Added all files

6 files changed, 2347 insertions, 0 deletions

diff --git a/xlators/features/locks/src/Makefile.am b/xlators/features/locks/src/Makefile.am
new file mode 100644
index 00000000000..ec4a953eb91
--- /dev/null
+++ b/xlators/features/locks/src/Makefile.am
@@ -0,0 +1,20 @@
+xlator_LTLIBRARIES = locks.la
+xlatordir = $(libdir)/glusterfs/$(PACKAGE_VERSION)/xlator/features
+
+locks_la_LDFLAGS = -module -avoidversion
+
+locks_la_SOURCES = common.c posix.c internal.c
+locks_la_LIBADD = $(top_builddir)/libglusterfs/src/libglusterfs.la 
+
+noinst_HEADERS = locks.h common.h
+
+AM_CFLAGS = -fPIC -D_FILE_OFFSET_BITS=64 -D_GNU_SOURCE -Wall -fno-strict-aliasing -D$(GF_HOST_OS) \
+	-I$(top_srcdir)/libglusterfs/src $(GF_CFLAGS) -shared -nostartfiles
+
+CLEANFILES = 
+
+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
diff --git a/xlators/features/locks/src/common.c b/xlators/features/locks/src/common.c
new file mode 100644
index 00000000000..9ac1250cc57
--- /dev/null
+++ b/xlators/features/locks/src/common.c
@@ -0,0 +1,561 @@
+/*
+  Copyright (c) 2006, 2007, 2008 Z RESEARCH, Inc. <http://www.zresearch.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 <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"
+
+
+int
+pl_is_lock_grantable (pl_inode_t *pl_inode, posix_lock_t *lock,
+		      gf_lk_domain_t dom);
+static void
+__insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock,
+		    gf_lk_domain_t dom);
+
+
+pl_inode_t *
+pl_inode_get (xlator_t *this, inode_t *inode)
+{
+	pl_inode_t *pl_inode     = NULL;
+	mode_t      st_mode      = 0;
+	uint64_t    tmp_pl_inode = 0;
+	int         ret          = 0;
+	
+	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 out;
+		}
+
+		pl_inode = CALLOC (1, sizeof (*pl_inode));
+		if (!pl_inode) {
+			gf_log (this->name, GF_LOG_ERROR,
+				"out of memory :(");
+			goto out;
+		}
+
+		st_mode  = inode->st_mode;
+		if ((st_mode & S_ISGID) && !(st_mode & S_IXGRP))
+			pl_inode->mandatory = 1;
+
+
+		pthread_mutex_init (&pl_inode->mutex, NULL);
+		
+		INIT_LIST_HEAD (&pl_inode->dir_list);
+		INIT_LIST_HEAD (&pl_inode->ext_list);
+		INIT_LIST_HEAD (&pl_inode->int_list);
+		INIT_LIST_HEAD (&pl_inode->rw_list);
+
+		ret = inode_ctx_put (inode, this, (uint64_t)(long)pl_inode);
+	}
+out:
+	UNLOCK (&inode->lock);
+	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)
+{
+	posix_lock_t *lock = NULL;
+
+	lock = CALLOC (1, sizeof (posix_lock_t));
+	if (!lock) {
+		return NULL;
+	}
+
+	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;
+
+	lock->transport  = transport;
+	lock->client_pid = client_pid;
+
+	INIT_LIST_HEAD (&lock->list);
+
+	return lock;
+}
+
+
+/* Delete a lock from the inode's lock list */
+void
+__delete_lock (pl_inode_t *pl_inode, posix_lock_t *lock)
+{
+	list_del_init (&lock->list);
+}
+
+
+/* Destroy a posix_lock */
+void
+__destroy_lock (posix_lock_t *lock)
+{
+	free (lock);
+}
+
+
+/* Convert a posix_lock to a struct flock */
+void
+posix_lock_to_flock (posix_lock_t *lock, struct flock *flock)
+{
+	flock->l_pid   = lock->client_pid;
+	flock->l_type  = lock->fl_type;
+	flock->l_start = lock->fl_start;
+
+	if (lock->fl_end == 0)
+		flock->l_len = LLONG_MAX;
+	else
+		flock->l_len = lock->fl_end - lock->fl_start + 1;
+}
+
+
+/* Insert the lock into the inode's lock list */
+void
+pl_insert_lock (pl_inode_t *pl_inode, posix_lock_t *lock, gf_lk_domain_t dom)
+{
+	list_add_tail (&lock->list, DOMAIN_HEAD (pl_inode, dom));
+
+	return;
+}
+
+
+/* Return true if the locks overlap, false otherwise */
+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));
+}
+
+
+/* Return true if the locks have the same owner */
+int
+same_owner (posix_lock_t *l1, posix_lock_t *l2)
+{
+	return ((l1->client_pid == l2->client_pid) &&
+		(l1->transport  == l2->transport));
+}
+
+
+/* Delete all F_UNLCK locks */
+void
+__delete_unlck_locks (pl_inode_t *pl_inode, gf_lk_domain_t dom)
+{
+	posix_lock_t *l = NULL;
+	posix_lock_t *tmp = NULL;
+
+	list_for_each_entry_safe (l, tmp, DOMAIN_HEAD (pl_inode, dom), list) {
+		if (l->fl_type == F_UNLCK) {
+			__delete_lock (pl_inode, l);
+			__destroy_lock (l);
+		}
+	}
+}
+
+
+/* Add two locks */
+static posix_lock_t *
+add_locks (posix_lock_t *l1, posix_lock_t *l2)
+{
+	posix_lock_t *sum = NULL;
+
+	sum = CALLOC (1, sizeof (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);
+
+	return sum;
+}
+
+/* Subtract two locks */
+struct _values {
+	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_DEBUG, 
+			"unexpected case in subtract_locks");
+	}
+
+	return v;
+}
+
+/* 
+   Start searching from {begin}, and return the first lock that
+   conflicts, NULL if no conflict
+   If {begin} is NULL, then start from the beginning of the list
+*/
+static posix_lock_t *
+first_overlap (pl_inode_t *pl_inode, posix_lock_t *lock,
+	       gf_lk_domain_t dom)
+{
+	posix_lock_t *l = NULL;
+
+	list_for_each_entry (l, DOMAIN_HEAD (pl_inode, dom), list) {
+		if (l->blocked)
+			continue;
+
+		if (locks_overlap (l, lock))
+			return l;
+	}
+
+	return NULL;
+}
+
+
+
+/* Return true if lock is grantable */
+int
+pl_is_lock_grantable (pl_inode_t *pl_inode, posix_lock_t *lock,
+		      gf_lk_domain_t dom)
+{
+	posix_lock_t *l = NULL;
+	int           ret = 1;
+
+	list_for_each_entry (l, DOMAIN_HEAD (pl_inode, dom), list) {
+		if (!l->blocked && locks_overlap (lock, l)) {
+			if (((l->fl_type == F_WRLCK)
+			     || (lock->fl_type == F_WRLCK))
+			    && (lock->fl_type != F_UNLCK)
+			    && !same_owner (l, lock)) {
+				ret = 0;
+				break;
+			}
+		}
+	}
+	return ret;
+}
+
+
+extern void do_blocked_rw (pl_inode_t *);
+
+
+static void
+__insert_and_merge (pl_inode_t *pl_inode, posix_lock_t *lock,
+		    gf_lk_domain_t dom)
+{
+	posix_lock_t  *conf = NULL;
+	posix_lock_t  *t = NULL;
+	posix_lock_t  *sum = NULL;
+	int            i = 0;
+	struct _values v = { .locks = {0, 0, 0} };
+
+	list_for_each_entry_safe (conf, t, DOMAIN_HEAD (pl_inode, dom), list) {
+		if (!locks_overlap (conf, lock))
+			continue;
+
+		if (same_owner (conf, lock)) {
+			if (conf->fl_type == lock->fl_type) {
+				sum = add_locks (lock, conf);
+
+				sum->fl_type    = lock->fl_type;
+				sum->transport  = lock->transport;
+				sum->client_pid = lock->client_pid;
+
+				__delete_lock (pl_inode, conf); 
