part of inode
+	 */
+	info->nr_minor = fmt->minor_id;
+
+	object->o_alg = fmt->alg_id;
+	object->o_dkey_size = fmt->dkey_factor << KEY_FACTOR_BITS;
+	object->o_block_bits = fmt->block_bits;
+	object->o_mode = fmt->mode_id;
+
+	return check_file_metadata(info);
+}
+
+/*
+ * perform metadata authentication against @loc->path;
+ * extract crypt-specific attribtes and populate @info
+ * with them (optional)
+ */
+int32_t open_format(unsigned char *str,
+		    int32_t len,
+		    loc_t *loc,
+		    struct crypt_inode_info *info,
+		    struct master_cipher_info *master,
+		    crypt_local_t *local,
+		    gf_boolean_t load_info)
+{
+	struct crypt_format *fmt;
+	if (len < sizeof(*fmt)) {
+		gf_log("crypt", GF_LOG_ERROR, "Bad core format");
+		return EIO;
+	}
+	fmt = (struct crypt_format *)str;
+
+	if (fmt->loader_id >= LAST_MTD_LOADER) {
+		gf_log("crypt", GF_LOG_ERROR,
+		       "Unsupported loader id %d", fmt->loader_id);
+		return EINVAL;
+	}
+	str += sizeof(*fmt);
+	len -= sizeof(*fmt);
+
+	return mtd_loaders[fmt->loader_id].open_format(str,
+						       len,
+						       loc,
+						       info,
+						       master,
+						       local,
+						       load_info);
+}
+
+struct crypt_mtd_loader mtd_loaders [LAST_MTD_LOADER] = {
+	[MTD_LOADER_V1] = 
+	{.format_size = format_size_v1,
+	 .create_format = create_format_v1,
+	 .open_format = open_format_v1,
+	 .update_format = update_format_v1
+	}
+};
+
+/*
+  Local variables:
+  c-indentation-style: "K&R"
+  mode-name: "LC"
+  c-basic-offset: 8
+  tab-width: 8
+  fill-column: 80
+  scroll-step: 1
+  End:
+*/
diff --git a/xlators/encryption/crypt/src/metadata.h b/xlators/encryption/crypt/src/metadata.h
new file mode 100644
index 000000000..a92f149ef
--- /dev/null
+++ b/xlators/encryption/crypt/src/metadata.h
@@ -0,0 +1,74 @@
+/*
+  Copyright (c) 2008-2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef __METADATA_H__
+#define __METADATA_H__
+
+#define NMTD_8_MAC_SIZE   (8)
+#define EMTD_8_MAC_SIZE   (8)
+
+typedef uint8_t nmtd_8_mac_t[NMTD_8_MAC_SIZE];
+typedef uint8_t emtd_8_mac_t[EMTD_8_MAC_SIZE] ;
+
+/*
+ * Version "v1" of file's metadata.
+ * Metadata of this version has 4 components:
+ * 
+ * 1) EMTD (Encrypted part of MeTaData);
+ * 2) NMTD (Non-encrypted part of MeTaData);
+ * 3) EMTD_MAC; (EMTD Message Authentication Code);
+ * 4) Array of per-link NMTD MACs (for every (hard)link it includes
+ *    exactly one MAC)
+ */
+struct mtd_format_v1 {
+	/* EMTD, encrypted part of meta-data */
+	uint8_t            alg_id;  /* cipher algorithm id (only AES for now) */
+        uint8_t            mode_id; /* cipher mode id; (only XTS for now) */
+	uint8_t            block_bits; /* encoded block size */
+	uint8_t            minor_id; /* client translator id */
+	uint8_t            dkey_factor; /* encoded size of the data key */
+	/* MACs */
+	emtd_8_mac_t       gmac; /* MAC of the encrypted meta-data, 8 bytes */
+	nmtd_8_mac_t       omac; /* per-link MACs of the non-encrypted
+				  * meta-data: at least one such MAC is always
+				  * present */
+} __attribute__((packed));
+
+/*
+ * NMTD, the non-encrypted part of metadata of version "v1"
+ * is file's gfid, which is generated on trusted machines.
+ */
+#define SIZE_OF_NMTD_V1 (sizeof(uuid_t))
+#define SIZE_OF_EMTD_V1 (offsetof(struct mtd_format_v1, gmac) -		\
+			 offsetof(struct mtd_format_v1, alg_id))
+#define SIZE_OF_NMTD_V1_MAC (NMTD_8_MAC_SIZE)
+#define SIZE_OF_EMTD_V1_MAC (EMTD_8_MAC_SIZE)
+
+static inline unsigned char *get_EMTD_V1(struct mtd_format_v1 *format)
+{
+        return &format->alg_id;
+}
+
+static inline unsigned char *get_NMTD_V1(struct crypt_inode_info *info)
+{
+        return info->oid;
+}
+
+static inline unsigned char *get_EMTD_V1_MAC(struct mtd_format_v1 *format)
+{
+        return format->gmac;
+}
+
+static inline unsigned char *get_NMTD_V1_MAC(struct mtd_format_v1 *format)
+{
+        return format->omac;
+}
+
+#endif /* __METADATA_H__ */
diff --git a/xlators/encryption/rot-13/src/Makefile.am b/xlators/encryption/rot-13/src/Makefile.am
index ba5e623d8..94e8d18e7 100644
--- a/xlators/encryption/rot-13/src/Makefile.am
+++ b/xlators/encryption/rot-13/src/Makefile.am
@@ -1,14 +1,15 @@
 xlator_LTLIBRARIES = rot-13.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/encryption
 
-rot_13_la_LDFLAGS = -module -avoidversion
+rot_13_la_LDFLAGS = -module -avoid-version
 
 rot_13_la_SOURCES = rot-13.c
 rot_13_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la 
 
 noinst_HEADERS = rot-13.h
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
diff --git a/xlators/encryption/rot-13/src/rot-13.c b/xlators/encryption/rot-13/src/rot-13.c
index d5c702882..1bcfe0192 100644
--- a/xlators/encryption/rot-13/src/rot-13.c
+++ b/xlators/encryption/rot-13/src/rot-13.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #include <ctype.h>
 #include <sys/uio.h>
 
@@ -32,13 +22,13 @@
 #include "rot-13.h"
 
 /*
- * This is a rot13 ``encryption'' xlator. It rot13's data when 
- * writing to disk and rot13's it back when reading it. 
+ * This is a rot13 ``encryption'' xlator. It rot13's data when
+ * writing to disk and rot13's it back when reading it.
  * This xlator is meant as an example, NOT FOR PRODUCTION
  * USE ;) (hence no error-checking)
  */
 
-void 
+void
 rot13 (char *buf, int len)
 {
 	int i;
@@ -67,15 +57,16 @@ rot13_readv_cbk (call_frame_t *frame,
                  int32_t op_errno,
                  struct iovec *vector,
                  int32_t count,
-		 struct stat *stbuf,
-                 struct iobref *iobref)
+		 struct iatt *stbuf,
+                 struct iobref *iobref, dict_t *xdata)
 {
 	rot_13_private_t *priv = (rot_13_private_t *)this->private;
-  
+
 	if (priv->decrypt_read)
 		rot13_iovec (vector, count);
 
-	STACK_UNWIND (frame, op_ret, op_errno, vector, count, stbuf, iobref);
+	STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector, count,
+                             stbuf, iobref, xdata);
 	return 0;
 }
 
@@ -84,13 +75,13 @@ rot13_readv (call_frame_t *frame,
              xlator_t *this,
              fd_t *fd,
              size_t size,
-             off_t offset)
+             off_t offset, uint32_t flags, dict_t *xdata)
 {
 	STACK_WIND (frame,
 		    rot13_readv_cbk,
 		    FIRST_CHILD (this),
 		    FIRST_CHILD (this)->fops->readv,
-		    fd, size, offset);
+		    fd, size, offset, flags, xdata);
 	return 0;
 }
 
@@ -100,10 +91,11 @@ rot13_writev_cbk (call_frame_t *frame,
                   xlator_t *this,
                   int32_t op_ret,
                   int32_t op_errno,
-                  struct stat *prebuf,
-		  struct stat *postbuf)
+                  struct iatt *prebuf,
+		  struct iatt *postbuf, dict_t *xdata)
 {
-	STACK_UNWIND (frame, op_ret, op_errno, prebuf, postbuf);
+	STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf,
+                             xdata);
 	return 0;
 }
 
@@ -112,20 +104,20 @@ rot13_writev (call_frame_t *frame,
               xlator_t *this,
               fd_t *fd,
               struct iovec *vector,
-              int32_t count, 
-              off_t offset,
-              struct iobref *iobref)
+              int32_t count,
+              off_t offset, uint32_t flags,
+              struct iobref *iobref, dict_t *xdata)
 {
 	rot_13_private_t *priv = (rot_13_private_t *)this->private;
 	if (priv->encrypt_write)
 		rot13_iovec (vector, count);
 
-	STACK_WIND (frame, 
+	STACK_WIND (frame,
 		    rot13_writev_cbk,
 		    FIRST_CHILD (this),
 		    FIRST_CHILD (this)->fops->writev,
-		    fd, vector, count, offset,
-                    iobref);
+		    fd, vector, count, offset, flags,
+                    iobref, xdata);
 	return 0;
 }
 
@@ -136,7 +128,7 @@ init (xlator_t *this)
 	rot_13_private_t *priv = NULL;
 
 	if (!this->children || this->children->next) {
-		gf_log ("rot13", GF_LOG_ERROR, 
+		gf_log ("rot13", GF_LOG_ERROR,
 			"FATAL: rot13 should have exactly one child");
 		return -1;
 	}
@@ -145,9 +137,11 @@ init (xlator_t *this)
 		gf_log (this->name, GF_LOG_WARNING,
 			"dangling volume. check volfile ");
 	}
-  
-	priv = CALLOC (sizeof (rot_13_private_t), 1);
-	ERR_ABORT (priv);
+
+	priv = GF_CALLOC (sizeof (rot_13_private_t), 1, 0);
+        if (!priv)
+                return -1;
+
 	priv->decrypt_read = 1;
 	priv->encrypt_write = 1;
 
@@ -156,6 +150,7 @@ init (xlator_t *this)
 		if (gf_string2boolean (data->data, &priv->encrypt_write) == -1) {
 			gf_log (this->name, GF_LOG_ERROR,
 				"encrypt-write takes only boolean options");
+			GF_FREE (priv);
 			return -1;
 		}
 	}
@@ -165,6 +160,7 @@ init (xlator_t *this)
 		if (gf_string2boolean (data->data, &priv->decrypt_read) == -1) {
 			gf_log (this->name, GF_LOG_ERROR,
 				"decrypt-read takes only boolean options");
+			GF_FREE (priv);
 			return -1;
 		}
 	}
@@ -174,13 +170,16 @@ init (xlator_t *this)
 	return 0;
 }
 
-void 
+void
 fini (xlator_t *this)
 {
 	rot_13_private_t *priv = this->private;
-	
-	FREE (priv);
-	
+
+        if (!priv)
+                return;
+        this->private = NULL;
+	GF_FREE (priv);
+
 	return;
 }
 
@@ -189,18 +188,14 @@ struct xlator_fops fops = {
 	.writev       = rot13_writev
 };
 
-struct xlator_mops mops = {
-};
-
-struct xlator_cbks cbks = {
-};
+struct xlator_cbks cbks;
 
 struct volume_options options[] = {
-	{ .key  = {"encrypt-write"}, 
+	{ .key  = {"encrypt-write"},
 	  .type = GF_OPTION_TYPE_BOOL
 	},
-	{ .key  = {"decrypt-read"}, 
-	  .type = GF_OPTION_TYPE_BOOL 
+	{ .key  = {"decrypt-read"},
+	  .type = GF_OPTION_TYPE_BOOL
 	},
 	{ .key  = {NULL} },
 };
diff --git a/xlators/encryption/rot-13/src/rot-13.h b/xlators/encryption/rot-13/src/rot-13.h
index 35a2df2da..3e9fc19c7 100644
--- a/xlators/encryption/rot-13/src/rot-13.h
+++ b/xlators/encryption/rot-13/src/rot-13.h
@@ -1,22 +1,12 @@
 /*
-   Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
    This file is part of GlusterFS.
 
-   GlusterFS is free software; you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published
-   by the Free Software Foundation; either version 3 of the License,
-   or (at your option) any later version.
-
-   GlusterFS is distributed in the hope that it will be useful, but
-   WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-   General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see
-   <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef __ROT_13_H__
 #define __ROT_13_H__
 
diff --git a/xlators/features/Makefile.am b/xlators/features/Makefile.am
index 9ac9b6f19..d2f5ef192 100644
--- a/xlators/features/Makefile.am
+++ b/xlators/features/Makefile.am
@@ -1,3 +1,4 @@
-SUBDIRS = locks trash path-convertor filter quota
+SUBDIRS = locks quota read-only mac-compat quiesce marker index \
+	  protect compress changelog gfid-access $(GLUPY_SUBDIR) qemu-block # trash path-converter # filter
 
-CLEANFILES = 
+CLEANFILES =
diff --git a/xlators/features/changelog/Makefile.am b/xlators/features/changelog/Makefile.am
new file mode 100644
index 000000000..153bb6850
--- /dev/null
+++ b/xlators/features/changelog/Makefile.am
@@ -0,0 +1,3 @@
+SUBDIRS = src lib
+
+CLEANFILES =
diff --git a/xlators/features/changelog/lib/Makefile.am b/xlators/features/changelog/lib/Makefile.am
new file mode 100644
index 000000000..a985f42a8
--- /dev/null
+++ b/xlators/features/changelog/lib/Makefile.am
@@ -0,0 +1,3 @@
+SUBDIRS = src
+
+CLEANFILES =
diff --git a/xlators/features/changelog/lib/examples/c/get-changes.c b/xlators/features/changelog/lib/examples/c/get-changes.c
new file mode 100644
index 000000000..6d0d0357d
--- /dev/null
+++ b/xlators/features/changelog/lib/examples/c/get-changes.c
@@ -0,0 +1,87 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+/**
+ * get set of new changes every 10 seconds (just print the file names)
+ *
+ * Compile it using:
+ *  gcc -o getchanges `pkg-config --cflags libgfchangelog` get-changes.c \
+ *  `pkg-config --libs libgfchangelog`
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <unistd.h>
+#include <sys/un.h>
+#include <limits.h>
+#include <sys/socket.h>
+#include <sys/types.h>
+#include <errno.h>
+
+#include "changelog.h"
+
+#define handle_error(fn)                                \
+        printf ("%s (reason: %s)\n", fn, strerror (errno))
+
+int
+main (int argc, char ** argv)
+{
+        int     i           = 0;
+        int     ret         = 0;
+        ssize_t nr_changes  = 0;
+        ssize_t changes     = 0;
+        char fbuf[PATH_MAX] = {0,};
+
+        /* get changes for brick "/home/vshankar/export/yow/yow-1" */
+        ret = gf_changelog_register ("/home/vshankar/exports/yow/yow-1",
+                                     "/tmp/scratch", "/tmp/change.log", 9, 5);
+        if (ret) {
+                handle_error ("register failed");
+                goto out;
+        }
+
+        while (1) {
+                i = 0;
+                nr_changes = gf_changelog_scan ();
+                if (nr_changes < 0) {
+                        handle_error ("scan(): ");
+                        break;
+                }
+
+                if (nr_changes == 0)
+                        goto next;
+
+                printf ("Got %ld changelog files\n", nr_changes);
+
+                while ( (changes =
+                         gf_changelog_next_change (fbuf, PATH_MAX)) > 0) {
+                        printf ("changelog file [%d]: %s\n", ++i, fbuf);
+
+                        /* process changelog */
+                        /* ... */
+                        /* ... */
+                        /* ... */
+                        /* done processing */
+
+                        ret = gf_changelog_done (fbuf);
+                        if (ret)
+                                handle_error ("gf_changelog_done");
+                }
+
+                if (changes == -1)
+                        handle_error ("gf_changelog_next_change");
+
+        next:
+                sleep (10);
+        }
+
+ out:
+        return ret;
+}
diff --git a/xlators/features/changelog/lib/examples/python/changes.py b/xlators/features/changelog/lib/examples/python/changes.py
new file mode 100644
index 000000000..d21db8eab
--- /dev/null
+++ b/xlators/features/changelog/lib/examples/python/changes.py
@@ -0,0 +1,32 @@
+#!/usr/bin/python
+
+import os
+import sys
+import time
+import libgfchangelog
+
+cl = libgfchangelog.Changes()
+
+def get_changes(brick, scratch_dir, log_file, log_level, interval):
+    change_list = []
+    try:
+        cl.cl_register(brick, scratch_dir, log_file, log_level)
+        while True:
+            cl.cl_scan()
+            change_list = cl.cl_getchanges()
+            if change_list:
+                print change_list
+            for change in change_list:
+                print('done with %s' % (change))
+                cl.cl_done(change)
+            time.sleep(interval)
+    except OSError:
+        ex = sys.exc_info()[1]
+        print ex
+
+if __name__ == '__main__':
+    if len(sys.argv) != 5:
+        print("usage: %s <brick> <scratch-dir> <log-file> <fetch-interval>"
+              % (sys.argv[0]))
+        sys.exit(1)
+    get_changes(sys.argv[1], sys.argv[2], sys.argv[3], 9, int(sys.argv[4]))
diff --git a/xlators/features/changelog/lib/examples/python/libgfchangelog.py b/xlators/features/changelog/lib/examples/python/libgfchangelog.py
new file mode 100644
index 000000000..68ec3baf1
--- /dev/null
+++ b/xlators/features/changelog/lib/examples/python/libgfchangelog.py
@@ -0,0 +1,64 @@
+import os
+from ctypes import *
+from ctypes.util import find_library
+
+class Changes(object):
+    libgfc = CDLL(find_library("gfchangelog"), use_errno=True)
+
+    @classmethod
+    def geterrno(cls):
+        return get_errno()
+
+    @classmethod
+    def raise_oserr(cls):
+        errn = cls.geterrno()
+        raise OSError(errn, os.strerror(errn))
+
+    @classmethod
+    def _get_api(cls, call):
+        return getattr(cls.libgfc, call)
+
+    @classmethod
+    def cl_register(cls, brick, path, log_file, log_level, retries = 0):
+        ret = cls._get_api('gf_changelog_register')(brick, path,
+                                                    log_file, log_level, retries)
+        if ret == -1:
+            cls.raise_oserr()
+
+    @classmethod
+    def cl_scan(cls):
+        ret = cls._get_api('gf_changelog_scan')()
+        if ret == -1:
+            cls.raise_oserr()
+
+    @classmethod
+    def cl_startfresh(cls):
+        ret = cls._get_api('gf_changelog_start_fresh')()
+        if ret == -1:
+            cls.raise_oserr()
+
+    @classmethod
+    def cl_getchanges(cls):
+        """ remove hardcoding for path name length """
+        def clsort(f):
+            return f.split('.')[-1]
+        changes = []
+        buf = create_string_buffer('\0', 4096)
+        call = cls._get_api('gf_changelog_next_change')
+
+        while True:
+            ret = call(buf, 4096)
+            if ret in (0, -1):
+                break;
+            changes.append(buf.raw[:ret-1])
+        if ret == -1:
+            cls.raise_oserr()
+        # cleanup tracker
+        cls.cl_startfresh()
+        return sorted(changes, key=clsort)
+
+    @classmethod
+    def cl_done(cls, clfile):
+        ret = cls._get_api('gf_changelog_done')(clfile)
+        if ret == -1:
+            cls.raise_oserr()
diff --git a/xlators/features/changelog/lib/src/Makefile.am b/xlators/features/changelog/lib/src/Makefile.am
new file mode 100644
index 000000000..775f026cf
--- /dev/null
+++ b/xlators/features/changelog/lib/src/Makefile.am
@@ -0,0 +1,37 @@
+libgfchangelog_la_CFLAGS = -Wall $(GF_CFLAGS) $(GF_DARWIN_LIBGLUSTERFS_CFLAGS) \
+			-DDATADIR=\"$(localstatedir)\"
+
+libgfchangelog_la_CPPFLAGS = $(GF_CPPFLAGS) -D__USE_FILE_OFFSET64 -fpic \
+				-I../../../src/ -I$(top_srcdir)/libglusterfs/src \
+				-I$(top_srcdir)/xlators/features/changelog/src \
+				-DDATADIR=\"$(localstatedir)\"
+
+libgfchangelog_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la \
+			$(GF_GLUSTERFS_LIBS)
+
+libgfchangelog_la_LDFLAGS = $(GF_LDFLAGS) -version-info $(LIBGFCHANGELOG_LT_VERSION)
+
+libgfchangelogdir = $(includedir)/glusterfs/gfchangelog
+lib_LTLIBRARIES = libgfchangelog.la
+
+CONTRIB_BUILDDIR = $(top_builddir)/contrib
+
+libgfchangelog_la_SOURCES = gf-changelog.c gf-changelog-process.c \
+		gf-changelog-helpers.c $(CONTRIBDIR)/uuid/clear.c \
+		$(CONTRIBDIR)/uuid/copy.c $(CONTRIBDIR)/uuid/gen_uuid.c \
+		$(CONTRIBDIR)/uuid/pack.c $(CONTRIBDIR)/uuid/parse.c \
+		$(CONTRIBDIR)/uuid/unparse.c $(CONTRIBDIR)/uuid/uuid_time.c \
+		$(CONTRIBDIR)/uuid/compare.c $(CONTRIBDIR)/uuid/isnull.c \
+		$(CONTRIBDIR)/uuid/unpack.c
+
+noinst_HEADERS = gf-changelog-helpers.h $(CONTRIBDIR)/uuid/uuidd.h \
+		$(CONTRIBDIR)/uuid/uuid.h $(CONTRIBDIR)/uuid/uuidP.h \
+		$(CONTRIB_BUILDDIR)/uuid/uuid_types.h
+
+libgfchangelog_HEADERS = changelog.h
+
+CLEANFILES =
+CONFIG_CLEAN_FILES = $(CONTRIB_BUILDDIR)/uuid/uuid_types.h
+
+$(top_builddir)/libglusterfs/src/libglusterfs.la:
+	$(MAKE) -C $(top_builddir)/libglusterfs/src/ all
diff --git a/xlators/features/changelog/lib/src/changelog.h b/xlators/features/changelog/lib/src/changelog.h
new file mode 100644
index 000000000..5cddfb583
--- /dev/null
+++ b/xlators/features/changelog/lib/src/changelog.h
@@ -0,0 +1,31 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _GF_CHANGELOG_H
+#define _GF_CHANGELOG_H
+
+/* API set */
+
+int
+gf_changelog_register (char *brick_path, char *scratch_dir,
+                       char *log_file, int log_levl, int max_reconnects);
+ssize_t
+gf_changelog_scan ();
+
+int
+gf_changelog_start_fresh ();
+
+ssize_t
+gf_changelog_next_change (char *bufptr, size_t maxlen);
+
+int
+gf_changelog_done (char *file);
+
+#endif
diff --git a/xlators/features/changelog/lib/src/gf-changelog-helpers.c b/xlators/features/changelog/lib/src/gf-changelog-helpers.c
new file mode 100644
index 000000000..1eef8bf04
--- /dev/null
+++ b/xlators/features/changelog/lib/src/gf-changelog-helpers.c
@@ -0,0 +1,180 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#include "changelog-mem-types.h"
+#include "gf-changelog-helpers.h"
+
+ssize_t gf_changelog_read_path (int fd, char *buffer, size_t bufsize)
+{
+        return read (fd, buffer, bufsize);
+}
+
+size_t
+gf_changelog_write (int fd, char *buffer, size_t len)
+{
+        ssize_t size = 0;
+        size_t writen = 0;
+
+        while (writen < len) {
+                size = write (fd,
+                              buffer + writen, len - writen);
+                if (size <= 0)
+                        break;
+
+                writen += size;
+        }
+
+        return writen;
+}
+
+void
+gf_rfc3986_encode (unsigned char *s, char *enc, char *estr)
+{
+        for (; *s; s++) {
+                if (estr[*s])
+                        sprintf(enc, "%c", estr[*s]);
+                else
+                        sprintf(enc, "%%%02X", *s);
+                while (*++enc);
+        }
+}
+
+/**
+ * thread safe version of readline with buffering
+ * (taken from Unix Network Programming Volume I, W.R. Stevens)
+ *
+ * This is favoured over fgets() as we'd need to ftruncate()
+ * (see gf_changelog_scan() API) to record new changelog files.
+ * stream open functions does have a truncate like api (although
+ * that can be done via @fflush(fp), @ftruncate(fd) and @fseek(fp),
+ * but this involves mixing POSIX file descriptors and stream FILE *).
+ *
+ * NOTE: This implmentation still does work with more than one fd's
+ *       used to perform gf_readline(). For this very reason it's not
+ *       made a part of libglusterfs.
+ */
+
+static pthread_key_t rl_key;
+static pthread_once_t rl_once = PTHREAD_ONCE_INIT;
+
+static void
+readline_destructor (void *ptr)
+{
+        GF_FREE (ptr);
+}
+
+static void
+readline_once (void)
+{
+        pthread_key_create (&rl_key, readline_destructor);
+}
+
+static ssize_t
+my_read (read_line_t *tsd, int fd, char *ptr)
+{
+        if (tsd->rl_cnt <= 0) {
+                if ( (tsd->rl_cnt = read (fd, tsd->rl_buf, MAXLINE)) < 0 )
+                        return -1;
+                else if (tsd->rl_cnt == 0)
+                        return 0;
+                tsd->rl_bufptr = tsd->rl_buf;
+        }
+
+        tsd->rl_cnt--;
+        *ptr = *tsd->rl_bufptr++;
+        return 1;
+}
+
+static int
+gf_readline_init_once (read_line_t **tsd)
+{
+        if (pthread_once (&rl_once, readline_once) != 0)
+                return -1;
+
+        *tsd = pthread_getspecific (rl_key);
+        if (*tsd)
+                goto out;
+
+        *tsd = GF_CALLOC (1, sizeof (**tsd),
+                          gf_changelog_mt_libgfchangelog_rl_t);
+        if (!*tsd)
+                return -1;
+
+        if (pthread_setspecific (rl_key, *tsd) != 0)
+                return -1;
+
+ out:
+        return 0;
+}
+
+ssize_t
+gf_readline (int fd, void *vptr, size_t maxlen)
+{
+        size_t       n   = 0;
+        size_t       rc  = 0;
+        char         c   = ' ';
+        char        *ptr = NULL;
+        read_line_t *tsd = NULL;
+
+        if (gf_readline_init_once (&tsd))
+                return -1;
+
+        ptr = vptr;
+        for (n = 1; n < maxlen; n++) {
+                if ( (rc = my_read (tsd, fd, &c)) == 1 ) {
+                        *ptr++ = c;
+                        if (c == '\n')
+                                break;
+                } else if (rc == 0) {
+                        *ptr = '\0';
+                        return (n - 1);
+                } else
+                        return -1;
+        }
+
+        *ptr = '\0';
+        return n;
+
+}
+
+off_t
+gf_lseek (int fd, off_t offset, int whence)
+{
+        off_t        off = 0;
+        read_line_t *tsd = NULL;
+
+        if (gf_readline_init_once (&tsd))
+                return -1;
+
+        if ( (off = lseek (fd, offset, whence)) == -1)
+                return -1;
+
+        tsd->rl_cnt = 0;
+        tsd->rl_bufptr = tsd->rl_buf;
+
+        return off;
+}
+
+int
+gf_ftruncate (int fd, off_t length)
+{
+        read_line_t *tsd = NULL;
+
+        if (gf_readline_init_once (&tsd))
+                return -1;
+
+        if (ftruncate (fd, 0))
+                return -1;
+
+        tsd->rl_cnt = 0;
+        tsd->rl_bufptr = tsd->rl_buf;
+
+        return 0;
+}
diff --git a/xlators/features/changelog/lib/src/gf-changelog-helpers.h b/xlators/features/changelog/lib/src/gf-changelog-helpers.h
new file mode 100644
index 000000000..7e13d9376
--- /dev/null
+++ b/xlators/features/changelog/lib/src/gf-changelog-helpers.h
@@ -0,0 +1,97 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _GF_CHANGELOG_HELPERS_H
+#define _GF_CHANGELOG_HELPERS_H
+
+#include <unistd.h>
+#include <dirent.h>
+#include <limits.h>
+#include <pthread.h>
+
+#include <xlator.h>
+
+#define GF_CHANGELOG_TRACKER  "tracker"
+
+#define GF_CHANGELOG_CURRENT_DIR    ".current"
+#define GF_CHANGELOG_PROCESSED_DIR  ".processed"
+#define GF_CHANGELOG_PROCESSING_DIR ".processing"
+
+#ifndef MAXLINE
+#define MAXLINE 4096
+#endif
+
+#define GF_CHANGELOG_FILL_BUFFER(ptr, ascii, off, len) do {     \
+                memcpy (ascii + off, ptr, len);                 \
+                off += len;                                     \
+        } while (0)
+
+typedef struct read_line {
+        int rl_cnt;
+        char *rl_bufptr;
+        char rl_buf[MAXLINE];
+} read_line_t;
+
+typedef struct gf_changelog {
+        xlator_t *this;
+
+        /* 'processing' directory stream */
+        DIR *gfc_dir;
+
+        /* fd to the tracker file */
+        int gfc_fd;
+
+        /* connection retries */
+        int gfc_connretries;
+
+        char gfc_sockpath[UNIX_PATH_MAX];
+
+        char gfc_brickpath[PATH_MAX];
+
+        /* socket for recieving notifications */
+        int gfc_sockfd;
+
+        char *gfc_working_dir;
+
+        /* RFC 3986 string encoding */
+        char rfc3986[256];
+
+        char gfc_current_dir[PATH_MAX];
+        char gfc_processed_dir[PATH_MAX];
+        char gfc_processing_dir[PATH_MAX];
+
+        pthread_t gfc_changelog_processor;
+} gf_changelog_t;
+
+int
+gf_changelog_notification_init (xlator_t *this, gf_changelog_t *gfc);
+
+void *
+gf_changelog_process (void *data);
+
+ssize_t
+gf_changelog_read_path (int fd, char *buffer, size_t bufsize);
+
+void
+gf_rfc3986_encode (unsigned char *s, char *enc, char *estr);
+
+size_t
+gf_changelog_write (int fd, char *buffer, size_t len);
+
+ssize_t
+gf_readline (int fd, void *vptr, size_t maxlen);
+
+int
+gf_ftruncate (int fd, off_t length);
+
+off_t
+gf_lseek (int fd, off_t offset, int whence);
+
+#endif
diff --git a/xlators/features/changelog/lib/src/gf-changelog-process.c b/xlators/features/changelog/lib/src/gf-changelog-process.c
new file mode 100644
index 000000000..3ea2700c6
--- /dev/null
+++ b/xlators/features/changelog/lib/src/gf-changelog-process.c
@@ -0,0 +1,618 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#include <unistd.h>
+#include <pthread.h>
+
+#include "uuid.h"
+#include "globals.h"
+#include "glusterfs.h"
+
+#include "gf-changelog-helpers.h"
+
+/* from the changelog translator */
+#include "changelog-misc.h"
+
+extern int byebye;
+
+/**
+ * number of gfid records after fop number
+ */
+int nr_gfids[] = {
+        [GF_FOP_MKNOD]   = 1,
+        [GF_FOP_MKDIR]   = 1,
+        [GF_FOP_UNLINK]  = 1,
+        [GF_FOP_RMDIR]   = 1,
+        [GF_FOP_SYMLINK] = 1,
+        [GF_FOP_RENAME]  = 2,
+        [GF_FOP_LINK]    = 1,
+        [GF_FOP_CREATE]  = 1,
+};
+
+int nr_extra_recs[] = {
+        [GF_FOP_MKNOD]   = 3,
+        [GF_FOP_MKDIR]   = 3,
+        [GF_FOP_UNLINK]  = 0,
+        [GF_FOP_RMDIR]   = 0,
+        [GF_FOP_SYMLINK] = 0,
+        [GF_FOP_RENAME]  = 0,
+        [GF_FOP_LINK]    = 0,
+        [GF_FOP_CREATE]  = 3,
+};
+
+static char *
+binary_to_ascii (uuid_t uuid)
+{
+        return uuid_utoa (uuid);
+}
+
+static char *
+conv_noop (char *ptr) { return ptr; }
+
+#define VERIFY_SEPARATOR(ptr, plen, perr)       \
+        {                                       \
+                if (*(ptr + plen) != '\0') {    \
+                        perr = 1;               \
+                        break;                  \
+                }                               \
+        }
+
+#define MOVER_MOVE(mover, nleft, bytes)         \
+        {                                       \
+                mover += bytes;                 \
+                nleft -= bytes;                 \
+        }                                       \
+
+#define PARSE_GFID(mov, ptr, le, fn, perr)                      \
+        {                                                       \
+                VERIFY_SEPARATOR (mov, le, perr);               \
+                ptr = fn (mov);                                 \
+                if (!ptr) {                                     \
+                        perr = 1;                               \
+                        break;                                  \
+                }                                               \
+        }
+
+#define FILL_AND_MOVE(pt, buf, of, mo, nl, le)                          \
+        {                                                               \
+                GF_CHANGELOG_FILL_BUFFER (pt, buf, of, strlen (pt));    \
+                MOVER_MOVE (mo, nl, le);                                \
+        }
+
+
+#define PARSE_GFID_MOVE(ptr, uuid, mover, nleft, perr)          \
+        {                                                       \
+                memcpy (uuid, mover, sizeof (uuid_t));          \
+                ptr = binary_to_ascii (uuid);                   \
+                if (!ptr) {                                     \
+                        perr = 1;                               \
+                        break;                                  \
+                }                                               \
+                MOVER_MOVE (mover, nleft, sizeof (uuid_t));     \
+        }                                                       \
+
+#define LINE_BUFSIZE  3*PATH_MAX /* enough buffer for extra chars too */
+
+/**
+ * using mmap() makes parsing easy. fgets() cannot be used here as
+ * the binary gfid could contain a line-feed (0x0A), in that case fgets()
+ * would read an incomplete line and parsing would fail. using POSIX fds
+ * would result is additional code to maintain state in case of partial
+ * reads of data (where multiple entries do not fit extirely in the buffer).
+ *
+ * mmap() gives the flexibility of pointing to an offset in the file
+ * without us worrying about reading it in memory (VM does that for us for
+ * free).
+ */
+
+static int
+gf_changelog_parse_binary (xlator_t *this,
+                           gf_changelog_t *gfc, int from_fd, int to_fd,
+                           size_t start_offset, struct stat *stbuf)
+
+{
+        int     ret              = -1;
+        off_t   off              = 0;
+        off_t   nleft            = 0;
+        uuid_t  uuid             = {0,};
+        char   *ptr              = NULL;
+        char   *bname_start      = NULL;
+        char   *bname_end        = NULL;
+        char   *mover            = NULL;
+        char   *start            = NULL;
+        char    current_mover    = ' ';
+        size_t  blen             = 0;
+        int     parse_err        = 0;
+        char ascii[LINE_BUFSIZE] = {0,};
+
+        nleft = stbuf->st_size;
+
+        start = (char *) mmap (NULL, nleft,
+                               PROT_READ, MAP_PRIVATE, from_fd, 0);
+        if (!start) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "mmap() error (reason: %s)", strerror (errno));
+                goto out;
+        }
+
+        mover = start;
+
+        MOVER_MOVE (mover, nleft, start_offset);
+
+        while (nleft > 0) {
+
+                off = blen = 0;
+                ptr = bname_start = bname_end = NULL;
+
+                current_mover = *mover;
+
+                switch (current_mover) {
+                case 'D':
+                case 'M':
+                        MOVER_MOVE (mover, nleft, 1);
+                        PARSE_GFID_MOVE (ptr, uuid, mover, nleft, parse_err);
+
+                        break;
+
+                case 'E':
+                        MOVER_MOVE (mover, nleft, 1);
+                        PARSE_GFID_MOVE (ptr, uuid, mover, nleft, parse_err);
+
+                        bname_start = mover;
+                        if ( (bname_end = strchr (mover, '\n')) == NULL ) {
+                                parse_err = 1;
+                                break;
+                        }
+
+                        blen = bname_end - bname_start;
+                        MOVER_MOVE (mover, nleft, blen);
+
+                        break;
+
+                default:
+                        parse_err = 1;
+                }
+
+                if (parse_err)
+                        break;
+
+                GF_CHANGELOG_FILL_BUFFER (&current_mover, ascii, off, 1);
+                GF_CHANGELOG_FILL_BUFFER (" ", ascii, off, 1);
+                GF_CHANGELOG_FILL_BUFFER (ptr, ascii, off, strlen (ptr));
+                if (blen)
+                        GF_CHANGELOG_FILL_BUFFER (bname_start,
+                                                  ascii, off, blen);
+                GF_CHANGELOG_FILL_BUFFER ("\n", ascii, off, 1);
+
+                if (gf_changelog_write (to_fd, ascii, off) != off) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "processing binary changelog failed due to "
+                                " error in writing ascii change (reason: %s)",
+                                strerror (errno));
+                        break;
+                }
+
+                MOVER_MOVE (mover, nleft, 1);
+        }
+
+        if ( (nleft == 0) && (!parse_err))
+                ret = 0;
+
+        if (munmap (start, stbuf->st_size))
+                gf_log (this->name, GF_LOG_ERROR,
+                        "munmap() error (reason: %s)", strerror (errno));
+ out:
+        return ret;
+}
+
+/**
+ * ascii decoder:
+ *  - separate out one entry from another
+ *  - use fop name rather than fop number
+ */
+static int
+gf_changelog_parse_ascii (xlator_t *this,
+                          gf_changelog_t *gfc, int from_fd, int to_fd,
+                          size_t start_offset, struct stat *stbuf)
+{
+        int           ng            = 0;
+        int           ret           = -1;
+        int           fop           = 0;
+        int           len           = 0;
+        off_t         off           = 0;
+        off_t         nleft         = 0;
+        char         *ptr           = NULL;
+        char         *eptr          = NULL;
+        char         *start         = NULL;
+        char         *mover         = NULL;
+        int           parse_err     = 0;
+        char          current_mover = ' ';
+        char ascii[LINE_BUFSIZE]    = {0,};
+        const char   *fopname       = NULL;
+
+        nleft = stbuf->st_size;
+
+        start = (char *) mmap (NULL, nleft,
+                               PROT_READ, MAP_PRIVATE, from_fd, 0);
+        if (!start) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "mmap() error (reason: %s)", strerror (errno));
+                goto out;
+        }
+
+        mover = start;
+
+        MOVER_MOVE (mover, nleft, start_offset);
+
+        while (nleft > 0) {
+                off = 0;
+                current_mover = *mover;
+
+                GF_CHANGELOG_FILL_BUFFER (&current_mover, ascii, off, 1);
+                GF_CHANGELOG_FILL_BUFFER (" ", ascii, off, 1);
+
+                switch (current_mover) {
+                case 'D':
+                        MOVER_MOVE (mover, nleft, 1);
+
+                        /* target gfid */
+                        PARSE_GFID (mover, ptr, UUID_CANONICAL_FORM_LEN,
+                                    conv_noop, parse_err);
+                        FILL_AND_MOVE(ptr, ascii, off,
+                                      mover, nleft, UUID_CANONICAL_FORM_LEN);
+                        break;
+                case 'M':
+                        MOVER_MOVE (mover, nleft, 1);
+
+                        /* target gfid */
+                        PARSE_GFID (mover, ptr, UUID_CANONICAL_FORM_LEN,
+                                    conv_noop, parse_err);
+                        FILL_AND_MOVE (ptr, ascii, off,
+                                       mover, nleft, UUID_CANONICAL_FORM_LEN);
+                        FILL_AND_MOVE (" ", ascii, off, mover, nleft, 1);
+
+                        /* fop */
+                        len = strlen (mover);
+                        VERIFY_SEPARATOR (mover, len, parse_err);
+
+                        fop = atoi (mover);
+                        if ( (fopname = gf_fop_list[fop]) == NULL) {
+                                parse_err = 1;
+                                break;
+                        }
+
+                        MOVER_MOVE (mover, nleft, len);
+
+                        len = strlen (fopname);
+                        GF_CHANGELOG_FILL_BUFFER (fopname, ascii, off, len);
+
+                        break;
+
+                case 'E':
+                        MOVER_MOVE (mover, nleft, 1);
+
+                        /* target gfid */
+                        PARSE_GFID (mover, ptr, UUID_CANONICAL_FORM_LEN,
+                                    conv_noop, parse_err);
+                        FILL_AND_MOVE (ptr, ascii, off,
+                                       mover, nleft, UUID_CANONICAL_FORM_LEN);
+                        FILL_AND_MOVE (" ", ascii, off,
+                                       mover, nleft, 1);
+
+                        /* fop */
+                        len = strlen (mover);
+                        VERIFY_SEPARATOR (mover, len, parse_err);
+
+                        fop = atoi (mover);
+                        if ( (fopname = gf_fop_list[fop]) == NULL) {
+                                parse_err = 1;
+                                break;
+                        }
+
+                        MOVER_MOVE (mover, nleft, len);
+
+                        len = strlen (fopname);
+                        GF_CHANGELOG_FILL_BUFFER (fopname, ascii, off, len);
+
+                        ng = nr_extra_recs[fop];
+                        for (;ng > 0; ng--) {
+                                MOVER_MOVE (mover, nleft, 1);
+                                len = strlen (mover);
+                                VERIFY_SEPARATOR (mover, len, parse_err);
+
+                                GF_CHANGELOG_FILL_BUFFER (" ", ascii, off, 1);
+                                FILL_AND_MOVE (mover, ascii,
+                                               off, mover, nleft, len);
+                        }
+
+                        /* pargfid + bname */
+                        ng = nr_gfids[fop];
+                        while (ng-- > 0) {
+                                MOVER_MOVE (mover, nleft, 1);
+                                len = strlen (mover);
+                                GF_CHANGELOG_FILL_BUFFER (" ", ascii, off, 1);
+
+                                PARSE_GFID (mover, ptr, len,
+                                            conv_noop, parse_err);
+                                eptr = calloc (3, strlen (ptr));
+                                if (!eptr) {
+                                        parse_err = 1;
+                                        break;
+                                }
+
+                                gf_rfc3986_encode ((unsigned char *) ptr,
+                                                   eptr, gfc->rfc3986);
+                                FILL_AND_MOVE (eptr, ascii, off,
+                                               mover, nleft, len);
+                                free (eptr);
+                        }
+
+                        break;
+                default:
+                        parse_err = 1;
+                }
+
+                if (parse_err)
+                        break;
+
+                GF_CHANGELOG_FILL_BUFFER ("\n", ascii, off, 1);
+
+                if (gf_changelog_write (to_fd, ascii, off) != off) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "processing ascii changelog failed due to "
+                                " error in writing change (reason: %s)",
+                                strerror (errno));
+                        break;
+                }
+
+                MOVER_MOVE (mover, nleft, 1);
+
+        }
+
+        if ( (nleft == 0) && (!parse_err))
+                ret = 0;
+
+        if (munmap (start, stbuf->st_size))
+                gf_log (this->name, GF_LOG_ERROR,
+                        "munmap() error (reason: %s)", strerror (errno));
+
+ out:
+        return ret;
+}
+
+#define COPY_BUFSIZE  8192
+static int
+gf_changelog_copy (xlator_t *this, int from_fd, int to_fd)
+{
+        ssize_t size                  = 0;
+        char   buffer[COPY_BUFSIZE+1] = {0,};
+
+        while (1) {
+                size = read (from_fd, buffer, COPY_BUFSIZE);
+                if (size <= 0)
+                        break;
+
+                if (gf_changelog_write (to_fd,
+                                        buffer, size) != size) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "error processing ascii changlog");
+                        size = -1;
+                        break;
+                }
+        }
+
+        return (size < 0 ? -1 : 0);
+}
+
+static int
+gf_changelog_decode (xlator_t *this, gf_changelog_t *gfc, int from_fd,
+                     int to_fd, struct stat *stbuf, int *zerob)
+{
+        int    ret        = -1;
+        int    encoding   = -1;
+        size_t elen       = 0;
+        char buffer[1024] = {0,};
+
+        CHANGELOG_GET_ENCODING (from_fd, buffer, 1024, encoding, elen);
+        if (encoding == -1) /* unknown encoding */
+                goto out;
+
+        if (!CHANGELOG_VALID_ENCODING (encoding))
+                goto out;
+
+        if (elen == stbuf->st_size) {
+                *zerob = 1;
+                goto out;
+        }
+
+        /**
+         * start processing after the header
+         */
+        lseek (from_fd, elen, SEEK_SET);
+
+        switch (encoding) {
+        case CHANGELOG_ENCODE_BINARY:
+                /**
+                 * this ideally should have been a part of changelog-encoders.c
+                 * (ie. part of the changelog translator).
+                 */
+                ret = gf_changelog_parse_binary (this, gfc, from_fd,
+                                                 to_fd, elen, stbuf);
+                break;
+
+        case CHANGELOG_ENCODE_ASCII:
+                ret = gf_changelog_parse_ascii (this, gfc, from_fd,
+                                                to_fd, elen, stbuf);
+                break;
+        default:
+                ret = gf_changelog_copy (this, from_fd, to_fd);
+        }
+
+ out:
+        return ret;
+}
+
+static int
+gf_changelog_consume (xlator_t *this, gf_changelog_t *gfc, char *from_path)
+{
+        int         ret        = -1;
+        int         fd1        = 0;
+        int         fd2        = 0;
+        int         zerob      = 0;
+        struct stat stbuf      = {0,};
+        char dest[PATH_MAX]    = {0,};
+        char to_path[PATH_MAX] = {0,};
+
+        ret = stat (from_path, &stbuf);
+        if (ret || !S_ISREG(stbuf.st_mode)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "stat failed on changelog file: %s", from_path);
+                goto out;
+        }
+
+        fd1 = open (from_path, O_RDONLY);
+        if (fd1 < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "cannot open changelog file: %s (reason: %s)",
+                        from_path, strerror (errno));
+                goto out;
+        }
+
+        (void) snprintf (to_path, PATH_MAX, "%s%s",
+                         gfc->gfc_current_dir, basename (from_path));
+        (void) snprintf (dest, PATH_MAX, "%s%s",
+                         gfc->gfc_processing_dir, basename (from_path));
+
+        fd2 = open (to_path, O_CREAT | O_TRUNC | O_RDWR,
+                    S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
+        if (fd2 < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "cannot create ascii changelog file %s (reason %s)",
+                        to_path, strerror (errno));
+                goto close_fd;
+        } else {
+                ret = gf_changelog_decode (this, gfc, fd1,
+                                           fd2, &stbuf, &zerob);
+
+                close (fd2);
+
+                if (!ret) {
+                        /* move it to processing on a successfull
+                           decode */
+                        ret = rename (to_path, dest);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "error moving %s to processing dir"
+                                        " (reason: %s)", to_path,
+                                        strerror (errno));
+                }
+
+                /* remove it from .current if it's an empty file */
+                if (zerob) {
+                        ret = unlink (to_path);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "could not unlink %s (reason: %s",
+                                        to_path, strerror (errno));
+                }
+        }
+
+ close_fd:
+        close (fd1);
+
+ out:
+        return ret;
+}
+
+static char *
+gf_changelog_ext_change (xlator_t *this,
+                         gf_changelog_t *gfc, char *path, size_t readlen)
+{
+        int     alo = 0;
+        int     ret = 0;
+        size_t  len = 0;
+        char   *buf = NULL;
+
+        buf = path;
+        while (len < readlen) {
+                if (*buf == '\0') {
+                        alo = 1;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "processing changelog: %s", path);
+                        ret = gf_changelog_consume (this, gfc, path);
+                }
+
+                if (ret)
+                        break;
+
+                len++; buf++;
+                if (alo) {
+                        alo = 0;
+                        path = buf;
+                }
+        }
+
+        return (ret) ? NULL : path;
+}
+
+void *
+gf_changelog_process (void *data)
+{
+        ssize_t         len      = 0;
+        ssize_t         offlen   = 0;
+        xlator_t       *this     = NULL;
+        char           *sbuf     = NULL;
+        gf_changelog_t *gfc      = NULL;
+        char from_path[PATH_MAX] = {0,};
+
+        gfc = (gf_changelog_t *) data;
+        this = gfc->this;
+
+        pthread_detach (pthread_self());
+
+        for (;;) {
+                len = gf_changelog_read_path (gfc->gfc_sockfd,
+                                              from_path + offlen,
+                                              PATH_MAX - offlen);
+                if (len < 0)
+                        continue; /* ignore it for now */
+
+                if (len == 0) { /* close() from the changelog translator */
+                        gf_log (this->name, GF_LOG_INFO, "close from changelog"
+                                " notification translator.");
+
+                        if (gfc->gfc_connretries != 1) {
+                                if (!gf_changelog_notification_init(this, gfc))
+                                        continue;
+                        }
+
+                        byebye = 1;
+                        break;
+                }
+
+                len += offlen;
+                sbuf = gf_changelog_ext_change (this, gfc, from_path, len);
+                if (!sbuf) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "could not extract changelog filename");
+                        continue;
+                }
+
+                offlen = 0;
+                if (sbuf != (from_path + len)) {
+                        offlen = from_path + len - sbuf;
+                        memmove (from_path, sbuf, offlen);
+                }
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "byebye (%d) from processing thread...", byebye);
+        return NULL;
+}
diff --git a/xlators/features/changelog/lib/src/gf-changelog.c b/xlators/features/changelog/lib/src/gf-changelog.c
new file mode 100644
index 000000000..4b2b25ad5
--- /dev/null
+++ b/xlators/features/changelog/lib/src/gf-changelog.c
@@ -0,0 +1,516 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#include <errno.h>
+#include <dirent.h>
+#include <stddef.h>
+#include <sys/types.h>
+#include <sys/socket.h>
+#include <sys/un.h>
+
+#ifndef _GNU_SOURCE
+#define _GNU_SOURCE
+#endif
+#include <string.h>
+
+#include "globals.h"
+#include "glusterfs.h"
+#include "logging.h"
+
+#include "gf-changelog-helpers.h"
+
+/* from the changelog translator */
+#include "changelog-misc.h"
+#include "changelog-mem-types.h"
+
+int byebye = 0;
+
+static void
+gf_changelog_cleanup (gf_changelog_t *gfc)
+{
+        /* socket */
+        if (gfc->gfc_sockfd != -1)
+                close (gfc->gfc_sockfd);
+        /* tracker fd */
+        if (gfc->gfc_fd != -1)
+                close (gfc->gfc_fd);
+        /* processing dir */
+        if (gfc->gfc_dir)
+                closedir (gfc->gfc_dir);
+
+        if (gfc->gfc_working_dir)
+                free (gfc->gfc_working_dir); /* allocated by realpath */
+}
+
+void
+__attribute__ ((constructor)) gf_changelog_ctor (void)
+{
+        glusterfs_ctx_t *ctx = NULL;
+
+        ctx = glusterfs_ctx_new ();
+        if (!ctx)
+                return;
+
+        if (glusterfs_globals_init (ctx)) {
+                free (ctx);
+                ctx = NULL;
+                return;
+        }
+
+        THIS->ctx = ctx;
+}
+
+void
+__attribute__ ((destructor)) gf_changelog_dtor (void)
+{
+        xlator_t        *this = NULL;
+        glusterfs_ctx_t *ctx  = NULL;
+        gf_changelog_t  *gfc  = NULL;
+
+        this = THIS;
+        if (!this)
+                return;
+
+        ctx = this->ctx;
+        gfc = this->private;
+
+        if (gfc) {
+                gf_changelog_cleanup (gfc);
+                GF_FREE (gfc);
+        }
+
+        if (ctx) {
+                pthread_mutex_destroy (&ctx->lock);
+                free (ctx);
+                ctx = NULL;
+        }
+}
+
+
+static int
+gf_changelog_open_dirs (gf_changelog_t *gfc)
+{
+        int  ret                    = -1;
+        DIR *dir                    = NULL;
+        int  tracker_fd             = 0;
+        char tracker_path[PATH_MAX] = {0,};
+
+        (void) snprintf (gfc->gfc_current_dir, PATH_MAX,
+                         "%s/"GF_CHANGELOG_CURRENT_DIR"/",
+                         gfc->gfc_working_dir);
+        ret = mkdir_p (gfc->gfc_current_dir, 0600, _gf_false);
+        if (ret)
+                goto out;
+
+        (void) snprintf (gfc->gfc_processed_dir, PATH_MAX,
+                         "%s/"GF_CHANGELOG_PROCESSED_DIR"/",
+                         gfc->gfc_working_dir);
+        ret = mkdir_p (gfc->gfc_processed_dir, 0600, _gf_false);
+        if (ret)
+                goto out;
+
+        (void) snprintf (gfc->gfc_processing_dir, PATH_MAX,
+                         "%s/"GF_CHANGELOG_PROCESSING_DIR"/",
+                         gfc->gfc_working_dir);
+        ret = mkdir_p (gfc->gfc_processing_dir, 0600, _gf_false);
+        if (ret)
+                goto out;
+
+        dir = opendir (gfc->gfc_processing_dir);
+        if (!dir) {
+                gf_log ("", GF_LOG_ERROR,
+                        "opendir() error [reason: %s]", strerror (errno));
+                goto out;
+        }
+
+        gfc->gfc_dir = dir;
+
+        (void) snprintf (tracker_path, PATH_MAX,
+                         "%s/"GF_CHANGELOG_TRACKER, gfc->gfc_working_dir);
+
+        tracker_fd = open (tracker_path, O_CREAT | O_APPEND | O_RDWR,
+                           S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
+        if (tracker_fd < 0) {
+                closedir (gfc->gfc_dir);
+                ret = -1;
+                goto out;
+        }
+
+        gfc->gfc_fd = tracker_fd;
+        ret = 0;
+ out:
+        return ret;
+}
+
+int
+gf_changelog_notification_init (xlator_t *this, gf_changelog_t *gfc)
+{
+        int                ret    = 0;
+        int                len    = 0;
+        int                tries  = 0;
+        int                sockfd = 0;
+        struct sockaddr_un remote;
+
+        this = gfc->this;
+
+        if (gfc->gfc_sockfd != -1) {
+                gf_log (this->name, GF_LOG_INFO,
+                        "Reconnecting...");
+                close (gfc->gfc_sockfd);
+        }
+
+        sockfd = socket (AF_UNIX, SOCK_STREAM, 0);
+        if (sockfd < 0) {
+                ret = -1;
+                goto out;
+        }
+
+        CHANGELOG_MAKE_SOCKET_PATH (gfc->gfc_brickpath,
+                                    gfc->gfc_sockpath, UNIX_PATH_MAX);
+        gf_log (this->name, GF_LOG_INFO,
+                "connecting to changelog socket: %s (brick: %s)",
+                gfc->gfc_sockpath, gfc->gfc_brickpath);
+
+        remote.sun_family = AF_UNIX;
+        strcpy (remote.sun_path, gfc->gfc_sockpath);
+
+        len = strlen (remote.sun_path) + sizeof (remote.sun_family);
+
+        while (tries < gfc->gfc_connretries) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "connection attempt %d/%d...",
+                        tries + 1, gfc->gfc_connretries);
+
+                /* initiate a connect */
+                if (connect (sockfd, (struct sockaddr *) &remote, len) == 0) {
+                        gfc->gfc_sockfd = sockfd;
+                        break;
+                }
+
+                tries++;
+                sleep (2);
+        }
+
+        if (tries == gfc->gfc_connretries) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "could not connect to changelog socket!"
+                        " bailing out...");
+                close (sockfd);
+                ret = -1;
+        } else
+                gf_log (this->name, GF_LOG_INFO,
+                        "connection successful");
+
+ out:
+        return ret;
+}
+
+int
+gf_changelog_done (char *file)
+{
+        int             ret    = -1;
+        char           *buffer = NULL;
+        xlator_t       *this   = NULL;
+        gf_changelog_t *gfc    = NULL;
+        char to_path[PATH_MAX] = {0,};
+
+        errno = EINVAL;
+
+        this = THIS;
+        if (!this)
+                goto out;
+
+        gfc = (gf_changelog_t *) this->private;
+        if (!gfc)
+                goto out;
+
+        if (!file || !strlen (file))
+                goto out;
+
+        /* make sure 'file' is inside ->gfc_working_dir */
+        buffer = realpath (file, NULL);
+        if (!buffer)
+                goto out;
+
+        if (strncmp (gfc->gfc_working_dir,
+                     buffer, strlen (gfc->gfc_working_dir)))
+                goto out;
+
+        (void) snprintf (to_path, PATH_MAX, "%s%s",
+                         gfc->gfc_processed_dir, basename (buffer));
+        gf_log (this->name, GF_LOG_DEBUG,
+                "moving %s to processed directory", file);
+        ret = rename (buffer, to_path);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "cannot move %s to %s (reason: %s)",
+                        file, to_path, strerror (errno));
+                goto out;
+        }
+
+        ret = 0;
+
+ out:
+        if (buffer)
+                free (buffer); /* allocated by realpath() */
+        return ret;
+}
+
+/**
+ * @API
+ *  for a set of changelogs, start from the begining
+ */
+int
+gf_changelog_start_fresh ()
+{
+        xlator_t *this = NULL;
+        gf_changelog_t *gfc = NULL;
+
+        this = THIS;
+        if (!this)
+                goto out;
+
+        errno = EINVAL;
+
+        gfc = (gf_changelog_t *) this->private;
+        if (!gfc)
+                goto out;
+
+        if (gf_ftruncate (gfc->gfc_fd, 0))
+                goto out;
+
+        return 0;
+
+ out:
+        return -1;
+}
+
+/**
+ * @API
+ * return the next changelog file entry. zero means all chanelogs
+ * consumed.
+ */
+ssize_t
+gf_changelog_next_change (char *bufptr, size_t maxlen)
+{
+        ssize_t         size       = 0;
+        int             tracker_fd = 0;
+        xlator_t       *this       = NULL;
+        gf_changelog_t *gfc        = NULL;
+        char buffer[PATH_MAX]      = {0,};
+
+        errno = EINVAL;
+
+        this = THIS;
+        if (!this)
+                goto out;
+
+        gfc = (gf_changelog_t *) this->private;
+        if (!gfc)
+                goto out;
+
+        tracker_fd = gfc->gfc_fd;
+
+        size = gf_readline (tracker_fd, buffer, maxlen);
+        if (size < 0)
+                goto out;
+        if (size == 0)
+                return 0;
+
+        memcpy (bufptr, buffer, size - 1);
+        *(buffer + size) = '\0';
+
+        return size;
+
+ out:
+        return -1;
+}
+
+/**
+ * @API
+ *  gf_changelog_scan() - scan and generate a list of change entries
+ *
+ * calling this api multiple times (without calling gf_changlog_done())
+ * would result new changelogs(s) being refreshed in the tracker file.
+ * This call also acts as a cancellation point for the consumer.
+ */
+ssize_t
+gf_changelog_scan ()
+{
+        int             ret        = 0;
+        int             tracker_fd = 0;
+        size_t          len        = 0;
+        size_t          off        = 0;
+        xlator_t       *this       = NULL;
+        size_t          nr_entries = 0;
+        gf_changelog_t *gfc        = NULL;
+        struct dirent  *entryp     = NULL;
+        struct dirent  *result     = NULL;
+        char buffer[PATH_MAX]      = {0,};
+
+        this = THIS;
+        if (!this)
+                goto out;
+
+        gfc = (gf_changelog_t *) this->private;
+        if (!gfc)
+                goto out;
+
+        /**
+         * do we need to protect 'byebye' with locks? worst, the
+         * consumer would get notified during next scan().
+         */
+        if (byebye) {
+                errno = ECONNREFUSED;
+                goto out;
+        }
+
+        errno = EINVAL;
+
+        tracker_fd = gfc->gfc_fd;
+
+        if (gf_ftruncate (tracker_fd, 0))
+                goto out;
+
+        len = offsetof(struct dirent, d_name)
+                + pathconf(gfc->gfc_processing_dir, _PC_NAME_MAX) + 1;
+        entryp = GF_CALLOC (1, len,
+                            gf_changelog_mt_libgfchangelog_dirent_t);
+        if (!entryp)
+                goto out;
+
+        rewinddir (gfc->gfc_dir);
+        while (1) {
+                ret = readdir_r (gfc->gfc_dir, entryp, &result);
+                if (ret || !result)
+                        break;
+
+                if ( !strcmp (basename (entryp->d_name), ".")
+                     || !strcmp (basename (entryp->d_name), "..") )
+                        continue;
+
+                nr_entries++;
+
+                GF_CHANGELOG_FILL_BUFFER (gfc->gfc_processing_dir,
+                                          buffer, off,
+                                          strlen (gfc->gfc_processing_dir));
+                GF_CHANGELOG_FILL_BUFFER (entryp->d_name, buffer,
+                                          off, strlen (entryp->d_name));
+                GF_CHANGELOG_FILL_BUFFER ("\n", buffer, off, 1);
+
+                if (gf_changelog_write (tracker_fd, buffer, off) != off) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "error writing changelog filename"
+                                " to tracker file");
+                        break;
+                }
+                off = 0;
+        }
+
+        GF_FREE (entryp);
+
+        if (!result) {
+                if (gf_lseek (tracker_fd, 0, SEEK_SET) != -1)
+                        return nr_entries;
+        }
+ out:
+        return -1;
+}
+
+/**
+ * @API
+ *  gf_changelog_register() - register a client for updates.
+ */
+int
+gf_changelog_register (char *brick_path, char *scratch_dir,
+                       char *log_file, int log_level, int max_reconnects)
+{
+        int             i    = 0;
+        int             ret  = -1;
+        int             errn = 0;
+        xlator_t       *this = NULL;
+        gf_changelog_t *gfc  = NULL;
+
+        this = THIS;
+        if (!this->ctx)
+                goto out;
+
+        errno = ENOMEM;
+
+        gfc = GF_CALLOC (1, sizeof (*gfc),
+                         gf_changelog_mt_libgfchangelog_t);
+        if (!gfc)
+                goto out;
+
+        gfc->this = this;
+
+        gfc->gfc_dir = NULL;
+        gfc->gfc_fd = gfc->gfc_sockfd = -1;
+
+        gfc->gfc_working_dir = realpath (scratch_dir, NULL);
+        if (!gfc->gfc_working_dir) {
+                errn = errno;
+                goto cleanup;
+        }
+
+        ret = gf_changelog_open_dirs (gfc);
+        if (ret) {
+                errn = errno;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "could not create entries in scratch dir");
+                goto cleanup;
+        }
+
+        /* passing ident as NULL means to use default ident for syslog */
+        if (gf_log_init (this->ctx, log_file, NULL))
+                goto cleanup;
+
+        gf_log_set_loglevel ((log_level == -1) ? GF_LOG_INFO :
+                             log_level);
+
+        gfc->gfc_connretries = (max_reconnects <= 0) ? 1 : max_reconnects;
+        (void) strncpy (gfc->gfc_brickpath, brick_path, PATH_MAX);
+
+        ret = gf_changelog_notification_init (this, gfc);
+        if (ret) {
+                errn = errno;
+                goto cleanup;
+        }
+
+        ret = gf_thread_create (&gfc->gfc_changelog_processor,
+				NULL, gf_changelog_process, gfc);
+        if (ret) {
+                errn = errno;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "error creating changelog processor thread"
+                        " new changes won't be recorded!!!");
+                goto cleanup;
+        }
+
+        for (; i < 256; i++) {
+                gfc->rfc3986[i] =
+                        (isalnum(i) || i == '~' ||
+                        i == '-' || i == '.' || i == '_') ? i : 0;
+        }
+
+        ret = 0;
+        this->private = gfc;
+
+        goto out;
+
+ cleanup:
+        gf_changelog_cleanup (gfc);
+        GF_FREE (gfc);
+        this->private = NULL;
+        errno = errn;
+
+ out:
+        return ret;
+}
diff --git a/xlators/features/changelog/src/Makefile.am b/xlators/features/changelog/src/Makefile.am
new file mode 100644
index 000000000..54c21ac21
--- /dev/null
+++ b/xlators/features/changelog/src/Makefile.am
@@ -0,0 +1,19 @@
+xlator_LTLIBRARIES = changelog.la
+
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+noinst_HEADERS = changelog-helpers.h changelog-mem-types.h changelog-rt.h \
+	changelog-misc.h changelog-encoders.h changelog-notifier.h
+
+changelog_la_LDFLAGS = -module -avoid-version
+
+changelog_la_SOURCES = changelog.c changelog-rt.c changelog-helpers.c \
+	changelog-encoders.c changelog-notifier.c
+changelog_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src -fPIC -D_FILE_OFFSET_BITS=64 \
+	-D_GNU_SOURCE -D$(GF_HOST_OS) -shared -nostartfiles -DDATADIR=\"$(localstatedir)\"
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+CLEANFILES =
diff --git a/xlators/features/changelog/src/changelog-encoders.c b/xlators/features/changelog/src/changelog-encoders.c
new file mode 100644
index 000000000..08626ee2f
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-encoders.c
@@ -0,0 +1,197 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "changelog-encoders.h"
+
+size_t
+entry_fn (void *data, char *buffer, gf_boolean_t encode)
+{
+        char    *tmpbuf = NULL;
+        size_t  bufsz  = 0;
+        struct changelog_entry_fields *ce = NULL;
+
+        ce = (struct changelog_entry_fields *) data;
+
+        if (encode) {
+                tmpbuf = uuid_utoa (ce->cef_uuid);
+                CHANGELOG_FILL_BUFFER (buffer, bufsz, tmpbuf, strlen (tmpbuf));
+        } else {
+                CHANGELOG_FILL_BUFFER (buffer, bufsz,
+                                       ce->cef_uuid, sizeof (uuid_t));
+        }
+
+        CHANGELOG_FILL_BUFFER (buffer, bufsz, "/", 1);
+        CHANGELOG_FILL_BUFFER (buffer, bufsz,
+                               ce->cef_bname, strlen (ce->cef_bname));
+        return bufsz;
+}
+
+size_t
+fop_fn (void *data, char *buffer, gf_boolean_t encode)
+{
+        char buf[10]          = {0,};
+        size_t         bufsz = 0;
+        glusterfs_fop_t fop   = 0;
+
+        fop = *(glusterfs_fop_t *) data;
+
+        if (encode) {
+                (void) snprintf (buf, sizeof (buf), "%d", fop);
+                CHANGELOG_FILL_BUFFER (buffer, bufsz, buf, strlen (buf));
+        } else
+                CHANGELOG_FILL_BUFFER (buffer, bufsz, &fop, sizeof (fop));
+
+        return bufsz;
+}
+
+size_t
+number_fn (void *data, char *buffer, gf_boolean_t encode)
+{
+        size_t       bufsz = 0;
+        unsigned int nr    = 0;
+        char buf[20]       = {0,};
+
+        nr = *(unsigned int *) data;
+
+        if (encode) {
+                (void) snprintf (buf, sizeof (buf), "%u", nr);
+                CHANGELOG_FILL_BUFFER (buffer, bufsz, buf, strlen (buf));
+        } else
+                CHANGELOG_FILL_BUFFER (buffer, bufsz, &nr, sizeof (unsigned int));
+
+        return bufsz;
+}
+
+void
+entry_free_fn (void *data)
+{
+        changelog_opt_t *co = data;
+
+        if (!co)
+                return;
+
+        GF_FREE (co->co_entry.cef_bname);
+}
+
+/**
+ * try to write all data in one shot
+ */
+
+static inline void
+changelog_encode_write_xtra (changelog_log_data_t *cld,
+                             char *buffer, size_t *off, gf_boolean_t encode)
+{
+        int              i      = 0;
+        size_t           offset = 0;
+        void            *data   = NULL;
+        changelog_opt_t *co     = NULL;
+
+        offset = *off;
+
+        co = (changelog_opt_t *) cld->cld_ptr;
+
+        for (; i < cld->cld_xtra_records; i++, co++) {
+                CHANGELOG_FILL_BUFFER (buffer, offset, "\0", 1);
+
+                switch (co->co_type) {
+                case CHANGELOG_OPT_REC_FOP:
+                        data = &co->co_fop;
+                        break;
+                case CHANGELOG_OPT_REC_ENTRY:
+                        data = &co->co_entry;
+                        break;
+                case CHANGELOG_OPT_REC_UINT32:
+                        data = &co->co_uint32;
+                        break;
+                }
+
+                if (co->co_convert)
+                        offset += co->co_convert (data,
+                                                  buffer + offset, encode);
+                else /* no coversion: write it out as it is */
+                        CHANGELOG_FILL_BUFFER (buffer, offset,
+                                               data, co->co_len);
+        }
+
+        *off = offset;
+}
+
+int
+changelog_encode_ascii (xlator_t *this, changelog_log_data_t *cld)
+{
+        size_t            off      = 0;
+        size_t            gfid_len = 0;
+        char             *gfid_str = NULL;
+        char             *buffer   = NULL;
+        changelog_priv_t *priv     = NULL;
+
+        priv = this->private;
+
+        gfid_str = uuid_utoa (cld->cld_gfid);
+        gfid_len = strlen (gfid_str);
+
+        /* extra bytes for decorations */
+        buffer = alloca (gfid_len + cld->cld_ptr_len + 10);
+        CHANGELOG_STORE_ASCII (priv, buffer,
+                               off, gfid_str, gfid_len, cld);
+
+        if (cld->cld_xtra_records)
+                changelog_encode_write_xtra (cld, buffer, &off, _gf_true);
+
+        CHANGELOG_FILL_BUFFER (buffer, off, "\0", 1);
+
+        return changelog_write_change (priv, buffer, off);
+}
+
+int
+changelog_encode_binary (xlator_t *this, changelog_log_data_t *cld)
+{
+        size_t            off    = 0;
+        char             *buffer = NULL;
+        changelog_priv_t *priv   = NULL;
+
+        priv = this->private;
+
+        /* extra bytes for decorations */
+        buffer = alloca (sizeof (uuid_t) + cld->cld_ptr_len + 10);
+        CHANGELOG_STORE_BINARY (priv, buffer, off, cld->cld_gfid, cld);
+
+        if (cld->cld_xtra_records)
+                changelog_encode_write_xtra (cld, buffer, &off, _gf_false);
+
+        CHANGELOG_FILL_BUFFER (buffer, off, "\0", 1);
+
+        return changelog_write_change (priv, buffer, off);
+}
+
+static struct changelog_encoder
+cb_encoder[] = {
+        [CHANGELOG_ENCODE_BINARY] =
+        {
+                .encoder = CHANGELOG_ENCODE_BINARY,
+                .encode = changelog_encode_binary,
+        },
+        [CHANGELOG_ENCODE_ASCII] =
+        {
+                .encoder = CHANGELOG_ENCODE_ASCII,
+                .encode = changelog_encode_ascii,
+        },
+};
+
+void
+changelog_encode_change( changelog_priv_t * priv)
+{
+        priv->ce = &cb_encoder[priv->encode_mode];
+}
diff --git a/xlators/features/changelog/src/changelog-encoders.h b/xlators/features/changelog/src/changelog-encoders.h
new file mode 100644
index 000000000..c5dcc8a77
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-encoders.h
@@ -0,0 +1,48 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CHANGELOG_ENCODERS_H
+#define _CHANGELOG_ENCODERS_H
+
+#include "xlator.h"
+#include "defaults.h"
+
+#include "changelog-helpers.h"
+
+#define CHANGELOG_STORE_ASCII(priv, buf, off, gfid, gfid_len, cld) do { \
+                CHANGELOG_FILL_BUFFER (buffer, off,                     \
+                                       priv->maps[cld->cld_type], 1);   \
+                CHANGELOG_FILL_BUFFER (buffer,                          \
+                                       off, gfid, gfid_len);            \
+        } while (0)
+
+#define CHANGELOG_STORE_BINARY(priv, buf, off, gfid, cld) do {          \
+                CHANGELOG_FILL_BUFFER (buffer, off,                     \
+                                       priv->maps[cld->cld_type], 1);   \
+                CHANGELOG_FILL_BUFFER (buffer,                          \
+                                       off, gfid, sizeof (uuid_t));     \
+        } while (0)
+
+size_t
+entry_fn (void *data, char *buffer, gf_boolean_t encode);
+size_t
+fop_fn (void *data, char *buffer, gf_boolean_t encode);
+size_t
+number_fn (void *data, char *buffer, gf_boolean_t encode);
+void
+entry_free_fn (void *data);
+int
+changelog_encode_binary (xlator_t *, changelog_log_data_t *);
+int
+changelog_encode_ascii (xlator_t *, changelog_log_data_t *);
+void
+changelog_encode_change(changelog_priv_t *);
+
+#endif /* _CHANGELOG_ENCODERS_H */
diff --git a/xlators/features/changelog/src/changelog-helpers.c b/xlators/features/changelog/src/changelog-helpers.c
new file mode 100644
index 000000000..91c43a16c
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-helpers.c
@@ -0,0 +1,696 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+#include "logging.h"
+#include "iobuf.h"
+
+#include "changelog-helpers.h"
+#include "changelog-mem-types.h"
+
+#include "changelog-encoders.h"
+#include <pthread.h>
+
+void
+changelog_thread_cleanup (xlator_t *this, pthread_t thr_id)
+{
+        int   ret    = 0;
+        void *retval = NULL;
+
+        /* send a cancel request to the thread */
+        ret = pthread_cancel (thr_id);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "could not cancel thread (reason: %s)",
+                        strerror (errno));
+                goto out;
+        }
+
+        ret = pthread_join (thr_id, &retval);
+        if (ret || (retval != PTHREAD_CANCELED)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "cancel request not adhered as expected"
+                        " (reason: %s)", strerror (errno));
+        }
+
+ out:
+        return;
+}
+
+inline void *
+changelog_get_usable_buffer (changelog_local_t *local)
+{
+        changelog_log_data_t *cld = NULL;
+
+        if (!local)
+                return NULL;
+
+        cld = &local->cld;
+        if (!cld->cld_iobuf)
+                return NULL;
+
+        return cld->cld_iobuf->ptr;
+}
+
+inline void
+changelog_set_usable_record_and_length (changelog_local_t *local,
+                                        size_t len, int xr)
+{
+        changelog_log_data_t *cld = NULL;
+
+        cld = &local->cld;
+
+        cld->cld_ptr_len = len;
+        cld->cld_xtra_records = xr;
+}
+
+void
+changelog_local_cleanup (xlator_t *xl, changelog_local_t *local)
+{
+        int                   i   = 0;
+        changelog_opt_t      *co  = NULL;
+        changelog_log_data_t *cld = NULL;
+
+        if (!local)
+                return;
+
+        cld = &local->cld;
+
+        /* cleanup dynamic allocation for extra records */
+        if (cld->cld_xtra_records) {
+                co = (changelog_opt_t *) cld->cld_ptr;
+                for (; i < cld->cld_xtra_records; i++, co++)
+                        if (co->co_free)
+                                co->co_free (co);
+        }
+
+        CHANGELOG_IOBUF_UNREF (cld->cld_iobuf);
+
+        if (local->inode)
+                inode_unref (local->inode);
+
+        mem_put (local);
+}
+
+inline int
+changelog_write (int fd, char *buffer, size_t len)
+{
+        ssize_t size = 0;
+        size_t writen = 0;
+
+        while (writen < len) {
+                size = write (fd,
+                              buffer + writen, len - writen);
+                if (size <= 0)
+                        break;
+
+                writen += size;
+        }
+
+        return (writen != len);
+}
+
+static int
+changelog_rollover_changelog (xlator_t *this,
+                              changelog_priv_t *priv, unsigned long ts)
+{
+        int   ret            = -1;
+        int   notify         = 0;
+        char *bname          = NULL;
+        char ofile[PATH_MAX] = {0,};
+        char nfile[PATH_MAX] = {0,};
+
+        if (priv->changelog_fd != -1) {
+                close (priv->changelog_fd);
+                priv->changelog_fd = -1;
+        }
+
+        (void) snprintf (ofile, PATH_MAX,
+                         "%s/"CHANGELOG_FILE_NAME, priv->changelog_dir);
+        (void) snprintf (nfile, PATH_MAX,
+                         "%s/"CHANGELOG_FILE_NAME".%lu",
+                         priv->changelog_dir, ts);
+
+        ret = rename (ofile, nfile);
+        if (!ret)
+                notify = 1;
+
+        if (ret && (errno == ENOENT)) {
+                ret = 0;
+        }
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "error renaming %s -> %s (reason %s)",
+                        ofile, nfile, strerror (errno));
+        }
+
+        if (notify) {
+                bname = basename (nfile);
+                gf_log (this->name, GF_LOG_DEBUG, "notifying: %s", bname);
+                ret = changelog_write (priv->wfd, bname, strlen (bname) + 1);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to send file name to notify thread"
+                                " (reason: %s)", strerror (errno));
+                }
+        }
+
+        return ret;
+}
+
+int
+changelog_open (xlator_t *this,
+                changelog_priv_t *priv)
+{
+        int fd                        = 0;
+        int ret                       = -1;
+        int flags                     = 0;
+        char buffer[1024]             = {0,};
+        char changelog_path[PATH_MAX] = {0,};
+
+        (void) snprintf (changelog_path, PATH_MAX,
+                         "%s/"CHANGELOG_FILE_NAME,
+                         priv->changelog_dir);
+
+        flags |= (O_CREAT | O_RDWR);
+        if (priv->fsync_interval == 0)
+                flags |= O_SYNC;
+
+        fd = open (changelog_path, flags,
+                   S_IRUSR | S_IWUSR | S_IRGRP | S_IROTH);
+        if (fd < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "unable to open/create changelog file %s"
+                        " (reason: %s). change-logging will be"
+                        " inactive", changelog_path, strerror (errno));
+                goto out;
+        }
+
+        priv->changelog_fd = fd;
+
+        (void) snprintf (buffer, 1024, CHANGELOG_HEADER,
+                         CHANGELOG_VERSION_MAJOR,
+                         CHANGELOG_VERSION_MINOR,
+                         priv->ce->encoder);
+        ret = changelog_write_change (priv, buffer, strlen (buffer));
+        if (ret) {
+                close (priv->changelog_fd);
+                priv->changelog_fd = -1;
+                goto out;
+        }
+
+        ret = 0;
+
+ out:
+        return ret;
+}
+
+int
+changelog_start_next_change (xlator_t *this,
+                             changelog_priv_t *priv,
+                             unsigned long ts, gf_boolean_t finale)
+{
+        int ret = -1;
+
+        ret = changelog_rollover_changelog (this, priv, ts);
+
+        if (!ret && !finale)
+                ret = changelog_open (this, priv);
+
+        return ret;
+}
+
+/**
+ * return the length of entry
+ */
+inline size_t
+changelog_entry_length ()
+{
+        return sizeof (changelog_log_data_t);
+}
+
+int
+changelog_fill_rollover_data (changelog_log_data_t *cld, gf_boolean_t is_last)
+{
+        struct timeval tv = {0,};
+
+        cld->cld_type = CHANGELOG_TYPE_ROLLOVER;
+
+        if (gettimeofday (&tv, NULL))
+                return -1;
+
+        cld->cld_roll_time = (unsigned long) tv.tv_sec;
+        cld->cld_finale = is_last;
+        return 0;
+}
+
+int
+changelog_write_change (changelog_priv_t *priv, char *buffer, size_t len)
+{
+        return changelog_write (priv->changelog_fd, buffer, len);
+}
+
+inline int
+changelog_handle_change (xlator_t *this,
+                         changelog_priv_t *priv, changelog_log_data_t *cld)
+{
+        int ret = 0;
+
+        if (CHANGELOG_TYPE_IS_ROLLOVER (cld->cld_type)) {
+                changelog_encode_change(priv);
+                ret = changelog_start_next_change (this, priv,
+                                                   cld->cld_roll_time,
+                                                   cld->cld_finale);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Problem rolling over changelog(s)");
+                goto out;
+        }
+
+        /**
+         * case when there is reconfigure done (disabling changelog) and there
+         * are still fops that have updates in prgress.
+         */
+        if (priv->changelog_fd == -1)
+                return 0;
+
+        if (CHANGELOG_TYPE_IS_FSYNC (cld->cld_type)) {
+                ret = fsync (priv->changelog_fd);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "fsync failed (reason: %s)",
+                                strerror (errno));
+                }
+                goto out;
+        }
+
+        ret = priv->ce->encode (this, cld);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "error writing changelog to disk");
+        }
+
+ out:
+        return ret;
+}
+
+changelog_local_t *
+changelog_local_init (xlator_t *this, inode_t *inode,
+                      uuid_t gfid, int xtra_records,
+                      gf_boolean_t update_flag)
+{
+        changelog_local_t *local = NULL;
+        struct iobuf      *iobuf = NULL;
+
+        /**
+         * We relax the presence of inode if @update_flag is true.
+         * The caller (implmentation of the fop) needs to be careful to
+         * not blindly use local->inode.
+         */
+        if (!update_flag && !inode) {
+                gf_log_callingfn (this->name, GF_LOG_WARNING,
+                                  "inode needed for version checking !!!");
+                goto out;
+        }
+
+        if (xtra_records) {
+                iobuf = iobuf_get2 (this->ctx->iobuf_pool,
+                                    xtra_records * CHANGELOG_OPT_RECORD_LEN);
+                if (!iobuf)
+                        goto out;
+        }
+
+        local = mem_get0 (this->local_pool);
+        if (!local) {
+                CHANGELOG_IOBUF_UNREF (iobuf);
+                goto out;
+        }
+
+        local->update_no_check = update_flag;
+
+        uuid_copy (local->cld.cld_gfid, gfid);
+
+        local->cld.cld_iobuf = iobuf;
+        local->cld.cld_xtra_records = 0; /* set by the caller */
+
+        if (inode)
+                local->inode = inode_ref (inode);
+
+ out:
+        return local;
+}
+
+int
+changelog_forget (xlator_t *this, inode_t *inode)
+{
+        uint64_t ctx_addr = 0;
+        changelog_inode_ctx_t *ctx = NULL;
+
+        inode_ctx_del (inode, this, &ctx_addr);
+        if (!ctx_addr)
+                return 0;
+
+        ctx = (changelog_inode_ctx_t *) (long) ctx_addr;
+        GF_FREE (ctx);
+
+        return 0;
+}
+
+int
+changelog_inject_single_event (xlator_t *this,
+                               changelog_priv_t *priv,
+                               changelog_log_data_t *cld)
+{
+        return priv->cd.dispatchfn (this, priv, priv->cd.cd_data, cld, NULL);
+}
+
+/**
+ * TODO: these threads have many thing in common (wake up after
+ * a certain time etc..). move them into separate routine.
+ */
+void *
+changelog_rollover (void *data)
+{
+        int                     ret   = 0;
+        xlator_t               *this  = NULL;
+        struct timeval          tv    = {0,};
+        changelog_log_data_t    cld   = {0,};
+        changelog_time_slice_t *slice = NULL;
+        changelog_priv_t       *priv  = data;
+
+        this = priv->cr.this;
+        slice = &priv->slice;
+
+        while (1) {
+                tv.tv_sec  = priv->rollover_time;
+                tv.tv_usec = 0;
+
+                ret = select (0, NULL, NULL, NULL, &tv);
+                if (ret)
+                        continue;
+
+                ret = changelog_fill_rollover_data (&cld, _gf_false);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to fill rollover data");
+                        continue;
+                }
+
+                LOCK (&priv->lock);
+                {
+                        ret = changelog_inject_single_event (this, priv, &cld);
+                        if (!ret)
+                                SLICE_VERSION_UPDATE (slice);
+                }
+                UNLOCK (&priv->lock);
+        }
+
+        return NULL;
+}
+
+void *
+changelog_fsync_thread (void *data)
+{
+        int                   ret  = 0;
+        xlator_t             *this = NULL;
+        struct timeval        tv   = {0,};
+        changelog_log_data_t  cld  = {0,};
+        changelog_priv_t     *priv = data;
+
+        this = priv->cf.this;
+        cld.cld_type = CHANGELOG_TYPE_FSYNC;
+
+        while (1) {
+                tv.tv_sec  = priv->fsync_interval;
+                tv.tv_usec = 0;
+
+                ret = select (0, NULL, NULL, NULL, &tv);
+                if (ret)
+                        continue;
+
+                ret = changelog_inject_single_event (this, priv, &cld);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "failed to inject fsync event");
+        }
+
+        return NULL;
+}
+
+/* macros for inode/changelog version checks */
+
+#define INODE_VERSION_UPDATE(priv, inode, iver, slice, type) do {       \
+                LOCK (&inode->lock);                                    \
+                {                                                       \
+                        LOCK (&priv->lock);                             \
+                        {                                               \
+                                *iver = slice->changelog_version[type]; \
+                        }                                               \
+                        UNLOCK (&priv->lock);                           \
+                }                                                       \
+                UNLOCK (&inode->lock);                                  \
+        } while (0)
+
+#define INODE_VERSION_EQUALS_SLICE(priv, ver, slice, type, upd) do {    \
+                LOCK (&priv->lock);                                     \
+                {                                                       \
+                        upd = (ver == slice->changelog_version[type])   \
+                                ? _gf_false : _gf_true;                 \
+                }                                                       \
+                UNLOCK (&priv->lock);                                   \
+        } while (0)
+
+static int
+__changelog_inode_ctx_set (xlator_t *this,
+                           inode_t *inode, changelog_inode_ctx_t *ctx)
+{
+        uint64_t ctx_addr = (uint64_t) ctx;
+        return __inode_ctx_set (inode, this, &ctx_addr);
+}
+
+/**
+ * one shot routine to get the address and the value of a inode version
+ * for a particular type.
+ */
+static changelog_inode_ctx_t *
+__changelog_inode_ctx_get (xlator_t *this,
+                           inode_t *inode, unsigned long **iver,
+                           unsigned long *version, changelog_log_type type)
+{
+        int                    ret      = 0;
+        uint64_t               ctx_addr = 0;
+        changelog_inode_ctx_t *ctx      = NULL;
+
+        ret = __inode_ctx_get (inode, this, &ctx_addr);
+        if (ret < 0)
+                ctx_addr = 0;
+        if (ctx_addr != 0) {
+                ctx = (changelog_inode_ctx_t *) (long)ctx_addr;
+                goto out;
+        }
+
+        ctx = GF_CALLOC (1, sizeof (*ctx), gf_changelog_mt_inode_ctx_t);
+        if (!ctx)
+                goto out;
+
+        ret = __changelog_inode_ctx_set (this, inode, ctx);
+        if (ret) {
+                GF_FREE (ctx);
+                ctx = NULL;
+        }
+
+ out:
+        if (ctx && iver && version) {
+                *iver = CHANGELOG_INODE_VERSION_TYPE (ctx, type);
+                *version = **iver;
+        }
+
+        return ctx;
+}
+
+static changelog_inode_ctx_t *
+changelog_inode_ctx_get (xlator_t *this,
+                         inode_t *inode, unsigned long **iver,
+                         unsigned long *version, changelog_log_type type)
+{
+        changelog_inode_ctx_t *ctx = NULL;
+
+        LOCK (&inode->lock);
+        {
+                ctx = __changelog_inode_ctx_get (this,
+                                                 inode, iver, version, type);
+        }
+        UNLOCK (&inode->lock);
+
+        return ctx;
+}
+
+/**
+ * This is the main update routine. Locking has been made granular so as to
+ * maximize parallelism of fops - I'll try to explain it below using execution
+ * timelines.
+ *
+ * Basically, the contention is between multiple execution threads of this
+ * routine and the roll-over thread. So, instead of having a big lock, we hold
+ * granular locks: inode->lock and priv->lock. Now I'll explain what happens
+ * when there is an update and a roll-over at just about the same time.
+ * NOTE:
+ *  - the dispatcher itself synchronizes updates via it's own lock
+ *  - the slice version in incremented by the roll-over thread
+ *
+ * Case 1: When the rollover thread wins before the inode version can be
+ * compared with the slice version.
+ *
+ *          [updater]                 |             [rollover]
+ *                                    |
+ *                                    |           <SLICE: 1, 1, 1>
+ * <changelog_update>                 |
+ *   <changelog_inode_ctx_get>        |
+ *      <CTX: 1, 1, 1>                |
+ *                                    |         <dispatch-rollover-event>
+ *                                    |         LOCK (&priv->lock)
+ *                                    |            <SLICE_VERSION_UPDATE>
+ *                                    |              <SLICE: 2, 2, 2>
+ *                                    |         UNLOCK (&priv->lock)
+ *                                    |
+ * LOCK (&priv->lock)                 |
+ *   <INODE_VERSION_EQUALS_SLICE>     |
+ *    I: 1 <-> S: 2                   |
+ *    update: true                    |
+ * UNLOCK (&priv->lock)               |
+ *                                    |
+ * <if update == true>                |
+ *  <dispath-update-event>            |
+ *  <INODE_VERSION_UPDATE>            |
+ *   LOCK (&inode->lock)              |
+ *    LOCK (&priv->lock)              |
+ *     <CTX: 2, 1, 1>                 |
+ *    UNLOCK (&priv->lock)            |
+ *   UNLOCK (&inode->lock)            |
+ *
+ * Therefore, the change gets recorded in the next change (no lost change). If
+ * the slice version was ahead of the inode version (say I:1, S: 2), then
+ * anyway the comparison would result in a update (I: 1, S: 3).
+ *
+ * If the rollover time is too less, then there is another contention when the
+ * updater tries to bring up inode version to the slice version (this is also
+ * the case when the roll-over thread wakes up during INODE_VERSION_UPDATE.
+ *
+ *   <CTX: 1, 1, 1>                   |       <SLICE: 2, 2, 2>
+ *                                    |
+ *                                    |
+ * <dispath-update-event>             |
+ * <INODE_VERSION_UPDATE>             |
+ *  LOCK (&inode->lock)               |
+ *   LOCK (&priv->lock)               |
+ *    <CTX: 2, 1, 1>                  |
+ *   UNLOCK (&priv->lock)             |
+ *  UNLOCK (&inode->lock)             |
+ *                                    |         <dispatch-rollover-event>
+ *                                    |         LOCK (&priv->lock)
+ *                                    |            <SLICE_VERSION_UPDATE>
+ *                                    |              <SLICE: 3, 3, 3>
+ *                                    |         UNLOCK (&priv->lock)
+ *
+ *
+ * Case 2: When the fop thread wins
+ *
+ *          [updater]                 |             [rollover]
+ *                                    |
+ *                                    |           <SLICE: 1, 1, 1>
+ * <changelog_update>                 |
+ *   <changelog_inode_ctx_get>        |
+ *      <CTX: 0, 0, 0>                |
+ *                                    |
+ * LOCK (&priv->lock)                 |
+ *   <INODE_VERSION_EQUALS_SLICE>     |
+ *    I: 0 <-> S: 1                   |
+ *    update: true                    |
+ * UNLOCK (&priv->lock)               |
+ *                                    |         <dispatch-rollover-event>
+ *                                    |         LOCK (&priv->lock)
+ *                                    |            <SLICE_VERSION_UPDATE>
+ *                                    |              <SLICE: 2, 2, 2>
+ *                                    |         UNLOCK (&priv->lock)
+ * <if update == true>                |
+ *  <dispath-update-event>            |
+ *  <INODE_VERSION_UPDATE>            |
+ *   LOCK (&inode->lock)              |
+ *    LOCK (&priv->lock)              |
+ *     <CTX: 2, 0, 0>                 |
+ *    UNLOCK (&priv->lock)            |
+ *   UNLOCK (&inode->lock)            |
+ *
+ * Here again, if the inode version was equal to the slice version (I: 1, S: 1)
+ * then there is no need to record an update (as the equality of the two version
+ * signifies an update was recorded in the current time slice).
+ */
+inline void
+changelog_update (xlator_t *this, changelog_priv_t *priv,
+                  changelog_local_t *local, changelog_log_type type)
+{
+        int                     ret        = 0;
+        unsigned long          *iver       = NULL;
+        unsigned long           version    = 0;
+        inode_t                *inode      = NULL;
+        changelog_time_slice_t *slice      = NULL;
+        changelog_inode_ctx_t  *ctx        = NULL;
+        changelog_log_data_t   *cld_0      = NULL;
+        changelog_log_data_t   *cld_1      = NULL;
+        changelog_local_t      *next_local = NULL;
+        gf_boolean_t            need_upd   = _gf_true;
+
+        slice = &priv->slice;
+
+        /**
+         * for fops that do not require inode version checking
+         */
+        if (local->update_no_check)
+                goto update;
+
+        inode = local->inode;
+
+        ctx = changelog_inode_ctx_get (this,
+                                       inode, &iver, &version, type);
+        if (!ctx)
+                goto update;
+
+        INODE_VERSION_EQUALS_SLICE (priv, version, slice, type, need_upd);
+
+ update:
+        if (need_upd) {
+                cld_0 = &local->cld;
+                cld_0->cld_type = type;
+
+                if ( (next_local = local->prev_entry) != NULL ) {
+                        cld_1 = &next_local->cld;
+                        cld_1->cld_type = type;
+                }
+
+                ret = priv->cd.dispatchfn (this, priv,
+                                           priv->cd.cd_data, cld_0, cld_1);
+
+                /**
+                 * update after the dispatcher has successfully done
+                 * it's job.
+                 */
+                if (!local->update_no_check && iver && !ret)
+                        INODE_VERSION_UPDATE (priv, inode, iver, slice, type);
+        }
+
+        return;
+}
diff --git a/xlators/features/changelog/src/changelog-helpers.h b/xlators/features/changelog/src/changelog-helpers.h
new file mode 100644
index 000000000..b441f7068
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-helpers.h
@@ -0,0 +1,405 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CHANGELOG_HELPERS_H
+#define _CHANGELOG_HELPERS_H
+
+#include "locking.h"
+#include "timer.h"
+#include "pthread.h"
+#include "iobuf.h"
+
+#include "changelog-misc.h"
+
+/**
+ * the changelog entry
+ */
+typedef struct changelog_log_data {
+        /* rollover related */
+        unsigned long cld_roll_time;
+
+        /* reopen changelog? */
+        gf_boolean_t cld_finale;
+
+        changelog_log_type cld_type;
+
+        /**
+         * sincd gfid is _always_ a necessity, it's not a part
+         * of the iobuf. by doing this we do not add any overhead
+         * for data and metadata related fops.
+         */
+        uuid_t        cld_gfid;
+
+        /**
+         * iobufs are used for optionals records: pargfid, path,
+         * write offsets etc.. It's the fop implementers job
+         * to allocate (iobuf_get() in the fop) and get unref'ed
+         * in the callback (CHANGELOG_STACK_UNWIND).
+         */
+        struct iobuf *cld_iobuf;
+
+#define cld_ptr cld_iobuf->ptr
+
+        /**
+         * after allocation you can point this to the length of
+         * usable data, but make sure it does not exceed the
+         * the size of the requested iobuf.
+         */
+        size_t        cld_iobuf_len;
+
+#define cld_ptr_len cld_iobuf_len
+
+        /**
+         * number of optional records
+         */
+        int cld_xtra_records;
+} changelog_log_data_t;
+
+/**
+ * holder for dispatch function and private data
+ */
+
+typedef struct changelog_priv changelog_priv_t;
+
+typedef struct changelog_dispatcher {
+        void *cd_data;
+        int (*dispatchfn) (xlator_t *, changelog_priv_t *, void *,
+                           changelog_log_data_t *, changelog_log_data_t *);
+} changelog_dispatcher_t;
+
+struct changelog_bootstrap {
+        changelog_mode_t mode;
+        int (*ctor) (xlator_t *, changelog_dispatcher_t *);
+        int (*dtor) (xlator_t *, changelog_dispatcher_t *);
+};
+
+struct changelog_encoder {
+        changelog_encoder_t encoder;
+        int (*encode) (xlator_t *, changelog_log_data_t *);
+};
+
+
+/* xlator private */
+
+typedef struct changelog_time_slice {
+        /**
+         * just in case we need nanosecond granularity some day.
+         * field is unused as of now (maybe we'd need it later).
+         */
+        struct timeval tv_start;
+
+        /**
+         * version of changelog file, incremented each time changes
+         * rollover.
+         */
+        unsigned long changelog_version[CHANGELOG_MAX_TYPE];
+} changelog_time_slice_t;
+
+typedef struct changelog_rollover {
+        /* rollover thread */
+        pthread_t rollover_th;
+
+        xlator_t *this;
+} changelog_rollover_t;
+
+typedef struct changelog_fsync {
+        /* fsync() thread */
+        pthread_t fsync_th;
+
+        xlator_t *this;
+} changelog_fsync_t;
+
+# define CHANGELOG_MAX_CLIENTS  5
+typedef struct changelog_notify {
+        /* reader end of the pipe */
+        int rfd;
+
+        /* notifier thread */
+        pthread_t notify_th;
+
+        /* unique socket path */
+        char sockpath[UNIX_PATH_MAX];
+
+        int socket_fd;
+
+        /**
+         * simple array of accept()'ed fds. Not scalable at all
+         * for large number of clients, but it's okay as we have
+         * a ahrd limit in this version (@CHANGELOG_MAX_CLIENTS).
+         */
+        int client_fd[CHANGELOG_MAX_CLIENTS];
+
+        xlator_t *this;
+} changelog_notify_t;
+
+struct changelog_priv {
+        gf_boolean_t active;
+
+        /* to generate unique socket file per brick */
+        char *changelog_brick;
+
+        /* logging directory */
+        char *changelog_dir;
+
+        /* one file for all changelog types */
+        int changelog_fd;
+
+        gf_lock_t lock;
+
+        /* writen end of the pipe */
+        int wfd;
+
+        /* rollover time */
+        int32_t rollover_time;
+
+        /* fsync() interval */
+        int32_t fsync_interval;
+
+        /* changelog type maps */
+        const char *maps[CHANGELOG_MAX_TYPE];
+
+        /* time slicer */
+        changelog_time_slice_t slice;
+
+        /* context of the updater */
+        changelog_dispatcher_t cd;
+
+        /* context of the rollover thread */
+        changelog_rollover_t cr;
+
+        /* context of fsync thread */
+        changelog_fsync_t cf;
+
+        /* context of the notifier thread */
+        changelog_notify_t cn;
+
+        /* operation mode */
+        changelog_mode_t op_mode;
+
+        /* bootstrap routine for 'current' logger */
+        struct changelog_bootstrap *cb;
+
+        /* encoder mode */
+        changelog_encoder_t encode_mode;
+
+        /* encoder */
+        struct changelog_encoder *ce;
+};
+
+struct changelog_local {
+        inode_t              *inode;
+        gf_boolean_t          update_no_check;
+
+        changelog_log_data_t  cld;
+
+        /**
+         * ->prev_entry is used in cases when there needs to be
+         * additional changelog entry for the parent (eg. rename)
+         * It's analogous to ->next in single linked list world,
+         * but we call it as ->prev_entry... ha ha ha
+         */
+        struct changelog_local *prev_entry;
+};
+
+typedef struct changelog_local changelog_local_t;
+
+/* inode version is stored in inode ctx */
+typedef struct changelog_inode_ctx {
+        unsigned long iversion[CHANGELOG_MAX_TYPE];
+} changelog_inode_ctx_t;
+
+#define CHANGELOG_INODE_VERSION_TYPE(ctx, type)  &(ctx->iversion[type])
+
+/**
+ * Optional Records:
+ *  fops that need to save additional information request a array of
+ *  @changelog_opt_t struct. The array is allocated via @iobufs.
+ */
+typedef enum {
+        CHANGELOG_OPT_REC_FOP,
+        CHANGELOG_OPT_REC_ENTRY,
+        CHANGELOG_OPT_REC_UINT32,
+} changelog_optional_rec_type_t;
+
+struct changelog_entry_fields {
+        uuid_t  cef_uuid;
+        char   *cef_bname;
+};
+
+typedef struct {
+        /**
+         * @co_covert can be used to do post-processing of the record before
+         * it's persisted to the CHANGELOG. If this is NULL, then the record
+         * is persisted as per it's in memory format.
+         */
+        size_t (*co_convert) (void *data, char *buffer, gf_boolean_t encode);
+
+        /* release routines */
+        void (*co_free) (void *data);
+
+        /* type of the field */
+        changelog_optional_rec_type_t co_type;
+
+        /**
+         * sizeof of the 'valid' field in the union. This field is not used if
+         * @co_convert is specified.
+         */
+        size_t co_len;
+
+        union {
+                unsigned int                  co_uint32;
+                glusterfs_fop_t               co_fop;
+                struct changelog_entry_fields co_entry;
+        };
+} changelog_opt_t;
+
+#define CHANGELOG_OPT_RECORD_LEN  sizeof (changelog_opt_t)
+
+/**
+ * helpers routines
+ */
+
+void
+changelog_thread_cleanup (xlator_t *this, pthread_t thr_id);
+inline void *
+changelog_get_usable_buffer (changelog_local_t *local);
+inline void
+changelog_set_usable_record_and_length (changelog_local_t *local,
+                                        size_t len, int xr);
+void
+changelog_local_cleanup (xlator_t *xl, changelog_local_t *local);
+changelog_local_t *
+changelog_local_init (xlator_t *this, inode_t *inode, uuid_t gfid,
+                      int xtra_records, gf_boolean_t update_flag);
+int
+changelog_start_next_change (xlator_t *this,
+                             changelog_priv_t *priv,
+                             unsigned long ts, gf_boolean_t finale);
+int
+changelog_open (xlator_t *this, changelog_priv_t *priv);
+int
+changelog_fill_rollover_data (changelog_log_data_t *cld, gf_boolean_t is_last);
+int
+changelog_inject_single_event (xlator_t *this,
+                               changelog_priv_t *priv,
+                               changelog_log_data_t *cld);
+inline size_t
+changelog_entry_length ();
+inline int
+changelog_write (int fd, char *buffer, size_t len);
+int
+changelog_write_change (changelog_priv_t *priv, char *buffer, size_t len);
+inline int
+changelog_handle_change (xlator_t *this,
+                         changelog_priv_t *priv, changelog_log_data_t *cld);
+inline void
+changelog_update (xlator_t *this, changelog_priv_t *priv,
+                  changelog_local_t *local, changelog_log_type type);
+void *
+changelog_rollover (void *data);
+void *
+changelog_fsync_thread (void *data);
+int
+changelog_forget (xlator_t *this, inode_t *inode);
+
+/* macros */
+
+#define CHANGELOG_STACK_UNWIND(fop, frame, params ...) do {             \
+                changelog_local_t *__local = NULL;                      \
+                xlator_t          *__xl    = NULL;                      \
+                if (frame) {                                            \
+                        __local      = frame->local;                    \
+                        __xl         = frame->this;                     \
+                        frame->local = NULL;                            \
+                }                                                       \
+                STACK_UNWIND_STRICT (fop, frame, params);               \
+                changelog_local_cleanup (__xl, __local);                \
+                if (__local && __local->prev_entry)                     \
+                        changelog_local_cleanup (__xl,                  \
+                                                 __local->prev_entry);  \
+        } while (0)
+
+#define CHANGELOG_IOBUF_REF(iobuf) do {         \
+                if (iobuf)                      \
+                        iobuf_ref (iobuf);      \
+        } while (0)
+
+#define CHANGELOG_IOBUF_UNREF(iobuf) do {       \
+                if (iobuf)                      \
+                        iobuf_unref (iobuf);    \
+        } while (0)
+
+#define CHANGELOG_FILL_BUFFER(buffer, off, val, len) do {       \
+                memcpy (buffer + off, val, len);                \
+                off += len;                                     \
+        } while (0)
+
+#define SLICE_VERSION_UPDATE(slice) do {                \
+                int i = 0;                              \
+                for (; i < CHANGELOG_MAX_TYPE; i++) {   \
+                        slice->changelog_version[i]++;  \
+                }                                       \
+        } while (0)
+
+#define CHANGELOG_FILL_UINT32(co, number, converter, xlen) do { \
+                co->co_convert = converter;                     \
+                co->co_free = NULL;                             \
+                co->co_type = CHANGELOG_OPT_REC_UINT32;         \
+                co->co_uint32 = number;                         \
+                xlen += sizeof (unsigned int);                 \
+        } while (0)
+
+#define CHANGLOG_FILL_FOP_NUMBER(co, fop, converter, xlen) do { \
+                co->co_convert = converter;                     \
+                co->co_free = NULL;                             \
+                co->co_type = CHANGELOG_OPT_REC_FOP;            \
+                co->co_fop = fop;                               \
+                xlen += sizeof (fop);                           \
+        } while (0)
+
+#define CHANGELOG_FILL_ENTRY(co, pargfid, bname,                        \
+                             converter, freefn, xlen, label)            \
+        do {                                                            \
+                co->co_convert = converter;                             \
+                co->co_free = freefn;                                   \
+                co->co_type = CHANGELOG_OPT_REC_ENTRY;                  \
+                uuid_copy (co->co_entry.cef_uuid, pargfid);             \
+                co->co_entry.cef_bname = gf_strdup(bname);              \
+                if (!co->co_entry.cef_bname)                            \
+                        goto label;                                     \
+                xlen += (UUID_CANONICAL_FORM_LEN + strlen (bname));     \
+        } while (0)
+
+#define CHANGELOG_INIT(this, local, inode, gfid, xrec)                  \
+        local = changelog_local_init (this, inode, gfid, xrec, _gf_false)
+
+#define CHANGELOG_INIT_NOCHECK(this, local, inode, gfid, xrec)          \
+        local = changelog_local_init (this, inode, gfid, xrec, _gf_true)
+
+#define CHANGELOG_NOT_ACTIVE_THEN_GOTO(frame, priv, label) do { \
+                if (!priv->active)                              \
+                        goto label;                             \
+                /* ignore rebalance process's activity. */      \
+                if (frame->root->pid == GF_CLIENT_PID_DEFRAG)   \
+                        goto label;                             \
+        } while (0)
+
+/* ignore internal fops */
+#define CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO(dict, label) do {           \
+                if (dict && dict_get (dict, GLUSTERFS_INTERNAL_FOP_KEY)) \
+                        goto label;                                     \
+        } while (0)
+
+#define CHANGELOG_COND_GOTO(priv, cond, label) do {                    \
+                if (!priv->active || cond)                             \
+                        goto label;                                    \
+        } while (0)
+
+#endif /* _CHANGELOG_HELPERS_H */
diff --git a/xlators/features/changelog/src/changelog-mem-types.h b/xlators/features/changelog/src/changelog-mem-types.h
new file mode 100644
index 000000000..d72464eab
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-mem-types.h
@@ -0,0 +1,29 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CHANGELOG_MEM_TYPES_H
+#define _CHANGELOG_MEM_TYPES_H
+
+#include "mem-types.h"
+
+enum gf_changelog_mem_types {
+        gf_changelog_mt_priv_t                  = gf_common_mt_end + 1,
+        gf_changelog_mt_str_t                   = gf_common_mt_end + 2,
+        gf_changelog_mt_batch_t                 = gf_common_mt_end + 3,
+        gf_changelog_mt_rt_t                    = gf_common_mt_end + 4,
+        gf_changelog_mt_inode_ctx_t             = gf_common_mt_end + 5,
+        gf_changelog_mt_libgfchangelog_t        = gf_common_mt_end + 6,
+        gf_changelog_mt_libgfchangelog_rl_t     = gf_common_mt_end + 7,
+        gf_changelog_mt_libgfchangelog_dirent_t = gf_common_mt_end + 8,
+        gf_changelog_mt_changelog_buffer_t      = gf_common_mt_end + 9,
+        gf_changelog_mt_end
+};
+
+#endif
diff --git a/xlators/features/changelog/src/changelog-misc.h b/xlators/features/changelog/src/changelog-misc.h
new file mode 100644
index 000000000..127b03e2e
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-misc.h
@@ -0,0 +1,101 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CHANGELOG_MISC_H
+#define _CHANGELOG_MISC_H
+
+#include "glusterfs.h"
+#include "common-utils.h"
+
+#define CHANGELOG_MAX_TYPE  3
+#define CHANGELOG_FILE_NAME "CHANGELOG"
+
+#define CHANGELOG_VERSION_MAJOR  1
+#define CHANGELOG_VERSION_MINOR  1
+
+#define CHANGELOG_UNIX_SOCK  DEFAULT_VAR_RUN_DIRECTORY"/changelog-%s.sock"
+
+/**
+ * header starts with the version and the format of the changelog.
+ * 'version' not much of a use now.
+ */
+#define CHANGELOG_HEADER                                                \
+        "GlusterFS Changelog | version: v%d.%d | encoding : %d\n"
+
+#define CHANGELOG_MAKE_SOCKET_PATH(brick_path, sockpath, len) do {      \
+                char md5_sum[MD5_DIGEST_LENGTH*2+1] = {0,};             \
+                md5_wrapper((unsigned char *) brick_path,               \
+                            strlen(brick_path),                         \
+                            md5_sum);                                   \
+                (void) snprintf (sockpath, len,                         \
+                                 CHANGELOG_UNIX_SOCK, md5_sum);         \
+        } while (0)
+
+/**
+ * ... used by libgfchangelog.
+ */
+#define CHANGELOG_GET_ENCODING(fd, buffer, len, enc, enc_len) do {      \
+                FILE *fp;                                               \
+                int fd_dup, maj, min;                                   \
+                                                                        \
+                enc = -1;                                               \
+                fd_dup = dup (fd);                                      \
+                                                                        \
+                if (fd_dup != -1) {                                     \
+                        fp = fdopen (fd_dup, "r");                      \
+                        if (fp) {                                       \
+                                if (fgets (buffer, len, fp)) {          \
+                                        elen = strlen (buffer);         \
+                                        sscanf (buffer,                 \
+                                                CHANGELOG_HEADER,       \
+                                                &maj, &min, &enc);      \
+                                }                                       \
+                                fclose (fp);                            \
+                        } else {                                        \
+                                close (fd_dup);                         \
+                        }                                               \
+                }                                                       \
+        } while (0)
+
+/**
+ * everything after 'CHANGELOG_TYPE_ENTRY' are internal types
+ * (ie. none of the fops trigger this type of event), hence
+ * CHANGELOG_MAX_TYPE = 3
+ */
+typedef enum {
+        CHANGELOG_TYPE_DATA = 0,
+        CHANGELOG_TYPE_METADATA,
+        CHANGELOG_TYPE_ENTRY,
+        CHANGELOG_TYPE_ROLLOVER,
+        CHANGELOG_TYPE_FSYNC,
+} changelog_log_type;
+
+/* operation modes - RT for now */
+typedef enum {
+        CHANGELOG_MODE_RT = 0,
+} changelog_mode_t;
+
+/* encoder types */
+
+typedef enum {
+        CHANGELOG_ENCODE_MIN = 0,
+        CHANGELOG_ENCODE_BINARY,
+        CHANGELOG_ENCODE_ASCII,
+        CHANGELOG_ENCODE_MAX,
+} changelog_encoder_t;
+
+#define CHANGELOG_VALID_ENCODING(enc)                                   \
+        (enc > CHANGELOG_ENCODE_MIN && enc < CHANGELOG_ENCODE_MAX)
+
+#define CHANGELOG_TYPE_IS_ENTRY(type)  (type == CHANGELOG_TYPE_ENTRY)
+#define CHANGELOG_TYPE_IS_ROLLOVER(type)  (type == CHANGELOG_TYPE_ROLLOVER)
+#define CHANGELOG_TYPE_IS_FSYNC(type)  (type == CHANGELOG_TYPE_FSYNC)
+
+#endif /* _CHANGELOG_MISC_H */
diff --git a/xlators/features/changelog/src/changelog-notifier.c b/xlators/features/changelog/src/changelog-notifier.c
new file mode 100644
index 000000000..5f3d063a8
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-notifier.c
@@ -0,0 +1,314 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#include "changelog-notifier.h"
+
+#include <pthread.h>
+
+inline static void
+changelog_notify_clear_fd (changelog_notify_t *cn, int i)
+{
+        cn->client_fd[i] = -1;
+}
+
+inline static void
+changelog_notify_save_fd (changelog_notify_t *cn, int i, int fd)
+{
+        cn->client_fd[i] = fd;
+}
+
+static int
+changelog_notify_insert_fd (xlator_t *this, changelog_notify_t *cn, int fd)
+{
+        int i   = 0;
+        int ret = 0;
+
+        for (; i < CHANGELOG_MAX_CLIENTS; i++) {
+                if (cn->client_fd[i] == -1)
+                        break;
+        }
+
+        if (i == CHANGELOG_MAX_CLIENTS) {
+                /**
+                 * this case should not be hit as listen() would limit
+                 * the number of completely established connections.
+                 */
+                gf_log (this->name, GF_LOG_WARNING,
+                        "hit max client limit (%d)", CHANGELOG_MAX_CLIENTS);
+                ret = -1;
+        }
+        else
+                changelog_notify_save_fd (cn, i, fd);
+
+        return ret;
+}
+
+static void
+changelog_notify_fill_rset (changelog_notify_t *cn, fd_set *rset, int *maxfd)
+{
+        int i = 0;
+
+        FD_ZERO (rset);
+
+        FD_SET (cn->socket_fd, rset);
+        *maxfd = cn->socket_fd;
+
+        FD_SET (cn->rfd, rset);
+        *maxfd = max (*maxfd, cn->rfd);
+
+        for (; i < CHANGELOG_MAX_CLIENTS; i++) {
+                if (cn->client_fd[i] != -1) {
+                        FD_SET (cn->client_fd[i], rset);
+                        *maxfd = max (*maxfd, cn->client_fd[i]);
+                }
+        }
+
+        *maxfd = *maxfd + 1;
+}
+
+static int
+changelog_notify_client (changelog_notify_t *cn, char *path, ssize_t len)
+{
+        int i = 0;
+        int ret = 0;
+
+        for (; i < CHANGELOG_MAX_CLIENTS; i++) {
+                if (cn->client_fd[i] == -1)
+                        continue;
+
+                if (changelog_write (cn->client_fd[i],
+                                     path, len)) {
+                        ret = -1;
+
+                        close (cn->client_fd[i]);
+                        changelog_notify_clear_fd (cn, i);
+                }
+        }
+
+        return ret;
+}
+
+static void
+changelog_notifier_init (changelog_notify_t *cn)
+{
+        int i = 0;
+
+        cn->socket_fd = -1;
+
+        for (; i < CHANGELOG_MAX_CLIENTS; i++) {
+                changelog_notify_clear_fd (cn, i);
+        }
+}
+
+static void
+changelog_close_client_conn (changelog_notify_t *cn)
+{
+        int i = 0;
+
+        for (; i < CHANGELOG_MAX_CLIENTS; i++) {
+                if (cn->client_fd[i] == -1)
+                        continue;
+
+                close (cn->client_fd[i]);
+                changelog_notify_clear_fd (cn, i);
+        }
+}
+
+static void
+changelog_notifier_cleanup (void *arg)
+{
+        changelog_notify_t *cn = NULL;
+
+        cn = (changelog_notify_t *) arg;
+
+        changelog_close_client_conn (cn);
+
+        if (cn->socket_fd != -1)
+                close (cn->socket_fd);
+
+        if (cn->rfd)
+                close (cn->rfd);
+
+        if (unlink (cn->sockpath))
+                gf_log ("", GF_LOG_WARNING,
+                        "could not unlink changelog socket file"
+                        " %s (reason: %s", cn->sockpath, strerror (errno));
+}
+
+void *
+changelog_notifier (void *data)
+{
+        int                 i         = 0;
+        int                 fd        = 0;
+        int                 max_fd    = 0;
+        int                 len       = 0;
+        ssize_t             readlen   = 0;
+        xlator_t           *this      = NULL;
+        changelog_priv_t   *priv      = NULL;
+        changelog_notify_t *cn        = NULL;
+        struct sockaddr_un  local     = {0,};
+        char path[PATH_MAX]           = {0,};
+        char abspath[PATH_MAX]        = {0,};
+
+        char buffer;
+        fd_set rset;
+
+        priv = (changelog_priv_t *) data;
+
+        cn = &priv->cn;
+        this = cn->this;
+
+        pthread_cleanup_push (changelog_notifier_cleanup, cn);
+
+        changelog_notifier_init (cn);
+
+        cn->socket_fd = socket (AF_UNIX, SOCK_STREAM, 0);
+        if (cn->socket_fd < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "changelog socket error (reason: %s)",
+                        strerror (errno));
+                goto out;
+        }
+
+        CHANGELOG_MAKE_SOCKET_PATH (priv->changelog_brick,
+                                    cn->sockpath, UNIX_PATH_MAX);
+        if (unlink (cn->sockpath) < 0) {
+                if (errno != ENOENT) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Could not unlink changelog socket file (%s)"
+                                " (reason: %s)",
+                                CHANGELOG_UNIX_SOCK, strerror (errno));
+                        goto cleanup;
+                }
+        }
+
+        local.sun_family = AF_UNIX;
+        strcpy (local.sun_path, cn->sockpath);
+
+        len = strlen (local.sun_path) + sizeof (local.sun_family);
+
+        /* bind to the unix domain socket */
+        if (bind (cn->socket_fd, (struct sockaddr *) &local, len) < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Could not bind to changelog socket (reason: %s)",
+                        strerror (errno));
+                goto cleanup;
+        }
+
+        /* listen for incoming connections */
+        if (listen (cn->socket_fd, CHANGELOG_MAX_CLIENTS) < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "listen() error on changelog socket (reason: %s)",
+                        strerror (errno));
+                goto cleanup;
+        }
+
+        /**
+         * simple select() on all to-be-read file descriptors. This method
+         * though old school works pretty well when you have a handfull of
+         * fd's to be watched (clients).
+         *
+         * Future TODO: move this to epoll based notification facility if
+         *              number of clients increase.
+         */
+        for (;;) {
+                changelog_notify_fill_rset (cn, &rset, &max_fd);
+
+                if (select (max_fd, &rset, NULL, NULL, NULL) < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "select() returned -1 (reason: %s)",
+                                strerror (errno));
+                        sleep (2);
+                        continue;
+                }
+
+                if (FD_ISSET (cn->socket_fd, &rset)) {
+                        fd = accept (cn->socket_fd, NULL, NULL);
+                        if (fd < 0) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "accept error on changelog socket"
+                                        " (reason: %s)", strerror (errno));
+                        } else if (changelog_notify_insert_fd (this, cn, fd)) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "hit max client limit");
+                        }
+                }
+
+                if (FD_ISSET (cn->rfd, &rset)) {
+                        /**
+                         * read changelog filename and notify all connected
+                         * clients.
+                         */
+                        readlen = 0;
+                        while (readlen < PATH_MAX) {
+                                len = read (cn->rfd, &path[readlen++], 1);
+                                if (len == -1) {
+                                        break;
+                                }
+
+                                if (len == 0) {
+                                        gf_log (this->name, GF_LOG_ERROR,
+                                                "rollover thread sent EOF"
+                                                " on pipe - possibly a crash.");
+                                        /* be blunt and close all connections */
+                                        pthread_exit(NULL);
+                                }
+
+                                if (path[readlen - 1] == '\0')
+                                        break;
+                        }
+
+                        /* should we close all client connections here too? */
+                        if (len < 0 || readlen == PATH_MAX) {
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "Could not get pathname from rollover"
+                                        " thread or pathname too long");
+                                goto process_rest;
+                        }
+
+                        (void) snprintf (abspath, PATH_MAX,
+                                         "%s/%s", priv->changelog_dir, path);
+                        if (changelog_notify_client (cn, abspath,
+                                                     strlen (abspath) + 1))
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "could not notify some clients with new"
+                                        " changelogs");
+                }
+
+        process_rest:
+                for (i = 0; i < CHANGELOG_MAX_CLIENTS; i++) {
+                        if ( (fd = cn->client_fd[i]) == -1 )
+                                continue;
+
+                        if (FD_ISSET (fd, &rset)) {
+                                /**
+                                 * the only data we accept from the client is a
+                                 * disconnect. Anything else is treated as bogus
+                                 * and is silently discarded (also warned!!!).
+                                 */
+                                if ( (readlen = read (fd, &buffer, 1)) <= 0 ) {
+                                        close (fd);
+                                        changelog_notify_clear_fd (cn, i);
+                                } else {
+                                        /* silently discard data and log */
+                                        gf_log (this->name, GF_LOG_WARNING,
+                                                "misbehaving changelog client");
+                                }
+                        }
+                }
+
+        }
+
+ cleanup:;
+        pthread_cleanup_pop (1);
+
+ out:
+        return NULL;
+}
diff --git a/xlators/features/changelog/src/changelog-notifier.h b/xlators/features/changelog/src/changelog-notifier.h
new file mode 100644
index 000000000..55e728356
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-notifier.h
@@ -0,0 +1,19 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CHANGELOG_NOTIFIER_H
+#define _CHANGELOG_NOTIFIER_H
+
+#include "changelog-helpers.h"
+
+void *
+changelog_notifier (void *data);
+
+#endif
diff --git a/xlators/features/changelog/src/changelog-rt.c b/xlators/features/changelog/src/changelog-rt.c
new file mode 100644
index 000000000..c147f68ca
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-rt.c
@@ -0,0 +1,72 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+#include "logging.h"
+
+#include "changelog-rt.h"
+#include "changelog-mem-types.h"
+
+int
+changelog_rt_init (xlator_t *this, changelog_dispatcher_t *cd)
+{
+        changelog_rt_t *crt = NULL;
+
+        crt = GF_CALLOC (1, sizeof (*crt),
+                         gf_changelog_mt_rt_t);
+        if (!crt)
+                return -1;
+
+        LOCK_INIT (&crt->lock);
+
+        cd->cd_data = crt;
+        cd->dispatchfn = &changelog_rt_enqueue;
+
+        return 0;
+}
+
+int
+changelog_rt_fini (xlator_t *this, changelog_dispatcher_t *cd)
+{
+        changelog_rt_t *crt = NULL;
+
+        crt = cd->cd_data;
+
+        LOCK_DESTROY (&crt->lock);
+        GF_FREE (crt);
+
+        return 0;
+}
+
+int
+changelog_rt_enqueue (xlator_t *this, changelog_priv_t *priv, void *cbatch,
+                      changelog_log_data_t *cld_0, changelog_log_data_t *cld_1)
+{
+        int             ret = 0;
+        changelog_rt_t *crt = NULL;
+
+        crt = (changelog_rt_t *) cbatch;
+
+        LOCK (&crt->lock);
+        {
+                ret = changelog_handle_change (this, priv, cld_0);
+                if (!ret && cld_1)
+                        ret = changelog_handle_change (this, priv, cld_1);
+        }
+        UNLOCK (&crt->lock);
+
+        return ret;
+}
diff --git a/xlators/features/changelog/src/changelog-rt.h b/xlators/features/changelog/src/changelog-rt.h
new file mode 100644
index 000000000..1fc2bbc5b
--- /dev/null
+++ b/xlators/features/changelog/src/changelog-rt.h
@@ -0,0 +1,33 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CHANGELOG_RT_H
+#define _CHANGELOG_RT_H
+
+#include "locking.h"
+#include "timer.h"
+#include "pthread.h"
+
+#include "changelog-helpers.h"
+
+/* unused as of now - may be you would need it later */
+typedef struct changelog_rt {
+        gf_lock_t lock;
+} changelog_rt_t;
+
+int
+changelog_rt_init (xlator_t *this, changelog_dispatcher_t *cd);
+int
+changelog_rt_fini (xlator_t *this, changelog_dispatcher_t *cd);
+int
+changelog_rt_enqueue (xlator_t *this, changelog_priv_t *priv, void *cbatch,
+                      changelog_log_data_t *cld_0, changelog_log_data_t *cld_1);
+
+#endif /* _CHANGELOG_RT_H */
diff --git a/xlators/features/changelog/src/changelog.c b/xlators/features/changelog/src/changelog.c
new file mode 100644
index 000000000..5fe3b4362
--- /dev/null
+++ b/xlators/features/changelog/src/changelog.c
@@ -0,0 +1,1568 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+#include "logging.h"
+#include "iobuf.h"
+
+#include "changelog-rt.h"
+
+#include "changelog-encoders.h"
+#include "changelog-mem-types.h"
+
+#include <pthread.h>
+
+#include "changelog-notifier.h"
+
+static struct changelog_bootstrap
+cb_bootstrap[] = {
+        {
+                .mode = CHANGELOG_MODE_RT,
+                .ctor = changelog_rt_init,
+                .dtor = changelog_rt_fini,
+        },
+};
+
+/* Entry operations - TYPE III */
+
+/**
+ * entry operations do not undergo inode version checking.
+ */
+
+/* {{{ */
+
+/* rmdir */
+
+int32_t
+changelog_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                     struct iatt *postparent, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (rmdir, frame, op_ret, op_errno,
+                                preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+changelog_rmdir (call_frame_t *frame, xlator_t *this,
+                 loc_t *loc, int xflags, dict_t *xdata)
+{
+        size_t            xtra_len = 0;
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT_NOCHECK (this, frame->local,
+                                NULL, loc->inode->gfid, 2);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        co++;
+        CHANGELOG_FILL_ENTRY (co, loc->pargfid, loc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 2);
+
+ wind:
+        STACK_WIND (frame, changelog_rmdir_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->rmdir,
+                    loc, xflags, xdata);
+        return 0;
+}
+
+/* unlink */
+
+int32_t
+changelog_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                      struct iatt *postparent, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (unlink, frame, op_ret, op_errno,
+                                preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+changelog_unlink (call_frame_t *frame, xlator_t *this,
+                  loc_t *loc, int xflags, dict_t *xdata)
+{
+        size_t            xtra_len = 0;
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+        CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO (xdata, wind);
+
+        CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, loc->inode->gfid, 2);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        co++;
+        CHANGELOG_FILL_ENTRY (co, loc->pargfid, loc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 2);
+
+ wind:
+        STACK_WIND (frame, changelog_unlink_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->unlink,
+                    loc, xflags, xdata);
+        return 0;
+}
+
+/* rename */
+
+int32_t
+changelog_rename_cbk (call_frame_t *frame,
+                      void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno,
+                      struct iatt *buf, struct iatt *preoldparent,
+                      struct iatt *postoldparent, struct iatt *prenewparent,
+                      struct iatt *postnewparent, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (rename, frame, op_ret, op_errno,
+                                buf, preoldparent, postoldparent,
+                                prenewparent, postnewparent, xdata);
+        return 0;
+}
+
+
+int32_t
+changelog_rename (call_frame_t *frame, xlator_t *this,
+                  loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+{
+        size_t            xtra_len  = 0;
+        changelog_priv_t *priv      = NULL;
+        changelog_opt_t  *co        = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        /* 3 == fop + oldloc + newloc */
+        CHANGELOG_INIT_NOCHECK (this, frame->local,
+                                NULL, oldloc->inode->gfid, 3);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        co++;
+        CHANGELOG_FILL_ENTRY (co, oldloc->pargfid, oldloc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        co++;
+        CHANGELOG_FILL_ENTRY (co, newloc->pargfid, newloc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 3);
+
+ wind:
+        STACK_WIND (frame, changelog_rename_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->rename,
+                    oldloc, newloc, xdata);
+        return 0;
+}
+
+/* link */
+
+int32_t
+changelog_link_cbk (call_frame_t *frame,
+                    void *cookie, xlator_t *this, int32_t op_ret,
+                    int32_t op_errno, inode_t *inode,
+                    struct iatt *buf, struct iatt *preparent,
+                    struct iatt *postparent, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (link, frame, op_ret, op_errno,
+                                inode, buf, preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+changelog_link (call_frame_t *frame,
+                xlator_t *this, loc_t *oldloc,
+                loc_t *newloc, dict_t *xdata)
+{
+        size_t            xtra_len = 0;
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+
+        priv = this->private;
+
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+        CHANGELOG_IF_INTERNAL_FOP_THEN_GOTO (xdata, wind);
+
+        CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, oldloc->gfid, 2);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        co++;
+        CHANGELOG_FILL_ENTRY (co, newloc->pargfid, newloc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 2);
+
+ wind:
+        STACK_WIND (frame, changelog_link_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->link,
+                    oldloc, newloc, xdata);
+        return 0;
+}
+
+/* mkdir */
+
+int32_t
+changelog_mkdir_cbk (call_frame_t *frame,
+                     void *cookie, xlator_t *this, int32_t op_ret,
+                     int32_t op_errno, inode_t *inode,
+                     struct iatt *buf, struct iatt *preparent,
+                     struct iatt *postparent, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (mkdir, frame, op_ret, op_errno,
+                                inode, buf, preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+changelog_mkdir (call_frame_t *frame, xlator_t *this,
+                 loc_t *loc, mode_t mode, mode_t umask, dict_t *xdata)
+{
+        int               ret      = -1;
+        uuid_t            gfid     = {0,};
+        void             *uuid_req = NULL;
+        size_t            xtra_len = 0;
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        ret = dict_get_ptr (xdata, "gfid-req", &uuid_req);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "failed to get gfid from dict");
+                goto wind;
+        }
+        uuid_copy (gfid, uuid_req);
+
+        CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, gfid, 5);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+        co++;
+
+        CHANGELOG_FILL_UINT32 (co, S_IFDIR | mode, number_fn, xtra_len);
+        co++;
+
+        CHANGELOG_FILL_UINT32 (co, frame->root->uid, number_fn, xtra_len);
+        co++;
+
+        CHANGELOG_FILL_UINT32 (co, frame->root->gid, number_fn, xtra_len);
+        co++;
+
+        CHANGELOG_FILL_ENTRY (co, loc->pargfid, loc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 5);
+
+ wind:
+        STACK_WIND (frame, changelog_mkdir_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->mkdir,
+                    loc, mode, umask, xdata);
+        return 0;
+}
+
+/* symlink */
+
+int32_t
+changelog_symlink_cbk (call_frame_t *frame,
+                       void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno,
+                       inode_t *inode, struct iatt *buf, struct iatt *preparent,
+                       struct iatt *postparent, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (symlink, frame, op_ret, op_errno,
+                                inode, buf, preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+changelog_symlink (call_frame_t *frame, xlator_t *this,
+                   const char *linkname, loc_t *loc,
+                   mode_t umask, dict_t *xdata)
+{
+        int               ret      = -1;
+        size_t            xtra_len = 0;
+        uuid_t            gfid     = {0,};
+        void             *uuid_req = NULL;
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        ret = dict_get_ptr (xdata, "gfid-req", &uuid_req);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "failed to get gfid from dict");
+                goto wind;
+        }
+        uuid_copy (gfid, uuid_req);
+
+        CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, gfid, 2);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+        co++;
+
+        CHANGELOG_FILL_ENTRY (co, loc->pargfid, loc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 2);
+
+ wind:
+        STACK_WIND (frame, changelog_symlink_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->symlink,
+                    linkname, loc, umask, xdata);
+        return 0;
+}
+
+/* mknod */
+
+int32_t
+changelog_mknod_cbk (call_frame_t *frame,
+                     void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, inode_t *inode,
+                     struct iatt *buf, struct iatt *preparent,
+                     struct iatt *postparent, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (mknod, frame, op_ret, op_errno,
+                                inode, buf, preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+changelog_mknod (call_frame_t *frame,
+                 xlator_t *this, loc_t *loc,
+                 mode_t mode, dev_t dev, mode_t umask, dict_t *xdata)
+{
+        int               ret      = -1;
+        uuid_t            gfid     = {0,};
+        void             *uuid_req = NULL;
+        size_t            xtra_len = 0;
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        ret = dict_get_ptr (xdata, "gfid-req", &uuid_req);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "failed to get gfid from dict");
+                goto wind;
+        }
+        uuid_copy (gfid, uuid_req);
+
+        CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, gfid, 5);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+        co++;
+
+        CHANGELOG_FILL_UINT32 (co, mode, number_fn, xtra_len);
+        co++;
+
+        CHANGELOG_FILL_UINT32 (co, frame->root->uid, number_fn, xtra_len);
+        co++;
+
+        CHANGELOG_FILL_UINT32 (co, frame->root->gid, number_fn, xtra_len);
+        co++;
+
+        CHANGELOG_FILL_ENTRY (co, loc->pargfid, loc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 5);
+
+ wind:
+        STACK_WIND (frame, changelog_mknod_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->mknod,
+                    loc, mode, dev, umask, xdata);
+        return 0;
+}
+
+/* creat */
+
+int32_t
+changelog_create_cbk (call_frame_t *frame,
+                      void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno,
+                      fd_t *fd, inode_t *inode, struct iatt *buf,
+                      struct iatt *preparent,
+                      struct iatt *postparent, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_ENTRY);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (create, frame,
+                                op_ret, op_errno, fd, inode,
+                                buf, preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+changelog_create (call_frame_t *frame, xlator_t *this,
+                  loc_t *loc, int32_t flags, mode_t mode,
+                  mode_t umask, fd_t *fd, dict_t *xdata)
+{
+        int               ret      = -1;
+        uuid_t            gfid     = {0,};
+        void             *uuid_req = NULL;
+        changelog_opt_t  *co       = NULL;
+        changelog_priv_t *priv     = NULL;
+        size_t            xtra_len = 0;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        ret = dict_get_ptr (xdata, "gfid-req", &uuid_req);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "failed to get gfid from dict");
+                goto wind;
+        }
+        uuid_copy (gfid, uuid_req);
+
+        /* init with two extra records */
+        CHANGELOG_INIT_NOCHECK (this, frame->local, NULL, gfid, 5);
+        if (!frame->local)
+                goto wind;
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+        co++;
+
+        CHANGELOG_FILL_UINT32 (co, mode, number_fn, xtra_len);
+        co++;
+
+        CHANGELOG_FILL_UINT32 (co, frame->root->uid, number_fn, xtra_len);
+        co++;
+
+        CHANGELOG_FILL_UINT32 (co, frame->root->gid, number_fn, xtra_len);
+        co++;
+
+        CHANGELOG_FILL_ENTRY (co, loc->pargfid, loc->name,
+                              entry_fn, entry_free_fn, xtra_len, wind);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 5);
+
+ wind:
+        STACK_WIND (frame, changelog_create_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->create,
+                    loc, flags, mode, umask, fd, xdata);
+        return 0;
+}
+
+/* }}} */
+
+
+/* Metadata modification fops - TYPE II */
+
+/* {{{ */
+
+/* {f}setattr */
+
+int32_t
+changelog_fsetattr_cbk (call_frame_t *frame,
+                        void *cookie, xlator_t *this, int32_t op_ret,
+                        int32_t op_errno, struct iatt *preop_stbuf,
+                        struct iatt *postop_stbuf, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (fsetattr, frame, op_ret, op_errno,
+                                preop_stbuf, postop_stbuf, xdata);
+
+        return 0;
+
+
+}
+
+int32_t
+changelog_fsetattr (call_frame_t *frame,
+                    xlator_t *this, fd_t *fd,
+                    struct iatt *stbuf, int32_t valid, dict_t *xdata)
+{
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+        size_t            xtra_len = 0;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        fd->inode, fd->inode->gfid, 1);
+        if (!frame->local)
+                goto wind;
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 1);
+
+ wind:
+        STACK_WIND (frame, changelog_fsetattr_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->fsetattr,
+                    fd, stbuf, valid, xdata);
+        return 0;
+
+
+}
+
+int32_t
+changelog_setattr_cbk (call_frame_t *frame,
+                       void *cookie, xlator_t *this, int32_t op_ret,
+                       int32_t op_errno, struct iatt *preop_stbuf,
+                       struct iatt *postop_stbuf, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (setattr, frame, op_ret, op_errno,
+                                preop_stbuf, postop_stbuf, xdata);
+
+        return 0;
+}
+
+int32_t
+changelog_setattr (call_frame_t *frame,
+                   xlator_t *this, loc_t *loc,
+                   struct iatt *stbuf, int32_t valid, dict_t *xdata)
+{
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+        size_t            xtra_len = 0;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        loc->inode, loc->inode->gfid, 1);
+        if (!frame->local)
+                goto wind;
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 1);
+
+ wind:
+        STACK_WIND (frame, changelog_setattr_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->setattr,
+                    loc, stbuf, valid, xdata);
+        return 0;
+}
+
+/* {f}removexattr */
+
+int32_t
+changelog_fremovexattr_cbk (call_frame_t *frame,
+                            void *cookie, xlator_t *this,
+                            int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (fremovexattr, frame, op_ret, op_errno, xdata);
+
+        return 0;
+}
+
+int32_t
+changelog_fremovexattr (call_frame_t *frame, xlator_t *this,
+                        fd_t *fd, const char *name, dict_t *xdata)
+{
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+        size_t            xtra_len = 0;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        fd->inode, fd->inode->gfid, 1);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 1);
+
+ wind:
+        STACK_WIND (frame, changelog_fremovexattr_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->fremovexattr,
+                    fd, name, xdata);
+        return 0;
+}
+
+int32_t
+changelog_removexattr_cbk (call_frame_t *frame,
+                           void *cookie, xlator_t *this,
+                           int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (removexattr, frame, op_ret, op_errno, xdata);
+
+        return 0;
+}
+
+int32_t
+changelog_removexattr (call_frame_t *frame, xlator_t *this,
+                       loc_t *loc, const char *name, dict_t *xdata)
+{
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+        size_t            xtra_len = 0;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        loc->inode, loc->inode->gfid, 1);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 1);
+
+ wind:
+        STACK_WIND (frame, changelog_removexattr_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->removexattr,
+                    loc, name, xdata);
+        return 0;
+}
+
+/* {f}setxattr */
+
+int32_t
+changelog_setxattr_cbk (call_frame_t *frame,
+                        void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (setxattr, frame, op_ret, op_errno, xdata);
+
+        return 0;
+}
+
+int32_t
+changelog_setxattr (call_frame_t *frame,
+                    xlator_t *this, loc_t *loc,
+                    dict_t *dict, int32_t flags, dict_t *xdata)
+{
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+        size_t            xtra_len = 0;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        loc->inode, loc->inode->gfid, 1);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 1);
+
+ wind:
+        STACK_WIND (frame, changelog_setxattr_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->setxattr,
+                    loc, dict, flags, xdata);
+        return 0;
+}
+
+int32_t
+changelog_fsetxattr_cbk (call_frame_t *frame,
+                         void *cookie, xlator_t *this, int32_t op_ret,
+                         int32_t op_errno, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_METADATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (fsetxattr, frame, op_ret, op_errno, xdata);
+
+        return 0;
+}
+
+int32_t
+changelog_fsetxattr (call_frame_t *frame,
+                     xlator_t *this, fd_t *fd, dict_t *dict,
+                     int32_t flags, dict_t *xdata)
+{
+        changelog_priv_t *priv     = NULL;
+        changelog_opt_t  *co       = NULL;
+        size_t            xtra_len = 0;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        fd->inode, fd->inode->gfid, 1);
+
+        co = changelog_get_usable_buffer (frame->local);
+        if (!co)
+                goto wind;
+
+        CHANGLOG_FILL_FOP_NUMBER (co, frame->root->op, fop_fn, xtra_len);
+
+        changelog_set_usable_record_and_length (frame->local, xtra_len, 1);
+
+ wind:
+        STACK_WIND (frame, changelog_fsetxattr_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->fsetxattr,
+                    fd, dict, flags, xdata);
+        return 0;
+}
+
+/* }}} */
+
+
+/* Data modification fops - TYPE I */
+
+/* {{{ */
+
+/* {f}truncate() */
+
+int32_t
+changelog_truncate_cbk (call_frame_t *frame,
+                        void *cookie, xlator_t *this, int32_t op_ret,
+                        int32_t op_errno, struct iatt *prebuf,
+                        struct iatt *postbuf, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_DATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (truncate, frame,
+                                op_ret, op_errno, prebuf, postbuf, xdata);
+        return 0;
+}
+
+int32_t
+changelog_truncate (call_frame_t *frame,
+                    xlator_t *this, loc_t *loc, off_t offset, dict_t *xdata)
+{
+        changelog_priv_t *priv = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        loc->inode, loc->inode->gfid, 0);
+
+ wind:
+        STACK_WIND (frame, changelog_truncate_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->truncate,
+                    loc, offset, xdata);
+        return 0;
+}
+
+int32_t
+changelog_ftruncate_cbk (call_frame_t *frame,
+                         void *cookie, xlator_t *this, int32_t op_ret,
+                         int32_t op_errno, struct iatt *prebuf,
+                         struct iatt *postbuf, dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret < 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_DATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (ftruncate, frame,
+                                op_ret, op_errno, prebuf, postbuf, xdata);
+        return 0;
+}
+
+int32_t
+changelog_ftruncate (call_frame_t *frame,
+                     xlator_t *this, fd_t *fd, off_t offset, dict_t *xdata)
+{
+        changelog_priv_t *priv = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        fd->inode, fd->inode->gfid, 0);
+
+ wind:
+        STACK_WIND (frame, changelog_ftruncate_cbk,
+                    FIRST_CHILD (this), FIRST_CHILD (this)->fops->ftruncate,
+                    fd, offset, xdata);
+        return 0;
+}
+
+/* writev() */
+
+int32_t
+changelog_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                    struct iatt *postbuf,
+                    dict_t *xdata)
+{
+        changelog_priv_t  *priv  = NULL;
+        changelog_local_t *local = NULL;
+
+        priv  = this->private;
+        local = frame->local;
+
+        CHANGELOG_COND_GOTO (priv, ((op_ret <= 0) || !local), unwind);
+
+        changelog_update (this, priv, local, CHANGELOG_TYPE_DATA);
+
+ unwind:
+        CHANGELOG_STACK_UNWIND (writev, frame,
+                                op_ret, op_errno, prebuf, postbuf, xdata);
+        return 0;
+}
+
+int32_t
+changelog_writev (call_frame_t *frame,
+                  xlator_t *this, fd_t *fd, struct iovec *vector,
+                  int32_t count, off_t offset, uint32_t flags,
+                  struct iobref *iobref, dict_t *xdata)
+{
+        changelog_priv_t *priv = NULL;
+
+        priv = this->private;
+        CHANGELOG_NOT_ACTIVE_THEN_GOTO (frame, priv, wind);
+
+        CHANGELOG_INIT (this, frame->local,
+                        fd->inode, fd->inode->gfid, 0);
+
+ wind:
+        STACK_WIND (frame, changelog_writev_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->writev, fd, vector,
+                    count, offset, flags, iobref, xdata);
+        return 0;
+}
+
+/* }}} */
+
+/**
+ * The
+ *   - @init ()
+ *   - @fini ()
+ *   - @reconfigure ()
+ *   ... and helper routines
+ */
+
+/**
+ * needed if there are more operation modes in the future.
+ */
+static void
+changelog_assign_opmode (changelog_priv_t *priv, char *mode)
+{
+        if ( strncmp (mode, "realtime", 8) == 0 ) {
+                priv->op_mode = CHANGELOG_MODE_RT;
+        }
+}
+
+static void
+changelog_assign_encoding (changelog_priv_t *priv, char *enc)
+{
+        if ( strncmp (enc, "binary", 6) == 0 ) {
+                priv->encode_mode = CHANGELOG_ENCODE_BINARY;
+        } else if ( strncmp (enc, "ascii", 5) == 0 ) {
+                priv->encode_mode = CHANGELOG_ENCODE_ASCII;
+        }
+}
+
+/* cleanup any helper threads that are running */
+static void
+changelog_cleanup_helper_threads (xlator_t *this, changelog_priv_t *priv)
+{
+        if (priv->cr.rollover_th) {
+                changelog_thread_cleanup (this, priv->cr.rollover_th);
+                priv->cr.rollover_th = 0;
+        }
+
+        if (priv->cf.fsync_th) {
+                changelog_thread_cleanup (this, priv->cf.fsync_th);
+                priv->cf.fsync_th = 0;
+        }
+}
+
+/* spawn helper thread; cleaning up in case of errors */
+static int
+changelog_spawn_helper_threads (xlator_t *this, changelog_priv_t *priv)
+{
+        int ret = 0;
+
+        priv->cr.this = this;
+        ret = gf_thread_create (&priv->cr.rollover_th,
+				NULL, changelog_rollover, priv);
+        if (ret)
+                goto out;
+
+        if (priv->fsync_interval) {
+                priv->cf.this = this;
+                ret = gf_thread_create (&priv->cf.fsync_th,
+					NULL, changelog_fsync_thread, priv);
+        }
+
+        if (ret)
+                changelog_cleanup_helper_threads (this, priv);
+
+ out:
+        return ret;
+}
+
+/* cleanup the notifier thread */
+static int
+changelog_cleanup_notifier (xlator_t *this, changelog_priv_t *priv)
+{
+        int ret = 0;
+
+        if (priv->cn.notify_th) {
+                changelog_thread_cleanup (this, priv->cn.notify_th);
+                priv->cn.notify_th = 0;
+
+                ret = close (priv->wfd);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "error closing writer end of notifier pipe"
+                                " (reason: %s)", strerror (errno));
+        }
+
+        return ret;
+}
+
+/* spawn the notifier thread - nop if already running */
+static int
+changelog_spawn_notifier (xlator_t *this, changelog_priv_t *priv)
+{
+        int ret        = 0;
+        int flags      = 0;
+        int pipe_fd[2] = {0, 0};
+
+        if (priv->cn.notify_th)
+                goto out; /* notifier thread already running */
+
+        ret = pipe (pipe_fd);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Cannot create pipe (reason: %s)", strerror (errno));
+                goto out;
+        }
+
+        /* writer is non-blocking */
+        flags = fcntl (pipe_fd[1], F_GETFL);
+        flags |= O_NONBLOCK;
+
+        ret = fcntl (pipe_fd[1], F_SETFL, flags);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to set O_NONBLOCK flag");
+                goto out;
+        }
+
+        priv->wfd = pipe_fd[1];
+
+        priv->cn.this = this;
+        priv->cn.rfd  = pipe_fd[0];
+
+        ret = gf_thread_create (&priv->cn.notify_th,
+				NULL, changelog_notifier, priv);
+
+ out:
+        return ret;
+}
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_changelog_mt_end + 1);
+
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_WARNING, "Memory accounting"
+                        " init failed");
+                return ret;
+        }
+
+        return ret;
+}
+
+static int
+changelog_init (xlator_t *this, changelog_priv_t *priv)
+{
+        int                  i   = 0;
+        int                  ret = -1;
+        struct timeval       tv  = {0,};
+        changelog_log_data_t cld = {0,};
+
+        ret = gettimeofday (&tv, NULL);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "gettimeofday() failure");
+                goto out;
+        }
+
+        priv->slice.tv_start = tv;
+
+        priv->maps[CHANGELOG_TYPE_DATA]     = "D ";
+        priv->maps[CHANGELOG_TYPE_METADATA] = "M ";
+        priv->maps[CHANGELOG_TYPE_ENTRY]    = "E ";
+
+        for (; i < CHANGELOG_MAX_TYPE; i++) {
+                /* start with version 1 */
+                priv->slice.changelog_version[i] = 1;
+        }
+
+        if (!priv->active)
+                return ret;
+
+        /* spawn the notifier thread */
+        ret = changelog_spawn_notifier (this, priv);
+        if (ret)
+                goto out;
+
+        /**
+         * start with a fresh changelog file every time. this is done
+         * in case there was an encoding change. so... things are kept
+         * simple here.
+         */
+        ret = changelog_fill_rollover_data (&cld, _gf_false);
+        if (ret)
+                goto out;
+
+        LOCK (&priv->lock);
+        {
+                ret = changelog_inject_single_event (this, priv, &cld);
+        }
+        UNLOCK (&priv->lock);
+
+        /* ... and finally spawn the helpers threads */
+        ret = changelog_spawn_helper_threads (this, priv);
+
+ out:
+        return ret;
+}
+
+int
+reconfigure (xlator_t *this, dict_t *options)
+{
+        int                     ret            = 0;
+        char                   *tmp            = NULL;
+        changelog_priv_t       *priv           = NULL;
+        gf_boolean_t            active_earlier = _gf_true;
+        gf_boolean_t            active_now     = _gf_true;
+        changelog_time_slice_t *slice          = NULL;
+        changelog_log_data_t    cld            = {0,};
+
+        priv = this->private;
+        if (!priv)
+                goto out;
+
+        ret = -1;
+        active_earlier = priv->active;
+
+        /* first stop the rollover and the fsync thread */
+        changelog_cleanup_helper_threads (this, priv);
+
+        GF_OPTION_RECONF ("changelog-dir", tmp, options, str, out);
+        if (!tmp) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "\"changelog-dir\" option is not set");
+                goto out;
+        }
+
+        GF_FREE (priv->changelog_dir);
+        priv->changelog_dir = gf_strdup (tmp);
+        if (!priv->changelog_dir)
+                goto out;
+
+        ret = mkdir_p (priv->changelog_dir, 0600, _gf_true);
+        if (ret)
+                goto out;
+
+        GF_OPTION_RECONF ("changelog", active_now, options, bool, out);
+
+        /**
+         * changelog_handle_change() handles changes that could possibly
+         * have been submit changes before changelog deactivation.
+         */
+        if (!active_now)
+                priv->active = _gf_false;
+
+        GF_OPTION_RECONF ("op-mode", tmp, options, str, out);
+        changelog_assign_opmode (priv, tmp);
+
+        tmp = NULL;
+
+        GF_OPTION_RECONF ("encoding", tmp, options, str, out);
+        changelog_assign_encoding (priv, tmp);
+
+        GF_OPTION_RECONF ("rollover-time",
+                          priv->rollover_time, options, int32, out);
+        GF_OPTION_RECONF ("fsync-interval",
+                          priv->fsync_interval, options, int32, out);
+
+        if (active_now || active_earlier) {
+                ret = changelog_fill_rollover_data (&cld, !active_now);
+                if (ret)
+                        goto out;
+
+                slice = &priv->slice;
+
+                LOCK (&priv->lock);
+                {
+                        ret = changelog_inject_single_event (this, priv, &cld);
+                        if (!ret && active_now)
+                                SLICE_VERSION_UPDATE (slice);
+                }
+                UNLOCK (&priv->lock);
+
+                if (ret)
+                        goto out;
+
+                if (active_now) {
+                        ret = changelog_spawn_notifier (this, priv);
+                        if (!ret)
+                                ret = changelog_spawn_helper_threads (this,
+                                                                      priv);
+                } else
+                        ret = changelog_cleanup_notifier (this, priv);
+        }
+
+ out:
+        if (ret) {
+                ret = changelog_cleanup_notifier (this, priv);
+        } else {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "changelog reconfigured");
+                if (active_now)
+                        priv->active = _gf_true;
+        }
+
+        return ret;
+}
+
+int32_t
+init (xlator_t *this)
+{
+        int               ret  = -1;
+        char             *tmp  = NULL;
+        changelog_priv_t *priv = NULL;
+
+        GF_VALIDATE_OR_GOTO ("changelog", this, out);
+
+        if (!this->children || this->children->next) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "translator needs a single subvolume");
+                goto out;
+        }
+
+        if (!this->parents) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "dangling volume. please check volfile");
+                goto out;
+        }
+
+        priv = GF_CALLOC (1, sizeof (*priv), gf_changelog_mt_priv_t);
+        if (!priv)
+                goto out;
+
+        this->local_pool = mem_pool_new (changelog_local_t, 64);
+        if (!this->local_pool) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create local memory pool");
+                goto out;
+        }
+
+        LOCK_INIT (&priv->lock);
+
+        GF_OPTION_INIT ("changelog-brick", tmp, str, out);
+        if (!tmp) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "\"changelog-brick\" option is not set");
+                goto out;
+        }
+
+        priv->changelog_brick = gf_strdup (tmp);
+        if (!priv->changelog_brick)
+                goto out;
+        tmp = NULL;
+
+        GF_OPTION_INIT ("changelog-dir", tmp, str, out);
+        if (!tmp) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "\"changelog-dir\" option is not set");
+                goto out;
+        }
+
+        priv->changelog_dir = gf_strdup (tmp);
+        if (!priv->changelog_dir)
+                goto out;
+        tmp = NULL;
+
+        /**
+         * create the directory even if change-logging would be inactive
+         * so that consumers can _look_ into it (finding nothing...)
+         */
+        ret = mkdir_p (priv->changelog_dir, 0600, _gf_true);
+        if (ret)
+                goto out;
+
+        GF_OPTION_INIT ("changelog", priv->active, bool, out);
+
+        GF_OPTION_INIT ("op-mode", tmp, str, out);
+        changelog_assign_opmode (priv, tmp);
+
+        tmp = NULL;
+
+        GF_OPTION_INIT ("encoding", tmp, str, out);
+        changelog_assign_encoding (priv, tmp);
+
+        GF_OPTION_INIT ("rollover-time", priv->rollover_time, int32, out);
+
+        GF_OPTION_INIT ("fsync-interval", priv->fsync_interval, int32, out);
+
+        changelog_encode_change(priv);
+
+        GF_ASSERT (cb_bootstrap[priv->op_mode].mode == priv->op_mode);
+        priv->cb = &cb_bootstrap[priv->op_mode];
+
+        /* ... now bootstrap the logger */
+        ret = priv->cb->ctor (this, &priv->cd);
+        if (ret)
+                goto out;
+
+        priv->changelog_fd = -1;
+        ret = changelog_init (this, priv);
+        if (ret)
+                goto out;
+
+        gf_log (this->name, GF_LOG_DEBUG, "changelog translator loaded");
+
+ out:
+        if (ret) {
+                if (this->local_pool)
+                        mem_pool_destroy (this->local_pool);
+                if (priv->cb) {
+                        ret = priv->cb->dtor (this, &priv->cd);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_ERROR,
+                                        "error in cleanup during init()");
+                }
+                GF_FREE (priv->changelog_brick);
+                GF_FREE (priv->changelog_dir);
+                GF_FREE (priv);
+                this->private = NULL;
+        } else
+                this->private = priv;
+
+        return ret;
+}
+
+void
+fini (xlator_t *this)
+{
+        int               ret  = -1;
+        changelog_priv_t *priv = NULL;
+
+        priv = this->private;
+
+        if (priv) {
+                ret = priv->cb->dtor (this, &priv->cd);
+                if (ret)
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "error in fini");
+                mem_pool_destroy (this->local_pool);
+                GF_FREE (priv->changelog_brick);
+                GF_FREE (priv->changelog_dir);
+                GF_FREE (priv);
+        }
+
+        this->private = NULL;
+
+        return;
+}
+
+struct xlator_fops fops = {
+        .mknod        = changelog_mknod,
+        .mkdir        = changelog_mkdir,
+        .create       = changelog_create,
+        .symlink      = changelog_symlink,
+        .writev       = changelog_writev,
+        .truncate     = changelog_truncate,
+        .ftruncate    = changelog_ftruncate,
+        .link         = changelog_link,
+        .rename       = changelog_rename,
+        .unlink       = changelog_unlink,
+        .rmdir        = changelog_rmdir,
+        .setattr      = changelog_setattr,
+        .fsetattr     = changelog_fsetattr,
+        .setxattr     = changelog_setxattr,
+        .fsetxattr    = changelog_fsetxattr,
+        .removexattr  = changelog_removexattr,
+        .fremovexattr = changelog_fremovexattr,
+};
+
+struct xlator_cbks cbks = {
+        .forget = changelog_forget,
+};
+
+struct volume_options options[] = {
+        {.key = {"changelog"},
+         .type = GF_OPTION_TYPE_BOOL,
+         .default_value = "off",
+         .description = "enable/disable change-logging"
+        },
+        {.key = {"changelog-brick"},
+         .type = GF_OPTION_TYPE_PATH,
+         .description = "brick path to generate unique socket file name."
+                       " should be the export directory of the volume strictly."
+        },
+        {.key = {"changelog-dir"},
+         .type = GF_OPTION_TYPE_PATH,
+         .description = "directory for the changelog files"
+        },
+        {.key = {"op-mode"},
+         .type = GF_OPTION_TYPE_STR,
+         .default_value = "realtime",
+         .value = {"realtime"},
+         .description = "operation mode - futuristic operation modes"
+        },
+        {.key = {"encoding"},
+         .type = GF_OPTION_TYPE_STR,
+         .default_value = "ascii",
+         .value = {"binary", "ascii"},
+         .description = "encoding type for changelogs"
+        },
+        {.key = {"rollover-time"},
+         .default_value = "60",
+         .type = GF_OPTION_TYPE_TIME,
+         .description = "time to switch to a new changelog file (in seconds)"
+        },
+        {.key = {"fsync-interval"},
+         .type = GF_OPTION_TYPE_TIME,
+         .default_value = "0",
+         .description = "do not open CHANGELOG file with O_SYNC mode."
+                        " instead perform fsync() at specified intervals"
+        },
+        {.key = {NULL}
+        },
+};
diff --git a/xlators/features/compress/Makefile.am b/xlators/features/compress/Makefile.am
new file mode 100644
index 000000000..a985f42a8
--- /dev/null
+++ b/xlators/features/compress/Makefile.am
@@ -0,0 +1,3 @@
+SUBDIRS = src
+
+CLEANFILES =
diff --git a/xlators/features/compress/src/Makefile.am b/xlators/features/compress/src/Makefile.am
new file mode 100644
index 000000000..0bf757c06
--- /dev/null
+++ b/xlators/features/compress/src/Makefile.am
@@ -0,0 +1,17 @@
+xlator_LTLIBRARIES = cdc.la
+
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+noinst_HEADERS = cdc.h cdc-mem-types.h
+
+cdc_la_LDFLAGS = -module -avoid-version $(LIBZ_LIBS)
+
+cdc_la_SOURCES = cdc.c cdc-helper.c
+cdc_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -D$(GF_HOST_OS) \
+-shared -nostartfiles  $(LIBZ_CFLAGS)
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+CLEANFILES =
diff --git a/xlators/features/compress/src/cdc-helper.c b/xlators/features/compress/src/cdc-helper.c
new file mode 100644
index 000000000..54432ff45
--- /dev/null
+++ b/xlators/features/compress/src/cdc-helper.c
@@ -0,0 +1,547 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "logging.h"
+
+#include "cdc.h"
+#include "cdc-mem-types.h"
+
+#ifdef HAVE_LIB_Z
+#include "zlib.h"
+#endif
+
+#ifdef HAVE_LIB_Z
+/* gzip header looks something like this
+ * (RFC 1950)
+ *
+ * +---+---+---+---+---+---+---+---+---+---+
+ * |ID1|ID2|CM |FLG|     MTIME     |XFL|OS |
+ * +---+---+---+---+---+---+---+---+---+---+
+ *
+ * Data is usually sent without this header i.e
+ * Data sent = <compressed-data> + trailer(8)
+ * The trailer contains the checksum.
+ *
+ * gzip_header is added only during debugging.
+ * Refer to the function cdc_dump_iovec_to_disk
+ */
+static const char gzip_header[10] =
+        {
+                '\037', '\213', Z_DEFLATED, 0,
+                0,      0,      0,          0,
+                0,      GF_CDC_OS_ID
+        };
+
+static int32_t
+cdc_next_iovec (xlator_t *this, cdc_info_t *ci)
+{
+        int ret = -1;
+
+        ci->ncount++;
+        /* check for iovec overflow -- should not happen */
+        if (ci->ncount == MAX_IOVEC) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Zlib output buffer overflow"
+                        " ->ncount (%d) | ->MAX_IOVEC (%d)",
+                        ci->ncount, MAX_IOVEC);
+                goto out;
+        }
+
+        ret = 0;
+
+ out:
+        return ret;
+}
+
+static void
+cdc_put_long (unsigned char *string, unsigned long x)
+{
+        string[0] = (unsigned char) (x & 0xff);
+        string[1] = (unsigned char) ((x & 0xff00) >> 8);
+        string[2] = (unsigned char) ((x & 0xff0000) >> 16);
+        string[3] = (unsigned char) ((x & 0xff000000) >> 24);
+}
+
+static unsigned long
+cdc_get_long (unsigned char *buf)
+{
+        return ((unsigned long) buf[0])
+                | (((unsigned long) buf[1]) << 8)
+                | (((unsigned long) buf[2]) << 16)
+                | (((unsigned long) buf[3]) << 24);
+}
+
+static int32_t
+cdc_init_gzip_trailer (xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci)
+{
+        int   ret = -1;
+        char *buf = NULL;
+
+        ret = cdc_next_iovec (this, ci);
+        if (ret)
+                goto out;
+
+        buf = CURR_VEC(ci).iov_base =
+                (char *) GF_CALLOC (1, GF_CDC_VALIDATION_SIZE,
+                                    gf_cdc_mt_gzip_trailer_t);
+
+        if (!CURR_VEC(ci).iov_base)
+                goto out;
+
+        CURR_VEC(ci).iov_len = GF_CDC_VALIDATION_SIZE;
+
+        cdc_put_long ((unsigned char *)&buf[0], ci->crc);
+        cdc_put_long ((unsigned char *)&buf[4], ci->stream.total_in);
+
+        ret = 0;
+
+ out:
+        return ret;
+}
+
+static int32_t
+cdc_alloc_iobuf_and_init_vec (xlator_t *this,
+                              cdc_priv_t *priv, cdc_info_t *ci,
+                              int size)
+{
+        int           ret       = -1;
+        int           alloc_len = 0;
+        struct iobuf *iobuf     = NULL;
+
+        ret = cdc_next_iovec (this, ci);
+        if (ret)
+                goto out;
+
+        alloc_len = size ? size : ci->buffer_size;
+
+        iobuf = iobuf_get2 (this->ctx->iobuf_pool, alloc_len);
+        if (!iobuf)
+                goto out;
+
+        ret = iobref_add (ci->iobref, iobuf);
+        if (ret)
+                goto out;
+
+        /* Initialize this iovec */
+        CURR_VEC(ci).iov_base = iobuf->ptr;
+        CURR_VEC(ci).iov_len  = alloc_len;
+
+        ret = 0;
+
+ out:
+        return ret;
+}
+
+static void
+cdc_init_zlib_output_stream (cdc_priv_t *priv, cdc_info_t *ci, int size)
+{
+        ci->stream.next_out  = (unsigned char *) CURR_VEC(ci).iov_base;
+        ci->stream.avail_out = size ? size : ci->buffer_size;
+}
+
+/* This routine is for testing and debugging only.
+ * Data written = header(10) + <compressed-data> + trailer(8)
+ * So each gzip dump file is at least 18 bytes in size.
+ */
+void
+cdc_dump_iovec_to_disk (xlator_t *this, cdc_info_t *ci, const char *file)
+{
+        int    i      = 0;
+        int    fd     = 0;
+        size_t writen = 0;
+        size_t total_writen = 0;
+
+        fd = open (file, O_WRONLY|O_CREAT|O_TRUNC, 0777 );
+        if (fd < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Cannot open file: %s", file);
+                return;
+        }
+
+        writen = write (fd, (char *) gzip_header, 10);
+        total_writen += writen;
+        for (i = 0; i < ci->ncount; i++) {
+                writen = write (fd, (char *) ci->vec[i].iov_base, ci->vec[i].iov_len);
+                total_writen += writen;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                        "dump'd %zu bytes to %s", total_writen, GF_CDC_DEBUG_DUMP_FILE );
+
+        close (fd);
+}
+
+static int32_t
+cdc_flush_libz_buffer (cdc_priv_t *priv, xlator_t *this, cdc_info_t *ci,
+                       int (*libz_func)(z_streamp, int),
+                       int flush)
+{
+        int32_t ret = Z_OK;
+        int done = 0;
+        unsigned int deflate_len = 0;
+
+        for (;;) {
+                deflate_len = ci->buffer_size - ci->stream.avail_out;
+
+                if (deflate_len != 0) {
+                        CURR_VEC(ci).iov_len = deflate_len;
+
+                        ret = cdc_alloc_iobuf_and_init_vec (this, priv, ci, 0);
+                        if (ret) {
+                                ret = Z_MEM_ERROR;
+                                break;
+                        }
+
+                        /* Re-position Zlib output buffer */
+                        cdc_init_zlib_output_stream (priv, ci, 0);
+                }
+
+                if (done) {
+                        ci->ncount--;
+                        break;
+                }
+
+                ret = libz_func (&ci->stream, flush);
+
+                if (ret == Z_BUF_ERROR) {
+                        ret = Z_OK;
+                        ci->ncount--;
+                        break;
+                }
+
+                done = (ci->stream.avail_out != 0 || ret == Z_STREAM_END);
+
+                if (ret != Z_OK && ret != Z_STREAM_END)
+                        break;
+        }
+
+        return ret;
+}
+
+static int32_t
+do_cdc_compress (struct iovec *vec, xlator_t *this, cdc_priv_t *priv,
+                 cdc_info_t *ci)
+{
+        int ret = -1;
+
+        /* Initialize defalte */
+        ret = deflateInit2 (&ci->stream, priv->cdc_level, Z_DEFLATED,
+                            priv->window_size, priv->mem_level,
+                            Z_DEFAULT_STRATEGY);
+
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "unable to init Zlib (retval: %d)", ret);
+                goto out;
+        }
+
+        ret = cdc_alloc_iobuf_and_init_vec (this, priv, ci, 0);
+        if (ret)
+                goto out;
+
+        /* setup output buffer */
+        cdc_init_zlib_output_stream (priv, ci, 0);
+
+        /* setup input buffer */
+        ci->stream.next_in  = (unsigned char *) vec->iov_base;
+        ci->stream.avail_in = vec->iov_len;
+
+        ci->crc = crc32 (ci->crc, (const Bytef *) vec->iov_base, vec->iov_len);
+
+        gf_log (this->name, GF_LOG_DEBUG, "crc=%lu len=%d buffer_size=%d",
+                ci->crc, ci->stream.avail_in, ci->buffer_size);
+
+        /* compress !! */
+        while (ci->stream.avail_in != 0) {
+                if (ci->stream.avail_out == 0) {
+
+                        CURR_VEC(ci).iov_len = ci->buffer_size;
+
+                        ret = cdc_alloc_iobuf_and_init_vec (this, priv, ci, 0);
+                        if (ret)
+                                break;
+
+                        /* Re-position Zlib output buffer */
+                        cdc_init_zlib_output_stream (priv, ci, 0);
+                }
+
+                ret = deflate (&ci->stream, Z_NO_FLUSH);
+                if (ret != Z_OK)
+                        break;
+        }
+
+ out:
+        return ret;
+}
+
+int32_t
+cdc_compress (xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci,
+              dict_t **xdata)
+{
+        int ret = -1;
+        int i   = 0;
+
+        ci->iobref = iobref_new ();
+        if (!ci->iobref)
+                goto out;
+
+        if (!*xdata) {
+                *xdata = dict_new ();
+                if (!*xdata) {
+                        gf_log (this->name, GF_LOG_ERROR, "Cannot allocate xdata"
+                                " dict");
+                        goto out;
+                }
+        }
+
+        /* data */
+        for (i = 0; i < ci->count; i++) {
+                ret = do_cdc_compress (&ci->vector[i], this, priv, ci);
+                if (ret != Z_OK)
+                        goto deflate_cleanup_out;
+        }
+
+        /* flush zlib buffer */
+        ret = cdc_flush_libz_buffer (priv, this, ci, deflate, Z_FINISH);
+        if (!(ret == Z_OK || ret == Z_STREAM_END)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Compression Error: ret (%d)", ret);
+                ret = -1;
+                goto deflate_cleanup_out;
+        }
+
+        /* trailer */
+        ret = cdc_init_gzip_trailer (this, priv, ci);
+        if (ret)
+                goto deflate_cleanup_out;
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Compressed %ld to %ld bytes",
+                ci->stream.total_in, ci->stream.total_out);
+
+        ci->nbytes = ci->stream.total_out + GF_CDC_VALIDATION_SIZE;
+
+        /* set deflated canary value for identification */
+        ret = dict_set_int32 (*xdata, GF_CDC_DEFLATE_CANARY_VAL, 1);
+        if (ret) {
+                /* Send uncompressed data if we can't _tell_ the client
+                 * that deflated data is on it's way. So, we just log
+                 * the faliure and continue as usual.
+                 */
+                 gf_log (this->name, GF_LOG_ERROR,
+                 "Data deflated, but could not set canary"
+                 " value in dict for identification");
+        }
+
+        /* This is to be used in testing */
+        if ( priv->debug ) {
+                cdc_dump_iovec_to_disk (this, ci, GF_CDC_DEBUG_DUMP_FILE );
+        }
+
+ deflate_cleanup_out:
+        (void) deflateEnd(&ci->stream);
+
+ out:
+        return ret;
+}
+
+
+/* deflate content is checked by the presence of a canary
+ * value in the dict as the key
+ */
+static int32_t
+cdc_check_content_for_deflate (dict_t *xdata)
+{
+        return dict_get (xdata, GF_CDC_DEFLATE_CANARY_VAL) ? -1 : 0;
+}
+
+static unsigned long
+cdc_extract_crc (char *trailer)
+{
+        return cdc_get_long ((unsigned char *) &trailer[0]);
+}
+
+static unsigned long
+cdc_extract_size (char *trailer)
+{
+        return cdc_get_long ((unsigned char *) &trailer[4]);
+}
+
+static int32_t
+cdc_validate_inflate (cdc_info_t *ci, unsigned long crc,
+                      unsigned long len)
+{
+        return !((crc == ci->crc)
+                 /* inflated length is hidden inside
+                  * Zlib stream struct */
+                 && (len == ci->stream.total_out));
+}
+
+static int32_t
+do_cdc_decompress (xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci)
+{
+        int            ret          = -1;
+        int            i            = 0;
+        int            len          = 0;
+        char          *inflte       = NULL;
+        char          *trailer      = NULL;
+        struct iovec   vec          = {0,};
+        unsigned long  computed_crc = 0;
+        unsigned long  computed_len = 0;
+
+        ret = inflateInit2 (&ci->stream, priv->window_size);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Zlib: Unable to initialize inflate");
+                goto out;
+        }
+
+        vec = THIS_VEC(ci, 0);
+
+        trailer = (char *) (((char *) vec.iov_base) + vec.iov_len
+                - GF_CDC_VALIDATION_SIZE);
+
+        /* CRC of uncompressed data */
+        computed_crc = cdc_extract_crc (trailer);
+
+        /* size of uncomrpessed data */
+        computed_len = cdc_extract_size (trailer);
+
+        gf_log (this->name, GF_LOG_DEBUG, "crc=%lu len=%lu buffer_size=%d",
+                computed_crc, computed_len, ci->buffer_size);
+
+        inflte = vec.iov_base ;
+        len = vec.iov_len - GF_CDC_VALIDATION_SIZE;
+
+        /* allocate buffer of the original length of the data */
+        ret = cdc_alloc_iobuf_and_init_vec (this, priv, ci, 0);
+        if (ret)
+                goto out;
+
+        /* setup output buffer */
+        cdc_init_zlib_output_stream (priv, ci, 0);
+
+        /* setup input buffer */
+        ci->stream.next_in = (unsigned char *) inflte;
+        ci->stream.avail_in = len;
+
+        while (ci->stream.avail_in != 0) {
+                if (ci->stream.avail_out == 0) {
+                        CURR_VEC(ci).iov_len = ci->buffer_size;
+
+                        ret = cdc_alloc_iobuf_and_init_vec (this, priv, ci, 0);
+                        if (ret)
+                                break;
+
+                        /* Re-position Zlib output buffer */
+                        cdc_init_zlib_output_stream (priv, ci, 0);
+                }
+
+                ret = inflate (&ci->stream, Z_NO_FLUSH);
+                if (ret == Z_STREAM_ERROR)
+                        break;
+        }
+
+        /* flush zlib buffer */
+        ret = cdc_flush_libz_buffer (priv, this, ci, inflate, Z_SYNC_FLUSH);
+        if (!(ret == Z_OK || ret == Z_STREAM_END)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Decompression Error: ret (%d)", ret);
+                ret = -1;
+                goto out;
+        }
+
+        /* compute CRC of the uncompresses data to check for
+         * correctness */
+
+        for (i = 0; i < ci->ncount; i++) {
+                ci->crc = crc32 (ci->crc,
+                                 (const Bytef *) ci->vec[i].iov_base,
+                                 ci->vec[i].iov_len);
+        }
+
+        /* validate inflated data */
+        ret = cdc_validate_inflate (ci, computed_crc, computed_len);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Checksum or length mismatched in inflated data");
+        }
+
+ out:
+        return ret;
+}
+
+int32_t
+cdc_decompress (xlator_t *this, cdc_priv_t *priv, cdc_info_t *ci,
+                dict_t *xdata)
+{
+        int32_t ret = -1;
+
+        /* check for deflate content */
+        if (!cdc_check_content_for_deflate (xdata)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Content not deflated, passing through ...");
+                goto passthrough_out;
+        }
+
+        ci->iobref = iobref_new ();
+        if (!ci->iobref)
+                goto passthrough_out;
+
+        /* do we need to do this? can we assume that one iovec
+         * will hold per request data everytime?
+         *
+         * server/client protocol seems to deal with a single
+         * iovec even if op_ret > 1M. So, it looks ok to
+         * assume that a single iovec will contain all the
+         * data (This saves us a lot from finding the trailer
+         * and the data since it could have been split-up onto
+         * two adjacent iovec's.
+         *
+         * But, in case this translator is loaded above quick-read
+         * for some reason, then it's entirely possible that we get
+         * multiple iovec's...
+         *
+         * This case (handled below) is not tested. (by loading the
+         * xlator below quick-read)
+         */
+
+        /* @@ I_HOPE_THIS_IS_NEVER_HIT */
+        if (ci->count > 1) {
+                gf_log (this->name, GF_LOG_WARNING, "unable to handle"
+                        " multiple iovecs (%d in number)", ci->count);
+                goto inflate_cleanup_out;
+                /* TODO: coallate all iovecs in one */
+        }
+
+        ret = do_cdc_decompress (this, priv, ci);
+        if (ret)
+                goto inflate_cleanup_out;
+
+        ci->nbytes = ci->stream.total_out;
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "Inflated %ld to %ld bytes",
+                ci->stream.total_in, ci->stream.total_out);
+
+ inflate_cleanup_out:
+        (void) inflateEnd (&ci->stream);
+
+ passthrough_out:
+        return ret;
+}
+
+#endif
diff --git a/xlators/features/compress/src/cdc-mem-types.h b/xlators/features/compress/src/cdc-mem-types.h
new file mode 100644
index 000000000..ead2c70ba
--- /dev/null
+++ b/xlators/features/compress/src/cdc-mem-types.h
@@ -0,0 +1,23 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef __CDC_MEM_TYPES_H
+#define __CDC_MEM_TYPES_H
+
+#include "mem-types.h"
+
+enum gf_cdc_mem_types {
+        gf_cdc_mt_priv_t         = gf_common_mt_end + 1,
+        gf_cdc_mt_vec_t          = gf_common_mt_end + 2,
+        gf_cdc_mt_gzip_trailer_t = gf_common_mt_end + 3,
+        gf_cdc_mt_end            = gf_common_mt_end + 4,
+};
+
+#endif
diff --git a/xlators/features/compress/src/cdc.c b/xlators/features/compress/src/cdc.c
new file mode 100644
index 000000000..67fc52505
--- /dev/null
+++ b/xlators/features/compress/src/cdc.c
@@ -0,0 +1,361 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#include <sys/uio.h>
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+#include "logging.h"
+
+#include "cdc.h"
+#include "cdc-mem-types.h"
+
+static void
+cdc_cleanup_iobref (cdc_info_t *ci)
+{
+        assert(ci->iobref != NULL);
+        iobref_clear (ci->iobref);
+}
+
+int32_t
+cdc_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno,
+               struct iovec *vector, int32_t count,
+               struct iatt *stbuf, struct iobref *iobref,
+               dict_t *xdata)
+{
+        int         ret   = -1;
+        cdc_priv_t *priv  = NULL;
+        cdc_info_t  ci    = {0,};
+
+        GF_VALIDATE_OR_GOTO ("cdc", this, default_out);
+        GF_VALIDATE_OR_GOTO (this->name, frame, default_out);
+
+        priv = this->private;
+
+        if (op_ret <= 0)
+                goto default_out;
+
+        if ( (priv->min_file_size != 0)
+             && (op_ret < priv->min_file_size) )
+                goto default_out;
+
+        ci.count       = count;
+        ci.ibytes      = op_ret;
+        ci.vector      = vector;
+        ci.buf         = NULL;
+        ci.iobref      = NULL;
+        ci.ncount      = 0;
+        ci.crc         = 0;
+        ci.buffer_size = GF_CDC_DEF_BUFFERSIZE;
+
+/* A readv compresses on the server side and decompresses on the client side
+ */
+        if (priv->op_mode == GF_CDC_MODE_SERVER) {
+                ret = cdc_compress (this, priv, &ci, &xdata);
+        } else if (priv->op_mode == GF_CDC_MODE_CLIENT) {
+                ret = cdc_decompress (this, priv, &ci, xdata);
+        } else {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Invalid operation mode (%d)", priv->op_mode);
+        }
+
+        if (ret)
+                goto default_out;
+
+        STACK_UNWIND_STRICT (readv, frame, ci.nbytes, op_errno,
+                             ci.vec, ci.ncount, stbuf, iobref,
+                             xdata);
+        cdc_cleanup_iobref (&ci);
+        return 0;
+
+ default_out:
+        STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno,
+                             vector, count, stbuf, iobref, xdata);
+        return 0;
+}
+
+int32_t
+cdc_readv (call_frame_t *frame, xlator_t *this,
+           fd_t *fd, size_t size, off_t offset, uint32_t flags,
+           dict_t *xdata)
+{
+        fop_readv_cbk_t cbk = NULL;
+
+#ifdef HAVE_LIB_Z
+        cbk = cdc_readv_cbk;
+#else
+        cbk = default_readv_cbk;
+#endif
+        STACK_WIND (frame, cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readv,
+                    fd, size, offset, flags, xdata);
+        return 0;
+}
+
+int32_t
+cdc_writev_cbk (call_frame_t *frame,
+                void *cookie,
+                xlator_t *this,
+                int32_t op_ret,
+                int32_t op_errno,
+                struct iatt *prebuf,
+                struct iatt *postbuf, dict_t *xdata)
+{
+
+        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+        return 0;
+}
+
+int32_t
+cdc_writev (call_frame_t *frame,
+            xlator_t *this,
+            fd_t *fd,
+            struct iovec *vector,
+            int32_t count,
+            off_t offset,
+            uint32_t flags,
+            struct iobref *iobref, dict_t *xdata)
+{
+        int          ret   = -1;
+        cdc_priv_t  *priv  = NULL;
+        cdc_info_t   ci    = {0,};
+        size_t       isize = 0;
+
+        GF_VALIDATE_OR_GOTO ("cdc", this, default_out);
+        GF_VALIDATE_OR_GOTO (this->name, frame, default_out);
+
+        priv = this->private;
+
+        isize =  iov_length(vector, count);
+
+        if (isize <= 0)
+            goto default_out;
+
+        if ( (priv->min_file_size != 0) && (isize < priv->min_file_size) )
+            goto default_out;
+
+        ci.count       = count;
+        ci.ibytes      = isize;
+        ci.vector      = vector;
+        ci.buf         = NULL;
+        ci.iobref      = NULL;
+        ci.ncount      = 0;
+        ci.crc         = 0;
+        ci.buffer_size = GF_CDC_DEF_BUFFERSIZE;
+
+/* A writev compresses on the client side and decompresses on the server side
+ */
+	    if (priv->op_mode == GF_CDC_MODE_CLIENT) {
+		    ret = cdc_compress (this, priv, &ci, &xdata);
+	    } else if (priv->op_mode == GF_CDC_MODE_SERVER) {
+		    ret = cdc_decompress (this, priv, &ci, xdata);
+	    } else {
+		    gf_log (this->name, GF_LOG_ERROR, "Invalid operation mode (%d) ", priv->op_mode);
+	    }
+
+	    if (ret)
+		    goto default_out;
+
+	    STACK_WIND (frame,
+                    cdc_writev_cbk,
+                    FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->writev,
+                    fd, ci.vec, ci.ncount, offset, flags,
+                    iobref, xdata);
+
+        cdc_cleanup_iobref (&ci);
+        return 0;
+
+ default_out:
+        STACK_WIND (frame,
+                    cdc_writev_cbk,
+                    FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->writev,
+                    fd, vector, count, offset, flags,
+                    iobref, xdata);
+        return 0;
+}
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_cdc_mt_end);
+
+        if (ret != 0) {
+                gf_log(this->name, GF_LOG_ERROR, "Memory accounting init"
+                       "failed");
+                return ret;
+        }
+
+        return ret;
+}
+
+int32_t
+init (xlator_t *this)
+{
+        int         ret      = -1;
+        char       *temp_str = NULL;
+        cdc_priv_t *priv     = NULL;
+
+        GF_VALIDATE_OR_GOTO ("cdc", this, err);
+
+        if (!this->children || this->children->next) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Need subvolume == 1");
+                goto err;
+        }
+
+        if (!this->parents) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Dangling volume. Check volfile");
+        }
+
+        priv = GF_CALLOC (1, sizeof (*priv), gf_cdc_mt_priv_t);
+        if (!priv) {
+                goto err;
+        }
+
+        /* Check if debug mode is turned on */
+        GF_OPTION_INIT ("debug", priv->debug, bool, err);
+        if( priv->debug ) {
+                gf_log (this->name, GF_LOG_DEBUG, "CDC debug option turned on");
+        }
+
+        /* Set Gzip Window Size */
+        GF_OPTION_INIT ("window-size", priv->window_size, int32, err);
+        if ( (priv->window_size > GF_CDC_MAX_WINDOWSIZE)
+             || (priv->window_size < GF_CDC_DEF_WINDOWSIZE) ) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Invalid gzip window size (%d), using default",
+                        priv->window_size);
+                priv->window_size = GF_CDC_DEF_WINDOWSIZE;
+        }
+
+        /* Set Gzip (De)Compression Level */
+        GF_OPTION_INIT ("compression-level", priv->cdc_level, int32, err);
+        if ( ((priv->cdc_level < 1) || (priv->cdc_level > 9))
+             && (priv->cdc_level != GF_CDC_DEF_COMPRESSION) ) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Invalid gzip (de)compression level (%d),"
+                        " using default", priv->cdc_level);
+                priv->cdc_level = GF_CDC_DEF_COMPRESSION;
+        }
+
+        /* Set Gzip Memory Level */
+        GF_OPTION_INIT ("mem-level", priv->mem_level, int32, err);
+        if ( (priv->mem_level < 1) || (priv->mem_level > 9) ) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Invalid gzip memory level, using the default");
+                priv->mem_level = GF_CDC_DEF_MEMLEVEL;
+        }
+
+        /* Set min file size to enable compression */
+        GF_OPTION_INIT ("min-size", priv->min_file_size, int32, err);
+
+        /* Mode of operation - Server/Client */
+        ret = dict_get_str (this->options, "mode", &temp_str);
+        if (ret) {
+                gf_log (this->name, GF_LOG_CRITICAL,
+                        "Operation mode not specified !!");
+                goto err;
+        }
+
+        if (GF_CDC_MODE_IS_CLIENT (temp_str)) {
+                priv->op_mode = GF_CDC_MODE_CLIENT;
+        } else if (GF_CDC_MODE_IS_SERVER (temp_str)) {
+                priv->op_mode = GF_CDC_MODE_SERVER;
+        } else {
+                gf_log (this->name, GF_LOG_CRITICAL,
+                        "Bogus operation mode (%s) specified", temp_str);
+                goto err;
+        }
+
+        this->private = priv;
+        gf_log (this->name, GF_LOG_DEBUG, "CDC xlator loaded in (%s) mode",temp_str);
+        return 0;
+
+ err:
+        if (priv)
+                GF_FREE (priv);
+
+        return -1;
+}
+
+void
+fini (xlator_t *this)
+{
+        cdc_priv_t *priv = this->private;
+
+        if (priv)
+                GF_FREE (priv);
+        this->private = NULL;
+        return;
+}
+
+struct xlator_fops fops = {
+        .readv  = cdc_readv,
+        .writev = cdc_writev,
+};
+
+struct xlator_cbks cbks = {
+};
+
+struct volume_options options[] = {
+        { .key  = {"window-size"},
+          .default_value = "-15",
+          .type = GF_OPTION_TYPE_INT,
+          .description = "Size of the zlib history buffer."
+        },
+        { .key  = {"mem-level"},
+          .default_value = "8",
+          .type = GF_OPTION_TYPE_INT,
+          .description = "Memory allocated for internal compression state. "
+                         "1 uses minimum memory but is slow and reduces "
+                         "compression ratio; memLevel=9 uses maximum memory "
+                         "for optimal speed. The default value is 8."
+        },
+        { .key  = {"compression-level"},
+          .default_value = "-1",
+          .type = GF_OPTION_TYPE_INT,
+          .description = "Compression levels \n"
+                         "0 : no compression, 1 : best speed, \n"
+                         "9 : best compression, -1 : default compression "
+        },
+        { .key  = {"min-size"},
+          .default_value = "0",
+          .type = GF_OPTION_TYPE_INT,
+          .description = "Data is compressed only when its size exceeds this."
+        },
+        { .key  = {"mode"},
+          .value = {"server", "client"},
+          .type = GF_OPTION_TYPE_STR,
+          .description = "Set on the basis of where the xlator is loaded. "
+                         "This option should NOT be configured by user."
+        },
+        { .key = {"debug"},
+          .default_value = "false",
+          .type = GF_OPTION_TYPE_BOOL,
+          .description = "This is used in testing. Will dump compressed data "
+                         "to disk as a gzip file."
+        },
+        { .key  = {NULL}
+        },
+};
diff --git a/xlators/features/compress/src/cdc.h b/xlators/features/compress/src/cdc.h
new file mode 100644
index 000000000..71f4d2317
--- /dev/null
+++ b/xlators/features/compress/src/cdc.h
@@ -0,0 +1,107 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef __CDC_H
+#define __CDC_H
+
+#ifdef HAVE_LIB_Z
+#include "zlib.h"
+#endif
+
+#include "xlator.h"
+
+#ifndef MAX_IOVEC
+#define MAX_IOVEC 16
+#endif
+
+typedef struct cdc_priv {
+        int window_size;
+        int mem_level;
+        int cdc_level;
+        int min_file_size;
+        int op_mode;
+        gf_boolean_t debug;
+        gf_lock_t lock;
+} cdc_priv_t;
+
+typedef struct cdc_info {
+        /* input bits */
+        int            count;
+        int32_t        ibytes;
+        struct iovec  *vector;
+        struct iatt   *buf;
+
+        /* output bits */
+        int            ncount;
+        int            nbytes;
+        int            buffer_size;
+        struct iovec   vec[MAX_IOVEC];
+        struct iobref *iobref;
+
+        /* zlib bits */
+#ifdef HAVE_LIB_Z
+        z_stream      stream;
+#endif
+        unsigned long crc;
+} cdc_info_t;
+
+#define NVEC(ci) (ci->ncount - 1)
+#define CURR_VEC(ci) ci->vec[ci->ncount - 1]
+#define THIS_VEC(ci, i) ci->vector[i]
+
+/* Gzip defaults */
+#define GF_CDC_DEF_WINDOWSIZE  -15 /* default value */
+#define GF_CDC_MAX_WINDOWSIZE  -8  /* max value     */
+
+#ifdef HAVE_LIB_Z
+#define GF_CDC_DEF_COMPRESSION Z_DEFAULT_COMPRESSION
+#else
+#define GF_CDC_DEF_COMPRESSION -1
+#endif
+
+#define GF_CDC_DEF_MEMLEVEL    8
+#define GF_CDC_DEF_BUFFERSIZE  262144 // 256K - default compression buffer size
+
+/* Operation mode
+ * If xlator is loaded on client, readv decompresses and writev compresses
+ * If xlator is loaded on server, readv compresses and writev decompresses
+ */
+#define GF_CDC_MODE_CLIENT   0
+#define GF_CDC_MODE_SERVER   1
+
+/* min size of data to do cmpression
+ * 0 == compress even 1byte
+ */
+#define GF_CDC_MIN_CHUNK_SIZE 0
+
+#define GF_CDC_VALIDATION_SIZE 8
+
+#define GF_CDC_OS_ID 0xFF
+#define GF_CDC_DEFLATE_CANARY_VAL "deflate"
+#define GF_CDC_DEBUG_DUMP_FILE "/tmp/cdcdump.gz"
+
+#define GF_CDC_MODE_IS_CLIENT(m) \
+        (strcmp (m, "client") == 0)
+
+#define GF_CDC_MODE_IS_SERVER(m) \
+        (strcmp (m, "server") == 0)
+
+int32_t
+cdc_compress (xlator_t *this,
+              cdc_priv_t *priv,
+              cdc_info_t *ci,
+              dict_t **xdata);
+int32_t
+cdc_decompress (xlator_t *this,
+                cdc_priv_t *priv,
+                cdc_info_t *ci,
+                dict_t *xdata);
+
+#endif
diff --git a/xlators/features/filter/src/Makefile.am b/xlators/features/filter/src/Makefile.am
index cda5f0767..d1fda8b0a 100644
--- a/xlators/features/filter/src/Makefile.am
+++ b/xlators/features/filter/src/Makefile.am
@@ -1,13 +1,16 @@
 xlator_LTLIBRARIES = filter.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/testing/features
 
