[PATCH] OCFS2: The Second Oracle Cluster Filesystem

The OCFS2 file system module.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
Signed-off-by: Kurt Hackel <kurt.hackel@oracle.com>

55 files changed, 24504 insertions, 0 deletions

diff --git a/Documentation/filesystems/00-INDEX b/Documentation/filesystems/00-INDEX
index d9b0a069186..2580ada100a 100644
--- a/Documentation/filesystems/00-INDEX
+++ b/Documentation/filesystems/00-INDEX
@@ -36,6 +36,8 @@ ntfs.txt
 	- info and mount options for the NTFS filesystem (Windows NT).
 proc.txt
 	- info on Linux's /proc filesystem.
+ocfs2.txt
+	- info and mount options for the OCFS2 clustered filesystem.
 romfs.txt
 	- Description of the ROMFS filesystem.
 smbfs.txt
diff --git a/Documentation/filesystems/ocfs2.txt b/Documentation/filesystems/ocfs2.txt
new file mode 100644
index 00000000000..f2595caf052
--- /dev/null
+++ b/Documentation/filesystems/ocfs2.txt
@@ -0,0 +1,55 @@
+OCFS2 filesystem
+==================
+OCFS2 is a general purpose extent based shared disk cluster file
+system with many similarities to ext3. It supports 64 bit inode
+numbers, and has automatically extending metadata groups which may
+also make it attractive for non-clustered use.
+
+You'll want to install the ocfs2-tools package in order to at least
+get "mount.ocfs2" and "ocfs2_hb_ctl".
+
+Project web page:    http://oss.oracle.com/projects/ocfs2
+Tools web page:      http://oss.oracle.com/projects/ocfs2-tools
+OCFS2 mailing lists: http://oss.oracle.com/projects/ocfs2/mailman/
+
+All code copyright 2005 Oracle except when otherwise noted.
+
+CREDITS:
+Lots of code taken from ext3 and other projects.
+
+Authors in alphabetical order:
+Joel Becker   <joel.becker@oracle.com>
+Zach Brown    <zach.brown@oracle.com>
+Mark Fasheh   <mark.fasheh@oracle.com>
+Kurt Hackel   <kurt.hackel@oracle.com>
+Sunil Mushran <sunil.mushran@oracle.com>
+Manish Singh  <manish.singh@oracle.com>
+
+Caveats
+=======
+Features which OCFS2 does not support yet:
+	- sparse files
+	- extended attributes
+	- shared writeable mmap
+	- loopback is supported, but data written will not
+	  be cluster coherent.
+	- quotas
+	- cluster aware flock
+	- Directory change notification (F_NOTIFY)
+	- Distributed Caching (F_SETLEASE/F_GETLEASE/break_lease)
+	- POSIX ACLs
+	- readpages / writepages (not user visible)
+
+Mount options
+=============
+
+OCFS2 supports the following mount options:
+(*) == default
+
+barrier=1		This enables/disables barriers. barrier=0 disables it,
+			barrier=1 enables it.
+errors=remount-ro(*)	Remount the filesystem read-only on an error.
+errors=panic		Panic and halt the machine if an error occurs.
+intr		(*)	Allow signals to interrupt cluster operations.
+nointr			Do not allow signals to interrupt cluster
+			operations.
diff --git a/MAINTAINERS b/MAINTAINERS
index 86ee06f4379..15888302025 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -1905,6 +1905,15 @@ M:	ajoshi@shell.unixbox.com
 L:	linux-nvidia@lists.surfsouth.com
 S:	Maintained
 
+ORACLE CLUSTER FILESYSTEM 2 (OCFS2)
+P:	Mark Fasheh
+M:	mark.fasheh@oracle.com
+P:	Kurt Hackel
+M:	kurt.hackel@oracle.com
+L:	ocfs2-devel@oss.oracle.com
+W:	http://oss.oracle.com/projects/ocfs2/
+S:	Supported	
+
 OLYMPIC NETWORK DRIVER
 P:	Peter De Shrijver
 M:	p2@ace.ulyssis.student.kuleuven.ac.be
diff --git a/fs/ocfs2/Makefile b/fs/ocfs2/Makefile
new file mode 100644
index 00000000000..7d3be845a61
--- /dev/null
+++ b/fs/ocfs2/Makefile
@@ -0,0 +1,33 @@
+EXTRA_CFLAGS += -Ifs/ocfs2
+
+EXTRA_CFLAGS += -DCATCH_BH_JBD_RACES
+
+obj-$(CONFIG_OCFS2_FS) += ocfs2.o
+
+ocfs2-objs := \
+	alloc.o 		\
+	aops.o 			\
+	buffer_head_io.o	\
+	dcache.o 		\
+	dir.o 			\
+	dlmglue.o 		\
+	export.o 		\
+	extent_map.o 		\
+	file.o 			\
+	heartbeat.o 		\
+	inode.o 		\
+	journal.o 		\
+	localalloc.o 		\
+	mmap.o 			\
+	namei.o 		\
+	slot_map.o 		\
+	suballoc.o 		\
+	super.o 		\
+	symlink.o 		\
+	sysfile.o 		\
+	uptodate.o		\
+	ver.o 			\
+	vote.o
+
+obj-$(CONFIG_OCFS2_FS) += cluster/
+obj-$(CONFIG_OCFS2_FS) += dlm/
diff --git a/fs/ocfs2/alloc.c b/fs/ocfs2/alloc.c
new file mode 100644
index 00000000000..465f797451e
--- /dev/null
+++ b/fs/ocfs2/alloc.c
@@ -0,0 +1,2040 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * alloc.c
+ *
+ * Extent allocs and frees
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+
+#define MLOG_MASK_PREFIX ML_DISK_ALLOC
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "dlmglue.h"
+#include "extent_map.h"
+#include "inode.h"
+#include "journal.h"
+#include "localalloc.h"
+#include "suballoc.h"
+#include "sysfile.h"
+#include "file.h"
+#include "super.h"
+#include "uptodate.h"
+
+#include "buffer_head_io.h"
+
+static int ocfs2_extent_contig(struct inode *inode,
+			       struct ocfs2_extent_rec *ext,
+			       u64 blkno);
+
+static int ocfs2_create_new_meta_bhs(struct ocfs2_super *osb,
+				     struct ocfs2_journal_handle *handle,
+				     struct inode *inode,
+				     int wanted,
+				     struct ocfs2_alloc_context *meta_ac,
+				     struct buffer_head *bhs[]);
+
+static int ocfs2_add_branch(struct ocfs2_super *osb,
+			    struct ocfs2_journal_handle *handle,
+			    struct inode *inode,
+			    struct buffer_head *fe_bh,
+			    struct buffer_head *eb_bh,
+			    struct buffer_head *last_eb_bh,
+			    struct ocfs2_alloc_context *meta_ac);
+
+static int ocfs2_shift_tree_depth(struct ocfs2_super *osb,
+				  struct ocfs2_journal_handle *handle,
+				  struct inode *inode,
+				  struct buffer_head *fe_bh,
+				  struct ocfs2_alloc_context *meta_ac,
+				  struct buffer_head **ret_new_eb_bh);
+
+static int ocfs2_do_insert_extent(struct ocfs2_super *osb,
+				  struct ocfs2_journal_handle *handle,
+				  struct inode *inode,
+				  struct buffer_head *fe_bh,
+				  u64 blkno,
+				  u32 new_clusters);
+
+static int ocfs2_find_branch_target(struct ocfs2_super *osb,
+				    struct inode *inode,
+				    struct buffer_head *fe_bh,
+				    struct buffer_head **target_bh);
+
+static int ocfs2_find_new_last_ext_blk(struct ocfs2_super *osb,
+				       struct inode *inode,
+				       struct ocfs2_dinode *fe,
+				       unsigned int new_i_clusters,
+				       struct buffer_head *old_last_eb,
+				       struct buffer_head **new_last_eb);
+
+static void ocfs2_free_truncate_context(struct ocfs2_truncate_context *tc);
+
+static int ocfs2_extent_contig(struct inode *inode,
+			       struct ocfs2_extent_rec *ext,
+			       u64 blkno)
+{
+	return blkno == (le64_to_cpu(ext->e_blkno) +
+			 ocfs2_clusters_to_blocks(inode->i_sb,
+						  le32_to_cpu(ext->e_clusters)));
+}
+
+/*
+ * How many free extents have we got before we need more meta data?
+ */
+int ocfs2_num_free_extents(struct ocfs2_super *osb,
+			   struct inode *inode,
+			   struct ocfs2_dinode *fe)
+{
+	int retval;
+	struct ocfs2_extent_list *el;
+	struct ocfs2_extent_block *eb;
+	struct buffer_head *eb_bh = NULL;
+
+	mlog_entry_void();
+
+	if (!OCFS2_IS_VALID_DINODE(fe)) {
+		OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe);
+		retval = -EIO;
+		goto bail;
+	}
+
+	if (fe->i_last_eb_blk) {
+		retval = ocfs2_read_block(osb, le64_to_cpu(fe->i_last_eb_blk),
+					  &eb_bh, OCFS2_BH_CACHED, inode);
+		if (retval < 0) {
+			mlog_errno(retval);
+			goto bail;
+		}
+		eb = (struct ocfs2_extent_block *) eb_bh->b_data;
+		el = &eb->h_list;
+	} else
+		el = &fe->id2.i_list;
+
+	BUG_ON(el->l_tree_depth != 0);
+
+	retval = le16_to_cpu(el->l_count) - le16_to_cpu(el->l_next_free_rec);
+bail:
+	if (eb_bh)
+		brelse(eb_bh);
+
+	mlog_exit(retval);
+	return retval;
+}
+
+/* expects array to already be allocated
+ *
+ * sets h_signature, h_blkno, h_suballoc_bit, h_suballoc_slot, and
+ * l_count for you
+ */
+static int ocfs2_create_new_meta_bhs(struct ocfs2_super *osb,
+				     struct ocfs2_journal_handle *handle,
+				     struct inode *inode,
+				     int wanted,
+				     struct ocfs2_alloc_context *meta_ac,
+				     struct buffer_head *bhs[])
+{
+	int count, status, i;
+	u16 suballoc_bit_start;
+	u32 num_got;
+	u64 first_blkno;
+	struct ocfs2_extent_block *eb;
+
+	mlog_entry_void();
+
+	count = 0;
+	while (count < wanted) {
+		status = ocfs2_claim_metadata(osb,
+					      handle,
+					      meta_ac,
+					      wanted - count,
+					      &suballoc_bit_start,
+					      &num_got,
+					      &first_blkno);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		for(i = count;  i < (num_got + count); i++) {
+			bhs[i] = sb_getblk(osb->sb, first_blkno);
+			if (bhs[i] == NULL) {
+				status = -EIO;
+				mlog_errno(status);
+				goto bail;
+			}
+			ocfs2_set_new_buffer_uptodate(inode, bhs[i]);
+
+			status = ocfs2_journal_access(handle, inode, bhs[i],
+						      OCFS2_JOURNAL_ACCESS_CREATE);
+			if (status < 0) {
+				mlog_errno(status);
+				goto bail;
+			}
+
+			memset(bhs[i]->b_data, 0, osb->sb->s_blocksize);
+			eb = (struct ocfs2_extent_block *) bhs[i]->b_data;
+			/* Ok, setup the minimal stuff here. */
+			strcpy(eb->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE);
+			eb->h_blkno = cpu_to_le64(first_blkno);
+			eb->h_fs_generation = cpu_to_le32(osb->fs_generation);
+
+#ifndef OCFS2_USE_ALL_METADATA_SUBALLOCATORS
+			/* we always use slot zero's suballocator */
+			eb->h_suballoc_slot = 0;
+#else
+			eb->h_suballoc_slot = cpu_to_le16(osb->slot_num);
+#endif
+			eb->h_suballoc_bit = cpu_to_le16(suballoc_bit_start);
+			eb->h_list.l_count =
+				cpu_to_le16(ocfs2_extent_recs_per_eb(osb->sb));
+
+			suballoc_bit_start++;
+			first_blkno++;
+
+			/* We'll also be dirtied by the caller, so
+			 * this isn't absolutely necessary. */
+			status = ocfs2_journal_dirty(handle, bhs[i]);
+			if (status < 0) {
+				mlog_errno(status);
+				goto bail;
+			}
+		}
+
+		count += num_got;
+	}
+
+	status = 0;
+bail:
+	if (status < 0) {
+		for(i = 0; i < wanted; i++) {
+			if (bhs[i])
+				brelse(bhs[i]);
+			bhs[i] = NULL;
+		}
+	}
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * Add an entire tree branch to our inode. eb_bh is the extent block
+ * to start at, if we don't want to start the branch at the dinode
+ * structure.
+ *
+ * last_eb_bh is required as we have to update it's next_leaf pointer
+ * for the new last extent block.
+ *
+ * the new branch will be 'empty' in the sense that every block will
+ * contain a single record with e_clusters == 0.
+ */
+static int ocfs2_add_branch(struct ocfs2_super *osb,
+			    struct ocfs2_journal_handle *handle,
+			    struct inode *inode,
+			    struct buffer_head *fe_bh,
+			    struct buffer_head *eb_bh,
+			    struct buffer_head *last_eb_bh,
+			    struct ocfs2_alloc_context *meta_ac)
+{
+	int status, new_blocks, i;
+	u64 next_blkno, new_last_eb_blk;
+	struct buffer_head *bh;
+	struct buffer_head **new_eb_bhs = NULL;
+	struct ocfs2_dinode *fe;
+	struct ocfs2_extent_block *eb;
+	struct ocfs2_extent_list  *eb_el;
+	struct ocfs2_extent_list  *el;
+
+	mlog_entry_void();
+
+	BUG_ON(!last_eb_bh);
+
+	fe = (struct ocfs2_dinode *) fe_bh->b_data;
+
+	if (eb_bh) {
+		eb = (struct ocfs2_extent_block *) eb_bh->b_data;
+		el = &eb->h_list;
+	} else
+		el = &fe->id2.i_list;
+
+	/* we never add a branch to a leaf. */
+	BUG_ON(!el->l_tree_depth);
+
+	new_blocks = le16_to_cpu(el->l_tree_depth);
+
+	/* allocate the number of new eb blocks we need */
+	new_eb_bhs = kcalloc(new_blocks, sizeof(struct buffer_head *),
+			     GFP_KERNEL);
+	if (!new_eb_bhs) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_create_new_meta_bhs(osb, handle, inode, new_blocks,
+					   meta_ac, new_eb_bhs);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* Note: new_eb_bhs[new_blocks - 1] is the guy which will be
+	 * linked with the rest of the tree.
+	 * conversly, new_eb_bhs[0] is the new bottommost leaf.
+	 *
+	 * when we leave the loop, new_last_eb_blk will point to the
+	 * newest leaf, and next_blkno will point to the topmost extent
+	 * block. */
+	next_blkno = new_last_eb_blk = 0;
+	for(i = 0; i < new_blocks; i++) {
+		bh = new_eb_bhs[i];
+		eb = (struct ocfs2_extent_block *) bh->b_data;
+		if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
+			OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
+			status = -EIO;
+			goto bail;
+		}
+		eb_el = &eb->h_list;
+
+		status = ocfs2_journal_access(handle, inode, bh,
+					      OCFS2_JOURNAL_ACCESS_CREATE);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		eb->h_next_leaf_blk = 0;
+		eb_el->l_tree_depth = cpu_to_le16(i);
+		eb_el->l_next_free_rec = cpu_to_le16(1);
+		eb_el->l_recs[0].e_cpos = fe->i_clusters;
+		eb_el->l_recs[0].e_blkno = cpu_to_le64(next_blkno);
+		eb_el->l_recs[0].e_clusters = cpu_to_le32(0);
+		if (!eb_el->l_tree_depth)
+			new_last_eb_blk = le64_to_cpu(eb->h_blkno);
+
+		status = ocfs2_journal_dirty(handle, bh);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		next_blkno = le64_to_cpu(eb->h_blkno);
+	}
+
+	/* This is a bit hairy. We want to update up to three blocks
+	 * here without leaving any of them in an inconsistent state
+	 * in case of error. We don't have to worry about
+	 * journal_dirty erroring as it won't unless we've aborted the
+	 * handle (in which case we would never be here) so reserving
+	 * the write with journal_access is all we need to do. */
+	status = ocfs2_journal_access(handle, inode, last_eb_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+	status = ocfs2_journal_access(handle, inode, fe_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+	if (eb_bh) {
+		status = ocfs2_journal_access(handle, inode, eb_bh,
+					      OCFS2_JOURNAL_ACCESS_WRITE);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+	}
+
+	/* Link the new branch into the rest of the tree (el will
+	 * either be on the fe, or the extent block passed in. */
+	i = le16_to_cpu(el->l_next_free_rec);
+	el->l_recs[i].e_blkno = cpu_to_le64(next_blkno);
+	el->l_recs[i].e_cpos = fe->i_clusters;
+	el->l_recs[i].e_clusters = 0;
+	le16_add_cpu(&el->l_next_free_rec, 1);
+
+	/* fe needs a new last extent block pointer, as does the
+	 * next_leaf on the previously last-extent-block. */
+	fe->i_last_eb_blk = cpu_to_le64(new_last_eb_blk);
+
+	eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
+	eb->h_next_leaf_blk = cpu_to_le64(new_last_eb_blk);
+
+	status = ocfs2_journal_dirty(handle, last_eb_bh);
+	if (status < 0)
+		mlog_errno(status);
+	status = ocfs2_journal_dirty(handle, fe_bh);
+	if (status < 0)
+		mlog_errno(status);
+	if (eb_bh) {
+		status = ocfs2_journal_dirty(handle, eb_bh);
+		if (status < 0)
+			mlog_errno(status);
+	}
+
+	status = 0;
+bail:
+	if (new_eb_bhs) {
+		for (i = 0; i < new_blocks; i++)
+			if (new_eb_bhs[i])
+				brelse(new_eb_bhs[i]);
+		kfree(new_eb_bhs);
+	}
+
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * adds another level to the allocation tree.
+ * returns back the new extent block so you can add a branch to it
+ * after this call.
+ */
+static int ocfs2_shift_tree_depth(struct ocfs2_super *osb,
+				  struct ocfs2_journal_handle *handle,
+				  struct inode *inode,
+				  struct buffer_head *fe_bh,
+				  struct ocfs2_alloc_context *meta_ac,
+				  struct buffer_head **ret_new_eb_bh)
+{
+	int status, i;
+	struct buffer_head *new_eb_bh = NULL;
+	struct ocfs2_dinode *fe;
+	struct ocfs2_extent_block *eb;
+	struct ocfs2_extent_list  *fe_el;
+	struct ocfs2_extent_list  *eb_el;
+
+	mlog_entry_void();
+
+	status = ocfs2_create_new_meta_bhs(osb, handle, inode, 1, meta_ac,
+					   &new_eb_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	eb = (struct ocfs2_extent_block *) new_eb_bh->b_data;
+	if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
+		OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
+		status = -EIO;
+		goto bail;
+	}
+
+	eb_el = &eb->h_list;
+	fe = (struct ocfs2_dinode *) fe_bh->b_data;
+	fe_el = &fe->id2.i_list;
+
+	status = ocfs2_journal_access(handle, inode, new_eb_bh,
+				      OCFS2_JOURNAL_ACCESS_CREATE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* copy the fe data into the new extent block */
+	eb_el->l_tree_depth = fe_el->l_tree_depth;
+	eb_el->l_next_free_rec = fe_el->l_next_free_rec;
+	for(i = 0; i < le16_to_cpu(fe_el->l_next_free_rec); i++) {
+		eb_el->l_recs[i].e_cpos = fe_el->l_recs[i].e_cpos;
+		eb_el->l_recs[i].e_clusters = fe_el->l_recs[i].e_clusters;
+		eb_el->l_recs[i].e_blkno = fe_el->l_recs[i].e_blkno;
+	}
+
+	status = ocfs2_journal_dirty(handle, new_eb_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_journal_access(handle, inode, fe_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* update fe now */
+	le16_add_cpu(&fe_el->l_tree_depth, 1);
+	fe_el->l_recs[0].e_cpos = 0;
+	fe_el->l_recs[0].e_blkno = eb->h_blkno;
+	fe_el->l_recs[0].e_clusters = fe->i_clusters;
+	for(i = 1; i < le16_to_cpu(fe_el->l_next_free_rec); i++) {
+		fe_el->l_recs[i].e_cpos = 0;
+		fe_el->l_recs[i].e_clusters = 0;
+		fe_el->l_recs[i].e_blkno = 0;
+	}
+	fe_el->l_next_free_rec = cpu_to_le16(1);
+
+	/* If this is our 1st tree depth shift, then last_eb_blk
+	 * becomes the allocated extent block */
+	if (fe_el->l_tree_depth == cpu_to_le16(1))
+		fe->i_last_eb_blk = eb->h_blkno;
+
+	status = ocfs2_journal_dirty(handle, fe_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	*ret_new_eb_bh = new_eb_bh;
+	new_eb_bh = NULL;
+	status = 0;
+bail:
+	if (new_eb_bh)
+		brelse(new_eb_bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * Expects the tree to already have room in the rightmost leaf for the
+ * extent.  Updates all the extent blocks (and the dinode) on the way
+ * down.
+ */
+static int ocfs2_do_insert_extent(struct ocfs2_super *osb,
+				  struct ocfs2_journal_handle *handle,
+				  struct inode *inode,
+				  struct buffer_head *fe_bh,
+				  u64 start_blk,
+				  u32 new_clusters)
+{
+	int status, i, num_bhs = 0;
+	u64 next_blkno;
+	u16 next_free;
+	struct buffer_head **eb_bhs = NULL;
+	struct ocfs2_dinode *fe;
+	struct ocfs2_extent_block *eb;
+	struct ocfs2_extent_list  *el;
+
+	mlog_entry_void();
+
+	status = ocfs2_journal_access(handle, inode, fe_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	fe = (struct ocfs2_dinode *) fe_bh->b_data;
+	el = &fe->id2.i_list;
+	if (el->l_tree_depth) {
+		/* This is another operation where we want to be
+		 * careful about our tree updates. An error here means
+		 * none of the previous changes we made should roll
+		 * forward. As a result, we have to record the buffers
+		 * for this part of the tree in an array and reserve a
+		 * journal write to them before making any changes. */
+		num_bhs = le16_to_cpu(fe->id2.i_list.l_tree_depth);
+		eb_bhs = kcalloc(num_bhs, sizeof(struct buffer_head *),
+				 GFP_KERNEL);
+		if (!eb_bhs) {
+			status = -ENOMEM;
+			mlog_errno(status);
+			goto bail;
+		}
+
+		i = 0;
+		while(el->l_tree_depth) {
+			next_free = le16_to_cpu(el->l_next_free_rec);
+			if (next_free == 0) {
+				ocfs2_error(inode->i_sb,
+					    "Dinode %"MLFu64" has a bad "
+					    "extent list",
+					    OCFS2_I(inode)->ip_blkno);
+				status = -EIO;
+				goto bail;
+			}
+			next_blkno = le64_to_cpu(el->l_recs[next_free - 1].e_blkno);
+
+			BUG_ON(i >= num_bhs);
+			status = ocfs2_read_block(osb, next_blkno, &eb_bhs[i],
+						  OCFS2_BH_CACHED, inode);
+			if (status < 0) {
+				mlog_errno(status);
+				goto bail;
+			}
+			eb = (struct ocfs2_extent_block *) eb_bhs[i]->b_data;
+			if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
+				OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb,
+								 eb);
+				status = -EIO;
+				goto bail;
+			}
+
+			status = ocfs2_journal_access(handle, inode, eb_bhs[i],
+						      OCFS2_JOURNAL_ACCESS_WRITE);
+			if (status < 0) {
+				mlog_errno(status);
+				goto bail;
+			}
+
+			el = &eb->h_list;
+			i++;
+			/* When we leave this loop, eb_bhs[num_bhs - 1] will
+			 * hold the bottom-most leaf extent block. */
+		}
+		BUG_ON(el->l_tree_depth);
+
+		el = &fe->id2.i_list;
+		/* If we have tree depth, then the fe update is
+		 * trivial, and we want to switch el out for the
+		 * bottom-most leaf in order to update it with the
+		 * actual extent data below. */
+		next_free = le16_to_cpu(el->l_next_free_rec);
+		if (next_free == 0) {
+			ocfs2_error(inode->i_sb,
+				    "Dinode %"MLFu64" has a bad "
+				    "extent list",
+				    OCFS2_I(inode)->ip_blkno);
+			status = -EIO;
+			goto bail;
+		}
+		le32_add_cpu(&el->l_recs[next_free - 1].e_clusters,
+			     new_clusters);
+		/* (num_bhs - 1) to avoid the leaf */
+		for(i = 0; i < (num_bhs - 1); i++) {
+			eb = (struct ocfs2_extent_block *) eb_bhs[i]->b_data;
+			el = &eb->h_list;
+
+			/* finally, make our actual change to the
+			 * intermediate extent blocks. */
+			next_free = le16_to_cpu(el->l_next_free_rec);
+			le32_add_cpu(&el->l_recs[next_free - 1].e_clusters,
+				     new_clusters);
+
+			status = ocfs2_journal_dirty(handle, eb_bhs[i]);
+			if (status < 0)
+				mlog_errno(status);
+		}
+		BUG_ON(i != (num_bhs - 1));
+		/* note that the leaf block wasn't touched in
+		 * the loop above */
+		eb = (struct ocfs2_extent_block *) eb_bhs[num_bhs - 1]->b_data;
+		el = &eb->h_list;
+		BUG_ON(el->l_tree_depth);
+	}
+
+	/* yay, we can finally add the actual extent now! */
+	i = le16_to_cpu(el->l_next_free_rec) - 1;
+	if (le16_to_cpu(el->l_next_free_rec) &&
+	    ocfs2_extent_contig(inode, &el->l_recs[i], start_blk)) {
+		le32_add_cpu(&el->l_recs[i].e_clusters, new_clusters);
+	} else if (le16_to_cpu(el->l_next_free_rec) &&
+		   (le32_to_cpu(el->l_recs[i].e_clusters) == 0)) {
+		/* having an empty extent at eof is legal. */
+		if (el->l_recs[i].e_cpos != fe->i_clusters) {
+			ocfs2_error(inode->i_sb,
+				    "Dinode %"MLFu64" trailing extent is bad: "
+				    "cpos (%u) != number of clusters (%u)",
+				    le32_to_cpu(el->l_recs[i].e_cpos),
+				    le32_to_cpu(fe->i_clusters));
+			status = -EIO;
+			goto bail;
+		}
+		el->l_recs[i].e_blkno = cpu_to_le64(start_blk);
+		el->l_recs[i].e_clusters = cpu_to_le32(new_clusters);
+	} else {
+		/* No contiguous record, or no empty record at eof, so
+		 * we add a new one. */
+
+		BUG_ON(le16_to_cpu(el->l_next_free_rec) >=
+		       le16_to_cpu(el->l_count));
+		i = le16_to_cpu(el->l_next_free_rec);
+
+		el->l_recs[i].e_blkno = cpu_to_le64(start_blk);
+		el->l_recs[i].e_clusters = cpu_to_le32(new_clusters);
+		el->l_recs[i].e_cpos = fe->i_clusters;
+		le16_add_cpu(&el->l_next_free_rec, 1);
+	}
+
+	/*
+	 * extent_map errors are not fatal, so they are ignored outside
+	 * of flushing the thing.
+	 */
+	status = ocfs2_extent_map_append(inode, &el->l_recs[i],
+					 new_clusters);
+	if (status) {
+		mlog_errno(status);
+		ocfs2_extent_map_drop(inode, le32_to_cpu(fe->i_clusters));
+	}
+
+	status = ocfs2_journal_dirty(handle, fe_bh);
+	if (status < 0)
+		mlog_errno(status);
+	if (fe->id2.i_list.l_tree_depth) {
+		status = ocfs2_journal_dirty(handle, eb_bhs[num_bhs - 1]);
+		if (status < 0)
+			mlog_errno(status);
+	}
+
+	status = 0;
+bail:
+	if (eb_bhs) {
+		for (i = 0; i < num_bhs; i++)
+			if (eb_bhs[i])
+				brelse(eb_bhs[i]);
+		kfree(eb_bhs);
+	}
+
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * Should only be called when there is no space left in any of the
+ * leaf nodes. What we want to do is find the lowest tree depth
+ * non-leaf extent block with room for new records. There are three
+ * valid results of this search:
+ *
+ * 1) a lowest extent block is found, then we pass it back in
+ *    *lowest_eb_bh and return '0'
+ *
+ * 2) the search fails to find anything, but the dinode has room. We
+ *    pass NULL back in *lowest_eb_bh, but still return '0'
+ *
+ * 3) the search fails to find anything AND the dinode is full, in
+ *    which case we return > 0
+ *
+ * return status < 0 indicates an error.
+ */
+static int ocfs2_find_branch_target(struct ocfs2_super *osb,
+				    struct inode *inode,
+				    struct buffer_head *fe_bh,
+				    struct buffer_head **target_bh)
+{
+	int status = 0, i;
+	u64 blkno;
+	struct ocfs2_dinode *fe;
+	struct ocfs2_extent_block *eb;
+	struct ocfs2_extent_list  *el;
+	struct buffer_head *bh = NULL;
+	struct buffer_head *lowest_bh = NULL;
+
+	mlog_entry_void();
+
+	*target_bh = NULL;
+
+	fe = (struct ocfs2_dinode *) fe_bh->b_data;
+	el = &fe->id2.i_list;
+
+	while(le16_to_cpu(el->l_tree_depth) > 1) {
+		if (le16_to_cpu(el->l_next_free_rec) == 0) {
+			ocfs2_error(inode->i_sb, "Dinode %"MLFu64" has empty "
+				    "extent list (next_free_rec == 0)",
+				    OCFS2_I(inode)->ip_blkno);
+			status = -EIO;
+			goto bail;
+		}
+		i = le16_to_cpu(el->l_next_free_rec) - 1;
+		blkno = le64_to_cpu(el->l_recs[i].e_blkno);
+		if (!blkno) {
+			ocfs2_error(inode->i_sb, "Dinode %"MLFu64" has extent "
+				    "list where extent # %d has no physical "
+				    "block start",
+				    OCFS2_I(inode)->ip_blkno, i);
+			status = -EIO;
+			goto bail;
+		}
+
+		if (bh) {
+			brelse(bh);
+			bh = NULL;
+		}
+
+		status = ocfs2_read_block(osb, blkno, &bh, OCFS2_BH_CACHED,
+					  inode);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		eb = (struct ocfs2_extent_block *) bh->b_data;
+		if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
+			OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
+			status = -EIO;
+			goto bail;
+		}
+		el = &eb->h_list;
+
+		if (le16_to_cpu(el->l_next_free_rec) <
+		    le16_to_cpu(el->l_count)) {
+			if (lowest_bh)
+				brelse(lowest_bh);
+			lowest_bh = bh;
+			get_bh(lowest_bh);
+		}
+	}
+
+	/* If we didn't find one and the fe doesn't have any room,
+	 * then return '1' */
+	if (!lowest_bh
+	    && (fe->id2.i_list.l_next_free_rec == fe->id2.i_list.l_count))
+		status = 1;
+
+	*target_bh = lowest_bh;
+bail:
+	if (bh)
+		brelse(bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+/* the caller needs to update fe->i_clusters */
+int ocfs2_insert_extent(struct ocfs2_super *osb,
+			struct ocfs2_journal_handle *handle,
+			struct inode *inode,
+			struct buffer_head *fe_bh,
+			u64 start_blk,
+			u32 new_clusters,
+			struct ocfs2_alloc_context *meta_ac)
+{
+	int status, i, shift;
+	struct buffer_head *last_eb_bh = NULL;
+	struct buffer_head *bh = NULL;
+	struct ocfs2_dinode *fe;
+	struct ocfs2_extent_block *eb;
+	struct ocfs2_extent_list  *el;
+
+	mlog_entry_void();
+
+	mlog(0, "add %u clusters starting at block %"MLFu64" to "
+		"inode %"MLFu64"\n",
+	     new_clusters, start_blk, OCFS2_I(inode)->ip_blkno);
+
+	fe = (struct ocfs2_dinode *) fe_bh->b_data;
+	el = &fe->id2.i_list;
+
+	if (el->l_tree_depth) {
+		/* jump to end of tree */
+		status = ocfs2_read_block(osb, le64_to_cpu(fe->i_last_eb_blk),
+					  &last_eb_bh, OCFS2_BH_CACHED, inode);
+		if (status < 0) {
+			mlog_exit(status);
+			goto bail;
+		}
+		eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
+		el = &eb->h_list;
+	}
+
+	/* Can we allocate without adding/shifting tree bits? */
+	i = le16_to_cpu(el->l_next_free_rec) - 1;
+	if (le16_to_cpu(el->l_next_free_rec) == 0
+	    || (le16_to_cpu(el->l_next_free_rec) < le16_to_cpu(el->l_count))
+	    || le32_to_cpu(el->l_recs[i].e_clusters) == 0
+	    || ocfs2_extent_contig(inode, &el->l_recs[i], start_blk))
+		goto out_add;
+
+	mlog(0, "ocfs2_allocate_extent: couldn't do a simple add, traversing "
+	     "tree now.\n");
+
+	shift = ocfs2_find_branch_target(osb, inode, fe_bh, &bh);
+	if (shift < 0) {
+		status = shift;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* We traveled all the way to the bottom of the allocation tree
+	 * and didn't find room for any more extents - we need to add
+	 * another tree level */
+	if (shift) {
+		/* if we hit a leaf, we'd better be empty :) */
+		BUG_ON(le16_to_cpu(el->l_next_free_rec) !=
+		       le16_to_cpu(el->l_count));
+		BUG_ON(bh);
+		mlog(0, "ocfs2_allocate_extent: need to shift tree depth "
+		     "(current = %u)\n",
+		     le16_to_cpu(fe->id2.i_list.l_tree_depth));
+
+		/* ocfs2_shift_tree_depth will return us a buffer with
+		 * the new extent block (so we can pass that to
+		 * ocfs2_add_branch). */
+		status = ocfs2_shift_tree_depth(osb, handle, inode, fe_bh,
+						meta_ac, &bh);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+		/* Special case: we have room now if we shifted from
+		 * tree_depth 0 */
+		if (fe->id2.i_list.l_tree_depth == cpu_to_le16(1))
+			goto out_add;
+	}
+
+	/* call ocfs2_add_branch to add the final part of the tree with
+	 * the new data. */
+	mlog(0, "ocfs2_allocate_extent: add branch. bh = %p\n", bh);
+	status = ocfs2_add_branch(osb, handle, inode, fe_bh, bh, last_eb_bh,
+				  meta_ac);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+out_add:
+	/* Finally, we can add clusters. */
+	status = ocfs2_do_insert_extent(osb, handle, inode, fe_bh,
+					start_blk, new_clusters);
+	if (status < 0)
+		mlog_errno(status);
+
+bail:
+	if (bh)
+		brelse(bh);
+
+	if (last_eb_bh)
+		brelse(last_eb_bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+static inline int ocfs2_truncate_log_needs_flush(struct ocfs2_super *osb)
+{
+	struct buffer_head *tl_bh = osb->osb_tl_bh;
+	struct ocfs2_dinode *di;
+	struct ocfs2_truncate_log *tl;
+
+	di = (struct ocfs2_dinode *) tl_bh->b_data;
+	tl = &di->id2.i_dealloc;
+
+	mlog_bug_on_msg(le16_to_cpu(tl->tl_used) > le16_to_cpu(tl->tl_count),
+			"slot %d, invalid truncate log parameters: used = "
+			"%u, count = %u\n", osb->slot_num,
+			le16_to_cpu(tl->tl_used), le16_to_cpu(tl->tl_count));
+	return le16_to_cpu(tl->tl_used) == le16_to_cpu(tl->tl_count);
+}
+
+static int ocfs2_truncate_log_can_coalesce(struct ocfs2_truncate_log *tl,
+					   unsigned int new_start)
+{
+	unsigned int tail_index;
+	unsigned int current_tail;
+
+	/* No records, nothing to coalesce */
+	if (!le16_to_cpu(tl->tl_used))
+		return 0;
+
+	tail_index = le16_to_cpu(tl->tl_used) - 1;
+	current_tail = le32_to_cpu(tl->tl_recs[tail_index].t_start);
+	current_tail += le32_to_cpu(tl->tl_recs[tail_index].t_clusters);
+
+	return current_tail == new_start;
+}
+
+static int ocfs2_truncate_log_append(struct ocfs2_super *osb,
+				     struct ocfs2_journal_handle *handle,
+				     u64 start_blk,
+				     unsigned int num_clusters)
+{
+	int status, index;
+	unsigned int start_cluster, tl_count;
+	struct inode *tl_inode = osb->osb_tl_inode;
+	struct buffer_head *tl_bh = osb->osb_tl_bh;
+	struct ocfs2_dinode *di;
+	struct ocfs2_truncate_log *tl;
+
+	mlog_entry("start_blk = %"MLFu64", num_clusters = %u\n", start_blk,
+		   num_clusters);
+
+	BUG_ON(!down_trylock(&tl_inode->i_sem));
+
+	start_cluster = ocfs2_blocks_to_clusters(osb->sb, start_blk);
+
+	di = (struct ocfs2_dinode *) tl_bh->b_data;
+	tl = &di->id2.i_dealloc;
+	if (!OCFS2_IS_VALID_DINODE(di)) {
+		OCFS2_RO_ON_INVALID_DINODE(osb->sb, di);
+		status = -EIO;
+		goto bail;
+	}
+
+	tl_count = le16_to_cpu(tl->tl_count);
+	mlog_bug_on_msg(tl_count > ocfs2_truncate_recs_per_inode(osb->sb) ||
+			tl_count == 0,
+			"Truncate record count on #%"MLFu64" invalid ("
+			"wanted %u, actual %u\n", OCFS2_I(tl_inode)->ip_blkno,
+			ocfs2_truncate_recs_per_inode(osb->sb),
+			le16_to_cpu(tl->tl_count));
+
+	/* Caller should have known to flush before calling us. */
+	index = le16_to_cpu(tl->tl_used);
+	if (index >= tl_count) {
+		status = -ENOSPC;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_journal_access(handle, tl_inode, tl_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	mlog(0, "Log truncate of %u clusters starting at cluster %u to "
+	     "%"MLFu64" (index = %d)\n", num_clusters, start_cluster,
+	     OCFS2_I(tl_inode)->ip_blkno, index);
+
+	if (ocfs2_truncate_log_can_coalesce(tl, start_cluster)) {
+		/*
+		 * Move index back to the record we are coalescing with.
+		 * ocfs2_truncate_log_can_coalesce() guarantees nonzero
+		 */
+		index--;
+
+		num_clusters += le32_to_cpu(tl->tl_recs[index].t_clusters);
+		mlog(0, "Coalesce with index %u (start = %u, clusters = %u)\n",
+		     index, le32_to_cpu(tl->tl_recs[index].t_start),
+		     num_clusters);
+	} else {
+		tl->tl_recs[index].t_start = cpu_to_le32(start_cluster);
+		tl->tl_used = cpu_to_le16(index + 1);
+	}
+	tl->tl_recs[index].t_clusters = cpu_to_le32(num_clusters);
+
+	status = ocfs2_journal_dirty(handle, tl_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_replay_truncate_records(struct ocfs2_super *osb,
+					 struct ocfs2_journal_handle *handle,
+					 struct inode *data_alloc_inode,
+					 struct buffer_head *data_alloc_bh)
+{
+	int status = 0;
+	int i;
+	unsigned int num_clusters;
+	u64 start_blk;
+	struct ocfs2_truncate_rec rec;
+	struct ocfs2_dinode *di;
+	struct ocfs2_truncate_log *tl;
+	struct inode *tl_inode = osb->osb_tl_inode;
+	struct buffer_head *tl_bh = osb->osb_tl_bh;
+
+	mlog_entry_void();
+
+	di = (struct ocfs2_dinode *) tl_bh->b_data;
+	tl = &di->id2.i_dealloc;
+	i = le16_to_cpu(tl->tl_used) - 1;
+	while (i >= 0) {
+		/* Caller has given us at least enough credits to
+		 * update the truncate log dinode */
+		status = ocfs2_journal_access(handle, tl_inode, tl_bh,
+					      OCFS2_JOURNAL_ACCESS_WRITE);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		tl->tl_used = cpu_to_le16(i);
+
+		status = ocfs2_journal_dirty(handle, tl_bh);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		/* TODO: Perhaps we can calculate the bulk of the
+		 * credits up front rather than extending like
+		 * this. */
+		status = ocfs2_extend_trans(handle,
+					    OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		rec = tl->tl_recs[i];
+		start_blk = ocfs2_clusters_to_blocks(data_alloc_inode->i_sb,
+						    le32_to_cpu(rec.t_start));
+		num_clusters = le32_to_cpu(rec.t_clusters);
+
+		/* if start_blk is not set, we ignore the record as
+		 * invalid. */
+		if (start_blk) {
+			mlog(0, "free record %d, start = %u, clusters = %u\n",
+			     i, le32_to_cpu(rec.t_start), num_clusters);
+
+			status = ocfs2_free_clusters(handle, data_alloc_inode,
+						     data_alloc_bh, start_blk,
+						     num_clusters);
+			if (status < 0) {
+				mlog_errno(status);
+				goto bail;
+			}
+		}
+		i--;
+	}
+
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+/* Expects you to already be holding tl_inode->i_sem */
+static int __ocfs2_flush_truncate_log(struct ocfs2_super *osb)
+{
+	int status;
+	unsigned int num_to_flush;
+	struct ocfs2_journal_handle *handle = NULL;
+	struct inode *tl_inode = osb->osb_tl_inode;
+	struct inode *data_alloc_inode = NULL;
+	struct buffer_head *tl_bh = osb->osb_tl_bh;
+	struct buffer_head *data_alloc_bh = NULL;
+	struct ocfs2_dinode *di;
+	struct ocfs2_truncate_log *tl;
+
+	mlog_entry_void();
+
+	BUG_ON(!down_trylock(&tl_inode->i_sem));
+
+	di = (struct ocfs2_dinode *) tl_bh->b_data;
+	tl = &di->id2.i_dealloc;
+	if (!OCFS2_IS_VALID_DINODE(di)) {
+		OCFS2_RO_ON_INVALID_DINODE(osb->sb, di);
+		status = -EIO;
+		goto bail;
+	}
+
+	num_to_flush = le16_to_cpu(tl->tl_used);
+	mlog(0, "Flush %u records from truncate log #%"MLFu64"\n",
+	     num_to_flush, OCFS2_I(tl_inode)->ip_blkno);
+	if (!num_to_flush) {
+		status = 0;
+		goto bail;
+	}
+
+	handle = ocfs2_alloc_handle(osb);
+	if (!handle) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	data_alloc_inode = ocfs2_get_system_file_inode(osb,
+						       GLOBAL_BITMAP_SYSTEM_INODE,
+						       OCFS2_INVALID_SLOT);
+	if (!data_alloc_inode) {
+		status = -EINVAL;
+		mlog(ML_ERROR, "Could not get bitmap inode!\n");
+		goto bail;
+	}
+
+	ocfs2_handle_add_inode(handle, data_alloc_inode);
+	status = ocfs2_meta_lock(data_alloc_inode, handle, &data_alloc_bh, 1);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	handle = ocfs2_start_trans(osb, handle, OCFS2_TRUNCATE_LOG_UPDATE);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		handle = NULL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_replay_truncate_records(osb, handle, data_alloc_inode,
+					       data_alloc_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+bail:
+	if (handle)
+		ocfs2_commit_trans(handle);
+
+	if (data_alloc_inode)
+		iput(data_alloc_inode);
+
+	if (data_alloc_bh)
+		brelse(data_alloc_bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+int ocfs2_flush_truncate_log(struct ocfs2_super *osb)
+{
+	int status;
+	struct inode *tl_inode = osb->osb_tl_inode;
+
+	down(&tl_inode->i_sem);
+	status = __ocfs2_flush_truncate_log(osb);
+	up(&tl_inode->i_sem);
+
+	return status;
+}
+
+static void ocfs2_truncate_log_worker(void *data)
+{
+	int status;
+	struct ocfs2_super *osb = data;
+
+	mlog_entry_void();
+
+	status = ocfs2_flush_truncate_log(osb);
+	if (status < 0)
+		mlog_errno(status);
+
+	mlog_exit(status);
+}
+
+#define OCFS2_TRUNCATE_LOG_FLUSH_INTERVAL (2 * HZ)
+void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb,
+				       int cancel)
+{
+	if (osb->osb_tl_inode) {
+		/* We want to push off log flushes while truncates are
+		 * still running. */
+		if (cancel)
+			cancel_delayed_work(&osb->osb_truncate_log_wq);
+
+		queue_delayed_work(ocfs2_wq, &osb->osb_truncate_log_wq,
+				   OCFS2_TRUNCATE_LOG_FLUSH_INTERVAL);
+	}
+}
+
+static int ocfs2_get_truncate_log_info(struct ocfs2_super *osb,
+				       int slot_num,
+				       struct inode **tl_inode,
+				       struct buffer_head **tl_bh)
+{
+	int status;
+	struct inode *inode = NULL;
+	struct buffer_head *bh = NULL;
+
+	inode = ocfs2_get_system_file_inode(osb,
+					   TRUNCATE_LOG_SYSTEM_INODE,
+					   slot_num);
+	if (!inode) {
+		status = -EINVAL;
+		mlog(ML_ERROR, "Could not get load truncate log inode!\n");
+		goto bail;
+	}
+
+	status = ocfs2_read_block(osb, OCFS2_I(inode)->ip_blkno, &bh,
+				  OCFS2_BH_CACHED, inode);
+	if (status < 0) {
+		iput(inode);
+		mlog_errno(status);
+		goto bail;
+	}
+
+	*tl_inode = inode;
+	*tl_bh    = bh;
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+/* called during the 1st stage of node recovery. we stamp a clean
+ * truncate log and pass back a copy for processing later. if the
+ * truncate log does not require processing, a *tl_copy is set to
+ * NULL. */
+int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb,
+				      int slot_num,
+				      struct ocfs2_dinode **tl_copy)
+{
+	int status;
+	struct inode *tl_inode = NULL;
+	struct buffer_head *tl_bh = NULL;
+	struct ocfs2_dinode *di;
+	struct ocfs2_truncate_log *tl;
+
+	*tl_copy = NULL;
+
+	mlog(0, "recover truncate log from slot %d\n", slot_num);
+
+	status = ocfs2_get_truncate_log_info(osb, slot_num, &tl_inode, &tl_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	di = (struct ocfs2_dinode *) tl_bh->b_data;
+	tl = &di->id2.i_dealloc;
+	if (!OCFS2_IS_VALID_DINODE(di)) {
+		OCFS2_RO_ON_INVALID_DINODE(tl_inode->i_sb, di);
+		status = -EIO;
+		goto bail;
+	}
+
+	if (le16_to_cpu(tl->tl_used)) {
+		mlog(0, "We'll have %u logs to recover\n",
+		     le16_to_cpu(tl->tl_used));
+
+		*tl_copy = kmalloc(tl_bh->b_size, GFP_KERNEL);
+		if (!(*tl_copy)) {
+			status = -ENOMEM;
+			mlog_errno(status);
+			goto bail;
+		}
+
+		/* Assuming the write-out below goes well, this copy
+		 * will be passed back to recovery for processing. */
+		memcpy(*tl_copy, tl_bh->b_data, tl_bh->b_size);
+
+		/* All we need to do to clear the truncate log is set
+		 * tl_used. */
+		tl->tl_used = 0;
+
+		status = ocfs2_write_block(osb, tl_bh, tl_inode);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+	}
+
+bail:
+	if (tl_inode)
+		iput(tl_inode);
+	if (tl_bh)
+		brelse(tl_bh);
+
+	if (status < 0 && (*tl_copy)) {
+		kfree(*tl_copy);
+		*tl_copy = NULL;
+	}
+
+	mlog_exit(status);
+	return status;
+}
+
+int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb,
+					 struct ocfs2_dinode *tl_copy)
+{
+	int status = 0;
+	int i;
+	unsigned int clusters, num_recs, start_cluster;
+	u64 start_blk;
+	struct ocfs2_journal_handle *handle;
+	struct inode *tl_inode = osb->osb_tl_inode;
+	struct ocfs2_truncate_log *tl;
+
+	mlog_entry_void();
+
+	if (OCFS2_I(tl_inode)->ip_blkno == le64_to_cpu(tl_copy->i_blkno)) {
+		mlog(ML_ERROR, "Asked to recover my own truncate log!\n");
+		return -EINVAL;
+	}
+
+	tl = &tl_copy->id2.i_dealloc;
+	num_recs = le16_to_cpu(tl->tl_used);
+	mlog(0, "cleanup %u records from %"MLFu64"\n", num_recs,
+	     tl_copy->i_blkno);
+
+	down(&tl_inode->i_sem);
+	for(i = 0; i < num_recs; i++) {
+		if (ocfs2_truncate_log_needs_flush(osb)) {
+			status = __ocfs2_flush_truncate_log(osb);
+			if (status < 0) {
+				mlog_errno(status);
+				goto bail_up;
+			}
+		}
+
+		handle = ocfs2_start_trans(osb, NULL,
+					   OCFS2_TRUNCATE_LOG_UPDATE);
+		if (IS_ERR(handle)) {
+			status = PTR_ERR(handle);
+			mlog_errno(status);
+			goto bail_up;
+		}
+
+		clusters = le32_to_cpu(tl->tl_recs[i].t_clusters);
+		start_cluster = le32_to_cpu(tl->tl_recs[i].t_start);
+		start_blk = ocfs2_clusters_to_blocks(osb->sb, start_cluster);
+
+		status = ocfs2_truncate_log_append(osb, handle,
+						   start_blk, clusters);
+		ocfs2_commit_trans(handle);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail_up;
+		}
+	}
+
+bail_up:
+	up(&tl_inode->i_sem);
+
+	mlog_exit(status);
+	return status;
+}
+
+void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb)
+{
+	int status;
+	struct inode *tl_inode = osb->osb_tl_inode;
+
+	mlog_entry_void();
+
+	if (tl_inode) {
+		cancel_delayed_work(&osb->osb_truncate_log_wq);
+		flush_workqueue(ocfs2_wq);
+
+		status = ocfs2_flush_truncate_log(osb);
+		if (status < 0)
+			mlog_errno(status);
+
+		brelse(osb->osb_tl_bh);
+		iput(osb->osb_tl_inode);
+	}
+
+	mlog_exit_void();
+}
+
+int ocfs2_truncate_log_init(struct ocfs2_super *osb)
+{
+	int status;
+	struct inode *tl_inode = NULL;
+	struct buffer_head *tl_bh = NULL;
+
+	mlog_entry_void();
+
+	status = ocfs2_get_truncate_log_info(osb,
+					     osb->slot_num,
+					     &tl_inode,
+					     &tl_bh);
+	if (status < 0)
+		mlog_errno(status);
+
+	/* ocfs2_truncate_log_shutdown keys on the existence of
+	 * osb->osb_tl_inode so we don't set any of the osb variables
+	 * until we're sure all is well. */
+	INIT_WORK(&osb->osb_truncate_log_wq, ocfs2_truncate_log_worker, osb);
+	osb->osb_tl_bh    = tl_bh;
+	osb->osb_tl_inode = tl_inode;
+
+	mlog_exit(status);
+	return status;
+}
+
+/* This function will figure out whether the currently last extent
+ * block will be deleted, and if it will, what the new last extent
+ * block will be so we can update his h_next_leaf_blk field, as well
+ * as the dinodes i_last_eb_blk */
+static int ocfs2_find_new_last_ext_blk(struct ocfs2_super *osb,
+				       struct inode *inode,
+				       struct ocfs2_dinode *fe,
+				       u32 new_i_clusters,
+				       struct buffer_head *old_last_eb,
+				       struct buffer_head **new_last_eb)
+{
+	int i, status = 0;
+	u64 block = 0;
+	struct ocfs2_extent_block *eb;
+	struct ocfs2_extent_list *el;
+	struct buffer_head *bh = NULL;
+
+	*new_last_eb = NULL;
+
+	if (!OCFS2_IS_VALID_DINODE(fe)) {
+		OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe);
+		status = -EIO;
+		goto bail;
+	}
+
+	/* we have no tree, so of course, no last_eb. */
+	if (!fe->id2.i_list.l_tree_depth)
+		goto bail;
+
+	/* trunc to zero special case - this makes tree_depth = 0
+	 * regardless of what it is.  */
+	if (!new_i_clusters)
+		goto bail;
+
+	eb = (struct ocfs2_extent_block *) old_last_eb->b_data;
+	el = &(eb->h_list);
+	BUG_ON(!el->l_next_free_rec);
+
+	/* Make sure that this guy will actually be empty after we
+	 * clear away the data. */
+	if (le32_to_cpu(el->l_recs[0].e_cpos) < new_i_clusters)
+		goto bail;
+
+	/* Ok, at this point, we know that last_eb will definitely
+	 * change, so lets traverse the tree and find the second to
+	 * last extent block. */
+	el = &(fe->id2.i_list);
+	/* go down the tree, */
+	do {
+		for(i = (le16_to_cpu(el->l_next_free_rec) - 1); i >= 0; i--) {
+			if (le32_to_cpu(el->l_recs[i].e_cpos) <
+			    new_i_clusters) {
+				block = le64_to_cpu(el->l_recs[i].e_blkno);
+				break;
+			}
+		}
+		BUG_ON(i < 0);
+
+		if (bh) {
+			brelse(bh);
+			bh = NULL;
+		}
+
+		status = ocfs2_read_block(osb, block, &bh, OCFS2_BH_CACHED,
+					 inode);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+		eb = (struct ocfs2_extent_block *) bh->b_data;
+		el = &eb->h_list;
+		if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
+			OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
+			status = -EIO;
+			goto bail;
+		}
+	} while (el->l_tree_depth);
+
+	*new_last_eb = bh;
+	get_bh(*new_last_eb);
+	mlog(0, "returning block %"MLFu64"\n", le64_to_cpu(eb->h_blkno));
+bail:
+	if (bh)
+		brelse(bh);
+
+	return status;
+}
+
+static int ocfs2_do_truncate(struct ocfs2_super *osb,
+			     unsigned int clusters_to_del,
+			     struct inode *inode,
+			     struct buffer_head *fe_bh,
+			     struct buffer_head *old_last_eb_bh,
+			     struct ocfs2_journal_handle *handle,
+			     struct ocfs2_truncate_context *tc)
+{
+	int status, i, depth;
+	struct ocfs2_dinode *fe;
+	struct ocfs2_extent_block *eb;
+	struct ocfs2_extent_block *last_eb = NULL;
+	struct ocfs2_extent_list *el;
+	struct buffer_head *eb_bh = NULL;
+	struct buffer_head *last_eb_bh = NULL;
+	u64 next_eb = 0;
+	u64 delete_blk = 0;
+
+	fe = (struct ocfs2_dinode *) fe_bh->b_data;
+
+	status = ocfs2_find_new_last_ext_blk(osb,
+					     inode,
+					     fe,
+					     le32_to_cpu(fe->i_clusters) -
+					     		clusters_to_del,
+					     old_last_eb_bh,
+					     &last_eb_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+	if (last_eb_bh)
+		last_eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
+
+	status = ocfs2_journal_access(handle, inode, fe_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+	el = &(fe->id2.i_list);
+
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+	OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters) -
+				      clusters_to_del;
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+	le32_add_cpu(&fe->i_clusters, -clusters_to_del);
+	fe->i_mtime = cpu_to_le64(CURRENT_TIME.tv_sec);
+	fe->i_mtime_nsec = cpu_to_le32(CURRENT_TIME.tv_nsec);
+
+	i = le16_to_cpu(el->l_next_free_rec) - 1;
+
+	BUG_ON(le32_to_cpu(el->l_recs[i].e_clusters) < clusters_to_del);
+	le32_add_cpu(&el->l_recs[i].e_clusters, -clusters_to_del);
+	/* tree depth zero, we can just delete the clusters, otherwise
+	 * we need to record the offset of the next level extent block
+	 * as we may overwrite it. */
+	if (!el->l_tree_depth)
+		delete_blk = le64_to_cpu(el->l_recs[i].e_blkno)
+			+ ocfs2_clusters_to_blocks(osb->sb,
+					le32_to_cpu(el->l_recs[i].e_clusters));
+	else
+		next_eb = le64_to_cpu(el->l_recs[i].e_blkno);
+
+	if (!el->l_recs[i].e_clusters) {
+		/* if we deleted the whole extent record, then clear
+		 * out the other fields and update the extent
+		 * list. For depth > 0 trees, we've already recorded
+		 * the extent block in 'next_eb' */
+		el->l_recs[i].e_cpos = 0;
+		el->l_recs[i].e_blkno = 0;
+		BUG_ON(!el->l_next_free_rec);
+		le16_add_cpu(&el->l_next_free_rec, -1);
+	}
+
+	depth = le16_to_cpu(el->l_tree_depth);
+	if (!fe->i_clusters) {
+		/* trunc to zero is a special case. */
+		el->l_tree_depth = 0;
+		fe->i_last_eb_blk = 0;
+	} else if (last_eb)
+		fe->i_last_eb_blk = last_eb->h_blkno;
+
+	status = ocfs2_journal_dirty(handle, fe_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	if (last_eb) {
+		/* If there will be a new last extent block, then by
+		 * definition, there cannot be any leaves to the right of
+		 * him. */
+		status = ocfs2_journal_access(handle, inode, last_eb_bh,
+					      OCFS2_JOURNAL_ACCESS_WRITE);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+		last_eb->h_next_leaf_blk = 0;
+		status = ocfs2_journal_dirty(handle, last_eb_bh);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+	}
+
+	/* if our tree depth > 0, update all the tree blocks below us. */
+	while (depth) {
+		mlog(0, "traveling tree (depth = %d, next_eb = %"MLFu64")\n",
+		     depth,  next_eb);
+		status = ocfs2_read_block(osb, next_eb, &eb_bh,
+					  OCFS2_BH_CACHED, inode);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+		eb = (struct ocfs2_extent_block *)eb_bh->b_data;
+		if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
+			OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
+			status = -EIO;
+			goto bail;
+		}
+		el = &(eb->h_list);
+
+		status = ocfs2_journal_access(handle, inode, eb_bh,
+					      OCFS2_JOURNAL_ACCESS_WRITE);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		BUG_ON(le16_to_cpu(el->l_next_free_rec) == 0);
+		BUG_ON(depth != (le16_to_cpu(el->l_tree_depth) + 1));
+
+		i = le16_to_cpu(el->l_next_free_rec) - 1;
+
+		mlog(0, "extent block %"MLFu64", before: record %d: "
+		     "(%u, %u, %"MLFu64"), next = %u\n",
+		     le64_to_cpu(eb->h_blkno), i,
+		     le32_to_cpu(el->l_recs[i].e_cpos),
+		     le32_to_cpu(el->l_recs[i].e_clusters),
+		     le64_to_cpu(el->l_recs[i].e_blkno),
+		     le16_to_cpu(el->l_next_free_rec));
+
+		BUG_ON(le32_to_cpu(el->l_recs[i].e_clusters) < clusters_to_del);
+		le32_add_cpu(&el->l_recs[i].e_clusters, -clusters_to_del);
+
+		next_eb = le64_to_cpu(el->l_recs[i].e_blkno);
+		/* bottom-most block requires us to delete data.*/
+		if (!el->l_tree_depth)
+			delete_blk = le64_to_cpu(el->l_recs[i].e_blkno)
+				+ ocfs2_clusters_to_blocks(osb->sb,
+					le32_to_cpu(el->l_recs[i].e_clusters));
+		if (!el->l_recs[i].e_clusters) {
+			el->l_recs[i].e_cpos = 0;
+			el->l_recs[i].e_blkno = 0;
+			BUG_ON(!el->l_next_free_rec);
+			le16_add_cpu(&el->l_next_free_rec, -1);
+		}
+		mlog(0, "extent block %"MLFu64", after: record %d: "
+		     "(%u, %u, %"MLFu64"), next = %u\n",
+		     le64_to_cpu(eb->h_blkno), i,
+		     le32_to_cpu(el->l_recs[i].e_cpos),
+		     le32_to_cpu(el->l_recs[i].e_clusters),
+		     le64_to_cpu(el->l_recs[i].e_blkno),
+		     le16_to_cpu(el->l_next_free_rec));
+
+		status = ocfs2_journal_dirty(handle, eb_bh);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		if (!el->l_next_free_rec) {
+			mlog(0, "deleting this extent block.\n");
+
+			ocfs2_remove_from_cache(inode, eb_bh);
+
+			BUG_ON(eb->h_suballoc_slot);
+			BUG_ON(el->l_recs[0].e_clusters);
+			BUG_ON(el->l_recs[0].e_cpos);
+			BUG_ON(el->l_recs[0].e_blkno);
+			status = ocfs2_free_extent_block(handle,
+							 tc->tc_ext_alloc_inode,
+							 tc->tc_ext_alloc_bh,
+							 eb);
+			if (status < 0) {
+				mlog_errno(status);
+				goto bail;
+			}
+		}
+		brelse(eb_bh);
+		eb_bh = NULL;
+		depth--;
+	}
+
+	BUG_ON(!delete_blk);
+	status = ocfs2_truncate_log_append(osb, handle, delete_blk,
+					   clusters_to_del);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+	status = 0;
+bail:
+	if (!status)
+		ocfs2_extent_map_trunc(inode, le32_to_cpu(fe->i_clusters));
+	else
+		ocfs2_extent_map_drop(inode, 0);
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * It is expected, that by the time you call this function,
+ * inode->i_size and fe->i_size have been adjusted.
+ *
+ * WARNING: This will kfree the truncate context
+ */
+int ocfs2_commit_truncate(struct ocfs2_super *osb,
+			  struct inode *inode,
+			  struct buffer_head *fe_bh,
+			  struct ocfs2_truncate_context *tc)
+{
+	int status, i, credits, tl_sem = 0;
+	u32 clusters_to_del, target_i_clusters;
+	u64 last_eb = 0;
+	struct ocfs2_dinode *fe;
+	struct ocfs2_extent_block *eb;
+	struct ocfs2_extent_list *el;
+	struct buffer_head *last_eb_bh;
+	struct ocfs2_journal_handle *handle = NULL;
+	struct inode *tl_inode = osb->osb_tl_inode;
+
+	mlog_entry_void();
+
+	down_write(&OCFS2_I(inode)->ip_alloc_sem);
+
+	target_i_clusters = ocfs2_clusters_for_bytes(osb->sb,
+						     i_size_read(inode));
+
+	last_eb_bh = tc->tc_last_eb_bh;
+	tc->tc_last_eb_bh = NULL;
+
+	fe = (struct ocfs2_dinode *) fe_bh->b_data;
+
+	if (fe->id2.i_list.l_tree_depth) {
+		eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
+		el = &eb->h_list;
+	} else
+		el = &fe->id2.i_list;
+	last_eb = le64_to_cpu(fe->i_last_eb_blk);
+start:
+	mlog(0, "ocfs2_commit_truncate: fe->i_clusters = %u, "
+	     "last_eb = %"MLFu64", fe->i_last_eb_blk = %"MLFu64", "
+	     "fe->id2.i_list.l_tree_depth = %u last_eb_bh = %p\n",
+	     le32_to_cpu(fe->i_clusters), last_eb,
+	     le64_to_cpu(fe->i_last_eb_blk),
+	     le16_to_cpu(fe->id2.i_list.l_tree_depth), last_eb_bh);
+
+	if (last_eb != le64_to_cpu(fe->i_last_eb_blk)) {
+		mlog(0, "last_eb changed!\n");
+		BUG_ON(!fe->id2.i_list.l_tree_depth);
+		last_eb = le64_to_cpu(fe->i_last_eb_blk);
+		/* i_last_eb_blk may have changed, read it if
+		 * necessary. We don't have to worry about the
+		 * truncate to zero case here (where there becomes no
+		 * last_eb) because we never loop back after our work
+		 * is done. */
+		if (last_eb_bh) {
+			brelse(last_eb_bh);
+			last_eb_bh = NULL;
+		}
+
+		status = ocfs2_read_block(osb, last_eb,
+					  &last_eb_bh, OCFS2_BH_CACHED,
+					  inode);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+		eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
+		if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
+			OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
+			status = -EIO;
+			goto bail;
+		}
+		el = &(eb->h_list);
+	}
+
+	/* by now, el will point to the extent list on the bottom most
+	 * portion of this tree. */
+	i = le16_to_cpu(el->l_next_free_rec) - 1;
+	if (le32_to_cpu(el->l_recs[i].e_cpos) >= target_i_clusters)
+		clusters_to_del = le32_to_cpu(el->l_recs[i].e_clusters);
+	else
+		clusters_to_del = (le32_to_cpu(el->l_recs[i].e_clusters) +
+				   le32_to_cpu(el->l_recs[i].e_cpos)) -
+				  target_i_clusters;
+
+	mlog(0, "clusters_to_del = %u in this pass\n", clusters_to_del);
+
+	down(&tl_inode->i_sem);
+	tl_sem = 1;
+	/* ocfs2_truncate_log_needs_flush guarantees us at least one
+	 * record is free for use. If there isn't any, we flush to get
+	 * an empty truncate log.  */
+	if (ocfs2_truncate_log_needs_flush(osb)) {
+		status = __ocfs2_flush_truncate_log(osb);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+	}
+
+	credits = ocfs2_calc_tree_trunc_credits(osb->sb, clusters_to_del,
+						fe, el);
+	handle = ocfs2_start_trans(osb, NULL, credits);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		handle = NULL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	inode->i_ctime = inode->i_mtime = CURRENT_TIME;
+	status = ocfs2_mark_inode_dirty(handle, inode, fe_bh);
+	if (status < 0)
+		mlog_errno(status);
+
+	status = ocfs2_do_truncate(osb, clusters_to_del, inode, fe_bh,
+				   last_eb_bh, handle, tc);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	up(&tl_inode->i_sem);
+	tl_sem = 0;
+
+	ocfs2_commit_trans(handle);
+	handle = NULL;
+
+	BUG_ON(le32_to_cpu(fe->i_clusters) < target_i_clusters);
+	if (le32_to_cpu(fe->i_clusters) > target_i_clusters)
+		goto start;
+bail:
+	up_write(&OCFS2_I(inode)->ip_alloc_sem);
+
+	ocfs2_schedule_truncate_log_flush(osb, 1);
+
+	if (tl_sem)
+		up(&tl_inode->i_sem);
+
+	if (handle)
+		ocfs2_commit_trans(handle);
+
+	if (last_eb_bh)
+		brelse(last_eb_bh);
+
+	/* This will drop the ext_alloc cluster lock for us */
+	ocfs2_free_truncate_context(tc);
+
+	mlog_exit(status);
+	return status;
+}
+
+
+/*
+ * Expects the inode to already be locked. This will figure out which
+ * inodes need to be locked and will put them on the returned truncate
+ * context.