+				__destroy_lock (conf);
+
+				__destroy_lock (lock);
+				__insert_and_merge (pl_inode, sum, dom);
+
+				return;
+			} else {
+				sum = add_locks (lock, conf);
+
+				sum->fl_type    = conf->fl_type;
+				sum->transport  = conf->transport;
+				sum->client_pid = conf->client_pid;
+
+				v = subtract_locks (sum, lock);
+	
+				__delete_lock (pl_inode, conf);
+				__destroy_lock (conf);
+
+				__delete_lock (pl_inode, lock);
+				__destroy_lock (lock);
+
+				__destroy_lock (sum);
+
+				for (i = 0; i < 3; i++) {
+					if (!v.locks[i])
+						continue;
+
+					if (v.locks[i]->fl_type == F_UNLCK) {
+						__destroy_lock (v.locks[i]);
+						continue;
+					}
+					__insert_and_merge (pl_inode,
+							    v.locks[i], dom);
+				}
+
+				__delete_unlck_locks (pl_inode, dom);
+				return; 
+			}
+		}
+
+		if (lock->fl_type == F_UNLCK) {
+			continue;
+		}
+
+		if ((conf->fl_type == F_RDLCK) && (lock->fl_type == F_RDLCK)) {
+			pl_insert_lock (pl_inode, lock, dom);
+			return;
+		}
+	}
+
+	/* no conflicts, so just insert */
+	if (lock->fl_type != F_UNLCK) {
+		pl_insert_lock (pl_inode, lock, dom);
+	} else {
+		__destroy_lock (lock);
+	}
+}
+
+
+void
+__grant_blocked_locks (xlator_t *this, pl_inode_t *pl_inode,
+		       gf_lk_domain_t dom, struct list_head *granted)
+{
+	struct list_head  tmp_list;
+	posix_lock_t     *l = NULL;
+	posix_lock_t     *tmp = NULL;
+	posix_lock_t     *conf = NULL;
+
+	INIT_LIST_HEAD (&tmp_list);
+
+	list_for_each_entry_safe (l, tmp, DOMAIN_HEAD (pl_inode, dom), list) {
+		if (l->blocked) {
+			conf = first_overlap (pl_inode, l, dom);
+			if (conf)
+				continue;
+
+			l->blocked = 0;
+			list_move_tail (&l->list, &tmp_list);
+		}
+	}
+
+	list_for_each_entry_safe (l, tmp, &tmp_list, list) {
+		list_del_init (&l->list);
+
+		if (pl_is_lock_grantable (pl_inode, l, dom)) {
+			conf = CALLOC (1, sizeof (*conf));
+
+			if (!conf) {
+				l->blocked = 1;
+				pl_insert_lock (pl_inode, l, dom);
+				continue;
+			}
+
+			conf->frame = l->frame;
+			l->frame = NULL;
+
+			posix_lock_to_flock (l, &conf->user_flock);
+
+			gf_log (this->name, GF_LOG_DEBUG,
+				"%s (pid=%d) %"PRId64" - %"PRId64" => Granted",
+				l->fl_type == F_UNLCK ? "Unlock" : "Lock",
+				l->client_pid,
+				l->user_flock.l_start,
+				l->user_flock.l_len);
+
+			__insert_and_merge (pl_inode, l, dom);
+
+			list_add (&conf->list, granted);
+		} else {
+			l->blocked = 1;
+			pl_insert_lock (pl_inode, l, dom);
+		}
+	}
+}
+
+
+void
+grant_blocked_locks (xlator_t *this, pl_inode_t *pl_inode, gf_lk_domain_t dom)
+{
+	struct list_head granted_list;
+	posix_lock_t     *tmp = NULL;
+	posix_lock_t     *lock = NULL;
+
+	INIT_LIST_HEAD (&granted_list);
+
+	pthread_mutex_lock (&pl_inode->mutex);
+	{
+		__grant_blocked_locks (this, pl_inode, dom, &granted_list);
+	}
+	pthread_mutex_unlock (&pl_inode->mutex);
+
+	list_for_each_entry_safe (lock, tmp, &granted_list, list) {
+		list_del_init (&lock->list);
+
+		STACK_UNWIND (lock->frame, 0, 0, &lock->user_flock);
+
+		FREE (lock);
+	}
+
+	return;
+}
+
+
+int
+pl_setlk (xlator_t *this, pl_inode_t *pl_inode, posix_lock_t *lock,
+	  int can_block,  gf_lk_domain_t dom)
+{
+	int              ret = 0;
+
+	errno = 0;
+
+	pthread_mutex_lock (&pl_inode->mutex);
+	{
+		if (pl_is_lock_grantable (pl_inode, lock, dom)) {
+			gf_log (this->name, GF_LOG_DEBUG,
+				"%s (pid=%d) %"PRId64" - %"PRId64" => OK",
+				lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+				lock->client_pid,
+				lock->user_flock.l_start,
+				lock->user_flock.l_len);
+			__insert_and_merge (pl_inode, lock, dom);
+		} else if (can_block) {
+			gf_log (this->name, GF_LOG_DEBUG,
+				"%s (pid=%d) %"PRId64" - %"PRId64" => Blocked",
+				lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+				lock->client_pid,
+				lock->user_flock.l_start,
+				lock->user_flock.l_len);
+			lock->blocked = 1;
+			pl_insert_lock (pl_inode, lock, dom);
+			ret = -1;
+		} else {
+			gf_log (this->name, GF_LOG_DEBUG,
+				"%s (pid=%d) %"PRId64" - %"PRId64" => NOK",
+				lock->fl_type == F_UNLCK ? "Unlock" : "Lock",
+				lock->client_pid,
+				lock->user_flock.l_start,
+				lock->user_flock.l_len);
+			errno = EAGAIN;
+			ret = -1;
+		}
+	}
+	pthread_mutex_unlock (&pl_inode->mutex);
+
+	grant_blocked_locks (this, pl_inode, dom);
+
+	do_blocked_rw (pl_inode);
+
+	return ret;
+}
+
+
+posix_lock_t *
+pl_getlk (pl_inode_t *pl_inode, posix_lock_t *lock, gf_lk_domain_t dom)
+{
+	posix_lock_t *conf = NULL;
+
+	conf = first_overlap (pl_inode, lock, dom);
+
+	if (conf == NULL) {
+		lock->fl_type = F_UNLCK;
+		return lock;
+	}
+
+	return conf;
+}
diff --git a/xlators/features/locks/src/common.h b/xlators/features/locks/src/common.h
new file mode 100644
index 00000000000..135f33011bf
--- /dev/null
+++ b/xlators/features/locks/src/common.h
@@ -0,0 +1,59 @@
+/*
+  Copyright (c) 2006, 2007, 2008 Z RESEARCH, Inc. <http://www.zresearch.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 __COMMON_H__
+#define __COMMON_H__
+
+posix_lock_t *
+new_posix_lock (struct flock *flock, transport_t *transport, pid_t client_pid);
+
+pl_inode_t *
+pl_inode_get (xlator_t *this, inode_t *inode);
+
+posix_lock_t *
+pl_getlk (pl_inode_t *inode, posix_lock_t *lock, gf_lk_domain_t domain);
+
+int
+pl_setlk (xlator_t *this, pl_inode_t *inode, posix_lock_t *lock,
+	  int can_block, gf_lk_domain_t domain);
+
+int
+pl_is_lock_grantable (pl_inode_t *pl_inode, posix_lock_t *lock,
+		      gf_lk_domain_t dom);
+
+void
+pl_insert_lock (pl_inode_t *pl_inode, posix_lock_t *lock, gf_lk_domain_t dom);
+
+void
+grant_blocked_locks (xlator_t *this, pl_inode_t *inode, gf_lk_domain_t domain);
+
+void
+posix_lock_to_flock (posix_lock_t *lock, struct flock *flock);
+
+int
+locks_overlap (posix_lock_t *l1, posix_lock_t *l2);
+
+int
+same_owner (posix_lock_t *l1, posix_lock_t *l2);
+
+void __delete_lock (pl_inode_t *, posix_lock_t *);
+
+void __destroy_lock (posix_lock_t *);
+
+#endif /* __COMMON_H__ */
diff --git a/xlators/features/locks/src/internal.c b/xlators/features/locks/src/internal.c
new file mode 100644
index 00000000000..7f454a78e22
--- /dev/null
+++ b/xlators/features/locks/src/internal.c
@@ -0,0 +1,762 @@
+/*
+  Copyright (c) 2006, 2007, 2008 Z RESEARCH, Inc. <http://www.zresearch.com>
+  This file is part of GlusterFS.