-filter_la_LDFLAGS = -module -avoidversion
+filter_la_LDFLAGS = -module -avoid-version
 
 filter_la_SOURCES = filter.c
 filter_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la 
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS) \
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+noinst_HEADERS = filter-mem-types.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/features/filter/src/filter-mem-types.h b/xlators/features/filter/src/filter-mem-types.h
new file mode 100644
index 000000000..47a17249b
--- /dev/null
+++ b/xlators/features/filter/src/filter-mem-types.h
@@ -0,0 +1,20 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef __FILTER_MEM_TYPES_H__
+#define __FILTER_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_filter_mem_types_ {
+        gf_filter_mt_gf_filter = gf_common_mt_end + 1,
+        gf_filter_mt_end
+};
+#endif
+
diff --git a/xlators/features/filter/src/filter.c b/xlators/features/filter/src/filter.c
index 1d5099113..1d4887b71 100644
--- a/xlators/features/filter/src/filter.c
+++ b/xlators/features/filter/src/filter.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -30,6 +20,7 @@
 #include "logging.h"
 #include "dict.h"
 #include "xlator.h"
+#include "filter-mem-types.h"
 
 #define GF_FILTER_NOBODY_UID         65534
 #define GF_FILTER_NOBODY_GID         65534
@@ -159,36 +150,36 @@ update_frame (call_frame_t *frame,
 
 /* if 'root' don't change the uid/gid */
 static int32_t
-update_stat (struct stat *stbuf,
+update_stat (struct iatt *stbuf,
 	     struct gf_filter *filter)
 {
 	int32_t idx = 0;
 	for (idx = 0; idx < filter->translate_num_uid_entries; idx++) {
-		if (stbuf->st_uid == GF_FILTER_ROOT_UID)
+		if (stbuf->ia_uid == GF_FILTER_ROOT_UID)
 			continue;
-		if ((stbuf->st_uid >= filter->translate_input_uid[idx][0]) &&
-		    (stbuf->st_uid <= filter->translate_input_uid[idx][1])) {
-			stbuf->st_uid = filter->translate_output_uid[idx];
+		if ((stbuf->ia_uid >= filter->translate_input_uid[idx][0]) &&
+		    (stbuf->ia_uid <= filter->translate_input_uid[idx][1])) {
+			stbuf->ia_uid = filter->translate_output_uid[idx];
 			break;
 		}
 	}
 	
 	for (idx = 0; idx < filter->translate_num_gid_entries; idx++) {
-		if (stbuf->st_gid == GF_FILTER_ROOT_GID)
+		if (stbuf->ia_gid == GF_FILTER_ROOT_GID)
 			continue;
-		if ((stbuf->st_gid >= filter->translate_input_gid[idx][0]) &&
-		    (stbuf->st_gid <= filter->translate_input_gid[idx][1])) {
-			stbuf->st_gid = filter->translate_output_gid[idx];
+		if ((stbuf->ia_gid >= filter->translate_input_gid[idx][0]) &&
+		    (stbuf->ia_gid <= filter->translate_input_gid[idx][1])) {
+			stbuf->ia_gid = filter->translate_output_gid[idx];
 			break;
 		}
 	}
 
 	if (filter->fixed_uid_set) {
-		stbuf->st_uid = filter->fixed_uid;
+		stbuf->ia_uid = filter->fixed_uid;
 	}
 
 	if (filter->fixed_gid_set) {
-		stbuf->st_gid = filter->fixed_gid;
+		stbuf->ia_gid = filter->fixed_gid;
 	}
 	
 	return 0;
@@ -201,14 +192,14 @@ filter_lookup_cbk (call_frame_t *frame,
 		   int32_t op_ret,
 		   int32_t op_errno,
 		   inode_t *inode,
-		   struct stat *buf,
+		   struct iatt *buf,
 		   dict_t *dict,
-                   struct stat *postparent)
+                   struct iatt *postparent)
 {
 	int ret = 0;
 	if (op_ret >= 0) {
 		update_stat (buf, this->private);
-		ret = inode_ctx_put (inode, this, (uint64_t)(long)buf->st_uid);
+		ret = inode_ctx_put (inode, this, (uint64_t)(long)buf->ia_uid);
 		if (ret == -1) {
 			gf_log (this->name, GF_LOG_ERROR,
 				"couldn't set context");
@@ -242,7 +233,7 @@ filter_stat_cbk (call_frame_t *frame,
 		 xlator_t *this,
 		 int32_t op_ret,
 		 int32_t op_errno,
-		 struct stat *buf)
+		 struct iatt *buf)
 {
 	if (op_ret >= 0) {
 		update_stat (buf, this->private);
@@ -270,8 +261,8 @@ filter_setattr_cbk (call_frame_t *frame,
                     xlator_t *this,
                     int32_t op_ret,
                     int32_t op_errno,
-                    struct stat *preop,
-                    struct stat *postop)
+                    struct iatt *preop,
+                    struct iatt *postop)
 {
 	if (op_ret >= 0) {
 		update_stat (preop, this->private);
@@ -285,7 +276,7 @@ int32_t
 filter_setattr (call_frame_t *frame,
                 xlator_t *this,
                 loc_t *loc,
-                struct stat *stbuf,
+                struct iatt *stbuf,
                 int32_t valid)
 {
 	int32_t ret = 0;
@@ -326,8 +317,8 @@ filter_fsetattr_cbk (call_frame_t *frame,
                      xlator_t *this,
                      int32_t op_ret,
                      int32_t op_errno,
-                     struct stat *preop,
-                     struct stat *postop)
+                     struct iatt *preop,
+                     struct iatt *postop)
 {
 	if (op_ret >= 0) {
                 update_stat (preop, this->private);
@@ -344,7 +335,7 @@ int32_t
 filter_fsetattr (call_frame_t *frame,
                  xlator_t *this,
                  fd_t *fd,
-                 struct stat *stbuf,
+                 struct iatt *stbuf,
                  int32_t valid)
 {
 	STACK_WIND (frame,
@@ -363,8 +354,8 @@ filter_truncate_cbk (call_frame_t *frame,
 		     xlator_t *this,
 		     int32_t op_ret,
 		     int32_t op_errno,
-		     struct stat *prebuf,
-                     struct stat *postbuf)
+		     struct iatt *prebuf,
+                     struct iatt *postbuf)
 {
 	if (op_ret >= 0) {
 		update_stat (prebuf, this->private);
@@ -415,8 +406,8 @@ filter_ftruncate_cbk (call_frame_t *frame,
 		      xlator_t *this,
 		      int32_t op_ret,
 		      int32_t op_errno,
-		      struct stat *prebuf,
-                      struct stat *postbuf)
+		      struct iatt *prebuf,
+                      struct iatt *postbuf)
 {
 	if (op_ret >= 0) {
 		update_stat (prebuf, this->private);
@@ -449,7 +440,7 @@ filter_readlink_cbk (call_frame_t *frame,
 		     int32_t op_ret,
 		     int32_t op_errno,
 		     const char *path,
-                     struct stat *sbuf)
+                     struct iatt *sbuf)
 {
         if (op_ret >= 0)
                 update_stat (sbuf, this->private);
@@ -494,15 +485,15 @@ filter_mknod_cbk (call_frame_t *frame,
 		  int32_t op_ret,
 		  int32_t op_errno,
 		  inode_t *inode,
-                  struct stat *buf,
-                  struct stat *preparent,
-                  struct stat *postparent)
+                  struct iatt *buf,
+                  struct iatt *preparent,
+                  struct iatt *postparent)
 {
 	int ret = 0;
 
 	if (op_ret >= 0) {
 		update_stat (buf, this->private);
-		ret = inode_ctx_put (inode, this, (uint64_t)(long)buf->st_uid);
+		ret = inode_ctx_put (inode, this, (uint64_t)(long)buf->ia_uid);
 		if (ret == -1) {
 			gf_log (this->name, GF_LOG_ERROR,
 				"couldn't set context");
@@ -560,14 +551,14 @@ filter_mkdir_cbk (call_frame_t *frame,
 		  int32_t op_ret,
 		  int32_t op_errno,
 		  inode_t *inode,
-                  struct stat *buf,
-                  struct stat *preparent,
-                  struct stat *postparent)
+                  struct iatt *buf,
+                  struct iatt *preparent,
+                  struct iatt *postparent)
 {
 	int ret = 0;
 	if (op_ret >= 0) {
 		update_stat (buf, this->private);
-		ret = inode_ctx_put (inode, this, (uint64_t)(long)buf->st_uid);
+		ret = inode_ctx_put (inode, this, (uint64_t)(long)buf->ia_uid);
 		if (ret == -1) {
 			gf_log (this->name, GF_LOG_ERROR,
 				"couldn't set context");
@@ -623,8 +614,8 @@ filter_unlink_cbk (call_frame_t *frame,
 		   xlator_t *this,
 		   int32_t op_ret,
 		   int32_t op_errno,
-                   struct stat *preparent,
-                   struct stat *postparent)
+                   struct iatt *preparent,
+                   struct iatt *postparent)
 {
         if (op_ret >= 0) {
 		update_stat (preparent, this->private);
@@ -679,8 +670,8 @@ filter_rmdir_cbk (call_frame_t *frame,
 		  xlator_t *this,
 		  int32_t op_ret,
 		  int32_t op_errno,
-                  struct stat *preparent,
-                  struct stat *postparent)
+                  struct iatt *preparent,
+                  struct iatt *postparent)
 {
         if (op_ret >= 0) {
 		update_stat (preparent, this->private);
@@ -736,14 +727,14 @@ filter_symlink_cbk (call_frame_t *frame,
 		    int32_t op_ret,
 		    int32_t op_errno,
 		    inode_t *inode,
-                    struct stat *buf,
-                    struct stat *preparent,
-                    struct stat *postparent)
+                    struct iatt *buf,
+                    struct iatt *preparent,
+                    struct iatt *postparent)
 {
 	int ret = 0;
 	if (op_ret >= 0) {
 		update_stat (buf, this->private);
-		ret = inode_ctx_put (inode, this, (uint64_t)(long)buf->st_uid);
+		ret = inode_ctx_put (inode, this, (uint64_t)(long)buf->ia_uid);
 		if (ret == -1) {
 			gf_log (this->name, GF_LOG_ERROR,
 				"couldn't set context");
@@ -800,11 +791,11 @@ filter_rename_cbk (call_frame_t *frame,
 		   xlator_t *this,
 		   int32_t op_ret,
 		   int32_t op_errno,
-		   struct stat *buf,
-                   struct stat *preoldparent,
-                   struct stat *postoldparent,
-                   struct stat *prenewparent,
-                   struct stat *postnewparent)
+		   struct iatt *buf,
+                   struct iatt *preoldparent,
+                   struct iatt *postoldparent,
+                   struct iatt *prenewparent,
+                   struct iatt *postnewparent)
 {
 	if (op_ret >= 0) {
 		update_stat (buf, this->private);
@@ -875,14 +866,14 @@ filter_link_cbk (call_frame_t *frame,
 		 int32_t op_ret,
 		 int32_t op_errno,
 		 inode_t *inode,
-                 struct stat *buf,
-                 struct stat *preparent,
-                 struct stat *postparent)
+                 struct iatt *buf,
+                 struct iatt *preparent,
+                 struct iatt *postparent)
 {
 	int ret = 0;
 	if (op_ret >= 0) {
 		update_stat (buf, this->private);
-		ret = inode_ctx_put (inode, this, (uint64_t)(long)buf->st_uid);
+		ret = inode_ctx_put (inode, this, (uint64_t)(long)buf->ia_uid);
 		if (ret == -1) {
 			gf_log (this->name, GF_LOG_ERROR,
 				"couldn't set context");
@@ -929,14 +920,14 @@ filter_create_cbk (call_frame_t *frame,
 		   int32_t op_errno,
 		   fd_t *fd,
 		   inode_t *inode,
-		   struct stat *buf,
-                   struct stat *preparent,
-                   struct stat *postparent)
+		   struct iatt *buf,
+                   struct iatt *preparent,
+                   struct iatt *postparent)
 {
 	int ret = 0;
 	if (op_ret >= 0) {
 		update_stat (buf, this->private);
-		ret = inode_ctx_put (inode, this, (uint64_t)(long)buf->st_uid);
+		ret = inode_ctx_put (inode, this, (uint64_t)(long)buf->ia_uid);
 		if (ret == -1) {
 			gf_log (this->name, GF_LOG_ERROR,
 				"couldn't set context");
@@ -1053,7 +1044,7 @@ filter_readv_cbk (call_frame_t *frame,
 		  int32_t op_errno,
 		  struct iovec *vector,
 		  int32_t count,
-		  struct stat *stbuf,
+		  struct iatt *stbuf,
                   struct iobref *iobref)
 {
 	if (op_ret >= 0) {
@@ -1093,8 +1084,8 @@ filter_writev_cbk (call_frame_t *frame,
 		   xlator_t *this,
 		   int32_t op_ret,
 		   int32_t op_errno,
-                   struct stat *prebuf,
-		   struct stat *postbuf)
+                   struct iatt *prebuf,
+		   struct iatt *postbuf)
 {
 	if (op_ret >= 0) {
 		update_stat (prebuf, this->private);
@@ -1145,7 +1136,7 @@ filter_fstat_cbk (call_frame_t *frame,
 		  xlator_t *this,
 		  int32_t op_ret,
 		  int32_t op_errno,
-		  struct stat *buf)
+		  struct iatt *buf)
 {
 	if (op_ret >= 0) {
 		update_stat (buf, this->private);
@@ -1355,6 +1346,25 @@ filter_removexattr (call_frame_t *frame,
 	return 0;
 }
 
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_filter_mt_end + 1);
+        
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
+                        "failed");
+                return ret;
+        }
+
+        return ret;
+}
+
 int32_t 
 init (xlator_t *this)
 {
@@ -1384,7 +1394,7 @@ init (xlator_t *this)
 			"dangling volume. check volfile ");
 	}
 
-	filter = CALLOC (sizeof (*filter), 1);
+	filter = GF_CALLOC (sizeof (*filter), 1, gf_filter_mt_gf_filter);
 	ERR_ABORT (filter);
 	
 	if (dict_get (this->options, "read-only")) {
@@ -1419,11 +1429,11 @@ init (xlator_t *this)
 		option_data = dict_get (this->options, "translate-uid");
 		value = strtok_r (option_data->data, ",", &tmp_str);
 		while (value) {
-			dup_str = strdup (value);
+			dup_str = gf_strdup (value);
 			input_value_str1 = strtok_r (dup_str, "=", &tmp_str1);
 			if (input_value_str1) {
 				/* Check for n-m */
-				char *temp_string = strdup (input_value_str1);
+				char *temp_string = gf_strdup (input_value_str1);
 				input_value_str2 = strtok_r (temp_string, "-", &tmp_str2);
 				if (gf_string2int (input_value_str2, &input_value) != 0) {
 					gf_log (this->name, GF_LOG_ERROR, 
@@ -1442,7 +1452,7 @@ init (xlator_t *this)
 					}
 				}
 				filter->translate_input_uid[filter->translate_num_uid_entries][1] = input_value;
-				FREE (temp_string);
+				GF_FREE (temp_string);
 				output_value_str = strtok_r (NULL, "=", &tmp_str1);
 				if (output_value_str) {
 					if (gf_string2int (output_value_str, &output_value) != 0) {
@@ -1471,7 +1481,7 @@ init (xlator_t *this)
 			if (filter->translate_num_uid_entries == GF_MAXIMUM_FILTERING_ALLOWED)
 				break;
 			value = strtok_r (NULL, ",", &tmp_str);
-			FREE (dup_str);
+			GF_FREE (dup_str);
 		}
 	}
 
@@ -1483,11 +1493,11 @@ init (xlator_t *this)
 		option_data = dict_get (this->options, "translate-gid");
 		value = strtok_r (option_data->data, ",", &tmp_str);
 		while (value) {
-			dup_str = strdup (value);
+			dup_str = gf_strdup (value);
 			input_value_str1 = strtok_r (dup_str, "=", &tmp_str1);
 			if (input_value_str1) {
 				/* Check for n-m */
-				char *temp_string = strdup (input_value_str1);
+				char *temp_string = gf_strdup (input_value_str1);
 				input_value_str2 = strtok_r (temp_string, "-", &tmp_str2);
 				if (gf_string2int (input_value_str2, &input_value) != 0) {
 					gf_log (this->name, GF_LOG_ERROR, 
@@ -1506,7 +1516,7 @@ init (xlator_t *this)
 					}
 				}
 				filter->translate_input_gid[filter->translate_num_gid_entries][1] = input_value;
-				FREE (temp_string);
+				GF_FREE (temp_string);
 				output_value_str = strtok_r (NULL, "=", &tmp_str1);
 				if (output_value_str) {
 					if (gf_string2int (output_value_str, &output_value) != 0) {
@@ -1536,7 +1546,7 @@ init (xlator_t *this)
 			if (filter->translate_num_gid_entries == GF_MAXIMUM_FILTERING_ALLOWED)
 				break;
 			value = strtok_r (NULL, ",", &tmp_str);
-			FREE (dup_str);
+			GF_FREE (dup_str);
 		}
 	}
 
@@ -1547,7 +1557,7 @@ init (xlator_t *this)
 		option_data = dict_get (this->options, "filter-uid");
 		value = strtok_r (option_data->data, ",", &tmp_str);
 		while (value) {
-			dup_str = strdup (value);
+			dup_str = gf_strdup (value);
 			/* Check for n-m */
 			input_value_str1 = strtok_r (dup_str, "-", &tmp_str1);
 			if (gf_string2int (input_value_str1, &input_value) != 0) {
@@ -1577,7 +1587,7 @@ init (xlator_t *this)
 			if (filter->filter_num_uid_entries == GF_MAXIMUM_FILTERING_ALLOWED)
 				break;
 			value = strtok_r (NULL, ",", &tmp_str);
-			FREE (dup_str);
+			GF_FREE (dup_str);
 		}
 		filter->partial_filter = 1;
 	}
@@ -1589,7 +1599,7 @@ init (xlator_t *this)
 		option_data = dict_get (this->options, "filter-gid");
 		value = strtok_r (option_data->data, ",", &tmp_str);
 		while (value) {
-			dup_str = strdup (value);
+			dup_str = gf_strdup (value);
 			/* Check for n-m */
 			input_value_str1 = strtok_r (dup_str, "-", &tmp_str1);
 			if (gf_string2int (input_value_str1, &input_value) != 0) {
@@ -1619,7 +1629,7 @@ init (xlator_t *this)
 			if (filter->filter_num_gid_entries == GF_MAXIMUM_FILTERING_ALLOWED)
 				break;
 			value = strtok_r (NULL, ",", &tmp_str);
-			FREE (dup_str);
+			GF_FREE (dup_str);
 		}
 		gf_log (this->name, GF_LOG_ERROR, "this option is not supported currently.. exiting");
 		return -1;
@@ -1660,7 +1670,7 @@ fini (xlator_t *this)
 {
 	struct gf_filter *filter = this->private;
 
-	FREE (filter);
+	GF_FREE (filter);
 
 	return;
 }
@@ -1692,9 +1702,6 @@ struct xlator_fops fops = {
         .fsetattr    = filter_fsetattr,
 };
 
-struct xlator_mops mops = {
-};
-
 struct xlator_cbks cbks = {
 };
 