+ */
+int ocfs2_prepare_truncate(struct ocfs2_super *osb,
+			   struct inode *inode,
+			   struct buffer_head *fe_bh,
+			   struct ocfs2_truncate_context **tc)
+{
+	int status, metadata_delete;
+	unsigned int new_i_clusters;
+	struct ocfs2_dinode *fe;
+	struct ocfs2_extent_block *eb;
+	struct ocfs2_extent_list *el;
+	struct buffer_head *last_eb_bh = NULL;
+	struct inode *ext_alloc_inode = NULL;
+	struct buffer_head *ext_alloc_bh = NULL;
+
+	mlog_entry_void();
+
+	*tc = NULL;
+
+	new_i_clusters = ocfs2_clusters_for_bytes(osb->sb,
+						  i_size_read(inode));
+	fe = (struct ocfs2_dinode *) fe_bh->b_data;
+
+	mlog(0, "fe->i_clusters = %u, new_i_clusters = %u, fe->i_size ="
+	     "%"MLFu64"\n", fe->i_clusters, new_i_clusters, fe->i_size);
+
+	if (le32_to_cpu(fe->i_clusters) <= new_i_clusters) {
+		ocfs2_error(inode->i_sb, "Dinode %"MLFu64" has cluster count "
+			    "%u and size %"MLFu64" whereas struct inode has "
+			    "cluster count %u and size %llu which caused an "
+			    "invalid truncate to %u clusters.",
+			    le64_to_cpu(fe->i_blkno),
+			    le32_to_cpu(fe->i_clusters),
+			    le64_to_cpu(fe->i_size),
+			    OCFS2_I(inode)->ip_clusters, i_size_read(inode),
+			    new_i_clusters);
+		mlog_meta_lvb(ML_ERROR, &OCFS2_I(inode)->ip_meta_lockres);
+		status = -EIO;
+		goto bail;
+	}
+
+	*tc = kcalloc(1, sizeof(struct ocfs2_truncate_context), GFP_KERNEL);
+	if (!(*tc)) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	metadata_delete = 0;
+	if (fe->id2.i_list.l_tree_depth) {
+		/* If we have a tree, then the truncate may result in
+		 * metadata deletes. Figure this out from the
+		 * rightmost leaf block.*/
+		status = ocfs2_read_block(osb, le64_to_cpu(fe->i_last_eb_blk),
+					  &last_eb_bh, OCFS2_BH_CACHED, inode);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+		eb = (struct ocfs2_extent_block *) last_eb_bh->b_data;
+		if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
+			OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
+
+			brelse(last_eb_bh);
+			status = -EIO;
+			goto bail;
+		}
+		el = &(eb->h_list);
+		if (le32_to_cpu(el->l_recs[0].e_cpos) >= new_i_clusters)
+			metadata_delete = 1;
+	}
+
+	(*tc)->tc_last_eb_bh = last_eb_bh;
+
+	if (metadata_delete) {
+		mlog(0, "Will have to delete metadata for this trunc. "
+		     "locking allocator.\n");
+		ext_alloc_inode = ocfs2_get_system_file_inode(osb, EXTENT_ALLOC_SYSTEM_INODE, 0);
+		if (!ext_alloc_inode) {
+			status = -ENOMEM;
+			mlog_errno(status);
+			goto bail;
+		}
+
+		down(&ext_alloc_inode->i_sem);
+		(*tc)->tc_ext_alloc_inode = ext_alloc_inode;
+
+		status = ocfs2_meta_lock(ext_alloc_inode,
+					 NULL,
+					 &ext_alloc_bh,
+					 1);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+		(*tc)->tc_ext_alloc_bh = ext_alloc_bh;
+		(*tc)->tc_ext_alloc_locked = 1;
+	}
+
+	status = 0;
+bail:
+	if (status < 0) {
+		if (*tc)
+			ocfs2_free_truncate_context(*tc);
+		*tc = NULL;
+	}
+	mlog_exit_void();
+	return status;
+}
+
+static void ocfs2_free_truncate_context(struct ocfs2_truncate_context *tc)
+{
+	if (tc->tc_ext_alloc_inode) {
+		if (tc->tc_ext_alloc_locked)
+			ocfs2_meta_unlock(tc->tc_ext_alloc_inode, 1);
+
+		up(&tc->tc_ext_alloc_inode->i_sem);
+		iput(tc->tc_ext_alloc_inode);
+	}
+
+	if (tc->tc_ext_alloc_bh)
+		brelse(tc->tc_ext_alloc_bh);
+
+	if (tc->tc_last_eb_bh)
+		brelse(tc->tc_last_eb_bh);
+
+	kfree(tc);
+}
diff --git a/fs/ocfs2/alloc.h b/fs/ocfs2/alloc.h
new file mode 100644
index 00000000000..12ba897743f
--- /dev/null
+++ b/fs/ocfs2/alloc.h
@@ -0,0 +1,82 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * alloc.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_ALLOC_H
+#define OCFS2_ALLOC_H
+
+struct ocfs2_alloc_context;
+int ocfs2_insert_extent(struct ocfs2_super *osb,
+			struct ocfs2_journal_handle *handle,
+			struct inode *inode,
+			struct buffer_head *fe_bh,
+			u64 blkno,
+			u32 new_clusters,
+			struct ocfs2_alloc_context *meta_ac);
+int ocfs2_num_free_extents(struct ocfs2_super *osb,
+			   struct inode *inode,
+			   struct ocfs2_dinode *fe);
+/* how many new metadata chunks would an allocation need at maximum? */
+static inline int ocfs2_extend_meta_needed(struct ocfs2_dinode *fe)
+{
+	/*
+	 * Rather than do all the work of determining how much we need
+	 * (involves a ton of reads and locks), just ask for the
+	 * maximal limit.  That's a tree depth shift.  So, one block for
+	 * level of the tree (current l_tree_depth), one block for the
+	 * new tree_depth==0 extent_block, and one block at the new
+	 * top-of-the tree.
+	 */
+	return le16_to_cpu(fe->id2.i_list.l_tree_depth) + 2;
+}
+
+int ocfs2_truncate_log_init(struct ocfs2_super *osb);
+void ocfs2_truncate_log_shutdown(struct ocfs2_super *osb);
+void ocfs2_schedule_truncate_log_flush(struct ocfs2_super *osb,
+				       int cancel);
+int ocfs2_flush_truncate_log(struct ocfs2_super *osb);
+int ocfs2_begin_truncate_log_recovery(struct ocfs2_super *osb,
+				      int slot_num,
+				      struct ocfs2_dinode **tl_copy);
+int ocfs2_complete_truncate_log_recovery(struct ocfs2_super *osb,
+					 struct ocfs2_dinode *tl_copy);
+
+struct ocfs2_truncate_context {
+	struct inode *tc_ext_alloc_inode;
+	struct buffer_head *tc_ext_alloc_bh;
+	int tc_ext_alloc_locked; /* is it cluster locked? */
+	/* these get destroyed once it's passed to ocfs2_commit_truncate. */
+	struct buffer_head *tc_last_eb_bh;
+};
+
+int ocfs2_prepare_truncate(struct ocfs2_super *osb,
+			   struct inode *inode,
+			   struct buffer_head *fe_bh,
+			   struct ocfs2_truncate_context **tc);
+int ocfs2_commit_truncate(struct ocfs2_super *osb,
+			  struct inode *inode,
+			  struct buffer_head *fe_bh,
+			  struct ocfs2_truncate_context *tc);
+
+#endif /* OCFS2_ALLOC_H */
diff --git a/fs/ocfs2/aops.c b/fs/ocfs2/aops.c
new file mode 100644
index 00000000000..8f4467a930a
--- /dev/null
+++ b/fs/ocfs2/aops.c
@@ -0,0 +1,643 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/pagemap.h>
+#include <asm/byteorder.h>
+
+#define MLOG_MASK_PREFIX ML_FILE_IO
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "aops.h"
+#include "dlmglue.h"
+#include "extent_map.h"
+#include "file.h"
+#include "inode.h"
+#include "journal.h"
+#include "super.h"
+#include "symlink.h"
+
+#include "buffer_head_io.h"
+
+static int ocfs2_symlink_get_block(struct inode *inode, sector_t iblock,
+				   struct buffer_head *bh_result, int create)
+{
+	int err = -EIO;
+	int status;
+	struct ocfs2_dinode *fe = NULL;
+	struct buffer_head *bh = NULL;
+	struct buffer_head *buffer_cache_bh = NULL;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	void *kaddr;
+
+	mlog_entry("(0x%p, %llu, 0x%p, %d)\n", inode,
+		   (unsigned long long)iblock, bh_result, create);
+
+	BUG_ON(ocfs2_inode_is_fast_symlink(inode));
+
+	if ((iblock << inode->i_sb->s_blocksize_bits) > PATH_MAX + 1) {
+		mlog(ML_ERROR, "block offset > PATH_MAX: %llu",
+		     (unsigned long long)iblock);
+		goto bail;
+	}
+
+	status = ocfs2_read_block(OCFS2_SB(inode->i_sb),
+				  OCFS2_I(inode)->ip_blkno,
+				  &bh, OCFS2_BH_CACHED, inode);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+	fe = (struct ocfs2_dinode *) bh->b_data;
+
+	if (!OCFS2_IS_VALID_DINODE(fe)) {
+		mlog(ML_ERROR, "Invalid dinode #%"MLFu64": signature = %.*s\n",
+		     fe->i_blkno, 7, fe->i_signature);
+		goto bail;
+	}
+
+	if ((u64)iblock >= ocfs2_clusters_to_blocks(inode->i_sb,
+						    le32_to_cpu(fe->i_clusters))) {
+		mlog(ML_ERROR, "block offset is outside the allocated size: "
+		     "%llu\n", (unsigned long long)iblock);
+		goto bail;
+	}
+
+	/* We don't use the page cache to create symlink data, so if
+	 * need be, copy it over from the buffer cache. */
+	if (!buffer_uptodate(bh_result) && ocfs2_inode_is_new(inode)) {
+		u64 blkno = le64_to_cpu(fe->id2.i_list.l_recs[0].e_blkno) +
+			    iblock;
+		buffer_cache_bh = sb_getblk(osb->sb, blkno);
+		if (!buffer_cache_bh) {
+			mlog(ML_ERROR, "couldn't getblock for symlink!\n");
+			goto bail;
+		}
+
+		/* we haven't locked out transactions, so a commit
+		 * could've happened. Since we've got a reference on
+		 * the bh, even if it commits while we're doing the
+		 * copy, the data is still good. */
+		if (buffer_jbd(buffer_cache_bh)
+		    && ocfs2_inode_is_new(inode)) {
+			kaddr = kmap_atomic(bh_result->b_page, KM_USER0);
+			if (!kaddr) {
+				mlog(ML_ERROR, "couldn't kmap!\n");
+				goto bail;
+			}
+			memcpy(kaddr + (bh_result->b_size * iblock),
+			       buffer_cache_bh->b_data,
+			       bh_result->b_size);
+			kunmap_atomic(kaddr, KM_USER0);
+			set_buffer_uptodate(bh_result);
+		}
+		brelse(buffer_cache_bh);
+	}
+
+	map_bh(bh_result, inode->i_sb,
+	       le64_to_cpu(fe->id2.i_list.l_recs[0].e_blkno) + iblock);
+
+	err = 0;
+
+bail:
+	if (bh)
+		brelse(bh);
+
+	mlog_exit(err);
+	return err;
+}
+
+static int ocfs2_get_block(struct inode *inode, sector_t iblock,
+			   struct buffer_head *bh_result, int create)
+{
+	int err = 0;
+	u64 p_blkno, past_eof;
+
+	mlog_entry("(0x%p, %llu, 0x%p, %d)\n", inode,
+		   (unsigned long long)iblock, bh_result, create);
+
+	if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE)
+		mlog(ML_NOTICE, "get_block on system inode 0x%p (%lu)\n",
+		     inode, inode->i_ino);
+
+	if (S_ISLNK(inode->i_mode)) {
+		/* this always does I/O for some reason. */
+		err = ocfs2_symlink_get_block(inode, iblock, bh_result, create);
+		goto bail;
+	}
+
+	/* this can happen if another node truncs after our extend! */
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+	if (iblock >= ocfs2_clusters_to_blocks(inode->i_sb,
+					       OCFS2_I(inode)->ip_clusters))
+		err = -EIO;
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+	if (err)
+		goto bail;
+
+	err = ocfs2_extent_map_get_blocks(inode, iblock, 1, &p_blkno,
+					  NULL);
+	if (err) {
+		mlog(ML_ERROR, "Error %d from get_blocks(0x%p, %llu, 1, "
+		     "%"MLFu64", NULL)\n", err, inode,
+		     (unsigned long long)iblock, p_blkno);
+		goto bail;
+	}
+
+	map_bh(bh_result, inode->i_sb, p_blkno);
+
+	if (bh_result->b_blocknr == 0) {
+		err = -EIO;
+		mlog(ML_ERROR, "iblock = %llu p_blkno = %"MLFu64" "
+		     "blkno=(%"MLFu64")\n", (unsigned long long)iblock,
+		     p_blkno, OCFS2_I(inode)->ip_blkno);
+	}
+
+	past_eof = ocfs2_blocks_for_bytes(inode->i_sb, i_size_read(inode));
+	mlog(0, "Inode %lu, past_eof = %"MLFu64"\n", inode->i_ino, past_eof);
+
+	if (create && (iblock >= past_eof))
+		set_buffer_new(bh_result);
+
+bail:
+	if (err < 0)
+		err = -EIO;
+
+	mlog_exit(err);
+	return err;
+}
+
+static int ocfs2_readpage(struct file *file, struct page *page)
+{
+	struct inode *inode = page->mapping->host;
+	loff_t start = (loff_t)page->index << PAGE_CACHE_SHIFT;
+	int ret, unlock = 1;
+
+	mlog_entry("(0x%p, %lu)\n", file, (page ? page->index : 0));
+
+	ret = ocfs2_meta_lock_with_page(inode, NULL, NULL, 0, page);
+	if (ret != 0) {
+		if (ret == AOP_TRUNCATED_PAGE)
+			unlock = 0;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	down_read(&OCFS2_I(inode)->ip_alloc_sem);
+
+	/*
+	 * i_size might have just been updated as we grabed the meta lock.  We
+	 * might now be discovering a truncate that hit on another node.
+	 * block_read_full_page->get_block freaks out if it is asked to read
+	 * beyond the end of a file, so we check here.  Callers
+	 * (generic_file_read, fault->nopage) are clever enough to check i_size
+	 * and notice that the page they just read isn't needed.
+	 *
+	 * XXX sys_readahead() seems to get that wrong?
+	 */
+	if (start >= i_size_read(inode)) {
+		char *addr = kmap(page);
+		memset(addr, 0, PAGE_SIZE);
+		flush_dcache_page(page);
+		kunmap(page);
+		SetPageUptodate(page);
+		ret = 0;
+		goto out_alloc;
+	}
+
+	ret = ocfs2_data_lock_with_page(inode, 0, page);
+	if (ret != 0) {
+		if (ret == AOP_TRUNCATED_PAGE)
+			unlock = 0;
+		mlog_errno(ret);
+		goto out_alloc;
+	}
+
+	ret = block_read_full_page(page, ocfs2_get_block);
+	unlock = 0;
+
+	ocfs2_data_unlock(inode, 0);
+out_alloc:
+	up_read(&OCFS2_I(inode)->ip_alloc_sem);
+	ocfs2_meta_unlock(inode, 0);
+out:
+	if (unlock)
+		unlock_page(page);
+	mlog_exit(ret);
+	return ret;
+}
+
+/* Note: Because we don't support holes, our allocation has
+ * already happened (allocation writes zeros to the file data)
+ * so we don't have to worry about ordered writes in
+ * ocfs2_writepage.
+ *
+ * ->writepage is called during the process of invalidating the page cache
+ * during blocked lock processing.  It can't block on any cluster locks
+ * to during block mapping.  It's relying on the fact that the block
+ * mapping can't have disappeared under the dirty pages that it is
+ * being asked to write back.
+ */
+static int ocfs2_writepage(struct page *page, struct writeback_control *wbc)
+{
+	int ret;
+
+	mlog_entry("(0x%p)\n", page);
+
+	ret = block_write_full_page(page, ocfs2_get_block, wbc);
+
+	mlog_exit(ret);
+
+	return ret;
+}
+
+/*
+ * ocfs2_prepare_write() can be an outer-most ocfs2 call when it is called
+ * from loopback.  It must be able to perform its own locking around
+ * ocfs2_get_block().
+ */
+int ocfs2_prepare_write(struct file *file, struct page *page,
+			unsigned from, unsigned to)
+{
+	struct inode *inode = page->mapping->host;
+	int ret;
+
+	mlog_entry("(0x%p, 0x%p, %u, %u)\n", file, page, from, to);
+
+	ret = ocfs2_meta_lock_with_page(inode, NULL, NULL, 0, page);
+	if (ret != 0) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	down_read(&OCFS2_I(inode)->ip_alloc_sem);
+
+	ret = block_prepare_write(page, from, to, ocfs2_get_block);
+
+	up_read(&OCFS2_I(inode)->ip_alloc_sem);
+
+	ocfs2_meta_unlock(inode, 0);
+out:
+	mlog_exit(ret);
+	return ret;
+}
+
+/* Taken from ext3. We don't necessarily need the full blown
+ * functionality yet, but IMHO it's better to cut and paste the whole
+ * thing so we can avoid introducing our own bugs (and easily pick up
+ * their fixes when they happen) --Mark */
+static int walk_page_buffers(	handle_t *handle,
+				struct buffer_head *head,
+				unsigned from,
+				unsigned to,
+				int *partial,
+				int (*fn)(	handle_t *handle,
+						struct buffer_head *bh))
+{
+	struct buffer_head *bh;
+	unsigned block_start, block_end;
+	unsigned blocksize = head->b_size;
+	int err, ret = 0;
+	struct buffer_head *next;
+
+	for (	bh = head, block_start = 0;
+		ret == 0 && (bh != head || !block_start);
+	    	block_start = block_end, bh = next)
+	{
+		next = bh->b_this_page;
+		block_end = block_start + blocksize;
+		if (block_end <= from || block_start >= to) {
+			if (partial && !buffer_uptodate(bh))
+				*partial = 1;
+			continue;
+		}
+		err = (*fn)(handle, bh);
+		if (!ret)
+			ret = err;
+	}
+	return ret;
+}
+
+struct ocfs2_journal_handle *ocfs2_start_walk_page_trans(struct inode *inode,
+							 struct page *page,
+							 unsigned from,
+							 unsigned to)
+{
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	struct ocfs2_journal_handle *handle = NULL;
+	int ret = 0;
+
+	handle = ocfs2_start_trans(osb, NULL, OCFS2_INODE_UPDATE_CREDITS);
+	if (!handle) {
+		ret = -ENOMEM;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	if (ocfs2_should_order_data(inode)) {
+		ret = walk_page_buffers(handle->k_handle,
+					page_buffers(page),
+					from, to, NULL,
+					ocfs2_journal_dirty_data);
+		if (ret < 0) 
+			mlog_errno(ret);
+	}
+out:
+	if (ret) {
+		if (handle)
+			ocfs2_commit_trans(handle);
+		handle = ERR_PTR(ret);
+	}
+	return handle;
+}
+
+static int ocfs2_commit_write(struct file *file, struct page *page,
+			      unsigned from, unsigned to)
+{
+	int ret, extending = 0, locklevel = 0;
+	loff_t new_i_size;
+	struct buffer_head *di_bh = NULL;
+	struct inode *inode = page->mapping->host;
+	struct ocfs2_journal_handle *handle = NULL;
+
+	mlog_entry("(0x%p, 0x%p, %u, %u)\n", file, page, from, to);
+
+	/* NOTE: ocfs2_file_aio_write has ensured that it's safe for
+	 * us to sample inode->i_size here without the metadata lock:
+	 *
+	 * 1) We're currently holding the inode alloc lock, so no
+	 *    nodes can change it underneath us.
+	 *
+	 * 2) We've had to take the metadata lock at least once
+	 *    already to check for extending writes, hence insuring
+	 *    that our current copy is also up to date.
+	 */
+	new_i_size = ((loff_t)page->index << PAGE_CACHE_SHIFT) + to;
+	if (new_i_size > i_size_read(inode)) {
+		extending = 1;
+		locklevel = 1;
+	}
+
+	ret = ocfs2_meta_lock_with_page(inode, NULL, &di_bh, locklevel, page);
+	if (ret != 0) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	ret = ocfs2_data_lock_with_page(inode, 1, page);
+	if (ret != 0) {
+		mlog_errno(ret);
+		goto out_unlock_meta;
+	}
+
+	if (extending) {
+		handle = ocfs2_start_walk_page_trans(inode, page, from, to);
+		if (IS_ERR(handle)) {
+			ret = PTR_ERR(handle);
+			handle = NULL;
+			goto out_unlock_data;
+		}
+
+		/* Mark our buffer early. We'd rather catch this error up here
+		 * as opposed to after a successful commit_write which would
+		 * require us to set back inode->i_size. */
+		ret = ocfs2_journal_access(handle, inode, di_bh,
+					   OCFS2_JOURNAL_ACCESS_WRITE);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out_commit;
+		}
+	}
+
+	/* might update i_size */
+	ret = generic_commit_write(file, page, from, to);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out_commit;
+	}
+
+	if (extending) {
+		loff_t size = (u64) i_size_read(inode);
+		struct ocfs2_dinode *di =
+			(struct ocfs2_dinode *)di_bh->b_data;
+
+		/* ocfs2_mark_inode_dirty is too heavy to use here. */
+		inode->i_blocks = ocfs2_align_bytes_to_sectors(size);
+		inode->i_ctime = inode->i_mtime = CURRENT_TIME;
+
+		di->i_size = cpu_to_le64(size);
+		di->i_ctime = di->i_mtime = 
+				cpu_to_le64(inode->i_mtime.tv_sec);
+		di->i_ctime_nsec = di->i_mtime_nsec = 
+				cpu_to_le32(inode->i_mtime.tv_nsec);
+
+		ret = ocfs2_journal_dirty(handle, di_bh);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out_commit;
+		}
+	}
+
+	BUG_ON(extending && (i_size_read(inode) != new_i_size));
+
+out_commit:
+	if (handle)
+		ocfs2_commit_trans(handle);
+out_unlock_data:
+	ocfs2_data_unlock(inode, 1);
+out_unlock_meta:
+	ocfs2_meta_unlock(inode, locklevel);
+out:
+	if (di_bh)
+		brelse(di_bh);
+
+	mlog_exit(ret);
+	return ret;
+}
+
+static sector_t ocfs2_bmap(struct address_space *mapping, sector_t block)
+{
+	sector_t status;
+	u64 p_blkno = 0;
+	int err = 0;
+	struct inode *inode = mapping->host;
+
+	mlog_entry("(block = %llu)\n", (unsigned long long)block);
+
+	/* We don't need to lock journal system files, since they aren't
+	 * accessed concurrently from multiple nodes.
+	 */
+	if (!INODE_JOURNAL(inode)) {
+		err = ocfs2_meta_lock(inode, NULL, NULL, 0);
+		if (err) {
+			if (err != -ENOENT)
+				mlog_errno(err);
+			goto bail;
+		}
+		down_read(&OCFS2_I(inode)->ip_alloc_sem);
+	}
+
+	err = ocfs2_extent_map_get_blocks(inode, block, 1, &p_blkno,
+					  NULL);
+
+	if (!INODE_JOURNAL(inode)) {
+		up_read(&OCFS2_I(inode)->ip_alloc_sem);
+		ocfs2_meta_unlock(inode, 0);
+	}
+
+	if (err) {
+		mlog(ML_ERROR, "get_blocks() failed, block = %llu\n",
+		     (unsigned long long)block);
+		mlog_errno(err);
+		goto bail;
+	}
+
+
+bail:
+	status = err ? 0 : p_blkno;
+
+	mlog_exit((int)status);
+
+	return status;
+}
+
+/*
+ * TODO: Make this into a generic get_blocks function.
+ *
+ * From do_direct_io in direct-io.c:
+ *  "So what we do is to permit the ->get_blocks function to populate
+ *   bh.b_size with the size of IO which is permitted at this offset and
+ *   this i_blkbits."
+ *
+ * This function is called directly from get_more_blocks in direct-io.c.
+ *
+ * called like this: dio->get_blocks(dio->inode, fs_startblk,
+ * 					fs_count, map_bh, dio->rw == WRITE);
+ */
+static int ocfs2_direct_IO_get_blocks(struct inode *inode, sector_t iblock,
+				     unsigned long max_blocks,
+				     struct buffer_head *bh_result, int create)
+{
+	int ret;
+	u64 vbo_max; /* file offset, max_blocks from iblock */
+	u64 p_blkno;
+	int contig_blocks;
+	unsigned char blocksize_bits;
+
+	if (!inode || !bh_result) {
+		mlog(ML_ERROR, "inode or bh_result is null\n");
+		return -EIO;
+	}
+
+	blocksize_bits = inode->i_sb->s_blocksize_bits;
+
+	/* This function won't even be called if the request isn't all
+	 * nicely aligned and of the right size, so there's no need
+	 * for us to check any of that. */
+
+	vbo_max = ((u64)iblock + max_blocks) << blocksize_bits;
+
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+	if ((iblock + max_blocks) >
+	    ocfs2_clusters_to_blocks(inode->i_sb,
+				     OCFS2_I(inode)->ip_clusters)) {
+		spin_unlock(&OCFS2_I(inode)->ip_lock);
+		ret = -EIO;
+		goto bail;
+	}
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+	/* This figures out the size of the next contiguous block, and
+	 * our logical offset */
+	ret = ocfs2_extent_map_get_blocks(inode, iblock, 1, &p_blkno,
+					  &contig_blocks);
+	if (ret) {
+		mlog(ML_ERROR, "get_blocks() failed iblock=%llu\n",
+		     (unsigned long long)iblock);
+		ret = -EIO;
+		goto bail;
+	}
+
+	map_bh(bh_result, inode->i_sb, p_blkno);
+
+	/* make sure we don't map more than max_blocks blocks here as
+	   that's all the kernel will handle at this point. */
+	if (max_blocks < contig_blocks)
+		contig_blocks = max_blocks;
+	bh_result->b_size = contig_blocks << blocksize_bits;
+bail:
+	return ret;
+}
+
+/* 
+ * ocfs2_dio_end_io is called by the dio core when a dio is finished.  We're
+ * particularly interested in the aio/dio case.  Like the core uses
+ * i_alloc_sem, we use the rw_lock DLM lock to protect io on one node from
+ * truncation on another.
+ */
+static void ocfs2_dio_end_io(struct kiocb *iocb,
+			     loff_t offset,
+			     ssize_t bytes,
+			     void *private)
+{
+	struct inode *inode = iocb->ki_filp->f_dentry->d_inode;
+
+	/* this io's submitter should not have unlocked this before we could */
+	BUG_ON(!ocfs2_iocb_is_rw_locked(iocb));
+	ocfs2_iocb_clear_rw_locked(iocb);
+	up_read(&inode->i_alloc_sem);
+	ocfs2_rw_unlock(inode, 0);
+}
+
+static ssize_t ocfs2_direct_IO(int rw,
+			       struct kiocb *iocb,
+			       const struct iovec *iov,
+			       loff_t offset,
+			       unsigned long nr_segs)
+{
+	struct file *file = iocb->ki_filp;
+	struct inode *inode = file->f_dentry->d_inode->i_mapping->host;
+	int ret;
+
+	mlog_entry_void();
+	ret = blockdev_direct_IO_no_locking(rw, iocb, inode,
+					    inode->i_sb->s_bdev, iov, offset,
+					    nr_segs, 
+					    ocfs2_direct_IO_get_blocks,
+					    ocfs2_dio_end_io);
+	mlog_exit(ret);
+	return ret;
+}
+
+struct address_space_operations ocfs2_aops = {
+	.readpage	= ocfs2_readpage,
+	.writepage	= ocfs2_writepage,
+	.prepare_write	= ocfs2_prepare_write,
+	.commit_write	= ocfs2_commit_write,
+	.bmap		= ocfs2_bmap,
+	.sync_page	= block_sync_page,
+	.direct_IO	= ocfs2_direct_IO
+};
diff --git a/fs/ocfs2/aops.h b/fs/ocfs2/aops.h
new file mode 100644
index 00000000000..d40456d509a
--- /dev/null
+++ b/fs/ocfs2/aops.h
@@ -0,0 +1,41 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2002, 2004, 2005 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_AOPS_H
+#define OCFS2_AOPS_H
+
+int ocfs2_prepare_write(struct file *file, struct page *page,
+			unsigned from, unsigned to);
+
+struct ocfs2_journal_handle *ocfs2_start_walk_page_trans(struct inode *inode,
+							 struct page *page,
+							 unsigned from,
+							 unsigned to);
+
+/* all ocfs2_dio_end_io()'s fault */
+#define ocfs2_iocb_is_rw_locked(iocb) \
+	test_bit(0, (unsigned long *)&iocb->private)
+#define ocfs2_iocb_set_rw_locked(iocb) \
+	set_bit(0, (unsigned long *)&iocb->private)
+#define ocfs2_iocb_clear_rw_locked(iocb) \
+	clear_bit(0, (unsigned long *)&iocb->private)
+
+#endif /* OCFS2_FILE_H */
diff --git a/fs/ocfs2/buffer_head_io.c b/fs/ocfs2/buffer_head_io.c
new file mode 100644
index 00000000000..d424041b38e
--- /dev/null
+++ b/fs/ocfs2/buffer_head_io.c
@@ -0,0 +1,232 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * io.c
+ *
+ * Buffer cache handling
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "inode.h"
+#include "journal.h"
+#include "uptodate.h"
+
+#include "buffer_head_io.h"
+
+int ocfs2_write_block(struct ocfs2_super *osb, struct buffer_head *bh,
+		      struct inode *inode)
+{
+	int ret = 0;
+
+	mlog_entry("(bh->b_blocknr = %llu, inode=%p)\n",
+		   (unsigned long long)bh->b_blocknr, inode);
+
+	BUG_ON(bh->b_blocknr < OCFS2_SUPER_BLOCK_BLKNO);
+	BUG_ON(buffer_jbd(bh));
+
+	/* No need to check for a soft readonly file system here. non
+	 * journalled writes are only ever done on system files which
+	 * can get modified during recovery even if read-only. */
+	if (ocfs2_is_hard_readonly(osb)) {
+		ret = -EROFS;
+		goto out;
+	}
+
+	down(&OCFS2_I(inode)->ip_io_sem);
+
+	lock_buffer(bh);
+	set_buffer_uptodate(bh);
+
+	/* remove from dirty list before I/O. */
+	clear_buffer_dirty(bh);
+
+	get_bh(bh); /* for end_buffer_write_sync() */                   
+	bh->b_end_io = end_buffer_write_sync;
+	submit_bh(WRITE, bh);
+
+	wait_on_buffer(bh);
+
+	if (buffer_uptodate(bh)) {
+		ocfs2_set_buffer_uptodate(inode, bh);
+	} else {
+		/* We don't need to remove the clustered uptodate
+		 * information for this bh as it's not marked locally
+		 * uptodate. */
+		ret = -EIO;
+		brelse(bh);
+	}
+
+	up(&OCFS2_I(inode)->ip_io_sem);
+out:
+	mlog_exit(ret);
+	return ret;
+}
+
+int ocfs2_read_blocks(struct ocfs2_super *osb, u64 block, int nr,
+		      struct buffer_head *bhs[], int flags,
+		      struct inode *inode)
+{
+	int status = 0;
+	struct super_block *sb;
+	int i, ignore_cache = 0;
+	struct buffer_head *bh;
+
+	mlog_entry("(block=(%"MLFu64"), nr=(%d), flags=%d, inode=%p)\n",
+		   block, nr, flags, inode);
+
+	if (osb == NULL || osb->sb == NULL || bhs == NULL) {
+		status = -EINVAL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	if (nr < 0) {
+		mlog(ML_ERROR, "asked to read %d blocks!\n", nr);
+		status = -EINVAL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	if (nr == 0) {
+		mlog(ML_BH_IO, "No buffers will be read!\n");
+		status = 0;
+		goto bail;
+	}
+
+	sb = osb->sb;
+
+	if (flags & OCFS2_BH_CACHED && !inode)
+		flags &= ~OCFS2_BH_CACHED;
+
+	if (inode)
+		down(&OCFS2_I(inode)->ip_io_sem);
+	for (i = 0 ; i < nr ; i++) {
+		if (bhs[i] == NULL) {
+			bhs[i] = sb_getblk(sb, block++);
+			if (bhs[i] == NULL) {
+				if (inode)
+					up(&OCFS2_I(inode)->ip_io_sem);
+				status = -EIO;
+				mlog_errno(status);
+				goto bail;
+			}
+		}
+		bh = bhs[i];
+		ignore_cache = 0;
+
+		if (flags & OCFS2_BH_CACHED &&
+		    !ocfs2_buffer_uptodate(inode, bh)) {
+			mlog(ML_UPTODATE,
+			     "bh (%llu), inode %"MLFu64" not uptodate\n",
+			     (unsigned long long)bh->b_blocknr,
+			     OCFS2_I(inode)->ip_blkno);
+			ignore_cache = 1;
+		}
+
+		/* XXX: Can we ever get this and *not* have the cached
+		 * flag set? */
+		if (buffer_jbd(bh)) {
+			if (!(flags & OCFS2_BH_CACHED) || ignore_cache)
+				mlog(ML_BH_IO, "trying to sync read a jbd "
+					       "managed bh (blocknr = %llu)\n",
+				     (unsigned long long)bh->b_blocknr);
+			continue;
+		}
+
+		if (!(flags & OCFS2_BH_CACHED) || ignore_cache) {
+			if (buffer_dirty(bh)) {
+				/* This should probably be a BUG, or
+				 * at least return an error. */
+				mlog(ML_BH_IO, "asking me to sync read a dirty "
+					       "buffer! (blocknr = %llu)\n",
+				     (unsigned long long)bh->b_blocknr);
+				continue;
+			}
+
+			lock_buffer(bh);
+			if (buffer_jbd(bh)) {
+#ifdef CATCH_BH_JBD_RACES
+				mlog(ML_ERROR, "block %llu had the JBD bit set "
+					       "while I was in lock_buffer!",
+				     (unsigned long long)bh->b_blocknr);
+				BUG();
+#else
+				unlock_buffer(bh);
+				continue;
+#endif
+			}
+			clear_buffer_uptodate(bh);
+			get_bh(bh); /* for end_buffer_read_sync() */
+			bh->b_end_io = end_buffer_read_sync;
+			if (flags & OCFS2_BH_READAHEAD)
+				submit_bh(READA, bh);
+			else
+				submit_bh(READ, bh);
+			continue;
+		}
+	}
+
+	status = 0;
+
+	for (i = (nr - 1); i >= 0; i--) {
+		bh = bhs[i];
+
+		/* We know this can't have changed as we hold the
+		 * inode sem. Avoid doing any work on the bh if the
+		 * journal has it. */
+		if (!buffer_jbd(bh))
+			wait_on_buffer(bh);
+
+		if (!buffer_uptodate(bh)) {
+			/* Status won't be cleared from here on out,
+			 * so we can safely record this and loop back
+			 * to cleanup the other buffers. Don't need to
+			 * remove the clustered uptodate information
+			 * for this bh as it's not marked locally
+			 * uptodate. */
+			status = -EIO;
+			brelse(bh);
+			bhs[i] = NULL;
+			continue;
+		}
+
+		if (inode)
+			ocfs2_set_buffer_uptodate(inode, bh);
+	}
+	if (inode)
+		up(&OCFS2_I(inode)->ip_io_sem);
+
+	mlog(ML_BH_IO, "block=(%"MLFu64"), nr=(%d), cached=%s\n", block, nr,
+	     (!(flags & OCFS2_BH_CACHED) || ignore_cache) ? "no" : "yes");
+
+bail:
+
+	mlog_exit(status);
+	return status;
+}
diff --git a/fs/ocfs2/buffer_head_io.h b/fs/ocfs2/buffer_head_io.h
new file mode 100644
index 00000000000..6ecb90937b6
--- /dev/null
+++ b/fs/ocfs2/buffer_head_io.h
@@ -0,0 +1,73 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs2_buffer_head.h
+ *
+ * Buffer cache handling functions defined
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_BUFFER_HEAD_IO_H
+#define OCFS2_BUFFER_HEAD_IO_H
+
+#include <linux/buffer_head.h>
+
+void ocfs2_end_buffer_io_sync(struct buffer_head *bh,
+			     int uptodate);
+
+static inline int ocfs2_read_block(struct ocfs2_super          *osb,
+				   u64                  off,
+				   struct buffer_head **bh,
+				   int                  flags,
+				   struct inode        *inode);
+
+int ocfs2_write_block(struct ocfs2_super          *osb,
+		      struct buffer_head  *bh,
+		      struct inode        *inode);
+int ocfs2_read_blocks(struct ocfs2_super          *osb,
+		      u64                  block,
+		      int                  nr,
+		      struct buffer_head  *bhs[],
+		      int                  flags,
+		      struct inode        *inode);
+
+
+#define OCFS2_BH_CACHED            1
+#define OCFS2_BH_READAHEAD         8	/* use this to pass READA down to submit_bh */
+
+static inline int ocfs2_read_block(struct ocfs2_super * osb, u64 off,
+				   struct buffer_head **bh, int flags,
+				   struct inode *inode)
+{
+	int status = 0;
+
+	if (bh == NULL) {
+		printk("ocfs2: bh == NULL\n");
+		status = -EINVAL;
+		goto bail;
+	}
+
+	status = ocfs2_read_blocks(osb, off, 1, bh,
+				   flags, inode);
+
+bail:
+	return status;
+}
+
+#endif /* OCFS2_BUFFER_HEAD_IO_H */
diff --git a/fs/ocfs2/dcache.c b/fs/ocfs2/dcache.c
new file mode 100644
index 00000000000..bd85182e97b
--- /dev/null
+++ b/fs/ocfs2/dcache.c
@@ -0,0 +1,91 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dcache.c
+ *
+ * dentry cache handling code
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/namei.h>
+
+#define MLOG_MASK_PREFIX ML_DCACHE
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "dcache.h"
+#include "file.h"
+#include "inode.h"
+
+static int ocfs2_dentry_revalidate(struct dentry *dentry,
+				   struct nameidata *nd)
+{
+	struct inode *inode = dentry->d_inode;
+	int ret = 0;    /* if all else fails, just return false */
+	struct ocfs2_super *osb;
+
+	mlog_entry("(0x%p, '%.*s')\n", dentry,
+		   dentry->d_name.len, dentry->d_name.name);
+
+	/* Never trust a negative dentry - force a new lookup. */
+	if (inode == NULL) {
+		mlog(0, "negative dentry: %.*s\n", dentry->d_name.len,
+		     dentry->d_name.name);
+		goto bail;
+	}
+
+	osb = OCFS2_SB(inode->i_sb);
+
+	BUG_ON(!osb);
+
+	if (inode != osb->root_inode) {
+		spin_lock(&OCFS2_I(inode)->ip_lock);
+		/* did we or someone else delete this inode? */
+		if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_DELETED) {
+			spin_unlock(&OCFS2_I(inode)->ip_lock);
+			mlog(0, "inode (%"MLFu64") deleted, returning false\n",
+			     OCFS2_I(inode)->ip_blkno);
+			goto bail;
+		}
+		spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+		if (!inode->i_nlink) {
+			mlog(0, "Inode %"MLFu64" orphaned, returning false "
+			     "dir = %d\n", OCFS2_I(inode)->ip_blkno,
+			     S_ISDIR(inode->i_mode));
+			goto bail;
+		}
+	}
+
+	ret = 1;
+
+bail:
+	mlog_exit(ret);
+
+	return ret;
+}
+
+struct dentry_operations ocfs2_dentry_ops = {
+	.d_revalidate		= ocfs2_dentry_revalidate,
+};
diff --git a/fs/ocfs2/dcache.h b/fs/ocfs2/dcache.h
new file mode 100644
index 00000000000..90072771114
--- /dev/null
+++ b/fs/ocfs2/dcache.h
@@ -0,0 +1,31 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dcache.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_DCACHE_H
+#define OCFS2_DCACHE_H
+
+extern struct dentry_operations ocfs2_dentry_ops;
+
+#endif /* OCFS2_DCACHE_H */
diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c
new file mode 100644
index 00000000000..856e20ae826
--- /dev/null
+++ b/fs/ocfs2/dir.c
@@ -0,0 +1,618 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dir.c
+ *
+ * Creates, reads, walks and deletes directory-nodes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ *  Portions of this code from linux/fs/ext3/dir.c
+ *
+ *  Copyright (C) 1992, 1993, 1994, 1995
+ *  Remy Card (card@masi.ibp.fr)
+ *  Laboratoire MASI - Institut Blaise pascal
+ *  Universite Pierre et Marie Curie (Paris VI)
+ *
+ *   from
+ *
+ *   linux/fs/minix/dir.c
+ *
+ *   Copyright (C) 1991, 1992 Linux Torvalds
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+
+#define MLOG_MASK_PREFIX ML_NAMEI
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "dir.h"
+#include "dlmglue.h"
+#include "extent_map.h"
+#include "file.h"
+#include "inode.h"
+#include "journal.h"
+#include "namei.h"
+#include "suballoc.h"
+#include "uptodate.h"
+
+#include "buffer_head_io.h"
+
+static unsigned char ocfs2_filetype_table[] = {
+	DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
+};
+
+static int ocfs2_extend_dir(struct ocfs2_super *osb,
+			    struct inode *dir,
+			    struct buffer_head *parent_fe_bh,
+			    struct buffer_head **new_de_bh);
+/*
+ * ocfs2_readdir()
+ *
+ */
+int ocfs2_readdir(struct file * filp, void * dirent, filldir_t filldir)
+{
+	int error = 0;
+	unsigned long offset, blk;
+	int i, num, stored;
+	struct buffer_head * bh, * tmp;
+	struct ocfs2_dir_entry * de;
+	int err;
+	struct inode *inode = filp->f_dentry->d_inode;
+	struct super_block * sb = inode->i_sb;
+	int have_disk_lock = 0;
+
+	mlog_entry("dirino=%"MLFu64"\n", OCFS2_I(inode)->ip_blkno);
+
+	stored = 0;
+	bh = NULL;
+
+	error = ocfs2_meta_lock(inode, NULL, NULL, 0);
+	if (error < 0) {
+		if (error != -ENOENT)
+			mlog_errno(error);
+		/* we haven't got any yet, so propagate the error. */
+		stored = error;
+		goto bail;
+	}
+	have_disk_lock = 1;
+
+	offset = filp->f_pos & (sb->s_blocksize - 1);
+
+	while (!error && !stored && filp->f_pos < i_size_read(inode)) {
+		blk = (filp->f_pos) >> sb->s_blocksize_bits;
+		bh = ocfs2_bread(inode, blk, &err, 0);
+		if (!bh) {
+			mlog(ML_ERROR, "directory #%"MLFu64" contains a hole "
+				       "at offset %lld\n",
+			     OCFS2_I(inode)->ip_blkno,
+			     filp->f_pos);
+			filp->f_pos += sb->s_blocksize - offset;
+			continue;
+		}
+
+		/*
+		 * Do the readahead (8k)
+		 */
+		if (!offset) {
+			for (i = 16 >> (sb->s_blocksize_bits - 9), num = 0;
+			     i > 0; i--) {
+				tmp = ocfs2_bread(inode, ++blk, &err, 1);
+				if (tmp)
+					brelse(tmp);
+			}
+		}
+
+revalidate:
+		/* If the dir block has changed since the last call to
+		 * readdir(2), then we might be pointing to an invalid
+		 * dirent right now.  Scan from the start of the block
+		 * to make sure. */
+		if (filp->f_version != inode->i_version) {
+			for (i = 0; i < sb->s_blocksize && i < offset; ) {
+				de = (struct ocfs2_dir_entry *) (bh->b_data + i);
+				/* It's too expensive to do a full
+				 * dirent test each time round this
+				 * loop, but we do have to test at
+				 * least that it is non-zero.  A
+				 * failure will be detected in the
+				 * dirent test below. */
+				if (le16_to_cpu(de->rec_len) <
+				    OCFS2_DIR_REC_LEN(1))
+					break;
+				i += le16_to_cpu(de->rec_len);
+			}
+			offset = i;
+			filp->f_pos = (filp->f_pos & ~(sb->s_blocksize - 1))
+				| offset;
+			filp->f_version = inode->i_version;
+		}
+
+		while (!error && filp->f_pos < i_size_read(inode)
+		       && offset < sb->s_blocksize) {
+			de = (struct ocfs2_dir_entry *) (bh->b_data + offset);
+			if (!ocfs2_check_dir_entry(inode, de, bh, offset)) {
+				/* On error, skip the f_pos to the
+				   next block. */
+				filp->f_pos = (filp->f_pos |
+					       (sb->s_blocksize - 1)) + 1;
+				brelse(bh);
+				goto bail;
+			}
+			offset += le16_to_cpu(de->rec_len);
+			if (le64_to_cpu(de->inode)) {
+				/* We might block in the next section
+				 * if the data destination is
+				 * currently swapped out.  So, use a
+				 * version stamp to detect whether or
+				 * not the directory has been modified
+				 * during the copy operation.
+				 */
+				unsigned long version = filp->f_version;
+				unsigned char d_type = DT_UNKNOWN;
+
+				if (de->file_type < OCFS2_FT_MAX)
+					d_type = ocfs2_filetype_table[de->file_type];
+				error = filldir(dirent, de->name,
+						de->name_len,
+						filp->f_pos,
+						ino_from_blkno(sb, le64_to_cpu(de->inode)),
+						d_type);
+				if (error)
+					break;
+				if (version != filp->f_version)
+					goto revalidate;
+				stored ++;
+			}
+			filp->f_pos += le16_to_cpu(de->rec_len);
+		}
+		offset = 0;
+		brelse(bh);
+	}
+
+	stored = 0;
+bail:
+	if (have_disk_lock)
+		ocfs2_meta_unlock(inode, 0);
+
+	mlog_exit(stored);
+
+	return stored;
+}
+
+/*
+ * NOTE: this should always be called with parent dir i_sem taken.
+ */
+int ocfs2_find_files_on_disk(const char *name,
+			     int namelen,
+			     u64 *blkno,
+			     struct inode *inode,
+			     struct buffer_head **dirent_bh,
+			     struct ocfs2_dir_entry **dirent)
+{
+	int status = -ENOENT;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	mlog_entry("(osb=%p, parent=%"MLFu64", name='%.*s', blkno=%p, "
+		   "inode=%p)\n",
+		   osb, OCFS2_I(inode)->ip_blkno, namelen, name, blkno, inode);
+
+	*dirent_bh = ocfs2_find_entry(name, namelen, inode, dirent);
+	if (!*dirent_bh || !*dirent) {
+		status = -ENOENT;
+		goto leave;
+	}
+
+	*blkno = le64_to_cpu((*dirent)->inode);
+
+	status = 0;
+leave:
+	if (status < 0) {
+		*dirent = NULL;
+		if (*dirent_bh) {
+			brelse(*dirent_bh);
+			*dirent_bh = NULL;
+		}
+	}
+
+	mlog_exit(status);
+	return status;
+}
+
+/* Check for a name within a directory.
+ *
+ * Return 0 if the name does not exist
+ * Return -EEXIST if the directory contains the name
+ *
+ * Callers should have i_sem + a cluster lock on dir
+ */
+int ocfs2_check_dir_for_entry(struct inode *dir,
+			      const char *name,
+			      int namelen)
+{
+	int ret;
+	struct buffer_head *dirent_bh = NULL;
+	struct ocfs2_dir_entry *dirent = NULL;
+
+	mlog_entry("dir %"MLFu64", name '%.*s'\n", OCFS2_I(dir)->ip_blkno,
+		   namelen, name);
+
+	ret = -EEXIST;
+	dirent_bh = ocfs2_find_entry(name, namelen, dir, &dirent);
+	if (dirent_bh)
+		goto bail;
+
+	ret = 0;
+bail:
+	if (dirent_bh)
+		brelse(dirent_bh);
+
+	mlog_exit(ret);
+	return ret;
+}
+
+/*
+ * routine to check that the specified directory is empty (for rmdir)
+ */
+int ocfs2_empty_dir(struct inode *inode)
+{
+	unsigned long offset;
+	struct buffer_head * bh;
+	struct ocfs2_dir_entry * de, * de1;
+	struct super_block * sb;
+	int err;
+
+	sb = inode->i_sb;
+	if ((i_size_read(inode) <
+	     (OCFS2_DIR_REC_LEN(1) + OCFS2_DIR_REC_LEN(2))) ||
+	    !(bh = ocfs2_bread(inode, 0, &err, 0))) {
+	    	mlog(ML_ERROR, "bad directory (dir #%"MLFu64") - "
+			       "no data block\n",
+		     OCFS2_I(inode)->ip_blkno);
+		return 1;
+	}
+
+	de = (struct ocfs2_dir_entry *) bh->b_data;
+	de1 = (struct ocfs2_dir_entry *)
+			((char *)de + le16_to_cpu(de->rec_len));
+	if ((le64_to_cpu(de->inode) != OCFS2_I(inode)->ip_blkno) ||
+			!le64_to_cpu(de1->inode) ||
+			strcmp(".", de->name) ||
+			strcmp("..", de1->name)) {
+	    	mlog(ML_ERROR, "bad directory (dir #%"MLFu64") - "
+			       "no `.' or `..'\n",
+		     OCFS2_I(inode)->ip_blkno);
+		brelse(bh);
+		return 1;
+	}
+	offset = le16_to_cpu(de->rec_len) + le16_to_cpu(de1->rec_len);
+	de = (struct ocfs2_dir_entry *)((char *)de1 + le16_to_cpu(de1->rec_len));
+	while (offset < i_size_read(inode) ) {
+		if (!bh || (void *)de >= (void *)(bh->b_data + sb->s_blocksize)) {
+			brelse(bh);
+			bh = ocfs2_bread(inode,
+					 offset >> sb->s_blocksize_bits, &err, 0);
+			if (!bh) {
+				mlog(ML_ERROR, "directory #%"MLFu64" contains "
+					       "a hole at offset %lu\n",
+				     OCFS2_I(inode)->ip_blkno, offset);
+				offset += sb->s_blocksize;
+				continue;
+			}
+			de = (struct ocfs2_dir_entry *) bh->b_data;
+		}
+		if (!ocfs2_check_dir_entry(inode, de, bh, offset)) {
+			brelse(bh);
+			return 1;
+		}
+		if (le64_to_cpu(de->inode)) {
+			brelse(bh);
+			return 0;
+		}
+		offset += le16_to_cpu(de->rec_len);
+		de = (struct ocfs2_dir_entry *)
+			((char *)de + le16_to_cpu(de->rec_len));
+	}
+	brelse(bh);
+	return 1;
+}
+
+/* returns a bh of the 1st new block in the allocation. */
+int ocfs2_do_extend_dir(struct super_block *sb,
+			struct ocfs2_journal_handle *handle,
+			struct inode *dir,
+			struct buffer_head *parent_fe_bh,
+			struct ocfs2_alloc_context *data_ac,
+			struct ocfs2_alloc_context *meta_ac,
+			struct buffer_head **new_bh)
+{
+	int status;
+	int extend;
+	u64 p_blkno;
+
+	spin_lock(&OCFS2_I(dir)->ip_lock);
+	extend = (i_size_read(dir) == ocfs2_clusters_to_bytes(sb, OCFS2_I(dir)->ip_clusters));
+	spin_unlock(&OCFS2_I(dir)->ip_lock);
+
+	if (extend) {
+		status = ocfs2_do_extend_allocation(OCFS2_SB(sb), dir, 1,
+						    parent_fe_bh, handle,
+						    data_ac, meta_ac, NULL);
+		BUG_ON(status == -EAGAIN);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+	}
+
+	status = ocfs2_extent_map_get_blocks(dir, (dir->i_blocks >>
+						   (sb->s_blocksize_bits - 9)),
+					     1, &p_blkno, NULL);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	*new_bh = sb_getblk(sb, p_blkno);
+	if (!*new_bh) {
+		status = -EIO;
+		mlog_errno(status);
+		goto bail;
+	}
+	status = 0;
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+/* assumes you already have a cluster lock on the directory. */
+static int ocfs2_extend_dir(struct ocfs2_super *osb,
+			    struct inode *dir,
+			    struct buffer_head *parent_fe_bh,
+			    struct buffer_head **new_de_bh)
+{
+	int status = 0;
+	int credits, num_free_extents;
+	loff_t dir_i_size;
+	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) parent_fe_bh->b_data;
+	struct ocfs2_alloc_context *data_ac = NULL;
+	struct ocfs2_alloc_context *meta_ac = NULL;
+	struct ocfs2_journal_handle *handle = NULL;
+	struct buffer_head *new_bh = NULL;
+	struct ocfs2_dir_entry * de;
+	struct super_block *sb = osb->sb;
+
+	mlog_entry_void();
+
+	dir_i_size = i_size_read(dir);
+	mlog(0, "extending dir %"MLFu64" (i_size = %lld)\n",
+	     OCFS2_I(dir)->ip_blkno, dir_i_size);
+
+	handle = ocfs2_alloc_handle(osb);
+	if (handle == NULL) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* dir->i_size is always block aligned. */
+	spin_lock(&OCFS2_I(dir)->ip_lock);
+	if (dir_i_size == ocfs2_clusters_to_bytes(sb, OCFS2_I(dir)->ip_clusters)) {
+		spin_unlock(&OCFS2_I(dir)->ip_lock);
+		num_free_extents = ocfs2_num_free_extents(osb, dir, fe);
+		if (num_free_extents < 0) {
+			status = num_free_extents;
+			mlog_errno(status);
+			goto bail;
+		}
+
+		if (!num_free_extents) {
+			status = ocfs2_reserve_new_metadata(osb, handle,
+							    fe, &meta_ac);
+			if (status < 0) {
+				if (status != -ENOSPC)
+					mlog_errno(status);
+				goto bail;
+			}
+		}
+
+		status = ocfs2_reserve_clusters(osb, handle, 1, &data_ac);
+		if (status < 0) {
+			if (status != -ENOSPC)
+				mlog_errno(status);
+			goto bail;
+		}
+
+		credits = ocfs2_calc_extend_credits(sb, fe, 1);
+	} else {
+		spin_unlock(&OCFS2_I(dir)->ip_lock);
+		credits = OCFS2_SIMPLE_DIR_EXTEND_CREDITS;
+	}
+
+	handle = ocfs2_start_trans(osb, handle, credits);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		handle = NULL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_do_extend_dir(osb->sb, handle, dir, parent_fe_bh,
+				     data_ac, meta_ac, &new_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	ocfs2_set_new_buffer_uptodate(dir, new_bh);
+
+	status = ocfs2_journal_access(handle, dir, new_bh,
+				      OCFS2_JOURNAL_ACCESS_CREATE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+	memset(new_bh->b_data, 0, sb->s_blocksize);
+	de = (struct ocfs2_dir_entry *) new_bh->b_data;
+	de->inode = 0;
+	de->rec_len = cpu_to_le16(sb->s_blocksize);
+	status = ocfs2_journal_dirty(handle, new_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	dir_i_size += dir->i_sb->s_blocksize;
+	i_size_write(dir, dir_i_size);
+	dir->i_blocks = ocfs2_align_bytes_to_sectors(dir_i_size);
+	status = ocfs2_mark_inode_dirty(handle, dir, parent_fe_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	*new_de_bh = new_bh;
+	get_bh(*new_de_bh);
+bail:
+	if (handle)
+		ocfs2_commit_trans(handle);
+
+	if (data_ac)
+		ocfs2_free_alloc_context(data_ac);
+	if (meta_ac)
+		ocfs2_free_alloc_context(meta_ac);
+
+	if (new_bh)
+		brelse(new_bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * Search the dir for a good spot, extending it if necessary. The
+ * block containing an appropriate record is returned in ret_de_bh.
+ */
+int ocfs2_prepare_dir_for_insert(struct ocfs2_super *osb,
+				 struct inode *dir,
+				 struct buffer_head *parent_fe_bh,
+				 const char *name,
+				 int namelen,
+				 struct buffer_head **ret_de_bh)
+{
+	unsigned long offset;
+	struct buffer_head * bh = NULL;
+	unsigned short rec_len;
+	struct ocfs2_dinode *fe;
+	struct ocfs2_dir_entry *de;
+	struct super_block *sb;
+	int status;
+
+	mlog_entry_void();
+
+	mlog(0, "getting ready to insert namelen %d into dir %"MLFu64"\n",
+	     namelen, OCFS2_I(dir)->ip_blkno);
+
+	BUG_ON(!S_ISDIR(dir->i_mode));
+	fe = (struct ocfs2_dinode *) parent_fe_bh->b_data;
+	BUG_ON(le64_to_cpu(fe->i_size) != i_size_read(dir));
+
+	sb = dir->i_sb;
+
+	if (!namelen) {
+		status = -EINVAL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	bh = ocfs2_bread(dir, 0, &status, 0);
+	if (!bh) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	rec_len = OCFS2_DIR_REC_LEN(namelen);
+	offset = 0;
+	de = (struct ocfs2_dir_entry *) bh->b_data;
+	while (1) {
+		if ((char *)de >= sb->s_blocksize + bh->b_data) {
+			brelse(bh);
+			bh = NULL;
+
+			if (i_size_read(dir) <= offset) {
+				status = ocfs2_extend_dir(osb,
+							  dir,
+							  parent_fe_bh,
+							  &bh);
+				if (status < 0) {
+					mlog_errno(status);
+					goto bail;
+				}
+				BUG_ON(!bh);
+				*ret_de_bh = bh;
+				get_bh(*ret_de_bh);
+				goto bail;
+			}
+			bh = ocfs2_bread(dir,
+					 offset >> sb->s_blocksize_bits,
+					 &status,
+					 0);
+			if (!bh) {
+				mlog_errno(status);
+				goto bail;
+			}
+			/* move to next block */
+			de = (struct ocfs2_dir_entry *) bh->b_data;
+		}
+		if (!ocfs2_check_dir_entry(dir, de, bh, offset)) {
+			status = -ENOENT;
+			goto bail;
+		}
+		if (ocfs2_match(namelen, name, de)) {
+			status = -EEXIST;
+			goto bail;
+		}
+		if (((le64_to_cpu(de->inode) == 0) &&
+		     (le16_to_cpu(de->rec_len) >= rec_len)) ||
+		    (le16_to_cpu(de->rec_len) >=
+		     (OCFS2_DIR_REC_LEN(de->name_len) + rec_len))) {
+			/* Ok, we found a spot. Return this bh and let
+			 * the caller actually fill it in. */
+			*ret_de_bh = bh;
+			get_bh(*ret_de_bh);
+			status = 0;
+			goto bail;
+		}
+		offset += le16_to_cpu(de->rec_len);
+		de = (struct ocfs2_dir_entry *)((char *) de + le16_to_cpu(de->rec_len));
+	}
+
+	status = 0;
+bail:
+	if (bh)
+		brelse(bh);
+
+	mlog_exit(status);
+	return status;
+}
diff --git a/fs/ocfs2/dir.h b/fs/ocfs2/dir.h
new file mode 100644
index 00000000000..5f614ec9649
--- /dev/null
+++ b/fs/ocfs2/dir.h
@@ -0,0 +1,54 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dir.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_DIR_H
+#define OCFS2_DIR_H
+
+int ocfs2_check_dir_for_entry(struct inode *dir,
+			      const char *name,
+			      int namelen);
+int ocfs2_empty_dir(struct inode *inode);  /* FIXME: to namei.c */
+int ocfs2_find_files_on_disk(const char *name,
+			     int namelen,
+			     u64 *blkno,
+			     struct inode *inode,
+			     struct buffer_head **dirent_bh,
+			     struct ocfs2_dir_entry **dirent);
+int ocfs2_readdir(struct file *filp, void *dirent, filldir_t filldir);
+int ocfs2_prepare_dir_for_insert(struct ocfs2_super *osb,
+				 struct inode *dir,
+				 struct buffer_head *parent_fe_bh,
+				 const char *name,
+				 int namelen,
+				 struct buffer_head **ret_de_bh);
+struct ocfs2_alloc_context;
+int ocfs2_do_extend_dir(struct super_block *sb,
+			struct ocfs2_journal_handle *handle,
+			struct inode *dir,
+			struct buffer_head *parent_fe_bh,
+			struct ocfs2_alloc_context *data_ac,
+			struct ocfs2_alloc_context *meta_ac,
+			struct buffer_head **new_bh);
+#endif /* OCFS2_DIR_H */
diff --git a/fs/ocfs2/dlmglue.c b/fs/ocfs2/dlmglue.c
new file mode 100644
index 00000000000..e971ec2f840
--- /dev/null
+++ b/fs/ocfs2/dlmglue.c
@@ -0,0 +1,2904 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dlmglue.c
+ *
+ * Code which implements an OCFS2 specific interface to our DLM.