+
+  GlusterFS is free software; you can redistribute it and/or modify
+  it under the terms of the GNU General Public License as published
+  by the Free Software Foundation; either version 3 of the License,
+  or (at your option) any later version.
+
+  GlusterFS is distributed in the hope that it will be useful, but
+  WITHOUT ANY WARRANTY; without even the implied warranty of
+  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+  General Public License for more details.
+
+  You should have received a copy of the GNU General Public License
+  along with this program.  If not, see
+  <http://www.gnu.org/licenses/>.
+*/
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "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"
+
+
+
+static int
+delete_locks_of_transport (pl_inode_t *pinode, transport_t *trans)
+{
+	posix_lock_t *tmp = NULL;
+	posix_lock_t *l = NULL;
+
+	list_for_each_entry_safe (l, tmp, &pinode->dir_list, list) {
+		if (l->transport == trans) {
+			__delete_lock (pinode, tmp);
+			__destroy_lock (tmp);
+		}
+	}
+
+	return 0;
+}
+
+
+static posix_lock_t *
+__find_exact_matching_lock (pl_inode_t *pinode, posix_lock_t *lock)
+{
+	posix_lock_t *l     = NULL;
+	posix_lock_t *match = NULL;
+
+	list_for_each_entry (l, DOMAIN_HEAD (pinode, GF_LOCK_INTERNAL), list) {
+		if (same_owner (l, lock)
+		    && (l->fl_start == lock->fl_start)
+		    && (l->fl_end   == lock->fl_end)) {
+			match = l;
+			break;
+		}
+	}
+
+	return match;
+}
+
+/**
+ * pl_inodelk: 
+ *
+ * This fop provides fcntl-style locking on files for internal
+ * purposes. Locks held through this fop reside in a domain different
+ * from those held by applications. This fop is for the use of AFR.
+ */
+
+
+static int
+pl_inodelk_common (call_frame_t *frame, xlator_t *this,
+		   inode_t *inode, int32_t cmd, struct flock *flock)
+{
+	int32_t op_ret   = -1;
+	int32_t op_errno = 0;
+	int     can_block = 0;
+
+	posix_locks_private_t * priv       = NULL;
+	transport_t *           transport  = NULL;
+	pid_t                   client_pid = -1;
+	pl_inode_t *            pinode     = NULL;
+	
+	posix_lock_t *          reqlock    = NULL;
+	posix_lock_t *          matchlock  = NULL;  /* steady, fire! */
+
+	VALIDATE_OR_GOTO (frame, unwind);
+	VALIDATE_OR_GOTO (inode, unwind);
+	VALIDATE_OR_GOTO (flock, unwind);
+	
+	if ((flock->l_start < 0) || (flock->l_len < 0)) {
+		op_errno = EINVAL;
+		goto unwind;
+	}
+
+	transport  = frame->root->trans;
+	client_pid = frame->root->pid;
+
+	priv = (posix_locks_private_t *) this->private;
+
+	VALIDATE_OR_GOTO (priv, unwind);
+
+	pinode = pl_inode_get (this, inode);
+	if (!pinode) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"out of memory :(");
+		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_DEBUG,
+			"releasing all locks from transport %p", transport);
+
+		delete_locks_of_transport (pinode, transport);
+		goto unwind;
+	}
+
+	reqlock = new_posix_lock (flock, transport, client_pid);
+	if (!reqlock) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"out of memory :(");
+		op_ret = -1;
+		op_errno = ENOMEM;
+		goto unwind;
+	}
+
+	pthread_mutex_lock (&pinode->mutex);
+	{
+		switch (cmd) {
+		case F_SETLKW:
+			can_block = 1;
+			reqlock->frame = frame;
+			reqlock->this  = this;
+			
+			/* fall through */
+			
+		case F_SETLK:
+			memcpy (&reqlock->user_flock, flock, sizeof (struct flock));
+			
+			switch (flock->l_type) {
+				
+			case F_WRLCK:
+				if (!pl_is_lock_grantable (pinode, reqlock, GF_LOCK_INTERNAL)) {
+					if (can_block) {
+						gf_log (this->name, GF_LOG_DEBUG,
+							"%s (pid=%d) %"PRId64" - %"PRId64" => blocked",
+							reqlock->fl_type == F_UNLCK ? "unlock" : "lock",
+							reqlock->client_pid,
+							reqlock->user_flock.l_start,
+							reqlock->user_flock.l_len);
+						pl_insert_lock (pinode, reqlock, GF_LOCK_INTERNAL);
+						
+						goto unlock;
+					}
+					
+					__destroy_lock (reqlock);
+					
+					
+					gf_log (this->name, GF_LOG_DEBUG,
+						"%s (pid=%d) %"PRId64" - %"PRId64" => NOK",
+						reqlock->fl_type == F_UNLCK ? "unlock" : "lock",
+						reqlock->client_pid, reqlock->user_flock.l_start,
+						reqlock->user_flock.l_len);
+					op_errno = EAGAIN;
+					
+					goto unlock;
+				}
+				
+				gf_log (this->name, GF_LOG_DEBUG,
+					"%s (pid=%d) %"PRId64" - %"PRId64" => OK",
+					reqlock->fl_type == F_UNLCK ? "unlock" : "lock",
+					reqlock->client_pid,
+					reqlock->user_flock.l_start,
+					reqlock->user_flock.l_len);
+				pl_insert_lock (pinode, reqlock, GF_LOCK_INTERNAL);
+				
+				break;
+				
+			case F_UNLCK:
+				matchlock = __find_exact_matching_lock (pinode, reqlock);
+
+				__destroy_lock (reqlock);
+				if (!matchlock) {
+					op_errno = EINVAL;
+					goto unlock;
+				}
+				
+				__delete_lock (pinode, matchlock);
+				__destroy_lock (matchlock);
+				
+				break;
+				
+			default:
+				op_errno = ENOTSUP;
+				gf_log (this->name, GF_LOG_ERROR,
+					"lock type %d not supported for [F]INODELK",
+					flock->l_type);
+				goto unlock;
+			}
+			
+			
+			break;
+			
+		default:
+			op_errno = ENOTSUP;
+			gf_log (this->name, GF_LOG_ERROR,
+				"lock command F_GETLK not supported for [F]INODELK (cmd=%d)", 
+				cmd);
+			goto unlock;
+		}
+		
+		op_ret = 0;
+		
+	unlock:
+		if (pinode)
+			pthread_mutex_unlock (&pinode->mutex);
+	}
+
+unwind:
+	STACK_UNWIND (frame, op_ret, op_errno);
+	return 0;
+}
+
+
+int
+pl_inodelk (call_frame_t *frame, xlator_t *this,
+	    loc_t *loc, int32_t cmd, struct flock *flock)
+{
+	return pl_inodelk_common (frame, this, loc->inode, cmd, flock);
+}
+
+
+int
+pl_finodelk (call_frame_t *frame, xlator_t *this,
+	     fd_t *fd, int32_t cmd, struct flock *flock)
+{
+	return pl_inodelk_common (frame, this, fd->inode, cmd, flock);
+}
+
+
+/**
+ * types_conflict - do two types of lock conflict?