diff --git a/xlators/bindings/python/Makefile.am b/xlators/features/gfid-access/Makefile.am
index af437a64d..af437a64d 100644
--- a/xlators/bindings/python/Makefile.am
+++ b/xlators/features/gfid-access/Makefile.am
diff --git a/xlators/features/gfid-access/src/Makefile.am b/xlators/features/gfid-access/src/Makefile.am
new file mode 100644
index 000000000..db53affaa
--- /dev/null
+++ b/xlators/features/gfid-access/src/Makefile.am
@@ -0,0 +1,15 @@
+xlator_LTLIBRARIES = gfid-access.la
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+gfid_access_la_LDFLAGS = -module -avoid-version
+
+gfid_access_la_SOURCES = gfid-access.c
+gfid_access_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+noinst_HEADERS = gfid-access.h gfid-access-mem-types.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+CLEANFILES =
diff --git a/xlators/features/gfid-access/src/gfid-access-mem-types.h b/xlators/features/gfid-access/src/gfid-access-mem-types.h
new file mode 100644
index 000000000..168d67b43
--- /dev/null
+++ b/xlators/features/gfid-access/src/gfid-access-mem-types.h
@@ -0,0 +1,23 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _GFID_ACCESS_MEM_TYPES_H
+#define _GFID_ACCESS_MEM_TYPES_H
+
+#include "mem-types.h"
+
+enum gf_changelog_mem_types {
+        gf_gfid_access_mt_priv_t = gf_common_mt_end + 1,
+        gf_gfid_access_mt_gfid_t,
+        gf_gfid_access_mt_end
+};
+
+#endif
+
diff --git a/xlators/features/gfid-access/src/gfid-access.c b/xlators/features/gfid-access/src/gfid-access.c
new file mode 100644
index 000000000..8e614397c
--- /dev/null
+++ b/xlators/features/gfid-access/src/gfid-access.c
@@ -0,0 +1,1287 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "gfid-access.h"
+#include "inode.h"
+#include "byte-order.h"
+
+
+
+void
+ga_newfile_args_free (ga_newfile_args_t *args)
+{
+        if (!args)
+                goto out;
+
+        GF_FREE (args->bname);
+
+        if (S_ISLNK (args->st_mode) && args->args.symlink.linkpath) {
+                GF_FREE (args->args.symlink.linkpath);
+                args->args.symlink.linkpath = NULL;
+        }
+
+        mem_put (args);
+out:
+        return;
+}
+
+
+void
+ga_heal_args_free (ga_heal_args_t *args)
+{
+        if (!args)
+                goto out;
+
+        GF_FREE (args->bname);
+
+        mem_put (args);
+out:
+        return;
+}
+
+
+ga_newfile_args_t *
+ga_newfile_parse_args (xlator_t *this, data_t *data)
+{
+        ga_newfile_args_t *args     = NULL;
+        ga_private_t      *priv     = NULL;
+        int                len      = 0;
+        int                blob_len = 0;
+        int                min_len  = 0;
+        void              *blob     = NULL;
+
+        priv = this->private;
+
+        blob = data->data;
+        blob_len = data->len;
+
+        min_len = sizeof (args->uid) + sizeof (args->gid) + sizeof (args->gfid)
+                + sizeof (args->st_mode) + 2 + 2;
+        if (blob_len < min_len) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Invalid length: Total length is less "
+                        "than minimum length.");
+                goto err;
+        }
+
+        args = mem_get0 (priv->newfile_args_pool);
+        if (args == NULL)
+                goto err;
+
+        args->uid = ntoh32 (*(uint32_t *)blob);
+        blob += sizeof (uint32_t);
+        blob_len -= sizeof (uint32_t);
+
+        args->gid = ntoh32 (*(uint32_t *)blob);
+        blob += sizeof (uint32_t);
+        blob_len -= sizeof (uint32_t);
+
+        memcpy (args->gfid, blob, sizeof (args->gfid));
+        blob += sizeof (args->gfid);
+        blob_len -= sizeof (args->gfid);
+
+        args->st_mode = ntoh32 (*(uint32_t *)blob);
+        blob += sizeof (uint32_t);
+        blob_len -= sizeof (uint32_t);
+
+        len = strnlen (blob, blob_len);
+        if (len == blob_len)
+        if (len == blob_len) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "gfid: %s. No null byte present.",
+                        args->gfid);
+                goto err;
+        }
+
+        args->bname = GF_CALLOC (1, (len + 1), gf_common_mt_char);
+        if (args->bname == NULL)
+                goto err;
+
+        memcpy (args->bname, blob, (len + 1));
+        blob += (len + 1);
+        blob_len -= (len + 1);
+
+        if (S_ISDIR (args->st_mode)) {
+                if (blob_len < sizeof (uint32_t)) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "gfid: %s. Invalid length",
+                                args->gfid);
+                        goto err;
+                }
+                args->args.mkdir.mode = ntoh32 (*(uint32_t *)blob);
+                blob += sizeof (uint32_t);
+                blob_len -= sizeof (uint32_t);
+
+                if (blob_len < sizeof (uint32_t)) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "gfid: %s. Invalid length",
+                                args->gfid);
+                        goto err;
+                }
+                args->args.mkdir.umask = ntoh32 (*(uint32_t *)blob);
+                blob += sizeof (uint32_t);
+                blob_len -= sizeof (uint32_t);
+                if (blob_len < 0) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "gfid: %s. Invalid length",
+                                args->gfid);
+                        goto err;
+                }
+        } else if (S_ISLNK (args->st_mode)) {
+                len = strnlen (blob, blob_len);
+                if (len == blob_len) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "gfid: %s. Invalid length",
+                                args->gfid);
+                        goto err;
+                }
+                args->args.symlink.linkpath = GF_CALLOC (1, len + 1,
+                                                         gf_common_mt_char);
+                if (args->args.symlink.linkpath == NULL)
+                        goto err;
+
+                memcpy (args->args.symlink.linkpath, blob, (len + 1));
+                blob += (len + 1);
+                blob_len -= (len + 1);
+        } else {
+                if (blob_len < sizeof (uint32_t)) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "gfid: %s. Invalid length",
+                                args->gfid);
+                        goto err;
+                }
+                args->args.mknod.mode = ntoh32 (*(uint32_t *)blob);
+                blob += sizeof (uint32_t);
+                blob_len -= sizeof (uint32_t);
+
+                if (blob_len < sizeof (uint32_t)) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "gfid: %s. Invalid length",
+                                args->gfid);
+                        goto err;
+                }
+                args->args.mknod.rdev = ntoh32 (*(uint32_t *)blob);
+                blob += sizeof (uint32_t);
+                blob_len -= sizeof (uint32_t);
+
+                if (blob_len < sizeof (uint32_t)) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "gfid: %s. Invalid length",
+                                args->gfid);
+                        goto err;
+                }
+                args->args.mknod.umask = ntoh32 (*(uint32_t *)blob);
+                blob += sizeof (uint32_t);
+                blob_len -= sizeof (uint32_t);
+        }
+
+        if (blob_len) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "gfid: %s. Invalid length",
+                        args->gfid);
+                goto err;
+        }
+
+        return args;
+
+err:
+        if (args)
+                ga_newfile_args_free (args);
+
+        return NULL;
+}
+
+ga_heal_args_t *
+ga_heal_parse_args (xlator_t *this, data_t *data)
+{
+        ga_heal_args_t *args     = NULL;
+        ga_private_t   *priv     = NULL;
+        void           *blob     = NULL;
+        int             len      = 0;
+        int             blob_len = 0;
+
+        blob = data->data;
+        blob_len = data->len;
+
+        priv = this->private;
+
+        /* bname should at least contain a character */
+        if (blob_len < (sizeof (args->gfid) + 2))
+                goto err;
+
+        args = mem_get0 (priv->heal_args_pool);
+        if (!args)
+                goto err;
+
+        memcpy (args->gfid, blob, sizeof (args->gfid));
+        blob += sizeof (args->gfid);
+        blob_len -= sizeof (args->gfid);
+
+        len = strnlen (blob, blob_len);
+        if (len == blob_len)
+                goto err;
+
+        args->bname = GF_CALLOC (1, len + 1, gf_common_mt_char);
+        if (!args->bname)
+                goto err;
+
+        memcpy (args->bname, blob, len);
+        blob_len -= (len + 1);
+
+        if (blob_len)
+                goto err;
+
+        return args;
+
+err:
+        if (args)
+                ga_heal_args_free (args);
+
+        return NULL;
+}
+
+static int32_t
+ga_fill_tmp_loc (loc_t *loc, xlator_t *this, uuid_t gfid,
+                 char *bname, dict_t *xdata, loc_t *new_loc)
+{
+        int       ret    = -1;
+        uint64_t  value  = 0;
+        inode_t  *parent = NULL;
+
+        parent = loc->inode;
+        ret = inode_ctx_get (loc->inode, this, &value);
+        if (!ret) {
+                parent = (void *)value;
+                if (uuid_is_null (parent->gfid))
+                        parent = loc->inode;
+        }
+
+        /* parent itself should be looked up */
+        uuid_copy (new_loc->pargfid, parent->gfid);
+        new_loc->parent = inode_ref (parent);
+
+        new_loc->inode = inode_grep (parent->table, parent, bname);
+        if (!new_loc->inode)
+                new_loc->inode = inode_new (parent->table);
+
+        loc_path (new_loc, bname);
+        new_loc->name = basename (new_loc->path);
+
+        /* As GFID would not be set on the entry yet, lets not send entry
+           gfid in the request */
+        /*uuid_copy (new_loc->gfid, (const unsigned char *)gfid); */
+
+        ret = dict_set_static_bin (xdata, "gfid-req", gfid, 16);
+        if (ret < 0)
+                goto out;
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+
+
+static gf_boolean_t
+__is_gfid_access_dir (uuid_t gfid)
+{
+        uuid_t  aux_gfid;
+
+        memset (aux_gfid, 0, 16);
+        aux_gfid[15] = GF_AUX_GFID;
+
+        if (uuid_compare (gfid, aux_gfid) == 0)
+                return _gf_true;
+
+        return _gf_false;
+}
+
+int32_t
+ga_forget (xlator_t *this, inode_t *inode)
+{
+        int       ret = -1;
+        uint64_t  value = 0;
+        inode_t  *tmp_inode = NULL;
+
+        ret = inode_ctx_del (inode, this, &value);
+        if (ret)
+                goto out;
+
+        tmp_inode = (void *)value;
+        inode_unref (tmp_inode);
+
+out:
+        return 0;
+}
+
+
+static int
+ga_heal_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+             int32_t op_ret, int32_t op_errno,
+             inode_t *inode, struct iatt *stat, dict_t *dict,
+             struct iatt *postparent)
+{
+        call_frame_t *orig_frame = NULL;
+
+        orig_frame = frame->local;
+        frame->local = NULL;
+
+        /* don't worry about inode linking and other stuff. They'll happen on
+         * the next lookup.
+         */
+        STACK_DESTROY (frame->root);
+
+        STACK_UNWIND_STRICT (setxattr, orig_frame, op_ret, op_errno, dict);
+
+        return 0;
+}
+
+static int32_t
+ga_newentry_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                         int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+                         struct iatt *statpost,
+                         dict_t *xdata)
+{
+        ga_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+
+        /* don't worry about inode linking and other stuff. They'll happen on
+         * the next lookup.
+         */
+        STACK_DESTROY (frame->root);
+
+        STACK_UNWIND_STRICT (setxattr, local->orig_frame, op_ret,
+                             op_errno, xdata);
+
+        loc_wipe (&local->loc);
+        mem_put (local);
+
+        return 0;
+}
+
+static int
+ga_newentry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno,
+                 inode_t *inode, struct iatt *buf,
+                 struct iatt *preparent, struct iatt *postparent,
+                 dict_t *xdata)
+{
+        ga_local_t *local = NULL;
+        struct iatt temp_stat = {0,};
+
+        local = frame->local;
+
+        if (!local->uid && !local->gid)
+                goto done;
+
+        temp_stat.ia_uid = local->uid;
+        temp_stat.ia_gid = local->gid;
+
+        STACK_WIND (frame, ga_newentry_setattr_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->setattr, &local->loc, &temp_stat,
+                    (GF_SET_ATTR_UID | GF_SET_ATTR_GID), xdata);
+
+        return 0;
+
+done:
+        /* don't worry about inode linking and other stuff. They'll happen on
+         * the next lookup.
+         */
+        frame->local = NULL;
+        STACK_DESTROY (frame->root);
+
+        STACK_UNWIND_STRICT (setxattr, local->orig_frame, op_ret,
+                             op_errno, xdata);
+
+        loc_wipe (&local->loc);
+        mem_put (local);
+
+        return 0;
+}
+
+int32_t
+ga_new_entry (call_frame_t *frame, xlator_t *this, loc_t *loc, data_t *data,
+              dict_t *xdata)
+{
+        int                ret       = -1;
+        ga_newfile_args_t *args      = NULL;
+        loc_t              tmp_loc   = {0,};
+        call_frame_t      *new_frame = NULL;
+        mode_t             mode      = 0;
+        ga_local_t        *local     = NULL;
+        uuid_t             gfid      = {0,};
+
+        args = ga_newfile_parse_args (this, data);
+        if (!args)
+                goto out;
+
+        ret = uuid_parse (args->gfid, gfid);
+        if (ret)
+                goto out;
+
+        if (!xdata)
+                xdata = dict_new ();
+
+        ret = ga_fill_tmp_loc (loc, this, gfid,
+                               args->bname, xdata, &tmp_loc);
+        if (ret)
+                goto out;
+
+        new_frame = copy_frame (frame);
+        if (!new_frame)
+                goto out;
+
+        local = mem_get0 (this->local_pool);
+        local->orig_frame = frame;
+
+        local->uid = args->uid;
+        local->gid = args->gid;
+
+        loc_copy (&local->loc, &tmp_loc);
+
+        new_frame->local = local;
+
+        if (S_ISDIR (args->st_mode)) {
+                STACK_WIND (new_frame, ga_newentry_cbk,
+                            FIRST_CHILD(this), FIRST_CHILD(this)->fops->mkdir,
+                            &tmp_loc, args->args.mkdir.mode,
+                            args->args.mkdir.umask, xdata);
+        } else if (S_ISLNK (args->st_mode)) {
+                STACK_WIND (new_frame, ga_newentry_cbk,
+                            FIRST_CHILD(this), FIRST_CHILD(this)->fops->symlink,
+                            args->args.symlink.linkpath,
+                            &tmp_loc, 0, xdata);
+        } else {
+                /* use 07777 (4 7s) for considering the Sticky bits etc) */
+                mode = (S_IFMT & args->st_mode) |
+                        (07777 & args->args.mknod.mode);;
+
+                STACK_WIND (new_frame, ga_newentry_cbk,
+                            FIRST_CHILD(this), FIRST_CHILD(this)->fops->mknod,
+                            &tmp_loc, mode,
+                            args->args.mknod.rdev, args->args.mknod.umask,
+                            xdata);
+        }
+
+        ret = 0;
+out:
+        ga_newfile_args_free (args);
+
+        return ret;
+}
+
+int32_t
+ga_heal_entry (call_frame_t *frame, xlator_t *this, loc_t *loc, data_t *data,
+               dict_t *xdata)
+{
+        int             ret       = -1;
+        ga_heal_args_t *args      = NULL;
+        loc_t           tmp_loc   = {0,};
+        call_frame_t   *new_frame = NULL;
+        uuid_t          gfid      = {0,};
+
+        args = ga_heal_parse_args (this, data);
+        if (!args)
+                goto out;
+
+        ret = uuid_parse (args->gfid, gfid);
+        if (ret)
+                goto out;
+
+        if (!xdata)
+                xdata = dict_new ();
+
+        ret = ga_fill_tmp_loc (loc, this, gfid, args->bname,
+                               xdata, &tmp_loc);
+        if (ret)
+                goto out;
+
+        new_frame = copy_frame (frame);
+        if (!new_frame)
+                goto out;
+        new_frame->local = (void *)frame;
+
+        STACK_WIND (new_frame, ga_heal_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD(this)->fops->lookup,
+                    &tmp_loc, xdata);
+
+        ret = 0;
+out:
+        if (args)
+                ga_heal_args_free (args);
+
+        return ret;
+}
+
+int32_t
+ga_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno,
+                 dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int32_t
+ga_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+             int32_t flags, dict_t *xdata)
+{
+
+        data_t  *data     = NULL;
+        int      op_errno = ENOMEM;
+        int      ret      = 0;
+        inode_t *unref    = NULL;
+
+        if ((loc->name && !strcmp (GF_GFID_DIR, loc->name)) &&
+            ((loc->parent &&
+              __is_root_gfid (loc->parent->gfid)) ||
+             __is_root_gfid (loc->pargfid))) {
+                op_errno = EPERM;
+                goto err;
+        }
+
+        data = dict_get (dict, GF_FUSE_AUX_GFID_NEWFILE);
+        if (data) {
+                ret = ga_new_entry (frame, this, loc, data, xdata);
+                if (ret)
+                        goto err;
+                return 0;
+        }
+
+        data = dict_get (dict, GF_FUSE_AUX_GFID_HEAL);
+        if (data) {
+                ret = ga_heal_entry (frame, this, loc, data, xdata);
+                if (ret)
+                        goto err;
+                return 0;
+        }
+
+        //If the inode is a virtual inode change the inode otherwise perform
+        //the operation on same inode
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, loc, unref, wind);
+
+wind:
+        STACK_WIND (frame, ga_setxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr, loc, dict, flags,
+                    xdata);
+        if (unref)
+                inode_unref (unref);
+
+        return 0;
+err:
+        STACK_UNWIND_STRICT (setxattr, frame, -1, op_errno, xdata);
+        return 0;
+}
+
+
+int32_t
+ga_virtual_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, inode_t *inode,
+                       struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+{
+        int       j           = 0;
+        int       i           = 0;
+        int       ret         = 0;
+        uint64_t  temp_ino    = 0;
+        inode_t  *cbk_inode   = NULL;
+        inode_t  *true_inode  = NULL;
+        uuid_t    random_gfid = {0,};
+
+        if (frame->local)
+                cbk_inode = frame->local;
+        else
+                cbk_inode = inode;
+
+        frame->local = NULL;
+        if (op_ret)
+                goto unwind;
+
+        if (!IA_ISDIR (buf->ia_type))
+                goto unwind;
+
+        /* need to send back a different inode for linking in itable */
+        if (cbk_inode == inode) {
+                /* check if the inode is in the 'itable' or
+                   if its just previously discover()'d inode */
+                true_inode = inode_find (inode->table, buf->ia_gfid);
+                if (!true_inode) {
+                        cbk_inode = inode_new (inode->table);
+
+                        if (!cbk_inode) {
+                                op_ret = -1;
+                                op_errno = ENOMEM;
+                                goto unwind;
+                        }
+                        /* the inode is not present in itable, ie, the actual
+                           path is not yet looked up. Use the current inode
+                           itself for now */
+
+                        inode_link (inode, NULL, NULL, buf);
+                } else {
+                        /* 'inode_ref()' has been done in inode_find() */
+                        inode = true_inode;
+                }
+
+                ret = inode_ctx_put (cbk_inode, this, (uint64_t)inode);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "failed to set the inode ctx with"
+                                "the actual inode");
+                        if (inode)
+                                inode_unref (inode);
+                }
+                inode = NULL;
+        }
+
+        if (!uuid_is_null (cbk_inode->gfid)) {
+                /* if the previous linked inode is used, use the
+                   same gfid */
+                uuid_copy (random_gfid, cbk_inode->gfid);
+        } else {
+                /* replace the buf->ia_gfid to a random gfid
+                   for directory, for files, what we received is fine */
+                uuid_generate (random_gfid);
+        }
+
+        uuid_copy (buf->ia_gfid, random_gfid);
+
+        for (i = 15; i > (15 - 8); i--) {
+                temp_ino += (uint64_t)(buf->ia_gfid[i]) << j;
+                j += 8;
+        }
+        buf->ia_ino = temp_ino;
+
+unwind:
+        /* Lookup on non-existing gfid returns ESTALE.
+           Convert into ENOENT for virtual lookup*/
+        if (op_errno == ESTALE)
+               op_errno = ENOENT;
+
+        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, cbk_inode, buf,
+                             xdata, postparent);
+
+        return 0;
+}
+
+int32_t
+ga_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno, inode_t *inode,
+               struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+{
+        ga_private_t *priv = NULL;
+
+        /* if the entry in question is not 'root',
+           then follow the normal path */
+        if (op_ret || !__is_root_gfid(buf->ia_gfid))
+                goto unwind;
+
+        priv = this->private;
+
+        /* do we need to copy root stbuf everytime? */
+        /* mostly yes, as we want to have the 'stat' info show latest
+           in every _cbk() */
+
+        /* keep the reference for root stat buf */
+        priv->root_stbuf = *buf;
+        priv->gfiddir_stbuf = priv->root_stbuf;
+        priv->gfiddir_stbuf.ia_gfid[15] = GF_AUX_GFID;
+        priv->gfiddir_stbuf.ia_ino = GF_AUX_GFID;
+
+unwind:
+        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
+                             xdata, postparent);
+        return 0;
+}
+
+int32_t
+ga_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+        ga_private_t *priv     = NULL;
+        int           ret      = -1;
+        uuid_t        tmp_gfid = {0,};
+        loc_t         tmp_loc  = {0,};
+        uint64_t      value    = 0;
+        inode_t      *inode    = NULL;
+        inode_t      *true_inode    = NULL;
+        int32_t       op_errno = ENOENT;
+
+        /* if its discover(), no need for any action here */
+        if (!loc->name)
+                goto wind;
+
+        /* if its revalidate, and inode is not of type directory,
+           proceed with 'wind' */
+        if (loc->inode && loc->inode->ia_type &&
+            !IA_ISDIR (loc->inode->ia_type)) {
+
+                /* a revalidate on ".gfid/<dentry>" is possible, check for it */
+                if (((loc->parent &&
+                      __is_gfid_access_dir (loc->parent->gfid)) ||
+                     __is_gfid_access_dir (loc->pargfid))) {
+
+                        /* here, just send 'loc->gfid' and 'loc->inode' */
+                        tmp_loc.inode = inode_ref (loc->inode);
+                        uuid_copy (tmp_loc.gfid, loc->inode->gfid);
+
+                        STACK_WIND (frame, default_lookup_cbk,
+                                    FIRST_CHILD(this),
+                                    FIRST_CHILD(this)->fops->lookup,
+                                    &tmp_loc, xdata);
+
+                        inode_unref (tmp_loc.inode);
+
+                        return 0;
+                }
+
+                /* not something to bother, continue the flow */
+                goto wind;
+        }
+
+        priv = this->private;
+
+        /* need to check if the lookup is on virtual dir */
+        if ((loc->name && !strcmp (GF_GFID_DIR, loc->name)) &&
+            ((loc->parent && __is_root_gfid (loc->parent->gfid)) ||
+             __is_root_gfid (loc->pargfid))) {
+                /* this means, the query is on '/.gfid', return the fake stat,
+                   and say success */
+
+                STACK_UNWIND_STRICT (lookup, frame, 0, 0, loc->inode,
+                                     &priv->gfiddir_stbuf, xdata,
+                                     &priv->root_stbuf);
+                return 0;
+        }
+
+        /* now, check if the lookup() is on an existing entry,
+           but on gfid-path */
+        if (!((loc->parent && __is_gfid_access_dir (loc->parent->gfid)) ||
+              __is_gfid_access_dir (loc->pargfid))) {
+                if (!loc->parent)
+                        goto wind;
+
+                ret = inode_ctx_get (loc->parent, this, &value);
+                if (ret)
+                        goto wind;
+
+                inode = (inode_t *) value;
+
+                ret = loc_copy_overload_parent (&tmp_loc, loc, inode);
+                if (ret)
+                        goto err;
+
+                STACK_WIND (frame, ga_lookup_cbk, FIRST_CHILD (this),
+                            FIRST_CHILD (this)->fops->lookup, &tmp_loc, xdata);
+
+                loc_wipe (&tmp_loc);
+                return 0;
+        }
+
+        /* make sure the 'basename' is actually a 'canonical-gfid',
+           otherwise, return error */
+        ret = uuid_parse (loc->name, tmp_gfid);
+        if (ret)
+                goto err;
+
+        /* if its fresh lookup, go ahead and send it down, if not,
+           for directory, we need indirection to actual dir inode */
+        if (!(loc->inode && loc->inode->ia_type))
+                goto discover;
+
+        /* revalidate on directory */
+        ret = inode_ctx_get (loc->inode, this, &value);
+        if (ret)
+                goto err;
+
+        inode = (void *)value;
+
+        /* valid inode, already looked up, work on that */
+        if (inode->ia_type)
+                goto discover;
+
+        /* check if the inode is in the 'itable' or
+           if its just previously discover()'d inode */
+        true_inode = inode_find (loc->inode->table, tmp_gfid);
+        if (true_inode) {
+                /* time do another lookup and update the context
+                   with proper inode */
+                op_errno = ESTALE;
+                goto err;
+        }
+
+discover:
+        /* for the virtual entries, we don't need to send 'gfid-req' key, as
+           for these entries, we don't want to 'set' a new gfid */
+        if (xdata)
+                dict_del (xdata, "gfid-req");
+
+        uuid_copy (tmp_loc.gfid, tmp_gfid);
+
+        /* if revalidate, then we need to have the proper reference */
+        if (inode) {
+                tmp_loc.inode = inode_ref (inode);
+                frame->local = loc->inode;
+        } else {
+                tmp_loc.inode = inode_ref (loc->inode);
+        }
+
+        STACK_WIND (frame, ga_virtual_lookup_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, &tmp_loc, xdata);
+
+        inode_unref (tmp_loc.inode);
+
+        return 0;
+
+wind:
+        /* used for all the normal lookup path */
+        STACK_WIND (frame, ga_lookup_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, loc, xdata);
+
+        return 0;
+
+err:
+        STACK_UNWIND_STRICT (lookup, frame, -1, op_errno, loc->inode,
+                             &priv->gfiddir_stbuf, xdata,
+                             &priv->root_stbuf);
+        return 0;
+}
+
+int
+ga_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+          mode_t umask, dict_t *xdata)
+{
+        int op_errno = 0;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+
+        STACK_WIND (frame, default_mkdir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->mkdir, loc, mode, umask,
+                    xdata);
+
+        return 0;
+
+err:
+        STACK_UNWIND_STRICT (mkdir, frame, -1, op_errno, loc->inode,
+                             NULL, NULL, NULL, xdata);
+        return 0;
+}
+
+
+int
+ga_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+           mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+{
+        int op_errno = 0;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+
+        STACK_WIND (frame, default_create_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->create,
+                    loc, flags, mode, umask, fd, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (create, frame, -1, op_errno, NULL,
+                             NULL, NULL, NULL, NULL, xdata);
+
+        return 0;
+
+}
+
+int
+ga_symlink (call_frame_t *frame, xlator_t *this, const char *linkname,
+            loc_t *loc, mode_t umask, dict_t *xdata)
+{
+        int op_errno = 0;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+
+        STACK_WIND (frame, default_symlink_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->symlink,
+                    linkname, loc, umask, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (symlink, frame, -1, op_errno, NULL,
+                             NULL, NULL, NULL, xdata);
+
+        return 0;
+}
+
+int
+ga_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+          dev_t rdev, mode_t umask, dict_t *xdata)
+{
+        int op_errno = 0;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+
+        STACK_WIND (frame, default_mknod_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->mknod, loc, mode, rdev,
+                    umask, xdata);
+
+        return 0;
+err:
+        STACK_UNWIND_STRICT (mknod, frame, -1, op_errno, NULL,
+                             NULL, NULL, NULL, xdata);
+
+        return 0;
+}
+
+int
+ga_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flag,
+          dict_t *xdata)
+{
+        int op_errno = 0;
+        inode_t *unref = NULL;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, loc, unref, wind);
+
+wind:
+        STACK_WIND (frame, default_rmdir_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->rmdir,
+                    loc, flag, xdata);
+        if (unref)
+                inode_unref (unref);
+
+        return 0;
+err:
+        STACK_UNWIND_STRICT (rmdir, frame, -1, op_errno, NULL,
+                             NULL, xdata);
+
+        return 0;
+}
+
+int
+ga_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t xflag,
+           dict_t *xdata)
+{
+        int op_errno = 0;
+        inode_t *unref = NULL;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, loc, unref, wind);
+
+wind:
+        STACK_WIND (frame, default_unlink_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->unlink,
+                    loc, xflag, xdata);
+
+        if (unref)
+                inode_unref (unref);
+
+        return 0;
+err:
+        STACK_UNWIND_STRICT (unlink, frame, -1, op_errno, NULL,
+                             NULL, xdata);
+
+        return 0;
+}
+
+int
+ga_rename (call_frame_t *frame, xlator_t *this,
+           loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+{
+        int op_errno = 0;
+        inode_t *oldloc_unref = NULL;
+        inode_t *newloc_unref = NULL;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (oldloc, op_errno, err);
+        GFID_ACCESS_ENTRY_OP_CHECK (newloc, op_errno, err);
+
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, oldloc, oldloc_unref,
+                                         handle_newloc);
+
+handle_newloc:
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, newloc, newloc_unref, wind);
+
+wind:
+        STACK_WIND (frame, default_rename_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->rename,
+                    oldloc, newloc, xdata);
+
+        if (oldloc_unref)
+                inode_unref (oldloc_unref);
+
+        if (newloc_unref)
+                inode_unref (newloc_unref);
+
+        return 0;
+err:
+        STACK_UNWIND_STRICT (rename, frame, -1, op_errno, NULL,
+                             NULL, NULL, NULL, NULL, xdata);
+
+        return 0;
+}
+
+
+int
+ga_link (call_frame_t *frame, xlator_t *this,
+         loc_t *oldloc, loc_t *newloc, dict_t *xdata)
+{
+        int op_errno = 0;
+        inode_t *oldloc_unref = NULL;
+        inode_t *newloc_unref = NULL;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (oldloc, op_errno, err);
+        GFID_ACCESS_ENTRY_OP_CHECK (newloc, op_errno, err);
+
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, oldloc, oldloc_unref,
+                                         handle_newloc);
+
+handle_newloc:
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, newloc, newloc_unref, wind);
+
+wind:
+        STACK_WIND (frame, default_link_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->link,
+                    oldloc, newloc, xdata);
+
+        if (oldloc_unref)
+                inode_unref (oldloc_unref);
+
+        if (newloc_unref)
+                inode_unref (newloc_unref);
+
+        return 0;
+err:
+        STACK_UNWIND_STRICT (link, frame, -1, op_errno, NULL,
+                             NULL, NULL, NULL, xdata);
+
+        return 0;
+}
+
+int32_t
+ga_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc,
+            fd_t *fd, dict_t *xdata)
+{
+        int op_errno = 0;
+
+        GFID_ACCESS_ENTRY_OP_CHECK (loc, op_errno, err);
+
+        /* also check if the loc->inode itself is virtual
+           inode, if yes, return with failure, mainly because we
+           can't handle all the readdirp and other things on it. */
+        if (inode_ctx_get (loc->inode, this, NULL) == 0) {
+                op_errno = ENOTSUP;
+                goto err;
+        }
+
+        STACK_WIND (frame, default_opendir_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->opendir,
+                    loc, fd, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (opendir, frame, -1, op_errno, NULL, xdata);
+
+        return 0;
+}
+
+int32_t
+ga_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+             const char *name, dict_t *xdata)
+{
+        inode_t *unref = NULL;
+
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, loc, unref, wind);
+
+wind:
+        STACK_WIND (frame, default_getxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+
+        if (unref)
+                inode_unref (unref);
+
+        return 0;
+}
+
+int32_t
+ga_stat (call_frame_t *frame, xlator_t *this, loc_t *loc,
+         dict_t *xdata)
+{
+        inode_t *unref = NULL;
+
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, loc, unref, wind);
+
+wind:
+        STACK_WIND (frame, default_stat_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->stat, loc, xdata);
+        if (unref)
+                inode_unref (unref);
+
+        return 0;
+}
+
+int32_t
+ga_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+            struct iatt *stbuf, int32_t valid,
+            dict_t *xdata)
+{
+        inode_t *unref = NULL;
+
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, loc, unref, wind);
+
+wind:
+        STACK_WIND (frame, default_setattr_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->setattr, loc, stbuf, valid,
+                    xdata);
+        if (unref)
+                inode_unref (unref);
+
+        return 0;
+}
+
+int32_t
+ga_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                const char *name, dict_t *xdata)
+{
+        inode_t *unref = NULL;
+
+        GFID_ACCESS_GET_VALID_DIR_INODE (this, loc, unref, wind);
+
+wind:
+        STACK_WIND (frame, default_removexattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->removexattr, loc, name,
+                    xdata);
+        if (unref)
+                inode_unref (unref);
+
+        return 0;
+}
+
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_gfid_access_mt_end + 1);
+
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_WARNING, "Memory accounting"
+                        " init failed");
+                return ret;
+        }
+
+        return ret;
+}
+
+int32_t
+init (xlator_t *this)
+{
+        ga_private_t *priv = NULL;
+        int ret = -1;
+
+        if (!this->children || this->children->next) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "not configured with exactly one child. exiting");
+                goto out;
+        }
+
+        /* This can be the top of graph in certain cases */
+        if (!this->parents) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "dangling volume. check volfile ");
+        }
+
+        /* TODO: define a mem-type structure */
+        priv = GF_CALLOC (1, sizeof (*priv), gf_gfid_access_mt_priv_t);
+        if (!priv)
+                goto out;
+
+        priv->newfile_args_pool = mem_pool_new (ga_newfile_args_t, 512);
+        if (!priv->newfile_args_pool)
+                goto out;
+
+        priv->heal_args_pool = mem_pool_new (ga_heal_args_t, 512);
+        if (!priv->heal_args_pool)
+                goto out;
+
+        this->local_pool = mem_pool_new (ga_local_t, 16);
+        if (!this->local_pool) {
+                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 && priv) {
+                if (priv->newfile_args_pool)
+                        mem_pool_destroy (priv->newfile_args_pool);
+                GF_FREE (priv);
+        }
+
+        return ret;
+}
+
+void
+fini (xlator_t *this)
+{
+        ga_private_t *priv = NULL;
+        priv = this->private;
+        this->private = NULL;
+
+        if (priv) {
+                if (priv->newfile_args_pool)
+                        mem_pool_destroy (priv->newfile_args_pool);
+                if (priv->heal_args_pool)
+                        mem_pool_destroy (priv->heal_args_pool);
+                GF_FREE (priv);
+        }
+
+        return;
+}
+
+
+struct xlator_fops fops = {
+        .lookup = ga_lookup,
+
+        /* entry fops */
+        .mkdir   = ga_mkdir,
+        .mknod   = ga_mknod,
+        .create  = ga_create,
+        .symlink = ga_symlink,
+        .link    = ga_link,
+        .unlink  = ga_unlink,
+        .rmdir   = ga_rmdir,
+        .rename  = ga_rename,
+
+        /* handle any other directory operations here */
+        .opendir  = ga_opendir,
+        .stat     = ga_stat,
+        .setattr  = ga_setattr,
+        .getxattr = ga_getxattr,
+        .removexattr = ga_removexattr,
+
+        /* special fop to handle more entry creations */
+        .setxattr = ga_setxattr,
+};
+
+struct xlator_cbks cbks = {
+        .forget = ga_forget,
+};
+
+struct volume_options options[] = {
+        /* This translator doesn't take any options, or provide any options */
+        { .key  = {NULL} },
+};
diff --git a/xlators/features/gfid-access/src/gfid-access.h b/xlators/features/gfid-access/src/gfid-access.h
new file mode 100644
index 000000000..e883eca69
--- /dev/null
+++ b/xlators/features/gfid-access/src/gfid-access.h
@@ -0,0 +1,134 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef __GFID_ACCESS_H__
+#define __GFID_ACCESS_H__
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "logging.h"
+#include "dict.h"
+#include "xlator.h"
+#include "defaults.h"
+#include "gfid-access-mem-types.h"
+
+#define UUID_CANONICAL_FORM_LEN 36
+
+#define GF_FUSE_AUX_GFID_NEWFILE "glusterfs.gfid.newfile"
+#define GF_FUSE_AUX_GFID_HEAL    "glusterfs.gfid.heal"
+
+#define GF_GFID_KEY "GLUSTERFS_GFID"
+#define GF_GFID_DIR ".gfid"
+#define GF_AUX_GFID 0xd
+
+#define GFID_ACCESS_GET_VALID_DIR_INODE(x,l,unref,lbl) do {             \
+                int       ret       = 0;                                \
+                uint64_t  value     = 0;                                \
+                inode_t  *tmp_inode = NULL;                             \
+                                                                        \
+                /* if its an entry operation, on the virtual */         \
+                /* directory inode as parent, we need to handle */      \
+                /* it properly */                                       \
+                if (l->parent) {                                        \
+                        ret = inode_ctx_get (l->parent, x, &value);     \
+                        if (ret)                                        \
+                                goto lbl;                               \
+                        tmp_inode = (inode_t *)value;                   \
+                        l->parent = inode_ref (tmp_inode);                          \
+                        /* if parent is virtual, no need to handle */   \
+                        /* loc->inode */                                \
+                        break;                                          \
+                }                                                       \
+                                                                        \
+                /* if its an inode operation, on the virtual */         \
+                /* directory inode itself, we need to handle */         \
+                /* it properly */                                       \
+                if (l->inode) {                                         \
+                        ret = inode_ctx_get (l->inode, x, &value);      \
+                        if (ret)                                        \
+                                goto lbl;                               \
+                        tmp_inode = (inode_t *)value;                   \
+                        l->inode = inode_ref (tmp_inode);                           \
+                }                                                       \
+                                                                        \
+        } while (0)
+
+#define GFID_ACCESS_ENTRY_OP_CHECK(loc,err,lbl)    do {                 \
+                /* need to check if the lookup is on virtual dir */     \
+                if ((loc->name && !strcmp (GF_GFID_DIR, loc->name)) &&  \
+                    ((loc->parent &&                                    \
+                      __is_root_gfid (loc->parent->gfid)) ||            \
+                      __is_root_gfid (loc->pargfid))) {                 \
+                        err = EEXIST;                                   \
+                        goto lbl;                                       \
+                }                                                       \
+                                                                        \
+                /* now, check if the lookup() is on an existing */      \
+                /* entry, but on gfid-path */                           \
+                if ((loc->parent &&                                     \
+                     __is_gfid_access_dir (loc->parent->gfid)) ||       \
+                    __is_gfid_access_dir (loc->pargfid)) {              \
+                        err = EPERM;                                    \
+                        goto lbl;                                       \
+                }                                                       \
+        } while (0)
+
+
+typedef struct {
+        unsigned int  uid;
+        unsigned int  gid;
+        char          gfid[UUID_CANONICAL_FORM_LEN + 1];
+        unsigned int  st_mode;
+        char         *bname;
+
+        union {
+                struct _symlink_in {
+                        char     *linkpath;
+                } __attribute__ ((__packed__)) symlink;
+
+                struct _mknod_in {
+                        unsigned int mode;
+                        unsigned int rdev;
+                        unsigned int umask;
+                } __attribute__ ((__packed__)) mknod;
+
+                struct _mkdir_in {
+                        unsigned int mode;
+                        unsigned int umask;
+                } __attribute__ ((__packed__)) mkdir;
+        } __attribute__ ((__packed__)) args;
+} __attribute__((__packed__)) ga_newfile_args_t;
+
+typedef struct {
+        char      gfid[UUID_CANONICAL_FORM_LEN + 1];
+        char     *bname; /* a null terminated basename */
+} __attribute__((__packed__)) ga_heal_args_t;
+
+struct ga_private {
+        /* root inode's stbuf */
+        struct iatt root_stbuf;
+        struct iatt gfiddir_stbuf;
+        struct mem_pool *newfile_args_pool;
+        struct mem_pool *heal_args_pool;
+};
+typedef struct ga_private ga_private_t;
+
+struct __ga_local {
+        call_frame_t *orig_frame;
+        unsigned int uid;
+        unsigned int gid;
+        loc_t loc;
+};
+typedef struct __ga_local ga_local_t;
+
+#endif /* __GFID_ACCESS_H__ */
diff --git a/xlators/features/glupy/Makefile.am b/xlators/features/glupy/Makefile.am
new file mode 100644
index 000000000..060429ecf
--- /dev/null
+++ b/xlators/features/glupy/Makefile.am
@@ -0,0 +1,3 @@
+SUBDIRS = src examples
+
+CLEANFILES =
diff --git a/xlators/features/glupy/doc/README.md b/xlators/features/glupy/doc/README.md
new file mode 100644
index 000000000..2d7b30ef6
--- /dev/null
+++ b/xlators/features/glupy/doc/README.md
@@ -0,0 +1,44 @@
+This is just the very start for a GlusterFS[1] meta-translator that will
+allow translator code to be written in Python.  It's based on the standard
+Python embedding (not extending) techniques, plus a dash of the ctypes module.
+The interface is a pretty minimal adaptation of the dispatches and callbacks
+from the C API[2] to Python, as follows:
+
+* Dispatch functions and callbacks must be defined on an "xlator" class
+  derived from gluster.Translator so that they'll be auto-registered with
+  the C translator during initialization.
+
+* For each dispatch or callback function you want to intercept, you define a
+  Python function using the xxx\_fop\_t or xxx\_cbk\_t decorator.
+
+* The arguments for each operation are different, so you'll need to refer to
+  the C API.  GlusterFS-specific types are used (though only loc\_t is fully
+  defined so far) and type correctness is enforced by ctypes.
+
+* If you do intercept a dispatch function, it is your responsibility to call
+  xxx\_wind (like STACK\_WIND in the C API but operation-specific) to pass
+  the request to the next translator.  If you do not intercept a function, it
+  will default the same way as for C (pass through to the same operation with
+  the same arguments on the first child translator).
+
+* If you intercept a callback function, it is your responsibility to call
+  xxx\_unwind (like STACK\_UNWIND\_STRICT in the C API) to pass the request back
+  to the caller.
+
+So far only the lookup and create operations are handled this way, to support
+the "negative lookup" example.  Now that the basic infrastructure is in place,
+adding more functions should be very quick, though with that much boilerplate I
+might pause to write a code generator.  I also plan to add structure
+definitions and interfaces for some of the utility functions in libglusterfs
+(especially those having to do with inode and fd context) in the fairly near
+future.  Note that you can also use ctypes to get at anything not explicitly
+exposed to Python already.
+
+_If you're coming here because of the Linux Journal article, please note that
+the code has evolved since that was written. The version that matches the
+article is here:_
+
+https://github.com/jdarcy/glupy/tree/4bbae91ba459ea46ef32f2966562492e4ca9187a
+
+[1] http://www.gluster.org
+[2] http://hekafs.org/dist/xlator_api_2.html
diff --git a/xlators/features/glupy/doc/TESTING b/xlators/features/glupy/doc/TESTING
new file mode 100644
index 000000000..e05f17f49
--- /dev/null
+++ b/xlators/features/glupy/doc/TESTING
@@ -0,0 +1,9 @@
+Loading a translator written in Python using the glupy meta translator
+-------------------------------------------------------------------------------
+'test.vol' is a simple volfile with the debug-trace Python translator on top
+of a brick. The volfile can be mounted using the following command.
+
+$ glusterfs --debug -f test.vol /path/to/mntpt
+
+If then file operations are performed on the newly mounted file system, log
+output would be printed by the Python translator on the standard output.
diff --git a/xlators/features/glupy/doc/test.vol b/xlators/features/glupy/doc/test.vol
new file mode 100644
index 000000000..0751a488c
--- /dev/null
+++ b/xlators/features/glupy/doc/test.vol
@@ -0,0 +1,10 @@
+volume vol-posix
+    type storage/posix
+    option directory /path/to/brick
+end-volume
+
+volume vol-glupy
+    type features/glupy
+    option module-name debug-trace
+    subvolumes vol-posix
+end-volume
diff --git a/xlators/features/glupy/examples/Makefile.am b/xlators/features/glupy/examples/Makefile.am
new file mode 100644
index 000000000..c26abeaaf
--- /dev/null
+++ b/xlators/features/glupy/examples/Makefile.am
@@ -0,0 +1,5 @@
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+glupyexamplesdir = $(xlatordir)/glupy
+
+glupyexamples_PYTHON = negative.py helloworld.py debug-trace.py
diff --git a/xlators/features/glupy/examples/debug-trace.py b/xlators/features/glupy/examples/debug-trace.py
new file mode 100644
index 000000000..6eef1b58b
--- /dev/null
+++ b/xlators/features/glupy/examples/debug-trace.py
@@ -0,0 +1,775 @@
+import sys
+import stat
+from uuid import UUID
+from time import strftime, localtime
+from gluster.glupy import *
+
+# This translator was written primarily to test the fop entry point definitions
+# and structure definitions in 'glupy.py'.
+
+# It is similar to the C language debug-trace translator, which logs the
+# arguments passed to the fops and their corresponding cbk functions.
+
+dl.get_id.restype = c_long
+dl.get_id.argtypes = [ POINTER(call_frame_t) ]
+
+dl.get_rootunique.restype = c_uint64
+dl.get_rootunique.argtypes = [ POINTER(call_frame_t) ]
+
+def uuid2str (gfid):
+        return str(UUID(''.join(map("{0:02x}".format, gfid))))
+
+
+def st_mode_from_ia (prot, filetype):
+        st_mode = 0
+        type_bit = 0
+        prot_bit = 0
+
+        if filetype == IA_IFREG:
+                type_bit = stat.S_IFREG
+        elif filetype == IA_IFDIR:
+                type_bit = stat.S_IFDIR
+        elif filetype == IA_IFLNK:
+                type_bit = stat.S_IFLNK
+        elif filetype == IA_IFBLK:
+                type_bit = stat.S_IFBLK
+        elif filetype == IA_IFCHR:
+                type_bit = stat.S_IFCHR
+        elif filetype == IA_IFIFO:
+                type_bit = stat.S_IFIFO
+        elif filetype == IA_IFSOCK:
+                type_bit = stat.S_IFSOCK
+        elif filetype == IA_INVAL:
+                pass
+
+
+        if prot.suid:
+                prot_bit |= stat.S_ISUID
+        if prot.sgid:
+                prot_bit |= stat.S_ISGID
+        if prot.sticky:
+                prot_bit |= stat.S_ISVTX
+
+        if prot.owner.read:
+                prot_bit |= stat.S_IRUSR
+        if prot.owner.write:
+                prot_bit |= stat.S_IWUSR
+        if prot.owner.execn:
+                prot_bit |= stat.S_IXUSR
+
+        if prot.group.read:
+                prot_bit |= stat.S_IRGRP
+        if prot.group.write:
+                prot_bit |= stat.S_IWGRP
+        if prot.group.execn:
+                prot_bit |= stat.S_IXGRP
+
+        if prot.other.read:
+                prot_bit |= stat.S_IROTH
+        if prot.other.write:
+                prot_bit |= stat.S_IWOTH
+        if prot.other.execn:
+                prot_bit |= stat.S_IXOTH
+
+        st_mode = (type_bit | prot_bit)
+
+        return st_mode
+
+
+def trace_stat2str (buf):
+        gfid = uuid2str(buf.contents.ia_gfid)
+        mode = st_mode_from_ia(buf.contents.ia_prot, buf.contents.ia_type)
+        atime_buf = strftime("[%b %d %H:%M:%S]",
+                             localtime(buf.contents.ia_atime))
+        mtime_buf = strftime("[%b %d %H:%M:%S]",
+                             localtime(buf.contents.ia_mtime))
+        ctime_buf = strftime("[%b %d %H:%M:%S]",
+                             localtime(buf.contents.ia_ctime))
+        return ("(gfid={0:s}, ino={1:d}, mode={2:o}, nlink={3:d}, uid ={4:d}, "+
+                "gid ={5:d}, size={6:d}, blocks={7:d}, atime={8:s}, mtime={9:s}, "+
+                "ctime={10:s})").format(gfid, buf.contents.ia_no, mode,
+                                        buf.contents.ia_nlink,
+                                        buf.contents.ia_uid,
+                                        buf.contents.ia_gid,
+                                        buf.contents.ia_size,
+                                        buf.contents.ia_blocks,
+                                        atime_buf, mtime_buf,
+                                        ctime_buf)
+
+class xlator(Translator):
+
+        def __init__(self, c_this):
+                Translator.__init__(self, c_this)
+                self.gfids = {}
+
+        def lookup_fop(self, frame, this, loc, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.gfid)
+                print("GLUPY TRACE LOOKUP FOP- {0:d}: gfid={1:s}; " +
+                      "path={2:s}").format(unique, gfid, loc.contents.path)
+                self.gfids[key] = gfid
+                dl.wind_lookup(frame, POINTER(xlator_t)(), loc, xdata)
+                return 0
+
+        def lookup_cbk(self, frame, cookie, this, op_ret, op_errno,
+                       inode, buf, xdata, postparent):
+                unique =dl.get_rootunique(frame)
+                key =dl.get_id(frame)
+                if op_ret == 0:
+                        gfid = uuid2str(buf.contents.ia_gfid)
+                        statstr = trace_stat2str(buf)
+                        postparentstr = trace_stat2str(postparent)
+                        print("GLUPY TRACE LOOKUP CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; *buf={3:s}; " +
+                              "*postparent={4:s}").format(unique, gfid,
+                                                          op_ret, statstr,
+                                                          postparentstr)
+                else:
+                        gfid = self.gfids[key]
+                        print("GLUPY TRACE LOOKUP CBK - {0:d}: gfid={1:s};" +
+                              " op_ret={2:d}; op_errno={3:d}").format(unique,
+                                                                      gfid,
+                                                                      op_ret,
+                                                                      op_errno)
+                del self.gfids[key]
+                dl.unwind_lookup(frame, cookie, this, op_ret, op_errno,
+                                 inode, buf, xdata, postparent)
+                return 0
+
+        def create_fop(self, frame, this, loc, flags, mode, umask, fd,
+                       xdata):
+                unique = dl.get_rootunique(frame)
+                gfid = uuid2str(loc.contents.gfid)
+                print("GLUPY TRACE CREATE FOP- {0:d}: gfid={1:s}; path={2:s}; " +
+                      "fd={3:s}; flags=0{4:o}; mode=0{5:o}; " +
+                      "umask=0{6:o}").format(unique, gfid, loc.contents.path,
+                                             fd, flags, mode, umask)
+                dl.wind_create(frame, POINTER(xlator_t)(), loc, flags,mode,
+                               umask, fd, xdata)
+                return 0
+
+        def create_cbk(self, frame, cookie, this, op_ret, op_errno, fd,
+                       inode, buf, preparent, postparent, xdata):
+                unique = dl.get_rootunique(frame)
+                if op_ret >= 0:
+                        gfid = uuid2str(inode.contents.gfid)
+                        statstr = trace_stat2str(buf)
+                        preparentstr = trace_stat2str(preparent)
+                        postparentstr = trace_stat2str(postparent)
+                        print("GLUPY TRACE CREATE CBK- {0:d}: gfid={1:s};" +
+                              " op_ret={2:d}; fd={3:s}; *stbuf={4:s}; " +
+                              "*preparent={5:s};" +
+                              " *postparent={6:s}").format(unique, gfid, op_ret,
+                                                           fd, statstr,
+                                                           preparentstr,
+                                                           postparentstr)
+                else:
+                        print ("GLUPY TRACE CREATE CBK- {0:d}: op_ret={1:d}; " +
+                              "op_errno={2:d}").format(unique, op_ret, op_errno)
+                dl.unwind_create(frame, cookie, this, op_ret, op_errno, fd,
+                                 inode, buf, preparent, postparent, xdata)
+                return 0
+
+        def open_fop(self, frame, this, loc, flags, fd, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE OPEN FOP- {0:d}: gfid={1:s}; path={2:s}; "+
+                      "flags={3:d}; fd={4:s}").format(unique, gfid,
+                                                      loc.contents.path, flags,
+                                                      fd)
+                self.gfids[key] = gfid
+                dl.wind_open(frame, POINTER(xlator_t)(), loc, flags, fd, xdata)
+                return 0
+
+        def open_cbk(self, frame, cookie, this, op_ret, op_errno, fd, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE OPEN CBK- {0:d}: gfid={1:s}; op_ret={2:d}; "
+                      "op_errno={3:d}; *fd={4:s}").format(unique, gfid,
+                                                          op_ret, op_errno, fd)
+                del self.gfids[key]
+                dl.unwind_open(frame, cookie, this, op_ret, op_errno, fd,
+                               xdata)
+                return 0
+
+        def readv_fop(self, frame, this, fd, size, offset, flags, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(fd.contents.inode.contents.gfid)
+                print("GLUPY TRACE READV FOP- {0:d}: gfid={1:s}; "+
+                      "fd={2:s}; size ={3:d}; offset={4:d}; " +
+                      "flags=0{5:x}").format(unique, gfid, fd, size, offset,
+                                             flags)
+                self.gfids[key] = gfid
+                dl.wind_readv (frame, POINTER(xlator_t)(), fd, size, offset,
+                               flags, xdata)
+                return 0
+
+        def readv_cbk(self, frame, cookie, this, op_ret, op_errno, vector,
+                      count, buf, iobref, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                if op_ret >= 0:
+                        statstr = trace_stat2str(buf)
+                        print("GLUPY TRACE READV CBK- {0:d}: gfid={1:s}, "+
+                              "op_ret={2:d}; *buf={3:s};").format(unique, gfid,
+                                                                  op_ret,
+                                                                  statstr)
+
+                else:
+                        print("GLUPY TRACE READV CBK- {0:d}: gfid={1:s}, "+
+                              "op_ret={2:d}; op_errno={3:d}").format(unique,
+                                                                     gfid,
+                                                                     op_ret,
+                                                                     op_errno)
+                del self.gfids[key]
+                dl.unwind_readv (frame, cookie, this, op_ret, op_errno,
+                                 vector, count, buf, iobref, xdata)
+                return 0
+
+        def writev_fop(self, frame, this, fd, vector, count, offset, flags,
+                       iobref, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(fd.contents.inode.contents.gfid)
+                print("GLUPY TRACE  WRITEV FOP- {0:d}: gfid={1:s}; " +
+                      "fd={2:s}; count={3:d}; offset={4:d}; " +
+                      "flags=0{5:x}").format(unique, gfid, fd, count, offset,
+                                             flags)
+                self.gfids[key] = gfid
+                dl.wind_writev(frame, POINTER(xlator_t)(), fd, vector, count,
+                               offset, flags, iobref, xdata)
+                return 0
+
+        def writev_cbk(self, frame, cookie, this, op_ret, op_errno, prebuf,
+                       postbuf, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                if op_ret >= 0:
+                        preopstr = trace_stat2str(prebuf)
+                        postopstr = trace_stat2str(postbuf)
+                        print("GLUPY TRACE WRITEV CBK- {0:d}: op_ret={1:d}; " +
+                              "*prebuf={2:s}; " +
+                              "*postbuf={3:s}").format(unique, op_ret, preopstr,
+                                                       postopstr)
+                else:
+                        gfid = self.gfids[key]
+                        print("GLUPY TRACE WRITEV CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; op_errno={3:d}").format(unique,
+                                                                     gfid,
+                                                                     op_ret,
+                                                                     op_errno)
+                del self.gfids[key]
+                dl.unwind_writev (frame, cookie, this, op_ret, op_errno,
+                                  prebuf, postbuf, xdata)
+                return 0
+
+        def opendir_fop(self, frame, this, loc, fd, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE OPENDIR FOP- {0:d}: gfid={1:s}; path={2:s}; "+
+                      "fd={3:s}").format(unique, gfid, loc.contents.path, fd)
+                self.gfids[key] = gfid
+                dl.wind_opendir(frame, POINTER(xlator_t)(), loc, fd, xdata)
+                return 0
+
+        def opendir_cbk(self, frame, cookie, this, op_ret, op_errno, fd,
+                        xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE OPENDIR CBK- {0:d}: gfid={1:s}; op_ret={2:d};"+
+                      " op_errno={3:d}; fd={4:s}").format(unique, gfid, op_ret,
+                                                          op_errno, fd)
+                del self.gfids[key]
+                dl.unwind_opendir(frame, cookie, this, op_ret, op_errno,
+                                  fd, xdata)
+                return 0
+
+        def readdir_fop(self, frame, this, fd, size, offset, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(fd.contents.inode.contents.gfid)
+                print("GLUPY TRACE READDIR FOP- {0:d}:  gfid={1:s}; fd={2:s}; " +
+                      "size={3:d}; offset={4:d}").format(unique, gfid, fd, size,
+                                                         offset)
+                self.gfids[key] = gfid
+                dl.wind_readdir(frame, POINTER(xlator_t)(), fd, size, offset,
+                                xdata)
+                return 0
+
+        def readdir_cbk(self, frame, cookie, this, op_ret, op_errno, buf,
+                        xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE READDIR CBK- {0:d}: gfid={1:s}; op_ret={2:d};"+
+                      " op_errno={3:d}").format(unique, gfid, op_ret, op_errno)
+                del self.gfids[key]
+                dl.unwind_readdir(frame, cookie, this, op_ret, op_errno, buf,
+                                  xdata)
+                return 0
+
+        def readdirp_fop(self, frame, this, fd, size, offset, dictionary):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(fd.contents.inode.contents.gfid)
+                print("GLUPY TRACE READDIRP FOP- {0:d}: gfid={1:s}; fd={2:s}; "+
+                      " size={3:d}; offset={4:d}").format(unique, gfid, fd, size,
+                                                          offset)
+                self.gfids[key] = gfid
+                dl.wind_readdirp(frame, POINTER(xlator_t)(), fd, size, offset,
+                                 dictionary)
+                return 0
+
+        def readdirp_cbk(self, frame, cookie, this, op_ret, op_errno, buf,
+                         xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE READDIRP CBK- {0:d}: gfid={1:s}; "+
+                      "op_ret={2:d}; op_errno={3:d}").format(unique, gfid,
+                                                             op_ret, op_errno)
+                del self.gfids[key]
+                dl.unwind_readdirp(frame, cookie, this, op_ret, op_errno, buf,
+                                  xdata)
+                return 0
+
+        def mkdir_fop(self, frame, this, loc, mode, umask, xdata):
+                unique = dl.get_rootunique(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE MKDIR FOP- {0:d}: gfid={1:s}; path={2:s}; " +
+                      "mode={3:d}; umask=0{4:o}").format(unique, gfid,
+                                                         loc.contents.path, mode,
+                                                         umask)
+                dl.wind_mkdir(frame, POINTER(xlator_t)(), loc, mode, umask,
+                              xdata)
+                return 0
+
+        def mkdir_cbk(self, frame, cookie, this, op_ret, op_errno, inode, buf,
+                      preparent, postparent,  xdata):
+                unique = dl.get_rootunique(frame)
+                if op_ret == 0:
+                        gfid = uuid2str(inode.contents.gfid)
+                        statstr = trace_stat2str(buf)
+                        preparentstr = trace_stat2str(preparent)
+                        postparentstr = trace_stat2str(postparent)
+                        print("GLUPY TRACE MKDIR CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; *stbuf={3:s}; *prebuf={4:s}; "+
+                              "*postbuf={5:s} ").format(unique, gfid, op_ret,
+                                                        statstr,
+                                                        preparentstr,
+                                                        postparentstr)
+                else:
+                        print("GLUPY TRACE MKDIR CBK- {0:d}:  op_ret={1:d}; "+
+                              "op_errno={2:d}").format(unique, op_ret, op_errno)
+                dl.unwind_mkdir(frame, cookie, this, op_ret, op_errno, inode,
+                                buf, preparent, postparent, xdata)
+                return 0
+
+        def rmdir_fop(self, frame, this, loc, flags, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE RMDIR FOP- {0:d}: gfid={1:s}; path={2:s}; "+
+                      "flags={3:d}").format(unique, gfid, loc.contents.path,
+                                            flags)
+                self.gfids[key] = gfid
+                dl.wind_rmdir(frame, POINTER(xlator_t)(), loc, flags, xdata)
+                return 0
+
+        def rmdir_cbk(self, frame, cookie, this, op_ret, op_errno, preparent,
+                      postparent, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                if op_ret == 0:
+                        preparentstr = trace_stat2str(preparent)
+                        postparentstr = trace_stat2str(postparent)
+                        print("GLUPY TRACE RMDIR CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; *prebuf={3:s}; "+
+                              "*postbuf={4:s}").format(unique, gfid, op_ret,
+                                                       preparentstr,
+                                                       postparentstr)
+                else:
+                        print("GLUPY TRACE RMDIR CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; op_errno={3:d}").format(unique,
+                                                                     gfid,
+                                                                     op_ret,
+                                                                     op_errno)
+                del self.gfids[key]
+                dl.unwind_rmdir(frame, cookie, this, op_ret, op_errno,
+                                preparent, postparent, xdata)
+                return 0
+
+        def stat_fop(self, frame, this, loc, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE STAT FOP- {0:d}: gfid={1:s}; " +
+                      " path={2:s}").format(unique, gfid, loc.contents.path)
+                self.gfids[key] = gfid
+                dl.wind_stat(frame, POINTER(xlator_t)(), loc, xdata)
+                return 0
+
+        def stat_cbk(self, frame, cookie, this, op_ret, op_errno, buf,
+                     xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                if op_ret == 0:
+                        statstr = trace_stat2str(buf)
+                        print("GLUPY TRACE STAT CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d};  *buf={3:s};").format(unique,
+                                                                   gfid,
+                                                                   op_ret,
+                                                                   statstr)
+                else:
+                        print("GLUPY TRACE STAT CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; op_errno={3:d}").format(unique,
+                                                                     gfid,
+                                                                     op_ret,
+                                                                     op_errno)
+                del self.gfids[key]
+                dl.unwind_stat(frame, cookie, this, op_ret, op_errno,
+                               buf, xdata)
+                return 0
+
+        def fstat_fop(self, frame, this, fd, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(fd.contents.inode.contents.gfid)
+                print("GLUPY TRACE FSTAT FOP- {0:d}:  gfid={1:s}; " +
+                      "fd={2:s}").format(unique, gfid, fd)
+                self.gfids[key] = gfid
+                dl.wind_fstat(frame, POINTER(xlator_t)(), fd, xdata)
+                return 0
+
+        def fstat_cbk(self, frame, cookie, this, op_ret, op_errno, buf,
+                      xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                if op_ret == 0:
+                        statstr = trace_stat2str(buf)
+                        print("GLUPY TRACE FSTAT CBK- {0:d}: gfid={1:s} "+
+                              " op_ret={2:d}; *buf={3:s}").format(unique,
+                                                                  gfid,
+                                                                  op_ret,
+                                                                  statstr)
+                else:
+                        print("GLUPY TRACE FSTAT CBK- {0:d}: gfid={1:s} "+
+                              "op_ret={2:d}; op_errno={3:d}").format(unique.
+                                                                     gfid,
+                                                                     op_ret,
+                                                                     op_errno)
+                del self.gfids[key]
+                dl.unwind_fstat(frame, cookie, this, op_ret, op_errno,
+                                buf, xdata)
+                return 0
+
+        def statfs_fop(self, frame, this, loc, xdata):
+                unique = dl.get_rootunique(frame)
+                if loc.contents.inode:
+                        gfid = uuid2str(loc.contents.inode.contents.gfid)
+                else:
+                        gfid = "0"
+                print("GLUPY TRACE STATFS FOP- {0:d}: gfid={1:s}; "+
+                      "path={2:s}").format(unique, gfid, loc.contents.path)
+                dl.wind_statfs(frame, POINTER(xlator_t)(), loc, xdata)
+                return 0
+
+        def statfs_cbk(self, frame, cookie, this, op_ret, op_errno, buf,
+                       xdata):
+                unique = dl.get_rootunique(frame)
+                if op_ret == 0:
+                        #TBD: print buf (pointer to an iovec type object)
+                        print("GLUPY TRACE STATFS CBK {0:d}: "+
+                              "op_ret={1:d}").format(unique, op_ret)
+                else:
+                        print("GLUPY TRACE STATFS CBK-  {0:d}"+
+                              "op_ret={1:d}; op_errno={2:d}").format(unique,
+                                                                     op_ret,
+                                                                     op_errno)
+                dl.unwind_statfs(frame, cookie, this, op_ret, op_errno,
+                                 buf, xdata)
+                return 0
+
+        def getxattr_fop(self, frame, this, loc, name, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE GETXATTR FOP- {0:d}: gfid={1:s}; path={2:s};"+
+                      " name={3:s}").format(unique, gfid, loc.contents.path,
+                                            name)
+                self.gfids[key]=gfid
+                dl.wind_getxattr(frame, POINTER(xlator_t)(), loc, name, xdata)
+                return 0
+
+        def getxattr_cbk(self, frame, cookie, this, op_ret, op_errno,
+                         dictionary, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE GETXATTR CBK- {0:d}: gfid={1:s}; "+
+                      "op_ret={2:d}; op_errno={3:d}; "+
+                      " dictionary={4:s}").format(unique, gfid, op_ret, op_errno,
+                                                  dictionary)
+                del self.gfids[key]
+                dl.unwind_getxattr(frame, cookie, this, op_ret, op_errno,
+                                   dictionary, xdata)
+                return 0
+
+        def fgetxattr_fop(self, frame, this, fd, name, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(fd.contents.inode.contents.gfid)
+                print("GLUPY TRACE FGETXATTR FOP- {0:d}: gfid={1:s}; fd={2:s}; "+
+                      "name={3:s}").format(unique, gfid, fd, name)
+                self.gfids[key] = gfid
+                dl.wind_fgetxattr(frame, POINTER(xlator_t)(), fd, name, xdata)
+                return 0
+
+        def fgetxattr_cbk(self, frame, cookie, this, op_ret, op_errno,
+                          dictionary, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE FGETXATTR CBK- {0:d}: gfid={1:s}; "+
+                      "op_ret={2:d}; op_errno={3:d};"+
+                      " dictionary={4:s}").format(unique, gfid, op_ret,
+                                                  op_errno, dictionary)
+                del self.gfids[key]
+                dl.unwind_fgetxattr(frame, cookie, this, op_ret, op_errno,
+                                    dictionary, xdata)
+                return 0
+
+        def setxattr_fop(self, frame, this, loc, dictionary, flags, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE SETXATTR FOP- {0:d}:  gfid={1:s}; path={2:s};"+
+                      " flags={3:d}").format(unique, gfid, loc.contents.path,
+                                             flags)
+                self.gfids[key] = gfid
+                dl.wind_setxattr(frame, POINTER(xlator_t)(), loc, dictionary,
+                                 flags, xdata)
+                return 0
+
+        def setxattr_cbk(self, frame, cookie, this, op_ret, op_errno, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE SETXATTR CBK- {0:d}: gfid={1:s}; "+
+                      "op_ret={2:d}; op_errno={3:d}").format(unique, gfid,
+                                                             op_ret, op_errno)
+                del self.gfids[key]
+                dl.unwind_setxattr(frame, cookie, this, op_ret, op_errno,
+                                   xdata)
+                return 0
+
+        def fsetxattr_fop(self, frame, this, fd, dictionary, flags, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(fd.contents.inode.contents.gfid)
+                print("GLUPY TRACE FSETXATTR FOP- {0:d}: gfid={1:s}; fd={2:p}; "+
+                      "flags={3:d}").format(unique, gfid, fd, flags)
+                self.gfids[key] = gfid
+                dl.wind_fsetxattr(frame, POINTER(xlator_t)(), fd, dictionary,
+                                  flags, xdata)
+                return 0
+
+        def fsetxattr_cbk(self, frame, cookie, this, op_ret, op_errno, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE FSETXATTR CBK- {0:d}: gfid={1:s};  "+
+                      "op_ret={2:d}; op_errno={3:d}").format(unique, gfid,
+                                                             op_ret, op_errno)
+                del self.gfids[key]
+                dl.unwind_fsetxattr(frame, cookie, this, op_ret, op_errno,
+                                   xdata)
+                return 0
+
+        def removexattr_fop(self, frame, this, loc, name, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE REMOVEXATTR FOP- {0:d}:  gfid={1:s}; "+
+                      "path={2:s}; name={3:s}").format(unique, gfid,
+                                                       loc.contents.path,
+                                                       name)
+                self.gfids[key] = gfid
+                dl.wind_removexattr(frame, POINTER(xlator_t)(), loc, name,
+                                    xdata)
+                return 0
+
+        def removexattr_cbk(self, frame, cookie, this, op_ret, op_errno,
+                            xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                print("GLUPY TRACE REMOVEXATTR CBK- {0:d}: gfid={1:s} "+
+                      " op_ret={2:d}; op_errno={3:d}").format(unique, gfid,
+                                                              op_ret, op_errno)
+                del self.gfids[key]
+                dl.unwind_removexattr(frame, cookie, this, op_ret, op_errno,
+                                      xdata)
+                return 0
+
+        def link_fop(self, frame, this, oldloc, newloc, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                if (newloc.contents.inode):
+                        newgfid = uuid2str(newloc.contents.inode.contents.gfid)
+                else:
+                        newgfid = "0"
+                oldgfid = uuid2str(oldloc.contents.inode.contents.gfid)
+                print("GLUPY TRACE LINK FOP-{0:d}: oldgfid={1:s}; oldpath={2:s};"+
+                      "newgfid={3:s};"+
+                      "newpath={4:s}").format(unique, oldgfid,
+                                              oldloc.contents.path,
+                                              newgfid,
+                                              newloc.contents.path)
+                self.gfids[key] =  oldgfid
+                dl.wind_link(frame, POINTER(xlator_t)(), oldloc, newloc,
+                             xdata)
+                return 0
+
+        def link_cbk(self, frame, cookie, this, op_ret, op_errno, inode, buf,
+                     preparent, postparent, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                if op_ret == 0:
+                        statstr = trace_stat2str(buf)
+                        preparentstr = trace_stat2str(preparent)
+                        postparentstr = trace_stat2str(postparent)
+                        print("GLUPY TRACE LINK CBK- {0:d}: op_ret={1:d} "+
+                              "*stbuf={2:s}; *prebuf={3:s}; "+
+                              "*postbuf={4:s} ").format(unique, op_ret, statstr,
+                                                        preparentstr,
+                                                        postparentstr)
+                else:
+                        print("GLUPY TRACE LINK CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; "+
+                              "op_errno={3:d}").format(unique, gfid,
+                                                       op_ret, op_errno)
+                del self.gfids[key]
+                dl.unwind_link(frame, cookie, this, op_ret, op_errno, inode,
+                               buf, preparent, postparent, xdata)
+                return 0
+
+        def unlink_fop(self, frame, this, loc, xflag, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE UNLINK FOP- {0:d}; gfid={1:s}; path={2:s}; "+
+                      "flag={3:d}").format(unique, gfid, loc.contents.path,
+                                           xflag)
+                self.gfids[key] = gfid
+                dl.wind_unlink(frame, POINTER(xlator_t)(), loc, xflag,
+                               xdata)
+                return 0
+
+        def unlink_cbk(self, frame, cookie, this, op_ret, op_errno,
+                       preparent, postparent, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                if op_ret == 0:
+                        preparentstr = trace_stat2str(preparent)
+                        postparentstr = trace_stat2str(postparent)
+                        print("GLUPY TRACE UNLINK CBK- {0:d}: gfid ={1:s}; "+
+                              "op_ret={2:d}; *prebuf={3:s}; "+
+                              "*postbuf={4:s} ").format(unique, gfid, op_ret,
+                                                        preparentstr,
+                                                        postparentstr)
+                else:
+                        print("GLUPY TRACE UNLINK CBK: {0:d}: gfid ={1:s}; "+
+                              "op_ret={2:d}; "+
+                              "op_errno={3:d}").format(unique, gfid, op_ret,
+                                                       op_errno)
+                del self.gfids[key]
+                dl.unwind_unlink(frame, cookie, this, op_ret, op_errno,
+                                 preparent, postparent, xdata)
+                return 0
+
+        def readlink_fop(self, frame, this, loc, size, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE READLINK FOP- {0:d}:  gfid={1:s}; path={2:s};"+
+                      " size={3:d}").format(unique, gfid, loc.contents.path,
+                                            size)
+                self.gfids[key] = gfid
+                dl.wind_readlink(frame, POINTER(xlator_t)(), loc, size,
+                               xdata)
+                return 0
+
+        def readlink_cbk(self, frame, cookie, this, op_ret, op_errno,
+                         buf, stbuf, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid  = self.gfids[key]
+                if op_ret == 0:
+                        statstr = trace_stat2str(stbuf)
+                        print("GLUPY TRACE READLINK CBK- {0:d}: gfid={1:s} "+
+                              " op_ret={2:d}; op_errno={3:d}; *prebuf={4:s}; "+
+                              "*postbuf={5:s} ").format(unique, gfid,
+                                                        op_ret, op_errno,
+                                                        buf, statstr)
+                else:
+                        print("GLUPY TRACE READLINK CBK- {0:d}: gfid={1:s} "+
+                              " op_ret={2:d}; op_errno={3:d}").format(unique,
+                                                                      gfid,
+                                                                      op_ret,
+                                                                      op_errno)
+                del self.gfids[key]
+                dl.unwind_readlink(frame, cookie, this, op_ret, op_errno, buf,
+                                   stbuf, xdata)
+                return 0
+
+        def symlink_fop(self, frame, this, linkpath, loc, umask, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = uuid2str(loc.contents.inode.contents.gfid)
+                print("GLUPY TRACE SYMLINK FOP- {0:d}: gfid={1:s}; "+
+                      "linkpath={2:s}; path={3:s};"+
+                      "umask=0{4:o}").format(unique, gfid, linkpath,
+                                             loc.contents.path, umask)
+                self.gfids[key] = gfid
+                dl.wind_symlink(frame, POINTER(xlator_t)(), linkpath, loc,
+                                umask, xdata)
+                return 0
+
+        def symlink_cbk(self, frame, cookie, this, op_ret, op_errno,
+                        inode, buf, preparent, postparent, xdata):
+                unique = dl.get_rootunique(frame)
+                key = dl.get_id(frame)
+                gfid = self.gfids[key]
+                if op_ret == 0:
+                        statstr = trace_stat2str(buf)
+                        preparentstr = trace_stat2str(preparent)
+                        postparentstr = trace_stat2str(postparent)
+                        print("GLUPY TRACE SYMLINK CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; *stbuf={3:s}; *preparent={4:s}; "+
+                              "*postparent={5:s}").format(unique, gfid,
+                                                          op_ret, statstr,
+                                                          preparentstr,
+                                                          postparentstr)
+                else:
+                        print("GLUPY TRACE SYMLINK CBK- {0:d}: gfid={1:s}; "+
+                              "op_ret={2:d}; op_errno={3:d}").format(unique,
+                                                                     gfid,
+                                                                     op_ret,
+                                                                     op_errno)
+                del self.gfids[key]
+                dl.unwind_symlink(frame, cookie, this, op_ret, op_errno,
+                                  inode, buf, preparent, postparent, xdata)
+                return 0
diff --git a/xlators/features/glupy/examples/helloworld.py b/xlators/features/glupy/examples/helloworld.py
new file mode 100644
index 000000000..b565a4e5b
--- /dev/null
+++ b/xlators/features/glupy/examples/helloworld.py
@@ -0,0 +1,19 @@
+import sys
+from gluster.glupy import *
+
+class xlator (Translator):
+
+    def __init__(self, c_this):
+        Translator.__init__(self, c_this)
+
+    def lookup_fop(self, frame, this, loc, xdata):
+        print "Python xlator: Hello!"
+        dl.wind_lookup(frame, POINTER(xlator_t)(), loc, xdata)
+        return 0
+
+    def lookup_cbk(self, frame, cookie, this, op_ret, op_errno, inode, buf,
+                   xdata, postparent):
+        print "Python xlator: Hello again!"
+        dl.unwind_lookup(frame, cookie, this, op_ret, op_errno, inode, buf,
+                         xdata, postparent)
+        return 0
diff --git a/xlators/features/glupy/examples/negative.py b/xlators/features/glupy/examples/negative.py
new file mode 100644
index 000000000..e7a4fc07c
--- /dev/null
+++ b/xlators/features/glupy/examples/negative.py
@@ -0,0 +1,91 @@
+import sys
+from uuid import UUID
+from gluster.glupy import *
+
+# Negative-lookup-caching example.  If a file wasn't there the last time we
+# looked, it's probably still not there.  This translator keeps track of
+# those failed lookups for us, and returns ENOENT without needing to pass the
+# call any further for repeated requests.
+
+# If we were doing this for real, we'd need separate caches for each xlator
+# instance.  The easiest way to do this would be to have xlator.__init__
+# "register" each instance in a module-global dict, with the key as the C
+# translator address and the value as the xlator object itself.  For testing
+# and teaching, it's sufficient just to have one cache.  The keys are parent
+# GFIDs, and the entries are lists of names within that parent that we know
+# don't exist.
+cache = {}
+
+# TBD: we need a better way of handling per-request data (frame->local in C).
+dl.get_id.restype = c_long
+dl.get_id.argtypes = [ POINTER(call_frame_t) ]
+
+def uuid2str (gfid):
+    return str(UUID(''.join(map("{0:02x}".format, gfid))))
+
+class xlator (Translator):
+
+    def __init__ (self, c_this):
+        self.requests = {}
+        Translator.__init__(self,c_this)
+
+    def lookup_fop (self, frame, this, loc, xdata):
+        pargfid = uuid2str(loc.contents.pargfid)
+        print "lookup FOP: %s:%s" % (pargfid, loc.contents.name)
+        # Check the cache.
+        if cache.has_key(pargfid):
+            if loc.contents.name in cache[pargfid]:
+                print "short-circuiting for %s:%s" % (pargfid,
+                    loc.contents.name)
+                dl.unwind_lookup(frame,0,this,-1,2,None,None,None,None)
+                return 0
+        key = dl.get_id(frame)
+        self.requests[key] = (pargfid, loc.contents.name[:])
+        # TBD: get real child xl from init, pass it here
+        dl.wind_lookup(frame,POINTER(xlator_t)(),loc,xdata)
+        return 0
+
+    def lookup_cbk (self, frame, cookie, this, op_ret, op_errno, inode, buf,
+                    xdata, postparent):
+        print "lookup CBK: %d (%d)" % (op_ret, op_errno)
+        key = dl.get_id(frame)
+        pargfid, name = self.requests[key]
+        # Update the cache.
+        if op_ret == 0:
+            print "found %s, removing from cache" % name
+            if cache.has_key(pargfid):
+                cache[pargfid].discard(name)
+        elif op_errno == 2:    # ENOENT
+            print "failed to find %s, adding to cache" % name
+            if cache.has_key(pargfid):
+                cache[pargfid].add(name)
+            else:
+                cache[pargfid] = set([name])
+        del self.requests[key]
+        dl.unwind_lookup(frame,cookie,this,op_ret,op_errno,
+                         inode,buf,xdata,postparent)
+        return 0
+
+    def create_fop (self, frame, this, loc, flags, mode, umask, fd, xdata):
+        pargfid = uuid2str(loc.contents.pargfid)
+        print "create FOP: %s:%s" % (pargfid, loc.contents.name)
+        key = dl.get_id(frame)
+        self.requests[key] = (pargfid, loc.contents.name[:])
+        # TBD: get real child xl from init, pass it here
+        dl.wind_create(frame,POINTER(xlator_t)(),loc,flags,mode,umask,fd,xdata)
+        return 0
+
+    def create_cbk (self, frame, cookie, this, op_ret, op_errno, fd, inode,
+                    buf, preparent, postparent, xdata):
+        print "create CBK: %d (%d)" % (op_ret, op_errno)
+        key = dl.get_id(frame)
+        pargfid, name = self.requests[key]
+        # Update the cache.
+        if op_ret == 0:
+            print "created %s, removing from cache" % name
+            if cache.has_key(pargfid):
+                cache[pargfid].discard(name)
+        del self.requests[key]
+        dl.unwind_create(frame,cookie,this,op_ret,op_errno,fd,inode,buf,
+                         preparent,postparent,xdata)
+        return 0
diff --git a/xlators/features/glupy/src/Makefile.am b/xlators/features/glupy/src/Makefile.am
new file mode 100644
index 000000000..ae7b6d14d
--- /dev/null
+++ b/xlators/features/glupy/src/Makefile.am
@@ -0,0 +1,21 @@
+xlator_LTLIBRARIES = glupy.la
+
+# Ensure GLUSTER_PYTHON_PATH is passed to glupy.so
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+glupydir = $(xlatordir)/glupy
+AM_CPPFLAGS = $(PYTHONDEV_CPPFLAGS) $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src -isystem $(BUILD_PYTHON_INC)
+AM_CFLAGS = $(PYTHONDEV_CFLAGS) -Wall -fno-strict-aliasing -DGLUSTER_PYTHON_PATH=\"$(glupydir)\" $(GF_CFLAGS)
+
+# Flags to build glupy.so with
+glupy_la_LDFLAGS = $(PYTHONDEV_LDFLAGS) -module -avoid-version -shared -nostartfiles
+glupy_la_SOURCES = glupy.c
+glupy_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la \
+	 -lpthread -l$(BUILD_PYTHON_LIB)
+
+noinst_HEADERS = glupy.h
+
+# Install glupy.py into the Python site-packages area
+pyglupydir = $(pythondir)/gluster
+pyglupy_PYTHON = glupy.py
+
+CLEANFILES =
diff --git a/xlators/features/glupy/src/glupy.c b/xlators/features/glupy/src/glupy.c
new file mode 100644
index 000000000..948b66f8d
--- /dev/null
+++ b/xlators/features/glupy/src/glupy.c
@@ -0,0 +1,2471 @@
+/*
+  Copyright (c) 2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  GlusterFS is free software; you can redistribute it and/or modify
+  it under the terms of the GNU General Public License as published
+  by the Free Software Foundation; either version 3 of the License,
+  or (at your option) any later version.
+
+  GlusterFS is distributed in the hope that it will be useful, but
+  WITHOUT ANY WARRANTY; without even the implied warranty of
+  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+  General Public License for more details.
+
+  You should have received a copy of the GNU General Public License
+  along with this program.  If not, see
+  <http://www.gnu.org/licenses/>.
+*/
+
+#include <ctype.h>
+#include <sys/uio.h>
+#include <Python.h>
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "xlator.h"
+#include "logging.h"
+#include "defaults.h"
+
+#include "glupy.h"
+
+/* UTILITY FUNCTIONS FOR FOP-SPECIFIC CODE */
+
+pthread_key_t gil_init_key;
+
+PyGILState_STATE
+glupy_enter (void)
+{
+#if 0
+        if (!pthread_getspecific(gil_init_key)) {
+                PyEval_ReleaseLock();
+                (void)pthread_setspecific(gil_init_key,(void *)1);
+        }
+#endif
+
+        return PyGILState_Ensure();
+}
+
+void
+glupy_leave (PyGILState_STATE gstate)
+{
+        PyGILState_Release(gstate);
+}
+
+/* FOP: LOOKUP */
+
+int32_t
+glupy_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, inode_t *inode,
+                  struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_LOOKUP]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_lookup_cbk_t)(priv->cbks[GLUPY_LOOKUP]))(
+                frame, cookie, this, op_ret, op_errno,
+                inode, buf, xdata, postparent);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
+                             xdata, postparent);
+        return 0;
+}
+
+int32_t
+glupy_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc,
+              dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_LOOKUP]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_lookup_t)(priv->fops[GLUPY_LOOKUP]))(
+                frame, this, loc, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_lookup_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, loc, xdata);
+        return 0;
+}
+
+void
+wind_lookup (call_frame_t *frame, xlator_t *xl, loc_t *loc, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND(frame,glupy_lookup_cbk,xl,xl->fops->lookup,loc,xdata);
+}
+
+void
+unwind_lookup (call_frame_t *frame, long cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno, inode_t *inode,
+               struct iatt *buf, dict_t *xdata, struct iatt *postparent)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT(lookup,frame,op_ret,op_errno,
+                            inode,buf,xdata,postparent);
+}
+
+void
+set_lookup_fop (long py_this, fop_lookup_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_LOOKUP] = (long)fop;
+}
+
+void
+set_lookup_cbk (long py_this, fop_lookup_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_LOOKUP] = (long)cbk;
+}
+
+/* FOP: CREATE */
+
+int32_t
+glupy_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+                  struct iatt *buf, struct iatt *preparent,
+                  struct iatt *postparent, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_CREATE]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_create_cbk_t)(priv->cbks[GLUPY_CREATE]))(
+                frame, cookie, this, op_ret, op_errno,
+                fd, inode, buf, preparent, postparent, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
+                             preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+glupy_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+              mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_CREATE]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_create_t)(priv->fops[GLUPY_CREATE]))(
+                frame, this, loc, flags, mode, umask, fd, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_create_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->create, loc, flags, mode, umask,
+                    fd, xdata);
+        return 0;
+}
+
+void
+wind_create (call_frame_t *frame, xlator_t *xl, loc_t *loc, int32_t flags,
+             mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_create_cbk,xl, xl->fops->create,
+                    loc, flags, mode, umask, fd, xdata);
+}
+
+void
+unwind_create (call_frame_t *frame, long cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+               struct iatt *buf, struct iatt *preparent,
+               struct iatt *postparent, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
+                             preparent, postparent, xdata);
+}
+
+void
+set_create_fop (long py_this, fop_create_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_CREATE] = (long)fop;
+}
+
+void
+set_create_cbk (long py_this, fop_create_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_CREATE] = (long)cbk;
+}
+
+/* FOP: OPEN */
+
+int32_t
+glupy_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_OPEN]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_open_cbk_t)(priv->cbks[GLUPY_OPEN]))(
+                frame, cookie, this, op_ret, op_errno,
+                fd, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
+        return 0;
+}
+
+int32_t
+glupy_open (call_frame_t *frame, xlator_t *this, loc_t *loc,
+            int32_t flags, fd_t *fd, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_OPEN]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_open_t)(priv->fops[GLUPY_OPEN]))(
+                frame, this, loc, flags, fd, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_open_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
+        return 0;
+}
+
+void
+wind_open (call_frame_t *frame, xlator_t *xl, loc_t *loc, int32_t flags,
+           fd_t *fd, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_open_cbk, xl, xl->fops->open, loc, flags,
+                    fd, xdata);
+}
+
+void
+unwind_open (call_frame_t *frame, long cookie, xlator_t *this,
+             int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
+}
+
+void
+set_open_fop (long py_this, fop_open_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+        priv->fops[GLUPY_OPEN] = (long)fop;
+}
+
+void
+set_open_cbk (long py_this, fop_open_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+        priv->cbks[GLUPY_OPEN] = (long)cbk;
+}
+
+/* FOP: READV */
+
+int32_t
+glupy_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, struct iovec *vector,
+                 int32_t count, struct iatt *stbuf, struct iobref *iobref,
+                 dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_READV]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_readv_cbk_t)(priv->cbks[GLUPY_READV]))(
+                frame, cookie, this, op_ret, op_errno,
+                vector, count, stbuf, iobref, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector,
+                             count, stbuf, iobref, xdata);
+        return 0;
+}
+
+int32_t
+glupy_readv (call_frame_t *frame, xlator_t *this, fd_t *fd,
+             size_t size, off_t offset, uint32_t flags, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_READV]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_readv_t)(priv->fops[GLUPY_READV]))(
+                frame, this, fd, size, offset, flags, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_readv_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readv, fd, size, offset,
+                    flags, xdata);
+        return 0;
+}
+
+void
+wind_readv (call_frame_t *frame, xlator_t *xl, fd_t *fd, size_t size,
+            off_t offset, uint32_t flags, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_readv_cbk, xl, xl->fops->readv, fd, size,
+                    offset, flags, xdata);
+}
+
+void
+unwind_readv (call_frame_t *frame, long cookie, xlator_t *this,
+              int32_t op_ret, int32_t op_errno, struct iovec *vector,
+              int32_t count, struct iatt *stbuf, struct iobref *iobref,
+              dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector,
+                             count, stbuf, iobref, xdata);
+}
+
+void
+set_readv_fop (long py_this, fop_readv_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+        priv->fops[GLUPY_READV] = (long)fop;
+}
+
+void
+set_readv_cbk (long py_this, fop_readv_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+        priv->cbks[GLUPY_READV] = (long)cbk;
+}
+
+/* FOP: WRITEV */
+
+int32_t
+glupy_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                  struct iatt *postbuf, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_WRITEV]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_writev_cbk_t)(priv->cbks[GLUPY_WRITEV]))(
+                frame, cookie, this, op_ret, op_errno,
+                prebuf, postbuf, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf,
+                             postbuf, xdata);
+        return 0;
+}
+
+int32_t
+glupy_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
+              struct iovec *vector, int32_t count, off_t offset,
+              uint32_t flags, struct iobref *iobref, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_WRITEV]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_writev_t)(priv->fops[GLUPY_WRITEV]))(
+                frame, this, fd, vector, count, offset, flags,
+                iobref, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_writev_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->writev, fd, vector, count,
+                    offset, flags, iobref, xdata);
+        return 0;
+}
+
+void
+wind_writev (call_frame_t *frame, xlator_t *xl, fd_t *fd, struct iovec *vector,
+             int32_t count, off_t offset, uint32_t flags, struct iobref *iobref,
+             dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_writev_cbk, xl, xl->fops->writev, fd, vector,
+                    count, offset, flags, iobref, xdata);
+}
+
+void
+unwind_writev (call_frame_t *frame, long cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+               struct iatt *postbuf, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf,
+                             postbuf, xdata);
+}
+
+void
+set_writev_fop (long py_this, fop_writev_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+        priv->fops[GLUPY_WRITEV] = (long)fop;
+}
+
+void
+set_writev_cbk (long py_this, fop_writev_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+        priv->cbks[GLUPY_WRITEV] = (long)cbk;
+}
+
+
+/* FOP: OPENDIR */
+
+int32_t
+glupy_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, fd_t *fd,
+                   dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_OPENDIR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_opendir_cbk_t)(priv->cbks[GLUPY_OPENDIR]))(
+                frame, cookie, this, op_ret, op_errno,
+                fd, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd, xdata);
+        return 0;
+}
+
+int32_t
+glupy_opendir (call_frame_t *frame, xlator_t *this, loc_t *loc,
+               fd_t *fd, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_OPENDIR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_opendir_t)(priv->fops[GLUPY_OPENDIR]))(
+                frame, this, loc, fd, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_opendir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->opendir, loc, fd, xdata);
+        return 0;
+}
+
+void
+wind_opendir (call_frame_t *frame, xlator_t *xl, loc_t *loc, fd_t *fd, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND(frame,glupy_opendir_cbk,xl,xl->fops->opendir,loc,fd,xdata);
+}
+
+void
+unwind_opendir (call_frame_t *frame, long cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT(opendir,frame,op_ret,op_errno,
+                            fd,xdata);
+}
+
+void
+set_opendir_fop (long py_this, fop_opendir_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_OPENDIR] = (long)fop;
+}
+
+void
+set_opendir_cbk (long py_this, fop_opendir_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_OPENDIR] = (long)cbk;
+}
+
+/* FOP: READDIR */
+
+int32_t
+glupy_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+                   dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_READDIR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_readdir_cbk_t)(priv->cbks[GLUPY_READDIR]))(
+                frame, cookie, this, op_ret, op_errno,
+                entries,  xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, entries,
+                             xdata);
+        return 0;
+}
+
+int32_t
+glupy_readdir (call_frame_t *frame, xlator_t *this, fd_t *fd,
+               size_t size, off_t offset, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_READDIR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_readdir_t)(priv->fops[GLUPY_READDIR]))(
+                frame, this, fd, size, offset, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_readdir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readdir,fd, size, offset, xdata);
+        return 0;
+}
+
+void
+wind_readdir(call_frame_t *frame, xlator_t *xl, fd_t *fd, size_t size,
+             off_t offset, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND(frame,glupy_readdir_cbk,xl,xl->fops->readdir,fd,size,offset,xdata);
+}
+
+void
+unwind_readdir (call_frame_t *frame, long cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+                dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT(readdir,frame,op_ret,op_errno,
+                            entries, xdata);
+}
+
+void
+set_readdir_fop (long py_this, fop_readdir_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_READDIR] = (long)fop;
+}
+
+void
+set_readdir_cbk (long py_this, fop_readdir_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_READDIR] = (long)cbk;
+}
+
+
+/* FOP: READDIRP */
+
+int32_t
+glupy_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+                    dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_READDIRP]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_readdirp_cbk_t)(priv->cbks[GLUPY_READDIRP]))(
+                frame, cookie, this, op_ret, op_errno,
+                entries,  xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries,
+                             xdata);
+        return 0;
+}
+
+int32_t
+glupy_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                size_t size, off_t offset, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_READDIRP]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_readdirp_t)(priv->fops[GLUPY_READDIRP]))(
+                frame, this, fd, size, offset, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_readdirp_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readdirp,fd, size, offset, xdata);
+        return 0;
+}
+
+void
+wind_readdirp (call_frame_t *frame, xlator_t *xl, fd_t *fd, size_t size,
+               off_t offset, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND(frame,glupy_readdirp_cbk,xl,xl->fops->readdirp,fd,size,offset,xdata);
+}
+
+void
+unwind_readdirp (call_frame_t *frame, long cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+                 dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT(readdirp,frame,op_ret,op_errno,
+                            entries, xdata);
+}
+
+void
+set_readdirp_fop (long py_this, fop_readdirp_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_READDIRP] = (long)fop;
+}
+
+void
+set_readdirp_cbk (long py_this, fop_readdirp_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_READDIRP] = (long)cbk;
+}
+
+
+/* FOP:STAT */
+
+int32_t
+glupy_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_STAT]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_stat_cbk_t)(priv->cbks[GLUPY_STAT]))(
+                frame, cookie, this, op_ret, op_errno,
+                buf, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (stat, frame, op_ret, op_errno, buf, xdata);
+        return 0;
+}
+
+int32_t
+glupy_stat (call_frame_t *frame, xlator_t *this, loc_t *loc,
+            dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_STAT]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_stat_t)(priv->fops[GLUPY_STAT]))(
+                frame, this, loc, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_stat_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->stat, loc, xdata);
+        return 0;
+}
+
+void
+wind_stat (call_frame_t *frame, xlator_t *xl, loc_t *loc, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND(frame,glupy_stat_cbk,xl,xl->fops->stat,loc,xdata);
+}
+
+void
+unwind_stat (call_frame_t *frame, long cookie, xlator_t *this,
+             int32_t op_ret, int32_t op_errno, struct iatt *buf,
+             dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT(stat,frame,op_ret,op_errno,
+                            buf,xdata);
+}
+
+void
+set_stat_fop (long py_this, fop_stat_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_STAT] = (long)fop;
+}
+
+void
+set_stat_cbk (long py_this, fop_stat_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_STAT] = (long)cbk;
+}
+
+
+/* FOP: FSTAT */
+
+int32_t
+glupy_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_FSTAT]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_fstat_cbk_t)(priv->cbks[GLUPY_FSTAT]))(
+                frame, cookie, this, op_ret, op_errno,
+                buf, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, buf, xdata);
+        return 0;
+}
+
+int32_t
+glupy_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd,
+             dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_FSTAT]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_fstat_t)(priv->fops[GLUPY_FSTAT]))(
+                frame, this, fd, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_fstat_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fstat, fd, xdata);
+        return 0;
+}
+
+void
+wind_fstat (call_frame_t *frame, xlator_t *xl, fd_t *fd, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND(frame,glupy_fstat_cbk,xl,xl->fops->fstat,fd,xdata);
+}
+
+void
+unwind_fstat (call_frame_t *frame, long cookie, xlator_t *this,
+              int32_t op_ret, int32_t op_errno, struct iatt *buf,
+              dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT(fstat,frame,op_ret,op_errno,
+                            buf,xdata);
+}
+
+void
+set_fstat_fop (long py_this, fop_fstat_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_FSTAT] = (long)fop;
+}
+
+void
+set_fstat_cbk (long py_this, fop_fstat_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_FSTAT] = (long)cbk;
+}
+
+/* FOP:STATFS */
+
+int32_t
+glupy_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, struct statvfs *buf, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_STATFS]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_statfs_cbk_t)(priv->cbks[GLUPY_STATFS]))(
+                frame, cookie, this, op_ret, op_errno,
+                buf, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, buf, xdata);
+        return 0;
+}
+
+int32_t
+glupy_statfs (call_frame_t *frame, xlator_t *this, loc_t *loc,
+              dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_STATFS]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_statfs_t)(priv->fops[GLUPY_STATFS]))(
+                frame, this, loc, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_statfs_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->statfs, loc, xdata);
+        return 0;
+}
+
+void
+wind_statfs (call_frame_t *frame, xlator_t *xl, loc_t *loc, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND(frame,glupy_statfs_cbk,xl,xl->fops->statfs,loc,xdata);
+}
+
+void
+unwind_statfs (call_frame_t *frame, long cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno, struct statvfs *buf,
+               dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT(statfs,frame,op_ret,op_errno,
+                            buf,xdata);
+}
+
+void
+set_statfs_fop (long py_this, fop_statfs_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_STATFS] = (long)fop;
+}
+
+void
+set_statfs_cbk (long py_this, fop_statfs_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_STATFS] = (long)cbk;
+}
+
+
+/* FOP: SETXATTR */
+
+int32_t
+glupy_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_SETXATTR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_setxattr_cbk_t)(priv->cbks[GLUPY_SETXATTR]))(
+                frame, cookie, this, op_ret, op_errno,
+                xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int32_t
+glupy_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                dict_t *dict, int32_t flags, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_SETXATTR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_setxattr_t)(priv->fops[GLUPY_SETXATTR]))(
+                frame, this, loc, dict, flags, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_setxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr, loc, dict,
+                    flags, xdata);
+        return 0;
+}
+
+void
+wind_setxattr (call_frame_t *frame, xlator_t *xl, loc_t *loc,
+               dict_t *dict, int32_t flags, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_setxattr_cbk, xl, xl->fops->setxattr,
+                    loc, dict, flags, xdata);
+}
+
+
+void
+unwind_setxattr (call_frame_t *frame, long cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
+
+}
+
+void
+set_setxattr_fop (long py_this, fop_setxattr_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_SETXATTR] = (long)fop;
+}
+
+void
+set_setxattr_cbk (long py_this, fop_setxattr_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_SETXATTR] = (long)cbk;
+}
+
+/* FOP: GETXATTR */
+
+int32_t
+glupy_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *dict,
+                    dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_GETXATTR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_getxattr_cbk_t)(priv->cbks[GLUPY_GETXATTR]))(
+                frame, cookie, this, op_ret, op_errno, dict,
+                xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict,
+                             xdata);
+        return 0;
+}
+
+int32_t
+glupy_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                const char *name, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_GETXATTR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_getxattr_t)(priv->fops[GLUPY_GETXATTR]))(
+                frame, this, loc, name, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_getxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->getxattr, loc, name,
+                    xdata);
+        return 0;
+}
+
+void
+wind_getxattr (call_frame_t *frame, xlator_t *xl, loc_t *loc,
+               const char *name, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_getxattr_cbk, xl, xl->fops->getxattr,
+                    loc, name, xdata);
+}
+
+
+void
+unwind_getxattr (call_frame_t *frame, long cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, dict_t *dict,
+                 dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict,
+                             xdata);
+
+}
+
+
+void
+set_getxattr_fop (long py_this, fop_getxattr_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_GETXATTR] = (long)fop;
+}
+
+
+void
+set_getxattr_cbk (long py_this, fop_getxattr_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_GETXATTR] = (long)cbk;
+}
+
+/* FOP: FSETXATTR */
+
+int32_t
+glupy_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_FSETXATTR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_fsetxattr_cbk_t)(priv->cbks[GLUPY_FSETXATTR]))(
+                frame, cookie, this, op_ret, op_errno,
+                xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int32_t
+glupy_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                 dict_t *dict, int32_t flags, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_FSETXATTR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_fsetxattr_t)(priv->fops[GLUPY_FSETXATTR]))(
+                frame, this, fd, dict, flags, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_fsetxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsetxattr, fd, dict,
+                    flags, xdata);
+        return 0;
+}
+
+void
+wind_fsetxattr (call_frame_t *frame, xlator_t *xl, fd_t *fd,
+                dict_t *dict, int32_t flags, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_fsetxattr_cbk, xl, xl->fops->fsetxattr,
+                    fd, dict, flags, xdata);
+}
+
+
+void
+unwind_fsetxattr (call_frame_t *frame, long cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
+
+}
+
+void
+set_fsetxattr_fop (long py_this, fop_fsetxattr_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_FSETXATTR] = (long)fop;
+}
+
+void
+set_fsetxattr_cbk (long py_this, fop_fsetxattr_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_FSETXATTR] = (long)cbk;
+}
+
+/* FOP: FGETXATTR */
+
+int32_t
+glupy_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *dict,
+                     dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_FGETXATTR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_fgetxattr_cbk_t)(priv->cbks[GLUPY_FGETXATTR]))(
+                frame, cookie, this, op_ret, op_errno, dict,
+                xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict,
+                             xdata);
+        return 0;
+}
+
+int32_t
+glupy_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                 const char *name, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_FGETXATTR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_fgetxattr_t)(priv->fops[GLUPY_FGETXATTR]))(
+                frame, this, fd, name, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_fgetxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fgetxattr, fd, name,
+                    xdata);
+        return 0;
+}
+
+void
+wind_fgetxattr (call_frame_t *frame, xlator_t *xl, fd_t *fd,
+                const char *name, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_fgetxattr_cbk, xl, xl->fops->fgetxattr,
+                    fd, name, xdata);
+}
+
+
+void
+unwind_fgetxattr (call_frame_t *frame, long cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, dict_t *dict,
+                  dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict,
+                             xdata);
+
+}
+
+
+void
+set_fgetxattr_fop (long py_this, fop_fgetxattr_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_FGETXATTR] = (long)fop;
+}
+
+
+void
+set_fgetxattr_cbk (long py_this, fop_fgetxattr_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_FGETXATTR] = (long)cbk;
+}
+
+/* FOP:REMOVEXATTR */
+
+int32_t
+glupy_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_REMOVEXATTR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_removexattr_cbk_t)(priv->cbks[GLUPY_REMOVEXATTR]))(
+                frame, cookie, this, op_ret, op_errno, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int32_t
+glupy_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                   const char *name, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_REMOVEXATTR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_removexattr_t)(priv->fops[GLUPY_REMOVEXATTR]))(
+                frame, this, loc, name, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_removexattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->removexattr, loc, name,
+                    xdata);
+        return 0;
+}
+
+void
+wind_removexattr (call_frame_t *frame, xlator_t *xl, loc_t *loc,
+                  const char *name, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_removexattr_cbk, xl, xl->fops->removexattr,
+                    loc, name, xdata);
+}
+
+
+void
+unwind_removexattr (call_frame_t *frame, long cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
+
+}
+
+void
+set_removexattr_fop (long py_this, fop_removexattr_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_REMOVEXATTR] = (long)fop;
+}
+
+void
+set_removexattr_cbk (long py_this, fop_removexattr_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_REMOVEXATTR] = (long)cbk;
+}
+
+
+/* FOP:FREMOVEXATTR */
+
+int32_t
+glupy_fremovexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_FREMOVEXATTR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_fremovexattr_cbk_t)(priv->cbks[GLUPY_FREMOVEXATTR]))(
+                frame, cookie, this, op_ret, op_errno, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (fremovexattr, frame, op_ret, op_errno, xdata);
+        return 0;
+}
+
+int32_t
+glupy_fremovexattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                   const char *name, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_FREMOVEXATTR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_fremovexattr_t)(priv->fops[GLUPY_FREMOVEXATTR]))(
+                frame, this, fd, name, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_fremovexattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fremovexattr, fd, name,
+                    xdata);
+        return 0;
+}
+
+void
+wind_fremovexattr (call_frame_t *frame, xlator_t *xl, fd_t *fd,
+                   const char *name, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_fremovexattr_cbk, xl, xl->fops->fremovexattr,
+                    fd, name, xdata);
+}
+
+
+void
+unwind_fremovexattr (call_frame_t *frame, long cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (fremovexattr, frame, op_ret, op_errno, xdata);
+
+}
+
+void
+set_fremovexattr_fop (long py_this, fop_fremovexattr_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_FREMOVEXATTR] = (long)fop;
+}
+
+void
+set_fremovexattr_cbk (long py_this, fop_fremovexattr_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_FREMOVEXATTR] = (long)cbk;
+}
+
+
+/* FOP: LINK*/
+int32_t
+glupy_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, inode_t *inode,
+                struct iatt *buf, struct iatt *preparent,
+                struct iatt *postparent, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_LINK]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_link_cbk_t)(priv->cbks[GLUPY_LINK]))(
+                frame, cookie, this, op_ret, op_errno,
+                inode, buf, preparent, postparent, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+glupy_link(call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+           dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_LINK]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_link_t)(priv->fops[GLUPY_LINK]))(
+                frame, this, oldloc, newloc, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_link_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->link, oldloc, newloc,
+                    xdata);
+        return 0;
+}
+
+void
+wind_link (call_frame_t *frame, xlator_t *xl, loc_t *oldloc, loc_t *newloc,
+           dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_link_cbk, xl, xl->fops->link,
+                    oldloc, newloc, xdata);
+}
+
+void
+unwind_link (call_frame_t *frame, long cookie, xlator_t *this,
+             int32_t op_ret, int32_t op_errno, inode_t *inode,
+             struct iatt *buf, struct iatt *preparent,
+             struct iatt *postparent, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+}
+
+void
+set_link_fop (long py_this, fop_link_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_LINK] = (long)fop;
+}
+
+void
+set_link_cbk (long py_this, fop_link_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_LINK] = (long)cbk;
+}
+
+/* FOP: SYMLINK*/
+int32_t
+glupy_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, inode_t *inode,
+                   struct iatt *buf, struct iatt *preparent,
+                   struct iatt *postparent, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_SYMLINK]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_symlink_cbk_t)(priv->cbks[GLUPY_SYMLINK]))(
+                frame, cookie, this, op_ret, op_errno,
+                inode, buf, preparent, postparent, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+glupy_symlink(call_frame_t *frame, xlator_t *this, const char *linkname,
+              loc_t *loc, mode_t umask, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_SYMLINK]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_symlink_t)(priv->fops[GLUPY_SYMLINK]))(
+                frame, this, linkname, loc, umask, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_symlink_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->symlink, linkname, loc,
+                    umask, xdata);
+        return 0;
+}
+
+void
+wind_symlink (call_frame_t *frame, xlator_t *xl, const char *linkname,
+              loc_t *loc, mode_t umask, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_symlink_cbk, xl, xl->fops->symlink,
+                    linkname, loc, umask, xdata);
+}
+
+void
+unwind_symlink (call_frame_t *frame, long cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, inode_t *inode,
+                struct iatt *buf, struct iatt *preparent,
+                struct iatt *postparent, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+}
+
+void
+set_symlink_fop (long py_this, fop_symlink_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_SYMLINK] = (long)fop;
+}
+
+void
+set_symlink_cbk (long py_this, fop_symlink_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_SYMLINK] = (long)cbk;
+}
+
+
+/* FOP: READLINK */
+int32_t
+glupy_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, const char *path,
+                    struct iatt *buf, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_READLINK]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_readlink_cbk_t)(priv->cbks[GLUPY_READLINK]))(
+                frame, cookie, this, op_ret, op_errno,
+                path, buf, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, path,
+                             buf, xdata);
+        return 0;
+}
+
+int32_t
+glupy_readlink (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                size_t size, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_READLINK]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_readlink_t)(priv->fops[GLUPY_READLINK]))(
+                frame, this, loc, size, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_readlink_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readlink, loc,
+                    size, xdata);
+        return 0;
+}
+
+void
+wind_readlink (call_frame_t *frame, xlator_t *xl, loc_t *loc,
+               size_t size, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_readlink_cbk, xl, xl->fops->readlink,
+                    loc, size, xdata);
+}
+
+void
+unwind_readlink (call_frame_t *frame, long cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, const char *path,
+                 struct iatt *buf, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, path, buf,
+                             xdata);
+}
+
+void
+set_readlink_fop (long py_this, fop_readlink_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_READLINK] = (long)fop;
+}
+
+void
+set_readlink_cbk (long py_this, fop_readlink_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_READLINK] = (long)cbk;
+}
+
+
+/* FOP: UNLINK */
+
+int32_t
+glupy_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno,
+                  struct iatt *preparent, struct iatt *postparent,
+                  dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_UNLINK]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_unlink_cbk_t)(priv->cbks[GLUPY_UNLINK]))(
+                frame, cookie, this, op_ret, op_errno,
+                preparent, postparent, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno, preparent,
+                             postparent, xdata);
+        return 0;
+}
+
+int32_t
+glupy_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc,
+              int xflags, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_UNLINK]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_unlink_t)(priv->fops[GLUPY_UNLINK]))(
+                frame, this, loc, xflags, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_unlink_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->unlink, loc,
+                    xflags, xdata);
+        return 0;
+}
+
+void
+wind_unlink (call_frame_t *frame, xlator_t *xl, loc_t *loc,
+             int xflags, dict_t *xdata)
+{
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_unlink_cbk, xl, xl->fops->unlink,
+                    loc, xflags, xdata);
+}
+
+void
+unwind_unlink (call_frame_t *frame, long cookie, xlator_t *this,
+               int32_t op_ret, int32_t op_errno,
+               struct iatt *preparent, struct iatt *postparent,
+               dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno,
+                             preparent, postparent, xdata);
+}
+
+void
+set_unlink_fop (long py_this, fop_unlink_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_UNLINK] = (long)fop;
+}
+
+void
+set_unlink_cbk (long py_this, fop_unlink_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_UNLINK] = (long)cbk;
+}
+
+
+/* FOP: MKDIR */
+
+int32_t
+glupy_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, inode_t *inode,
+                 struct iatt *buf, struct iatt *preparent,
+                 struct iatt *postparent, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_MKDIR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_mkdir_cbk_t)(priv->cbks[GLUPY_MKDIR]))(
+                frame, cookie, this, op_ret, op_errno,
+                inode, buf, preparent, postparent, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+        return 0;
+}
+
+int32_t
+glupy_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+             mode_t umask, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_MKDIR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_mkdir_t)(priv->fops[GLUPY_MKDIR]))(
+                frame, this, loc, mode, umask, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_mkdir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->mkdir, loc, mode, umask,
+                    xdata);
+        return 0;
+}
+
+void
+wind_mkdir (call_frame_t *frame, xlator_t *xl, loc_t *loc, mode_t mode,
+            mode_t umask, dict_t *xdata)
+{
+
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_mkdir_cbk, xl, xl->fops->mkdir,
+                    loc, mode, umask, xdata);
+}
+
+void
+unwind_mkdir (call_frame_t *frame, long cookie, xlator_t *this,
+              int32_t op_ret, int32_t op_errno, inode_t *inode,
+              struct iatt *buf, struct iatt *preparent,
+              struct iatt *postparent, dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+}
+
+void
+set_mkdir_fop (long py_this, fop_mkdir_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_MKDIR] = (long)fop;
+}
+
+void
+set_mkdir_cbk (long py_this, fop_mkdir_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_MKDIR] = (long)cbk;
+}
+
+
+/* FOP: RMDIR */
+
+int32_t
+glupy_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno,
+                 struct iatt *preparent, struct iatt *postparent,
+                 dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+
+        if (!priv->cbks[GLUPY_RMDIR]) {
+                goto unwind;
+        }
+
+        gstate = glupy_enter();
+        ret = ((fop_rmdir_cbk_t)(priv->cbks[GLUPY_RMDIR]))(
+                frame, cookie, this, op_ret, op_errno,
+                preparent, postparent, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno, preparent,
+                             postparent, xdata);
+        return 0;
+}
+
+int32_t
+glupy_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc,
+             int xflags, dict_t *xdata)
+{
+        glupy_private_t *priv = this->private;
+        PyGILState_STATE gstate;
+        int32_t ret;
+        static long next_id = 0;
+
+        if (!priv->fops[GLUPY_RMDIR]) {
+                goto wind;
+        }
+
+        gstate = glupy_enter();
+        frame->local = (void *)++next_id;
+        ret = ((fop_rmdir_t)(priv->fops[GLUPY_RMDIR]))(
+                frame, this, loc, xflags, xdata);
+        glupy_leave(gstate);
+
+        return ret;
+
+wind:
+        STACK_WIND (frame, glupy_rmdir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->rmdir, loc,
+                    xflags, xdata);
+        return 0;
+}
+
+void
+wind_rmdir (call_frame_t *frame, xlator_t *xl, loc_t *loc,
+            int xflags, dict_t *xdata)
+{
+
+        xlator_t        *this = THIS;
+
+        if (!xl || (xl == this)) {
+                xl = FIRST_CHILD(this);
+        }
+
+        STACK_WIND (frame, glupy_rmdir_cbk, xl, xl->fops->rmdir,
+                    loc, xflags, xdata);
+}
+
+void
+unwind_rmdir (call_frame_t *frame, long cookie, xlator_t *this,
+              int32_t op_ret, int32_t op_errno,
+              struct iatt *preparent, struct iatt *postparent,
+              dict_t *xdata)
+{
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno,
+                             preparent, postparent, xdata);
+}
+
+void
+set_rmdir_fop (long py_this, fop_rmdir_t fop)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->fops[GLUPY_RMDIR] = (long)fop;
+}
+
+void
+set_rmdir_cbk (long py_this, fop_rmdir_cbk_t cbk)
+{
+        glupy_private_t *priv   = ((xlator_t *)py_this)->private;
+
+        priv->cbks[GLUPY_RMDIR] = (long)cbk;
+}
+
+
+/* NON-FOP-SPECIFIC CODE */
+
+
+long
+get_id (call_frame_t *frame)
+{
+        return (long)(frame->local);
+}
+
+uint64_t
+get_rootunique (call_frame_t *frame)
+{
+        return frame->root->unique;
+}
+
+int32_t
+init (xlator_t *this)
+{
+        glupy_private_t         *priv           = NULL;
+        char                    *module_name    = NULL;
+        PyObject                *py_mod_name    = NULL;
+        PyObject                *py_init_func   = NULL;
+        PyObject                *py_args        = NULL;
+        PyObject                *syspath        = NULL;
+	PyObject                *path           = NULL;
+	static gf_boolean_t      py_inited      = _gf_false;
+        void *                   err_cleanup    = &&err_return;
+
+        if (dict_get_str(this->options,"module-name",&module_name) != 0) {
+                gf_log (this->name, GF_LOG_ERROR, "missing module-name");
+                return -1;
+        }
+
+	priv = GF_CALLOC (1, sizeof (glupy_private_t), gf_glupy_mt_priv);
+        if (!priv) {
+                goto *err_cleanup;
+        }
+        this->private = priv;
+        err_cleanup = &&err_free_priv;
+
+	if (!py_inited) {
+                Py_Initialize();
+                PyEval_InitThreads();
+#if 0
+                (void)pthread_key_create(&gil_init_key,NULL);
+                (void)pthread_setspecific(gil_init_key,(void *)1);
+#endif
+                /* PyEval_InitThreads takes this "for" us.  No thanks. */
+                PyEval_ReleaseLock();
+                py_inited = _gf_true;
+        }
+
+	/* Adjust python's path */
+	syspath = PySys_GetObject("path");
+	path = PyString_FromString(GLUSTER_PYTHON_PATH);
+	PyList_Append(syspath, path);
+	Py_DECREF(path);
+
+	py_mod_name = PyString_FromString(module_name);
+        if (!py_mod_name) {
+                gf_log (this->name, GF_LOG_ERROR, "could not create name");
+                if (PyErr_Occurred()) {
+                        PyErr_Print();
+                }
+                goto *err_cleanup;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "py_mod_name = %s", module_name);
+        priv->py_module = PyImport_Import(py_mod_name);
+        Py_DECREF(py_mod_name);
+        if (!priv->py_module) {
+                gf_log (this->name, GF_LOG_ERROR, "Python import failed");
+                if (PyErr_Occurred()) {
+                        PyErr_Print();
+                }
+                goto *err_cleanup;
+        }
+        gf_log (this->name, GF_LOG_INFO, "Import of %s succeeded", module_name);
+        err_cleanup = &&err_deref_module;
+
+        py_init_func = PyObject_GetAttrString(priv->py_module, "xlator");
+        if (!py_init_func || !PyCallable_Check(py_init_func)) {
+                gf_log (this->name, GF_LOG_ERROR, "missing init func");
+                if (PyErr_Occurred()) {
+                        PyErr_Print();
+                }
+                goto *err_cleanup;
+        }
+        err_cleanup = &&err_deref_init;
+
+        py_args = PyTuple_New(1);
+        if (!py_args) {
+                gf_log (this->name, GF_LOG_ERROR, "could not create args");
+                if (PyErr_Occurred()) {
+                        PyErr_Print();
+                }
+                goto *err_cleanup;
+        }
+        PyTuple_SetItem(py_args,0,PyLong_FromLong((long)this));
+
+        /* TBD: pass in list of children */
+        priv->py_xlator = PyObject_CallObject(py_init_func, py_args);
+        Py_DECREF(py_args);
+        if (!priv->py_xlator) {
+                gf_log (this->name, GF_LOG_ERROR, "Python init failed");
+                if (PyErr_Occurred()) {
+                        PyErr_Print();
+                }
+                goto *err_cleanup;
+        }
+        gf_log (this->name, GF_LOG_DEBUG, "init returned %p", priv->py_xlator);
+
+        return 0;
+
+err_deref_init:
+        Py_DECREF(py_init_func);
+err_deref_module:
+        Py_DECREF(priv->py_module);
+err_free_priv:
+        GF_FREE(priv);
+err_return:
+        return -1;
+}
+
+void
+fini (xlator_t *this)
+{
+        glupy_private_t *priv = this->private;
+
+        if (!priv)
+                return;
+        Py_DECREF(priv->py_xlator);
+        Py_DECREF(priv->py_module);
+        this->private = NULL;
+        GF_FREE (priv);
+
+        return;
+}
+
+struct xlator_fops fops = {
+        .lookup       = glupy_lookup,
+        .create       = glupy_create,
+        .open         = glupy_open,
+        .readv        = glupy_readv,
+        .writev       = glupy_writev,
+        .opendir      = glupy_opendir,
+        .readdir      = glupy_readdir,
+        .stat         = glupy_stat,
+        .fstat        = glupy_fstat,
+        .setxattr     = glupy_setxattr,
+        .getxattr     = glupy_getxattr,
+        .fsetxattr    = glupy_fsetxattr,
+        .fgetxattr    = glupy_fgetxattr,
+        .removexattr  = glupy_removexattr,
+        .fremovexattr = glupy_fremovexattr,
+        .link         = glupy_link,
+        .unlink       = glupy_unlink,
+        .readlink     = glupy_readlink,
+        .symlink      = glupy_symlink,
+        .mkdir        = glupy_mkdir,
+        .rmdir        = glupy_rmdir,
+        .statfs       = glupy_statfs,
+        .readdirp     = glupy_readdirp
+};
+
+struct xlator_cbks cbks = {
+};
+
+struct volume_options options[] = {
+        { .key  = {NULL} },
+};
diff --git a/xlators/features/glupy/src/glupy.h b/xlators/features/glupy/src/glupy.h
new file mode 100644
index 000000000..8661fce88
--- /dev/null
+++ b/xlators/features/glupy/src/glupy.h
@@ -0,0 +1,69 @@
+/*
+   Copyright (c) 2006-2011 Gluster, Inc. <http://www.gluster.com>
+   This file is part of GlusterFS.
+
+   GlusterFS is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published
+   by the Free Software Foundation; either version 3 of the License,
+   or (at your option) any later version.
+
+   GlusterFS is distributed in the hope that it will be useful, but
+   WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   General Public License for more details.
+
+   You should have received a copy of the GNU General Public License
+   along with this program.  If not, see
+   <http://www.gnu.org/licenses/>.
+*/
+
+#ifndef __GLUPY_H__
+#define __GLUPY_H__
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+#include "mem-types.h"
+
+enum {
+        GLUPY_LOOKUP = 0,
+        GLUPY_CREATE,
+        GLUPY_OPEN,
+        GLUPY_READV,
+        GLUPY_WRITEV,
+        GLUPY_OPENDIR,
+        GLUPY_READDIR,
+        GLUPY_READDIRP,
+        GLUPY_STAT,
+        GLUPY_FSTAT,
+        GLUPY_STATFS,
+        GLUPY_SETXATTR,
+        GLUPY_GETXATTR,
+        GLUPY_FSETXATTR,
+        GLUPY_FGETXATTR,
+        GLUPY_REMOVEXATTR,
+        GLUPY_FREMOVEXATTR,
+        GLUPY_LINK,
+        GLUPY_UNLINK,
+        GLUPY_READLINK,
+        GLUPY_SYMLINK,
+        GLUPY_MKNOD,
+        GLUPY_MKDIR,
+        GLUPY_RMDIR,
+        GLUPY_N_FUNCS
+};
+
+typedef struct {
+        PyObject        *py_module;
+        PyObject        *py_xlator;
+        long            fops[GLUPY_N_FUNCS];
+        long            cbks[GLUPY_N_FUNCS];
+} glupy_private_t;
+
+enum gf_glupy_mem_types_ {
+        gf_glupy_mt_priv = gf_common_mt_end + 1,
+        gf_glupy_mt_end
+};
+
+#endif /* __GLUPY_H__ */
diff --git a/xlators/features/glupy/src/glupy.py b/xlators/features/glupy/src/glupy.py
new file mode 100644
index 000000000..a5daa77d3
--- /dev/null
+++ b/xlators/features/glupy/src/glupy.py
@@ -0,0 +1,841 @@
+import sys
+from ctypes import *
+
+dl = CDLL("",RTLD_GLOBAL)
+
+
+class call_frame_t (Structure):
+        pass
+
+class dev_t (Structure):
+        pass
+
+
+class dict_t (Structure):
+        pass
+
+
+class gf_dirent_t (Structure):
+        pass
+
+
+class iobref_t (Structure):
+	pass
+
+
+class iovec_t (Structure):
+        pass
+
+
+class list_head (Structure):
+        pass
+
+list_head._fields_ = [
+                ("next", POINTER(list_head)),
+                ("prev", POINTER(list_head))
+        ]
+
+
+class rwxperm_t (Structure):
+        _fields_ = [
+                ("read", c_uint8, 1),
+                ("write", c_uint8, 1),
+                ("execn", c_uint8, 1)
+        ]
+
+
+class statvfs_t (Structure):
+        pass
+
+
+class xlator_t (Structure):
+        pass
+
+
+class ia_prot_t (Structure):
+        _fields_ = [
+                ("suid", c_uint8, 1),
+                ("sgid", c_uint8, 1),
+                ("sticky", c_uint8, 1),
+                ("owner", rwxperm_t),
+                ("group", rwxperm_t),
+                ("other", rwxperm_t)
+        ]
+
+# For checking file type.
+(IA_INVAL, IA_IFREG, IA_IFDIR, IA_IFLNK, IA_IFBLK, IA_IFCHR, IA_IFIFO,
+ IA_IFSOCK) = xrange(8)
+
+
+class iatt_t (Structure):
+        _fields_ = [
+                ("ia_no", c_uint64),
+                ("ia_gfid", c_ubyte * 16),
+                ("ia_dev", c_uint64),
+                ("ia_type", c_uint),
+                ("ia_prot", ia_prot_t),
+                ("ia_nlink", c_uint32),
+                ("ia_uid", c_uint32),
+                ("ia_gid", c_uint32),
+                ("ia_rdev", c_uint64),
+                ("ia_size", c_uint64),
+                ("ia_blksize", c_uint32),
+                ("ia_blocks", c_uint64),
+                ("ia_atime", c_uint32 ),
+                ("ia_atime_nsec", c_uint32),
+                ("ia_mtime", c_uint32),
+                ("ia_mtime_nsec", c_uint32),
+                ("ia_ctime", c_uint32),
+                ("ia_ctime_nsec", c_uint32)
+        ]
+
+
+class mem_pool (Structure):
+        _fields_ = [
+                ("list", list_head),
+                ("hot_count", c_int),
+                ("cold_count", c_int),
+                ("lock", c_void_p),
+                ("padded_sizeof_type", c_ulong),
+                ("pool", c_void_p),
+                ("pool_end", c_void_p),
+                ("real_sizeof_type", c_int),
+                ("alloc_count", c_uint64),
+                ("pool_misses", c_uint64),
+                ("max_alloc", c_int),
+                ("curr_stdalloc", c_int),
+                ("max_stdalloc", c_int),
+                ("name", c_char_p),
+                ("global_list", list_head)
+        ]
+
+
+class U_ctx_key_inode (Union):
+        _fields_ = [
+                ("key", c_uint64),
+                ("xl_key", POINTER(xlator_t))
+        ]
+
+
+class U_ctx_value1 (Union):
+        _fields_ = [
+                ("value1", c_uint64),
+                ("ptr1", c_void_p)
+        ]
+
+
+class U_ctx_value2 (Union):
+        _fields_ = [
+                ("value2", c_uint64),
+                ("ptr2", c_void_p)
+        ]
+
+class inode_ctx (Structure):
+        _anonymous_ = ("u_key","u_value1","u_value2",)
+        _fields_ = [
+                ("u_key", U_ctx_key_inode),
+                ("u_value1", U_ctx_value1),
+                ("u_value2", U_ctx_value2)
+        ]
+
+class inode_t (Structure):
+        pass
+
+class inode_table_t (Structure):
+        _fields_ = [
+                ("lock", c_void_p),
+                ("hashsize", c_size_t),
+                ("name", c_char_p),
+                ("root", POINTER(inode_t)),
+                ("xl", POINTER(xlator_t)),
+                ("lru_limit", c_uint32),
+                ("inode_hash", POINTER(list_head)),
+                ("name_hash", POINTER(list_head)),
+                ("active", list_head),
+                ("active_size", c_uint32),
+                ("lru", list_head),
+                ("lru_size", c_uint32),
+                ("purge", list_head),
+                ("purge_size", c_uint32),
+                ("inode_pool", POINTER(mem_pool)),
+                ("dentry_pool", POINTER(mem_pool)),
+                ("fd_mem_pool", POINTER(mem_pool))
+        ]
+
+inode_t._fields_ = [
+                ("table", POINTER(inode_table_t)),
+                ("gfid", c_ubyte * 16),
+                ("lock", c_void_p),
+                ("nlookup", c_uint64),
+                ("fd_count", c_uint32),
+                ("ref", c_uint32),
+                ("ia_type", c_uint),
+                ("fd_list", list_head),
+                ("dentry_list", list_head),
+                ("hashv", list_head),
+                ("listv", list_head),
+                ("ctx", POINTER(inode_ctx))
+        ]
+
+
+
+class U_ctx_key_fd (Union):
+        _fields_ = [
+                ("key", c_uint64),
+                ("xl_key", c_void_p)
+        ]
+
+class fd_lk_ctx (Structure):
+        _fields_ = [
+                ("lk_list", list_head),
+                ("ref", c_int),
+                ("lock", c_void_p)
+        ]
+
+class fd_ctx (Structure):
+        _anonymous_ = ("u_key","u_value1")
+        _fields_ = [
+                ("u_key", U_ctx_key_fd),
+                ("u_value1", U_ctx_value1)
+        ]
+
+class fd_t (Structure):
+        _fields_ = [
+                ("pid", c_uint64),
+                ("flags", c_int32),
+                ("refcount", c_int32),
+                ("inode_list", list_head),
+                ("inode", POINTER(inode_t)),
+                ("lock", c_void_p),
+                ("ctx", POINTER(fd_ctx)),
+                ("xl_count", c_int),
+                ("lk_ctx", POINTER(fd_lk_ctx)),
+                ("anonymous", c_uint)
+        ]
+
+class loc_t (Structure):
+        _fields_ = [
+                ("path", c_char_p),
+                ("name", c_char_p),
+                ("inode", POINTER(inode_t)),
+                ("parent", POINTER(inode_t)),
+                ("gfid", c_ubyte * 16),
+                ("pargfid", c_ubyte * 16),
+        ]
+
+
+
+def _init_op (a_class, fop, cbk, wind, unwind):
+        # Decorators, used by translators. We could pass the signatures as
+        # parameters, but it's actually kind of nice to keep them around for
+        # inspection.
+        a_class.fop_type = apply(CFUNCTYPE,a_class.fop_sig)
+        a_class.cbk_type = apply(CFUNCTYPE,a_class.cbk_sig)
+        # Dispatch-function registration.
+        fop.restype = None
+        fop.argtypes = [ c_long, a_class.fop_type ]
+        # Callback-function registration.
+        cbk.restype = None
+        cbk.argtypes = [ c_long, a_class.cbk_type ]
+        # STACK_WIND function.
+        wind.restype = None
+        wind.argtypes = list(a_class.fop_sig[1:])
+        # STACK_UNWIND function.
+        unwind.restype = None
+        unwind.argtypes = list(a_class.cbk_sig[1:])
+
+class OpLookup:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(inode_t), POINTER(iatt_t),
+                   POINTER(dict_t), POINTER(iatt_t))
+_init_op (OpLookup, dl.set_lookup_fop, dl.set_lookup_cbk,
+                    dl.wind_lookup,    dl.unwind_lookup)
+
+class OpCreate:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), c_int, c_uint, c_uint, POINTER(fd_t),
+                   POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(fd_t), POINTER(inode_t),
+                   POINTER(iatt_t), POINTER(iatt_t), POINTER(iatt_t),
+                   POINTER(dict_t))
+_init_op (OpCreate, dl.set_create_fop, dl.set_create_cbk,
+                    dl.wind_create,    dl.unwind_create)
+
+class OpOpen:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), c_int, POINTER(fd_t), POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(fd_t), POINTER(dict_t))
+_init_op (OpOpen, dl.set_open_fop, dl.set_open_cbk,
+                  dl.wind_open,    dl.unwind_open)
+
+class OpReadv:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(fd_t), c_size_t, c_long, c_uint32, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(iovec_t), c_int, POINTER(iatt_t),
+                   POINTER(iobref_t), POINTER(dict_t))
+_init_op (OpReadv, dl.set_readv_fop, dl.set_readv_cbk,
+                   dl.wind_readv,    dl.unwind_readv)
+class OpWritev:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(fd_t), POINTER(iovec_t), c_int, c_long, c_uint32,
+                   POINTER(iobref_t), POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(iatt_t), POINTER(iatt_t),
+                   POINTER(dict_t))
+_init_op (OpWritev, dl.set_writev_fop, dl.set_writev_cbk,
+                    dl.wind_writev,    dl.unwind_writev)
+
+class OpOpendir:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), POINTER(fd_t) ,POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(fd_t), POINTER(dict_t))
+_init_op (OpOpendir, dl.set_opendir_fop, dl.set_opendir_cbk,
+                     dl.wind_opendir,    dl.unwind_opendir)
+
+class OpReaddir:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(fd_t), c_size_t, c_long, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(gf_dirent_t), POINTER(dict_t))
+_init_op (OpReaddir, dl.set_readdir_fop, dl.set_readdir_cbk,
+                     dl.wind_readdir,    dl.unwind_readdir)
+
+class OpReaddirp:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(fd_t), c_size_t, c_long, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(gf_dirent_t), POINTER(dict_t))
+_init_op (OpReaddirp, dl.set_readdirp_fop, dl.set_readdirp_cbk,
+                      dl.wind_readdirp,    dl.unwind_readdirp)
+
+class OpStat:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(iatt_t), POINTER(dict_t))
+_init_op (OpStat, dl.set_stat_fop, dl.set_stat_cbk,
+                  dl.wind_stat,    dl.unwind_stat)
+
+class OpFstat:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(fd_t), POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(iatt_t), POINTER(dict_t))
+_init_op (OpFstat, dl.set_fstat_fop, dl.set_fstat_cbk,
+                   dl.wind_fstat,    dl.unwind_fstat)
+
+class OpStatfs:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(statvfs_t), POINTER(dict_t))
+_init_op (OpStatfs, dl.set_statfs_fop, dl.set_statfs_cbk,
+                    dl.wind_statfs,    dl.unwind_statfs)
+
+
+class OpSetxattr:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), POINTER(dict_t), c_int32,
+                   POINTER (dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(dict_t))
+_init_op (OpSetxattr, dl.set_setxattr_fop, dl.set_setxattr_cbk,
+                      dl.wind_setxattr,    dl.unwind_setxattr)
+
+class OpGetxattr:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), c_char_p, POINTER (dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(dict_t), POINTER(dict_t))
+_init_op (OpGetxattr, dl.set_getxattr_fop, dl.set_getxattr_cbk,
+                      dl.wind_getxattr,    dl.unwind_getxattr)
+
+class OpFsetxattr:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(fd_t), POINTER(dict_t), c_int32,
+                   POINTER (dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(dict_t))
+_init_op (OpFsetxattr, dl.set_fsetxattr_fop, dl.set_fsetxattr_cbk,
+                       dl.wind_fsetxattr,    dl.unwind_fsetxattr)
+
+class OpFgetxattr:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(fd_t), c_char_p, POINTER (dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(dict_t), POINTER(dict_t))
+_init_op (OpFgetxattr, dl.set_fgetxattr_fop, dl.set_fgetxattr_cbk,
+                       dl.wind_fgetxattr,    dl.unwind_fgetxattr)
+
+class OpRemovexattr:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), c_char_p, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(dict_t))
+_init_op (OpRemovexattr, dl.set_removexattr_fop, dl.set_removexattr_cbk,
+                         dl.wind_removexattr,    dl.unwind_removexattr)
+
+
+class OpFremovexattr:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(fd_t), c_char_p, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(dict_t))
+_init_op (OpFremovexattr, dl.set_fremovexattr_fop, dl.set_fremovexattr_cbk,
+                          dl.wind_fremovexattr,    dl.unwind_fremovexattr)
+
+class OpLink:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), POINTER(loc_t), POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(inode_t), POINTER(iatt_t),
+                   POINTER(iatt_t), POINTER(iatt_t), POINTER(dict_t))
+_init_op (OpLink, dl.set_link_fop, dl.set_link_cbk,
+                  dl.wind_link,    dl.unwind_link)
+
+class OpSymlink:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   c_char_p, POINTER(loc_t), c_uint, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(inode_t), POINTER(iatt_t),
+                   POINTER(iatt_t), POINTER(iatt_t), POINTER(dict_t))
+_init_op (OpSymlink, dl.set_symlink_fop, dl.set_symlink_cbk,
+                     dl.wind_symlink,    dl.unwind_symlink)
+
+class OpUnlink:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), c_int, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(iatt_t), POINTER(iatt_t),
+                   POINTER(dict_t))
+_init_op (OpUnlink, dl.set_unlink_fop, dl.set_unlink_cbk,
+                    dl.wind_unlink,    dl.unwind_unlink)
+
+class OpReadlink:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), c_size_t, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, c_char_p, POINTER(iatt_t), POINTER(dict_t))
+_init_op (OpReadlink, dl.set_readlink_fop, dl.set_readlink_cbk,
+                      dl.wind_readlink,    dl.unwind_readlink)
+
+class OpMkdir:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), c_uint, c_uint, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(inode_t), POINTER(iatt_t),
+                   POINTER(iatt_t), POINTER(iatt_t), POINTER(dict_t))
+_init_op (OpMkdir, dl.set_mkdir_fop, dl.set_mkdir_cbk,
+                   dl.wind_mkdir,    dl.unwind_mkdir)
+
+class OpRmdir:
+        fop_sig = (c_int, POINTER(call_frame_t), POINTER(xlator_t),
+                   POINTER(loc_t), c_int, POINTER(dict_t))
+        cbk_sig = (c_int, POINTER(call_frame_t), c_long, POINTER(xlator_t),
+                   c_int, c_int, POINTER(iatt_t), POINTER(iatt_t),
+                   POINTER(dict_t))
+_init_op (OpRmdir, dl.set_rmdir_fop, dl.set_rmdir_cbk,
+                   dl.wind_rmdir,    dl.unwind_rmdir)
+
+
+class Translator:
+        def __init__ (self, c_this):
+                # This is only here to keep references to the stubs we create,
+                # because ctypes doesn't and glupy.so can't because it doesn't
+                # get a pointer to the actual Python object. It's a dictionary
+                # instead of a list in case we ever allow changing fops/cbks
+                # after initialization and need to look them up.
+                self.stub_refs = {}
+                funcs = dir(self.__class__)
+                if "lookup_fop" in funcs:
+                        @OpLookup.fop_type
+                        def stub (frame, this, loc, xdata, s=self):
+                                return s.lookup_fop (frame, this, loc, xdata)
+                        self.stub_refs["lookup_fop"] = stub
+                        dl.set_lookup_fop(c_this,stub)
+                if "lookup_cbk" in funcs:
+                        @OpLookup.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, inode,
+                                  buf, xdata, postparent, s=self):
+                                return s.lookup_cbk(frame, cookie, this, op_ret,
+                                                    op_errno, inode, buf, xdata,
+                                                    postparent)
+                        self.stub_refs["lookup_cbk"] = stub
+                        dl.set_lookup_cbk(c_this,stub)
+                if "create_fop" in funcs:
+                        @OpCreate.fop_type
+                        def stub (frame, this, loc, flags, mode, umask, fd,
+                                  xdata, s=self):
+                                return s.create_fop (frame, this, loc, flags,
+                                                     mode, umask, fd, xdata)
+                        self.stub_refs["create_fop"] = stub
+                        dl.set_create_fop(c_this,stub)
+                if "create_cbk" in funcs:
+                        @OpCreate.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, fd,
+                                  inode, buf, preparent, postparent, xdata,
+                                  s=self):
+                                return s.create_cbk (frame, cookie, this,
+                                                     op_ret, op_errno, fd,
+                                                     inode, buf, preparent,
+                                                     postparent, xdata)
+                        self.stub_refs["create_cbk"] = stub
+                        dl.set_create_cbk(c_this,stub)
+                if "open_fop" in funcs:
+                        @OpOpen.fop_type
+                        def stub (frame, this, loc, flags, fd,
+                                  xdata, s=self):
+                                return s.open_fop (frame, this, loc, flags,
+                                                   fd, xdata)
+                        self.stub_refs["open_fop"] = stub
+                        dl.set_open_fop(c_this,stub)
+                if "open_cbk" in funcs:
+                        @OpOpen.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, fd,
+                                  xdata, s=self):
+                                return s.open_cbk (frame, cookie, this,
+                                                   op_ret, op_errno, fd,
+                                                   xdata)
+                        self.stub_refs["open_cbk"] = stub
+                        dl.set_open_cbk(c_this,stub)
+                if "readv_fop" in funcs:
+                        @OpReadv.fop_type
+                        def stub (frame, this, fd, size, offset, flags,
+                                  xdata, s=self):
+                                return s.readv_fop (frame, this, fd, size,
+                                                    offset, flags, xdata)
+                        self.stub_refs["readv_fop"] = stub
+                        dl.set_readv_fop(c_this,stub)
+                if "readv_cbk" in funcs:
+                        @OpReadv.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  vector, count, stbuf, iobref, xdata,
+                                  s=self):
+                                return s.readv_cbk (frame, cookie, this,
+                                                    op_ret, op_errno, vector,
+                                                    count, stbuf, iobref,
+                                                    xdata)
+                        self.stub_refs["readv_cbk"] = stub
+                        dl.set_readv_cbk(c_this,stub)
+                if "writev_fop" in funcs:
+                        @OpWritev.fop_type
+                        def stub (frame, this, fd, vector, count,
+                                  offset, flags, iobref, xdata, s=self):
+                                return s.writev_fop (frame, this, fd, vector,
+                                                     count, offset, flags,
+                                                     iobref, xdata)
+                        self.stub_refs["writev_fop"] = stub
+                        dl.set_writev_fop(c_this,stub)
+                if "writev_cbk" in funcs:
+                        @OpWritev.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  prebuf, postbuf, xdata, s=self):
+                                return s.writev_cbk (frame, cookie, this,
+                                                     op_ret, op_errno, prebuf,
+                                                     postbuf, xdata)
+                        self.stub_refs["writev_cbk"] = stub
+                        dl.set_writev_cbk(c_this,stub)
+                if "opendir_fop" in funcs:
+                        @OpOpendir.fop_type
+                        def stub (frame, this, loc, fd, xdata, s=self):
+                                return s.opendir_fop (frame, this, loc, fd,
+                                                      xdata)
+                        self.stub_refs["opendir_fop"] = stub
+                        dl.set_opendir_fop(c_this,stub)
+                if "opendir_cbk" in funcs:
+                        @OpOpendir.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, fd,
+                                  xdata, s=self):
+                                return s.opendir_cbk(frame, cookie, this,
+                                                     op_ret, op_errno, fd,
+                                                     xdata)
+                        self.stub_refs["opendir_cbk"] = stub
+                        dl.set_opendir_cbk(c_this,stub)
+                if "readdir_fop" in funcs:
+                        @OpReaddir.fop_type
+                        def stub (frame, this, fd, size, offset, xdata, s=self):
+                                return s.readdir_fop (frame, this, fd, size,
+                                                      offset, xdata)
+                        self.stub_refs["readdir_fop"] = stub
+                        dl.set_readdir_fop(c_this,stub)
+                if "readdir_cbk" in funcs:
+                        @OpReaddir.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  entries, xdata, s=self):
+                                return s.readdir_cbk(frame, cookie, this,
+                                                     op_ret, op_errno, entries,
+                                                     xdata)
+                        self.stub_refs["readdir_cbk"] = stub
+                        dl.set_readdir_cbk(c_this,stub)
+                if "readdirp_fop" in funcs:
+                        @OpReaddirp.fop_type
+                        def stub (frame, this, fd, size, offset, xdata, s=self):
+                                return s.readdirp_fop (frame, this, fd, size,
+                                                       offset, xdata)
+                        self.stub_refs["readdirp_fop"] = stub
+                        dl.set_readdirp_fop(c_this,stub)
+                if "readdirp_cbk" in funcs:
+                        @OpReaddirp.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  entries, xdata, s=self):
+                                return s.readdirp_cbk (frame, cookie, this,
+                                                       op_ret, op_errno,
+                                                       entries, xdata)
+                        self.stub_refs["readdirp_cbk"] = stub
+                        dl.set_readdirp_cbk(c_this,stub)
+                if "stat_fop" in funcs:
+                        @OpStat.fop_type
+                        def stub (frame, this, loc, xdata, s=self):
+                                return s.stat_fop (frame, this, loc, xdata)
+                        self.stub_refs["stat_fop"] = stub
+                        dl.set_stat_fop(c_this,stub)
+                if "stat_cbk" in funcs:
+                        @OpStat.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, buf,
+                                  xdata, s=self):
+                                return s.stat_cbk(frame, cookie, this, op_ret,
+                                                  op_errno, buf, xdata)
+                        self.stub_refs["stat_cbk"] = stub
+                        dl.set_stat_cbk(c_this,stub)
+                if "fstat_fop" in funcs:
+                        @OpFstat.fop_type
+                        def stub (frame, this, fd, xdata, s=self):
+                                return s.fstat_fop (frame, this, fd, xdata)
+                        self.stub_refs["fstat_fop"] = stub
+                        dl.set_fstat_fop(c_this,stub)
+                if "fstat_cbk" in funcs:
+                        @OpFstat.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, buf,
+                                  xdata, s=self):
+                                return s.fstat_cbk(frame, cookie, this, op_ret,
+                                                   op_errno, buf, xdata)
+                        self.stub_refs["fstat_cbk"] = stub
+                        dl.set_fstat_cbk(c_this,stub)
+                if "statfs_fop" in funcs:
+                        @OpStatfs.fop_type
+                        def stub (frame, this, loc, xdata, s=self):
+                                return s.statfs_fop (frame, this, loc, xdata)
+                        self.stub_refs["statfs_fop"] = stub
+                        dl.set_statfs_fop(c_this,stub)
+                if "statfs_cbk" in funcs:
+                        @OpStatfs.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, buf,
+                                  xdata, s=self):
+                                return s.statfs_cbk (frame, cookie, this,
+                                                     op_ret, op_errno, buf,
+                                                     xdata)
+                        self.stub_refs["statfs_cbk"] = stub
+                        dl.set_statfs_cbk(c_this,stub)
+                if "setxattr_fop" in funcs:
+                        @OpSetxattr.fop_type
+                        def stub (frame, this, loc, dictionary, flags, xdata,
+                                  s=self):
+                                return s.setxattr_fop (frame, this, loc,
+                                                       dictionary, flags,
+                                                       xdata)
+                        self.stub_refs["setxattr_fop"] = stub
+                        dl.set_setxattr_fop(c_this,stub)
+                if "setxattr_cbk" in funcs:
+                        @OpSetxattr.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, xdata,
+                                  s=self):
+                                return s.setxattr_cbk(frame, cookie, this,
+                                                      op_ret, op_errno, xdata)
+                        self.stub_refs["setxattr_cbk"] = stub
+                        dl.set_setxattr_cbk(c_this,stub)
+                if "getxattr_fop" in funcs:
+                        @OpGetxattr.fop_type
+                        def stub (frame, this, loc, name, xdata, s=self):
+                                return s.getxattr_fop (frame, this, loc, name,
+                                                       xdata)
+                        self.stub_refs["getxattr_fop"] = stub
+                        dl.set_getxattr_fop(c_this,stub)
+                if "getxattr_cbk" in funcs:
+                        @OpGetxattr.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  dictionary, xdata, s=self):
+                                return s.getxattr_cbk(frame, cookie, this,
+                                                      op_ret, op_errno,
+                                                      dictionary, xdata)
+                        self.stub_refs["getxattr_cbk"] = stub
+                        dl.set_getxattr_cbk(c_this,stub)
+                if "fsetxattr_fop" in funcs:
+                        @OpFsetxattr.fop_type
+                        def stub (frame, this, fd, dictionary, flags, xdata,
+                                  s=self):
+                                return s.fsetxattr_fop (frame, this, fd,
+                                                        dictionary, flags,
+                                                        xdata)
+                        self.stub_refs["fsetxattr_fop"] = stub
+                        dl.set_fsetxattr_fop(c_this,stub)
+                if "fsetxattr_cbk" in funcs:
+                        @OpFsetxattr.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, xdata,
+                                  s=self):
+                                return s.fsetxattr_cbk(frame, cookie, this,
+                                                       op_ret, op_errno, xdata)
+                        self.stub_refs["fsetxattr_cbk"] = stub
+                        dl.set_fsetxattr_cbk(c_this,stub)
+                if "fgetxattr_fop" in funcs:
+                        @OpFgetxattr.fop_type
+                        def stub (frame, this, fd, name, xdata, s=self):
+                                return s.fgetxattr_fop (frame, this, fd, name,
+                                                        xdata)
+                        self.stub_refs["fgetxattr_fop"] = stub
+                        dl.set_fgetxattr_fop(c_this,stub)
+                if "fgetxattr_cbk" in funcs:
+                        @OpFgetxattr.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  dictionary, xdata, s=self):
+                                return s.fgetxattr_cbk(frame, cookie, this,
+                                                       op_ret, op_errno,
+                                                       dictionary, xdata)
+                        self.stub_refs["fgetxattr_cbk"] = stub
+                        dl.set_fgetxattr_cbk(c_this,stub)
+                if "removexattr_fop" in funcs:
+                        @OpRemovexattr.fop_type
+                        def stub (frame, this, loc, name, xdata, s=self):
+                                return s.removexattr_fop (frame, this, loc,
+                                                          name, xdata)
+                        self.stub_refs["removexattr_fop"] = stub
+                        dl.set_removexattr_fop(c_this,stub)
+                if "removexattr_cbk" in funcs:
+                        @OpRemovexattr.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  xdata, s=self):
+                                return s.removexattr_cbk(frame, cookie, this,
+                                                         op_ret, op_errno,
+                                                         xdata)
+                        self.stub_refs["removexattr_cbk"] = stub
+                        dl.set_removexattr_cbk(c_this,stub)
+                if "fremovexattr_fop" in funcs:
+                        @OpFremovexattr.fop_type
+                        def stub (frame, this, fd, name, xdata, s=self):
+                                return s.fremovexattr_fop (frame, this, fd,
+                                                           name, xdata)
+                        self.stub_refs["fremovexattr_fop"] = stub
+                        dl.set_fremovexattr_fop(c_this,stub)
+                if "fremovexattr_cbk" in funcs:
+                        @OpFremovexattr.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  xdata, s=self):
+                                return s.fremovexattr_cbk(frame, cookie, this,
+                                                          op_ret, op_errno,
+                                                          xdata)
+                        self.stub_refs["fremovexattr_cbk"] = stub
+                        dl.set_fremovexattr_cbk(c_this,stub)
+                if "link_fop" in funcs:
+                        @OpLink.fop_type
+                        def stub (frame, this, oldloc, newloc,
+                                  xdata, s=self):
+                                return s.link_fop (frame, this, oldloc,
+                                                   newloc, xdata)
+                        self.stub_refs["link_fop"] = stub
+                        dl.set_link_fop(c_this,stub)
+                if "link_cbk" in funcs:
+                        @OpLink.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  inode, buf, preparent, postparent, xdata,
+                                  s=self):
+                                return s.link_cbk (frame, cookie, this,
+                                                   op_ret, op_errno, inode,
+                                                   buf, preparent,
+                                                   postparent, xdata)
+                        self.stub_refs["link_cbk"] = stub
+                        dl.set_link_cbk(c_this,stub)
+                if "symlink_fop" in funcs:
+                        @OpSymlink.fop_type
+                        def stub (frame, this, linkname, loc,
+                                  umask, xdata, s=self):
+                                return s.symlink_fop (frame, this, linkname,
+                                                      loc, umask, xdata)
+                        self.stub_refs["symlink_fop"] = stub
+                        dl.set_symlink_fop(c_this,stub)
+                if "symlink_cbk" in funcs:
+                        @OpSymlink.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  inode, buf, preparent, postparent, xdata,
+                                  s=self):
+                                return s.symlink_cbk (frame, cookie, this,
+                                                      op_ret, op_errno, inode,
+                                                      buf, preparent,
+                                                      postparent, xdata)
+                        self.stub_refs["symlink_cbk"] = stub
+                        dl.set_symlink_cbk(c_this,stub)
+                if "unlink_fop" in funcs:
+                        @OpUnlink.fop_type
+                        def stub (frame, this, loc, xflags,
+                                  xdata, s=self):
+                                return s.unlink_fop (frame, this, loc,
+                                                     xflags, xdata)
+                        self.stub_refs["unlink_fop"] = stub
+                        dl.set_unlink_fop(c_this,stub)
+                if "unlink_cbk" in funcs:
+                        @OpUnlink.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  preparent, postparent, xdata, s=self):
+                                return s.unlink_cbk (frame, cookie, this,
+                                                     op_ret, op_errno,
+                                                     preparent, postparent,
+                                                     xdata)
+                        self.stub_refs["unlink_cbk"] = stub
+                        dl.set_unlink_cbk(c_this,stub)
+                if "readlink_fop" in funcs:
+                        @OpReadlink.fop_type
+                        def stub (frame, this, loc, size,
+                                  xdata, s=self):
+                                return s.readlink_fop (frame, this, loc,
+                                                       size, xdata)
+                        self.stub_refs["readlink_fop"] = stub
+                        dl.set_readlink_fop(c_this,stub)
+                if "readlink_cbk" in funcs:
+                        @OpReadlink.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  path, buf, xdata, s=self):
+                                return s.readlink_cbk (frame, cookie, this,
+                                                       op_ret, op_errno,
+                                                       path, buf, xdata)
+                        self.stub_refs["readlink_cbk"] = stub
+                        dl.set_readlink_cbk(c_this,stub)
+                if "mkdir_fop" in funcs:
+                        @OpMkdir.fop_type
+                        def stub (frame, this, loc, mode, umask, xdata,
+                                  s=self):
+                                return s.mkdir_fop (frame, this, loc, mode,
+                                                    umask, xdata)
+                        self.stub_refs["mkdir_fop"] = stub
+                        dl.set_mkdir_fop(c_this,stub)
+                if "mkdir_cbk" in funcs:
+                        @OpMkdir.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno, inode,
+                                  buf, preparent, postparent, xdata, s=self):
+                                return s.mkdir_cbk (frame, cookie, this,
+                                                    op_ret, op_errno, inode,
+                                                    buf, preparent,
+                                                    postparent, xdata)
+                        self.stub_refs["mkdir_cbk"] = stub
+                        dl.set_mkdir_cbk(c_this,stub)
+                if "rmdir_fop" in funcs:
+                        @OpRmdir.fop_type
+                        def stub (frame, this, loc, xflags,
+                                  xdata, s=self):
+                                return s.rmdir_fop (frame, this, loc,
+                                                    xflags, xdata)
+                        self.stub_refs["rmdir_fop"] = stub
+                        dl.set_rmdir_fop(c_this,stub)
+                if "rmdir_cbk" in funcs:
+                        @OpRmdir.cbk_type
+                        def stub (frame, cookie, this, op_ret, op_errno,
+                                  preparent, postparent, xdata, s=self):
+                                return s.rmdir_cbk (frame, cookie, this,
+                                                    op_ret, op_errno,
+                                                    preparent, postparent,
+                                                    xdata)
+                        self.stub_refs["rmdir_cbk"] = stub
+                        dl.set_rmdir_cbk(c_this,stub)
diff --git a/xlators/features/glupy/src/setup.py.in b/xlators/features/glupy/src/setup.py.in
new file mode 100644
index 000000000..1aea9875f
--- /dev/null
+++ b/xlators/features/glupy/src/setup.py.in
@@ -0,0 +1,24 @@
+from distutils.core import setup
+
+DESC = """GlusterFS is a clustered file-system capable of scaling to
+several petabytes. It aggregates various storage bricks over Infiniband
+RDMA or TCP/IP interconnect into one large parallel network file system.
+GlusterFS is one of the most sophisticated file systems in terms of
+features and extensibility.  It borrows a powerful concept called
+Translators from GNU Hurd kernel. Much of the code in GlusterFS is in
+user space and easily manageable.
+
+This package contains Glupy, the Python translator interface for GlusterFS."""
+
+setup(
+    name='glusterfs-glupy',
+    version='@PACKAGE_VERSION@',
+    description='Glupy is the Python translator interface for GlusterFS',
+    long_description=DESC,
+    author='Gluster Community',
+    author_email='gluster-devel@nongnu.org',
+    license='LGPLv3',
+    url='http://gluster.org/',
+    package_dir={'gluster':''},
+    packages=['gluster']
+)
diff --git a/xlators/features/index/Makefile.am b/xlators/features/index/Makefile.am
new file mode 100644
index 000000000..a985f42a8
--- /dev/null
+++ b/xlators/features/index/Makefile.am
@@ -0,0 +1,3 @@
+SUBDIRS = src
+
+CLEANFILES =
diff --git a/xlators/features/index/src/Makefile.am b/xlators/features/index/src/Makefile.am
new file mode 100644
index 000000000..73bb8972e
--- /dev/null
+++ b/xlators/features/index/src/Makefile.am
@@ -0,0 +1,17 @@
+xlator_LTLIBRARIES = index.la
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+index_la_LDFLAGS = -module -avoid-version
+
+index_la_SOURCES = index.c
+index_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+noinst_HEADERS = index.h index-mem-types.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) \
+	-I$(top_srcdir)/libglusterfs/src -I$(top_srcdir)/rpc/xdr/src \
+	-I$(top_srcdir)/rpc/rpc-lib/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+CLEANFILES =
diff --git a/xlators/features/index/src/index-mem-types.h b/xlators/features/index/src/index-mem-types.h
new file mode 100644
index 000000000..553d492df
--- /dev/null
+++ b/xlators/features/index/src/index-mem-types.h
@@ -0,0 +1,22 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef __QUIESCE_MEM_TYPES_H__
+#define __QUIESCE_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_index_mem_types_ {
+        gf_index_mt_priv_t = gf_common_mt_end + 1,
+        gf_index_inode_ctx_t = gf_common_mt_end + 2,
+        gf_index_fd_ctx_t = gf_common_mt_end + 3,
+        gf_index_mt_end
+};
+#endif
diff --git a/xlators/features/index/src/index.c b/xlators/features/index/src/index.c
new file mode 100644
index 000000000..5c1c65fbd
--- /dev/null
+++ b/xlators/features/index/src/index.c
@@ -0,0 +1,1275 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "index.h"
+#include "options.h"
+#include "glusterfs3-xdr.h"
+#include "syscall.h"
+
+#define XATTROP_SUBDIR "xattrop"
+
+call_stub_t *
+__index_dequeue (struct list_head *callstubs)
+{
+        call_stub_t *stub = NULL;
+
+        if (!list_empty (callstubs)) {
+                stub = list_entry (callstubs->next, call_stub_t, list);
+                list_del_init (&stub->list);
+        }
+
+        return stub;
+}
+
+inline static void
+__index_enqueue (struct list_head *callstubs, call_stub_t *stub)
+{
+        list_add_tail (&stub->list, callstubs);
+}
+
+static void
+worker_enqueue (xlator_t *this, call_stub_t *stub)
+{
+        index_priv_t    *priv = NULL;
+
+        priv = this->private;
+        pthread_mutex_lock (&priv->mutex);
+        {
+                __index_enqueue (&priv->callstubs, stub);
+                pthread_cond_signal (&priv->cond);
+        }
+        pthread_mutex_unlock (&priv->mutex);
+}
+
+void *
+index_worker (void *data)
+{
+        index_priv_t     *priv = NULL;
+        xlator_t         *this = NULL;
+        call_stub_t      *stub = NULL;
+        int               ret = 0;
+
+        THIS = data;
+        this = data;
+        priv = this->private;
+
+        for (;;) {
+                pthread_mutex_lock (&priv->mutex);
+                {
+                        while (list_empty (&priv->callstubs)) {
+                                ret = pthread_cond_wait (&priv->cond,
+                                                         &priv->mutex);
+                        }
+
+                        stub = __index_dequeue (&priv->callstubs);
+                }
+                pthread_mutex_unlock (&priv->mutex);
+
+                if (stub) /* guard against spurious wakeups */
+                        call_resume (stub);
+        }
+
+        return NULL;
+}
+int
+__index_inode_ctx_get (inode_t *inode, xlator_t *this, index_inode_ctx_t **ctx)
+{
+        int               ret = 0;
+        index_inode_ctx_t *ictx = NULL;
+        uint64_t          tmpctx = 0;
+
+        ret = __inode_ctx_get (inode, this, &tmpctx);
+        if (!ret) {
+                ictx = (index_inode_ctx_t*) (long) tmpctx;
+                goto out;
+        }
+        ictx = GF_CALLOC (1, sizeof (*ictx), gf_index_inode_ctx_t);
+        if (!ictx) {
+                ret = -1;
+                goto out;
+        }
+
+        INIT_LIST_HEAD (&ictx->callstubs);
+        ret = __inode_ctx_put (inode, this, (uint64_t)ictx);
+        if (ret) {
+                GF_FREE (ictx);
+                ictx = NULL;
+                goto out;
+        }
+out:
+        if (ictx)
+                *ctx = ictx;
+        return ret;
+}
+
+int
+index_inode_ctx_get (inode_t *inode, xlator_t *this, index_inode_ctx_t **ctx)
+{
+        int               ret = 0;
+
+        LOCK (&inode->lock);
+        {
+                ret = __index_inode_ctx_get (inode, this, ctx);
+        }
+        UNLOCK (&inode->lock);
+
+        return ret;
+}
+
+static void
+make_index_dir_path (char *base, const char *subdir,
+                     char *index_dir, size_t len)
+{
+        snprintf (index_dir, len, "%s/%s", base, subdir);
+}
+
+int
+index_dir_create (xlator_t *this, const char *subdir)
+{
+        int          ret = 0;
+        struct       stat st = {0};
+        char         fullpath[PATH_MAX] = {0};
+        char         path[PATH_MAX] = {0};
+        char         *dir = NULL;
+        index_priv_t *priv = NULL;
+        size_t       len = 0;
+        size_t       pathlen = 0;
+
+        priv = this->private;
+        make_index_dir_path (priv->index_basepath, subdir, fullpath,
+                             sizeof (fullpath));
+        ret = stat (fullpath, &st);
+        if (!ret) {
+                if (!S_ISDIR (st.st_mode))
+                        ret = -2;
+                goto out;
+        }
+
+        pathlen = strlen (fullpath);
+        if ((pathlen > 1) && fullpath[pathlen - 1] == '/')
+                fullpath[pathlen - 1] = '\0';
+        dir = strchr (fullpath, '/');
+        while (dir) {
+                dir = strchr (dir + 1, '/');
+                if (dir)
+                        len = pathlen - strlen (dir);
+                else
+                        len = pathlen;
+                strncpy (path, fullpath, len);
+                path[len] = '\0';
+                ret = mkdir (path, 0600);
+                if (ret && (errno != EEXIST))
+                        goto out;
+        }
+        ret = 0;
+out:
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR, "%s/%s: Failed to "
+                        "create (%s)", priv->index_basepath, subdir,
+                        strerror (errno));
+        } else if (ret == -2) {
+                gf_log (this->name, GF_LOG_ERROR, "%s/%s: Failed to create, "
+                        "path exists, not a directory ", priv->index_basepath,
+                        subdir);
+        }
+        return ret;
+}
+
+void
+index_get_index (index_priv_t *priv, uuid_t index)
+{
+        LOCK (&priv->lock);
+        {
+                uuid_copy (index, priv->index);
+        }
+        UNLOCK (&priv->lock);
+}
+
+void
+index_generate_index (index_priv_t *priv, uuid_t index)
+{
+        LOCK (&priv->lock);
+        {
+                //To prevent duplicate generates.
+                //This method fails if number of contending threads is greater
+                //than MAX_LINK count of the fs
+                if (!uuid_compare (priv->index, index))
+                        uuid_generate (priv->index);
+                uuid_copy (index, priv->index);
+        }
+        UNLOCK (&priv->lock);
+}
+
+static void
+make_index_path (char *base, const char *subdir, uuid_t index,
+                 char *index_path, size_t len)
+{
+        make_index_dir_path (base, subdir, index_path, len);
+        snprintf (index_path + strlen (index_path), len - strlen (index_path),
+                  "/%s-%s", subdir, uuid_utoa (index));
+}
+
+static void
+make_gfid_path (char *base, const char *subdir, uuid_t gfid,
+                char *gfid_path, size_t len)
+{
+        make_index_dir_path (base, subdir, gfid_path, len);
+        snprintf (gfid_path + strlen (gfid_path), len - strlen (gfid_path),
+                  "/%s", uuid_utoa (gfid));
+}
+
+static void
+make_file_path (char *base, const char *subdir, const char *filename,
+                char *file_path, size_t len)
+{
+        make_index_dir_path (base, subdir, file_path, len);
+        snprintf (file_path + strlen (file_path), len - strlen (file_path),
+                  "/%s", filename);
+}
+
+static void
+check_delete_stale_index_file (xlator_t *this, char *filename)
+{
+        int             ret = 0;
+        struct stat     st = {0};
+        char            filepath[PATH_MAX] = {0};
+        index_priv_t    *priv = NULL;
+
+        priv = this->private;
+        make_file_path (priv->index_basepath, XATTROP_SUBDIR,
+                        filename, filepath, sizeof (filepath));
+        ret = stat (filepath, &st);
+        if (!ret && st.st_nlink == 1)
+                unlink (filepath);
+}
+
+static int
+index_fill_readdir (fd_t *fd, DIR *dir, off_t off,
+                    size_t size, gf_dirent_t *entries)
+{
+        off_t     in_case = -1;
+        size_t    filled = 0;
+        int       count = 0;
+        char      entrybuf[sizeof(struct dirent) + 256 + 8];
+        struct dirent  *entry          = NULL;
+        int32_t              this_size      = -1;
+        gf_dirent_t          *this_entry     = NULL;
+        xlator_t             *this = NULL;
+
+        this = THIS;
+        if (!off) {
+                rewinddir (dir);
+        } else {
+                seekdir (dir, off);
+        }
+
+        while (filled <= size) {
+                in_case = telldir (dir);
+
+                if (in_case == -1) {
+                        gf_log (THIS->name, GF_LOG_ERROR,
+                                "telldir failed on dir=%p: %s",
+                                dir, strerror (errno));
+                        goto out;
+                }
+
+                errno = 0;
+                entry = NULL;
+                readdir_r (dir, (struct dirent *)entrybuf, &entry);
+
+                if (!entry) {
+                        if (errno == EBADF) {
+                                gf_log (THIS->name, GF_LOG_WARNING,
+                                        "readdir failed on dir=%p: %s",
+                                        dir, strerror (errno));
+                                goto out;
+                        }
+                        break;
+                }
+
+                if (!strncmp (entry->d_name, XATTROP_SUBDIR"-",
+                              strlen (XATTROP_SUBDIR"-"))) {
+                        check_delete_stale_index_file (this, entry->d_name);
+                        continue;
+                }
+
+                this_size = max (sizeof (gf_dirent_t),
+                                 sizeof (gfs3_dirplist))
+                        + strlen (entry->d_name) + 1;
+
+                if (this_size + filled > size) {
+                        seekdir (dir, in_case);
+                        break;
+                }
+
+                this_entry = gf_dirent_for_name (entry->d_name);
+
+                if (!this_entry) {
+                        gf_log (THIS->name, GF_LOG_ERROR,
+                                "could not create gf_dirent for entry %s: (%s)",
+                                entry->d_name, strerror (errno));
+                        goto out;
+                }
+                this_entry->d_off = telldir (dir);
+                this_entry->d_ino = entry->d_ino;
+
+                list_add_tail (&this_entry->list, &entries->list);
+
+                filled += this_size;
+                count ++;
+        }
+
+        if ((!readdir (dir) && (errno == 0)))
+                /* Indicate EOF */
+                errno = ENOENT;
+out:
+        return count;
+}
+
+int
+index_add (xlator_t *this, uuid_t gfid, const char *subdir)
+{
+        int32_t           op_errno = 0;
+        char              gfid_path[PATH_MAX] = {0};
+        char              index_path[PATH_MAX] = {0};
+        int               ret = 0;
+        uuid_t            index = {0};
+        index_priv_t      *priv = NULL;
+        struct stat       st = {0};
+        int               fd = 0;
+
+        priv = this->private;
+        GF_ASSERT_AND_GOTO_WITH_ERROR (this->name, !uuid_is_null (gfid),
+                                       out, op_errno, EINVAL);
+
+        make_gfid_path (priv->index_basepath, subdir, gfid,
+                        gfid_path, sizeof (gfid_path));
+
+        ret = stat (gfid_path, &st);
+        if (!ret)
+                goto out;
+        index_get_index (priv, index);
+        make_index_path (priv->index_basepath, subdir,
+                         index, index_path, sizeof (index_path));
+        ret = sys_link (index_path, gfid_path);
+        if (!ret || (errno == EEXIST))  {
+                ret = 0;
+                goto out;
+        }
+
+        op_errno = errno;
+        if (op_errno == ENOENT) {
+                ret = index_dir_create (this, subdir);
+                if (ret)
+                        goto out;
+        } else if (op_errno == EMLINK) {
+                index_generate_index (priv, index);
+                make_index_path (priv->index_basepath, subdir,
+                                 index, index_path, sizeof (index_path));
+        } else {
+                goto out;
+        }
+
+        fd = creat (index_path, 0);
+        if ((fd < 0) && (errno != EEXIST)) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR, "%s: Not able to "
+                        "create index (%s)", uuid_utoa (gfid),
+                        strerror (errno));
+                goto out;
+        }
+
+        if (fd >= 0)
+                close (fd);
+
+        ret = sys_link (index_path, gfid_path);
+        if (ret && (errno != EEXIST)) {
+                gf_log (this->name, GF_LOG_ERROR, "%s: Not able to "
+                        "add to index (%s)", uuid_utoa (gfid),
+                        strerror (errno));
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int
+index_del (xlator_t *this, uuid_t gfid, const char *subdir)
+{
+        int32_t      op_errno __attribute__((unused)) = 0;
+        index_priv_t *priv = NULL;
+        int          ret = 0;
+        char         gfid_path[PATH_MAX] = {0};
+
+        priv = this->private;
+        GF_ASSERT_AND_GOTO_WITH_ERROR (this->name, !uuid_is_null (gfid),
+                                       out, op_errno, EINVAL);
+        make_gfid_path (priv->index_basepath, subdir, gfid,
+                        gfid_path, sizeof (gfid_path));
+        ret = unlink (gfid_path);
+        if (ret && (errno != ENOENT)) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "%s: failed to delete from index (%s)",
+                        gfid_path, strerror (errno));
+                ret = -errno;
+                goto out;
+        }
+        ret = 0;
+out:
+        return ret;
+}
+
+static int
+_check_key_is_zero_filled (dict_t *d, char *k, data_t *v,
+                           void *tmp)
+{
+        if (mem_0filled ((const char*)v->data, v->len)) {
+                /* -1 means, no more iterations, treat as 'break' */
+                return -1;
+        }
+        return 0;
+}
+
+void
+_index_action (xlator_t *this, inode_t *inode, gf_boolean_t zero_xattr)
+{
+        int               ret  = 0;
+        index_inode_ctx_t *ctx = NULL;
+
+        ret = index_inode_ctx_get (inode, this, &ctx);
+        if (ret) {
+                gf_log (this->name, GF_LOG_ERROR, "Not able to %s %s -> index",
+                        zero_xattr?"del":"add", uuid_utoa (inode->gfid));
+                goto out;
+        }
+        if (zero_xattr) {
+                if (ctx->state == NOTIN)
+                        goto out;
+                ret = index_del (this, inode->gfid, XATTROP_SUBDIR);
+                if (!ret)
+                        ctx->state = NOTIN;
+        } else {
+                if (ctx->state == IN)
+                        goto out;
+                ret = index_add (this, inode->gfid, XATTROP_SUBDIR);
+                if (!ret)
+                        ctx->state = IN;
+        }
+out:
+        return;
+}
+
+void
+_xattrop_index_action (xlator_t *this, inode_t *inode,  dict_t *xattr)
+{
+        gf_boolean_t      zero_xattr = _gf_true;
+        int               ret = 0;
+
+        ret = dict_foreach (xattr, _check_key_is_zero_filled, NULL);
+        if (ret == -1)
+                zero_xattr = _gf_false;
+        _index_action (this, inode, zero_xattr);
+        return;
+}
+
+void
+fop_xattrop_index_action (xlator_t *this, inode_t *inode, dict_t *xattr)
+{
+        _xattrop_index_action (this, inode, xattr);
+}
+
+void
+fop_fxattrop_index_action (xlator_t *this, inode_t *inode, dict_t *xattr)
+{
+        _xattrop_index_action (this, inode, xattr);
+}
+
+inline gf_boolean_t
+index_xattrop_track (loc_t *loc, gf_xattrop_flags_t flags, dict_t *dict)
+{
+        return (flags == GF_XATTROP_ADD_ARRAY);
+}
+
+inline gf_boolean_t
+index_fxattrop_track (fd_t *fd, gf_xattrop_flags_t flags, dict_t *dict)
+{
+        return (flags == GF_XATTROP_ADD_ARRAY);
+}
+
+int
+__index_fd_ctx_get (fd_t *fd, xlator_t *this, index_fd_ctx_t **ctx)
+{
+        int               ret = 0;
+        index_fd_ctx_t    *fctx = NULL;
+        uint64_t          tmpctx = 0;
+        char              index_dir[PATH_MAX] = {0};
+        index_priv_t      *priv = NULL;
+
+        priv = this->private;
+        if (uuid_compare (fd->inode->gfid, priv->xattrop_vgfid)) {
+                ret = -EINVAL;
+                goto out;
+        }
+
+        ret = __fd_ctx_get (fd, this, &tmpctx);
+        if (!ret) {
+                fctx = (index_fd_ctx_t*) (long) tmpctx;
+                goto out;
+        }
+
+        fctx = GF_CALLOC (1, sizeof (*fctx), gf_index_fd_ctx_t);
+        if (!fctx) {
+                ret = -ENOMEM;
+                goto out;
+        }
+
+        make_index_dir_path (priv->index_basepath, XATTROP_SUBDIR,
+                             index_dir, sizeof (index_dir));
+        fctx->dir = opendir (index_dir);
+        if (!fctx->dir) {
+                ret = -errno;
+                GF_FREE (fctx);
+                fctx = NULL;
+                goto out;
+        }
+
+        ret = __fd_ctx_set (fd, this, (uint64_t)(long)fctx);
+        if (ret) {
+                GF_FREE (fctx);
+                fctx = NULL;
+                ret = -EINVAL;
+                goto out;
+        }
+out:
+        if (fctx)
+                *ctx = fctx;
+        return ret;
+}
+
+int
+index_fd_ctx_get (fd_t *fd, xlator_t *this, index_fd_ctx_t **ctx)
+{
+        int     ret = 0;
+        LOCK (&fd->lock);
+        {
+                ret = __index_fd_ctx_get (fd, this, ctx);
+        }
+        UNLOCK (&fd->lock);
+        return ret;
+}
+
+//new - Not NULL means start a fop
+//new - NULL means done processing the fop
+void
+index_queue_process (xlator_t *this, inode_t *inode, call_stub_t *new)
+{
+        call_stub_t       *stub = NULL;
+        index_inode_ctx_t *ctx = NULL;
+        int               ret = 0;
+        call_frame_t      *frame = NULL;
+
+        LOCK (&inode->lock);
+        {
+                ret = __index_inode_ctx_get (inode, this, &ctx);
+                if (ret)
+                        goto unlock;
+
+                if (new) {
+                        __index_enqueue (&ctx->callstubs, new);
+                        new = NULL;
+                } else {
+                        ctx->processing = _gf_false;
+                }
+
+                if (!ctx->processing) {
+                        stub = __index_dequeue (&ctx->callstubs);
+                        if (stub)
+                                ctx->processing = _gf_true;
+                        else
+                                ctx->processing = _gf_false;
+                }
+        }
+unlock:
+        UNLOCK (&inode->lock);
+
+        if (ret && new) {
+                frame = new->frame;
+                if (new->fop == GF_FOP_XATTROP) {
+                        INDEX_STACK_UNWIND (xattrop, frame, -1, ENOMEM,
+                                            NULL, NULL);
+                } else if (new->fop == GF_FOP_FXATTROP) {
+                        INDEX_STACK_UNWIND (fxattrop, frame, -1, ENOMEM,
+                                            NULL, NULL);
+                }
+                call_stub_destroy (new);
+        } else if (stub) {
+                call_resume (stub);
+        }
+        return;
+}
+
+int32_t
+index_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, dict_t *xattr, dict_t *xdata)
+{
+        inode_t *inode = NULL;
+
+        inode = inode_ref (frame->local);
+        if (op_ret < 0)
+                goto out;
+        fop_xattrop_index_action (this, frame->local, xattr);
+out:
+        INDEX_STACK_UNWIND (xattrop, frame, op_ret, op_errno, xattr, xdata);
+        index_queue_process (this, inode, NULL);
+        inode_unref (inode);
+
+        return 0;
+}
+
+int32_t
+index_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xattr,
+                    dict_t *xdata)
+{
+        inode_t *inode = NULL;
+
+        inode = inode_ref (frame->local);
+        if (op_ret < 0)
+                goto out;
+
+        fop_fxattrop_index_action (this, frame->local, xattr);
+out:
+        INDEX_STACK_UNWIND (fxattrop, frame, op_ret, op_errno, xattr, xdata);
+        index_queue_process (this, inode, NULL);
+        inode_unref (inode);
+
+        return 0;
+}
+
+int
+index_xattrop_wrapper (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                       gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
+{
+        //In wind phase bring the gfid into index. This way if the brick crashes
+        //just after posix performs xattrop before _cbk reaches index xlator
+        //we will still have the gfid in index.
+        _index_action (this, frame->local, _gf_false);
+
+        STACK_WIND (frame, index_xattrop_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->xattrop, loc, optype, xattr,
+                    xdata);
+        return 0;
+}
+
+int
+index_fxattrop_wrapper (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                        gf_xattrop_flags_t optype, dict_t *xattr, dict_t *xdata)
+{
+        //In wind phase bring the gfid into index. This way if the brick crashes
+        //just after posix performs xattrop before _cbk reaches index xlator
+        //we will still have the gfid in index.
+        _index_action (this, frame->local, _gf_false);
+        STACK_WIND (frame, index_fxattrop_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->fxattrop, fd, optype, xattr,
+                    xdata);
+        return 0;
+}
+
+int32_t
+index_xattrop (call_frame_t *frame, xlator_t *this, loc_t *loc,
+	       gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+{
+        call_stub_t     *stub = NULL;
+
+        if (!index_xattrop_track (loc, flags, dict))
+                goto out;
+
+        frame->local = inode_ref (loc->inode);
+        stub = fop_xattrop_stub (frame, index_xattrop_wrapper,
+                                 loc, flags, dict, xdata);
+        if (!stub) {
+                INDEX_STACK_UNWIND (xattrop, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+
+        index_queue_process (this, loc->inode, stub);
+        return 0;
+out:
+        STACK_WIND (frame, default_xattrop_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->xattrop, loc, flags, dict, xdata);
+        return 0;
+}
+
+int32_t
+index_fxattrop (call_frame_t *frame, xlator_t *this, fd_t *fd,
+		gf_xattrop_flags_t flags, dict_t *dict, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+
+        if (!index_fxattrop_track (fd, flags, dict))
+                goto out;
+
+        frame->local = inode_ref (fd->inode);
+        stub = fop_fxattrop_stub (frame, index_fxattrop_wrapper,
+                                  fd, flags, dict, xdata);
+        if (!stub) {
+                INDEX_STACK_UNWIND (fxattrop, frame, -1, ENOMEM, NULL, xdata);
+                return 0;
+        }
+
+        index_queue_process (this, fd->inode, stub);
+        return 0;
+out:
+        STACK_WIND (frame, default_fxattrop_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fxattrop, fd, flags, dict, xdata);
+        return 0;
+}
+
+uint64_t
+index_entry_count (xlator_t *this, char *subdir)
+{
+	index_priv_t *priv = NULL;
+	char index_dir[PATH_MAX];
+	DIR *dirp = NULL;
+	uint64_t count = 0;
+	struct dirent buf;
+	struct dirent *entry = NULL;
+
+	priv = this->private;
+
+	make_index_dir_path (priv->index_basepath, subdir,
+			     index_dir, sizeof (index_dir));
+
+	dirp = opendir (index_dir);
+	if (!dirp)
+		return 0;
+
+	while (readdir_r (dirp, &buf, &entry) == 0) {
+		if (!entry)
+			break;
+		if (!strcmp (entry->d_name, ".") ||
+		    !strcmp (entry->d_name, ".."))
+			continue;
+                if (!strncmp (entry->d_name, subdir, strlen (subdir)))
+			continue;
+		count++;
+	}
+	closedir (dirp);
+
+	return count;
+}
+
+
+int32_t
+index_getxattr_wrapper (call_frame_t *frame, xlator_t *this,
+                        loc_t *loc, const char *name, dict_t *xdata)
+{
+        index_priv_t    *priv = NULL;
+        dict_t          *xattr = NULL;
+        int             ret = 0;
+	uint64_t        count = 0;
+
+        priv = this->private;
+
+        xattr = dict_new ();
+        if (!xattr) {
+                ret = -ENOMEM;
+                goto done;
+        }
+
+	if (strcmp (name, GF_XATTROP_INDEX_GFID) == 0) {
+		ret = dict_set_static_bin (xattr, (char*)name, priv->xattrop_vgfid,
+					   sizeof (priv->xattrop_vgfid));
+		if (ret) {
+			ret = -ENOMEM;
+			gf_log (this->name, GF_LOG_ERROR, "xattrop index "
+				"gfid set failed");
+			goto done;
+		}
+	} else if (strcmp (name, GF_XATTROP_INDEX_COUNT) == 0) {
+		count = index_entry_count (this, XATTROP_SUBDIR);
+
+		ret = dict_set_uint64 (xattr, (char *)name, count);
+		if (ret) {
+			ret = -ENOMEM;
+			gf_log (this->name, GF_LOG_ERROR, "xattrop index "
+				"count set failed");
+			goto done;
+		}
+	}
+done:
+        if (ret)
+                STACK_UNWIND_STRICT (getxattr, frame, -1, -ret, xattr, xdata);
+        else
+                STACK_UNWIND_STRICT (getxattr, frame, 0, 0, xattr, xdata);
+
+        if (xattr)
+                dict_unref (xattr);
+
+        return 0;
+}
+
+int32_t
+index_lookup_wrapper (call_frame_t *frame, xlator_t *this,
+                      loc_t *loc, dict_t *xattr_req)
+{
+        index_priv_t    *priv = NULL;
+        struct stat     lstatbuf = {0};
+        int             ret = 0;
+        int32_t         op_errno = EINVAL;
+        int32_t         op_ret = -1;
+        char            path[PATH_MAX] = {0};
+        struct iatt     stbuf        = {0, };
+        struct iatt     postparent = {0,};
+        dict_t          *xattr = NULL;
+        gf_boolean_t    is_dir = _gf_false;
+
+        priv = this->private;
+
+        VALIDATE_OR_GOTO (loc, done);
+        if (!uuid_compare (loc->gfid, priv->xattrop_vgfid)) {
+                make_index_dir_path (priv->index_basepath, XATTROP_SUBDIR,
+                                     path, sizeof (path));
+                is_dir = _gf_true;
+        } else if (!uuid_compare (loc->pargfid, priv->xattrop_vgfid)) {
+                make_file_path (priv->index_basepath, XATTROP_SUBDIR,
+                                loc->name, path, sizeof (path));
+        }
+
+        ret = lstat (path, &lstatbuf);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG, "Stat failed on index dir "
+                        "(%s)", strerror (errno));
+                op_errno = errno;
+                goto done;
+        } else if (!S_ISDIR (lstatbuf.st_mode) && is_dir) {
+                gf_log (this->name, GF_LOG_DEBUG, "Stat failed on index dir, "
+                        "not a directory");
+                op_errno = ENOENT;
+                goto done;
+        }
+        xattr = dict_new ();
+        if (!xattr) {
+                op_errno = ENOMEM;
+                goto done;
+        }
+
+        iatt_from_stat (&stbuf, &lstatbuf);
+        if (is_dir)
+                uuid_copy (stbuf.ia_gfid, priv->xattrop_vgfid);
+        else
+                uuid_generate (stbuf.ia_gfid);
+        stbuf.ia_ino = -1;
+        op_ret = 0;
+done:
+        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno,
+                             loc->inode, &stbuf, xattr, &postparent);
+        if (xattr)
+                dict_unref (xattr);
+        return 0;
+}
+
+int32_t
+index_readdir_wrapper (call_frame_t *frame, xlator_t *this,
+                       fd_t *fd, size_t size, off_t off, dict_t *xdata)
+{
+        index_fd_ctx_t       *fctx           = NULL;
+        DIR                  *dir            = NULL;
+        int                   ret            = -1;
+        int32_t               op_ret         = -1;
+        int32_t               op_errno       = 0;
+        int                   count          = 0;
+        gf_dirent_t           entries;
+
+        INIT_LIST_HEAD (&entries.list);
+
+        ret = index_fd_ctx_get (fd, this, &fctx);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "pfd is NULL, fd=%p", fd);
+                op_errno = -ret;
+                goto done;
+        }
+
+        dir = fctx->dir;
+
+        if (!dir) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "dir is NULL for fd=%p", fd);
+                op_errno = EINVAL;
+                goto done;
+        }
+
+        count = index_fill_readdir (fd, dir, off, size, &entries);
+
+        /* pick ENOENT to indicate EOF */
+        op_errno = errno;
+        op_ret = count;
+done:
+        STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, &entries, xdata);
+        gf_dirent_free (&entries);
+        return 0;
+}
+
+int
+index_unlink_wrapper (call_frame_t *frame, xlator_t *this, loc_t *loc, int flag,
+                      dict_t *xdata)
+{
+        index_priv_t    *priv = NULL;
+        int32_t         op_ret = 0;
+        int32_t         op_errno = 0;
+        int             ret = 0;
+        struct  iatt    preparent = {0};
+        struct  iatt    postparent = {0};
+        char            index_dir[PATH_MAX] = {0};
+        struct  stat    lstatbuf = {0};
+        uuid_t          gfid = {0};
+
+        priv = this->private;
+        make_index_dir_path (priv->index_basepath, XATTROP_SUBDIR,
+                             index_dir, sizeof (index_dir));
+        ret = lstat (index_dir, &lstatbuf);
+        if (ret < 0) {
+                op_ret = -1;
+                op_errno = errno;
+                goto done;
+        }
+
+        iatt_from_stat (&preparent, &lstatbuf);
+        uuid_copy (preparent.ia_gfid, priv->xattrop_vgfid);
+        preparent.ia_ino = -1;
+        uuid_parse (loc->name, gfid);
+        ret = index_del (this, gfid, XATTROP_SUBDIR);
+        if (ret < 0) {
+                op_ret = -1;
+                op_errno = -ret;
+                goto done;
+        }
+        memset (&lstatbuf, 0, sizeof (lstatbuf));
+        ret = lstat (index_dir, &lstatbuf);
+        if (ret < 0) {
+                op_ret = -1;
+                op_errno = errno;
+                goto done;
+        }
+        iatt_from_stat (&postparent, &lstatbuf);
+        uuid_copy (postparent.ia_gfid, priv->xattrop_vgfid);
+        postparent.ia_ino = -1;
+done:
+        INDEX_STACK_UNWIND (unlink, frame, op_ret, op_errno, &preparent,
+                            &postparent, xdata);
+        return 0;
+}
+
+int32_t
+index_getxattr (call_frame_t *frame, xlator_t *this,
+                loc_t *loc, const char *name, dict_t *xdata)
+{
+        call_stub_t     *stub = NULL;
+        index_priv_t    *priv = NULL;
+
+        priv = this->private;
+
+        if (!name || (strcmp (GF_XATTROP_INDEX_GFID, name) &&
+		      strcmp (GF_XATTROP_INDEX_COUNT, name)))
+                goto out;
+
+        stub = fop_getxattr_stub (frame, index_getxattr_wrapper, loc, name,
+                                  xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (getxattr, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+        worker_enqueue (this, stub);
+        return 0;
+out:
+        STACK_WIND (frame, default_getxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
+        return 0;
+}
+
+int32_t
+index_lookup (call_frame_t *frame, xlator_t *this,
+              loc_t *loc, dict_t *xattr_req)
+{
+        call_stub_t     *stub = NULL;
+        index_priv_t    *priv = NULL;
+
+        priv = this->private;
+
+        if (uuid_compare (loc->gfid, priv->xattrop_vgfid) &&
+            uuid_compare (loc->pargfid, priv->xattrop_vgfid))
+                goto normal;
+
+        stub = fop_lookup_stub (frame, index_lookup_wrapper, loc, xattr_req);
+        if (!stub) {
+                STACK_UNWIND_STRICT (lookup, frame, -1, ENOMEM, loc->inode,
+                                     NULL, NULL, NULL);
+                return 0;
+        }
+        worker_enqueue (this, stub);
+        return 0;
+normal:
+        STACK_WIND (frame, default_lookup_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
+
+        return 0;
+}
+
+int32_t
+index_readdir (call_frame_t *frame, xlator_t *this,
+               fd_t *fd, size_t size, off_t off, dict_t *xdata)
+{
+        call_stub_t     *stub = NULL;
+        index_priv_t    *priv = NULL;
+
+        priv = this->private;
+        if (uuid_compare (fd->inode->gfid, priv->xattrop_vgfid))
+                goto out;
+        stub = fop_readdir_stub (frame, index_readdir_wrapper, fd, size, off,
+                                 xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (readdir, frame, -1, ENOMEM, NULL, NULL);
+                return 0;
+        }
+        worker_enqueue (this, stub);
+        return 0;
+out:
+        STACK_WIND (frame, default_readdir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readdir, fd, size, off, xdata);
+        return 0;
+}
+
+int
+index_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+              dict_t *xdata)
+{
+        call_stub_t     *stub = NULL;
+        index_priv_t    *priv = NULL;
+
+        priv = this->private;
+        if (uuid_compare (loc->pargfid, priv->xattrop_vgfid))
+                goto out;
+
+        stub = fop_unlink_stub (frame, index_unlink_wrapper, loc, xflag, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (unlink, frame, -1, ENOMEM, NULL, NULL,
+                                     NULL);
+                return 0;
+        }
+        worker_enqueue (this, stub);
+        return 0;
+out:
+        STACK_WIND (frame, default_unlink_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
+        return 0;
+}
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        ret = xlator_mem_acct_init (this, gf_index_mt_end + 1);
+
+        return ret;
+}
+
+int
+init (xlator_t *this)
+{
+        int ret = -1;
+        index_priv_t *priv = NULL;
+        pthread_t thread;
+        pthread_attr_t  w_attr;
+        gf_boolean_t    mutex_inited = _gf_false;
+        gf_boolean_t    cond_inited  = _gf_false;
+        gf_boolean_t    attr_inited  = _gf_false;
+
+	if (!this->children || this->children->next) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"'index' not configured with exactly one child");
+                goto out;
+	}
+
+	if (!this->parents) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"dangling volume. check volfile ");
+	}
+
+        priv = GF_CALLOC (1, sizeof (*priv), gf_index_mt_priv_t);
+        if (!priv)
+                goto out;
+
+        LOCK_INIT (&priv->lock);
+        if ((ret = pthread_cond_init(&priv->cond, NULL)) != 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "pthread_cond_init failed (%d)", ret);
+                goto out;
+        }
+        cond_inited = _gf_true;
+
+        if ((ret = pthread_mutex_init(&priv->mutex, NULL)) != 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "pthread_mutex_init failed (%d)", ret);
+                goto out;
+        }
+        mutex_inited = _gf_true;
+
+        if ((ret = pthread_attr_init (&w_attr)) != 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "pthread_attr_init failed (%d)", ret);
+                goto out;
+        }
+        attr_inited = _gf_true;
+
+        ret = pthread_attr_setstacksize (&w_attr, INDEX_THREAD_STACK_SIZE);
+        if (ret == EINVAL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Using default thread stack size");
+        }
+        GF_OPTION_INIT ("index-base", priv->index_basepath, path, out);
+        uuid_generate (priv->index);
+        uuid_generate (priv->xattrop_vgfid);
+        INIT_LIST_HEAD (&priv->callstubs);
+
+        this->private = priv;
+
+        ret = index_dir_create (this, XATTROP_SUBDIR);
+        if (ret < 0)
+                goto out;
+
+        ret = gf_thread_create (&thread, &w_attr, index_worker, this);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING, "Failed to create "
+                        "worker thread, aborting");
+                goto out;
+        }
+
+        ret = 0;
+out:
+        if (ret) {
+                if (cond_inited)
+                        pthread_cond_destroy (&priv->cond);
+                if (mutex_inited)
+                        pthread_mutex_destroy (&priv->mutex);
+                if (priv)
+                        GF_FREE (priv);
+                this->private = NULL;
+        }
+        if (attr_inited)
+                pthread_attr_destroy (&w_attr);
+        return ret;
+}
+
+void
+fini (xlator_t *this)
+{
+        index_priv_t *priv = NULL;
+
+        priv = this->private;
+        if (!priv)
+                goto out;
+        this->private = NULL;
+        LOCK_DESTROY (&priv->lock);
+        pthread_cond_destroy (&priv->cond);
+        pthread_mutex_destroy (&priv->mutex);
+        GF_FREE (priv);
+out:
+        return;
+}
+
+int
+index_forget (xlator_t *this, inode_t *inode)
+{
+        uint64_t tmp_cache = 0;
+        if (!inode_ctx_del (inode, this, &tmp_cache))
+                GF_FREE ((index_inode_ctx_t*) (long)tmp_cache);
+
+        return 0;
+}
+
+int32_t
+index_releasedir (xlator_t *this, fd_t *fd)
+{
+        index_fd_ctx_t *fctx = NULL;
+        uint64_t        ctx = 0;
+        int             ret = 0;
+
+        ret = fd_ctx_del (fd, this, &ctx);
+        if (ret < 0)
+                goto out;
+
+        fctx = (index_fd_ctx_t*) (long) ctx;
+        if (fctx->dir)
+                closedir (fctx->dir);
+
+        GF_FREE (fctx);
+out:
+        return 0;
+}
+
+int32_t
+index_release (xlator_t *this, fd_t *fd)
+{
+        index_fd_ctx_t *fctx = NULL;
+        uint64_t        ctx = 0;
+        int             ret = 0;
+
+        ret = fd_ctx_del (fd, this, &ctx);
+        if (ret < 0)
+                goto out;
+
+        fctx = (index_fd_ctx_t*) (long) ctx;
+        GF_FREE (fctx);
+out:
+        return 0;
+}
+
+int
+notify (xlator_t *this, int event, void *data, ...)
+{
+        int     ret = 0;
+        ret = default_notify (this, event, data);
+        return ret;
+}
+
+struct xlator_fops fops = {
+	.xattrop     = index_xattrop,
+	.fxattrop    = index_fxattrop,
+
+        //interface functions follow
+        .getxattr    = index_getxattr,
+        .lookup      = index_lookup,
+        .readdir     = index_readdir,
+        .unlink      = index_unlink
+};
+
+struct xlator_dumpops dumpops;
+
+struct xlator_cbks cbks = {
+        .forget         = index_forget,
+        .release        = index_release,
+        .releasedir     = index_releasedir
+};
+
+struct volume_options options[] = {
+        { .key  = {"index-base" },
+          .type = GF_OPTION_TYPE_PATH,
+          .description = "path where the index files need to be stored",
+        },
+        { .key  = {NULL} },
+};
diff --git a/xlators/features/index/src/index.h b/xlators/features/index/src/index.h
new file mode 100644
index 000000000..661dcdbc4
--- /dev/null
+++ b/xlators/features/index/src/index.h
@@ -0,0 +1,59 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef __INDEX_H__
+#define __INDEX_H__
+
+#include "xlator.h"
+#include "call-stub.h"
+#include "defaults.h"
+#include "byte-order.h"
+#include "common-utils.h"
+#include "index-mem-types.h"
+
+#define INDEX_THREAD_STACK_SIZE   ((size_t)(1024*1024))
+
+typedef enum {
+        UNKNOWN,
+        IN,
+        NOTIN
+} index_state_t;
+
+typedef struct index_inode_ctx {
+        gf_boolean_t processing;
+        struct list_head callstubs;
+        index_state_t state;
+} index_inode_ctx_t;
+
+typedef struct index_fd_ctx {
+        DIR *dir;
+} index_fd_ctx_t;
+
+typedef struct index_priv {
+        char *index_basepath;
+        uuid_t index;
+        gf_lock_t lock;
+        uuid_t xattrop_vgfid;//virtual gfid of the xattrop index dir
+        struct list_head callstubs;
+        pthread_mutex_t mutex;
+        pthread_cond_t  cond;
+} index_priv_t;
+
+#define INDEX_STACK_UNWIND(fop, frame, params ...)      \
+do {                                                    \
+        if (frame) {                                    \
+                inode_t *_inode = frame->local;         \
+                frame->local = NULL;                    \
+                inode_unref (_inode);                   \
+        }                                               \
+        STACK_UNWIND_STRICT (fop, frame, params);       \
+} while (0)
+
+#endif
diff --git a/xlators/features/locks/src/Makefile.am b/xlators/features/locks/src/Makefile.am
index ab545cb1c..0f79731b4 100644
--- a/xlators/features/locks/src/Makefile.am
+++ b/xlators/features/locks/src/Makefile.am
@@ -1,15 +1,18 @@
 xlator_LTLIBRARIES = locks.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
 