+ *
+ * Copyright (C) 2003, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/mm.h>
+#include <linux/smp_lock.h>
+#include <linux/crc32.h>
+#include <linux/kthread.h>
+#include <linux/pagemap.h>
+#include <linux/debugfs.h>
+#include <linux/seq_file.h>
+
+#include <cluster/heartbeat.h>
+#include <cluster/nodemanager.h>
+#include <cluster/tcp.h>
+
+#include <dlm/dlmapi.h>
+
+#define MLOG_MASK_PREFIX ML_DLM_GLUE
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "dlmglue.h"
+#include "extent_map.h"
+#include "heartbeat.h"
+#include "inode.h"
+#include "journal.h"
+#include "slot_map.h"
+#include "super.h"
+#include "uptodate.h"
+#include "vote.h"
+
+#include "buffer_head_io.h"
+
+struct ocfs2_mask_waiter {
+	struct list_head	mw_item;
+	int			mw_status;
+	struct completion	mw_complete;
+	unsigned long		mw_mask;
+	unsigned long		mw_goal;
+};
+
+static void ocfs2_inode_ast_func(void *opaque);
+static void ocfs2_inode_bast_func(void *opaque,
+				  int level);
+static void ocfs2_super_ast_func(void *opaque);
+static void ocfs2_super_bast_func(void *opaque,
+				  int level);
+static void ocfs2_rename_ast_func(void *opaque);
+static void ocfs2_rename_bast_func(void *opaque,
+				   int level);
+
+/* so far, all locks have gotten along with the same unlock ast */
+static void ocfs2_unlock_ast_func(void *opaque,
+				  enum dlm_status status);
+static int ocfs2_do_unblock_meta(struct inode *inode,
+				 int *requeue);
+static int ocfs2_unblock_meta(struct ocfs2_lock_res *lockres,
+			      int *requeue);
+static int ocfs2_unblock_data(struct ocfs2_lock_res *lockres,
+			      int *requeue);
+static int ocfs2_unblock_inode_lock(struct ocfs2_lock_res *lockres,
+			      int *requeue);
+static int ocfs2_unblock_osb_lock(struct ocfs2_lock_res *lockres,
+				  int *requeue);
+typedef void (ocfs2_convert_worker_t)(struct ocfs2_lock_res *, int);
+static int ocfs2_generic_unblock_lock(struct ocfs2_super *osb,
+				      struct ocfs2_lock_res *lockres,
+				      int *requeue,
+				      ocfs2_convert_worker_t *worker);
+
+struct ocfs2_lock_res_ops {
+	void (*ast)(void *);
+	void (*bast)(void *, int);
+	void (*unlock_ast)(void *, enum dlm_status);
+	int  (*unblock)(struct ocfs2_lock_res *, int *);
+};
+
+static struct ocfs2_lock_res_ops ocfs2_inode_rw_lops = {
+	.ast		= ocfs2_inode_ast_func,
+	.bast		= ocfs2_inode_bast_func,
+	.unlock_ast	= ocfs2_unlock_ast_func,
+	.unblock	= ocfs2_unblock_inode_lock,
+};
+
+static struct ocfs2_lock_res_ops ocfs2_inode_meta_lops = {
+	.ast		= ocfs2_inode_ast_func,
+	.bast		= ocfs2_inode_bast_func,
+	.unlock_ast	= ocfs2_unlock_ast_func,
+	.unblock	= ocfs2_unblock_meta,
+};
+
+static void ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres,
+				      int blocking);
+
+static struct ocfs2_lock_res_ops ocfs2_inode_data_lops = {
+	.ast		= ocfs2_inode_ast_func,
+	.bast		= ocfs2_inode_bast_func,
+	.unlock_ast	= ocfs2_unlock_ast_func,
+	.unblock	= ocfs2_unblock_data,
+};
+
+static struct ocfs2_lock_res_ops ocfs2_super_lops = {
+	.ast		= ocfs2_super_ast_func,
+	.bast		= ocfs2_super_bast_func,
+	.unlock_ast	= ocfs2_unlock_ast_func,
+	.unblock	= ocfs2_unblock_osb_lock,
+};
+
+static struct ocfs2_lock_res_ops ocfs2_rename_lops = {
+	.ast		= ocfs2_rename_ast_func,
+	.bast		= ocfs2_rename_bast_func,
+	.unlock_ast	= ocfs2_unlock_ast_func,
+	.unblock	= ocfs2_unblock_osb_lock,
+};
+
+static inline int ocfs2_is_inode_lock(struct ocfs2_lock_res *lockres)
+{
+	return lockres->l_type == OCFS2_LOCK_TYPE_META ||
+		lockres->l_type == OCFS2_LOCK_TYPE_DATA ||
+		lockres->l_type == OCFS2_LOCK_TYPE_RW;
+}
+
+static inline int ocfs2_is_super_lock(struct ocfs2_lock_res *lockres)
+{
+	return lockres->l_type == OCFS2_LOCK_TYPE_SUPER;
+}
+
+static inline int ocfs2_is_rename_lock(struct ocfs2_lock_res *lockres)
+{
+	return lockres->l_type == OCFS2_LOCK_TYPE_RENAME;
+}
+
+static inline struct ocfs2_super *ocfs2_lock_res_super(struct ocfs2_lock_res *lockres)
+{
+	BUG_ON(!ocfs2_is_super_lock(lockres)
+	       && !ocfs2_is_rename_lock(lockres));
+
+	return (struct ocfs2_super *) lockres->l_priv;
+}
+
+static inline struct inode *ocfs2_lock_res_inode(struct ocfs2_lock_res *lockres)
+{
+	BUG_ON(!ocfs2_is_inode_lock(lockres));
+
+	return (struct inode *) lockres->l_priv;
+}
+
+static int ocfs2_lock_create(struct ocfs2_super *osb,
+			     struct ocfs2_lock_res *lockres,
+			     int level,
+			     int dlm_flags);
+static inline int ocfs2_may_continue_on_blocked_lock(struct ocfs2_lock_res *lockres,
+						     int wanted);
+static void ocfs2_cluster_unlock(struct ocfs2_super *osb,
+				 struct ocfs2_lock_res *lockres,
+				 int level);
+static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres);
+static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres);
+static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres);
+static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres, int level);
+static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb,
+					struct ocfs2_lock_res *lockres);
+static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres,
+						int convert);
+#define ocfs2_log_dlm_error(_func, _stat, _lockres) do {	\
+	mlog(ML_ERROR, "Dlm error \"%s\" while calling %s on "	\
+		"resource %s: %s\n", dlm_errname(_stat), _func,	\
+		_lockres->l_name, dlm_errmsg(_stat));		\
+} while (0)
+static void ocfs2_vote_on_unlock(struct ocfs2_super *osb,
+				 struct ocfs2_lock_res *lockres);
+static int ocfs2_meta_lock_update(struct inode *inode,
+				  struct buffer_head **bh);
+static void ocfs2_drop_osb_locks(struct ocfs2_super *osb);
+static inline int ocfs2_highest_compat_lock_level(int level);
+static inline int ocfs2_can_downconvert_meta_lock(struct inode *inode,
+						  struct ocfs2_lock_res *lockres,
+						  int new_level);
+
+static char *ocfs2_lock_type_strings[] = {
+	[OCFS2_LOCK_TYPE_META] = "Meta",
+	[OCFS2_LOCK_TYPE_DATA] = "Data",
+	[OCFS2_LOCK_TYPE_SUPER] = "Super",
+	[OCFS2_LOCK_TYPE_RENAME] = "Rename",
+	/* Need to differntiate from [R]ename.. serializing writes is the
+	 * important job it does, anyway. */
+	[OCFS2_LOCK_TYPE_RW] = "Write/Read",
+};
+
+static char *ocfs2_lock_type_string(enum ocfs2_lock_type type)
+{
+	mlog_bug_on_msg(type >= OCFS2_NUM_LOCK_TYPES, "%d\n", type);
+	return ocfs2_lock_type_strings[type];
+}
+
+static void ocfs2_build_lock_name(enum ocfs2_lock_type type,
+				  u64 blkno,
+				  u32 generation,
+				  char *name)
+{
+	int len;
+
+	mlog_entry_void();
+
+	BUG_ON(type >= OCFS2_NUM_LOCK_TYPES);
+
+	len = snprintf(name, OCFS2_LOCK_ID_MAX_LEN, "%c%s%016"MLFx64"%08x",
+		       ocfs2_lock_type_char(type), OCFS2_LOCK_ID_PAD, blkno,
+		       generation);
+
+	BUG_ON(len != (OCFS2_LOCK_ID_MAX_LEN - 1));
+
+	mlog(0, "built lock resource with name: %s\n", name);
+
+	mlog_exit_void();
+}
+
+static spinlock_t ocfs2_dlm_tracking_lock = SPIN_LOCK_UNLOCKED;
+
+static void ocfs2_add_lockres_tracking(struct ocfs2_lock_res *res,
+				       struct ocfs2_dlm_debug *dlm_debug)
+{
+	mlog(0, "Add tracking for lockres %s\n", res->l_name);
+
+	spin_lock(&ocfs2_dlm_tracking_lock);
+	list_add(&res->l_debug_list, &dlm_debug->d_lockres_tracking);
+	spin_unlock(&ocfs2_dlm_tracking_lock);
+}
+
+static void ocfs2_remove_lockres_tracking(struct ocfs2_lock_res *res)
+{
+	spin_lock(&ocfs2_dlm_tracking_lock);
+	if (!list_empty(&res->l_debug_list))
+		list_del_init(&res->l_debug_list);
+	spin_unlock(&ocfs2_dlm_tracking_lock);
+}
+
+static void ocfs2_lock_res_init_common(struct ocfs2_super *osb,
+				       struct ocfs2_lock_res *res,
+				       enum ocfs2_lock_type type,
+				       u64 blkno,
+				       u32 generation,
+				       struct ocfs2_lock_res_ops *ops,
+				       void *priv)
+{
+	ocfs2_build_lock_name(type, blkno, generation, res->l_name);
+
+	res->l_type          = type;
+	res->l_ops           = ops;
+	res->l_priv          = priv;
+
+	res->l_level         = LKM_IVMODE;
+	res->l_requested     = LKM_IVMODE;
+	res->l_blocking      = LKM_IVMODE;
+	res->l_action        = OCFS2_AST_INVALID;
+	res->l_unlock_action = OCFS2_UNLOCK_INVALID;
+
+	res->l_flags         = OCFS2_LOCK_INITIALIZED;
+
+	ocfs2_add_lockres_tracking(res, osb->osb_dlm_debug);
+}
+
+void ocfs2_lock_res_init_once(struct ocfs2_lock_res *res)
+{
+	/* This also clears out the lock status block */
+	memset(res, 0, sizeof(struct ocfs2_lock_res));
+	spin_lock_init(&res->l_lock);
+	init_waitqueue_head(&res->l_event);
+	INIT_LIST_HEAD(&res->l_blocked_list);
+	INIT_LIST_HEAD(&res->l_mask_waiters);
+}
+
+void ocfs2_inode_lock_res_init(struct ocfs2_lock_res *res,
+			       enum ocfs2_lock_type type,
+			       struct inode *inode)
+{
+	struct ocfs2_lock_res_ops *ops;
+
+	switch(type) {
+		case OCFS2_LOCK_TYPE_RW:
+			ops = &ocfs2_inode_rw_lops;
+			break;
+		case OCFS2_LOCK_TYPE_META:
+			ops = &ocfs2_inode_meta_lops;
+			break;
+		case OCFS2_LOCK_TYPE_DATA:
+			ops = &ocfs2_inode_data_lops;
+			break;
+		default:
+			mlog_bug_on_msg(1, "type: %d\n", type);
+			ops = NULL; /* thanks, gcc */
+			break;
+	};
+
+	ocfs2_lock_res_init_common(OCFS2_SB(inode->i_sb), res, type,
+				   OCFS2_I(inode)->ip_blkno,
+				   inode->i_generation, ops, inode);
+}
+
+static void ocfs2_super_lock_res_init(struct ocfs2_lock_res *res,
+				      struct ocfs2_super *osb)
+{
+	/* Superblock lockres doesn't come from a slab so we call init
+	 * once on it manually.  */
+	ocfs2_lock_res_init_once(res);
+	ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_SUPER,
+				   OCFS2_SUPER_BLOCK_BLKNO, 0,
+				   &ocfs2_super_lops, osb);
+}
+
+static void ocfs2_rename_lock_res_init(struct ocfs2_lock_res *res,
+				       struct ocfs2_super *osb)
+{
+	/* Rename lockres doesn't come from a slab so we call init
+	 * once on it manually.  */
+	ocfs2_lock_res_init_once(res);
+	ocfs2_lock_res_init_common(osb, res, OCFS2_LOCK_TYPE_RENAME, 0, 0,
+				   &ocfs2_rename_lops, osb);
+}
+
+void ocfs2_lock_res_free(struct ocfs2_lock_res *res)
+{
+	mlog_entry_void();
+
+	if (!(res->l_flags & OCFS2_LOCK_INITIALIZED))
+		return;
+
+	ocfs2_remove_lockres_tracking(res);
+
+	mlog_bug_on_msg(!list_empty(&res->l_blocked_list),
+			"Lockres %s is on the blocked list\n",
+			res->l_name);
+	mlog_bug_on_msg(!list_empty(&res->l_mask_waiters),
+			"Lockres %s has mask waiters pending\n",
+			res->l_name);
+	mlog_bug_on_msg(spin_is_locked(&res->l_lock),
+			"Lockres %s is locked\n",
+			res->l_name);
+	mlog_bug_on_msg(res->l_ro_holders,
+			"Lockres %s has %u ro holders\n",
+			res->l_name, res->l_ro_holders);
+	mlog_bug_on_msg(res->l_ex_holders,
+			"Lockres %s has %u ex holders\n",
+			res->l_name, res->l_ex_holders);
+
+	/* Need to clear out the lock status block for the dlm */
+	memset(&res->l_lksb, 0, sizeof(res->l_lksb));
+
+	res->l_flags = 0UL;
+	mlog_exit_void();
+}
+
+static inline void ocfs2_inc_holders(struct ocfs2_lock_res *lockres,
+				     int level)
+{
+	mlog_entry_void();
+
+	BUG_ON(!lockres);
+
+	switch(level) {
+	case LKM_EXMODE:
+		lockres->l_ex_holders++;
+		break;
+	case LKM_PRMODE:
+		lockres->l_ro_holders++;
+		break;
+	default:
+		BUG();
+	}
+
+	mlog_exit_void();
+}
+
+static inline void ocfs2_dec_holders(struct ocfs2_lock_res *lockres,
+				     int level)
+{
+	mlog_entry_void();
+
+	BUG_ON(!lockres);
+
+	switch(level) {
+	case LKM_EXMODE:
+		BUG_ON(!lockres->l_ex_holders);
+		lockres->l_ex_holders--;
+		break;
+	case LKM_PRMODE:
+		BUG_ON(!lockres->l_ro_holders);
+		lockres->l_ro_holders--;
+		break;
+	default:
+		BUG();
+	}
+	mlog_exit_void();
+}
+
+/* WARNING: This function lives in a world where the only three lock
+ * levels are EX, PR, and NL. It *will* have to be adjusted when more
+ * lock types are added. */
+static inline int ocfs2_highest_compat_lock_level(int level)
+{
+	int new_level = LKM_EXMODE;
+
+	if (level == LKM_EXMODE)
+		new_level = LKM_NLMODE;
+	else if (level == LKM_PRMODE)
+		new_level = LKM_PRMODE;
+	return new_level;
+}
+
+static void lockres_set_flags(struct ocfs2_lock_res *lockres,
+			      unsigned long newflags)
+{
+	struct list_head *pos, *tmp;
+	struct ocfs2_mask_waiter *mw;
+
+ 	assert_spin_locked(&lockres->l_lock);
+
+	lockres->l_flags = newflags;
+
+	list_for_each_safe(pos, tmp, &lockres->l_mask_waiters) {
+		mw = list_entry(pos, struct ocfs2_mask_waiter, mw_item);
+		if ((lockres->l_flags & mw->mw_mask) != mw->mw_goal)
+			continue;
+
+		list_del_init(&mw->mw_item);
+		mw->mw_status = 0;
+		complete(&mw->mw_complete);
+	}
+}
+static void lockres_or_flags(struct ocfs2_lock_res *lockres, unsigned long or)
+{
+	lockres_set_flags(lockres, lockres->l_flags | or);
+}
+static void lockres_clear_flags(struct ocfs2_lock_res *lockres,
+				unsigned long clear)
+{
+	lockres_set_flags(lockres, lockres->l_flags & ~clear);
+}
+
+static inline void ocfs2_generic_handle_downconvert_action(struct ocfs2_lock_res *lockres)
+{
+	mlog_entry_void();
+
+	BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY));
+	BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED));
+	BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED));
+	BUG_ON(lockres->l_blocking <= LKM_NLMODE);
+
+	lockres->l_level = lockres->l_requested;
+	if (lockres->l_level <=
+	    ocfs2_highest_compat_lock_level(lockres->l_blocking)) {
+		lockres->l_blocking = LKM_NLMODE;
+		lockres_clear_flags(lockres, OCFS2_LOCK_BLOCKED);
+	}
+	lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
+
+	mlog_exit_void();
+}
+
+static inline void ocfs2_generic_handle_convert_action(struct ocfs2_lock_res *lockres)
+{
+	mlog_entry_void();
+
+	BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BUSY));
+	BUG_ON(!(lockres->l_flags & OCFS2_LOCK_ATTACHED));
+
+	/* Convert from RO to EX doesn't really need anything as our
+	 * information is already up to data. Convert from NL to
+	 * *anything* however should mark ourselves as needing an
+	 * update */
+	if (lockres->l_level == LKM_NLMODE)
+		lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);
+
+	lockres->l_level = lockres->l_requested;
+	lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
+
+	mlog_exit_void();
+}
+
+static inline void ocfs2_generic_handle_attach_action(struct ocfs2_lock_res *lockres)
+{
+	mlog_entry_void();
+
+	BUG_ON((!lockres->l_flags & OCFS2_LOCK_BUSY));
+	BUG_ON(lockres->l_flags & OCFS2_LOCK_ATTACHED);
+
+	if (lockres->l_requested > LKM_NLMODE &&
+	    !(lockres->l_flags & OCFS2_LOCK_LOCAL))
+		lockres_or_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);
+
+	lockres->l_level = lockres->l_requested;
+	lockres_or_flags(lockres, OCFS2_LOCK_ATTACHED);
+	lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
+
+	mlog_exit_void();
+}
+
+static void ocfs2_inode_ast_func(void *opaque)
+{
+	struct ocfs2_lock_res *lockres = opaque;
+	struct inode *inode;
+	struct dlm_lockstatus *lksb;
+	unsigned long flags;
+
+	mlog_entry_void();
+
+	inode = ocfs2_lock_res_inode(lockres);
+
+	mlog(0, "AST fired for inode %"MLFu64", l_action = %u, type = %s\n",
+	     OCFS2_I(inode)->ip_blkno, lockres->l_action,
+	     ocfs2_lock_type_string(lockres->l_type));
+
+	BUG_ON(!ocfs2_is_inode_lock(lockres));
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+
+	lksb = &(lockres->l_lksb);
+	if (lksb->status != DLM_NORMAL) {
+		mlog(ML_ERROR, "ocfs2_inode_ast_func: lksb status value of %u "
+		     "on inode %"MLFu64"\n", lksb->status,
+		     OCFS2_I(inode)->ip_blkno);
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+		mlog_exit_void();
+		return;
+	}
+
+	switch(lockres->l_action) {
+	case OCFS2_AST_ATTACH:
+		ocfs2_generic_handle_attach_action(lockres);
+		lockres_clear_flags(lockres, OCFS2_LOCK_LOCAL);
+		break;
+	case OCFS2_AST_CONVERT:
+		ocfs2_generic_handle_convert_action(lockres);
+		break;
+	case OCFS2_AST_DOWNCONVERT:
+		ocfs2_generic_handle_downconvert_action(lockres);
+		break;
+	default:
+		mlog(ML_ERROR, "lockres %s: ast fired with invalid action: %u "
+		     "lockres flags = 0x%lx, unlock action: %u\n",
+		     lockres->l_name, lockres->l_action, lockres->l_flags,
+		     lockres->l_unlock_action);
+
+		BUG();
+	}
+
+	/* data and rw locking ignores refresh flag for now. */
+	if (lockres->l_type != OCFS2_LOCK_TYPE_META)
+		lockres_clear_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);
+
+	/* set it to something invalid so if we get called again we
+	 * can catch it. */
+	lockres->l_action = OCFS2_AST_INVALID;
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+	wake_up(&lockres->l_event);
+
+	mlog_exit_void();
+}
+
+static int ocfs2_generic_handle_bast(struct ocfs2_lock_res *lockres,
+				     int level)
+{
+	int needs_downconvert = 0;
+	mlog_entry_void();
+
+	assert_spin_locked(&lockres->l_lock);
+
+	lockres_or_flags(lockres, OCFS2_LOCK_BLOCKED);
+
+	if (level > lockres->l_blocking) {
+		/* only schedule a downconvert if we haven't already scheduled
+		 * one that goes low enough to satisfy the level we're
+		 * blocking.  this also catches the case where we get
+		 * duplicate BASTs */
+		if (ocfs2_highest_compat_lock_level(level) <
+		    ocfs2_highest_compat_lock_level(lockres->l_blocking))
+			needs_downconvert = 1;
+
+		lockres->l_blocking = level;
+	}
+
+	mlog_exit(needs_downconvert);
+	return needs_downconvert;
+}
+
+static void ocfs2_generic_bast_func(struct ocfs2_super *osb,
+				    struct ocfs2_lock_res *lockres,
+				    int level)
+{
+	int needs_downconvert;
+	unsigned long flags;
+
+	mlog_entry_void();
+
+	BUG_ON(level <= LKM_NLMODE);
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+	needs_downconvert = ocfs2_generic_handle_bast(lockres, level);
+	if (needs_downconvert)
+		ocfs2_schedule_blocked_lock(osb, lockres);
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+	ocfs2_kick_vote_thread(osb);
+
+	wake_up(&lockres->l_event);
+	mlog_exit_void();
+}
+
+static void ocfs2_inode_bast_func(void *opaque, int level)
+{
+	struct ocfs2_lock_res *lockres = opaque;
+	struct inode *inode;
+	struct ocfs2_super *osb;
+
+	mlog_entry_void();
+
+	BUG_ON(!ocfs2_is_inode_lock(lockres));
+
+	inode = ocfs2_lock_res_inode(lockres);
+	osb = OCFS2_SB(inode->i_sb);
+
+	mlog(0, "BAST fired for inode %"MLFu64", blocking = %d, level = %d "
+	     "type = %s\n", OCFS2_I(inode)->ip_blkno, level,
+	     lockres->l_level,
+	     ocfs2_lock_type_string(lockres->l_type));
+
+	ocfs2_generic_bast_func(osb, lockres, level);
+
+	mlog_exit_void();
+}
+
+static void ocfs2_generic_ast_func(struct ocfs2_lock_res *lockres,
+				   int ignore_refresh)
+{
+	struct dlm_lockstatus *lksb = &lockres->l_lksb;
+	unsigned long flags;
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+
+	if (lksb->status != DLM_NORMAL) {
+		mlog(ML_ERROR, "lockres %s: lksb status value of %u!\n",
+		     lockres->l_name, lksb->status);
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+		return;
+	}
+
+	switch(lockres->l_action) {
+	case OCFS2_AST_ATTACH:
+		ocfs2_generic_handle_attach_action(lockres);
+		break;
+	case OCFS2_AST_CONVERT:
+		ocfs2_generic_handle_convert_action(lockres);
+		break;
+	case OCFS2_AST_DOWNCONVERT:
+		ocfs2_generic_handle_downconvert_action(lockres);
+		break;
+	default:
+		BUG();
+	}
+
+	if (ignore_refresh)
+		lockres_clear_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);
+
+	/* set it to something invalid so if we get called again we
+	 * can catch it. */
+	lockres->l_action = OCFS2_AST_INVALID;
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+	wake_up(&lockres->l_event);
+}
+
+static void ocfs2_super_ast_func(void *opaque)
+{
+	struct ocfs2_lock_res *lockres = opaque;
+
+	mlog_entry_void();
+	mlog(0, "Superblock AST fired\n");
+
+	BUG_ON(!ocfs2_is_super_lock(lockres));
+	ocfs2_generic_ast_func(lockres, 0);
+
+	mlog_exit_void();
+}
+
+static void ocfs2_super_bast_func(void *opaque,
+				  int level)
+{
+	struct ocfs2_lock_res *lockres = opaque;
+	struct ocfs2_super *osb;
+
+	mlog_entry_void();
+	mlog(0, "Superblock BAST fired\n");
+
+	BUG_ON(!ocfs2_is_super_lock(lockres));
+       	osb = ocfs2_lock_res_super(lockres);
+	ocfs2_generic_bast_func(osb, lockres, level);
+
+	mlog_exit_void();
+}
+
+static void ocfs2_rename_ast_func(void *opaque)
+{
+	struct ocfs2_lock_res *lockres = opaque;
+
+	mlog_entry_void();
+
+	mlog(0, "Rename AST fired\n");
+
+	BUG_ON(!ocfs2_is_rename_lock(lockres));
+
+	ocfs2_generic_ast_func(lockres, 1);
+
+	mlog_exit_void();
+}
+
+static void ocfs2_rename_bast_func(void *opaque,
+				   int level)
+{
+	struct ocfs2_lock_res *lockres = opaque;
+	struct ocfs2_super *osb;
+
+	mlog_entry_void();
+
+	mlog(0, "Rename BAST fired\n");
+
+	BUG_ON(!ocfs2_is_rename_lock(lockres));
+
+	osb = ocfs2_lock_res_super(lockres);
+	ocfs2_generic_bast_func(osb, lockres, level);
+
+	mlog_exit_void();
+}
+
+static inline void ocfs2_recover_from_dlm_error(struct ocfs2_lock_res *lockres,
+						int convert)
+{
+	unsigned long flags;
+
+	mlog_entry_void();
+	spin_lock_irqsave(&lockres->l_lock, flags);
+	lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
+	if (convert)
+		lockres->l_action = OCFS2_AST_INVALID;
+	else
+		lockres->l_unlock_action = OCFS2_UNLOCK_INVALID;
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+	wake_up(&lockres->l_event);
+	mlog_exit_void();
+}
+
+/* Note: If we detect another process working on the lock (i.e.,
+ * OCFS2_LOCK_BUSY), we'll bail out returning 0. It's up to the caller
+ * to do the right thing in that case.
+ */
+static int ocfs2_lock_create(struct ocfs2_super *osb,
+			     struct ocfs2_lock_res *lockres,
+			     int level,
+			     int dlm_flags)
+{
+	int ret = 0;
+	enum dlm_status status;
+	unsigned long flags;
+
+	mlog_entry_void();
+
+	mlog(0, "lock %s, level = %d, flags = %d\n", lockres->l_name, level,
+	     dlm_flags);
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+	if ((lockres->l_flags & OCFS2_LOCK_ATTACHED) ||
+	    (lockres->l_flags & OCFS2_LOCK_BUSY)) {
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+		goto bail;
+	}
+
+	lockres->l_action = OCFS2_AST_ATTACH;
+	lockres->l_requested = level;
+	lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+	status = dlmlock(osb->dlm,
+			 level,
+			 &lockres->l_lksb,
+			 dlm_flags,
+			 lockres->l_name,
+			 lockres->l_ops->ast,
+			 lockres,
+			 lockres->l_ops->bast);
+	if (status != DLM_NORMAL) {
+		ocfs2_log_dlm_error("dlmlock", status, lockres);
+		ret = -EINVAL;
+		ocfs2_recover_from_dlm_error(lockres, 1);
+	}
+
+	mlog(0, "lock %s, successfull return from dlmlock\n", lockres->l_name);
+
+bail:
+	mlog_exit(ret);
+	return ret;
+}
+
+static inline int ocfs2_check_wait_flag(struct ocfs2_lock_res *lockres,
+					int flag)
+{
+	unsigned long flags;
+	int ret;
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+	ret = lockres->l_flags & flag;
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+	return ret;
+}
+
+static inline void ocfs2_wait_on_busy_lock(struct ocfs2_lock_res *lockres)
+
+{
+	wait_event(lockres->l_event,
+		   !ocfs2_check_wait_flag(lockres, OCFS2_LOCK_BUSY));
+}
+
+static inline void ocfs2_wait_on_refreshing_lock(struct ocfs2_lock_res *lockres)
+
+{
+	wait_event(lockres->l_event,
+		   !ocfs2_check_wait_flag(lockres, OCFS2_LOCK_REFRESHING));
+}
+
+/* predict what lock level we'll be dropping down to on behalf
+ * of another node, and return true if the currently wanted
+ * level will be compatible with it. */
+static inline int ocfs2_may_continue_on_blocked_lock(struct ocfs2_lock_res *lockres,
+						     int wanted)
+{
+	BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED));
+
+	return wanted <= ocfs2_highest_compat_lock_level(lockres->l_blocking);
+}
+
+static void ocfs2_init_mask_waiter(struct ocfs2_mask_waiter *mw)
+{
+	INIT_LIST_HEAD(&mw->mw_item);
+	init_completion(&mw->mw_complete);
+}
+
+static int ocfs2_wait_for_mask(struct ocfs2_mask_waiter *mw)
+{
+	wait_for_completion(&mw->mw_complete);
+	/* Re-arm the completion in case we want to wait on it again */
+	INIT_COMPLETION(mw->mw_complete);
+	return mw->mw_status;
+}
+
+static void lockres_add_mask_waiter(struct ocfs2_lock_res *lockres,
+				    struct ocfs2_mask_waiter *mw,
+				    unsigned long mask,
+				    unsigned long goal)
+{
+	BUG_ON(!list_empty(&mw->mw_item));
+
+	assert_spin_locked(&lockres->l_lock);
+
+	list_add_tail(&mw->mw_item, &lockres->l_mask_waiters);
+	mw->mw_mask = mask;
+	mw->mw_goal = goal;
+}
+
+/* returns 0 if the mw that was removed was already satisfied, -EBUSY
+ * if the mask still hadn't reached its goal */
+static int lockres_remove_mask_waiter(struct ocfs2_lock_res *lockres,
+				      struct ocfs2_mask_waiter *mw)
+{
+	unsigned long flags;
+	int ret = 0;
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+	if (!list_empty(&mw->mw_item)) {
+		if ((lockres->l_flags & mw->mw_mask) != mw->mw_goal)
+			ret = -EBUSY;
+
+		list_del_init(&mw->mw_item);
+		init_completion(&mw->mw_complete);
+	}
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+	return ret;
+
+}
+
+static int ocfs2_cluster_lock(struct ocfs2_super *osb,
+			      struct ocfs2_lock_res *lockres,
+			      int level,
+			      int lkm_flags,
+			      int arg_flags)
+{
+	struct ocfs2_mask_waiter mw;
+	enum dlm_status status;
+	int wait, catch_signals = !(osb->s_mount_opt & OCFS2_MOUNT_NOINTR);
+	int ret = 0; /* gcc doesn't realize wait = 1 guarantees ret is set */
+	unsigned long flags;
+
+	mlog_entry_void();
+
+	ocfs2_init_mask_waiter(&mw);
+
+again:
+	wait = 0;
+
+	if (catch_signals && signal_pending(current)) {
+		ret = -ERESTARTSYS;
+		goto out;
+	}
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+
+	mlog_bug_on_msg(lockres->l_flags & OCFS2_LOCK_FREEING,
+			"Cluster lock called on freeing lockres %s! flags "
+			"0x%lx\n", lockres->l_name, lockres->l_flags);
+
+	/* We only compare against the currently granted level
+	 * here. If the lock is blocked waiting on a downconvert,
+	 * we'll get caught below. */
+	if (lockres->l_flags & OCFS2_LOCK_BUSY &&
+	    level > lockres->l_level) {
+		/* is someone sitting in dlm_lock? If so, wait on
+		 * them. */
+		lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BUSY, 0);
+		wait = 1;
+		goto unlock;
+	}
+
+	if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) {
+		/* lock has not been created yet. */
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+		ret = ocfs2_lock_create(osb, lockres, LKM_NLMODE, 0);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out;
+		}
+		goto again;
+	}
+
+	if (lockres->l_flags & OCFS2_LOCK_BLOCKED &&
+	    !ocfs2_may_continue_on_blocked_lock(lockres, level)) {
+		/* is the lock is currently blocked on behalf of
+		 * another node */
+		lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_BLOCKED, 0);
+		wait = 1;
+		goto unlock;
+	}
+
+	if (level > lockres->l_level) {
+		if (lockres->l_action != OCFS2_AST_INVALID)
+			mlog(ML_ERROR, "lockres %s has action %u pending\n",
+			     lockres->l_name, lockres->l_action);
+
+		lockres->l_action = OCFS2_AST_CONVERT;
+		lockres->l_requested = level;
+		lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+		BUG_ON(level == LKM_IVMODE);
+		BUG_ON(level == LKM_NLMODE);
+
+		mlog(0, "lock %s, convert from %d to level = %d\n",
+		     lockres->l_name, lockres->l_level, level);
+
+		/* call dlm_lock to upgrade lock now */
+		status = dlmlock(osb->dlm,
+				 level,
+				 &lockres->l_lksb,
+				 lkm_flags|LKM_CONVERT|LKM_VALBLK,
+				 lockres->l_name,
+				 lockres->l_ops->ast,
+				 lockres,
+				 lockres->l_ops->bast);
+		if (status != DLM_NORMAL) {
+			if ((lkm_flags & LKM_NOQUEUE) &&
+			    (status == DLM_NOTQUEUED))
+				ret = -EAGAIN;
+			else {
+				ocfs2_log_dlm_error("dlmlock", status,
+						    lockres);
+				ret = -EINVAL;
+			}
+			ocfs2_recover_from_dlm_error(lockres, 1);
+			goto out;
+		}
+
+		mlog(0, "lock %s, successfull return from dlmlock\n",
+		     lockres->l_name);
+
+		/* At this point we've gone inside the dlm and need to
+		 * complete our work regardless. */
+		catch_signals = 0;
+
+		/* wait for busy to clear and carry on */
+		goto again;
+	}
+
+	/* Ok, if we get here then we're good to go. */
+	ocfs2_inc_holders(lockres, level);
+
+	ret = 0;
+unlock:
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+out:
+	/*
+	 * This is helping work around a lock inversion between the page lock
+	 * and dlm locks.  One path holds the page lock while calling aops
+	 * which block acquiring dlm locks.  The voting thread holds dlm
+	 * locks while acquiring page locks while down converting data locks.
+	 * This block is helping an aop path notice the inversion and back
+	 * off to unlock its page lock before trying the dlm lock again.
+	 */
+	if (wait && arg_flags & OCFS2_LOCK_NONBLOCK &&
+	    mw.mw_mask & (OCFS2_LOCK_BUSY|OCFS2_LOCK_BLOCKED)) {
+		wait = 0;
+		if (lockres_remove_mask_waiter(lockres, &mw))
+			ret = -EAGAIN;
+		else
+			goto again;
+	}
+	if (wait) {
+		ret = ocfs2_wait_for_mask(&mw);
+		if (ret == 0)
+			goto again;
+		mlog_errno(ret);
+	}
+
+	mlog_exit(ret);
+	return ret;
+}
+
+static void ocfs2_cluster_unlock(struct ocfs2_super *osb,
+				 struct ocfs2_lock_res *lockres,
+				 int level)
+{
+	unsigned long flags;
+
+	mlog_entry_void();
+	spin_lock_irqsave(&lockres->l_lock, flags);
+	ocfs2_dec_holders(lockres, level);
+	ocfs2_vote_on_unlock(osb, lockres);
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+	mlog_exit_void();
+}
+
+static int ocfs2_create_new_inode_lock(struct inode *inode,
+				       struct ocfs2_lock_res *lockres)
+{
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	unsigned long flags;
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+	BUG_ON(lockres->l_flags & OCFS2_LOCK_ATTACHED);
+	lockres_or_flags(lockres, OCFS2_LOCK_LOCAL);
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+	return ocfs2_lock_create(osb, lockres, LKM_EXMODE, LKM_LOCAL);
+}
+
+/* Grants us an EX lock on the data and metadata resources, skipping
+ * the normal cluster directory lookup. Use this ONLY on newly created
+ * inodes which other nodes can't possibly see, and which haven't been
+ * hashed in the inode hash yet. This can give us a good performance
+ * increase as it'll skip the network broadcast normally associated
+ * with creating a new lock resource. */
+int ocfs2_create_new_inode_locks(struct inode *inode)
+{
+	int ret;
+
+	BUG_ON(!inode);
+	BUG_ON(!ocfs2_inode_is_new(inode));
+
+	mlog_entry_void();
+
+	mlog(0, "Inode %"MLFu64"\n", OCFS2_I(inode)->ip_blkno);
+
+	/* NOTE: That we don't increment any of the holder counts, nor
+	 * do we add anything to a journal handle. Since this is
+	 * supposed to be a new inode which the cluster doesn't know
+	 * about yet, there is no need to.  As far as the LVB handling
+	 * is concerned, this is basically like acquiring an EX lock
+	 * on a resource which has an invalid one -- we'll set it
+	 * valid when we release the EX. */
+
+	ret = ocfs2_create_new_inode_lock(inode,
+					  &OCFS2_I(inode)->ip_rw_lockres);
+	if (ret) {
+		mlog_errno(ret);
+		goto bail;
+	}
+
+	ret = ocfs2_create_new_inode_lock(inode,
+					  &OCFS2_I(inode)->ip_meta_lockres);
+	if (ret) {
+		mlog_errno(ret);
+		goto bail;
+	}
+
+	ret = ocfs2_create_new_inode_lock(inode,
+					  &OCFS2_I(inode)->ip_data_lockres);
+	if (ret) {
+		mlog_errno(ret);
+		goto bail;
+	}
+
+bail:
+	mlog_exit(ret);
+	return ret;
+}
+
+int ocfs2_rw_lock(struct inode *inode, int write)
+{
+	int status, level;
+	struct ocfs2_lock_res *lockres;
+
+	BUG_ON(!inode);
+
+	mlog_entry_void();
+
+	mlog(0, "inode %"MLFu64" take %s RW lock\n",
+	     OCFS2_I(inode)->ip_blkno,
+	     write ? "EXMODE" : "PRMODE");
+
+	lockres = &OCFS2_I(inode)->ip_rw_lockres;
+
+	level = write ? LKM_EXMODE : LKM_PRMODE;
+
+	status = ocfs2_cluster_lock(OCFS2_SB(inode->i_sb), lockres, level, 0,
+				    0);
+	if (status < 0)
+		mlog_errno(status);
+
+	mlog_exit(status);
+	return status;
+}
+
+void ocfs2_rw_unlock(struct inode *inode, int write)
+{
+	int level = write ? LKM_EXMODE : LKM_PRMODE;
+	struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_rw_lockres;
+
+	mlog_entry_void();
+
+	mlog(0, "inode %"MLFu64" drop %s RW lock\n",
+	     OCFS2_I(inode)->ip_blkno,
+	     write ? "EXMODE" : "PRMODE");
+
+	ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level);
+
+	mlog_exit_void();
+}
+
+int ocfs2_data_lock_full(struct inode *inode,
+			 int write,
+			 int arg_flags)
+{
+	int status = 0, level;
+	struct ocfs2_lock_res *lockres;
+
+	BUG_ON(!inode);
+
+	mlog_entry_void();
+
+	mlog(0, "inode %"MLFu64" take %s DATA lock\n",
+	     OCFS2_I(inode)->ip_blkno,
+	     write ? "EXMODE" : "PRMODE");
+
+	/* We'll allow faking a readonly data lock for
+	 * rodevices. */
+	if (ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb))) {
+		if (write) {
+			status = -EROFS;
+			mlog_errno(status);
+		}
+		goto out;
+	}
+
+	lockres = &OCFS2_I(inode)->ip_data_lockres;
+
+	level = write ? LKM_EXMODE : LKM_PRMODE;
+
+	status = ocfs2_cluster_lock(OCFS2_SB(inode->i_sb), lockres, level,
+				    0, arg_flags);
+	if (status < 0 && status != -EAGAIN)
+		mlog_errno(status);
+
+out:
+	mlog_exit(status);
+	return status;
+}
+
+/* see ocfs2_meta_lock_with_page() */
+int ocfs2_data_lock_with_page(struct inode *inode,
+			      int write,
+			      struct page *page)
+{
+	int ret;
+
+	ret = ocfs2_data_lock_full(inode, write, OCFS2_LOCK_NONBLOCK);
+	if (ret == -EAGAIN) {
+		unlock_page(page);
+		if (ocfs2_data_lock(inode, write) == 0)
+			ocfs2_data_unlock(inode, write);
+		ret = AOP_TRUNCATED_PAGE;
+	}
+
+	return ret;
+}
+
+static void ocfs2_vote_on_unlock(struct ocfs2_super *osb,
+				 struct ocfs2_lock_res *lockres)
+{
+	int kick = 0;
+
+	mlog_entry_void();
+
+	/* If we know that another node is waiting on our lock, kick
+	 * the vote thread * pre-emptively when we reach a release
+	 * condition. */
+	if (lockres->l_flags & OCFS2_LOCK_BLOCKED) {
+		switch(lockres->l_blocking) {
+		case LKM_EXMODE:
+			if (!lockres->l_ex_holders && !lockres->l_ro_holders)
+				kick = 1;
+			break;
+		case LKM_PRMODE:
+			if (!lockres->l_ex_holders)
+				kick = 1;
+			break;
+		default:
+			BUG();
+		}
+	}
+
+	if (kick)
+		ocfs2_kick_vote_thread(osb);
+
+	mlog_exit_void();
+}
+
+void ocfs2_data_unlock(struct inode *inode,
+		       int write)
+{
+	int level = write ? LKM_EXMODE : LKM_PRMODE;
+	struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_data_lockres;
+
+	mlog_entry_void();
+
+	mlog(0, "inode %"MLFu64" drop %s DATA lock\n",
+	     OCFS2_I(inode)->ip_blkno,
+	     write ? "EXMODE" : "PRMODE");
+
+	if (!ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb)))
+		ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level);
+
+	mlog_exit_void();
+}
+
+#define OCFS2_SEC_BITS   34
+#define OCFS2_SEC_SHIFT  (64 - 34)
+#define OCFS2_NSEC_MASK  ((1ULL << OCFS2_SEC_SHIFT) - 1)
+
+/* LVB only has room for 64 bits of time here so we pack it for
+ * now. */
+static u64 ocfs2_pack_timespec(struct timespec *spec)
+{
+	u64 res;
+	u64 sec = spec->tv_sec;
+	u32 nsec = spec->tv_nsec;
+
+	res = (sec << OCFS2_SEC_SHIFT) | (nsec & OCFS2_NSEC_MASK);
+
+	return res;
+}
+
+/* Call this with the lockres locked. I am reasonably sure we don't
+ * need ip_lock in this function as anyone who would be changing those
+ * values is supposed to be blocked in ocfs2_meta_lock right now. */
+static void __ocfs2_stuff_meta_lvb(struct inode *inode)
+{
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+	struct ocfs2_lock_res *lockres = &oi->ip_meta_lockres;
+	struct ocfs2_meta_lvb *lvb;
+
+	mlog_entry_void();
+
+	lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb;
+
+	lvb->lvb_version   = cpu_to_be32(OCFS2_LVB_VERSION);
+	lvb->lvb_isize	   = cpu_to_be64(i_size_read(inode));
+	lvb->lvb_iclusters = cpu_to_be32(oi->ip_clusters);
+	lvb->lvb_iuid      = cpu_to_be32(inode->i_uid);
+	lvb->lvb_igid      = cpu_to_be32(inode->i_gid);
+	lvb->lvb_imode     = cpu_to_be16(inode->i_mode);
+	lvb->lvb_inlink    = cpu_to_be16(inode->i_nlink);
+	lvb->lvb_iatime_packed  =
+		cpu_to_be64(ocfs2_pack_timespec(&inode->i_atime));
+	lvb->lvb_ictime_packed =
+		cpu_to_be64(ocfs2_pack_timespec(&inode->i_ctime));
+	lvb->lvb_imtime_packed =
+		cpu_to_be64(ocfs2_pack_timespec(&inode->i_mtime));
+
+	mlog_meta_lvb(0, lockres);
+
+	mlog_exit_void();
+}
+
+static void ocfs2_unpack_timespec(struct timespec *spec,
+				  u64 packed_time)
+{
+	spec->tv_sec = packed_time >> OCFS2_SEC_SHIFT;
+	spec->tv_nsec = packed_time & OCFS2_NSEC_MASK;
+}
+
+static void ocfs2_refresh_inode_from_lvb(struct inode *inode)
+{
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+	struct ocfs2_lock_res *lockres = &oi->ip_meta_lockres;
+	struct ocfs2_meta_lvb *lvb;
+
+	mlog_entry_void();
+
+	mlog_meta_lvb(0, lockres);
+
+	lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb;
+
+	/* We're safe here without the lockres lock... */
+	spin_lock(&oi->ip_lock);
+	oi->ip_clusters = be32_to_cpu(lvb->lvb_iclusters);
+	i_size_write(inode, be64_to_cpu(lvb->lvb_isize));
+
+	/* fast-symlinks are a special case */
+	if (S_ISLNK(inode->i_mode) && !oi->ip_clusters)
+		inode->i_blocks = 0;
+	else
+		inode->i_blocks =
+			ocfs2_align_bytes_to_sectors(i_size_read(inode));
+
+	inode->i_uid     = be32_to_cpu(lvb->lvb_iuid);
+	inode->i_gid     = be32_to_cpu(lvb->lvb_igid);
+	inode->i_mode    = be16_to_cpu(lvb->lvb_imode);
+	inode->i_nlink   = be16_to_cpu(lvb->lvb_inlink);
+	ocfs2_unpack_timespec(&inode->i_atime,
+			      be64_to_cpu(lvb->lvb_iatime_packed));
+	ocfs2_unpack_timespec(&inode->i_mtime,
+			      be64_to_cpu(lvb->lvb_imtime_packed));
+	ocfs2_unpack_timespec(&inode->i_ctime,
+			      be64_to_cpu(lvb->lvb_ictime_packed));
+	spin_unlock(&oi->ip_lock);
+
+	mlog_exit_void();
+}
+
+static inline int ocfs2_meta_lvb_is_trustable(struct ocfs2_lock_res *lockres)
+{
+	struct ocfs2_meta_lvb *lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb;
+
+	if (be32_to_cpu(lvb->lvb_version) == OCFS2_LVB_VERSION)
+		return 1;
+	return 0;
+}
+
+/* Determine whether a lock resource needs to be refreshed, and
+ * arbitrate who gets to refresh it.
+ *
+ *   0 means no refresh needed.
+ *
+ *   > 0 means you need to refresh this and you MUST call
+ *   ocfs2_complete_lock_res_refresh afterwards. */
+static int ocfs2_should_refresh_lock_res(struct ocfs2_lock_res *lockres)
+{
+	unsigned long flags;
+	int status = 0;
+
+	mlog_entry_void();
+
+refresh_check:
+	spin_lock_irqsave(&lockres->l_lock, flags);
+	if (!(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH)) {
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+		goto bail;
+	}
+
+	if (lockres->l_flags & OCFS2_LOCK_REFRESHING) {
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+		ocfs2_wait_on_refreshing_lock(lockres);
+		goto refresh_check;
+	}
+
+	/* Ok, I'll be the one to refresh this lock. */
+	lockres_or_flags(lockres, OCFS2_LOCK_REFRESHING);
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+	status = 1;
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+/* If status is non zero, I'll mark it as not being in refresh
+ * anymroe, but i won't clear the needs refresh flag. */
+static inline void ocfs2_complete_lock_res_refresh(struct ocfs2_lock_res *lockres,
+						   int status)
+{
+	unsigned long flags;
+	mlog_entry_void();
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+	lockres_clear_flags(lockres, OCFS2_LOCK_REFRESHING);
+	if (!status)
+		lockres_clear_flags(lockres, OCFS2_LOCK_NEEDS_REFRESH);
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+	wake_up(&lockres->l_event);
+
+	mlog_exit_void();
+}
+
+/* may or may not return a bh if it went to disk. */
+static int ocfs2_meta_lock_update(struct inode *inode,
+				  struct buffer_head **bh)
+{
+	int status = 0;
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+	struct ocfs2_lock_res *lockres;
+	struct ocfs2_dinode *fe;
+
+	mlog_entry_void();
+
+	spin_lock(&oi->ip_lock);
+	if (oi->ip_flags & OCFS2_INODE_DELETED) {
+		mlog(0, "Orphaned inode %"MLFu64" was deleted while we "
+		     "were waiting on a lock. ip_flags = 0x%x\n",
+		     oi->ip_blkno, oi->ip_flags);
+		spin_unlock(&oi->ip_lock);
+		status = -ENOENT;
+		goto bail;
+	}
+	spin_unlock(&oi->ip_lock);
+
+	lockres = &oi->ip_meta_lockres;
+
+	if (!ocfs2_should_refresh_lock_res(lockres))
+		goto bail;
+
+	/* This will discard any caching information we might have had
+	 * for the inode metadata. */
+	ocfs2_metadata_cache_purge(inode);
+
+	/* will do nothing for inode types that don't use the extent
+	 * map (directories, bitmap files, etc) */
+	ocfs2_extent_map_trunc(inode, 0);
+
+	if (ocfs2_meta_lvb_is_trustable(lockres)) {
+		mlog(0, "Trusting LVB on inode %"MLFu64"\n",
+		     oi->ip_blkno);
+		ocfs2_refresh_inode_from_lvb(inode);
+	} else {
+		/* Boo, we have to go to disk. */
+		/* read bh, cast, ocfs2_refresh_inode */
+		status = ocfs2_read_block(OCFS2_SB(inode->i_sb), oi->ip_blkno,
+					  bh, OCFS2_BH_CACHED, inode);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail_refresh;
+		}
+		fe = (struct ocfs2_dinode *) (*bh)->b_data;
+
+		/* This is a good chance to make sure we're not
+		 * locking an invalid object.
+		 *
+		 * We bug on a stale inode here because we checked
+		 * above whether it was wiped from disk. The wiping
+		 * node provides a guarantee that we receive that
+		 * message and can mark the inode before dropping any
+		 * locks associated with it. */
+		if (!OCFS2_IS_VALID_DINODE(fe)) {
+			OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe);
+			status = -EIO;
+			goto bail_refresh;
+		}
+		mlog_bug_on_msg(inode->i_generation !=
+				le32_to_cpu(fe->i_generation),
+				"Invalid dinode %"MLFu64" disk generation: %u "
+				"inode->i_generation: %u\n",
+				oi->ip_blkno, le32_to_cpu(fe->i_generation),
+				inode->i_generation);
+		mlog_bug_on_msg(le64_to_cpu(fe->i_dtime) ||
+				!(fe->i_flags & cpu_to_le32(OCFS2_VALID_FL)),
+				"Stale dinode %"MLFu64" dtime: %"MLFu64" "
+				"flags: 0x%x\n", oi->ip_blkno,
+				le64_to_cpu(fe->i_dtime),
+				le32_to_cpu(fe->i_flags));
+
+		ocfs2_refresh_inode(inode, fe);
+	}
+
+	status = 0;
+bail_refresh:
+	ocfs2_complete_lock_res_refresh(lockres, status);
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_assign_bh(struct inode *inode,
+			   struct buffer_head **ret_bh,
+			   struct buffer_head *passed_bh)
+{
+	int status;
+
+	if (passed_bh) {
+		/* Ok, the update went to disk for us, use the
+		 * returned bh. */
+		*ret_bh = passed_bh;
+		get_bh(*ret_bh);
+
+		return 0;
+	}
+
+	status = ocfs2_read_block(OCFS2_SB(inode->i_sb),
+				  OCFS2_I(inode)->ip_blkno,
+				  ret_bh,
+				  OCFS2_BH_CACHED,
+				  inode);
+	if (status < 0)
+		mlog_errno(status);
+
+	return status;
+}
+
+/*
+ * returns < 0 error if the callback will never be called, otherwise
+ * the result of the lock will be communicated via the callback.
+ */
+int ocfs2_meta_lock_full(struct inode *inode,
+			 struct ocfs2_journal_handle *handle,
+			 struct buffer_head **ret_bh,
+			 int ex,
+			 int arg_flags)
+{
+	int status, level, dlm_flags, acquired;
+	struct ocfs2_lock_res *lockres;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	struct buffer_head *local_bh = NULL;
+
+	BUG_ON(!inode);
+
+	mlog_entry_void();
+
+	mlog(0, "inode %"MLFu64", take %s META lock\n",
+	     OCFS2_I(inode)->ip_blkno,
+	     ex ? "EXMODE" : "PRMODE");
+
+	status = 0;
+	acquired = 0;
+	/* We'll allow faking a readonly metadata lock for
+	 * rodevices. */
+	if (ocfs2_is_hard_readonly(osb)) {
+		if (ex)
+			status = -EROFS;
+		goto bail;
+	}
+
+	if (!(arg_flags & OCFS2_META_LOCK_RECOVERY))
+		wait_event(osb->recovery_event,
+			   ocfs2_node_map_is_empty(osb, &osb->recovery_map));
+
+	acquired = 0;
+	lockres = &OCFS2_I(inode)->ip_meta_lockres;
+	level = ex ? LKM_EXMODE : LKM_PRMODE;
+	dlm_flags = 0;
+	if (arg_flags & OCFS2_META_LOCK_NOQUEUE)
+		dlm_flags |= LKM_NOQUEUE;
+
+	status = ocfs2_cluster_lock(osb, lockres, level, dlm_flags, arg_flags);
+	if (status < 0) {
+		if (status != -EAGAIN && status != -EIOCBRETRY)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	/* Notify the error cleanup path to drop the cluster lock. */
+	acquired = 1;
+
+	/* We wait twice because a node may have died while we were in
+	 * the lower dlm layers. The second time though, we've
+	 * committed to owning this lock so we don't allow signals to
+	 * abort the operation. */
+	if (!(arg_flags & OCFS2_META_LOCK_RECOVERY))
+		wait_event(osb->recovery_event,
+			   ocfs2_node_map_is_empty(osb, &osb->recovery_map));
+
+	/* This is fun. The caller may want a bh back, or it may
+	 * not. ocfs2_meta_lock_update definitely wants one in, but
+	 * may or may not read one, depending on what's in the
+	 * LVB. The result of all of this is that we've *only* gone to
+	 * disk if we have to, so the complexity is worthwhile. */
+	status = ocfs2_meta_lock_update(inode, &local_bh);
+	if (status < 0) {
+		if (status != -ENOENT)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	if (ret_bh) {
+		status = ocfs2_assign_bh(inode, ret_bh, local_bh);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+	}
+
+	if (handle) {
+		status = ocfs2_handle_add_lock(handle, inode);
+		if (status < 0)
+			mlog_errno(status);
+	}
+
+bail:
+	if (status < 0) {
+		if (ret_bh && (*ret_bh)) {
+			brelse(*ret_bh);
+			*ret_bh = NULL;
+		}
+		if (acquired)
+			ocfs2_meta_unlock(inode, ex);
+	}
+
+	if (local_bh)
+		brelse(local_bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * This is working around a lock inversion between tasks acquiring DLM locks
+ * while holding a page lock and the vote thread which blocks dlm lock acquiry
+ * while acquiring page locks.
+ *
+ * ** These _with_page variantes are only intended to be called from aop
+ * methods that hold page locks and return a very specific *positive* error
+ * code that aop methods pass up to the VFS -- test for errors with != 0. **
+ *
+ * The DLM is called such that it returns -EAGAIN if it would have blocked
+ * waiting for the vote thread.  In that case we unlock our page so the vote
+ * thread can make progress.  Once we've done this we have to return
+ * AOP_TRUNCATED_PAGE so the aop method that called us can bubble that back up
+ * into the VFS who will then immediately retry the aop call.
+ *
+ * We do a blocking lock and immediate unlock before returning, though, so that
+ * the lock has a great chance of being cached on this node by the time the VFS
+ * calls back to retry the aop.    This has a potential to livelock as nodes
+ * ping locks back and forth, but that's a risk we're willing to take to avoid
+ * the lock inversion simply.
+ */
+int ocfs2_meta_lock_with_page(struct inode *inode,
+			      struct ocfs2_journal_handle *handle,
+			      struct buffer_head **ret_bh,
+			      int ex,
+			      struct page *page)
+{
+	int ret;
+
+	ret = ocfs2_meta_lock_full(inode, handle, ret_bh, ex,
+				   OCFS2_LOCK_NONBLOCK);
+	if (ret == -EAGAIN) {
+		unlock_page(page);
+		if (ocfs2_meta_lock(inode, handle, ret_bh, ex) == 0)
+			ocfs2_meta_unlock(inode, ex);
+		ret = AOP_TRUNCATED_PAGE;
+	}
+
+	return ret;
+}
+
+void ocfs2_meta_unlock(struct inode *inode,
+		       int ex)
+{
+	int level = ex ? LKM_EXMODE : LKM_PRMODE;
+	struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_meta_lockres;
+
+	mlog_entry_void();
+
+	mlog(0, "inode %"MLFu64" drop %s META lock\n",
+	     OCFS2_I(inode)->ip_blkno,
+	     ex ? "EXMODE" : "PRMODE");
+
+	if (!ocfs2_is_hard_readonly(OCFS2_SB(inode->i_sb)))
+		ocfs2_cluster_unlock(OCFS2_SB(inode->i_sb), lockres, level);
+
+	mlog_exit_void();
+}
+
+int ocfs2_super_lock(struct ocfs2_super *osb,
+		     int ex)
+{
+	int status;
+	int level = ex ? LKM_EXMODE : LKM_PRMODE;
+	struct ocfs2_lock_res *lockres = &osb->osb_super_lockres;
+	struct buffer_head *bh;
+	struct ocfs2_slot_info *si = osb->slot_info;
+
+	mlog_entry_void();
+
+	if (ocfs2_is_hard_readonly(osb))
+		return -EROFS;
+
+	status = ocfs2_cluster_lock(osb, lockres, level, 0, 0);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* The super block lock path is really in the best position to
+	 * know when resources covered by the lock need to be
+	 * refreshed, so we do it here. Of course, making sense of
+	 * everything is up to the caller :) */
+	status = ocfs2_should_refresh_lock_res(lockres);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+	if (status) {
+		bh = si->si_bh;
+		status = ocfs2_read_block(osb, bh->b_blocknr, &bh, 0,
+					  si->si_inode);
+		if (status == 0)
+			ocfs2_update_slot_info(si);
+
+		ocfs2_complete_lock_res_refresh(lockres, status);
+
+		if (status < 0)
+			mlog_errno(status);
+	}
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+void ocfs2_super_unlock(struct ocfs2_super *osb,
+			int ex)
+{
+	int level = ex ? LKM_EXMODE : LKM_PRMODE;
+	struct ocfs2_lock_res *lockres = &osb->osb_super_lockres;
+
+	ocfs2_cluster_unlock(osb, lockres, level);
+}
+
+int ocfs2_rename_lock(struct ocfs2_super *osb)
+{
+	int status;
+	struct ocfs2_lock_res *lockres = &osb->osb_rename_lockres;
+
+	if (ocfs2_is_hard_readonly(osb))
+		return -EROFS;
+
+	status = ocfs2_cluster_lock(osb, lockres, LKM_EXMODE, 0, 0);
+	if (status < 0)
+		mlog_errno(status);
+
+	return status;
+}
+
+void ocfs2_rename_unlock(struct ocfs2_super *osb)
+{
+	struct ocfs2_lock_res *lockres = &osb->osb_rename_lockres;
+
+	ocfs2_cluster_unlock(osb, lockres, LKM_EXMODE);
+}
+
+/* Reference counting of the dlm debug structure. We want this because
+ * open references on the debug inodes can live on after a mount, so
+ * we can't rely on the ocfs2_super to always exist. */
+static void ocfs2_dlm_debug_free(struct kref *kref)
+{
+	struct ocfs2_dlm_debug *dlm_debug;
+
+	dlm_debug = container_of(kref, struct ocfs2_dlm_debug, d_refcnt);
+
+	kfree(dlm_debug);
+}
+
+void ocfs2_put_dlm_debug(struct ocfs2_dlm_debug *dlm_debug)
+{
+	if (dlm_debug)
+		kref_put(&dlm_debug->d_refcnt, ocfs2_dlm_debug_free);
+}
+
+static void ocfs2_get_dlm_debug(struct ocfs2_dlm_debug *debug)
+{
+	kref_get(&debug->d_refcnt);
+}
+
+struct ocfs2_dlm_debug *ocfs2_new_dlm_debug(void)
+{
+	struct ocfs2_dlm_debug *dlm_debug;
+
+	dlm_debug = kmalloc(sizeof(struct ocfs2_dlm_debug), GFP_KERNEL);
+	if (!dlm_debug) {
+		mlog_errno(-ENOMEM);
+		goto out;
+	}
+
+	kref_init(&dlm_debug->d_refcnt);
+	INIT_LIST_HEAD(&dlm_debug->d_lockres_tracking);
+	dlm_debug->d_locking_state = NULL;
+out:
+	return dlm_debug;
+}
+
+/* Access to this is arbitrated for us via seq_file->sem. */
+struct ocfs2_dlm_seq_priv {
+	struct ocfs2_dlm_debug *p_dlm_debug;
+	struct ocfs2_lock_res p_iter_res;
+	struct ocfs2_lock_res p_tmp_res;
+};
+
+static struct ocfs2_lock_res *ocfs2_dlm_next_res(struct ocfs2_lock_res *start,
+						 struct ocfs2_dlm_seq_priv *priv)
+{
+	struct ocfs2_lock_res *iter, *ret = NULL;
+	struct ocfs2_dlm_debug *dlm_debug = priv->p_dlm_debug;
+
+	assert_spin_locked(&ocfs2_dlm_tracking_lock);
+
+	list_for_each_entry(iter, &start->l_debug_list, l_debug_list) {
+		/* discover the head of the list */
+		if (&iter->l_debug_list == &dlm_debug->d_lockres_tracking) {
+			mlog(0, "End of list found, %p\n", ret);
+			break;
+		}
+
+		/* We track our "dummy" iteration lockres' by a NULL
+		 * l_ops field. */
+		if (iter->l_ops != NULL) {
+			ret = iter;
+			break;
+		}
+	}
+
+	return ret;
+}
+
+static void *ocfs2_dlm_seq_start(struct seq_file *m, loff_t *pos)
+{
+	struct ocfs2_dlm_seq_priv *priv = m->private;
+	struct ocfs2_lock_res *iter;
+
+	spin_lock(&ocfs2_dlm_tracking_lock);
+	iter = ocfs2_dlm_next_res(&priv->p_iter_res, priv);
+	if (iter) {
+		/* Since lockres' have the lifetime of their container
+		 * (which can be inodes, ocfs2_supers, etc) we want to
+		 * copy this out to a temporary lockres while still
+		 * under the spinlock. Obviously after this we can't
+		 * trust any pointers on the copy returned, but that's
+		 * ok as the information we want isn't typically held
+		 * in them. */
+		priv->p_tmp_res = *iter;
+		iter = &priv->p_tmp_res;
+	}
+	spin_unlock(&ocfs2_dlm_tracking_lock);
+
+	return iter;
+}
+
+static void ocfs2_dlm_seq_stop(struct seq_file *m, void *v)
+{
+}
+
+static void *ocfs2_dlm_seq_next(struct seq_file *m, void *v, loff_t *pos)
+{
+	struct ocfs2_dlm_seq_priv *priv = m->private;
+	struct ocfs2_lock_res *iter = v;
+	struct ocfs2_lock_res *dummy = &priv->p_iter_res;
+
+	spin_lock(&ocfs2_dlm_tracking_lock);
+	iter = ocfs2_dlm_next_res(iter, priv);
+	list_del_init(&dummy->l_debug_list);
+	if (iter) {
+		list_add(&dummy->l_debug_list, &iter->l_debug_list);
+		priv->p_tmp_res = *iter;
+		iter = &priv->p_tmp_res;
+	}
+	spin_unlock(&ocfs2_dlm_tracking_lock);
+
+	return iter;
+}
+
+/* So that debugfs.ocfs2 can determine which format is being used */
+#define OCFS2_DLM_DEBUG_STR_VERSION 1
+static int ocfs2_dlm_seq_show(struct seq_file *m, void *v)
+{
+	int i;
+	char *lvb;
+	struct ocfs2_lock_res *lockres = v;
+
+	if (!lockres)
+		return -EINVAL;
+
+	seq_printf(m, "0x%x\t"
+		   "%.*s\t"
+		   "%d\t"
+		   "0x%lx\t"
+		   "0x%x\t"
+		   "0x%x\t"
+		   "%u\t"
+		   "%u\t"
+		   "%d\t"
+		   "%d\t",
+		   OCFS2_DLM_DEBUG_STR_VERSION,
+		   OCFS2_LOCK_ID_MAX_LEN, lockres->l_name,
+		   lockres->l_level,
+		   lockres->l_flags,
+		   lockres->l_action,
+		   lockres->l_unlock_action,
+		   lockres->l_ro_holders,
+		   lockres->l_ex_holders,
+		   lockres->l_requested,
+		   lockres->l_blocking);
+
+	/* Dump the raw LVB */
+	lvb = lockres->l_lksb.lvb;
+	for(i = 0; i < DLM_LVB_LEN; i++)
+		seq_printf(m, "0x%x\t", lvb[i]);
+
+	/* End the line */
+	seq_printf(m, "\n");
+	return 0;
+}
+
+static struct seq_operations ocfs2_dlm_seq_ops = {
+	.start =	ocfs2_dlm_seq_start,
+	.stop =		ocfs2_dlm_seq_stop,
+	.next =		ocfs2_dlm_seq_next,
+	.show =		ocfs2_dlm_seq_show,
+};
+
+static int ocfs2_dlm_debug_release(struct inode *inode, struct file *file)
+{
+	struct seq_file *seq = (struct seq_file *) file->private_data;
+	struct ocfs2_dlm_seq_priv *priv = seq->private;
+	struct ocfs2_lock_res *res = &priv->p_iter_res;
+
+	ocfs2_remove_lockres_tracking(res);
+	ocfs2_put_dlm_debug(priv->p_dlm_debug);
+	return seq_release_private(inode, file);
+}
+
+static int ocfs2_dlm_debug_open(struct inode *inode, struct file *file)
+{
+	int ret;
+	struct ocfs2_dlm_seq_priv *priv;
+	struct seq_file *seq;
+	struct ocfs2_super *osb;
+
+	priv = kzalloc(sizeof(struct ocfs2_dlm_seq_priv), GFP_KERNEL);
+	if (!priv) {
+		ret = -ENOMEM;
+		mlog_errno(ret);
+		goto out;
+	}
+	osb = (struct ocfs2_super *) inode->u.generic_ip;
+	ocfs2_get_dlm_debug(osb->osb_dlm_debug);
+	priv->p_dlm_debug = osb->osb_dlm_debug;
+	INIT_LIST_HEAD(&priv->p_iter_res.l_debug_list);
+
+	ret = seq_open(file, &ocfs2_dlm_seq_ops);
+	if (ret) {
+		kfree(priv);
+		mlog_errno(ret);
+		goto out;
+	}
+
+	seq = (struct seq_file *) file->private_data;
+	seq->private = priv;
+
+	ocfs2_add_lockres_tracking(&priv->p_iter_res,
+				   priv->p_dlm_debug);
+
+out:
+	return ret;
+}
+
+static struct file_operations ocfs2_dlm_debug_fops = {
+	.open =		ocfs2_dlm_debug_open,
+	.release =	ocfs2_dlm_debug_release,
+	.read =		seq_read,
+	.llseek =	seq_lseek,
+};
+
+static int ocfs2_dlm_init_debug(struct ocfs2_super *osb)
+{
+	int ret = 0;
+	struct ocfs2_dlm_debug *dlm_debug = osb->osb_dlm_debug;
+
+	dlm_debug->d_locking_state = debugfs_create_file("locking_state",
+							 S_IFREG|S_IRUSR,
+							 osb->osb_debug_root,
+							 osb,
+							 &ocfs2_dlm_debug_fops);
+	if (!dlm_debug->d_locking_state) {
+		ret = -EINVAL;
+		mlog(ML_ERROR,
+		     "Unable to create locking state debugfs file.\n");
+		goto out;
+	}
+
+	ocfs2_get_dlm_debug(dlm_debug);
+out:
+	return ret;
+}
+
+static void ocfs2_dlm_shutdown_debug(struct ocfs2_super *osb)
+{
+	struct ocfs2_dlm_debug *dlm_debug = osb->osb_dlm_debug;
+
+	if (dlm_debug) {
+		debugfs_remove(dlm_debug->d_locking_state);
+		ocfs2_put_dlm_debug(dlm_debug);
+	}
+}
+
+int ocfs2_dlm_init(struct ocfs2_super *osb)
+{
+	int status;
+	u32 dlm_key;
+	struct dlm_ctxt *dlm;
+
+	mlog_entry_void();
+
+	status = ocfs2_dlm_init_debug(osb);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* launch vote thread */
+	osb->vote_task = kthread_run(ocfs2_vote_thread, osb, "ocfs2vote-%d",
+				     osb->osb_id);
+	if (IS_ERR(osb->vote_task)) {
+		status = PTR_ERR(osb->vote_task);
+		osb->vote_task = NULL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* used by the dlm code to make message headers unique, each
+	 * node in this domain must agree on this. */
+	dlm_key = crc32_le(0, osb->uuid_str, strlen(osb->uuid_str));
+
+	/* for now, uuid == domain */
+	dlm = dlm_register_domain(osb->uuid_str, dlm_key);
+	if (IS_ERR(dlm)) {
+		status = PTR_ERR(dlm);
+		mlog_errno(status);
+		goto bail;
+	}
+
+	ocfs2_super_lock_res_init(&osb->osb_super_lockres, osb);
+	ocfs2_rename_lock_res_init(&osb->osb_rename_lockres, osb);
+
+	dlm_register_eviction_cb(dlm, &osb->osb_eviction_cb);
+
+	osb->dlm = dlm;
+
+	status = 0;
+bail:
+	if (status < 0) {
+		ocfs2_dlm_shutdown_debug(osb);
+		if (osb->vote_task)
+			kthread_stop(osb->vote_task);
+	}
+
+	mlog_exit(status);
+	return status;
+}
+
+void ocfs2_dlm_shutdown(struct ocfs2_super *osb)
+{
+	mlog_entry_void();
+
+	dlm_unregister_eviction_cb(&osb->osb_eviction_cb);
+
+	ocfs2_drop_osb_locks(osb);
+
+	if (osb->vote_task) {
+		kthread_stop(osb->vote_task);
+		osb->vote_task = NULL;
+	}
+
+	ocfs2_lock_res_free(&osb->osb_super_lockres);
+	ocfs2_lock_res_free(&osb->osb_rename_lockres);
+
+	dlm_unregister_domain(osb->dlm);
+	osb->dlm = NULL;
+
+	ocfs2_dlm_shutdown_debug(osb);
+
+	mlog_exit_void();
+}
+
+static void ocfs2_unlock_ast_func(void *opaque, enum dlm_status status)
+{
+	struct ocfs2_lock_res *lockres = opaque;
+	unsigned long flags;
+
+	mlog_entry_void();
+
+	mlog(0, "UNLOCK AST called on lock %s, action = %d\n", lockres->l_name,
+	     lockres->l_unlock_action);
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+	/* We tried to cancel a convert request, but it was already
+	 * granted. All we want to do here is clear our unlock
+	 * state. The wake_up call done at the bottom is redundant
+	 * (ocfs2_prepare_cancel_convert doesn't sleep on this) but doesn't
+	 * hurt anything anyway */
+	if (status == DLM_CANCELGRANT &&
+	    lockres->l_unlock_action == OCFS2_UNLOCK_CANCEL_CONVERT) {
+		mlog(0, "Got cancelgrant for %s\n", lockres->l_name);
+
+		/* We don't clear the busy flag in this case as it
+		 * should have been cleared by the ast which the dlm
+		 * has called. */
+		goto complete_unlock;
+	}
+
+	if (status != DLM_NORMAL) {
+		mlog(ML_ERROR, "Dlm passes status %d for lock %s, "
+		     "unlock_action %d\n", status, lockres->l_name,
+		     lockres->l_unlock_action);
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+		return;
+	}
+
+	switch(lockres->l_unlock_action) {
+	case OCFS2_UNLOCK_CANCEL_CONVERT:
+		mlog(0, "Cancel convert success for %s\n", lockres->l_name);
+		lockres->l_action = OCFS2_AST_INVALID;
+		break;
+	case OCFS2_UNLOCK_DROP_LOCK:
+		lockres->l_level = LKM_IVMODE;
+		break;
+	default:
+		BUG();
+	}
+
+	lockres_clear_flags(lockres, OCFS2_LOCK_BUSY);
+complete_unlock:
+	lockres->l_unlock_action = OCFS2_UNLOCK_INVALID;
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+	wake_up(&lockres->l_event);
+
+	mlog_exit_void();
+}
+
+typedef void (ocfs2_pre_drop_cb_t)(struct ocfs2_lock_res *, void *);
+
+struct drop_lock_cb {
+	ocfs2_pre_drop_cb_t	*drop_func;
+	void			*drop_data;
+};
+
+static int ocfs2_drop_lock(struct ocfs2_super *osb,
+			   struct ocfs2_lock_res *lockres,
+			   struct drop_lock_cb *dcb)
+{
+	enum dlm_status status;
+	unsigned long flags;
+
+	/* We didn't get anywhere near actually using this lockres. */
+	if (!(lockres->l_flags & OCFS2_LOCK_INITIALIZED))
+		goto out;
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+
+	mlog_bug_on_msg(!(lockres->l_flags & OCFS2_LOCK_FREEING),
+			"lockres %s, flags 0x%lx\n",
+			lockres->l_name, lockres->l_flags);
+
+	while (lockres->l_flags & OCFS2_LOCK_BUSY) {
+		mlog(0, "waiting on busy lock \"%s\": flags = %lx, action = "
+		     "%u, unlock_action = %u\n",
+		     lockres->l_name, lockres->l_flags, lockres->l_action,
+		     lockres->l_unlock_action);
+
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+		/* XXX: Today we just wait on any busy
+		 * locks... Perhaps we need to cancel converts in the
+		 * future? */
+		ocfs2_wait_on_busy_lock(lockres);
+
+		spin_lock_irqsave(&lockres->l_lock, flags);
+	}
+
+	if (dcb)
+		dcb->drop_func(lockres, dcb->drop_data);
+
+	if (lockres->l_flags & OCFS2_LOCK_BUSY)
+		mlog(ML_ERROR, "destroying busy lock: \"%s\"\n",
+		     lockres->l_name);
+	if (lockres->l_flags & OCFS2_LOCK_BLOCKED)
+		mlog(0, "destroying blocked lock: \"%s\"\n", lockres->l_name);
+
+	if (!(lockres->l_flags & OCFS2_LOCK_ATTACHED)) {
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+		goto out;
+	}
+
+	lockres_clear_flags(lockres, OCFS2_LOCK_ATTACHED);
+
+	/* make sure we never get here while waiting for an ast to
+	 * fire. */
+	BUG_ON(lockres->l_action != OCFS2_AST_INVALID);
+
+	/* is this necessary? */
+	lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
+	lockres->l_unlock_action = OCFS2_UNLOCK_DROP_LOCK;
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+	mlog(0, "lock %s\n", lockres->l_name);
+
+	status = dlmunlock(osb->dlm, &lockres->l_lksb, LKM_VALBLK,
+			   lockres->l_ops->unlock_ast, lockres);
+	if (status != DLM_NORMAL) {
+		ocfs2_log_dlm_error("dlmunlock", status, lockres);
+		mlog(ML_ERROR, "lockres flags: %lu\n", lockres->l_flags);
+		dlm_print_one_lock(lockres->l_lksb.lockid);
+		BUG();
+	}
+	mlog(0, "lock %s, successfull return from dlmunlock\n",
+	     lockres->l_name);
+
+	ocfs2_wait_on_busy_lock(lockres);
+out:
+	mlog_exit(0);
+	return 0;
+}
+
+/* Mark the lockres as being dropped. It will no longer be
+ * queued if blocking, but we still may have to wait on it
+ * being dequeued from the vote thread before we can consider
+ * it safe to drop. 