+ * @t1: type
+ * @t2: type
+ *
+ * two read locks do not conflict
+ * any other case conflicts
+ */
+
+static int
+types_conflict (entrylk_type t1, entrylk_type t2)
+{
+	return !((t1 == ENTRYLK_RDLCK) && (t2 == ENTRYLK_RDLCK));
+}
+
+/**
+ * all_names - does a basename represent all names?
+ * @basename: name to check
+ */
+
+#define all_names(basename) ((basename == NULL) ? 1 : 0)
+
+/**
+ * names_conflict - do two names conflict?
+ * @n1: name
+ * @n2: name
+ */
+
+static int 
+names_conflict (const char *n1, const char *n2)
+{
+	return all_names (n1) || all_names (n2) || !strcmp (n1, n2);
+}
+
+
+static int 
+names_equal (const char *n1, const char *n2)
+{
+	return (n1 == NULL && n2 == NULL) || (n1 && n2 && !strcmp (n1, n2));
+}
+
+/**
+ * lock_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_inode_t *pinode, const char *basename, entrylk_type type)
+{
+	pl_entry_lock_t *lock = NULL;
+
+	if (list_empty (&pinode->dir_list))
+		return NULL;
+
+	list_for_each_entry (lock, &pinode->dir_list, inode_list) {
+		if (names_conflict (lock->basename, basename) &&
+		    types_conflict (lock->type, type))
+			return lock;
+	}
+
+	return NULL;
+}
+
+/**
+ * find_most_matching_lock - find the lock struct which most matches in order of:
+ *                           lock on the exact basename ||
+ *                           an all_names lock
+ *                      
+ *
+ * @inode: inode in which to look
+ * @basename: name to search for
+ */
+
+static pl_entry_lock_t * 
+__find_most_matching_lock (pl_inode_t *pinode, const char *basename)
+{
+	pl_entry_lock_t *lock;
+	pl_entry_lock_t *all = NULL;
+	pl_entry_lock_t *exact = NULL;
+
+	if (list_empty (&pinode->dir_list)) 
+		return NULL;
+
+	list_for_each_entry (lock, &pinode->dir_list, inode_list) {
+		if (all_names (lock->basename))
+			all = lock;
+		else if (names_equal (lock->basename, basename))
+			exact = lock;
+	}
+
+	return (exact ? exact : all);
+}
+
+
+/**
+ * insert_new_lock - insert a new dir lock into the inode with the given parameters
+ * @pinode: inode to insert into
+ * @basename: basename for the lock
+ * @type: type of the lock
+ */
+
+static pl_entry_lock_t *
+new_entrylk_lock (pl_inode_t *pinode, const char *basename, entrylk_type type,
+		  transport_t *trans)
+{
+	pl_entry_lock_t *newlock = NULL;
+
+	newlock = CALLOC (sizeof (pl_entry_lock_t), 1);
+	if (!newlock) {
+		goto out;
+	}
+
+	newlock->basename = basename ? strdup (basename) : NULL;
+	newlock->type     = type;
+	newlock->trans    = trans;
+
+	if (type == ENTRYLK_RDLCK)
+		newlock->read_count = 1;
+
+	INIT_LIST_HEAD (&newlock->inode_list);
+	INIT_LIST_HEAD (&newlock->blocked_locks);
+
+out:
+	return newlock;
+}
+
+/**
+ * lock_name - 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
+ *            
+ * the entire directory being locked is represented as: a single
+ * pl_entry_lock_t present in the entrylk_locks list with its
+ * basename = NULL
+ */
+
+int
+__lock_name (pl_inode_t *pinode, const char *basename, entrylk_type type,
+	     call_frame_t *frame, xlator_t *this, int nonblock)
+{
+	pl_entry_lock_t *lock    = NULL;
+	pl_entry_lock_t *conf    = NULL;
+
+	transport_t *trans = NULL;
+
+	int ret = -EINVAL;
+
+	trans = frame->root->trans;
+
+	conf = __lock_grantable (pinode, basename, type);
+	if (conf) {
+		ret = -EAGAIN;
+		if (nonblock)
+			goto out;
+
+		lock = new_entrylk_lock (pinode, basename, type, trans);
+
+		if (!lock) {
+			ret = -ENOMEM;
+			goto out;
+		}
+
+		gf_log (this->name, GF_LOG_DEBUG,
+			"blocking lock: {pinode=%p, basename=%s}",
+			pinode, basename);
+
+		lock->frame   = frame;
+		lock->this    = this;
+		lock->blocked = 1;
+
+		list_add (&lock->blocked_locks, &conf->blocked_locks);
+
+
+		goto out;
+	}
+		
+	switch (type) {
+	case ENTRYLK_RDLCK:
+		lock = __find_most_matching_lock (pinode, basename);
+
+		if (lock && names_equal (lock->basename, basename)) {
+			lock->read_count++;
+
+			FREE (lock->basename);
+			FREE (lock);
+
+			lock = NULL;
+		} else {
+			lock = new_entrylk_lock (pinode, basename, type, trans);
+
+			if (!lock) {
+				ret = -ENOMEM;
+				goto out;
+			}
+
+			list_add (&lock->inode_list, &pinode->dir_list);
+		}
+		break;
+
+	case ENTRYLK_WRLCK:
+		lock = new_entrylk_lock (pinode, basename, type, trans);
+			
+		if (!lock) {
+			ret = -ENOMEM;
+			goto out;
+		}
+
+		list_add (&lock->inode_list, &pinode->dir_list);
+		break;
+	}
+
+	ret = 0;
+out:
+	return ret;
+}
+
+
+/**
+ * unlock_name - 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_inode_t *pinode, const char *basename, entrylk_type type)
+{
+	pl_entry_lock_t *lock = NULL;
+	pl_entry_lock_t *ret_lock = NULL;
+
+	lock = __find_most_matching_lock (pinode, basename);
+	
+	if (!lock) {
+		gf_log ("locks", GF_LOG_DEBUG,
+			"unlock on %s (type=%s) attempted but no matching lock found",
+			basename, type == ENTRYLK_RDLCK ? "ENTRYLK_RDLCK" : 
+			"ENTRYLK_WRLCK");
+		goto out;
+	}
+	
+	if (names_equal (lock->basename, basename)
+	    && lock->type == type) {
+		if (type == ENTRYLK_RDLCK) {
+			lock->read_count--;
+		}
+		if (type == ENTRYLK_WRLCK || lock->read_count == 0) {
+			list_del (&lock->inode_list);
+			ret_lock = lock;
+		}
+	} else {
+		gf_log ("locks", GF_LOG_ERROR,
+			"unlock for a non-existing lock!");
+		goto out;
+	}
+
+out:
+	return ret_lock;
+}
+
+
+void
+__grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode,
+			     pl_entry_lock_t *lock,
+			     struct list_head *granted)
+{
+	int              bl_ret = 0;
+	pl_entry_lock_t *bl   = NULL;
+	pl_entry_lock_t *tmp  = NULL;
+
+	list_for_each_entry_safe (bl, tmp, &lock->blocked_locks, 
+				  blocked_locks) {
+		list_del_init (&bl->blocked_locks);
+
+		/* TODO: error checking */