-locks_la_LDFLAGS = -module -avoidversion
+locks_la_LDFLAGS = -module -avoid-version
 
-locks_la_SOURCES = common.c posix.c entrylk.c inodelk.c
-locks_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la 
+locks_la_SOURCES = common.c posix.c entrylk.c inodelk.c reservelk.c \
+		   clear.c
+locks_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
 
-noinst_HEADERS = locks.h common.h
+noinst_HEADERS = locks.h common.h locks-mem-types.h clear.h
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -fno-strict-aliasing -D$(GF_HOST_OS) \
-	-I$(top_srcdir)/libglusterfs/src $(GF_CFLAGS) -shared -nostartfiles
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+
+AM_CFLAGS = -Wall -fno-strict-aliasing $(GF_CFLAGS)
 
 CLEANFILES = 
 
@@ -17,4 +20,4 @@ uninstall-local:
 	rm -f $(DESTDIR)$(xlatordir)/posix-locks.so
 
 install-data-hook:
-	ln -sf locks.so $(DESTDIR)$(xlatordir)/posix-locks.so
-\ No newline at end of file
+	ln -sf locks.so $(DESTDIR)$(xlatordir)/posix-locks.so
diff --git a/xlators/features/locks/src/clear.c b/xlators/features/locks/src/clear.c
new file mode 100644
index 000000000..75593b898
--- /dev/null
+++ b/xlators/features/locks/src/clear.c
@@ -0,0 +1,423 @@
+/*
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#include <unistd.h>
+#include <fcntl.h>
+#include <limits.h>
+#include <pthread.h>
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "compat.h"
+#include "xlator.h"
+#include "inode.h"
+#include "logging.h"
+#include "common-utils.h"
+
+#include "locks.h"
+#include "common.h"
+#include "statedump.h"
+#include "clear.h"
+
+int
+clrlk_get_kind (char *kind)
+{
+        char  *clrlk_kinds[CLRLK_KIND_MAX] = {"dummy", "blocked", "granted",
+                                              "all"};
+        int   ret_kind                     = CLRLK_KIND_MAX;
+        int   i                            = 0;
+
+        for (i = CLRLK_BLOCKED; i < CLRLK_KIND_MAX; i++) {
+                if (!strcmp (clrlk_kinds[i], kind)) {
+                        ret_kind = i;
+                        break;
+                }
+        }
+
+        return ret_kind;
+}
+
+int
+clrlk_get_type (char *type)
+{
+        char    *clrlk_types[CLRLK_TYPE_MAX]    = {"inode", "entry", "posix"};
+        int     ret_type                        = CLRLK_TYPE_MAX;
+        int     i                               = 0;
+
+        for (i = CLRLK_INODE; i < CLRLK_TYPE_MAX; i++) {
+                if (!strcmp (clrlk_types[i], type)) {
+                        ret_type = i;
+                        break;
+                }
+        }
+
+        return ret_type;
+}
+
+int
+clrlk_get_lock_range (char *range_str, struct gf_flock *ulock,
+                      gf_boolean_t *chk_range)
+{
+        int     ret     =       -1;
+
+        if (!chk_range)
+                goto out;
+
+        if (!range_str) {
+                ret = 0;
+                *chk_range = _gf_false;
+                goto out;
+        }
+
+        if (sscanf (range_str, "%hd,%"PRId64"-""%"PRId64, &ulock->l_whence,
+                    &ulock->l_start, &ulock->l_len) != 3) {
+                goto out;
+        }
+
+        ret = 0;
+        *chk_range = _gf_true;
+out:
+        return ret;
+}
+
+int
+clrlk_parse_args (const char* cmd, clrlk_args *args)
+{
+        char            *opts           = NULL;
+        char            *cur            = NULL;
+        char            *tok            = NULL;
+        char            *sptr           = NULL;
+        char            *free_ptr       = NULL;
+        char            kw[KW_MAX]     = {[KW_TYPE]     = 't',
+                                          [KW_KIND]     = 'k',
+                                          };
+        int             ret             = -1;
+        int             i               = 0;
+
+        GF_ASSERT (cmd);
+        free_ptr = opts = GF_CALLOC (1, strlen (cmd), gf_common_mt_char);
+        if (!opts)
+                goto out;
+
+        if (sscanf (cmd, GF_XATTR_CLRLK_CMD".%s", opts) < 1) {
+                ret = -1;
+                goto out;
+        }
+
+        /*clr_lk_prefix.ttype.kkind.args, args - type specific*/
+        cur = opts;
+        for (i = 0; i < KW_MAX && (tok = strtok_r (cur, ".", &sptr));
+             cur = NULL, i++) {
+                if (tok[0] != kw[i]) {
+                        ret = -1;
+                        goto out;
+                }
+                if (i == KW_TYPE)
+                        args->type = clrlk_get_type (tok+1);
+                if (i == KW_KIND)
+                        args->kind = clrlk_get_kind (tok+1);
+        }
+
+        if ((args->type == CLRLK_TYPE_MAX) || (args->kind == CLRLK_KIND_MAX))
+                goto out;
+
+        /*optional args, neither range nor basename can 'legally' contain
+         * "/" in them*/
+        tok = strtok_r (NULL, "/", &sptr);
+        if (tok)
+                args->opts = gf_strdup (tok);
+
+        ret = 0;
+out:
+        GF_FREE (free_ptr);
+        return ret;
+}
+
+int
+clrlk_clear_posixlk (xlator_t *this, pl_inode_t *pl_inode, clrlk_args *args,
+                     int *blkd, int *granted, int *op_errno)
+{
+        posix_lock_t            *plock          = NULL;
+        posix_lock_t            *tmp            = NULL;
+        struct gf_flock         ulock           = {0, };
+        int                     ret             = -1;
+        int                     bcount          = 0;
+        int                     gcount          = 0;
+        gf_boolean_t            chk_range       = _gf_false;
+
+        if (clrlk_get_lock_range (args->opts, &ulock, &chk_range)) {
+                *op_errno = EINVAL;
+                goto out;
+        }
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry_safe (plock, tmp, &pl_inode->ext_list,
+                                          list) {
+                        if ((plock->blocked &&
+                             !(args->kind & CLRLK_BLOCKED)) ||
+                            (!plock->blocked &&
+                             !(args->kind & CLRLK_GRANTED)))
+                                continue;
+
+                        if (chk_range &&
+                            (plock->user_flock.l_whence != ulock.l_whence
+                            || plock->user_flock.l_start != ulock.l_start
+                            || plock->user_flock.l_len != ulock.l_len))
+                                continue;
+
+                        list_del_init (&plock->list);
+                        if (plock->blocked) {
+                                bcount++;
+                                pl_trace_out (this, plock->frame, NULL, NULL,
+                                              F_SETLKW, &plock->user_flock,
+                                              -1, EAGAIN, NULL);
+
+                                STACK_UNWIND_STRICT (lk, plock->frame, -1, EAGAIN,
+                                                     &plock->user_flock, NULL);
+
+                        } else {
+                                gcount++;
+                        }
+                        GF_FREE (plock);
+                }
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+        grant_blocked_locks (this, pl_inode);
+        ret = 0;
+out:
+        *blkd    = bcount;
+        *granted = gcount;
+        return ret;
+}
+
+/* Returns 0 on success and -1 on failure */
+int
+clrlk_clear_inodelk (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom,
+                     clrlk_args *args, int *blkd, int *granted, int *op_errno)
+{
+        pl_inode_lock_t         *ilock          = NULL;
+        pl_inode_lock_t         *tmp            = NULL;
+        struct gf_flock         ulock           = {0, };
+        int                     ret             = -1;
+        int                     bcount          = 0;
+        int                     gcount          = 0;
+        gf_boolean_t            chk_range       = _gf_false;
+        struct list_head        released;
+
+        INIT_LIST_HEAD (&released);
+        if (clrlk_get_lock_range (args->opts, &ulock, &chk_range)) {
+                *op_errno = EINVAL;
+                goto out;
+        }
+
+        if (args->kind & CLRLK_BLOCKED)
+                goto blkd;
+
+        if (args->kind & CLRLK_GRANTED)
+                goto granted;
+
+blkd:
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry_safe (ilock, tmp, &dom->blocked_inodelks,
+                                          blocked_locks) {
+                        if (chk_range &&
+                            (ilock->user_flock.l_whence != ulock.l_whence
+                            || ilock->user_flock.l_start != ulock.l_start
+                            || ilock->user_flock.l_len != ulock.l_len))
+                                continue;
+
+                        bcount++;
+                        list_del_init (&ilock->blocked_locks);
+                        list_add (&ilock->blocked_locks, &released);
+                }
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        list_for_each_entry_safe (ilock, tmp, &released, blocked_locks) {
+                list_del_init (&ilock->blocked_locks);
+                pl_trace_out (this, ilock->frame, NULL, NULL, F_SETLKW,
+                              &ilock->user_flock, -1, EAGAIN,
+                              ilock->volume);
+                STACK_UNWIND_STRICT (inodelk, ilock->frame, -1,
+                                     EAGAIN, NULL);
+                //No need to take lock as the locks are only in one list
+                __pl_inodelk_unref (ilock);
+        }
+
+        if (!(args->kind & CLRLK_GRANTED)) {
+                ret = 0;
+                goto out;
+        }
+
+granted:
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry_safe (ilock, tmp, &dom->inodelk_list,
+                                          list) {
+                        if (chk_range &&
+                            (ilock->user_flock.l_whence != ulock.l_whence
+                            || ilock->user_flock.l_start != ulock.l_start
+                            || ilock->user_flock.l_len != ulock.l_len))
+                                continue;
+
+                        gcount++;
+                        list_del_init (&ilock->list);
+                        list_add (&ilock->list, &released);
+                }
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        list_for_each_entry_safe (ilock, tmp, &released, list) {
+                list_del_init (&ilock->list);
+                //No need to take lock as the locks are only in one list
+                __pl_inodelk_unref (ilock);
+        }
+
+        ret = 0;
+out:
+        grant_blocked_inode_locks (this, pl_inode, dom);
+        *blkd    = bcount;
+        *granted = gcount;
+        return ret;
+}
+
+/* Returns 0 on success and -1 on failure */
+int
+clrlk_clear_entrylk (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom,
+                     clrlk_args *args, int *blkd, int *granted, int *op_errno)
+{
+        pl_entry_lock_t         *elock          = NULL;
+        pl_entry_lock_t         *tmp            = NULL;
+        int                     bcount          = 0;
+        int                     gcount          = 0;
+        int                     ret             = -1;
+        struct list_head        removed;
+        struct list_head        released;
+
+        INIT_LIST_HEAD (&released);
+        if (args->kind & CLRLK_BLOCKED)
+                goto blkd;
+
+        if (args->kind & CLRLK_GRANTED)
+                goto granted;
+
+blkd:
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry_safe (elock, tmp, &dom->blocked_entrylks,
+                                          blocked_locks) {
+                        if (args->opts) {
+                                if (!elock->basename ||
+                                    strcmp (elock->basename, args->opts))
+                                        continue;
+                        }
+
+                        bcount++;
+
+                        list_del_init (&elock->blocked_locks);
+                        list_add_tail (&elock->blocked_locks, &released);
+                }
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        list_for_each_entry_safe (elock, tmp, &released, blocked_locks) {
+                list_del_init (&elock->blocked_locks);
+                entrylk_trace_out (this, elock->frame, elock->volume, NULL, NULL,
+                                   elock->basename, ENTRYLK_LOCK, elock->type,
+                                   -1, EAGAIN);
+                STACK_UNWIND_STRICT (entrylk, elock->frame, -1, EAGAIN, NULL);
+
+		__pl_entrylk_unref (elock);
+        }
+
+        if (!(args->kind & CLRLK_GRANTED)) {
+                ret = 0;
+                goto out;
+        }
+
+granted:
+        INIT_LIST_HEAD (&removed);
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry_safe (elock, tmp, &dom->entrylk_list,
+                                          domain_list) {
+                        if (args->opts) {
+                                if (!elock->basename ||
+                                    strcmp (elock->basename, args->opts))
+                                        continue;
+                        }
+
+                        gcount++;
+                        list_del_init (&elock->domain_list);
+                        list_add_tail (&elock->domain_list, &removed);
+
+			__pl_entrylk_unref (elock);
+                }
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+	grant_blocked_entry_locks (this, pl_inode, dom);
+
+        ret = 0;
+out:
+        *blkd    = bcount;
+        *granted = gcount;
+        return ret;
+}
+
+int
+clrlk_clear_lks_in_all_domains (xlator_t *this, pl_inode_t *pl_inode,
+                                clrlk_args *args, int *blkd, int *granted,
+                                int *op_errno)
+{
+        pl_dom_list_t   *dom            = NULL;
+        int             ret             = -1;
+        int             tmp_bcount      = 0;
+        int             tmp_gcount      = 0;
+
+        if (list_empty (&pl_inode->dom_list)) {
+                ret = 0;
+                goto out;
+        }
+
+        list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
+                tmp_bcount = tmp_gcount = 0;
+
+                switch (args->type)
+                {
+                case CLRLK_INODE:
+                        ret = clrlk_clear_inodelk (this, pl_inode, dom, args,
+                                             &tmp_bcount, &tmp_gcount,
+                                             op_errno);
+                        if (ret)
+                                goto out;
+                        break;
+                case CLRLK_ENTRY:
+                        ret = clrlk_clear_entrylk (this, pl_inode, dom, args,
+                                             &tmp_bcount, &tmp_gcount,
+                                             op_errno);
+                        if (ret)
+                                goto out;
+                        break;
+                }
+
+                *blkd    += tmp_bcount;
+                *granted += tmp_gcount;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
diff --git a/xlators/features/locks/src/clear.h b/xlators/features/locks/src/clear.h
new file mode 100644
index 000000000..511f3f74a
--- /dev/null
+++ b/xlators/features/locks/src/clear.h
@@ -0,0 +1,76 @@
+/*
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef __CLEAR_H__
+#define __CLEAR_H__
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "compat-errno.h"
+#include "stack.h"
+#include "call-stub.h"
+#include "locks.h"
+
+typedef enum {
+        CLRLK_INODE,
+        CLRLK_ENTRY,
+        CLRLK_POSIX,
+        CLRLK_TYPE_MAX
+} clrlk_type;
+
+typedef enum {
+        CLRLK_BLOCKED = 1,
+        CLRLK_GRANTED,
+        CLRLK_ALL,
+        CLRLK_KIND_MAX
+} clrlk_kind;
+
+typedef enum {
+        KW_TYPE,
+        KW_KIND,
+        /*add new keywords here*/
+        KW_MAX
+} clrlk_opts;
+
+struct _clrlk_args;
+typedef struct _clrlk_args clrlk_args;
+
+struct _clrlk_args {
+        int type;
+        int kind;
+        char *opts;
+};
+
+int
+clrlk_get__kind (char *kind);
+int
+clrlk_get_type (char *type);
+int
+clrlk_get_lock_range (char *range_str, struct gf_flock *ulock,
+                      gf_boolean_t *chk_range);
+int
+clrlk_parse_args (const char* cmd, clrlk_args *args);
+
+int
+clrlk_clear_posixlk (xlator_t *this, pl_inode_t *pl_inode, clrlk_args *args,
+                     int *blkd, int *granted, int *op_errno);
+int
+clrlk_clear_inodelk (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom,
+                     clrlk_args *args, int *blkd, int *granted, int *op_errno);
+int
+clrlk_clear_entrylk (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom,
+                     clrlk_args *args, int *blkd, int *granted, int *op_errno);
+int
+clrlk_clear_lks_in_all_domains (xlator_t *this, pl_inode_t *pl_inode,
+                                clrlk_args *args, int *blkd, int *granted,
+                                int *op_errno);
+#endif /* __CLEAR_H__ */
diff --git a/xlators/features/locks/src/common.c b/xlators/features/locks/src/common.c
index 58b95d207..f6c71c1cf 100644
--- a/xlators/features/locks/src/common.c
+++ b/xlators/features/locks/src/common.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006, 2007, 2008 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #include <unistd.h>
 #include <fcntl.h>
 #include <limits.h>
@@ -42,23 +32,23 @@ static int
 __is_lock_grantable (pl_inode_t *pl_inode, posix_lock_t *lock);
 static void
 __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock);
+static int
+pl_send_prelock_unlock (xlator_t *this, pl_inode_t *pl_inode,
+                        posix_lock_t *old_lock);
 
 static pl_dom_list_t *
-allocate_domain (const char *volume)
+__allocate_domain (const char *volume)
 {
         pl_dom_list_t *dom = NULL;
 
-        dom = CALLOC (1, sizeof (*dom));
+        dom = GF_CALLOC (1, sizeof (*dom),
+                         gf_locks_mt_pl_dom_list_t);
         if (!dom)
-                return NULL;
-
+                goto out;
 
-        dom->domain = strdup(volume);
-	if (!dom->domain) {
-		gf_log ("posix-locks", GF_LOG_TRACE,
-			"Out of Memory");
-		return NULL;
-	}
+        dom->domain = gf_strdup(volume);
+        if (!dom->domain)
+                goto out;
 
         gf_log ("posix-locks", GF_LOG_TRACE,
                 "New domain allocated: %s", dom->domain);
@@ -69,6 +59,12 @@ allocate_domain (const char *volume)
         INIT_LIST_HEAD (&dom->inodelk_list);
         INIT_LIST_HEAD (&dom->blocked_inodelks);
 
+out:
+        if (dom && (NULL == dom->domain)) {
+                GF_FREE (dom);
+                dom = NULL;
+        }
+
         return dom;
 }
 
@@ -80,22 +76,42 @@ get_domain (pl_inode_t *pl_inode, const char *volume)
 {
         pl_dom_list_t *dom = NULL;
 
-        list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
-                if (strcmp (dom->domain, volume) == 0)
-                        goto found;
+        GF_VALIDATE_OR_GOTO ("posix-locks", pl_inode, out);
+        GF_VALIDATE_OR_GOTO ("posix-locks", volume, out);
 
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
+                        if (strcmp (dom->domain, volume) == 0)
+                                goto unlock;
+                }
 
+                dom = __allocate_domain (volume);
+                if (dom)
+                        list_add (&dom->inode_list, &pl_inode->dom_list);
         }
-
-        dom = allocate_domain(volume);
-
-        if (dom)
-                list_add (&dom->inode_list, &pl_inode->dom_list);
-found:
-
+unlock:
+        pthread_mutex_unlock (&pl_inode->mutex);
+        if (dom) {
+                gf_log ("posix-locks", GF_LOG_TRACE, "Domain %s found", volume);
+        } else {
+                gf_log ("posix-locks", GF_LOG_TRACE, "Domain %s not found", volume);
+        }
+out:
         return dom;
 }
 
+unsigned long
+fd_to_fdnum (fd_t *fd)
+{
+        return ((unsigned long) fd);
+}
+
+fd_t *
+fd_from_fdnum (posix_lock_t *lock)
+{
+        return ((fd_t *) lock->fd_num);
+}
 
 int
 __pl_inode_is_empty (pl_inode_t *pl_inode)
@@ -120,10 +136,10 @@ __pl_inode_is_empty (pl_inode_t *pl_inode)
 void
 pl_print_locker (char *str, int size, xlator_t *this, call_frame_t *frame)
 {
-        snprintf (str, size, "Pid=%llu, lk-owner=%llu, Transport=%p, Frame=%llu",
+        snprintf (str, size, "Pid=%llu, lk-owner=%s, Client=%p, Frame=%llu",
                   (unsigned long long) frame->root->pid,
-                  (unsigned long long) frame->root->lk_owner,
-                  (void *)frame->root->trans,
+                  lkowner_utoa (&frame->root->lk_owner),
+                  frame->root->client,
                   (unsigned long long) frame->root->unique);
 }
 
@@ -146,25 +162,24 @@ pl_print_lockee (char *str, int size, fd_t *fd, loc_t *loc)
         }
 
         if (loc && loc->path) {
-                ipath = strdup (loc->path);
+                ipath = gf_strdup (loc->path);
         } else {
                 ret = inode_path (inode, NULL, &ipath);
                 if (ret <= 0)
                         ipath = NULL;
         }
 
-        snprintf (str, size, "ino=%llu, fd=%p, path=%s",
-                  (unsigned long long) inode->ino, fd,
+        snprintf (str, size, "gfid=%s, fd=%p, path=%s",
+                  uuid_utoa (inode->gfid), fd,
                   ipath ? ipath : "<nul>");
 
-        if (ipath)
-                FREE (ipath);
+        GF_FREE (ipath);
 }
 
 
 void
 pl_print_lock (char *str, int size, int cmd,
-               struct flock *flock, uint64_t owner)
+               struct gf_flock *flock, gf_lkowner_t *owner)
 {
         char *cmd_str = NULL;
         char *type_str = NULL;
@@ -212,17 +227,17 @@ pl_print_lock (char *str, int size, int cmd,
         }
 
         snprintf (str, size, "lock=FCNTL, cmd=%s, type=%s, "
-                  "start=%llu, len=%llu, pid=%llu, lk-owner=%llu",
+                  "start=%llu, len=%llu, pid=%llu, lk-owner=%s",
                   cmd_str, type_str, (unsigned long long) flock->l_start,
                   (unsigned long long) flock->l_len,
                   (unsigned long long) flock->l_pid,
-                  (unsigned long long) owner);
+                  lkowner_utoa (owner));
 }
 
 
 void
 pl_trace_in (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
-             int cmd, struct flock *flock, const char *domain)
+             int cmd, struct gf_flock *flock, const char *domain)
 {
         posix_locks_private_t  *priv = NULL;
         char                    pl_locker[256];
@@ -239,9 +254,9 @@ pl_trace_in (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
         if (domain)
                 pl_print_inodelk (pl_lock, 256, cmd, flock, domain);
         else
-                pl_print_lock (pl_lock, 256, cmd, flock, frame->root->lk_owner);
+                pl_print_lock (pl_lock, 256, cmd, flock, &frame->root->lk_owner);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[REQUEST] Locker = {%s} Lockee = {%s} Lock = {%s}",
                 pl_locker, pl_lockee, pl_lock);
 }
@@ -264,13 +279,13 @@ pl_print_verdict (char *str, int size, int op_ret, int op_errno)
                 }
         }
 
-        snprintf (str, size, verdict);
+        snprintf (str, size, "%s", verdict);
 }
 
 
 void
 pl_trace_out (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
-              int cmd, struct flock *flock, int op_ret, int op_errno, const char *domain)
+              int cmd, struct gf_flock *flock, int op_ret, int op_errno, const char *domain)
 
 {
         posix_locks_private_t  *priv = NULL;
@@ -289,11 +304,11 @@ pl_trace_out (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
         if (domain)
                 pl_print_inodelk (pl_lock, 256, cmd, flock, domain);
         else
-                pl_print_lock (pl_lock, 256, cmd, flock, frame->root->lk_owner);
+                pl_print_lock (pl_lock, 256, cmd, flock, &frame->root->lk_owner);
 
         pl_print_verdict (verdict, 32, op_ret, op_errno);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[%s] Locker = {%s} Lockee = {%s} Lock = {%s}",
                 verdict, pl_locker, pl_lockee, pl_lock);
 }
@@ -301,7 +316,7 @@ pl_trace_out (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
 
 void
 pl_trace_block (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
-                int cmd, struct flock *flock, const char *domain)
+                int cmd, struct gf_flock *flock, const char *domain)
 
 {
         posix_locks_private_t  *priv = NULL;
@@ -319,9 +334,9 @@ pl_trace_block (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
         if (domain)
                 pl_print_inodelk (pl_lock, 256, cmd, flock, domain);
         else
-                pl_print_lock (pl_lock, 256, cmd, flock, frame->root->lk_owner);
+                pl_print_lock (pl_lock, 256, cmd, flock, &frame->root->lk_owner);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[BLOCKED] Locker = {%s} Lockee = {%s} Lock = {%s}",
                 pl_locker, pl_lockee, pl_lock);
 }
@@ -348,17 +363,36 @@ pl_trace_flush (xlator_t *this, call_frame_t *frame, fd_t *fd)
         pl_print_locker (pl_locker, 256, this, frame);
         pl_print_lockee (pl_lockee, 256, fd, NULL);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[FLUSH] Locker = {%s} Lockee = {%s}",
                 pl_locker, pl_lockee);
 }
 
 void
+pl_trace_release (xlator_t *this, fd_t *fd)
+{
+        posix_locks_private_t  *priv = NULL;
+        char                    pl_lockee[256];
+
+        priv = this->private;
+
+        if (!priv->trace)
+                return;
+
+        pl_print_lockee (pl_lockee, 256, fd, NULL);
+
+        gf_log (this->name, GF_LOG_INFO,
+                "[RELEASE] Lockee = {%s}", pl_lockee);
+}
+
+
+void
 pl_update_refkeeper (xlator_t *this, inode_t *inode)
 {
         pl_inode_t *pl_inode  = NULL;
         int         is_empty  = 0;
         int         need_unref = 0;
+        int         need_ref = 0;
 
         pl_inode = pl_inode_get (this, inode);
 
@@ -372,13 +406,17 @@ pl_update_refkeeper (xlator_t *this, inode_t *inode)
                 }
 
                 if (!is_empty && !pl_inode->refkeeper) {
-                        pl_inode->refkeeper = inode_ref (inode);
+                        need_ref = 1;
+                        pl_inode->refkeeper = inode;
                 }
         }
         pthread_mutex_unlock (&pl_inode->mutex);
 
         if (need_unref)
                 inode_unref (inode);
+
+        if (need_ref)
+                inode_ref (inode);
 }
 
 
@@ -386,69 +424,78 @@ pl_inode_t *
 pl_inode_get (xlator_t *this, inode_t *inode)
 {
         uint64_t    tmp_pl_inode = 0;
-	pl_inode_t *pl_inode = NULL;
-	mode_t      st_mode = 0;
-	int         ret = 0;
+        pl_inode_t *pl_inode = NULL;
+        int         ret = 0;
 
-	ret = inode_ctx_get (inode, this,&tmp_pl_inode);
-	if (ret == 0) {
-                pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
-		goto out;
-        }
-	pl_inode = CALLOC (1, sizeof (*pl_inode));
-	if (!pl_inode) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		goto out;
-	}
-
-        gf_log (this->name, GF_LOG_TRACE,
-                "Allocating new pl inode");
-
-	st_mode  = inode->st_mode;
-	if ((st_mode & S_ISGID) && !(st_mode & S_IXGRP))
-		pl_inode->mandatory = 1;
+        LOCK (&inode->lock);
+        {
+                ret = __inode_ctx_get (inode, this, &tmp_pl_inode);
+                if (ret == 0) {
+                        pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
+                        goto unlock;
+                }
+                pl_inode = GF_CALLOC (1, sizeof (*pl_inode),
+                                      gf_locks_mt_pl_inode_t);
+                if (!pl_inode) {
+                        goto unlock;
+                }
 
+                gf_log (this->name, GF_LOG_TRACE,
+                        "Allocating new pl inode");
 
-	pthread_mutex_init (&pl_inode->mutex, NULL);
+                pthread_mutex_init (&pl_inode->mutex, NULL);
 
-	INIT_LIST_HEAD (&pl_inode->dom_list);
-	INIT_LIST_HEAD (&pl_inode->ext_list);
-	INIT_LIST_HEAD (&pl_inode->rw_list);
+                INIT_LIST_HEAD (&pl_inode->dom_list);
+                INIT_LIST_HEAD (&pl_inode->ext_list);
+                INIT_LIST_HEAD (&pl_inode->rw_list);
+                INIT_LIST_HEAD (&pl_inode->reservelk_list);
+                INIT_LIST_HEAD (&pl_inode->blocked_reservelks);
+                INIT_LIST_HEAD (&pl_inode->blocked_calls);
 
-	ret = inode_ctx_put (inode, this, (uint64_t)(long)(pl_inode));
+                __inode_ctx_put (inode, this, (uint64_t)(long)(pl_inode));
+        }
+unlock:
+        UNLOCK (&inode->lock);
 
-out:
-	return pl_inode;
+        return pl_inode;
 }
 
 
 /* Create a new posix_lock_t */
 posix_lock_t *
-new_posix_lock (struct flock *flock, transport_t *transport, pid_t client_pid, uint64_t owner)
+new_posix_lock (struct gf_flock *flock, client_t *client, pid_t client_pid,
+                gf_lkowner_t *owner, fd_t *fd)
 {
-	posix_lock_t *lock = NULL;
+        posix_lock_t *lock = NULL;
 
-	lock = CALLOC (1, sizeof (posix_lock_t));
-	if (!lock) {
-		return NULL;
-	}
+        GF_VALIDATE_OR_GOTO ("posix-locks", flock, out);
+        GF_VALIDATE_OR_GOTO ("posix-locks", client, out);
+        GF_VALIDATE_OR_GOTO ("posix-locks", fd, out);
 
-	lock->fl_start = flock->l_start;
-	lock->fl_type  = flock->l_type;
+        lock = GF_CALLOC (1, sizeof (posix_lock_t),
+                          gf_locks_mt_posix_lock_t);
+        if (!lock) {
+                goto out;
+        }
 
-	if (flock->l_len == 0)
-		lock->fl_end = LLONG_MAX;
-	else
-		lock->fl_end = flock->l_start + flock->l_len - 1;
+        lock->fl_start = flock->l_start;
+        lock->fl_type  = flock->l_type;
 
-	lock->transport  = transport;
-	lock->client_pid = client_pid;
-        lock->owner      = owner;
+        if (flock->l_len == 0)
+                lock->fl_end = LLONG_MAX;
+        else
+                lock->fl_end = flock->l_start + flock->l_len - 1;
+
+        lock->client     = client;
+        lock->fd_num     = fd_to_fdnum (fd);
+        lock->fd         = fd;
+        lock->client_pid = client_pid;
+        lock->owner      = *owner;
 
-	INIT_LIST_HEAD (&lock->list);
+        INIT_LIST_HEAD (&lock->list);
 
-	return lock;
+out:
+        return lock;
 }
 
 
@@ -456,7 +503,7 @@ new_posix_lock (struct flock *flock, transport_t *transport, pid_t client_pid, u
 void
 __delete_lock (pl_inode_t *pl_inode, posix_lock_t *lock)
 {
-	list_del_init (&lock->list);
+        list_del_init (&lock->list);
 }
 
 
@@ -464,32 +511,37 @@ __delete_lock (pl_inode_t *pl_inode, posix_lock_t *lock)
 void
 __destroy_lock (posix_lock_t *lock)
 {
-	free (lock);
+        GF_FREE (lock);
 }
 
 
-/* Convert a posix_lock to a struct flock */
+/* Convert a posix_lock to a struct gf_flock */
 void
-posix_lock_to_flock (posix_lock_t *lock, struct flock *flock)
+posix_lock_to_flock (posix_lock_t *lock, struct gf_flock *flock)
 {
-	flock->l_pid   = lock->client_pid;
-	flock->l_type  = lock->fl_type;
-	flock->l_start = lock->fl_start;
+        flock->l_pid   = lock->client_pid;
+        flock->l_type  = lock->fl_type;
+        flock->l_start = lock->fl_start;
+        flock->l_owner = lock->owner;
 
-	if (lock->fl_end == LLONG_MAX)
-		flock->l_len = 0;
-	else
-		flock->l_len = lock->fl_end - lock->fl_start + 1;
+        if (lock->fl_end == LLONG_MAX)
+                flock->l_len = 0;
+        else
+                flock->l_len = lock->fl_end - lock->fl_start + 1;
 }
 
-
 /* Insert the lock into the inode's lock list */
 static void
 __insert_lock (pl_inode_t *pl_inode, posix_lock_t *lock)
 {
-	list_add_tail (&lock->list, &pl_inode->ext_list);
+        if (lock->blocked)
+                gettimeofday (&lock->blkd_time, NULL);
+        else
+                gettimeofday (&lock->granted_time, NULL);
+
+        list_add_tail (&lock->list, &pl_inode->ext_list);
 
-	return;
+        return;
 }
 
 
@@ -497,14 +549,14 @@ __insert_lock (pl_inode_t *pl_inode, posix_lock_t *lock)
 int
 locks_overlap (posix_lock_t *l1, posix_lock_t *l2)
 {
-	/* 
-	   Note:
-	   FUSE always gives us absolute offsets, so no need to worry 
-	   about SEEK_CUR or SEEK_END
-	*/
-
-	return ((l1->fl_end >= l2->fl_start) &&
-		(l2->fl_end >= l1->fl_start));
+        /*
+           Note:
+           FUSE always gives us absolute offsets, so no need to worry
+           about SEEK_CUR or SEEK_END
+        */
+
+        return ((l1->fl_end >= l2->fl_start) &&
+                (l2->fl_end >= l1->fl_start));
 }
 
 
@@ -513,8 +565,8 @@ int
 same_owner (posix_lock_t *l1, posix_lock_t *l2)
 {
 
-                return ((l1->owner == l2->owner) &&
-                        (l1->transport  == l2->transport));
+        return (is_same_lkowner (&l1->owner, &l2->owner) &&
+                (l1->client == l2->client));
 
 }
 
@@ -523,15 +575,15 @@ same_owner (posix_lock_t *l1, posix_lock_t *l2)
 void
 __delete_unlck_locks (pl_inode_t *pl_inode)
 {
-	posix_lock_t *l = NULL;
-	posix_lock_t *tmp = NULL;
+        posix_lock_t *l = NULL;
+        posix_lock_t *tmp = NULL;
 
-	list_for_each_entry_safe (l, tmp, &pl_inode->ext_list, list) {
-		if (l->fl_type == F_UNLCK) {
-			__delete_lock (pl_inode, l);
-			__destroy_lock (l);
-		}
-	}
+        list_for_each_entry_safe (l, tmp, &pl_inode->ext_list, list) {
+                if (l->fl_type == F_UNLCK) {
+                        __delete_lock (pl_inode, l);
+                        __destroy_lock (l);
+                }
+        }
 }
 
 
@@ -539,86 +591,160 @@ __delete_unlck_locks (pl_inode_t *pl_inode)
 static posix_lock_t *
 add_locks (posix_lock_t *l1, posix_lock_t *l2)
 {
-	posix_lock_t *sum = NULL;
+        posix_lock_t *sum = NULL;
 
-	sum = CALLOC (1, sizeof (posix_lock_t));
-	if (!sum)
-		return NULL;
+        sum = GF_CALLOC (1, sizeof (posix_lock_t),
+                         gf_locks_mt_posix_lock_t);
+        if (!sum)
+                return NULL;
 
-	sum->fl_start = min (l1->fl_start, l2->fl_start);
-	sum->fl_end   = max (l1->fl_end, l2->fl_end);
+        sum->fl_start = min (l1->fl_start, l2->fl_start);
+        sum->fl_end   = max (l1->fl_end, l2->fl_end);
 
-	return sum;
+        return sum;
 }
 
 /* Subtract two locks */
 struct _values {
-	posix_lock_t *locks[3];
+        posix_lock_t *locks[3];
 };
 
 /* {big} must always be contained inside {small} */
 static struct _values
 subtract_locks (posix_lock_t *big, posix_lock_t *small)
 {
-	struct _values v = { .locks = {0, 0, 0} };
-  
-	if ((big->fl_start == small->fl_start) && 
-	    (big->fl_end   == small->fl_end)) {  
-		/* both edges coincide with big */
-		v.locks[0] = CALLOC (1, sizeof (posix_lock_t));
-		ERR_ABORT (v.locks[0]);
-		memcpy (v.locks[0], big, sizeof (posix_lock_t));
-		v.locks[0]->fl_type = small->fl_type;
-	}
-	else if ((small->fl_start > big->fl_start) &&
-		 (small->fl_end   < big->fl_end)) {
-		/* both edges lie inside big */
-		v.locks[0] = CALLOC (1, sizeof (posix_lock_t));
-		ERR_ABORT (v.locks[0]);
-		v.locks[1] = CALLOC (1, sizeof (posix_lock_t));
-		ERR_ABORT (v.locks[1]);
-		v.locks[2] = CALLOC (1, sizeof (posix_lock_t));
-		ERR_ABORT (v.locks[2]);
-
-		memcpy (v.locks[0], big, sizeof (posix_lock_t));
-		v.locks[0]->fl_end = small->fl_start - 1;
-
-		memcpy (v.locks[1], small, sizeof (posix_lock_t));
-		memcpy (v.locks[2], big, sizeof (posix_lock_t));
-		v.locks[2]->fl_start = small->fl_end + 1;
-	}
-	/* one edge coincides with big */
-	else if (small->fl_start == big->fl_start) {
-		v.locks[0] = CALLOC (1, sizeof (posix_lock_t));
-		ERR_ABORT (v.locks[0]);
-		v.locks[1] = CALLOC (1, sizeof (posix_lock_t));
-		ERR_ABORT (v.locks[1]);
-    
-		memcpy (v.locks[0], big, sizeof (posix_lock_t));
-		v.locks[0]->fl_start = small->fl_end + 1;
-    
-		memcpy (v.locks[1], small, sizeof (posix_lock_t));
-	}
-	else if (small->fl_end   == big->fl_end) {
-		v.locks[0] = CALLOC (1, sizeof (posix_lock_t));
-		ERR_ABORT (v.locks[0]);
-		v.locks[1] = CALLOC (1, sizeof (posix_lock_t));
-		ERR_ABORT (v.locks[1]);
-
-		memcpy (v.locks[0], big, sizeof (posix_lock_t));
-		v.locks[0]->fl_end = small->fl_start - 1;
-    
-		memcpy (v.locks[1], small, sizeof (posix_lock_t));
-	}
-        else {
-                gf_log ("posix-locks", GF_LOG_ERROR,
-                        "Unexpected case in subtract_locks. Please send "
-                        "a bug report to gluster-devel@nongnu.org");
+
+        struct _values v = { .locks = {0, 0, 0} };
+
+        if ((big->fl_start == small->fl_start) &&
+            (big->fl_end   == small->fl_end)) {
+                /* both edges coincide with big */
+                v.locks[0] = GF_CALLOC (1, sizeof (posix_lock_t),
+                                        gf_locks_mt_posix_lock_t);
+                if (!v.locks[0])
+                        goto out;
+                memcpy (v.locks[0], big, sizeof (posix_lock_t));
+                v.locks[0]->fl_type = small->fl_type;
+                goto done;
+        }
+
+        if ((small->fl_start > big->fl_start) &&
+            (small->fl_end   < big->fl_end)) {
+                /* both edges lie inside big */
+                v.locks[0] = GF_CALLOC (1, sizeof (posix_lock_t),
+                                        gf_locks_mt_posix_lock_t);
+                if (!v.locks[0])
+                        goto out;
+
+                v.locks[1] = GF_CALLOC (1, sizeof (posix_lock_t),
+                                        gf_locks_mt_posix_lock_t);
+                if (!v.locks[1])
+                        goto out;
+
+                v.locks[2] = GF_CALLOC (1, sizeof (posix_lock_t),
+                                        gf_locks_mt_posix_lock_t);
+                if (!v.locks[1])
+                        goto out;
+
+                memcpy (v.locks[0], big, sizeof (posix_lock_t));
+                v.locks[0]->fl_end = small->fl_start - 1;
+
+                memcpy (v.locks[1], small, sizeof (posix_lock_t));
+
+                memcpy (v.locks[2], big, sizeof (posix_lock_t));
+                v.locks[2]->fl_start = small->fl_end + 1;
+                goto done;
+
+        }
+
+        /* one edge coincides with big */
+        if (small->fl_start == big->fl_start) {
+                v.locks[0] = GF_CALLOC (1, sizeof (posix_lock_t),
+                                        gf_locks_mt_posix_lock_t);
+                if (!v.locks[0])
+                        goto out;
+
+                v.locks[1] = GF_CALLOC (1, sizeof (posix_lock_t),
+                                        gf_locks_mt_posix_lock_t);
+                if (!v.locks[1])
+                        goto out;
+
+                memcpy (v.locks[0], big, sizeof (posix_lock_t));
+                v.locks[0]->fl_start = small->fl_end + 1;
+
+                memcpy (v.locks[1], small, sizeof (posix_lock_t));
+                goto done;
         }
 
+        if (small->fl_end  == big->fl_end) {
+                v.locks[0] = GF_CALLOC (1, sizeof (posix_lock_t),
+                                        gf_locks_mt_posix_lock_t);
+                if (!v.locks[0])
+                        goto out;
+
+                v.locks[1] = GF_CALLOC (1, sizeof (posix_lock_t),
+                                        gf_locks_mt_posix_lock_t);
+                if (!v.locks[1])
+                        goto out;
+
+                memcpy (v.locks[0], big, sizeof (posix_lock_t));
+                v.locks[0]->fl_end = small->fl_start - 1;
+
+                memcpy (v.locks[1], small, sizeof (posix_lock_t));
+                goto done;
+        }
+
+        GF_ASSERT (0);
+        gf_log ("posix-locks", GF_LOG_ERROR, "Unexpected case in subtract_locks");
+
+out:
+        if (v.locks[0]) {
+                GF_FREE (v.locks[0]);
+                v.locks[0] = NULL;
+        }
+        if (v.locks[1]) {
+                GF_FREE (v.locks[1]);
+                v.locks[1] = NULL;
+        }
+        if (v.locks[2]) {
+                GF_FREE (v.locks[2]);
+                v.locks[2] = NULL;
+        }
+
+done:
         return v;
 }
 
+static posix_lock_t *
+first_conflicting_overlap (pl_inode_t *pl_inode, posix_lock_t *lock)
+{
+        posix_lock_t *l = NULL;
+        posix_lock_t *conf = NULL;
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry (l, &pl_inode->ext_list, list) {
+                        if (l->blocked)
+                                continue;
+
+                        if (locks_overlap (l, lock)) {
+                                if (same_owner (l, lock))
+                                        continue;
+
+                                if ((l->fl_type == F_WRLCK) ||
+                                    (lock->fl_type == F_WRLCK)) {
+                                        conf = l;
+                                        goto unlock;
+                                }
+                        }
+                }
+        }
+unlock:
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        return conf;
+}
+
 /*
   Start searching from {begin}, and return the first lock that
   conflicts, NULL if no conflict
@@ -677,6 +803,8 @@ __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock)
         struct _values v = { .locks = {0, 0, 0} };
 
         list_for_each_entry_safe (conf, t, &pl_inode->ext_list, list) {
+                if (conf->blocked)
+                        continue;
                 if (!locks_overlap (conf, lock))
                         continue;
 
@@ -685,7 +813,8 @@ __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock)
                                 sum = add_locks (lock, conf);
 
                                 sum->fl_type    = lock->fl_type;
-                                sum->transport  = lock->transport;
+                                sum->client     = lock->client;
+                                sum->fd_num     = lock->fd_num;
                                 sum->client_pid = lock->client_pid;
                                 sum->owner      = lock->owner;
 
@@ -693,6 +822,8 @@ __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock)
                                 __destroy_lock (conf);
 
                                 __destroy_lock (lock);
+                                INIT_LIST_HEAD (&sum->list);
+                                posix_lock_to_flock (sum, &sum->user_flock);
                                 __insert_and_merge (pl_inode, sum);
 
                                 return;
@@ -700,7 +831,8 @@ __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock)
                                 sum = add_locks (lock, conf);
 
                                 sum->fl_type    = conf->fl_type;
-                                sum->transport  = conf->transport;
+                                sum->client     = conf->client;
+                                sum->fd_num     = conf->fd_num;
                                 sum->client_pid = conf->client_pid;
                                 sum->owner      = conf->owner;
 
@@ -719,6 +851,8 @@ __insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock)
                                                 continue;
 
                                         INIT_LIST_HEAD (&v.locks[i]->list);
+                                        posix_lock_to_flock (v.locks[i],
+                                                       &v.locks[i]->user_flock);
                                         __insert_and_merge (pl_inode,
                                                             v.locks[i]);
                                 }
@@ -772,7 +906,8 @@ __grant_blocked_locks (xlator_t *this, pl_inode_t *pl_inode, struct list_head *g
                 list_del_init (&l->list);
 
                 if (__is_lock_grantable (pl_inode, l)) {
-                        conf = CALLOC (1, sizeof (*conf));
+                        conf = GF_CALLOC (1, sizeof (*conf),
+                                          gf_locks_mt_posix_lock_t);
 
                         if (!conf) {
                                 l->blocked = 1;
@@ -786,10 +921,9 @@ __grant_blocked_locks (xlator_t *this, pl_inode_t *pl_inode, struct list_head *g
                         posix_lock_to_flock (l, &conf->user_flock);
 
                         gf_log (this->name, GF_LOG_TRACE,
-                                "%s (pid=%d) lk-owner:%"PRIu64" %"PRId64" - %"PRId64" => Granted",
+                                "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => Granted",
                                 l->fl_type == F_UNLCK ? "Unlock" : "Lock",
-                                l->client_pid,
-                                l->owner,
+                                l->client_pid, lkowner_utoa (&l->owner),
                                 l->user_flock.l_start,
                                 l->user_flock.l_len);
 
@@ -825,14 +959,61 @@ grant_blocked_locks (xlator_t *this, pl_inode_t *pl_inode)
                 pl_trace_out (this, lock->frame, NULL, NULL, F_SETLKW,
                               &lock->user_flock, 0, 0, NULL);
 
-                STACK_UNWIND (lock->frame, 0, 0, &lock->user_flock);
+                STACK_UNWIND_STRICT (lk, lock->frame, 0, 0,
+                                     &lock->user_flock, NULL);
 
-                FREE (lock);
+                GF_FREE (lock);
         }
 
         return;
 }
 
+static int
+pl_send_prelock_unlock (xlator_t *this, pl_inode_t *pl_inode,
+                        posix_lock_t *old_lock)
+{
+        struct gf_flock  flock       = {0,};
+        posix_lock_t *unlock_lock = NULL;
+
+        struct list_head granted_list;
+        posix_lock_t     *tmp = NULL;
+        posix_lock_t     *lock = NULL;
+
+        int ret = -1;
+
+        INIT_LIST_HEAD (&granted_list);
+
+        flock.l_type   = F_UNLCK;
+        flock.l_whence = old_lock->user_flock.l_whence;
+        flock.l_start  = old_lock->user_flock.l_start;
+        flock.l_len    = old_lock->user_flock.l_len;
+
+
+        unlock_lock = new_posix_lock (&flock, old_lock->client,
+                                      old_lock->client_pid, &old_lock->owner,
+                                      old_lock->fd);
+        GF_VALIDATE_OR_GOTO (this->name, unlock_lock, out);
+        ret = 0;
+
+        __insert_and_merge (pl_inode, unlock_lock);
+
+        __grant_blocked_locks (this, pl_inode, &granted_list);
+
+        list_for_each_entry_safe (lock, tmp, &granted_list, list) {
+                list_del_init (&lock->list);
+
+                pl_trace_out (this, lock->frame, NULL, NULL, F_SETLKW,
+                              &lock->user_flock, 0, 0, NULL);
+
+                STACK_UNWIND_STRICT (lk, lock->frame, 0, 0,
+                                     &lock->user_flock, NULL);
+
+                GF_FREE (lock);
+        }
+
+out:
+        return ret;
+}
 
 int
 pl_setlk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
@@ -844,21 +1025,38 @@ pl_setlk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
 
         pthread_mutex_lock (&pl_inode->mutex);
         {
+                /* Send unlock before the actual lock to
+                   prevent lock upgrade / downgrade
+                   problems only if:
+                   - it is a blocking call
+                   - it has other conflicting locks
+                */
+
+                if (can_block &&
+                    !(__is_lock_grantable (pl_inode, lock))) {
+                        ret = pl_send_prelock_unlock (this, pl_inode,
+                                                      lock);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_DEBUG,
+                                "Could not send pre-lock "
+                                        "unlock");
+                }
+
                 if (__is_lock_grantable (pl_inode, lock)) {
                         gf_log (this->name, GF_LOG_TRACE,
-                                "%s (pid=%d) lk-owner:%"PRIu64" %"PRId64" - %"PRId64" => OK",
+                                "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => OK",
                                 lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
                                 lock->client_pid,
-                                lock->owner,
+                                lkowner_utoa (&lock->owner),
                                 lock->user_flock.l_start,
                                 lock->user_flock.l_len);
                         __insert_and_merge (pl_inode, lock);
                 } else if (can_block) {
                         gf_log (this->name, GF_LOG_TRACE,
-                                "%s (pid=%d) lk-owner:%"PRIu64" %"PRId64" - %"PRId64" => Blocked",
+                                "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => Blocked",
                                 lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
                                 lock->client_pid,
-                                lock->owner,
+                                lkowner_utoa (&lock->owner),
                                 lock->user_flock.l_start,
                                 lock->user_flock.l_len);
                         lock->blocked = 1;
@@ -866,10 +1064,10 @@ pl_setlk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
                         ret = -1;
                 } else {
                         gf_log (this->name, GF_LOG_TRACE,
-                                "%s (pid=%d) lk-owner:%"PRIu64" %"PRId64" - %"PRId64" => NOK",
+                                "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => NOK",
                                 lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
                                 lock->client_pid,
-                                lock->owner,
+                                lkowner_utoa (&lock->owner),
                                 lock->user_flock.l_start,
                                 lock->user_flock.l_len);
                         errno = EAGAIN;
@@ -891,7 +1089,7 @@ pl_getlk (pl_inode_t *pl_inode, posix_lock_t *lock)
 {
         posix_lock_t *conf = NULL;
 
-        conf = first_overlap (pl_inode, lock);
+        conf = first_conflicting_overlap (pl_inode, lock);
 
         if (conf == NULL) {
                 lock->fl_type = F_UNLCK;
@@ -900,3 +1098,4 @@ pl_getlk (pl_inode_t *pl_inode, posix_lock_t *lock)
 
         return conf;
 }
+
diff --git a/xlators/features/locks/src/common.h b/xlators/features/locks/src/common.h
index 0d847f770..5ec630ee8 100644
--- a/xlators/features/locks/src/common.h
+++ b/xlators/features/locks/src/common.h
@@ -1,27 +1,41 @@
 /*
-  Copyright (c) 2006, 2007, 2008 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef __COMMON_H__
 #define __COMMON_H__
 
+#include "lkowner.h"
+/*dump locks format strings */
+#define RANGE_FMT               "type=%s, whence=%hd, start=%llu, len=%llu"
+#define ENTRY_FMT               "type=%s on basename=%s"
+#define DUMP_GEN_FMT            "pid = %llu, owner=%s, client=%p"
+#define GRNTD_AT                "granted at %s"
+#define BLKD_AT                 "blocked at %s"
+#define CONN_ID                 "connection-id=%s"
+#define DUMP_BLKD_FMT           DUMP_GEN_FMT", "CONN_ID", "BLKD_AT
+#define DUMP_GRNTD_FMT          DUMP_GEN_FMT", "CONN_ID", "GRNTD_AT
+#define DUMP_BLKD_GRNTD_FMT     DUMP_GEN_FMT", "CONN_ID", "BLKD_AT", "GRNTD_AT
+
+#define ENTRY_BLKD_FMT          ENTRY_FMT", "DUMP_BLKD_FMT
+#define ENTRY_GRNTD_FMT         ENTRY_FMT", "DUMP_GRNTD_FMT
+#define ENTRY_BLKD_GRNTD_FMT    ENTRY_FMT", "DUMP_BLKD_GRNTD_FMT
+
+#define RANGE_BLKD_FMT          RANGE_FMT", "DUMP_BLKD_FMT
+#define RANGE_GRNTD_FMT         RANGE_FMT", "DUMP_GRNTD_FMT
+#define RANGE_BLKD_GRNTD_FMT    RANGE_FMT", "DUMP_BLKD_GRNTD_FMT
+
+#define SET_FLOCK_PID(flock, lock) ((flock)->l_pid = lock->client_pid)
+
+
 posix_lock_t *
-new_posix_lock (struct flock *flock, transport_t *transport, pid_t client_pid, uint64_t owner);
+new_posix_lock (struct gf_flock *flock, client_t *client, pid_t client_pid,
+                gf_lkowner_t *owner, fd_t *fd);
 
 pl_inode_t *
 pl_inode_get (xlator_t *this, inode_t *inode);
@@ -37,7 +51,7 @@ void
 grant_blocked_locks (xlator_t *this, pl_inode_t *inode);
 
 void
-posix_lock_to_flock (posix_lock_t *lock, struct flock *flock);
+posix_lock_to_flock (posix_lock_t *lock, struct gf_flock *flock);
 
 int
 locks_overlap (posix_lock_t *l1, posix_lock_t *l2);
@@ -53,34 +67,39 @@ pl_dom_list_t *
 get_domain (pl_inode_t *pl_inode, const char *volume);
 
 void
-grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode, pl_inode_lock_t *lock, pl_dom_list_t *dom);
+grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode,
+                           pl_dom_list_t *dom);
 
 void
 __delete_inode_lock (pl_inode_lock_t *lock);
 
 void
-__destroy_inode_lock (pl_inode_lock_t *lock);
+__pl_inodelk_unref (pl_inode_lock_t *lock);
 
 void
 grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode,
-                           pl_entry_lock_t *unlocked, pl_dom_list_t *dom);
+                           pl_dom_list_t *dom);
 
 void pl_update_refkeeper (xlator_t *this, inode_t *inode);
 
 int32_t
-get_inodelk_count (xlator_t *this, inode_t *inode);
+__get_inodelk_count (xlator_t *this, pl_inode_t *pl_inode, char *domname);
+int32_t
+get_inodelk_count (xlator_t *this, inode_t *inode, char *domname);
 
 int32_t
+__get_entrylk_count (xlator_t *this, pl_inode_t *pl_inode);
+int32_t
 get_entrylk_count (xlator_t *this, inode_t *inode);
 
 void pl_trace_in (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
-                  int cmd, struct flock *flock, const char *domain);
+                  int cmd, struct gf_flock *flock, const char *domain);
 
 void pl_trace_out (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
-                   int cmd, struct flock *flock, int op_ret, int op_errno, const char *domain);
+                   int cmd, struct gf_flock *flock, int op_ret, int op_errno, const char *domain);
 
 void pl_trace_block (xlator_t *this, call_frame_t *frame, fd_t *fd, loc_t *loc,
-                     int cmd, struct flock *flock, const char *domain);
+                     int cmd, struct gf_flock *flock, const char *domain);
 
 void pl_trace_flush (xlator_t *this, call_frame_t *frame, fd_t *fd);
 
@@ -107,6 +126,33 @@ void
 pl_print_locker (char *str, int size, xlator_t *this, call_frame_t *frame);
 
 void
-pl_print_inodelk (char *str, int size, int cmd, struct flock *flock, const char *domain);
+pl_print_inodelk (char *str, int size, int cmd, struct gf_flock *flock, const char *domain);
+
+void
+pl_trace_release (xlator_t *this, fd_t *fd);
+
+unsigned long
+fd_to_fdnum (fd_t *fd);
+
+fd_t *
+fd_from_fdnum (posix_lock_t *lock);
+
+int
+pl_reserve_setlk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
+                  int can_block);
+int
+reservelks_equal (posix_lock_t *l1, posix_lock_t *l2);
+
+int
+pl_verify_reservelk (xlator_t *this, pl_inode_t *pl_inode,
+                     posix_lock_t *lock, int can_block);
+int
+pl_reserve_unlock (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *reqlock);
+
+uint32_t
+check_entrylk_on_basename (xlator_t *this, inode_t *parent, char *basename);
+
+void __pl_inodelk_unref (pl_inode_lock_t *lock);
+void __pl_entrylk_unref (pl_entry_lock_t *lock);
 
 #endif /* __COMMON_H__ */
diff --git a/xlators/features/locks/src/entrylk.c b/xlators/features/locks/src/entrylk.c
index b83044b7f..ea6995627 100644
--- a/xlators/features/locks/src/entrylk.c
+++ b/xlators/features/locks/src/entrylk.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006, 2007, 2008 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -33,30 +23,57 @@
 #include "locks.h"
 #include "common.h"
 
+
+void
+__pl_entrylk_unref (pl_entry_lock_t *lock)
+{
+        lock->ref--;
+        if (!lock->ref) {
+		GF_FREE ((char *)lock->basename);
+                GF_FREE (lock->connection_id);
+                GF_FREE (lock);
+        }
+}
+
+
+static void
+__pl_entrylk_ref (pl_entry_lock_t *lock)
+{
+        lock->ref++;
+}
+
+
 static pl_entry_lock_t *
 new_entrylk_lock (pl_inode_t *pinode, const char *basename, entrylk_type type,
-		  transport_t *trans, pid_t client_pid, uint64_t owner, const char *volume)
-
+		  const char *domain, call_frame_t *frame, char *conn_id)
 {
-	pl_entry_lock_t *newlock = NULL;
+        pl_entry_lock_t *newlock = NULL;
 
-	newlock = CALLOC (1, sizeof (pl_entry_lock_t));
-	if (!newlock) {
-		goto out;
-	}
+        newlock = GF_CALLOC (1, sizeof (pl_entry_lock_t),
+                             gf_locks_mt_pl_entry_lock_t);
+        if (!newlock) {
+                goto out;
+        }
 
-	newlock->basename       = basename ? strdup (basename) : NULL;
-	newlock->type           = type;
-	newlock->trans          = trans;
-	newlock->volume         = volume;
-        newlock->client_pid     = client_pid;
-        newlock->owner          = owner;
+        newlock->basename   = basename ? gf_strdup (basename) : NULL;
+        newlock->type       = type;
+        newlock->client     = frame->root->client;
+        newlock->client_pid = frame->root->pid;
+        newlock->volume     = domain;
+        newlock->owner      = frame->root->lk_owner;
+	newlock->frame      = frame;
 
-	INIT_LIST_HEAD (&newlock->domain_list);
-	INIT_LIST_HEAD (&newlock->blocked_locks);
+        if (conn_id) {
+                newlock->connection_id = gf_strdup (conn_id);
+        }
+
+        INIT_LIST_HEAD (&newlock->domain_list);
+        INIT_LIST_HEAD (&newlock->blocked_locks);
+	INIT_LIST_HEAD (&newlock->client_list);
 
+	__pl_entrylk_ref (newlock);
 out:
-	return newlock;
+        return newlock;
 }
 