+ *
+ * You can *not* attempt to call cluster_lock on this lockres anymore. */
+void ocfs2_mark_lockres_freeing(struct ocfs2_lock_res *lockres)
+{
+	int status;
+	struct ocfs2_mask_waiter mw;
+	unsigned long flags;
+
+	ocfs2_init_mask_waiter(&mw);
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+	lockres->l_flags |= OCFS2_LOCK_FREEING;
+	while (lockres->l_flags & OCFS2_LOCK_QUEUED) {
+		lockres_add_mask_waiter(lockres, &mw, OCFS2_LOCK_QUEUED, 0);
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+		mlog(0, "Waiting on lockres %s\n", lockres->l_name);
+
+		status = ocfs2_wait_for_mask(&mw);
+		if (status)
+			mlog_errno(status);
+
+		spin_lock_irqsave(&lockres->l_lock, flags);
+	}
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+}
+
+static void ocfs2_drop_osb_locks(struct ocfs2_super *osb)
+{
+	int status;
+
+	mlog_entry_void();
+
+	ocfs2_mark_lockres_freeing(&osb->osb_super_lockres);
+
+	status = ocfs2_drop_lock(osb, &osb->osb_super_lockres, NULL);
+	if (status < 0)
+		mlog_errno(status);
+
+	ocfs2_mark_lockres_freeing(&osb->osb_rename_lockres);
+
+	status = ocfs2_drop_lock(osb, &osb->osb_rename_lockres, NULL);
+	if (status < 0)
+		mlog_errno(status);
+
+	mlog_exit(status);
+}
+
+static void ocfs2_meta_pre_drop(struct ocfs2_lock_res *lockres, void *data)
+{
+	struct inode *inode = data;
+
+	/* the metadata lock requires a bit more work as we have an
+	 * LVB to worry about. */
+	if (lockres->l_flags & OCFS2_LOCK_ATTACHED &&
+	    lockres->l_level == LKM_EXMODE &&
+	    !(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH))
+		__ocfs2_stuff_meta_lvb(inode);
+}
+
+int ocfs2_drop_inode_locks(struct inode *inode)
+{
+	int status, err;
+	struct drop_lock_cb meta_dcb = { ocfs2_meta_pre_drop, inode, };
+
+	mlog_entry_void();
+
+	/* No need to call ocfs2_mark_lockres_freeing here -
+	 * ocfs2_clear_inode has done it for us. */
+
+	err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb),
+			      &OCFS2_I(inode)->ip_data_lockres,
+			      NULL);
+	if (err < 0)
+		mlog_errno(err);
+
+	status = err;
+
+	err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb),
+			      &OCFS2_I(inode)->ip_meta_lockres,
+			      &meta_dcb);
+	if (err < 0)
+		mlog_errno(err);
+	if (err < 0 && !status)
+		status = err;
+
+	err = ocfs2_drop_lock(OCFS2_SB(inode->i_sb),
+			      &OCFS2_I(inode)->ip_rw_lockres,
+			      NULL);
+	if (err < 0)
+		mlog_errno(err);
+	if (err < 0 && !status)
+		status = err;
+
+	mlog_exit(status);
+	return status;
+}
+
+static void ocfs2_prepare_downconvert(struct ocfs2_lock_res *lockres,
+				      int new_level)
+{
+	assert_spin_locked(&lockres->l_lock);
+
+	BUG_ON(lockres->l_blocking <= LKM_NLMODE);
+
+	if (lockres->l_level <= new_level) {
+		mlog(ML_ERROR, "lockres->l_level (%u) <= new_level (%u)\n",
+		     lockres->l_level, new_level);
+		BUG();
+	}
+
+	mlog(0, "lock %s, new_level = %d, l_blocking = %d\n",
+	     lockres->l_name, new_level, lockres->l_blocking);
+
+	lockres->l_action = OCFS2_AST_DOWNCONVERT;
+	lockres->l_requested = new_level;
+	lockres_or_flags(lockres, OCFS2_LOCK_BUSY);
+}
+
+static int ocfs2_downconvert_lock(struct ocfs2_super *osb,
+				  struct ocfs2_lock_res *lockres,
+				  int new_level,
+				  int lvb)
+{
+	int ret, dlm_flags = LKM_CONVERT;
+	enum dlm_status status;
+
+	mlog_entry_void();
+
+	if (lvb)
+		dlm_flags |= LKM_VALBLK;
+
+	status = dlmlock(osb->dlm,
+			 new_level,
+			 &lockres->l_lksb,
+			 dlm_flags,
+			 lockres->l_name,
+			 lockres->l_ops->ast,
+			 lockres,
+			 lockres->l_ops->bast);
+	if (status != DLM_NORMAL) {
+		ocfs2_log_dlm_error("dlmlock", status, lockres);
+		ret = -EINVAL;
+		ocfs2_recover_from_dlm_error(lockres, 1);
+		goto bail;
+	}
+
+	ret = 0;
+bail:
+	mlog_exit(ret);
+	return ret;
+}
+
+/* returns 1 when the caller should unlock and call dlmunlock */
+static int ocfs2_prepare_cancel_convert(struct ocfs2_super *osb,
+				        struct ocfs2_lock_res *lockres)
+{
+	assert_spin_locked(&lockres->l_lock);
+
+	mlog_entry_void();
+	mlog(0, "lock %s\n", lockres->l_name);
+
+	if (lockres->l_unlock_action == OCFS2_UNLOCK_CANCEL_CONVERT) {
+		/* If we're already trying to cancel a lock conversion
+		 * then just drop the spinlock and allow the caller to
+		 * requeue this lock. */
+
+		mlog(0, "Lockres %s, skip convert\n", lockres->l_name);
+		return 0;
+	}
+
+	/* were we in a convert when we got the bast fire? */
+	BUG_ON(lockres->l_action != OCFS2_AST_CONVERT &&
+	       lockres->l_action != OCFS2_AST_DOWNCONVERT);
+	/* set things up for the unlockast to know to just
+	 * clear out the ast_action and unset busy, etc. */
+	lockres->l_unlock_action = OCFS2_UNLOCK_CANCEL_CONVERT;
+
+	mlog_bug_on_msg(!(lockres->l_flags & OCFS2_LOCK_BUSY),
+			"lock %s, invalid flags: 0x%lx\n",
+			lockres->l_name, lockres->l_flags);
+
+	return 1;
+}
+
+static int ocfs2_cancel_convert(struct ocfs2_super *osb,
+				struct ocfs2_lock_res *lockres)
+{
+	int ret;
+	enum dlm_status status;
+
+	mlog_entry_void();
+	mlog(0, "lock %s\n", lockres->l_name);
+
+	ret = 0;
+	status = dlmunlock(osb->dlm,
+			   &lockres->l_lksb,
+			   LKM_CANCEL,
+			   lockres->l_ops->unlock_ast,
+			   lockres);
+	if (status != DLM_NORMAL) {
+		ocfs2_log_dlm_error("dlmunlock", status, lockres);
+		ret = -EINVAL;
+		ocfs2_recover_from_dlm_error(lockres, 0);
+	}
+
+	mlog(0, "lock %s return from dlmunlock\n", lockres->l_name);
+
+	mlog_exit(ret);
+	return ret;
+}
+
+static inline int ocfs2_can_downconvert_meta_lock(struct inode *inode,
+						  struct ocfs2_lock_res *lockres,
+						  int new_level)
+{
+	int ret;
+
+	mlog_entry_void();
+
+	BUG_ON(new_level != LKM_NLMODE && new_level != LKM_PRMODE);
+
+	if (lockres->l_flags & OCFS2_LOCK_REFRESHING) {
+		ret = 0;
+		mlog(0, "lockres %s currently being refreshed -- backing "
+		     "off!\n", lockres->l_name);
+	} else if (new_level == LKM_PRMODE)
+		ret = !lockres->l_ex_holders &&
+			ocfs2_inode_fully_checkpointed(inode);
+	else /* Must be NLMODE we're converting to. */
+		ret = !lockres->l_ro_holders && !lockres->l_ex_holders &&
+			ocfs2_inode_fully_checkpointed(inode);
+
+	mlog_exit(ret);
+	return ret;
+}
+
+static int ocfs2_do_unblock_meta(struct inode *inode,
+				 int *requeue)
+{
+	int new_level;
+	int set_lvb = 0;
+	int ret = 0;
+	struct ocfs2_lock_res *lockres = &OCFS2_I(inode)->ip_meta_lockres;
+	unsigned long flags;
+
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	mlog_entry_void();
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+
+	BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED));
+
+	mlog(0, "l_level=%d, l_blocking=%d\n", lockres->l_level,
+	     lockres->l_blocking);
+
+	BUG_ON(lockres->l_level != LKM_EXMODE &&
+	       lockres->l_level != LKM_PRMODE);
+
+	if (lockres->l_flags & OCFS2_LOCK_BUSY) {
+		*requeue = 1;
+		ret = ocfs2_prepare_cancel_convert(osb, lockres);
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+		if (ret) {
+			ret = ocfs2_cancel_convert(osb, lockres);
+			if (ret < 0)
+				mlog_errno(ret);
+		}
+		goto leave;
+	}
+
+	new_level = ocfs2_highest_compat_lock_level(lockres->l_blocking);
+
+	mlog(0, "l_level=%d, l_blocking=%d, new_level=%d\n",
+	     lockres->l_level, lockres->l_blocking, new_level);
+
+	if (ocfs2_can_downconvert_meta_lock(inode, lockres, new_level)) {
+		if (lockres->l_level == LKM_EXMODE)
+			set_lvb = 1;
+
+		/* If the lock hasn't been refreshed yet (rare), then
+		 * our memory inode values are old and we skip
+		 * stuffing the lvb. There's no need to actually clear
+		 * out the lvb here as it's value is still valid. */
+		if (!(lockres->l_flags & OCFS2_LOCK_NEEDS_REFRESH)) {
+			if (set_lvb)
+				__ocfs2_stuff_meta_lvb(inode);
+		} else
+			mlog(0, "lockres %s: downconverting stale lock!\n",
+			     lockres->l_name);
+
+		mlog(0, "calling ocfs2_downconvert_lock with l_level=%d, "
+		     "l_blocking=%d, new_level=%d\n",
+		     lockres->l_level, lockres->l_blocking, new_level);
+
+		ocfs2_prepare_downconvert(lockres, new_level);
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+		ret = ocfs2_downconvert_lock(osb, lockres, new_level, set_lvb);
+		goto leave;
+	}
+	if (!ocfs2_inode_fully_checkpointed(inode))
+		ocfs2_start_checkpoint(osb);
+
+	*requeue = 1;
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+	ret = 0;
+leave:
+	mlog_exit(ret);
+	return ret;
+}
+
+static int ocfs2_generic_unblock_lock(struct ocfs2_super *osb,
+				      struct ocfs2_lock_res *lockres,
+				      int *requeue,
+				      ocfs2_convert_worker_t *worker)
+{
+	unsigned long flags;
+	int blocking;
+	int new_level;
+	int ret = 0;
+
+	mlog_entry_void();
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+
+	BUG_ON(!(lockres->l_flags & OCFS2_LOCK_BLOCKED));
+
+recheck:
+	if (lockres->l_flags & OCFS2_LOCK_BUSY) {
+		*requeue = 1;
+		ret = ocfs2_prepare_cancel_convert(osb, lockres);
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+		if (ret) {
+			ret = ocfs2_cancel_convert(osb, lockres);
+			if (ret < 0)
+				mlog_errno(ret);
+		}
+		goto leave;
+	}
+
+	/* if we're blocking an exclusive and we have *any* holders,
+	 * then requeue. */
+	if ((lockres->l_blocking == LKM_EXMODE)
+	    && (lockres->l_ex_holders || lockres->l_ro_holders)) {
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+		*requeue = 1;
+		ret = 0;
+		goto leave;
+	}
+
+	/* If it's a PR we're blocking, then only
+	 * requeue if we've got any EX holders */
+	if (lockres->l_blocking == LKM_PRMODE &&
+	    lockres->l_ex_holders) {
+		spin_unlock_irqrestore(&lockres->l_lock, flags);
+		*requeue = 1;
+		ret = 0;
+		goto leave;
+	}
+
+	/* If we get here, then we know that there are no more
+	 * incompatible holders (and anyone asking for an incompatible
+	 * lock is blocked). We can now downconvert the lock */
+	if (!worker)
+		goto downconvert;
+
+	/* Some lockres types want to do a bit of work before
+	 * downconverting a lock. Allow that here. The worker function
+	 * may sleep, so we save off a copy of what we're blocking as
+	 * it may change while we're not holding the spin lock. */
+	blocking = lockres->l_blocking;
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+	worker(lockres, blocking);
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+	if (blocking != lockres->l_blocking) {
+		/* If this changed underneath us, then we can't drop
+		 * it just yet. */
+		goto recheck;
+	}
+
+downconvert:
+	*requeue = 0;
+	new_level = ocfs2_highest_compat_lock_level(lockres->l_blocking);
+
+	ocfs2_prepare_downconvert(lockres, new_level);
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+	ret = ocfs2_downconvert_lock(osb, lockres, new_level, 0);
+leave:
+	mlog_exit(ret);
+	return ret;
+}
+
+static void ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres,
+				      int blocking)
+{
+	struct inode *inode;
+	struct address_space *mapping;
+
+	mlog_entry_void();
+
+       	inode = ocfs2_lock_res_inode(lockres);
+	mapping = inode->i_mapping;
+
+	if (filemap_fdatawrite(mapping)) {
+		mlog(ML_ERROR, "Could not sync inode %"MLFu64" for downconvert!",
+		     OCFS2_I(inode)->ip_blkno);
+	}
+	sync_mapping_buffers(mapping);
+	if (blocking == LKM_EXMODE) {
+		truncate_inode_pages(mapping, 0);
+		unmap_mapping_range(mapping, 0, 0, 0);
+	} else {
+		/* We only need to wait on the I/O if we're not also
+		 * truncating pages because truncate_inode_pages waits
+		 * for us above. We don't truncate pages if we're
+		 * blocking anything < EXMODE because we want to keep
+		 * them around in that case. */
+		filemap_fdatawait(mapping);
+	}
+
+	mlog_exit_void();
+}
+
+int ocfs2_unblock_data(struct ocfs2_lock_res *lockres,
+		       int *requeue)
+{
+	int status;
+	struct inode *inode;
+	struct ocfs2_super *osb;
+
+	mlog_entry_void();
+
+	inode = ocfs2_lock_res_inode(lockres);
+	osb = OCFS2_SB(inode->i_sb);
+
+	mlog(0, "unblock inode %"MLFu64"\n", OCFS2_I(inode)->ip_blkno);
+
+	status = ocfs2_generic_unblock_lock(osb,
+					    lockres,
+					    requeue,
+					    ocfs2_data_convert_worker);
+	if (status < 0)
+		mlog_errno(status);
+
+	mlog(0, "inode %"MLFu64", requeue = %d\n",
+	     OCFS2_I(inode)->ip_blkno, *requeue);
+
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_unblock_inode_lock(struct ocfs2_lock_res *lockres,
+				    int *requeue)
+{
+	int status;
+	struct inode *inode;
+
+	mlog_entry_void();
+
+	mlog(0, "Unblock lockres %s\n", lockres->l_name);
+
+	inode  = ocfs2_lock_res_inode(lockres);
+
+	status = ocfs2_generic_unblock_lock(OCFS2_SB(inode->i_sb),
+					    lockres,
+					    requeue,
+					    NULL);
+	if (status < 0)
+		mlog_errno(status);
+
+	mlog_exit(status);
+	return status;
+}
+
+
+int ocfs2_unblock_meta(struct ocfs2_lock_res *lockres,
+		       int *requeue)
+{
+	int status;
+	struct inode *inode;
+
+	mlog_entry_void();
+
+       	inode = ocfs2_lock_res_inode(lockres);
+
+	mlog(0, "unblock inode %"MLFu64"\n", OCFS2_I(inode)->ip_blkno);
+
+	status = ocfs2_do_unblock_meta(inode, requeue);
+	if (status < 0)
+		mlog_errno(status);
+
+	mlog(0, "inode %"MLFu64", requeue = %d\n",
+	     OCFS2_I(inode)->ip_blkno, *requeue);
+
+	mlog_exit(status);
+	return status;
+}
+
+/* Generic unblock function for any lockres whose private data is an
+ * ocfs2_super pointer. */
+static int ocfs2_unblock_osb_lock(struct ocfs2_lock_res *lockres,
+				  int *requeue)
+{
+	int status;
+	struct ocfs2_super *osb;
+
+	mlog_entry_void();
+
+	mlog(0, "Unblock lockres %s\n", lockres->l_name);
+
+	osb = ocfs2_lock_res_super(lockres);
+
+	status = ocfs2_generic_unblock_lock(osb,
+					    lockres,
+					    requeue,
+					    NULL);
+	if (status < 0)
+		mlog_errno(status);
+
+	mlog_exit(status);
+	return status;
+}
+
+void ocfs2_process_blocked_lock(struct ocfs2_super *osb,
+				struct ocfs2_lock_res *lockres)
+{
+	int status;
+	int requeue = 0;
+	unsigned long flags;
+
+	/* Our reference to the lockres in this function can be
+	 * considered valid until we remove the OCFS2_LOCK_QUEUED
+	 * flag. */
+
+	mlog_entry_void();
+
+	BUG_ON(!lockres);
+	BUG_ON(!lockres->l_ops);
+	BUG_ON(!lockres->l_ops->unblock);
+
+	mlog(0, "lockres %s blocked.\n", lockres->l_name);
+
+	/* Detect whether a lock has been marked as going away while
+	 * the vote thread was processing other things. A lock can
+	 * still be marked with OCFS2_LOCK_FREEING after this check,
+	 * but short circuiting here will still save us some
+	 * performance. */
+	spin_lock_irqsave(&lockres->l_lock, flags);
+	if (lockres->l_flags & OCFS2_LOCK_FREEING)
+		goto unqueue;
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+	status = lockres->l_ops->unblock(lockres, &requeue);
+	if (status < 0)
+		mlog_errno(status);
+
+	spin_lock_irqsave(&lockres->l_lock, flags);
+unqueue:
+	if (lockres->l_flags & OCFS2_LOCK_FREEING || !requeue) {
+		lockres_clear_flags(lockres, OCFS2_LOCK_QUEUED);
+	} else
+		ocfs2_schedule_blocked_lock(osb, lockres);
+
+	mlog(0, "lockres %s, requeue = %s.\n", lockres->l_name,
+	     requeue ? "yes" : "no");
+	spin_unlock_irqrestore(&lockres->l_lock, flags);
+
+	mlog_exit_void();
+}
+
+static void ocfs2_schedule_blocked_lock(struct ocfs2_super *osb,
+					struct ocfs2_lock_res *lockres)
+{
+	mlog_entry_void();
+
+	assert_spin_locked(&lockres->l_lock);
+
+	if (lockres->l_flags & OCFS2_LOCK_FREEING) {
+		/* Do not schedule a lock for downconvert when it's on
+		 * the way to destruction - any nodes wanting access
+		 * to the resource will get it soon. */
+		mlog(0, "Lockres %s won't be scheduled: flags 0x%lx\n",
+		     lockres->l_name, lockres->l_flags);
+		return;
+	}
+
+	lockres_or_flags(lockres, OCFS2_LOCK_QUEUED);
+
+	spin_lock(&osb->vote_task_lock);
+	if (list_empty(&lockres->l_blocked_list)) {
+		list_add_tail(&lockres->l_blocked_list,
+			      &osb->blocked_lock_list);
+		osb->blocked_lock_count++;
+	}
+	spin_unlock(&osb->vote_task_lock);
+
+	mlog_exit_void();
+}
+
+/* This aids in debugging situations where a bad LVB might be involved. */
+void ocfs2_dump_meta_lvb_info(u64 level,
+			      const char *function,
+			      unsigned int line,
+			      struct ocfs2_lock_res *lockres)
+{
+	struct ocfs2_meta_lvb *lvb = (struct ocfs2_meta_lvb *) lockres->l_lksb.lvb;
+
+	mlog(level, "LVB information for %s (called from %s:%u):\n",
+	     lockres->l_name, function, line);
+	mlog(level, "version: %u, clusters: %u\n",
+	     be32_to_cpu(lvb->lvb_version), be32_to_cpu(lvb->lvb_iclusters));
+	mlog(level, "size: %"MLFu64", uid %u, gid %u, mode 0x%x\n",
+	     be64_to_cpu(lvb->lvb_isize), be32_to_cpu(lvb->lvb_iuid),
+	     be32_to_cpu(lvb->lvb_igid), be16_to_cpu(lvb->lvb_imode));
+	mlog(level, "nlink %u, atime_packed 0x%"MLFx64", "
+	     "ctime_packed 0x%"MLFx64", mtime_packed 0x%"MLFx64"\n",
+	     be16_to_cpu(lvb->lvb_inlink), be64_to_cpu(lvb->lvb_iatime_packed),
+	     be64_to_cpu(lvb->lvb_ictime_packed),
+	     be64_to_cpu(lvb->lvb_imtime_packed));
+}
diff --git a/fs/ocfs2/dlmglue.h b/fs/ocfs2/dlmglue.h
new file mode 100644
index 00000000000..8f2d1db2d9e
--- /dev/null
+++ b/fs/ocfs2/dlmglue.h
@@ -0,0 +1,111 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * dlmglue.h
+ *
+ * description here
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+
+#ifndef DLMGLUE_H
+#define DLMGLUE_H
+
+#define OCFS2_LVB_VERSION 2
+
+struct ocfs2_meta_lvb {
+	__be32       lvb_version;
+	__be32       lvb_iclusters;
+	__be32       lvb_iuid;
+	__be32       lvb_igid;
+	__be64       lvb_iatime_packed;
+	__be64       lvb_ictime_packed;
+	__be64       lvb_imtime_packed;
+	__be64       lvb_isize;
+	__be16       lvb_imode;
+	__be16       lvb_inlink;
+	__be32       lvb_reserved[3];
+};
+
+/* ocfs2_meta_lock_full() and ocfs2_data_lock_full() 'arg_flags' flags */
+/* don't wait on recovery. */
+#define OCFS2_META_LOCK_RECOVERY	(0x01)
+/* Instruct the dlm not to queue ourselves on the other node. */
+#define OCFS2_META_LOCK_NOQUEUE		(0x02)
+/* don't block waiting for the vote thread, instead return -EAGAIN */
+#define OCFS2_LOCK_NONBLOCK		(0x04)
+
+int ocfs2_dlm_init(struct ocfs2_super *osb);
+void ocfs2_dlm_shutdown(struct ocfs2_super *osb);
+void ocfs2_lock_res_init_once(struct ocfs2_lock_res *res);
+void ocfs2_inode_lock_res_init(struct ocfs2_lock_res *res,
+			       enum ocfs2_lock_type type,
+			       struct inode *inode);
+void ocfs2_lock_res_free(struct ocfs2_lock_res *res);
+int ocfs2_create_new_inode_locks(struct inode *inode);
+int ocfs2_drop_inode_locks(struct inode *inode);
+int ocfs2_data_lock_full(struct inode *inode,
+			 int write,
+			 int arg_flags);
+#define ocfs2_data_lock(inode, write) ocfs2_data_lock_full(inode, write, 0)
+int ocfs2_data_lock_with_page(struct inode *inode,
+			      int write,
+			      struct page *page);
+void ocfs2_data_unlock(struct inode *inode,
+		       int write);
+int ocfs2_rw_lock(struct inode *inode, int write);
+void ocfs2_rw_unlock(struct inode *inode, int write);
+int ocfs2_meta_lock_full(struct inode *inode,
+			 struct ocfs2_journal_handle *handle,
+			 struct buffer_head **ret_bh,
+			 int ex,
+			 int arg_flags);
+int ocfs2_meta_lock_with_page(struct inode *inode,
+			      struct ocfs2_journal_handle *handle,
+			      struct buffer_head **ret_bh,
+			      int ex,
+			      struct page *page);
+/* 99% of the time we don't want to supply any additional flags --
+ * those are for very specific cases only. */
+#define ocfs2_meta_lock(i, h, b, e) ocfs2_meta_lock_full(i, h, b, e, 0)
+void ocfs2_meta_unlock(struct inode *inode,
+		       int ex);
+int ocfs2_super_lock(struct ocfs2_super *osb,
+		     int ex);
+void ocfs2_super_unlock(struct ocfs2_super *osb,
+			int ex);
+int ocfs2_rename_lock(struct ocfs2_super *osb);
+void ocfs2_rename_unlock(struct ocfs2_super *osb);
+void ocfs2_mark_lockres_freeing(struct ocfs2_lock_res *lockres);
+
+/* for the vote thread */
+void ocfs2_process_blocked_lock(struct ocfs2_super *osb,
+				struct ocfs2_lock_res *lockres);
+
+struct ocfs2_dlm_debug *ocfs2_new_dlm_debug(void);
+void ocfs2_put_dlm_debug(struct ocfs2_dlm_debug *dlm_debug);
+
+/* aids in debugging and tracking lvbs */
+void ocfs2_dump_meta_lvb_info(u64 level,
+			      const char *function,
+			      unsigned int line,
+			      struct ocfs2_lock_res *lockres);
+#define mlog_meta_lvb(__level, __lockres) ocfs2_dump_meta_lvb_info(__level, __PRETTY_FUNCTION__, __LINE__, __lockres)
+
+#endif	/* DLMGLUE_H */
diff --git a/fs/ocfs2/endian.h b/fs/ocfs2/endian.h
new file mode 100644
index 00000000000..f226b220762
--- /dev/null
+++ b/fs/ocfs2/endian.h
@@ -0,0 +1,45 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * Copyright (C) 2005 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_ENDIAN_H
+#define OCFS2_ENDIAN_H
+
+static inline void le16_add_cpu(__le16 *var, u16 val)
+{
+	*var = cpu_to_le16(le16_to_cpu(*var) + val);
+}
+
+static inline void le32_add_cpu(__le32 *var, u32 val)
+{
+	*var = cpu_to_le32(le32_to_cpu(*var) + val);
+}
+
+static inline void le32_and_cpu(__le32 *var, u32 val)
+{
+	*var = cpu_to_le32(le32_to_cpu(*var) & val);
+}
+
+static inline void be32_add_cpu(__be32 *var, u32 val)
+{
+	*var = cpu_to_be32(be32_to_cpu(*var) + val);
+}
+
+#endif /* OCFS2_ENDIAN_H */
diff --git a/fs/ocfs2/export.c b/fs/ocfs2/export.c
new file mode 100644
index 00000000000..5810160d92a
--- /dev/null
+++ b/fs/ocfs2/export.c
@@ -0,0 +1,248 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * export.c
+ *
+ * Functions to facilitate NFS exporting
+ *
+ * Copyright (C) 2002, 2005 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+
+#define MLOG_MASK_PREFIX ML_EXPORT
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "dir.h"
+#include "dlmglue.h"
+#include "export.h"
+#include "inode.h"
+
+#include "buffer_head_io.h"
+
+struct ocfs2_inode_handle
+{
+	u64 ih_blkno;
+	u32 ih_generation;
+};
+
+static struct dentry *ocfs2_get_dentry(struct super_block *sb, void *vobjp)
+{
+	struct ocfs2_inode_handle *handle = vobjp;
+	struct inode *inode;
+	struct dentry *result;
+
+	mlog_entry("(0x%p, 0x%p)\n", sb, handle);
+
+	if (handle->ih_blkno == 0) {
+		mlog_errno(-ESTALE);
+		return ERR_PTR(-ESTALE);
+	}
+
+	inode = ocfs2_iget(OCFS2_SB(sb), handle->ih_blkno);
+
+	if (IS_ERR(inode)) {
+		mlog_errno(PTR_ERR(inode));
+		return (void *)inode;
+	}
+
+	if (handle->ih_generation != inode->i_generation) {
+		iput(inode);
+		mlog_errno(-ESTALE);
+		return ERR_PTR(-ESTALE);
+	}
+
+	result = d_alloc_anon(inode);
+
+	if (!result) {
+		iput(inode);
+		mlog_errno(-ENOMEM);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	mlog_exit_ptr(result);
+	return result;
+}
+
+static struct dentry *ocfs2_get_parent(struct dentry *child)
+{
+	int status;
+	u64 blkno;
+	struct dentry *parent;
+	struct inode *inode;
+	struct inode *dir = child->d_inode;
+	struct buffer_head *dirent_bh = NULL;
+	struct ocfs2_dir_entry *dirent;
+
+	mlog_entry("(0x%p, '%.*s')\n", child,
+		   child->d_name.len, child->d_name.name);
+
+	mlog(0, "find parent of directory %"MLFu64"\n",
+	     OCFS2_I(dir)->ip_blkno);
+
+	status = ocfs2_meta_lock(dir, NULL, NULL, 0);
+	if (status < 0) {
+		if (status != -ENOENT)
+			mlog_errno(status);
+		parent = ERR_PTR(status);
+		goto bail;
+	}
+
+	status = ocfs2_find_files_on_disk("..", 2, &blkno, dir, &dirent_bh,
+					  &dirent);
+	if (status < 0) {
+		parent = ERR_PTR(-ENOENT);
+		goto bail_unlock;
+	}
+
+	inode = ocfs2_iget(OCFS2_SB(dir->i_sb), blkno);
+	if (IS_ERR(inode)) {
+		mlog(ML_ERROR, "Unable to create inode %"MLFu64"\n", blkno);
+		parent = ERR_PTR(-EACCES);
+		goto bail_unlock;
+	}
+
+	parent = d_alloc_anon(inode);
+	if (!parent) {
+		iput(inode);
+		parent = ERR_PTR(-ENOMEM);
+	}
+
+bail_unlock:
+	ocfs2_meta_unlock(dir, 0);
+
+	if (dirent_bh)
+		brelse(dirent_bh);
+
+bail:
+	mlog_exit_ptr(parent);
+
+	return parent;
+}
+
+static int ocfs2_encode_fh(struct dentry *dentry, __be32 *fh, int *max_len,
+			   int connectable)
+{
+	struct inode *inode = dentry->d_inode;
+	int len = *max_len;
+	int type = 1;
+	u64 blkno;
+	u32 generation;
+
+	mlog_entry("(0x%p, '%.*s', 0x%p, %d, %d)\n", dentry,
+		   dentry->d_name.len, dentry->d_name.name,
+		   fh, len, connectable);
+
+	if (len < 3 || (connectable && len < 6)) {
+		mlog(ML_ERROR, "fh buffer is too small for encoding\n");
+		type = 255;
+		goto bail;
+	}
+
+	blkno = OCFS2_I(inode)->ip_blkno;
+	generation = inode->i_generation;
+
+	mlog(0, "Encoding fh: blkno: %"MLFu64", generation: %u\n",
+	     blkno, generation);
+
+	len = 3;
+	fh[0] = cpu_to_le32((u32)(blkno >> 32));
+	fh[1] = cpu_to_le32((u32)(blkno & 0xffffffff));
+	fh[2] = cpu_to_le32(generation);
+
+	if (connectable && !S_ISDIR(inode->i_mode)) {
+		struct inode *parent;
+
+		spin_lock(&dentry->d_lock);
+
+		parent = dentry->d_parent->d_inode;
+		blkno = OCFS2_I(parent)->ip_blkno;
+		generation = parent->i_generation;
+
+		fh[3] = cpu_to_le32((u32)(blkno >> 32));
+		fh[4] = cpu_to_le32((u32)(blkno & 0xffffffff));
+		fh[5] = cpu_to_le32(generation);
+
+		spin_unlock(&dentry->d_lock);
+
+		len = 6;
+		type = 2;
+
+		mlog(0, "Encoding parent: blkno: %"MLFu64", generation: %u\n",
+		     blkno, generation);
+	}
+	
+	*max_len = len;
+
+bail:
+	mlog_exit(type);
+	return type;
+}
+
+static struct dentry *ocfs2_decode_fh(struct super_block *sb, __be32 *fh,
+				      int fh_len, int fileid_type,
+				      int (*acceptable)(void *context,
+						        struct dentry *de),
+				      void *context)
+{
+	struct ocfs2_inode_handle handle, parent;
+	struct dentry *ret = NULL;
+
+	mlog_entry("(0x%p, 0x%p, %d, %d, 0x%p, 0x%p)\n",
+		   sb, fh, fh_len, fileid_type, acceptable, context);
+
+	if (fh_len < 3 || fileid_type > 2)
+		goto bail;
+
+	if (fileid_type == 2) {
+		if (fh_len < 6)
+			goto bail;
+
+		parent.ih_blkno = (u64)le32_to_cpu(fh[3]) << 32;
+		parent.ih_blkno |= (u64)le32_to_cpu(fh[4]);
+		parent.ih_generation = le32_to_cpu(fh[5]);
+
+		mlog(0, "Decoding parent: blkno: %"MLFu64", generation: %u\n",
+		     parent.ih_blkno, parent.ih_generation);
+	}
+
+	handle.ih_blkno = (u64)le32_to_cpu(fh[0]) << 32;
+	handle.ih_blkno |= (u64)le32_to_cpu(fh[1]);
+	handle.ih_generation = le32_to_cpu(fh[2]);
+
+	mlog(0, "Encoding fh: blkno: %"MLFu64", generation: %u\n",
+	     handle.ih_blkno, handle.ih_generation);
+
+	ret = ocfs2_export_ops.find_exported_dentry(sb, &handle, &parent,
+						    acceptable, context);
+
+bail:
+	mlog_exit_ptr(ret);
+	return ret;
+}
+
+struct export_operations ocfs2_export_ops = {
+	.decode_fh	= ocfs2_decode_fh,
+	.encode_fh	= ocfs2_encode_fh,
+
+	.get_parent	= ocfs2_get_parent,
+	.get_dentry	= ocfs2_get_dentry,
+};
diff --git a/fs/ocfs2/export.h b/fs/ocfs2/export.h
new file mode 100644
index 00000000000..5b77ee7866e
--- /dev/null
+++ b/fs/ocfs2/export.h
@@ -0,0 +1,31 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * export.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2005 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_EXPORT_H
+#define OCFS2_EXPORT_H
+
+extern struct export_operations ocfs2_export_ops;
+
+#endif /* OCFS2_EXPORT_H */
diff --git a/fs/ocfs2/extent_map.c b/fs/ocfs2/extent_map.c
new file mode 100644
index 00000000000..f2fb40cd296
--- /dev/null
+++ b/fs/ocfs2/extent_map.c
@@ -0,0 +1,994 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * extent_map.c
+ *
+ * In-memory extent map for OCFS2.  Man, this code was prettier in
+ * the library.
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License, version 2,  as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/init.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/rbtree.h>
+
+#define MLOG_MASK_PREFIX ML_EXTENT_MAP
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "extent_map.h"
+#include "inode.h"
+#include "super.h"
+
+#include "buffer_head_io.h"
+
+
+/*
+ * SUCK SUCK SUCK
+ * Our headers are so bad that struct ocfs2_extent_map is in ocfs.h
+ */
+
+struct ocfs2_extent_map_entry {
+	struct rb_node e_node;
+	int e_tree_depth;
+	struct ocfs2_extent_rec e_rec;
+};
+
+struct ocfs2_em_insert_context {
+	int need_left;
+	int need_right;
+	struct ocfs2_extent_map_entry *new_ent;
+	struct ocfs2_extent_map_entry *old_ent;
+	struct ocfs2_extent_map_entry *left_ent;
+	struct ocfs2_extent_map_entry *right_ent;
+};
+
+static kmem_cache_t *ocfs2_em_ent_cachep = NULL;
+
+
+static struct ocfs2_extent_map_entry *
+ocfs2_extent_map_lookup(struct ocfs2_extent_map *em,
+			u32 cpos, u32 clusters,
+			struct rb_node ***ret_p,
+			struct rb_node **ret_parent);
+static int ocfs2_extent_map_insert(struct inode *inode,
+				   struct ocfs2_extent_rec *rec,
+				   int tree_depth);
+static int ocfs2_extent_map_insert_entry(struct ocfs2_extent_map *em,
+					 struct ocfs2_extent_map_entry *ent);
+static int ocfs2_extent_map_find_leaf(struct inode *inode,
+				      u32 cpos, u32 clusters,
+				      struct ocfs2_extent_list *el);
+static int ocfs2_extent_map_lookup_read(struct inode *inode,
+					u32 cpos, u32 clusters,
+					struct ocfs2_extent_map_entry **ret_ent);
+static int ocfs2_extent_map_try_insert(struct inode *inode,
+				       struct ocfs2_extent_rec *rec,
+				       int tree_depth,
+				       struct ocfs2_em_insert_context *ctxt);
+
+/* returns 1 only if the rec contains all the given clusters -- that is that
+ * rec's cpos is <= the cluster cpos and that the rec endpoint (cpos +
+ * clusters) is >= the argument's endpoint */
+static int ocfs2_extent_rec_contains_clusters(struct ocfs2_extent_rec *rec,
+					      u32 cpos, u32 clusters)
+{
+	if (le32_to_cpu(rec->e_cpos) > cpos)
+		return 0;
+	if (cpos + clusters > le32_to_cpu(rec->e_cpos) + 
+			      le32_to_cpu(rec->e_clusters))
+		return 0;
+	return 1;
+}
+
+
+/*
+ * Find an entry in the tree that intersects the region passed in.
+ * Note that this will find straddled intervals, it is up to the
+ * callers to enforce any boundary conditions.
+ *
+ * Callers must hold ip_lock.  This lookup is not guaranteed to return
+ * a tree_depth 0 match, and as such can race inserts if the lock
+ * were not held.
+ *
+ * The rb_node garbage lets insertion share the search.  Trivial
+ * callers pass NULL.
+ */
+static struct ocfs2_extent_map_entry *
+ocfs2_extent_map_lookup(struct ocfs2_extent_map *em,
+			u32 cpos, u32 clusters,
+			struct rb_node ***ret_p,
+			struct rb_node **ret_parent)
+{
+	struct rb_node **p = &em->em_extents.rb_node;
+	struct rb_node *parent = NULL;
+	struct ocfs2_extent_map_entry *ent = NULL;
+
+	while (*p)
+	{
+		parent = *p;
+		ent = rb_entry(parent, struct ocfs2_extent_map_entry,
+			       e_node);
+		if ((cpos + clusters) <= le32_to_cpu(ent->e_rec.e_cpos)) {
+			p = &(*p)->rb_left;
+			ent = NULL;
+		} else if (cpos >= (le32_to_cpu(ent->e_rec.e_cpos) +
+				    le32_to_cpu(ent->e_rec.e_clusters))) {
+			p = &(*p)->rb_right;
+			ent = NULL;
+		} else
+			break;
+	}
+
+	if (ret_p != NULL)
+		*ret_p = p;
+	if (ret_parent != NULL)
+		*ret_parent = parent;
+	return ent;
+}
+
+/*
+ * Find the leaf containing the interval we want.  While we're on our
+ * way down the tree, fill in every record we see at any depth, because
+ * we might want it later.
+ *
+ * Note that this code is run without ip_lock.  That's because it
+ * sleeps while reading.  If someone is also filling the extent list at
+ * the same time we are, we might have to restart.
+ */
+static int ocfs2_extent_map_find_leaf(struct inode *inode,
+				      u32 cpos, u32 clusters,
+				      struct ocfs2_extent_list *el)
+{
+	int i, ret;
+	struct buffer_head *eb_bh = NULL;
+	u64 blkno;
+	u32 rec_end;
+	struct ocfs2_extent_block *eb;
+	struct ocfs2_extent_rec *rec;
+
+	/*
+	 * The bh data containing the el cannot change here, because
+	 * we hold alloc_sem.  So we can do this without other
+	 * locks.
+	 */
+	while (el->l_tree_depth)
+	{
+		blkno = 0;
+		for (i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) {
+			rec = &el->l_recs[i];
+			rec_end = (le32_to_cpu(rec->e_cpos) +
+				   le32_to_cpu(rec->e_clusters));
+
+			ret = -EBADR;
+			if (rec_end > OCFS2_I(inode)->ip_clusters) {
+				mlog_errno(ret);
+				goto out_free;
+			}
+
+			if (rec_end <= cpos) {
+				ret = ocfs2_extent_map_insert(inode, rec,
+						le16_to_cpu(el->l_tree_depth));
+				if (ret && (ret != -EEXIST)) {
+					mlog_errno(ret);
+					goto out_free;
+				}
+				continue;
+			}
+			if ((cpos + clusters) <= le32_to_cpu(rec->e_cpos)) {
+				ret = ocfs2_extent_map_insert(inode, rec,
+						le16_to_cpu(el->l_tree_depth));
+				if (ret && (ret != -EEXIST)) {
+					mlog_errno(ret);
+					goto out_free;
+				}
+				continue;
+			}
+
+			/*
+			 * We've found a record that matches our
+			 * interval.  We don't insert it because we're
+			 * about to traverse it.
+			 */
+
+			/* Check to see if we're stradling */
+			ret = -ESRCH;
+			if (!ocfs2_extent_rec_contains_clusters(rec,
+							        cpos,
+								clusters)) {
+				mlog_errno(ret);
+				goto out_free;
+			}
+
+			/*
+			 * If we've already found a record, the el has
+			 * two records covering the same interval.
+			 * EEEK!
+			 */
+			ret = -EBADR;
+			if (blkno) {
+				mlog_errno(ret);
+				goto out_free;
+			}
+
+			blkno = le64_to_cpu(rec->e_blkno);
+		}
+
+		/*
+		 * We don't support holes, and we're still up
+		 * in the branches, so we'd better have found someone
+		 */
+		ret = -EBADR;
+		if (!blkno) {
+			mlog_errno(ret);
+			goto out_free;
+		}
+
+		if (eb_bh) {
+			brelse(eb_bh);
+			eb_bh = NULL;
+		}
+		ret = ocfs2_read_block(OCFS2_SB(inode->i_sb),
+				       blkno, &eb_bh, OCFS2_BH_CACHED,
+				       inode);
+		if (ret) {
+			mlog_errno(ret);
+			goto out_free;
+		}
+		eb = (struct ocfs2_extent_block *)eb_bh->b_data;
+		if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
+			OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
+			ret = -EIO;
+			goto out_free;
+		}
+		el = &eb->h_list;
+	}
+
+	if (el->l_tree_depth)
+		BUG();
+
+	for (i = 0; i < le16_to_cpu(el->l_next_free_rec); i++) {
+		rec = &el->l_recs[i];
+		ret = ocfs2_extent_map_insert(inode, rec,
+					      le16_to_cpu(el->l_tree_depth));
+		if (ret) {
+			mlog_errno(ret);
+			goto out_free;
+		}
+	}
+
+	ret = 0;
+
+out_free:
+	if (eb_bh)
+		brelse(eb_bh);
+
+	return ret;
+}
+
+/*
+ * This lookup actually will read from disk.  It has one invariant:
+ * It will never re-traverse blocks.  This means that all inserts should
+ * be new regions or more granular regions (both allowed by insert).
+ */
+static int ocfs2_extent_map_lookup_read(struct inode *inode,
+					u32 cpos,
+					u32 clusters,
+					struct ocfs2_extent_map_entry **ret_ent)
+{
+	int ret;
+	u64 blkno;
+	struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
+	struct ocfs2_extent_map_entry *ent;
+	struct buffer_head *bh = NULL;
+	struct ocfs2_extent_block *eb;
+	struct ocfs2_dinode *di;
+	struct ocfs2_extent_list *el;
+
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+	ent = ocfs2_extent_map_lookup(em, cpos, clusters, NULL, NULL);
+	if (ent) {
+		if (!ent->e_tree_depth) {
+			spin_unlock(&OCFS2_I(inode)->ip_lock);
+			*ret_ent = ent;
+			return 0;
+		}
+		blkno = le64_to_cpu(ent->e_rec.e_blkno);
+		spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+		ret = ocfs2_read_block(OCFS2_SB(inode->i_sb), blkno, &bh,
+				       OCFS2_BH_CACHED, inode);
+		if (ret) {
+			mlog_errno(ret);
+			if (bh)
+				brelse(bh);
+			return ret;
+		}
+		eb = (struct ocfs2_extent_block *)bh->b_data;
+		if (!OCFS2_IS_VALID_EXTENT_BLOCK(eb)) {
+			OCFS2_RO_ON_INVALID_EXTENT_BLOCK(inode->i_sb, eb);
+			brelse(bh);
+			return -EIO;
+		}
+		el = &eb->h_list;
+	} else {
+		spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+		ret = ocfs2_read_block(OCFS2_SB(inode->i_sb),
+				       OCFS2_I(inode)->ip_blkno, &bh,
+				       OCFS2_BH_CACHED, inode);
+		if (ret) {
+			mlog_errno(ret);
+			if (bh)
+				brelse(bh);
+			return ret;
+		}
+		di = (struct ocfs2_dinode *)bh->b_data;
+		if (!OCFS2_IS_VALID_DINODE(di)) {
+			brelse(bh);
+			OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, di);
+			return -EIO;
+		}
+		el = &di->id2.i_list;
+	}
+
+	ret = ocfs2_extent_map_find_leaf(inode, cpos, clusters, el);
+	brelse(bh);
+	if (ret) {
+		mlog_errno(ret);
+		return ret;
+	}
+
+	ent = ocfs2_extent_map_lookup(em, cpos, clusters, NULL, NULL);
+	if (!ent) {
+		ret = -ESRCH;
+		mlog_errno(ret);
+		return ret;
+	}
+
+	if (ent->e_tree_depth)
+		BUG();  /* FIXME: Make sure this isn't a corruption */
+
+	*ret_ent = ent;
+
+	return 0;
+}
+
+/*
+ * Callers must hold ip_lock.  This can insert pieces of the tree,
+ * thus racing lookup if the lock weren't held.
+ */
+static int ocfs2_extent_map_insert_entry(struct ocfs2_extent_map *em,
+					 struct ocfs2_extent_map_entry *ent)
+{
+	struct rb_node **p, *parent;
+	struct ocfs2_extent_map_entry *old_ent;
+
+	old_ent = ocfs2_extent_map_lookup(em, le32_to_cpu(ent->e_rec.e_cpos),
+					  le32_to_cpu(ent->e_rec.e_clusters),
+					  &p, &parent);
+	if (old_ent)
+		return -EEXIST;
+
+	rb_link_node(&ent->e_node, parent, p);
+	rb_insert_color(&ent->e_node, &em->em_extents);
+
+	return 0;
+}
+
+
+/*
+ * Simple rule: on any return code other than -EAGAIN, anything left
+ * in the insert_context will be freed.
+ */
+static int ocfs2_extent_map_try_insert(struct inode *inode,
+				       struct ocfs2_extent_rec *rec,
+				       int tree_depth,
+				       struct ocfs2_em_insert_context *ctxt)
+{
+	int ret;
+	struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
+	struct ocfs2_extent_map_entry *old_ent;
+
+	ctxt->need_left = 0;
+	ctxt->need_right = 0;
+	ctxt->old_ent = NULL;
+
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+	ret = ocfs2_extent_map_insert_entry(em, ctxt->new_ent);
+	if (!ret) {
+		ctxt->new_ent = NULL;
+		goto out_unlock;
+	}
+
+	old_ent = ocfs2_extent_map_lookup(em, le32_to_cpu(rec->e_cpos),
+					  le32_to_cpu(rec->e_clusters), NULL,
+					  NULL);
+
+	if (!old_ent)
+		BUG();
+
+	ret = -EEXIST;
+	if (old_ent->e_tree_depth < tree_depth)
+		goto out_unlock;
+
+	if (old_ent->e_tree_depth == tree_depth) {
+		if (!memcmp(rec, &old_ent->e_rec,
+			    sizeof(struct ocfs2_extent_rec)))
+			ret = 0;
+
+		/* FIXME: Should this be ESRCH/EBADR??? */
+		goto out_unlock;
+	}
+
+	/*
+	 * We do it in this order specifically so that no actual tree
+	 * changes occur until we have all the pieces we need.  We
+	 * don't want malloc failures to leave an inconsistent tree.
+	 * Whenever we drop the lock, another process could be
+	 * inserting.  Also note that, if another process just beat us
+	 * to an insert, we might not need the same pieces we needed
+	 * the first go round.  In the end, the pieces we need will
+	 * be used, and the pieces we don't will be freed.
+	 */
+	ctxt->need_left = !!(le32_to_cpu(rec->e_cpos) >
+			     le32_to_cpu(old_ent->e_rec.e_cpos));
+	ctxt->need_right = !!((le32_to_cpu(old_ent->e_rec.e_cpos) +
+			       le32_to_cpu(old_ent->e_rec.e_clusters)) >
+			      (le32_to_cpu(rec->e_cpos) + le32_to_cpu(rec->e_clusters)));
+	ret = -EAGAIN;
+	if (ctxt->need_left) {
+		if (!ctxt->left_ent)
+			goto out_unlock;
+		*(ctxt->left_ent) = *old_ent;
+		ctxt->left_ent->e_rec.e_clusters =
+			cpu_to_le32(le32_to_cpu(rec->e_cpos) -
+				    le32_to_cpu(ctxt->left_ent->e_rec.e_cpos));
+	}
+	if (ctxt->need_right) {
+		if (!ctxt->right_ent)
+			goto out_unlock;
+		*(ctxt->right_ent) = *old_ent;
+		ctxt->right_ent->e_rec.e_cpos =
+			cpu_to_le32(le32_to_cpu(rec->e_cpos) +
+				    le32_to_cpu(rec->e_clusters));
+		ctxt->right_ent->e_rec.e_clusters =
+			cpu_to_le32((le32_to_cpu(old_ent->e_rec.e_cpos) +
+				     le32_to_cpu(old_ent->e_rec.e_clusters)) -
+				    le32_to_cpu(ctxt->right_ent->e_rec.e_cpos));
+	}
+
+	rb_erase(&old_ent->e_node, &em->em_extents);
+	/* Now that he's erased, set him up for deletion */
+	ctxt->old_ent = old_ent;
+
+	if (ctxt->need_left) {
+		ret = ocfs2_extent_map_insert_entry(em,
+						    ctxt->left_ent);
+		if (ret)
+			goto out_unlock;
+		ctxt->left_ent = NULL;
+	}
+
+	if (ctxt->need_right) {
+		ret = ocfs2_extent_map_insert_entry(em,
+						    ctxt->right_ent);
+		if (ret)
+			goto out_unlock;
+		ctxt->right_ent = NULL;
+	}
+
+	ret = ocfs2_extent_map_insert_entry(em, ctxt->new_ent);
+
+	if (!ret)
+		ctxt->new_ent = NULL;
+
+out_unlock:
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+	return ret;
+}
+
+
+static int ocfs2_extent_map_insert(struct inode *inode,
+				   struct ocfs2_extent_rec *rec,
+				   int tree_depth)
+{
+	int ret;
+	struct ocfs2_em_insert_context ctxt = {0, };
+
+	if ((le32_to_cpu(rec->e_cpos) + le32_to_cpu(rec->e_clusters)) >
+	    OCFS2_I(inode)->ip_map.em_clusters) {
+		ret = -EBADR;
+		mlog_errno(ret);
+		return ret;
+	}
+
+	/* Zero e_clusters means a truncated tail record.  It better be EOF */
+	if (!rec->e_clusters) {
+		if ((le32_to_cpu(rec->e_cpos) + le32_to_cpu(rec->e_clusters)) !=
+		    OCFS2_I(inode)->ip_map.em_clusters) {
+			ret = -EBADR;
+			mlog_errno(ret);
+			return ret;
+		}
+
+		/* Ignore the truncated tail */
+		return 0;
+	}
+
+	ret = -ENOMEM;
+	ctxt.new_ent = kmem_cache_alloc(ocfs2_em_ent_cachep,
+					GFP_KERNEL);
+	if (!ctxt.new_ent) {
+		mlog_errno(ret);
+		return ret;
+	}
+
+	ctxt.new_ent->e_rec = *rec;
+	ctxt.new_ent->e_tree_depth = tree_depth;
+
+	do {
+		ret = -ENOMEM;
+		if (ctxt.need_left && !ctxt.left_ent) {
+			ctxt.left_ent =
+				kmem_cache_alloc(ocfs2_em_ent_cachep,
+						 GFP_KERNEL);
+			if (!ctxt.left_ent)
+				break;
+		}
+		if (ctxt.need_right && !ctxt.right_ent) {
+			ctxt.right_ent =
+				kmem_cache_alloc(ocfs2_em_ent_cachep,
+						 GFP_KERNEL);
+			if (!ctxt.right_ent)
+				break;
+		}
+
+		ret = ocfs2_extent_map_try_insert(inode, rec,
+						  tree_depth, &ctxt);
+	} while (ret == -EAGAIN);
+
+	if (ret < 0)
+		mlog_errno(ret);
+
+	if (ctxt.left_ent)
+		kmem_cache_free(ocfs2_em_ent_cachep, ctxt.left_ent);
+	if (ctxt.right_ent)
+		kmem_cache_free(ocfs2_em_ent_cachep, ctxt.right_ent);
+	if (ctxt.old_ent)
+		kmem_cache_free(ocfs2_em_ent_cachep, ctxt.old_ent);
+	if (ctxt.new_ent)
+		kmem_cache_free(ocfs2_em_ent_cachep, ctxt.new_ent);
+
+	return ret;
+}
+
+/*
+ * Append this record to the tail of the extent map.  It must be
+ * tree_depth 0.  The record might be an extension of an existing
+ * record, and as such that needs to be handled.  eg:
+ *
+ * Existing record in the extent map:
+ *
+ *	cpos = 10, len = 10
+ * 	|---------|
+ *
+ * New Record:
+ *
+ *	cpos = 10, len = 20
+ * 	|------------------|
+ *
+ * The passed record is the new on-disk record.  The new_clusters value
+ * is how many clusters were added to the file.  If the append is a
+ * contiguous append, the new_clusters has been added to
+ * rec->e_clusters.  If the append is an entirely new extent, then
+ * rec->e_clusters is == new_clusters.