+
+		gf_log ("locks", GF_LOG_DEBUG,
+			"trying to unblock: {pinode=%p, basename=%s}",
+			pl_inode, bl->basename);
+					
+		bl_ret = __lock_name (pl_inode, bl->basename, bl->type,
+				      bl->frame, bl->this, 0);
+
+		if (bl_ret == 0) {
+			list_add (&bl->blocked_locks, granted);
+		} else {
+			if (bl->basename)
+				FREE (bl->basename);
+			FREE (bl);
+		}
+	}	
+	return;
+}
+
+
+void
+grant_blocked_entry_locks (xlator_t *this, pl_inode_t *pl_inode,
+			   pl_entry_lock_t *unlocked)
+{
+	struct list_head  granted_list;
+	pl_entry_lock_t  *tmp = NULL;
+	pl_entry_lock_t  *lock = NULL;
+
+	INIT_LIST_HEAD (&granted_list);
+
+	pthread_mutex_lock (&pl_inode->mutex);
+	{
+		__grant_blocked_entry_locks (this, pl_inode, unlocked,
+					     &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);
+
+		STACK_UNWIND (lock->frame, 0, 0);
+
+		FREE (lock->basename);
+		FREE (lock);
+	}
+
+	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,
+				   transport_t *trans)
+{
+	pl_entry_lock_t  *lock;
+	pl_entry_lock_t  *tmp;
+	struct list_head  granted;
+
+	INIT_LIST_HEAD (&granted);
+
+	pthread_mutex_lock (&pinode->mutex);
+	{
+		if (list_empty (&pinode->dir_list)) {
+			goto unlock;
+		}
+
+		list_for_each_entry_safe (lock, tmp, &pinode->dir_list,
+					  inode_list) {
+			if (lock->trans != trans)
+				continue;
+
+			list_del_init (&lock->inode_list);
+			__grant_blocked_entry_locks (this, pinode, lock,
+						     &granted);
+
+			FREE (lock->basename);
+			FREE (lock);
+		}
+	}
+unlock:
+	pthread_mutex_unlock (&pinode->mutex);
+
+	list_for_each_entry_safe (lock, tmp, &granted, blocked_locks) {
+		list_del_init (&lock->blocked_locks);
+
+		STACK_UNWIND (lock->frame, 0, 0);
+
+		FREE (lock->basename);
+		FREE (lock);
+	}
+
+	return 0;
+}
+
+
+/**
+ * pl_entrylk:
+ * 
+ * Locking on names (directory entries)
+ */
+
+int
+pl_entrylk_common (call_frame_t *frame, xlator_t *this,
+		   inode_t *inode, const char *basename, 
+		   entrylk_cmd cmd, entrylk_type type)
+{
+	int32_t op_ret   = -1;
+	int32_t op_errno = 0;
+
+	transport_t * transport = NULL;
+	pid_t pid = -1;
+
+	pl_inode_t *       pinode = NULL; 
+	int                ret    = -1;
+	pl_entry_lock_t   *unlocked = NULL;
+	char               unwind = 1;
+
+	pinode = pl_inode_get (this, inode);
+	if (!pinode) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"out of memory :(");
+		op_errno = ENOMEM;
+		goto out;
+	}
+
+	pid       = frame->root->pid;
+	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_DEBUG,
+			"releasing locks for transport %p", transport);
+
+		release_entry_locks_for_transport (this, pinode, transport);
+		op_ret = 0;
+
+		goto out;
+	}
+
+	switch (cmd) {
+	case ENTRYLK_LOCK:
+		pthread_mutex_lock (&pinode->mutex);
+		{
+			ret = __lock_name (pinode, basename, type,
+					   frame, this, 0);
+		}
+		pthread_mutex_unlock (&pinode->mutex);
+
+		if (ret < 0) {
+			if (ret == -EAGAIN)
+				unwind = 0;
+			op_errno = -ret;
+			goto out;
+		}
+
+		break;
+
+	case ENTRYLK_LOCK_NB:
+		pthread_mutex_lock (&pinode->mutex);
+		{
+			ret = __lock_name (pinode, basename, type,
+					   frame, this, 1);
+		}
+		pthread_mutex_unlock (&pinode->mutex);
+
+		if (ret < 0) {
+			op_errno = -ret;
+			goto out;
+		}
+
+		break;
+
+	case ENTRYLK_UNLOCK:
+		pthread_mutex_lock (&pinode->mutex);
+		{
+			unlocked = __unlock_name (pinode, basename, type);
+		}
+		pthread_mutex_unlock (&pinode->mutex);
+
+		if (unlocked)
+			grant_blocked_entry_locks (this, pinode, unlocked);
+
+		break;
+
+	default:
+		gf_log (this->name, GF_LOG_ERROR,
+			"unexpected case!");
+		goto out;
+	}
+
+	op_ret = 0;
+out:
+	if (unwind) {
+		STACK_UNWIND (frame, op_ret, op_errno);
+	}
+	
+	return 0;
+}
+
+
+int
+pl_entrylk (call_frame_t *frame, xlator_t *this,
+	    loc_t *loc, const char *basename, 
+	    entrylk_cmd cmd, entrylk_type type)
+{
+	return pl_entrylk_common (frame, this, loc->inode, basename, cmd, type);
+}
+
+
+int
+pl_fentrylk (call_frame_t *frame, xlator_t *this,
+	     fd_t *fd, const char *basename, 
+	     entrylk_cmd cmd, entrylk_type type)
+{
+	return pl_entrylk_common (frame, this, fd->inode, basename, cmd, type);	
+}
diff --git a/xlators/features/locks/src/locks.h b/xlators/features/locks/src/locks.h
new file mode 100644
index 00000000000..8ed7bb63f1c
--- /dev/null
+++ b/xlators/features/locks/src/locks.h
@@ -0,0 +1,111 @@
+/*
+   Copyright (c) 2006, 2007, 2008 Z RESEARCH, Inc. <http://www.zresearch.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 __POSIX_LOCKS_H__
+#define __POSIX_LOCKS_H__
+
+#ifndef _CONFIG_H
+#define _CONFIG_H
+#include "config.h"
+#endif
+
+#include "compat-errno.h"
+#include "transport.h"
+#include "stack.h"
+#include "call-stub.h"
+
+struct __pl_fd;
+
+struct __posix_lock {
+	struct list_head   list;
+
+	short              fl_type;
+	off_t              fl_start;
+	off_t              fl_end;
+
+	short              blocked;    /* waiting to acquire */
+	struct flock       user_flock; /* the flock supplied by the user */
+	xlator_t          *this;       /* required for blocked locks */
+	fd_t              *fd;
+
+	call_frame_t      *frame;
+
+	/* These two together serve to uniquely identify each process
+	   across nodes */
+
+	transport_t       *transport;     /* to identify client node */
+	pid_t              client_pid;    /* pid of client process */
+};
+typedef struct __posix_lock posix_lock_t;
+
+struct __pl_rw_req_t {
+	struct list_head      list;
+	call_stub_t          *stub;
+	posix_lock_t          region;
+};
+typedef struct __pl_rw_req_t pl_rw_req_t;
+
+
+struct __entry_lock {
+	struct list_head  inode_list;    /* list_head back to pl_inode_t */
+	struct list_head  blocked_locks; /* locks blocked due to this lock */