 
@@ -76,55 +93,55 @@ out:
 static int
 names_conflict (const char *n1, const char *n2)
 {
-	return all_names (n1) || all_names (n2) || !strcmp (n1, n2);
+        return all_names (n1) || all_names (n2) || !strcmp (n1, n2);
 }
 
 
-static int
+static inline int
 __same_entrylk_owner (pl_entry_lock_t *l1, pl_entry_lock_t *l2)
 {
 
-        return ((l1->owner == l2->owner) &&
-		(l1->trans  == l2->trans));
+        return (is_same_lkowner (&l1->owner, &l2->owner) &&
+                (l1->client  == l2->client));
 }
 
 
 /**
- * lock_grantable - is this lock grantable?
+ * entrylk_grantable - is this lock grantable?
  * @inode: inode in which to look
  * @basename: name we're trying to lock
  * @type: type of lock
  */
 static pl_entry_lock_t *
-__lock_grantable (pl_dom_list_t *dom, const char *basename, entrylk_type type)
+__entrylk_grantable (pl_dom_list_t *dom, pl_entry_lock_t *lock)
 {
-	pl_entry_lock_t *lock = NULL;
+        pl_entry_lock_t *tmp = NULL;
 
-	if (list_empty (&dom->entrylk_list))
-		return NULL;
+        if (list_empty (&dom->entrylk_list))
+                return NULL;
 
-	list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
-		if (names_conflict (lock->basename, basename))
-			return lock;
-	}
+        list_for_each_entry (tmp, &dom->entrylk_list, domain_list) {
+                if (names_conflict (tmp->basename, lock->basename))
+                        return tmp;
+        }
 
-	return NULL;
+        return NULL;
 }
 
 static pl_entry_lock_t *
-__blocked_lock_conflict (pl_dom_list_t *dom, const char *basename, entrylk_type type)
+__blocked_entrylk_conflict (pl_dom_list_t *dom, pl_entry_lock_t *lock)
 {
-	pl_entry_lock_t *lock = NULL;
+        pl_entry_lock_t *tmp = NULL;
 
-	if (list_empty (&dom->blocked_entrylks))
-		return NULL;
+        if (list_empty (&dom->blocked_entrylks))
+                return NULL;
 
-	list_for_each_entry (lock, &dom->blocked_entrylks, blocked_locks) {
-		if (names_conflict (lock->basename, basename))
-			return lock;
-	}
+        list_for_each_entry (tmp, &dom->blocked_entrylks, blocked_locks) {
+                if (names_conflict (tmp->basename, lock->basename))
+                        return lock;
+        }
 
-	return NULL;
+        return NULL;
 }
 
 static int
@@ -132,23 +149,23 @@ __owner_has_lock (pl_dom_list_t *dom, pl_entry_lock_t *newlock)
 {
         pl_entry_lock_t *lock = NULL;
 
-	list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
-		if (__same_entrylk_owner (lock, newlock))
-			return 1;
-	}
+        list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
+                if (__same_entrylk_owner (lock, newlock))
+                        return 1;
+        }
 
-	list_for_each_entry (lock, &dom->blocked_entrylks, blocked_locks) {
-		if (__same_entrylk_owner (lock, newlock))
-			return 1;
-	}
+        list_for_each_entry (lock, &dom->blocked_entrylks, blocked_locks) {
+                if (__same_entrylk_owner (lock, newlock))
+                        return 1;
+        }
 
-	return 0;
+        return 0;
 }
 
 static int
 names_equal (const char *n1, const char *n2)
 {
-	return (n1 == NULL && n2 == NULL) || (n1 && n2 && !strcmp (n1, n2));
+        return (n1 == NULL && n2 == NULL) || (n1 && n2 && !strcmp (n1, n2));
 }
 
 void
@@ -212,7 +229,7 @@ entrylk_trace_in (xlator_t *this, call_frame_t *frame, const char *domain,
         pl_print_lockee (pl_lockee, 256, fd, loc);
         pl_print_entrylk (pl_entrylk, 256, cmd, type, basename, domain);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[REQUEST] Locker = {%s} Lockee = {%s} Lock = {%s}",
                 pl_locker, pl_lockee, pl_entrylk);
 }
@@ -239,7 +256,7 @@ entrylk_trace_out (xlator_t *this, call_frame_t *frame, const char *domain,
         pl_print_entrylk (pl_entrylk, 256, cmd, type, basename, domain);
         pl_print_verdict (verdict, 32, op_ret, op_errno);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[%s] Locker = {%s} Lockee = {%s} Lock = {%s}",
                 verdict, pl_locker, pl_lockee, pl_entrylk);
 }
@@ -265,7 +282,7 @@ entrylk_trace_block (xlator_t *this, call_frame_t *frame, const char *volume,
         pl_print_lockee (pl_lockee, 256, fd, loc);
         pl_print_entrylk (pl_entrylk, 256, cmd, type, basename, volume);
 
-        gf_log (this->name, GF_LOG_NORMAL,
+        gf_log (this->name, GF_LOG_INFO,
                 "[BLOCKED] Locker = {%s} Lockee = {%s} Lock = {%s}",
                 pl_locker, pl_lockee, pl_entrylk);
 }
@@ -283,25 +300,25 @@ entrylk_trace_block (xlator_t *this, call_frame_t *frame, const char *volume,
 static pl_entry_lock_t *
 __find_most_matching_lock (pl_dom_list_t *dom, const char *basename)
 {
-	pl_entry_lock_t *lock;
-	pl_entry_lock_t *all = NULL;
-	pl_entry_lock_t *exact = NULL;
-
-	if (list_empty (&dom->entrylk_list))
-		return NULL;
-
-	list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
-		if (all_names (lock->basename))
-			all = lock;
-		else if (names_equal (lock->basename, basename))
-			exact = lock;
-	}
+        pl_entry_lock_t *lock;
+        pl_entry_lock_t *all = NULL;
+        pl_entry_lock_t *exact = NULL;
+
+        if (list_empty (&dom->entrylk_list))
+                return NULL;
+
+        list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
+                if (all_names (lock->basename))
+                        all = lock;
+                else if (names_equal (lock->basename, basename))
+                        exact = lock;
+        }
 
-	return (exact ? exact : all);
+        return (exact ? exact : all);
 }
 
 /**
- * __lock_name - lock a name in a directory
+ * __lock_entrylk - lock a name in a directory
  * @inode: inode for the directory in which to lock
  * @basename: name of the entry to lock
  *            if null, lock the entire directory
@@ -312,397 +329,318 @@ __find_most_matching_lock (pl_dom_list_t *dom, const char *basename)
  */
 
 int
-__lock_name (pl_inode_t *pinode, const char *basename, entrylk_type type,
-	     call_frame_t *frame, pl_dom_list_t *dom, xlator_t *this, int nonblock)
+__lock_entrylk (xlator_t *this, pl_inode_t *pinode, pl_entry_lock_t *lock,
+		int nonblock, pl_dom_list_t *dom)
 {
-	pl_entry_lock_t *lock       = NULL;
-	pl_entry_lock_t *conf       = NULL;
-	transport_t     *trans      = NULL;
-        pid_t            client_pid = 0;
-        uint64_t         owner      = 0;
-
-	int ret = -EINVAL;
-
-	trans = frame->root->trans;
-        client_pid = frame->root->pid;
-        owner      = (uint64_t)(long)frame->root;
-
-	lock = new_entrylk_lock (pinode, basename, type, trans, client_pid, owner, dom->domain);
-	if (!lock) {
-		ret = -ENOMEM;
-		goto out;
-	}
-
-	conf = __lock_grantable (dom, basename, type);
-	if (conf) {
-		ret = -EAGAIN;
-		if (nonblock)
-			goto out;
+        pl_entry_lock_t *conf = NULL;
+        int              ret  = -EAGAIN;
 
-		lock->frame   = frame;
-		lock->this    = this;
-
-		list_add_tail (&lock->blocked_locks, &dom->blocked_entrylks);
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"Blocking lock: {pinode=%p, basename=%s}",
-			pinode, basename);
-
-		goto out;
-	}
-
-        if ( __blocked_lock_conflict (dom, basename, type) && !(__owner_has_lock (dom, lock))) {
+        conf = __entrylk_grantable (dom, lock);
+        if (conf) {
                 ret = -EAGAIN;
                 if (nonblock)
                         goto out;
-                lock->frame     = frame;
-                lock->this      = this;
 
+                gettimeofday (&lock->blkd_time, NULL);
                 list_add_tail (&lock->blocked_locks, &dom->blocked_entrylks);
 
                 gf_log (this->name, GF_LOG_TRACE,
-                        "Lock is grantable, but blocking to prevent starvation");
-		gf_log (this->name, GF_LOG_TRACE,
-			"Blocking lock: {pinode=%p, basename=%s}",
-			pinode, basename);
+                        "Blocking lock: {pinode=%p, basename=%s}",
+                        pinode, lock->basename);
 
-		goto out;
+                goto out;
         }
-        switch (type) {
 
-	case ENTRYLK_WRLCK:
-		list_add (&lock->domain_list, &dom->entrylk_list);
-		break;
+        if (__blocked_entrylk_conflict (dom, lock) && !(__owner_has_lock (dom, lock))) {
+                ret = -EAGAIN;
+                if (nonblock)
+                        goto out;
 
-        default:
+                gettimeofday (&lock->blkd_time, NULL);
+                list_add_tail (&lock->blocked_locks, &dom->blocked_entrylks);
 
                 gf_log (this->name, GF_LOG_DEBUG,
-                        "Invalid type for entrylk specified: %d", type);
-                ret = -EINVAL;
+                        "Lock is grantable, but blocking to prevent starvation");
+                gf_log (this->name, GF_LOG_TRACE,
+                        "Blocking lock: {pinode=%p, basename=%s}",
+                        pinode, lock->basename);
+
                 goto out;
-	}
+        }
 
-	ret = 0;
+        __pl_entrylk_ref (lock);
+        gettimeofday (&lock->granted_time, NULL);
+        list_add (&lock->domain_list, &dom->entrylk_list);
+
+        ret = 0;
 out:
-	return ret;
+        return ret;
 }
 
 /**
- * __unlock_name - unlock a name in a directory
+ * __unlock_entrylk - unlock a name in a directory
  * @inode: inode for the directory to unlock in
  * @basename: name of the entry to unlock
  *            if null, unlock the entire directory
  */
 
 pl_entry_lock_t *
-__unlock_name (pl_dom_list_t *dom, const char *basename, entrylk_type type)
+__unlock_entrylk (pl_dom_list_t *dom, pl_entry_lock_t *lock)
 {
-	pl_entry_lock_t *lock = NULL;
-	pl_entry_lock_t *ret_lock = NULL;
+        pl_entry_lock_t *tmp = NULL;
+        pl_entry_lock_t *ret_lock = NULL;
 
-	lock = __find_most_matching_lock (dom, basename);
+        tmp = __find_most_matching_lock (dom, lock->basename);
 
-	if (!lock) {
-		gf_log ("locks", GF_LOG_DEBUG,
-			"unlock on %s (type=ENTRYLK_WRLCK) attempted but no matching lock found",
-			basename);
-		goto out;
-	}
+        if (!tmp) {
+                gf_log ("locks", GF_LOG_ERROR,
+                        "unlock on %s (type=ENTRYLK_WRLCK) attempted but no matching lock found",
+                        lock->basename);
+                goto out;
+        }
 
-	if (names_equal (lock->basename, basename)
-	    && lock->type == type) {
+        if (names_equal (tmp->basename, lock->basename)
+            && tmp->type == lock->type) {
 
-		if (type == ENTRYLK_WRLCK) {
-			list_del (&lock->domain_list);
-			ret_lock = lock;
-		}
-	} else {
-		gf_log ("locks", GF_LOG_DEBUG,
-			"Unlock for a non-existing lock!");
-		goto out;
-	}
+		list_del_init (&tmp->domain_list);
+		ret_lock = tmp;
+
+        } else {
+                gf_log ("locks", GF_LOG_ERROR,
+                        "Unlock on %s for a non-existing lock!", lock->basename);
+                goto out;
+        }
 
 out:
-	return ret_lock;
+        return ret_lock;
 }
 
+uint32_t
+check_entrylk_on_basename (xlator_t *this, inode_t *parent, char *basename)
+{
+        uint32_t        entrylk = 0;
+        pl_inode_t      *pinode = 0;
+        pl_dom_list_t   *dom = NULL;
+        pl_entry_lock_t *conf       = NULL;
+
+        pinode = pl_inode_get (this, parent);
+        if (!pinode)
+                goto out;
+        pthread_mutex_lock (&pinode->mutex);
+        {
+                list_for_each_entry (dom, &pinode->dom_list, inode_list) {
+                        conf = __find_most_matching_lock (dom, basename);
+                        if (conf && conf->basename) {
+                                entrylk = 1;
+                                break;
+                        }
+                }
+        }
+        pthread_mutex_unlock (&pinode->mutex);
+
+out:
+        return entrylk;
+}
 
 void
 __grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode,
-			     pl_dom_list_t *dom, struct list_head *granted)
+                             pl_dom_list_t *dom, struct list_head *granted)
 {
-	int              bl_ret = 0;
-	pl_entry_lock_t *bl   = NULL;
-	pl_entry_lock_t *tmp  = NULL;
+        int              bl_ret = 0;
+        pl_entry_lock_t *bl   = NULL;
+        pl_entry_lock_t *tmp  = NULL;
 
         struct list_head blocked_list;
 
         INIT_LIST_HEAD (&blocked_list);
         list_splice_init (&dom->blocked_entrylks, &blocked_list);
-        
-	list_for_each_entry_safe (bl, tmp, &blocked_list,
-				  blocked_locks) {
-
-		list_del_init (&bl->blocked_locks);
 
+        list_for_each_entry_safe (bl, tmp, &blocked_list, blocked_locks) {
 
-		gf_log ("locks", GF_LOG_TRACE,
-			"Trying to unblock: {pinode=%p, basename=%s}",
-			pl_inode, bl->basename);
+                list_del_init (&bl->blocked_locks);
 
-		bl_ret = __lock_name (pl_inode, bl->basename, bl->type,
-				      bl->frame, dom, bl->this, 0);
+                bl_ret = __lock_entrylk (bl->this, pl_inode, bl, 0, dom);
 
-		if (bl_ret == 0) {
-			list_add (&bl->blocked_locks, granted);
-		} else {
-			if (bl->basename)
-				FREE (bl->basename);
-			FREE (bl);
-		}
-	}
-	return;
+                if (bl_ret == 0) {
+                        list_add (&bl->blocked_locks, granted);
+                }
+        }
+        return;
 }
 
 /* Grants locks if possible which are blocked on a lock */
 void
 grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode,
-			   pl_entry_lock_t *unlocked, pl_dom_list_t *dom)
+			   pl_dom_list_t *dom)
 {
-	struct list_head  granted_list;
-	pl_entry_lock_t  *tmp = NULL;
-	pl_entry_lock_t  *lock = NULL;
+        struct list_head  granted_list;
+        pl_entry_lock_t  *tmp = NULL;
+        pl_entry_lock_t  *lock = NULL;
 
-	INIT_LIST_HEAD (&granted_list);
+        INIT_LIST_HEAD (&granted_list);
 
-	pthread_mutex_lock (&pl_inode->mutex);
-	{
-		__grant_blocked_entry_locks (this, pl_inode, dom, &granted_list);
-	}
-	pthread_mutex_unlock (&pl_inode->mutex);
-
-	list_for_each_entry_safe (lock, tmp, &granted_list, blocked_locks) {
-		list_del_init (&lock->blocked_locks);
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                __grant_blocked_entry_locks (this, pl_inode, dom,
+                                             &granted_list);
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
 
+        list_for_each_entry_safe (lock, tmp, &granted_list, blocked_locks) {
                 entrylk_trace_out (this, lock->frame, NULL, NULL, NULL,
                                    lock->basename, ENTRYLK_LOCK, lock->type,
                                    0, 0);
 
-		STACK_UNWIND_STRICT (entrylk, lock->frame, 0, 0);
-
-		FREE (lock->basename);
-		FREE (lock);
+                STACK_UNWIND_STRICT (entrylk, lock->frame, 0, 0, NULL);
+		lock->frame = NULL;
 	}
 
-	FREE (unlocked->basename);
-	FREE (unlocked);
-
-	return;
-}
-
-/**
- * release_entry_locks_for_transport: release all entry locks from this
- * transport for this loc_t
- */
-
-static int
-release_entry_locks_for_transport (xlator_t *this, pl_inode_t *pinode,
-				   pl_dom_list_t *dom, transport_t *trans)
-{
-	pl_entry_lock_t  *lock = NULL;
-	pl_entry_lock_t  *tmp = NULL;
-	struct list_head  granted;
-        struct list_head  released;
-
-	INIT_LIST_HEAD (&granted);
-        INIT_LIST_HEAD (&released);
-
-	pthread_mutex_lock (&pinode->mutex);
-	{
-                list_for_each_entry_safe (lock, tmp, &dom->blocked_entrylks,
-                                          blocked_locks) {
-                        if (lock->trans != trans)
-                                continue;
-
-                        list_del_init (&lock->blocked_locks);
-
-			gf_log (this->name, GF_LOG_TRACE,
-                                "releasing lock on  held by "
-                                "{transport=%p}",trans);
-
-                        list_add (&lock->blocked_locks, &released);
-
-                }
-
-		list_for_each_entry_safe (lock, tmp, &dom->entrylk_list,
-					  domain_list) {
-			if (lock->trans != trans)
-				continue;
-
-			list_del_init (&lock->domain_list);
-
-			gf_log (this->name, GF_LOG_TRACE,
-                                "releasing lock on  held by "
-                                "{transport=%p}",trans);
-
-			FREE (lock->basename);
-			FREE (lock);
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+		list_for_each_entry_safe (lock, tmp, &granted_list, blocked_locks) {
+			list_del_init (&lock->blocked_locks);
+			__pl_entrylk_unref (lock);
 		}
-
-		__grant_blocked_entry_locks (this, pinode, dom, &granted);
-
-	}
-
-	pthread_mutex_unlock (&pinode->mutex);
-
-        list_for_each_entry_safe (lock, tmp, &released, blocked_locks) {
-                list_del_init (&lock->blocked_locks);
-
-                STACK_UNWIND_STRICT (entrylk, lock->frame, -1, EAGAIN);
-
-                if (lock->basename)
-                        FREE (lock->basename);
-                FREE (lock);
-
-        }
-
-	list_for_each_entry_safe (lock, tmp, &granted, blocked_locks) {
-		list_del_init (&lock->blocked_locks);
-
-		STACK_UNWIND_STRICT (entrylk, lock->frame, 0, 0);
-
-		if (lock->basename)
-			FREE (lock->basename);
-		FREE (lock);
 	}
+        pthread_mutex_unlock (&pl_inode->mutex);
 
-	return 0;
+        return;
 }
 
+
 /* Common entrylk code called by pl_entrylk and pl_fentrylk */
 int
 pl_common_entrylk (call_frame_t *frame, xlator_t *this,
                    const char *volume, inode_t *inode, const char *basename,
-                   entrylk_cmd cmd, entrylk_type type, loc_t *loc, fd_t *fd)
+                   entrylk_cmd cmd, entrylk_type type, loc_t *loc, fd_t *fd,
+                   dict_t *xdata)
+
 {
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
-
-	transport_t * transport = NULL;
-	pid_t         pid       = -1;
-        uint64_t      owner     = -1;
-
-	pl_inode_t *     pinode   = NULL;
-	int              ret      = -1;
-	pl_entry_lock_t *unlocked = NULL;
-	char             unwind   = 1;
-
-	pl_dom_list_t	  *dom = NULL;
-
-	pinode = pl_inode_get (this, inode);
-	if (!pinode) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		op_errno = ENOMEM;
-		goto out;
-	}
+        int32_t          op_ret   = -1;
+        int32_t          op_errno = 0;
+        int              ret      = -1;
+        char             unwind   = 1;
+        GF_UNUSED int    dict_ret = -1;
+        pl_inode_t      *pinode   = NULL;
+        pl_entry_lock_t *reqlock  = NULL;
+        pl_entry_lock_t *unlocked = NULL;
+        pl_dom_list_t   *dom      = NULL;
+        char            *conn_id  = NULL;
+        pl_ctx_t        *ctx      = NULL;
+	int              nonblock = 0;
+
+        if (xdata)
+                dict_ret = dict_get_str (xdata, "connection-id", &conn_id);
+
+        pinode = pl_inode_get (this, inode);
+        if (!pinode) {
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-	dom = get_domain (pinode, volume);
-	if (!dom){
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory");
-		op_errno = ENOMEM;
-		goto out;
+	if (frame->root->client) {
+		ctx = pl_ctx_get (frame->root->client, this);
+		if (!ctx) {
+			op_errno = ENOMEM;
+			gf_log (this->name, GF_LOG_INFO, "pl_ctx_get() failed");
+			goto unwind;
+		}
 	}
 
-        entrylk_trace_in (this, frame, volume, fd, loc, basename, cmd, type);
-
-	pid       = frame->root->pid;
-        owner     = (uint64_t)(long) frame->root;
-	transport = frame->root->trans;
-
-	if (pid == 0) {
-		/*
-                  this is a special case that means release
-                  all locks from this transport
-		*/
-
-		gf_log (this->name, GF_LOG_TRACE,
-			"Releasing locks for transport %p", transport);
+        dom = get_domain (pinode, volume);
+        if (!dom){
+                op_errno = ENOMEM;
+                goto out;
+        }
 
-		release_entry_locks_for_transport (this, pinode, dom, transport);
-		op_ret = 0;
+        entrylk_trace_in (this, frame, volume, fd, loc, basename, cmd, type);
 
-		goto out;
-	}
+        reqlock = new_entrylk_lock (pinode, basename, type, dom->domain, frame,
+				    conn_id);
+        if (!reqlock) {
+                op_ret = -1;
+                op_errno = ENOMEM;
+                goto unwind;
+        }
 
-	switch (cmd) {
+        switch (cmd) {
+        case ENTRYLK_LOCK_NB:
+		nonblock = 1;
+		/* fall through */
 	case ENTRYLK_LOCK:
-		pthread_mutex_lock (&pinode->mutex);
-		{
-			ret = __lock_name (pinode, basename, type,
-					   frame, dom, this, 0);
-		}
-		pthread_mutex_unlock (&pinode->mutex);
-
-		if (ret < 0) {
-			if (ret == -EAGAIN)
+		if (ctx)
+			pthread_mutex_lock (&ctx->lock);
+                pthread_mutex_lock (&pinode->mutex);
+                {
+			reqlock->pinode = pinode;
+
+                        ret = __lock_entrylk (this, pinode, reqlock, nonblock, dom);
+			if (ret == 0) {
+				reqlock->frame = NULL;
+				op_ret = 0;
+			} else {
+				op_errno = -ret;
+			}
+
+			if (ctx && (!ret || !nonblock))
+				list_add (&reqlock->client_list,
+					  &ctx->entrylk_lockers);
+
+			if (ret == -EAGAIN && !nonblock) {
+				/* blocked */
 				unwind = 0;
-			op_errno = -ret;
-			goto out;
-		}
-
-		break;
-
-	case ENTRYLK_LOCK_NB:
-		pthread_mutex_lock (&pinode->mutex);
-		{
-			ret = __lock_name (pinode, basename, type,
-					   frame, dom, this, 1);
-		}
-		pthread_mutex_unlock (&pinode->mutex);
-
-		if (ret < 0) {
-			op_errno = -ret;
-			goto out;
-		}
-
+			} else {
+				__pl_entrylk_unref (reqlock);
+			}
+                }
+                pthread_mutex_unlock (&pinode->mutex);
+		if (ctx)
+			pthread_mutex_unlock (&ctx->lock);
 		break;
 
-	case ENTRYLK_UNLOCK:
-		pthread_mutex_lock (&pinode->mutex);
-		{
-                        unlocked = __unlock_name (dom, basename, type);
-		}
-		pthread_mutex_unlock (&pinode->mutex);
+        case ENTRYLK_UNLOCK:
+		if (ctx)
+			pthread_mutex_lock (&ctx->lock);
+                pthread_mutex_lock (&pinode->mutex);
+                {
+                        unlocked = __unlock_entrylk (dom, reqlock);
+			if (unlocked) {
+				list_del_init (&unlocked->client_list);
+				__pl_entrylk_unref (unlocked);
+				op_ret = 0;
+			} else {
+				op_errno = EINVAL;
+			}
+			__pl_entrylk_unref (reqlock);
+                }
+                pthread_mutex_unlock (&pinode->mutex);
+		if (ctx)
+			pthread_mutex_unlock (&ctx->lock);
 
-		if (unlocked)
-			grant_blocked_entry_locks (this, pinode, unlocked, dom);
+		grant_blocked_entry_locks (this, pinode, dom);
 
-		break;
+                break;
 
-	default:
-		gf_log (this->name, GF_LOG_ERROR,
-			"Unexpected case in entrylk (cmd=%d). Please file"
+        default:
+                gf_log (this->name, GF_LOG_ERROR,
+                        "Unexpected case in entrylk (cmd=%d). Please file"
                         "a bug report at http://bugs.gluster.com", cmd);
-		goto out;
-	}
-
-	op_ret = 0;
+                goto out;
+        }
 out:
         pl_update_refkeeper (this, inode);
-	if (unwind) {
+
+        if (unwind) {
                 entrylk_trace_out (this, frame, volume, fd, loc, basename,
                                    cmd, type, op_ret, op_errno);
-
-		STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno);
-	} else {
+unwind:
+                STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno, NULL);
+        } else {
                 entrylk_trace_block (this, frame, volume, fd, loc, basename,
                                      cmd, type);
         }
 
-
-	return 0;
+        return 0;
 }
 
 /**
@@ -713,11 +651,11 @@ out:
 
 int
 pl_entrylk (call_frame_t *frame, xlator_t *this,
-	    const char *volume, loc_t *loc, const char *basename,
-	    entrylk_cmd cmd, entrylk_type type)
+            const char *volume, loc_t *loc, const char *basename,
+            entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
-
-        pl_common_entrylk (frame, this, volume, loc->inode, basename, cmd, type, loc, NULL);
+        pl_common_entrylk (frame, this, volume, loc->inode, basename, cmd,
+                           type, loc, NULL, xdata);
 
         return 0;
 }
@@ -732,16 +670,98 @@ pl_entrylk (call_frame_t *frame, xlator_t *this,
 int
 pl_fentrylk (call_frame_t *frame, xlator_t *this,
              const char *volume, fd_t *fd, const char *basename,
-             entrylk_cmd cmd, entrylk_type type)
+             entrylk_cmd cmd, entrylk_type type, dict_t *xdata)
 {
+        pl_common_entrylk (frame, this, volume, fd->inode, basename, cmd,
+                           type, NULL, fd, xdata);
 
-        pl_common_entrylk (frame, this, volume, fd->inode, basename, cmd, type, NULL, fd);
+        return 0;
+}
+
+
+static void
+pl_entrylk_log_cleanup (pl_entry_lock_t *lock)
+{
+	pl_inode_t *pinode = NULL;
+        char *path = NULL;
+        char *file = NULL;
+
+	pinode = lock->pinode;
+
+	inode_path (pinode->refkeeper, NULL, &path);
+
+	if (path)
+		file = path;
+	else
+		file = uuid_utoa (pinode->refkeeper->gfid);
+
+	gf_log (THIS->name, GF_LOG_WARNING,
+		"releasing lock on %s held by "
+		"{client=%p, pid=%"PRId64" lk-owner=%s}",
+		file, lock->client, (uint64_t) lock->client_pid,
+		lkowner_utoa (&lock->owner));
+	GF_FREE (path);
+}
+
+
+/* Release all entrylks from this client */
+int
+pl_entrylk_client_cleanup (xlator_t *this, pl_ctx_t *ctx)
+{
+        pl_entry_lock_t *tmp = NULL;
+        pl_entry_lock_t *l = NULL;
+	pl_dom_list_t *dom = NULL;
+        pl_inode_t *pinode = NULL;
+
+        struct list_head released;
+
+        INIT_LIST_HEAD (&released);
+
+	pthread_mutex_lock (&ctx->lock);
+        {
+                list_for_each_entry_safe (l, tmp, &ctx->entrylk_lockers,
+					  client_list) {
+                        list_del_init (&l->client_list);
+			list_add_tail (&l->client_list, &released);
+
+			pl_entrylk_log_cleanup (l);
+
+			pinode = l->pinode;
+
+			pthread_mutex_lock (&pinode->mutex);
+			{
+				list_del_init (&l->domain_list);
+                        }
+			pthread_mutex_unlock (&pinode->mutex);
+                }
+	}
+        pthread_mutex_unlock (&ctx->lock);
+
+        list_for_each_entry_safe (l, tmp, &released, client_list) {
+                list_del_init (&l->client_list);
+
+		if (l->frame)
+			STACK_UNWIND_STRICT (entrylk, l->frame, -1, EAGAIN,
+					     NULL);
+
+		pinode = l->pinode;
+
+		dom = get_domain (pinode, l->volume);
+
+		grant_blocked_inode_locks (this, pinode, dom);
+
+		pthread_mutex_lock (&pinode->mutex);
+		{
+			__pl_entrylk_unref (l);
+		}
+		pthread_mutex_unlock (&pinode->mutex);
+        }
 
         return 0;
 }
 
 
-static int32_t
+int32_t
 __get_entrylk_count (xlator_t *this, pl_inode_t *pl_inode)
 {
         int32_t            count = 0;
@@ -750,24 +770,10 @@ __get_entrylk_count (xlator_t *this, pl_inode_t *pl_inode)
 
         list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
                 list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
-
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                " XATTR DEBUG"
-                                " domain: %s  %s on %s state = Active",
-                                dom->domain,
-                                lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" : 
-                                "ENTRYLK_WRLCK", lock->basename);
                         count++;
                 }
 
                 list_for_each_entry (lock, &dom->blocked_entrylks, blocked_locks) {
-
-                        gf_log (this->name, GF_LOG_DEBUG,
-                                " XATTR DEBUG"
-                                " domain: %s  %s on %s state = Blocked",
-                                dom->domain,
-                                lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" : 
-                                "ENTRYLK_WRLCK", lock->basename);
                         count++;
                 }
 
diff --git a/xlators/features/locks/src/inodelk.c b/xlators/features/locks/src/inodelk.c
index 019604484..e7093e60e 100644
--- a/xlators/features/locks/src/inodelk.c
+++ b/xlators/features/locks/src/inodelk.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006, 2007, 2008 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
@@ -33,30 +23,40 @@
 #include "locks.h"
 #include "common.h"
 
-void
+inline void
 __delete_inode_lock (pl_inode_lock_t *lock)
 {
-	list_del (&lock->list);
+        list_del (&lock->list);
+}
+
+static inline void
+__pl_inodelk_ref (pl_inode_lock_t *lock)
+{
+        lock->ref++;
 }
 
 void
-__destroy_inode_lock (pl_inode_lock_t *lock)
+__pl_inodelk_unref (pl_inode_lock_t *lock)
 {
-	FREE (lock);
+        lock->ref--;
+        if (!lock->ref) {
+                GF_FREE (lock->connection_id);
+                GF_FREE (lock);
+        }
 }
 
 /* Check if 2 inodelks are conflicting on type. Only 2 shared locks don't conflict */
-static int
+static inline int
 inodelk_type_conflict (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
 {
-	if (l2->fl_type == F_WRLCK || l1->fl_type == F_WRLCK)
-		return 1;
+        if (l2->fl_type == F_WRLCK || l1->fl_type == F_WRLCK)
+                return 1;
 
-	return 0;
+        return 0;
 }
 
 void
-pl_print_inodelk (char *str, int size, int cmd, struct flock *flock, const char *domain)
+pl_print_inodelk (char *str, int size, int cmd, struct gf_flock *flock, const char *domain)
 {
         char *cmd_str = NULL;
         char *type_str = NULL;
@@ -115,46 +115,43 @@ pl_print_inodelk (char *str, int size, int cmd, struct flock *flock, const char
 static int
 inodelk_overlap (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
 {
-	return ((l1->fl_end >= l2->fl_start) &&
-		(l2->fl_end >= l1->fl_start));
+        return ((l1->fl_end >= l2->fl_start) &&
+                (l2->fl_end >= l1->fl_start));
 }
 
 /* Returns true if the 2 inodelks have the same owner */
-static int same_inodelk_owner (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
+static inline int
+same_inodelk_owner (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
 {
-	return ((l1->owner == l2->owner) &&
-                (l1->transport  == l2->transport));
+        return (is_same_lkowner (&l1->owner, &l2->owner) &&
+                (l1->client == l2->client));
 }
 
 /* Returns true if the 2 inodelks conflict with each other */
 static int
 inodelk_conflict (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
 {
-	if (same_inodelk_owner (l1, l2))
-		return 0;
-
-	if (!inodelk_overlap (l1, l2))
-		return 0;
-
-	return (inodelk_type_conflict(l1, l2));
+        return (inodelk_overlap (l1, l2) &&
+                inodelk_type_conflict (l1, l2));
 }
 
 /* Determine if lock is grantable or not */
 static pl_inode_lock_t *
 __inodelk_grantable (pl_dom_list_t *dom, pl_inode_lock_t *lock)
 {
-	pl_inode_lock_t *l = NULL;
-	pl_inode_lock_t *ret = NULL;
-	if (list_empty (&dom->inodelk_list))
-		goto out;
-	list_for_each_entry (l, &dom->inodelk_list, list){
-		if (inodelk_conflict (lock, l)) {
-			ret = l;
-			goto out;
-		}
-	}
+        pl_inode_lock_t *l = NULL;
+        pl_inode_lock_t *ret = NULL;
+        if (list_empty (&dom->inodelk_list))
+                goto out;
+        list_for_each_entry (l, &dom->inodelk_list, list){
+                if (inodelk_conflict (lock, l) &&
+                    !same_inodelk_owner (lock, l)) {
+                        ret = l;
+                        goto out;
+                }
+        }
 out:
-	return ret;
+        return ret;
 }
 
 static pl_inode_lock_t *
@@ -163,18 +160,18 @@ __blocked_lock_conflict (pl_dom_list_t *dom, pl_inode_lock_t *lock)
         pl_inode_lock_t *l   = NULL;
         pl_inode_lock_t *ret = NULL;
 
-	if (list_empty (&dom->blocked_entrylks))
-		return NULL;
+        if (list_empty (&dom->blocked_inodelks))
+                return NULL;
 
-	list_for_each_entry (l, &dom->blocked_inodelks, blocked_locks) {
-		if (inodelk_conflict (lock, l)) {
-			ret = l;
-			goto out;
+        list_for_each_entry (l, &dom->blocked_inodelks, blocked_locks) {
+                if (inodelk_conflict (lock, l)) {
+                        ret = l;
+                        goto out;
                 }
-	}
+        }
 
 out:
-	return ret;
+        return ret;
 }
 
 static int
@@ -182,17 +179,17 @@ __owner_has_lock (pl_dom_list_t *dom, pl_inode_lock_t *newlock)
 {
         pl_inode_lock_t *lock = NULL;
 
-	list_for_each_entry (lock, &dom->entrylk_list, list) {
-		if (same_inodelk_owner (lock, newlock))
-			return 1;
-	}
+        list_for_each_entry (lock, &dom->inodelk_list, list) {
+                if (same_inodelk_owner (lock, newlock))
+                        return 1;
+        }
 
-	list_for_each_entry (lock, &dom->blocked_entrylks, blocked_locks) {
-		if (same_inodelk_owner (lock, newlock))
-			return 1;
-	}
+        list_for_each_entry (lock, &dom->blocked_inodelks, blocked_locks) {
+                if (same_inodelk_owner (lock, newlock))
+                        return 1;
+        }
 
-	return 0;
+        return 0;
 }
 
 
@@ -201,80 +198,85 @@ __owner_has_lock (pl_dom_list_t *dom, pl_inode_lock_t *newlock)
  */
 static int
 __lock_inodelk (xlator_t *this, pl_inode_t *pl_inode, pl_inode_lock_t *lock,
-		int can_block,  pl_dom_list_t *dom)
+                int can_block,  pl_dom_list_t *dom)
 {
-	pl_inode_lock_t *conf = NULL;
-	int ret = -EINVAL;
+        pl_inode_lock_t *conf = NULL;
+        int ret = -EINVAL;
 
-	conf = __inodelk_grantable (dom, lock);
-	if (conf){
-		ret = -EAGAIN;
-		if (can_block == 0)
-			goto out;
+        conf = __inodelk_grantable (dom, lock);
+        if (conf) {
+                ret = -EAGAIN;
+                if (can_block == 0)
+                        goto out;
 
-		list_add_tail (&lock->blocked_locks, &dom->blocked_inodelks);
+                gettimeofday (&lock->blkd_time, NULL);
+                list_add_tail (&lock->blocked_locks, &dom->blocked_inodelks);
 
                 gf_log (this->name, GF_LOG_TRACE,
-                        "%s (pid=%d) lk-owner:%"PRIu64" %"PRId64" - %"PRId64" => Blocked",
+                        "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => Blocked",
                         lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
                         lock->client_pid,
-                        lock->owner,
+                        lkowner_utoa (&lock->owner),
                         lock->user_flock.l_start,
                         lock->user_flock.l_len);
 
 
-		goto out;
-	}
+                goto out;
+        }
 
         if (__blocked_lock_conflict (dom, lock) && !(__owner_has_lock (dom, lock))) {
                 ret = -EAGAIN;
                 if (can_block == 0)
                         goto out;
 
+                gettimeofday (&lock->blkd_time, NULL);
                 list_add_tail (&lock->blocked_locks, &dom->blocked_inodelks);
 
-                gf_log (this->name, GF_LOG_TRACE,
+                gf_log (this->name, GF_LOG_DEBUG,
                         "Lock is grantable, but blocking to prevent starvation");
-		gf_log (this->name, GF_LOG_TRACE,
-                        "%s (pid=%d) (lk-owner=%"PRIu64") %"PRId64" - %"PRId64" => Blocked",
+                gf_log (this->name, GF_LOG_TRACE,
+                        "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => Blocked",
                         lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
                         lock->client_pid,
-                        lock->owner,
+                        lkowner_utoa (&lock->owner),
                         lock->user_flock.l_start,
                         lock->user_flock.l_len);
 
 
-		goto out;
+                goto out;
         }
-	list_add (&lock->list, &dom->inodelk_list);
+        __pl_inodelk_ref (lock);
+        gettimeofday (&lock->granted_time, NULL);
+        list_add (&lock->list, &dom->inodelk_list);
 
-	ret = 0;
+        ret = 0;
 
 out:
-	return ret;
+        return ret;
 }
 
 /* Return true if the two inodelks have exactly same lock boundaries */
 static int
 inodelks_equal (pl_inode_lock_t *l1, pl_inode_lock_t *l2)
 {
-	if ((l1->fl_start == l2->fl_start) &&
-	    (l1->fl_end == l2->fl_end))
-		return 1;
+        if ((l1->fl_start == l2->fl_start) &&
+            (l1->fl_end == l2->fl_end))
+                return 1;
 
-	return 0;
+        return 0;
 }
 
 
 static pl_inode_lock_t *
 find_matching_inodelk (pl_inode_lock_t *lock, pl_dom_list_t *dom)
 {
-	pl_inode_lock_t *l = NULL;
-	list_for_each_entry (l, &dom->inodelk_list, list) {
-		if (inodelks_equal (l, lock))
-			return l;
-	}
-	return NULL;
+        pl_inode_lock_t *l = NULL;
+        list_for_each_entry (l, &dom->inodelk_list, list) {
+                if (inodelks_equal (l, lock) &&
+                    same_inodelk_owner (l, lock))
+                        return l;
+        }
+        return NULL;
 }
 
 /* Set F_UNLCK removes a lock which has the exact same lock boundaries
@@ -284,341 +286,416 @@ static pl_inode_lock_t *
 __inode_unlock_lock (xlator_t *this, pl_inode_lock_t *lock, pl_dom_list_t *dom)
 {
 
-	pl_inode_lock_t *conf = NULL;
+        pl_inode_lock_t *conf = NULL;
 
-	conf = find_matching_inodelk (lock, dom);
-	if (!conf) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        " Matching lock not found for unlock");
-		goto out;
+        conf = find_matching_inodelk (lock, dom);
+        if (!conf) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        " Matching lock not found for unlock %llu-%llu, by %s "
+                        "on %p", (unsigned long long)lock->fl_start,
+                        (unsigned long long)lock->fl_end,
+                        lkowner_utoa (&lock->owner), lock->client);
+                goto out;
         }
-	__delete_inode_lock (conf);
+        __delete_inode_lock (conf);
         gf_log (this->name, GF_LOG_DEBUG,
-                " Matching lock found for unlock");
-        __destroy_inode_lock (lock);
-
+                " Matching lock found for unlock %llu-%llu, by %s on %p",
+                (unsigned long long)lock->fl_start,
+                (unsigned long long)lock->fl_end, lkowner_utoa (&lock->owner),
+                lock->client);
 
 out:
-	return conf;
+        return conf;
+}
 
 
-}
 static void
-__grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode, pl_dom_list_t *dom)
+__grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode,
+                             struct list_head *granted, pl_dom_list_t *dom)
 {
-	int	      bl_ret = 0;
-	pl_inode_lock_t *bl = NULL;
-	pl_inode_lock_t *tmp = NULL;
+        int              bl_ret = 0;
+        pl_inode_lock_t *bl = NULL;
+        pl_inode_lock_t *tmp = NULL;
 
         struct list_head blocked_list;
 
         INIT_LIST_HEAD (&blocked_list);
         list_splice_init (&dom->blocked_inodelks, &blocked_list);
 
-	list_for_each_entry_safe (bl, tmp, &blocked_list, blocked_locks) {
-
-		list_del_init (&bl->blocked_locks);
+        list_for_each_entry_safe (bl, tmp, &blocked_list, blocked_locks) {
 
-		bl_ret = __lock_inodelk (this, pl_inode, bl, 1, dom);
+                list_del_init (&bl->blocked_locks);
 
-		if (bl_ret == 0) {
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "%s (pid=%d) (lk-owner=%"PRIu64") %"PRId64" - %"PRId64" => Granted",
-                                bl->fl_type == F_UNLCK ? "Unlock" : "Lock",
-                                bl->client_pid,
-                                bl->owner,
-                                bl->user_flock.l_start,
-                                bl->user_flock.l_len);
+                bl_ret = __lock_inodelk (this, pl_inode, bl, 1, dom);
 
-                        pl_trace_out (this, bl->frame, NULL, NULL, F_SETLKW,
-                                      &bl->user_flock, 0, 0, bl->volume);
-
-
-                        STACK_UNWIND_STRICT (inodelk, bl->frame, 0, 0);
+                if (bl_ret == 0) {
+                        list_add (&bl->blocked_locks, granted);
                 }
         }
-	return;
+        return;
 }
 
 /* Grant all inodelks blocked on a lock */
 void
-grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode, pl_inode_lock_t *lock, pl_dom_list_t *dom)
+grant_blocked_inode_locks (xlator_t *this, pl_inode_t *pl_inode,
+                           pl_dom_list_t *dom)
 {
+        struct list_head granted;
+        pl_inode_lock_t *lock;
+        pl_inode_lock_t *tmp;
 
-        if (list_empty (&dom->blocked_inodelks)) {
-                gf_log (this->name, GF_LOG_TRACE,
-                        "No locks for domain: %s", dom->domain);
+        INIT_LIST_HEAD (&granted);
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                __grant_blocked_inode_locks (this, pl_inode, &granted, dom);
         }
+        pthread_mutex_unlock (&pl_inode->mutex);
 
-	__grant_blocked_inode_locks (this, pl_inode, dom);
+        list_for_each_entry_safe (lock, tmp, &granted, blocked_locks) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => Granted",
+                        lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+                        lock->client_pid,
+                        lkowner_utoa (&lock->owner),
+                        lock->user_flock.l_start,
+                        lock->user_flock.l_len);
 
-}
+                pl_trace_out (this, lock->frame, NULL, NULL, F_SETLKW,
+                              &lock->user_flock, 0, 0, lock->volume);
 
-/* Release all inodelks from this transport */
-static int
-release_inode_locks_of_transport (xlator_t *this, pl_dom_list_t *dom,
-                                  inode_t *inode, transport_t *trans)
-{
-	pl_inode_lock_t *tmp = NULL;
-	pl_inode_lock_t *l = NULL;
+                STACK_UNWIND_STRICT (inodelk, lock->frame, 0, 0, NULL);
+		lock->frame = NULL;
+        }
 
-        pl_inode_t * pinode = NULL;
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry_safe (lock, tmp, &granted, blocked_locks) {
+                        list_del_init (&lock->blocked_locks);
+                        __pl_inodelk_unref (lock);
+                }
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+}
 
-        struct list_head granted;
-        struct list_head released;
 
+static void
+pl_inodelk_log_cleanup (pl_inode_lock_t *lock)
+{
+	pl_inode_t *pl_inode = NULL;
         char *path = NULL;
+        char *file = NULL;
 
-        INIT_LIST_HEAD (&granted);
-        INIT_LIST_HEAD (&released);
+	pl_inode = lock->pl_inode;
 
-        pinode = pl_inode_get (this, inode);
+	inode_path (pl_inode->refkeeper, NULL, &path);
 
-        pthread_mutex_lock (&pinode->mutex);
-        {
-
-                list_for_each_entry_safe (l, tmp, &dom->blocked_inodelks, blocked_locks) {
-                        if (l->transport != trans)
-                                continue;
-
-                        list_del_init (&l->blocked_locks);
+	if (path)
+		file = path;
+	else
+		file = uuid_utoa (pl_inode->refkeeper->gfid);
+
+	gf_log (THIS->name, GF_LOG_WARNING,
+		"releasing lock on %s held by "
+		"{client=%p, pid=%"PRId64" lk-owner=%s}",
+		file, lock->client, (uint64_t) lock->client_pid,
+		lkowner_utoa (&lock->owner));
+	GF_FREE (path);
+}
 
-                        if (inode_path (inode, NULL, &path) < 0) {
-                                gf_log (this->name, GF_LOG_TRACE,
-                                        "inode_path failed");
-                                goto unlock;
-                        }
 
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "releasing lock on %s held by "
-                                "{transport=%p, pid=%"PRId64" lk-owner=%"PRIu64"}",
-                                path, trans,
-                                (uint64_t) l->client_pid,
-                                l->owner);
+/* Release all entrylks from this client */
+int
+pl_inodelk_client_cleanup (xlator_t *this, pl_ctx_t *ctx)
+{
+        pl_inode_lock_t *tmp = NULL;
+        pl_inode_lock_t *l = NULL;
+	pl_dom_list_t *dom = NULL;
+        pl_inode_t *pl_inode = NULL;
 
-                        list_add (&l->blocked_locks, &released);
+        struct list_head released;
 
-                }
+        INIT_LIST_HEAD (&released);
 
-                list_for_each_entry_safe (l, tmp, &dom->inodelk_list, list) {
-                        if (l->transport != trans)
-                                continue;
+	pthread_mutex_lock (&ctx->lock);
+        {
+                list_for_each_entry_safe (l, tmp, &ctx->inodelk_lockers,
+					  client_list) {
+                        list_del_init (&l->client_list);
+			list_add_tail (&l->client_list, &released);
 
-                        __delete_inode_lock (l);
+			pl_inodelk_log_cleanup (l);
 
-			grant_blocked_inode_locks (this, pinode, l, dom);
+			pl_inode = l->pl_inode;
 
-                        if (inode_path (inode, NULL, &path) < 0) {
-                                gf_log (this->name, GF_LOG_TRACE,
-                                        "inode_path failed");
-                                goto unlock;
+			pthread_mutex_lock (&pl_inode->mutex);
+			{
+				__delete_inode_lock (l);
                         }
+			pthread_mutex_unlock (&pl_inode->mutex);
+                }
+	}
+        pthread_mutex_unlock (&ctx->lock);
 
-                        gf_log (this->name, GF_LOG_TRACE,
-                                "releasing lock on %s held by "
-                                "{transport=%p, pid=%"PRId64" lk-owner=%"PRIu64"}",
-                                path, trans,
-                                (uint64_t) l->client_pid,
-                                l->owner);
-
-
-                        __destroy_inode_lock (l);
+        list_for_each_entry_safe (l, tmp, &released, client_list) {
+                list_del_init (&l->client_list);
 
+		if (l->frame)
+			STACK_UNWIND_STRICT (inodelk, l->frame, -1, EAGAIN,
+					     NULL);
 
-                }
-        }
-unlock:
-        if (path)
-                FREE (path);
+		pl_inode = l->pl_inode;
 
-        pthread_mutex_unlock (&pinode->mutex);
+		dom = get_domain (pl_inode, l->volume);
 
-        list_for_each_entry_safe (l, tmp, &released, blocked_locks) {
-                list_del_init (&l->blocked_locks);
+		grant_blocked_inode_locks (this, pl_inode, dom);
 
-                STACK_UNWIND_STRICT (inodelk, l->frame, -1, EAGAIN);
-                FREE (l);
+		pthread_mutex_lock (&pl_inode->mutex);
+		{
+			__pl_inodelk_unref (l);
+		}
+		pthread_mutex_unlock (&pl_inode->mutex);
         }
 
-	return 0;
+        return 0;
 }
 
 
 static int
-pl_inode_setlk (xlator_t *this, pl_inode_t *pl_inode, pl_inode_lock_t *lock,
-		int can_block,  pl_dom_list_t *dom)
+pl_inode_setlk (xlator_t *this, pl_ctx_t *ctx, pl_inode_t *pl_inode,
+		pl_inode_lock_t *lock, int can_block, pl_dom_list_t *dom)
 {
-	int ret = -EINVAL;
+        int ret = -EINVAL;
         pl_inode_lock_t *retlock = NULL;
+        gf_boolean_t    unref = _gf_true;
+
+	lock->pl_inode = pl_inode;
 
-	pthread_mutex_lock (&pl_inode->mutex);
-	{
-		if (lock->fl_type != F_UNLCK) {
-			ret = __lock_inodelk (this, pl_inode, lock, can_block, dom);
-			if (ret == 0)
-				gf_log (this->name, GF_LOG_TRACE,
-                                        "%s (pid=%d) (lk-owner=%"PRIu64") %"PRId64" - %"PRId64" => OK",
+	if (ctx)
+		pthread_mutex_lock (&ctx->lock);
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                if (lock->fl_type != F_UNLCK) {
+                        ret = __lock_inodelk (this, pl_inode, lock, can_block, dom);
+                        if (ret == 0) {
+				lock->frame = NULL;
+                                gf_log (this->name, GF_LOG_TRACE,
+                                        "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => OK",
                                         lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
                                         lock->client_pid,
-                                        lock->owner,
+                                        lkowner_utoa (&lock->owner),
                                         lock->fl_start,
                                         lock->fl_end);
+                        } else if (ret == -EAGAIN) {
+                                gf_log (this->name, GF_LOG_TRACE,
+                                        "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => NOK",
+                                        lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+                                        lock->client_pid,
+                                        lkowner_utoa (&lock->owner),
+                                        lock->user_flock.l_start,
+                                        lock->user_flock.l_len);
+                                if (can_block)
+                                        unref = _gf_false;
+                        }
 
-			if (ret == -EAGAIN)
-				gf_log (this->name, GF_LOG_TRACE,
-                                        "%s (pid=%d) (lk-owner=%"PRIu64") %"PRId64" - %"PRId64" => NOK",
-					lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
-					lock->client_pid,
-                                        lock->owner,
-					lock->user_flock.l_start,
-					lock->user_flock.l_len);
-
-			goto out;
-		}
-
+			if (ctx && (!ret || can_block))
+				list_add_tail (&lock->client_list,
+					       &ctx->inodelk_lockers);
+                } else {
+                        retlock = __inode_unlock_lock (this, lock, dom);
+                        if (!retlock) {
+                                gf_log (this->name, GF_LOG_DEBUG,
+                                        "Bad Unlock issued on Inode lock");
+                                ret = -EINVAL;
+                                goto out;
+                        }
+			list_del_init (&retlock->client_list);
+			__pl_inodelk_unref (retlock);
 
-		retlock = __inode_unlock_lock (this, lock, dom);
-		if (!retlock) {
-			gf_log (this->name, GF_LOG_DEBUG,
-				"Bad Unlock issued on Inode lock");
-                        ret = -EINVAL;
-                        goto out;
+                        ret = 0;
                 }
+out:
+		if (unref)
+			__pl_inodelk_unref (lock);
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+	if (ctx)
+		pthread_mutex_unlock (&ctx->lock);
 
-                ret = 0;
+        grant_blocked_inode_locks (this, pl_inode, dom);
 
-                grant_blocked_inode_locks (this, pl_inode, retlock, dom);
-	}
-out:
-	pthread_mutex_unlock (&pl_inode->mutex);
         return ret;
 }
 
 /* Create a new inode_lock_t */
 pl_inode_lock_t *
-new_inode_lock (struct flock *flock, transport_t *transport, pid_t client_pid,
-                uint64_t owner, const char *volume)
+new_inode_lock (struct gf_flock *flock, client_t *client, pid_t client_pid,
+                call_frame_t *frame, xlator_t *this, const char *volume,
+                char *conn_id)
 
 {
-	pl_inode_lock_t *lock = NULL;
+        pl_inode_lock_t *lock = NULL;
 
-	lock = CALLOC (1, sizeof (*lock));
-	if (!lock) {
-		return NULL;
-	}
+        lock = GF_CALLOC (1, sizeof (*lock),
+                          gf_locks_mt_pl_inode_lock_t);
+        if (!lock) {
+                return NULL;
+        }
 
-	lock->fl_start = flock->l_start;
-	lock->fl_type  = flock->l_type;
+        lock->fl_start = flock->l_start;
+        lock->fl_type  = flock->l_type;
 
-	if (flock->l_len == 0)
-		lock->fl_end = LLONG_MAX;
-	else
-		lock->fl_end = flock->l_start + flock->l_len - 1;
+        if (flock->l_len == 0)
+                lock->fl_end = LLONG_MAX;
+        else
+                lock->fl_end = flock->l_start + flock->l_len - 1;
 
-	lock->transport  = transport;
-	lock->client_pid = client_pid;
-        lock->owner      = owner;
-	lock->volume     = volume;
+        lock->client     = client;
+        lock->client_pid = client_pid;
+        lock->volume     = volume;
+        lock->owner      = frame->root->lk_owner;
+        lock->frame      = frame;
+        lock->this       = this;
 
-	INIT_LIST_HEAD (&lock->list);
-	INIT_LIST_HEAD (&lock->blocked_locks);
+        if (conn_id) {
+                lock->connection_id = gf_strdup (conn_id);
+        }
+
+        INIT_LIST_HEAD (&lock->list);
+        INIT_LIST_HEAD (&lock->blocked_locks);
+	INIT_LIST_HEAD (&lock->client_list);
+        __pl_inodelk_ref (lock);
 
-	return lock;
+        return lock;
 }
 
-/* Common inodelk code called form pl_inodelk and pl_finodelk */
+int32_t
+_pl_convert_volume (const char *volume, char **res)
+{
+        char    *mdata_vol = NULL;
+        int     ret = 0;
+
+        mdata_vol = strrchr (volume, ':');
+        //if the volume already ends with :metadata don't bother
+        if (mdata_vol && (strcmp (mdata_vol, ":metadata") == 0))
+                return 0;
+
+        ret = gf_asprintf (res, "%s:metadata", volume);
+        if (ret <= 0)
+                return ENOMEM;
+        return 0;
+}
+
+int32_t
+_pl_convert_volume_for_special_range (struct gf_flock *flock,
+                                      const char *volume, char **res)
+{
+        int32_t     ret = 0;
+
+        if ((flock->l_start == LLONG_MAX -1) &&
+            (flock->l_len == 0)) {
+                ret = _pl_convert_volume (volume, res);
+        }
+
+        return ret;
+}
+
+/* Common inodelk code called from pl_inodelk and pl_finodelk */
 int
 pl_common_inodelk (call_frame_t *frame, xlator_t *this,
                    const char *volume, inode_t *inode, int32_t cmd,
-                   struct flock *flock, loc_t *loc, fd_t *fd)
+                   struct gf_flock *flock, loc_t *loc, fd_t *fd, dict_t *xdata)
 {
-	int32_t op_ret   = -1;
-	int32_t op_errno = 0;
-	int     ret      = -1;
-	int     can_block = 0;
-	transport_t *           transport  = NULL;
-	pid_t                   client_pid = -1;
-        uint64_t                owner      = -1;
-	pl_inode_t *            pinode     = NULL;
-	pl_inode_lock_t *          reqlock    = NULL;
-	pl_dom_list_t *		dom	   = NULL;
-
-	VALIDATE_OR_GOTO (frame, out);
-	VALIDATE_OR_GOTO (inode, out);
-	VALIDATE_OR_GOTO (flock, out);
-
-	if ((flock->l_start < 0) || (flock->l_len < 0)) {
-		op_errno = EINVAL;
-		goto unwind;
-	}
+        int32_t           op_ret     = -1;
+        int32_t           op_errno   = 0;
+        int               ret        = -1;
+        GF_UNUSED int     dict_ret   = -1;
+        int               can_block  = 0;
+        pl_inode_t *      pinode     = NULL;
+        pl_inode_lock_t * reqlock    = NULL;
+        pl_dom_list_t *   dom        = NULL;
+        char             *res        = NULL;
+        char             *res1       = NULL;
+        char             *conn_id    = NULL;
+        pl_ctx_t         *ctx        = NULL;
+
+        if (xdata)
+                dict_ret = dict_get_str (xdata, "connection-id", &conn_id);
+
+        VALIDATE_OR_GOTO (frame, out);
+        VALIDATE_OR_GOTO (inode, unwind);
+        VALIDATE_OR_GOTO (flock, unwind);
+
+        if ((flock->l_start < 0) || (flock->l_len < 0)) {
+                op_errno = EINVAL;
+                goto unwind;
+        }
 
-        pl_trace_in (this, frame, fd, loc, cmd, flock, volume);
+        op_errno = _pl_convert_volume_for_special_range (flock, volume, &res);
+        if (op_errno)
+                goto unwind;
+        if (res)
+                volume = res;
 
-	transport  = frame->root->trans;
-	client_pid = frame->root->pid;
-        owner      = (uint64_t) (long)frame->root;
+        pl_trace_in (this, frame, fd, loc, cmd, flock, volume);
 
-	pinode = pl_inode_get (this, inode);
-	if (!pinode) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		op_errno = ENOMEM;
-		goto unwind;
+	if (frame->root->client) {
+		ctx = pl_ctx_get (frame->root->client, this);
+		if (!ctx) {
+			op_errno = ENOMEM;
+			gf_log (this->name, GF_LOG_INFO, "pl_ctx_get() failed");
+			goto unwind;
+		}
 	}
 
-	dom = get_domain (pinode, volume);
+        pinode = pl_inode_get (this, inode);
+        if (!pinode) {
+                op_errno = ENOMEM;
+                goto unwind;
+        }
 
-	if (client_pid == 0) {
-		/*
-                  special case: this means release all locks
-                  from this transport
-		*/
-		gf_log (this->name, GF_LOG_TRACE,
-			"Releasing all locks from transport %p", transport);
+        dom = get_domain (pinode, volume);
+        if (!dom) {
+                op_errno = ENOMEM;
+                goto unwind;
+        }
 
-		release_inode_locks_of_transport (this, dom, inode, transport);
-		goto unwind;
-	}
+        reqlock = new_inode_lock (flock, frame->root->client, frame->root->pid,
+                                  frame, this, volume, conn_id);
 
-	reqlock = new_inode_lock (flock, transport, client_pid, owner, volume);
+        if (!reqlock) {
+                op_ret = -1;
+                op_errno = ENOMEM;
+                goto unwind;
+        }
 
-	if (!reqlock) {
-		gf_log (this->name, GF_LOG_ERROR,
-			"Out of memory.");
-		op_ret = -1;
-		op_errno = ENOMEM;
-		goto unwind;
-	}
 
-	switch (cmd) {
-	case F_SETLKW:
-		can_block = 1;
-		reqlock->frame = frame;
-		reqlock->this  = this;
+        switch (cmd) {
+        case F_SETLKW:
+                can_block = 1;
 
-		/* fall through */
+                /* fall through */
 
-	case F_SETLK:
-		memcpy (&reqlock->user_flock, flock, sizeof (struct flock));
-		ret = pl_inode_setlk (this, pinode, reqlock,
-                                      can_block, dom);
+        case F_SETLK:
+                memcpy (&reqlock->user_flock, flock, sizeof (struct gf_flock));
+                ret = pl_inode_setlk (this, ctx, pinode, reqlock, can_block,
+				      dom);
 
-		if (ret < 0) {
-                        if (can_block) {
+                if (ret < 0) {
+                        if ((can_block) && (F_UNLCK != flock->l_type)) {
                                 pl_trace_block (this, frame, fd, loc,
                                                 cmd, flock, volume);
-				goto out;
+                                goto out;
                         }
-			gf_log (this->name, GF_LOG_TRACE, "returning EAGAIN");
-			op_errno = -ret;
-			__destroy_inode_lock (reqlock);
-			goto unwind;
-		}
-		break;
+                        gf_log (this->name, GF_LOG_TRACE, "returning EAGAIN");
+                        op_errno = -ret;
+                        goto unwind;
+                }
+                break;
 
-	default:
-		op_errno = ENOTSUP;
-		gf_log (this->name, GF_LOG_DEBUG,
+        default:
+                op_errno = ENOTSUP;
+                gf_log (this->name, GF_LOG_DEBUG,
                         "Lock command F_GETLK not supported for [f]inodelk "
                         "(cmd=%d)",
                         cmd);
@@ -628,82 +705,85 @@ pl_common_inodelk (call_frame_t *frame, xlator_t *this,
         op_ret = 0;
 
 unwind:
-        pl_update_refkeeper (this, inode);
-        pl_trace_out (this, frame, fd, loc, cmd, flock, op_ret, op_errno, volume);
-        STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno);
+        if ((inode != NULL) && (flock !=NULL)) {
+                pl_update_refkeeper (this, inode);
+                pl_trace_out (this, frame, fd, loc, cmd, flock, op_ret, op_errno, volume);
+        }
+
+        STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno, NULL);
 out:
+        GF_FREE (res);
+        GF_FREE (res1);
         return 0;
 }
 
 int
 pl_inodelk (call_frame_t *frame, xlator_t *this,
-            const char *volume, loc_t *loc, int32_t cmd, struct flock *flock)
+            const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *flock,
+            dict_t *xdata)
 {
-
-        pl_common_inodelk (frame, this, volume, loc->inode, cmd, flock, loc, NULL);
+        pl_common_inodelk (frame, this, volume, loc->inode, cmd, flock,
+                           loc, NULL, xdata);
 
         return 0;
 }
 
 int
 pl_finodelk (call_frame_t *frame, xlator_t *this,
-             const char *volume, fd_t *fd, int32_t cmd, struct flock *flock)
+             const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *flock,
+             dict_t *xdata)
 {
-
-        pl_common_inodelk (frame, this, volume, fd->inode, cmd, flock, NULL, fd);
+        pl_common_inodelk (frame, this, volume, fd->inode, cmd, flock,
+                           NULL, fd, xdata);
 
         return 0;
 
 }
 
+static inline int32_t
+__get_inodelk_dom_count (pl_dom_list_t *dom)
+{
+        pl_inode_lock_t     *lock   = NULL;
+        int32_t             count   = 0;
+
+        list_for_each_entry (lock, &dom->inodelk_list, list) {
+                count++;
+        }
+        list_for_each_entry (lock, &dom->blocked_inodelks, blocked_locks) {
+                count++;
+        }
+        return count;
+}
 
-static int32_t
-__get_inodelk_count (xlator_t *this, pl_inode_t *pl_inode)
+/* Returns the no. of locks (blocked/granted) held on a given domain name
+ * If @domname is NULL, returns the no. of locks in all the domains present.
+ * If @domname is non-NULL and non-existent, returns 0 */
+int32_t
+__get_inodelk_count (xlator_t *this, pl_inode_t *pl_inode, char *domname)
 {
         int32_t            count  = 0;
-        pl_inode_lock_t   *lock   = NULL;
         pl_dom_list_t     *dom    = NULL;
 
         list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
-                list_for_each_entry (lock, &dom->inodelk_list, list) {
-
-			gf_log (this->name, GF_LOG_DEBUG,
-                                " XATTR DEBUG"
-				" domain: %s %s (pid=%d) (lk-owner=%"PRIu64") %"PRId64" - %"PRId64" "
-                                "state = Active",
-                                dom->domain,
-				lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
-				lock->client_pid,
-                                lock->owner,
-				lock->user_flock.l_start,
-				lock->user_flock.l_len);
-
-                        count++;
-                }
-
-                list_for_each_entry (lock, &dom->blocked_inodelks, blocked_locks) {
+                if (domname) {
+                        if (strcmp (domname, dom->domain) == 0) {
+                                count = __get_inodelk_dom_count (dom);
+                                goto out;
+                        }
 
-			gf_log (this->name, GF_LOG_DEBUG,
-                                " XATTR DEBUG"
-				" domain: %s %s (pid=%d) (lk-owner=%"PRIu64") %"PRId64" - %"PRId64" "
-                                "state = Blocked",
-                                dom->domain,
-				lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
-				lock->client_pid,
-                                lock->owner,
-				lock->user_flock.l_start,
-				lock->user_flock.l_len);
+                } else {
+                    /* Counting locks from all domains */
+                        count += __get_inodelk_dom_count (dom);
 
-                        count++;
                 }
-
         }
 
+out:
         return count;
 }
 
 int32_t
-get_inodelk_count (xlator_t *this, inode_t *inode)
+get_inodelk_count (xlator_t *this, inode_t *inode, char *domname)
 {
         pl_inode_t   *pl_inode = NULL;
         uint64_t      tmp_pl_inode = 0;
@@ -719,7 +799,7 @@ get_inodelk_count (xlator_t *this, inode_t *inode)
 
         pthread_mutex_lock (&pl_inode->mutex);
         {
-                count = __get_inodelk_count (this, pl_inode);
+                count = __get_inodelk_count (this, pl_inode, domname);
         }
         pthread_mutex_unlock (&pl_inode->mutex);
 
diff --git a/xlators/features/locks/src/locks-mem-types.h b/xlators/features/locks/src/locks-mem-types.h
new file mode 100644
index 000000000..08aeb0a79
--- /dev/null
+++ b/xlators/features/locks/src/locks-mem-types.h
@@ -0,0 +1,29 @@
+/*
+   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 __LOCKS_MEM_TYPES_H__
+#define __LOCKS_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_locks_mem_types_ {
+        gf_locks_mt_pl_dom_list_t = gf_common_mt_end + 1,
+        gf_locks_mt_pl_inode_t,
+        gf_locks_mt_posix_lock_t,
+        gf_locks_mt_pl_entry_lock_t,
+        gf_locks_mt_pl_inode_lock_t,
+        gf_locks_mt_truncate_ops,
+        gf_locks_mt_pl_rw_req_t,
+        gf_locks_mt_posix_locks_private_t,
+        gf_locks_mt_pl_fdctx_t,
+        gf_locks_mt_end
+};
+#endif
+
diff --git a/xlators/features/locks/src/locks.h b/xlators/features/locks/src/locks.h
index 7dc8e2d4f..8c2a6f867 100644
--- a/xlators/features/locks/src/locks.h
+++ b/xlators/features/locks/src/locks.h
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006, 2007, 2008 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #ifndef __POSIX_LOCKS_H__
 #define __POSIX_LOCKS_H__
 
@@ -26,9 +16,12 @@
 #endif
 
 #include "compat-errno.h"
-#include "transport.h"
 #include "stack.h"
 #include "call-stub.h"
+#include "locks-mem-types.h"
+#include "client_t.h"
+
+#include "lkowner.h"
 
 struct __pl_fd;
 
@@ -40,24 +33,29 @@ struct __posix_lock {
         off_t              fl_end;
 
         short              blocked;    /* waiting to acquire */
-        struct flock       user_flock; /* the flock supplied by the user */
+        struct gf_flock    user_flock; /* the flock supplied by the user */
         xlator_t          *this;       /* required for blocked locks */
-        fd_t              *fd;
+        unsigned long      fd_num;
 
+        fd_t              *fd;
         call_frame_t      *frame;
 
+        struct timeval     blkd_time;   /*time at which lock was queued into blkd list*/
+        struct timeval     granted_time; /*time at which lock was queued into active list*/
+
         /* These two together serve to uniquely identify each process
            across nodes */
 
-        transport_t       *transport;     /* to identify client node */
+        void              *client;     /* to identify client node */
+        gf_lkowner_t       owner;
         pid_t              client_pid;    /* pid of client process */
-        uint64_t           owner;         /* lock owner from fuse */
 };
 typedef struct __posix_lock posix_lock_t;
 
 struct __pl_inode_lock {
         struct list_head   list;
         struct list_head   blocked_locks; /* list_head pointing to blocked_inodelks */
+        int                ref;
 
         short              fl_type;
         off_t              fl_start;
@@ -65,18 +63,25 @@ struct __pl_inode_lock {
 
         const char        *volume;
 
-        struct flock       user_flock; /* the flock supplied by the user */
+        struct gf_flock    user_flock; /* the flock supplied by the user */
         xlator_t          *this;       /* required for blocked locks */
-        fd_t              *fd;
+	struct __pl_inode *pl_inode;
 
         call_frame_t      *frame;
 
+        struct timeval     blkd_time;   /*time at which lock was queued into blkd list*/
+        struct timeval     granted_time; /*time at which lock was queued into active list*/
+
         /* These two together serve to uniquely identify each process
            across nodes */
 
-        transport_t       *transport;     /* to identify client node */
+        void              *client;     /* to identify client node */
+        gf_lkowner_t       owner;
         pid_t              client_pid;    /* pid of client process */
-        uint64_t           owner;
+
+        char              *connection_id; /* stores the client connection id */
+
+	struct list_head   client_list; /* list of all locks from a client */
 };
 typedef struct __pl_inode_lock pl_inode_lock_t;
 
@@ -100,18 +105,27 @@ typedef struct __pl_dom_list_t pl_dom_list_t;
 struct __entry_lock {
         struct list_head  domain_list;    /* list_head back to pl_dom_list_t */
         struct list_head  blocked_locks; /* list_head back to blocked_entrylks */
+	int ref;
 
         call_frame_t     *frame;
         xlator_t         *this;
+	struct __pl_inode *pinode;
 
         const char       *volume;
 
         const char       *basename;
         entrylk_type      type;
 
-        transport_t      *trans;
-        pid_t             client_pid;    /* pid of client process */
-        uint64_t          owner;
+        struct timeval     blkd_time;   /*time at which lock was queued into blkd list*/
+        struct timeval     granted_time; /*time at which lock was queued into active list*/
+
+        void             *client;
+        gf_lkowner_t      owner;
+	pid_t             client_pid;    /* pid of client process */
+
+        char             *connection_id; /* stores the client connection id */
+
+	struct list_head   client_list; /* list of all locks from a client */
 };
 typedef struct __entry_lock pl_entry_lock_t;
 
@@ -125,6 +139,9 @@ struct __pl_inode {
         struct list_head dom_list;       /* list of domains */
         struct list_head ext_list;       /* list of fcntl locks */
         struct list_head rw_list;        /* list of waiting r/w requests */
+        struct list_head reservelk_list;        /* list of reservelks */
+        struct list_head blocked_reservelks;        /* list of blocked reservelks */
+        struct list_head blocked_calls;  /* List of blocked lock calls while a reserve is held*/
         int              mandatory;      /* if mandatory locking is enabled */
 
         inode_t          *refkeeper;     /* hold refs on an inode while locks are
@@ -133,21 +150,55 @@ struct __pl_inode {
 typedef struct __pl_inode pl_inode_t;
 
 
-struct __pl_fd {
-        gf_boolean_t nonblocking;       /* whether O_NONBLOCK has been set */
-};
-typedef struct __pl_fd pl_fd_t;
-
-
 typedef struct {
         gf_boolean_t    mandatory;      /* if mandatory locking is enabled */
         gf_boolean_t    trace;          /* trace lock requests in and out */
+        char           *brickname;
 } posix_locks_private_t;
 
+
 typedef struct {
         gf_boolean_t   entrylk_count_req;
         gf_boolean_t   inodelk_count_req;
+        gf_boolean_t   inodelk_dom_count_req;
         gf_boolean_t   posixlk_count_req;
+        gf_boolean_t   parent_entrylk_req;
+
+        /* used by {f,}truncate */
+        loc_t  loc;
+        fd_t  *fd;
+        off_t  offset;
+        dict_t *xdata;
+        enum {TRUNCATE, FTRUNCATE} op;
 } pl_local_t;
 
+
+typedef struct {
+        struct list_head locks_list;
+} pl_fdctx_t;
+
+
+struct _locker {
+        struct list_head  lockers;
+        char             *volume;
+        inode_t          *inode;
+        gf_lkowner_t      owner;
+};
+
+typedef struct _locks_ctx {
+        pthread_mutex_t      lock;
+        struct list_head     inodelk_lockers;
+        struct list_head     entrylk_lockers;
+} pl_ctx_t;
+
+
+pl_ctx_t *
+pl_ctx_get (client_t *client, xlator_t *xlator);
+
+int
+pl_inodelk_client_cleanup (xlator_t *this, pl_ctx_t *ctx);
+
+int
+pl_entrylk_client_cleanup (xlator_t *this, pl_ctx_t *ctx);
+
 #endif /* __POSIX_LOCKS_H__ */
diff --git a/xlators/features/locks/src/posix.c b/xlators/features/locks/src/posix.c
index f3b95e059..2db327687 100644
--- a/xlators/features/locks/src/posix.c
+++ b/xlators/features/locks/src/posix.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006, 2007, 2008 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 #include <unistd.h>
 #include <fcntl.h>
 #include <limits.h>
@@ -37,6 +27,9 @@
 #include "locks.h"
 #include "common.h"
 #include "statedump.h"
+#include "clear.h"
+#include "defaults.h"
+#include "syncop.h"
 
 #ifndef LLONG_MAX
 #define LLONG_MAX LONG_LONG_MAX /* compat with old gcc */
@@ -47,37 +40,87 @@
 
 void do_blocked_rw (pl_inode_t *);
 static int __rw_allowable (pl_inode_t *, posix_lock_t *, glusterfs_fop_t);
+static int format_brickname(char *);
+int pl_lockinfo_get_brickname (xlator_t *, inode_t *, int32_t *);
+static int fetch_pathinfo(xlator_t *, inode_t *, int32_t *, char **);
 
-struct _truncate_ops {
-        loc_t  loc;
-        fd_t  *fd;
-        off_t  offset;
-        enum {TRUNCATE, FTRUNCATE} op;
-};
+static pl_fdctx_t *
+pl_new_fdctx ()
+{
+        pl_fdctx_t *fdctx = NULL;
+
+        fdctx = GF_CALLOC (1, sizeof (*fdctx),
+                           gf_locks_mt_pl_fdctx_t);
+        GF_VALIDATE_OR_GOTO ("posix-locks", fdctx, out);
+
+        INIT_LIST_HEAD (&fdctx->locks_list);
+
+out:
+        return fdctx;
+}
+
+static pl_fdctx_t *
+pl_check_n_create_fdctx (xlator_t *this, fd_t *fd)
+{
+        int         ret   = 0;
+        uint64_t    tmp   = 0;
+        pl_fdctx_t *fdctx = NULL;
+
+        GF_VALIDATE_OR_GOTO ("posix-locks", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, fd, out);
+
+        LOCK (&fd->lock);
+        {
+                ret = __fd_ctx_get (fd, this, &tmp);
+                if ((ret != 0) || (tmp == 0)) {
+                        fdctx = pl_new_fdctx ();
+                        if (fdctx == NULL) {
+                                goto unlock;
+                        }
+                }
 
+                ret = __fd_ctx_set (fd, this, (uint64_t)(long)fdctx);
+                if (ret != 0) {
+                        GF_FREE (fdctx);
+                        fdctx = NULL;
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "failed to set fd ctx");
+                }
+        }
+unlock:
+        UNLOCK (&fd->lock);
+
+out:
+        return fdctx;
+}
 
 int
 pl_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                 int32_t op_ret, int32_t op_errno, struct stat *prebuf,
-                 struct stat *postbuf)
+                 int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                 struct iatt *postbuf, dict_t *xdata)
 {
-        struct _truncate_ops *local = NULL;
+        pl_local_t *local = NULL;
 
         local = frame->local;
 
         if (local->op == TRUNCATE)
                 loc_wipe (&local->loc);
 
+        if (local->xdata)
+                dict_unref (local->xdata);
+        if (local->fd)
+                fd_unref (local->fd);
+
         STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno,
-                             prebuf, postbuf);
+                             prebuf, postbuf, xdata);
         return 0;
 }
 
 
 static int
 truncate_allowed (pl_inode_t *pl_inode,
-                  transport_t *transport, pid_t client_pid,
-                  uint64_t owner, off_t offset)
+                  client_t *client, pid_t client_pid,
+                  gf_lkowner_t *owner, off_t offset)
 {
         posix_lock_t *l = NULL;
         posix_lock_t  region = {.list = {0, }, };
@@ -85,9 +128,9 @@ truncate_allowed (pl_inode_t *pl_inode,
 
         region.fl_start   = offset;
         region.fl_end     = LLONG_MAX;
-        region.transport  = transport;
+        region.client     = client;
         region.client_pid = client_pid;
-        region.owner      = owner;
+        region.owner      = *owner;
 
         pthread_mutex_lock (&pl_inode->mutex);
         {
@@ -96,6 +139,8 @@ truncate_allowed (pl_inode_t *pl_inode,
                             && locks_overlap (&region, l)
                             && !same_owner (&region, l)) {
                                 ret = 0;
+                                gf_log ("posix-locks", GF_LOG_TRACE, "Truncate "
+                                        "allowed");
                                 break;
                         }
                 }
@@ -108,10 +153,11 @@ truncate_allowed (pl_inode_t *pl_inode,
 
 static int
 truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno, struct stat *buf)
+                   int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                   dict_t *xdata)
 {
         posix_locks_private_t *priv = NULL;
-        struct _truncate_ops  *local = NULL;
+        pl_local_t            *local = NULL;
         inode_t               *inode = NULL;
         pl_inode_t            *pl_inode = NULL;
 
@@ -133,8 +179,6 @@ truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         pl_inode = pl_inode_get (this, inode);
         if (!pl_inode) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
                 op_ret   = -1;
                 op_errno = ENOMEM;
                 goto unwind;
@@ -142,8 +186,8 @@ truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
         if (priv->mandatory
             && pl_inode->mandatory
-            && !truncate_allowed (pl_inode, frame->root->trans,
-                                  frame->root->pid, frame->root->lk_owner,
+            && !truncate_allowed (pl_inode, frame->root->client,
+                                  frame->root->pid, &frame->root->lk_owner,
                                   local->offset)) {
                 op_ret   = -1;
                 op_errno = EAGAIN;
@@ -154,52 +198,58 @@ truncate_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
         case TRUNCATE:
                 STACK_WIND (frame, pl_truncate_cbk, FIRST_CHILD (this),
                             FIRST_CHILD (this)->fops->truncate,
-                            &local->loc, local->offset);
+                            &local->loc, local->offset, local->xdata);
                 break;
         case FTRUNCATE:
                 STACK_WIND (frame, pl_truncate_cbk, FIRST_CHILD (this),
                             FIRST_CHILD (this)->fops->ftruncate,
-                            local->fd, local->offset);
+                            local->fd, local->offset, local->xdata);
                 break;
         }
 
         return 0;
 
 unwind:
+        gf_log (this->name, GF_LOG_ERROR, "truncate failed with ret: %d, "
+                "error: %s", op_ret, strerror (op_errno));
         if (local->op == TRUNCATE)
                 loc_wipe (&local->loc);
+        if (local->xdata)
+                dict_unref (local->xdata);
+        if (local->fd)
+                fd_unref (local->fd);
 
-        STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, buf, NULL);
+        STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, buf, NULL, xdata);
         return 0;
 }
 
 
 int
 pl_truncate (call_frame_t *frame, xlator_t *this,
-             loc_t *loc, off_t offset)
+             loc_t *loc, off_t offset, dict_t *xdata)
 {
-        struct _truncate_ops *local = NULL;
+        pl_local_t *local = NULL;
 
-        local = CALLOC (1, sizeof (struct _truncate_ops));
-        if (!local) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
-                goto unwind;
-        }
+        local = mem_get0 (this->local_pool);
+        GF_VALIDATE_OR_GOTO (this->name, local, unwind);
 
         local->op         = TRUNCATE;
         local->offset     = offset;
         loc_copy (&local->loc, loc);
+        if (xdata)
+                local->xdata = dict_ref (xdata);
 
         frame->local = local;
 
         STACK_WIND (frame, truncate_stat_cbk, FIRST_CHILD (this),
-                    FIRST_CHILD (this)->fops->stat, loc);
+                    FIRST_CHILD (this)->fops->stat, loc, NULL);
 
         return 0;
 
 unwind:
-        STACK_UNWIND_STRICT (truncate, frame, -1, ENOMEM, NULL, NULL);
+        gf_log (this->name, GF_LOG_ERROR, "truncate for %s failed with ret: %d, "
+                "error: %s", loc->path, -1, strerror (ENOMEM));
+        STACK_UNWIND_STRICT (truncate, frame, -1, ENOMEM, NULL, NULL, NULL);
 
         return 0;
 }
@@ -207,37 +257,97 @@ unwind:
 
 int
 pl_ftruncate (call_frame_t *frame, xlator_t *this,
-              fd_t *fd, off_t offset)
+              fd_t *fd, off_t offset, dict_t *xdata)
 {
-        struct _truncate_ops *local = NULL;
+        pl_local_t *local = NULL;
 
-        local = CALLOC (1, sizeof (struct _truncate_ops));
-        if (!local) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
-                goto unwind;
-        }
+        local = mem_get0 (this->local_pool);
+        GF_VALIDATE_OR_GOTO (this->name, local, unwind);
 
         local->op         = FTRUNCATE;
         local->offset     = offset;
-        local->fd         = fd;
+        local->fd         = fd_ref (fd);
+        if (xdata)
+                local->xdata = dict_ref (xdata);
 
         frame->local = local;
 
         STACK_WIND (frame, truncate_stat_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->fstat, fd);
+                    FIRST_CHILD(this)->fops->fstat, fd, xdata);
         return 0;
 
 unwind:
-        STACK_UNWIND_STRICT (ftruncate, frame, -1, ENOMEM, NULL, NULL);
+        gf_log (this->name, GF_LOG_ERROR, "ftruncate failed with ret: %d, "
+                "error: %s", -1, strerror (ENOMEM));
+        STACK_UNWIND_STRICT (ftruncate, frame, -1, ENOMEM, NULL, NULL, NULL);
 
         return 0;
 }
 
+int
+pl_locks_by_fd (pl_inode_t *pl_inode, fd_t *fd)
+{
+       posix_lock_t *l = NULL;
+       int found = 0;
+
+       pthread_mutex_lock (&pl_inode->mutex);
+       {
+
+               list_for_each_entry (l, &pl_inode->ext_list, list) {
+                       if ((l->fd_num == fd_to_fdnum(fd))) {
+                               found = 1;
+                               break;
+                       }
+               }
+
+       }
+       pthread_mutex_unlock (&pl_inode->mutex);
+       return found;
+}
+
+static void
+delete_locks_of_fd (xlator_t *this, pl_inode_t *pl_inode, fd_t *fd)
+{
+       posix_lock_t *tmp = NULL;
+       posix_lock_t *l = NULL;
+
+       struct list_head blocked_list;
+
+       INIT_LIST_HEAD (&blocked_list);
+
+       pthread_mutex_lock (&pl_inode->mutex);
+       {
+
+               list_for_each_entry_safe (l, tmp, &pl_inode->ext_list, list) {
+                       if ((l->fd_num == fd_to_fdnum(fd))) {
+                               if (l->blocked) {
+                                       list_move_tail (&l->list, &blocked_list);
+                                       continue;
+                               }
+                               __delete_lock (pl_inode, l);
+                               __destroy_lock (l);
+                       }
+               }
+
+       }
+       pthread_mutex_unlock (&pl_inode->mutex);
+
+       list_for_each_entry_safe (l, tmp, &blocked_list, list) {
+               list_del_init(&l->list);
+               STACK_UNWIND_STRICT (lk, l->frame, -1, EAGAIN, &l->user_flock,
+                                    NULL);
+               __destroy_lock (l);
+       }
+
+        grant_blocked_locks (this, pl_inode);
+
+        do_blocked_rw (pl_inode);
+
+}
 
 static void
 __delete_locks_of_owner (pl_inode_t *pl_inode,
-                         transport_t *transport, uint64_t owner)
+                         client_t *client, gf_lkowner_t *owner)
 {
         posix_lock_t *tmp = NULL;
         posix_lock_t *l = NULL;
@@ -245,8 +355,20 @@ __delete_locks_of_owner (pl_inode_t *pl_inode,
         /* TODO: what if it is a blocked lock with pending l->frame */
 
         list_for_each_entry_safe (l, tmp, &pl_inode->ext_list, list) {
-                if ((l->transport == transport)
-                    && (l->owner == owner)) {
+                if (l->blocked)
+                        continue;
+                if ((l->client == client) &&
+                    is_same_lkowner (&l->owner, owner)) {
+                        gf_log ("posix-locks", GF_LOG_TRACE,
+                                " Flushing lock"
+                                "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" state: %s",
+                                l->fl_type == F_UNLCK ? "Unlock" : "Lock",
+                                l->client_pid,
+                                lkowner_utoa (&l->owner),
+                                l->user_flock.l_start,
+                                l->user_flock.l_len,
+                                l->blocked == 1 ? "Blocked" : "Active");
+
                         __delete_lock (pl_inode, l);
                         __destroy_lock (l);
                 }
@@ -256,39 +378,614 @@ __delete_locks_of_owner (pl_inode_t *pl_inode,
 }
 
 
-int
-pl_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-              int32_t op_ret, int32_t op_errno)
+int32_t
+pl_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno);
+        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
+        return 0;
+
+}
+
+int32_t
+pl_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+             const char *name, dict_t *xdata)
+{
+        int32_t                 op_errno        = EINVAL;
+        int                     op_ret          = -1;
+        int32_t                 bcount          = 0;
+        int32_t                 gcount          = 0;
+        char                    key[PATH_MAX]   = {0, };
+        char                    *lk_summary     = NULL;
+        pl_inode_t              *pl_inode       = NULL;
+        dict_t                  *dict           = NULL;
+        clrlk_args              args            = {0,};
+        char                    *brickname      = NULL;
+
+        if (!name)
+                goto usual;
+
+        if (strncmp (name, GF_XATTR_CLRLK_CMD, strlen (GF_XATTR_CLRLK_CMD)))
+                goto usual;
+
+        if (clrlk_parse_args (name, &args)) {
+                op_errno = EINVAL;
+                goto out;
+        }
+
+        dict = dict_new ();
+        if (!dict) {
+                op_errno = ENOMEM;
+                goto out;
+        }
+
+        pl_inode = pl_inode_get (this, loc->inode);
+        if (!pl_inode) {
+                op_errno = ENOMEM;
+                goto out;
+        }
+
+        switch (args.type) {
+                case CLRLK_INODE:
+                case CLRLK_ENTRY:
+                        op_ret = clrlk_clear_lks_in_all_domains (this, pl_inode,
+                                                                 &args, &bcount,
+                                                                 &gcount,
+                                                                 &op_errno);
+                        if (op_ret)
+                                goto out;
+                        break;
+                case CLRLK_POSIX:
+                        op_ret = clrlk_clear_posixlk (this, pl_inode, &args,
+                                                      &bcount, &gcount,
+                                                      &op_errno);
+                        if (op_ret)
+                                goto out;
+                        break;
+                case CLRLK_TYPE_MAX:
+                        op_errno = EINVAL;
+                        goto out;
+        }
 
+        op_ret = fetch_pathinfo (this, loc->inode, &op_errno, &brickname);
+        if (op_ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Couldn't get brickname");
+        } else {
+                op_ret = format_brickname(brickname);
+                if (op_ret) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "Couldn't format brickname");
+                        GF_FREE(brickname);
+                        brickname = NULL;
+                }
+        }
+
+        if (!gcount && !bcount) {
+                if (gf_asprintf (&lk_summary, "No locks cleared.") == -1) {
+                        op_ret = -1;
+                        op_errno = ENOMEM;
+                        goto out;
+                }
+        } else if (gf_asprintf (&lk_summary, "%s: %s blocked locks=%d "
+                                "granted locks=%d",
+                                (brickname == NULL)? this->name : brickname,
+                                (args.type == CLRLK_INODE)? "inode":
+                                (args.type == CLRLK_ENTRY)? "entry":
+                                (args.type == CLRLK_POSIX)? "posix": " ",
+                                bcount, gcount) == -1) {
+                op_ret = -1;
+                op_errno = ENOMEM;
+                goto out;
+        }
+
+        strncpy (key, name, strlen (name));
+        if (dict_set_dynstr (dict, key, lk_summary)) {
+                op_ret = -1;
+                op_errno = ENOMEM;
+                goto out;
+        }
+
+        op_ret = 0;
+out:
+        GF_FREE(brickname);
+        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
+
+        GF_FREE (args.opts);
+        if (op_ret && lk_summary)
+                GF_FREE (lk_summary);
+        if (dict)
+                dict_unref (dict);
+        return 0;
+
+usual:
+        STACK_WIND (frame, pl_getxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->getxattr, loc, name, xdata);
         return 0;
 }
 
+static int
+format_brickname(char *brickname)
+{
+       int   ret             = -1;
+       char *hostname        = NULL;
+       char *volume          = NULL;
+       char *saveptr         = NULL;
+
+       if (!brickname)
+               goto out;
+
+       strtok_r(brickname, ":", &saveptr);
+       hostname = gf_strdup(strtok_r(NULL, ":", &saveptr));
+       if (hostname == NULL)
+               goto out;
+       volume = gf_strdup(strtok_r(NULL, ".", &saveptr));
+       if (volume == NULL)
+               goto out;
+
+       sprintf(brickname, "%s:%s", hostname, volume);
+
+       ret = 0;
+out:
+       GF_FREE(hostname);
+       GF_FREE(volume);
+       return ret;
+}
+
+static int
+fetch_pathinfo (xlator_t *this, inode_t *inode, int32_t *op_errno,
+                char **brickname)
+{
+        int                    ret       = -1;
+        loc_t                  loc       = {0, };
+        dict_t                *dict      = NULL;
+
+        if (!brickname)
+                goto out;
+
+        if (!op_errno)
+                goto out;
+
+        uuid_copy (loc.gfid, inode->gfid);
+        loc.inode = inode_ref (inode);
+
+        ret = syncop_getxattr (FIRST_CHILD(this), &loc, &dict,
+                               GF_XATTR_PATHINFO_KEY);
+        if (ret < 0) {
+                *op_errno = -ret;
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_get_str (dict, GF_XATTR_PATHINFO_KEY, brickname);
+        if (ret)
+                goto out;
+
+        *brickname = gf_strdup(*brickname);
+        if (*brickname == NULL) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = 0;
+out:
+        if (dict != NULL) {
+                dict_unref (dict);
+        }
+        loc_wipe(&loc);
+
+        return ret;
+}
+
 
 int
-pl_flush (call_frame_t *frame, xlator_t *this,
-          fd_t *fd)
+pl_lockinfo_get_brickname (xlator_t *this, inode_t *inode, int32_t *op_errno)
+{
+        int                    ret       = -1;
+        posix_locks_private_t *priv      = NULL;
+        char                  *brickname = NULL;
+        char                  *end       = NULL;
+        char                  *tmp       = NULL;
+
+        priv = this->private;
+
+        ret = fetch_pathinfo (this, inode, op_errno, &brickname);
+        if (ret)
+                goto out;
+
+        end = strrchr (brickname, ':');
+        if (!end) {
+                GF_FREE(brickname);
+                ret = -1;
+                goto out;
+        }
+
+        tmp = brickname;
+        brickname = gf_strndup (brickname, (end - brickname));
+        if (brickname == NULL) {
+                ret = -1;
+                goto out;
+        }
+
+        priv->brickname = brickname;
+        ret = 0;
+out:
+        GF_FREE(tmp);
+        return ret;
+}
+
+char *
+pl_lockinfo_key (xlator_t *this, inode_t *inode, int32_t *op_errno)
 {
         posix_locks_private_t *priv = NULL;
-        pl_inode_t            *pl_inode = NULL;
+        char                  *key  = NULL;
+        int                    ret = 0;
 
         priv = this->private;
 
+        if (priv->brickname == NULL) {
+                ret = pl_lockinfo_get_brickname (this, inode, op_errno);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "cannot get brickname");
+                        goto out;
+                }
+        }
+
+        key = priv->brickname;
+out:
+        return key;
+}
+
+int32_t
+pl_fgetxattr_handle_lockinfo (xlator_t *this, fd_t *fd,
+                              dict_t *dict, int32_t *op_errno)
+{
+        pl_inode_t    *pl_inode = NULL;
+        char          *key      = NULL, *buf = NULL;
+        int32_t        op_ret   = 0;
+        unsigned long  fdnum    = 0, len = 0;
+        dict_t        *tmp      = NULL;
+
+        pl_inode = pl_inode_get (this, fd->inode);
+
+        if (!pl_inode) {
+                gf_log (this->name, GF_LOG_DEBUG, "Could not get inode.");
+                *op_errno = EBADFD;
+                op_ret = -1;
+                goto out;
+        }
+
+        if (!pl_locks_by_fd (pl_inode, fd)) {
+                op_ret = 0;
+                goto out;
+        }
+
+        fdnum = fd_to_fdnum (fd);
+
+        key = pl_lockinfo_key (this, fd->inode, op_errno);
+        if (key == NULL) {
+                op_ret = -1;
+                goto out;
+        }
+
+        tmp = dict_new ();
+        if (tmp == NULL) {
+                op_ret = -1;
+                *op_errno = ENOMEM;
+                goto out;
+        }
+
+        op_ret = dict_set_uint64 (tmp, key, fdnum);
+        if (op_ret < 0) {
+                *op_errno = -op_ret;
+                op_ret = -1;
+                gf_log (this->name, GF_LOG_WARNING, "setting lockinfo value "
+                        "(%lu) for fd (ptr:%p inode-gfid:%s) failed (%s)",
+                        fdnum, fd, uuid_utoa (fd->inode->gfid),
+                        strerror (*op_errno));
+                goto out;
+        }
+
+        len = dict_serialized_length (tmp);
+        if (len < 0) {
+                *op_errno = -op_ret;
+                op_ret = -1;
+                gf_log (this->name, GF_LOG_WARNING,
+                        "dict_serialized_length failed (%s) while handling "
+                        "lockinfo for fd (ptr:%p inode-gfid:%s)",
+                        strerror (*op_errno), fd, uuid_utoa (fd->inode->gfid));
+                goto out;
+        }
+
+        buf = GF_CALLOC (1, len, gf_common_mt_char);
+        if (buf == NULL) {
+                op_ret = -1;
+                *op_errno = ENOMEM;
+                goto out;
+        }
+
+        op_ret = dict_serialize (tmp, buf);
+        if (op_ret < 0) {
+                *op_errno = -op_ret;
+                op_ret = -1;
+                gf_log (this->name, GF_LOG_WARNING,
+                        "dict_serialize failed (%s) while handling lockinfo "
+                        "for fd (ptr: %p inode-gfid:%s)", strerror (*op_errno),
+                        fd, uuid_utoa (fd->inode->gfid));
+                goto out;
+        }
+
+        op_ret = dict_set_dynptr (dict, GF_XATTR_LOCKINFO_KEY, buf, len);
+        if (op_ret < 0) {
+                *op_errno = -op_ret;
+                op_ret = -1;
+                gf_log (this->name, GF_LOG_WARNING, "setting lockinfo value "
+                        "(%lu) for fd (ptr:%p inode-gfid:%s) failed (%s)",
+                        fdnum, fd, uuid_utoa (fd->inode->gfid),
+                        strerror (*op_errno));
+                goto out;
+        }
+
+        buf = NULL;
+out:
+        if (tmp != NULL) {
+                dict_unref (tmp);
+        }
+
+        if (buf != NULL) {
+                GF_FREE (buf);
+        }
+
+        return op_ret;
+}
+
+
+int32_t
+pl_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+              const char *name, dict_t *xdata)
+{
+        int32_t  op_ret = 0, op_errno = 0;
+        dict_t  *dict   = NULL;
+
+        if (!name) {
+                goto usual;
+        }
+
+        if (strcmp (name, GF_XATTR_LOCKINFO_KEY) == 0) {
+                dict = dict_new ();
+                if (dict == NULL) {
+                        op_ret = -1;
+                        op_errno = ENOMEM;
+                        goto unwind;
+                }
+
+                op_ret = pl_fgetxattr_handle_lockinfo (this, fd, dict,
+                                                       &op_errno);
+                if (op_ret < 0) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "getting lockinfo on fd (ptr:%p inode-gfid:%s) "
+                                "failed (%s)", fd, uuid_utoa (fd->inode->gfid),
+                                strerror (op_errno));
+                }
+
+                goto unwind;
+        } else {
+                goto usual;
+        }
+
+unwind:
+        STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict, NULL);
+        if (dict != NULL) {
+                dict_unref (dict);
+        }
+
+        return 0;
+
+usual:
+        STACK_WIND (frame, default_fgetxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fgetxattr, fd, name, xdata);
+        return 0;
+}
+
+int32_t
+pl_migrate_locks (call_frame_t *frame, fd_t *newfd, uint64_t oldfd_num,
+                  int32_t *op_errno)
+{
+        pl_inode_t   *pl_inode  = NULL;
+        uint64_t      newfd_num = 0;
+        posix_lock_t *l         = NULL;
+        int32_t       op_ret    = 0;
+
+        newfd_num = fd_to_fdnum (newfd);
+
+        pl_inode = pl_inode_get (frame->this, newfd->inode);
+        if (pl_inode == NULL) {
+                op_ret = -1;
+                *op_errno = EBADFD;
+                goto out;
+        }
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                list_for_each_entry (l, &pl_inode->ext_list, list) {
+                        if (l->fd_num == oldfd_num) {
+                                l->fd_num = newfd_num;
+                                l->client = frame->root->client;
+                        }
+                }
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        op_ret = 0;
+out:
+        return op_ret;
+}
+
+int32_t
+pl_fsetxattr_handle_lockinfo (call_frame_t *frame, fd_t *fd, char *lockinfo_buf,
+                              int len, int32_t *op_errno)
+{
+        int32_t   op_ret    = -1;
+        dict_t   *lockinfo  = NULL;
+        uint64_t  oldfd_num = 0;
+        char     *key       = NULL;
+
+        lockinfo = dict_new ();
+        if (lockinfo == NULL) {
+                op_ret = -1;
+                *op_errno = ENOMEM;
+                goto out;
+        }
+
+        op_ret = dict_unserialize (lockinfo_buf, len, &lockinfo);
+        if (op_ret < 0) {
+                *op_errno = -op_ret;
+                op_ret = -1;
+                goto out;
+        }
+
+        key = pl_lockinfo_key (frame->this, fd->inode, op_errno);
+        if (key == NULL) {
+                op_ret = -1;
+                goto out;
+        }
+
+        op_ret = dict_get_uint64 (lockinfo, key, &oldfd_num);
+
+        if (oldfd_num == 0) {
+                op_ret = 0;
+                goto out;
+        }
+
+        op_ret = pl_migrate_locks (frame, fd, oldfd_num, op_errno);
+        if (op_ret < 0) {
+                gf_log (frame->this->name, GF_LOG_WARNING,
+                        "migration of locks from oldfd (ptr:%p) to newfd "
+                        "(ptr:%p) (inode-gfid:%s)", (void *)oldfd_num, fd,
+                        uuid_utoa (fd->inode->gfid));
+                goto out;
+        }
+
+out:
+        dict_unref (lockinfo);
+
+        return op_ret;
+}
+
+int32_t
+pl_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+              int32_t flags, dict_t *xdata)
+{
+        int32_t  op_ret       = 0, op_errno = 0;
+        void    *lockinfo_buf = NULL;
+        int      len          = 0;
+
+        op_ret = dict_get_ptr_and_len (dict, GF_XATTR_LOCKINFO_KEY,
+                                       &lockinfo_buf, &len);
+        if (lockinfo_buf == NULL) {
+                goto usual;
+        }
+
+        op_ret = pl_fsetxattr_handle_lockinfo (frame, fd, lockinfo_buf, len,
+                                               &op_errno);
+        if (op_ret < 0) {
+                goto unwind;
+        }
+
+usual:
+        STACK_WIND (frame, default_fsetxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+        return 0;
+
+unwind:
+        STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, NULL);
+        return 0;
+}
+
+int32_t
+pl_opendir_cbk (call_frame_t *frame,
+                void *cookie,
+                xlator_t *this,
+                int32_t op_ret,
+                int32_t op_errno,
+                fd_t *fd, dict_t *xdata)
+{
+        pl_fdctx_t *fdctx = NULL;
+
+        if (op_ret < 0)
+                goto unwind;
+
+        fdctx = pl_check_n_create_fdctx (this, fd);
+        if (!fdctx) {
+                op_errno = ENOMEM;
+                op_ret   = -1;
+                goto unwind;
+        }
+
+unwind:
+        STACK_UNWIND_STRICT (opendir,
+                             frame,
+                             op_ret,
+                             op_errno,
+                             fd, xdata);
+        return 0;
+}
+
+int32_t
+pl_opendir (call_frame_t *frame, xlator_t *this,
+             loc_t *loc, fd_t *fd, dict_t *xdata)
+{
+        STACK_WIND (frame,
+                    pl_opendir_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->opendir,
+                    loc, fd, xdata);
+        return 0;
+
+}
+
+int
+pl_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+              int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno, xdata);
+
+        return 0;
+}
+
+
+int
+pl_flush (call_frame_t *frame, xlator_t *this,
+          fd_t *fd, dict_t *xdata)
+{
+        pl_inode_t *pl_inode = NULL;
+
         pl_inode = pl_inode_get (this, fd->inode);
 
         if (!pl_inode) {
                 gf_log (this->name, GF_LOG_DEBUG, "Could not get inode.");
-                STACK_UNWIND_STRICT (flush, frame, -1, EBADFD);
+                STACK_UNWIND_STRICT (flush, frame, -1, EBADFD, NULL);
                 return 0;
         }
 
         pl_trace_flush (this, frame, fd);
 
+        if (frame->root->lk_owner.len == 0) {
+                /* Handle special case when protocol/server sets lk-owner to zero.
+                 * This usually happens due to a client disconnection. Hence, free
+                 * all locks opened with this fd.
+                 */
+                gf_log (this->name, GF_LOG_TRACE,
+                        "Releasing all locks with fd %p", fd);
+                delete_locks_of_fd (this, pl_inode, fd);
+                goto wind;
+
+        }
         pthread_mutex_lock (&pl_inode->mutex);
         {
-                __delete_locks_of_owner (pl_inode, frame->root->trans,
-                                         frame->root->lk_owner);
+                __delete_locks_of_owner (pl_inode, frame->root->client,
+                                         &frame->root->lk_owner);
         }
         pthread_mutex_unlock (&pl_inode->mutex);
 
@@ -296,17 +993,31 @@ pl_flush (call_frame_t *frame, xlator_t *this,
 
         do_blocked_rw (pl_inode);
 
+wind:
         STACK_WIND (frame, pl_flush_cbk, FIRST_CHILD(this),
-                    FIRST_CHILD(this)->fops->flush, fd);
+                    FIRST_CHILD(this)->fops->flush, fd, xdata);
         return 0;
 }
 
 
 int
 pl_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-             int32_t op_ret, int32_t op_errno, fd_t *fd)
+             int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd);
+        pl_fdctx_t *fdctx = NULL;
+
+        if (op_ret < 0)
+                goto unwind;
+
+        fdctx = pl_check_n_create_fdctx (this, fd);
+        if (!fdctx) {
+                op_errno = ENOMEM;
+                op_ret   = -1;
+                goto unwind;
+        }
+
+unwind:
+        STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
 
         return 0;
 }
@@ -314,12 +1025,11 @@ pl_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 int
 pl_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
-         fd_t *fd, int32_t wbflags)
+         fd_t *fd, dict_t *xdata)
 {
-        /* why isn't O_TRUNC being handled ? */
         STACK_WIND (frame, pl_open_cbk,
                     FIRST_CHILD(this), FIRST_CHILD(this)->fops->open,
-                    loc, flags & ~O_TRUNC, fd, wbflags);
+                    loc, flags, fd, xdata);
 
         return 0;
 }
@@ -328,11 +1038,24 @@ pl_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
 int
 pl_create_cbk (call_frame_t *frame, void *cookie,
                xlator_t *this, int32_t op_ret, int32_t op_errno,
-               fd_t *fd, inode_t *inode, struct stat *buf,
-               struct stat *preparent, struct stat *postparent)
+               fd_t *fd, inode_t *inode, struct iatt *buf,
+               struct iatt *preparent, struct iatt *postparent, dict_t *xdata)
 {
+        pl_fdctx_t *fdctx = NULL;
+
+        if (op_ret < 0)
+                goto unwind;
+
+        fdctx = pl_check_n_create_fdctx (this, fd);
+        if (!fdctx) {
+                op_errno = ENOMEM;
+                op_ret   = -1;
+                goto unwind;
+        }
+
+unwind:
         STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
-                             preparent, postparent);
+                             preparent, postparent, xdata);
 
         return 0;
 }
@@ -340,11 +1063,12 @@ pl_create_cbk (call_frame_t *frame, void *cookie,
 
 int
 pl_create (call_frame_t *frame, xlator_t *this,
-           loc_t *loc, int32_t flags, mode_t mode, fd_t *fd)
+           loc_t *loc, int32_t flags, mode_t mode, mode_t umask, fd_t *fd,
+           dict_t *xdata)
 {
         STACK_WIND (frame, pl_create_cbk,
                     FIRST_CHILD (this), FIRST_CHILD (this)->fops->create,
-                    loc, flags, mode, fd);
+                    loc, flags, mode, umask, fd, xdata);
         return 0;
 }
 
@@ -352,21 +1076,22 @@ pl_create (call_frame_t *frame, xlator_t *this,
 int
 pl_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
               int32_t op_ret, int32_t op_errno,
-              struct iovec *vector, int32_t count, struct stat *stbuf,
-              struct iobref *iobref)
+              struct iovec *vector, int32_t count, struct iatt *stbuf,
+              struct iobref *iobref, dict_t *xdata)
 {
         STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno,
-                             vector, count, stbuf, iobref);
+                             vector, count, stbuf, iobref, xdata);
 
         return 0;
 }
 
 int
 pl_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-               int32_t op_ret, int32_t op_errno, struct stat *prebuf,
-               struct stat *postbuf)
+               int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+               struct iatt *postbuf, dict_t *xdata)
 {
-        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf);
+        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf,
+                             xdata);
 
         return 0;
 }
@@ -396,7 +1121,7 @@ do_blocked_rw (pl_inode_t *pl_inode)
         list_for_each_entry_safe (rw, tmp, &wind_list, list) {
                 list_del_init (&rw->list);
                 call_resume (rw->stub);
-                free (rw);
+                GF_FREE (rw);
         }
 
         return;
@@ -424,12 +1149,12 @@ __rw_allowable (pl_inode_t *pl_inode, posix_lock_t *region,
 
 
 int
-pl_readv_cont (call_frame_t *frame, xlator_t *this,
-               fd_t *fd, size_t size, off_t offset)
+pl_readv_cont (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+               off_t offset, uint32_t flags, dict_t *xdata)
 {
         STACK_WIND (frame, pl_readv_cbk,
                     FIRST_CHILD (this), FIRST_CHILD (this)->fops->readv,
-                    fd, size, offset);
+                    fd, size, offset, flags, xdata);
 
         return 0;
 }
@@ -437,7 +1162,7 @@ pl_readv_cont (call_frame_t *frame, xlator_t *this,
 
 int
 pl_readv (call_frame_t *frame, xlator_t *this,
-          fd_t *fd, size_t size, off_t offset)
+          fd_t *fd, size_t size, off_t offset, uint32_t flags, dict_t *xdata)
 {
         posix_locks_private_t *priv = NULL;
         pl_inode_t            *pl_inode = NULL;
@@ -454,7 +1179,8 @@ pl_readv (call_frame_t *frame, xlator_t *this,
         if (priv->mandatory && pl_inode->mandatory) {
                 region.fl_start   = offset;
                 region.fl_end     = offset + size - 1;
-                region.transport  = frame->root->trans;
+                region.client     = frame->root->client;
+                region.fd_num     = fd_to_fdnum(fd);
                 region.client_pid = frame->root->pid;
                 region.owner      = frame->root->lk_owner;
 
@@ -474,23 +1200,21 @@ pl_readv (call_frame_t *frame, xlator_t *this,
                                 goto unlock;
                         }
 
-                        rw = CALLOC (1, sizeof (*rw));
+                        rw = GF_CALLOC (1, sizeof (*rw),
+                                        gf_locks_mt_pl_rw_req_t);
                         if (!rw) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "Out of memory.");
                                 op_errno = ENOMEM;
                                 op_ret = -1;
                                 goto unlock;
                         }
 
                         rw->stub = fop_readv_stub (frame, pl_readv_cont,
-                                                   fd, size, offset);
+                                                   fd, size, offset, flags,
+                                                   xdata);
                         if (!rw->stub) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "Out of memory.");
                                 op_errno = ENOMEM;
                                 op_ret = -1;
-                                free (rw);
+                                GF_FREE (rw);
                                 goto unlock;
                         }
 
@@ -506,12 +1230,12 @@ pl_readv (call_frame_t *frame, xlator_t *this,
         if (wind_needed) {
                 STACK_WIND (frame, pl_readv_cbk,
                             FIRST_CHILD (this), FIRST_CHILD (this)->fops->readv,
-                            fd, size, offset);
+                            fd, size, offset, flags, xdata);
         }
 
         if (op_ret == -1)
                 STACK_UNWIND_STRICT (readv, frame, -1, op_errno,
-                                     NULL, 0, NULL, NULL);
+                                     NULL, 0, NULL, NULL, NULL);
 
         return 0;
 }
@@ -520,11 +1244,11 @@ pl_readv (call_frame_t *frame, xlator_t *this,
 int
 pl_writev_cont (call_frame_t *frame, xlator_t *this, fd_t *fd,
                 struct iovec *vector, int count, off_t offset,
-                struct iobref *iobref)
+                uint32_t flags, struct iobref *iobref, dict_t *xdata)
 {
         STACK_WIND (frame, pl_writev_cbk,
                     FIRST_CHILD (this), FIRST_CHILD (this)->fops->writev,
-                    fd, vector, count, offset, iobref);
+                    fd, vector, count, offset, flags, iobref, xdata);
 
         return 0;
 }
@@ -533,7 +1257,7 @@ pl_writev_cont (call_frame_t *frame, xlator_t *this, fd_t *fd,
 int
 pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
            struct iovec *vector, int32_t count, off_t offset,
-           struct iobref *iobref)
+           uint32_t flags, struct iobref *iobref, dict_t *xdata)
 {
         posix_locks_private_t *priv = NULL;
         pl_inode_t            *pl_inode = NULL;
@@ -543,14 +1267,14 @@ pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
         int                    op_errno = 0;
         char                   wind_needed = 1;
 
-
         priv = this->private;
         pl_inode = pl_inode_get (this, fd->inode);
 
         if (priv->mandatory && pl_inode->mandatory) {
                 region.fl_start   = offset;
                 region.fl_end     = offset + iov_length (vector, count) - 1;
-                region.transport  = frame->root->trans;
+                region.client     = frame->root->client;
+                region.fd_num     = fd_to_fdnum(fd);
                 region.client_pid = frame->root->pid;
                 region.owner      = frame->root->lk_owner;
 
@@ -570,10 +1294,9 @@ pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
                                 goto unlock;
                         }
 
-                        rw = CALLOC (1, sizeof (*rw));
+                        rw = GF_CALLOC (1, sizeof (*rw),
+                                        gf_locks_mt_pl_rw_req_t);
                         if (!rw) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "Out of memory.");
                                 op_errno = ENOMEM;
                                 op_ret = -1;
                                 goto unlock;
@@ -581,13 +1304,11 @@ pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
 
                         rw->stub = fop_writev_stub (frame, pl_writev_cont,
                                                     fd, vector, count, offset,
-                                                    iobref);
+                                                    flags, iobref, xdata);
                         if (!rw->stub) {
-                                gf_log (this->name, GF_LOG_ERROR,
-                                        "Out of memory.");
                                 op_errno = ENOMEM;
                                 op_ret = -1;
-                                free (rw);
+                                GF_FREE (rw);
                                 goto unlock;
                         }
 