+ */
+int ocfs2_extent_map_append(struct inode *inode,
+			    struct ocfs2_extent_rec *rec,
+			    u32 new_clusters)
+{
+	int ret;
+	struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
+	struct ocfs2_extent_map_entry *ent;
+	struct ocfs2_extent_rec *old;
+
+	BUG_ON(!new_clusters);
+	BUG_ON(le32_to_cpu(rec->e_clusters) < new_clusters);
+
+	if (em->em_clusters < OCFS2_I(inode)->ip_clusters) {
+		/*
+		 * Size changed underneath us on disk.  Drop any
+		 * straddling records and update our idea of
+		 * i_clusters
+		 */
+		ocfs2_extent_map_drop(inode, em->em_clusters - 1);
+		em->em_clusters = OCFS2_I(inode)->ip_clusters;
+	}
+
+	mlog_bug_on_msg((le32_to_cpu(rec->e_cpos) +
+			 le32_to_cpu(rec->e_clusters)) !=
+			(em->em_clusters + new_clusters),
+			"Inode %"MLFu64":\n"
+			"rec->e_cpos = %u + rec->e_clusters = %u = %u\n"
+			"em->em_clusters = %u + new_clusters = %u = %u\n",
+			OCFS2_I(inode)->ip_blkno,
+			le32_to_cpu(rec->e_cpos), le32_to_cpu(rec->e_clusters),
+			le32_to_cpu(rec->e_cpos) + le32_to_cpu(rec->e_clusters),
+			em->em_clusters, new_clusters,
+			em->em_clusters + new_clusters);
+
+	em->em_clusters += new_clusters;
+
+	ret = -ENOENT;
+	if (le32_to_cpu(rec->e_clusters) > new_clusters) {
+		/* This is a contiguous append */
+		ent = ocfs2_extent_map_lookup(em, le32_to_cpu(rec->e_cpos), 1,
+					      NULL, NULL);
+		if (ent) {
+			old = &ent->e_rec;
+			BUG_ON((le32_to_cpu(rec->e_cpos) +
+				le32_to_cpu(rec->e_clusters)) !=
+				 (le32_to_cpu(old->e_cpos) +
+				  le32_to_cpu(old->e_clusters) +
+				  new_clusters));
+			if (ent->e_tree_depth == 0) {
+				BUG_ON(le32_to_cpu(old->e_cpos) !=
+				       le32_to_cpu(rec->e_cpos));
+				BUG_ON(le64_to_cpu(old->e_blkno) !=
+				       le64_to_cpu(rec->e_blkno));
+				ret = 0;
+			}
+			/*
+			 * Let non-leafs fall through as -ENOENT to
+			 * force insertion of the new leaf.
+			 */
+			le32_add_cpu(&old->e_clusters, new_clusters);
+		}
+	}
+
+	if (ret == -ENOENT)
+		ret = ocfs2_extent_map_insert(inode, rec, 0);
+	if (ret < 0)
+		mlog_errno(ret);
+	return ret;
+}
+
+#if 0
+/* Code here is included but defined out as it completes the extent
+ * map api and may be used in the future. */
+
+/*
+ * Look up the record containing this cluster offset.  This record is
+ * part of the extent map.  Do not free it.  Any changes you make to
+ * it will reflect in the extent map.  So, if your last extent
+ * is (cpos = 10, clusters = 10) and you truncate the file by 5
+ * clusters, you can do:
+ *
+ * ret = ocfs2_extent_map_get_rec(em, orig_size - 5, &rec);
+ * rec->e_clusters -= 5;
+ *
+ * The lookup does not read from disk.  If the map isn't filled in for
+ * an entry, you won't find it.
+ *
+ * Also note that the returned record is valid until alloc_sem is
+ * dropped.  After that, truncate and extend can happen.  Caveat Emptor.
+ */
+int ocfs2_extent_map_get_rec(struct inode *inode, u32 cpos,
+			     struct ocfs2_extent_rec **rec,
+			     int *tree_depth)
+{
+	int ret = -ENOENT;
+	struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
+	struct ocfs2_extent_map_entry *ent;
+
+	*rec = NULL;
+
+	if (cpos >= OCFS2_I(inode)->ip_clusters)
+		return -EINVAL;
+
+	if (cpos >= em->em_clusters) {
+		/*
+		 * Size changed underneath us on disk.  Drop any
+		 * straddling records and update our idea of
+		 * i_clusters
+		 */
+		ocfs2_extent_map_drop(inode, em->em_clusters - 1);
+		em->em_clusters = OCFS2_I(inode)->ip_clusters ;
+	}
+
+	ent = ocfs2_extent_map_lookup(&OCFS2_I(inode)->ip_map, cpos, 1,
+				      NULL, NULL);
+
+	if (ent) {
+		*rec = &ent->e_rec;
+		if (tree_depth)
+			*tree_depth = ent->e_tree_depth;
+		ret = 0;
+	}
+
+	return ret;
+}
+
+int ocfs2_extent_map_get_clusters(struct inode *inode,
+				  u32 v_cpos, int count,
+				  u32 *p_cpos, int *ret_count)
+{
+	int ret;
+	u32 coff, ccount;
+	struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
+	struct ocfs2_extent_map_entry *ent = NULL;
+
+	*p_cpos = ccount = 0;
+
+	if ((v_cpos + count) > OCFS2_I(inode)->ip_clusters)
+		return -EINVAL;
+
+	if ((v_cpos + count) > em->em_clusters) {
+		/*
+		 * Size changed underneath us on disk.  Drop any
+		 * straddling records and update our idea of
+		 * i_clusters
+		 */
+		ocfs2_extent_map_drop(inode, em->em_clusters - 1);
+		em->em_clusters = OCFS2_I(inode)->ip_clusters;
+	}
+
+
+	ret = ocfs2_extent_map_lookup_read(inode, v_cpos, count, &ent);
+	if (ret)
+		return ret;
+
+	if (ent) {
+		/* We should never find ourselves straddling an interval */
+		if (!ocfs2_extent_rec_contains_clusters(&ent->e_rec,
+							v_cpos,
+							count))
+			return -ESRCH;
+
+		coff = v_cpos - le32_to_cpu(ent->e_rec.e_cpos);
+		*p_cpos = ocfs2_blocks_to_clusters(inode->i_sb,
+				le64_to_cpu(ent->e_rec.e_blkno)) +
+			  coff;
+
+		if (ret_count)
+			*ret_count = le32_to_cpu(ent->e_rec.e_clusters) - coff;
+
+		return 0;
+	}
+
+
+	return -ENOENT;
+}
+
+#endif  /*  0  */
+
+int ocfs2_extent_map_get_blocks(struct inode *inode,
+				u64 v_blkno, int count,
+				u64 *p_blkno, int *ret_count)
+{
+	int ret;
+	u64 boff;
+	u32 cpos, clusters;
+	int bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
+	struct ocfs2_extent_map_entry *ent = NULL;
+	struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
+	struct ocfs2_extent_rec *rec;
+
+	*p_blkno = 0;
+
+	cpos = ocfs2_blocks_to_clusters(inode->i_sb, v_blkno);
+	clusters = ocfs2_blocks_to_clusters(inode->i_sb,
+					    (u64)count + bpc - 1);
+	if ((cpos + clusters) > OCFS2_I(inode)->ip_clusters) {
+		ret = -EINVAL;
+		mlog_errno(ret);
+		return ret;
+	}
+
+	if ((cpos + clusters) > em->em_clusters) {
+		/*
+		 * Size changed underneath us on disk.  Drop any
+		 * straddling records and update our idea of
+		 * i_clusters
+		 */
+		ocfs2_extent_map_drop(inode, em->em_clusters - 1);
+		em->em_clusters = OCFS2_I(inode)->ip_clusters;
+	}
+
+	ret = ocfs2_extent_map_lookup_read(inode, cpos, clusters, &ent);
+	if (ret) {
+		mlog_errno(ret);
+		return ret;
+	}
+
+	if (ent)
+	{
+		rec = &ent->e_rec;
+
+		/* We should never find ourselves straddling an interval */
+		if (!ocfs2_extent_rec_contains_clusters(rec, cpos, clusters)) {
+			ret = -ESRCH;
+			mlog_errno(ret);
+			return ret;
+		}
+
+		boff = ocfs2_clusters_to_blocks(inode->i_sb, cpos -
+						le32_to_cpu(rec->e_cpos));
+		boff += (v_blkno & (u64)(bpc - 1));
+		*p_blkno = le64_to_cpu(rec->e_blkno) + boff;
+
+		if (ret_count) {
+			*ret_count = ocfs2_clusters_to_blocks(inode->i_sb,
+					le32_to_cpu(rec->e_clusters)) - boff;
+		}
+
+		return 0;
+	}
+
+	return -ENOENT;
+}
+
+int ocfs2_extent_map_init(struct inode *inode)
+{
+	struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
+
+	em->em_extents = RB_ROOT;
+	em->em_clusters = 0;
+
+	return 0;
+}
+
+/* Needs the lock */
+static void __ocfs2_extent_map_drop(struct inode *inode,
+				    u32 new_clusters,
+				    struct rb_node **free_head,
+				    struct ocfs2_extent_map_entry **tail_ent)
+{
+	struct rb_node *node, *next;
+	struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
+	struct ocfs2_extent_map_entry *ent;
+
+	*free_head = NULL;
+
+	ent = NULL;
+	node = rb_last(&em->em_extents);
+	while (node)
+	{
+		next = rb_prev(node);
+
+		ent = rb_entry(node, struct ocfs2_extent_map_entry,
+			       e_node);
+		if (le32_to_cpu(ent->e_rec.e_cpos) < new_clusters)
+			break;
+
+		rb_erase(&ent->e_node, &em->em_extents);
+
+		node->rb_right = *free_head;
+		*free_head = node;
+
+		ent = NULL;
+		node = next;
+	}
+
+	/* Do we have an entry straddling new_clusters? */
+	if (tail_ent) {
+		if (ent &&
+		    ((le32_to_cpu(ent->e_rec.e_cpos) +
+		      le32_to_cpu(ent->e_rec.e_clusters)) > new_clusters))
+			*tail_ent = ent;
+		else
+			*tail_ent = NULL;
+	}
+}
+
+static void __ocfs2_extent_map_drop_cleanup(struct rb_node *free_head)
+{
+	struct rb_node *node;
+	struct ocfs2_extent_map_entry *ent;
+
+	while (free_head) {
+		node = free_head;
+		free_head = node->rb_right;
+
+		ent = rb_entry(node, struct ocfs2_extent_map_entry,
+			       e_node);
+		kmem_cache_free(ocfs2_em_ent_cachep, ent);
+	}
+}
+
+/*
+ * Remove all entries past new_clusters, inclusive of an entry that
+ * contains new_clusters.  This is effectively a cache forget.
+ *
+ * If you want to also clip the last extent by some number of clusters,
+ * you need to call ocfs2_extent_map_trunc().
+ * This code does not check or modify ip_clusters.
+ */
+int ocfs2_extent_map_drop(struct inode *inode, u32 new_clusters)
+{
+	struct rb_node *free_head = NULL;
+	struct ocfs2_extent_map *em = &OCFS2_I(inode)->ip_map;
+	struct ocfs2_extent_map_entry *ent;
+
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+
+	__ocfs2_extent_map_drop(inode, new_clusters, &free_head, &ent);
+
+	if (ent) {
+		rb_erase(&ent->e_node, &em->em_extents);
+		ent->e_node.rb_right = free_head;
+		free_head = &ent->e_node;
+	}
+
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+	if (free_head)
+		__ocfs2_extent_map_drop_cleanup(free_head);
+
+	return 0;
+}
+
+/*
+ * Remove all entries past new_clusters and also clip any extent
+ * straddling new_clusters, if there is one.  This does not check
+ * or modify ip_clusters
+ */
+int ocfs2_extent_map_trunc(struct inode *inode, u32 new_clusters)
+{
+	struct rb_node *free_head = NULL;
+	struct ocfs2_extent_map_entry *ent = NULL;
+
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+
+	__ocfs2_extent_map_drop(inode, new_clusters, &free_head, &ent);
+
+	if (ent)
+		ent->e_rec.e_clusters = cpu_to_le32(new_clusters -
+					       le32_to_cpu(ent->e_rec.e_cpos));
+
+	OCFS2_I(inode)->ip_map.em_clusters = new_clusters;
+
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+	if (free_head)
+		__ocfs2_extent_map_drop_cleanup(free_head);
+
+	return 0;
+}
+
+int __init init_ocfs2_extent_maps(void)
+{
+	ocfs2_em_ent_cachep =
+		kmem_cache_create("ocfs2_em_ent",
+				  sizeof(struct ocfs2_extent_map_entry),
+				  0, SLAB_HWCACHE_ALIGN, NULL, NULL);
+	if (!ocfs2_em_ent_cachep)
+		return -ENOMEM;
+
+	return 0;
+}
+
+void __exit exit_ocfs2_extent_maps(void)
+{
+	kmem_cache_destroy(ocfs2_em_ent_cachep);
+}
diff --git a/fs/ocfs2/extent_map.h b/fs/ocfs2/extent_map.h
new file mode 100644
index 00000000000..fa3745efa88
--- /dev/null
+++ b/fs/ocfs2/extent_map.h
@@ -0,0 +1,46 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * extent_map.h
+ *
+ * In-memory file extent mappings for OCFS2.
+ *
+ * Copyright (C) 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License, version 2,  as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef _EXTENT_MAP_H
+#define _EXTENT_MAP_H
+
+int init_ocfs2_extent_maps(void);
+void exit_ocfs2_extent_maps(void);
+
+/*
+ * EVERY CALL here except _init, _trunc, and _drop expects alloc_sem
+ * to be held.  The allocation cannot change at all while the map is
+ * in the process of being updated.
+ */
+int ocfs2_extent_map_init(struct inode *inode);
+int ocfs2_extent_map_append(struct inode *inode,
+			    struct ocfs2_extent_rec *rec,
+			    u32 new_clusters);
+int ocfs2_extent_map_get_blocks(struct inode *inode,
+				u64 v_blkno, int count,
+				u64 *p_blkno, int *ret_count);
+int ocfs2_extent_map_drop(struct inode *inode, u32 new_clusters);
+int ocfs2_extent_map_trunc(struct inode *inode, u32 new_clusters);
+
+#endif  /* _EXTENT_MAP_H */
diff --git a/fs/ocfs2/file.c b/fs/ocfs2/file.c
new file mode 100644
index 00000000000..72ae9e3306f
--- /dev/null
+++ b/fs/ocfs2/file.c
@@ -0,0 +1,1237 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * file.c
+ *
+ * File open, close, extend, truncate
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/pagemap.h>
+#include <linux/uio.h>
+
+#define MLOG_MASK_PREFIX ML_INODE
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "aops.h"
+#include "dir.h"
+#include "dlmglue.h"
+#include "extent_map.h"
+#include "file.h"
+#include "sysfile.h"
+#include "inode.h"
+#include "journal.h"
+#include "mmap.h"
+#include "suballoc.h"
+#include "super.h"
+
+#include "buffer_head_io.h"
+
+static int ocfs2_sync_inode(struct inode *inode)
+{
+	filemap_fdatawrite(inode->i_mapping);
+	return sync_mapping_buffers(inode->i_mapping);
+}
+
+static int ocfs2_file_open(struct inode *inode, struct file *file)
+{
+	int status;
+	int mode = file->f_flags;
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+
+	mlog_entry("(0x%p, 0x%p, '%.*s')\n", inode, file,
+		   file->f_dentry->d_name.len, file->f_dentry->d_name.name);
+
+	spin_lock(&oi->ip_lock);
+
+	/* Check that the inode hasn't been wiped from disk by another
+	 * node. If it hasn't then we're safe as long as we hold the
+	 * spin lock until our increment of open count. */
+	if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_DELETED) {
+		spin_unlock(&oi->ip_lock);
+
+		status = -ENOENT;
+		goto leave;
+	}
+
+	if (mode & O_DIRECT)
+		oi->ip_flags |= OCFS2_INODE_OPEN_DIRECT;
+
+	oi->ip_open_count++;
+	spin_unlock(&oi->ip_lock);
+	status = 0;
+leave:
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_file_release(struct inode *inode, struct file *file)
+{
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+
+	mlog_entry("(0x%p, 0x%p, '%.*s')\n", inode, file,
+		       file->f_dentry->d_name.len,
+		       file->f_dentry->d_name.name);
+
+	spin_lock(&oi->ip_lock);
+	if (!--oi->ip_open_count)
+		oi->ip_flags &= ~OCFS2_INODE_OPEN_DIRECT;
+	spin_unlock(&oi->ip_lock);
+
+	mlog_exit(0);
+
+	return 0;
+}
+
+static int ocfs2_sync_file(struct file *file,
+			   struct dentry *dentry,
+			   int datasync)
+{
+	int err = 0;
+	journal_t *journal;
+	struct inode *inode = dentry->d_inode;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	mlog_entry("(0x%p, 0x%p, %d, '%.*s')\n", file, dentry, datasync,
+		   dentry->d_name.len, dentry->d_name.name);
+
+	err = ocfs2_sync_inode(dentry->d_inode);
+	if (err)
+		goto bail;
+
+	journal = osb->journal->j_journal;
+	err = journal_force_commit(journal);
+
+bail:
+	mlog_exit(err);
+
+	return (err < 0) ? -EIO : 0;
+}
+
+int ocfs2_set_inode_size(struct ocfs2_journal_handle *handle,
+			 struct inode *inode,
+			 struct buffer_head *fe_bh,
+			 u64 new_i_size)
+{
+	int status;
+
+	mlog_entry_void();
+	i_size_write(inode, new_i_size);
+	inode->i_blocks = ocfs2_align_bytes_to_sectors(new_i_size);
+	inode->i_ctime = inode->i_mtime = CURRENT_TIME;
+
+	status = ocfs2_mark_inode_dirty(handle, inode, fe_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_simple_size_update(struct inode *inode,
+				    struct buffer_head *di_bh,
+				    u64 new_i_size)
+{
+	int ret;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	struct ocfs2_journal_handle *handle = NULL;
+
+	handle = ocfs2_start_trans(osb, NULL,
+				   OCFS2_INODE_UPDATE_CREDITS);
+	if (handle == NULL) {
+		ret = -ENOMEM;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	ret = ocfs2_set_inode_size(handle, inode, di_bh,
+				   new_i_size);
+	if (ret < 0)
+		mlog_errno(ret);
+
+	ocfs2_commit_trans(handle);
+out:
+	return ret;
+}
+
+static int ocfs2_orphan_for_truncate(struct ocfs2_super *osb,
+				     struct inode *inode,
+				     struct buffer_head *fe_bh,
+				     u64 new_i_size)
+{
+	int status;
+	struct ocfs2_journal_handle *handle;
+
+	mlog_entry_void();
+
+	/* TODO: This needs to actually orphan the inode in this
+	 * transaction. */
+
+	handle = ocfs2_start_trans(osb, NULL, OCFS2_INODE_UPDATE_CREDITS);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		mlog_errno(status);
+		goto out;
+	}
+
+	status = ocfs2_set_inode_size(handle, inode, fe_bh, new_i_size);
+	if (status < 0)
+		mlog_errno(status);
+
+	ocfs2_commit_trans(handle);
+out:
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_truncate_file(struct inode *inode,
+			       struct buffer_head *di_bh,
+			       u64 new_i_size)
+{
+	int status = 0;
+	struct ocfs2_dinode *fe = NULL;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	struct ocfs2_truncate_context *tc = NULL;
+
+	mlog_entry("(inode = %"MLFu64", new_i_size = %"MLFu64"\n",
+		   OCFS2_I(inode)->ip_blkno, new_i_size);
+
+	truncate_inode_pages(inode->i_mapping, new_i_size);
+
+	fe = (struct ocfs2_dinode *) di_bh->b_data;
+	if (!OCFS2_IS_VALID_DINODE(fe)) {
+		OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe);
+		status = -EIO;
+		goto bail;
+	}
+
+	mlog_bug_on_msg(le64_to_cpu(fe->i_size) != i_size_read(inode),
+			"Inode %"MLFu64", inode i_size = %lld != di "
+			"i_size = %"MLFu64", i_flags = 0x%x\n",
+			OCFS2_I(inode)->ip_blkno,
+			i_size_read(inode),
+			le64_to_cpu(fe->i_size), le32_to_cpu(fe->i_flags));
+
+	if (new_i_size > le64_to_cpu(fe->i_size)) {
+		mlog(0, "asked to truncate file with size (%"MLFu64") "
+		     "to size (%"MLFu64")!\n",
+		     le64_to_cpu(fe->i_size), new_i_size);
+		status = -EINVAL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	mlog(0, "inode %"MLFu64", i_size = %"MLFu64", new_i_size = %"MLFu64"\n",
+	     le64_to_cpu(fe->i_blkno), le64_to_cpu(fe->i_size), new_i_size);
+
+	/* lets handle the simple truncate cases before doing any more
+	 * cluster locking. */
+	if (new_i_size == le64_to_cpu(fe->i_size))
+		goto bail;
+
+	if (le32_to_cpu(fe->i_clusters) ==
+	    ocfs2_clusters_for_bytes(osb->sb, new_i_size)) {
+		mlog(0, "fe->i_clusters = %u, so we do a simple truncate\n",
+		     fe->i_clusters);
+		/* No allocation change is required, so lets fast path
+		 * this truncate. */
+		status = ocfs2_simple_size_update(inode, di_bh, new_i_size);
+		if (status < 0)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	/* This forces other nodes to sync and drop their pages */
+	status = ocfs2_data_lock(inode, 1);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+	ocfs2_data_unlock(inode, 1);
+
+	/* alright, we're going to need to do a full blown alloc size
+	 * change. Orphan the inode so that recovery can complete the
+	 * truncate if necessary. This does the task of marking
+	 * i_size. */
+	status = ocfs2_orphan_for_truncate(osb, inode, di_bh, new_i_size);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_prepare_truncate(osb, inode, di_bh, &tc);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_commit_truncate(osb, inode, di_bh, tc);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* TODO: orphan dir cleanup here. */
+bail:
+
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * extend allocation only here.
+ * we'll update all the disk stuff, and oip->alloc_size
+ *
+ * expect stuff to be locked, a transaction started and enough data /
+ * metadata reservations in the contexts.
+ *
+ * Will return -EAGAIN, and a reason if a restart is needed.
+ * If passed in, *reason will always be set, even in error.
+ */
+int ocfs2_do_extend_allocation(struct ocfs2_super *osb,
+			       struct inode *inode,
+			       u32 clusters_to_add,
+			       struct buffer_head *fe_bh,
+			       struct ocfs2_journal_handle *handle,
+			       struct ocfs2_alloc_context *data_ac,
+			       struct ocfs2_alloc_context *meta_ac,
+			       enum ocfs2_alloc_restarted *reason_ret)
+{
+	int status = 0;
+	int free_extents;
+	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) fe_bh->b_data;
+	enum ocfs2_alloc_restarted reason = RESTART_NONE;
+	u32 bit_off, num_bits;
+	u64 block;
+
+	BUG_ON(!clusters_to_add);
+
+	free_extents = ocfs2_num_free_extents(osb, inode, fe);
+	if (free_extents < 0) {
+		status = free_extents;
+		mlog_errno(status);
+		goto leave;
+	}
+
+	/* there are two cases which could cause us to EAGAIN in the
+	 * we-need-more-metadata case:
+	 * 1) we haven't reserved *any*
+	 * 2) we are so fragmented, we've needed to add metadata too
+	 *    many times. */
+	if (!free_extents && !meta_ac) {
+		mlog(0, "we haven't reserved any metadata!\n");
+		status = -EAGAIN;
+		reason = RESTART_META;
+		goto leave;
+	} else if ((!free_extents)
+		   && (ocfs2_alloc_context_bits_left(meta_ac)
+		       < ocfs2_extend_meta_needed(fe))) {
+		mlog(0, "filesystem is really fragmented...\n");
+		status = -EAGAIN;
+		reason = RESTART_META;
+		goto leave;
+	}
+
+	status = ocfs2_claim_clusters(osb, handle, data_ac, 1,
+				      &bit_off, &num_bits);
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto leave;
+	}
+
+	BUG_ON(num_bits > clusters_to_add);
+
+	/* reserve our write early -- insert_extent may update the inode */
+	status = ocfs2_journal_access(handle, inode, fe_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	block = ocfs2_clusters_to_blocks(osb->sb, bit_off);
+	mlog(0, "Allocating %u clusters at block %u for inode %"MLFu64"\n",
+	     num_bits, bit_off, OCFS2_I(inode)->ip_blkno);
+	status = ocfs2_insert_extent(osb, handle, inode, fe_bh, block,
+				     num_bits, meta_ac);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	le32_add_cpu(&fe->i_clusters, num_bits);
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+	OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+	status = ocfs2_journal_dirty(handle, fe_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	clusters_to_add -= num_bits;
+
+	if (clusters_to_add) {
+		mlog(0, "need to alloc once more, clusters = %u, wanted = "
+		     "%u\n", fe->i_clusters, clusters_to_add);
+		status = -EAGAIN;
+		reason = RESTART_TRANS;
+	}
+
+leave:
+	mlog_exit(status);
+	if (reason_ret)
+		*reason_ret = reason;
+	return status;
+}
+
+static int ocfs2_extend_allocation(struct inode *inode,
+				   u32 clusters_to_add)
+{
+	int status = 0;
+	int restart_func = 0;
+	int drop_alloc_sem = 0;
+	int credits, num_free_extents;
+	u32 prev_clusters;
+	struct buffer_head *bh = NULL;
+	struct ocfs2_dinode *fe = NULL;
+	struct ocfs2_journal_handle *handle = NULL;
+	struct ocfs2_alloc_context *data_ac = NULL;
+	struct ocfs2_alloc_context *meta_ac = NULL;
+	enum ocfs2_alloc_restarted why;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	mlog_entry("(clusters_to_add = %u)\n", clusters_to_add);
+
+	status = ocfs2_read_block(osb, OCFS2_I(inode)->ip_blkno, &bh,
+				  OCFS2_BH_CACHED, inode);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	fe = (struct ocfs2_dinode *) bh->b_data;
+	if (!OCFS2_IS_VALID_DINODE(fe)) {
+		OCFS2_RO_ON_INVALID_DINODE(inode->i_sb, fe);
+		status = -EIO;
+		goto leave;
+	}
+
+restart_all:
+	BUG_ON(le32_to_cpu(fe->i_clusters) != OCFS2_I(inode)->ip_clusters);
+
+	mlog(0, "extend inode %"MLFu64", i_size = %lld, fe->i_clusters = %u, "
+	     "clusters_to_add = %u\n",
+	     OCFS2_I(inode)->ip_blkno, i_size_read(inode),
+	     fe->i_clusters, clusters_to_add);
+
+	handle = ocfs2_alloc_handle(osb);
+	if (handle == NULL) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto leave;
+	}
+
+	num_free_extents = ocfs2_num_free_extents(osb,
+						  inode,
+						  fe);
+	if (num_free_extents < 0) {
+		status = num_free_extents;
+		mlog_errno(status);
+		goto leave;
+	}
+
+	if (!num_free_extents) {
+		status = ocfs2_reserve_new_metadata(osb,
+						    handle,
+						    fe,
+						    &meta_ac);
+		if (status < 0) {
+			if (status != -ENOSPC)
+				mlog_errno(status);
+			goto leave;
+		}
+	}
+
+	status = ocfs2_reserve_clusters(osb,
+					handle,
+					clusters_to_add,
+					&data_ac);
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto leave;
+	}
+
+	/* blocks peope in read/write from reading our allocation
+	 * until we're done changing it. We depend on i_sem to block
+	 * other extend/truncate calls while we're here. Ordering wrt
+	 * start_trans is important here -- always do it before! */
+	down_write(&OCFS2_I(inode)->ip_alloc_sem);
+	drop_alloc_sem = 1;
+
+	credits = ocfs2_calc_extend_credits(osb->sb, fe, clusters_to_add);
+	handle = ocfs2_start_trans(osb, handle, credits);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		handle = NULL;
+		mlog_errno(status);
+		goto leave;
+	}
+
+restarted_transaction:
+	/* reserve a write to the file entry early on - that we if we
+	 * run out of credits in the allocation path, we can still
+	 * update i_size. */
+	status = ocfs2_journal_access(handle, inode, bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	prev_clusters = OCFS2_I(inode)->ip_clusters;
+
+	status = ocfs2_do_extend_allocation(osb,
+					    inode,
+					    clusters_to_add,
+					    bh,
+					    handle,
+					    data_ac,
+					    meta_ac,
+					    &why);
+	if ((status < 0) && (status != -EAGAIN)) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto leave;
+	}
+
+	status = ocfs2_journal_dirty(handle, bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+	clusters_to_add -= (OCFS2_I(inode)->ip_clusters - prev_clusters);
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+	if (why != RESTART_NONE && clusters_to_add) {
+		if (why == RESTART_META) {
+			mlog(0, "restarting function.\n");
+			restart_func = 1;
+		} else {
+			BUG_ON(why != RESTART_TRANS);
+
+			mlog(0, "restarting transaction.\n");
+			/* TODO: This can be more intelligent. */
+			credits = ocfs2_calc_extend_credits(osb->sb,
+							    fe,
+							    clusters_to_add);
+			status = ocfs2_extend_trans(handle, credits);
+			if (status < 0) {
+				/* handle still has to be committed at
+				 * this point. */
+				status = -ENOMEM;
+				mlog_errno(status);
+				goto leave;
+			}
+			goto restarted_transaction;
+		}
+	}
+
+	mlog(0, "fe: i_clusters = %u, i_size=%"MLFu64"\n",
+	     fe->i_clusters, fe->i_size);
+	mlog(0, "inode: ip_clusters=%u, i_size=%lld\n",
+	     OCFS2_I(inode)->ip_clusters, i_size_read(inode));
+
+leave:
+	if (drop_alloc_sem) {
+		up_write(&OCFS2_I(inode)->ip_alloc_sem);
+		drop_alloc_sem = 0;
+	}
+	if (handle) {
+		ocfs2_commit_trans(handle);
+		handle = NULL;
+	}
+	if (data_ac) {
+		ocfs2_free_alloc_context(data_ac);
+		data_ac = NULL;
+	}
+	if (meta_ac) {
+		ocfs2_free_alloc_context(meta_ac);
+		meta_ac = NULL;
+	}
+	if ((!status) && restart_func) {
+		restart_func = 0;
+		goto restart_all;
+	}
+	if (bh) {
+		brelse(bh);
+		bh = NULL;
+	}
+
+	mlog_exit(status);
+	return status;
+}
+
+/* Some parts of this taken from generic_cont_expand, which turned out
+ * to be too fragile to do exactly what we need without us having to
+ * worry about recursive locking in ->commit_write(). */
+static int ocfs2_write_zero_page(struct inode *inode,
+				 u64 size)
+{
+	struct address_space *mapping = inode->i_mapping;
+	struct page *page;
+	unsigned long index;
+	unsigned int offset;
+	struct ocfs2_journal_handle *handle = NULL;
+	int ret;
+
+	offset = (size & (PAGE_CACHE_SIZE-1)); /* Within page */
+	/* ugh.  in prepare/commit_write, if from==to==start of block, we 
+	** skip the prepare.  make sure we never send an offset for the start
+	** of a block
+	*/
+	if ((offset & (inode->i_sb->s_blocksize - 1)) == 0) {
+		offset++;
+	}
+	index = size >> PAGE_CACHE_SHIFT;
+
+	page = grab_cache_page(mapping, index);
+	if (!page) {
+		ret = -ENOMEM;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	ret = ocfs2_prepare_write(NULL, page, offset, offset);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out_unlock;
+	}
+
+	if (ocfs2_should_order_data(inode)) {
+		handle = ocfs2_start_walk_page_trans(inode, page, offset,
+						     offset);
+		if (IS_ERR(handle)) {
+			ret = PTR_ERR(handle);
+			handle = NULL;
+			goto out_unlock;
+		}
+	}
+
+	/* must not update i_size! */
+	ret = block_commit_write(page, offset, offset);
+	if (ret < 0)
+		mlog_errno(ret);
+	else
+		ret = 0;
+
+	if (handle)
+		ocfs2_commit_trans(handle);
+out_unlock:
+	unlock_page(page);
+	page_cache_release(page);
+out:
+	return ret;
+}
+
+static int ocfs2_zero_extend(struct inode *inode,
+			     u64 zero_to_size)
+{
+	int ret = 0;
+	u64 start_off;
+	struct super_block *sb = inode->i_sb;
+
+	start_off = ocfs2_align_bytes_to_blocks(sb, i_size_read(inode));
+	while (start_off < zero_to_size) {
+		ret = ocfs2_write_zero_page(inode, start_off);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out;
+		}
+
+		start_off += sb->s_blocksize;
+	}
+
+out:
+	return ret;
+}
+
+static int ocfs2_extend_file(struct inode *inode,
+			     struct buffer_head *di_bh,
+			     u64 new_i_size)
+{
+	int ret = 0;
+	u32 clusters_to_add;
+
+	/* setattr sometimes calls us like this. */
+	if (new_i_size == 0)
+		goto out;
+
+	if (i_size_read(inode) == new_i_size)
+  		goto out;
+	BUG_ON(new_i_size < i_size_read(inode));
+
+	clusters_to_add = ocfs2_clusters_for_bytes(inode->i_sb, new_i_size) - 
+		OCFS2_I(inode)->ip_clusters;
+
+	if (clusters_to_add) {
+		ret = ocfs2_extend_allocation(inode, clusters_to_add);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out;
+		}
+
+		ret = ocfs2_zero_extend(inode, new_i_size);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out;
+		}
+	} 
+
+	/* No allocation required, we just use this helper to
+	 * do a trivial update of i_size. */
+	ret = ocfs2_simple_size_update(inode, di_bh, new_i_size);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+out:
+	return ret;
+}
+
+int ocfs2_setattr(struct dentry *dentry, struct iattr *attr)
+{
+	int status = 0, size_change;
+	struct inode *inode = dentry->d_inode;
+	struct super_block *sb = inode->i_sb;
+	struct ocfs2_super *osb = OCFS2_SB(sb);
+	struct buffer_head *bh = NULL;
+	struct ocfs2_journal_handle *handle = NULL;
+
+	mlog_entry("(0x%p, '%.*s')\n", dentry,
+	           dentry->d_name.len, dentry->d_name.name);
+
+	if (attr->ia_valid & ATTR_MODE)
+		mlog(0, "mode change: %d\n", attr->ia_mode);
+	if (attr->ia_valid & ATTR_UID)
+		mlog(0, "uid change: %d\n", attr->ia_uid);
+	if (attr->ia_valid & ATTR_GID)
+		mlog(0, "gid change: %d\n", attr->ia_gid);
+	if (attr->ia_valid & ATTR_SIZE)
+		mlog(0, "size change...\n");
+	if (attr->ia_valid & (ATTR_ATIME | ATTR_MTIME | ATTR_CTIME))
+		mlog(0, "time change...\n");
+
+#define OCFS2_VALID_ATTRS (ATTR_ATIME | ATTR_MTIME | ATTR_CTIME | ATTR_SIZE \
+			   | ATTR_GID | ATTR_UID | ATTR_MODE)
+	if (!(attr->ia_valid & OCFS2_VALID_ATTRS)) {
+		mlog(0, "can't handle attrs: 0x%x\n", attr->ia_valid);
+		return 0;
+	}
+
+	status = inode_change_ok(inode, attr);
+	if (status)
+		return status;
+
+	size_change = S_ISREG(inode->i_mode) && attr->ia_valid & ATTR_SIZE;
+	if (size_change) {
+		status = ocfs2_rw_lock(inode, 1);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+	}
+
+	status = ocfs2_meta_lock(inode, NULL, &bh, 1);
+	if (status < 0) {
+		if (status != -ENOENT)
+			mlog_errno(status);
+		goto bail_unlock_rw;
+	}
+
+	if (size_change && attr->ia_size != i_size_read(inode)) {
+		if (i_size_read(inode) > attr->ia_size)
+			status = ocfs2_truncate_file(inode, bh, attr->ia_size);
+		else
+			status = ocfs2_extend_file(inode, bh, attr->ia_size);
+		if (status < 0) {
+			if (status != -ENOSPC)
+				mlog_errno(status);
+			status = -ENOSPC;
+			goto bail_unlock;
+		}
+	}
+
+	handle = ocfs2_start_trans(osb, NULL, OCFS2_INODE_UPDATE_CREDITS);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		mlog_errno(status);
+		goto bail_unlock;
+	}
+
+	status = inode_setattr(inode, attr);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail_commit;
+	}
+
+	status = ocfs2_mark_inode_dirty(handle, inode, bh);
+	if (status < 0)
+		mlog_errno(status);
+
+bail_commit:
+	ocfs2_commit_trans(handle);
+bail_unlock:
+	ocfs2_meta_unlock(inode, 1);
+bail_unlock_rw:
+	if (size_change)
+		ocfs2_rw_unlock(inode, 1);
+bail:
+	if (bh)
+		brelse(bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+int ocfs2_getattr(struct vfsmount *mnt,
+		  struct dentry *dentry,
+		  struct kstat *stat)
+{
+	struct inode *inode = dentry->d_inode;
+	struct super_block *sb = dentry->d_inode->i_sb;
+	struct ocfs2_super *osb = sb->s_fs_info;
+	int err;
+
+	mlog_entry_void();
+
+	err = ocfs2_inode_revalidate(dentry);
+	if (err) {
+		if (err != -ENOENT)
+			mlog_errno(err);
+		goto bail;
+	}
+
+	generic_fillattr(inode, stat);
+
+	/* We set the blksize from the cluster size for performance */
+	stat->blksize = osb->s_clustersize;
+
+bail:
+	mlog_exit(err);
+
+	return err;
+}
+
+static int ocfs2_write_remove_suid(struct inode *inode)
+{
+	int ret;
+	struct buffer_head *bh = NULL;
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+	struct ocfs2_journal_handle *handle;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	struct ocfs2_dinode *di;
+
+	mlog_entry("(Inode %"MLFu64", mode 0%o)\n", oi->ip_blkno,
+		   inode->i_mode);
+
+	handle = ocfs2_start_trans(osb, NULL, OCFS2_INODE_UPDATE_CREDITS);
+	if (handle == NULL) {
+		ret = -ENOMEM;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	ret = ocfs2_read_block(osb, oi->ip_blkno, &bh, OCFS2_BH_CACHED, inode);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out_trans;
+	}
+
+	ret = ocfs2_journal_access(handle, inode, bh,
+				   OCFS2_JOURNAL_ACCESS_WRITE);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out_bh;
+	}
+
+	inode->i_mode &= ~S_ISUID;
+	if ((inode->i_mode & S_ISGID) && (inode->i_mode & S_IXGRP))
+		inode->i_mode &= ~S_ISGID;
+
+	di = (struct ocfs2_dinode *) bh->b_data;
+	di->i_mode = cpu_to_le16(inode->i_mode);
+
+	ret = ocfs2_journal_dirty(handle, bh);
+	if (ret < 0)
+		mlog_errno(ret);
+out_bh:
+	brelse(bh);
+out_trans:
+	ocfs2_commit_trans(handle);
+out:
+	mlog_exit(ret);
+	return ret;
+}
+
+static inline int ocfs2_write_should_remove_suid(struct inode *inode)
+{
+	mode_t mode = inode->i_mode;
+
+	if (!capable(CAP_FSETID)) {
+		if (unlikely(mode & S_ISUID))
+			return 1;
+
+		if (unlikely((mode & S_ISGID) && (mode & S_IXGRP)))
+			return 1;
+	}
+	return 0;
+}
+
+static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
+				    const char __user *buf,
+				    size_t count,
+				    loff_t pos)
+{
+	struct iovec local_iov = { .iov_base = (void __user *)buf,
+				   .iov_len = count };
+	int ret, rw_level = -1, meta_level = -1, have_alloc_sem = 0;
+	u32 clusters;
+	struct file *filp = iocb->ki_filp;
+	struct inode *inode = filp->f_dentry->d_inode;
+	loff_t newsize, saved_pos;
+#ifdef OCFS2_ORACORE_WORKAROUNDS
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+#endif
+
+	mlog_entry("(0x%p, 0x%p, %u, '%.*s')\n", filp, buf,
+		   (unsigned int)count,
+		   filp->f_dentry->d_name.len,
+		   filp->f_dentry->d_name.name);
+
+	/* happy write of zero bytes */
+	if (count == 0)
+		return 0;
+
+	if (!inode) {
+		mlog(0, "bad inode\n");
+		return -EIO;
+	}
+
+#ifdef OCFS2_ORACORE_WORKAROUNDS
+	/* ugh, work around some applications which open everything O_DIRECT +
+	 * O_APPEND and really don't mean to use O_DIRECT. */
+	if (osb->s_mount_opt & OCFS2_MOUNT_COMPAT_OCFS &&
+	    (filp->f_flags & O_APPEND) && (filp->f_flags & O_DIRECT)) 
+		filp->f_flags &= ~O_DIRECT;
+#endif
+
+	down(&inode->i_sem);
+	/* to match setattr's i_sem -> i_alloc_sem -> rw_lock ordering */
+	if (filp->f_flags & O_DIRECT) {
+		have_alloc_sem = 1;
+		down_read(&inode->i_alloc_sem);
+	}
+
+	/* concurrent O_DIRECT writes are allowed */
+	rw_level = (filp->f_flags & O_DIRECT) ? 0 : 1;
+	ret = ocfs2_rw_lock(inode, rw_level);
+	if (ret < 0) {
+		rw_level = -1;
+		mlog_errno(ret);
+		goto out;
+	}
+
+	/* 
+	 * We sample i_size under a read level meta lock to see if our write
+	 * is extending the file, if it is we back off and get a write level
+	 * meta lock.
+	 */
+	meta_level = (filp->f_flags & O_APPEND) ? 1 : 0;
+	for(;;) {
+		ret = ocfs2_meta_lock(inode, NULL, NULL, meta_level);
+		if (ret < 0) {
+			meta_level = -1;
+			mlog_errno(ret);
+			goto out;
+		}
+
+		/* Clear suid / sgid if necessary. We do this here
+		 * instead of later in the write path because
+		 * remove_suid() calls ->setattr without any hint that
+		 * we may have already done our cluster locking. Since
+		 * ocfs2_setattr() *must* take cluster locks to
+		 * proceeed, this will lead us to recursively lock the
+		 * inode. There's also the dinode i_size state which
+		 * can be lost via setattr during extending writes (we
+		 * set inode->i_size at the end of a write. */
+		if (ocfs2_write_should_remove_suid(inode)) {
+			if (meta_level == 0) {
+				ocfs2_meta_unlock(inode, meta_level);
+				meta_level = 1;
+				continue;
+			}
+
+			ret = ocfs2_write_remove_suid(inode);
+			if (ret < 0) {
+				mlog_errno(ret);
+				goto out;
+			}
+		}
+
+		/* work on a copy of ppos until we're sure that we won't have
+		 * to recalculate it due to relocking. */
+		if (filp->f_flags & O_APPEND) {
+			saved_pos = i_size_read(inode);
+			mlog(0, "O_APPEND: inode->i_size=%llu\n", saved_pos);
+		} else {
+			saved_pos = iocb->ki_pos;
+		}
+		newsize = count + saved_pos;
+
+		mlog(0, "pos=%lld newsize=%"MLFu64" cursize=%lld\n",
+		     saved_pos, newsize, i_size_read(inode));
+
+		/* No need for a higher level metadata lock if we're
+		 * never going past i_size. */
+		if (newsize <= i_size_read(inode))
+			break;
+
+		if (meta_level == 0) {
+			ocfs2_meta_unlock(inode, meta_level);
+			meta_level = 1;
+			continue;
+		}
+
+		spin_lock(&OCFS2_I(inode)->ip_lock);
+		clusters = ocfs2_clusters_for_bytes(inode->i_sb, newsize) -
+			OCFS2_I(inode)->ip_clusters;
+		spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+		mlog(0, "Writing at EOF, may need more allocation: "
+		     "i_size = %lld, newsize = %"MLFu64", need %u clusters\n",
+		     i_size_read(inode), newsize, clusters);
+
+		/* We only want to continue the rest of this loop if
+		 * our extend will actually require more
+		 * allocation. */
+		if (!clusters)
+			break;
+
+		ret = ocfs2_extend_allocation(inode, clusters);
+		if (ret < 0) {
+			if (ret != -ENOSPC)
+				mlog_errno(ret);
+			goto out;
+		}
+
+		/* Fill any holes which would've been created by this
+		 * write. If we're O_APPEND, this will wind up
+		 * (correctly) being a noop. */
+		ret = ocfs2_zero_extend(inode, (u64) newsize - count);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out;
+		}
+		break;
+	}
+
+	/* ok, we're done with i_size and alloc work */
+	iocb->ki_pos = saved_pos;
+	ocfs2_meta_unlock(inode, meta_level);
+	meta_level = -1;
+
+	/* communicate with ocfs2_dio_end_io */
+	ocfs2_iocb_set_rw_locked(iocb);
+
+#ifdef OCFS2_ORACORE_WORKAROUNDS
+	if (osb->s_mount_opt & OCFS2_MOUNT_COMPAT_OCFS &&
+	    filp->f_flags & O_DIRECT) {
+		unsigned int saved_flags = filp->f_flags;
+		int sector_size = 1 << osb->s_sectsize_bits;
+
+		if ((saved_pos & (sector_size - 1)) ||
+		    (count & (sector_size - 1)) ||
+		    ((unsigned long)buf & (sector_size - 1))) {
+			filp->f_flags |= O_SYNC;
+			filp->f_flags &= ~O_DIRECT;
+		}
+
+		ret = generic_file_aio_write_nolock(iocb, &local_iov, 1,
+						    &iocb->ki_pos);
+
+		filp->f_flags = saved_flags;
+	} else
+#endif
+		ret = generic_file_aio_write_nolock(iocb, &local_iov, 1,
+						    &iocb->ki_pos);
+
+	/* buffered aio wouldn't have proper lock coverage today */
+	BUG_ON(ret == -EIOCBQUEUED && !(filp->f_flags & O_DIRECT));
+
+	/* 
+	 * deep in g_f_a_w_n()->ocfs2_direct_IO we pass in a ocfs2_dio_end_io
+	 * function pointer which is called when o_direct io completes so that
+	 * it can unlock our rw lock.  (it's the clustered equivalent of
+	 * i_alloc_sem; protects truncate from racing with pending ios).
+	 * Unfortunately there are error cases which call end_io and others
+	 * that don't.  so we don't have to unlock the rw_lock if either an
+	 * async dio is going to do it in the future or an end_io after an
+	 * error has already done it.
+	 */
+	if (ret == -EIOCBQUEUED || !ocfs2_iocb_is_rw_locked(iocb)) {
+		rw_level = -1;
+		have_alloc_sem = 0;
+	}
+
+out:
+	if (meta_level != -1)
+		ocfs2_meta_unlock(inode, meta_level);
+	if (have_alloc_sem)
+		up_read(&inode->i_alloc_sem);
+	if (rw_level != -1) 
+		ocfs2_rw_unlock(inode, rw_level);
+	up(&inode->i_sem);
+
+	mlog_exit(ret);
+	return ret;
+}
+
+static ssize_t ocfs2_file_aio_read(struct kiocb *iocb,
+				   char __user *buf,
+				   size_t count,
+				   loff_t pos)
+{
+	int ret = 0, rw_level = -1, have_alloc_sem = 0;
+	struct file *filp = iocb->ki_filp;
+	struct inode *inode = filp->f_dentry->d_inode;
+#ifdef OCFS2_ORACORE_WORKAROUNDS
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+#endif
+
+	mlog_entry("(0x%p, 0x%p, %u, '%.*s')\n", filp, buf,
+		   (unsigned int)count,
+		   filp->f_dentry->d_name.len,
+		   filp->f_dentry->d_name.name);
+
+	if (!inode) {
+		ret = -EINVAL;
+		mlog_errno(ret);
+		goto bail;
+	}
+
+#ifdef OCFS2_ORACORE_WORKAROUNDS
+	if (osb->s_mount_opt & OCFS2_MOUNT_COMPAT_OCFS) {
+		if (filp->f_flags & O_DIRECT) {
+			int sector_size = 1 << osb->s_sectsize_bits;
+
+			if ((pos & (sector_size - 1)) ||
+			    (count & (sector_size - 1)) ||
+			    ((unsigned long)buf & (sector_size - 1)) ||
+			    (i_size_read(inode) & (sector_size -1))) {
+				filp->f_flags &= ~O_DIRECT;
+			}
+		}
+	}
+#endif
+
+	/* 
+	 * buffered reads protect themselves in ->readpage().  O_DIRECT reads
+	 * need locks to protect pending reads from racing with truncate.
+	 */
+	if (filp->f_flags & O_DIRECT) {
+		down_read(&inode->i_alloc_sem);
+		have_alloc_sem = 1;
+
+		ret = ocfs2_rw_lock(inode, 0);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto bail;
+		}
+		rw_level = 0;
+		/* communicate with ocfs2_dio_end_io */
+		ocfs2_iocb_set_rw_locked(iocb);
+	}
+
+	ret = generic_file_aio_read(iocb, buf, count, iocb->ki_pos);
+	if (ret == -EINVAL)
+		mlog(ML_ERROR, "generic_file_aio_read returned -EINVAL\n");
+
+	/* buffered aio wouldn't have proper lock coverage today */
+	BUG_ON(ret == -EIOCBQUEUED && !(filp->f_flags & O_DIRECT));
+
+	/* see ocfs2_file_aio_write */
+	if (ret == -EIOCBQUEUED || !ocfs2_iocb_is_rw_locked(iocb)) {
+		rw_level = -1;
+		have_alloc_sem = 0;
+	}
+
+bail:
+	if (have_alloc_sem)
+		up_read(&inode->i_alloc_sem);
+	if (rw_level != -1) 
+		ocfs2_rw_unlock(inode, rw_level);
+	mlog_exit(ret);
+
+	return ret;
+}
+
+struct inode_operations ocfs2_file_iops = {
+	.setattr	= ocfs2_setattr,
+	.getattr	= ocfs2_getattr,
+};
+
+struct inode_operations ocfs2_special_file_iops = {
+	.setattr	= ocfs2_setattr,
+	.getattr	= ocfs2_getattr,
+};
+
+struct file_operations ocfs2_fops = {
+	.read		= do_sync_read,
+	.write		= do_sync_write,
+	.sendfile	= generic_file_sendfile,
+	.mmap		= ocfs2_mmap,
+	.fsync		= ocfs2_sync_file,
+	.release	= ocfs2_file_release,
+	.open		= ocfs2_file_open,
+	.aio_read	= ocfs2_file_aio_read,
+	.aio_write	= ocfs2_file_aio_write,
+};
+
+struct file_operations ocfs2_dops = {
+	.read		= generic_read_dir,
+	.readdir	= ocfs2_readdir,
+	.fsync		= ocfs2_sync_file,
+};
diff --git a/fs/ocfs2/file.h b/fs/ocfs2/file.h
new file mode 100644
index 00000000000..a5ea33b2406
--- /dev/null
+++ b/fs/ocfs2/file.h
@@ -0,0 +1,57 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * file.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_FILE_H
+#define OCFS2_FILE_H
+
+extern struct file_operations ocfs2_fops;
+extern struct file_operations ocfs2_dops;
+extern struct inode_operations ocfs2_file_iops;
+extern struct inode_operations ocfs2_special_file_iops;
+struct ocfs2_alloc_context;
+
+enum ocfs2_alloc_restarted {
+	RESTART_NONE = 0,
+	RESTART_TRANS,
+	RESTART_META
+};
+int ocfs2_do_extend_allocation(struct ocfs2_super *osb,
+			       struct inode *inode,
+			       u32 clusters_to_add,
+			       struct buffer_head *fe_bh,
+			       struct ocfs2_journal_handle *handle,
+			       struct ocfs2_alloc_context *data_ac,
+			       struct ocfs2_alloc_context *meta_ac,
+			       enum ocfs2_alloc_restarted *reason);
+int ocfs2_setattr(struct dentry *dentry, struct iattr *attr);
+int ocfs2_getattr(struct vfsmount *mnt, struct dentry *dentry,
+		  struct kstat *stat);
+
+int ocfs2_set_inode_size(struct ocfs2_journal_handle *handle,
+			 struct inode *inode,
+			 struct buffer_head *fe_bh,
+			 u64 new_i_size);
+
+#endif /* OCFS2_FILE_H */
diff --git a/fs/ocfs2/heartbeat.c b/fs/ocfs2/heartbeat.c
new file mode 100644
index 00000000000..0bbd22f46c8
--- /dev/null
+++ b/fs/ocfs2/heartbeat.c
@@ -0,0 +1,378 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * heartbeat.c
+ *
+ * Register ourselves with the heartbaet service, keep our node maps
+ * up to date, and fire off recovery when needed.
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/kmod.h>
+
+#include <cluster/heartbeat.h>
+#include <cluster/nodemanager.h>
+
+#include <dlm/dlmapi.h>
+
+#define MLOG_MASK_PREFIX ML_SUPER
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "heartbeat.h"
+#include "inode.h"
+#include "journal.h"
+#include "vote.h"
+
+#include "buffer_head_io.h"
+
+#define OCFS2_HB_NODE_DOWN_PRI     (0x0000002)
+#define OCFS2_HB_NODE_UP_PRI	   OCFS2_HB_NODE_DOWN_PRI
+
+static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map *map,
+					    int bit);
+static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map *map,
+					      int bit);
+static inline int __ocfs2_node_map_is_empty(struct ocfs2_node_map *map);
+static void __ocfs2_node_map_dup(struct ocfs2_node_map *target,
+				 struct ocfs2_node_map *from);
+static void __ocfs2_node_map_set(struct ocfs2_node_map *target,
+				 struct ocfs2_node_map *from);
+
+void ocfs2_init_node_maps(struct ocfs2_super *osb)
+{
+	spin_lock_init(&osb->node_map_lock);
+	ocfs2_node_map_init(&osb->mounted_map);
+	ocfs2_node_map_init(&osb->recovery_map);
+	ocfs2_node_map_init(&osb->umount_map);
+}
+
+static void ocfs2_do_node_down(int node_num,
+			       struct ocfs2_super *osb)
+{
+	BUG_ON(osb->node_num == node_num);
+
+	mlog(0, "ocfs2: node down event for %d\n", node_num);
+
+	if (!osb->dlm) {
+		/*
+		 * No DLM means we're not even ready to participate yet.
+		 * We check the slots after the DLM comes up, so we will
+		 * notice the node death then.  We can safely ignore it
+		 * here.
+		 */
+		return;
+	}
+
+	if (ocfs2_node_map_test_bit(osb, &osb->umount_map, node_num)) {
+		/* If a node is in the umount map, then we've been
+		 * expecting him to go down and we know ahead of time
+		 * that recovery is not necessary. */
+		ocfs2_node_map_clear_bit(osb, &osb->umount_map, node_num);
+		return;
+	}
+
+	ocfs2_recovery_thread(osb, node_num);
+
+	ocfs2_remove_node_from_vote_queues(osb, node_num);
+}
+
+static void ocfs2_hb_node_down_cb(struct o2nm_node *node,
+				  int node_num,
+				  void *data)
+{
+	ocfs2_do_node_down(node_num, (struct ocfs2_super *) data);
+}
+
+/* Called from the dlm when it's about to evict a node. We may also
+ * get a heartbeat callback later. */
+static void ocfs2_dlm_eviction_cb(int node_num,
+				  void *data)
+{
+	struct ocfs2_super *osb = (struct ocfs2_super *) data;
+	struct super_block *sb = osb->sb;
+
+	mlog(ML_NOTICE, "device (%u,%u): dlm has evicted node %d\n",
+	     MAJOR(sb->s_dev), MINOR(sb->s_dev), node_num);
+
+	ocfs2_do_node_down(node_num, osb);
+}
+
+static void ocfs2_hb_node_up_cb(struct o2nm_node *node,
+				int node_num,
+				void *data)
+{
+	struct ocfs2_super *osb = data;
+
+	BUG_ON(osb->node_num == node_num);
+
+	mlog(0, "node up event for %d\n", node_num);
+	ocfs2_node_map_clear_bit(osb, &osb->umount_map, node_num);
+}
+
+void ocfs2_setup_hb_callbacks(struct ocfs2_super *osb)
+{
+	o2hb_setup_callback(&osb->osb_hb_down, O2HB_NODE_DOWN_CB,
+			    ocfs2_hb_node_down_cb, osb,
+			    OCFS2_HB_NODE_DOWN_PRI);
+
+	o2hb_setup_callback(&osb->osb_hb_up, O2HB_NODE_UP_CB,
+			    ocfs2_hb_node_up_cb, osb, OCFS2_HB_NODE_UP_PRI);
+
+	/* Not exactly a heartbeat callback, but leads to essentially
+	 * the same path so we set it up here. */
+	dlm_setup_eviction_cb(&osb->osb_eviction_cb,
+			      ocfs2_dlm_eviction_cb,
+			      osb);
+}
+
+/* Most functions here are just stubs for now... */
+int ocfs2_register_hb_callbacks(struct ocfs2_super *osb)
+{
+	int status;
+
+	status = o2hb_register_callback(&osb->osb_hb_down);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = o2hb_register_callback(&osb->osb_hb_up);
+	if (status < 0)
+		mlog_errno(status);
+
+bail:
+	return status;
+}
+
+void ocfs2_clear_hb_callbacks(struct ocfs2_super *osb)
+{
+	int status;
+
+	status = o2hb_unregister_callback(&osb->osb_hb_down);
+	if (status < 0)
+		mlog_errno(status);
+
+	status = o2hb_unregister_callback(&osb->osb_hb_up);
+	if (status < 0)
+		mlog_errno(status);
+}
+
+void ocfs2_stop_heartbeat(struct ocfs2_super *osb)
+{
+	int ret;
+	char *argv[5], *envp[3];
+
+	if (!osb->uuid_str) {
+		/* This can happen if we don't get far enough in mount... */
+		mlog(0, "No UUID with which to stop heartbeat!\n\n");
+		return;
+	}
+
+	argv[0] = (char *)o2nm_get_hb_ctl_path();
+	argv[1] = "-K";
+	argv[2] = "-u";
+	argv[3] = osb->uuid_str;
+	argv[4] = NULL;
+
+	mlog(0, "Run: %s %s %s %s\n", argv[0], argv[1], argv[2], argv[3]);
+
+	/* minimal command environment taken from cpu_run_sbin_hotplug */
+	envp[0] = "HOME=/";
+	envp[1] = "PATH=/sbin:/bin:/usr/sbin:/usr/bin";
+	envp[2] = NULL;
+
+	ret = call_usermodehelper(argv[0], argv, envp, 1);
+	if (ret < 0)
+		mlog_errno(ret);
+}
+
+/* special case -1 for now
+ * TODO: should *really* make sure the calling func never passes -1!!  */
+void ocfs2_node_map_init(struct ocfs2_node_map *map)
+{
+	map->num_nodes = OCFS2_NODE_MAP_MAX_NODES;
+	memset(map->map, 0, BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES) *
+	       sizeof(unsigned long));
+}
+
+static inline void __ocfs2_node_map_set_bit(struct ocfs2_node_map *map,
+					    int bit)
+{
+	set_bit(bit, map->map);
+}
+
+void ocfs2_node_map_set_bit(struct ocfs2_super *osb,
+			    struct ocfs2_node_map *map,
+			    int bit)
+{
+	if (bit==-1)
+		return;
+	BUG_ON(bit >= map->num_nodes);
+	spin_lock(&osb->node_map_lock);
+	__ocfs2_node_map_set_bit(map, bit);
+	spin_unlock(&osb->node_map_lock);
+}
+
+static inline void __ocfs2_node_map_clear_bit(struct ocfs2_node_map *map,
+					      int bit)
+{
+	clear_bit(bit, map->map);
+}
+
+void ocfs2_node_map_clear_bit(struct ocfs2_super *osb,
+			      struct ocfs2_node_map *map,
+			      int bit)
+{
+	if (bit==-1)
+		return;
+	BUG_ON(bit >= map->num_nodes);
+	spin_lock(&osb->node_map_lock);
+	__ocfs2_node_map_clear_bit(map, bit);
+	spin_unlock(&osb->node_map_lock);
+}
+
+int ocfs2_node_map_test_bit(struct ocfs2_super *osb,
+			    struct ocfs2_node_map *map,
+			    int bit)
+{
+	int ret;
+	if (bit >= map->num_nodes) {
+		mlog(ML_ERROR, "bit=%d map->num_nodes=%d\n", bit, map->num_nodes);
+		BUG();
+	}
+	spin_lock(&osb->node_map_lock);
+	ret = test_bit(bit, map->map);
+	spin_unlock(&osb->node_map_lock);
+	return ret;
+}
+
+static inline int __ocfs2_node_map_is_empty(struct ocfs2_node_map *map)
+{
+	int bit;
+	bit = find_next_bit(map->map, map->num_nodes, 0);
+	if (bit < map->num_nodes)
+		return 0;
+	return 1;
+}
+
+int ocfs2_node_map_is_empty(struct ocfs2_super *osb,
+			    struct ocfs2_node_map *map)
+{
+	int ret;
+	BUG_ON(map->num_nodes == 0);
+	spin_lock(&osb->node_map_lock);
+	ret = __ocfs2_node_map_is_empty(map);
+	spin_unlock(&osb->node_map_lock);
+	return ret;
+}
+
+static void __ocfs2_node_map_dup(struct ocfs2_node_map *target,
+				 struct ocfs2_node_map *from)
+{
+	BUG_ON(from->num_nodes == 0);
+	ocfs2_node_map_init(target);
+	__ocfs2_node_map_set(target, from);
+}
+
+/* returns 1 if bit is the only bit set in target, 0 otherwise */
+int ocfs2_node_map_is_only(struct ocfs2_super *osb,
+			   struct ocfs2_node_map *target,
+			   int bit)
+{
+	struct ocfs2_node_map temp;
+	int ret;
+
+	spin_lock(&osb->node_map_lock);
+	__ocfs2_node_map_dup(&temp, target);
+	__ocfs2_node_map_clear_bit(&temp, bit);
+	ret = __ocfs2_node_map_is_empty(&temp);
+	spin_unlock(&osb->node_map_lock);
+
+	return ret;
+}
+
+static void __ocfs2_node_map_set(struct ocfs2_node_map *target,
+				 struct ocfs2_node_map *from)
+{
+	int num_longs, i;
+
+	BUG_ON(target->num_nodes != from->num_nodes);
+	BUG_ON(target->num_nodes == 0);
+
+	num_longs = BITS_TO_LONGS(target->num_nodes);
+	for (i = 0; i < num_longs; i++)
+		target->map[i] = from->map[i];
+}
+
+/* Returns whether the recovery bit was actually set - it may not be
+ * if a node is still marked as needing recovery */
+int ocfs2_recovery_map_set(struct ocfs2_super *osb,
+			   int num)
+{
+	int set = 0;
+
+	spin_lock(&osb->node_map_lock);
+
+	__ocfs2_node_map_clear_bit(&osb->mounted_map, num);
+
+	if (!test_bit(num, osb->recovery_map.map)) {
+	    __ocfs2_node_map_set_bit(&osb->recovery_map, num);
+	    set = 1;
+	}
+
+	spin_unlock(&osb->node_map_lock);
+
+	return set;
+}
+
+void ocfs2_recovery_map_clear(struct ocfs2_super *osb,
+			      int num)
+{
+	ocfs2_node_map_clear_bit(osb, &osb->recovery_map, num);
+}
+
+int ocfs2_node_map_iterate(struct ocfs2_super *osb,
+			   struct ocfs2_node_map *map,
+			   int idx)
+{
+	int i = idx;
+
+	idx = O2NM_INVALID_NODE_NUM;
+	spin_lock(&osb->node_map_lock);
+	if ((i != O2NM_INVALID_NODE_NUM) &&
+	    (i >= 0) &&
+	    (i < map->num_nodes)) {
+		while(i < map->num_nodes) {
+			if (test_bit(i, map->map)) {
+				idx = i;
+				break;
+			}
+			i++;
+		}
+	}
+	spin_unlock(&osb->node_map_lock);
+	return idx;
+}
diff --git a/fs/ocfs2/heartbeat.h b/fs/ocfs2/heartbeat.h
new file mode 100644
index 00000000000..e8fb079122e
--- /dev/null
+++ b/fs/ocfs2/heartbeat.h
@@ -0,0 +1,67 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * heartbeat.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_HEARTBEAT_H
+#define OCFS2_HEARTBEAT_H
+
+void ocfs2_init_node_maps(struct ocfs2_super *osb);
+
+void ocfs2_setup_hb_callbacks(struct ocfs2_super *osb);
+int ocfs2_register_hb_callbacks(struct ocfs2_super *osb);
+void ocfs2_clear_hb_callbacks(struct ocfs2_super *osb);
+void ocfs2_stop_heartbeat(struct ocfs2_super *osb);
+
+/* node map functions - used to keep track of mounted and in-recovery
+ * nodes. */
+void ocfs2_node_map_init(struct ocfs2_node_map *map);
+int ocfs2_node_map_is_empty(struct ocfs2_super *osb,
+			    struct ocfs2_node_map *map);
+void ocfs2_node_map_set_bit(struct ocfs2_super *osb,
+			    struct ocfs2_node_map *map,
+			    int bit);
+void ocfs2_node_map_clear_bit(struct ocfs2_super *osb,
+			      struct ocfs2_node_map *map,
+			      int bit);
+int ocfs2_node_map_test_bit(struct ocfs2_super *osb,
+			    struct ocfs2_node_map *map,
+			    int bit);
+int ocfs2_node_map_iterate(struct ocfs2_super *osb,
+			   struct ocfs2_node_map *map,
+			   int idx);
+static inline int ocfs2_node_map_first_set_bit(struct ocfs2_super *osb,
+					       struct ocfs2_node_map *map)
+{
+	return ocfs2_node_map_iterate(osb, map, 0);
+}
+int ocfs2_recovery_map_set(struct ocfs2_super *osb,
+			   int num);
+void ocfs2_recovery_map_clear(struct ocfs2_super *osb,
+			      int num);
+/* returns 1 if bit is the only bit set in target, 0 otherwise */
+int ocfs2_node_map_is_only(struct ocfs2_super *osb,
+			   struct ocfs2_node_map *target,
+			   int bit);
+
+#endif /* OCFS2_HEARTBEAT_H */
diff --git a/fs/ocfs2/inode.c b/fs/ocfs2/inode.c
new file mode 100644
index 00000000000..a91ba4dec93
--- /dev/null
+++ b/fs/ocfs2/inode.c
@@ -0,0 +1,1140 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * inode.c
+ *
+ * vfs' aops, fops, dops and iops
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/pagemap.h>
+#include <linux/smp_lock.h>
+
+#include <asm/byteorder.h>
+
+#define MLOG_MASK_PREFIX ML_INODE
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "dlmglue.h"
+#include "extent_map.h"
+#include "file.h"
+#include "inode.h"
+#include "journal.h"
+#include "namei.h"
+#include "suballoc.h"
+#include "super.h"
+#include "symlink.h"
+#include "sysfile.h"
+#include "uptodate.h"
+#include "vote.h"
+
+#include "buffer_head_io.h"
+
+#define OCFS2_FI_FLAG_NOWAIT	0x1
+#define OCFS2_FI_FLAG_DELETE	0x2
+struct ocfs2_find_inode_args
+{
+	u64		fi_blkno;
+	unsigned long	fi_ino;
+	unsigned int	fi_flags;
+};
+
+static int ocfs2_read_locked_inode(struct inode *inode,
+				   struct ocfs2_find_inode_args *args);
+static int ocfs2_init_locked_inode(struct inode *inode, void *opaque);
+static int ocfs2_find_actor(struct inode *inode, void *opaque);
+static int ocfs2_truncate_for_delete(struct ocfs2_super *osb,
+				    struct inode *inode,
+				    struct buffer_head *fe_bh);
+
+struct inode *ocfs2_ilookup_for_vote(struct ocfs2_super *osb,
+				     u64 blkno,
+				     int delete_vote)
+{
+	struct ocfs2_find_inode_args args;
+
+	/* ocfs2_ilookup_for_vote should *only* be called from the
+	 * vote thread */
+	BUG_ON(current != osb->vote_task);
+
+	args.fi_blkno = blkno;
+	args.fi_flags = OCFS2_FI_FLAG_NOWAIT;
+	if (delete_vote)
+		args.fi_flags |= OCFS2_FI_FLAG_DELETE;
+	args.fi_ino = ino_from_blkno(osb->sb, blkno);
+	return ilookup5(osb->sb, args.fi_ino, ocfs2_find_actor, &args);
+}
+
+struct inode *ocfs2_iget(struct ocfs2_super *osb, u64 blkno)
+{
+	struct inode *inode = NULL;
+	struct super_block *sb = osb->sb;
+	struct ocfs2_find_inode_args args;
+
+	mlog_entry("(blkno = %"MLFu64")\n", blkno);
+
+	/* Ok. By now we've either got the offsets passed to us by the
+	 * caller, or we just pulled them off the bh. Lets do some
+	 * sanity checks to make sure they're OK. */
+	if (blkno == 0) {
+		inode = ERR_PTR(-EINVAL);
+		mlog_errno(PTR_ERR(inode));
+		goto bail;
+	}
+
+	args.fi_blkno = blkno;
+	args.fi_flags = 0;
+	args.fi_ino = ino_from_blkno(sb, blkno);
+
+	inode = iget5_locked(sb, args.fi_ino, ocfs2_find_actor,
+			     ocfs2_init_locked_inode, &args);
+	/* inode was *not* in the inode cache. 2.6.x requires
+	 * us to do our own read_inode call and unlock it
+	 * afterwards. */
+	if (inode && inode->i_state & I_NEW) {
+		mlog(0, "Inode was not in inode cache, reading it.\n");
+		ocfs2_read_locked_inode(inode, &args);
+		unlock_new_inode(inode);
+	}
+	if (inode == NULL) {
+		inode = ERR_PTR(-ENOMEM);
+		mlog_errno(PTR_ERR(inode));
+		goto bail;
+	}
+	if (is_bad_inode(inode)) {
+		iput(inode);
+		inode = ERR_PTR(-ESTALE);
+		mlog_errno(PTR_ERR(inode));
+		goto bail;
+	}
+
+bail:
+	if (!IS_ERR(inode)) {
+		mlog(0, "returning inode with number %"MLFu64"\n",
+		     OCFS2_I(inode)->ip_blkno);
+		mlog_exit_ptr(inode);
+	} else
+		mlog_errno(PTR_ERR(inode));
+
+	return inode;
+}
+
+
+/*
+ * here's how inodes get read from disk:
+ * iget5_locked -> find_actor -> OCFS2_FIND_ACTOR
+ * found? : return the in-memory inode
+ * not found? : get_new_inode -> OCFS2_INIT_LOCKED_INODE
+ */
+
+static int ocfs2_find_actor(struct inode *inode, void *opaque)
+{
+	struct ocfs2_find_inode_args *args = NULL;
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+	int ret = 0;
+
+	mlog_entry("(0x%p, %lu, 0x%p)\n", inode, inode->i_ino, opaque);
+
+	args = opaque;
+
+	mlog_bug_on_msg(!inode, "No inode in find actor!\n");
+
+	if (oi->ip_blkno != args->fi_blkno)
+		goto bail;
+
+	/* OCFS2_FI_FLAG_NOWAIT is *only* set from
+	 * ocfs2_ilookup_for_vote which won't create an inode for one
+	 * that isn't found. The vote thread which doesn't want to get
+	 * an inode which is in the process of going away - otherwise
+	 * the call to __wait_on_freeing_inode in find_inode_fast will
+	 * cause it to deadlock on an inode which may be waiting on a
+	 * vote (or lock release) in delete_inode */
+	if ((args->fi_flags & OCFS2_FI_FLAG_NOWAIT) &&
+	    (inode->i_state & (I_FREEING|I_CLEAR))) {
+		/* As stated above, we're not going to return an
+		 * inode.  In the case of a delete vote, the voting
+		 * code is going to signal the other node to go
+		 * ahead. Mark that state here, so this freeing inode
+		 * has the state when it gets to delete_inode. */
+		if (args->fi_flags & OCFS2_FI_FLAG_DELETE) {
+			spin_lock(&oi->ip_lock);
+			ocfs2_mark_inode_remotely_deleted(inode);
+			spin_unlock(&oi->ip_lock);
+		}
+		goto bail;
+	}
+
+	ret = 1;
+bail:
+	mlog_exit(ret);
+	return ret;
+}
+
+/*
+ * initialize the new inode, but don't do anything that would cause
+ * us to sleep.