+
+	call_frame_t     *frame;
+	xlator_t         *this;
+	int               blocked;
+	
+	const char       *basename;
+	entrylk_type      type;
+	unsigned int      read_count;    /* number of read locks */
+	transport_t      *trans;
+};
+typedef struct __entry_lock pl_entry_lock_t;
+
+
+/* The "simulated" inode. This contains a list of all the locks associated 
+   with this file */
+
+struct __pl_inode {
+	pthread_mutex_t  mutex;
+
+	struct list_head dir_list;       /* list of entry locks */
+	struct list_head ext_list;       /* list of fcntl locks */
+	struct list_head int_list;       /* list of internal locks */
+	struct list_head rw_list;        /* list of waiting r/w requests */
+	int              mandatory;      /* if mandatory locking is enabled */
+};
+typedef struct __pl_inode pl_inode_t;
+
+#define DOMAIN_HEAD(pl_inode, dom) (dom == GF_LOCK_POSIX	\
+				    ? &pl_inode->ext_list	\
+				    : &pl_inode->int_list)
+
+
+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 */
+} posix_locks_private_t;
+
+
+#endif /* __POSIX_LOCKS_H__ */
diff --git a/xlators/features/locks/src/posix.c b/xlators/features/locks/src/posix.c
new file mode 100644
index 00000000000..e2b336607c4
--- /dev/null
+++ b/xlators/features/locks/src/posix.c
@@ -0,0 +1,834 @@
+/*
+  Copyright (c) 2006, 2007, 2008 Z RESEARCH, Inc. <http://www.zresearch.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 <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"
+
+#ifndef LLONG_MAX
+#define LLONG_MAX LONG_LONG_MAX /* compat with old gcc */
+#endif /* LLONG_MAX */
+
+/* Forward declarations */
+
+
+void do_blocked_rw (pl_inode_t *);
+static int __rw_allowable (pl_inode_t *, posix_lock_t *, glusterfs_fop_t);
+
+struct _truncate_ops {
+	loc_t  loc;
+	fd_t  *fd;
+	off_t  offset;
+	enum {TRUNCATE, FTRUNCATE} op;
+};
+
+
+int
+pl_truncate_cbk (call_frame_t *frame, void *cookie, xlator_t *this,
+		 int32_t op_ret, int32_t op_errno, struct stat *buf)
+{
+	struct _truncate_ops *local = NULL;
+
+	local = frame->local;
+
+	if (local->op == TRUNCATE)
+		loc_wipe (&local->loc);
+
+	STACK_UNWIND (frame, op_ret, op_errno, buf);
+	return 0;
+}
+
+
+static int
+truncate_allowed (pl_inode_t *pl_inode, 
+		  transport_t *transport, pid_t client_pid, 
+		  off_t offset)
+{
+	posix_lock_t *l = NULL;
+	posix_lock_t  region = {.list = {0, }, };
+	int           ret = 1;
+
+	region.fl_start   = offset;
+	region.fl_end     = LLONG_MAX;
+	region.transport  = transport;
+	region.client_pid = client_pid;
+
+	pthread_mutex_lock (&pl_inode->mutex);
+	{
+		list_for_each_entry (l, &pl_inode->ext_list, list) {
+			if (!l->blocked
+			    && locks_overlap (&region, l)
+			    && !same_owner (&region, l)) {
+				ret = 0;
+				break;
+			}
+		}
+	}
+	pthread_mutex_unlock (&pl_inode->mutex);
+
+	return ret;
+}
+
+
+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)
+{
+	posix_locks_private_t *priv = NULL;
+	struct _truncate_ops  *local = NULL;
+	inode_t               *inode = NULL;
+	pl_inode_t            *pl_inode = NULL;
+
+
+	priv = this->private;
+	local = frame->local;
+
+	if (op_ret != 0) {
+		gf_log (this->name, GF_LOG_ERROR, 
+			"got error (errno=%d, stderror=%s) from child", 
+			op_errno, strerror (op_errno));
+		goto unwind;
+	}
+
+	if (local->op == TRUNCATE)
+		inode = local->loc.inode;
+	else
+		inode = local->fd->inode;
+
+	pl_inode = pl_inode_get (this, inode);
+	if (!pl_inode) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"unable to get pl_inode from %p", inode);
+		op_errno = ENOMEM;
+		goto unwind;
+	}
+
+	if (priv->mandatory
+	    && pl_inode->mandatory
+	    && !truncate_allowed (pl_inode, frame->root->trans,
+				  frame->root->pid, local->offset)) {
+		op_errno = EAGAIN;
+		goto unwind;
+	}
+
+	switch (local->op) {
+	case TRUNCATE:
+		STACK_WIND (frame, pl_truncate_cbk, FIRST_CHILD (this),
+			    FIRST_CHILD (this)->fops->truncate,
+			    &local->loc, local->offset);
+		break;
+	case FTRUNCATE:
+		STACK_WIND (frame, pl_truncate_cbk, FIRST_CHILD (this),
+			    FIRST_CHILD (this)->fops->ftruncate,
+			    local->fd, local->offset);
+		break;
+	}
+
+	return 0;
+
+unwind:
+	if (local->op == TRUNCATE)
+		loc_wipe (&local->loc);
+
+	STACK_UNWIND (frame, -1, ENOMEM, buf);
+	return 0;
+}
+
+
+int
+pl_truncate (call_frame_t *frame, xlator_t *this,
+	     loc_t *loc, off_t offset)
+{
+	struct _truncate_ops *local = NULL;
+
+	local = CALLOC (1, sizeof (struct _truncate_ops));
+	if (!local) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"out of memory :(");
+		goto unwind;
+	}
+
+	local->op         = TRUNCATE;
+	local->offset     = offset;
+	loc_copy (&local->loc, loc);
+
+	frame->local = local;
+
+	STACK_WIND (frame, truncate_stat_cbk, FIRST_CHILD (this),
+		    FIRST_CHILD (this)->fops->stat, loc);
+
+	return 0;
+
+unwind:
+	STACK_UNWIND (frame, -1, ENOMEM, NULL);
+
+	return 0;
+}
+
+
+int
+pl_ftruncate (call_frame_t *frame, xlator_t *this,
+	      fd_t *fd, off_t offset)
+{
+	struct _truncate_ops *local = NULL;
+
+	local = CALLOC (1, sizeof (struct _truncate_ops));
+	if (!local) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"out of memory :(");
+		goto unwind;
+	}
+
+	local->op         = FTRUNCATE;
+	local->offset     = offset;
+	local->fd         = fd;
+
+	frame->local = local;
+
+	STACK_WIND (frame, truncate_stat_cbk, FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->fstat, fd);
+	return 0;
+
+unwind:
+	STACK_UNWIND (frame, -1, ENOMEM, NULL);
+
+	return 0;
+}
+
+
+static void
+__delete_locks_of_owner (pl_inode_t *pl_inode,
+			 transport_t *transport, pid_t pid)
+{
+	posix_lock_t *tmp = NULL;
+	posix_lock_t *l = NULL;
+
+	/* 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->client_pid == pid)) {