@@ -603,50 +1324,221 @@ pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
         if (wind_needed)
                 STACK_WIND (frame, pl_writev_cbk,
                             FIRST_CHILD (this), FIRST_CHILD (this)->fops->writev,
-                            fd, vector, count, offset, iobref);
+                            fd, vector, count, offset, flags, iobref, xdata);
 
         if (op_ret == -1)
-                STACK_UNWIND_STRICT (writev, frame, -1, op_errno, NULL, NULL);
+                STACK_UNWIND_STRICT (writev, frame, -1, op_errno, NULL, NULL,
+                                     NULL);
 
         return 0;
 }
 
+static int
+__fd_has_locks (pl_inode_t *pl_inode, fd_t *fd)
+{
+        int           found = 0;
+        posix_lock_t *l     = NULL;
+
+        list_for_each_entry (l, &pl_inode->ext_list, list) {
+                if ((l->fd_num == fd_to_fdnum(fd))) {
+                        found = 1;
+                        break;
+                }
+        }
+
+        return found;
+}
+
+static posix_lock_t *
+lock_dup (posix_lock_t *lock)
+{
+        posix_lock_t *new_lock = NULL;
+
+        new_lock = new_posix_lock (&lock->user_flock, lock->client,
+                                   lock->client_pid, &lock->owner,
+                                   (fd_t *)lock->fd_num);
+        return new_lock;
+}
+
+static int
+__dup_locks_to_fdctx (pl_inode_t *pl_inode, fd_t *fd,
+                      pl_fdctx_t *fdctx)
+{
+        posix_lock_t *l        = NULL;
+        posix_lock_t *duplock = NULL;
+        int ret = 0;
+
+        list_for_each_entry (l, &pl_inode->ext_list, list) {
+                if ((l->fd_num == fd_to_fdnum(fd))) {
+                        duplock = lock_dup (l);
+                        if (!duplock) {
+                                ret = -1;
+                                break;
+                        }
+
+                        list_add_tail (&duplock->list, &fdctx->locks_list);
+                }
+        }
+
+        return ret;
+}
+
+static int
+__copy_locks_to_fdctx (pl_inode_t *pl_inode, fd_t *fd,
+                      pl_fdctx_t *fdctx)
+{
+        int ret = 0;
+
+        ret = __dup_locks_to_fdctx (pl_inode, fd, fdctx);
+        if (ret)
+                goto out;
+
+out:
+        return ret;
+
+}
+
+static void
+pl_mark_eol_lock (posix_lock_t *lock)
+{
+        lock->user_flock.l_type = GF_LK_EOL;
+        return;
+}
+
+static posix_lock_t *
+__get_next_fdctx_lock (pl_fdctx_t *fdctx)
+{
+        posix_lock_t *lock = NULL;
+
+        GF_ASSERT (fdctx);
+
+        if (list_empty (&fdctx->locks_list)) {
+                gf_log (THIS->name, GF_LOG_DEBUG,
+                        "fdctx lock list empty");
+                goto out;
+        }
+
+        lock = list_entry (fdctx->locks_list.next, typeof (*lock),
+                           list);
+
+        GF_ASSERT (lock);
+
+        list_del_init (&lock->list);
+
+out:
+        return lock;
+}
+
+static int
+__set_next_lock_fd (pl_fdctx_t *fdctx, posix_lock_t *reqlock)
+{
+        posix_lock_t *lock  = NULL;
+        int           ret   = 0;
+
+        GF_ASSERT (fdctx);
+
+        lock = __get_next_fdctx_lock (fdctx);
+        if (!lock) {
+                gf_log (THIS->name, GF_LOG_DEBUG,
+                        "marking EOL in reqlock");
+                pl_mark_eol_lock (reqlock);
+                goto out;
+        }
+
+        reqlock->user_flock  = lock->user_flock;
+        reqlock->fl_start    = lock->fl_start;
+        reqlock->fl_type     = lock->fl_type;
+        reqlock->fl_end      = lock->fl_end;
+        reqlock->owner       = lock->owner;
+
+out:
+        if (lock)
+                __destroy_lock (lock);
+
+        return ret;
+}
+
+static int
+pl_getlk_fd (xlator_t *this, pl_inode_t *pl_inode,
+             fd_t *fd, posix_lock_t *reqlock)
+{
+        uint64_t    tmp   = 0;
+        pl_fdctx_t *fdctx = NULL;
+        int         ret   = 0;
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                if (!__fd_has_locks (pl_inode, fd)) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "fd=%p has no active locks", fd);
+                        ret = 0;
+                        goto unlock;
+                }
+
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "There are active locks on fd");
+
+                ret = fd_ctx_get (fd, this, &tmp);
+                fdctx = (pl_fdctx_t *)(long) tmp;
+
+                if (list_empty (&fdctx->locks_list)) {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "no fdctx -> copying all locks on fd");
+
+                        ret = __copy_locks_to_fdctx (pl_inode, fd, fdctx);
+                        if (ret) {
+                                goto unlock;
+                        }
+
+                        ret = __set_next_lock_fd (fdctx, reqlock);
+
+                } else {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "fdctx present -> returning the next lock");
+                        ret = __set_next_lock_fd (fdctx, reqlock);
+                        if (ret) {
+                                gf_log (this->name, GF_LOG_DEBUG,
+                                        "could not get next lock of fd");
+                                goto unlock;
+                        }
+                }
+        }
+
+unlock:
+        pthread_mutex_unlock (&pl_inode->mutex);
+        return ret;
+
+}
 
 int
 pl_lk (call_frame_t *frame, xlator_t *this,
-       fd_t *fd, int32_t cmd, struct flock *flock)
+       fd_t *fd, int32_t cmd, struct gf_flock *flock, dict_t *xdata)
 {
-        transport_t           *transport = NULL;
-        pid_t                  client_pid = 0;
-        uint64_t               owner      = 0;
-        posix_locks_private_t *priv = NULL;
-        pl_inode_t            *pl_inode = NULL;
-        int                    op_ret = 0;
-        int                    op_errno = 0;
-        int                    can_block = 0;
-        posix_lock_t          *reqlock = NULL;
-        posix_lock_t          *conf = NULL;
-        int                    ret = 0;
-
-        transport  = frame->root->trans;
-        client_pid = frame->root->pid;
-        owner      = frame->root->lk_owner;
-        priv       = this->private;
+        pl_inode_t   *pl_inode   = NULL;
+        int           op_ret     = 0;
+        int           op_errno   = 0;
+        int           can_block  = 0;
+        posix_lock_t *reqlock    = NULL;
+        posix_lock_t *conf       = NULL;
+        int           ret        = 0;
+
+        if ((flock->l_start < 0) || (flock->l_len < 0)) {
+                op_ret = -1;
+                op_errno = EINVAL;
+                goto unwind;
+        }
 
         pl_inode = pl_inode_get (this, fd->inode);
         if (!pl_inode) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
                 op_ret = -1;
                 op_errno = ENOMEM;
                 goto unwind;
         }
 
-        reqlock = new_posix_lock (flock, transport, client_pid, owner);
+        reqlock = new_posix_lock (flock, frame->root->client, frame->root->pid,
+                                  &frame->root->lk_owner, fd);
 
         if (!reqlock) {
-                gf_log (this->name, GF_LOG_ERROR,
-                        "Out of memory.");
                 op_ret = -1;
                 op_errno = ENOMEM;
                 goto unwind;
@@ -656,6 +1548,68 @@ pl_lk (call_frame_t *frame, xlator_t *this,
 
         switch (cmd) {
 
+        case F_RESLK_LCKW:
+                can_block = 1;
+
+                /* fall through */
+        case F_RESLK_LCK:
+                memcpy (&reqlock->user_flock, flock, sizeof (struct gf_flock));
+                reqlock->frame = frame;
+                reqlock->this = this;
+
+                ret = pl_reserve_setlk (this, pl_inode, reqlock,
+                                        can_block);
+                if (ret < 0) {
+                        if (can_block)
+                                goto out;
+
+                        op_ret = -1;
+                        op_errno = -ret;
+                        __destroy_lock (reqlock);
+                        goto unwind;
+                }
+                /* Finally a getlk and return the call */
+                conf = pl_getlk (pl_inode, reqlock);
+                if (conf)
+                        posix_lock_to_flock (conf, flock);
+                break;
+
+        case F_RESLK_UNLCK:
+                reqlock->frame = frame;
+                reqlock->this = this;
+                ret = pl_reserve_unlock (this, pl_inode, reqlock);
+                if (ret < 0) {
+                        op_ret = -1;
+                        op_errno = -ret;
+                }
+                __destroy_lock (reqlock);
+                goto unwind;
+
+                break;
+
+        case F_GETLK_FD:
+                reqlock->frame = frame;
+                reqlock->this = this;
+                ret = pl_verify_reservelk (this, pl_inode, reqlock, can_block);
+                GF_ASSERT (ret >= 0);
+
+                ret = pl_getlk_fd (this, pl_inode, fd, reqlock);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "getting locks on fd failed");
+                        op_ret = -1;
+                        op_errno = ENOLCK;
+                        goto unwind;
+                }
+
+                gf_log (this->name, GF_LOG_TRACE,
+                        "Replying with a lock on fd for healing");
+
+                posix_lock_to_flock (reqlock, flock);
+                __destroy_lock (reqlock);
+
+                break;
+
 #if F_GETLK != F_GETLK64
         case F_GETLK64:
 #endif
@@ -671,9 +1625,8 @@ pl_lk (call_frame_t *frame, xlator_t *this,
 #endif
         case F_SETLKW:
                 can_block = 1;
-                reqlock->frame = frame;
-                reqlock->this  = this;
-                reqlock->fd    = fd;
+                reqlock->frame  = frame;
+                reqlock->this   = this;
 
                 /* fall through */
 
@@ -681,12 +1634,18 @@ pl_lk (call_frame_t *frame, xlator_t *this,
         case F_SETLK64:
 #endif
         case F_SETLK:
-                memcpy (&reqlock->user_flock, flock, sizeof (struct flock));
+                memcpy (&reqlock->user_flock, flock, sizeof (struct gf_flock));
+                ret = pl_verify_reservelk (this, pl_inode, reqlock, can_block);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "Lock blocked due to conflicting reserve lock");
+                        goto out;
+                }
                 ret = pl_setlk (this, pl_inode, reqlock,
                                 can_block);
 
                 if (ret == -1) {
-                        if (can_block) {
+                        if ((can_block) && (F_UNLCK != flock->l_type)) {
                                 pl_trace_block (this, frame, fd, NULL, cmd, flock, NULL);
                                 goto out;
                         }
@@ -694,13 +1653,22 @@ pl_lk (call_frame_t *frame, xlator_t *this,
                         op_ret = -1;
                         op_errno = EAGAIN;
                         __destroy_lock (reqlock);
+
+                } else if ((0 == ret) && (F_UNLCK == flock->l_type)) {
+                        /* For NLM's last "unlock on fd" detection */
+                        if (pl_locks_by_fd (pl_inode, fd))
+                                flock->l_type = F_RDLCK;
+                        else
+                                flock->l_type = F_UNLCK;
                 }
         }
 
 unwind:
         pl_trace_out (this, frame, fd, NULL, cmd, flock, op_ret, op_errno, NULL);
         pl_update_refkeeper (this, fd->inode);
-        STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, flock);
+
+
+        STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, flock, xdata);
 out:
         return 0;
 }
@@ -715,86 +1683,194 @@ pl_forget (xlator_t *this,
 
         posix_lock_t *ext_tmp = NULL;
         posix_lock_t *ext_l   = NULL;
+        struct list_head posixlks_released;
 
         pl_inode_lock_t *ino_tmp = NULL;
         pl_inode_lock_t *ino_l   = NULL;
+        struct list_head inodelks_released;
 
         pl_rw_req_t *rw_tmp = NULL;
         pl_rw_req_t *rw_req = NULL;
 
         pl_entry_lock_t *entry_tmp = NULL;
         pl_entry_lock_t *entry_l   = NULL;
+        struct list_head entrylks_released;
 
         pl_dom_list_t *dom = NULL;
         pl_dom_list_t *dom_tmp = NULL;
 
+        INIT_LIST_HEAD (&posixlks_released);
+        INIT_LIST_HEAD (&inodelks_released);
+        INIT_LIST_HEAD (&entrylks_released);
+
         pl_inode = pl_inode_get (this, inode);
 
-        if (!list_empty (&pl_inode->rw_list)) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Pending R/W requests found, releasing.");
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+
+                if (!list_empty (&pl_inode->rw_list)) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "Pending R/W requests found, releasing.");
 
-                list_for_each_entry_safe (rw_req, rw_tmp, &pl_inode->rw_list,
-                                          list) {
+                        list_for_each_entry_safe (rw_req, rw_tmp, &pl_inode->rw_list,
+                                                  list) {
 
-                        list_del (&rw_req->list);
-                        FREE (rw_req);
+                                list_del (&rw_req->list);
+                                GF_FREE (rw_req);
+                        }
                 }
-        }
 
-        if (!list_empty (&pl_inode->ext_list)) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        "Pending fcntl locks found, releasing.");
+                if (!list_empty (&pl_inode->ext_list)) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "Pending fcntl locks found, releasing.");
 
-                list_for_each_entry_safe (ext_l, ext_tmp, &pl_inode->ext_list,
-                                          list) {
+                        list_for_each_entry_safe (ext_l, ext_tmp, &pl_inode->ext_list,
+                                                  list) {
 
-                        __delete_lock (pl_inode, ext_l);
-                        __destroy_lock (ext_l);
+                                __delete_lock (pl_inode, ext_l);
+                                if (ext_l->blocked) {
+                                        list_add_tail (&ext_l->list, &posixlks_released);
+                                        continue;
+                                }
+                                __destroy_lock (ext_l);
+                        }
                 }
-        }
 
 
-        list_for_each_entry_safe (dom, dom_tmp, &pl_inode->dom_list, inode_list) {
+                list_for_each_entry_safe (dom, dom_tmp, &pl_inode->dom_list, inode_list) {
+
+                        if (!list_empty (&dom->inodelk_list)) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "Pending inode locks found, releasing.");
+
+                                list_for_each_entry_safe (ino_l, ino_tmp, &dom->inodelk_list, list) {
+                                        __delete_inode_lock (ino_l);
+                                        __pl_inodelk_unref (ino_l);
+                                }
+
+                                list_splice_init (&dom->blocked_inodelks, &inodelks_released);
 
-                if (!list_empty (&dom->inodelk_list)) {
-                        gf_log (this->name, GF_LOG_WARNING,
-                                "Pending inode locks found, releasing.");
 
-                        list_for_each_entry_safe (ino_l, ino_tmp, &dom->inodelk_list, list) {
-                                __delete_inode_lock (ino_l);
-                                grant_blocked_inode_locks (this, pl_inode, ino_l, dom);
                         }
+                        if (!list_empty (&dom->entrylk_list)) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "Pending entry locks found, releasing.");
 
-                }
-                if (!list_empty (&dom->entrylk_list)) {
-                        gf_log (this->name, GF_LOG_WARNING,
-                                "Pending entry locks found, releasing.");
+                                list_for_each_entry_safe (entry_l, entry_tmp, &dom->entrylk_list, domain_list) {
+                                        list_del_init (&entry_l->domain_list);
 
-                        list_for_each_entry_safe (entry_l, entry_tmp, &dom->entrylk_list, domain_list) {
-                                list_del_init (&entry_l->domain_list);
-                                grant_blocked_entry_locks (this, pl_inode, entry_l, dom);
+                                        GF_FREE ((char *)entry_l->basename);
+                                        GF_FREE (entry_l->connection_id);
+                                        GF_FREE (entry_l);
+                                }
 
-                                if (entry_l->basename)
-                                        FREE (entry_l->basename);
-                                FREE (entry_l);
+                                list_splice_init (&dom->blocked_entrylks, &entrylks_released);
                         }
 
+                        list_del (&dom->inode_list);
+                        gf_log ("posix-locks", GF_LOG_TRACE,
+                                " Cleaning up domain: %s", dom->domain);
+                        GF_FREE ((char *)(dom->domain));
+                        GF_FREE (dom);
                 }
 
-                list_del (&dom->inode_list);
-                gf_log ("posix-locks", GF_LOG_TRACE,
-                        " Cleaning up domain: %s", dom->domain);
-                FREE (dom->domain);
-                FREE (dom);
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        list_for_each_entry_safe (ext_l, ext_tmp, &posixlks_released, list) {
+
+                STACK_UNWIND_STRICT (lk, ext_l->frame, -1, 0,
+                                     &ext_l->user_flock, NULL);
+                __destroy_lock (ext_l);
+        }
+
+        list_for_each_entry_safe (ino_l, ino_tmp, &inodelks_released, blocked_locks) {
+
+                STACK_UNWIND_STRICT (inodelk, ino_l->frame, -1, 0, NULL);
+                __pl_inodelk_unref (ino_l);
         }
 
-        FREE (pl_inode);
+        list_for_each_entry_safe (entry_l, entry_tmp, &entrylks_released, blocked_locks) {
+
+                STACK_UNWIND_STRICT (entrylk, entry_l->frame, -1, 0, NULL);
+                GF_FREE ((char *)entry_l->basename);
+                GF_FREE (entry_l->connection_id);
+                GF_FREE (entry_l);
+
+        }
+
+        GF_FREE (pl_inode);
 
         return 0;
 }
 
-static int32_t
+int
+pl_release (xlator_t *this, fd_t *fd)
+{
+        pl_inode_t *pl_inode     = NULL;
+        uint64_t    tmp_pl_inode = 0;
+        int         ret          = -1;
+        uint64_t    tmp          = 0;
+        pl_fdctx_t *fdctx        = NULL;
+
+        if (fd == NULL) {
+                goto out;
+        }
+
+        ret = inode_ctx_get (fd->inode, this, &tmp_pl_inode);
+        if (ret != 0)
+                goto out;
+
+        pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
+
+        pl_trace_release (this, fd);
+
+        gf_log (this->name, GF_LOG_TRACE,
+                "Releasing all locks with fd %p", fd);
+
+        delete_locks_of_fd (this, pl_inode, fd);
+        pl_update_refkeeper (this, fd->inode);
+
+        ret = fd_ctx_del (fd, this, &tmp);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Could not get fdctx");
+                goto out;
+        }
+
+        fdctx = (pl_fdctx_t *)(long)tmp;
+
+        GF_FREE (fdctx);
+out:
+        return ret;
+}
+
+int
+pl_releasedir (xlator_t *this, fd_t *fd)
+{
+        int         ret          = -1;
+        uint64_t    tmp          = 0;
+        pl_fdctx_t *fdctx        = NULL;
+
+        if (fd == NULL) {
+                goto out;
+        }
+
+        ret = fd_ctx_del (fd, this, &tmp);
+        if (ret) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Could not get fdctx");
+                goto out;
+        }
+
+        fdctx = (pl_fdctx_t *)(long)tmp;
+
+        GF_FREE (fdctx);
+out:
+        return ret;
+}
+
+int32_t
 __get_posixlk_count (xlator_t *this, pl_inode_t *pl_inode)
 {
         posix_lock_t *lock   = NULL;
@@ -802,16 +1878,6 @@ __get_posixlk_count (xlator_t *this, pl_inode_t *pl_inode)
 
         list_for_each_entry (lock, &pl_inode->ext_list, list) {
 
-			gf_log (this->name, GF_LOG_DEBUG,
-                                " XATTR DEBUG"
-				"%s (pid=%d) (lk-owner=%"PRIu64") %"PRId64" - %"PRId64" state: %s",
-				lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
-				lock->client_pid,
-                                lock->owner,
-				lock->user_flock.l_start,
-				lock->user_flock.l_len,
-                                lock->blocked == 1 ? "Blocked" : "Active");
-
                 count++;
         }
 
@@ -844,6 +1910,24 @@ out:
 }
 
 void
+pl_parent_entrylk_xattr_fill (xlator_t *this, inode_t *parent,
+                              char *basename, dict_t *dict)
+{
+        uint32_t         entrylk = 0;
+        int             ret     = -1;
+
+        if (!parent || !basename || !strlen (basename))
+                goto out;
+        entrylk = check_entrylk_on_basename (this, parent, basename);
+out:
+        ret = dict_set_uint32 (dict, GLUSTERFS_PARENT_ENTRYLK, entrylk);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        " dict_set failed on key %s", GLUSTERFS_PARENT_ENTRYLK);
+        }
+}
+
+void
 pl_entrylk_xattr_fill (xlator_t *this, inode_t *inode,
                        dict_t *dict)
 {
@@ -860,19 +1944,34 @@ pl_entrylk_xattr_fill (xlator_t *this, inode_t *inode,
 }
 
 void
-pl_inodelk_xattr_fill (xlator_t *this, inode_t *inode,
-                       dict_t *dict)
+pl_inodelk_xattr_fill (xlator_t *this, inode_t *inode, dict_t *dict,
+                       gf_boolean_t per_dom)
 {
         int32_t     count = 0;
         int         ret   = -1;
+        char        *domname = NULL;
+
+
+        if (per_dom){
+                ret = dict_get_str (dict, GLUSTERFS_INODELK_DOM_COUNT,
+                                    &domname);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_ERROR, "Failed to get "
+                                "value for key %s",GLUSTERFS_INODELK_DOM_COUNT);
+                        goto out;
+                }
+        }
+
+        count = get_inodelk_count (this, inode, domname);
 
-        count = get_inodelk_count (this, inode);
         ret = dict_set_int32 (dict, GLUSTERFS_INODELK_COUNT, count);
         if (ret < 0) {
-                gf_log (this->name, GF_LOG_DEBUG,
-                        " dict_set failed on key %s", GLUSTERFS_INODELK_COUNT);
+                gf_log (this->name, GF_LOG_DEBUG, "Failed to set count for "
+                        "key %s", GLUSTERFS_INODELK_COUNT);
         }
 
+out:
+        return;
 }
 
 void
@@ -893,56 +1992,63 @@ pl_posixlk_xattr_fill (xlator_t *this, inode_t *inode,
 
 int32_t
 pl_lookup_cbk (call_frame_t *frame,
-		    void *cookie,
-		    xlator_t *this,
-		    int32_t op_ret,
-		    int32_t op_errno,
-		    inode_t *inode,
-		    struct stat *buf,
-                    dict_t *dict,
-                    struct stat *postparent)
+                    void *cookie,
+                    xlator_t *this,
+                    int32_t op_ret,
+                    int32_t op_errno,
+                    inode_t *inode,
+                    struct iatt *buf,
+                    dict_t *xdata,
+                    struct iatt *postparent)
 {
         pl_local_t *local = NULL;
 
-        if (!frame->local) {
-                goto out;
-        }
+        GF_VALIDATE_OR_GOTO (this->name, frame->local, out);
 
-        if (op_ret) {
+        if (op_ret)
                 goto out;
-        }
 
         local = frame->local;
 
+        if (local->parent_entrylk_req)
+                pl_parent_entrylk_xattr_fill (this, local->loc.parent,
+                                              (char*)local->loc.name, xdata);
         if (local->entrylk_count_req)
-                pl_entrylk_xattr_fill (this, inode, dict);
+                pl_entrylk_xattr_fill (this, inode, xdata);
         if (local->inodelk_count_req)
-                pl_inodelk_xattr_fill (this, inode, dict);
+                pl_inodelk_xattr_fill (this, inode, xdata, _gf_false);
+        if (local->inodelk_dom_count_req)
+                pl_inodelk_xattr_fill (this, inode, xdata, _gf_true);
         if (local->posixlk_count_req)
-                pl_posixlk_xattr_fill (this, inode, dict);
+                pl_posixlk_xattr_fill (this, inode, xdata);
 
 
+out:
+        local = frame->local;
         frame->local = NULL;
 
-        if (local != NULL)
-                FREE (local);
+        if (local != NULL) {
+                loc_wipe (&local->loc);
+                mem_put (local);
+        }
 
-out:
-	STACK_UNWIND (frame,
-		      op_ret,
-		      op_errno,
-		      inode,
-		      buf,
-                      dict,
-                      postparent);
-	return 0;
+        STACK_UNWIND_STRICT (
+                     lookup,
+                     frame,
+                     op_ret,
+                     op_errno,
+                     inode,
+                     buf,
+                     xdata,
+                     postparent);
+        return 0;
 }
 
 int32_t
 pl_lookup (call_frame_t *frame,
-		xlator_t *this,
-		loc_t *loc,
-		dict_t *xattr_req)
+           xlator_t *this,
+           loc_t *loc,
+           dict_t *xdata)
 {
         pl_local_t *local  = NULL;
         int         ret    = -1;
@@ -951,35 +2057,349 @@ pl_lookup (call_frame_t *frame,
         VALIDATE_OR_GOTO (this, out);
         VALIDATE_OR_GOTO (loc, out);
 
-        local = CALLOC (1, sizeof (*local));
-        if (!local) {
-                ret = -1;
-                gf_log (this->name, GF_LOG_ERROR,
-                        " Out of memory");
-                goto out;
+        local = mem_get0 (this->local_pool);
+        GF_VALIDATE_OR_GOTO (this->name, local, out);
+
+        if (xdata) {
+                if (dict_get (xdata, GLUSTERFS_ENTRYLK_COUNT))
+                        local->entrylk_count_req = 1;
+                if (dict_get (xdata, GLUSTERFS_INODELK_COUNT))
+                        local->inodelk_count_req = 1;
+                if (dict_get (xdata, GLUSTERFS_INODELK_DOM_COUNT))
+                        local->inodelk_dom_count_req = 1;
+                if (dict_get (xdata, GLUSTERFS_POSIXLK_COUNT))
+                        local->posixlk_count_req = 1;
+                if (dict_get (xdata, GLUSTERFS_PARENT_ENTRYLK))
+                        local->parent_entrylk_req = 1;
         }
 
-        if (dict_get (xattr_req, GLUSTERFS_ENTRYLK_COUNT))
-                local->entrylk_count_req = 1;
-        if (dict_get (xattr_req, GLUSTERFS_INODELK_COUNT))
-                local->inodelk_count_req = 1;
-        if (dict_get (xattr_req, GLUSTERFS_POSIXLK_COUNT))
-                local->posixlk_count_req = 1;
-
         frame->local = local;
+        loc_copy (&local->loc, loc);
 
-	STACK_WIND (frame,
-		    pl_lookup_cbk,
-		    FIRST_CHILD(this),
-		    FIRST_CHILD(this)->fops->lookup,
-		    loc,
-		    xattr_req);
+        STACK_WIND (frame,
+                    pl_lookup_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup,
+                    loc, xdata);
         ret = 0;
 out:
         if (ret == -1)
-                STACK_UNWIND_STRICT (lookup, frame, -1, 0, NULL, NULL, NULL, NULL);
+                STACK_UNWIND_STRICT (lookup, frame, -1, 0, NULL,
+                                     NULL, NULL, NULL);
+
+        return 0;
+}
+int
+pl_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int op_ret, int op_errno, gf_dirent_t *entries, dict_t *xdata)
+{
+        pl_local_t *local  = NULL;
+        gf_dirent_t *entry = NULL;
+
+        local = frame->local;
+
+        if (op_ret <= 0)
+                goto unwind;
+
+        list_for_each_entry (entry, &entries->list, list) {
+                if (local->entrylk_count_req)
+                        pl_entrylk_xattr_fill (this, entry->inode, entry->dict);
+                if (local->inodelk_count_req)
+                        pl_inodelk_xattr_fill (this, entry->inode, entry->dict,
+                                               _gf_false);
+                if (local->inodelk_dom_count_req)
+                        pl_inodelk_xattr_fill (this, entry->inode, entry->dict,
+                                               _gf_true);
+                if (local->posixlk_count_req)
+                        pl_posixlk_xattr_fill (this, entry->inode, entry->dict);
+        }
+
+unwind:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries, xdata);
+
+        if (local)
+                mem_put (local);
+
+        return 0;
+}
+
+int
+pl_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+             off_t offset, dict_t *dict)
+{
+        pl_local_t *local  = NULL;
+
+        local = mem_get0 (this->local_pool);
+        GF_VALIDATE_OR_GOTO (this->name, local, out);
+
+        if (dict) {
+                if (dict_get (dict, GLUSTERFS_ENTRYLK_COUNT))
+                        local->entrylk_count_req = 1;
+                if (dict_get (dict, GLUSTERFS_INODELK_COUNT))
+                        local->inodelk_count_req = 1;
+                if (dict_get (dict, GLUSTERFS_INODELK_DOM_COUNT))
+                        local->inodelk_dom_count_req = 1;
+                if (dict_get (dict, GLUSTERFS_POSIXLK_COUNT))
+                        local->posixlk_count_req = 1;
+        }
+
+        frame->local = local;
+
+        STACK_WIND (frame, pl_readdirp_cbk,
+                    FIRST_CHILD(this), FIRST_CHILD(this)->fops->readdirp,
+                    fd, size, offset, dict);
+
+        return 0;
+out:
+        STACK_UNWIND_STRICT (readdirp, frame, -1, ENOMEM, NULL, NULL);
+        return 0;
+}
+
+
+void
+pl_dump_lock (char *str, int size, struct gf_flock *flock,
+              gf_lkowner_t *owner, void *trans, char *conn_id,
+              time_t *granted_time, time_t *blkd_time, gf_boolean_t active)
+{
+        char  *type_str    = NULL;
+        char   granted[32] = {0,};
+        char   blocked[32] = {0,};
+
+        switch (flock->l_type) {
+        case F_RDLCK:
+                type_str = "READ";
+                break;
+        case F_WRLCK:
+                type_str = "WRITE";
+                break;
+        case F_UNLCK:
+                type_str = "UNLOCK";
+                break;
+        default:
+                type_str = "UNKNOWN";
+                break;
+        }
+
+        if (active) {
+                if (blkd_time && *blkd_time == 0) {
+                        snprintf (str, size, RANGE_GRNTD_FMT,
+                                  type_str, flock->l_whence,
+                                  (unsigned long long) flock->l_start,
+                                  (unsigned long long) flock->l_len,
+                                  (unsigned long long) flock->l_pid,
+                                  lkowner_utoa (owner), trans, conn_id,
+                                  ctime_r (granted_time, granted));
+                } else {
+                        snprintf (str, size, RANGE_BLKD_GRNTD_FMT,
+                                  type_str, flock->l_whence,
+                                  (unsigned long long) flock->l_start,
+                                  (unsigned long long) flock->l_len,
+                                  (unsigned long long) flock->l_pid,
+                                  lkowner_utoa (owner), trans, conn_id,
+                                  ctime_r (blkd_time, blocked),
+                                  ctime_r (granted_time, granted));
+                }
+        }
+        else {
+                snprintf (str, size, RANGE_BLKD_FMT,
+                          type_str, flock->l_whence,
+                          (unsigned long long) flock->l_start,
+                          (unsigned long long) flock->l_len,
+                          (unsigned long long) flock->l_pid,
+                          lkowner_utoa (owner), trans, conn_id,
+                          ctime_r (blkd_time, blocked));
+        }
+
+}
+
+void
+__dump_entrylks (pl_inode_t *pl_inode)
+{
+        pl_dom_list_t   *dom  = NULL;
+        pl_entry_lock_t *lock = NULL;
+        char             blocked[32] = {0,};
+        char             granted[32] = {0,};
+        int              count = 0;
+        char             key[GF_DUMP_MAX_BUF_LEN] = {0,};
+
+        char tmp[256];
+
+        list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
+
+                count = 0;
+
+                gf_proc_dump_build_key(key,
+                                       "lock-dump.domain",
+                                       "domain");
+                gf_proc_dump_write(key, "%s", dom->domain);
+
+                list_for_each_entry (lock, &dom->entrylk_list, domain_list) {
+
+                        gf_proc_dump_build_key(key,
+                                               "xlator.feature.locks.lock-dump.domain.entrylk",
+                                               "entrylk[%d](ACTIVE)", count );
+                        if (lock->blkd_time.tv_sec == 0 && lock->blkd_time.tv_usec == 0) {
+                                snprintf (tmp, 256, ENTRY_GRNTD_FMT,
+                                          lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" :
+                                          "ENTRYLK_WRLCK", lock->basename,
+                                          (unsigned long long) lock->client_pid,
+                                          lkowner_utoa (&lock->owner), lock->client,
+                                          lock->connection_id,
+                                          ctime_r (&lock->granted_time.tv_sec, granted));
+                        } else {
+                                snprintf (tmp, 256, ENTRY_BLKD_GRNTD_FMT,
+                                          lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" :
+                                          "ENTRYLK_WRLCK", lock->basename,
+                                          (unsigned long long) lock->client_pid,
+                                          lkowner_utoa (&lock->owner), lock->client,
+                                          lock->connection_id,
+                                          ctime_r (&lock->blkd_time.tv_sec, blocked),
+                                          ctime_r (&lock->granted_time.tv_sec, granted));
+                        }
+
+                        gf_proc_dump_write(key, tmp);
+
+                        count++;
+                }
+
+                list_for_each_entry (lock, &dom->blocked_entrylks, blocked_locks) {
+
+                        gf_proc_dump_build_key(key,
+                                               "xlator.feature.locks.lock-dump.domain.entrylk",
+                                               "entrylk[%d](BLOCKED)", count );
+                        snprintf (tmp, 256, ENTRY_BLKD_FMT,
+                                  lock->type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" :
+                                  "ENTRYLK_WRLCK", lock->basename,
+                                  (unsigned long long) lock->client_pid,
+                                  lkowner_utoa (&lock->owner), lock->client,
+                                  lock->connection_id,
+                                  ctime_r (&lock->blkd_time.tv_sec, blocked));
+
+                        gf_proc_dump_write(key, tmp);
+
+                        count++;
+                }
+
+        }
+
+}
+
+void
+dump_entrylks (pl_inode_t *pl_inode)
+{
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                __dump_entrylks (pl_inode);
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+}
+
+void
+__dump_inodelks (pl_inode_t *pl_inode)
+{
+        pl_dom_list_t   *dom  = NULL;
+        pl_inode_lock_t *lock = NULL;
+        int             count = 0;
+        char            key[GF_DUMP_MAX_BUF_LEN];
+
+        char tmp[256];
+
+        list_for_each_entry (dom, &pl_inode->dom_list, inode_list) {
+
+                count = 0;
+
+                gf_proc_dump_build_key(key,
+                                       "lock-dump.domain",
+                                       "domain");
+                gf_proc_dump_write(key, "%s", dom->domain);
+
+                list_for_each_entry (lock, &dom->inodelk_list, list) {
+
+                        gf_proc_dump_build_key(key,
+                                               "inodelk",
+                                               "inodelk[%d](ACTIVE)",count );
+
+                        SET_FLOCK_PID (&lock->user_flock, lock);
+                        pl_dump_lock (tmp, 256, &lock->user_flock,
+                                      &lock->owner,
+                                      lock->client, lock->connection_id,
+                                      &lock->granted_time.tv_sec,
+                                      &lock->blkd_time.tv_sec,
+                                      _gf_true);
+                        gf_proc_dump_write(key, tmp);
+
+                        count++;
+                }
+
+                list_for_each_entry (lock, &dom->blocked_inodelks, blocked_locks) {
+
+                        gf_proc_dump_build_key(key,
+                                               "inodelk",
+                                               "inodelk[%d](BLOCKED)",count );
+                        SET_FLOCK_PID (&lock->user_flock, lock);
+                        pl_dump_lock (tmp, 256, &lock->user_flock,
+                                      &lock->owner,
+                                      lock->client, lock->connection_id,
+                                      0, &lock->blkd_time.tv_sec,
+                                      _gf_false);
+                        gf_proc_dump_write(key, tmp);
+
+                        count++;
+                }
+
+        }
+
+}
+
+void
+dump_inodelks (pl_inode_t *pl_inode)
+{
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                __dump_inodelks (pl_inode);
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+}
+
+void
+__dump_posixlks (pl_inode_t *pl_inode)
+{
+        posix_lock_t    *lock = NULL;
+        int             count = 0;
+        char            key[GF_DUMP_MAX_BUF_LEN];
+
+        char tmp[256];
+
+      list_for_each_entry (lock, &pl_inode->ext_list, list) {
+
+              SET_FLOCK_PID (&lock->user_flock, lock);
+              gf_proc_dump_build_key(key,
+                                     "posixlk",
+                                     "posixlk[%d](%s)",
+                                     count,
+                                     lock->blocked ? "BLOCKED" : "ACTIVE");
+              pl_dump_lock (tmp, 256, &lock->user_flock,
+                            &lock->owner, lock->client, NULL,
+                            &lock->granted_time.tv_sec, &lock->blkd_time.tv_sec,
+                            (lock->blocked)? _gf_false: _gf_true);
+              gf_proc_dump_write(key, tmp);
+
+              count++;
+        }
+}
+
+void
+dump_posixlks (pl_inode_t *pl_inode)
+{
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                __dump_posixlks (pl_inode);
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
 
-	return 0;
 }
 
 int32_t
@@ -989,50 +2409,175 @@ pl_dump_inode_priv (xlator_t *this, inode_t *inode)
         int             ret = -1;
         uint64_t        tmp_pl_inode = 0;
         pl_inode_t      *pl_inode = NULL;
-        char            key[GF_DUMP_MAX_BUF_LEN];
+        char            *pathname = NULL;
+        gf_boolean_t    section_added = _gf_false;
 
-        if (!inode)
-                return -1;
+        int count      = 0;
 
-        ret = inode_ctx_get (inode, this, &tmp_pl_inode);
+        if (!inode) {
+                errno = EINVAL;
+                goto out;
+        }
 
-        if (ret != 0)
-                return ret;
+        ret = TRY_LOCK (&inode->lock);
+        if (ret)
+                goto out;
+        {
+                ret = __inode_ctx_get (inode, this, &tmp_pl_inode);
+                if (ret)
+                        goto unlock;
+        }
+unlock:
+        UNLOCK (&inode->lock);
+        if (ret)
+                goto out;
 
         pl_inode = (pl_inode_t *)(long)tmp_pl_inode;
+        if (!pl_inode) {
+                ret = -1;
+                goto out;
+        }
 
-        if (!pl_inode)
-                return -1;
+        gf_proc_dump_add_section("xlator.features.locks.%s.inode", this->name);
+        section_added = _gf_true;
 
-        gf_proc_dump_build_key(key,
-                               "xlator.feature.locks.inode",
-                               "%ld.mandatory",inode->ino);
-        gf_proc_dump_write(key, "%d", pl_inode->mandatory);
+        /*We are safe to call __inode_path since we have the
+         * inode->table->lock */
+        __inode_path (inode, NULL, &pathname);
+        if (pathname)
+                gf_proc_dump_write ("path", "%s", pathname);
 
-        return 0;
+        gf_proc_dump_write("mandatory", "%d", pl_inode->mandatory);
+
+        ret = pthread_mutex_trylock (&pl_inode->mutex);
+        if (ret)
+                goto out;
+        {
+                count = __get_entrylk_count (this, pl_inode);
+                if (count) {
+                        gf_proc_dump_write("entrylk-count", "%d", count);
+                        __dump_entrylks (pl_inode);
+                }
+
+                count = __get_inodelk_count (this, pl_inode, NULL);
+                if (count) {
+                        gf_proc_dump_write("inodelk-count", "%d", count);
+                        __dump_inodelks (pl_inode);
+                }
+
+                count = __get_posixlk_count (this, pl_inode);
+                if (count) {
+                        gf_proc_dump_write("posixlk-count", "%d", count);
+                        __dump_posixlks (pl_inode);
+                }
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+out:
+        GF_FREE (pathname);
+
+        if (ret && inode) {
+                if (!section_added)
+                        gf_proc_dump_add_section ("xlator.features.locks.%s."
+                                                  "inode", this->name);
+                gf_proc_dump_write ("Unable to print lock state", "(Lock "
+                                    "acquisition failure) %s",
+                                    uuid_utoa (inode->gfid));
+        }
+        return ret;
 }
 
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
 
+        if (!this)
+                return ret;
 
-/*
- * pl_dump_inode - inode dump function for posix locks
- *
- */
-int
-pl_dump_inode (xlator_t *this)
+        ret = xlator_mem_acct_init (this, gf_locks_mt_end + 1);
+
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
+                                "failed");
+                return ret;
+        }
+
+        return ret;
+}
+
+
+pl_ctx_t*
+pl_ctx_get (client_t *client, xlator_t *xlator)
 {
+        void *tmp = NULL;
+        pl_ctx_t *ctx = NULL;
+
+        client_ctx_get (client, xlator, &tmp);
 
-        assert(this);
+        ctx = tmp;
 
-        if (this->itable) {
-                inode_table_dump(this->itable,
-                                 "xlator.features.locks.inode_table");
+        if (ctx != NULL)
+                goto out;
+
+        ctx = GF_CALLOC (1, sizeof (pl_ctx_t), gf_locks_mt_posix_lock_t);
+
+        if (ctx == NULL)
+                goto out;
+
+        pthread_mutex_init (&ctx->lock, NULL);
+	INIT_LIST_HEAD (&ctx->inodelk_lockers);
+	INIT_LIST_HEAD (&ctx->entrylk_lockers);
+
+        if (client_ctx_set (client, xlator, ctx) != 0) {
+                pthread_mutex_destroy (&ctx->lock);
+                GF_FREE (ctx);
+                ctx = NULL;
         }
+out:
+        return ctx;
+}
 
-        return 0;
+
+static int
+pl_client_disconnect_cbk (xlator_t *this, client_t *client)
+{
+        pl_ctx_t *pl_ctx = NULL;
+
+	pl_ctx = pl_ctx_get (client, this);
+
+	pl_inodelk_client_cleanup (this, pl_ctx);
+
+	pl_entrylk_client_cleanup (this, pl_ctx);
+
+	return 0;
 }
 
 
+static int
+pl_client_destroy_cbk (xlator_t *this, client_t *client)
+{
+        void     *tmp    = NULL;
+        pl_ctx_t *pl_ctx = NULL;
+
+	pl_client_disconnect_cbk (this, client);
+
+        client_ctx_del (client, this, &tmp);
+
+        if (tmp == NULL)
+                return 0;
+
+        pl_ctx = tmp;
+
+	GF_ASSERT (list_empty(&pl_ctx->inodelk_lockers));
+	GF_ASSERT (list_empty(&pl_ctx->entrylk_lockers));
+
+	pthread_mutex_destroy (&pl_ctx->lock);
+        GF_FREE (pl_ctx);
+
+        return 0;
+}
+
 
 int
 init (xlator_t *this)
@@ -1040,12 +2585,13 @@ init (xlator_t *this)
         posix_locks_private_t *priv = NULL;
         xlator_list_t         *trav = NULL;
         data_t                *mandatory = NULL;
-	data_t                *trace = NULL;
+        data_t                *trace = NULL;
+        int                   ret = -1;
 
         if (!this->children || this->children->next) {
                 gf_log (this->name, GF_LOG_CRITICAL,
                         "FATAL: posix-locks should have exactly one child");
-                return -1;
+                goto out;
         }
 
         if (!this->parents) {
@@ -1061,34 +2607,43 @@ init (xlator_t *this)
                 gf_log (this->name, GF_LOG_CRITICAL,
                         "'locks' translator is not loaded over a storage "
                         "translator");
-                return -1;
+                goto out;
         }
 
-        priv = CALLOC (1, sizeof (*priv));
+        priv = GF_CALLOC (1, sizeof (*priv),
+                          gf_locks_mt_posix_locks_private_t);
 
         mandatory = dict_get (this->options, "mandatory-locks");
-        if (mandatory) {
-                if (gf_string2boolean (mandatory->data,
-                                       &priv->mandatory) == -1) {
+        if (mandatory)
+                gf_log (this->name, GF_LOG_WARNING,
+                        "mandatory locks not supported in this minor release.");
+
+        trace = dict_get (this->options, "trace");
+        if (trace) {
+                if (gf_string2boolean (trace->data,
+                                       &priv->trace) == -1) {
                         gf_log (this->name, GF_LOG_ERROR,
-                                "'mandatory-locks' takes on only boolean "
-                                "values.");
-                        return -1;
+                                "'trace' takes on only boolean values.");
+                        goto out;
                 }
         }
 
-	trace = dict_get (this->options, "trace");
-	if (trace) {
-		if (gf_string2boolean (trace->data,
-				       &priv->trace) == -1) {
-			gf_log (this->name, GF_LOG_ERROR,
-				"'trace' takes on only boolean values.");
-			return -1;
-		}
-	}
+        this->local_pool = mem_pool_new (pl_local_t, 32);
+        if (!this->local_pool) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create local_t's memory pool");
+                goto out;
+        }
 
         this->private = priv;
-        return 0;
+        ret = 0;
+
+out:
+        if (ret) {
+                GF_FREE (priv);
+        }
+        return ret;
 }
 
 
@@ -1098,7 +2653,11 @@ fini (xlator_t *this)
         posix_locks_private_t *priv = NULL;
 
         priv = this->private;
-        free (priv);
+        if (!priv)
+                return 0;
+        this->private = NULL;
+        GF_FREE (priv->brickname);
+        GF_FREE (priv);
 
         return 0;
 }
@@ -1106,21 +2665,23 @@ fini (xlator_t *this)
 
 int
 pl_inodelk (call_frame_t *frame, xlator_t *this,
-            const char *volume, loc_t *loc, int32_t cmd, struct flock *flock);
+            const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *flock,
+            dict_t *xdata);
 
 int
 pl_finodelk (call_frame_t *frame, xlator_t *this,
-             const char *volume, fd_t *fd, int32_t cmd, struct flock *flock);
+             const char *volume, fd_t *fd, int32_t cmd, struct gf_flock *flock,
+             dict_t *xdata);
 
 int
 pl_entrylk (call_frame_t *frame, xlator_t *this,
             const char *volume, loc_t *loc, const char *basename,
-            entrylk_cmd cmd, entrylk_type type);
+            entrylk_cmd cmd, entrylk_type type, dict_t *xdata);
 
 int
 pl_fentrylk (call_frame_t *frame, xlator_t *this,
              const char *volume, fd_t *fd, const char *basename,
-             entrylk_cmd cmd, entrylk_type type);
+             entrylk_cmd cmd, entrylk_type type, dict_t *xdata);
 
 struct xlator_fops fops = {
         .lookup      = pl_lookup,
@@ -1136,10 +2697,11 @@ struct xlator_fops fops = {
         .entrylk     = pl_entrylk,
         .fentrylk    = pl_fentrylk,
         .flush       = pl_flush,
-};
-
-
-struct xlator_mops mops = {
+        .opendir     = pl_opendir,
+        .readdirp    = pl_readdirp,
+        .getxattr    = pl_getxattr,
+        .fgetxattr   = pl_fgetxattr,
+        .fsetxattr   = pl_fsetxattr,
 };
 
 struct xlator_dumpops dumpops = {
@@ -1147,7 +2709,11 @@ struct xlator_dumpops dumpops = {
 };
 
 struct xlator_cbks cbks = {
-        .forget      = pl_forget,
+        .forget            = pl_forget,
+        .release           = pl_release,
+        .releasedir        = pl_releasedir,
+        .client_destroy    = pl_client_destroy_cbk,
+        .client_disconnect = pl_client_disconnect_cbk,
 };
 