+ * return 0 on success, 1 on failure
+ */
+static int ocfs2_init_locked_inode(struct inode *inode, void *opaque)
+{
+	struct ocfs2_find_inode_args *args = opaque;
+
+	mlog_entry("inode = %p, opaque = %p\n", inode, opaque);
+
+	inode->i_ino = args->fi_ino;
+	OCFS2_I(inode)->ip_blkno = args->fi_blkno;
+
+	mlog_exit(0);
+	return 0;
+}
+
+int ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe,
+		     	 int create_ino)
+{
+	struct super_block *sb;
+	struct ocfs2_super *osb;
+	int status = -EINVAL;
+
+	mlog_entry("(0x%p, size:%"MLFu64")\n", inode, fe->i_size);
+
+	sb = inode->i_sb;
+	osb = OCFS2_SB(sb);
+
+	/* this means that read_inode cannot create a superblock inode
+	 * today.  change if needed. */
+	if (!OCFS2_IS_VALID_DINODE(fe) ||
+	    !(fe->i_flags & cpu_to_le32(OCFS2_VALID_FL))) {
+		mlog(ML_ERROR, "Invalid dinode: i_ino=%lu, i_blkno=%"MLFu64", "
+		     "signature = %.*s, flags = 0x%x\n",
+		     inode->i_ino, le64_to_cpu(fe->i_blkno), 7,
+		     fe->i_signature, le32_to_cpu(fe->i_flags));
+		goto bail;
+	}
+
+	if (le32_to_cpu(fe->i_fs_generation) != osb->fs_generation) {
+		mlog(ML_ERROR, "file entry generation does not match "
+		     "superblock! osb->fs_generation=%x, "
+		     "fe->i_fs_generation=%x\n",
+		     osb->fs_generation, le32_to_cpu(fe->i_fs_generation));
+		goto bail;
+	}
+
+	inode->i_version = 1;
+	inode->i_generation = le32_to_cpu(fe->i_generation);
+	inode->i_rdev = huge_decode_dev(le64_to_cpu(fe->id1.dev1.i_rdev));
+	inode->i_mode = le16_to_cpu(fe->i_mode);
+	inode->i_uid = le32_to_cpu(fe->i_uid);
+	inode->i_gid = le32_to_cpu(fe->i_gid);
+	inode->i_blksize = (u32)osb->s_clustersize;
+
+	/* Fast symlinks will have i_size but no allocated clusters. */
+	if (S_ISLNK(inode->i_mode) && !fe->i_clusters)
+		inode->i_blocks = 0;
+	else
+		inode->i_blocks =
+			ocfs2_align_bytes_to_sectors(le64_to_cpu(fe->i_size));
+	inode->i_mapping->a_ops = &ocfs2_aops;
+	inode->i_flags |= S_NOATIME;
+	inode->i_atime.tv_sec = le64_to_cpu(fe->i_atime);
+	inode->i_atime.tv_nsec = le32_to_cpu(fe->i_atime_nsec);
+	inode->i_mtime.tv_sec = le64_to_cpu(fe->i_mtime);
+	inode->i_mtime.tv_nsec = le32_to_cpu(fe->i_mtime_nsec);
+	inode->i_ctime.tv_sec = le64_to_cpu(fe->i_ctime);
+	inode->i_ctime.tv_nsec = le32_to_cpu(fe->i_ctime_nsec);
+
+	if (OCFS2_I(inode)->ip_blkno != le64_to_cpu(fe->i_blkno))
+		mlog(ML_ERROR,
+		     "ip_blkno %"MLFu64" != i_blkno %"MLFu64"!\n",
+		     OCFS2_I(inode)->ip_blkno, fe->i_blkno);
+
+	OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
+	OCFS2_I(inode)->ip_orphaned_slot = OCFS2_INVALID_SLOT;
+
+	if (create_ino)
+		inode->i_ino = ino_from_blkno(inode->i_sb,
+			       le64_to_cpu(fe->i_blkno));
+
+	mlog(0, "blkno = %"MLFu64", ino = %lu, create_ino = %s\n",
+	     fe->i_blkno, inode->i_ino, create_ino ? "true" : "false");
+
+	inode->i_nlink = le16_to_cpu(fe->i_links_count);
+
+	if (fe->i_flags & cpu_to_le32(OCFS2_LOCAL_ALLOC_FL)) {
+		OCFS2_I(inode)->ip_flags |= OCFS2_INODE_BITMAP;
+		mlog(0, "local alloc inode: i_ino=%lu\n", inode->i_ino);
+	} else if (fe->i_flags & cpu_to_le32(OCFS2_BITMAP_FL)) {
+		OCFS2_I(inode)->ip_flags |= OCFS2_INODE_BITMAP;
+	} else if (fe->i_flags & cpu_to_le32(OCFS2_SUPER_BLOCK_FL)) {
+		mlog(0, "superblock inode: i_ino=%lu\n", inode->i_ino);
+		/* we can't actually hit this as read_inode can't
+		 * handle superblocks today ;-) */
+		BUG();
+	}
+
+	switch (inode->i_mode & S_IFMT) {
+	    case S_IFREG:
+		    inode->i_fop = &ocfs2_fops;
+		    inode->i_op = &ocfs2_file_iops;
+		    i_size_write(inode, le64_to_cpu(fe->i_size));
+		    break;
+	    case S_IFDIR:
+		    inode->i_op = &ocfs2_dir_iops;
+		    inode->i_fop = &ocfs2_dops;
+		    i_size_write(inode, le64_to_cpu(fe->i_size));
+		    break;
+	    case S_IFLNK:
+		    if (ocfs2_inode_is_fast_symlink(inode))
+			inode->i_op = &ocfs2_fast_symlink_inode_operations;
+		    else
+			inode->i_op = &ocfs2_symlink_inode_operations;
+		    i_size_write(inode, le64_to_cpu(fe->i_size));
+		    break;
+	    default:
+		    inode->i_op = &ocfs2_special_file_iops;
+		    init_special_inode(inode, inode->i_mode,
+				       inode->i_rdev);
+		    break;
+	}
+
+	ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_rw_lockres,
+				  OCFS2_LOCK_TYPE_RW, inode);
+	ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_meta_lockres,
+				  OCFS2_LOCK_TYPE_META, inode);
+	ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_data_lockres,
+				  OCFS2_LOCK_TYPE_DATA, inode);
+
+	status = 0;
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_read_locked_inode(struct inode *inode,
+				   struct ocfs2_find_inode_args *args)
+{
+	struct super_block *sb;
+	struct ocfs2_super *osb;
+	struct ocfs2_dinode *fe;
+	struct buffer_head *bh = NULL;
+	int status;
+	int sysfile = 0;
+
+	mlog_entry("(0x%p, 0x%p)\n", inode, args);
+
+	status = -EINVAL;
+	if (inode == NULL || inode->i_sb == NULL) {
+		mlog(ML_ERROR, "bad inode\n");
+		goto bail;
+	}
+	sb = inode->i_sb;
+	osb = OCFS2_SB(sb);
+
+	if (!args) {
+		mlog(ML_ERROR, "bad inode args\n");
+		make_bad_inode(inode);
+		goto bail;
+	}
+
+	/* Read the FE off disk. This is safe because the kernel only
+	 * does one read_inode2 for a new inode, and if it doesn't
+	 * exist yet then nobody can be working on it! */
+	status = ocfs2_read_block(osb, args->fi_blkno, &bh, 0, NULL);
+	if (status < 0) {
+		mlog_errno(status);
+		make_bad_inode(inode);
+		goto bail;
+	}
+
+	fe = (struct ocfs2_dinode *) bh->b_data;
+	if (!OCFS2_IS_VALID_DINODE(fe)) {
+		mlog(ML_ERROR, "Invalid dinode #%"MLFu64": signature = %.*s\n",
+		     fe->i_blkno, 7, fe->i_signature);
+		make_bad_inode(inode);
+		goto bail;
+	}
+
+	if (fe->i_flags & cpu_to_le32(OCFS2_SYSTEM_FL))
+		sysfile = 1;
+
+	if (S_ISCHR(le16_to_cpu(fe->i_mode)) ||
+	    S_ISBLK(le16_to_cpu(fe->i_mode)))
+    		inode->i_rdev = huge_decode_dev(le64_to_cpu(fe->id1.dev1.i_rdev));
+
+	status = -EINVAL;
+	if (ocfs2_populate_inode(inode, fe, 0) < 0) {
+		mlog(ML_ERROR, "populate inode failed! i_blkno=%"MLFu64", "
+		     "i_ino=%lu\n", fe->i_blkno, inode->i_ino);
+		make_bad_inode(inode);
+		goto bail;
+	}
+
+	BUG_ON(args->fi_blkno != le64_to_cpu(fe->i_blkno));
+
+	if (sysfile)
+	       OCFS2_I(inode)->ip_flags |= OCFS2_INODE_SYSTEM_FILE;
+
+	status = 0;
+
+bail:
+	if (args && bh)
+		brelse(bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+void ocfs2_sync_blockdev(struct super_block *sb)
+{
+	sync_blockdev(sb->s_bdev);
+}
+
+static int ocfs2_truncate_for_delete(struct ocfs2_super *osb,
+				     struct inode *inode,
+				     struct buffer_head *fe_bh)
+{
+	int status = 0;
+	struct ocfs2_journal_handle *handle = NULL;
+	struct ocfs2_truncate_context *tc = NULL;
+	struct ocfs2_dinode *fe;
+
+	mlog_entry_void();
+
+	fe = (struct ocfs2_dinode *) fe_bh->b_data;
+
+	/* zero allocation, zero truncate :) */
+	if (!fe->i_clusters)
+		goto bail;
+
+	handle = ocfs2_start_trans(osb, handle, OCFS2_INODE_UPDATE_CREDITS);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		handle = NULL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_set_inode_size(handle, inode, fe_bh, 0ULL);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	ocfs2_commit_trans(handle);
+	handle = NULL;
+
+	status = ocfs2_prepare_truncate(osb, inode, fe_bh, &tc);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_commit_truncate(osb, inode, fe_bh, tc);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+bail:
+	if (handle)
+		ocfs2_commit_trans(handle);
+
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_remove_inode(struct inode *inode,
+			      struct buffer_head *di_bh,
+			      struct inode *orphan_dir_inode,
+			      struct buffer_head *orphan_dir_bh)
+{
+	int status;
+	struct inode *inode_alloc_inode = NULL;
+	struct buffer_head *inode_alloc_bh = NULL;
+	struct ocfs2_journal_handle *handle;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	struct ocfs2_dinode *di = (struct ocfs2_dinode *) di_bh->b_data;
+
+	inode_alloc_inode =
+		ocfs2_get_system_file_inode(osb, INODE_ALLOC_SYSTEM_INODE,
+					    le16_to_cpu(di->i_suballoc_slot));
+	if (!inode_alloc_inode) {
+		status = -EEXIST;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	down(&inode_alloc_inode->i_sem);
+	status = ocfs2_meta_lock(inode_alloc_inode, NULL, &inode_alloc_bh, 1);
+	if (status < 0) {
+		up(&inode_alloc_inode->i_sem);
+
+		mlog_errno(status);
+		goto bail;
+	}
+
+	handle = ocfs2_start_trans(osb, NULL, OCFS2_DELETE_INODE_CREDITS);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		mlog_errno(status);
+		goto bail_unlock;
+	}
+
+	status = ocfs2_orphan_del(osb, handle, orphan_dir_inode, inode,
+				  orphan_dir_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail_commit;
+	}
+
+	/* set the inodes dtime */
+	status = ocfs2_journal_access(handle, inode, di_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail_commit;
+	}
+
+	di->i_dtime = cpu_to_le64(CURRENT_TIME.tv_sec);
+	le32_and_cpu(&di->i_flags, ~(OCFS2_VALID_FL | OCFS2_ORPHANED_FL));
+
+	status = ocfs2_journal_dirty(handle, di_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail_commit;
+	}
+
+	ocfs2_remove_from_cache(inode, di_bh);
+
+	status = ocfs2_free_dinode(handle, inode_alloc_inode,
+				   inode_alloc_bh, di);
+	if (status < 0)
+		mlog_errno(status);
+
+bail_commit:
+	ocfs2_commit_trans(handle);
+bail_unlock:
+	ocfs2_meta_unlock(inode_alloc_inode, 1);
+	up(&inode_alloc_inode->i_sem);
+	brelse(inode_alloc_bh);
+bail:
+	iput(inode_alloc_inode);
+
+	return status;
+}
+
+static int ocfs2_wipe_inode(struct inode *inode,
+			    struct buffer_head *di_bh)
+{
+	int status, orphaned_slot;
+	struct inode *orphan_dir_inode = NULL;
+	struct buffer_head *orphan_dir_bh = NULL;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	/* We've already voted on this so it should be readonly - no
+	 * spinlock needed. */
+	orphaned_slot = OCFS2_I(inode)->ip_orphaned_slot;
+	orphan_dir_inode = ocfs2_get_system_file_inode(osb,
+						       ORPHAN_DIR_SYSTEM_INODE,
+						       orphaned_slot);
+	if (!orphan_dir_inode) {
+		status = -EEXIST;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* Lock the orphan dir. The lock will be held for the entire
+	 * delete_inode operation. We do this now to avoid races with
+	 * recovery completion on other nodes. */
+	down(&orphan_dir_inode->i_sem);
+	status = ocfs2_meta_lock(orphan_dir_inode, NULL, &orphan_dir_bh, 1);
+	if (status < 0) {
+		up(&orphan_dir_inode->i_sem);
+
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* we do this while holding the orphan dir lock because we
+	 * don't want recovery being run from another node to vote for
+	 * an inode delete on us -- this will result in two nodes
+	 * truncating the same file! */
+	status = ocfs2_truncate_for_delete(osb, inode, di_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail_unlock_dir;
+	}
+
+	status = ocfs2_remove_inode(inode, di_bh, orphan_dir_inode,
+				    orphan_dir_bh);
+	if (status < 0)
+		mlog_errno(status);
+
+bail_unlock_dir:
+	ocfs2_meta_unlock(orphan_dir_inode, 1);
+	up(&orphan_dir_inode->i_sem);
+	brelse(orphan_dir_bh);
+bail:
+	iput(orphan_dir_inode);
+
+	return status;
+}
+
+/* There is a series of simple checks that should be done before a
+ * vote is even considered. Encapsulate those in this function. */
+static int ocfs2_inode_is_valid_to_delete(struct inode *inode)
+{
+	int ret = 0;
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	/* We shouldn't be getting here for the root directory
+	 * inode.. */
+	if (inode == osb->root_inode) {
+		mlog(ML_ERROR, "Skipping delete of root inode.\n");
+		goto bail;
+	}
+
+	/* If we're coming from process_vote we can't go into our own
+	 * voting [hello, deadlock city!], so unforuntately we just
+	 * have to skip deleting this guy. That's OK though because
+	 * the node who's doing the actual deleting should handle it
+	 * anyway. */
+	if (current == osb->vote_task) {
+		mlog(0, "Skipping delete of %lu because we're currently "
+		     "in process_vote\n", inode->i_ino);
+		goto bail;
+	}
+
+	spin_lock(&oi->ip_lock);
+	/* OCFS2 *never* deletes system files. This should technically
+	 * never get here as system file inodes should always have a
+	 * positive link count. */
+	if (oi->ip_flags & OCFS2_INODE_SYSTEM_FILE) {
+		mlog(ML_ERROR, "Skipping delete of system file %"MLFu64".\n",
+		     oi->ip_blkno);
+		goto bail_unlock;
+	}
+
+	/* If we have voted "yes" on the wipe of this inode for
+	 * another node, it will be marked here so we can safely skip
+	 * it. Recovery will cleanup any inodes we might inadvertantly
+	 * skip here. */
+	if (oi->ip_flags & OCFS2_INODE_SKIP_DELETE) {
+		mlog(0, "Skipping delete of %lu because another node "
+		     "has done this for us.\n", inode->i_ino);
+		goto bail_unlock;
+	}
+
+	ret = 1;
+bail_unlock:
+	spin_unlock(&oi->ip_lock);
+bail:
+	return ret;
+}
+
+/* Query the cluster to determine whether we should wipe an inode from
+ * disk or not.
+ *
+ * Requires the inode to have the cluster lock. */
+static int ocfs2_query_inode_wipe(struct inode *inode,
+				  struct buffer_head *di_bh,
+				  int *wipe)
+{
+	int status = 0;
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+	struct ocfs2_dinode *di;
+
+	*wipe = 0;
+
+	/* While we were waiting for the cluster lock in
+	 * ocfs2_delete_inode, another node might have asked to delete
+	 * the inode. Recheck our flags to catch this. */
+	if (!ocfs2_inode_is_valid_to_delete(inode)) {
+		mlog(0, "Skipping delete of %"MLFu64" because flags changed\n",
+		     oi->ip_blkno);
+		goto bail;
+	}
+
+	/* Now that we have an up to date inode, we can double check
+	 * the link count. */
+	if (inode->i_nlink) {
+		mlog(0, "Skipping delete of %"MLFu64" because nlink = %u\n",
+		     oi->ip_blkno, inode->i_nlink);
+		goto bail;
+	}
+
+	/* Do some basic inode verification... */
+	di = (struct ocfs2_dinode *) di_bh->b_data;
+	if (!(di->i_flags & cpu_to_le32(OCFS2_ORPHANED_FL))) {
+		/* for lack of a better error? */
+		status = -EEXIST;
+		mlog(ML_ERROR,
+		     "Inode %"MLFu64" (on-disk %"MLFu64") not orphaned! "
+		     "Disk flags  0x%x, inode flags 0x%x\n",
+		     oi->ip_blkno, di->i_blkno, di->i_flags, oi->ip_flags);
+		goto bail;
+	}
+
+	/* has someone already deleted us?! baaad... */
+	if (di->i_dtime) {
+		status = -EEXIST;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_request_delete_vote(inode);
+	/* -EBUSY means that other nodes are still using the
+	 * inode. We're done here though, so avoid doing anything on
+	 * disk and let them worry about deleting it. */
+	if (status == -EBUSY) {
+		status = 0;
+		mlog(0, "Skipping delete of %"MLFu64" because it is in use on"
+		     "other nodes\n", oi->ip_blkno);
+		goto bail;
+	}
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	spin_lock(&oi->ip_lock);
+	if (oi->ip_orphaned_slot == OCFS2_INVALID_SLOT) {
+		/* Nobody knew which slot this inode was orphaned
+		 * into. This may happen during node death and
+		 * recovery knows how to clean it up so we can safely
+		 * ignore this inode for now on. */
+		mlog(0, "Nobody knew where inode %"MLFu64" was orphaned!\n",
+		     oi->ip_blkno);
+	} else {
+		*wipe = 1;
+
+		mlog(0, "Inode %"MLFu64" is ok to wipe from orphan dir %d\n",
+		     oi->ip_blkno, oi->ip_orphaned_slot);
+	}
+	spin_unlock(&oi->ip_lock);
+
+bail:
+	return status;
+}
+
+/* Support function for ocfs2_delete_inode. Will help us keep the
+ * inode data in a consistent state for clear_inode. Always truncates
+ * pages, optionally sync's them first. */
+static void ocfs2_cleanup_delete_inode(struct inode *inode,
+				       int sync_data)
+{
+	mlog(0, "Cleanup inode %"MLFu64", sync = %d\n",
+	     OCFS2_I(inode)->ip_blkno, sync_data);
+	if (sync_data)
+		write_inode_now(inode, 1);
+	truncate_inode_pages(&inode->i_data, 0);
+}
+
+void ocfs2_delete_inode(struct inode *inode)
+{
+	int wipe, status;
+	sigset_t blocked, oldset;
+	struct buffer_head *di_bh = NULL;
+
+	mlog_entry("(inode->i_ino = %lu)\n", inode->i_ino);
+
+	if (is_bad_inode(inode)) {
+		mlog(0, "Skipping delete of bad inode\n");
+		goto bail;
+	}
+
+	if (!ocfs2_inode_is_valid_to_delete(inode)) {
+		/* It's probably not necessary to truncate_inode_pages
+		 * here but we do it for safety anyway (it will most
+		 * likely be a no-op anyway) */
+		ocfs2_cleanup_delete_inode(inode, 0);
+		goto bail;
+	}
+
+	/* We want to block signals in delete_inode as the lock and
+	 * messaging paths may return us -ERESTARTSYS. Which would
+	 * cause us to exit early, resulting in inodes being orphaned
+	 * forever. */
+	sigfillset(&blocked);
+	status = sigprocmask(SIG_BLOCK, &blocked, &oldset);
+	if (status < 0) {
+		mlog_errno(status);
+		ocfs2_cleanup_delete_inode(inode, 1);
+		goto bail;
+	}
+
+	/* Lock down the inode. This gives us an up to date view of
+	 * it's metadata (for verification), and allows us to
+	 * serialize delete_inode votes. 
+	 *
+	 * Even though we might be doing a truncate, we don't take the
+	 * allocation lock here as it won't be needed - nobody will
+	 * have the file open.
+	 */
+	status = ocfs2_meta_lock(inode, NULL, &di_bh, 1);
+	if (status < 0) {
+		if (status != -ENOENT)
+			mlog_errno(status);
+		ocfs2_cleanup_delete_inode(inode, 0);
+		goto bail_unblock;
+	}
+
+	/* Query the cluster. This will be the final decision made
+	 * before we go ahead and wipe the inode. */
+	status = ocfs2_query_inode_wipe(inode, di_bh, &wipe);
+	if (!wipe || status < 0) {
+		/* Error and inode busy vote both mean we won't be
+		 * removing the inode, so they take almost the same
+		 * path. */
+		if (status < 0)
+			mlog_errno(status);
+
+		/* Someone in the cluster has voted to not wipe this
+		 * inode, or it was never completely orphaned. Write
+		 * out the pages and exit now. */
+		ocfs2_cleanup_delete_inode(inode, 1);
+		goto bail_unlock_inode;
+	}
+
+	ocfs2_cleanup_delete_inode(inode, 0);
+
+	status = ocfs2_wipe_inode(inode, di_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail_unlock_inode;
+	}
+
+	/* Mark the inode as successfully deleted. This is important
+	 * for ocfs2_clear_inode as it will check this flag and skip
+	 * any checkpointing work */
+	OCFS2_I(inode)->ip_flags |= OCFS2_INODE_DELETED;
+
+bail_unlock_inode:
+	ocfs2_meta_unlock(inode, 1);
+	brelse(di_bh);
+bail_unblock:
+	status = sigprocmask(SIG_SETMASK, &oldset, NULL);
+	if (status < 0)
+		mlog_errno(status);
+bail:
+	clear_inode(inode);
+	mlog_exit_void();
+}
+
+void ocfs2_clear_inode(struct inode *inode)
+{
+	int status;
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+
+	mlog_entry_void();
+
+	if (!inode)
+		goto bail;
+
+	mlog(0, "Clearing inode: %"MLFu64", nlink = %u\n",
+	     OCFS2_I(inode)->ip_blkno, inode->i_nlink);
+
+	mlog_bug_on_msg(OCFS2_SB(inode->i_sb) == NULL,
+			"Inode=%lu\n", inode->i_ino);
+
+	/* Do these before all the other work so that we don't bounce
+	 * the vote thread while waiting to destroy the locks. */
+	ocfs2_mark_lockres_freeing(&oi->ip_rw_lockres);
+	ocfs2_mark_lockres_freeing(&oi->ip_meta_lockres);
+	ocfs2_mark_lockres_freeing(&oi->ip_data_lockres);
+
+	/* We very well may get a clear_inode before all an inodes
+	 * metadata has hit disk. Of course, we can't drop any cluster
+	 * locks until the journal has finished with it. The only
+	 * exception here are successfully wiped inodes - their
+	 * metadata can now be considered to be part of the system
+	 * inodes from which it came. */
+	if (!(OCFS2_I(inode)->ip_flags & OCFS2_INODE_DELETED))
+		ocfs2_checkpoint_inode(inode);
+
+	mlog_bug_on_msg(!list_empty(&oi->ip_io_markers),
+			"Clear inode of %"MLFu64", inode has io markers\n",
+			oi->ip_blkno);
+
+	ocfs2_extent_map_drop(inode, 0);
+	ocfs2_extent_map_init(inode);
+
+	status = ocfs2_drop_inode_locks(inode);
+	if (status < 0)
+		mlog_errno(status);
+
+	ocfs2_lock_res_free(&oi->ip_rw_lockres);
+	ocfs2_lock_res_free(&oi->ip_meta_lockres);
+	ocfs2_lock_res_free(&oi->ip_data_lockres);
+
+	ocfs2_metadata_cache_purge(inode);
+
+	mlog_bug_on_msg(oi->ip_metadata_cache.ci_num_cached,
+			"Clear inode of %"MLFu64", inode has %u cache items\n",
+			oi->ip_blkno, oi->ip_metadata_cache.ci_num_cached);
+
+	mlog_bug_on_msg(!(oi->ip_flags & OCFS2_INODE_CACHE_INLINE),
+			"Clear inode of %"MLFu64", inode has a bad flag\n",
+			oi->ip_blkno);
+
+	mlog_bug_on_msg(spin_is_locked(&oi->ip_lock),
+			"Clear inode of %"MLFu64", inode is locked\n",
+			oi->ip_blkno);
+
+	mlog_bug_on_msg(down_trylock(&oi->ip_io_sem),
+			"Clear inode of %"MLFu64", io_sem is locked\n",
+			oi->ip_blkno);
+	up(&oi->ip_io_sem);
+
+	/*
+	 * down_trylock() returns 0, down_write_trylock() returns 1
+	 * kernel 1, world 0
+	 */
+	mlog_bug_on_msg(!down_write_trylock(&oi->ip_alloc_sem),
+			"Clear inode of %"MLFu64", alloc_sem is locked\n",
+			oi->ip_blkno);
+	up_write(&oi->ip_alloc_sem);
+
+	mlog_bug_on_msg(oi->ip_open_count,
+			"Clear inode of %"MLFu64" has open count %d\n",
+			oi->ip_blkno, oi->ip_open_count);
+	mlog_bug_on_msg(!list_empty(&oi->ip_handle_list),
+			"Clear inode of %"MLFu64" has non empty handle list\n",
+			oi->ip_blkno);
+	mlog_bug_on_msg(oi->ip_handle,
+			"Clear inode of %"MLFu64" has non empty handle pointer\n",
+			oi->ip_blkno);
+
+	/* Clear all other flags. */
+	oi->ip_flags = OCFS2_INODE_CACHE_INLINE;
+	oi->ip_created_trans = 0;
+	oi->ip_last_trans = 0;
+	oi->ip_dir_start_lookup = 0;
+	oi->ip_blkno = 0ULL;
+
+bail:
+	mlog_exit_void();
+}
+
+/* Called under inode_lock, with no more references on the
+ * struct inode, so it's safe here to check the flags field
+ * and to manipulate i_nlink without any other locks. */
+void ocfs2_drop_inode(struct inode *inode)
+{
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+
+	mlog_entry_void();
+
+	mlog(0, "Drop inode %"MLFu64", nlink = %u, ip_flags = 0x%x\n",
+	     oi->ip_blkno, inode->i_nlink, oi->ip_flags);
+
+	/* Testing ip_orphaned_slot here wouldn't work because we may
+	 * not have gotten a delete_inode vote from any other nodes
+	 * yet. */
+	if (oi->ip_flags & OCFS2_INODE_MAYBE_ORPHANED) {
+		mlog(0, "Inode was orphaned on another node, clearing nlink.\n");
+		inode->i_nlink = 0;
+	}
+
+	generic_drop_inode(inode);
+
+	mlog_exit_void();
+}
+
+/*
+ * TODO: this should probably be merged into ocfs2_get_block
+ *
+ * However, you now need to pay attention to the cont_prepare_write()
+ * stuff in ocfs2_get_block (that is, ocfs2_get_block pretty much
+ * expects never to extend).
+ */
+struct buffer_head *ocfs2_bread(struct inode *inode,
+				int block, int *err, int reada)
+{
+	struct buffer_head *bh = NULL;
+	int tmperr;
+	u64 p_blkno;
+	int readflags = OCFS2_BH_CACHED;
+
+#if 0
+	/* only turn this on if we know we can deal with read_block
+	 * returning nothing */
+	if (reada)
+		readflags |= OCFS2_BH_READAHEAD;
+#endif
+
+	if (((u64)block << inode->i_sb->s_blocksize_bits) >=
+	    i_size_read(inode)) {
+		BUG_ON(!reada);
+		return NULL;
+	}
+
+	tmperr = ocfs2_extent_map_get_blocks(inode, block, 1,
+					     &p_blkno, NULL);
+	if (tmperr < 0) {
+		mlog_errno(tmperr);
+		goto fail;
+	}
+
+	tmperr = ocfs2_read_block(OCFS2_SB(inode->i_sb), p_blkno, &bh,
+				  readflags, inode);
+	if (tmperr < 0)
+		goto fail;
+
+	tmperr = 0;
+
+	*err = 0;
+	return bh;
+
+fail:
+	if (bh) {
+		brelse(bh);
+		bh = NULL;
+	}
+	*err = -EIO;
+	return NULL;
+}
+
+/*
+ * This is called from our getattr.
+ */
+int ocfs2_inode_revalidate(struct dentry *dentry)
+{
+	struct inode *inode = dentry->d_inode;
+	int status = 0;
+
+	mlog_entry("(inode = 0x%p, ino = %"MLFu64")\n", inode,
+		   inode ? OCFS2_I(inode)->ip_blkno : 0ULL);
+
+	if (!inode) {
+		mlog(0, "eep, no inode!\n");
+		status = -ENOENT;
+		goto bail;
+	}
+
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+	if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_DELETED) {
+		spin_unlock(&OCFS2_I(inode)->ip_lock);
+		mlog(0, "inode deleted!\n");
+		status = -ENOENT;
+		goto bail;
+	}
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+	/* Let ocfs2_meta_lock do the work of updating our struct
+	 * inode for us. */
+	status = ocfs2_meta_lock(inode, NULL, NULL, 0);
+	if (status < 0) {
+		if (status != -ENOENT)
+			mlog_errno(status);
+		goto bail;
+	}
+	ocfs2_meta_unlock(inode, 0);
+bail:
+	mlog_exit(status);
+
+	return status;
+}
+
+/*
+ * Updates a disk inode from a
+ * struct inode.
+ * Only takes ip_lock.
+ */
+int ocfs2_mark_inode_dirty(struct ocfs2_journal_handle *handle,
+			   struct inode *inode,
+			   struct buffer_head *bh)
+{
+	int status;
+	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) bh->b_data;
+
+	mlog_entry("(inode %"MLFu64")\n", OCFS2_I(inode)->ip_blkno);
+
+	status = ocfs2_journal_access(handle, inode, bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+	fe->i_clusters = cpu_to_le32(OCFS2_I(inode)->ip_clusters);
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+	fe->i_size = cpu_to_le64(i_size_read(inode));
+	fe->i_links_count = cpu_to_le16(inode->i_nlink);
+	fe->i_uid = cpu_to_le32(inode->i_uid);
+	fe->i_gid = cpu_to_le32(inode->i_gid);
+	fe->i_mode = cpu_to_le16(inode->i_mode);
+	fe->i_atime = cpu_to_le64(inode->i_atime.tv_sec);
+	fe->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec);
+	fe->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
+	fe->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
+	fe->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec);
+	fe->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec);
+
+	status = ocfs2_journal_dirty(handle, bh);
+	if (status < 0)
+		mlog_errno(status);
+
+	status = 0;
+leave:
+
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ *
+ * Updates a struct inode from a disk inode.
+ * does no i/o, only takes ip_lock.
+ */
+void ocfs2_refresh_inode(struct inode *inode,
+			 struct ocfs2_dinode *fe)
+{
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+
+	OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
+	i_size_write(inode, le64_to_cpu(fe->i_size));
+	inode->i_nlink = le16_to_cpu(fe->i_links_count);
+	inode->i_uid = le32_to_cpu(fe->i_uid);
+	inode->i_gid = le32_to_cpu(fe->i_gid);
+	inode->i_mode = le16_to_cpu(fe->i_mode);
+	inode->i_blksize = (u32) osb->s_clustersize;
+	if (S_ISLNK(inode->i_mode) && le32_to_cpu(fe->i_clusters) == 0)
+		inode->i_blocks = 0;
+	else
+		inode->i_blocks = ocfs2_align_bytes_to_sectors(i_size_read(inode));
+	inode->i_atime.tv_sec = le64_to_cpu(fe->i_atime);
+	inode->i_atime.tv_nsec = le32_to_cpu(fe->i_atime_nsec);
+	inode->i_mtime.tv_sec = le64_to_cpu(fe->i_mtime);
+	inode->i_mtime.tv_nsec = le32_to_cpu(fe->i_mtime_nsec);
+	inode->i_ctime.tv_sec = le64_to_cpu(fe->i_ctime);
+	inode->i_ctime.tv_nsec = le32_to_cpu(fe->i_ctime_nsec);
+
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+}
diff --git a/fs/ocfs2/inode.h b/fs/ocfs2/inode.h
new file mode 100644
index 00000000000..9b017743365
--- /dev/null
+++ b/fs/ocfs2/inode.h
@@ -0,0 +1,145 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * inode.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_INODE_H
+#define OCFS2_INODE_H
+
+/* OCFS2 Inode Private Data */
+struct ocfs2_inode_info
+{
+	u64			ip_blkno;
+
+	struct ocfs2_lock_res		ip_rw_lockres;
+	struct ocfs2_lock_res		ip_meta_lockres;
+	struct ocfs2_lock_res		ip_data_lockres;
+
+	/* protects allocation changes on this inode. */
+	struct rw_semaphore		ip_alloc_sem;
+
+	/* These fields are protected by ip_lock */
+	spinlock_t			ip_lock;
+	u32				ip_open_count;
+	u32				ip_clusters;
+	struct ocfs2_extent_map		ip_map;
+	struct list_head		ip_io_markers;
+	int				ip_orphaned_slot;
+
+	struct semaphore		ip_io_sem;
+
+	/* Used by the journalling code to attach an inode to a
+	 * handle.  These are protected by ip_io_sem in order to lock
+	 * out other I/O to the inode until we either commit or
+	 * abort. */
+	struct list_head		ip_handle_list;
+	struct ocfs2_journal_handle	*ip_handle;
+
+	u32				ip_flags; /* see below */
+
+	/* protected by recovery_lock. */
+	struct inode			*ip_next_orphan;
+
+	u32				ip_dir_start_lookup;
+
+	/* next two are protected by trans_inc_lock */
+	/* which transaction were we created on? Zero if none. */
+	unsigned long			ip_created_trans;
+	/* last transaction we were a part of. */
+	unsigned long			ip_last_trans;
+
+	struct ocfs2_caching_info	ip_metadata_cache;
+
+	struct inode			vfs_inode;
+};
+
+/*
+ * Flags for the ip_flags field
+ */
+/* System file inodes  */
+#define OCFS2_INODE_SYSTEM_FILE		0x00000001
+#define OCFS2_INODE_JOURNAL		0x00000002
+#define OCFS2_INODE_BITMAP		0x00000004
+/* This inode has been wiped from disk */
+#define OCFS2_INODE_DELETED		0x00000008
+/* Another node is deleting, so our delete is a nop */
+#define OCFS2_INODE_SKIP_DELETE		0x00000010
+/* Has the inode been orphaned on another node?
+ *
+ * This hints to ocfs2_drop_inode that it should clear i_nlink before
+ * continuing.
+ *
+ * We *only* set this on unlink vote from another node. If the inode
+ * was locally orphaned, then we're sure of the state and don't need
+ * to twiddle i_nlink later - it's either zero or not depending on
+ * whether our unlink succeeded. Otherwise we got this from a node
+ * whose intention was to orphan the inode, however he may have
+ * crashed, failed etc, so we let ocfs2_drop_inode zero the value and
+ * rely on ocfs2_delete_inode to sort things out under the proper
+ * cluster locks.
+ */
+#define OCFS2_INODE_MAYBE_ORPHANED	0x00000020
+/* Does someone have the file open O_DIRECT */
+#define OCFS2_INODE_OPEN_DIRECT		0x00000040
+/* Indicates that the metadata cache should be used as an array. */
+#define OCFS2_INODE_CACHE_INLINE	0x00000080
+
+static inline struct ocfs2_inode_info *OCFS2_I(struct inode *inode)
+{
+	return container_of(inode, struct ocfs2_inode_info, vfs_inode);
+}
+
+#define INODE_JOURNAL(i) (OCFS2_I(i)->ip_flags & OCFS2_INODE_JOURNAL)
+#define SET_INODE_JOURNAL(i) (OCFS2_I(i)->ip_flags |= OCFS2_INODE_JOURNAL)
+
+extern kmem_cache_t *ocfs2_inode_cache;
+
+extern struct address_space_operations ocfs2_aops;
+
+struct buffer_head *ocfs2_bread(struct inode *inode, int block,
+				int *err, int reada);
+void ocfs2_clear_inode(struct inode *inode);
+void ocfs2_delete_inode(struct inode *inode);
+void ocfs2_drop_inode(struct inode *inode);
+struct inode *ocfs2_iget(struct ocfs2_super *osb, u64 feoff);
+struct inode *ocfs2_ilookup_for_vote(struct ocfs2_super *osb,
+				     u64 blkno,
+				     int delete_vote);
+int ocfs2_inode_init_private(struct inode *inode);
+int ocfs2_inode_revalidate(struct dentry *dentry);
+int ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe,
+			 int create_ino);
+void ocfs2_read_inode(struct inode *inode);
+void ocfs2_read_inode2(struct inode *inode, void *opaque);
+ssize_t ocfs2_rw_direct(int rw, struct file *filp, char *buf,
+			size_t size, loff_t *offp);
+void ocfs2_sync_blockdev(struct super_block *sb);
+void ocfs2_refresh_inode(struct inode *inode,
+			 struct ocfs2_dinode *fe);
+int ocfs2_mark_inode_dirty(struct ocfs2_journal_handle *handle,
+			   struct inode *inode,
+			   struct buffer_head *bh);
+int ocfs2_aio_read(struct file *file, struct kiocb *req, struct iocb *iocb);
+int ocfs2_aio_write(struct file *file, struct kiocb *req, struct iocb *iocb);
+
+#endif /* OCFS2_INODE_H */
diff --git a/fs/ocfs2/journal.c b/fs/ocfs2/journal.c
new file mode 100644
index 00000000000..04428042e5e
--- /dev/null
+++ b/fs/ocfs2/journal.c
@@ -0,0 +1,1652 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * journal.c
+ *
+ * Defines functions of journalling api
+ *
+ * Copyright (C) 2003, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/kthread.h>
+
+#define MLOG_MASK_PREFIX ML_JOURNAL
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "dlmglue.h"
+#include "extent_map.h"
+#include "heartbeat.h"
+#include "inode.h"
+#include "journal.h"
+#include "localalloc.h"
+#include "namei.h"
+#include "slot_map.h"
+#include "super.h"
+#include "vote.h"
+#include "sysfile.h"
+
+#include "buffer_head_io.h"
+
+spinlock_t trans_inc_lock = SPIN_LOCK_UNLOCKED;
+
+static int ocfs2_force_read_journal(struct inode *inode);
+static int ocfs2_recover_node(struct ocfs2_super *osb,
+			      int node_num);
+static int __ocfs2_recovery_thread(void *arg);
+static int ocfs2_commit_cache(struct ocfs2_super *osb);
+static int ocfs2_wait_on_mount(struct ocfs2_super *osb);
+static void ocfs2_handle_cleanup_locks(struct ocfs2_journal *journal,
+				       struct ocfs2_journal_handle *handle);
+static void ocfs2_commit_unstarted_handle(struct ocfs2_journal_handle *handle);
+static int ocfs2_journal_toggle_dirty(struct ocfs2_super *osb,
+				      int dirty);
+static int ocfs2_trylock_journal(struct ocfs2_super *osb,
+				 int slot_num);
+static int ocfs2_recover_orphans(struct ocfs2_super *osb,
+				 int slot);
+static int ocfs2_commit_thread(void *arg);
+
+static int ocfs2_commit_cache(struct ocfs2_super *osb)
+{
+	int status = 0;
+	unsigned int flushed;
+	unsigned long old_id;
+	struct ocfs2_journal *journal = NULL;
+
+	mlog_entry_void();
+
+	journal = osb->journal;
+
+	/* Flush all pending commits and checkpoint the journal. */
+	down_write(&journal->j_trans_barrier);
+
+	if (atomic_read(&journal->j_num_trans) == 0) {
+		up_write(&journal->j_trans_barrier);
+		mlog(0, "No transactions for me to flush!\n");
+		goto finally;
+	}
+
+	journal_lock_updates(journal->j_journal);
+	status = journal_flush(journal->j_journal);
+	journal_unlock_updates(journal->j_journal);
+	if (status < 0) {
+		up_write(&journal->j_trans_barrier);
+		mlog_errno(status);
+		goto finally;
+	}
+
+	old_id = ocfs2_inc_trans_id(journal);
+
+	flushed = atomic_read(&journal->j_num_trans);
+	atomic_set(&journal->j_num_trans, 0);
+	up_write(&journal->j_trans_barrier);
+
+	mlog(0, "commit_thread: flushed transaction %lu (%u handles)\n",
+	     journal->j_trans_id, flushed);
+
+	ocfs2_kick_vote_thread(osb);
+	wake_up(&journal->j_checkpointed);
+finally:
+	mlog_exit(status);
+	return status;
+}
+
+struct ocfs2_journal_handle *ocfs2_alloc_handle(struct ocfs2_super *osb)
+{
+	struct ocfs2_journal_handle *retval = NULL;
+
+	retval = kcalloc(1, sizeof(*retval), GFP_KERNEL);
+	if (!retval) {
+		mlog(ML_ERROR, "Failed to allocate memory for journal "
+		     "handle!\n");
+		return NULL;
+	}
+
+	retval->max_buffs = 0;
+	retval->num_locks = 0;
+	retval->k_handle = NULL;
+
+	INIT_LIST_HEAD(&retval->locks);
+	INIT_LIST_HEAD(&retval->inode_list);
+	retval->journal = osb->journal;
+
+	return retval;
+}
+
+/* pass it NULL and it will allocate a new handle object for you.  If
+ * you pass it a handle however, it may still return error, in which
+ * case it has free'd the passed handle for you. */
+struct ocfs2_journal_handle *ocfs2_start_trans(struct ocfs2_super *osb,
+					       struct ocfs2_journal_handle *handle,
+					       int max_buffs)
+{
+	int ret;
+	journal_t *journal = osb->journal->j_journal;
+
+	mlog_entry("(max_buffs = %d)\n", max_buffs);
+
+	if (!osb || !osb->journal->j_journal)
+		BUG();
+
+	if (ocfs2_is_hard_readonly(osb)) {
+		ret = -EROFS;
+		goto done_free;
+	}
+
+	BUG_ON(osb->journal->j_state == OCFS2_JOURNAL_FREE);
+	BUG_ON(max_buffs <= 0);
+
+	/* JBD might support this, but our journalling code doesn't yet. */
+	if (journal_current_handle()) {
+		mlog(ML_ERROR, "Recursive transaction attempted!\n");
+		BUG();
+	}
+
+	if (!handle)
+		handle = ocfs2_alloc_handle(osb);
+	if (!handle) {
+		ret = -ENOMEM;
+		mlog(ML_ERROR, "Failed to allocate memory for journal "
+		     "handle!\n");
+		goto done_free;
+	}
+
+	handle->max_buffs = max_buffs;
+
+	down_read(&osb->journal->j_trans_barrier);
+
+	/* actually start the transaction now */
+	handle->k_handle = journal_start(journal, max_buffs);
+	if (IS_ERR(handle->k_handle)) {
+		up_read(&osb->journal->j_trans_barrier);
+
+		ret = PTR_ERR(handle->k_handle);
+		handle->k_handle = NULL;
+		mlog_errno(ret);
+
+		if (is_journal_aborted(journal)) {
+			ocfs2_abort(osb->sb, "Detected aborted journal");
+			ret = -EROFS;
+		}
+		goto done_free;
+	}
+
+	atomic_inc(&(osb->journal->j_num_trans));
+	handle->flags |= OCFS2_HANDLE_STARTED;
+
+	mlog_exit_ptr(handle);
+	return handle;
+
+done_free:
+	if (handle)
+		ocfs2_commit_unstarted_handle(handle); /* will kfree handle */
+
+	mlog_exit(ret);
+	return ERR_PTR(ret);
+}
+
+void ocfs2_handle_add_inode(struct ocfs2_journal_handle *handle,
+			    struct inode *inode)
+{
+	BUG_ON(!handle);
+	BUG_ON(!inode);
+
+	atomic_inc(&inode->i_count);
+
+	/* we're obviously changing it... */
+	down(&inode->i_sem);
+
+	/* sanity check */
+	BUG_ON(OCFS2_I(inode)->ip_handle);
+	BUG_ON(!list_empty(&OCFS2_I(inode)->ip_handle_list));
+
+	OCFS2_I(inode)->ip_handle = handle;
+	list_del(&(OCFS2_I(inode)->ip_handle_list));
+	list_add_tail(&(OCFS2_I(inode)->ip_handle_list), &(handle->inode_list));
+}
+
+static void ocfs2_handle_unlock_inodes(struct ocfs2_journal_handle *handle)
+{
+	struct list_head *p, *n;
+	struct inode *inode;
+	struct ocfs2_inode_info *oi;
+
+	list_for_each_safe(p, n, &handle->inode_list) {
+		oi = list_entry(p, struct ocfs2_inode_info,
+				ip_handle_list);
+		inode = &oi->vfs_inode;
+
+		OCFS2_I(inode)->ip_handle = NULL;
+		list_del_init(&OCFS2_I(inode)->ip_handle_list);
+
+		up(&inode->i_sem);
+		iput(inode);
+	}
+}
+
+/* This is trivial so we do it out of the main commit
+ * paths. Beware, it can be called from start_trans too! */
+static void ocfs2_commit_unstarted_handle(struct ocfs2_journal_handle *handle)
+{
+	mlog_entry_void();
+
+	BUG_ON(handle->flags & OCFS2_HANDLE_STARTED);
+
+	ocfs2_handle_unlock_inodes(handle);
+	/* You are allowed to add journal locks before the transaction
+	 * has started. */
+	ocfs2_handle_cleanup_locks(handle->journal, handle);
+
+	kfree(handle);
+
+	mlog_exit_void();
+}
+
+void ocfs2_commit_trans(struct ocfs2_journal_handle *handle)
+{
+	handle_t *jbd_handle;
+	int retval;
+	struct ocfs2_journal *journal = handle->journal;
+
+	mlog_entry_void();
+
+	BUG_ON(!handle);
+
+	if (!(handle->flags & OCFS2_HANDLE_STARTED)) {
+		ocfs2_commit_unstarted_handle(handle);
+		mlog_exit_void();
+		return;
+	}
+
+	/* release inode semaphores we took during this transaction */
+	ocfs2_handle_unlock_inodes(handle);
+
+	/* ocfs2_extend_trans may have had to call journal_restart
+	 * which will always commit the transaction, but may return
+	 * error for any number of reasons. If this is the case, we
+	 * clear k_handle as it's not valid any more. */
+	if (handle->k_handle) {
+		jbd_handle = handle->k_handle;
+
+		if (handle->flags & OCFS2_HANDLE_SYNC)
+			jbd_handle->h_sync = 1;
+		else
+			jbd_handle->h_sync = 0;
+
+		/* actually stop the transaction. if we've set h_sync,
+		 * it'll have been committed when we return */
+		retval = journal_stop(jbd_handle);
+		if (retval < 0) {
+			mlog_errno(retval);
+			mlog(ML_ERROR, "Could not commit transaction\n");
+			BUG();
+		}
+
+		handle->k_handle = NULL; /* it's been free'd in journal_stop */
+	}
+
+	ocfs2_handle_cleanup_locks(journal, handle);
+
+	up_read(&journal->j_trans_barrier);
+
+	kfree(handle);
+	mlog_exit_void();
+}
+
+/*
+ * 'nblocks' is what you want to add to the current
+ * transaction. extend_trans will either extend the current handle by
+ * nblocks, or commit it and start a new one with nblocks credits.
+ *
+ * WARNING: This will not release any semaphores or disk locks taken
+ * during the transaction, so make sure they were taken *before*
+ * start_trans or we'll have ordering deadlocks.
+ *
+ * WARNING2: Note that we do *not* drop j_trans_barrier here. This is
+ * good because transaction ids haven't yet been recorded on the
+ * cluster locks associated with this handle.
+ */
+int ocfs2_extend_trans(struct ocfs2_journal_handle *handle,
+		       int nblocks)
+{
+	int status;
+
+	BUG_ON(!handle);
+	BUG_ON(!(handle->flags & OCFS2_HANDLE_STARTED));
+	BUG_ON(!nblocks);
+
+	mlog_entry_void();
+
+	mlog(0, "Trying to extend transaction by %d blocks\n", nblocks);
+
+	status = journal_extend(handle->k_handle, nblocks);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	if (status > 0) {
+		mlog(0, "journal_extend failed, trying journal_restart\n");
+		status = journal_restart(handle->k_handle, nblocks);
+		if (status < 0) {
+			handle->k_handle = NULL;
+			mlog_errno(status);
+			goto bail;
+		}
+		handle->max_buffs = nblocks;
+	} else
+		handle->max_buffs += nblocks;
+
+	status = 0;
+bail:
+
+	mlog_exit(status);
+	return status;
+}
+
+int ocfs2_journal_access(struct ocfs2_journal_handle *handle,
+			 struct inode *inode,
+			 struct buffer_head *bh,
+			 int type)
+{
+	int status;
+
+	BUG_ON(!inode);
+	BUG_ON(!handle);
+	BUG_ON(!bh);
+	BUG_ON(!(handle->flags & OCFS2_HANDLE_STARTED));
+
+	mlog_entry("bh->b_blocknr=%llu, type=%d (\"%s\"), bh->b_size = %hu\n",
+		   (unsigned long long)bh->b_blocknr, type,
+		   (type == OCFS2_JOURNAL_ACCESS_CREATE) ?
+		   "OCFS2_JOURNAL_ACCESS_CREATE" :
+		   "OCFS2_JOURNAL_ACCESS_WRITE",
+		   bh->b_size);
+
+	/* we can safely remove this assertion after testing. */
+	if (!buffer_uptodate(bh)) {
+		mlog(ML_ERROR, "giving me a buffer that's not uptodate!\n");
+		mlog(ML_ERROR, "b_blocknr=%llu\n",
+		     (unsigned long long)bh->b_blocknr);
+		BUG();
+	}
+
+	/* Set the current transaction information on the inode so
+	 * that the locking code knows whether it can drop it's locks
+	 * on this inode or not. We're protected from the commit
+	 * thread updating the current transaction id until
+	 * ocfs2_commit_trans() because ocfs2_start_trans() took
+	 * j_trans_barrier for us. */
+	ocfs2_set_inode_lock_trans(OCFS2_SB(inode->i_sb)->journal, inode);
+
+	down(&OCFS2_I(inode)->ip_io_sem);
+	switch (type) {
+	case OCFS2_JOURNAL_ACCESS_CREATE:
+	case OCFS2_JOURNAL_ACCESS_WRITE:
+		status = journal_get_write_access(handle->k_handle, bh);
+		break;
+
+	case OCFS2_JOURNAL_ACCESS_UNDO:
+		status = journal_get_undo_access(handle->k_handle, bh);
+		break;
+
+	default:
+		status = -EINVAL;
+		mlog(ML_ERROR, "Uknown access type!\n");
+	}
+	up(&OCFS2_I(inode)->ip_io_sem);
+
+	if (status < 0)
+		mlog(ML_ERROR, "Error %d getting %d access to buffer!\n",
+		     status, type);
+
+	mlog_exit(status);
+	return status;
+}
+
+int ocfs2_journal_dirty(struct ocfs2_journal_handle *handle,
+			struct buffer_head *bh)
+{
+	int status;
+
+	BUG_ON(!(handle->flags & OCFS2_HANDLE_STARTED));
+
+	mlog_entry("(bh->b_blocknr=%llu)\n",
+		   (unsigned long long)bh->b_blocknr);
+
+	status = journal_dirty_metadata(handle->k_handle, bh);
+	if (status < 0)
+		mlog(ML_ERROR, "Could not dirty metadata buffer. "
+		     "(bh->b_blocknr=%llu)\n",
+		     (unsigned long long)bh->b_blocknr);
+
+	mlog_exit(status);
+	return status;
+}
+
+int ocfs2_journal_dirty_data(handle_t *handle,
+			     struct buffer_head *bh)
+{
+	int err = journal_dirty_data(handle, bh);
+	if (err)
+		mlog_errno(err);
+	/* TODO: When we can handle it, abort the handle and go RO on
+	 * error here. */
+
+	return err;
+}
+
+/* We always assume you're adding a metadata lock at level 'ex' */
+int ocfs2_handle_add_lock(struct ocfs2_journal_handle *handle,
+			  struct inode *inode)
+{
+	int status;
+	struct ocfs2_journal_lock *lock;
+
+	BUG_ON(!inode);
+
+	lock = kmem_cache_alloc(ocfs2_lock_cache, GFP_NOFS);
+	if (!lock) {
+		status = -ENOMEM;
+		mlog_errno(-ENOMEM);
+		goto bail;
+	}
+
+	if (!igrab(inode))
+		BUG();
+	lock->jl_inode = inode;
+
+	list_add_tail(&(lock->jl_lock_list), &(handle->locks));
+	handle->num_locks++;
+
+	status = 0;
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+static void ocfs2_handle_cleanup_locks(struct ocfs2_journal *journal,
+				       struct ocfs2_journal_handle *handle)
+{
+	struct list_head *p, *n;
+	struct ocfs2_journal_lock *lock;
+	struct inode *inode;
+
+	list_for_each_safe(p, n, &(handle->locks)) {
+		lock = list_entry(p, struct ocfs2_journal_lock,
+				  jl_lock_list);
+		list_del(&lock->jl_lock_list);
+		handle->num_locks--;
+
+		inode = lock->jl_inode;
+		ocfs2_meta_unlock(inode, 1);
+		if (atomic_read(&inode->i_count) == 1)
+			mlog(ML_ERROR,
+			     "Inode %"MLFu64", I'm doing a last iput for!",
+			     OCFS2_I(inode)->ip_blkno);
+		iput(inode);
+		kmem_cache_free(ocfs2_lock_cache, lock);
+	}
+}
+
+#define OCFS2_DEFAULT_COMMIT_INTERVAL 	(HZ * 5)
+
+void ocfs2_set_journal_params(struct ocfs2_super *osb)
+{
+	journal_t *journal = osb->journal->j_journal;
+
+	spin_lock(&journal->j_state_lock);
+	journal->j_commit_interval = OCFS2_DEFAULT_COMMIT_INTERVAL;
+	if (osb->s_mount_opt & OCFS2_MOUNT_BARRIER)
+		journal->j_flags |= JFS_BARRIER;
+	else
+		journal->j_flags &= ~JFS_BARRIER;
+	spin_unlock(&journal->j_state_lock);
+}
+
+int ocfs2_journal_init(struct ocfs2_journal *journal, int *dirty)
+{
+	int status = -1;
+	struct inode *inode = NULL; /* the journal inode */
+	journal_t *j_journal = NULL;
+	struct ocfs2_dinode *di = NULL;
+	struct buffer_head *bh = NULL;
+	struct ocfs2_super *osb;
+	int meta_lock = 0;
+
+	mlog_entry_void();
+
+	BUG_ON(!journal);
+
+	osb = journal->j_osb;
+
+	/* already have the inode for our journal */
+	inode = ocfs2_get_system_file_inode(osb, JOURNAL_SYSTEM_INODE,
+					    osb->slot_num);
+	if (inode == NULL) {
+		status = -EACCES;
+		mlog_errno(status);
+		goto done;
+	}
+	if (is_bad_inode(inode)) {
+		mlog(ML_ERROR, "access error (bad inode)\n");
+		iput(inode);
+		inode = NULL;
+		status = -EACCES;
+		goto done;
+	}
+
+	SET_INODE_JOURNAL(inode);
+	OCFS2_I(inode)->ip_open_count++;
+
+	status = ocfs2_meta_lock(inode, NULL, &bh, 1);
+	if (status < 0) {
+		if (status != -ERESTARTSYS)
+			mlog(ML_ERROR, "Could not get lock on journal!\n");
+		goto done;
+	}
+
+	meta_lock = 1;
+	di = (struct ocfs2_dinode *)bh->b_data;
+
+	if (inode->i_size <  OCFS2_MIN_JOURNAL_SIZE) {
+		mlog(ML_ERROR, "Journal file size (%lld) is too small!\n",
+		     inode->i_size);
+		status = -EINVAL;
+		goto done;
+	}
+
+	mlog(0, "inode->i_size = %lld\n", inode->i_size);
+	mlog(0, "inode->i_blocks = %lu\n", inode->i_blocks);
+	mlog(0, "inode->ip_clusters = %u\n", OCFS2_I(inode)->ip_clusters);
+
+	/* call the kernels journal init function now */
+	j_journal = journal_init_inode(inode);
+	if (j_journal == NULL) {
+		mlog(ML_ERROR, "Linux journal layer error\n");
+		status = -EINVAL;
+		goto done;
+	}
+
+	mlog(0, "Returned from journal_init_inode\n");
+	mlog(0, "j_journal->j_maxlen = %u\n", j_journal->j_maxlen);
+
+	*dirty = (le32_to_cpu(di->id1.journal1.ij_flags) &
+		  OCFS2_JOURNAL_DIRTY_FL);
+
+	journal->j_journal = j_journal;
+	journal->j_inode = inode;
+	journal->j_bh = bh;
+
+	ocfs2_set_journal_params(osb);
+
+	journal->j_state = OCFS2_JOURNAL_LOADED;
+
+	status = 0;
+done:
+	if (status < 0) {
+		if (meta_lock)
+			ocfs2_meta_unlock(inode, 1);
+		if (bh != NULL)
+			brelse(bh);
+		if (inode) {
+			OCFS2_I(inode)->ip_open_count--;
+			iput(inode);
+		}
+	}
+
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_journal_toggle_dirty(struct ocfs2_super *osb,
+				      int dirty)
+{
+	int status;
+	unsigned int flags;
+	struct ocfs2_journal *journal = osb->journal;
+	struct buffer_head *bh = journal->j_bh;
+	struct ocfs2_dinode *fe;
+
+	mlog_entry_void();
+
+	fe = (struct ocfs2_dinode *)bh->b_data;
+	if (!OCFS2_IS_VALID_DINODE(fe)) {
+		/* This is called from startup/shutdown which will
+		 * handle the errors in a specific manner, so no need
+		 * to call ocfs2_error() here. */
+		mlog(ML_ERROR, "Journal dinode %"MLFu64"  has invalid "
+		     "signature: %.*s", fe->i_blkno, 7, fe->i_signature);
+		status = -EIO;
+		goto out;
+	}
+
+	flags = le32_to_cpu(fe->id1.journal1.ij_flags);
+	if (dirty)
+		flags |= OCFS2_JOURNAL_DIRTY_FL;
+	else
+		flags &= ~OCFS2_JOURNAL_DIRTY_FL;
+	fe->id1.journal1.ij_flags = cpu_to_le32(flags);
+
+	status = ocfs2_write_block(osb, bh, journal->j_inode);
+	if (status < 0)
+		mlog_errno(status);
+
+out:
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * If the journal has been kmalloc'd it needs to be freed after this
+ * call.