+			__delete_lock (pl_inode, l);
+			__destroy_lock (l);
+		}
+	}
+
+	list_for_each_entry_safe (l, tmp, &pl_inode->int_list, list) {
+		if ((l->transport == transport)
+		    && (l->client_pid == pid)) {
+			__delete_lock (pl_inode, l);
+			__destroy_lock (l);
+		}
+	}
+
+	return;
+}
+
+
+int
+pl_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
+pl_flush (call_frame_t *frame, xlator_t *this,
+	  fd_t *fd)
+{
+	posix_locks_private_t *priv = NULL;
+	pl_inode_t            *pl_inode = NULL;
+
+	priv = this->private;
+
+	pl_inode = pl_inode_get (this, fd->inode);
+	if (!pl_inode) {
+		gf_log (this->name, GF_LOG_ERROR, "returning EBADFD");
+		STACK_UNWIND (frame, -1, EBADFD);
+		return 0;
+	}
+
+	pthread_mutex_lock (&pl_inode->mutex);
+	{
+		__delete_locks_of_owner (pl_inode, frame->root->trans,
+					 frame->root->pid);
+	}
+	pthread_mutex_unlock (&pl_inode->mutex);
+
+	grant_blocked_locks (this, pl_inode, GF_LOCK_POSIX);
+	grant_blocked_locks (this, pl_inode, GF_LOCK_INTERNAL);
+
+	do_blocked_rw (pl_inode);
+
+	STACK_WIND (frame, pl_flush_cbk, FIRST_CHILD(this),
+		    FIRST_CHILD(this)->fops->flush, fd);
+	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)
+{
+	STACK_UNWIND (frame, op_ret, op_errno, fd);
+
+	return 0;
+}
+
+
+int
+pl_open (call_frame_t *frame, xlator_t *this,
+	 loc_t *loc, int32_t flags, fd_t *fd)
+{
+	/* 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);
+
+	return 0;
+}
+
+
+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)
+{
+	STACK_UNWIND (frame, op_ret, op_errno, fd, inode, buf);
+
+	return 0;
+}
+
+
+int
+pl_create (call_frame_t *frame, xlator_t *this,
+	   loc_t *loc, int32_t flags, mode_t mode, fd_t *fd)
+{
+	STACK_WIND (frame, pl_create_cbk,
+		    FIRST_CHILD (this), FIRST_CHILD (this)->fops->create, 
+		    loc, flags, mode, fd);
+	return 0;
+}
+
+
+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)
+{
+	STACK_UNWIND (frame, op_ret, op_errno, vector, count, stbuf);
+
+	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 *stbuf)
+{
+	STACK_UNWIND (frame, op_ret, op_errno, stbuf);
+
+	return 0;
+}
+
+
+void
+do_blocked_rw (pl_inode_t *pl_inode)
+{
+	struct list_head  wind_list;
+	pl_rw_req_t      *rw = NULL;
+	pl_rw_req_t      *tmp = NULL;
+
+	INIT_LIST_HEAD (&wind_list);
+
+	pthread_mutex_lock (&pl_inode->mutex);
+	{
+		list_for_each_entry_safe (rw, tmp, &pl_inode->rw_list, list) {
+			if (__rw_allowable (pl_inode, &rw->region,
+					    rw->stub->fop)) {
+				list_del_init (&rw->list);
+				list_add_tail (&rw->list, &wind_list);
+			}
+		}
+	}
+	pthread_mutex_unlock (&pl_inode->mutex);
+
+	list_for_each_entry_safe (rw, tmp, &wind_list, list) {
+		list_del_init (&rw->list);
+		call_resume (rw->stub);
+		free (rw);
+	}
+
+	return;
+}
+
+
+static int
+__rw_allowable (pl_inode_t *pl_inode, posix_lock_t *region,
+		glusterfs_fop_t op)
+{
+	posix_lock_t *l = NULL;
+	int           ret = 1;
+
+	list_for_each_entry (l, &pl_inode->ext_list, list) {
+		if (locks_overlap (l, region) && !same_owner (l, region)) {
+			if ((op == GF_FOP_READ) && (l->fl_type != F_WRLCK))
+				continue;
+			ret = 0;
+			break;
+		}
+	}
+
+	return ret;
+}
+
+
+int
+pl_readv_cont (call_frame_t *frame, xlator_t *this,
+	       fd_t *fd, size_t size, off_t offset)
+{
+	STACK_WIND (frame, pl_readv_cbk,
+		    FIRST_CHILD (this), FIRST_CHILD (this)->fops->readv,
+		    fd, size, offset);
+
+	return 0;
+}
+
+
+int
+pl_readv (call_frame_t *frame, xlator_t *this,
+	  fd_t *fd, size_t size, off_t offset)
+{
+	posix_locks_private_t *priv = NULL;
+	pl_inode_t            *pl_inode = NULL;
+	pl_rw_req_t           *rw = NULL;
+	posix_lock_t           region = {.list = {0, }, };
+	int                    op_ret = 0;
+	int                    op_errno = 0;
+	char                   allowable = 0;
+
+
+	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 + size - 1;
+		region.transport  = frame->root->trans;
+		region.client_pid = frame->root->pid;
+    
+		pthread_mutex_lock (&pl_inode->mutex);
+		{
+			allowable = __rw_allowable (pl_inode, &region,
+						    GF_FOP_READ);
+			if (allowable)
+				goto unlock;
+
+			if (fd->flags & O_NONBLOCK) {
+				gf_log (this->name, GF_LOG_DEBUG,
+					"returning EWOULDBLOCK");
+				op_errno = EWOULDBLOCK;
+				op_ret = -1;
+				goto unlock;
+			}
+
+			rw = CALLOC (1, sizeof (*rw));
+			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);
+			if (!rw->stub) {
+				gf_log (this->name, GF_LOG_ERROR,
+					"out of memory :(");
+				op_errno = ENOMEM;
+				op_ret = -1;
+				free (rw);
+				goto unlock;
+			}
+
+			rw->region = region;
+
+			list_add_tail (&rw->list, &pl_inode->rw_list);
+		}
+	unlock:
+		pthread_mutex_unlock (&pl_inode->mutex);
+
+		goto unwind;
+	}
+
+
+	STACK_WIND (frame, pl_readv_cbk, 
+		    FIRST_CHILD (this), FIRST_CHILD (this)->fops->readv,
+		    fd, size, offset);
+	return 0;
+
+unwind:
+	if (op_ret == -1)
+		STACK_UNWIND (frame, -1, op_errno, NULL, 0, NULL);
+
+	return 0;
+}
+
+
+int
+pl_writev_cont (call_frame_t *frame, xlator_t *this, fd_t *fd,
+		struct iovec *vector, int count, off_t offset)
+{
+	STACK_WIND (frame, pl_writev_cbk,
+		    FIRST_CHILD (this), FIRST_CHILD (this)->fops->writev,
+		    fd, vector, count, offset);
+
+	return 0;
+}
+
+
+int
+pl_writev (call_frame_t *frame, xlator_t *this, fd_t *fd,
+	   struct iovec *vector, int32_t count, off_t offset)
+{
+	posix_locks_private_t *priv = NULL;
+	pl_inode_t            *pl_inode = NULL;
+	pl_rw_req_t           *rw = NULL;
+	posix_lock_t           region = {.list = {0, }, };
+	int                    op_ret = 0;
+	int                    op_errno = 0;
+	char                   allowable = 0;
+
+