 
@@ -1155,8 +2721,8 @@ struct volume_options options[] = {
         { .key  = { "mandatory-locks", "mandatory" },
           .type = GF_OPTION_TYPE_BOOL
         },
-	{ .key  = { "trace" },
-	  .type = GF_OPTION_TYPE_BOOL
-	},
+        { .key  = { "trace" },
+          .type = GF_OPTION_TYPE_BOOL
+        },
         { .key = {NULL} },
 };
diff --git a/xlators/features/locks/src/reservelk.c b/xlators/features/locks/src/reservelk.c
new file mode 100644
index 000000000..11abd26d8
--- /dev/null
+++ b/xlators/features/locks/src/reservelk.c
@@ -0,0 +1,443 @@
+/*
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "compat.h"
+#include "xlator.h"
+#include "inode.h"
+#include "logging.h"
+#include "common-utils.h"
+#include "list.h"
+
+#include "locks.h"
+#include "common.h"
+
+void
+__delete_reserve_lock (posix_lock_t *lock)
+{
+        list_del (&lock->list);
+}
+
+void
+__destroy_reserve_lock (posix_lock_t *lock)
+{
+        GF_FREE (lock);
+}
+
+/* Return true if the two reservelks have exactly same lock boundaries */
+int
+reservelks_equal (posix_lock_t *l1, posix_lock_t *l2)
+{
+        if ((l1->fl_start == l2->fl_start) &&
+            (l1->fl_end == l2->fl_end))
+                return 1;
+
+        return 0;
+}
+
+/* Determine if lock is grantable or not */
+static posix_lock_t *
+__reservelk_grantable (pl_inode_t *pl_inode, posix_lock_t *lock)
+{
+        xlator_t     *this     = NULL;
+        posix_lock_t *l        = NULL;
+        posix_lock_t *ret_lock = NULL;
+
+        this = THIS;
+
+        if (list_empty (&pl_inode->reservelk_list)) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "No reservelks in list");
+                goto out;
+        }
+        list_for_each_entry (l, &pl_inode->reservelk_list, list){
+                if (reservelks_equal (lock, l)) {
+                        ret_lock = l;
+                        break;
+                }
+        }
+out:
+        return ret_lock;
+}
+
+static inline int
+__same_owner_reservelk (posix_lock_t *l1, posix_lock_t *l2)
+{
+        return (is_same_lkowner (&l1->owner, &l2->owner));
+
+}
+
+static posix_lock_t *
+__matching_reservelk (pl_inode_t *pl_inode, posix_lock_t *lock)
+{
+        posix_lock_t *l = NULL;
+
+        if (list_empty (&pl_inode->reservelk_list)) {
+            gf_log ("posix-locks", GF_LOG_TRACE,
+                    "reservelk list empty");
+            return NULL;
+        }
+
+        list_for_each_entry (l, &pl_inode->reservelk_list, list) {
+                if (reservelks_equal (l, lock)) {
+                        gf_log ("posix-locks", GF_LOG_TRACE,
+                                "equal reservelk found");
+                        break;
+                }
+        }
+
+        return l;
+}
+
+static int
+__reservelk_conflict (xlator_t *this, pl_inode_t *pl_inode,
+                      posix_lock_t *lock)
+{
+        posix_lock_t *conf = NULL;
+        int ret = 0;
+
+        conf = __matching_reservelk (pl_inode, lock);
+        if (conf) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "Matching reservelk found");
+                if (__same_owner_reservelk (lock, conf)) {
+                        list_del_init (&conf->list);
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "Removing the matching reservelk for setlk to progress");
+                        GF_FREE (conf);
+                        ret = 0;
+                } else {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "Conflicting reservelk found");
+                        ret = 1;
+                }
+
+        }
+        return ret;
+
+}
+
+int
+pl_verify_reservelk (xlator_t *this, pl_inode_t *pl_inode,
+                     posix_lock_t *lock, int can_block)
+{
+        int ret = 0;
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                if (__reservelk_conflict (this, pl_inode, lock)) {
+                        gf_log (this->name, GF_LOG_TRACE,
+                                "Found conflicting reservelk. Blocking until reservelk is unlocked.");
+                        lock->blocked = can_block;
+                        list_add_tail (&lock->list, &pl_inode->blocked_calls);
+                        ret = -1;
+                        goto unlock;
+                }
+
+                gf_log (this->name, GF_LOG_TRACE,
+                        "no conflicting reservelk found. Call continuing");
+                ret = 0;
+
+        }
+unlock:
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        return ret;
+
+}
+
+
+/* Determines if lock can be granted and adds the lock. If the lock
+ * is blocking, adds it to the blocked_reservelks.
+ */
+static int
+__lock_reservelk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
+                  int can_block)
+{
+        posix_lock_t *conf = NULL;
+        int ret = -EINVAL;
+
+        conf = __reservelk_grantable (pl_inode, lock);
+        if (conf){
+                ret = -EAGAIN;
+                if (can_block == 0)
+                        goto out;
+
+                list_add_tail (&lock->list, &pl_inode->blocked_reservelks);
+
+                gf_log (this->name, GF_LOG_TRACE,
+                        "%s (pid=%d) lk-owner:%s %"PRId64" - %"PRId64" => Blocked",
+                        lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+                        lock->client_pid,
+                        lkowner_utoa (&lock->owner),
+                        lock->user_flock.l_start,
+                        lock->user_flock.l_len);
+
+
+                goto out;
+        }
+
+        list_add (&lock->list, &pl_inode->reservelk_list);
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+static posix_lock_t *
+find_matching_reservelk (posix_lock_t *lock, pl_inode_t *pl_inode)
+{
+        posix_lock_t *l = NULL;
+        list_for_each_entry (l, &pl_inode->reservelk_list, list) {
+                if (reservelks_equal (l, lock))
+                        return l;
+        }
+        return NULL;
+}
+
+/* Set F_UNLCK removes a lock which has the exact same lock boundaries
+ * as the UNLCK lock specifies. If such a lock is not found, returns invalid
+ */
+static posix_lock_t *
+__reserve_unlock_lock (xlator_t *this, posix_lock_t *lock, pl_inode_t *pl_inode)
+{
+
+        posix_lock_t *conf = NULL;
+
+        conf = find_matching_reservelk (lock, pl_inode);
+        if (!conf) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        " Matching lock not found for unlock");
+                goto out;
+        }
+        __delete_reserve_lock (conf);
+        gf_log (this->name, GF_LOG_DEBUG,
+                " Matching lock found for unlock");
+
+out:
+        return conf;
+
+
+}
+
+static void
+__grant_blocked_reserve_locks (xlator_t *this, pl_inode_t *pl_inode,
+                               struct list_head *granted)
+{
+        int              bl_ret = 0;
+        posix_lock_t *bl = NULL;
+        posix_lock_t *tmp = NULL;
+
+        struct list_head blocked_list;
+
+        INIT_LIST_HEAD (&blocked_list);
+        list_splice_init (&pl_inode->blocked_reservelks, &blocked_list);
+
+        list_for_each_entry_safe (bl, tmp, &blocked_list, list) {
+
+                list_del_init (&bl->list);
+
+                bl_ret = __lock_reservelk (this, pl_inode, bl, 1);
+
+                if (bl_ret == 0) {
+                        list_add (&bl->list, granted);
+                }
+        }
+        return;
+}
+
+/* Grant all reservelks blocked on lock(s) */
+void
+grant_blocked_reserve_locks (xlator_t *this, pl_inode_t *pl_inode)
+{
+        struct list_head granted;
+        posix_lock_t *lock = NULL;
+        posix_lock_t *tmp = NULL;
+
+        INIT_LIST_HEAD (&granted);
+
+        if (list_empty (&pl_inode->blocked_reservelks)) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "No blocked locks to be granted");
+                return;
+        }
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                __grant_blocked_reserve_locks (this, pl_inode, &granted);
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        list_for_each_entry_safe (lock, tmp, &granted, list) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => Granted",
+                        lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+                        lock->client_pid,
+                        lkowner_utoa (&lock->owner),
+                        lock->user_flock.l_start,
+                        lock->user_flock.l_len);
+
+                STACK_UNWIND_STRICT (lk, lock->frame, 0, 0, &lock->user_flock,
+                                     NULL);
+        }
+
+}
+
+static void
+__grant_blocked_lock_calls (xlator_t *this, pl_inode_t *pl_inode,
+                            struct list_head *granted)
+{
+        int              bl_ret = 0;
+        posix_lock_t *bl = NULL;
+        posix_lock_t *tmp = NULL;
+
+        struct list_head blocked_list;
+
+        INIT_LIST_HEAD (&blocked_list);
+        list_splice_init (&pl_inode->blocked_reservelks, &blocked_list);
+
+        list_for_each_entry_safe (bl, tmp, &blocked_list, list) {
+
+                list_del_init (&bl->list);
+
+                bl_ret = pl_verify_reservelk (this, pl_inode, bl, bl->blocked);
+
+                if (bl_ret == 0) {
+                        list_add_tail (&bl->list, granted);
+                }
+        }
+        return;
+}
+
+void
+grant_blocked_lock_calls (xlator_t *this, pl_inode_t *pl_inode)
+{
+        struct list_head granted;
+        posix_lock_t *lock = NULL;
+        posix_lock_t *tmp = NULL;
+        fd_t         *fd  = NULL;
+
+        int can_block = 0;
+        int32_t cmd = 0;
+        int ret = 0;
+
+        if (list_empty (&pl_inode->blocked_calls)) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "No blocked lock calls to be granted");
+                return;
+        }
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                __grant_blocked_lock_calls (this, pl_inode, &granted);
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        list_for_each_entry_safe (lock, tmp, &granted, list) {
+                fd = fd_from_fdnum (lock);
+
+                if (lock->blocked) {
+                        can_block = 1;
+                        cmd = F_SETLKW;
+                }
+                else
+                        cmd = F_SETLK;
+
+                lock->blocked = 0;
+                ret = pl_setlk (this, pl_inode, lock, can_block);
+                if (ret == -1) {
+                        if (can_block) {
+                                pl_trace_block (this, lock->frame, fd, NULL,
+                                                cmd, &lock->user_flock, NULL);
+                                continue;
+                        } else {
+                                gf_log (this->name, GF_LOG_DEBUG, "returning EAGAIN");
+                                pl_trace_out (this, lock->frame, fd, NULL, cmd,
+                                              &lock->user_flock, -1, EAGAIN, NULL);
+                                pl_update_refkeeper (this, fd->inode);
+                                STACK_UNWIND_STRICT (lk, lock->frame, -1,
+                                                     EAGAIN, &lock->user_flock,
+                                                     NULL);
+                                __destroy_lock (lock);
+                        }
+                }
+
+        }
+
+}
+
+
+int
+pl_reserve_unlock (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock)
+{
+        posix_lock_t *retlock = NULL;
+        int ret = -1;
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+                retlock = __reserve_unlock_lock (this, lock, pl_inode);
+                if (!retlock) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "Bad Unlock issued on Inode lock");
+                        ret = -EINVAL;
+                        goto out;
+                }
+
+                gf_log (this->name, GF_LOG_TRACE,
+                        "Reservelk Unlock successful");
+                __destroy_reserve_lock (retlock);
+                ret = 0;
+        }
+out:
+        pthread_mutex_unlock (&pl_inode->mutex);
+
+        grant_blocked_reserve_locks (this, pl_inode);
+        grant_blocked_lock_calls (this, pl_inode);
+
+        return ret;
+
+}
+
+int
+pl_reserve_setlk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
+                  int can_block)
+{
+        int ret = -EINVAL;
+
+        pthread_mutex_lock (&pl_inode->mutex);
+        {
+
+                        ret = __lock_reservelk (this, pl_inode, lock, can_block);
+                        if (ret < 0)
+                                gf_log (this->name, GF_LOG_TRACE,
+                                        "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => NOK",
+                                        lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+                                        lock->client_pid,
+                                        lkowner_utoa (&lock->owner),
+                                        lock->user_flock.l_start,
+                                        lock->user_flock.l_len);
+                        else
+                                gf_log (this->name, GF_LOG_TRACE,
+                                        "%s (pid=%d) (lk-owner=%s) %"PRId64" - %"PRId64" => OK",
+                                        lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+                                        lock->client_pid,
+                                        lkowner_utoa (&lock->owner),
+                                        lock->fl_start,
+                                        lock->fl_end);
+
+        }
+        pthread_mutex_unlock (&pl_inode->mutex);
+        return ret;
+}
diff --git a/xlators/features/locks/tests/unit-test.c b/xlators/features/locks/tests/unit-test.c
index fc69ce8a9..d2cca32de 100644
--- a/xlators/features/locks/tests/unit-test.c
+++ b/xlators/features/locks/tests/unit-test.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2006-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
+   Copyright (c) 2006-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
 */
-
 #ifndef _CONFIG_H
 #define _CONFIG_H
 #include "config.h"
diff --git a/xlators/storage/bdb/Makefile.am b/xlators/features/mac-compat/Makefile.am
index d471a3f92..d471a3f92 100644
--- a/xlators/storage/bdb/Makefile.am
+++ b/xlators/features/mac-compat/Makefile.am
diff --git a/xlators/features/mac-compat/src/Makefile.am b/xlators/features/mac-compat/src/Makefile.am
new file mode 100644
index 000000000..f8567edce
--- /dev/null
+++ b/xlators/features/mac-compat/src/Makefile.am
@@ -0,0 +1,14 @@
+xlator_LTLIBRARIES = mac-compat.la
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+mac_compat_la_LDFLAGS = -module -avoid-version
+
+mac_compat_la_SOURCES = mac-compat.c
+mac_compat_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+CLEANFILES = 
+
diff --git a/xlators/features/mac-compat/src/mac-compat.c b/xlators/features/mac-compat/src/mac-compat.c
new file mode 100644
index 000000000..7cb550ad5
--- /dev/null
+++ b/xlators/features/mac-compat/src/mac-compat.c
@@ -0,0 +1,237 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+#include "compat-errno.h"
+
+
+enum apple_xattr {
+        GF_FINDER_INFO_XATTR,
+        GF_RESOURCE_FORK_XATTR,
+        GF_XATTR_ALL,
+        GF_XATTR_NONE
+};
+
+static char *apple_xattr_name[] = {
+        [GF_FINDER_INFO_XATTR]   = "com.apple.FinderInfo",
+        [GF_RESOURCE_FORK_XATTR] = "com.apple.ResourceFork"
+};
+
+static const char *apple_xattr_value[] = {
+        [GF_FINDER_INFO_XATTR]   =
+        /* 1 2 3 4 5 6 7 8 */
+         "\0\0\0\0\0\0\0\0"
+         "\0\0\0\0\0\0\0\0"
+         "\0\0\0\0\0\0\0\0"
+         "\0\0\0\0\0\0\0\0",
+        [GF_RESOURCE_FORK_XATTR] = ""
+};
+
+static int32_t apple_xattr_len[] = {
+        [GF_FINDER_INFO_XATTR]   = 32,
+        [GF_RESOURCE_FORK_XATTR] = 1
+};
+
+
+int32_t
+maccomp_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *dict,
+                      dict_t *xdata)
+{
+        intptr_t ax = (intptr_t)this->private;
+        int i = 0;
+
+        if ((ax == GF_XATTR_ALL && op_ret >= 0) || ax != GF_XATTR_NONE) {
+                op_ret = op_errno = 0;
+
+                for (i = 0; i < GF_XATTR_ALL; i++) {
+                        if (dict_get (dict, apple_xattr_name[i]))
+                                continue;
+
+                        if (dict_set (dict, apple_xattr_name[i],
+                                      bin_to_data ((void *)apple_xattr_value[i],
+                                                   apple_xattr_len[i])) == -1) {
+                                op_ret = -1;
+                                op_errno = ENOMEM;
+
+                                break;
+                        }
+                }
+         }
+
+        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
+
+        return 0;
+}
+
+
+int32_t
+maccomp_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                  const char *name, dict_t *xdata)
+{
+        intptr_t ax = GF_XATTR_NONE;
+        int i = 0;
+
+        if (name) {
+                for (i = 0; i < GF_XATTR_ALL; i++) {
+                        if (strcmp (apple_xattr_name[i], name) == 0) {
+                                ax = i;
+
+                                break;
+                        }
+                }
+        } else
+                ax = GF_XATTR_ALL;
+
+        this->private = (void *)ax;
+
+        STACK_WIND (frame, maccomp_getxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->getxattr,
+                    loc, name, xdata);
+        return 0;
+}
+
+
+int32_t
+maccomp_fgetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                   const char *name, dict_t *xdata)
+{
+        intptr_t ax = GF_XATTR_NONE;
+        int i = 0;
+
+        if (name) {
+                for (i = 0; i < GF_XATTR_ALL; i++) {
+                        if (strcmp (apple_xattr_name[i], name) == 0) {
+                                ax = i;
+
+                                break;
+                        }
+                }
+        } else
+                ax = GF_XATTR_ALL;
+
+        this->private = (void *)ax;
+
+        STACK_WIND (frame, maccomp_getxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fgetxattr,
+                    fd, name, xdata);
+        return 0;
+}
+
+
+int32_t
+maccomp_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        intptr_t ax = (intptr_t)this->private;
+
+        if (op_ret == -1 && ax != GF_XATTR_NONE)
+                op_ret = op_errno = 0;
+
+        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
+
+        return 0;
+}
+
+
+int32_t
+maccomp_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+                  int32_t flags, dict_t *xdata)
+{
+        intptr_t ax = GF_XATTR_NONE;
+        int i = 0;
+
+        for (i = 0; i < GF_XATTR_ALL; i++) {
+                if (dict_get (dict, apple_xattr_name[i])) {
+                        ax = i;
+
+                        break;
+                }
+        }
+
+        this->private = (void *)ax;
+
+        STACK_WIND (frame, maccomp_setxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr,
+                    loc, dict, flags, xdata);
+        return 0;
+}
+
+
+int32_t
+maccomp_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+                   int32_t flags, dict_t *xdata)
+{
+        intptr_t ax = GF_XATTR_NONE;
+        int i = 0;
+
+        for (i = 0; i < GF_XATTR_ALL; i++) {
+                if (dict_get (dict, apple_xattr_name[i])) {
+                        ax = i;
+
+                        break;
+                }
+        }
+
+        this->private = (void *)ax;
+
+        STACK_WIND (frame, maccomp_setxattr_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsetxattr,
+                    fd, dict, flags, xdata);
+        return 0;
+}
+
+
+int32_t
+init (xlator_t *this)
+{
+        if (!this->children || this->children->next) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "translator not configured with exactly one child");
+                return -1;
+        }
+
+        if (!this->parents) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "dangling volume. check volfile ");
+        }
+
+        return 0;
+}
+
+
+void
+fini (xlator_t *this)
+{
+        return;
+}
+
+
+struct xlator_fops fops = {
+        .getxattr    = maccomp_getxattr,
+        .fgetxattr   = maccomp_fgetxattr,
+        .setxattr    = maccomp_setxattr,
+        .fsetxattr   = maccomp_fsetxattr,
+};
+
+struct xlator_cbks cbks;
+
+struct volume_options options[] = {
+        { .key = {NULL} },
+};
diff --git a/xlators/features/marker/Makefile.am b/xlators/features/marker/Makefile.am
new file mode 100644
index 000000000..a985f42a8
--- /dev/null
+++ b/xlators/features/marker/Makefile.am
@@ -0,0 +1,3 @@
+SUBDIRS = src
+
+CLEANFILES =
diff --git a/xlators/features/marker/src/Makefile.am b/xlators/features/marker/src/Makefile.am
new file mode 100644
index 000000000..a7c676472
--- /dev/null
+++ b/xlators/features/marker/src/Makefile.am
@@ -0,0 +1,17 @@
+xlator_LTLIBRARIES = marker.la
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+marker_la_LDFLAGS = -module -avoid-version
+
+marker_la_SOURCES = marker.c marker-quota.c marker-quota-helper.c marker-common.c
+marker_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+noinst_HEADERS = marker-mem-types.h marker.h marker-quota.h marker-quota-helper.h marker-common.h $(top_builddir)/xlators/lib/src/libxlator.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
+	-I$(top_srcdir)/xlators/lib/src
+
+AM_CFLAGS = -Wall -fno-strict-aliasing $(GF_CFLAGS)
+
+CLEANFILES =
+
diff --git a/xlators/features/marker/src/marker-common.c b/xlators/features/marker/src/marker-common.c
new file mode 100644
index 000000000..84a718add
--- /dev/null
+++ b/xlators/features/marker/src/marker-common.c
@@ -0,0 +1,69 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+#include <fnmatch.h>
+#include "marker-common.h"
+
+marker_inode_ctx_t *
+marker_inode_ctx_new ()
+{
+        marker_inode_ctx_t *ctx = NULL;
+
+        ctx = GF_CALLOC (1, sizeof (marker_inode_ctx_t),
+                         gf_marker_mt_marker_inode_ctx_t);
+        if (ctx == NULL)
+                goto out;
+
+        ctx->quota_ctx = NULL;
+out:
+        return ctx;
+}
+
+int32_t
+marker_force_inode_ctx_get (inode_t *inode, xlator_t *this,
+                            marker_inode_ctx_t **ctx)
+{
+        int32_t  ret     = -1;
+        uint64_t ctx_int = 0;
+
+        LOCK (&inode->lock);
+        {
+                ret = __inode_ctx_get (inode, this, &ctx_int);
+                if (ret == 0)
+                        *ctx = (marker_inode_ctx_t *) (unsigned long)ctx_int;
+                else {
+                        *ctx = marker_inode_ctx_new ();
+                        if (*ctx == NULL)
+                                goto unlock;
+
+                        ret = __inode_ctx_put (inode, this,
+                                               (uint64_t )(unsigned long) *ctx);
+                        if (ret == -1) {
+                                GF_FREE (*ctx);
+                                goto unlock;
+                        }
+                        ret = 0;
+                }
+        }
+unlock: UNLOCK (&inode->lock);
+
+        return ret;
+}
+
+int
+marker_filter_quota_xattr (dict_t *dict, char *key,
+                           data_t *value, void *data)
+{
+        dict_del (dict, key);
+        return 0;
+}
diff --git a/xlators/features/marker/src/marker-common.h b/xlators/features/marker/src/marker-common.h
new file mode 100644
index 000000000..23dd846cb
--- /dev/null
+++ b/xlators/features/marker/src/marker-common.h
@@ -0,0 +1,27 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _MARKER_COMMON_H
+#define _MARKER_COMMON_H
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "inode.h"
+#include "xlator.h"
+#include "marker.h"
+
+int32_t
+marker_force_inode_ctx_get (inode_t *, xlator_t *, marker_inode_ctx_t **);
+
+int
+marker_filter_quota_xattr (dict_t *, char *, data_t *, void *);
+#endif
diff --git a/xlators/features/marker/src/marker-mem-types.h b/xlators/features/marker/src/marker-mem-types.h
new file mode 100644
index 000000000..1f74d5048
--- /dev/null
+++ b/xlators/features/marker/src/marker-mem-types.h
@@ -0,0 +1,25 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef __MARKER_MEM_TYPES_H__
+#define __MARKER_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_marker_mem_types_ {
+        gf_marker_mt_marker_conf_t = gf_common_mt_end + 1,
+        gf_marker_mt_loc_t,
+        gf_marker_mt_volume_mark,
+        gf_marker_mt_int64_t,
+        gf_marker_mt_quota_inode_ctx_t,
+        gf_marker_mt_marker_inode_ctx_t,
+        gf_marker_mt_inode_contribution_t,
+        gf_marker_mt_end
+};
+#endif
diff --git a/xlators/features/marker/src/marker-quota-helper.c b/xlators/features/marker/src/marker-quota-helper.c
new file mode 100644
index 000000000..ec0d83316
--- /dev/null
+++ b/xlators/features/marker/src/marker-quota-helper.c
@@ -0,0 +1,423 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "locking.h"
+#include "marker-quota.h"
+#include "marker-common.h"
+#include "marker-quota-helper.h"
+#include "marker-mem-types.h"
+
+int
+mq_loc_fill (loc_t *loc, inode_t *inode, inode_t *parent, char *path)
+{
+        int ret = -1;
+
+        GF_VALIDATE_OR_GOTO ("marker", loc, out);
+        GF_VALIDATE_OR_GOTO ("marker", inode, out);
+        GF_VALIDATE_OR_GOTO ("marker", path, out);
+        /* Not checking for parent because while filling
+         * loc of root, parent will be NULL
+         */
+
+        if (inode) {
+                loc->inode = inode_ref (inode);
+        }
+
+        if (parent)
+                loc->parent = inode_ref (parent);
+
+        loc->path = gf_strdup (path);
+        if (!loc->path) {
+                gf_log ("loc fill", GF_LOG_ERROR, "strdup failed");
+                goto loc_wipe;
+        }
+
+        loc->name = strrchr (loc->path, '/');
+        if (loc->name)
+                loc->name++;
+        else
+                goto loc_wipe;
+
+        ret = 0;
+loc_wipe:
+        if (ret < 0)
+                loc_wipe (loc);
+out:
+        return ret;
+}
+
+
+int32_t
+mq_inode_loc_fill (const char *parent_gfid, inode_t *inode, loc_t *loc)
+{
+        char            *resolvedpath = NULL;
+        inode_t         *parent       = NULL;
+        int              ret          = -1;
+
+        if ((!inode) || (!loc))
+                return ret;
+
+        if ((inode) && __is_root_gfid (inode->gfid)) {
+                loc->parent = NULL;
+                goto ignore_parent;
+        }
+
+        if (parent_gfid == NULL)
+                parent = inode_parent (inode, 0, NULL);
+        else
+                parent = inode_find (inode->table,
+                                     (unsigned char *) parent_gfid);
+
+        if (parent == NULL)
+                goto err;
+
+ignore_parent:
+        ret = inode_path (inode, NULL, &resolvedpath);
+        if (ret < 0)
+                goto err;
+
+        ret = mq_loc_fill (loc, inode, parent, resolvedpath);
+        if (ret < 0)
+                goto err;
+
+err:
+        if (parent)
+                inode_unref (parent);
+
+        GF_FREE (resolvedpath);
+
+        return ret;
+}
+
+
+quota_inode_ctx_t *
+mq_alloc_inode_ctx ()
+{
+        int32_t                  ret    = -1;
+        quota_inode_ctx_t       *ctx    = NULL;
+
+        QUOTA_ALLOC (ctx, quota_inode_ctx_t, ret);
+        if (ret == -1)
+                goto out;
+
+        ctx->size = 0;
+        ctx->dirty = 0;
+        ctx->updation_status = _gf_false;
+        LOCK_INIT (&ctx->lock);
+        INIT_LIST_HEAD (&ctx->contribution_head);
+out:
+        return ctx;
+}
+
+inode_contribution_t *
+mq_get_contribution_node (inode_t *inode, quota_inode_ctx_t *ctx)
+{
+        inode_contribution_t    *contri = NULL;
+        inode_contribution_t    *temp   = NULL;
+
+        if (!inode || !ctx)
+                goto out;
+
+        list_for_each_entry (temp, &ctx->contribution_head, contri_list) {
+                if (uuid_compare (temp->gfid, inode->gfid) == 0) {
+                        contri = temp;
+                        goto out;
+                }
+        }
+out:
+        return contri;
+}
+
+
+int32_t
+mq_delete_contribution_node (dict_t *dict, char *key,
+                             inode_contribution_t *contribution)
+{
+        if (dict_get (dict, key) != NULL)
+                goto out;
+
+        QUOTA_FREE_CONTRIBUTION_NODE (contribution);
+out:
+        return 0;
+}
+
+
+inode_contribution_t *
+__mq_add_new_contribution_node (xlator_t *this, quota_inode_ctx_t *ctx,
+                                loc_t *loc)
+{
+        int32_t               ret          = 0;
+        inode_contribution_t *contribution = NULL;
+
+        if (!loc->parent) {
+                if (!uuid_is_null (loc->pargfid))
+                        loc->parent = inode_find (loc->inode->table,
+                                                  loc->pargfid);
+
+                if (!loc->parent)
+                        loc->parent = inode_parent (loc->inode, loc->pargfid,
+                                                    loc->name);
+                if (!loc->parent)
+                        goto out;
+        }
+
+        list_for_each_entry (contribution, &ctx->contribution_head,
+                             contri_list) {
+                if (loc->parent &&
+                    uuid_compare (contribution->gfid, loc->parent->gfid) == 0) {
+                        goto out;
+                }
+        }
+
+        QUOTA_ALLOC (contribution, inode_contribution_t, ret);
+        if (ret == -1)
+                goto out;
+
+        contribution->contribution = 0;
+
+        uuid_copy (contribution->gfid, loc->parent->gfid);
+
+	LOCK_INIT (&contribution->lock);
+        INIT_LIST_HEAD (&contribution->contri_list);
+
+        list_add_tail (&contribution->contri_list, &ctx->contribution_head);
+
+out:
+        return contribution;
+}
+
+
+inode_contribution_t *
+mq_add_new_contribution_node (xlator_t *this, quota_inode_ctx_t *ctx,
+                              loc_t *loc)
+{
+        inode_contribution_t *contribution = NULL;
+
+        if ((ctx == NULL) || (loc == NULL))
+                return NULL;
+
+        if (((loc->path) && (strcmp (loc->path, "/") == 0))
+            || (!loc->path && uuid_is_null (loc->pargfid)))
+                return NULL;
+
+        LOCK (&ctx->lock);
+        {
+                contribution = __mq_add_new_contribution_node (this, ctx, loc);
+        }
+        UNLOCK (&ctx->lock);
+
+        return contribution;
+}
+
+
+int32_t
+mq_dict_set_contribution (xlator_t *this, dict_t *dict,
+                          loc_t *loc)
+{
+        int32_t ret              = -1;
+        char    contri_key [512] = {0, };
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", dict, out);
+        GF_VALIDATE_OR_GOTO ("marker", loc, out);
+
+        if (loc->parent) {
+                GET_CONTRI_KEY (contri_key, loc->parent->gfid, ret);
+                if (ret < 0) {
+                        ret = -1;
+                        goto out;
+                }
+        } else {
+                /* nameless lookup, fetch contributions to all parents */
+                GET_CONTRI_KEY (contri_key, NULL, ret);
+        }
+
+        ret = dict_set_int64 (dict, contri_key, 0);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "unable to set dict value on %s.",
+                        loc->path);
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+
+int32_t
+mq_inode_ctx_get (inode_t *inode, xlator_t *this,
+                  quota_inode_ctx_t **ctx)
+{
+        int32_t             ret      = -1;
+        uint64_t            ctx_int  = 0;
+        marker_inode_ctx_t *mark_ctx = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", inode, out);
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", ctx, out);
+
+        ret = inode_ctx_get (inode, this, &ctx_int);
+        if (ret < 0) {
+                ret = -1;
+                *ctx = NULL;
+                goto out;
+        }
+
+        mark_ctx = (marker_inode_ctx_t *) (unsigned long)ctx_int;
+        if (mark_ctx->quota_ctx == NULL) {
+                ret = -1;
+                goto out;
+        }
+
+        *ctx = mark_ctx->quota_ctx;
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+
+quota_inode_ctx_t *
+__mq_inode_ctx_new (inode_t *inode, xlator_t *this)
+{
+        int32_t               ret        = -1;
+        quota_inode_ctx_t    *quota_ctx  = NULL;
+        marker_inode_ctx_t   *mark_ctx   = NULL;
+
+        ret = marker_force_inode_ctx_get (inode, this, &mark_ctx);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "marker_force_inode_ctx_get() failed");
+                goto out;
+        }
+
+        LOCK (&inode->lock);
+        {
+                if (mark_ctx->quota_ctx == NULL) {
+                        quota_ctx = mq_alloc_inode_ctx ();
+                        if (quota_ctx == NULL) {
+                                ret = -1;
+                                goto unlock;
+                        }
+                        mark_ctx->quota_ctx = quota_ctx;
+                } else {
+                        quota_ctx = mark_ctx->quota_ctx;
+                }
+
+                ret = 0;
+        }
+unlock:
+        UNLOCK (&inode->lock);
+out:
+        return quota_ctx;
+}
+
+
+quota_inode_ctx_t *
+mq_inode_ctx_new (inode_t * inode, xlator_t *this)
+{
+        return __mq_inode_ctx_new (inode, this);
+}
+
+quota_local_t *
+mq_local_new ()
+{
+        quota_local_t  *local   = NULL;
+
+        local = mem_get0 (THIS->local_pool);
+        if (!local)
+                goto out;
+
+        local->ref = 1;
+        LOCK_INIT (&local->lock);
+
+        local->ctx = NULL;
+        local->contri = NULL;
+
+out:
+        return local;
+}
+
+quota_local_t *
+mq_local_ref (quota_local_t *local)
+{
+        LOCK (&local->lock);
+        {
+                local->ref ++;
+        }
+        UNLOCK (&local->lock);
+
+        return local;
+}
+
+
+int32_t
+mq_local_unref (xlator_t *this, quota_local_t *local)
+{
+        int32_t ref = 0;
+        if (local == NULL)
+                goto out;
+
+        QUOTA_SAFE_DECREMENT (&local->lock, local->ref, ref);
+
+        if (ref != 0)
+                goto out;
+
+        if (local->fd != NULL)
+                fd_unref (local->fd);
+
+        loc_wipe (&local->loc);
+
+        loc_wipe (&local->parent_loc);
+
+        LOCK_DESTROY (&local->lock);
+
+        mem_put (local);
+out:
+        return 0;
+}
+
+
+inode_contribution_t *
+mq_get_contribution_from_loc (xlator_t *this, loc_t *loc)
+{
+        int32_t               ret          = 0;
+        quota_inode_ctx_t    *ctx          = NULL;
+        inode_contribution_t *contribution = NULL;
+
+        ret = mq_inode_ctx_get (loc->inode, this, &ctx);
+        if (ret < 0) {
+                gf_log_callingfn (this->name, GF_LOG_WARNING,
+                                  "cannot get marker-quota context from inode "
+                                  "(gfid:%s, path:%s)",
+                                  uuid_utoa (loc->inode->gfid), loc->path);
+                goto err;
+        }
+
+        contribution = mq_get_contribution_node (loc->parent, ctx);
+        if (contribution == NULL) {
+                gf_log_callingfn (this->name, GF_LOG_WARNING,
+                                  "inode (gfid:%s, path:%s) has "
+                                  "no contribution towards parent (gfid:%s)",
+                                  uuid_utoa (loc->inode->gfid),
+                                  loc->path, uuid_utoa (loc->parent->gfid));
+                goto err;
+        }
+
+err:
+        return contribution;
+}
diff --git a/xlators/features/marker/src/marker-quota-helper.h b/xlators/features/marker/src/marker-quota-helper.h
new file mode 100644
index 000000000..6cdd14881
--- /dev/null
+++ b/xlators/features/marker/src/marker-quota-helper.h
@@ -0,0 +1,76 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef _MARKER_QUOTA_HELPER_H
+#define _MARKER_QUOTA_HELPER
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "marker.h"
+
+#define QUOTA_FREE_CONTRIBUTION_NODE(_contribution)     \
+        do {                                            \
+                list_del (&_contribution->contri_list); \
+                GF_FREE (_contribution);                \
+        } while (0)
+
+#define QUOTA_SAFE_INCREMENT(lock, var)                 \
+        do {                                            \
+                LOCK (lock);                            \
+                        var ++;                         \
+                UNLOCK (lock);                          \
+        } while (0)
+
+#define QUOTA_SAFE_DECREMENT(lock, var, value)  \
+        do {                                    \
+                LOCK (lock);                    \
+                {                               \
+                      value = --var;            \
+                }                               \
+                UNLOCK (lock);                  \
+        } while (0)
+
+inode_contribution_t *
+mq_add_new_contribution_node (xlator_t *, quota_inode_ctx_t *, loc_t *);
+
+int32_t
+mq_dict_set_contribution (xlator_t *, dict_t *, loc_t *);
+
+quota_inode_ctx_t *
+mq_inode_ctx_new (inode_t *, xlator_t *);
+
+int32_t
+mq_inode_ctx_get (inode_t *, xlator_t *, quota_inode_ctx_t **);
+
+int32_t
+mq_delete_contribution_node (dict_t *, char *, inode_contribution_t *);
+
+int32_t
+mq_inode_loc_fill (const char *, inode_t *, loc_t *);
+
+quota_local_t *
+mq_local_new ();
+
+quota_local_t *
+mq_local_ref (quota_local_t *);
+
+int32_t
+mq_local_unref (xlator_t *, quota_local_t *);
+
+inode_contribution_t *
+mq_get_contribution_node (inode_t *, quota_inode_ctx_t *);
+
+inode_contribution_t *
+mq_get_contribution_from_loc (xlator_t *this, loc_t *loc);
+
+#endif
diff --git a/xlators/features/marker/src/marker-quota.c b/xlators/features/marker/src/marker-quota.c
new file mode 100644
index 000000000..a758e938f
--- /dev/null
+++ b/xlators/features/marker/src/marker-quota.c
@@ -0,0 +1,2703 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "dict.h"
+#include "xlator.h"
+#include "defaults.h"
+#include "libxlator.h"
+#include "common-utils.h"
+#include "byte-order.h"
+#include "marker-quota.h"
+#include "marker-quota-helper.h"
+
+int
+mq_loc_copy (loc_t *dst, loc_t *src)
+{
+        int ret = -1;
+
+        GF_VALIDATE_OR_GOTO ("marker", dst, out);
+        GF_VALIDATE_OR_GOTO ("marker", src, out);
+
+        if (src->inode == NULL ||
+            ((src->parent == NULL) && (uuid_is_null (src->pargfid))
+             && !__is_root_gfid (src->inode->gfid))) {
+                gf_log ("marker", GF_LOG_WARNING,
+                        "src loc is not valid");
+                goto out;
+        }
+
+        ret = loc_copy (dst, src);
+out:
+        return ret;
+}
+
+int32_t
+mq_get_local_err (quota_local_t *local,
+                  int32_t *val)
+{
+        int32_t ret = -1;
+
+        GF_VALIDATE_OR_GOTO ("marker", local, out);
+        GF_VALIDATE_OR_GOTO ("marker", val, out);
+
+        LOCK (&local->lock);
+        {
+                *val = local->err;
+        }
+        UNLOCK (&local->lock);
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int32_t
+mq_get_ctx_updation_status (quota_inode_ctx_t *ctx,
+                            gf_boolean_t *status)
+{
+        int32_t   ret = -1;
+
+        GF_VALIDATE_OR_GOTO ("marker", ctx, out);
+        GF_VALIDATE_OR_GOTO ("marker", status, out);
+
+        LOCK (&ctx->lock);
+        {
+                *status = ctx->updation_status;
+        }
+        UNLOCK (&ctx->lock);
+
+        ret = 0;
+out:
+        return ret;
+}
+
+
+int32_t
+mq_set_ctx_updation_status (quota_inode_ctx_t *ctx,
+                            gf_boolean_t status)
+{
+        int32_t   ret = -1;
+
+        if (ctx == NULL)
+                goto out;
+
+        LOCK (&ctx->lock);
+        {
+                ctx->updation_status = status;
+        }
+        UNLOCK (&ctx->lock);
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int32_t
+mq_test_and_set_ctx_updation_status (quota_inode_ctx_t *ctx,
+                                     gf_boolean_t *status)
+{
+        int32_t         ret     = -1;
+        gf_boolean_t    temp    = _gf_false;
+
+        GF_VALIDATE_OR_GOTO ("marker", ctx, out);
+        GF_VALIDATE_OR_GOTO ("marker", status, out);
+
+        LOCK (&ctx->lock);
+        {
+                temp = *status;
+                *status = ctx->updation_status;
+                ctx->updation_status = temp;
+        }
+        UNLOCK (&ctx->lock);
+
+        ret = 0;
+out:
+        return ret;
+}
+
+void
+mq_assign_lk_owner (xlator_t *this, call_frame_t *frame)
+{
+        marker_conf_t *conf     = NULL;
+        uint64_t       lk_owner = 0;
+
+        conf = this->private;
+
+        LOCK (&conf->lock);
+        {
+                if (++conf->quota_lk_owner == 0) {
+                        ++conf->quota_lk_owner;
+                }
+
+                lk_owner = conf->quota_lk_owner;
+        }
+        UNLOCK (&conf->lock);
+
+        set_lk_owner_from_uint64 (&frame->root->lk_owner, lk_owner);
+
+        return;
+}
+
+
+int32_t
+mq_loc_fill_from_name (xlator_t *this, loc_t *newloc, loc_t *oldloc,
+                       uint64_t ino, char *name)
+{
+        int32_t   ret  = -1;
+        int32_t   len  = 0;
+        char     *path = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", newloc, out);
+        GF_VALIDATE_OR_GOTO ("marker", oldloc, out);
+        GF_VALIDATE_OR_GOTO ("marker", name, out);
+
+        newloc->inode = inode_new (oldloc->inode->table);
+
+        if (!newloc->inode) {
+                ret = -1;
+                goto out;
+        }
+
+        newloc->parent = inode_ref (oldloc->inode);
+        uuid_copy (newloc->pargfid, oldloc->inode->gfid);
+
+        len = strlen (oldloc->path);
+
+        if (oldloc->path [len - 1] == '/')
+                ret = gf_asprintf ((char **) &path, "%s%s",
+                                   oldloc->path, name);
+        else
+                ret = gf_asprintf ((char **) &path, "%s/%s",
+                                   oldloc->path, name);
+
+        if (ret < 0)
+                goto out;
+
+        newloc->path = path;
+
+        newloc->name = strrchr (newloc->path, '/');
+
+        if (newloc->name)
+                newloc->name++;
+
+        gf_log (this->name, GF_LOG_DEBUG, "path = %s name =%s",
+                newloc->path, newloc->name);
+out:
+        return ret;
+}
+
+int32_t
+mq_dirty_inode_updation_done (call_frame_t *frame, void *cookie, xlator_t *this,
+                              int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        QUOTA_STACK_DESTROY (frame, this);
+
+        return 0;
+}
+
+int32_t
+mq_release_lock_on_dirty_inode (call_frame_t *frame, void *cookie, xlator_t *this,
+                                int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        struct gf_flock   lock  = {0, };
+        quota_local_t    *local = NULL;
+        loc_t             loc = {0, };
+	int               ret = -1;
+
+        local = frame->local;
+
+        if (op_ret == -1) {
+                local->err = -1;
+
+                mq_dirty_inode_updation_done (frame, NULL, this, 0, 0, NULL);
+
+                return 0;
+        }
+
+        if (op_ret == 0)
+                local->ctx->dirty = 0;
+
+        lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+        ret = loc_copy (&loc, &local->loc);
+	if (ret == -1) {
+                local->err = -1;
+                frame->local = NULL;
+                mq_dirty_inode_updation_done (frame, NULL, this, 0, 0, NULL);
+                return 0;
+        }
+
+        if (local->loc.inode == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Inode is NULL, so can't stackwind.");
+                goto out;
+        }
+
+        STACK_WIND (frame,
+                    mq_dirty_inode_updation_done,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &loc, F_SETLKW, &lock, NULL);
+
+        loc_wipe (&loc);
+
+        return 0;
+out:
+        mq_dirty_inode_updation_done (frame, NULL, this, -1, 0, NULL);
+
+        return 0;
+}
+
+int32_t
+mq_mark_inode_undirty (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *dict,
+                       dict_t *xdata)
+{
+        int32_t        ret     = -1;
+        int64_t       *size    = NULL;
+        dict_t        *newdict = NULL;
+        quota_local_t *local   = NULL;
+
+        local = (quota_local_t *) frame->local;
+
+        if (op_ret == -1)
+                goto err;
+
+        if (!dict)
+                goto wind;
+
+        ret = dict_get_bin (dict, QUOTA_SIZE_KEY, (void **) &size);
+        if (ret)
+                goto wind;
+
+        LOCK (&local->ctx->lock);
+        {
+                local->ctx->size = ntoh64 (*size);
+        }
+        UNLOCK (&local->ctx->lock);
+
+wind:
+        newdict = dict_new ();
+        if (!newdict)
+                goto err;
+
+        ret = dict_set_int8 (newdict, QUOTA_DIRTY_KEY, 0);
+        if (ret)
+                goto err;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+        STACK_WIND (frame, mq_release_lock_on_dirty_inode,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr,
+                    &local->loc, newdict, 0, NULL);
+        ret = 0;
+
+err:
+        if (op_ret == -1 || ret == -1) {
+                local->err = -1;
+
+                mq_release_lock_on_dirty_inode (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (newdict)
+                dict_unref (newdict);
+
+        return 0;
+}
+
+int32_t
+mq_update_size_xattr (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, inode_t *inode,
+                      struct iatt *buf, dict_t *dict, struct iatt *postparent)
+{
+        int32_t          ret      = -1;
+        dict_t          *new_dict = NULL;
+        int64_t         *size     = NULL;
+        int64_t         *delta    = NULL;
+        quota_local_t   *local    = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1)
+                goto err;
+
+        if (dict == NULL) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Dict is null while updating the size xattr %s",
+                        local->loc.path?local->loc.path:"");
+                goto err;
+        }
+
+        ret = dict_get_bin (dict, QUOTA_SIZE_KEY, (void **) &size);
+        if (!size) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to get the size, %s",
+                        local->loc.path?local->loc.path:"");
+                goto err;
+        }
+
+        QUOTA_ALLOC_OR_GOTO (delta, int64_t, ret, err);
+
+        *delta = hton64 (local->sum - ntoh64 (*size));
+
+        gf_log (this->name, GF_LOG_DEBUG, "calculated size = %"PRId64", "
+                "original size = %"PRIu64
+                " path = %s diff = %"PRIu64, local->sum, ntoh64 (*size),
+                local->loc.path, ntoh64 (*delta));
+
+        new_dict = dict_new ();
+        if (!new_dict);
+
+        ret = dict_set_bin (new_dict, QUOTA_SIZE_KEY, delta, 8);
+        if (ret)
+                goto err;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+
+        STACK_WIND (frame, mq_mark_inode_undirty, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->xattrop, &local->loc,
+                    GF_XATTROP_ADD_ARRAY64, new_dict, NULL);
+
+        ret = 0;
+
+err:
+        if (op_ret == -1 || ret == -1) {
+                local->err = -1;
+
+                mq_release_lock_on_dirty_inode (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (new_dict)
+                dict_unref (new_dict);
+
+        return 0;
+}
+
+int32_t
+mq_test_and_set_local_err(quota_local_t *local,
+                          int32_t *val)
+{
+        int     tmp = 0;
+        int32_t ret = -1;
+
+        GF_VALIDATE_OR_GOTO ("marker", local, out);
+        GF_VALIDATE_OR_GOTO ("marker", val, out);
+
+        LOCK (&local->lock);
+        {
+                tmp = local->err;
+                local->err = *val;
+                *val = tmp;
+        }
+        UNLOCK (&local->lock);
+
+        ret = 0;
+out:
+        return ret;
+}
+
+int32_t
+mq_get_dirty_inode_size (call_frame_t *frame, xlator_t *this)
+{
+        int32_t        ret   = -1;
+        dict_t        *dict  = NULL;
+        quota_local_t *local = NULL;
+
+        local = (quota_local_t *) frame->local;
+
+        dict = dict_new ();
+        if (!dict) {
+                ret = -1;
+                goto err;
+        }
+
+        ret = dict_set_int64 (dict, QUOTA_SIZE_KEY, 0);
+        if (ret)
+                goto err;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+
+        STACK_WIND (frame, mq_update_size_xattr, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, &local->loc, dict);
+        ret =0;
+
+err:
+        if (ret) {
+                local->err = -1;
+
+                mq_release_lock_on_dirty_inode (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (dict)
+                dict_unref (dict);
+
+        return 0;
+}
+
+int32_t
+mq_get_child_contribution (call_frame_t *frame,
+                           void *cookie,
+                           xlator_t *this,
+                           int32_t op_ret,
+                           int32_t op_errno,
+                           inode_t *inode,
+                           struct iatt *buf,
+                           dict_t *dict,
+                           struct iatt *postparent)
+{
+        int32_t        ret                = -1;
+        int32_t        val                = 0;
+        char           contri_key [512]   = {0, };
+        int64_t       *contri             = NULL;
+        quota_local_t *local              = NULL;
+
+        local = frame->local;
+
+        frame->local = NULL;
+
+        QUOTA_STACK_DESTROY (frame, this);
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR, "%s",
+                        strerror (op_errno));
+                val = -2;
+                if (!mq_test_and_set_local_err (local, &val) &&
+                    val != -2)
+                        mq_release_lock_on_dirty_inode (local->frame, NULL,
+                                                        this, 0, 0, NULL);
+
+                goto exit;
+        }
+
+        ret = mq_get_local_err (local, &val);
+        if (!ret && val == -2)
+                goto exit;
+
+        GET_CONTRI_KEY (contri_key, local->loc.inode->gfid, ret);
+        if (ret < 0)
+                goto out;
+
+        if (!dict)
+                goto out;
+
+        if (dict_get_bin (dict, contri_key, (void **) &contri) == 0)
+                local->sum += ntoh64 (*contri);
+
+out:
+        LOCK (&local->lock);
+        {
+                val = --local->dentry_child_count;
+        }
+        UNLOCK (&local->lock);
+
+        if (val == 0) {
+                mq_dirty_inode_readdir (local->frame, NULL, this,
+                                        0, 0, NULL, NULL);
+        }
+        mq_local_unref (this, local);
+
+        return 0;
+exit:
+        mq_local_unref (this, local);
+        return 0;
+}
+
+int32_t
+mq_readdir_cbk (call_frame_t *frame,
+                void *cookie,
+                xlator_t *this,
+                int32_t op_ret,
+                int32_t op_errno,
+                gf_dirent_t *entries, dict_t *xdata)
+{
+        char           contri_key [512]   = {0, };
+        int32_t        ret                = 0;
+        int32_t        val                = 0;
+        off_t          offset             = 0;
+        int32_t        count              = 0;
+        dict_t        *dict               = NULL;
+        quota_local_t *local              = NULL;
+        gf_dirent_t   *entry              = NULL;
+        call_frame_t  *newframe           = NULL;
+        loc_t          loc                = {0, };
+
+        local = mq_local_ref (frame->local);
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "readdir failed %s", strerror (op_errno));
+                local->err = -1;
+
+                mq_release_lock_on_dirty_inode (frame, NULL, this, 0, 0, NULL);
+
+                goto end;
+        } else if (op_ret == 0) {
+                mq_get_dirty_inode_size (frame, this);
+
+                goto end;
+        }
+
+        local->dentry_child_count =  0;
+
+        list_for_each_entry (entry, (&entries->list), list) {
+                gf_log (this->name, GF_LOG_DEBUG, "entry  = %s", entry->d_name);
+
+                if ((!strcmp (entry->d_name, ".")) || (!strcmp (entry->d_name,
+                                                                ".."))) {
+                        gf_log (this->name, GF_LOG_DEBUG, "entry  = %s",
+                                entry->d_name);
+                        continue;
+                }
+
+                offset = entry->d_off;
+                count++;
+        }
+
+        if (count == 0) {
+                mq_get_dirty_inode_size (frame, this);
+                goto end;
+
+        }
+
+        local->frame = frame;
+
+        LOCK (&local->lock);
+        {
+                local->dentry_child_count = count;
+                local->d_off = offset;
+        }
+        UNLOCK (&local->lock);
+
+
+        list_for_each_entry (entry, (&entries->list), list) {
+                gf_log (this->name, GF_LOG_DEBUG, "entry  = %s", entry->d_name);
+
+                if ((!strcmp (entry->d_name, ".")) || (!strcmp (entry->d_name,
+                                                                ".."))) {
+                        gf_log (this->name, GF_LOG_DEBUG, "entry  = %s",
+                                entry->d_name);
+                        continue;
+                }
+
+                ret = mq_loc_fill_from_name (this, &loc, &local->loc,
+                                             entry->d_ino, entry->d_name);
+                if (ret < 0)
+                        goto out;
+
+                ret = 0;
+
+                LOCK (&local->lock);
+                {
+                        if (local->err != -2) {
+                                newframe = copy_frame (frame);
+                                if (!newframe) {
+                                        ret = -1;
+                                }
+                        } else
+                                ret = -1;
+                }
+                UNLOCK (&local->lock);
+
+                if (ret == -1)
+                        goto out;
+
+                newframe->local = mq_local_ref (local);
+
+                dict = dict_new ();
+                if (!dict) {
+                        ret = -1;
+                        goto out;
+                }
+
+                GET_CONTRI_KEY (contri_key, local->loc.inode->gfid, ret);
+                if (ret < 0)
+                        goto out;
+
+                ret = dict_set_int64 (dict, contri_key, 0);
+                if (ret)
+                        goto out;
+
+                STACK_WIND (newframe,
+                            mq_get_child_contribution,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->lookup,
+                            &loc, dict);
+
+                offset = entry->d_off;
+
+                loc_wipe (&loc);
+
+                newframe = NULL;
+
+        out:
+                if (dict) {
+                        dict_unref (dict);
+                        dict = NULL;
+                }
+
+                if (ret) {
+                        val = -2;
+                        mq_test_and_set_local_err (local, &val);
+
+                        if (newframe) {
+                                newframe->local = NULL;
+                                mq_local_unref(this, local);
+                                QUOTA_STACK_DESTROY (newframe, this);
+                        }
+
+                        break;
+                }
+        }
+
+        if (ret && val != -2) {
+                mq_release_lock_on_dirty_inode (frame, NULL, this, 0, 0, NULL);
+        }
+end:
+        mq_local_unref (this, local);
+
+        return 0;
+}
+
+int32_t
+mq_dirty_inode_readdir (call_frame_t *frame,
+                        void *cookie,
+                        xlator_t *this,
+                        int32_t op_ret,
+                        int32_t op_errno,
+                        fd_t *fd, dict_t *xdata)
+{
+        quota_local_t *local = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1) {
+                local->err = -1;
+                mq_release_lock_on_dirty_inode (frame, NULL, this, 0, 0, NULL);
+                return 0;
+        }
+
+        if (local->fd == NULL)
+                local->fd = fd_ref (fd);
+
+        STACK_WIND (frame,
+                    mq_readdir_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->readdir,
+                    local->fd, READDIR_BUF, local->d_off, xdata);
+
+        return 0;
+}
+
+int32_t
+mq_check_if_still_dirty (call_frame_t *frame,
+                         void *cookie,
+                         xlator_t *this,
+                         int32_t op_ret,
+                         int32_t op_errno,
+                         inode_t *inode,
+                         struct iatt *buf,
+                         dict_t *dict,
+                         struct iatt *postparent)
+{
+        int8_t           dirty          = -1;
+        int32_t          ret            = -1;
+        fd_t            *fd             = NULL;
+        quota_local_t   *local          = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR, "failed to get "
+                        "the dirty xattr for %s", local->loc.path);
+                goto err;
+        }
+
+        if (!dict) {
+                ret = -1;
+                goto err;
+        }
+
+        ret = dict_get_int8 (dict, QUOTA_DIRTY_KEY, &dirty);
+        if (ret)
+                goto err;
+
+        //the inode is not dirty anymore
+        if (dirty == 0) {
+                mq_release_lock_on_dirty_inode (frame, NULL, this, 0, 0, NULL);
+
+                return 0;
+        }
+
+        fd = fd_create (local->loc.inode, frame->root->pid);
+
+        local->d_off = 0;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+        STACK_WIND(frame,
+                   mq_dirty_inode_readdir,
+                   FIRST_CHILD(this),
+                   FIRST_CHILD(this)->fops->opendir,
+                   &local->loc, fd, NULL);
+
+        ret = 0;
+
+err:
+        if (op_ret == -1 || ret == -1) {
+                local->err = -1;
+                mq_release_lock_on_dirty_inode (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (fd != NULL) {
+                fd_unref (fd);
+        }
+
+        return 0;
+}
+
+int32_t
+mq_get_dirty_xattr (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        int32_t        ret       = -1;
+        dict_t        *xattr_req = NULL;
+        quota_local_t *local     = NULL;
+
+        if (op_ret == -1) {
+                mq_dirty_inode_updation_done (frame, NULL, this, 0, 0, NULL);
+                return 0;
+        }
+
+        local = frame->local;
+
+        xattr_req = dict_new ();
+        if (xattr_req == NULL) {
+                ret = -1;
+                goto err;
+        }
+
+        ret = dict_set_int8 (xattr_req, QUOTA_DIRTY_KEY, 0);
+        if (ret)
+                goto err;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+
+        STACK_WIND (frame,
+                    mq_check_if_still_dirty,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup,
+                    &local->loc,
+                    xattr_req);
+        ret = 0;
+
+err:
+        if (ret) {
+                local->err = -1;
+                mq_release_lock_on_dirty_inode(frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (xattr_req)
+                dict_unref (xattr_req);
+
+        return 0;
+}
+
+/* return 1 when dirty updation started
+ * 0 other wise
+ */
+int32_t
+mq_update_dirty_inode (xlator_t *this,
+                       loc_t *loc,
+                       quota_inode_ctx_t *ctx,
+                       inode_contribution_t *contribution)
+{
+        int32_t          ret        = -1;
+        quota_local_t   *local      = NULL;
+        gf_boolean_t    status     = _gf_false;
+        struct gf_flock  lock       = {0, };
+        call_frame_t    *frame      = NULL;
+
+        ret = mq_get_ctx_updation_status (ctx, &status);
+        if (ret == -1 || status == _gf_true) {
+                ret = 0;
+                goto out;
+        }
+
+        frame = create_frame (this, this->ctx->pool);
+        if (frame == NULL) {
+                ret = -1;
+                goto out;
+        }
+
+        mq_assign_lk_owner (this, frame);
+
+        local = mq_local_new ();
+        if (local == NULL)
+                goto fr_destroy;
+
+        frame->local = local;
+        ret = mq_loc_copy (&local->loc, loc);
+        if (ret < 0)
+                goto fr_destroy;
+
+        local->ctx = ctx;
+
+        local->contri = contribution;
+
+        lock.l_type = F_WRLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start = 0;
+        lock.l_len = 0;
+
+        if (local->loc.inode == NULL) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Inode is NULL, so can't stackwind.");
+                goto fr_destroy;
+        }
+
+        STACK_WIND (frame,
+                    mq_get_dirty_xattr,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &local->loc, F_SETLKW, &lock, NULL);
+        return 1;
+
+fr_destroy:
+        QUOTA_STACK_DESTROY (frame, this);
+out:
+
+        return 0;
+}
+
+
+int32_t
+mq_inode_creation_done (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        quota_local_t *local = NULL;
+
+        if (frame == NULL)
+                return 0;
+
+        local = frame->local;
+
+        if (local != NULL) {
+                mq_initiate_quota_txn (this, &local->loc);
+        }
+
+        QUOTA_STACK_DESTROY (frame, this);
+
+        return 0;
+}
+
+
+int32_t
+mq_xattr_creation_release_lock (call_frame_t *frame, void *cookie,
+                                xlator_t *this, int32_t op_ret,
+                                int32_t op_errno, dict_t *xdata)
+{
+        struct gf_flock  lock  = {0, };
+        quota_local_t   *local = NULL;
+
+        local = frame->local;
+
+        lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+        STACK_WIND (frame,
+                    mq_inode_creation_done,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &local->loc,
+                    F_SETLKW, &lock, NULL);
+
+        return 0;
+}
+
+
+int32_t
+mq_create_dirty_xattr (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *dict,
+                       dict_t *xdata)
+{
+        int32_t          ret       = -1;
+        dict_t          *newdict   = NULL;
+        quota_local_t   *local     = NULL;
+
+        if (op_ret < 0) {
+                goto err;
+        }
+
+        local = frame->local;
+
+        if (local->loc.inode->ia_type == IA_IFDIR) {
+                newdict = dict_new ();
+                if (!newdict) {
+                        goto err;
+                }
+
+                ret = dict_set_int8 (newdict, QUOTA_DIRTY_KEY, 0);
+                if (ret == -1) {
+                        goto err;
+                }
+
+                uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+                GF_UUID_ASSERT (local->loc.gfid);
+
+                STACK_WIND (frame, mq_xattr_creation_release_lock,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->setxattr,
+                            &local->loc, newdict, 0, NULL);
+        } else {
+                mq_xattr_creation_release_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        ret = 0;
+
+err:
+        if (ret < 0) {
+                mq_xattr_creation_release_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (newdict != NULL)
+                dict_unref (newdict);
+
+        return 0;
+}
+
+
+int32_t
+mq_create_xattr (xlator_t *this, call_frame_t *frame)
+{
+        int32_t               ret       = 0;
+        int64_t              *value     = NULL;
+        int64_t              *size      = NULL;
+        dict_t               *dict      = NULL;
+        char                  key[512]  = {0, };
+        quota_local_t        *local     = NULL;
+        quota_inode_ctx_t    *ctx       = NULL;
+        inode_contribution_t *contri    = NULL;
+
+        if (frame == NULL || this == NULL)
+                return 0;
+
+        local = frame->local;
+
+        ret = mq_inode_ctx_get (local->loc.inode, this, &ctx);
+        if (ret < 0) {
+                ctx = mq_inode_ctx_new (local->loc.inode, this);
+                if (ctx == NULL) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "mq_inode_ctx_new failed");
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        dict = dict_new ();
+        if (!dict)
+                goto out;
+
+        if (local->loc.inode->ia_type == IA_IFDIR) {
+                QUOTA_ALLOC_OR_GOTO (size, int64_t, ret, err);
+                ret = dict_set_bin (dict, QUOTA_SIZE_KEY, size, 8);
+                if (ret < 0)
+                        goto free_size;
+        }
+
+        if ((local->loc.path && strcmp (local->loc.path, "/") != 0)
+            || (local->loc.inode && !uuid_is_null (local->loc.inode->gfid) &&
+                !__is_root_gfid (local->loc.inode->gfid))
+            || (!uuid_is_null (local->loc.gfid)
+                && !__is_root_gfid (local->loc.gfid))) {
+                contri = mq_add_new_contribution_node (this, ctx, &local->loc);
+                if (contri == NULL)
+                        goto err;
+
+                QUOTA_ALLOC_OR_GOTO (value, int64_t, ret, err);
+                GET_CONTRI_KEY (key, local->loc.parent->gfid, ret);
+
+                ret = dict_set_bin (dict, key, value, 8);
+                if (ret < 0)
+                        goto free_value;
+        }
+
+        GF_UUID_ASSERT (local->loc.gfid);
+
+        STACK_WIND (frame, mq_create_dirty_xattr, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->xattrop, &local->loc,
+                    GF_XATTROP_ADD_ARRAY64, dict, NULL);
+        ret = 0;
+
+free_size:
+        if (ret < 0) {
+                GF_FREE (size);
+        }
+
+free_value:
+        if (ret < 0) {
+                GF_FREE (value);
+        }
+
+err:
+        dict_unref (dict);
+
+out:
+        if (ret < 0) {
+                mq_xattr_creation_release_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        return 0;
+}
+
+
+int32_t
+mq_check_n_set_inode_xattr (call_frame_t *frame, void *cookie,
+                            xlator_t *this, int32_t op_ret, int32_t op_errno,
+                            inode_t *inode, struct iatt *buf, dict_t *dict,
+                            struct iatt *postparent)
+{
+        quota_local_t        *local           = NULL;
+        int64_t              *size            = NULL, *contri = NULL;
+        int8_t                dirty           = 0;
+        int32_t               ret             = 0;
+        char                  contri_key[512] = {0, };
+
+        if (op_ret < 0) {
+                goto out;
+        }
+
+        local = frame->local;
+
+        ret = dict_get_bin (dict, QUOTA_SIZE_KEY, (void **) &size);
+        if (ret < 0)
+                goto create_xattr;
+
+        ret = dict_get_int8 (dict, QUOTA_DIRTY_KEY, &dirty);
+        if (ret < 0)
+                goto create_xattr;
+
+        //check contribution xattr if not root
+        if ((local->loc.path && strcmp (local->loc.path, "/") != 0)
+            || (!uuid_is_null (local->loc.gfid)
+                && !__is_root_gfid (local->loc.gfid))
+            || (local->loc.inode
+                && !uuid_is_null (local->loc.inode->gfid)
+                && !__is_root_gfid (local->loc.inode->gfid))) {
+                GET_CONTRI_KEY (contri_key, local->loc.parent->gfid, ret);
+                if (ret < 0)
+                        goto out;
+
+                ret = dict_get_bin (dict, contri_key, (void **) &contri);
+                if (ret < 0)
+                        goto create_xattr;
+        }
+
+out:
+        mq_xattr_creation_release_lock (frame, NULL, this, 0, 0, NULL);
+        return 0;
+
+create_xattr:
+        if (uuid_is_null (local->loc.gfid)) {
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+        }
+
+        mq_create_xattr (this, frame);
+        return 0;
+}
+
+
+int32_t
+mq_get_xattr (call_frame_t *frame, void *cookie, xlator_t *this,
+              int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        dict_t        *xattr_req = NULL;
+        quota_local_t *local     = NULL;
+        int32_t        ret       = 0;
+
+        if (op_ret < 0) {
+                goto lock_err;
+        }
+
+        local = frame->local;
+
+        xattr_req = dict_new ();
+        if (xattr_req == NULL) {
+                goto err;
+        }
+
+        ret = mq_req_xattr (this, &local->loc, xattr_req);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING, "cannot request xattr");
+                goto err;
+        }
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+
+        STACK_WIND (frame, mq_check_n_set_inode_xattr, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, &local->loc, xattr_req);
+
+        dict_unref (xattr_req);
+
+        return 0;
+
+err:
+        mq_xattr_creation_release_lock (frame, NULL, this, 0, 0, NULL);
+
+        if (xattr_req)
+                dict_unref (xattr_req);
+        return 0;
+
+lock_err:
+        mq_inode_creation_done (frame, NULL, this, 0, 0, NULL);
+        return 0;
+}
+
+
+int32_t
+mq_set_inode_xattr (xlator_t *this, loc_t *loc)
+{
+        struct gf_flock  lock  = {0, };
+        quota_local_t   *local = NULL;
+        int32_t          ret   = 0;
+        call_frame_t    *frame = NULL;
+
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame) {
+                ret = -1;
+                goto err;
+        }
+
+        local = mq_local_new ();
+        if (local == NULL) {
+                goto err;
+        }
+
+        frame->local = local;
+
+        ret = loc_copy (&local->loc, loc);
+        if (ret < 0) {
+                goto err;
+        }
+
+        frame->local = local;
+
+        lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = F_WRLCK;
+        lock.l_whence = SEEK_SET;
+
+        STACK_WIND (frame,
+                    mq_get_xattr,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &local->loc, F_SETLKW, &lock, NULL);
+
+        return 0;
+
+err:
+        QUOTA_STACK_DESTROY (frame, this);
+
+        return 0;
+}
+
+
+int32_t
+mq_get_parent_inode_local (xlator_t *this, quota_local_t *local)
+{
+        int32_t            ret = -1;
+        quota_inode_ctx_t *ctx = NULL;
+        inode_contribution_t *contribution = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", local, out);
+
+        local->contri = NULL;
+
+        loc_wipe (&local->loc);
+
+        ret = mq_loc_copy (&local->loc, &local->parent_loc);
+        if (ret < 0) {
+                gf_log_callingfn (this->name, GF_LOG_WARNING,
+                        "loc copy failed");
+                goto out;
+        }
+
+        loc_wipe (&local->parent_loc);
+
+        ret = mq_inode_loc_fill (NULL, local->loc.parent,
+                                 &local->parent_loc);
+        if (ret < 0) {
+                gf_log_callingfn (this->name, GF_LOG_WARNING,
+                        "failed to build parent loc of %s",
+                        local->loc.path);
+                goto out;
+        }
+
+        ret = mq_inode_ctx_get (local->loc.inode, this, &ctx);
+        if (ret < 0) {
+                gf_log_callingfn (this->name, GF_LOG_WARNING,
+                                  "inode ctx get failed");
+                goto out;
+        }
+
+        local->ctx = ctx;
+
+        if (list_empty (&ctx->contribution_head)) {
+                gf_log_callingfn (this->name, GF_LOG_WARNING,
+                        "contribution node list is empty which "
+                        "is an error");
+                ret = -1;
+                goto out;
+        }
+
+        /* Earlier we used to get the next entry in the list maintained
+           by the context. In a good situation it works. i.e the next
+           parent in the directory hierarchy for this path is obtained.
+
+           But consider the below situation:
+           mount-point: /mnt/point
+           quota enabled directories within mount point: /a, /b, /c
+
+           Now when some file (file1) in the directory /c is written some data,
+           then to update the directories, marker has to get the contribution
+           object for the parent inode, i.e /c.
+           Beefore, it was being done by
+           local->contri = (inode_contribution_t *) ctx->contribution_head.next;
+           It works in the normal situations. But suppose /c is moved to /b.
+           Now /b's contribution object is added to the end of the list of
+           parents that the file file1 within /b/c is maintaining. Now if
+           the file /b/c/file1 is copied to /b/c/new, to update the parent in
+           the order c, b and / we cannot go to the next element in the list,
+           as in this case the next contribution object would be / and /b's
+           contribution will be at the end of the list. So get the proper
+           parent's contribution, by searching the entire list.
+        */
+        contribution = mq_get_contribution_node (local->loc.parent, ctx);
+        GF_ASSERT (contribution != NULL);
+        local->contri = contribution;
+
+        ret = 0;
+out:
+        return ret;
+}
+
+
+int32_t
+mq_xattr_updation_done (call_frame_t *frame,
+                        void *cookie,
+                        xlator_t *this,
+                        int32_t op_ret,
+                        int32_t op_errno,
+                        dict_t *dict, dict_t *xdata)
+{
+        QUOTA_STACK_DESTROY (frame, this);
+        return 0;
+}
+
+
+int32_t
+mq_inodelk_cbk (call_frame_t *frame, void *cookie,
+                xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        int32_t         ret    = 0;
+        gf_boolean_t    status = _gf_false;
+        quota_local_t  *local  = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1 || local->err) {
+                if (op_ret == -1) {
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "unlocking failed on path (%s)(%s)",
+                                local->parent_loc.path, strerror (op_errno));
+                }
+                mq_xattr_updation_done (frame, NULL, this, 0, 0, NULL, NULL);
+
+                return 0;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG,
+                "inodelk released on %s", local->parent_loc.path);
+
+        if ((strcmp (local->parent_loc.path, "/") == 0)
+            || (local->delta == 0)) {
+                mq_xattr_updation_done (frame, NULL, this, 0, 0, NULL, NULL);
+        } else {
+                ret = mq_get_parent_inode_local (this, local);
+                if (ret < 0) {
+                        mq_xattr_updation_done (frame, NULL, this, 0, 0, NULL,
+                                                NULL);
+                        goto out;
+                }
+                status = _gf_true;
+
+                ret = mq_test_and_set_ctx_updation_status (local->ctx, &status);
+                if (ret == 0 && status == _gf_false) {
+                        mq_get_lock_on_parent (frame, this);
+                } else {
+                        mq_xattr_updation_done (frame, NULL, this, 0, 0, NULL,
+                                                NULL);
+                }
+        }
+out:
+        return 0;
+}
+
+
+//now release lock on the parent inode
+int32_t
+mq_release_parent_lock (call_frame_t *frame, void *cookie,
+                        xlator_t *this, int32_t op_ret,
+                        int32_t op_errno, dict_t *xdata)
+{
+        int32_t            ret      = 0;
+        quota_local_t     *local    = NULL;
+        quota_inode_ctx_t *ctx      = NULL;
+        struct gf_flock    lock     = {0, };
+
+        local = frame->local;
+
+        if (local->err != 0) {
+                gf_log_callingfn (this->name,
+                                  (local->err == ENOENT) ? GF_LOG_DEBUG
+                                  : GF_LOG_WARNING,
+                                  "An operation during quota updation "
+                                  "of path (%s) failed (%s)", local->loc.path,
+                                  strerror (local->err));
+        }
+
+        ret = mq_inode_ctx_get (local->parent_loc.inode, this, &ctx);
+        if (ret < 0)
+                goto wind;
+
+        LOCK (&ctx->lock);
+        {
+                ctx->dirty = 0;
+        }
+        UNLOCK (&ctx->lock);
+
+        if (local->parent_loc.inode == NULL) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Invalid parent inode.");
+                goto err;
+        }
+
+wind:
+        lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+        STACK_WIND (frame,
+                    mq_inodelk_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &local->parent_loc,
+                    F_SETLKW, &lock, NULL);
+
+        return 0;
+err:
+        mq_xattr_updation_done (frame, NULL, this,
+                                0, 0 , NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+mq_mark_undirty (call_frame_t *frame,
+                 void *cookie,
+                 xlator_t *this,
+                 int32_t op_ret,
+                 int32_t op_errno,
+                 dict_t *dict, dict_t *xdata)
+{
+        int32_t            ret          = -1;
+        int64_t           *size         = NULL;
+        dict_t            *newdict      = NULL;
+        quota_local_t     *local        = NULL;
+        quota_inode_ctx_t *ctx          = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING, "%s occurred while"
+                        " updating the size of %s", strerror (op_errno),
+                        local->parent_loc.path);
+
+                goto err;
+        }
+
+        //update the size of the parent inode
+        if (dict != NULL) {
+                ret = mq_inode_ctx_get (local->parent_loc.inode, this, &ctx);
+                if (ret < 0) {
+                        op_errno = EINVAL;
+                        goto err;
+                }
+
+                ret = dict_get_bin (dict, QUOTA_SIZE_KEY, (void **) &size);
+                if (ret < 0) {
+                        op_errno = EINVAL;
+                        goto err;
+                }
+
+                LOCK (&ctx->lock);
+                {
+                        if (size)
+                                ctx->size = ntoh64 (*size);
+                        gf_log (this->name, GF_LOG_DEBUG, "%s %"PRId64,
+                                local->parent_loc.path, ctx->size);
+                }
+                UNLOCK (&ctx->lock);
+        }
+
+        newdict = dict_new ();
+        if (!newdict) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        ret = dict_set_int8 (newdict, QUOTA_DIRTY_KEY, 0);
+
+        if (ret == -1) {
+                op_errno = -ret;
+                goto err;
+        }
+
+        uuid_copy (local->parent_loc.gfid, local->parent_loc.inode->gfid);
+        GF_UUID_ASSERT (local->parent_loc.gfid);
+
+        STACK_WIND (frame, mq_release_parent_lock,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr,
+                    &local->parent_loc, newdict, 0, NULL);
+
+        ret = 0;
+err:
+        if (op_ret == -1 || ret == -1) {
+                local->err = op_errno;
+
+                mq_release_parent_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (newdict)
+                dict_unref (newdict);
+
+        return 0;
+}
+
+
+int32_t
+mq_update_parent_size (call_frame_t *frame,
+                       void *cookie,
+                       xlator_t *this,
+                       int32_t op_ret,
+                       int32_t op_errno,
+                       dict_t *dict, dict_t *xdata)
+{
+        int64_t             *size       = NULL;
+        int32_t              ret        = -1;
+        dict_t              *newdict    = NULL;
+        quota_local_t       *local      = NULL;
+        quota_inode_ctx_t   *ctx        = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1) {
+                gf_log (this->name, ((op_errno == ENOENT) ? GF_LOG_DEBUG :
+                                     GF_LOG_WARNING),
+                        "xattrop call failed: %s", strerror (op_errno));
+
+                goto err;
+        }
+
+        LOCK (&local->contri->lock);
+        {
+                local->contri->contribution += local->delta;
+        }
+        UNLOCK (&local->contri->lock);
+
+        gf_log_callingfn (this->name, GF_LOG_DEBUG, "path: %s size: %"PRId64
+                          " contribution:%"PRId64,
+                          local->loc.path, local->ctx->size,
+                          local->contri->contribution);
+
+        if (dict == NULL) {
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        ret = mq_inode_ctx_get (local->parent_loc.inode, this, &ctx);
+        if (ret < 0) {
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        newdict = dict_new ();
+        if (!newdict) {
+                op_errno = ENOMEM;
+                ret = -1;
+                goto err;
+        }
+
+        QUOTA_ALLOC_OR_GOTO (size, int64_t, ret, err);
+
+        *size = hton64 (local->delta);
+
+        ret = dict_set_bin (newdict, QUOTA_SIZE_KEY, size, 8);
+        if (ret < 0) {
+                op_errno = -ret;
+                goto err;
+        }
+
+        if (uuid_is_null (local->parent_loc.gfid))
+                        uuid_copy (local->parent_loc.gfid,
+                                   local->parent_loc.inode->gfid);
+        GF_UUID_ASSERT (local->parent_loc.gfid);
+
+        STACK_WIND (frame,
+                    mq_mark_undirty,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->xattrop,
+                    &local->parent_loc,
+                    GF_XATTROP_ADD_ARRAY64,
+                    newdict, NULL);
+        ret = 0;
+err:
+        if (op_ret == -1 || ret < 0) {
+                local->err = op_errno;
+                mq_release_parent_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (newdict)
+                dict_unref (newdict);
+
+        return 0;
+}
+
+int32_t
+mq_update_inode_contribution (call_frame_t *frame, void *cookie,
+                              xlator_t *this, int32_t op_ret,
+                              int32_t op_errno, inode_t *inode,
+                              struct iatt *buf, dict_t *dict,
+                              struct iatt *postparent)
+{
+        int32_t               ret              = -1;
+        int64_t              *size             = NULL, size_int = 0, contri_int = 0;
+        int64_t              *contri           = NULL;
+        int64_t              *delta            = NULL;
+        char                  contri_key [512] = {0, };
+        dict_t               *newdict          = NULL;
+        quota_local_t        *local            = NULL;
+        quota_inode_ctx_t    *ctx              = NULL;
+        inode_contribution_t *contribution     = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1) {
+                gf_log (this->name, ((op_errno == ENOENT) ? GF_LOG_DEBUG :
+                                     GF_LOG_WARNING),
+                        "failed to get size and contribution of path (%s)(%s)",
+                        local->loc.path, strerror (op_errno));
+                goto err;
+        }
+
+        ctx = local->ctx;
+        contribution = local->contri;
+
+        //prepare to update size & contribution of the inode
+        GET_CONTRI_KEY (contri_key, contribution->gfid, ret);
+        if (ret == -1) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                if (local->loc.inode->ia_type == IA_IFDIR ) {
+                        ret = dict_get_bin (dict, QUOTA_SIZE_KEY,
+                                            (void **) &size);
+                        if (ret < 0) {
+                                op_errno = EINVAL;
+                                goto unlock;
+                        }
+
+                        ctx->size = ntoh64 (*size);
+                } else
+                        ctx->size = buf->ia_blocks * 512;
+
+                size_int = ctx->size;
+        }
+unlock:
+        UNLOCK  (&ctx->lock);
+
+        if (ret < 0) {
+                goto err;
+        }
+
+        ret = dict_get_bin (dict, contri_key, (void **) &contri);
+
+        LOCK (&contribution->lock);
+        {
+                if (ret < 0)
+                        contribution->contribution = 0;
+                else
+                        contribution->contribution = ntoh64 (*contri);
+
+                contri_int = contribution->contribution;
+        }
+        UNLOCK (&contribution->lock);
+
+        gf_log (this->name, GF_LOG_DEBUG, "%s %"PRId64 " %"PRId64,
+                local->loc.path, size_int, contri_int);
+
+        local->delta = size_int - contri_int;
+
+        if (local->delta == 0) {
+                mq_mark_undirty (frame, NULL, this, 0, 0, NULL, NULL);
+                return 0;
+        }
+
+        newdict = dict_new ();
+        if (newdict == NULL) {
+                op_errno = ENOMEM;
+                ret = -1;
+                goto err;
+        }
+
+        QUOTA_ALLOC_OR_GOTO (delta, int64_t, ret, err);
+
+        *delta = hton64 (local->delta);
+
+        ret = dict_set_bin (newdict, contri_key, delta, 8);
+        if (ret < 0) {
+                op_errno = -ret;
+                ret = -1;
+                goto err;
+        }
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+
+        STACK_WIND (frame,
+                    mq_update_parent_size,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->xattrop,
+                    &local->loc,
+                    GF_XATTROP_ADD_ARRAY64,
+                    newdict, NULL);
+        ret = 0;
+
+err:
+        if (op_ret == -1 || ret < 0) {
+                local->err = op_errno;
+
+                mq_release_parent_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (newdict)
+                dict_unref (newdict);
+
+        return 0;
+}
+
+int32_t
+mq_fetch_child_size_and_contri (call_frame_t *frame, void *cookie,
+                                xlator_t *this, int32_t op_ret,
+                                int32_t op_errno, dict_t *xdata)
+{
+        int32_t            ret              = -1;
+        char               contri_key [512] = {0, };
+        dict_t            *newdict          = NULL;
+        quota_local_t     *local            = NULL;
+        quota_inode_ctx_t *ctx              = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1) {
+                gf_log (this->name, (op_errno == ENOENT) ? GF_LOG_DEBUG
+                        : GF_LOG_WARNING,
+                        "couldnt mark inode corresponding to path (%s) dirty "
+                        "(%s)", local->parent_loc.path, strerror (op_errno));
+                goto err;
+        }
+
+        VALIDATE_OR_GOTO (local->ctx, err);
+        VALIDATE_OR_GOTO (local->contri, err);
+
+        gf_log (this->name, GF_LOG_DEBUG, "%s marked dirty",
+                local->parent_loc.path);
+
+        //update parent ctx
+        ret = mq_inode_ctx_get (local->parent_loc.inode, this, &ctx);
+        if (ret == -1) {
+                op_errno = EINVAL;
+                goto err;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->dirty = 1;
+        }
+        UNLOCK (&ctx->lock);
+
+        newdict = dict_new ();
+        if (newdict == NULL) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        if (local->loc.inode->ia_type == IA_IFDIR) {
+                ret = dict_set_int64 (newdict, QUOTA_SIZE_KEY, 0);
+                if (ret < 0) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "dict_set failed.");
+                        goto err;
+                }
+        }
+
+        GET_CONTRI_KEY (contri_key, local->contri->gfid, ret);
+        if (ret < 0) {
+                op_errno = ENOMEM;
+                goto err;
+        }
+
+        ret = dict_set_int64 (newdict, contri_key, 0);
+        if (ret < 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "dict_set failed.");
+                goto err;
+        }
+
+        mq_set_ctx_updation_status (local->ctx, _gf_false);
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+
+        STACK_WIND (frame, mq_update_inode_contribution, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, &local->loc, newdict);
+
+        ret = 0;
+
+err:
+        if ((op_ret == -1) || (ret < 0)) {
+                local->err = op_errno;
+
+                mq_set_ctx_updation_status (local->ctx, _gf_false);
+
+                mq_release_parent_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (newdict)
+                dict_unref (newdict);
+
+        return 0;
+}
+
+int32_t
+mq_markdirty (call_frame_t *frame, void *cookie,
+              xlator_t *this, int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        int32_t        ret      = -1;
+        dict_t        *dict     = NULL;
+        quota_local_t *local    = NULL;
+
+        local = frame->local;
+
+        if (op_ret == -1){
+                gf_log (this->name, (op_errno == ENOENT) ? GF_LOG_DEBUG
+                        : GF_LOG_WARNING, "acquiring locks failed on %s (%s)",
+                        local->parent_loc.path, strerror (op_errno));
+
+                local->err = op_errno;
+
+                mq_set_ctx_updation_status (local->ctx, _gf_false);
+
+                mq_inodelk_cbk (frame, NULL, this, 0, 0, NULL);
+
+                return 0;
+        }
+
+        gf_log (this->name, GF_LOG_TRACE,
+                "inodelk succeeded on  %s", local->parent_loc.path);
+
+        dict = dict_new ();
+        if (!dict) {
+                ret = -1;
+                goto err;
+        }
+
+        ret = dict_set_int8 (dict, QUOTA_DIRTY_KEY, 1);
+        if (ret == -1)
+                goto err;
+
+        uuid_copy (local->parent_loc.gfid,
+                   local->parent_loc.inode->gfid);
+        GF_UUID_ASSERT (local->parent_loc.gfid);
+
+        STACK_WIND (frame, mq_fetch_child_size_and_contri,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr,
+                    &local->parent_loc, dict, 0, NULL);
+
+        ret = 0;
+err:
+        if (ret == -1) {
+                local->err = 1;
+
+                mq_set_ctx_updation_status (local->ctx, _gf_false);
+
+                mq_release_parent_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        if (dict)
+                dict_unref (dict);
+
+        return 0;
+}
+
+
+int32_t
+mq_get_lock_on_parent (call_frame_t *frame, xlator_t *this)
+{
+        struct gf_flock  lock  = {0, };
+        quota_local_t   *local = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", frame, fr_destroy);
+
+        local = frame->local;
+        gf_log (this->name, GF_LOG_DEBUG, "taking lock on %s",
+                local->parent_loc.path);
+
+        if (local->parent_loc.inode == NULL) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "parent inode is not valid, aborting "
+                        "transaction.");
+                goto fr_destroy;
+        }
+
+        lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = F_WRLCK;
+        lock.l_whence = SEEK_SET;
+
+        STACK_WIND (frame,
+                    mq_markdirty,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &local->parent_loc, F_SETLKW, &lock, NULL);
+
+        return 0;
+
+fr_destroy:
+        QUOTA_STACK_DESTROY (frame, this);
+
+        return -1;
+}
+
+int
+mq_prepare_txn_frame (xlator_t *this, loc_t *loc,
+                      quota_inode_ctx_t *ctx,
+                      inode_contribution_t *contri,
+                      call_frame_t **new_frame)
+{
+        call_frame_t    *frame = NULL;
+        int              ret   = -1;
+        quota_local_t   *local = NULL;
+
+        if (!this || !loc || !new_frame)
+                goto err;
+
+        frame = create_frame (this, this->ctx->pool);
+        if (frame == NULL)
+                goto err;
+
+        mq_assign_lk_owner (this, frame);
+
+        local = mq_local_new ();
+        if (local == NULL)
+                goto fr_destroy;
+
+        frame->local = local;
+
+        ret = mq_loc_copy (&local->loc, loc);
+        if (ret < 0)
+                goto fr_destroy;
+
+        ret = mq_inode_loc_fill (NULL, local->loc.parent,
+                                 &local->parent_loc);
+        if (ret < 0)
+                goto fr_destroy;
+
+        local->ctx = ctx;
+        local->contri = contri;
+
+        ret = 0;
+        *new_frame = frame;
+
+        return ret;
+
+fr_destroy:
+        QUOTA_STACK_DESTROY (frame, this);
+err:
+        return ret;
+}
+
+int
+mq_start_quota_txn (xlator_t *this, loc_t *loc,
+                    quota_inode_ctx_t *ctx,
+                    inode_contribution_t *contri)
+{
+        int32_t        ret      = -1;
+        call_frame_t  *frame    = NULL;
+
+        ret = mq_prepare_txn_frame (this, loc, ctx,
+                                    contri, &frame);
+        if (ret)
+                goto err;
+
+        ret = mq_get_lock_on_parent (frame, this);
+        if (ret == -1)
+                goto err;
+
+        return 0;
+
+err:
+        mq_set_ctx_updation_status (ctx, _gf_false);
+
+        return -1;
+}
+
+
+int
+mq_initiate_quota_txn (xlator_t *this, loc_t *loc)
+{
+        int32_t               ret          = -1;
+        gf_boolean_t          status       = _gf_false;
+        quota_inode_ctx_t    *ctx          = NULL;
+        inode_contribution_t *contribution = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", loc, out);
+        GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
+
+        ret = mq_inode_ctx_get (loc->inode, this, &ctx);
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "inode ctx get failed, aborting quota txn");
+                ret = -1;
+                goto out;
+        }
+
+        /* Create the contribution node if its absent. Is it right to
+           assume that if the contribution node is not there, then
+           create one and proceed instead of returning?
+           Reason for this assumption is for hard links. Suppose
+           hard link for a file f1 present in a directory d1 is
+           created in the directory d2 (as f2). Now, since d2's
+           contribution is not there in f1's inode ctx, d2's
+           contribution xattr wont be created and will create problems
+           for quota operations.
+        */
+        contribution = mq_get_contribution_node (loc->parent, ctx);
+        if (!contribution) {
+                if ((loc->path && strcmp (loc->path, "/"))
+                    || (!uuid_is_null (loc->gfid)
+                        && !__is_root_gfid (loc->gfid))
+                    || (loc->inode && !uuid_is_null (loc->inode->gfid)
+                        && !__is_root_gfid (loc->inode->gfid)))
+                        gf_log_callingfn (this->name, GF_LOG_TRACE,
+                                          "contribution node for the "
+                                          "path (%s) with parent (%s) "
+                                          "not found", loc->path,
+                                          loc->parent?
+                                          uuid_utoa (loc->parent->gfid):
+                                          NULL);
+
+                contribution = mq_add_new_contribution_node (this, ctx, loc);
+                if (!contribution) {
+                        if(loc->path && strcmp (loc->path, "/"))
+                                gf_log_callingfn (this->name, GF_LOG_WARNING,
+                                                  "could not allocate "
+                                                  " contribution node for (%s) "
+                                                  "parent: (%s)", loc->path,
+                                                  loc->parent?
+                                                  uuid_utoa (loc->parent->gfid):
+                                                  NULL);
+                        goto out;
+                }
+        }
+
+        /* To improve performance, do not start another transaction
+         * if one is already in progress for same inode
+         */
+        status = _gf_true;
+
+        ret = mq_test_and_set_ctx_updation_status (ctx, &status);
+        if (ret < 0)
+                goto out;
+
+        if (status == _gf_false) {
+                mq_start_quota_txn (this, loc, ctx, contribution);
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+
+
+
+
+int32_t
+mq_inspect_directory_xattr (xlator_t *this,
+                            loc_t *loc,
+                            dict_t *dict,
+                            struct iatt buf)
+{
+        int32_t               ret                 = 0;
+        int8_t                dirty               = -1;
+        int64_t              *size                = NULL, size_int = 0;
+        int64_t              *contri              = NULL, contri_int = 0;
+        char                  contri_key [512]    = {0, };
+        gf_boolean_t          not_root            = _gf_false;
+        quota_inode_ctx_t    *ctx                 = NULL;
+        inode_contribution_t *contribution        = NULL;
+
+        ret = mq_inode_ctx_get (loc->inode, this, &ctx);
+        if (ret < 0) {
+                ctx = mq_inode_ctx_new (loc->inode, this);
+                if (ctx == NULL) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "mq_inode_ctx_new failed");
+                        ret = -1;
+                        goto err;
+                }
+        }
+
+        if (!loc->path || (loc->path && strcmp (loc->path, "/") != 0)) {
+                contribution = mq_add_new_contribution_node (this, ctx, loc);
+                if (contribution == NULL) {
+                        if (!uuid_is_null (loc->inode->gfid))
+                                gf_log (this->name, GF_LOG_DEBUG,
+                                        "cannot add a new contribution node "
+                                        "(%s)", uuid_utoa (loc->inode->gfid));
+                        ret = -1;
+                        goto err;
+                }
+        }
+
+        ret = dict_get_bin (dict, QUOTA_SIZE_KEY, (void **) &size);
+        if (ret < 0)
+                goto out;
+
+        ret = dict_get_int8 (dict, QUOTA_DIRTY_KEY, &dirty);
+        if (ret < 0)
+                goto out;
+
+        if ((loc->path && strcmp (loc->path, "/") != 0)
+            || (!uuid_is_null (loc->gfid) && !__is_root_gfid (loc->gfid))
+            || (loc->inode && !uuid_is_null (loc->inode->gfid) &&
+                !__is_root_gfid (loc->inode->gfid))) {
+                not_root = _gf_true;
+
+                GET_CONTRI_KEY (contri_key, contribution->gfid, ret);
+                if (ret < 0)
+                        goto out;
+
+                ret = dict_get_bin (dict, contri_key, (void **) &contri);
+                if (ret < 0)
+                        goto out;
+
+                LOCK (&contribution->lock);
+                {
+                        contribution->contribution = ntoh64 (*contri);
+                        contri_int = contribution->contribution;
+                }
+                UNLOCK (&contribution->lock);
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->size = ntoh64 (*size);
+                ctx->dirty = dirty;
+                size_int = ctx->size;
+        }
+        UNLOCK (&ctx->lock);
+
+        gf_log (this->name, GF_LOG_DEBUG, "size=%"PRId64
+                " contri=%"PRId64, size_int, contri_int);
+
+        if (dirty) {
+                ret = mq_update_dirty_inode (this, loc, ctx, contribution);
+        }
+
+        if ((!dirty || ret == 0) && (not_root == _gf_true) &&
+            (size_int != contri_int)) {
+                mq_initiate_quota_txn (this, loc);
+        }
+
+        ret = 0;
+out:
+        if (ret)
+                mq_set_inode_xattr (this, loc);
+err:
+        return ret;
+}
+
+int32_t
+mq_inspect_file_xattr (xlator_t *this,
+                       loc_t *loc,
+                       dict_t *dict,
+                       struct iatt buf)
+{
+        int32_t               ret              = -1;
+        uint64_t              contri_int       = 0, size = 0;
+        int64_t              *contri_ptr       = NULL;
+        char                  contri_key [512] = {0, };
+        quota_inode_ctx_t    *ctx              = NULL;
+        inode_contribution_t *contribution     = NULL;
+
+        ret = mq_inode_ctx_get (loc->inode, this, &ctx);
+        if (ret < 0) {
+                ctx = mq_inode_ctx_new (loc->inode, this);
+                if (ctx == NULL) {
+                        gf_log (this->name, GF_LOG_WARNING,
+                                "mq_inode_ctx_new failed");
+                        ret = -1;
+                        goto out;
+                }
+        }
+
+        contribution = mq_add_new_contribution_node (this, ctx, loc);
+        if (contribution == NULL) {
+                gf_log_callingfn (this->name, GF_LOG_DEBUG, "cannot allocate "
+                                  "contribution node (path:%s)", loc->path);
+                goto out;
+        }
+
+        LOCK (&ctx->lock);
+        {
+                ctx->size = 512 * buf.ia_blocks;
+                size = ctx->size;
+        }
+        UNLOCK (&ctx->lock);
+
+        list_for_each_entry (contribution, &ctx->contribution_head,
+                             contri_list) {
+                GET_CONTRI_KEY (contri_key, contribution->gfid, ret);
+                if (ret < 0)
+                        continue;
+
+                ret = dict_get_bin (dict, contri_key, (void **) &contri_int);
+                if (ret == 0) {
+                        contri_ptr = (int64_t *)(unsigned long)contri_int;
+
+                        LOCK (&contribution->lock);
+                        {
+                                contribution->contribution = ntoh64 (*contri_ptr);
+                                contri_int = contribution->contribution;
+                        }
+                        UNLOCK (&contribution->lock);
+
+                        gf_log (this->name, GF_LOG_DEBUG,
+                                "size=%"PRId64 " contri=%"PRId64, size, contri_int);
+
+                        if (size != contri_int) {
+                                mq_initiate_quota_txn (this, loc);
+                        }
+                } else {
+                        if (size)
+                                mq_initiate_quota_txn (this, loc);
+                        else
+                                mq_set_inode_xattr (this, loc);
+                }
+        }
+
+out:
+        return ret;
+}
+
+int32_t
+mq_xattr_state (xlator_t *this,
+                loc_t *loc,
+                dict_t *dict,
+                struct iatt buf)
+{
+        if (((buf.ia_type == IA_IFREG) && !dht_is_linkfile (&buf, dict))
+            || (buf.ia_type == IA_IFLNK))  {
+                mq_inspect_file_xattr (this, loc, dict, buf);
+        } else if (buf.ia_type == IA_IFDIR)
+                mq_inspect_directory_xattr (this, loc, dict, buf);
+
+        return 0;
+}
+
+int32_t
+mq_req_xattr (xlator_t *this,
+              loc_t *loc,
+              dict_t *dict)
+{
+        int32_t               ret       = -1;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", dict, out);
+
+        if (!loc)
+                goto set_size;
+
+        //if not "/" then request contribution
+        if (loc->path && strcmp (loc->path, "/") == 0)
+                goto set_size;
+
+        ret = mq_dict_set_contribution (this, dict, loc);
+        if (ret == -1)
+                goto out;
+
+set_size:
+        ret = dict_set_uint64 (dict, QUOTA_SIZE_KEY, 0);
+        if (ret < 0) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = dict_set_int8 (dict, QUOTA_DIRTY_KEY, 0);
+        if (ret < 0) {
+                ret = -1;
+                goto out;
+        }
+
+        ret = 0;
+
+out:
+        return ret;
+}
+
+
+int32_t
+mq_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        QUOTA_STACK_DESTROY (frame, this);
+
+        return 0;
+}
+
+int32_t
+_mq_inode_remove_done (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        int32_t        ret                = 0;
+        char           contri_key [512]   = {0, };
+        quota_local_t *local              = NULL;
+        inode_t       *inode              = NULL;
+        dentry_t      *tmp                = NULL;
+        gf_boolean_t  last_dentry         = _gf_true;
+        loc_t         loc                 = {0, };
+        dentry_t      *other_dentry       = NULL;
+        gf_boolean_t  remove              = _gf_false;
+
+        local = (quota_local_t *) frame->local;
+
+        if (op_ret == -1 || local->err == -1) {
+                mq_removexattr_cbk (frame, NULL, this, -1, 0, NULL);
+                return 0;
+        }
+
+        frame->local = NULL;
+
+        GET_CONTRI_KEY (contri_key, local->contri->gfid, ret);
+
+        if (!local->loc.inode)
+                inode = inode_grep (local->loc.parent->table, local->loc.parent,
+                                    local->loc.name);
+        else
+                inode = inode_ref (local->loc.inode);
+
+        /* Suppose there are 2 directories dir1 and dir2. Quota limit is set on
+           both the directories. There is a file (f1) in dir1. A hark link is
+           created for that file inside the directory dir2 (say f2). Now one
+           more xattr is set in the inode as a new hard link is created in a
+           separate directory.
+           i.e trusted.glusterfs.quota.<gfid of dir2>.contri=<contribution>
+
+           Now when the hardlink f2 is removed, then the new xattr added (i.e
+           the xattr indicating its contribution to ITS parent directory) should
+           be removed (IFF there is not another hardlink for that file in the
+           same directory).
+
+           To do that upon getting unlink first check whether any other hard
+           links for the same inode exists in the same directory. If so do not
+           do anything and proceed for quota transaction.
+           Otherwise, if the removed entry was the only link for that inode
+           within that directory, then get another dentry for the inode
+           (by traversing the list of dentries for the inode) and using the
+           the dentry's parent and name, send removexattr so that the xattr
+           is removed.
+
+           If it is not done, then if the volume is restarted or the brick
+           process is restarted, then wrong quota usage will be shown for the
+           directory dir2.
+        */
+        if (inode) {
+                tmp = NULL;
+                list_for_each_entry (tmp, &inode->dentry_list, inode_list) {
+                        if (local->loc.parent == tmp->parent) {
+                                if (strcmp (local->loc.name, local->loc.name)) {
+                                        last_dentry = _gf_false;
+                                        break;
+                                }
+                        }
+                }
+                remove = last_dentry;
+        }
+
+        if (remove) {
+                if (!other_dentry)    {
+                        list_for_each_entry (tmp, &inode->dentry_list,
+                                             inode_list) {
+                                if (local->loc.parent != tmp->parent) {
+                                        other_dentry = tmp;
+                                        break;
+                                }
+                        }
+                }
+
+                if (!other_dentry)
+                        mq_removexattr_cbk (frame, NULL, this, 0, 0, NULL);
+                else {
+                        loc.parent = inode_ref (other_dentry->parent);
+                        loc.name = gf_strdup (other_dentry->name);
+                        uuid_copy (loc.pargfid , other_dentry->parent->gfid);
+                        loc.inode = inode_ref (inode);
+                        uuid_copy (loc.gfid, inode->gfid);
+                        inode_path (other_dentry->parent, other_dentry->name,
+                                    (char **)&loc.path);
+
+                        STACK_WIND (frame, mq_removexattr_cbk,
+                                    FIRST_CHILD(this),
+                                    FIRST_CHILD(this)->fops->removexattr,
+                                    &loc, contri_key, NULL);
+                }
+        } else
+                mq_removexattr_cbk (frame, NULL, this, 0, 0, NULL);
+
+        ret = 0;
+
+        if (strcmp (local->parent_loc.path, "/") != 0) {
+                ret = mq_get_parent_inode_local (this, local);
+                if (ret < 0)
+                        goto out;
+
+                mq_start_quota_txn (this, &local->loc, local->ctx, local->contri);
+        }
+out:
+        mq_local_unref (this, local);
+
+        loc_wipe (&loc);
+        inode_unref (inode);
+        return 0;
+}
+
+int32_t
+mq_inode_remove_done (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *dict,
+                      dict_t *xdata)
+{
+        int32_t            ret   = -1;
+        struct gf_flock    lock  = {0, };
+        quota_inode_ctx_t *ctx   = NULL;
+        quota_local_t     *local = NULL;
+        int64_t contribution = 0;
+
+        local = frame->local;
+        if (op_ret == -1)
+                local->err = -1;
+
+        ret = mq_inode_ctx_get (local->parent_loc.inode, this, &ctx);
+
+        LOCK (&local->contri->lock);
+        {
+                contribution = local->contri->contribution;
+        }
+        UNLOCK (&local->contri->lock);
+
+        if (contribution == local->size) {
+                if (ret == 0) {
+                        LOCK (&ctx->lock);
+                        {
+                                ctx->size -= contribution;
+                        }
+                        UNLOCK (&ctx->lock);
+
+                        LOCK (&local->contri->lock);
+                        {
+                                local->contri->contribution = 0;
+                        }
+                        UNLOCK (&local->contri->lock);
+                }
+        }
+
+        lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+        STACK_WIND (frame,
+                    _mq_inode_remove_done,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &local->parent_loc,
+                    F_SETLKW, &lock, NULL);
+        return 0;
+}
+
+int32_t
+mq_reduce_parent_size_xattr (call_frame_t *frame, void *cookie, xlator_t *this,
+                             int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        int32_t                  ret               = -1;
+        int64_t                 *size              = NULL;
+        dict_t                  *dict              = NULL;
+        quota_local_t           *local             = NULL;
+
+        local = frame->local;
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "inodelk set failed on %s", local->parent_loc.path);
+                QUOTA_STACK_DESTROY (frame, this);
+                return 0;
+        }
+
+        VALIDATE_OR_GOTO (local->contri, err);
+
+        dict = dict_new ();
+        if (dict == NULL) {
+                ret = -1;
+                goto err;
+        }
+
+        QUOTA_ALLOC_OR_GOTO (size, int64_t, ret, err);
+
+        *size = hton64 (-local->size);
+
+        ret = dict_set_bin (dict, QUOTA_SIZE_KEY, size, 8);
+        if (ret < 0)
+                goto err;
+
+        uuid_copy (local->parent_loc.gfid,
+                   local->parent_loc.inode->gfid);
+        GF_UUID_ASSERT (local->parent_loc.gfid);
+
+        STACK_WIND (frame, mq_inode_remove_done, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->xattrop, &local->parent_loc,
+                    GF_XATTROP_ADD_ARRAY64, dict, NULL);
+        dict_unref (dict);
+        return 0;
+
+err:
+        local->err = 1;
+        mq_inode_remove_done (frame, NULL, this, -1, 0, NULL, NULL);
+        if (dict)
+                dict_unref (dict);
+        return 0;
+}
+
+int32_t
+mq_reduce_parent_size (xlator_t *this, loc_t *loc, int64_t contri)
+{
+        int32_t                  ret           = -1;
+        struct gf_flock          lock          = {0,};
+        call_frame_t            *frame         = NULL;
+        quota_local_t           *local         = NULL;
+        quota_inode_ctx_t       *ctx           = NULL;
+        inode_contribution_t    *contribution  = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", loc, out);
+
+        ret = mq_inode_ctx_get (loc->inode, this, &ctx);
+        if (ret < 0)
+                goto out;
+
+        contribution = mq_get_contribution_node (loc->parent, ctx);
+        if (contribution == NULL) {
+                gf_log_callingfn (this->name, GF_LOG_WARNING, "contribution for"
+                                  " the node %s is NULL", loc->path);
+                goto out;
+        }
+
+        local = mq_local_new ();
+        if (local == NULL) {
+                ret = -1;
+                goto out;
+        }
+
+        if (contri >= 0) {
+                local->size = contri;
+        } else {
+                LOCK (&contribution->lock);
+                {
+                        local->size = contribution->contribution;
+                }
+                UNLOCK (&contribution->lock);
+        }
+
+        if (local->size == 0) {
+                gf_log_callingfn (this->name, GF_LOG_TRACE,
+                                  "local->size is 0 " "path: (%s)", loc->path);
+                ret = 0;
+                goto out;
+        }
+
+        ret = mq_loc_copy (&local->loc, loc);
+        if (ret < 0)
+                goto out;
+
+        local->ctx = ctx;
+        local->contri = contribution;
+
+        ret = mq_inode_loc_fill (NULL, loc->parent, &local->parent_loc);
+        if (ret < 0) {
+                gf_log_callingfn (this->name, GF_LOG_INFO, "building parent loc"
+                                  " failed. (gfid: %s)",
+                                  uuid_utoa (loc->parent->gfid));
+                goto out;
+        }
+
+        frame = create_frame (this, this->ctx->pool);
+        if (!frame) {
+                ret = -1;
+                goto out;
+        }
+
+        mq_assign_lk_owner (this, frame);
+
+        frame->local = local;
+
+        lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = F_WRLCK;
+        lock.l_whence = SEEK_SET;
+
+        if (local->parent_loc.inode == NULL) {
+                ret = -1;
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Inode is NULL, so can't stackwind.");
+                goto out;
+        }
+
+        STACK_WIND (frame,
+                    mq_reduce_parent_size_xattr,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &local->parent_loc, F_SETLKW, &lock, NULL);
+        local = NULL;
+        ret = 0;
+
+out:
+        if (local != NULL)
+                mq_local_unref (this, local);
+
+        return ret;
+}
+
+
+int32_t
+init_quota_priv (xlator_t *this)
+{
+        return 0;
+}
+
+
+int32_t
+mq_rename_update_newpath (xlator_t *this, loc_t *loc)
+{
+        int32_t               ret          = -1;
+        quota_inode_ctx_t    *ctx          = NULL;
+        inode_contribution_t *contribution = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", loc, out);
+        GF_VALIDATE_OR_GOTO ("marker", loc->inode, out);
+
+        ret = mq_inode_ctx_get (loc->inode, this, &ctx);
+        if (ret < 0)
+                goto out;
+
+        contribution = mq_add_new_contribution_node (this, ctx, loc);
+        if (contribution == NULL) {
+                ret = -1;
+                goto out;
+        }
+
+        mq_initiate_quota_txn (this, loc);
+out:
+        return ret;
+}
+
+int32_t
+mq_forget (xlator_t *this, quota_inode_ctx_t *ctx)
+{
+        inode_contribution_t *contri = NULL;
+        inode_contribution_t *next   = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO ("marker", ctx, out);
+
+        list_for_each_entry_safe (contri, next, &ctx->contribution_head,
+                                  contri_list) {
+                list_del (&contri->contri_list);
+                GF_FREE (contri);
+        }
+
+        LOCK_DESTROY (&ctx->lock);
+        GF_FREE (ctx);
+out:
+        return 0;
+}
diff --git a/xlators/features/marker/src/marker-quota.h b/xlators/features/marker/src/marker-quota.h
new file mode 100644
index 000000000..42def9d22
--- /dev/null
+++ b/xlators/features/marker/src/marker-quota.h
@@ -0,0 +1,135 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _MARKER_QUOTA_H
+#define _MARKER_QUOTA_H
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "marker-mem-types.h"
+
+#define QUOTA_XATTR_PREFIX "trusted.glusterfs"
+#define QUOTA_DIRTY_KEY "trusted.glusterfs.quota.dirty"
+
+#define CONTRIBUTION "contri"
+#define CONTRI_KEY_MAX 512
+#define READDIR_BUF 4096
+
+
+#define QUOTA_STACK_DESTROY(_frame, _this)              \
+        do {                                            \
+                quota_local_t *_local = NULL;           \
+                _local = _frame->local;                 \
+                _frame->local = NULL;                   \
+                STACK_DESTROY (_frame->root);           \
+                mq_local_unref (_this, _local);         \
+        } while (0)
+
+
+#define QUOTA_ALLOC(var, type, ret)                     \
+        do {                                            \
+                ret = 0;                                \
+                var = GF_CALLOC (sizeof (type), 1,      \
+                                 gf_marker_mt_##type);  \
+                if (!var) {                             \
+                        ret = -1;                       \
+                }                                       \
+        } while (0);
+
+#define QUOTA_ALLOC_OR_GOTO(var, type, ret, label)      \
+        do {                                            \
+                var = GF_CALLOC (sizeof (type), 1,      \
+                                 gf_marker_mt_##type);  \
+                if (!var) {                             \
+                        gf_log ("", GF_LOG_ERROR,       \
+                                "out of memory");       \
+                        ret = -1;                       \
+                        goto label;                     \
+                }                                       \
+                ret = 0;                                \
+        } while (0);
+
+#define GET_CONTRI_KEY(var, _gfid, _ret)              \
+        do {                                          \
+                if (_gfid != NULL) {                  \
+                        char _gfid_unparsed[40];      \
+                        uuid_unparse (_gfid, _gfid_unparsed);           \
+                        _ret = snprintf (var, CONTRI_KEY_MAX,           \
+                                         QUOTA_XATTR_PREFIX             \
+                                         ".%s.%s." CONTRIBUTION, "quota", \
+                                         _gfid_unparsed);               \
+                } else {                                                \
+                        _ret = snprintf (var, CONTRI_KEY_MAX,           \
+                                         QUOTA_XATTR_PREFIX             \
+                                         ".%s.." CONTRIBUTION, "quota"); \
+                }                                                       \
+        } while (0);
+
+#define QUOTA_SAFE_INCREMENT(lock, var)         \
+        do {                                    \
+                LOCK (lock);                    \
+                var ++;                         \
+                UNLOCK (lock);                  \
+        } while (0)
+
+struct quota_inode_ctx {
+        int64_t                size;
+        int8_t                 dirty;
+        gf_boolean_t           updation_status;
+        gf_lock_t              lock;
+        struct list_head       contribution_head;
+};
+typedef struct quota_inode_ctx quota_inode_ctx_t;
+
+struct inode_contribution {
+        struct list_head contri_list;
+        int64_t          contribution;
+        uuid_t           gfid;
+  gf_lock_t lock;
+};
+typedef struct inode_contribution inode_contribution_t;
+
+int32_t
+mq_get_lock_on_parent (call_frame_t *, xlator_t *);
+
+int32_t
+mq_req_xattr (xlator_t *, loc_t *, dict_t *);
+
+int32_t
+init_quota_priv (xlator_t *);
+
+int32_t
+mq_xattr_state (xlator_t *, loc_t *, dict_t *, struct iatt);
+
+int32_t
+mq_set_inode_xattr (xlator_t *, loc_t *);
+
+int
+mq_initiate_quota_txn (xlator_t *, loc_t *);
+
+int32_t
+mq_dirty_inode_readdir (call_frame_t *, void *, xlator_t *,
+                        int32_t, int32_t, fd_t *, dict_t *);
+
+int32_t
+mq_reduce_parent_size (xlator_t *, loc_t *, int64_t);
+
+int32_t
+mq_rename_update_newpath (xlator_t *, loc_t *);
+
+int32_t
+mq_inspect_file_xattr (xlator_t *this, loc_t *loc, dict_t *dict, struct iatt buf);
+
+int32_t
+mq_forget (xlator_t *, quota_inode_ctx_t *);
+#endif
diff --git a/xlators/features/marker/src/marker.c b/xlators/features/marker/src/marker.c
new file mode 100644
index 000000000..a27a266f0
--- /dev/null
+++ b/xlators/features/marker/src/marker.c
@@ -0,0 +1,3130 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+#include "libxlator.h"
+#include "marker.h"
+#include "marker-mem-types.h"
+#include "marker-quota.h"
+#include "marker-quota-helper.h"
+#include "marker-common.h"
+#include "byte-order.h"
+#include "syncop.h"
+
+#define _GF_UID_GID_CHANGED 1
+
+void
+fini (xlator_t *this);
+
+int32_t
+marker_start_setxattr (call_frame_t *, xlator_t *);
+
+marker_local_t *
+marker_local_ref (marker_local_t *local)
+{
+        GF_VALIDATE_OR_GOTO ("marker", local, err);
+
+        LOCK (&local->lock);
+        {
+                local->ref++;
+        }
+        UNLOCK (&local->lock);
+
+        return local;
+err:
+        return NULL;
+}
+
+int
+marker_loc_fill (loc_t *loc, inode_t *inode, inode_t *parent, char *path)
+{
+        int     ret = -1;
+
+        if (!loc)
+                return ret;
+
+        if (inode) {
+                loc->inode = inode_ref (inode);
+                if (uuid_is_null (loc->gfid)) {
+                        uuid_copy (loc->gfid, loc->inode->gfid);
+                }
+        }
+
+        if (parent)
+                loc->parent = inode_ref (parent);
+
+        if (path) {
+                loc->path = gf_strdup (path);
+                if (!loc->path) {
+                        gf_log ("loc fill", GF_LOG_ERROR, "strdup failed");
+                        goto loc_wipe;
+                }
+
+                loc->name = strrchr (loc->path, '/');
+                if (loc->name)
+                        loc->name++;
+        }
+
+        ret = 0;
+loc_wipe:
+        if (ret < 0)
+                loc_wipe (loc);
+
+        return ret;
+}
+
+int
+marker_inode_loc_fill (inode_t *inode, loc_t *loc)
+{
+        char            *resolvedpath = NULL;
+        int              ret          = -1;
+	inode_t         *parent = NULL;
+
+        if ((!inode) || (!loc))
+                return ret;
+
+	parent = inode_parent (inode, NULL, NULL);
+
+        ret = inode_path (inode, NULL, &resolvedpath);
+        if (ret < 0)
+                goto err;
+
+        ret = marker_loc_fill (loc, inode, parent, resolvedpath);
+        if (ret < 0)
+                goto err;
+
+err:
+	if (parent)
+		inode_unref (parent);
+
+        GF_FREE (resolvedpath);
+
+        return ret;
+}
+
+int32_t
+marker_trav_parent (marker_local_t *local)
+{
+        int32_t ret = 0;
+        loc_t   loc = {0, };
+        inode_t *parent = NULL;
+        int8_t  need_unref = 0;
+
+        if (!local->loc.parent) {
+                parent = inode_parent (local->loc.inode, NULL, NULL);
+                if (parent)
+                        need_unref = 1;
+        } else
+                parent = local->loc.parent;
+
+        ret = marker_inode_loc_fill (parent, &loc);
+
+        if (ret < 0) {
+                ret = -1;
+                goto out;
+        }
+
+        loc_wipe (&local->loc);
+
+        local->loc = loc;
+out:
+        if (need_unref)
+                inode_unref (parent);
+
+        return ret;
+}
+
+int32_t
+marker_error_handler (xlator_t *this, marker_local_t *local, int32_t op_errno)
+{
+         marker_conf_t *priv = NULL;
+         const char    *path = NULL;
+
+         priv = (marker_conf_t *) this->private;
+         path = local
+                 ? (local->loc.path
+                    ? local->loc.path : uuid_utoa(local->loc.gfid))
+                 : "<nul>";
+
+         gf_log (this->name, GF_LOG_CRITICAL,
+                 "Indexing gone corrupt at %s (reason: %s)."
+                 " Geo-replication slave content needs to be revalidated",
+                 path, strerror (op_errno));
+        unlink (priv->timestamp_file);
+
+        return 0;
+}
+
+int32_t
+marker_local_unref (marker_local_t *local)
+{
+        int32_t   var = 0;
+
+        if (local == NULL)
+                return -1;
+
+        LOCK (&local->lock);
+        {
+                var = --local->ref;
+        }
+        UNLOCK (&local->lock);
+
+        if (var != 0)
+                goto out;
+
+        loc_wipe (&local->loc);
+        loc_wipe (&local->parent_loc);
+        if (local->xdata)
+                dict_unref (local->xdata);
+
+        if (local->oplocal) {
+                marker_local_unref (local->oplocal);
+                local->oplocal = NULL;
+        }
+        mem_put (local);
+out:
+        return 0;
+}
+
+int32_t
+stat_stampfile (xlator_t *this, marker_conf_t *priv,
+                struct volume_mark **status)
+{
+        struct stat          buf      = {0, };
+        struct volume_mark  *vol_mark = NULL;
+
+        vol_mark = GF_CALLOC (sizeof (struct volume_mark), 1,
+                              gf_marker_mt_volume_mark);
+
+        vol_mark->major = 1;
+        vol_mark->minor = 0;
+
+        GF_ASSERT (sizeof (priv->volume_uuid_bin) == 16);
+        memcpy (vol_mark->uuid, priv->volume_uuid_bin, 16);
+
+        if (stat (priv->timestamp_file, &buf) != -1) {
+                vol_mark->retval = 0;
+                vol_mark->sec = htonl (buf.st_ctime);
+                vol_mark->usec = htonl (ST_CTIM_NSEC (&buf)/1000);
+        } else
+                vol_mark->retval = 1;
+
+        *status = vol_mark;
+
+        return 0;
+}
+
+int32_t
+marker_getxattr_stampfile_cbk (call_frame_t *frame, xlator_t *this,
+                               const char *name, struct volume_mark *vol_mark,
+                               dict_t *xdata)
+{
+        int32_t   ret  = -1;
+        dict_t   *dict = NULL;
+
+        if (vol_mark == NULL){
+                STACK_UNWIND_STRICT (getxattr, frame, -1, ENOMEM, NULL, NULL);
+
+                goto out;
+        }
+
+        dict = dict_new ();
+
+        ret = dict_set_bin (dict, (char *)name, vol_mark,
+                            sizeof (struct volume_mark));
+        if (ret)
+                gf_log (this->name, GF_LOG_WARNING, "failed to set key %s",
+                        name);
+
+        STACK_UNWIND_STRICT (getxattr, frame, 0, 0, dict, xdata);
+
+        dict_unref (dict);
+out:
+        return 0;
+}
+
+int32_t
+call_from_special_client (call_frame_t *frame, xlator_t *this, const char *name)
+{
+        struct volume_mark     *vol_mark   = NULL;
+        marker_conf_t          *priv       = NULL;
+        gf_boolean_t            ret        = _gf_true;
+
+        priv = (marker_conf_t *)this->private;
+
+        if (frame->root->pid != GF_CLIENT_PID_GSYNCD || name == NULL ||
+            strcmp (name, MARKER_XATTR_PREFIX "." VOLUME_MARK) != 0) {
+                ret = _gf_false;
+                goto out;
+        }
+
+        stat_stampfile (this, priv, &vol_mark);
+
+        marker_getxattr_stampfile_cbk (frame, this, name, vol_mark, NULL);
+out:
+        return ret;
+}
+
+int32_t
+marker_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *dict,
+                     dict_t *xdata)
+{
+        int            ret    = 0;
+        char           *src   = NULL;
+        char           *dst   = NULL;
+        int            len    = 0;
+        marker_local_t *local = NULL;
+
+        local = frame->local;
+
+
+        if (cookie) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "Filtering the quota extended attributes");
+
+                /* If the getxattr is from a non special client, then do not
+                   copy the quota related xattrs (except the quota limit key
+                   i.e trusted.glusterfs.quota.limit-set which has been set by
+                   glusterd on the directory on which quota limit is set.) for
+                   directories. Let the healing of xattrs happen upon lookup.
+                   NOTE: setting of trusted.glusterfs.quota.limit-set as of now
+                   happens from glusterd. It should be moved to quotad. Also
+                   trusted.glusterfs.quota.limit-set is set on directory which
+                   is permanent till quota is removed on that directory or limit
+                   is changed. So let that xattr be healed by other xlators
+                   properly whenever directory healing is done.
+                */
+                ret = dict_get_ptr_and_len (dict, QUOTA_LIMIT_KEY,
+                                            (void **)&src, &len);
+                if (ret) {
+                        gf_log (this->name, GF_LOG_DEBUG, "dict_get on %s "
+                                "failed", QUOTA_LIMIT_KEY);
+                } else {
+                        dst = GF_CALLOC (len, sizeof (char), gf_common_mt_char);
+                        if (dst)
+                                memcpy (dst, src, len);
+                }
+
+                /*
+                 * Except limit-set xattr, rest of the xattrs are maintained
+                 * by quota xlator. Don't expose them to other xlators.
+                 * This filter makes sure quota xattrs are not healed as part of
+                 * metadata self-heal
+                 */
+                GF_REMOVE_INTERNAL_XATTR ("trusted.glusterfs.quota*", dict);
+                if (!ret && IA_ISDIR (local->loc.inode->ia_type) && dst) {
+                        ret = dict_set_dynptr (dict, QUOTA_LIMIT_KEY,
+                                               dst, len);
+                        if (ret)
+                                gf_log (this->name, GF_LOG_WARNING, "setting "
+                                        "key %s failed", QUOTA_LIMIT_KEY);
+                        else
+                                dst = NULL;
+                }
+        }
+
+        GF_FREE (dst);
+
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
+        marker_local_unref (local);
+        return 0;
+}
+
+int32_t
+marker_getxattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                 const char *name, dict_t *xdata)
+{
+        gf_boolean_t   ret    = _gf_false;
+        marker_conf_t *priv   = NULL;
+        unsigned long  cookie = 0;
+        marker_local_t *local = NULL;
+
+        priv = this->private;
+
+        frame->local = mem_get0 (this->local_pool);
+        local = frame->local;
+        if (local == NULL)
+                goto out;
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+        if (ret < 0)
+                goto out;
+
+        gf_log (this->name, GF_LOG_DEBUG, "USER:PID = %d", frame->root->pid);
+
+        if (priv && priv->feature_enabled & GF_XTIME)
+                ret = call_from_special_client (frame, this, name);
+
+        if (ret == _gf_false) {
+                if (name == NULL) {
+                        /* Signifies that marker translator
+                         * has to filter the quota's xattr's,
+                         * this is to prevent afr from performing
+                         * self healing on marker-quota xattrs'
+                         */
+                        cookie = 1;
+                }
+                STACK_WIND_COOKIE (frame, marker_getxattr_cbk, (void *)cookie,
+                                   FIRST_CHILD(this),
+                                   FIRST_CHILD(this)->fops->getxattr, loc,
+                                   name, xdata);
+        }
+
+        return 0;
+out:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (getxattr, frame, -1, ENOMEM, NULL, NULL);
+        marker_local_unref (local);
+        return 0;
+}
+
+
+int32_t
+marker_setxattr_done (call_frame_t *frame)
+{
+        marker_local_t *local = NULL;
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_DESTROY (frame->root);
+
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int
+marker_specific_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                              int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        int32_t         ret   = 0;
+        int32_t         done  = 0;
+        marker_local_t *local = NULL;
+
+        local = (marker_local_t*) frame->local;
+
+        if (op_ret == -1 && op_errno == ENOSPC) {
+                marker_error_handler (this, local, op_errno);
+                done = 1;
+                goto out;
+        }
+
+        if (local) {
+                if (local->loc.path && strcmp (local->loc.path, "/") == 0) {
+                        done = 1;
+                        goto out;
+                }
+                if (__is_root_gfid (local->loc.gfid)) {
+                        done = 1;
+                        goto out;
+                }
+        }
+
+        ret = marker_trav_parent (local);
+
+        if (ret == -1) {
+                gf_log (this->name, GF_LOG_DEBUG, "Error occurred "
+                        "while traversing to the parent, stopping marker");
+
+                done = 1;
+
+                goto out;
+        }
+
+        marker_start_setxattr (frame, this);
+
+out:
+        if (done) {
+                marker_setxattr_done (frame);
+        }
+
+        return 0;
+}
+
+int32_t
+marker_start_setxattr (call_frame_t *frame, xlator_t *this)
+{
+        int32_t          ret   = -1;
+        dict_t          *dict  = NULL;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        local = (marker_local_t*) frame->local;
+
+        if (!local)
+                goto out;
+
+        dict = dict_new ();
+
+        if (!dict)
+                goto out;
+
+        if (local->loc.inode && uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+
+        GF_UUID_ASSERT (local->loc.gfid);
+
+        ret = dict_set_static_bin (dict, priv->marker_xattr,
+                                   (void *)local->timebuf, 8);
+        if (ret) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to set marker xattr (%s)", local->loc.path);
+                goto out;
+        }
+
+        STACK_WIND (frame, marker_specific_setxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr, &local->loc, dict, 0,
+                    NULL);
+
+        ret = 0;
+out:
+        if (dict)
+                dict_unref (dict);
+
+        return ret;
+}
+
+void
+marker_gettimeofday (marker_local_t *local)
+{
+        struct timeval tv = {0, };
+
+        gettimeofday (&tv, NULL);
+
+        local->timebuf [0] = htonl (tv.tv_sec);
+        local->timebuf [1] = htonl (tv.tv_usec);
+
+        return;
+}
+
+int32_t
+marker_create_frame (xlator_t *this, marker_local_t *local)
+{
+        call_frame_t *frame = NULL;
+
+        frame = create_frame (this, this->ctx->pool);
+
+        frame->local = (void *) local;
+
+        marker_start_setxattr (frame, this);
+
+        return 0;
+}
+
+int32_t
+marker_xtime_update_marks (xlator_t *this, marker_local_t *local)
+{
+        marker_conf_t *priv = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, local, out);
+
+        priv = this->private;
+
+        if ((local->pid == GF_CLIENT_PID_GSYNCD
+             && !(priv->feature_enabled & GF_XTIME_GSYNC_FORCE))
+            || (local->pid == GF_CLIENT_PID_DEFRAG))
+                goto out;
+
+        marker_gettimeofday (local);
+
+        marker_local_ref (local);
+
+        marker_create_frame (this, local);
+out:
+        return 0;
+}
+
+
+int32_t
+marker_mkdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, inode_t *inode,
+                  struct iatt *buf, struct iatt *preparent,
+                  struct iatt *postparent, dict_t *xdata)
+{
+        marker_conf_t      *priv    = NULL;
+        marker_local_t     *local   = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "error occurred "
+                        "while Creating a file %s", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (mkdir, frame, op_ret, op_errno, inode,
+                             buf, preparent, postparent, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_set_inode_xattr (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int
+marker_mkdir (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+              mode_t umask, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_mkdir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->mkdir, loc, mode, umask, xdata);
+
+        return 0;
+err:
+        STACK_UNWIND_STRICT (mkdir, frame, -1, ENOMEM, NULL,
+                             NULL, NULL, NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+marker_create_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, fd_t *fd, inode_t *inode,
+                   struct iatt *buf, struct iatt *preparent,
+                   struct iatt *postparent, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "error occurred "
+                        "while Creating a file %s", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (create, frame, op_ret, op_errno, fd, inode, buf,
+                             preparent, postparent, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_set_inode_xattr (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_create (call_frame_t *frame, xlator_t *this, loc_t *loc, int32_t flags,
+               mode_t mode, mode_t umask, fd_t *fd, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_create_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->create, loc, flags, mode, umask,
+                    fd, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (create, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
+                             NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                   struct iatt *postbuf, dict_t *xdata)
+{
+        marker_conf_t      *priv    = NULL;
+        marker_local_t     *local   = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "error occurred "
+                        "while write, %s", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf,
+                             xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_initiate_quota_txn (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_writev (call_frame_t *frame,
+               xlator_t *this,
+               fd_t *fd,
+               struct iovec *vector,
+               int32_t count,
+               off_t offset, uint32_t flags,
+               struct iobref *iobref, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = marker_inode_loc_fill (fd->inode, &local->loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_writev_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->writev, fd, vector, count, offset,
+                    flags, iobref, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (writev, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_rmdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                  struct iatt *postparent, dict_t *xdata)
+{
+        marker_conf_t      *priv    = NULL;
+        marker_local_t     *local   = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "error occurred "
+                        "rmdir %s", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (rmdir, frame, op_ret, op_errno, preparent,
+                             postparent, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_reduce_parent_size (this, &local->loc, -1);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_rmdir (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+              dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_rmdir_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->rmdir, loc, flags, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (rmdir, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_unlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *preparent,
+                   struct iatt *postparent, dict_t *xdata)
+{
+        marker_conf_t      *priv    = NULL;
+        marker_local_t     *local   = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE,
+                        "%s occurred in unlink", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (unlink, frame, op_ret, op_errno, preparent,
+                             postparent, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA) {
+                if (!local->skip_txn)
+                        mq_reduce_parent_size (this, &local->loc, -1);
+        }
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+
+int32_t
+marker_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+               dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto unlink_wind;
+
+        local = mem_get0 (this->local_pool);
+        local->xflag = xflag;
+        if (xdata)
+                local->xdata = dict_ref (xdata);
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+
+        if (xdata && dict_get (xdata, GLUSTERFS_MARKER_DONT_ACCOUNT_KEY)) {
+                local->skip_txn = 1;
+                goto unlink_wind;
+        }
+
+unlink_wind:
+        STACK_WIND (frame, marker_unlink_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
+        return 0;
+err:
+        frame->local = NULL;
+        STACK_UNWIND_STRICT (unlink, frame, -1, ENOMEM, NULL, NULL, NULL);
+        marker_local_unref (local);
+        return 0;
+}
+
+
+int32_t
+marker_link_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                 int32_t op_ret, int32_t op_errno, inode_t *inode,
+                 struct iatt *buf, struct iatt *preparent,
+                 struct iatt *postparent, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
+                        "linking a file ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (link, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA) {
+                if (!local->skip_txn)
+                        mq_set_inode_xattr (this, &local->loc);
+        }
+
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_link (call_frame_t *frame, xlator_t *this, loc_t *oldloc, loc_t *newloc,
+             dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, newloc);
+
+        if (ret == -1)
+                goto err;
+
+        if (xdata && dict_get (xdata, GLUSTERFS_MARKER_DONT_ACCOUNT_KEY))
+                local->skip_txn = 1;
+wind:
+        STACK_WIND (frame, marker_link_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->link, oldloc, newloc, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (link, frame, -1, ENOMEM, NULL, NULL, NULL, NULL,
+                             NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_rename_done (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        marker_local_t *local  = NULL, *oplocal = NULL;
+        loc_t           newloc = {0, };
+        marker_conf_t  *priv   = NULL;
+
+        local = frame->local;
+        oplocal = local->oplocal;
+
+        priv = this->private;
+
+        frame->local = NULL;
+
+        if (op_ret < 0) {
+                if (local->err == 0) {
+                        local->err =  op_errno ? op_errno : EINVAL;
+                }
+
+                gf_log (this->name, GF_LOG_WARNING,
+                        "inodelk (UNLOCK) failed on path:%s (gfid:%s) (%s)",
+                        local->parent_loc.path,
+                        uuid_utoa (local->parent_loc.inode->gfid),
+                        strerror (op_errno));
+        }
+
+        if (local->stub != NULL) {
+                call_resume (local->stub);
+                local->stub = NULL;
+        } else if (local->err != 0) {
+                STACK_UNWIND_STRICT (rename, frame, -1, local->err, NULL, NULL,
+                                     NULL, NULL, NULL, NULL);
+        } else {
+                gf_log (this->name, GF_LOG_CRITICAL,
+                        "continuation stub to unwind the call is absent, hence "
+                        "call will be hung (call-stack id = %"PRIu64")",
+                        frame->root->unique);
+        }
+
+        mq_reduce_parent_size (this, &oplocal->loc, oplocal->contribution);
+
+        if (local->loc.inode != NULL) {
+                mq_reduce_parent_size (this, &local->loc, local->contribution);
+        }
+
+        newloc.inode = inode_ref (oplocal->loc.inode);
+        newloc.path = gf_strdup (local->loc.path);
+        newloc.name = strrchr (newloc.path, '/');
+        if (newloc.name)
+                newloc.name++;
+        newloc.parent = inode_ref (local->loc.parent);
+
+        mq_set_inode_xattr (this, &newloc);
+
+        loc_wipe (&newloc);
+
+        if (priv->feature_enabled & GF_XTIME) {
+                //update marks on oldpath
+                uuid_copy (local->loc.gfid, oplocal->loc.inode->gfid);
+                marker_xtime_update_marks (this, oplocal);
+                marker_xtime_update_marks (this, local);
+        }
+
+        marker_local_unref (local);
+        marker_local_unref (oplocal);
+        return 0;
+}
+
+
+int32_t
+marker_rename_release_newp_lock (call_frame_t *frame, void *cookie,
+                                 xlator_t *this, int32_t op_ret,
+                                 int32_t op_errno, dict_t *xdata)
+{
+        marker_local_t  *local = NULL, *oplocal = NULL;
+        struct gf_flock  lock  = {0, };
+
+        local = frame->local;
+        oplocal = local->oplocal;
+
+        if (op_ret < 0) {
+                if (local->err == 0) {
+                        local->err = op_errno ? op_errno : EINVAL;
+                }
+
+                gf_log (this->name, GF_LOG_WARNING,
+                        "inodelk (UNLOCK) failed on %s (gfid:%s) (%s)",
+                        oplocal->parent_loc.path,
+                        uuid_utoa (oplocal->parent_loc.inode->gfid),
+                        strerror (op_errno));
+        }
+
+        if (local->next_lock_on == NULL) {
+                marker_rename_done (frame, NULL, this, 0, 0, NULL);
+                goto out;
+        }
+
+        lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+        STACK_WIND (frame,
+                    marker_rename_done,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &local->parent_loc, F_SETLKW, &lock, NULL);
+
+out:
+        return 0;
+}
+
+
+int32_t
+marker_rename_release_oldp_lock (call_frame_t *frame, void *cookie,
+                                 xlator_t *this, int32_t op_ret,
+                                 int32_t op_errno, dict_t *xdata)
+{
+        marker_local_t  *local = NULL, *oplocal = NULL;
+        struct gf_flock  lock  = {0, };
+
+        local = frame->local;
+        oplocal = local->oplocal;
+
+        if ((op_ret < 0) && (op_errno != ENOATTR)) {
+                local->err = op_errno;
+        }
+
+        //Reset frame uid and gid if set.
+        if (cookie == (void *) _GF_UID_GID_CHANGED)
+                MARKER_RESET_UID_GID (frame, frame->root, local);
+
+        lock.l_type   = F_UNLCK;
+        lock.l_whence = SEEK_SET;
+        lock.l_start  = 0;
+        lock.l_len    = 0;
+        lock.l_pid    = 0;
+
+        STACK_WIND (frame,
+                    marker_rename_release_newp_lock,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, &oplocal->parent_loc, F_SETLKW, &lock, NULL);
+        return 0;
+}
+
+
+int32_t
+marker_rename_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                   struct iatt *preoldparent, struct iatt *postoldparent,
+                   struct iatt *prenewparent, struct iatt *postnewparent,
+                   dict_t *xdata)
+{
+        marker_conf_t  *priv                 = NULL;
+        marker_local_t *local                = NULL;
+        marker_local_t *oplocal              = NULL;
+        call_stub_t    *stub                 = NULL;
+        int32_t         ret                  = 0;
+        char            contri_key [512]     = {0, };
+        loc_t           newloc               = {0, };
+
+        local = (marker_local_t *) frame->local;
+
+        if (local != NULL) {
+                oplocal = local->oplocal;
+        }
+
+        priv = this->private;
+
+        if (op_ret < 0) {
+                if (local != NULL) {
+                        local->err = op_errno;
+                }
+
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
+                        "renaming a file ", strerror (op_errno));
+        }
+
+        if (priv->feature_enabled & GF_QUOTA) {
+                if ((op_ret < 0) || (local == NULL)) {
+                        goto quota_err;
+                }
+
+                stub = fop_rename_cbk_stub (frame, default_rename_cbk, op_ret,
+                                            op_errno, buf, preoldparent,
+                                            postoldparent, prenewparent,
+                                            postnewparent, xdata);
+                if (stub == NULL) {
+                        local->err = ENOMEM;
+                        goto quota_err;
+                }
+
+                local->stub = stub;
+
+                GET_CONTRI_KEY (contri_key, oplocal->loc.parent->gfid, ret);
+                if (ret < 0) {
+                        local->err = ENOMEM;
+                        goto quota_err;
+                }
+
+                /* Removexattr requires uid and gid to be 0,
+                 * reset them in the callback.
+                 */
+                MARKER_SET_UID_GID (frame, local, frame->root);
+
+                newloc.inode = inode_ref (oplocal->loc.inode);
+                newloc.path = gf_strdup (local->loc.path);
+                newloc.name = strrchr (newloc.path, '/');
+                if (newloc.name)
+                        newloc.name++;
+                newloc.parent = inode_ref (local->loc.parent);
+                uuid_copy (newloc.gfid, oplocal->loc.inode->gfid);
+
+                STACK_WIND_COOKIE (frame, marker_rename_release_oldp_lock,
+                                   frame->cookie, FIRST_CHILD(this),
+                                   FIRST_CHILD(this)->fops->removexattr,
+                                   &newloc, contri_key, NULL);
+
+                loc_wipe (&newloc);
+        } else {
+                frame->local = NULL;
+
+                STACK_UNWIND_STRICT (rename, frame, op_ret, op_errno, buf,
+                                     preoldparent, postoldparent, prenewparent,
+                                     postnewparent, xdata);
+
+                if ((op_ret < 0) || (local == NULL)) {
+                        goto out;
+                }
+
+                if (priv->feature_enabled & GF_XTIME) {
+                        //update marks on oldpath
+                        uuid_copy (local->loc.gfid, oplocal->loc.inode->gfid);
+                        marker_xtime_update_marks (this, oplocal);
+                        marker_xtime_update_marks (this, local);
+                }
+        }
+
+out:
+        if (!(priv->feature_enabled & GF_QUOTA)) {
+                marker_local_unref (local);
+                marker_local_unref (oplocal);
+        }
+
+        return 0;
+
+quota_err:
+        marker_rename_release_oldp_lock (frame, NULL, this, 0, 0, NULL);
+        return 0;
+}
+
+
+int32_t
+marker_do_rename (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        marker_local_t       *local           = NULL, *oplocal = NULL;
+        char                  contri_key[512] = {0, };
+        int32_t               ret             = 0;
+        int64_t              *contribution    = 0;
+
+        local = frame->local;
+        oplocal = local->oplocal;
+
+        //Reset frame uid and gid if set.
+        if (cookie == (void *) _GF_UID_GID_CHANGED)
+                MARKER_RESET_UID_GID (frame, frame->root, local);
+
+        if ((op_ret < 0) && (op_errno != ENOATTR)) {
+                local->err = op_errno ? op_errno : EINVAL;
+                gf_log (this->name, GF_LOG_WARNING,
+                        "fetching contribution values from %s (gfid:%s) "
+                        "failed (%s)", local->loc.path,
+                        uuid_utoa (local->loc.inode->gfid),
+                        strerror (op_errno));
+                goto err;
+        }
+
+        if (local->loc.inode != NULL) {
+                GET_CONTRI_KEY (contri_key, local->loc.parent->gfid, ret);
+                if (ret < 0) {
+                        local->err = errno ? errno : ENOMEM;
+                        goto err;
+                }
+
+                if (dict_get_bin (dict, contri_key,
+                                  (void **) &contribution) == 0) {
+                        local->contribution = ntoh64 (*contribution);
+                }
+        }
+
+        STACK_WIND (frame, marker_rename_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->rename, &oplocal->loc,
+                    &local->loc, NULL);
+
+        return 0;
+
+err:
+        marker_rename_release_oldp_lock (frame, NULL, this, 0, 0, NULL);
+        return 0;
+}
+
+
+int32_t
+marker_get_newpath_contribution (call_frame_t *frame, void *cookie,
+                                 xlator_t *this, int32_t op_ret,
+                                 int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        marker_local_t *local           = NULL, *oplocal = NULL;
+        char            contri_key[512] = {0, };
+        int32_t         ret             = 0;
+        int64_t        *contribution    = 0;
+
+        local = frame->local;
+        oplocal = local->oplocal;
+
+        //Reset frame uid and gid if set.
+        if (cookie == (void *) _GF_UID_GID_CHANGED)
+                MARKER_RESET_UID_GID (frame, frame->root, local);
+
+        if ((op_ret < 0) && (op_errno != ENOATTR)) {
+                local->err = op_errno ? op_errno : EINVAL;
+                gf_log (this->name, GF_LOG_WARNING,
+                        "fetching contribution values from %s (gfid:%s) "
+                        "failed (%s)", oplocal->loc.path,
+                        uuid_utoa (oplocal->loc.inode->gfid),
+                        strerror (op_errno));
+                goto err;
+        }
+
+        GET_CONTRI_KEY (contri_key, oplocal->loc.parent->gfid, ret);
+        if (ret < 0) {
+                local->err = errno ? errno : ENOMEM;
+                goto err;
+        }
+
+        if (dict_get_bin (dict, contri_key, (void **) &contribution) == 0)
+                oplocal->contribution = ntoh64 (*contribution);
+
+        if (local->loc.inode != NULL) {
+                GET_CONTRI_KEY (contri_key, local->loc.parent->gfid, ret);
+                if (ret < 0) {
+                        local->err = errno ? errno : ENOMEM;
+                        goto err;
+                }
+
+                /* getxattr requires uid and gid to be 0,
+                 * reset them in the callback.
+                 */
+                MARKER_SET_UID_GID (frame, local, frame->root);
+                if (uuid_is_null (local->loc.gfid))
+                        uuid_copy (local->loc.gfid, local->loc.inode->gfid);
+
+                GF_UUID_ASSERT (local->loc.gfid);
+
+                STACK_WIND_COOKIE (frame, marker_do_rename,
+                                   frame->cookie, FIRST_CHILD(this),
+                                   FIRST_CHILD(this)->fops->getxattr,
+                                   &local->loc, contri_key, NULL);
+        } else {
+                marker_do_rename (frame, NULL, this, 0, 0, NULL, NULL);
+        }
+
+        return 0;
+err:
+        marker_rename_release_oldp_lock (frame, NULL, this, 0, 0, NULL);
+        return 0;
+}
+
+
+int32_t
+marker_get_oldpath_contribution (call_frame_t *frame, void *cookie,
+                                 xlator_t *this, int32_t op_ret,
+                                 int32_t op_errno, dict_t *xdata)
+{
+        marker_local_t *local           = NULL, *oplocal = NULL;
+        char            contri_key[512] = {0, };
+        int32_t         ret             = 0;
+
+        local = frame->local;
+        oplocal = local->oplocal;
+
+        if (op_ret < 0) {
+                local->err = op_errno ? op_errno : EINVAL;
+                gf_log (this->name, GF_LOG_WARNING,
+                        "cannot hold inodelk on %s (gfid:%s) (%s)",
+                        local->next_lock_on->path,
+                        uuid_utoa (local->next_lock_on->inode->gfid),
+                        strerror (op_errno));
+                goto lock_err;
+        }
+
+        GET_CONTRI_KEY (contri_key, oplocal->loc.parent->gfid, ret);
+        if (ret < 0) {
+                local->err = errno ? errno : ENOMEM;
+                goto quota_err;
+        }
+
+        /* getxattr requires uid and gid to be 0,
+         * reset them in the callback.
+         */
+        MARKER_SET_UID_GID (frame, local, frame->root);
+
+        if (uuid_is_null (oplocal->loc.gfid))
+                        uuid_copy (oplocal->loc.gfid,
+                                   oplocal->loc.inode->gfid);
+
+        GF_UUID_ASSERT (oplocal->loc.gfid);
+
+        STACK_WIND_COOKIE (frame, marker_get_newpath_contribution,
+                           frame->cookie, FIRST_CHILD(this),
+                           FIRST_CHILD(this)->fops->getxattr,
+                           &oplocal->loc, contri_key, NULL);
+        return 0;
+
+quota_err:
+        marker_rename_release_oldp_lock (frame, NULL, this, 0, 0, NULL);
+        return 0;
+
+lock_err:
+        if ((local->next_lock_on == NULL)
+            || (local->next_lock_on == &local->parent_loc)) {
+                local->next_lock_on = NULL;
+                marker_rename_release_oldp_lock (frame, NULL, this, 0, 0, NULL);
+        } else {
+                marker_rename_release_newp_lock (frame, NULL, this, 0, 0, NULL);
+        }
+
+        return 0;
+}
+
+
+int32_t
+marker_rename_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                           int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        marker_local_t  *local = NULL, *oplocal = NULL;
+        loc_t           *loc   = NULL;
+        struct gf_flock  lock  = {0, };
+
+        local = frame->local;
+        oplocal = local->oplocal;
+
+        if (op_ret < 0) {
+                if (local->next_lock_on != &oplocal->parent_loc) {
+                        loc = &oplocal->parent_loc;
+                } else {
+                        loc = &local->parent_loc;
+                }
+
+                local->err = op_errno ? op_errno : EINVAL;
+                gf_log (this->name, GF_LOG_WARNING,
+                        "cannot hold inodelk on %s (gfid:%s) (%s)",
+                        loc->path, uuid_utoa (loc->inode->gfid),
+                        strerror (op_errno));
+                goto err;
+        }
+
+        if (local->next_lock_on != NULL) {
+                lock.l_len    = 0;
+                lock.l_start  = 0;
+                lock.l_type   = F_WRLCK;
+                lock.l_whence = SEEK_SET;
+
+                STACK_WIND (frame,
+                            marker_get_oldpath_contribution,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->inodelk,
+                            this->name, local->next_lock_on,
+                            F_SETLKW, &lock, NULL);
+        } else {
+                marker_get_oldpath_contribution (frame, 0, this, 0, 0, NULL);
+        }
+
+        return 0;
+
+err:
+        marker_rename_done (frame, NULL, this, 0, 0, NULL);
+        return 0;
+}
+
+
+int32_t
+marker_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+               loc_t *newloc, dict_t *xdata)
+{
+        int32_t         ret              = 0;
+        marker_local_t *local            = NULL;
+        marker_local_t *oplocal          = NULL;
+        marker_conf_t  *priv             = NULL;
+        struct gf_flock lock             = {0, };
+        loc_t          *lock_on          = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto rename_wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        oplocal = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, oplocal);
+
+        frame->local = local;
+
+        local->oplocal = marker_local_ref (oplocal);
+
+        ret = loc_copy (&local->loc, newloc);
+        if (ret < 0)
+                goto err;
+
+        ret = loc_copy (&oplocal->loc, oldloc);
+        if (ret < 0)
+                goto err;
+
+        if (!(priv->feature_enabled & GF_QUOTA)) {
+                goto rename_wind;
+        }
+
+        ret = mq_inode_loc_fill (NULL, newloc->parent, &local->parent_loc);
+        if (ret < 0)
+                goto err;
+
+        ret = mq_inode_loc_fill (NULL, oldloc->parent, &oplocal->parent_loc);
+        if (ret < 0)
+                goto err;
+
+        if ((newloc->inode != NULL) && (newloc->parent != oldloc->parent)
+            && (uuid_compare (newloc->parent->gfid,
+                              oldloc->parent->gfid) < 0)) {
+                lock_on = &local->parent_loc;
+                local->next_lock_on = &oplocal->parent_loc;
+        } else {
+                lock_on = &oplocal->parent_loc;
+                if ((newloc->inode != NULL) && (newloc->parent
+                                                != oldloc->parent)) {
+                        local->next_lock_on = &local->parent_loc;
+                }
+        }
+
+        lock.l_len    = 0;
+        lock.l_start  = 0;
+        lock.l_type   = F_WRLCK;
+        lock.l_whence = SEEK_SET;
+
+        STACK_WIND (frame,
+                    marker_rename_inodelk_cbk,
+                    FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->inodelk,
+                    this->name, lock_on,
+                    F_SETLKW, &lock, NULL);
+
+        return 0;
+
+rename_wind:
+        STACK_WIND (frame, marker_rename_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->rename, oldloc, newloc, xdata);
+
+        return 0;
+err:
+        STACK_UNWIND_STRICT (rename, frame, -1, ENOMEM, NULL,
+                             NULL, NULL, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                     struct iatt *postbuf, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
+                        "truncating a file ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (truncate, frame, op_ret, op_errno, prebuf,
+                             postbuf, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_initiate_quota_txn (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+                 dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_truncate_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->truncate, loc, offset, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (truncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_ftruncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                      struct iatt *postbuf, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
+                        "truncating a file ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (ftruncate, frame, op_ret, op_errno, prebuf,
+                             postbuf, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_initiate_quota_txn (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+                  dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = marker_inode_loc_fill (fd->inode, &local->loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_ftruncate_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->ftruncate, fd, offset, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (ftruncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_symlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, inode_t *inode,
+                    struct iatt *buf, struct iatt *preparent,
+                    struct iatt *postparent, dict_t *xdata)
+{
+        marker_conf_t      *priv    = NULL;
+        marker_local_t     *local   = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
+                        "creating symlinks ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (symlink, frame, op_ret, op_errno, inode, buf,
+                             preparent, postparent, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_set_inode_xattr (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int
+marker_symlink (call_frame_t *frame, xlator_t *this, const char *linkpath,
+                loc_t *loc, mode_t umask, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_symlink_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->symlink, linkpath, loc, umask,
+                    xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (symlink, frame, -1, ENOMEM, NULL,
+                             NULL, NULL, NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+marker_mknod_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, inode_t *inode,
+                  struct iatt *buf, struct iatt *preparent,
+                  struct iatt *postparent, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
+                        "creating symlinks ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (mknod, frame, op_ret, op_errno, inode,
+                             buf, preparent, postparent, xdata);
+
+        if (op_ret == -1 ||  local == NULL)
+                goto out;
+
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+
+        priv = this->private;
+
+        if ((priv->feature_enabled & GF_QUOTA) && (S_ISREG (local->mode))) {
+                mq_set_inode_xattr (this, &local->loc);
+        }
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int
+marker_mknod (call_frame_t *frame, xlator_t *this, loc_t *loc, mode_t mode,
+              dev_t rdev, mode_t umask, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        local->mode = mode;
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_mknod_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->mknod, loc, mode, rdev, umask,
+                    xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (mknod, frame, -1, ENOMEM, NULL,
+                             NULL, NULL, NULL, NULL);
+        return 0;
+}
+
+
+int32_t
+marker_fallocate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                      struct iatt *postbuf, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
+                        "fallocating a file ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (fallocate, frame, op_ret, op_errno, prebuf,
+                             postbuf, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_initiate_quota_txn (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_fallocate(call_frame_t *frame, xlator_t *this, fd_t *fd, int32_t mode,
+		 off_t offset, size_t len, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = marker_inode_loc_fill (fd->inode, &local->loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_fallocate_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fallocate, fd, mode, offset, len,
+		    xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (fallocate, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_discard_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                   struct iatt *postbuf, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred during discard",
+                        strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (discard, frame, op_ret, op_errno, prebuf,
+                             postbuf, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_initiate_quota_txn (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_discard(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+	       size_t len, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = marker_inode_loc_fill (fd->inode, &local->loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_discard_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->discard, fd, offset, len, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (discard, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+int32_t
+marker_zerofill_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                   struct iatt *postbuf, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred during zerofill",
+                        strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (zerofill, frame, op_ret, op_errno, prebuf,
+                             postbuf, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA)
+                mq_initiate_quota_txn (this, &local->loc);
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_zerofill(call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+               off_t len, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = marker_inode_loc_fill (fd->inode, &local->loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_zerofill_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->zerofill, fd, offset, len, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (zerofill, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+/* when a call from the special client is received on
+ * key trusted.glusterfs.volume-mark with value "RESET"
+ * or if the value is 0length, update the change the
+ * access time and modification time via touching the
+ * timestamp file.
+ */
+int32_t
+call_from_sp_client_to_reset_tmfile (call_frame_t *frame,
+                                     xlator_t     *this,
+                                     dict_t       *dict)
+{
+        int32_t          fd       = 0;
+        int32_t          op_ret   = 0;
+        int32_t          op_errno = 0;
+        data_t          *data     = NULL;
+        marker_conf_t   *priv     = NULL;
+
+        if (frame == NULL || this == NULL || dict == NULL)
+                return -1;
+
+        priv = this->private;
+
+        data = dict_get (dict, "trusted.glusterfs.volume-mark");
+        if (data == NULL)
+                return -1;
+
+        if (frame->root->pid != GF_CLIENT_PID_GSYNCD) {
+                op_ret = -1;
+                op_errno = EPERM;
+
+                goto out;
+        }
+
+        if (data->len == 0 || (data->len == 5 &&
+                               memcmp (data->data, "RESET", 5) == 0)) {
+                fd = open (priv->timestamp_file, O_WRONLY|O_TRUNC);
+                if (fd != -1) {
+                        /* TODO check  whether the O_TRUNC would update the
+                         * timestamps on a zero length file on all machies.
+                         */
+                        close (fd);
+                }
+
+                if (fd != -1 || errno == ENOENT) {
+                        op_ret = 0;
+                        op_errno = 0;
+                } else {
+                        op_ret = -1;
+                        op_errno = errno;
+                }
+        } else {
+                op_ret = -1;
+                op_errno = EINVAL;
+        }
+out:
+        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_setxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred in "
+                        "setxattr ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int
+remove_quota_keys (dict_t *dict, char *k, data_t *v, void *data)
+{
+	call_frame_t 	*frame = data;
+	marker_local_t 	*local = frame->local;
+	xlator_t 	*this = frame->this;
+	int 		ret = -1;
+
+	ret = syncop_removexattr (FIRST_CHILD (this), &local->loc, k, 0);
+	if (ret) {
+		gf_log (this->name, GF_LOG_ERROR, "%s: Failed to remove "
+			"extended attribute: %s", local->loc.path, k);
+                return -1;
+	}
+	return 0;
+}
+
+int
+quota_xattr_cleaner_cbk (int ret, call_frame_t *frame, void *args)
+{
+        dict_t *xdata = args;
+        int op_ret = -1;
+        int op_errno = 0;
+	marker_local_t *local = NULL;
+
+	local = frame->local;
+	frame->local = NULL;
+
+        op_ret   = (ret < 0)? -1: 0;
+        op_errno = -ret;
+
+        STACK_UNWIND_STRICT (setxattr, frame, op_ret, op_errno, xdata);
+	marker_local_unref (local);
+        return ret;
+}
+
+int
+quota_xattr_cleaner (void *args)
+{
+        struct synctask *task  = NULL;
+        call_frame_t    *frame = NULL;
+        xlator_t        *this  = NULL;
+        marker_local_t   *local = NULL;
+        dict_t          *xdata = NULL;
+        int             ret    = -1;
+
+        task = synctask_get ();
+        if (!task)
+                goto out;
+
+        frame = task->frame;
+        this  = frame->this;
+        local = frame->local;
+
+        ret = syncop_listxattr (FIRST_CHILD(this), &local->loc, &xdata);
+        if (ret == -1) {
+                ret = -errno;
+                goto out;
+        }
+
+	ret = dict_foreach_fnmatch (xdata, "trusted.glusterfs.quota.*",
+                                    remove_quota_keys, frame);
+        if (ret == -1) {
+                ret = -errno;
+                goto out;
+        }
+	ret = dict_foreach_fnmatch (xdata, PGFID_XATTR_KEY_PREFIX"*",
+                                    remove_quota_keys, frame);
+        if (ret == -1) {
+                ret = -errno;
+                goto out;
+        }
+
+        ret = 0;
+out:
+        if (xdata)
+                dict_unref (xdata);
+
+        return ret;
+}
+
+int
+marker_do_xattr_cleanup (call_frame_t *frame, xlator_t *this, dict_t *xdata,
+                        loc_t *loc)
+{
+        int           ret       = -1;
+        marker_local_t *local    = NULL;
+
+        local = mem_get0 (this->local_pool);
+	if (!local)
+		goto out;
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        loc_copy (&local->loc, loc);
+        ret = synctask_new (this->ctx->env, quota_xattr_cleaner,
+			    quota_xattr_cleaner_cbk, frame, xdata);
+        if (ret) {
+		gf_log (this->name, GF_LOG_ERROR, "Failed to create synctask "
+			"for cleaning up quota extended attributes");
+                goto out;
+	}
+
+        ret = 0;
+out:
+        if (ret) {
+		frame->local = NULL;
+                STACK_UNWIND_STRICT (setxattr, frame, -1, ENOMEM, xdata);
+		marker_local_unref (local);
+        }
+        return ret;
+}
+
+static inline gf_boolean_t
+marker_xattr_cleanup_cmd (dict_t *dict)
+{
+        return (dict_get (dict, VIRTUAL_QUOTA_XATTR_CLEANUP_KEY) != NULL);
+}
+
+int32_t
+marker_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+                 int32_t flags, dict_t *xdata)
+{
+        int32_t          ret            = 0;
+        marker_local_t  *local          = NULL;
+        marker_conf_t   *priv           = NULL;
+        int              op_errno       = ENOMEM;
+
+        priv = this->private;
+
+        if (marker_xattr_cleanup_cmd (dict)) {
+                if (frame->root->uid != 0 || frame->root->gid != 0) {
+                        op_errno = EPERM;
+                        ret = -1;
+                        goto err;
+                }
+
+                /* The following function does the cleanup and then unwinds the
+                 * corresponding call*/
+                loc_path (loc, NULL);
+                marker_do_xattr_cleanup (frame, this, xdata, loc);
+                return 0;
+        }
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        ret = call_from_sp_client_to_reset_tmfile (frame, this, dict);
+        if (ret == 0)
+                return 0;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_setxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->setxattr, loc, dict, flags, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (setxattr, frame, -1, op_errno, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_fsetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "%s occurred while "
+                        "creating symlinks ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (fsetxattr, frame, op_ret, op_errno, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *dict,
+                  int32_t flags, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        ret = call_from_sp_client_to_reset_tmfile (frame, this, dict);
+        if (ret == 0)
+                return 0;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = marker_inode_loc_fill (fd->inode, &local->loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_fsetxattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->fsetxattr, fd, dict, flags, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (fsetxattr, frame, -1, ENOMEM, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_fsetattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+                     struct iatt *statpost, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR, "%s occurred while "
+                        "creating symlinks ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (fsetattr, frame, op_ret, op_errno, statpre,
+                             statpost, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+
+int32_t
+marker_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd,
+                 struct iatt *stbuf, int32_t valid, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = marker_inode_loc_fill (fd->inode, &local->loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_fsetattr_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->fsetattr, fd, stbuf, valid, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (fsetattr, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+                    struct iatt *statpost, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, ((op_errno == ENOENT) ? GF_LOG_DEBUG :
+                                     GF_LOG_ERROR),
+                        "%s occurred during setattr of %s",
+                        strerror (op_errno),
+                        (local ? local->loc.path : "<nul>"));
+        }
+
+        STACK_UNWIND_STRICT (setattr, frame, op_ret, op_errno, statpre,
+                             statpost, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                struct iatt *stbuf, int32_t valid, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_setattr_cbk, FIRST_CHILD (this),
+                    FIRST_CHILD (this)->fops->setattr, loc, stbuf, valid, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (setattr, frame, -1, ENOMEM, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_removexattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                        int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        marker_local_t     *local   = NULL;
+        marker_conf_t      *priv    = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_ERROR, "%s occurred while "
+                        "creating symlinks ", strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (removexattr, frame, op_ret, op_errno, xdata);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_XTIME)
+                marker_xtime_update_marks (this, local);
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_removexattr (call_frame_t *frame, xlator_t *this, loc_t *loc,
+                    const char *name, dict_t *xdata)
+{
+        int32_t          ret   = 0;
+        marker_local_t  *local = NULL;
+        marker_conf_t   *priv  = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+
+        if (ret == -1)
+                goto err;
+wind:
+        STACK_WIND (frame, marker_removexattr_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->removexattr, loc, name, xdata);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (removexattr, frame, -1, ENOMEM, NULL);
+
+        return 0;
+}
+
+
+int32_t
+marker_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, inode_t *inode,
+                   struct iatt *buf, dict_t *dict, struct iatt *postparent)
+{
+        marker_conf_t  *priv    = NULL;
+        marker_local_t *local   = NULL;
+
+        if (op_ret == -1) {
+                gf_log (this->name, GF_LOG_TRACE, "lookup failed with %s",
+                        strerror (op_errno));
+        }
+
+        local = (marker_local_t *) frame->local;
+
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
+                             dict, postparent);
+
+        if (op_ret == -1 || local == NULL)
+                goto out;
+
+        /* copy the gfid from the stat structure instead of inode,
+         * since if the lookup is fresh lookup, then the inode
+         * would have not yet linked to the inode table which happens
+         * in protocol/server.
+         */
+        if (uuid_is_null (local->loc.gfid))
+                uuid_copy (local->loc.gfid, buf->ia_gfid);
+
+
+        priv = this->private;
+
+        if (priv->feature_enabled & GF_QUOTA) {
+                mq_xattr_state (this, &local->loc, dict, *buf);
+        }
+
+out:
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int32_t
+marker_lookup (call_frame_t *frame, xlator_t *this,
+               loc_t *loc, dict_t *xattr_req)
+{
+        int32_t         ret     = 0;
+        marker_local_t *local   = NULL;
+        marker_conf_t  *priv    = NULL;
+
+        priv = this->private;
+
+        if (priv->feature_enabled == 0)
+                goto wind;
+
+        local = mem_get0 (this->local_pool);
+
+        MARKER_INIT_LOCAL (frame, local);
+
+        ret = loc_copy (&local->loc, loc);
+        if (ret == -1)
+                goto err;
+
+        if ((priv->feature_enabled & GF_QUOTA) && xattr_req)
+                mq_req_xattr (this, loc, xattr_req);
+wind:
+        STACK_WIND (frame, marker_lookup_cbk, FIRST_CHILD(this),
+                    FIRST_CHILD(this)->fops->lookup, loc, xattr_req);
+        return 0;
+err:
+        STACK_UNWIND_STRICT (lookup, frame, -1, 0, NULL, NULL, NULL, NULL);
+
+        return 0;
+}
+
+
+int
+marker_build_ancestry_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                           int op_ret, int op_errno, gf_dirent_t *entries,
+                           dict_t *xdata)
+{
+        gf_dirent_t *entry = NULL;
+        loc_t        loc    = {0, };
+        inode_t     *parent = NULL;
+
+        if ((op_ret <= 0) || (entries == NULL)) {
+                goto out;
+        }
+
+
+        list_for_each_entry (entry, &entries->list, list) {
+                if (entry->inode == entry->inode->table->root) {
+                        loc.path = gf_strdup ("/");
+                        inode_unref (parent);
+                        parent = NULL;
+                }
+
+                loc.inode = inode_ref (entry->inode);
+
+                if (parent != NULL) {
+                        loc.parent = inode_ref (parent);
+                        uuid_copy (loc.pargfid, parent->gfid);
+                }
+
+                uuid_copy (loc.gfid, entry->d_stat.ia_gfid);
+
+                mq_xattr_state (this, &loc, entry->dict, entry->d_stat);
+
+                inode_unref (parent);
+                parent = inode_ref (entry->inode);
+                loc_wipe (&loc);
+        }
+
+        if (parent)
+                inode_unref (parent);
+
+out:
+        STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries, xdata);
+        return 0;
+}
+
+int
+marker_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int op_ret, int op_errno, gf_dirent_t *entries,
+                     dict_t *xdata)
+{
+        gf_dirent_t    *entry = NULL;
+        marker_conf_t  *priv  = NULL;
+        marker_local_t *local = NULL;
+        loc_t           loc   = {0, };
+
+        if (op_ret <= 0)
+                goto unwind;
+
+        priv = this->private;
+        local = frame->local;
+
+        if (!(priv->feature_enabled & GF_QUOTA) || (local == NULL)) {
+                goto unwind;
+        }
+
+        list_for_each_entry (entry, &entries->list, list) {
+                if ((strcmp (entry->d_name, ".") == 0) ||
+                    (strcmp (entry->d_name, "..") == 0))
+                        continue;
+
+                loc.inode = inode_ref (entry->inode);
+                loc.parent = inode_ref (local->loc.inode);
+
+                uuid_copy (loc.gfid, entry->d_stat.ia_gfid);
+                uuid_copy (loc.pargfid, loc.parent->gfid);
+
+                mq_xattr_state (this, &loc, entry->dict, entry->d_stat);
+
+                loc_wipe (&loc);
+        }
+
+unwind:
+        local = frame->local;
+        frame->local = NULL;
+
+        STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries, xdata);
+        marker_local_unref (local);
+
+        return 0;
+}
+
+int
+marker_readdirp (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+                 off_t offset, dict_t *dict)
+{
+        marker_conf_t  *priv  = NULL;
+        loc_t           loc   = {0, };
+        marker_local_t *local = NULL;
+
+        priv = this->private;
+
+        if ((dict != NULL) && dict_get (dict, GET_ANCESTRY_DENTRY_KEY)) {
+                STACK_WIND (frame, marker_build_ancestry_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->readdirp,
+                            fd, size, offset, dict);
+        } else {
+                if (priv->feature_enabled & GF_QUOTA) {
+                        local = mem_get0 (this->local_pool);
+
+                        MARKER_INIT_LOCAL (frame, local);
+
+                        loc.parent = local->loc.inode = inode_ref (fd->inode);
+
+                        if (dict == NULL)
+                                dict = dict_new ();
+
+                        mq_req_xattr (this, &loc, dict);
+                }
+
+                STACK_WIND (frame, marker_readdirp_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->readdirp,
+                            fd, size, offset, dict);
+        }
+
+        return 0;
+}
+
+
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_marker_mt_end + 1);
+
+        if (ret != 0) {
+                gf_log(this->name, GF_LOG_ERROR, "Memory accounting init"
+                       "failed");
+                return ret;
+        }
+
+        return ret;
+}
+
+
+int32_t
+init_xtime_priv (xlator_t *this, dict_t *options)
+{
+        data_t          *data    = NULL;
+        int32_t          ret     = -1;
+        marker_conf_t   *priv    = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+        GF_VALIDATE_OR_GOTO (this->name, options, out);
+        GF_VALIDATE_OR_GOTO (this->name, this->private, out);
+
+        priv = this->private;
+
+        if((data = dict_get (options, VOLUME_UUID)) != NULL) {
+                priv->volume_uuid = data->data;
+
+                ret = uuid_parse (priv->volume_uuid, priv->volume_uuid_bin);
+                if (ret == -1) {
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "invalid volume uuid %s", priv->volume_uuid);
+                        goto out;
+                }
+
+                ret = gf_asprintf (& (priv->marker_xattr), "%s.%s.%s",
+                                   MARKER_XATTR_PREFIX, priv->volume_uuid,
+                                   XTIME);
+
+                if (ret == -1){
+                        priv->marker_xattr = NULL;
+
+                        gf_log (this->name, GF_LOG_ERROR,
+                                "Failed to allocate memory");
+                        goto out;
+                }
+
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "the volume-uuid = %s", priv->volume_uuid);
+        } else {
+                priv->volume_uuid = NULL;
+
+                gf_log (this->name, GF_LOG_ERROR,
+                        "please specify the volume-uuid"
+                        "in the translator options");
+
+                return -1;
+        }
+
+        if ((data = dict_get (options, TIMESTAMP_FILE)) != NULL) {
+                priv->timestamp_file = data->data;
+
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "the timestamp-file is = %s",
+                        priv->timestamp_file);
+
+        } else {
+                priv->timestamp_file = NULL;
+
+                gf_log (this->name, GF_LOG_ERROR,
+                        "please specify the timestamp-file"
+                        "in the translator options");
+
+                goto out;
+        }
+
+        ret = 0;
+out:
+        return ret;
+}
+
+void
+marker_xtime_priv_cleanup (xlator_t *this)
+{
+        marker_conf_t *priv = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+
+        priv = (marker_conf_t *) this->private;
+
+        GF_VALIDATE_OR_GOTO (this->name, priv, out);
+
+        GF_FREE (priv->volume_uuid);
+
+        GF_FREE (priv->timestamp_file);
+
+        GF_FREE (priv->marker_xattr);
+out:
+        return;
+}
+
+void
+marker_priv_cleanup (xlator_t *this)
+{
+        marker_conf_t *priv = NULL;
+
+        GF_VALIDATE_OR_GOTO ("marker", this, out);
+
+        priv = (marker_conf_t *) this->private;
+
+        GF_VALIDATE_OR_GOTO (this->name, priv, out);
+
+        marker_xtime_priv_cleanup (this);
+
+        LOCK_DESTROY (&priv->lock);
+
+        GF_FREE (priv);
+out:
+        return;
+}
+
+int32_t
+reconfigure (xlator_t *this, dict_t *options)
+{
+        int32_t         ret     = 0;
+        data_t         *data    = NULL;
+        gf_boolean_t    flag    = _gf_false;
+        marker_conf_t  *priv    = NULL;
+
+        GF_ASSERT (this);
+        GF_ASSERT (this->private);
+
+        priv = this->private;
+
+        priv->feature_enabled = 0;
+
+        GF_VALIDATE_OR_GOTO (this->name, options, out);
+
+        data = dict_get (options, "quota");
+        if (data) {
+                ret = gf_string2boolean (data->data, &flag);
+                if (ret == 0 && flag == _gf_true) {
+                        ret = init_quota_priv (this);
+                        if (ret < 0) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "failed to initialize quota private");
+                        } else {
+                                priv->feature_enabled |= GF_QUOTA;
+                        }
+                }
+        }
+
+        data = dict_get (options, "xtime");
+        if (data) {
+                ret = gf_string2boolean (data->data, &flag);
+                if (ret == 0 && flag == _gf_true) {
+                        marker_xtime_priv_cleanup (this);
+
+                        ret = init_xtime_priv (this, options);
+                        if (ret < 0) {
+                                gf_log (this->name, GF_LOG_WARNING,
+                                        "failed to initialize xtime private, "
+                                        "xtime updation will fail");
+                        } else {
+                                priv->feature_enabled |= GF_XTIME;
+                                data = dict_get (options, "gsync-force-xtime");
+                                if (!data)
+                                        goto out;
+                                ret = gf_string2boolean (data->data, &flag);
+                                if (ret == 0 && flag)
+                                        priv->feature_enabled |= GF_XTIME_GSYNC_FORCE;
+                        }
+                }
+        }
+out:
+        return ret;
+}
+
+
+int32_t
+init (xlator_t *this)
+{
+        dict_t        *options = NULL;
+        data_t        *data    = NULL;
+        int32_t        ret     = 0;
+        gf_boolean_t   flag    = _gf_false;
+        marker_conf_t *priv    = NULL;
+
+        if (!this->children) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "marker translator needs subvolume defined.");
+                return -1;
+        }
+
+        if (!this->parents) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "Volume is dangling.");
+                return -1;
+        }
+
+        options = this->options;
+
+        ALLOCATE_OR_GOTO (this->private, marker_conf_t, err);
+
+        priv = this->private;
+
+        priv->feature_enabled = 0;
+
+        LOCK_INIT (&priv->lock);
+
+        data = dict_get (options, "quota");
+        if (data) {
+                ret = gf_string2boolean (data->data, &flag);
+                if (ret == 0 && flag == _gf_true) {
+                        ret = init_quota_priv (this);
+                        if (ret < 0)
+                                goto err;
+
+                        priv->feature_enabled |= GF_QUOTA;
+                }
+        }
+
+        data = dict_get (options, "xtime");
+        if (data) {
+                ret = gf_string2boolean (data->data, &flag);
+                if (ret == 0 && flag == _gf_true) {
+                        ret = init_xtime_priv (this, options);
+                        if (ret < 0)
+                                goto err;
+
+                        priv->feature_enabled |= GF_XTIME;
+                        data = dict_get (options, "gsync-force-xtime");
+                        if (!data)
+                                goto cont;
+                        ret = gf_string2boolean (data->data, &flag);
+                        if (ret == 0 && flag)
+                                priv->feature_enabled |= GF_XTIME_GSYNC_FORCE;
+                }
+        }
+
+ cont:
+        this->local_pool = mem_pool_new (marker_local_t, 128);
+        if (!this->local_pool) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "failed to create local_t's memory pool");
+                goto err;
+        }
+
+        return 0;
+err:
+        marker_priv_cleanup (this);
+
+        return -1;
+}
+
+int32_t
+marker_forget (xlator_t *this, inode_t *inode)
+{
+        marker_inode_ctx_t *ctx   = NULL;
+        uint64_t            value = 0;
+
+        if (inode_ctx_del (inode, this, &value) != 0)
+                goto out;
+
+        ctx = (marker_inode_ctx_t *)(unsigned long)value;
+        if (ctx == NULL) {
+                goto out;
+        }
+
+        mq_forget (this, ctx->quota_ctx);
+
+        GF_FREE (ctx);
+out:
+        return 0;
+}
+
+void
+fini (xlator_t *this)
+{
+        marker_priv_cleanup (this);
+}
+
+struct xlator_fops fops = {
+        .lookup      = marker_lookup,
+        .create      = marker_create,
+        .mkdir       = marker_mkdir,
+        .writev      = marker_writev,
+        .truncate    = marker_truncate,
+        .ftruncate   = marker_ftruncate,
+        .symlink     = marker_symlink,
+        .link        = marker_link,
+        .unlink      = marker_unlink,
+        .rmdir       = marker_rmdir,
+        .rename      = marker_rename,
+        .mknod       = marker_mknod,
+        .setxattr    = marker_setxattr,
+        .fsetxattr   = marker_fsetxattr,
+        .setattr     = marker_setattr,
+        .fsetattr    = marker_fsetattr,
+        .removexattr = marker_removexattr,
+        .getxattr    = marker_getxattr,
+        .readdirp    = marker_readdirp,
+	.fallocate   = marker_fallocate,
+	.discard     = marker_discard,
+        .zerofill    = marker_zerofill,
+};
+
+struct xlator_cbks cbks = {
+        .forget = marker_forget
+};
+
+struct volume_options options[] = {
+        {.key = {"volume-uuid"}},
+        {.key = {"timestamp-file"}},
+        {.key = {"quota"}},
+        {.key = {"xtime"}},
+        {.key = {"gsync-force-xtime"}},
+        {.key = {NULL}}
+};
diff --git a/xlators/features/marker/src/marker.h b/xlators/features/marker/src/marker.h
new file mode 100644
index 000000000..23d1580f0
--- /dev/null
+++ b/xlators/features/marker/src/marker.h
@@ -0,0 +1,139 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _MARKER_H
+#define _MARKER_H
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "marker-quota.h"
+#include "xlator.h"
+#include "defaults.h"
+#include "uuid.h"
+#include "call-stub.h"
+
+#define MARKER_XATTR_PREFIX "trusted.glusterfs"
+#define XTIME               "xtime"
+#define VOLUME_MARK         "volume-mark"
+#define VOLUME_UUID         "volume-uuid"
+#define TIMESTAMP_FILE      "timestamp-file"
+
+enum {
+        GF_QUOTA             = 1,
+        GF_XTIME             = 2,
+        GF_XTIME_GSYNC_FORCE = 4,
+};
+
+/*initialize the local variable*/
+#define MARKER_INIT_LOCAL(_frame,_local) do {                   \
+                _frame->local = _local;                         \
+                _local->pid = _frame->root->pid;                \
+                memset (&_local->loc, 0, sizeof (loc_t));       \
+                _local->ref = 1;                                \
+                _local->uid = -1;                               \
+                _local->gid = -1;                               \
+                LOCK_INIT (&_local->lock);                      \
+                _local->oplocal = NULL;                         \
+        } while (0)
+
+/* try alloc and if it fails, goto label */
+#define ALLOCATE_OR_GOTO(var, type, label) do {                  \
+                var = GF_CALLOC (sizeof (type), 1,               \
+                                 gf_marker_mt_##type);           \
+                if (!var) {                                      \
+                        gf_log (this->name, GF_LOG_ERROR,        \
+                                "out of memory :(");             \
+                        goto label;                              \
+                }                                                \
+        } while (0)
+
+#define _MARKER_SET_UID_GID(dest, src)          \
+        do {                                    \
+                if (src->uid != -1 &&           \
+                    src->gid != -1) {           \
+                        dest->uid = src->uid;   \
+                        dest->gid = src->gid;   \
+                }                               \
+        } while (0)
+
+#define MARKER_SET_UID_GID(frame, dest, src)                    \
+        do {                                                    \
+                _MARKER_SET_UID_GID (dest, src);                \
+                frame->root->uid = 0;                           \
+                frame->root->gid = 0;                           \
+                frame->cookie = (void *) _GF_UID_GID_CHANGED;   \
+        } while (0)
+
+#define MARKER_RESET_UID_GID(frame, dest, src)                  \
+        do {                                                    \
+                _MARKER_SET_UID_GID (dest, src);                \
+                frame->cookie = NULL;                           \
+        } while (0)
+
+struct marker_local{
+        uint32_t        timebuf[2];
+        pid_t           pid;
+        loc_t           loc;
+        loc_t           parent_loc;
+        loc_t          *next_lock_on;
+        uid_t           uid;
+        gid_t           gid;
+        int32_t         ref;
+        int32_t         ia_nlink;
+        gf_lock_t       lock;
+        mode_t          mode;
+        int32_t         err;
+        call_stub_t    *stub;
+        int64_t         contribution;
+        struct marker_local *oplocal;
+
+        /* marker quota specific */
+        int64_t delta;
+        int64_t d_off;
+        int64_t sum;
+        int64_t size;
+        int32_t hl_count;
+        int32_t dentry_child_count;
+
+        fd_t         *fd;
+        call_frame_t *frame;
+
+        quota_inode_ctx_t    *ctx;
+        inode_contribution_t *contri;
+
+        int xflag;
+        dict_t *xdata;
+        gf_boolean_t skip_txn;
+};
+typedef struct marker_local marker_local_t;
+
+#define quota_local_t marker_local_t
+
+struct marker_inode_ctx {
+        struct quota_inode_ctx *quota_ctx;
+};
+typedef struct marker_inode_ctx marker_inode_ctx_t;
+
+struct marker_conf{
+        char         feature_enabled;
+        char        *size_key;
+        char        *dirty_key;
+        char        *volume_uuid;
+        uuid_t      volume_uuid_bin;
+        char        *timestamp_file;
+        char        *marker_xattr;
+        uint64_t     quota_lk_owner;
+        gf_lock_t    lock;
+};
+typedef struct marker_conf marker_conf_t;
+
+#endif
diff --git a/xlators/features/path-convertor/src/Makefile.am b/xlators/features/path-convertor/src/Makefile.am
index 58cfed0f9..393a7bd08 100644
--- a/xlators/features/path-convertor/src/Makefile.am
+++ b/xlators/features/path-convertor/src/Makefile.am
@@ -2,13 +2,14 @@
 xlator_LTLIBRARIES = path-converter.la
 xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/testing/features
 
-path_converter_la_LDFLAGS = -module -avoidversion 
+path_converter_la_LDFLAGS = -module -avoid-version 
 
 path_converter_la_SOURCES = path.c
 path_converter_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la 
 
-AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -D$(GF_HOST_OS)\
-	-I$(top_srcdir)/libglusterfs/src -shared -nostartfiles $(GF_CFLAGS)
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
 
 CLEANFILES = 
 
diff --git a/xlators/features/path-convertor/src/path-mem-types.h b/xlators/features/path-convertor/src/path-mem-types.h
new file mode 100644
index 000000000..77ada8d53
--- /dev/null
+++ b/xlators/features/path-convertor/src/path-mem-types.h
@@ -0,0 +1,22 @@
+/*
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef __PATH_MEM_TYPES_H__
+#define __PATH_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_path_mem_types_ {
+        gf_path_mt_path_private_t = gf_common_mt_end + 1,
+        gf_path_mt_char,
+        gf_path_mt_regex_t,
+        gf_path_mt_end
+};
+#endif
+
diff --git a/xlators/features/path-convertor/src/path.c b/xlators/features/path-convertor/src/path.c
index d58f0f3cb..5c52e0a8d 100644
--- a/xlators/features/path-convertor/src/path.c
+++ b/xlators/features/path-convertor/src/path.c
@@ -1,22 +1,12 @@
 /*
-  Copyright (c) 2008-2009 Gluster, Inc. <http://www.gluster.com>
-  This file is part of GlusterFS.
-
-  GlusterFS is free software; you can redistribute it and/or modify
-  it under the terms of the GNU General Public License as published
-  by the Free Software Foundation; either version 3 of the License,
-  or (at your option) any later version.
-
-  GlusterFS is distributed in the hope that it will be useful, but
-  WITHOUT ANY WARRANTY; without even the implied warranty of
-  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
-  General Public License for more details.
-
-  You should have received a copy of the GNU General Public License
-  along with this program.  If not, see
-  <http://www.gnu.org/licenses/>.
-*/
+   Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
 
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
 /* TODO: add gf_log to all the cases returning errors */
 
 #ifndef _CONFIG_H
@@ -35,6 +25,7 @@
 #include <errno.h>
 #include "glusterfs.h"
 #include "xlator.h"
+#include "path-mem-types.h"
 
 typedef struct path_private
 {
@@ -51,7 +42,7 @@ static char *
 name_this_to_that (xlator_t *xl, const char *path, const char *name)
 {
 	path_private_t *priv = xl->private;
-	char priv_path[ZR_PATH_MAX] = {0,};
+	char priv_path[PATH_MAX] = {0,};
 	char *tmp_name = NULL;
 	int32_t path_len = strlen (path);
 	int32_t name_len = strlen (name) - ZR_FILE_CONTENT_STRLEN;
@@ -63,7 +54,9 @@ name_this_to_that (xlator_t *xl, const char *path, const char *name)
 
 	if (priv->end_off && (total_len > priv->end_off)) {
 		j = priv->start_off;
-		tmp_name = CALLOC (1, (total_len + ZR_FILE_CONTENT_STRLEN));
+		tmp_name = GF_CALLOC (1, (total_len +
+                                          ZR_FILE_CONTENT_STRLEN),
+                                      gf_path_mt_char);
 		ERR_ABORT (tmp_name);
 
 		/* Get the complete path for the file first */
@@ -104,7 +97,7 @@ path_this_to_that (xlator_t *xl, const char *path)
 	int32_t i = 0, j = 0;
 
 	if (priv->end_off && (path_len > priv->start_off)) {
-		priv_path = CALLOC (1, path_len);
+		priv_path = GF_CALLOC (1, path_len, gf_path_mt_char);
 		ERR_ABORT (priv_path);
 
 		if (priv->start_off && (path_len > priv->start_off))
@@ -134,9 +127,9 @@ path_create_cbk (call_frame_t *frame,
 		 int32_t op_errno,
 		 fd_t *fd,
 		 inode_t *inode,
-		 struct stat *buf,
-                 struct stat *preparent,
-                 struct stat *postparent)
+		 struct iatt *buf,
+                 struct iatt *preparent,
+                 struct iatt *postparent)
 {
 	STACK_UNWIND (frame, op_ret, op_errno, fd, inode, buf);
 	return 0;
@@ -187,7 +180,7 @@ path_readlink_cbk (call_frame_t *frame,
 		   int32_t op_ret,
 		   int32_t op_errno,
 		   const char *buf,
-                   struct stat *sbuf)
+                   struct iatt *sbuf)
 {
 	STACK_UNWIND (frame, op_ret, op_errno, buf, sbuf);
 	return 0;
@@ -200,9 +193,9 @@ path_lookup_cbk (call_frame_t *frame,
 		 int32_t op_ret,
 		 int32_t op_errno,
 		 inode_t *inode,
-		 struct stat *buf,
+		 struct iatt *buf,
 		 dict_t *xattr,
-                 struct stat *postparent)
+                 struct iatt *postparent)
 {
 	STACK_UNWIND (frame, op_ret, op_errno, inode, buf, xattr);
 	return 0;
@@ -216,9 +209,9 @@ path_symlink_cbk (call_frame_t *frame,
 		  int32_t op_ret,
 		  int32_t op_errno,
 		  inode_t *inode,
-                  struct stat *buf,
-                  struct stat *preparent,
-                  struct stat *postparent)
+                  struct iatt *buf,
+                  struct iatt *preparent,
+                  struct iatt *postparent)
 {
 	STACK_UNWIND (frame, op_ret, op_errno, inode, buf);
 	return 0;
@@ -231,9 +224,9 @@ path_mknod_cbk (call_frame_t *frame,
 		int32_t op_ret,
 		int32_t op_errno,
 		inode_t *inode,
-                struct stat *buf,
-                struct stat *preparent,
-                struct stat *postparent)
+                struct iatt *buf,
+                struct iatt *preparent,
+                struct iatt *postparent)
 {
 	STACK_UNWIND (frame, op_ret, op_errno, inode, buf);
 	return 0;
@@ -247,9 +240,9 @@ path_mkdir_cbk (call_frame_t *frame,
 		int32_t op_ret,
 		int32_t op_errno,
 		inode_t *inode,
-                struct stat *buf,
-                struct stat *preparent,
-                struct stat *postparent)
+                struct iatt *buf,
+                struct iatt *preparent,
+                struct iatt *postparent)
 {
 	STACK_UNWIND (frame, op_ret, op_errno, inode, buf);
 	return 0;
@@ -262,9 +255,9 @@ path_link_cbk (call_frame_t *frame,
 	       int32_t op_ret,
 	       int32_t op_errno,
 	       inode_t *inode,
-               struct stat *buf,
-               struct stat *preparent,
-               struct stat *postparent)
+               struct iatt *buf,
+               struct iatt *preparent,
+               struct iatt *postparent)
 {
 	STACK_UNWIND (frame, op_ret, op_errno, inode, buf);
 	return 0;
@@ -289,11 +282,11 @@ path_rename_buf_cbk (call_frame_t *frame,
 		     xlator_t *this,
 		     int32_t op_ret,
 		     int32_t op_errno,
-		     struct stat *buf,
-                     struct stat *preoldparent,
-                     struct stat *postoldparent,
-                     struct stat *prenewparent,
-                     struct stat *postnewparent)
+		     struct iatt *buf,
+                     struct iatt *preoldparent,
+                     struct iatt *postoldparent,
+                     struct iatt *prenewparent,
+                     struct iatt *postnewparent)
 {
 	STACK_UNWIND (frame, op_ret, op_errno, buf);
 	return 0;
@@ -307,7 +300,7 @@ path_common_buf_cbk (call_frame_t *frame,
 		     xlator_t *this,
 		     int32_t op_ret,
 		     int32_t op_errno,
-		     struct stat *buf)
+		     struct iatt *buf)
 {
 	STACK_UNWIND (frame, op_ret, op_errno, buf);
 	return 0;
@@ -327,8 +320,8 @@ path_common_dict_cbk (call_frame_t *frame,
 
 int32_t 
 path_common_remove_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                        int32_t op_ret, int32_t op_errno,struct stat *preparent,
-                        struct stat *postparent)
+                        int32_t op_ret, int32_t op_errno,struct iatt *preparent,
+                        struct iatt *postparent)
 {
 	STACK_UNWIND (frame, op_ret, op_errno);
 	return 0;
@@ -336,8 +329,8 @@ path_common_remove_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
 
 int32_t
 path_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
-                   int32_t op_ret, int32_t op_errno,struct stat *prebuf,
-                   struct stat *postbuf)
+                   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;
@@ -378,7 +371,7 @@ path_lookup (call_frame_t *frame,
 