+ */
+void ocfs2_journal_shutdown(struct ocfs2_super *osb)
+{
+	struct ocfs2_journal *journal = NULL;
+	int status = 0;
+	struct inode *inode = NULL;
+	int num_running_trans = 0;
+
+	mlog_entry_void();
+
+	if (!osb)
+		BUG();
+
+	journal = osb->journal;
+	if (!journal)
+		goto done;
+
+	inode = journal->j_inode;
+
+	if (journal->j_state != OCFS2_JOURNAL_LOADED)
+		goto done;
+
+	/* need to inc inode use count as journal_destroy will iput. */
+	if (!igrab(inode))
+		BUG();
+
+	num_running_trans = atomic_read(&(osb->journal->j_num_trans));
+	if (num_running_trans > 0)
+		mlog(0, "Shutting down journal: must wait on %d "
+		     "running transactions!\n",
+		     num_running_trans);
+
+	/* Do a commit_cache here. It will flush our journal, *and*
+	 * release any locks that are still held.
+	 * set the SHUTDOWN flag and release the trans lock.
+	 * the commit thread will take the trans lock for us below. */
+	journal->j_state = OCFS2_JOURNAL_IN_SHUTDOWN;
+
+	/* The OCFS2_JOURNAL_IN_SHUTDOWN will signal to commit_cache to not
+	 * drop the trans_lock (which we want to hold until we
+	 * completely destroy the journal. */
+	if (osb->commit_task) {
+		/* Wait for the commit thread */
+		mlog(0, "Waiting for ocfs2commit to exit....\n");
+		kthread_stop(osb->commit_task);
+		osb->commit_task = NULL;
+	}
+
+	BUG_ON(atomic_read(&(osb->journal->j_num_trans)) != 0);
+
+	status = ocfs2_journal_toggle_dirty(osb, 0);
+	if (status < 0)
+		mlog_errno(status);
+
+	/* Shutdown the kernel journal system */
+	journal_destroy(journal->j_journal);
+
+	OCFS2_I(inode)->ip_open_count--;
+
+	/* unlock our journal */
+	ocfs2_meta_unlock(inode, 1);
+
+	brelse(journal->j_bh);
+	journal->j_bh = NULL;
+
+	journal->j_state = OCFS2_JOURNAL_FREE;
+
+//	up_write(&journal->j_trans_barrier);
+done:
+	if (inode)
+		iput(inode);
+	mlog_exit_void();
+}
+
+static void ocfs2_clear_journal_error(struct super_block *sb,
+				      journal_t *journal,
+				      int slot)
+{
+	int olderr;
+
+	olderr = journal_errno(journal);
+	if (olderr) {
+		mlog(ML_ERROR, "File system error %d recorded in "
+		     "journal %u.\n", olderr, slot);
+		mlog(ML_ERROR, "File system on device %s needs checking.\n",
+		     sb->s_id);
+
+		journal_ack_err(journal);
+		journal_clear_err(journal);
+	}
+}
+
+int ocfs2_journal_load(struct ocfs2_journal *journal)
+{
+	int status = 0;
+	struct ocfs2_super *osb;
+
+	mlog_entry_void();
+
+	if (!journal)
+		BUG();
+
+	osb = journal->j_osb;
+
+	status = journal_load(journal->j_journal);
+	if (status < 0) {
+		mlog(ML_ERROR, "Failed to load journal!\n");
+		goto done;
+	}
+
+	ocfs2_clear_journal_error(osb->sb, journal->j_journal, osb->slot_num);
+
+	status = ocfs2_journal_toggle_dirty(osb, 1);
+	if (status < 0) {
+		mlog_errno(status);
+		goto done;
+	}
+
+	/* Launch the commit thread */
+	osb->commit_task = kthread_run(ocfs2_commit_thread, osb, "ocfs2cmt-%d",
+				       osb->osb_id);
+	if (IS_ERR(osb->commit_task)) {
+		status = PTR_ERR(osb->commit_task);
+		osb->commit_task = NULL;
+		mlog(ML_ERROR, "unable to launch ocfs2commit thread, error=%d",
+		     status);
+		goto done;
+	}
+
+done:
+	mlog_exit(status);
+	return status;
+}
+
+
+/* 'full' flag tells us whether we clear out all blocks or if we just
+ * mark the journal clean */
+int ocfs2_journal_wipe(struct ocfs2_journal *journal, int full)
+{
+	int status;
+
+	mlog_entry_void();
+
+	if (!journal)
+		BUG();
+
+	status = journal_wipe(journal->j_journal, full);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_journal_toggle_dirty(journal->j_osb, 0);
+	if (status < 0)
+		mlog_errno(status);
+
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * JBD Might read a cached version of another nodes journal file. We
+ * don't want this as this file changes often and we get no
+ * notification on those changes. The only way to be sure that we've
+ * got the most up to date version of those blocks then is to force
+ * read them off disk. Just searching through the buffer cache won't
+ * work as there may be pages backing this file which are still marked
+ * up to date. We know things can't change on this file underneath us
+ * as we have the lock by now :)
+ */
+static int ocfs2_force_read_journal(struct inode *inode)
+{
+	int status = 0;
+	int i, p_blocks;
+	u64 v_blkno, p_blkno;
+#define CONCURRENT_JOURNAL_FILL 32
+	struct buffer_head *bhs[CONCURRENT_JOURNAL_FILL];
+
+	mlog_entry_void();
+
+	BUG_ON(inode->i_blocks !=
+		     ocfs2_align_bytes_to_sectors(i_size_read(inode)));
+
+	memset(bhs, 0, sizeof(struct buffer_head *) * CONCURRENT_JOURNAL_FILL);
+
+	mlog(0, "Force reading %lu blocks\n",
+	     (inode->i_blocks >> (inode->i_sb->s_blocksize_bits - 9)));
+
+	v_blkno = 0;
+	while (v_blkno <
+	       (inode->i_blocks >> (inode->i_sb->s_blocksize_bits - 9))) {
+
+		status = ocfs2_extent_map_get_blocks(inode, v_blkno,
+						     1, &p_blkno,
+						     &p_blocks);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		if (p_blocks > CONCURRENT_JOURNAL_FILL)
+			p_blocks = CONCURRENT_JOURNAL_FILL;
+
+		status = ocfs2_read_blocks(OCFS2_SB(inode->i_sb),
+					   p_blkno, p_blocks, bhs, 0,
+					   inode);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		for(i = 0; i < p_blocks; i++) {
+			brelse(bhs[i]);
+			bhs[i] = NULL;
+		}
+
+		v_blkno += p_blocks;
+	}
+
+bail:
+	for(i = 0; i < CONCURRENT_JOURNAL_FILL; i++)
+		if (bhs[i])
+			brelse(bhs[i]);
+	mlog_exit(status);
+	return status;
+}
+
+struct ocfs2_la_recovery_item {
+	struct list_head	lri_list;
+	int			lri_slot;
+	struct ocfs2_dinode	*lri_la_dinode;
+	struct ocfs2_dinode	*lri_tl_dinode;
+};
+
+/* Does the second half of the recovery process. By this point, the
+ * node is marked clean and can actually be considered recovered,
+ * hence it's no longer in the recovery map, but there's still some
+ * cleanup we can do which shouldn't happen within the recovery thread
+ * as locking in that context becomes very difficult if we are to take
+ * recovering nodes into account.
+ *
+ * NOTE: This function can and will sleep on recovery of other nodes
+ * during cluster locking, just like any other ocfs2 process.
+ */
+void ocfs2_complete_recovery(void *data)
+{
+	int ret;
+	struct ocfs2_super *osb = data;
+	struct ocfs2_journal *journal = osb->journal;
+	struct ocfs2_dinode *la_dinode, *tl_dinode;
+	struct ocfs2_la_recovery_item *item;
+	struct list_head *p, *n;
+	LIST_HEAD(tmp_la_list);
+
+	mlog_entry_void();
+
+	mlog(0, "completing recovery from keventd\n");
+
+	spin_lock(&journal->j_lock);
+	list_splice_init(&journal->j_la_cleanups, &tmp_la_list);
+	spin_unlock(&journal->j_lock);
+
+	list_for_each_safe(p, n, &tmp_la_list) {
+		item = list_entry(p, struct ocfs2_la_recovery_item, lri_list);
+		list_del_init(&item->lri_list);
+
+		mlog(0, "Complete recovery for slot %d\n", item->lri_slot);
+
+		la_dinode = item->lri_la_dinode;
+		if (la_dinode) {
+			mlog(0, "Clean up local alloc %"MLFu64"\n",
+			     la_dinode->i_blkno);
+
+			ret = ocfs2_complete_local_alloc_recovery(osb,
+								  la_dinode);
+			if (ret < 0)
+				mlog_errno(ret);
+
+			kfree(la_dinode);
+		}
+
+		tl_dinode = item->lri_tl_dinode;
+		if (tl_dinode) {
+			mlog(0, "Clean up truncate log %"MLFu64"\n",
+			     tl_dinode->i_blkno);
+
+			ret = ocfs2_complete_truncate_log_recovery(osb,
+								   tl_dinode);
+			if (ret < 0)
+				mlog_errno(ret);
+
+			kfree(tl_dinode);
+		}
+
+		ret = ocfs2_recover_orphans(osb, item->lri_slot);
+		if (ret < 0)
+			mlog_errno(ret);
+
+		kfree(item);
+	}
+
+	mlog(0, "Recovery completion\n");
+	mlog_exit_void();
+}
+
+/* NOTE: This function always eats your references to la_dinode and
+ * tl_dinode, either manually on error, or by passing them to
+ * ocfs2_complete_recovery */
+static void ocfs2_queue_recovery_completion(struct ocfs2_journal *journal,
+					    int slot_num,
+					    struct ocfs2_dinode *la_dinode,
+					    struct ocfs2_dinode *tl_dinode)
+{
+	struct ocfs2_la_recovery_item *item;
+
+	item = kmalloc(sizeof(struct ocfs2_la_recovery_item), GFP_KERNEL);
+	if (!item) {
+		/* Though we wish to avoid it, we are in fact safe in
+		 * skipping local alloc cleanup as fsck.ocfs2 is more
+		 * than capable of reclaiming unused space. */
+		if (la_dinode)
+			kfree(la_dinode);
+
+		if (tl_dinode)
+			kfree(tl_dinode);
+
+		mlog_errno(-ENOMEM);
+		return;
+	}
+
+	INIT_LIST_HEAD(&item->lri_list);
+	item->lri_la_dinode = la_dinode;
+	item->lri_slot = slot_num;
+	item->lri_tl_dinode = tl_dinode;
+
+	spin_lock(&journal->j_lock);
+	list_add_tail(&item->lri_list, &journal->j_la_cleanups);
+	queue_work(ocfs2_wq, &journal->j_recovery_work);
+	spin_unlock(&journal->j_lock);
+}
+
+/* Called by the mount code to queue recovery the last part of
+ * recovery for it's own slot. */
+void ocfs2_complete_mount_recovery(struct ocfs2_super *osb)
+{
+	struct ocfs2_journal *journal = osb->journal;
+
+	if (osb->dirty) {
+		/* No need to queue up our truncate_log as regular
+		 * cleanup will catch that. */
+		ocfs2_queue_recovery_completion(journal,
+						osb->slot_num,
+						osb->local_alloc_copy,
+						NULL);
+		ocfs2_schedule_truncate_log_flush(osb, 0);
+
+		osb->local_alloc_copy = NULL;
+		osb->dirty = 0;
+	}
+}
+
+static int __ocfs2_recovery_thread(void *arg)
+{
+	int status, node_num;
+	struct ocfs2_super *osb = arg;
+
+	mlog_entry_void();
+
+	status = ocfs2_wait_on_mount(osb);
+	if (status < 0) {
+		goto bail;
+	}
+
+restart:
+	status = ocfs2_super_lock(osb, 1);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	while(!ocfs2_node_map_is_empty(osb, &osb->recovery_map)) {
+		node_num = ocfs2_node_map_first_set_bit(osb,
+							&osb->recovery_map);
+		if (node_num == O2NM_INVALID_NODE_NUM) {
+			mlog(0, "Out of nodes to recover.\n");
+			break;
+		}
+
+		status = ocfs2_recover_node(osb, node_num);
+		if (status < 0) {
+			mlog(ML_ERROR,
+			     "Error %d recovering node %d on device (%u,%u)!\n",
+			     status, node_num,
+			     MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
+			mlog(ML_ERROR, "Volume requires unmount.\n");
+			continue;
+		}
+
+		ocfs2_recovery_map_clear(osb, node_num);
+	}
+	ocfs2_super_unlock(osb, 1);
+
+	/* We always run recovery on our own orphan dir - the dead
+	 * node(s) may have voted "no" on an inode delete earlier. A
+	 * revote is therefore required. */
+	ocfs2_queue_recovery_completion(osb->journal, osb->slot_num, NULL,
+					NULL);
+
+bail:
+	down(&osb->recovery_lock);
+	if (!status &&
+	    !ocfs2_node_map_is_empty(osb, &osb->recovery_map)) {
+		up(&osb->recovery_lock);
+		goto restart;
+	}
+
+	osb->recovery_thread_task = NULL;
+	mb(); /* sync with ocfs2_recovery_thread_running */
+	wake_up(&osb->recovery_event);
+
+	up(&osb->recovery_lock);
+
+	mlog_exit(status);
+	/* no one is callint kthread_stop() for us so the kthread() api
+	 * requires that we call do_exit().  And it isn't exported, but
+	 * complete_and_exit() seems to be a minimal wrapper around it. */
+	complete_and_exit(NULL, status);
+	return status;
+}
+
+void ocfs2_recovery_thread(struct ocfs2_super *osb, int node_num)
+{
+	mlog_entry("(node_num=%d, osb->node_num = %d)\n",
+		   node_num, osb->node_num);
+
+	down(&osb->recovery_lock);
+	if (osb->disable_recovery)
+		goto out;
+
+	/* People waiting on recovery will wait on
+	 * the recovery map to empty. */
+	if (!ocfs2_recovery_map_set(osb, node_num))
+		mlog(0, "node %d already be in recovery.\n", node_num);
+
+	mlog(0, "starting recovery thread...\n");
+
+	if (osb->recovery_thread_task)
+		goto out;
+
+	osb->recovery_thread_task =  kthread_run(__ocfs2_recovery_thread, osb,
+						 "ocfs2rec-%d", osb->osb_id);
+	if (IS_ERR(osb->recovery_thread_task)) {
+		mlog_errno((int)PTR_ERR(osb->recovery_thread_task));
+		osb->recovery_thread_task = NULL;
+	}
+
+out:
+	up(&osb->recovery_lock);
+	wake_up(&osb->recovery_event);
+
+	mlog_exit_void();
+}
+
+/* Does the actual journal replay and marks the journal inode as
+ * clean. Will only replay if the journal inode is marked dirty. */
+static int ocfs2_replay_journal(struct ocfs2_super *osb,
+				int node_num,
+				int slot_num)
+{
+	int status;
+	int got_lock = 0;
+	unsigned int flags;
+	struct inode *inode = NULL;
+	struct ocfs2_dinode *fe;
+	journal_t *journal = NULL;
+	struct buffer_head *bh = NULL;
+
+	inode = ocfs2_get_system_file_inode(osb, JOURNAL_SYSTEM_INODE,
+					    slot_num);
+	if (inode == NULL) {
+		status = -EACCES;
+		mlog_errno(status);
+		goto done;
+	}
+	if (is_bad_inode(inode)) {
+		status = -EACCES;
+		iput(inode);
+		inode = NULL;
+		mlog_errno(status);
+		goto done;
+	}
+	SET_INODE_JOURNAL(inode);
+
+	status = ocfs2_meta_lock_full(inode, NULL, &bh, 1,
+				      OCFS2_META_LOCK_RECOVERY);
+	if (status < 0) {
+		mlog(0, "status returned from ocfs2_meta_lock=%d\n", status);
+		if (status != -ERESTARTSYS)
+			mlog(ML_ERROR, "Could not lock journal!\n");
+		goto done;
+	}
+	got_lock = 1;
+
+	fe = (struct ocfs2_dinode *) bh->b_data;
+
+	flags = le32_to_cpu(fe->id1.journal1.ij_flags);
+
+	if (!(flags & OCFS2_JOURNAL_DIRTY_FL)) {
+		mlog(0, "No recovery required for node %d\n", node_num);
+		goto done;
+	}
+
+	mlog(ML_NOTICE, "Recovering node %d from slot %d on device (%u,%u)\n",
+	     node_num, slot_num,
+	     MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
+
+	OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
+
+	status = ocfs2_force_read_journal(inode);
+	if (status < 0) {
+		mlog_errno(status);
+		goto done;
+	}
+
+	mlog(0, "calling journal_init_inode\n");
+	journal = journal_init_inode(inode);
+	if (journal == NULL) {
+		mlog(ML_ERROR, "Linux journal layer error\n");
+		status = -EIO;
+		goto done;
+	}
+
+	status = journal_load(journal);
+	if (status < 0) {
+		mlog_errno(status);
+		if (!igrab(inode))
+			BUG();
+		journal_destroy(journal);
+		goto done;
+	}
+
+	ocfs2_clear_journal_error(osb->sb, journal, slot_num);
+
+	/* wipe the journal */
+	mlog(0, "flushing the journal.\n");
+	journal_lock_updates(journal);
+	status = journal_flush(journal);
+	journal_unlock_updates(journal);
+	if (status < 0)
+		mlog_errno(status);
+
+	/* This will mark the node clean */
+	flags = le32_to_cpu(fe->id1.journal1.ij_flags);
+	flags &= ~OCFS2_JOURNAL_DIRTY_FL;
+	fe->id1.journal1.ij_flags = cpu_to_le32(flags);
+
+	status = ocfs2_write_block(osb, bh, inode);
+	if (status < 0)
+		mlog_errno(status);
+
+	if (!igrab(inode))
+		BUG();
+
+	journal_destroy(journal);
+
+done:
+	/* drop the lock on this nodes journal */
+	if (got_lock)
+		ocfs2_meta_unlock(inode, 1);
+
+	if (inode)
+		iput(inode);
+
+	if (bh)
+		brelse(bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * Do the most important parts of node recovery:
+ *  - Replay it's journal
+ *  - Stamp a clean local allocator file
+ *  - Stamp a clean truncate log
+ *  - Mark the node clean
+ *
+ * If this function completes without error, a node in OCFS2 can be
+ * said to have been safely recovered. As a result, failure during the
+ * second part of a nodes recovery process (local alloc recovery) is
+ * far less concerning.
+ */
+static int ocfs2_recover_node(struct ocfs2_super *osb,
+			      int node_num)
+{
+	int status = 0;
+	int slot_num;
+	struct ocfs2_slot_info *si = osb->slot_info;
+	struct ocfs2_dinode *la_copy = NULL;
+	struct ocfs2_dinode *tl_copy = NULL;
+
+	mlog_entry("(node_num=%d, osb->node_num = %d)\n",
+		   node_num, osb->node_num);
+
+	mlog(0, "checking node %d\n", node_num);
+
+	/* Should not ever be called to recover ourselves -- in that
+	 * case we should've called ocfs2_journal_load instead. */
+	if (osb->node_num == node_num)
+		BUG();
+
+	slot_num = ocfs2_node_num_to_slot(si, node_num);
+	if (slot_num == OCFS2_INVALID_SLOT) {
+		status = 0;
+		mlog(0, "no slot for this node, so no recovery required.\n");
+		goto done;
+	}
+
+	mlog(0, "node %d was using slot %d\n", node_num, slot_num);
+
+	status = ocfs2_replay_journal(osb, node_num, slot_num);
+	if (status < 0) {
+		mlog_errno(status);
+		goto done;
+	}
+
+	/* Stamp a clean local alloc file AFTER recovering the journal... */
+	status = ocfs2_begin_local_alloc_recovery(osb, slot_num, &la_copy);
+	if (status < 0) {
+		mlog_errno(status);
+		goto done;
+	}
+
+	/* An error from begin_truncate_log_recovery is not
+	 * serious enough to warrant halting the rest of
+	 * recovery. */
+	status = ocfs2_begin_truncate_log_recovery(osb, slot_num, &tl_copy);
+	if (status < 0)
+		mlog_errno(status);
+
+	/* Likewise, this would be a strange but ultimately not so
+	 * harmful place to get an error... */
+	ocfs2_clear_slot(si, slot_num);
+	status = ocfs2_update_disk_slots(osb, si);
+	if (status < 0)
+		mlog_errno(status);
+
+	/* This will kfree the memory pointed to by la_copy and tl_copy */
+	ocfs2_queue_recovery_completion(osb->journal, slot_num, la_copy,
+					tl_copy);
+
+	status = 0;
+done:
+
+	mlog_exit(status);
+	return status;
+}
+
+/* Test node liveness by trylocking his journal. If we get the lock,
+ * we drop it here. Return 0 if we got the lock, -EAGAIN if node is
+ * still alive (we couldn't get the lock) and < 0 on error. */
+static int ocfs2_trylock_journal(struct ocfs2_super *osb,
+				 int slot_num)
+{
+	int status, flags;
+	struct inode *inode = NULL;
+
+	inode = ocfs2_get_system_file_inode(osb, JOURNAL_SYSTEM_INODE,
+					    slot_num);
+	if (inode == NULL) {
+		mlog(ML_ERROR, "access error\n");
+		status = -EACCES;
+		goto bail;
+	}
+	if (is_bad_inode(inode)) {
+		mlog(ML_ERROR, "access error (bad inode)\n");
+		iput(inode);
+		inode = NULL;
+		status = -EACCES;
+		goto bail;
+	}
+	SET_INODE_JOURNAL(inode);
+
+	flags = OCFS2_META_LOCK_RECOVERY | OCFS2_META_LOCK_NOQUEUE;
+	status = ocfs2_meta_lock_full(inode, NULL, NULL, 1, flags);
+	if (status < 0) {
+		if (status != -EAGAIN)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	ocfs2_meta_unlock(inode, 1);
+bail:
+	if (inode)
+		iput(inode);
+
+	return status;
+}
+
+/* Call this underneath ocfs2_super_lock. It also assumes that the
+ * slot info struct has been updated from disk. */
+int ocfs2_mark_dead_nodes(struct ocfs2_super *osb)
+{
+	int status, i, node_num;
+	struct ocfs2_slot_info *si = osb->slot_info;
+
+	/* This is called with the super block cluster lock, so we
+	 * know that the slot map can't change underneath us. */
+
+	spin_lock(&si->si_lock);
+	for(i = 0; i < si->si_num_slots; i++) {
+		if (i == osb->slot_num)
+			continue;
+		if (ocfs2_is_empty_slot(si, i))
+			continue;
+
+		node_num = si->si_global_node_nums[i];
+		if (ocfs2_node_map_test_bit(osb, &osb->recovery_map, node_num))
+			continue;
+		spin_unlock(&si->si_lock);
+
+		/* Ok, we have a slot occupied by another node which
+		 * is not in the recovery map. We trylock his journal
+		 * file here to test if he's alive. */
+		status = ocfs2_trylock_journal(osb, i);
+		if (!status) {
+			/* Since we're called from mount, we know that
+			 * the recovery thread can't race us on
+			 * setting / checking the recovery bits. */
+			ocfs2_recovery_thread(osb, node_num);
+		} else if ((status < 0) && (status != -EAGAIN)) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		spin_lock(&si->si_lock);
+	}
+	spin_unlock(&si->si_lock);
+
+	status = 0;
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_recover_orphans(struct ocfs2_super *osb,
+				 int slot)
+{
+	int status = 0;
+	int have_disk_lock = 0;
+	struct inode *inode = NULL;
+	struct inode *iter;
+	struct inode *orphan_dir_inode = NULL;
+	unsigned long offset, blk, local;
+	struct buffer_head *bh = NULL;
+	struct ocfs2_dir_entry *de;
+	struct super_block *sb = osb->sb;
+	struct ocfs2_inode_info *oi;
+
+	mlog(0, "Recover inodes from orphan dir in slot %d\n", slot);
+
+	orphan_dir_inode = ocfs2_get_system_file_inode(osb,
+						       ORPHAN_DIR_SYSTEM_INODE,
+						       slot);
+	if  (!orphan_dir_inode) {
+		status = -ENOENT;
+		mlog_errno(status);
+		goto out;
+	}
+
+	down(&orphan_dir_inode->i_sem);
+	status = ocfs2_meta_lock(orphan_dir_inode, NULL, NULL, 0);
+	if (status < 0) {
+		up(&orphan_dir_inode->i_sem);
+		mlog_errno(status);
+		goto out;
+	}
+	have_disk_lock = 1;
+
+	offset = 0;
+	iter = NULL;
+	while(offset < i_size_read(orphan_dir_inode)) {
+		blk = offset >> sb->s_blocksize_bits;
+
+		bh = ocfs2_bread(orphan_dir_inode, blk, &status, 0);
+		if (!bh)
+			status = -EINVAL;
+		if (status < 0) {
+			up(&orphan_dir_inode->i_sem);
+			if (bh)
+				brelse(bh);
+			mlog_errno(status);
+			goto out;
+		}
+
+		local = 0;
+		while(offset < i_size_read(orphan_dir_inode)
+		      && local < sb->s_blocksize) {
+			de = (struct ocfs2_dir_entry *) (bh->b_data + local);
+
+			if (!ocfs2_check_dir_entry(orphan_dir_inode,
+						  de, bh, local)) {
+				up(&orphan_dir_inode->i_sem);
+				status = -EINVAL;
+				mlog_errno(status);
+				brelse(bh);
+				goto out;
+			}
+
+			local += le16_to_cpu(de->rec_len);
+			offset += le16_to_cpu(de->rec_len);
+
+			/* I guess we silently fail on no inode? */
+			if (!le64_to_cpu(de->inode))
+				continue;
+			if (de->file_type > OCFS2_FT_MAX) {
+				mlog(ML_ERROR,
+				     "block %llu contains invalid de: "
+				     "inode = %"MLFu64", rec_len = %u, "
+				     "name_len = %u, file_type = %u, "
+				     "name='%.*s'\n",
+				     (unsigned long long)bh->b_blocknr,
+				     le64_to_cpu(de->inode),
+				     le16_to_cpu(de->rec_len),
+				     de->name_len,
+				     de->file_type,
+				     de->name_len,
+				     de->name);
+				continue;
+			}
+			if (de->name_len == 1 && !strncmp(".", de->name, 1))
+				continue;
+			if (de->name_len == 2 && !strncmp("..", de->name, 2))
+				continue;
+
+			iter = ocfs2_iget(osb, le64_to_cpu(de->inode));
+			if (IS_ERR(iter))
+				continue;
+
+			mlog(0, "queue orphan %"MLFu64"\n",
+			     OCFS2_I(iter)->ip_blkno);
+			OCFS2_I(iter)->ip_next_orphan = inode;
+			inode = iter;
+		}
+		brelse(bh);
+	}
+	up(&orphan_dir_inode->i_sem);
+
+	ocfs2_meta_unlock(orphan_dir_inode, 0);
+	have_disk_lock = 0;
+
+	iput(orphan_dir_inode);
+	orphan_dir_inode = NULL;
+
+	while (inode) {
+		oi = OCFS2_I(inode);
+		mlog(0, "iput orphan %"MLFu64"\n", oi->ip_blkno);
+
+		iter = oi->ip_next_orphan;
+
+		spin_lock(&oi->ip_lock);
+		/* Delete voting may have set these on the assumption
+		 * that the other node would wipe them successfully.
+		 * If they are still in the node's orphan dir, we need
+		 * to reset that state. */
+		oi->ip_flags &= ~(OCFS2_INODE_DELETED|OCFS2_INODE_SKIP_DELETE);
+
+		/* Set the proper information to get us going into
+		 * ocfs2_delete_inode. */
+		oi->ip_flags |= OCFS2_INODE_MAYBE_ORPHANED;
+		oi->ip_orphaned_slot = slot;
+		spin_unlock(&oi->ip_lock);
+
+		iput(inode);
+
+		inode = iter;
+	}
+
+out:
+	if (have_disk_lock)
+		ocfs2_meta_unlock(orphan_dir_inode, 0);
+
+	if (orphan_dir_inode)
+		iput(orphan_dir_inode);
+
+	return status;
+}
+
+static int ocfs2_wait_on_mount(struct ocfs2_super *osb)
+{
+	/* This check is good because ocfs2 will wait on our recovery
+	 * thread before changing it to something other than MOUNTED
+	 * or DISABLED. */
+	wait_event(osb->osb_mount_event,
+		   atomic_read(&osb->vol_state) == VOLUME_MOUNTED ||
+		   atomic_read(&osb->vol_state) == VOLUME_DISABLED);
+
+	/* If there's an error on mount, then we may never get to the
+	 * MOUNTED flag, but this is set right before
+	 * dismount_volume() so we can trust it. */
+	if (atomic_read(&osb->vol_state) == VOLUME_DISABLED) {
+		mlog(0, "mount error, exiting!\n");
+		return -EBUSY;
+	}
+
+	return 0;
+}
+
+static int ocfs2_commit_thread(void *arg)
+{
+	int status;
+	struct ocfs2_super *osb = arg;
+	struct ocfs2_journal *journal = osb->journal;
+
+	/* we can trust j_num_trans here because _should_stop() is only set in
+	 * shutdown and nobody other than ourselves should be able to start
+	 * transactions.  committing on shutdown might take a few iterations
+	 * as final transactions put deleted inodes on the list */
+	while (!(kthread_should_stop() &&
+		 atomic_read(&journal->j_num_trans) == 0)) {
+
+		wait_event_interruptible_timeout(osb->checkpoint_event,
+						 atomic_read(&journal->j_num_trans)
+						 || kthread_should_stop(),
+						 OCFS2_CHECKPOINT_INTERVAL);
+
+		status = ocfs2_commit_cache(osb);
+		if (status < 0)
+			mlog_errno(status);
+
+		if (kthread_should_stop() && atomic_read(&journal->j_num_trans)){
+			mlog(ML_KTHREAD,
+			     "commit_thread: %u transactions pending on "
+			     "shutdown\n",
+			     atomic_read(&journal->j_num_trans));
+		}
+	}
+
+	return 0;
+}
+
+/* Look for a dirty journal without taking any cluster locks. Used for
+ * hard readonly access to determine whether the file system journals
+ * require recovery. */
+int ocfs2_check_journals_nolocks(struct ocfs2_super *osb)
+{
+	int ret = 0;
+	unsigned int slot;
+	struct buffer_head *di_bh;
+	struct ocfs2_dinode *di;
+	struct inode *journal = NULL;
+
+	for(slot = 0; slot < osb->max_slots; slot++) {
+		journal = ocfs2_get_system_file_inode(osb,
+						      JOURNAL_SYSTEM_INODE,
+						      slot);
+		if (!journal || is_bad_inode(journal)) {
+			ret = -EACCES;
+			mlog_errno(ret);
+			goto out;
+		}
+
+		di_bh = NULL;
+		ret = ocfs2_read_block(osb, OCFS2_I(journal)->ip_blkno, &di_bh,
+				       0, journal);
+		if (ret < 0) {
+			mlog_errno(ret);
+			goto out;
+		}
+
+		di = (struct ocfs2_dinode *) di_bh->b_data;
+
+		if (le32_to_cpu(di->id1.journal1.ij_flags) &
+		    OCFS2_JOURNAL_DIRTY_FL)
+			ret = -EROFS;
+
+		brelse(di_bh);
+		if (ret)
+			break;
+	}
+
+out:
+	if (journal)
+		iput(journal);
+
+	return ret;
+}
diff --git a/fs/ocfs2/journal.h b/fs/ocfs2/journal.h
new file mode 100644
index 00000000000..7d0a816184f
--- /dev/null
+++ b/fs/ocfs2/journal.h
@@ -0,0 +1,457 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * journal.h
+ *
+ * Defines journalling api and structures.
+ *
+ * Copyright (C) 2003, 2005 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_JOURNAL_H
+#define OCFS2_JOURNAL_H
+
+#include <linux/fs.h>
+#include <linux/jbd.h>
+
+#define OCFS2_CHECKPOINT_INTERVAL        (8 * HZ)
+
+enum ocfs2_journal_state {
+	OCFS2_JOURNAL_FREE = 0,
+	OCFS2_JOURNAL_LOADED,
+	OCFS2_JOURNAL_IN_SHUTDOWN,
+};
+
+struct ocfs2_super;
+struct ocfs2_dinode;
+struct ocfs2_journal_handle;
+
+struct ocfs2_journal {
+	enum ocfs2_journal_state   j_state;    /* Journals current state   */
+
+	journal_t                 *j_journal; /* The kernels journal type */
+	struct inode              *j_inode;   /* Kernel inode pointing to
+					       * this journal             */
+	struct ocfs2_super        *j_osb;     /* pointer to the super
+					       * block for the node
+					       * we're currently
+					       * running on -- not
+					       * necessarily the super
+					       * block from the node
+					       * which we usually run
+					       * from (recovery,
+					       * etc)                     */
+	struct buffer_head        *j_bh;      /* Journal disk inode block */
+	atomic_t                  j_num_trans; /* Number of transactions
+					        * currently in the system. */
+	unsigned long             j_trans_id;
+	struct rw_semaphore       j_trans_barrier;
+	wait_queue_head_t         j_checkpointed;
+
+	spinlock_t                j_lock;
+	struct list_head          j_la_cleanups;
+	struct work_struct        j_recovery_work;
+};
+
+extern spinlock_t trans_inc_lock;
+
+/* wrap j_trans_id so we never have it equal to zero. */
+static inline unsigned long ocfs2_inc_trans_id(struct ocfs2_journal *j)
+{
+	unsigned long old_id;
+	spin_lock(&trans_inc_lock);
+	old_id = j->j_trans_id++;
+	if (unlikely(!j->j_trans_id))
+		j->j_trans_id = 1;
+	spin_unlock(&trans_inc_lock);
+	return old_id;
+}
+
+static inline void ocfs2_set_inode_lock_trans(struct ocfs2_journal *journal,
+					      struct inode *inode)
+{
+	spin_lock(&trans_inc_lock);
+	OCFS2_I(inode)->ip_last_trans = journal->j_trans_id;
+	spin_unlock(&trans_inc_lock);
+}
+
+/* Used to figure out whether it's safe to drop a metadata lock on an
+ * inode. Returns true if all the inodes changes have been
+ * checkpointed to disk. You should be holding the spinlock on the
+ * metadata lock while calling this to be sure that nobody can take
+ * the lock and put it on another transaction. */
+static inline int ocfs2_inode_fully_checkpointed(struct inode *inode)
+{
+	int ret;
+	struct ocfs2_journal *journal = OCFS2_SB(inode->i_sb)->journal;
+
+	spin_lock(&trans_inc_lock);
+	ret = time_after(journal->j_trans_id, OCFS2_I(inode)->ip_last_trans);
+	spin_unlock(&trans_inc_lock);
+	return ret;
+}
+
+/* convenience function to check if an inode is still new (has never
+ * hit disk) Will do you a favor and set created_trans = 0 when you've
+ * been checkpointed.  returns '1' if the inode is still new. */
+static inline int ocfs2_inode_is_new(struct inode *inode)
+{
+	int ret;
+
+	/* System files are never "new" as they're written out by
+	 * mkfs. This helps us early during mount, before we have the
+	 * journal open and j_trans_id could be junk. */
+	if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SYSTEM_FILE)
+		return 0;
+	spin_lock(&trans_inc_lock);
+	ret = !(time_after(OCFS2_SB(inode->i_sb)->journal->j_trans_id,
+			   OCFS2_I(inode)->ip_created_trans));
+	if (!ret)
+		OCFS2_I(inode)->ip_created_trans = 0;
+	spin_unlock(&trans_inc_lock);
+	return ret;
+}
+
+static inline void ocfs2_inode_set_new(struct ocfs2_super *osb,
+				       struct inode *inode)
+{
+	spin_lock(&trans_inc_lock);
+	OCFS2_I(inode)->ip_created_trans = osb->journal->j_trans_id;
+	spin_unlock(&trans_inc_lock);
+}
+
+extern kmem_cache_t *ocfs2_lock_cache;
+
+struct ocfs2_journal_lock {
+	struct inode     *jl_inode;
+	struct list_head  jl_lock_list;
+};
+
+struct ocfs2_journal_handle {
+	handle_t            *k_handle; /* kernel handle.                */
+	struct ocfs2_journal        *journal;
+	u32                 flags;     /* see flags below.              */
+	int                 max_buffs; /* Buffs reserved by this handle */
+
+	/* The following two fields are for ocfs2_handle_add_lock */
+	int                 num_locks;
+	struct list_head    locks;     /* A bunch of locks to
+					* release on commit. This
+					* should be a list_head */
+
+	struct list_head     inode_list;
+};
+
+#define OCFS2_HANDLE_STARTED			1
+/* should we sync-commit this handle? */
+#define OCFS2_HANDLE_SYNC			2
+static inline int ocfs2_handle_started(struct ocfs2_journal_handle *handle)
+{
+	return handle->flags & OCFS2_HANDLE_STARTED;
+}
+
+static inline void ocfs2_handle_set_sync(struct ocfs2_journal_handle *handle, int sync)
+{
+	if (sync)
+		handle->flags |= OCFS2_HANDLE_SYNC;
+	else
+		handle->flags &= ~OCFS2_HANDLE_SYNC;
+}
+
+/* Exported only for the journal struct init code in super.c. Do not call. */
+void ocfs2_complete_recovery(void *data);
+
+/*
+ *  Journal Control:
+ *  Initialize, Load, Shutdown, Wipe a journal.
+ *
+ *  ocfs2_journal_init     - Initialize journal structures in the OSB.
+ *  ocfs2_journal_load     - Load the given journal off disk. Replay it if
+ *                          there's transactions still in there.
+ *  ocfs2_journal_shutdown - Shutdown a journal, this will flush all
+ *                          uncommitted, uncheckpointed transactions.
+ *  ocfs2_journal_wipe     - Wipe transactions from a journal. Optionally
+ *                          zero out each block.
+ *  ocfs2_recovery_thread  - Perform recovery on a node. osb is our own osb.
+ *  ocfs2_mark_dead_nodes - Start recovery on nodes we won't get a heartbeat
+ *                          event on.
+ *  ocfs2_start_checkpoint - Kick the commit thread to do a checkpoint.
+ */
+void   ocfs2_set_journal_params(struct ocfs2_super *osb);
+int    ocfs2_journal_init(struct ocfs2_journal *journal,
+			  int *dirty);
+void   ocfs2_journal_shutdown(struct ocfs2_super *osb);
+int    ocfs2_journal_wipe(struct ocfs2_journal *journal,
+			  int full);
+int    ocfs2_journal_load(struct ocfs2_journal *journal);
+int    ocfs2_check_journals_nolocks(struct ocfs2_super *osb);
+void   ocfs2_recovery_thread(struct ocfs2_super *osb,
+			     int node_num);
+int    ocfs2_mark_dead_nodes(struct ocfs2_super *osb);
+void   ocfs2_complete_mount_recovery(struct ocfs2_super *osb);
+
+static inline void ocfs2_start_checkpoint(struct ocfs2_super *osb)
+{
+	atomic_set(&osb->needs_checkpoint, 1);
+	wake_up(&osb->checkpoint_event);
+}
+
+static inline void ocfs2_checkpoint_inode(struct inode *inode)
+{
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	if (!ocfs2_inode_fully_checkpointed(inode)) {
+		/* WARNING: This only kicks off a single
+		 * checkpoint. If someone races you and adds more
+		 * metadata to the journal, you won't know, and will
+		 * wind up waiting *alot* longer than necessary. Right
+		 * now we only use this in clear_inode so that's
+		 * OK. */
+		ocfs2_start_checkpoint(osb);
+
+		wait_event(osb->journal->j_checkpointed,
+			   ocfs2_inode_fully_checkpointed(inode));
+	}
+}
+
+/*
+ *  Transaction Handling:
+ *  Manage the lifetime of a transaction handle.
+ *
+ *  ocfs2_alloc_handle     - Only allocate a handle so we can start putting
+ *                          cluster locks on it. To actually change blocks,
+ *                          call ocfs2_start_trans with the handle returned
+ *                          from this function. You may call ocfs2_commit_trans
+ *                           at any time in the lifetime of a handle.
+ *  ocfs2_start_trans      - Begin a transaction. Give it an upper estimate of
+ *                          the number of blocks that will be changed during
+ *                          this handle.
+ *  ocfs2_commit_trans     - Complete a handle.
+ *  ocfs2_extend_trans     - Extend a handle by nblocks credits. This may
+ *                          commit the handle to disk in the process, but will
+ *                          not release any locks taken during the transaction.
+ *  ocfs2_journal_access   - Notify the handle that we want to journal this
+ *                          buffer. Will have to call ocfs2_journal_dirty once
+ *                          we've actually dirtied it. Type is one of . or .
+ *  ocfs2_journal_dirty    - Mark a journalled buffer as having dirty data.
+ *  ocfs2_journal_dirty_data - Indicate that a data buffer should go out before
+ *                             the current handle commits.
+ *  ocfs2_handle_add_lock  - Sometimes we need to delay lock release
+ *                          until after a transaction has been completed. Use
+ *                          ocfs2_handle_add_lock to indicate that a lock needs
+ *                          to be released at the end of that handle. Locks
+ *                          will be released in the order that they are added.
+ *  ocfs2_handle_add_inode - Add a locked inode to a transaction.
+ */
+
+/* You must always start_trans with a number of buffs > 0, but it's
+ * perfectly legal to go through an entire transaction without having
+ * dirtied any buffers. */
+struct ocfs2_journal_handle *ocfs2_alloc_handle(struct ocfs2_super *osb);
+struct ocfs2_journal_handle *ocfs2_start_trans(struct ocfs2_super *osb,
+					       struct ocfs2_journal_handle *handle,
+					       int max_buffs);
+void			     ocfs2_commit_trans(struct ocfs2_journal_handle *handle);
+int			     ocfs2_extend_trans(struct ocfs2_journal_handle *handle,
+						int nblocks);
+
+/*
+ * Create access is for when we get a newly created buffer and we're
+ * not gonna read it off disk, but rather fill it ourselves.  Right
+ * now, we don't do anything special with this (it turns into a write
+ * request), but this is a good placeholder in case we do...
+ *
+ * Write access is for when we read a block off disk and are going to
+ * modify it. This way the journalling layer knows it may need to make
+ * a copy of that block (if it's part of another, uncommitted
+ * transaction) before we do so.
+ */
+#define OCFS2_JOURNAL_ACCESS_CREATE 0
+#define OCFS2_JOURNAL_ACCESS_WRITE  1
+#define OCFS2_JOURNAL_ACCESS_UNDO   2
+
+int                  ocfs2_journal_access(struct ocfs2_journal_handle *handle,
+					  struct inode *inode,
+					  struct buffer_head *bh,
+					  int type);
+/*
+ * A word about the journal_access/journal_dirty "dance". It is
+ * entirely legal to journal_access a buffer more than once (as long
+ * as the access type is the same -- I'm not sure what will happen if
+ * access type is different but this should never happen anyway) It is
+ * also legal to journal_dirty a buffer more than once. In fact, you
+ * can even journal_access a buffer after you've done a
+ * journal_access/journal_dirty pair. The only thing you cannot do
+ * however, is journal_dirty a buffer which you haven't yet passed to
+ * journal_access at least once.
+ *
+ * That said, 99% of the time this doesn't matter and this is what the
+ * path looks like:
+ *
+ *	<read a bh>
+ *	ocfs2_journal_access(handle, bh,	OCFS2_JOURNAL_ACCESS_WRITE);
+ *	<modify the bh>
+ * 	ocfs2_journal_dirty(handle, bh);
+ */
+int                  ocfs2_journal_dirty(struct ocfs2_journal_handle *handle,
+					 struct buffer_head *bh);
+int                  ocfs2_journal_dirty_data(handle_t *handle,
+					      struct buffer_head *bh);
+int                  ocfs2_handle_add_lock(struct ocfs2_journal_handle *handle,
+					   struct inode *inode);
+/*
+ * Use this to protect from other processes reading buffer state while
+ * it's in flight.
+ */
+void                 ocfs2_handle_add_inode(struct ocfs2_journal_handle *handle,
+					    struct inode *inode);
+
+/*
+ *  Credit Macros:
+ *  Convenience macros to calculate number of credits needed.
+ *
+ *  For convenience sake, I have a set of macros here which calculate
+ *  the *maximum* number of sectors which will be changed for various
+ *  metadata updates.
+ */
+
+/* simple file updates like chmod, etc. */
+#define OCFS2_INODE_UPDATE_CREDITS 1
+
+/* get one bit out of a suballocator: dinode + group descriptor +
+ * prev. group desc. if we relink. */
+#define OCFS2_SUBALLOC_ALLOC (3)
+
+/* dinode + group descriptor update. We don't relink on free yet. */
+#define OCFS2_SUBALLOC_FREE  (2)
+
+#define OCFS2_TRUNCATE_LOG_UPDATE OCFS2_INODE_UPDATE_CREDITS
+#define OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC (OCFS2_SUBALLOC_FREE 		      \
+					 + OCFS2_TRUNCATE_LOG_UPDATE)
+
+/* data block for new dir/symlink, 2 for bitmap updates (bitmap fe +
+ * bitmap block for the new bit) */
+#define OCFS2_DIR_LINK_ADDITIONAL_CREDITS (1 + 2)
+
+/* parent fe, parent block, new file entry, inode alloc fe, inode alloc
+ * group descriptor + mkdir/symlink blocks */
+#define OCFS2_MKNOD_CREDITS (3 + OCFS2_SUBALLOC_ALLOC                         \
+			    + OCFS2_DIR_LINK_ADDITIONAL_CREDITS)
+
+/* local alloc metadata change + main bitmap updates */
+#define OCFS2_WINDOW_MOVE_CREDITS (OCFS2_INODE_UPDATE_CREDITS                 \
+				  + OCFS2_SUBALLOC_ALLOC + OCFS2_SUBALLOC_FREE)
+
+/* used when we don't need an allocation change for a dir extend. One
+ * for the dinode, one for the new block. */
+#define OCFS2_SIMPLE_DIR_EXTEND_CREDITS (2)
+
+/* file update (nlink, etc) + dir entry block */
+#define OCFS2_LINK_CREDITS  (OCFS2_INODE_UPDATE_CREDITS + 1)
+
+/* inode + dir inode (if we unlink a dir), + dir entry block + orphan
+ * dir inode link */
+#define OCFS2_UNLINK_CREDITS  (2 * OCFS2_INODE_UPDATE_CREDITS + 1             \
+			      + OCFS2_LINK_CREDITS)
+
+/* dinode + orphan dir dinode + inode alloc dinode + orphan dir entry +
+ * inode alloc group descriptor */
+#define OCFS2_DELETE_INODE_CREDITS (3 * OCFS2_INODE_UPDATE_CREDITS + 1 + 1)
+
+/* dinode update, old dir dinode update, new dir dinode update, old
+ * dir dir entry, new dir dir entry, dir entry update for renaming
+ * directory + target unlink */
+#define OCFS2_RENAME_CREDITS (3 * OCFS2_INODE_UPDATE_CREDITS + 3              \
+			     + OCFS2_UNLINK_CREDITS)
+
+static inline int ocfs2_calc_extend_credits(struct super_block *sb,
+					    struct ocfs2_dinode *fe,
+					    u32 bits_wanted)
+{
+	int bitmap_blocks, sysfile_bitmap_blocks, dinode_blocks;
+
+	/* bitmap dinode, group desc. + relinked group. */
+	bitmap_blocks = OCFS2_SUBALLOC_ALLOC;
+
+	/* we might need to shift tree depth so lets assume an
+	 * absolute worst case of complete fragmentation.  Even with
+	 * that, we only need one update for the dinode, and then
+	 * however many metadata chunks needed * a remaining suballoc
+	 * alloc. */
+	sysfile_bitmap_blocks = 1 +
+		(OCFS2_SUBALLOC_ALLOC - 1) * ocfs2_extend_meta_needed(fe);
+
+	/* this does not include *new* metadata blocks, which are
+	 * accounted for in sysfile_bitmap_blocks. fe +
+	 * prev. last_eb_blk + blocks along edge of tree.
+	 * calc_symlink_credits passes because we just need 1
+	 * credit for the dinode there. */
+	dinode_blocks = 1 + 1 + le16_to_cpu(fe->id2.i_list.l_tree_depth);
+
+	return bitmap_blocks + sysfile_bitmap_blocks + dinode_blocks;
+}
+
+static inline int ocfs2_calc_symlink_credits(struct super_block *sb)
+{
+	int blocks = OCFS2_MKNOD_CREDITS;
+
+	/* links can be longer than one block so we may update many
+	 * within our single allocated extent. */
+	blocks += ocfs2_clusters_to_blocks(sb, 1);
+
+	return blocks;
+}
+
+static inline int ocfs2_calc_group_alloc_credits(struct super_block *sb,
+						 unsigned int cpg)
+{
+	int blocks;
+	int bitmap_blocks = OCFS2_SUBALLOC_ALLOC + 1;
+	/* parent inode update + new block group header + bitmap inode update
+	   + bitmap blocks affected */
+	blocks = 1 + 1 + 1 + bitmap_blocks;
+	return blocks;
+}
+
+static inline int ocfs2_calc_tree_trunc_credits(struct super_block *sb,
+						unsigned int clusters_to_del,
+						struct ocfs2_dinode *fe,
+						struct ocfs2_extent_list *last_el)
+{
+ 	/* for dinode + all headers in this pass + update to next leaf */
+	u16 next_free = le16_to_cpu(last_el->l_next_free_rec);
+	u16 tree_depth = le16_to_cpu(fe->id2.i_list.l_tree_depth);
+	int credits = 1 + tree_depth + 1;
+	int i;
+
+	i = next_free - 1;
+	BUG_ON(i < 0);
+
+	/* We may be deleting metadata blocks, so metadata alloc dinode +
+	   one desc. block for each possible delete. */
+	if (tree_depth && next_free == 1 &&
+	    le32_to_cpu(last_el->l_recs[i].e_clusters) == clusters_to_del)
+		credits += 1 + tree_depth;
+
+	/* update to the truncate log. */
+	credits += OCFS2_TRUNCATE_LOG_UPDATE;
+
+	return credits;
+}
+
+#endif /* OCFS2_JOURNAL_H */
diff --git a/fs/ocfs2/localalloc.c b/fs/ocfs2/localalloc.c
new file mode 100644
index 00000000000..fe373a2101d
--- /dev/null
+++ b/fs/ocfs2/localalloc.c
@@ -0,0 +1,983 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * localalloc.c
+ *
+ * Node local data allocation
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/bitops.h>
+
+#define MLOG_MASK_PREFIX ML_DISK_ALLOC
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "dlmglue.h"
+#include "inode.h"
+#include "journal.h"
+#include "localalloc.h"
+#include "suballoc.h"
+#include "super.h"
+#include "sysfile.h"
+
+#include "buffer_head_io.h"
+
+#define OCFS2_LOCAL_ALLOC(dinode)	(&((dinode)->id2.i_lab))
+
+static inline int ocfs2_local_alloc_window_bits(struct ocfs2_super *osb);
+
+static u32 ocfs2_local_alloc_count_bits(struct ocfs2_dinode *alloc);
+
+static int ocfs2_local_alloc_find_clear_bits(struct ocfs2_super *osb,
+					     struct ocfs2_dinode *alloc,
+					     u32 numbits);
+
+static void ocfs2_clear_local_alloc(struct ocfs2_dinode *alloc);
+
+static int ocfs2_sync_local_to_main(struct ocfs2_super *osb,
+				    struct ocfs2_journal_handle *handle,
+				    struct ocfs2_dinode *alloc,
+				    struct inode *main_bm_inode,
+				    struct buffer_head *main_bm_bh);
+
+static int ocfs2_local_alloc_reserve_for_window(struct ocfs2_super *osb,
+						struct ocfs2_journal_handle *handle,
+						struct ocfs2_alloc_context **ac,
+						struct inode **bitmap_inode,
+						struct buffer_head **bitmap_bh);
+
+static int ocfs2_local_alloc_new_window(struct ocfs2_super *osb,
+					struct ocfs2_journal_handle *handle,
+					struct ocfs2_alloc_context *ac);
+
+static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb,
+					  struct inode *local_alloc_inode);
+
+/*
+ * Determine how large our local alloc window should be, in bits.
+ *
+ * These values (and the behavior in ocfs2_alloc_should_use_local) have
+ * been chosen so that most allocations, including new block groups go
+ * through local alloc.
+ */
+static inline int ocfs2_local_alloc_window_bits(struct ocfs2_super *osb)
+{
+	BUG_ON(osb->s_clustersize_bits < 12);
+
+	return 2048 >> (osb->s_clustersize_bits - 12);
+}
+
+/*
+ * Tell us whether a given allocation should use the local alloc
+ * file. Otherwise, it has to go to the main bitmap.
+ */
+int ocfs2_alloc_should_use_local(struct ocfs2_super *osb, u64 bits)
+{
+	int la_bits = ocfs2_local_alloc_window_bits(osb);
+
+	if (osb->local_alloc_state != OCFS2_LA_ENABLED)
+		return 0;
+
+	/* la_bits should be at least twice the size (in clusters) of
+	 * a new block group. We want to be sure block group
+	 * allocations go through the local alloc, so allow an
+	 * allocation to take up to half the bitmap. */
+	if (bits > (la_bits / 2))
+		return 0;
+
+	return 1;
+}
+
+int ocfs2_load_local_alloc(struct ocfs2_super *osb)
+{
+	int status = 0;
+	struct ocfs2_dinode *alloc = NULL;
+	struct buffer_head *alloc_bh = NULL;
+	u32 num_used;
+	struct inode *inode = NULL;
+	struct ocfs2_local_alloc *la;
+
+	mlog_entry_void();
+
+	/* read the alloc off disk */
+	inode = ocfs2_get_system_file_inode(osb, LOCAL_ALLOC_SYSTEM_INODE,
+					    osb->slot_num);
+	if (!inode) {
+		status = -EINVAL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_read_block(osb, OCFS2_I(inode)->ip_blkno,
+				  &alloc_bh, 0, inode);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	alloc = (struct ocfs2_dinode *) alloc_bh->b_data;
+	la = OCFS2_LOCAL_ALLOC(alloc);
+
+	if (!(le32_to_cpu(alloc->i_flags) &
+	    (OCFS2_LOCAL_ALLOC_FL|OCFS2_BITMAP_FL))) {
+		mlog(ML_ERROR, "Invalid local alloc inode, %"MLFu64"\n",
+		     OCFS2_I(inode)->ip_blkno);
+		status = -EINVAL;
+		goto bail;
+	}
+
+	if ((la->la_size == 0) ||
+	    (le16_to_cpu(la->la_size) > ocfs2_local_alloc_size(inode->i_sb))) {
+		mlog(ML_ERROR, "Local alloc size is invalid (la_size = %u)\n",
+		     le16_to_cpu(la->la_size));
+		status = -EINVAL;
+		goto bail;
+	}
+
+	/* do a little verification. */
+	num_used = ocfs2_local_alloc_count_bits(alloc);
+
+	/* hopefully the local alloc has always been recovered before
+	 * we load it. */
+	if (num_used
+	    || alloc->id1.bitmap1.i_used
+	    || alloc->id1.bitmap1.i_total
+	    || la->la_bm_off)
+		mlog(ML_ERROR, "Local alloc hasn't been recovered!\n"
+		     "found = %u, set = %u, taken = %u, off = %u\n",
+		     num_used, le32_to_cpu(alloc->id1.bitmap1.i_used),
+		     le32_to_cpu(alloc->id1.bitmap1.i_total),
+		     OCFS2_LOCAL_ALLOC(alloc)->la_bm_off);
+
+	osb->local_alloc_bh = alloc_bh;
+	osb->local_alloc_state = OCFS2_LA_ENABLED;
+
+bail:
+	if (status < 0)
+		if (alloc_bh)
+			brelse(alloc_bh);
+	if (inode)
+		iput(inode);
+
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * return any unused bits to the bitmap and write out a clean
+ * local_alloc.
+ *
+ * local_alloc_bh is optional. If not passed, we will simply use the
+ * one off osb. If you do pass it however, be warned that it *will* be
+ * returned brelse'd and NULL'd out.*/
+void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb)
+{
+	int status;
+	struct ocfs2_journal_handle *handle = NULL;
+	struct inode *local_alloc_inode = NULL;
+	struct buffer_head *bh = NULL;
+	struct buffer_head *main_bm_bh = NULL;
+	struct inode *main_bm_inode = NULL;
+	struct ocfs2_dinode *alloc_copy = NULL;
+	struct ocfs2_dinode *alloc = NULL;
+
+	mlog_entry_void();
+
+	if (osb->local_alloc_state == OCFS2_LA_UNUSED)
+		goto bail;
+
+	local_alloc_inode =
+		ocfs2_get_system_file_inode(osb,
+					    LOCAL_ALLOC_SYSTEM_INODE,
+					    osb->slot_num);
+	if (!local_alloc_inode) {
+		status = -ENOENT;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	osb->local_alloc_state = OCFS2_LA_DISABLED;
+
+	handle = ocfs2_alloc_handle(osb);
+	if (!handle) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	main_bm_inode = ocfs2_get_system_file_inode(osb,
+						    GLOBAL_BITMAP_SYSTEM_INODE,
+						    OCFS2_INVALID_SLOT);
+	if (!main_bm_inode) {
+		status = -EINVAL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	ocfs2_handle_add_inode(handle, main_bm_inode);
+	status = ocfs2_meta_lock(main_bm_inode, handle, &main_bm_bh, 1);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* WINDOW_MOVE_CREDITS is a bit heavy... */
+	handle = ocfs2_start_trans(osb, handle, OCFS2_WINDOW_MOVE_CREDITS);
+	if (IS_ERR(handle)) {
+		mlog_errno(PTR_ERR(handle));
+		handle = NULL;
+		goto bail;
+	}
+
+	bh = osb->local_alloc_bh;
+	alloc = (struct ocfs2_dinode *) bh->b_data;
+
+	alloc_copy = kmalloc(bh->b_size, GFP_KERNEL);
+	if (!alloc_copy) {
+		status = -ENOMEM;
+		goto bail;
+	}
+	memcpy(alloc_copy, alloc, bh->b_size);
+
+	status = ocfs2_journal_access(handle, local_alloc_inode, bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	ocfs2_clear_local_alloc(alloc);
+
+	status = ocfs2_journal_dirty(handle, bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	brelse(bh);
+	osb->local_alloc_bh = NULL;
+	osb->local_alloc_state = OCFS2_LA_UNUSED;
+
+	status = ocfs2_sync_local_to_main(osb, handle, alloc_copy,
+					  main_bm_inode, main_bm_bh);
+	if (status < 0)
+		mlog_errno(status);
+
+bail:
+	if (handle)
+		ocfs2_commit_trans(handle);
+
+	if (main_bm_bh)
+		brelse(main_bm_bh);
+
+	if (main_bm_inode)
+		iput(main_bm_inode);
+
+	if (local_alloc_inode)
+		iput(local_alloc_inode);
+
+	if (alloc_copy)
+		kfree(alloc_copy);
+
+	mlog_exit_void();
+}
+
+/*
+ * We want to free the bitmap bits outside of any recovery context as
+ * we'll need a cluster lock to do so, but we must clear the local
+ * alloc before giving up the recovered nodes journal. To solve this,
+ * we kmalloc a copy of the local alloc before it's change for the
+ * caller to process with ocfs2_complete_local_alloc_recovery
+ */
+int ocfs2_begin_local_alloc_recovery(struct ocfs2_super *osb,
+				     int slot_num,
+				     struct ocfs2_dinode **alloc_copy)
+{
+	int status = 0;
+	struct buffer_head *alloc_bh = NULL;
+	struct inode *inode = NULL;
+	struct ocfs2_dinode *alloc;
+
+	mlog_entry("(slot_num = %d)\n", slot_num);
+
+	*alloc_copy = NULL;
+
+	inode = ocfs2_get_system_file_inode(osb,
+					    LOCAL_ALLOC_SYSTEM_INODE,
+					    slot_num);
+	if (!inode) {
+		status = -EINVAL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	down(&inode->i_sem);
+
+	status = ocfs2_read_block(osb, OCFS2_I(inode)->ip_blkno,
+				  &alloc_bh, 0, inode);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	*alloc_copy = kmalloc(alloc_bh->b_size, GFP_KERNEL);
+	if (!(*alloc_copy)) {
+		status = -ENOMEM;
+		goto bail;
+	}
+	memcpy((*alloc_copy), alloc_bh->b_data, alloc_bh->b_size);
+
+	alloc = (struct ocfs2_dinode *) alloc_bh->b_data;
+	ocfs2_clear_local_alloc(alloc);
+
+	status = ocfs2_write_block(osb, alloc_bh, inode);
+	if (status < 0)
+		mlog_errno(status);
+
+bail:
+	if ((status < 0) && (*alloc_copy)) {
+		kfree(*alloc_copy);
+		*alloc_copy = NULL;
+	}
+
+	if (alloc_bh)
+		brelse(alloc_bh);
+
+	if (inode) {
+		up(&inode->i_sem);
+		iput(inode);
+	}
+
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * Step 2: By now, we've completed the journal recovery, we've stamped
+ * a clean local alloc on disk and dropped the node out of the
+ * recovery map. Dlm locks will no longer stall, so lets clear out the
+ * main bitmap.
+ */
+int ocfs2_complete_local_alloc_recovery(struct ocfs2_super *osb,
+					struct ocfs2_dinode *alloc)
+{
+	int status;
+	struct ocfs2_journal_handle *handle = NULL;
+	struct buffer_head *main_bm_bh = NULL;
+	struct inode *main_bm_inode = NULL;
+
+	mlog_entry_void();
+
+	handle = ocfs2_alloc_handle(osb);
+	if (!handle) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	main_bm_inode = ocfs2_get_system_file_inode(osb,
+						    GLOBAL_BITMAP_SYSTEM_INODE,
+						    OCFS2_INVALID_SLOT);
+	if (!main_bm_inode) {
+		status = -EINVAL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	ocfs2_handle_add_inode(handle, main_bm_inode);
+	status = ocfs2_meta_lock(main_bm_inode, handle, &main_bm_bh, 1);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	handle = ocfs2_start_trans(osb, handle, OCFS2_WINDOW_MOVE_CREDITS);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		handle = NULL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* we want the bitmap change to be recorded on disk asap */
+	ocfs2_handle_set_sync(handle, 1);
+
+	status = ocfs2_sync_local_to_main(osb, handle, alloc,
+					  main_bm_inode, main_bm_bh);
+	if (status < 0)
+		mlog_errno(status);
+
+bail:
+	if (handle)
+		ocfs2_commit_trans(handle);
+
+	if (main_bm_bh)
+		brelse(main_bm_bh);
+
+	if (main_bm_inode)
+		iput(main_bm_inode);
+
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * make sure we've got at least bitswanted contiguous bits in the
+ * local alloc. You lose them when you drop i_sem.
+ *
+ * We will add ourselves to the transaction passed in, but may start
+ * our own in order to shift windows.