+	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_pid = frame->root->pid;
+    
+		pthread_mutex_lock (&pl_inode->mutex);
+		{
+			allowable = __rw_allowable (pl_inode, &region,
+						    GF_FOP_WRITE);
+			if (allowable)
+				goto unlock;
+
+			if (fd->flags & O_NONBLOCK) {
+				gf_log (this->name, GF_LOG_DEBUG,
+					"returning EWOULDBLOCK");
+				op_errno = EWOULDBLOCK;
+				op_ret = -1;
+				goto unlock;
+			}
+
+			rw = CALLOC (1, sizeof (*rw));
+			if (!rw) {
+				gf_log (this->name, GF_LOG_ERROR,
+					"out of memory :(");
+				op_errno = ENOMEM;
+				op_ret = -1;
+				goto unlock;
+			}
+
+			rw->stub = fop_writev_stub (frame, pl_writev_cont,
+						    fd, vector, count, offset);
+			if (!rw->stub) {
+				gf_log (this->name, GF_LOG_ERROR,
+					"out of memory :(");
+				op_errno = ENOMEM;
+				op_ret = -1;
+				free (rw);
+				goto unlock;
+			}
+
+			rw->region = region;
+
+			list_add_tail (&rw->list, &pl_inode->rw_list);
+		}
+	unlock:
+		pthread_mutex_unlock (&pl_inode->mutex);
+
+		goto unwind;
+	}
+
+
+	STACK_WIND (frame, pl_writev_cbk, 
+		    FIRST_CHILD (this), FIRST_CHILD (this)->fops->writev,
+		    fd, vector, count, offset);
+	return 0;
+
+unwind:
+	if (op_ret == -1)
+		STACK_UNWIND (frame, -1, op_errno, NULL, 0, NULL);
+
+	return 0;
+}
+
+
+int
+pl_lk (call_frame_t *frame, xlator_t *this,
+       fd_t *fd, int32_t cmd, struct flock *flock)
+{
+	transport_t           *transport = NULL;
+	pid_t                  client_pid = 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;
+	priv       = this->private;
+
+	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);
+	if (!reqlock) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"out of memory :(");
+		op_ret = -1;
+		op_errno = ENOMEM;
+		goto unwind;
+	}
+
+	switch (cmd) {
+
+#if F_GETLK != F_GETLK64
+	case F_GETLK64:
+#endif
+	case F_GETLK:
+		conf = pl_getlk (pl_inode, reqlock, GF_LOCK_POSIX);
+		posix_lock_to_flock (conf, flock);
+		__destroy_lock (reqlock);
+
+		break;
+
+#if F_SETLKW != F_SETLKW64
+	case F_SETLKW64:
+#endif
+	case F_SETLKW:
+		can_block = 1;
+		reqlock->frame = frame;
+		reqlock->this  = this;
+		reqlock->fd    = fd;
+
+		/* fall through */
+
+#if F_SETLK != F_SETLK64
+	case F_SETLK64:
+#endif
+	case F_SETLK:
+		memcpy (&reqlock->user_flock, flock, sizeof (struct flock));
+		ret = pl_setlk (this, pl_inode, reqlock,
+				can_block, GF_LOCK_POSIX);
+
+		if (ret == -1) {
+			if (can_block)
+				goto out;
+
+			gf_log (this->name, GF_LOG_DEBUG, "returning EAGAIN");
+			op_ret = -1;
+			op_errno = EAGAIN;
+			__destroy_lock (reqlock);
+		}
+	}
+
+unwind:
+	STACK_UNWIND (frame, op_ret, op_errno, flock);
+out:
+	return 0;
+}
+
+
+/* TODO: this function just logs, no action required?? */
+int
+pl_forget (xlator_t *this,
+	   inode_t *inode)
+{
+	pl_inode_t   *pl_inode = NULL;
+
+	pl_inode = pl_inode_get (this, inode);
+
+	if (!list_empty (&pl_inode->rw_list)) {
+		gf_log (this->name, GF_LOG_CRITICAL,
+			"pending R/W requests found!");
+	}
+
+	if (!list_empty (&pl_inode->ext_list)) {
+		gf_log (this->name, GF_LOG_CRITICAL,
+			"Pending fcntl locks found!");
+	}
+
+	if (!list_empty (&pl_inode->int_list)) {
+		gf_log (this->name, GF_LOG_CRITICAL,
+			"Pending internal locks found!");
+	}
+
+	if (!list_empty (&pl_inode->dir_list)) {
+		gf_log (this->name, GF_LOG_CRITICAL,
+			"Pending entry locks found!");
+	}
+
+	FREE (pl_inode);
+
+	return 0;
+}
+
+
+int
+init (xlator_t *this)
+{
+	posix_locks_private_t *priv = NULL;
+	xlator_list_t         *trav = NULL;
+	data_t                *mandatory = NULL;
+
+	if (!this->children || this->children->next) {
+		gf_log (this->name, GF_LOG_ERROR, 
+			"FATAL: posix-locks should have exactly one child");
+		return -1;
+	}
+
+	if (!this->parents) {
+		gf_log (this->name, GF_LOG_WARNING,
+			"dangling volume. check volfile ");
+	}
+
+	trav = this->children;
+	while (trav->xlator->children)
+		trav = trav->xlator->children;
+
+	if (strncmp ("storage/", trav->xlator->type, 8)) {
+		gf_log (this->name, GF_LOG_ERROR,
+			"'posix-locks' not loaded over storage translator");
+		return -1;
+	}
+
+	priv = CALLOC (1, sizeof (*priv));
+
+	mandatory = dict_get (this->options, "mandatory-locks");
+	if (mandatory) {
+		if (gf_string2boolean (mandatory->data,
+				       &priv->mandatory) == -1) {
+			gf_log (this->name, GF_LOG_ERROR,
+				"'mandatory-locks' takes only boolean "
+				"options");
+			return -1;
+		}
+	}
+
+	this->private = priv;
+	return 0;
+}
+
+
+int
+fini (xlator_t *this)
+{
+	posix_locks_private_t *priv = NULL;
+
+	priv = this->private;
+	free (priv);
+
+	return 0;
+}
+
+
+int
+pl_inodelk (call_frame_t *frame, xlator_t *this, 
+	    loc_t *loc, int32_t cmd, struct flock *flock);
+
+int
+pl_finodelk (call_frame_t *frame, xlator_t *this, 
+	     fd_t *fd, int32_t cmd, struct flock *flock);
+
+int
+pl_entrylk (call_frame_t *frame, xlator_t *this, 
+	    loc_t *loc, const char *basename, 
+	    entrylk_cmd cmd, entrylk_type type);
+
+int
+pl_fentrylk (call_frame_t *frame, xlator_t *this, 
+	     fd_t *fd, const char *basename, 
+	     entrylk_cmd cmd, entrylk_type type);
+
+struct xlator_fops fops = {
+	.create      = pl_create,
+	.truncate    = pl_truncate,
+	.ftruncate   = pl_ftruncate,
+	.open        = pl_open,
+	.readv       = pl_readv,
+	.writev      = pl_writev,
+	.lk          = pl_lk,
+	.inodelk     = pl_inodelk,
+	.finodelk    = pl_finodelk,
+	.entrylk     = pl_entrylk,
+	.fentrylk    = pl_fentrylk,
+	.flush       = pl_flush,
+};
+
+
+struct xlator_mops mops = {
+};
+
+
+struct xlator_cbks cbks = {
+	.forget      = pl_forget,
+};
+
+
+struct volume_options options[] = {
+	{ .key  = { "mandatory-locks", "mandatory" }, 
+	  .type = GF_OPTION_TYPE_BOOL 
+	},
+	{ .key = {NULL} },
+};