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
@@ -405,7 +398,7 @@ path_stat (call_frame_t *frame,
   
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
@@ -434,7 +427,7 @@ path_readlink (call_frame_t *frame,
   
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
@@ -465,7 +458,7 @@ path_mknod (call_frame_t *frame,
 
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
@@ -494,7 +487,7 @@ path_mkdir (call_frame_t *frame,
 
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
@@ -521,7 +514,7 @@ path_unlink (call_frame_t *frame,
 
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
@@ -548,7 +541,7 @@ path_rmdir (call_frame_t *frame,
   
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
@@ -577,7 +570,7 @@ path_symlink (call_frame_t *frame,
   
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
@@ -615,11 +608,11 @@ path_rename (call_frame_t *frame,
   
 	oldloc->path = oldloc_path;	
 	if (tmp_oldloc_path != oldloc_path)
-		FREE (tmp_oldloc_path);
+		GF_FREE (tmp_oldloc_path);
 
 	newloc->path = newloc_path;	
 	if (tmp_newloc_path != newloc_path)
-		FREE (tmp_newloc_path);
+		GF_FREE (tmp_newloc_path);
 
 	return 0;
 }
@@ -657,11 +650,11 @@ path_link (call_frame_t *frame,
 
 	oldloc->path = oldloc_path;	
 	if (tmp_oldloc_path != oldloc_path)
-		FREE (tmp_oldloc_path);
+		GF_FREE (tmp_oldloc_path);
 
 	newloc->path = newloc_path;	
 	if (tmp_newloc_path != newloc_path)
-		FREE (tmp_newloc_path);
+		GF_FREE (tmp_newloc_path);
 
 	return 0;
 }
@@ -672,8 +665,8 @@ path_setattr_cbk (call_frame_t *frame,
                   xlator_t *this,
                   int32_t op_ret,
                   int32_t op_errno,
-                  struct stat *preop,
-                  struct stat *postop)
+                  struct iatt *preop,
+                  struct iatt *postop)
 {
 	STACK_UNWIND (frame, op_ret, op_errno, preop, postop);
 	return 0;
@@ -683,7 +676,7 @@ int32_t
 path_setattr (call_frame_t *frame,
               xlator_t *this,
               loc_t *loc,
-              struct stat *stbuf,
+              struct iatt *stbuf,
               int32_t valid)
 {
 	char *loc_path = (char *)loc->path;
@@ -704,7 +697,7 @@ path_setattr (call_frame_t *frame,
 
 	loc->path = loc_path;
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
@@ -734,7 +727,7 @@ path_truncate (call_frame_t *frame,
   
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
@@ -768,7 +761,7 @@ path_open (call_frame_t *frame,
 
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
@@ -801,7 +794,7 @@ path_create (call_frame_t *frame,
 
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
@@ -843,10 +836,9 @@ path_setxattr (call_frame_t *frame,
 
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
-	if (tmp_name)
-		FREE (tmp_name);
+	GF_FREE (tmp_name);
 
 	return 0;
 }
@@ -880,10 +872,10 @@ path_getxattr (call_frame_t *frame,
 
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	if (tmp_name != name)
-		FREE (tmp_name);
+		GF_FREE (tmp_name);
 
 	return 0;
 }
@@ -917,10 +909,10 @@ path_removexattr (call_frame_t *frame,
 
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	if (tmp_name != name)
-		FREE (tmp_name);
+		GF_FREE (tmp_name);
 
 	return 0;
 }
@@ -949,7 +941,7 @@ path_opendir (call_frame_t *frame,
 
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
@@ -978,7 +970,7 @@ path_access (call_frame_t *frame,
 
 	loc->path = loc_path;
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
@@ -1020,7 +1012,7 @@ path_checksum (call_frame_t *frame,
 
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
@@ -1047,14 +1039,14 @@ path_entrylk (call_frame_t *frame, xlator_t *this,
 
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
 
 int32_t
 path_inodelk (call_frame_t *frame, xlator_t *this,
-              const char *volume, loc_t *loc, int32_t cmd, struct flock *lock)
+              const char *volume, loc_t *loc, int32_t cmd, struct gf_flock *lock)
 {
 	char *loc_path = (char *)loc->path;
 	char *tmp_path = NULL;
@@ -1073,7 +1065,7 @@ path_inodelk (call_frame_t *frame, xlator_t *this,
 
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
@@ -1105,11 +1097,29 @@ path_xattrop (call_frame_t *frame,
 
 	loc->path = loc_path;	
 	if (tmp_path != loc_path)
-		FREE (tmp_path);
+		GF_FREE (tmp_path);
 
 	return 0;
 }
 
+int32_t
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        if (!this)
+                return ret;
+
+        ret = xlator_mem_acct_init (this, gf_path_mt_end + 1);
+        
+        if (ret != 0) {
+                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init"
+                        "failed");
+                return ret;
+        }
+
+        return ret;
+}
 
 int32_t 
 init (xlator_t *this)
@@ -1128,7 +1138,7 @@ init (xlator_t *this)
 			"dangling volume. check volfile ");
 	}
   
-	priv = CALLOC (1, sizeof (*priv));
+	priv = GF_CALLOC (1, sizeof (*priv), gf_path_mt_path_private_t);
 	ERR_ABORT (priv);
 	if (dict_get (options, "start-offset")) {
 		priv->start_off = data_to_int32 (dict_get (options, 
@@ -1141,7 +1151,8 @@ init (xlator_t *this)
 
 	if (dict_get (options, "regex")) {
 		int32_t ret = 0;
-		priv->preg = CALLOC (1, sizeof (regex_t));
+		priv->preg = GF_CALLOC (1, sizeof (regex_t),
+                                        gf_path_mt_regex_t);
 		ERR_ABORT (priv->preg);
 		ret = regcomp (priv->preg, 
 			       data_to_str (dict_get (options, "regex")), 
@@ -1149,7 +1160,7 @@ init (xlator_t *this)
 		if (ret) {
 			gf_log (this->name, GF_LOG_ERROR, 
 				"Failed to compile the 'option regex'");
-			FREE (priv);
+			GF_FREE (priv);
 			return -1;
 		}
 		if (dict_get (options, "replace-with")) {
@@ -1196,11 +1207,6 @@ struct xlator_fops fops = {
         .setattr     = path_setattr,
 };
 
-
-struct xlator_mops mops = {
-};
-
-
 struct xlator_cbks cbks = {
 };
 
diff --git a/xlators/features/protect/Makefile.am b/xlators/features/protect/Makefile.am
new file mode 100644
index 000000000..d471a3f92
--- /dev/null
+++ b/xlators/features/protect/Makefile.am
@@ -0,0 +1,3 @@
+SUBDIRS = src
+
+CLEANFILES = 
diff --git a/xlators/features/protect/src/Makefile.am b/xlators/features/protect/src/Makefile.am
new file mode 100644
index 000000000..968e88c45
--- /dev/null
+++ b/xlators/features/protect/src/Makefile.am
@@ -0,0 +1,21 @@
+xlator_LTLIBRARIES = prot_dht.la prot_client.la prot_server.la
+
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+prot_dht_la_LDFLAGS = -module -avoid-version
+prot_dht_la_SOURCES = prot_dht.c
+prot_dht_la_LIBADD  = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+prot_client_la_LDFLAGS = -module -avoid-version
+prot_client_la_SOURCES = prot_client.c
+prot_client_la_LIBADD  = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+prot_server_la_LDFLAGS = -module -avoid-version
+prot_server_la_SOURCES = prot_server.c
+prot_server_la_LIBADD  = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+CLEANFILES = 
+
diff --git a/xlators/features/protect/src/prot_client.c b/xlators/features/protect/src/prot_client.c
new file mode 100644
index 000000000..d09715067
--- /dev/null
+++ b/xlators/features/protect/src/prot_client.c
@@ -0,0 +1,217 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+
+#ifndef __NetBSD__
+#include <execinfo.h>
+#endif
+
+#define NUM_FRAMES 20
+
+static char PROTECT_KEY[] = "trusted.glusterfs.protect";
+
+enum {
+        PROT_ACT_NONE = 0,
+        PROT_ACT_LOG,
+        PROT_ACT_REJECT,
+};
+
+void
+pcli_print_trace (char *name, call_frame_t *frame)
+{
+        void    *frames[NUM_FRAMES];
+        char    **symbols;
+        int     size;
+        int     i;
+
+        gf_log (name, GF_LOG_INFO, "Translator stack:");
+        while (frame) {
+                gf_log (name, GF_LOG_INFO, "%s (%s)",
+                        frame->wind_from, frame->this->name);
+                frame = frame->next;
+        }
+
+        size = backtrace(frames,NUM_FRAMES);
+        if (size <= 0) {
+                return;
+        }
+        symbols = backtrace_symbols(frames,size);
+        if (!symbols) {
+                return;
+        }
+
+        gf_log(name, GF_LOG_INFO, "Processor stack:");
+        for (i = 0; i < size; ++i) {
+                gf_log (name, GF_LOG_INFO, "%s", symbols[i]);
+        }
+        free(symbols);
+}
+
+int32_t
+pcli_rename (call_frame_t *frame, xlator_t *this, loc_t *oldloc,
+                loc_t *newloc, dict_t *xdata)
+{
+        uint64_t        value;
+
+        if (newloc->parent == oldloc->parent) {
+                gf_log (this->name, GF_LOG_DEBUG, "rename in same directory");
+                goto simple_unwind;
+        }
+        if (!oldloc->parent) {
+                goto simple_unwind;
+        }
+        if (inode_ctx_get(oldloc->parent,this,&value) != 0) {
+                goto simple_unwind;
+        }
+
+        if (value != PROT_ACT_NONE) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "got rename for protected %s", oldloc->path);
+                pcli_print_trace(this->name,frame->next);
+                if (value == PROT_ACT_REJECT) {
+                        STACK_UNWIND_STRICT (rename, frame, -1, EPERM,
+                                             NULL, NULL, NULL, NULL, NULL,
+                                             xdata);
+                        return 0;
+                }
+        }
+
+simple_unwind:
+        STACK_WIND_TAIL (frame, FIRST_CHILD(this),
+                         FIRST_CHILD(this)->fops->rename, oldloc, newloc,
+                         xdata);
+        return 0;
+}
+
+int32_t
+pcli_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+               int32_t flags, dict_t *xdata)
+{
+        data_t          *data;
+        uint64_t        value;
+
+        /*
+         * We can't use dict_get_str and strcmp here, because the value comes
+         * directly from the user and might not be NUL-terminated (it would
+         * be if we had set it ourselves.
+         */
+
+        data = dict_get(dict,PROTECT_KEY);
+        if (!data) {
+                goto simple_wind;
+        }
+
+        if (dict->count > 1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "attempted to mix %s with other keys", PROTECT_KEY);
+                goto simple_wind;
+        }
+
+        gf_log (this->name, GF_LOG_DEBUG, "got %s request", PROTECT_KEY);
+        if (!strncmp(data->data,"log",data->len)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "logging removals on %s", loc->path);
+                value = PROT_ACT_LOG;
+        }
+        else if (!strncmp(data->data,"reject",data->len)) {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "rejecting removals on %s", loc->path);
+                value = PROT_ACT_REJECT;
+        }
+        else {
+                gf_log (this->name, GF_LOG_DEBUG,
+                        "removing protection on %s", loc->path);
+                value = PROT_ACT_NONE;
+        }
+        /* Right now the value doesn't matter - just the presence. */
+        if (inode_ctx_set(loc->inode,this,&value) != 0) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "failed to set protection status for %s", loc->path);
+        }
+        STACK_UNWIND_STRICT (setxattr, frame, 0, 0, NULL);
+        return 0;
+
+simple_wind:
+        STACK_WIND_TAIL (frame,
+                         FIRST_CHILD(this), FIRST_CHILD(this)->fops->setxattr,
+                         loc, dict, flags, xdata);
+        return 0;
+}
+
+int32_t
+pcli_unlink (call_frame_t *frame, xlator_t *this, loc_t *loc, int xflag,
+             dict_t *xdata)
+{
+        uint64_t        value;
+
+        if (!loc->parent || (inode_ctx_get(loc->parent,this,&value) != 0)) {
+                goto simple_unwind;
+        }
+
+        if (value != PROT_ACT_NONE) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "got unlink for protected %s", loc->path);
+                pcli_print_trace(this->name,frame->next);
+                if (value == PROT_ACT_REJECT) {
+                        STACK_UNWIND_STRICT (unlink, frame, -1, EPERM,
+                                             NULL, NULL, NULL);
+                        return 0;
+                }
+        }
+
+simple_unwind:
+        STACK_WIND_TAIL (frame, FIRST_CHILD(this),
+                         FIRST_CHILD(this)->fops->unlink, loc, xflag, xdata);
+        return 0;
+}
+
+int32_t
+init (xlator_t *this)
+{
+	if (!this->children || this->children->next) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"translator not configured with exactly one child");
+		return -1;
+	}
+
+	if (!this->parents) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"dangling volume. check volfile ");
+	}
+
+	return 0;
+}
+
+
+void
+fini (xlator_t *this)
+{
+	return;
+}
+
+
+struct xlator_fops fops = {
+        .rename         = pcli_rename,
+        .setxattr       = pcli_setxattr,
+        .unlink         = pcli_unlink,
+};
+
+struct xlator_cbks cbks = {
+};
+
+struct volume_options options[] = {
+	{ .key = {NULL} },
+};
diff --git a/xlators/features/protect/src/prot_dht.c b/xlators/features/protect/src/prot_dht.c
new file mode 100644
index 000000000..feec6ffd6
--- /dev/null
+++ b/xlators/features/protect/src/prot_dht.c
@@ -0,0 +1,168 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+
+enum gf_pdht_mem_types_ {
+        gf_pdht_mt_coord_t = gf_common_mt_end + 1,
+        gf_pdht_mt_end
+};
+
+typedef struct {
+        pthread_mutex_t         lock;
+        uint16_t                refs;
+        int32_t                 op_ret;
+        int32_t                 op_errno;
+        dict_t                  *xdata;
+} pdht_coord_t;
+
+static char PROTECT_KEY[] = "trusted.glusterfs.protect";
+
+void
+pdht_unref_and_unlock (call_frame_t *frame, xlator_t *this,
+                       pdht_coord_t *coord)
+{
+        gf_boolean_t    should_unwind;
+
+        should_unwind = (--(coord->refs) == 0);
+        pthread_mutex_unlock(&coord->lock);
+
+        if (should_unwind) {
+                STACK_UNWIND_STRICT (setxattr, frame, 
+                                     coord->op_ret, coord->op_errno,
+                                     coord->xdata);
+                if (coord->xdata) {
+                        dict_unref(coord->xdata);
+                }
+                GF_FREE(coord);
+        }
+}
+
+int32_t
+pdht_recurse_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        pdht_coord_t    *coord = cookie;
+
+        pthread_mutex_lock(&coord->lock);
+        if (op_ret) {
+                coord->op_ret = op_ret;
+                coord->op_errno = op_errno;
+        }
+        if (xdata) {
+                if (coord->xdata) {
+                        dict_unref(coord->xdata);
+                }
+                coord->xdata = dict_ref(xdata);
+        }
+        pdht_unref_and_unlock(frame,this,coord);
+                
+        return 0;
+}
+
+void
+pdht_recurse (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+              int32_t flags, dict_t *xdata, xlator_t *xl, pdht_coord_t *coord)
+{
+        xlator_list_t   *iter;
+
+        if (!strcmp(xl->type,"features/prot_client")) {
+                pthread_mutex_lock(&coord->lock);
+                ++(coord->refs);
+                pthread_mutex_unlock(&coord->lock);
+                STACK_WIND_COOKIE (frame, pdht_recurse_cbk, coord, xl,
+                                   xl->fops->setxattr, loc, dict, flags, xdata);
+        }
+
+        else for (iter = xl->children; iter; iter = iter->next) {
+                pdht_recurse (frame, this, loc, dict, flags, xdata,
+                              iter->xlator, coord);
+        }
+}
+
+int32_t
+pdht_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *dict,
+               int32_t flags, dict_t *xdata)
+{
+        pdht_coord_t    *coord;
+
+        if (!dict_get(dict,PROTECT_KEY)) {
+                goto simple_wind;
+        }
+
+        if (dict->count > 1) {
+                gf_log (this->name, GF_LOG_WARNING,
+                        "attempted to mix %s with other keys", PROTECT_KEY);
+                goto simple_wind;
+        }
+
+        coord = GF_CALLOC(1,sizeof(*coord),gf_pdht_mt_coord_t);
+        if (!coord) {
+                gf_log (this->name, GF_LOG_WARNING, "allocation failed");
+                goto simple_wind;
+        }
+
+        pthread_mutex_init(&coord->lock,NULL);
+        coord->refs = 1;
+        coord->op_ret = 0;
+        coord->xdata = NULL;
+
+        pdht_recurse(frame,this,loc,dict,flags,xdata,this,coord);
+        pthread_mutex_lock(&coord->lock);
+        pdht_unref_and_unlock(frame,this,coord);
+
+        return 0;
+
+simple_wind:
+        STACK_WIND_TAIL (frame,
+                         FIRST_CHILD(this), FIRST_CHILD(this)->fops->setxattr,
+                         loc, dict, flags, xdata);
+        return 0;
+}
+
+int32_t
+init (xlator_t *this)
+{
+	if (!this->children || this->children->next) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"translator not configured with exactly one child");
+		return -1;
+	}
+
+	if (!this->parents) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"dangling volume. check volfile ");
+	}
+
+	return 0;
+}
+
+
+void
+fini (xlator_t *this)
+{
+	return;
+}
+
+struct xlator_fops fops = {
+        .setxattr = pdht_setxattr,
+};
+
+struct xlator_cbks cbks = {
+};
+
+struct volume_options options[] = {
+	{ .key = {NULL} },
+};
diff --git a/xlators/features/protect/src/prot_server.c b/xlators/features/protect/src/prot_server.c
new file mode 100644
index 000000000..beaee0889
--- /dev/null
+++ b/xlators/features/protect/src/prot_server.c
@@ -0,0 +1,51 @@
+/*
+   Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "xlator.h"
+#include "defaults.h"
+
+int32_t
+init (xlator_t *this)
+{
+	if (!this->children || this->children->next) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"translator not configured with exactly one child");
+		return -1;
+	}
+
+	if (!this->parents) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"dangling volume. check volfile ");
+	}
+
+	return 0;
+}
+
+
+void
+fini (xlator_t *this)
+{
+	return;
+}
+
+
+struct xlator_fops fops = {
+};
+
+struct xlator_cbks cbks = {
+};
+
+struct volume_options options[] = {
+	{ .key = {NULL} },
+};
diff --git a/xlators/performance/stat-prefetch/Makefile.am b/xlators/features/qemu-block/Makefile.am
index af437a64d..af437a64d 100644
--- a/xlators/performance/stat-prefetch/Makefile.am
+++ b/xlators/features/qemu-block/Makefile.am
diff --git a/xlators/features/qemu-block/src/Makefile.am b/xlators/features/qemu-block/src/Makefile.am
new file mode 100644
index 000000000..08a7b62a0
--- /dev/null
+++ b/xlators/features/qemu-block/src/Makefile.am
@@ -0,0 +1,155 @@
+if ENABLE_QEMU_BLOCK
+xlator_LTLIBRARIES = qemu-block.la
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+qemu_block_la_LDFLAGS = -module -avoid-version
+qemu_block_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la $(GLIB_LIBS) -lz -lrt
+
+qemu_block_la_SOURCES_qemu = \
+	$(CONTRIBDIR)/qemu/qemu-coroutine.c \
+	$(CONTRIBDIR)/qemu/qemu-coroutine-lock.c \
+	$(CONTRIBDIR)/qemu/qemu-coroutine-sleep.c \
+	$(CONTRIBDIR)/qemu/coroutine-ucontext.c \
+	$(CONTRIBDIR)/qemu/block.c \
+	$(CONTRIBDIR)/qemu/nop-symbols.c
+
+qemu_block_la_SOURCES_qemu_util = \
+	$(CONTRIBDIR)/qemu/util/aes.c \
+	$(CONTRIBDIR)/qemu/util/bitmap.c \
+	$(CONTRIBDIR)/qemu/util/bitops.c \
+	$(CONTRIBDIR)/qemu/util/cutils.c \
+	$(CONTRIBDIR)/qemu/util/error.c \
+	$(CONTRIBDIR)/qemu/util/hbitmap.c \
+	$(CONTRIBDIR)/qemu/util/iov.c \
+	$(CONTRIBDIR)/qemu/util/module.c \
+	$(CONTRIBDIR)/qemu/util/oslib-posix.c \
+	$(CONTRIBDIR)/qemu/util/qemu-option.c \
+	$(CONTRIBDIR)/qemu/util/qemu-error.c \
+	$(CONTRIBDIR)/qemu/util/qemu-thread-posix.c \
+	$(CONTRIBDIR)/qemu/util/unicode.c \
+	$(CONTRIBDIR)/qemu/util/hexdump.c
+
+qemu_block_la_SOURCES_qemu_block = \
+	$(CONTRIBDIR)/qemu/block/snapshot.c \
+	$(CONTRIBDIR)/qemu/block/qcow2-cache.c \
+	$(CONTRIBDIR)/qemu/block/qcow2-cluster.c \
+	$(CONTRIBDIR)/qemu/block/qcow2-refcount.c \
+	$(CONTRIBDIR)/qemu/block/qcow2-snapshot.c \
+	$(CONTRIBDIR)/qemu/block/qcow2.c \
+	$(CONTRIBDIR)/qemu/block/qed-check.c \
+	$(CONTRIBDIR)/qemu/block/qed-cluster.c \
+	$(CONTRIBDIR)/qemu/block/qed-gencb.c \
+	$(CONTRIBDIR)/qemu/block/qed-l2-cache.c \
+	$(CONTRIBDIR)/qemu/block/qed-table.c \
+	$(CONTRIBDIR)/qemu/block/qed.c
+
+qemu_block_la_SOURCES_qemu_qobject = \
+	$(CONTRIBDIR)/qemu/qobject/json-lexer.c \
+	$(CONTRIBDIR)/qemu/qobject/json-parser.c \
+	$(CONTRIBDIR)/qemu/qobject/json-streamer.c \
+	$(CONTRIBDIR)/qemu/qobject/qbool.c \
+	$(CONTRIBDIR)/qemu/qobject/qdict.c \
+	$(CONTRIBDIR)/qemu/qobject/qerror.c \
+	$(CONTRIBDIR)/qemu/qobject/qfloat.c \
+	$(CONTRIBDIR)/qemu/qobject/qint.c \
+	$(CONTRIBDIR)/qemu/qobject/qjson.c \
+	$(CONTRIBDIR)/qemu/qobject/qlist.c \
+	$(CONTRIBDIR)/qemu/qobject/qstring.c
+
+qemu_block_la_SOURCES = \
+	$(qemu_block_la_SOURCES_qemu) \
+	$(qemu_block_la_SOURCES_qemu_util) \
+	$(qemu_block_la_SOURCES_qemu_block) \
+	$(qemu_block_la_SOURCES_qemu_qobject) \
+	bdrv-xlator.c \
+	coroutine-synctask.c \
+	bh-syncop.c \
+	monitor-logging.c \
+	clock-timer.c \
+	qemu-block.c \
+	qb-coroutines.c
+
+noinst_HEADERS_qemu = \
+	$(CONTRIBDIR)/qemu/config-host.h \
+	$(CONTRIBDIR)/qemu/qapi-types.h \
+	$(CONTRIBDIR)/qemu/qmp-commands.h \
+	$(CONTRIBDIR)/qemu/trace/generated-tracers.h \
+	$(CONTRIBDIR)/qemu/include/config.h \
+	$(CONTRIBDIR)/qemu/include/glib-compat.h \
+	$(CONTRIBDIR)/qemu/include/qemu-common.h \
+	$(CONTRIBDIR)/qemu/include/trace.h \
+	$(CONTRIBDIR)/qemu/include/block/coroutine.h \
+	$(CONTRIBDIR)/qemu/include/block/aio.h \
+	$(CONTRIBDIR)/qemu/include/block/block.h \
+	$(CONTRIBDIR)/qemu/include/block/block_int.h \
+	$(CONTRIBDIR)/qemu/include/block/blockjob.h \
+	$(CONTRIBDIR)/qemu/include/block/coroutine.h \
+	$(CONTRIBDIR)/qemu/include/block/coroutine_int.h \
+	$(CONTRIBDIR)/qemu/include/block/snapshot.h \
+	$(CONTRIBDIR)/qemu/include/exec/cpu-common.h \
+	$(CONTRIBDIR)/qemu/include/exec/hwaddr.h \
+	$(CONTRIBDIR)/qemu/include/exec/poison.h \
+	$(CONTRIBDIR)/qemu/include/fpu/softfloat.h \
+	$(CONTRIBDIR)/qemu/include/migration/migration.h \
+	$(CONTRIBDIR)/qemu/include/migration/qemu-file.h \
+	$(CONTRIBDIR)/qemu/include/migration/vmstate.h \
+	$(CONTRIBDIR)/qemu/include/monitor/monitor.h \
+	$(CONTRIBDIR)/qemu/include/monitor/readline.h \
+	$(CONTRIBDIR)/qemu/include/qapi/error.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/json-lexer.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/json-parser.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/json-streamer.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qbool.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qdict.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qerror.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qfloat.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qint.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qjson.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qlist.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qobject.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/qstring.h \
+	$(CONTRIBDIR)/qemu/include/qapi/qmp/types.h \
+	$(CONTRIBDIR)/qemu/include/qemu/aes.h \
+	$(CONTRIBDIR)/qemu/include/qemu/atomic.h \
+	$(CONTRIBDIR)/qemu/include/qemu/bitmap.h \
+	$(CONTRIBDIR)/qemu/include/qemu/bitops.h \
+	$(CONTRIBDIR)/qemu/include/qemu/bswap.h \
+	$(CONTRIBDIR)/qemu/include/qemu/compiler.h \
+	$(CONTRIBDIR)/qemu/include/qemu/error-report.h \
+	$(CONTRIBDIR)/qemu/include/qemu/event_notifier.h \
+	$(CONTRIBDIR)/qemu/include/qemu/hbitmap.h \
+	$(CONTRIBDIR)/qemu/include/qemu/host-utils.h \
+	$(CONTRIBDIR)/qemu/include/qemu/iov.h \
+	$(CONTRIBDIR)/qemu/include/qemu/main-loop.h \
+	$(CONTRIBDIR)/qemu/include/qemu/module.h \
+	$(CONTRIBDIR)/qemu/include/qemu/notify.h \
+	$(CONTRIBDIR)/qemu/include/qemu/option.h \
+	$(CONTRIBDIR)/qemu/include/qemu/option_int.h \
+	$(CONTRIBDIR)/qemu/include/qemu/osdep.h \
+	$(CONTRIBDIR)/qemu/include/qemu/queue.h \
+	$(CONTRIBDIR)/qemu/include/qemu/sockets.h \
+	$(CONTRIBDIR)/qemu/include/qemu/thread-posix.h \
+	$(CONTRIBDIR)/qemu/include/qemu/thread.h \
+	$(CONTRIBDIR)/qemu/include/qemu/timer.h \
+	$(CONTRIBDIR)/qemu/include/qemu/typedefs.h \
+	$(CONTRIBDIR)/qemu/include/sysemu/sysemu.h \
+	$(CONTRIBDIR)/qemu/include/sysemu/os-posix.h \
+	$(CONTRIBDIR)/qemu/block/qcow2.h \
+	$(CONTRIBDIR)/qemu/block/qed.h
+
+noinst_HEADERS = \
+	$(noinst_HEADERS_qemu) \
+	qemu-block.h \
+	qemu-block-memory-types.h \
+	qb-coroutines.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src \
+            -I$(CONTRIBDIR)/qemu \
+            -I$(CONTRIBDIR)/qemu/include \
+            -DGLUSTER_XLATOR
+
+AM_CFLAGS = -fno-strict-aliasing -Wall $(GF_CFLAGS) $(GLIB_CFLAGS)
+
+CLEANFILES =
+
+endif
diff --git a/xlators/features/qemu-block/src/bdrv-xlator.c b/xlators/features/qemu-block/src/bdrv-xlator.c
new file mode 100644
index 000000000..1e55b5fb7
--- /dev/null
+++ b/xlators/features/qemu-block/src/bdrv-xlator.c
@@ -0,0 +1,389 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "inode.h"
+#include "syncop.h"
+#include "qemu-block.h"
+#include "block/block_int.h"
+
+typedef struct BDRVGlusterState {
+	inode_t *inode;
+} BDRVGlusterState;
+
+static QemuOptsList runtime_opts = {
+	.name = "gluster",
+	.head = QTAILQ_HEAD_INITIALIZER(runtime_opts.head),
+	.desc = {
+		{
+			.name = "filename",
+			.type = QEMU_OPT_STRING,
+			.help = "GFID of file",
+		},
+		{ /* end of list */ }
+	},
+};
+
+inode_t *
+qb_inode_from_filename (const char *filename)
+{
+	const char *iptr = NULL;
+	inode_t *inode = NULL;
+
+	iptr = filename + 17;
+	sscanf (iptr, "%p", &inode);
+
+	return inode;
+}
+
+
+int
+qb_inode_to_filename (inode_t *inode, char *filename, int size)
+{
+	return snprintf (filename, size, "gluster://inodep:%p", inode);
+}
+
+
+static fd_t *
+fd_from_bs (BlockDriverState *bs)
+{
+	BDRVGlusterState *s = bs->opaque;
+
+	return fd_anonymous (s->inode);
+}
+
+
+static int
+qemu_gluster_open (BlockDriverState *bs, QDict *options, int bdrv_flags)
+{
+	inode_t *inode = NULL;
+	BDRVGlusterState *s = bs->opaque;
+	QemuOpts *opts = NULL;
+	Error *local_err = NULL;
+	const char *filename = NULL;
+	char gfid_str[128];
+	int ret;
+	qb_conf_t *conf = THIS->private;
+
+	opts = qemu_opts_create_nofail(&runtime_opts);
+	qemu_opts_absorb_qdict(opts, options, &local_err);
+	if (error_is_set(&local_err)) {
+		qerror_report_err(local_err);
+		error_free(local_err);
+		return -EINVAL;
+	}
+
+	filename = qemu_opt_get(opts, "filename");
+
+	/*
+	 * gfid:<gfid> format means we're opening a backing image.
+	 */
+	ret = sscanf(filename, "gluster://gfid:%s", gfid_str);
+	if (ret) {
+		loc_t loc = {0,};
+		struct iatt buf = {0,};
+		uuid_t gfid;
+
+		uuid_parse(gfid_str, gfid);
+
+		loc.inode = inode_find(conf->root_inode->table, gfid);
+		if (!loc.inode) {
+			loc.inode = inode_new(conf->root_inode->table);
+			uuid_copy(loc.inode->gfid, gfid);
+		}
+
+		uuid_copy(loc.gfid, loc.inode->gfid);
+		ret = syncop_lookup(FIRST_CHILD(THIS), &loc, NULL, &buf, NULL,
+				    NULL);
+		if (ret) {
+			loc_wipe(&loc);
+			return ret;
+		}
+
+		s->inode = inode_ref(loc.inode);
+		loc_wipe(&loc);
+	} else {
+		inode = qb_inode_from_filename (filename);
+		if (!inode)
+			return -EINVAL;
+
+		s->inode = inode_ref(inode);
+	}
+
+	return 0;
+}
+
+
+static int
+qemu_gluster_create (const char *filename, QEMUOptionParameter *options)
+{
+	uint64_t total_size = 0;
+	inode_t *inode = NULL;
+	fd_t *fd = NULL;
+	struct iatt stat = {0, };
+	int ret = 0;
+
+	inode = qb_inode_from_filename (filename);
+	if (!inode)
+		return -EINVAL;
+
+	while (options && options->name) {
+		if (!strcmp(options->name, BLOCK_OPT_SIZE)) {
+			total_size = options->value.n / BDRV_SECTOR_SIZE;
+		}
+		options++;
+	}
+
+	fd = fd_anonymous (inode);
+	if (!fd)
+		return -ENOMEM;
+
+	ret = syncop_fstat (FIRST_CHILD(THIS), fd, &stat);
+	if (ret) {
+		fd_unref (fd);
+		return ret;
+	}
+
+	if (stat.ia_size) {
+		/* format ONLY if the filesize is 0 bytes */
+		fd_unref (fd);
+		return -EFBIG;
+	}
+
+	if (total_size) {
+		ret = syncop_ftruncate (FIRST_CHILD(THIS), fd, total_size);
+		if (ret) {
+			fd_unref (fd);
+			return ret;
+		}
+	}
+
+	fd_unref (fd);
+	return 0;
+}
+
+
+static int
+qemu_gluster_co_readv (BlockDriverState *bs, int64_t sector_num, int nb_sectors,
+		       QEMUIOVector *qiov)
+{
+	fd_t *fd = NULL;
+	off_t offset = 0;
+	size_t size = 0;
+	struct iovec *iov = NULL;
+	int count = 0;
+	struct iobref *iobref = NULL;
+	int ret = 0;
+
+	fd = fd_from_bs (bs);
+	if (!fd)
+		return -EIO;
+
+	offset = sector_num * BDRV_SECTOR_SIZE;
+	size = nb_sectors * BDRV_SECTOR_SIZE;
+
+	ret = syncop_readv (FIRST_CHILD(THIS), fd, size, offset, 0,
+			    &iov, &count, &iobref);
+	if (ret < 0)
+		goto out;
+
+	iov_copy (qiov->iov, qiov->niov, iov, count); /* *choke!* */
+
+out:
+	GF_FREE (iov);
+	if (iobref)
+		iobref_unref (iobref);
+	fd_unref (fd);
+	return ret;
+}
+
+
+static int
+qemu_gluster_co_writev (BlockDriverState *bs, int64_t sector_num, int nb_sectors,
+			QEMUIOVector *qiov)
+{
+	fd_t *fd = NULL;
+	off_t offset = 0;
+	size_t size = 0;
+	struct iobref *iobref = NULL;
+	struct iobuf *iobuf = NULL;
+	struct iovec iov = {0, };
+	int ret = -ENOMEM;
+
+	fd = fd_from_bs (bs);
+	if (!fd)
+		return -EIO;
+
+	offset = sector_num * BDRV_SECTOR_SIZE;
+	size = nb_sectors * BDRV_SECTOR_SIZE;
+
+	iobuf = iobuf_get2 (THIS->ctx->iobuf_pool, size);
+	if (!iobuf)
+		goto out;
+
+	iobref = iobref_new ();
+	if (!iobref) {
+		goto out;
+	}
+
+	iobref_add (iobref, iobuf);
+
+	iov_unload (iobuf_ptr (iobuf), qiov->iov, qiov->niov); /* *choke!* */
+
+	iov.iov_base = iobuf_ptr (iobuf);
+	iov.iov_len = size;
+
+	ret = syncop_writev (FIRST_CHILD(THIS), fd, &iov, 1, offset, iobref, 0);
+
+out:
+	if (iobuf)
+		iobuf_unref (iobuf);
+	if (iobref)
+		iobref_unref (iobref);
+	fd_unref (fd);
+	return ret;
+}
+
+
+static int
+qemu_gluster_co_flush (BlockDriverState *bs)
+{
+	fd_t *fd = NULL;
+	int ret = 0;
+
+	fd = fd_from_bs (bs);
+
+	ret = syncop_flush (FIRST_CHILD(THIS), fd);
+
+	fd_unref (fd);
+
+	return ret;
+}
+
+
+static int
+qemu_gluster_co_fsync (BlockDriverState *bs)
+{
+	fd_t *fd = NULL;
+	int ret = 0;
+
+	fd = fd_from_bs (bs);
+
+	ret = syncop_fsync (FIRST_CHILD(THIS), fd, 0);
+
+	fd_unref (fd);
+
+	return ret;
+}
+
+
+static int
+qemu_gluster_truncate (BlockDriverState *bs, int64_t offset)
+{
+	fd_t *fd = NULL;
+	int ret = 0;
+
+	fd = fd_from_bs (bs);
+
+	ret = syncop_ftruncate (FIRST_CHILD(THIS), fd, offset);
+
+	fd_unref (fd);
+
+	return ret;
+}
+
+
+static int64_t
+qemu_gluster_getlength (BlockDriverState *bs)
+{
+	fd_t *fd = NULL;
+	int ret = 0;
+	struct iatt iatt = {0, };
+
+	fd = fd_from_bs (bs);
+
+	ret = syncop_fstat (FIRST_CHILD(THIS), fd, &iatt);
+	if (ret < 0)
+		return -1;
+
+	return iatt.ia_size;
+}
+
+
+static int64_t
+qemu_gluster_allocated_file_size (BlockDriverState *bs)
+{
+	fd_t *fd = NULL;
+	int ret = 0;
+	struct iatt iatt = {0, };
+
+	fd = fd_from_bs (bs);
+
+	ret = syncop_fstat (FIRST_CHILD(THIS), fd, &iatt);
+	if (ret < 0)
+		return -1;
+
+	return iatt.ia_blocks * 512;
+}
+
+
+static void
+qemu_gluster_close (BlockDriverState *bs)
+{
+	BDRVGlusterState *s = NULL;
+
+	s = bs->opaque;
+
+	inode_unref (s->inode);
+
+	return;
+}
+
+
+static QEMUOptionParameter qemu_gluster_create_options[] = {
+	{
+		.name = BLOCK_OPT_SIZE,
+		.type = OPT_SIZE,
+		.help = "Virtual disk size"
+	},
+	{ NULL }
+};
+
+
+static BlockDriver bdrv_gluster = {
+	.format_name                  = "gluster",
+	.protocol_name                = "gluster",
+	.instance_size                = sizeof(BDRVGlusterState),
+	.bdrv_file_open               = qemu_gluster_open,
+	.bdrv_close                   = qemu_gluster_close,
+	.bdrv_create                  = qemu_gluster_create,
+	.bdrv_getlength               = qemu_gluster_getlength,
+	.bdrv_get_allocated_file_size = qemu_gluster_allocated_file_size,
+	.bdrv_co_readv                = qemu_gluster_co_readv,
+	.bdrv_co_writev               = qemu_gluster_co_writev,
+	.bdrv_co_flush_to_os          = qemu_gluster_co_flush,
+	.bdrv_co_flush_to_disk        = qemu_gluster_co_fsync,
+	.bdrv_truncate                = qemu_gluster_truncate,
+	.create_options               = qemu_gluster_create_options,
+};
+
+
+static void bdrv_gluster_init(void)
+{
+	bdrv_register(&bdrv_gluster);
+}
+
+
+block_init(bdrv_gluster_init);
diff --git a/xlators/features/qemu-block/src/bh-syncop.c b/xlators/features/qemu-block/src/bh-syncop.c
new file mode 100644
index 000000000..e8686f6d4
--- /dev/null
+++ b/xlators/features/qemu-block/src/bh-syncop.c
@@ -0,0 +1,48 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "logging.h"
+#include "dict.h"
+#include "xlator.h"
+#include "syncop.h"
+#include "qemu-block-memory-types.h"
+
+#include "block/aio.h"
+
+void
+qemu_bh_schedule (QEMUBH *bh)
+{
+	return;
+}
+
+void
+qemu_bh_cancel (QEMUBH *bh)
+{
+	return;
+}
+
+void
+qemu_bh_delete (QEMUBH *bh)
+{
+
+}
+
+QEMUBH *
+qemu_bh_new (QEMUBHFunc *cb, void *opaque)
+{
+	return NULL;
+}
diff --git a/xlators/features/qemu-block/src/clock-timer.c b/xlators/features/qemu-block/src/clock-timer.c
new file mode 100644
index 000000000..fcbec6ad1
--- /dev/null
+++ b/xlators/features/qemu-block/src/clock-timer.c
@@ -0,0 +1,60 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "logging.h"
+#include "dict.h"
+#include "xlator.h"
+#include "syncop.h"
+#include "qemu-block-memory-types.h"
+
+#include "qemu/timer.h"
+
+QEMUClock *vm_clock;
+int use_rt_clock = 0;
+
+QEMUTimer *qemu_new_timer (QEMUClock *clock, int scale,
+			   QEMUTimerCB *cb, void *opaque)
+{
+	return NULL;
+}
+
+int64_t qemu_get_clock_ns (QEMUClock *clock)
+{
+	return 0;
+}
+
+void qemu_mod_timer (QEMUTimer *ts, int64_t expire_time)
+{
+	return;
+}
+
+void qemu_free_timer (QEMUTimer *ts)
+{
+
+}
+
+void qemu_del_timer (QEMUTimer *ts)
+{
+
+}
+
+bool qemu_aio_wait()
+{
+	synctask_wake (synctask_get());
+	synctask_yield (synctask_get());
+	return 0;
+}
diff --git a/xlators/features/qemu-block/src/coroutine-synctask.c b/xlators/features/qemu-block/src/coroutine-synctask.c
new file mode 100644
index 000000000..e43988a95
--- /dev/null
+++ b/xlators/features/qemu-block/src/coroutine-synctask.c
@@ -0,0 +1,116 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "logging.h"
+#include "dict.h"
+#include "xlator.h"
+#include "syncop.h"
+#include "qemu-block-memory-types.h"
+
+#include "qemu-block.h"
+
+/*
+ * This code serves as the bridge from the main glusterfs context to the qemu
+ * coroutine context via synctask. We create a single threaded syncenv with a
+ * single synctask responsible for processing a queue of coroutines. The qemu
+ * code invoked from within the synctask function handlers uses the ucontext
+ * coroutine implementation and scheduling logic internal to qemu. This
+ * effectively donates a thread of execution to qemu and its internal coroutine
+ * management.
+ *
+ * NOTE: The existence of concurrent synctasks has proven quite racy with regard
+ * to qemu coroutine management, particularly related to the lifecycle
+ * differences with top-level synctasks and internally created coroutines and
+ * interactions with qemu-internal queues (and locks, in turn). We explicitly
+ * disallow this scenario, via the queue, until it is more well supported.
+ */
+
+static struct {
+	struct list_head	queue;
+	gf_lock_t		lock;
+	struct synctask		*task;
+} qb_co;
+
+static void
+init_qbco()
+{
+	INIT_LIST_HEAD(&qb_co.queue);
+	LOCK_INIT(&qb_co.lock);
+}
+
+static int
+synctask_nop_cbk (int ret, call_frame_t *frame, void *opaque)
+{
+	return 0;
+}
+
+static int
+qb_synctask_wrap (void *opaque)
+{
+	qb_local_t *qb_local, *tmp;
+
+	LOCK(&qb_co.lock);
+
+	while (!list_empty(&qb_co.queue)) {
+		list_for_each_entry_safe(qb_local, tmp, &qb_co.queue, list) {
+			list_del_init(&qb_local->list);
+			break;
+		}
+
+		UNLOCK(&qb_co.lock);
+
+		qb_local->synctask_fn(qb_local);
+		/* qb_local is now unwound and gone! */
+
+		LOCK(&qb_co.lock);
+	}
+
+	qb_co.task = NULL;
+
+	UNLOCK(&qb_co.lock);
+
+	return 0;
+}
+
+int
+qb_coroutine (call_frame_t *frame, synctask_fn_t fn)
+{
+	qb_local_t *qb_local = NULL;
+	qb_conf_t *qb_conf = NULL;
+	static int init = 0;
+
+	qb_local = frame->local;
+	qb_local->synctask_fn = fn;
+	qb_conf = frame->this->private;
+
+	if (!init) {
+		init = 1;
+		init_qbco();
+	}
+
+	LOCK(&qb_co.lock);
+
+	if (!qb_co.task)
+		qb_co.task = synctask_create(qb_conf->env, qb_synctask_wrap,
+					     synctask_nop_cbk, frame, NULL);
+
+	list_add_tail(&qb_local->list, &qb_co.queue);
+
+	UNLOCK(&qb_co.lock);
+
+	return 0;
+}
diff --git a/xlators/features/qemu-block/src/monitor-logging.c b/xlators/features/qemu-block/src/monitor-logging.c
new file mode 100644
index 000000000..d37c37f0f
--- /dev/null
+++ b/xlators/features/qemu-block/src/monitor-logging.c
@@ -0,0 +1,50 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "logging.h"
+#include "dict.h"
+#include "xlator.h"
+#include "qemu-block-memory-types.h"
+
+#include "block/block_int.h"
+
+Monitor *cur_mon;
+
+int
+monitor_cur_is_qmp()
+{
+	/* No QMP support here */
+	return 0;
+}
+
+void
+monitor_set_error (Monitor *mon, QError *qerror)
+{
+	/* NOP here */
+	return;
+}
+
+
+void
+monitor_vprintf(Monitor *mon, const char *fmt, va_list ap)
+{
+	char buf[4096];
+
+	vsnprintf(buf, sizeof(buf), fmt, ap);
+
+	gf_log (THIS->name, GF_LOG_ERROR, "%s", buf);
+}
diff --git a/xlators/features/qemu-block/src/qb-coroutines.c b/xlators/features/qemu-block/src/qb-coroutines.c
new file mode 100644
index 000000000..974312f12
--- /dev/null
+++ b/xlators/features/qemu-block/src/qb-coroutines.c
@@ -0,0 +1,667 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "logging.h"
+#include "dict.h"
+#include "xlator.h"
+#include "inode.h"
+#include "call-stub.h"
+#include "defaults.h"
+#include "qemu-block-memory-types.h"
+#include "qemu-block.h"
+#include "qb-coroutines.h"
+
+
+int
+qb_format_and_resume (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	char filename[64];
+	char base_filename[128];
+	int use_base = 0;
+	qb_inode_t *qb_inode = NULL;
+	Error *local_err = NULL;
+	fd_t *fd = NULL;
+	dict_t *xattr = NULL;
+	qb_conf_t *qb_conf = NULL;
+	int ret = -1;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+	qb_conf = frame->this->private;
+
+	qb_inode_to_filename (inode, filename, 64);
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+
+	/*
+	 * See if the caller specified a backing image.
+	 */
+	if (!uuid_is_null(qb_inode->backing_gfid) || qb_inode->backing_fname) {
+		loc_t loc = {0,};
+		char gfid_str[64];
+		struct iatt buf;
+
+		if (!uuid_is_null(qb_inode->backing_gfid)) {
+			loc.inode = inode_find(qb_conf->root_inode->table,
+					qb_inode->backing_gfid);
+			if (!loc.inode) {
+				loc.inode = inode_new(qb_conf->root_inode->table);
+				uuid_copy(loc.inode->gfid,
+					qb_inode->backing_gfid);
+			}
+			uuid_copy(loc.gfid, loc.inode->gfid);
+		} else if (qb_inode->backing_fname) {
+			loc.inode = inode_new(qb_conf->root_inode->table);
+			loc.name = qb_inode->backing_fname;
+			loc.parent = inode_parent(inode, NULL, NULL);
+			loc_path(&loc, loc.name);
+		}
+
+		/*
+		 * Lookup the backing image. Verify existence and/or get the
+		 * gfid if we don't already have it.
+		 */
+		ret = syncop_lookup(FIRST_CHILD(frame->this), &loc, NULL, &buf,
+				    NULL, NULL);
+		GF_FREE(qb_inode->backing_fname);
+		if (ret) {
+			loc_wipe(&loc);
+			ret = -ret;
+			goto err;
+		}
+
+		uuid_copy(qb_inode->backing_gfid, buf.ia_gfid);
+		loc_wipe(&loc);
+
+		/*
+		 * We pass the filename of the backing image into the qemu block
+		 * subsystem as the associated gfid. This is embedded into the
+		 * clone image and passed along to the gluster bdrv backend when
+		 * the block subsystem needs to operate on the backing image on
+		 * behalf of the clone.
+		 */
+		uuid_unparse(qb_inode->backing_gfid, gfid_str);
+		snprintf(base_filename, sizeof(base_filename),
+			 "gluster://gfid:%s", gfid_str);
+		use_base = 1;
+	}
+
+	bdrv_img_create (filename, qb_inode->fmt,
+			 use_base ? base_filename : NULL, 0, 0, qb_inode->size,
+			 0, &local_err, true);
+
+	if (error_is_set (&local_err)) {
+		gf_log (frame->this->name, GF_LOG_ERROR, "%s",
+			error_get_pretty (local_err));
+		error_free (local_err);
+		QB_STUB_UNWIND (stub, -1, EIO);
+		return 0;
+	}
+
+	fd = fd_anonymous (inode);
+	if (!fd) {
+		gf_log (frame->this->name, GF_LOG_ERROR,
+			"could not create anonymous fd for %s",
+			uuid_utoa (inode->gfid));
+		QB_STUB_UNWIND (stub, -1, ENOMEM);
+		return 0;
+	}
+
+	xattr = dict_new ();
+	if (!xattr) {
+		gf_log (frame->this->name, GF_LOG_ERROR,
+			"could not allocate xattr dict for %s",
+			uuid_utoa (inode->gfid));
+		QB_STUB_UNWIND (stub, -1, ENOMEM);
+		fd_unref (fd);
+		return 0;
+	}
+
+	ret = dict_set_str (xattr, qb_conf->qb_xattr_key, local->fmt);
+	if (ret) {
+		gf_log (frame->this->name, GF_LOG_ERROR,
+			"could not dict_set for %s",
+			uuid_utoa (inode->gfid));
+		QB_STUB_UNWIND (stub, -1, ENOMEM);
+		fd_unref (fd);
+		dict_unref (xattr);
+		return 0;
+	}
+
+	ret = syncop_fsetxattr (FIRST_CHILD(THIS), fd, xattr, 0);
+	if (ret) {
+		gf_log (frame->this->name, GF_LOG_ERROR,
+			"failed to setxattr for %s",
+			uuid_utoa (inode->gfid));
+		QB_STUB_UNWIND (stub, -1, -ret);
+		fd_unref (fd);
+		dict_unref (xattr);
+		return 0;
+	}
+
+	fd_unref (fd);
+	dict_unref (xattr);
+
+	QB_STUB_UNWIND (stub, 0, 0);
+
+	return 0;
+
+err:
+	QB_STUB_UNWIND(stub, -1, ret);
+	return 0;
+}
+
+
+static BlockDriverState *
+qb_bs_create (inode_t *inode, const char *fmt)
+{
+	char filename[64];
+	BlockDriverState *bs = NULL;
+	BlockDriver *drv = NULL;
+	int op_errno = 0;
+	int ret = 0;
+
+	bs = bdrv_new (uuid_utoa (inode->gfid));
+	if (!bs) {
+		op_errno = ENOMEM;
+		gf_log (THIS->name, GF_LOG_ERROR,
+			"could not allocate @bdrv for gfid:%s",
+			uuid_utoa (inode->gfid));
+		goto err;
+	}
+
+	drv = bdrv_find_format (fmt);
+	if (!drv) {
+		op_errno = EINVAL;
+		gf_log (THIS->name, GF_LOG_ERROR,
+			"Unknown file format: %s for gfid:%s",
+			fmt, uuid_utoa (inode->gfid));
+		goto err;
+	}
+
+	qb_inode_to_filename (inode, filename, 64);
+
+	ret = bdrv_open (bs, filename, NULL, BDRV_O_RDWR, drv);
+	if (ret < 0) {
+		op_errno = -ret;
+		gf_log (THIS->name, GF_LOG_ERROR,
+			"Unable to bdrv_open() gfid:%s (%s)",
+			uuid_utoa (inode->gfid), strerror (op_errno));
+		goto err;
+	}
+
+	return bs;
+err:
+	errno = op_errno;
+	return NULL;
+}
+
+
+int
+qb_co_open (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+	if (!qb_inode->bs) {
+		/* FIXME: we need locks around this when
+		   enabling multithreaded syncop/coroutine
+		   for qemu-block
+		*/
+
+		qb_inode->bs = qb_bs_create (inode, qb_inode->fmt);
+		if (!qb_inode->bs) {
+			QB_STUB_UNWIND (stub, -1, errno);
+			return 0;
+		}
+	}
+	qb_inode->refcnt++;
+
+	QB_STUB_RESUME (stub);
+
+	return 0;
+}
+
+
+int
+qb_co_writev (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+	QEMUIOVector qiov = {0, };
+	int ret = 0;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+	if (!qb_inode->bs) {
+		/* FIXME: we need locks around this when
+		   enabling multithreaded syncop/coroutine
+		   for qemu-block
+		*/
+
+		qb_inode->bs = qb_bs_create (inode, qb_inode->fmt);
+		if (!qb_inode->bs) {
+			QB_STUB_UNWIND (stub, -1, errno);
+			return 0;
+		}
+	}
+
+	qemu_iovec_init_external (&qiov, stub->args.vector, stub->args.count);
+
+	ret = bdrv_pwritev (qb_inode->bs, stub->args.offset, &qiov);
+
+	if (ret < 0) {
+		QB_STUB_UNWIND (stub, -1, -ret);
+	} else {
+		QB_STUB_UNWIND (stub, ret, 0);
+	}
+
+	return 0;
+}
+
+
+int
+qb_co_readv (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+	struct iobuf *iobuf = NULL;
+	struct iobref *iobref = NULL;
+	struct iovec iov = {0, };
+	int ret = 0;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+	if (!qb_inode->bs) {
+		/* FIXME: we need locks around this when
+		   enabling multithreaded syncop/coroutine
+		   for qemu-block
+		*/
+
+		qb_inode->bs = qb_bs_create (inode, qb_inode->fmt);
+		if (!qb_inode->bs) {
+			QB_STUB_UNWIND (stub, -1, errno);
+			return 0;
+		}
+	}
+
+	if (stub->args.offset >= qb_inode->size) {
+		QB_STUB_UNWIND (stub, 0, 0);
+		return 0;
+	}
+
+	iobuf = iobuf_get2 (frame->this->ctx->iobuf_pool, stub->args.size);
+	if (!iobuf) {
+		QB_STUB_UNWIND (stub, -1, ENOMEM);
+		return 0;
+	}
+
+	iobref = iobref_new ();
+	if (!iobref) {
+		QB_STUB_UNWIND (stub, -1, ENOMEM);
+		iobuf_unref (iobuf);
+		return 0;
+	}
+
+	if (iobref_add (iobref, iobuf) < 0) {
+		iobuf_unref (iobuf);
+		iobref_unref (iobref);
+		QB_STUB_UNWIND (stub, -1, ENOMEM);
+		return 0;
+	}
+
+	ret = bdrv_pread (qb_inode->bs, stub->args.offset, iobuf_ptr (iobuf),
+			  stub->args.size);
+
+	if (ret < 0) {
+		QB_STUB_UNWIND (stub, -1, -ret);
+		iobref_unref (iobref);
+		return 0;
+	}
+
+	iov.iov_base = iobuf_ptr (iobuf);
+	iov.iov_len = ret;
+
+	stub->args_cbk.vector = iov_dup (&iov, 1);
+	stub->args_cbk.count = 1;
+	stub->args_cbk.iobref = iobref;
+
+	QB_STUB_UNWIND (stub, ret, 0);
+
+	return 0;
+}
+
+
+int
+qb_co_fsync (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+	int ret = 0;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+	if (!qb_inode->bs) {
+		/* FIXME: we need locks around this when
+		   enabling multithreaded syncop/coroutine
+		   for qemu-block
+		*/
+
+		qb_inode->bs = qb_bs_create (inode, qb_inode->fmt);
+		if (!qb_inode->bs) {
+			QB_STUB_UNWIND (stub, -1, errno);
+			return 0;
+		}
+	}
+
+	ret = bdrv_flush (qb_inode->bs);
+
+	if (ret < 0) {
+		QB_STUB_UNWIND (stub, -1, -ret);
+	} else {
+		QB_STUB_UNWIND (stub, ret, 0);
+	}
+
+	return 0;
+}
+
+
+static void
+qb_update_size_xattr (xlator_t *this, fd_t *fd, const char *fmt, off_t offset)
+{
+	char val[QB_XATTR_VAL_MAX];
+	qb_conf_t *qb_conf = NULL;
+	dict_t *xattr = NULL;
+
+	qb_conf = this->private;
+
+	snprintf (val, QB_XATTR_VAL_MAX, "%s:%llu",
+		  fmt, (long long unsigned) offset);
+
+	xattr = dict_new ();
+	if (!xattr)
+		return;
+
+	if (dict_set_str (xattr, qb_conf->qb_xattr_key, val) != 0) {
+		dict_unref (xattr);
+		return;
+	}
+
+	syncop_fsetxattr (FIRST_CHILD(this), fd, xattr, 0);
+	dict_unref (xattr);
+}
+
+
+int
+qb_co_truncate (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+	int ret = 0;
+	off_t offset = 0;
+	xlator_t *this = NULL;
+
+	this = THIS;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+	if (!qb_inode->bs) {
+		/* FIXME: we need locks around this when
+		   enabling multithreaded syncop/coroutine
+		   for qemu-block
+		*/
+
+		qb_inode->bs = qb_bs_create (inode, qb_inode->fmt);
+		if (!qb_inode->bs) {
+			QB_STUB_UNWIND (stub, -1, errno);
+			return 0;
+		}
+	}
+
+	ret = syncop_fstat (FIRST_CHILD(this), local->fd,
+                            &stub->args_cbk.prestat);
+        if (ret < 0)
+                goto out;
+	stub->args_cbk.prestat.ia_size = qb_inode->size;
+
+	ret = bdrv_truncate (qb_inode->bs, stub->args.offset);
+	if (ret < 0)
+		goto out;
+
+	offset = bdrv_getlength (qb_inode->bs);
+
+	qb_inode->size = offset;
+
+	ret = syncop_fstat (FIRST_CHILD(this), local->fd,
+                            &stub->args_cbk.poststat);
+        if (ret < 0)
+                goto out;
+	stub->args_cbk.poststat.ia_size = qb_inode->size;
+
+	qb_update_size_xattr (this, local->fd, qb_inode->fmt, qb_inode->size);
+
+out:
+	if (ret < 0) {
+		QB_STUB_UNWIND (stub, -1, -ret);
+	} else {
+		QB_STUB_UNWIND (stub, ret, 0);
+	}
+
+	return 0;
+}
+
+
+int
+qb_co_close (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+	BlockDriverState *bs = NULL;
+
+	local = opaque;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (THIS, inode);
+
+	if (!--qb_inode->refcnt) {
+		bs = qb_inode->bs;
+		qb_inode->bs = NULL;
+		bdrv_delete (bs);
+	}
+
+	frame = local->frame;
+	frame->local = NULL;
+	qb_local_free (THIS, local);
+	STACK_DESTROY (frame->root);
+
+	return 0;
+}
+
+
+int
+qb_snapshot_create (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+	QEMUSnapshotInfo sn;
+	struct timeval tv = {0, };
+	int ret = 0;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+	if (!qb_inode->bs) {
+		/* FIXME: we need locks around this when
+		   enabling multithreaded syncop/coroutine
+		   for qemu-block
+		*/
+
+		qb_inode->bs = qb_bs_create (inode, qb_inode->fmt);
+		if (!qb_inode->bs) {
+			QB_STUB_UNWIND (stub, -1, errno);
+			return 0;
+		}
+	}
+
+	memset (&sn, 0, sizeof (sn));
+        pstrcpy (sn.name, sizeof(sn.name), local->name);
+        gettimeofday (&tv, NULL);
+        sn.date_sec = tv.tv_sec;
+        sn.date_nsec = tv.tv_usec * 1000;
+
+        ret = bdrv_snapshot_create (qb_inode->bs, &sn);
+	if (ret < 0) {
+		QB_STUB_UNWIND (stub, -1, -ret);
+	} else {
+		QB_STUB_UNWIND (stub, ret, 0);
+	}
+
+	return 0;
+}
+
+
+int
+qb_snapshot_delete (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+	int ret = 0;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+	if (!qb_inode->bs) {
+		/* FIXME: we need locks around this when
+		   enabling multithreaded syncop/coroutine
+		   for qemu-block
+		*/
+
+		qb_inode->bs = qb_bs_create (inode, qb_inode->fmt);
+		if (!qb_inode->bs) {
+			QB_STUB_UNWIND (stub, -1, errno);
+			return 0;
+		}
+	}
+
+	ret = bdrv_snapshot_delete (qb_inode->bs, local->name);
+
+	if (ret < 0) {
+		QB_STUB_UNWIND (stub, -1, -ret);
+	} else {
+		QB_STUB_UNWIND (stub, ret, 0);
+	}
+
+	return 0;
+}
+
+
+int
+qb_snapshot_goto (void *opaque)
+{
+	qb_local_t *local = NULL;
+	call_frame_t *frame = NULL;
+	call_stub_t *stub = NULL;
+	inode_t *inode = NULL;
+	qb_inode_t *qb_inode = NULL;
+	int ret = 0;
+
+	local = opaque;
+	frame = local->frame;
+	stub = local->stub;
+	inode = local->inode;
+
+	qb_inode = qb_inode_ctx_get (frame->this, inode);
+	if (!qb_inode->bs) {
+		/* FIXME: we need locks around this when
+		   enabling multithreaded syncop/coroutine
+		   for qemu-block
+		*/
+
+		qb_inode->bs = qb_bs_create (inode, qb_inode->fmt);
+		if (!qb_inode->bs) {
+			QB_STUB_UNWIND (stub, -1, errno);
+			return 0;
+		}
+	}
+
+	ret = bdrv_snapshot_goto (qb_inode->bs, local->name);
+
+	if (ret < 0) {
+		QB_STUB_UNWIND (stub, -1, -ret);
+	} else {
+		QB_STUB_UNWIND (stub, ret, 0);
+	}
+
+	return 0;
+}
diff --git a/xlators/features/qemu-block/src/qb-coroutines.h b/xlators/features/qemu-block/src/qb-coroutines.h
new file mode 100644
index 000000000..583319f3b
--- /dev/null
+++ b/xlators/features/qemu-block/src/qb-coroutines.h
@@ -0,0 +1,30 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef __QB_COROUTINES_H
+#define __QB_COROUTINES_H
+
+#include "syncop.h"
+#include "call-stub.h"
+#include "block/block_int.h"
+#include "monitor/monitor.h"
+
+int qb_format_and_resume (void *opaque);
+int qb_snapshot_create (void *opaque);
+int qb_snapshot_delete (void *opaque);
+int qb_snapshot_goto (void *opaque);
+int qb_co_open (void *opaque);
+int qb_co_close (void *opaque);
+int qb_co_writev (void *opaque);
+int qb_co_readv (void *opaque);
+int qb_co_fsync (void *opaque);
+int qb_co_truncate (void *opaque);
+
+#endif /* __QB_COROUTINES_H */
diff --git a/xlators/features/qemu-block/src/qemu-block-memory-types.h b/xlators/features/qemu-block/src/qemu-block-memory-types.h
new file mode 100644
index 000000000..267b3893f
--- /dev/null
+++ b/xlators/features/qemu-block/src/qemu-block-memory-types.h
@@ -0,0 +1,25 @@
+/*
+  Copyright (c) 2008-2012 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef __QB_MEM_TYPES_H__
+#define __QB_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_qb_mem_types_ {
+        gf_qb_mt_qb_conf_t   = gf_common_mt_end + 1,
+        gf_qb_mt_qb_inode_t,
+        gf_qb_mt_qb_local_t,
+	gf_qb_mt_coroutinesynctask_t,
+        gf_qb_mt_end
+};
+#endif
+
diff --git a/xlators/features/qemu-block/src/qemu-block.c b/xlators/features/qemu-block/src/qemu-block.c
new file mode 100644
index 000000000..48bbf3140
--- /dev/null
+++ b/xlators/features/qemu-block/src/qemu-block.c
@@ -0,0 +1,1140 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "glusterfs.h"
+#include "logging.h"
+#include "dict.h"
+#include "xlator.h"
+#include "inode.h"
+#include "call-stub.h"
+#include "defaults.h"
+#include "qemu-block-memory-types.h"
+#include "qemu-block.h"
+#include "qb-coroutines.h"
+
+
+qb_inode_t *
+__qb_inode_ctx_get (xlator_t *this, inode_t *inode)
+{
+        uint64_t    value    = 0;
+        qb_inode_t *qb_inode = NULL;
+
+        __inode_ctx_get (inode, this, &value);
+        qb_inode = (qb_inode_t *)(unsigned long) value;
+
+        return qb_inode;
+}
+
+
+qb_inode_t *
+qb_inode_ctx_get (xlator_t *this, inode_t *inode)
+{
+        qb_inode_t *qb_inode = NULL;
+
+        LOCK (&inode->lock);
+        {
+                qb_inode = __qb_inode_ctx_get (this, inode);
+        }
+        UNLOCK (&inode->lock);
+
+        return qb_inode;
+}
+
+
+qb_inode_t *
+qb_inode_ctx_del (xlator_t *this, inode_t *inode)
+{
+        uint64_t    value    = 0;
+        qb_inode_t *qb_inode = NULL;
+
+        inode_ctx_del (inode, this, &value);
+        qb_inode = (qb_inode_t *)(unsigned long) value;
+
+        return qb_inode;
+}
+
+
+int
+qb_inode_cleanup (xlator_t *this, inode_t *inode, int warn)
+{
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_del (this, inode);
+
+	if (!qb_inode)
+		return 0;
+
+	if (warn)
+		gf_log (this->name, GF_LOG_WARNING,
+			"inode %s no longer block formatted",
+			uuid_utoa (inode->gfid));
+
+	/* free (qb_inode->bs); */
+
+	GF_FREE (qb_inode);
+
+	return 0;
+}
+
+
+int
+qb_iatt_fixup (xlator_t *this, inode_t *inode, struct iatt *iatt)
+{
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_get (this, inode);
+	if (!qb_inode)
+		return 0;
+
+	iatt->ia_size = qb_inode->size;
+
+	return 0;
+}
+
+
+int
+qb_format_extract (xlator_t *this, char *format, inode_t *inode)
+{
+	char       *s, *save;
+	uint64_t    size = 0;
+	char        fmt[QB_XATTR_VAL_MAX+1] = {0, };
+	qb_inode_t *qb_inode = NULL;
+	char *formatstr = NULL;
+	uuid_t gfid = {0,};
+	char gfid_str[64] = {0,};
+	int ret;
+
+	strncpy(fmt, format, QB_XATTR_VAL_MAX);
+
+	s = strtok_r(fmt, ":", &save);
+	if (!s)
+		goto invalid;
+	formatstr = gf_strdup(s);
+
+	s = strtok_r(NULL, ":", &save);
+	if (!s)
+		goto invalid;
+	if (gf_string2bytesize (s, &size))
+		goto invalid;
+	if (!size)
+		goto invalid;
+
+	s = strtok_r(NULL, "\0", &save);
+	if (s && !strncmp(s, "<gfid:", strlen("<gfid:"))) {
+		/*
+		 * Check for valid gfid backing image specifier.
+		 */
+		if (strlen(s) + 1 > sizeof(gfid_str))
+			goto invalid;
+		ret = sscanf(s, "<gfid:%[^>]s", gfid_str);
+		if (ret == 1) {
+			ret = uuid_parse(gfid_str, gfid);
+			if (ret < 0)
+				goto invalid;
+		}
+	}
+
+	qb_inode = qb_inode_ctx_get (this, inode);
+	if (!qb_inode)
+		qb_inode = GF_CALLOC (1, sizeof (*qb_inode),
+				      gf_qb_mt_qb_inode_t);
+	if (!qb_inode) {
+		GF_FREE(formatstr);
+		return ENOMEM;
+	}
+
+	strncpy(qb_inode->fmt, formatstr, QB_XATTR_VAL_MAX);
+	qb_inode->size = size;
+
+	/*
+	 * If a backing gfid was not specified, interpret any remaining bytes
+	 * associated with a backing image as a filename local to the parent
+	 * directory. The format processing will validate further.
+	 */
+	if (!uuid_is_null(gfid))
+		uuid_copy(qb_inode->backing_gfid, gfid);
+	else if (s)
+		qb_inode->backing_fname = gf_strdup(s);
+
+	inode_ctx_set (inode, this, (void *)&qb_inode);
+
+	GF_FREE(formatstr);
+
+	return 0;
+
+invalid:
+	GF_FREE(formatstr);
+
+	gf_log (this->name, GF_LOG_WARNING,
+		"invalid format '%s' in inode %s", format,
+		uuid_utoa (inode->gfid));
+	return EINVAL;
+}
+
+
+void
+qb_local_free (xlator_t *this, qb_local_t *local)
+{
+	if (local->inode)
+		inode_unref (local->inode);
+	if (local->fd)
+		fd_unref (local->fd);
+	GF_FREE (local);
+}
+
+
+int
+qb_local_init (call_frame_t *frame)
+{
+	qb_local_t *qb_local = NULL;
+
+	qb_local = GF_CALLOC (1, sizeof (*qb_local), gf_qb_mt_qb_local_t);
+	if (!qb_local)
+		return -1;
+	INIT_LIST_HEAD(&qb_local->list);
+
+	qb_local->frame = frame;
+	frame->local = qb_local;
+
+	return 0;
+}
+
+
+int
+qb_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+	       int op_ret, int op_errno, inode_t *inode, struct iatt *buf,
+	       dict_t *xdata, struct iatt *postparent)
+{
+	char *format = NULL;
+	qb_conf_t *conf = NULL;
+
+	conf = this->private;
+
+	if (op_ret == -1)
+		goto out;
+
+	/*
+	 * Cache the root inode for dealing with backing images. The format
+	 * coroutine and the gluster qemu backend driver both use the root inode
+	 * table to verify and/or redirect I/O to the backing image via
+	 * anonymous fd's.
+	 */
+	if (!conf->root_inode && __is_root_gfid(inode->gfid))
+		conf->root_inode = inode_ref(inode);
+
+	if (!xdata)
+		goto out;
+
+	if (dict_get_str (xdata, conf->qb_xattr_key, &format))
+		goto out;
+
+	if (!format) {
+		qb_inode_cleanup (this, inode, 1);
+		goto out;
+	}
+
+	op_errno = qb_format_extract (this, format, inode);
+	if (op_errno)
+		op_ret = -1;
+
+	qb_iatt_fixup (this, inode, buf);
+out:
+	QB_STACK_UNWIND (lookup, frame, op_ret, op_errno, inode, buf,
+			 xdata, postparent);
+	return 0;
+}
+
+
+int
+qb_lookup (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+	qb_conf_t *conf = NULL;
+
+	conf = this->private;
+
+	xdata = xdata ? dict_ref (xdata) : dict_new ();
+
+	if (!xdata)
+		goto enomem;
+
+	if (dict_set_int32 (xdata, conf->qb_xattr_key, 0))
+		goto enomem;
+
+	STACK_WIND (frame, qb_lookup_cbk, FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->lookup, loc, xdata);
+	dict_unref (xdata);
+	return 0;
+enomem:
+	QB_STACK_UNWIND (lookup, frame, -1, ENOMEM, 0, 0, 0, 0);
+	if (xdata)
+		dict_unref (xdata);
+	return 0;
+}
+
+
+int
+qb_setxattr_format (call_frame_t *frame, xlator_t *this, call_stub_t *stub,
+		    dict_t *xattr, inode_t *inode)
+{
+	char *format = NULL;
+	int op_errno = 0;
+	qb_local_t *qb_local = NULL;
+	data_t *data = NULL;
+	qb_inode_t *qb_inode;
+
+	if (!(data = dict_get (xattr, "trusted.glusterfs.block-format"))) {
+		QB_STUB_RESUME (stub);
+		return 0;
+	}
+
+	format = alloca (data->len + 1);
+	memcpy (format, data->data, data->len);
+	format[data->len] = 0;
+
+	op_errno = qb_format_extract (this, format, inode);
+	if (op_errno) {
+		QB_STUB_UNWIND (stub, -1, op_errno);
+		return 0;
+	}
+	qb_inode = qb_inode_ctx_get(this, inode);
+
+	qb_local = frame->local;
+
+	qb_local->stub = stub;
+	qb_local->inode = inode_ref (inode);
+
+	snprintf(qb_local->fmt, QB_XATTR_VAL_MAX, "%s:%lu", qb_inode->fmt,
+		 qb_inode->size);
+
+	qb_coroutine (frame, qb_format_and_resume);
+
+	return 0;
+}
+
+
+int
+qb_setxattr_snapshot_create (call_frame_t *frame, xlator_t *this,
+			     call_stub_t *stub, dict_t *xattr, inode_t *inode)
+{
+	qb_local_t *qb_local = NULL;
+	char *name = NULL;
+	data_t *data = NULL;
+
+	if (!(data = dict_get (xattr, "trusted.glusterfs.block-snapshot-create"))) {
+		QB_STUB_RESUME (stub);
+		return 0;
+	}
+
+	name = alloca (data->len + 1);
+	memcpy (name, data->data, data->len);
+	name[data->len] = 0;
+
+	qb_local = frame->local;
+
+	qb_local->stub = stub;
+	qb_local->inode = inode_ref (inode);
+	strncpy (qb_local->name, name, 128);
+
+	qb_coroutine (frame, qb_snapshot_create);
+
+	return 0;
+}
+
+
+int
+qb_setxattr_snapshot_delete (call_frame_t *frame, xlator_t *this,
+			     call_stub_t *stub, dict_t *xattr, inode_t *inode)
+{
+	qb_local_t *qb_local = NULL;
+	char *name = NULL;
+	data_t *data = NULL;
+
+	if (!(data = dict_get (xattr, "trusted.glusterfs.block-snapshot-delete"))) {
+		QB_STUB_RESUME (stub);
+		return 0;
+	}
+
+	name = alloca (data->len + 1);
+	memcpy (name, data->data, data->len);
+	name[data->len] = 0;
+
+	qb_local = frame->local;
+
+	qb_local->stub = stub;
+	qb_local->inode = inode_ref (inode);
+	strncpy (qb_local->name, name, 128);
+
+	qb_coroutine (frame, qb_snapshot_delete);
+
+	return 0;
+}
+
+int
+qb_setxattr_snapshot_goto (call_frame_t *frame, xlator_t *this,
+			   call_stub_t *stub, dict_t *xattr, inode_t *inode)
+{
+	qb_local_t *qb_local = NULL;
+	char *name = NULL;
+	data_t *data = NULL;
+
+	if (!(data = dict_get (xattr, "trusted.glusterfs.block-snapshot-goto"))) {
+		QB_STUB_RESUME (stub);
+		return 0;
+	}
+
+	name = alloca (data->len + 1);
+	memcpy (name, data->data, data->len);
+	name[data->len] = 0;
+
+	qb_local = frame->local;
+
+	qb_local->stub = stub;
+	qb_local->inode = inode_ref (inode);
+	strncpy (qb_local->name, name, 128);
+
+	qb_coroutine (frame, qb_snapshot_goto);
+
+	return 0;
+}
+
+
+int
+qb_setxattr_common (call_frame_t *frame, xlator_t *this, call_stub_t *stub,
+		    dict_t *xattr, inode_t *inode)
+{
+	data_t *data = NULL;
+
+	if ((data = dict_get (xattr, "trusted.glusterfs.block-format"))) {
+		qb_setxattr_format (frame, this, stub, xattr, inode);
+		return 0;
+	}
+
+	if ((data = dict_get (xattr, "trusted.glusterfs.block-snapshot-create"))) {
+		qb_setxattr_snapshot_create (frame, this, stub, xattr, inode);
+		return 0;
+	}
+
+	if ((data = dict_get (xattr, "trusted.glusterfs.block-snapshot-delete"))) {
+		qb_setxattr_snapshot_delete (frame, this, stub, xattr, inode);
+		return 0;
+	}
+
+	if ((data = dict_get (xattr, "trusted.glusterfs.block-snapshot-goto"))) {
+		qb_setxattr_snapshot_goto (frame, this, stub, xattr, inode);
+		return 0;
+	}
+
+	QB_STUB_RESUME (stub);
+
+	return 0;
+}
+
+
+int
+qb_setxattr (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xattr,
+	     int flags, dict_t *xdata)
+{
+	call_stub_t *stub = NULL;
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	stub = fop_setxattr_stub (frame, default_setxattr_resume, loc, xattr,
+				  flags, xdata);
+	if (!stub)
+		goto enomem;
+
+	qb_setxattr_common (frame, this, stub, xattr, loc->inode);
+
+	return 0;
+enomem:
+	QB_STACK_UNWIND (setxattr, frame, -1, ENOMEM, 0);
+	return 0;
+}
+
+
+int
+qb_fsetxattr (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xattr,
+	      int flags, dict_t *xdata)
+{
+	call_stub_t *stub = NULL;
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	stub = fop_fsetxattr_stub (frame, default_fsetxattr_resume, fd, xattr,
+				   flags, xdata);
+	if (!stub)
+		goto enomem;
+
+	qb_setxattr_common (frame, this, stub, xattr, fd->inode);
+
+	return 0;
+enomem:
+	QB_STACK_UNWIND (fsetxattr, frame, -1, ENOMEM, 0);
+	return 0;
+}
+
+
+int
+qb_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+	     int op_ret, int op_errno, fd_t *fd, dict_t *xdata)
+{
+	call_stub_t *stub = NULL;
+	qb_local_t *qb_local = NULL;
+
+	qb_local = frame->local;
+
+	if (op_ret < 0)
+		goto unwind;
+
+	if (!qb_inode_ctx_get (this, qb_local->inode))
+		goto unwind;
+
+	stub = fop_open_cbk_stub (frame, NULL, op_ret, op_errno, fd, xdata);
+	if (!stub) {
+		op_ret = -1;
+		op_errno = ENOMEM;
+		goto unwind;
+	}
+
+	qb_local->stub = stub;
+
+	qb_coroutine (frame, qb_co_open);
+
+	return 0;
+unwind:
+	QB_STACK_UNWIND (open, frame, op_ret, op_errno, fd, xdata);
+	return 0;
+}
+
+
+int
+qb_open (call_frame_t *frame, xlator_t *this, loc_t *loc, int flags,
+	 fd_t *fd, dict_t *xdata)
+{
+	qb_local_t *qb_local = NULL;
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_get (this, loc->inode);
+	if (!qb_inode) {
+		STACK_WIND (frame, default_open_cbk, FIRST_CHILD(this),
+			    FIRST_CHILD(this)->fops->open, loc, flags, fd,
+			    xdata);
+		return 0;
+	}
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	qb_local = frame->local;
+
+	qb_local->inode = inode_ref (loc->inode);
+	qb_local->fd = fd_ref (fd);
+
+	STACK_WIND (frame, qb_open_cbk, FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->open, loc, flags, fd, xdata);
+	return 0;
+enomem:
+	QB_STACK_UNWIND (open, frame, -1, ENOMEM, 0, 0);
+	return 0;
+}
+
+
+int
+qb_writev (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iovec *vector,
+	   int count, off_t offset, uint32_t flags, struct iobref *iobref,
+	   dict_t *xdata)
+{
+	qb_local_t *qb_local = NULL;
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_get (this, fd->inode);
+	if (!qb_inode) {
+		STACK_WIND (frame, default_writev_cbk, FIRST_CHILD(this),
+			    FIRST_CHILD(this)->fops->writev, fd, vector, count,
+			    offset, flags, iobref, xdata);
+		return 0;
+	}
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	qb_local = frame->local;
+
+	qb_local->inode = inode_ref (fd->inode);
+	qb_local->fd = fd_ref (fd);
+
+	qb_local->stub = fop_writev_stub (frame, NULL, fd, vector, count,
+					  offset, flags, iobref, xdata);
+	if (!qb_local->stub)
+		goto enomem;
+
+	qb_coroutine (frame, qb_co_writev);
+
+	return 0;
+enomem:
+	QB_STACK_UNWIND (writev, frame, -1, ENOMEM, 0, 0, 0);
+	return 0;
+}
+
+
+int
+qb_readv (call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+	  off_t offset, uint32_t flags, dict_t *xdata)
+{
+	qb_local_t *qb_local = NULL;
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_get (this, fd->inode);
+	if (!qb_inode) {
+		STACK_WIND (frame, default_readv_cbk, FIRST_CHILD(this),
+			    FIRST_CHILD(this)->fops->readv, fd, size, offset,
+			    flags, xdata);
+		return 0;
+	}
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	qb_local = frame->local;
+
+	qb_local->inode = inode_ref (fd->inode);
+	qb_local->fd = fd_ref (fd);
+
+	qb_local->stub = fop_readv_stub (frame, NULL, fd, size, offset,
+					  flags, xdata);
+	if (!qb_local->stub)
+		goto enomem;
+
+	qb_coroutine (frame, qb_co_readv);
+
+	return 0;
+enomem:
+	QB_STACK_UNWIND (readv, frame, -1, ENOMEM, 0, 0, 0, 0, 0);
+	return 0;
+}
+
+
+int
+qb_fsync (call_frame_t *frame, xlator_t *this, fd_t *fd, int dsync,
+	  dict_t *xdata)
+{
+	qb_local_t *qb_local = NULL;
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_get (this, fd->inode);
+	if (!qb_inode) {
+		STACK_WIND (frame, default_fsync_cbk, FIRST_CHILD(this),
+			    FIRST_CHILD(this)->fops->fsync, fd, dsync, xdata);
+		return 0;
+	}
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	qb_local = frame->local;
+
+	qb_local->inode = inode_ref (fd->inode);
+	qb_local->fd = fd_ref (fd);
+
+	qb_local->stub = fop_fsync_stub (frame, NULL, fd, dsync, xdata);
+
+	if (!qb_local->stub)
+		goto enomem;
+
+	qb_coroutine (frame, qb_co_fsync);
+
+	return 0;
+enomem:
+	QB_STACK_UNWIND (fsync, frame, -1, ENOMEM, 0, 0, 0);
+	return 0;
+}
+
+
+int
+qb_flush (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+{
+	qb_local_t *qb_local = NULL;
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_get (this, fd->inode);
+	if (!qb_inode) {
+		STACK_WIND (frame, default_flush_cbk, FIRST_CHILD(this),
+			    FIRST_CHILD(this)->fops->flush, fd, xdata);
+		return 0;
+	}
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	qb_local = frame->local;
+
+	qb_local->inode = inode_ref (fd->inode);
+	qb_local->fd = fd_ref (fd);
+
+	qb_local->stub = fop_flush_stub (frame, NULL, fd, xdata);
+
+	if (!qb_local->stub)
+		goto enomem;
+
+	qb_coroutine (frame, qb_co_fsync);
+
+	return 0;
+enomem:
+	QB_STACK_UNWIND (flush, frame, -1, ENOMEM, 0);
+	return 0;
+}
+
+static int32_t
+qb_readdirp_cbk(call_frame_t *frame, void *cookie, xlator_t *this,
+		int32_t op_ret, int32_t op_errno, gf_dirent_t *entries,
+		dict_t *xdata)
+{
+	qb_conf_t *conf = this->private;
+	gf_dirent_t *entry;
+	char *format;
+
+	list_for_each_entry(entry, &entries->list, list) {
+		if (!entry->inode || !entry->dict)
+			continue;
+
+		format = NULL;
+		if (dict_get_str(entry->dict, conf->qb_xattr_key, &format))
+			continue;
+
+		if (!format) {
+			qb_inode_cleanup(this, entry->inode, 1);
+			continue;
+		}
+
+		if (qb_format_extract(this, format, entry->inode))
+			continue;
+
+		qb_iatt_fixup(this, entry->inode, &entry->d_stat);
+	}
+
+	STACK_UNWIND_STRICT(readdirp, frame, op_ret, op_errno, entries, xdata);
+	return 0;
+}
+
+static int32_t
+qb_readdirp(call_frame_t *frame, xlator_t *this, fd_t *fd, size_t size,
+	    off_t off, dict_t *xdata)
+{
+	qb_conf_t *conf = this->private;
+
+	xdata = xdata ? dict_ref(xdata) : dict_new();
+	if (!xdata)
+		goto enomem;
+
+	if (dict_set_int32 (xdata, conf->qb_xattr_key, 0))
+		goto enomem;
+
+	STACK_WIND(frame, qb_readdirp_cbk, FIRST_CHILD(this),
+		   FIRST_CHILD(this)->fops->readdirp, fd, size, off, xdata);
+
+	dict_unref(xdata);
+	return 0;
+
+enomem:
+	QB_STACK_UNWIND(readdirp, frame, -1, ENOMEM, NULL, NULL);
+	if (xdata)
+		dict_unref(xdata);
+	return 0;
+}
+
+int
+qb_truncate (call_frame_t *frame, xlator_t *this, loc_t *loc, off_t offset,
+	     dict_t *xdata)
+{
+	qb_local_t *qb_local = NULL;
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_get (this, loc->inode);
+	if (!qb_inode) {
+		STACK_WIND (frame, default_truncate_cbk, FIRST_CHILD(this),
+			    FIRST_CHILD(this)->fops->truncate, loc, offset,
+			    xdata);
+		return 0;
+	}
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	qb_local = frame->local;
+
+	qb_local->inode = inode_ref (loc->inode);
+	qb_local->fd = fd_anonymous (loc->inode);
+
+	qb_local->stub = fop_truncate_stub (frame, NULL, loc, offset, xdata);
+
+	if (!qb_local->stub)
+		goto enomem;
+
+	qb_coroutine (frame, qb_co_truncate);
+
+	return 0;
+enomem:
+	QB_STACK_UNWIND (truncate, frame, -1, ENOMEM, 0, 0, 0);
+	return 0;
+}
+
+
+int
+qb_ftruncate (call_frame_t *frame, xlator_t *this, fd_t *fd, off_t offset,
+	     dict_t *xdata)
+{
+	qb_local_t *qb_local = NULL;
+	qb_inode_t *qb_inode = NULL;
+
+	qb_inode = qb_inode_ctx_get (this, fd->inode);
+	if (!qb_inode) {
+		STACK_WIND (frame, default_ftruncate_cbk, FIRST_CHILD(this),
+			    FIRST_CHILD(this)->fops->ftruncate, fd, offset,
+			    xdata);
+		return 0;
+	}
+
+	if (qb_local_init (frame) != 0)
+		goto enomem;
+
+	qb_local = frame->local;
+
+	qb_local->inode = inode_ref (fd->inode);
+	qb_local->fd = fd_ref (fd);
+
+	qb_local->stub = fop_ftruncate_stub (frame, NULL, fd, offset, xdata);
+
+	if (!qb_local->stub)
+		goto enomem;
+
+	qb_coroutine (frame, qb_co_truncate);
+
+	return 0;
+enomem:
+	QB_STACK_UNWIND (ftruncate, frame, -1, ENOMEM, 0, 0, 0);
+	return 0;
+}
+
+
+int
+qb_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+	     int op_ret, int op_errno, struct iatt *iatt, dict_t *xdata)
+{
+	inode_t *inode = NULL;
+
+	inode = frame->local;
+	frame->local = NULL;
+
+	if (inode) {
+		qb_iatt_fixup (this, inode, iatt);
+		inode_unref (inode);
+	}
+
+	QB_STACK_UNWIND (stat, frame, op_ret, op_errno, iatt, xdata);
+
+	return 0;
+}
+
+int
+qb_stat (call_frame_t *frame, xlator_t *this, loc_t *loc, dict_t *xdata)
+{
+	if (qb_inode_ctx_get (this, loc->inode))
+		frame->local = inode_ref (loc->inode);
+
+	STACK_WIND (frame, qb_stat_cbk, FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->stat, loc, xdata);
+	return 0;
+}
+
+
+int
+qb_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+	      int op_ret, int op_errno, struct iatt *iatt, dict_t *xdata)
+{
+	inode_t *inode = NULL;
+
+	inode = frame->local;
+	frame->local = NULL;
+
+	if (inode) {
+		qb_iatt_fixup (this, inode, iatt);
+		inode_unref (inode);
+	}
+
+	QB_STACK_UNWIND (fstat, frame, op_ret, op_errno, iatt, xdata);
+
+	return 0;
+}
+
+
+int
+qb_fstat (call_frame_t *frame, xlator_t *this, fd_t *fd, dict_t *xdata)
+{
+	if (qb_inode_ctx_get (this, fd->inode))
+		frame->local = inode_ref (fd->inode);
+
+	STACK_WIND (frame, qb_fstat_cbk, FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->fstat, fd, xdata);
+	return 0;
+}
+
+
+int
+qb_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		int op_ret, int op_errno, struct iatt *pre, struct iatt *post,
+		dict_t *xdata)
+{
+	inode_t *inode = NULL;
+
+	inode = frame->local;
+	frame->local = NULL;
+
+	if (inode) {
+		qb_iatt_fixup (this, inode, pre);
+		qb_iatt_fixup (this, inode, post);
+		inode_unref (inode);
+	}
+
+	QB_STACK_UNWIND (setattr, frame, op_ret, op_errno, pre, post, xdata);
+
+	return 0;
+}
+
+
+int
+qb_setattr (call_frame_t *frame, xlator_t *this, loc_t *loc, struct iatt *buf,
+	    int valid, dict_t *xdata)
+{
+	if (qb_inode_ctx_get (this, loc->inode))
+		frame->local = inode_ref (loc->inode);
+
+	STACK_WIND (frame, qb_setattr_cbk, FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->setattr, loc, buf, valid, xdata);
+	return 0;
+}
+
+
+int
+qb_fsetattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		 int op_ret, int op_errno, struct iatt *pre, struct iatt *post,
+		 dict_t *xdata)
+{
+	inode_t *inode = NULL;
+
+	inode = frame->local;
+	frame->local = NULL;
+
+	if (inode) {
+		qb_iatt_fixup (this, inode, pre);
+		qb_iatt_fixup (this, inode, post);
+		inode_unref (inode);
+	}
+
+	QB_STACK_UNWIND (fsetattr, frame, op_ret, op_errno, pre, post, xdata);
+
+	return 0;
+}
+
+
+int
+qb_fsetattr (call_frame_t *frame, xlator_t *this, fd_t *fd, struct iatt *buf,
+	     int valid, dict_t *xdata)
+{
+	if (qb_inode_ctx_get (this, fd->inode))
+		frame->local = inode_ref (fd->inode);
+
+	STACK_WIND (frame, qb_setattr_cbk, FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->fsetattr, fd, buf, valid, xdata);
+	return 0;
+}
+
+
+int
+qb_forget (xlator_t *this, inode_t *inode)
+{
+	return qb_inode_cleanup (this, inode, 0);
+}
+
+
+int
+qb_release (xlator_t *this, fd_t *fd)
+{
+	call_frame_t *frame = NULL;
+
+	frame = create_frame (this, this->ctx->pool);
+	if (!frame) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"Could not allocate frame. "
+			"Leaking QEMU BlockDriverState");
+		return -1;
+	}
+
+	if (qb_local_init (frame) != 0) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"Could not allocate local. "
+			"Leaking QEMU BlockDriverState");
+		STACK_DESTROY (frame->root);
+		return -1;
+	}
+
+	if (qb_coroutine (frame, qb_co_close) != 0) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"Could not allocate coroutine. "
+			"Leaking QEMU BlockDriverState");
+		qb_local_free (this, frame->local);
+		frame->local = NULL;
+		STACK_DESTROY (frame->root);
+	}
+
+	return 0;
+}
+
+int
+mem_acct_init (xlator_t *this)
+{
+        int     ret = -1;
+
+        ret = xlator_mem_acct_init (this, gf_qb_mt_end + 1);
+
+        if (ret)
+                gf_log (this->name, GF_LOG_ERROR, "Memory accounting init "
+                        "failed");
+        return ret;
+}
+
+
+int
+reconfigure (xlator_t *this, dict_t *options)
+{
+	return 0;
+}
+
+
+int
+init (xlator_t *this)
+{
+        qb_conf_t *conf    = NULL;
+        int32_t    ret     = -1;
+	static int bdrv_inited = 0;
+
+        if (!this->children || this->children->next) {
+                gf_log (this->name, GF_LOG_ERROR,
+                        "FATAL: qemu-block (%s) not configured with exactly "
+                        "one child", this->name);
+                goto out;
+        }
+
+        conf = GF_CALLOC (1, sizeof (*conf), gf_qb_mt_qb_conf_t);
+        if (!conf)
+                goto out;
+
+        /* configure 'option window-size <size>' */
+        GF_OPTION_INIT ("default-password", conf->default_password, str, out);
+
+	/* qemu coroutines use "co_mutex" for synchronizing among themselves.
+	   However "co_mutex" itself is not threadsafe if the coroutine framework
+	   is multithreaded (which usually is not). However synctasks are
+	   fundamentally multithreaded, so for now create a syncenv which has
+	   scaling limits set to max 1 thread so that the qemu coroutines can
+	   execute "safely".
+
+	   Future work: provide an implementation of "co_mutex" which is
+	   threadsafe and use the global multithreaded ctx->env syncenv.
+	*/
+	conf->env = syncenv_new (0, 1, 1);
+
+        this->private = conf;
+
+        ret = 0;
+
+	snprintf (conf->qb_xattr_key, QB_XATTR_KEY_MAX, QB_XATTR_KEY_FMT,
+		  this->name);
+
+	cur_mon = (void *) 1;
+
+	if (!bdrv_inited) {
+		bdrv_init ();
+		bdrv_inited = 1;
+	}
+
+out:
+        if (ret)
+                GF_FREE (conf);
+
+        return ret;
+}
+
+
+void
+fini (xlator_t *this)
+{
+        qb_conf_t *conf = NULL;
+
+        conf = this->private;
+
+        this->private = NULL;
+
+	if (conf->root_inode)
+		inode_unref(conf->root_inode);
+        GF_FREE (conf);
+
+	return;
+}
+
+
+struct xlator_fops fops = {
+	.lookup      = qb_lookup,
+	.fsetxattr   = qb_fsetxattr,
+	.setxattr    = qb_setxattr,
+	.open        = qb_open,
+        .writev      = qb_writev,
+        .readv       = qb_readv,
+        .fsync       = qb_fsync,
+        .truncate    = qb_truncate,
+        .ftruncate   = qb_ftruncate,
+        .stat        = qb_stat,
+        .fstat       = qb_fstat,
+	.setattr     = qb_setattr,
+	.fsetattr    = qb_fsetattr,
+        .flush       = qb_flush,
+/*
+	.getxattr    = qb_getxattr,
+	.fgetxattr   = qb_fgetxattr
+*/
+	.readdirp    = qb_readdirp,
+};
+
+
+struct xlator_cbks cbks = {
+        .forget   = qb_forget,
+	.release  = qb_release,
+};
+
+
+struct xlator_dumpops dumpops = {
+};
+
+
+struct volume_options options[] = {
+        { .key  = {"default-password"},
+          .type = GF_OPTION_TYPE_STR,
+          .default_value = "",
+          .description = "Default password for the AES encrypted block images."
+        },
+        { .key = {NULL} },
+};
diff --git a/xlators/features/qemu-block/src/qemu-block.h b/xlators/features/qemu-block/src/qemu-block.h
new file mode 100644
index 000000000..c95f2799a
--- /dev/null
+++ b/xlators/features/qemu-block/src/qemu-block.h
@@ -0,0 +1,109 @@
+/*
+  Copyright (c) 2013 Red Hat, Inc. <http://www.redhat.com>
+  This file is part of GlusterFS.
+
+  This file is licensed to you under your choice of the GNU Lesser
+  General Public License, version 3 or any later version (LGPLv3 or
+  later), or the GNU General Public License, version 2 (GPLv2), in all
+  cases as published by the Free Software Foundation.
+*/
+
+#ifndef __QEMU_BLOCK_H
+#define __QEMU_BLOCK_H
+
+#include "syncop.h"
+#include "call-stub.h"
+#include "block/block_int.h"
+#include "monitor/monitor.h"
+
+/* QB_XATTR_KEY_FMT is the on-disk xattr stored in the inode which
+   indicates that the file must be "interpreted" by the block format
+   logic. The value of the key is of the pattern:
+
+   "format:virtual_size"
+
+   e.g
+
+   "qcow2:20GB" or "qed:100GB"
+
+   The format and virtual size are colon separated. The format is
+   a case sensitive string which qemu recognizes. virtual_size is
+   specified as a size which glusterfs recognizes as size (i.e.,
+   value accepted by gf_string2bytesize())
+*/
+#define QB_XATTR_KEY_FMT "trusted.glusterfs.%s.format"
+
+#define QB_XATTR_KEY_MAX 64
+
+#define QB_XATTR_VAL_MAX 64
+
+
+typedef struct qb_inode {
+	char fmt[QB_XATTR_VAL_MAX]; /* this is only the format, not "format:size" */
+	size_t size; /* virtual size in bytes */
+	BlockDriverState *bs;
+	int refcnt;
+	uuid_t backing_gfid;
+	char *backing_fname;
+} qb_inode_t;
+
+
+typedef struct qb_conf {
+	Monitor *mon;
+	struct syncenv *env;
+	char qb_xattr_key[QB_XATTR_KEY_MAX];
+	char *default_password;
+	inode_t *root_inode;
+} qb_conf_t;
+
+
+typedef struct qb_local {
+	call_frame_t      *frame; /* backpointer */
+	call_stub_t       *stub;
+	inode_t           *inode;
+	fd_t              *fd;
+	char               fmt[QB_XATTR_VAL_MAX+1];
+	char               name[256];
+	synctask_fn_t      synctask_fn;
+	struct list_head   list;
+} qb_local_t;
+
+void qb_local_free (xlator_t *this, qb_local_t *local);
+int qb_coroutine (call_frame_t *frame, synctask_fn_t fn);
+inode_t *qb_inode_from_filename (const char *filename);
+int qb_inode_to_filename (inode_t *inode, char *filename, int size);
+int qb_format_extract (xlator_t *this, char *format, inode_t *inode);
+
+qb_inode_t *qb_inode_ctx_get (xlator_t *this, inode_t *inode);
+
+#define QB_STACK_UNWIND(typ, frame, args ...) do {	\
+	qb_local_t *__local = frame->local;		\
+	xlator_t *__this = frame->this;			\
+							\
+	frame->local = NULL;				\
+	STACK_UNWIND_STRICT (typ, frame, args);		\
+	if (__local)					\
+		qb_local_free (__this, __local);	\
+	} while (0)
+
+#define QB_STUB_UNWIND(stub, op_ret, op_errno) do {	\
+	qb_local_t *__local = stub->frame->local;	\
+	xlator_t *__this = stub->frame->this;		\
+							\
+	stub->frame->local = NULL;			\
+	call_unwind_error (stub, op_ret, op_errno);	\
+	if (__local)					\
+		qb_local_free (__this, __local);	\
+	} while (0)
+
+#define QB_STUB_RESUME(stub_errno) do {			\
+	qb_local_t *__local = stub->frame->local;	\
+	xlator_t *__this = stub->frame->this;		\
+							\
+	stub->frame->local = NULL;			\
+	call_resume (stub);				\
+	if (__local)					\
+		qb_local_free (__this, __local);	\
+	} while (0)
+
+#endif /* !__QEMU_BLOCK_H */
diff --git a/xlators/features/quiesce/Makefile.am b/xlators/features/quiesce/Makefile.am
new file mode 100644
index 000000000..a985f42a8
--- /dev/null
+++ b/xlators/features/quiesce/Makefile.am
@@ -0,0 +1,3 @@
+SUBDIRS = src
+
+CLEANFILES =
diff --git a/xlators/features/quiesce/src/Makefile.am b/xlators/features/quiesce/src/Makefile.am
new file mode 100644
index 000000000..15e46629e
--- /dev/null
+++ b/xlators/features/quiesce/src/Makefile.am
@@ -0,0 +1,15 @@
+xlator_LTLIBRARIES = quiesce.la
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+quiesce_la_LDFLAGS = -module -avoid-version
+
+quiesce_la_SOURCES = quiesce.c
+quiesce_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la
+
+noinst_HEADERS = quiesce.h quiesce-mem-types.h
+
+AM_CPPFLAGS = $(GF_CPPFLAGS) -I$(top_srcdir)/libglusterfs/src
+
+AM_CFLAGS = -Wall $(GF_CFLAGS)
+
+CLEANFILES =
diff --git a/xlators/features/quiesce/src/quiesce-mem-types.h b/xlators/features/quiesce/src/quiesce-mem-types.h
new file mode 100644
index 000000000..6e582f424
--- /dev/null
+++ b/xlators/features/quiesce/src/quiesce-mem-types.h
@@ -0,0 +1,20 @@
+/*
+   Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+
+#ifndef __QUIESCE_MEM_TYPES_H__
+#define __QUIESCE_MEM_TYPES_H__
+
+#include "mem-types.h"
+
+enum gf_quiesce_mem_types_ {
+        gf_quiesce_mt_priv_t = gf_common_mt_end + 1,
+        gf_quiesce_mt_end
+};
+#endif
diff --git a/xlators/features/quiesce/src/quiesce.c b/xlators/features/quiesce/src/quiesce.c
new file mode 100644
index 000000000..24c7dc6ed
--- /dev/null
+++ b/xlators/features/quiesce/src/quiesce.c
@@ -0,0 +1,2610 @@
+/*
+   Copyright (c) 2010-2012 Red Hat, Inc. <http://www.redhat.com>
+   This file is part of GlusterFS.
+
+   This file is licensed to you under your choice of the GNU Lesser
+   General Public License, version 3 or any later version (LGPLv3 or
+   later), or the GNU General Public License, version 2 (GPLv2), in all
+   cases as published by the Free Software Foundation.
+*/
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "quiesce.h"
+#include "defaults.h"
+#include "call-stub.h"
+
+/* TODO: */
+/* Think about 'writev/_*_lk/setattr/xattrop/' fops to do re-transmittion */
+
+
+/* Quiesce Specific Functions */
+void
+gf_quiesce_local_wipe (xlator_t *this, quiesce_local_t *local)
+{
+        if (!local || !this || !this->private)
+                return;
+
+        if (local->loc.inode)
+                loc_wipe (&local->loc);
+        if (local->fd)
+                fd_unref (local->fd);
+        GF_FREE (local->name);
+        GF_FREE (local->volname);
+        if (local->dict)
+                dict_unref (local->dict);
+        if (local->iobref)
+                iobref_unref (local->iobref);
+        GF_FREE (local->vector);
+
+        mem_put (local);
+}
+
+call_stub_t *
+gf_quiesce_dequeue (xlator_t *this)
+{
+        call_stub_t  *stub = NULL;
+        quiesce_priv_t *priv = NULL;
+
+        priv = this->private;
+
+        if (!priv || list_empty (&priv->req))
+                return NULL;
+
+        LOCK (&priv->lock);
+        {
+                stub = list_entry (priv->req.next, call_stub_t, list);
+                list_del_init (&stub->list);
+                priv->queue_size--;
+        }
+        UNLOCK (&priv->lock);
+
+        return stub;
+}
+
+void *
+gf_quiesce_dequeue_start (void *data)
+{
+        xlator_t       *this = NULL;
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        this = data;
+        priv = this->private;
+        THIS = this;
+
+        while (!list_empty (&priv->req)) {
+                stub = gf_quiesce_dequeue (this);
+                if (stub) {
+                        call_resume (stub);
+                }
+        }
+
+        return 0;
+}
+
+
+void
+gf_quiesce_timeout (void *data)
+{
+        xlator_t       *this = NULL;
+        quiesce_priv_t *priv = NULL;
+
+        this = data;
+        priv = this->private;
+        THIS = this;
+
+        LOCK (&priv->lock);
+        {
+                priv->pass_through = _gf_true;
+        }
+        UNLOCK (&priv->lock);
+
+        gf_quiesce_dequeue_start (this);
+
+        return;
+}
+
+void
+gf_quiesce_enqueue (xlator_t *this, call_stub_t *stub)
+{
+        quiesce_priv_t *priv    = NULL;
+        struct timespec timeout = {0,};
+
+        priv = this->private;
+        if (!priv) {
+                gf_log_callingfn (this->name, GF_LOG_ERROR,
+                                  "this->private == NULL");
+                return;
+        }
+
+        LOCK (&priv->lock);
+        {
+                list_add_tail (&stub->list, &priv->req);
+                priv->queue_size++;
+        }
+        UNLOCK (&priv->lock);
+
+        if (!priv->timer) {
+                timeout.tv_sec = 20;
+                timeout.tv_nsec = 0;
+
+                priv->timer = gf_timer_call_after (this->ctx,
+                                                   timeout,
+                                                   gf_quiesce_timeout,
+                                                   (void *) this);
+        }
+
+        return;
+}
+
+
+
+/* _CBK function section */
+
+int32_t
+quiesce_lookup_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, inode_t *inode,
+                    struct iatt *buf, dict_t *dict, struct iatt *postparent)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_lookup_stub (frame, default_lookup_resume,
+                                        &local->loc, local->dict);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (lookup, frame, -1, ENOMEM,
+                                             NULL, NULL, NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (lookup, frame, op_ret, op_errno, inode, buf,
+                             dict, postparent);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_stat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, struct iatt *buf,
+                  dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_stat_stub (frame, default_stat_resume,
+                                      &local->loc, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (stat, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (stat, frame, op_ret, op_errno, buf, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_access_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_access_stub (frame, default_access_resume,
+                                        &local->loc, local->flag, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (access, frame, -1, ENOMEM, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (access, frame, op_ret, op_errno, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_readlink_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, const char *path,
+                      struct iatt *buf, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_readlink_stub (frame, default_readlink_resume,
+                                          &local->loc, local->size, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (readlink, frame, -1, ENOMEM,
+                                             NULL, NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (readlink, frame, op_ret, op_errno, path, buf, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_open_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                  int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_open_stub (frame, default_open_resume,
+                                      &local->loc, local->flag, local->fd,
+                                      xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (open, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (open, frame, op_ret, op_errno, fd, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_readv_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iovec *vector,
+                   int32_t count, struct iatt *stbuf, struct iobref *iobref, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_readv_stub (frame, default_readv_resume,
+                                       local->fd, local->size, local->offset,
+                                       local->io_flag, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (readv, frame, -1, ENOMEM,
+                                             NULL, 0, NULL, NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (readv, frame, op_ret, op_errno, vector, count,
+                             stbuf, iobref, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_flush_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_flush_stub (frame, default_flush_resume,
+                                       local->fd, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (flush, frame, -1, ENOMEM, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (flush, frame, op_ret, op_errno, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+
+
+int32_t
+quiesce_fsync_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                   struct iatt *postbuf, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_fsync_stub (frame, default_fsync_resume,
+                                       local->fd, local->flag, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (fsync, frame, -1, ENOMEM,
+                                             NULL, NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (fsync, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_fstat_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                   int32_t op_ret, int32_t op_errno, struct iatt *buf, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_fstat_stub (frame, default_fstat_resume,
+                                       local->fd, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (fstat, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (fstat, frame, op_ret, op_errno, buf, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_opendir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, fd_t *fd, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_opendir_stub (frame, default_opendir_resume,
+                                         &local->loc, local->fd, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (opendir, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (opendir, frame, op_ret, op_errno, fd, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_fsyncdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_fsyncdir_stub (frame, default_fsyncdir_resume,
+                                          local->fd, local->flag, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (fsyncdir, frame, -1, ENOMEM, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (fsyncdir, frame, op_ret, op_errno, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_statfs_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, struct statvfs *buf, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_statfs_stub (frame, default_statfs_resume,
+                                        &local->loc, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (statfs, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (statfs, frame, op_ret, op_errno, buf, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_fgetxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_fgetxattr_stub (frame, default_fgetxattr_resume,
+                                           local->fd, local->name, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (fgetxattr, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (fgetxattr, frame, op_ret, op_errno, dict, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_getxattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_getxattr_stub (frame, default_getxattr_resume,
+                                          &local->loc, local->name, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (getxattr, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (getxattr, frame, op_ret, op_errno, dict, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_rchecksum_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                       int32_t op_ret, int32_t op_errno, uint32_t weak_checksum,
+                       uint8_t *strong_checksum, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_rchecksum_stub (frame, default_rchecksum_resume,
+                                           local->fd, local->offset, local->flag, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (rchecksum, frame, -1, ENOMEM,
+                                             0, NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (rchecksum, frame, op_ret, op_errno, weak_checksum,
+                             strong_checksum, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_readdir_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, gf_dirent_t *entries, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_readdir_stub (frame, default_readdir_resume,
+                                         local->fd, local->size, local->offset, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (readdir, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (readdir, frame, op_ret, op_errno, entries, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_readdirp_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, gf_dirent_t *entries, dict_t *xdata)
+{
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_readdirp_stub (frame, default_readdirp_resume,
+                                          local->fd, local->size, local->offset,
+                                          local->dict);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (readdirp, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (readdirp, frame, op_ret, op_errno, entries, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+
+#if 0
+
+int32_t
+quiesce_writev_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                    int32_t op_ret, int32_t op_errno, struct iatt *prebuf,
+                    struct iatt *postbuf, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_writev_stub (frame, default_writev_resume,
+                                        local->fd, local->vector, local->flag,
+                                        local->offset, local->io_flags,
+                                        local->iobref, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (writev, frame, -1, ENOMEM,
+                                             NULL, NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (writev, frame, op_ret, op_errno, prebuf, postbuf, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_xattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_xattrop_stub (frame, default_xattrop_resume,
+                                         &local->loc, local->xattrop_flags,
+                                         local->dict, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (xattrop, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (xattrop, frame, op_ret, op_errno, dict, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_fxattrop_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *dict, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_fxattrop_stub (frame, default_fxattrop_resume,
+                                          local->fd, local->xattrop_flags,
+                                          local->dict, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (fxattrop, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (fxattrop, frame, op_ret, op_errno, dict, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_lk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                int32_t op_ret, int32_t op_errno, struct gf_flock *lock, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_lk_stub (frame, default_lk_resume,
+                                    local->fd, local->flag, &local->flock, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (lk, frame, -1, ENOMEM,
+                                             NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (lk, frame, op_ret, op_errno, lock, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_inodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_inodelk_stub (frame, default_inodelk_resume,
+                                         local->volname, &local->loc,
+                                         local->flag, &local->flock, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (inodelk, frame, -1, ENOMEM, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (inodelk, frame, op_ret, op_errno, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+
+int32_t
+quiesce_finodelk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_finodelk_stub (frame, default_finodelk_resume,
+                                         local->volname, local->fd,
+                                         local->flag, &local->flock, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (finodelk, frame, -1, ENOMEM, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (finodelk, frame, op_ret, op_errno, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_entrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_entrylk_stub (frame, default_entrylk_resume,
+                                         local->volname, &local->loc,
+                                         local->name, local->cmd, local->type, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (entrylk, frame, -1, ENOMEM, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (entrylk, frame, op_ret, op_errno, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_fentrylk_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_fentrylk_stub (frame, default_fentrylk_resume,
+                                          local->volname, local->fd,
+                                          local->name, local->cmd, local->type, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (fentrylk, frame, -1, ENOMEM, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (fentrylk, frame, op_ret, op_errno, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_setattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                     int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+                     struct iatt *statpost, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_setattr_stub (frame, default_setattr_resume,
+                                         &local->loc, &local->stbuf, local->flag, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (setattr, frame, -1, ENOMEM,
+                                             NULL, NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (setattr, frame, op_ret, op_errno, statpre,
+                             statpost, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+int32_t
+quiesce_fsetattr_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+                      int32_t op_ret, int32_t op_errno, struct iatt *statpre,
+                      struct iatt *statpost, dict_t *xdata)
+{
+        quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+        quiesce_local_t *local = NULL;
+
+        priv = this->private;
+
+        local = frame->local;
+        frame->local = NULL;
+
+        if ((op_ret == -1) && (op_errno == ENOTCONN)) {
+                /* Re-transmit (by putting in the queue) */
+                stub = fop_fsetattr_stub (frame, default_fsetattr_resume,
+                                          local->fd, &local->stbuf, local->flag, xdata);
+                if (!stub) {
+                        STACK_UNWIND_STRICT (fsetattr, frame, -1, ENOMEM,
+                                             NULL, NULL, NULL);
+                        goto out;
+                }
+
+                gf_quiesce_enqueue (this, stub);
+                goto out;
+        }
+
+        STACK_UNWIND_STRICT (fsetattr, frame, op_ret, op_errno, statpre,
+                             statpost, xdata);
+out:
+        gf_quiesce_local_wipe (this, local);
+
+        return 0;
+}
+
+#endif /* if 0 */
+
+
+/* FOP */
+
+/* No retransmittion */
+
+int32_t
+quiesce_removexattr (call_frame_t *frame,
+		     xlator_t *this,
+		     loc_t *loc,
+		     const char *name, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                STACK_WIND (frame,
+                            default_removexattr_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->removexattr,
+                            loc,
+                            name, xdata);
+                return 0;
+        }
+
+        stub = fop_removexattr_stub (frame, default_removexattr_resume,
+                                     loc, name, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (removexattr, frame, -1, ENOMEM, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_truncate (call_frame_t *frame,
+		  xlator_t *this,
+		  loc_t *loc,
+		  off_t offset, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                STACK_WIND (frame,
+                            default_truncate_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->truncate,
+                            loc,
+                            offset, xdata);
+                return 0;
+        }
+
+        stub = fop_truncate_stub (frame, default_truncate_resume, loc, offset, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (truncate, frame, -1, ENOMEM, NULL, NULL, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_fsetxattr (call_frame_t *frame,
+                   xlator_t *this,
+                   fd_t *fd,
+                   dict_t *dict,
+                   int32_t flags, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                STACK_WIND (frame,
+                            default_fsetxattr_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->fsetxattr,
+                            fd,
+                            dict,
+                            flags, xdata);
+                return 0;
+        }
+
+        stub = fop_fsetxattr_stub (frame, default_fsetxattr_resume,
+                                   fd, dict, flags, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (fsetxattr, frame, -1, ENOMEM, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_setxattr (call_frame_t *frame,
+		  xlator_t *this,
+		  loc_t *loc,
+		  dict_t *dict,
+		  int32_t flags, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                STACK_WIND (frame,
+                            default_setxattr_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->setxattr,
+                            loc,
+                            dict,
+                            flags, xdata);
+                return 0;
+        }
+
+        stub = fop_setxattr_stub (frame, default_setxattr_resume,
+                                  loc, dict, flags, xdata);
+        if (!stub) {
+                STACK_UNWIND_STRICT (setxattr, frame, -1, ENOMEM, NULL);
+                return 0;
+        }
+
+        gf_quiesce_enqueue (this, stub);
+
+        return 0;
+}
+
+int32_t
+quiesce_create (call_frame_t *frame, xlator_t *this,
+		loc_t *loc, int32_t flags, mode_t mode,
+                mode_t umask, fd_t *fd, dict_t *xdata)
+{
+	quiesce_priv_t *priv = NULL;
+        call_stub_t    *stub = NULL;
+
+        priv = this->private;
+
+        if (priv->pass_through) {
+                /* Don't send O_APPEND below, as write() re-transmittions can
+                   fail with O_APPEND */
+                STACK_WIND (frame, default_create_cbk,
+                            FIRST_CHILD(this),
+                            FIRST_CHILD(this)->fops->create,