+ */
+int ocfs2_reserve_local_alloc_bits(struct ocfs2_super *osb,
+				   struct ocfs2_journal_handle *passed_handle,
+				   u32 bits_wanted,
+				   struct ocfs2_alloc_context *ac)
+{
+	int status;
+	struct ocfs2_dinode *alloc;
+	struct inode *local_alloc_inode;
+	unsigned int free_bits;
+
+	mlog_entry_void();
+
+	BUG_ON(!passed_handle);
+	BUG_ON(!ac);
+	BUG_ON(passed_handle->flags & OCFS2_HANDLE_STARTED);
+
+	local_alloc_inode =
+		ocfs2_get_system_file_inode(osb,
+					    LOCAL_ALLOC_SYSTEM_INODE,
+					    osb->slot_num);
+	if (!local_alloc_inode) {
+		status = -ENOENT;
+		mlog_errno(status);
+		goto bail;
+	}
+	ocfs2_handle_add_inode(passed_handle, local_alloc_inode);
+
+	if (osb->local_alloc_state != OCFS2_LA_ENABLED) {
+		status = -ENOSPC;
+		goto bail;
+	}
+
+	if (bits_wanted > ocfs2_local_alloc_window_bits(osb)) {
+		mlog(0, "Asking for more than my max window size!\n");
+		status = -ENOSPC;
+		goto bail;
+	}
+
+	alloc = (struct ocfs2_dinode *) osb->local_alloc_bh->b_data;
+
+	if (le32_to_cpu(alloc->id1.bitmap1.i_used) !=
+	    ocfs2_local_alloc_count_bits(alloc)) {
+		ocfs2_error(osb->sb, "local alloc inode %"MLFu64" says it has "
+			    "%u free bits, but a count shows %u",
+			    le64_to_cpu(alloc->i_blkno),
+			    le32_to_cpu(alloc->id1.bitmap1.i_used),
+			    ocfs2_local_alloc_count_bits(alloc));
+		status = -EIO;
+		goto bail;
+	}
+
+	free_bits = le32_to_cpu(alloc->id1.bitmap1.i_total) -
+		le32_to_cpu(alloc->id1.bitmap1.i_used);
+	if (bits_wanted > free_bits) {
+		/* uhoh, window change time. */
+		status =
+			ocfs2_local_alloc_slide_window(osb, local_alloc_inode);
+		if (status < 0) {
+			if (status != -ENOSPC)
+				mlog_errno(status);
+			goto bail;
+		}
+	}
+
+	ac->ac_inode = igrab(local_alloc_inode);
+	get_bh(osb->local_alloc_bh);
+	ac->ac_bh = osb->local_alloc_bh;
+	ac->ac_which = OCFS2_AC_USE_LOCAL;
+	status = 0;
+bail:
+	if (local_alloc_inode)
+		iput(local_alloc_inode);
+
+	mlog_exit(status);
+	return status;
+}
+
+int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb,
+				 struct ocfs2_journal_handle *handle,
+				 struct ocfs2_alloc_context *ac,
+				 u32 min_bits,
+				 u32 *bit_off,
+				 u32 *num_bits)
+{
+	int status, start;
+	struct inode *local_alloc_inode;
+	u32 bits_wanted;
+	void *bitmap;
+	struct ocfs2_dinode *alloc;
+	struct ocfs2_local_alloc *la;
+
+	mlog_entry_void();
+	BUG_ON(ac->ac_which != OCFS2_AC_USE_LOCAL);
+
+	bits_wanted = ac->ac_bits_wanted - ac->ac_bits_given;
+	local_alloc_inode = ac->ac_inode;
+	alloc = (struct ocfs2_dinode *) osb->local_alloc_bh->b_data;
+	la = OCFS2_LOCAL_ALLOC(alloc);
+
+	start = ocfs2_local_alloc_find_clear_bits(osb, alloc, bits_wanted);
+	if (start == -1) {
+		/* TODO: Shouldn't we just BUG here? */
+		status = -ENOSPC;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	bitmap = la->la_bitmap;
+	*bit_off = le32_to_cpu(la->la_bm_off) + start;
+	/* local alloc is always contiguous by nature -- we never
+	 * delete bits from it! */
+	*num_bits = bits_wanted;
+
+	status = ocfs2_journal_access(handle, local_alloc_inode,
+				      osb->local_alloc_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	while(bits_wanted--)
+		ocfs2_set_bit(start++, bitmap);
+
+	alloc->id1.bitmap1.i_used = cpu_to_le32(*num_bits +
+				le32_to_cpu(alloc->id1.bitmap1.i_used));
+
+	status = ocfs2_journal_dirty(handle, osb->local_alloc_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = 0;
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+static u32 ocfs2_local_alloc_count_bits(struct ocfs2_dinode *alloc)
+{
+	int i;
+	u8 *buffer;
+	u32 count = 0;
+	struct ocfs2_local_alloc *la = OCFS2_LOCAL_ALLOC(alloc);
+
+	mlog_entry_void();
+
+	buffer = la->la_bitmap;
+	for (i = 0; i < le16_to_cpu(la->la_size); i++)
+		count += hweight8(buffer[i]);
+
+	mlog_exit(count);
+	return count;
+}
+
+static int ocfs2_local_alloc_find_clear_bits(struct ocfs2_super *osb,
+					     struct ocfs2_dinode *alloc,
+					     u32 numbits)
+{
+	int numfound, bitoff, left, startoff, lastzero;
+	void *bitmap = NULL;
+
+	mlog_entry("(numbits wanted = %u)\n", numbits);
+
+	if (!alloc->id1.bitmap1.i_total) {
+		mlog(0, "No bits in my window!\n");
+		bitoff = -1;
+		goto bail;
+	}
+
+	bitmap = OCFS2_LOCAL_ALLOC(alloc)->la_bitmap;
+
+	numfound = bitoff = startoff = 0;
+	lastzero = -1;
+	left = le32_to_cpu(alloc->id1.bitmap1.i_total);
+	while ((bitoff = ocfs2_find_next_zero_bit(bitmap, left, startoff)) != -1) {
+		if (bitoff == left) {
+			/* mlog(0, "bitoff (%d) == left", bitoff); */
+			break;
+		}
+		/* mlog(0, "Found a zero: bitoff = %d, startoff = %d, "
+		   "numfound = %d\n", bitoff, startoff, numfound);*/
+
+		/* Ok, we found a zero bit... is it contig. or do we
+		 * start over?*/
+		if (bitoff == startoff) {
+			/* we found a zero */
+			numfound++;
+			startoff++;
+		} else {
+			/* got a zero after some ones */
+			numfound = 1;
+			startoff = bitoff+1;
+		}
+		/* we got everything we needed */
+		if (numfound == numbits) {
+			/* mlog(0, "Found it all!\n"); */
+			break;
+		}
+	}
+
+	mlog(0, "Exiting loop, bitoff = %d, numfound = %d\n", bitoff,
+	     numfound);
+
+	if (numfound == numbits)
+		bitoff = startoff - numfound;
+	else
+		bitoff = -1;
+
+bail:
+	mlog_exit(bitoff);
+	return bitoff;
+}
+
+static void ocfs2_clear_local_alloc(struct ocfs2_dinode *alloc)
+{
+	struct ocfs2_local_alloc *la = OCFS2_LOCAL_ALLOC(alloc);
+	int i;
+	mlog_entry_void();
+
+	alloc->id1.bitmap1.i_total = 0;
+	alloc->id1.bitmap1.i_used = 0;
+	la->la_bm_off = 0;
+	for(i = 0; i < le16_to_cpu(la->la_size); i++)
+		la->la_bitmap[i] = 0;
+
+	mlog_exit_void();
+}
+
+#if 0
+/* turn this on and uncomment below to aid debugging window shifts. */
+static void ocfs2_verify_zero_bits(unsigned long *bitmap,
+				   unsigned int start,
+				   unsigned int count)
+{
+	unsigned int tmp = count;
+	while(tmp--) {
+		if (ocfs2_test_bit(start + tmp, bitmap)) {
+			printk("ocfs2_verify_zero_bits: start = %u, count = "
+			       "%u\n", start, count);
+			printk("ocfs2_verify_zero_bits: bit %u is set!",
+			       start + tmp);
+			BUG();
+		}
+	}
+}
+#endif
+
+/*
+ * sync the local alloc to main bitmap.
+ *
+ * assumes you've already locked the main bitmap -- the bitmap inode
+ * passed is used for caching.
+ */
+static int ocfs2_sync_local_to_main(struct ocfs2_super *osb,
+				    struct ocfs2_journal_handle *handle,
+				    struct ocfs2_dinode *alloc,
+				    struct inode *main_bm_inode,
+				    struct buffer_head *main_bm_bh)
+{
+	int status = 0;
+	int bit_off, left, count, start;
+	u64 la_start_blk;
+	u64 blkno;
+	void *bitmap;
+	struct ocfs2_local_alloc *la = OCFS2_LOCAL_ALLOC(alloc);
+
+	mlog_entry("total = %u, COUNT = %u, used = %u\n",
+		   le32_to_cpu(alloc->id1.bitmap1.i_total),
+		   ocfs2_local_alloc_count_bits(alloc),
+		   le32_to_cpu(alloc->id1.bitmap1.i_used));
+
+	if (!alloc->id1.bitmap1.i_total) {
+		mlog(0, "nothing to sync!\n");
+		goto bail;
+	}
+
+	if (le32_to_cpu(alloc->id1.bitmap1.i_used) ==
+	    le32_to_cpu(alloc->id1.bitmap1.i_total)) {
+		mlog(0, "all bits were taken!\n");
+		goto bail;
+	}
+
+	la_start_blk = ocfs2_clusters_to_blocks(osb->sb,
+						le32_to_cpu(la->la_bm_off));
+	bitmap = la->la_bitmap;
+	start = count = bit_off = 0;
+	left = le32_to_cpu(alloc->id1.bitmap1.i_total);
+
+	while ((bit_off = ocfs2_find_next_zero_bit(bitmap, left, start))
+	       != -1) {
+		if ((bit_off < left) && (bit_off == start)) {
+			count++;
+			start++;
+			continue;
+		}
+		if (count) {
+			blkno = la_start_blk +
+				ocfs2_clusters_to_blocks(osb->sb,
+							 start - count);
+
+			mlog(0, "freeing %u bits starting at local "
+			     "alloc bit %u (la_start_blk = %"MLFu64", "
+			     "blkno = %"MLFu64")\n", count, start - count,
+			     la_start_blk, blkno);
+
+			status = ocfs2_free_clusters(handle, main_bm_inode,
+						     main_bm_bh, blkno, count);
+			if (status < 0) {
+				mlog_errno(status);
+				goto bail;
+			}
+		}
+		if (bit_off >= left)
+			break;
+		count = 1;
+		start = bit_off + 1;
+	}
+
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_local_alloc_reserve_for_window(struct ocfs2_super *osb,
+						struct ocfs2_journal_handle *handle,
+						struct ocfs2_alloc_context **ac,
+						struct inode **bitmap_inode,
+						struct buffer_head **bitmap_bh)
+{
+	int status;
+
+	*ac = kcalloc(1, sizeof(struct ocfs2_alloc_context), GFP_KERNEL);
+	if (!(*ac)) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	(*ac)->ac_handle = handle;
+	(*ac)->ac_bits_wanted = ocfs2_local_alloc_window_bits(osb);
+
+	status = ocfs2_reserve_cluster_bitmap_bits(osb, *ac);
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	*bitmap_inode = (*ac)->ac_inode;
+	igrab(*bitmap_inode);
+	*bitmap_bh = (*ac)->ac_bh;
+	get_bh(*bitmap_bh);
+	status = 0;
+bail:
+	if ((status < 0) && *ac) {
+		ocfs2_free_alloc_context(*ac);
+		*ac = NULL;
+	}
+
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * pass it the bitmap lock in lock_bh if you have it.
+ */
+static int ocfs2_local_alloc_new_window(struct ocfs2_super *osb,
+					struct ocfs2_journal_handle *handle,
+					struct ocfs2_alloc_context *ac)
+{
+	int status = 0;
+	u32 cluster_off, cluster_count;
+	struct ocfs2_dinode *alloc = NULL;
+	struct ocfs2_local_alloc *la;
+
+	mlog_entry_void();
+
+	alloc = (struct ocfs2_dinode *) osb->local_alloc_bh->b_data;
+	la = OCFS2_LOCAL_ALLOC(alloc);
+
+	if (alloc->id1.bitmap1.i_total)
+		mlog(0, "asking me to alloc a new window over a non-empty "
+		     "one\n");
+
+	mlog(0, "Allocating %u clusters for a new window.\n",
+	     ocfs2_local_alloc_window_bits(osb));
+	/* we used the generic suballoc reserve function, but we set
+	 * everything up nicely, so there's no reason why we can't use
+	 * the more specific cluster api to claim bits. */
+	status = ocfs2_claim_clusters(osb, handle, ac,
+				      ocfs2_local_alloc_window_bits(osb),
+				      &cluster_off, &cluster_count);
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	la->la_bm_off = cpu_to_le32(cluster_off);
+	alloc->id1.bitmap1.i_total = cpu_to_le32(cluster_count);
+	/* just in case... In the future when we find space ourselves,
+	 * we don't have to get all contiguous -- but we'll have to
+	 * set all previously used bits in bitmap and update
+	 * la_bits_set before setting the bits in the main bitmap. */
+	alloc->id1.bitmap1.i_used = 0;
+	memset(OCFS2_LOCAL_ALLOC(alloc)->la_bitmap, 0,
+	       le16_to_cpu(la->la_size));
+
+	mlog(0, "New window allocated:\n");
+	mlog(0, "window la_bm_off = %u\n",
+	     OCFS2_LOCAL_ALLOC(alloc)->la_bm_off);
+	mlog(0, "window bits = %u\n", le32_to_cpu(alloc->id1.bitmap1.i_total));
+
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+/* Note that we do *NOT* lock the local alloc inode here as
+ * it's been locked already for us. */
+static int ocfs2_local_alloc_slide_window(struct ocfs2_super *osb,
+					  struct inode *local_alloc_inode)
+{
+	int status = 0;
+	struct buffer_head *main_bm_bh = NULL;
+	struct inode *main_bm_inode = NULL;
+	struct ocfs2_journal_handle *handle = NULL;
+	struct ocfs2_dinode *alloc;
+	struct ocfs2_dinode *alloc_copy = NULL;
+	struct ocfs2_alloc_context *ac = NULL;
+
+	mlog_entry_void();
+
+	handle = ocfs2_alloc_handle(osb);
+	if (!handle) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* This will lock the main bitmap for us. */
+	status = ocfs2_local_alloc_reserve_for_window(osb,
+						      handle,
+						      &ac,
+						      &main_bm_inode,
+						      &main_bm_bh);
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	handle = ocfs2_start_trans(osb, handle, OCFS2_WINDOW_MOVE_CREDITS);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		handle = NULL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	alloc = (struct ocfs2_dinode *) osb->local_alloc_bh->b_data;
+
+	/* We want to clear the local alloc before doing anything
+	 * else, so that if we error later during this operation,
+	 * local alloc shutdown won't try to double free main bitmap
+	 * bits. Make a copy so the sync function knows which bits to
+	 * free. */
+	alloc_copy = kmalloc(osb->local_alloc_bh->b_size, GFP_KERNEL);
+	if (!alloc_copy) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+	memcpy(alloc_copy, alloc, osb->local_alloc_bh->b_size);
+
+	status = ocfs2_journal_access(handle, local_alloc_inode,
+				      osb->local_alloc_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	ocfs2_clear_local_alloc(alloc);
+
+	status = ocfs2_journal_dirty(handle, osb->local_alloc_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_sync_local_to_main(osb, handle, alloc_copy,
+					  main_bm_inode, main_bm_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_local_alloc_new_window(osb, handle, ac);
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	atomic_inc(&osb->alloc_stats.moves);
+
+	status = 0;
+bail:
+	if (handle)
+		ocfs2_commit_trans(handle);
+
+	if (main_bm_bh)
+		brelse(main_bm_bh);
+
+	if (main_bm_inode)
+		iput(main_bm_inode);
+
+	if (alloc_copy)
+		kfree(alloc_copy);
+
+	if (ac)
+		ocfs2_free_alloc_context(ac);
+
+	mlog_exit(status);
+	return status;
+}
+
diff --git a/fs/ocfs2/localalloc.h b/fs/ocfs2/localalloc.h
new file mode 100644
index 00000000000..30f88ce14e4
--- /dev/null
+++ b/fs/ocfs2/localalloc.h
@@ -0,0 +1,56 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * localalloc.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_LOCALALLOC_H
+#define OCFS2_LOCALALLOC_H
+
+int ocfs2_load_local_alloc(struct ocfs2_super *osb);
+
+void ocfs2_shutdown_local_alloc(struct ocfs2_super *osb);
+
+int ocfs2_begin_local_alloc_recovery(struct ocfs2_super *osb,
+				     int node_num,
+				     struct ocfs2_dinode **alloc_copy);
+
+int ocfs2_complete_local_alloc_recovery(struct ocfs2_super *osb,
+					struct ocfs2_dinode *alloc);
+
+int ocfs2_alloc_should_use_local(struct ocfs2_super *osb,
+				 u64 bits);
+
+struct ocfs2_alloc_context;
+int ocfs2_reserve_local_alloc_bits(struct ocfs2_super *osb,
+				   struct ocfs2_journal_handle *passed_handle,
+				   u32 bits_wanted,
+				   struct ocfs2_alloc_context *ac);
+
+int ocfs2_claim_local_alloc_bits(struct ocfs2_super *osb,
+				 struct ocfs2_journal_handle *handle,
+				 struct ocfs2_alloc_context *ac,
+				 u32 min_bits,
+				 u32 *bit_off,
+				 u32 *num_bits);
+
+#endif /* OCFS2_LOCALALLOC_H */
diff --git a/fs/ocfs2/mmap.c b/fs/ocfs2/mmap.c
new file mode 100644
index 00000000000..afdeec4b0ee
--- /dev/null
+++ b/fs/ocfs2/mmap.c
@@ -0,0 +1,102 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * mmap.c
+ *
+ * Code to deal with the mess that is clustered mmap.
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/pagemap.h>
+#include <linux/uio.h>
+#include <linux/signal.h>
+#include <linux/rbtree.h>
+
+#define MLOG_MASK_PREFIX ML_FILE_IO
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "dlmglue.h"
+#include "file.h"
+#include "inode.h"
+#include "mmap.h"
+
+static struct page *ocfs2_nopage(struct vm_area_struct * area,
+				 unsigned long address,
+				 int *type)
+{
+	struct inode *inode = area->vm_file->f_dentry->d_inode;
+	struct page *page = NOPAGE_SIGBUS;
+	sigset_t blocked, oldset;
+	int ret;
+
+	mlog_entry("(inode %lu, address %lu)\n", inode->i_ino, address);
+
+	/* The best way to deal with signals in this path is
+	 * to block them upfront, rather than allowing the
+	 * locking paths to return -ERESTARTSYS. */
+	sigfillset(&blocked);
+
+	/* We should technically never get a bad ret return
+	 * from sigprocmask */
+	ret = sigprocmask(SIG_BLOCK, &blocked, &oldset);
+	if (ret < 0) {
+		mlog_errno(ret);
+		goto out;
+	}
+
+	page = filemap_nopage(area, address, type);
+
+	ret = sigprocmask(SIG_SETMASK, &oldset, NULL);
+	if (ret < 0)
+		mlog_errno(ret);
+out:
+	mlog_exit_ptr(page);
+	return page;
+}
+
+static struct vm_operations_struct ocfs2_file_vm_ops = {
+	.nopage = ocfs2_nopage,
+};
+
+int ocfs2_mmap(struct file *file,
+	       struct vm_area_struct *vma)
+{
+	struct address_space *mapping = file->f_dentry->d_inode->i_mapping;
+	struct inode *inode = mapping->host;
+
+	/* We don't want to support shared writable mappings yet. */
+	if (((vma->vm_flags & VM_SHARED) || (vma->vm_flags & VM_MAYSHARE))
+	    && ((vma->vm_flags & VM_WRITE) || (vma->vm_flags & VM_MAYWRITE))) {
+		mlog(0, "disallow shared writable mmaps %lx\n", vma->vm_flags);
+		/* This is -EINVAL because generic_file_readonly_mmap
+		 * returns it in a similar situation. */
+		return -EINVAL;
+	}
+
+	update_atime(inode);
+	vma->vm_ops = &ocfs2_file_vm_ops;
+	return 0;
+}
+
diff --git a/fs/ocfs2/mmap.h b/fs/ocfs2/mmap.h
new file mode 100644
index 00000000000..1274ee0f1fe
--- /dev/null
+++ b/fs/ocfs2/mmap.h
@@ -0,0 +1,6 @@
+#ifndef OCFS2_MMAP_H
+#define OCFS2_MMAP_H
+
+int ocfs2_mmap(struct file *file, struct vm_area_struct *vma);
+
+#endif  /* OCFS2_MMAP_H */
diff --git a/fs/ocfs2/namei.c b/fs/ocfs2/namei.c
new file mode 100644
index 00000000000..f6b77ff1d2b
--- /dev/null
+++ b/fs/ocfs2/namei.c
@@ -0,0 +1,2264 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * namei.c
+ *
+ * Create and rename file, directory, symlinks
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ *  Portions of this code from linux/fs/ext3/dir.c
+ *
+ *  Copyright (C) 1992, 1993, 1994, 1995
+ *  Remy Card (card@masi.ibp.fr)
+ *  Laboratoire MASI - Institut Blaise pascal
+ *  Universite Pierre et Marie Curie (Paris VI)
+ *
+ *   from
+ *
+ *   linux/fs/minix/dir.c
+ *
+ *   Copyright (C) 1991, 1992 Linux Torvalds
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+
+#define MLOG_MASK_PREFIX ML_NAMEI
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "dcache.h"
+#include "dir.h"
+#include "dlmglue.h"
+#include "extent_map.h"
+#include "file.h"
+#include "inode.h"
+#include "journal.h"
+#include "namei.h"
+#include "suballoc.h"
+#include "symlink.h"
+#include "sysfile.h"
+#include "uptodate.h"
+#include "vote.h"
+
+#include "buffer_head_io.h"
+
+#define NAMEI_RA_CHUNKS  2
+#define NAMEI_RA_BLOCKS  4
+#define NAMEI_RA_SIZE        (NAMEI_RA_CHUNKS * NAMEI_RA_BLOCKS)
+#define NAMEI_RA_INDEX(c,b)  (((c) * NAMEI_RA_BLOCKS) + (b))
+
+static int inline ocfs2_search_dirblock(struct buffer_head *bh,
+					struct inode *dir,
+					const char *name, int namelen,
+					unsigned long offset,
+					struct ocfs2_dir_entry **res_dir);
+
+static int ocfs2_delete_entry(struct ocfs2_journal_handle *handle,
+			      struct inode *dir,
+			      struct ocfs2_dir_entry *de_del,
+			      struct buffer_head *bh);
+
+static int __ocfs2_add_entry(struct ocfs2_journal_handle *handle,
+			     struct inode *dir,
+			     const char *name, int namelen,
+			     struct inode *inode, u64 blkno,
+			     struct buffer_head *parent_fe_bh,
+			     struct buffer_head *insert_bh);
+
+static int ocfs2_mknod_locked(struct ocfs2_super *osb,
+			      struct inode *dir,
+			      struct dentry *dentry, int mode,
+			      dev_t dev,
+			      struct buffer_head **new_fe_bh,
+			      struct buffer_head *parent_fe_bh,
+			      struct ocfs2_journal_handle *handle,
+			      struct inode **ret_inode,
+			      struct ocfs2_alloc_context *inode_ac);
+
+static int ocfs2_fill_new_dir(struct ocfs2_super *osb,
+			      struct ocfs2_journal_handle *handle,
+			      struct inode *parent,
+			      struct inode *inode,
+			      struct buffer_head *fe_bh,
+			      struct ocfs2_alloc_context *data_ac);
+
+static int ocfs2_double_lock(struct ocfs2_super *osb,
+			     struct ocfs2_journal_handle *handle,
+			     struct buffer_head **bh1,
+			     struct inode *inode1,
+			     struct buffer_head **bh2,
+			     struct inode *inode2);
+
+static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb,
+				    struct ocfs2_journal_handle *handle,
+				    struct inode *inode,
+				    char *name,
+				    struct buffer_head **de_bh);
+
+static int ocfs2_orphan_add(struct ocfs2_super *osb,
+			    struct ocfs2_journal_handle *handle,
+			    struct inode *inode,
+			    struct ocfs2_dinode *fe,
+			    char *name,
+			    struct buffer_head *de_bh);
+
+static int ocfs2_create_symlink_data(struct ocfs2_super *osb,
+				     struct ocfs2_journal_handle *handle,
+				     struct inode *inode,
+				     const char *symname);
+
+static inline int ocfs2_add_entry(struct ocfs2_journal_handle *handle,
+				  struct dentry *dentry,
+				  struct inode *inode, u64 blkno,
+				  struct buffer_head *parent_fe_bh,
+				  struct buffer_head *insert_bh)
+{
+	return __ocfs2_add_entry(handle, dentry->d_parent->d_inode,
+				 dentry->d_name.name, dentry->d_name.len,
+				 inode, blkno, parent_fe_bh, insert_bh);
+}
+
+/* An orphan dir name is an 8 byte value, printed as a hex string */
+#define OCFS2_ORPHAN_NAMELEN ((int)(2 * sizeof(u64)))
+
+static struct dentry *ocfs2_lookup(struct inode *dir, struct dentry *dentry,
+				   struct nameidata *nd)
+{
+	int status;
+	u64 blkno;
+	struct buffer_head *dirent_bh = NULL;
+	struct inode *inode = NULL;
+	struct dentry *ret;
+	struct ocfs2_dir_entry *dirent;
+	struct ocfs2_inode_info *oi;
+
+	mlog_entry("(0x%p, 0x%p, '%.*s')\n", dir, dentry,
+		   dentry->d_name.len, dentry->d_name.name);
+
+	if (dentry->d_name.len > OCFS2_MAX_FILENAME_LEN) {
+		ret = ERR_PTR(-ENAMETOOLONG);
+		goto bail;
+	}
+
+	mlog(0, "find name %.*s in directory %"MLFu64"\n", dentry->d_name.len,
+	     dentry->d_name.name, OCFS2_I(dir)->ip_blkno);
+
+	status = ocfs2_meta_lock(dir, NULL, NULL, 0);
+	if (status < 0) {
+		if (status != -ENOENT)
+			mlog_errno(status);
+		ret = ERR_PTR(status);
+		goto bail;
+	}
+
+	status = ocfs2_find_files_on_disk(dentry->d_name.name,
+					  dentry->d_name.len, &blkno,
+					  dir, &dirent_bh, &dirent);
+	if (status < 0)
+		goto bail_add;
+
+	inode = ocfs2_iget(OCFS2_SB(dir->i_sb), blkno);
+	if (IS_ERR(inode)) {
+		mlog(ML_ERROR, "Unable to create inode %"MLFu64"\n", blkno);
+		ret = ERR_PTR(-EACCES);
+		goto bail_unlock;
+	}
+
+	oi = OCFS2_I(inode);
+	/* Clear any orphaned state... If we were able to look up the
+	 * inode from a directory, it certainly can't be orphaned. We
+	 * might have the bad state from a node which intended to
+	 * orphan this inode but crashed before it could commit the
+	 * unlink. */
+	spin_lock(&oi->ip_lock);
+	oi->ip_flags &= ~OCFS2_INODE_MAYBE_ORPHANED;
+	oi->ip_orphaned_slot = OCFS2_INVALID_SLOT;
+	spin_unlock(&oi->ip_lock);
+
+bail_add:
+
+	dentry->d_op = &ocfs2_dentry_ops;
+	ret = d_splice_alias(inode, dentry);
+
+bail_unlock:
+	/* Don't drop the cluster lock until *after* the d_add --
+	 * unlink on another node will message us to remove that
+	 * dentry under this lock so otherwise we can race this with
+	 * the vote thread and have a stale dentry. */
+	ocfs2_meta_unlock(dir, 0);
+
+bail:
+	if (dirent_bh)
+		brelse(dirent_bh);
+
+	mlog_exit_ptr(ret);
+
+	return ret;
+}
+
+static int ocfs2_fill_new_dir(struct ocfs2_super *osb,
+			      struct ocfs2_journal_handle *handle,
+			      struct inode *parent,
+			      struct inode *inode,
+			      struct buffer_head *fe_bh,
+			      struct ocfs2_alloc_context *data_ac)
+{
+	int status;
+	struct buffer_head *new_bh = NULL;
+	struct ocfs2_dir_entry *de = NULL;
+
+	mlog_entry_void();
+
+	status = ocfs2_do_extend_dir(osb->sb, handle, inode, fe_bh,
+				     data_ac, NULL, &new_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	ocfs2_set_new_buffer_uptodate(inode, new_bh);
+
+	status = ocfs2_journal_access(handle, inode, new_bh,
+				      OCFS2_JOURNAL_ACCESS_CREATE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+	memset(new_bh->b_data, 0, osb->sb->s_blocksize);
+
+	de = (struct ocfs2_dir_entry *) new_bh->b_data;
+	de->inode = cpu_to_le64(OCFS2_I(inode)->ip_blkno);
+	de->name_len = 1;
+	de->rec_len =
+		cpu_to_le16(OCFS2_DIR_REC_LEN(de->name_len));
+	strcpy(de->name, ".");
+	ocfs2_set_de_type(de, S_IFDIR);
+	de = (struct ocfs2_dir_entry *) ((char *)de + le16_to_cpu(de->rec_len));
+	de->inode = cpu_to_le64(OCFS2_I(parent)->ip_blkno);
+	de->rec_len = cpu_to_le16(inode->i_sb->s_blocksize -
+				  OCFS2_DIR_REC_LEN(1));
+	de->name_len = 2;
+	strcpy(de->name, "..");
+	ocfs2_set_de_type(de, S_IFDIR);
+
+	status = ocfs2_journal_dirty(handle, new_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	i_size_write(inode, inode->i_sb->s_blocksize);
+	inode->i_nlink = 2;
+	inode->i_blocks = ocfs2_align_bytes_to_sectors(inode->i_sb->s_blocksize);
+	status = ocfs2_mark_inode_dirty(handle, inode, fe_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = 0;
+bail:
+	if (new_bh)
+		brelse(new_bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_mknod(struct inode *dir,
+		       struct dentry *dentry,
+		       int mode,
+		       dev_t dev)
+{
+	int status = 0;
+	struct buffer_head *parent_fe_bh = NULL;
+	struct ocfs2_journal_handle *handle = NULL;
+	struct ocfs2_super *osb;
+	struct ocfs2_dinode *dirfe;
+	struct buffer_head *new_fe_bh = NULL;
+	struct buffer_head *de_bh = NULL;
+	struct inode *inode = NULL;
+	struct ocfs2_alloc_context *inode_ac = NULL;
+	struct ocfs2_alloc_context *data_ac = NULL;
+
+	mlog_entry("(0x%p, 0x%p, %d, %lu, '%.*s')\n", dir, dentry, mode,
+		   (unsigned long)dev, dentry->d_name.len,
+		   dentry->d_name.name);
+
+	/* get our super block */
+	osb = OCFS2_SB(dir->i_sb);
+
+	if (S_ISDIR(mode) && (dir->i_nlink >= OCFS2_LINK_MAX)) {
+		mlog(ML_ERROR, "inode %"MLFu64" has i_nlink of %u\n",
+		     OCFS2_I(dir)->ip_blkno, dir->i_nlink);
+		status = -EMLINK;
+		goto leave;
+	}
+
+	handle = ocfs2_alloc_handle(osb);
+	if (handle == NULL) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto leave;
+	}
+
+	status = ocfs2_meta_lock(dir, handle, &parent_fe_bh, 1);
+	if (status < 0) {
+		if (status != -ENOENT)
+			mlog_errno(status);
+		goto leave;
+	}
+
+	dirfe = (struct ocfs2_dinode *) parent_fe_bh->b_data;
+	if (!dirfe->i_links_count) {
+		/* can't make a file in a deleted directory. */
+		status = -ENOENT;
+		goto leave;
+	}
+
+	status = ocfs2_check_dir_for_entry(dir, dentry->d_name.name,
+					   dentry->d_name.len);
+	if (status)
+		goto leave;
+
+	/* get a spot inside the dir. */
+	status = ocfs2_prepare_dir_for_insert(osb, dir, parent_fe_bh,
+					      dentry->d_name.name,
+					      dentry->d_name.len, &de_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	/* reserve an inode spot */
+	status = ocfs2_reserve_new_inode(osb, handle, &inode_ac);
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto leave;
+	}
+
+	/* are we making a directory? If so, reserve a cluster for his
+	 * 1st extent. */
+	if (S_ISDIR(mode)) {
+		status = ocfs2_reserve_clusters(osb, handle, 1, &data_ac);
+		if (status < 0) {
+			if (status != -ENOSPC)
+				mlog_errno(status);
+			goto leave;
+		}
+	}
+
+	handle = ocfs2_start_trans(osb, handle, OCFS2_MKNOD_CREDITS);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		handle = NULL;
+		mlog_errno(status);
+		goto leave;
+	}
+
+	/* do the real work now. */
+	status = ocfs2_mknod_locked(osb, dir, dentry, mode, dev,
+				    &new_fe_bh, parent_fe_bh, handle,
+				    &inode, inode_ac);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	if (S_ISDIR(mode)) {
+		status = ocfs2_fill_new_dir(osb, handle, dir, inode,
+					    new_fe_bh, data_ac);
+		if (status < 0) {
+			mlog_errno(status);
+			goto leave;
+		}
+
+		status = ocfs2_journal_access(handle, dir, parent_fe_bh,
+					      OCFS2_JOURNAL_ACCESS_WRITE);
+		if (status < 0) {
+			mlog_errno(status);
+			goto leave;
+		}
+		le16_add_cpu(&dirfe->i_links_count, 1);
+		status = ocfs2_journal_dirty(handle, parent_fe_bh);
+		if (status < 0) {
+			mlog_errno(status);
+			goto leave;
+		}
+		dir->i_nlink++;
+	}
+
+	status = ocfs2_add_entry(handle, dentry, inode,
+				 OCFS2_I(inode)->ip_blkno, parent_fe_bh,
+				 de_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	insert_inode_hash(inode);
+	dentry->d_op = &ocfs2_dentry_ops;
+	d_instantiate(dentry, inode);
+	status = 0;
+leave:
+	if (handle)
+		ocfs2_commit_trans(handle);
+
+	if (status == -ENOSPC)
+		mlog(0, "Disk is full\n");
+
+	if (new_fe_bh)
+		brelse(new_fe_bh);
+
+	if (de_bh)
+		brelse(de_bh);
+
+	if (parent_fe_bh)
+		brelse(parent_fe_bh);
+
+	if ((status < 0) && inode)
+		iput(inode);
+
+	if (inode_ac)
+		ocfs2_free_alloc_context(inode_ac);
+
+	if (data_ac)
+		ocfs2_free_alloc_context(data_ac);
+
+	mlog_exit(status);
+
+	return status;
+}
+
+static int ocfs2_mknod_locked(struct ocfs2_super *osb,
+			      struct inode *dir,
+			      struct dentry *dentry, int mode,
+			      dev_t dev,
+			      struct buffer_head **new_fe_bh,
+			      struct buffer_head *parent_fe_bh,
+			      struct ocfs2_journal_handle *handle,
+			      struct inode **ret_inode,
+			      struct ocfs2_alloc_context *inode_ac)
+{
+	int status = 0;
+	struct ocfs2_dinode *fe = NULL;
+	struct ocfs2_extent_list *fel;
+	u64 fe_blkno = 0;
+	u16 suballoc_bit;
+	struct inode *inode = NULL;
+
+	mlog_entry("(0x%p, 0x%p, %d, %lu, '%.*s')\n", dir, dentry, mode,
+		   (unsigned long)dev, dentry->d_name.len,
+		   dentry->d_name.name);
+
+	*new_fe_bh = NULL;
+	*ret_inode = NULL;
+
+	status = ocfs2_claim_new_inode(osb, handle, inode_ac, &suballoc_bit,
+				       &fe_blkno);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	inode = new_inode(dir->i_sb);
+	if (IS_ERR(inode)) {
+		status = PTR_ERR(inode);
+		mlog(ML_ERROR, "new_inode failed!\n");
+		goto leave;
+	}
+
+	/* populate as many fields early on as possible - many of
+	 * these are used by the support functions here and in
+	 * callers. */
+	inode->i_ino = ino_from_blkno(osb->sb, fe_blkno);
+	OCFS2_I(inode)->ip_blkno = fe_blkno;
+	if (S_ISDIR(mode))
+		inode->i_nlink = 2;
+	else
+		inode->i_nlink = 1;
+	inode->i_mode = mode;
+	spin_lock(&osb->osb_lock);
+	inode->i_generation = osb->s_next_generation++;
+	spin_unlock(&osb->osb_lock);
+
+	*new_fe_bh = sb_getblk(osb->sb, fe_blkno);
+	if (!*new_fe_bh) {
+		status = -EIO;
+		mlog_errno(status);
+		goto leave;
+	}
+	ocfs2_set_new_buffer_uptodate(inode, *new_fe_bh);
+
+	status = ocfs2_journal_access(handle, inode, *new_fe_bh,
+				      OCFS2_JOURNAL_ACCESS_CREATE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	fe = (struct ocfs2_dinode *) (*new_fe_bh)->b_data;
+	memset(fe, 0, osb->sb->s_blocksize);
+
+	fe->i_generation = cpu_to_le32(inode->i_generation);
+	fe->i_fs_generation = cpu_to_le32(osb->fs_generation);
+	fe->i_blkno = cpu_to_le64(fe_blkno);
+	fe->i_suballoc_bit = cpu_to_le16(suballoc_bit);
+	fe->i_suballoc_slot = cpu_to_le16(osb->slot_num);
+	fe->i_uid = cpu_to_le32(current->fsuid);
+	if (dir->i_mode & S_ISGID) {
+		fe->i_gid = cpu_to_le32(dir->i_gid);
+		if (S_ISDIR(mode))
+			mode |= S_ISGID;
+	} else
+		fe->i_gid = cpu_to_le32(current->fsgid);
+	fe->i_mode = cpu_to_le16(mode);
+	if (S_ISCHR(mode) || S_ISBLK(mode))
+		fe->id1.dev1.i_rdev = cpu_to_le64(huge_encode_dev(dev));
+
+	fe->i_links_count = cpu_to_le16(inode->i_nlink);
+
+	fe->i_last_eb_blk = 0;
+	strcpy(fe->i_signature, OCFS2_INODE_SIGNATURE);
+	le32_add_cpu(&fe->i_flags, OCFS2_VALID_FL);
+	fe->i_atime = fe->i_ctime = fe->i_mtime =
+		cpu_to_le64(CURRENT_TIME.tv_sec);
+	fe->i_mtime_nsec = fe->i_ctime_nsec = fe->i_atime_nsec =
+		cpu_to_le32(CURRENT_TIME.tv_nsec);
+	fe->i_dtime = 0;
+
+	fel = &fe->id2.i_list;
+	fel->l_tree_depth = 0;
+	fel->l_next_free_rec = 0;
+	fel->l_count = cpu_to_le16(ocfs2_extent_recs_per_inode(osb->sb));
+
+	status = ocfs2_journal_dirty(handle, *new_fe_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	if (ocfs2_populate_inode(inode, fe, 1) < 0) {
+		mlog(ML_ERROR, "populate inode failed! bh->b_blocknr=%llu, "
+		     "i_blkno=%"MLFu64", i_ino=%lu\n",
+		     (unsigned long long) (*new_fe_bh)->b_blocknr,
+		     fe->i_blkno, inode->i_ino);
+		BUG();
+	}
+
+	ocfs2_inode_set_new(osb, inode);
+	status = ocfs2_create_new_inode_locks(inode);
+	if (status < 0)
+		mlog_errno(status);
+
+	status = 0; /* error in ocfs2_create_new_inode_locks is not
+		     * critical */
+
+	*ret_inode = inode;
+leave:
+	if (status < 0) {
+		if (*new_fe_bh) {
+			brelse(*new_fe_bh);
+			*new_fe_bh = NULL;
+		}
+		if (inode)
+			iput(inode);
+	}
+
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_mkdir(struct inode *dir,
+		       struct dentry *dentry,
+		       int mode)
+{
+	int ret;
+
+	mlog_entry("(0x%p, 0x%p, %d, '%.*s')\n", dir, dentry, mode,
+		   dentry->d_name.len, dentry->d_name.name);
+	ret = ocfs2_mknod(dir, dentry, mode | S_IFDIR, 0);
+	mlog_exit(ret);
+
+	return ret;
+}
+
+static int ocfs2_create(struct inode *dir,
+			struct dentry *dentry,
+			int mode,
+			struct nameidata *nd)
+{
+	int ret;
+
+	mlog_entry("(0x%p, 0x%p, %d, '%.*s')\n", dir, dentry, mode,
+		   dentry->d_name.len, dentry->d_name.name);
+	ret = ocfs2_mknod(dir, dentry, mode | S_IFREG, 0);
+	mlog_exit(ret);
+
+	return ret;
+}
+
+static int ocfs2_link(struct dentry *old_dentry,
+		      struct inode *dir,
+		      struct dentry *dentry)
+{
+	struct ocfs2_journal_handle *handle = NULL;
+	struct inode *inode = old_dentry->d_inode;
+	int err;
+	struct buffer_head *fe_bh = NULL;
+	struct buffer_head *parent_fe_bh = NULL;
+	struct buffer_head *de_bh = NULL;
+	struct ocfs2_dinode *fe = NULL;
+	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
+
+	mlog_entry("(inode=%lu, old='%.*s' new='%.*s')\n", inode->i_ino,
+		   old_dentry->d_name.len, old_dentry->d_name.name,
+		   dentry->d_name.len, dentry->d_name.name);
+
+	if (S_ISDIR(inode->i_mode)) {
+		err = -EPERM;
+		goto bail;
+	}
+
+	if (inode->i_nlink >= OCFS2_LINK_MAX) {
+		err = -EMLINK;
+		goto bail;
+	}
+
+	handle = ocfs2_alloc_handle(osb);
+	if (handle == NULL) {
+		err = -ENOMEM;
+		goto bail;
+	}
+
+	err = ocfs2_meta_lock(dir, handle, &parent_fe_bh, 1);
+	if (err < 0) {
+		if (err != -ENOENT)
+			mlog_errno(err);
+		goto bail;
+	}
+
+	err = ocfs2_check_dir_for_entry(dir, dentry->d_name.name,
+					dentry->d_name.len);
+	if (err)
+		goto bail;
+
+	err = ocfs2_prepare_dir_for_insert(osb, dir, parent_fe_bh,
+					   dentry->d_name.name,
+					   dentry->d_name.len, &de_bh);
+	if (err < 0) {
+		mlog_errno(err);
+		goto bail;
+	}
+
+	err = ocfs2_meta_lock(inode, handle, &fe_bh, 1);
+	if (err < 0) {
+		if (err != -ENOENT)
+			mlog_errno(err);
+		goto bail;
+	}
+
+	fe = (struct ocfs2_dinode *) fe_bh->b_data;
+	if (le16_to_cpu(fe->i_links_count) >= OCFS2_LINK_MAX) {
+		err = -EMLINK;
+		goto bail;
+	}
+
+	handle = ocfs2_start_trans(osb, handle, OCFS2_LINK_CREDITS);
+	if (IS_ERR(handle)) {
+		err = PTR_ERR(handle);
+		handle = NULL;
+		mlog_errno(err);
+		goto bail;
+	}
+
+	err = ocfs2_journal_access(handle, inode, fe_bh,
+				   OCFS2_JOURNAL_ACCESS_WRITE);
+	if (err < 0) {
+		mlog_errno(err);
+		goto bail;
+	}
+
+	inode->i_nlink++;
+	inode->i_ctime = CURRENT_TIME;
+	fe->i_links_count = cpu_to_le16(inode->i_nlink);
+	fe->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
+	fe->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
+
+	err = ocfs2_journal_dirty(handle, fe_bh);
+	if (err < 0) {
+		le16_add_cpu(&fe->i_links_count, -1);
+		inode->i_nlink--;
+		mlog_errno(err);
+		goto bail;
+	}
+
+	err = ocfs2_add_entry(handle, dentry, inode,
+			      OCFS2_I(inode)->ip_blkno,
+			      parent_fe_bh, de_bh);
+	if (err) {
+		le16_add_cpu(&fe->i_links_count, -1);
+		inode->i_nlink--;
+		mlog_errno(err);
+		goto bail;
+	}
+
+	atomic_inc(&inode->i_count);
+	dentry->d_op = &ocfs2_dentry_ops;
+	d_instantiate(dentry, inode);
+bail:
+	if (handle)
+		ocfs2_commit_trans(handle);
+	if (de_bh)
+		brelse(de_bh);
+	if (fe_bh)
+		brelse(fe_bh);
+	if (parent_fe_bh)
+		brelse(parent_fe_bh);
+
+	mlog_exit(err);
+
+	return err;
+}
+
+static int ocfs2_unlink(struct inode *dir,
+			struct dentry *dentry)
+{
+	int status;
+	unsigned int saved_nlink = 0;
+	struct inode *inode = dentry->d_inode;
+	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
+	u64 blkno;
+	struct ocfs2_dinode *fe = NULL;
+	struct buffer_head *fe_bh = NULL;
+	struct buffer_head *parent_node_bh = NULL;
+	struct ocfs2_journal_handle *handle = NULL;
+	struct ocfs2_dir_entry *dirent = NULL;
+	struct buffer_head *dirent_bh = NULL;
+	char orphan_name[OCFS2_ORPHAN_NAMELEN + 1];
+	struct buffer_head *orphan_entry_bh = NULL;
+
+	mlog_entry("(0x%p, 0x%p, '%.*s')\n", dir, dentry,
+		   dentry->d_name.len, dentry->d_name.name);
+
+	BUG_ON(dentry->d_parent->d_inode != dir);
+
+	mlog(0, "ino = %"MLFu64"\n", OCFS2_I(inode)->ip_blkno);
+
+	if (inode == osb->root_inode) {
+		mlog(0, "Cannot delete the root directory\n");
+		status = -EPERM;
+		goto leave;
+	}
+
+	handle = ocfs2_alloc_handle(osb);
+	if (handle == NULL) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto leave;
+	}
+
+	status = ocfs2_meta_lock(dir, handle, &parent_node_bh, 1);
+	if (status < 0) {
+		if (status != -ENOENT)
+			mlog_errno(status);
+		goto leave;
+	}
+
+	status = ocfs2_find_files_on_disk(dentry->d_name.name,
+					  dentry->d_name.len, &blkno,
+					  dir, &dirent_bh, &dirent);
+	if (status < 0) {
+		if (status != -ENOENT)
+			mlog_errno(status);
+		goto leave;
+	}
+
+	if (OCFS2_I(inode)->ip_blkno != blkno) {
+		status = -ENOENT;
+
+		mlog(0, "ip_blkno (%"MLFu64") != dirent blkno (%"MLFu64") "
+		     "ip_flags = %x\n", OCFS2_I(inode)->ip_blkno, blkno,
+		     OCFS2_I(inode)->ip_flags);
+		goto leave;
+	}
+
+	status = ocfs2_meta_lock(inode, handle, &fe_bh, 1);
+	if (status < 0) {
+		if (status != -ENOENT)
+			mlog_errno(status);
+		goto leave;
+	}
+
+	if (S_ISDIR(inode->i_mode)) {
+	       	if (!ocfs2_empty_dir(inode)) {
+			status = -ENOTEMPTY;
+			goto leave;
+		} else if (inode->i_nlink != 2) {
+			status = -ENOTEMPTY;
+			goto leave;
+		}
+	}
+
+	/* There are still a few steps left until we can consider the
+	 * unlink to have succeeded. Save off nlink here before
+	 * modification so we can set it back in case we hit an issue
+	 * before commit. */
+	saved_nlink = inode->i_nlink;
+	if (S_ISDIR(inode->i_mode))
+		inode->i_nlink = 0;
+	else
+		inode->i_nlink--;
+
+	status = ocfs2_request_unlink_vote(inode, dentry,
+					   (unsigned int) inode->i_nlink);
+	if (status < 0) {
+		/* This vote should succeed under all normal
+		 * circumstances. */
+		mlog_errno(status);
+		goto leave;
+	}
+
+	if (!inode->i_nlink) {
+		status = ocfs2_prepare_orphan_dir(osb, handle, inode,
+						  orphan_name,
+						  &orphan_entry_bh);
+		if (status < 0) {
+			mlog_errno(status);
+			goto leave;
+		}
+	}
+
+	handle = ocfs2_start_trans(osb, handle, OCFS2_UNLINK_CREDITS);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		handle = NULL;
+		mlog_errno(status);
+		goto leave;
+	}
+
+	status = ocfs2_journal_access(handle, inode, fe_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	fe = (struct ocfs2_dinode *) fe_bh->b_data;
+
+	if (!inode->i_nlink) {
+		status = ocfs2_orphan_add(osb, handle, inode, fe, orphan_name,
+					  orphan_entry_bh);
+		if (status < 0) {
+			mlog_errno(status);
+			goto leave;
+		}
+	}
+
+	/* delete the name from the parent dir */
+	status = ocfs2_delete_entry(handle, dir, dirent, dirent_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	/* We can set nlink on the dinode now. clear the saved version
+	 * so that it doesn't get set later. */
+	fe->i_links_count = cpu_to_le16(inode->i_nlink);
+	saved_nlink = 0;
+
+	status = ocfs2_journal_dirty(handle, fe_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	if (S_ISDIR(inode->i_mode)) {
+		dir->i_nlink--;
+		status = ocfs2_mark_inode_dirty(handle, dir,
+						parent_node_bh);
+		if (status < 0) {
+			mlog_errno(status);
+			dir->i_nlink++;
+		}
+	}
+
+leave:
+	if (status < 0 && saved_nlink)
+		inode->i_nlink = saved_nlink;
+
+	if (handle)
+		ocfs2_commit_trans(handle);
+
+	if (fe_bh)
+		brelse(fe_bh);
+
+	if (dirent_bh)
+		brelse(dirent_bh);
+
+	if (parent_node_bh)
+		brelse(parent_node_bh);
+
+	if (orphan_entry_bh)
+		brelse(orphan_entry_bh);
+
+	mlog_exit(status);
+
+	return status;
+}
+
+/*
+ * The only place this should be used is rename!
+ * if they have the same id, then the 1st one is the only one locked.
+ */
+static int ocfs2_double_lock(struct ocfs2_super *osb,
+			     struct ocfs2_journal_handle *handle,
+			     struct buffer_head **bh1,
+			     struct inode *inode1,
+			     struct buffer_head **bh2,
+			     struct inode *inode2)
+{
+	int status;
+	struct ocfs2_inode_info *oi1 = OCFS2_I(inode1);
+	struct ocfs2_inode_info *oi2 = OCFS2_I(inode2);
+	struct buffer_head **tmpbh;
+	struct inode *tmpinode;
+
+	mlog_entry("(inode1 = %"MLFu64", inode2 = %"MLFu64")\n",
+		   oi1->ip_blkno, oi2->ip_blkno);
+
+	BUG_ON(!handle);
+
+	if (*bh1)
+		*bh1 = NULL;
+	if (*bh2)
+		*bh2 = NULL;
+
+	/* we always want to lock the one with the lower lockid first. */
+	if (oi1->ip_blkno != oi2->ip_blkno) {
+		if (oi1->ip_blkno < oi2->ip_blkno) {
+			/* switch id1 and id2 around */
+			mlog(0, "switching them around...\n");
+			tmpbh = bh2;
+			bh2 = bh1;
+			bh1 = tmpbh;
+
+			tmpinode = inode2;
+			inode2 = inode1;
+			inode1 = tmpinode;
+		}
+		/* lock id2 */
+		status = ocfs2_meta_lock(inode2, handle, bh2, 1);
+		if (status < 0) {
+			if (status != -ENOENT)
+				mlog_errno(status);
+			goto bail;
+		}
+	}
+	/* lock id1 */
+	status = ocfs2_meta_lock(inode1, handle, bh1, 1);
+	if (status < 0) {
+		if (status != -ENOENT)
+			mlog_errno(status);
+		goto bail;
+	}
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+#define PARENT_INO(buffer) \
+	((struct ocfs2_dir_entry *) \
+	 ((char *)buffer + \
+	  le16_to_cpu(((struct ocfs2_dir_entry *)buffer)->rec_len)))->inode
+
+static int ocfs2_rename(struct inode *old_dir,
+			struct dentry *old_dentry,
+			struct inode *new_dir,
+			struct dentry *new_dentry)
+{
+	int status = 0, rename_lock = 0;
+	struct inode *old_inode = old_dentry->d_inode;
+	struct inode *new_inode = new_dentry->d_inode;
+	struct ocfs2_dinode *newfe = NULL;
+	char orphan_name[OCFS2_ORPHAN_NAMELEN + 1];
+	struct buffer_head *orphan_entry_bh = NULL;
+	struct buffer_head *newfe_bh = NULL;
+	struct buffer_head *insert_entry_bh = NULL;
+	struct ocfs2_super *osb = NULL;
+	u64 newfe_blkno;
+	struct ocfs2_journal_handle *handle = NULL;
+	struct buffer_head *old_dir_bh = NULL;
+	struct buffer_head *new_dir_bh = NULL;
+	struct ocfs2_dir_entry *old_de = NULL, *new_de = NULL; // dirent for old_dentry
+							       // and new_dentry
+	struct buffer_head *new_de_bh = NULL, *old_de_bh = NULL; // bhs for above
+	struct buffer_head *old_inode_de_bh = NULL; // if old_dentry is a dir,
+						    // this is the 1st dirent bh
+	nlink_t old_dir_nlink = old_dir->i_nlink, new_dir_nlink = new_dir->i_nlink;
+	unsigned int links_count;
+
+	/* At some point it might be nice to break this function up a
+	 * bit. */
+
+	mlog_entry("(0x%p, 0x%p, 0x%p, 0x%p, from='%.*s' to='%.*s')\n",
+		   old_dir, old_dentry, new_dir, new_dentry,
+		   old_dentry->d_name.len, old_dentry->d_name.name,
+		   new_dentry->d_name.len, new_dentry->d_name.name);
+
+	osb = OCFS2_SB(old_dir->i_sb);
+
+	if (new_inode) {
+		if (!igrab(new_inode))
+			BUG();
+	}
+
+	if (atomic_read(&old_dentry->d_count) > 2) {
+		shrink_dcache_parent(old_dentry);
+		if (atomic_read(&old_dentry->d_count) > 2) {
+			status = -EBUSY;
+			goto bail;
+		}
+	}
+
+	/* Assume a directory heirarchy thusly:
+	 * a/b/c
+	 * a/d
+	 * a,b,c, and d are all directories.
+	 *
+	 * from cwd of 'a' on both nodes:
+	 * node1: mv b/c d
+	 * node2: mv d   b/c
+	 *
+	 * And that's why, just like the VFS, we need a file system
+	 * rename lock. */
+	if (old_dentry != new_dentry) {
+		status = ocfs2_rename_lock(osb);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+		rename_lock = 1;
+	}
+
+	handle = ocfs2_alloc_handle(osb);
+	if (handle == NULL) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* if old and new are the same, this'll just do one lock. */
+	status = ocfs2_double_lock(osb, handle,
+				  &old_dir_bh, old_dir,
+				  &new_dir_bh, new_dir);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* make sure both dirs have bhs
+	 * get an extra ref on old_dir_bh if old==new */
+	if (!new_dir_bh) {
+		if (old_dir_bh) {
+			new_dir_bh = old_dir_bh;
+			get_bh(new_dir_bh);
+		} else {
+			mlog(ML_ERROR, "no old_dir_bh!\n");
+			status = -EIO;
+			goto bail;
+		}
+	}
+
+	if (S_ISDIR(old_inode->i_mode)) {
+		/* Directories actually require metadata updates to
+		 * the directory info so we can't get away with not
+		 * doing node locking on it. */
+		status = ocfs2_meta_lock(old_inode, handle, NULL, 1);
+		if (status < 0) {
+			if (status != -ENOENT)
+				mlog_errno(status);
+			goto bail;
+		}
+
+		status = ocfs2_request_rename_vote(old_inode, old_dentry);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		status = -EIO;
+		old_inode_de_bh = ocfs2_bread(old_inode, 0, &status, 0);
+		if (!old_inode_de_bh)
+			goto bail;
+
+		status = -EIO;
+		if (le64_to_cpu(PARENT_INO(old_inode_de_bh->b_data)) !=
+		    OCFS2_I(old_dir)->ip_blkno)
+			goto bail;
+		status = -EMLINK;
+		if (!new_inode && new_dir!=old_dir &&
+		    new_dir->i_nlink >= OCFS2_LINK_MAX)
+			goto bail;
+	} else {
+		/* Ah, the simple case - we're a file so just send a
+		 * message. */
+		status = ocfs2_request_rename_vote(old_inode, old_dentry);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+	}
+
+	status = -ENOENT;
+	old_de_bh = ocfs2_find_entry(old_dentry->d_name.name,
+				     old_dentry->d_name.len,
+				     old_dir, &old_de);
+	if (!old_de_bh)
+		goto bail;
+
+	/*
+	 *  Check for inode number is _not_ due to possible IO errors.
+	 *  We might rmdir the source, keep it as pwd of some process
+	 *  and merrily kill the link to whatever was created under the
+	 *  same name. Goodbye sticky bit ;-<
+	 */
+	if (le64_to_cpu(old_de->inode) != OCFS2_I(old_inode)->ip_blkno)
+		goto bail;
+
+	/* check if the target already exists (in which case we need
+	 * to delete it */
+	status = ocfs2_find_files_on_disk(new_dentry->d_name.name,
+					  new_dentry->d_name.len,
+					  &newfe_blkno, new_dir, &new_de_bh,
+					  &new_de);
+	/* The only error we allow here is -ENOENT because the new
+	 * file not existing is perfectly valid. */
+	if ((status < 0) && (status != -ENOENT)) {
+		/* If we cannot find the file specified we should just */
+		/* return the error... */
+		mlog_errno(status);
+		goto bail;
+	}
+
+	if (!new_de && new_inode)
+		mlog(ML_ERROR, "inode %lu does not exist in it's parent "
+		     "directory!", new_inode->i_ino);
+
+	/* In case we need to overwrite an existing file, we blow it
+	 * away first */
+	if (new_de) {
+		/* VFS didn't think there existed an inode here, but
+		 * someone else in the cluster must have raced our
+		 * rename to create one. Today we error cleanly, in
+		 * the future we should consider calling iget to build
+		 * a new struct inode for this entry. */
+		if (!new_inode) {
+			status = -EACCES;
+
+			mlog(0, "We found an inode for name %.*s but VFS "
+			     "didn't give us one.\n", new_dentry->d_name.len,
+			     new_dentry->d_name.name);
+			goto bail;
+		}
+
+		if (OCFS2_I(new_inode)->ip_blkno != newfe_blkno) {
+			status = -EACCES;
+
+			mlog(0, "Inode blkno (%"MLFu64") and dir (%"MLFu64") "
+			     "disagree. ip_flags = %x\n",
+			     OCFS2_I(new_inode)->ip_blkno, newfe_blkno,
+			     OCFS2_I(new_inode)->ip_flags);
+			goto bail;
+		}
+
+		status = ocfs2_meta_lock(new_inode, handle, &newfe_bh, 1);
+		if (status < 0) {
+			if (status != -ENOENT)
+				mlog_errno(status);
+			goto bail;
+		}
+
+		if (S_ISDIR(new_inode->i_mode))
+			links_count = 0;
+		else
+			links_count = (unsigned int) (new_inode->i_nlink - 1);
+
+		status = ocfs2_request_unlink_vote(new_inode, new_dentry,
+						   links_count);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		newfe = (struct ocfs2_dinode *) newfe_bh->b_data;
+
+		mlog(0, "aha rename over existing... new_de=%p "
+		     "new_blkno=%"MLFu64" newfebh=%p bhblocknr=%llu\n",
+		     new_de, newfe_blkno, newfe_bh, newfe_bh ?
+		     (unsigned long long)newfe_bh->b_blocknr : 0ULL);
+
+		if (S_ISDIR(new_inode->i_mode) || (new_inode->i_nlink == 1)) {
+			status = ocfs2_prepare_orphan_dir(osb, handle,
+							  new_inode,
+							  orphan_name,
+							  &orphan_entry_bh);
+			if (status < 0) {
+				mlog_errno(status);
+				goto bail;
+			}
+		}
+	} else {
+		BUG_ON(new_dentry->d_parent->d_inode != new_dir);
+
+		status = ocfs2_check_dir_for_entry(new_dir,
+						   new_dentry->d_name.name,
+						   new_dentry->d_name.len);
+		if (status)
+			goto bail;
+
+		status = ocfs2_prepare_dir_for_insert(osb, new_dir, new_dir_bh,
+						      new_dentry->d_name.name,
+						      new_dentry->d_name.len,
+						      &insert_entry_bh);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+	}
+
+	handle = ocfs2_start_trans(osb, handle, OCFS2_RENAME_CREDITS);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		handle = NULL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	if (new_de) {
+		if (S_ISDIR(new_inode->i_mode)) {
+			if (!ocfs2_empty_dir(new_inode) ||
+			    new_inode->i_nlink != 2) {
+				status = -ENOTEMPTY;
+				goto bail;
+			}
+		}
+		status = ocfs2_journal_access(handle, new_inode, newfe_bh,
+					      OCFS2_JOURNAL_ACCESS_WRITE);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		if (S_ISDIR(new_inode->i_mode) ||
+		    (newfe->i_links_count == cpu_to_le16(1))){
+			status = ocfs2_orphan_add(osb, handle, new_inode,
+						  newfe, orphan_name,
+						  orphan_entry_bh);
+			if (status < 0) {
+				mlog_errno(status);
+				goto bail;
+			}
+		}
+
+		/* change the dirent to point to the correct inode */
+		status = ocfs2_journal_access(handle, new_dir, new_de_bh,
+					      OCFS2_JOURNAL_ACCESS_WRITE);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+		new_de->inode = cpu_to_le64(OCFS2_I(old_inode)->ip_blkno);
+		new_de->file_type = old_de->file_type;
+		new_dir->i_version++;
+		status = ocfs2_journal_dirty(handle, new_de_bh);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		if (S_ISDIR(new_inode->i_mode))
+			newfe->i_links_count = 0;
+		else
+			le16_add_cpu(&newfe->i_links_count, -1);
+
+		status = ocfs2_journal_dirty(handle, newfe_bh);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+	} else {
+		/* if the name was not found in new_dir, add it now */
+		status = ocfs2_add_entry(handle, new_dentry, old_inode,
+					 OCFS2_I(old_inode)->ip_blkno,
+					 new_dir_bh, insert_entry_bh);
+	}
+
+	old_inode->i_ctime = CURRENT_TIME;
+	mark_inode_dirty(old_inode);
+
+	/* now that the name has been added to new_dir, remove the old name */
+	status = ocfs2_delete_entry(handle, old_dir, old_de, old_de_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	if (new_inode) {
+		new_inode->i_nlink--;
+		new_inode->i_ctime = CURRENT_TIME;
+	}
+	old_dir->i_ctime = old_dir->i_mtime = CURRENT_TIME;
+	if (old_inode_de_bh) {
+		status = ocfs2_journal_access(handle, old_inode,
+					     old_inode_de_bh,
+					     OCFS2_JOURNAL_ACCESS_WRITE);
+		PARENT_INO(old_inode_de_bh->b_data) =
+			cpu_to_le64(OCFS2_I(new_dir)->ip_blkno);
+		status = ocfs2_journal_dirty(handle, old_inode_de_bh);
+		old_dir->i_nlink--;
+		if (new_inode) {
+			new_inode->i_nlink--;
+		} else {
+			new_dir->i_nlink++;
+			mark_inode_dirty(new_dir);
+		}
+	}
+	mark_inode_dirty(old_dir);
+	if (new_inode)
+		mark_inode_dirty(new_inode);
+
+	if (old_dir != new_dir)
+		if (new_dir_nlink != new_dir->i_nlink) {
+			if (!new_dir_bh) {
+				mlog(ML_ERROR, "need to change nlink for new "
+				     "dir %"MLFu64" from %d to %d but bh is "
+				     "NULL\n", OCFS2_I(new_dir)->ip_blkno,
+				     (int)new_dir_nlink, new_dir->i_nlink);
+			} else {
+				struct ocfs2_dinode *fe;
+				status = ocfs2_journal_access(handle,
+							      new_dir,
+							      new_dir_bh,
+							      OCFS2_JOURNAL_ACCESS_WRITE);
+				fe = (struct ocfs2_dinode *) new_dir_bh->b_data;
+				fe->i_links_count = cpu_to_le16(new_dir->i_nlink);
+				status = ocfs2_journal_dirty(handle, new_dir_bh);
+			}
+		}
+
+	if (old_dir_nlink != old_dir->i_nlink) {
+		if (!old_dir_bh) {
+			mlog(ML_ERROR, "need to change nlink for old dir "
+			     "%"MLFu64" from %d to %d but bh is NULL!\n",
+			     OCFS2_I(old_dir)->ip_blkno,
+			     (int)old_dir_nlink,
+			     old_dir->i_nlink);
+		} else {
+			struct ocfs2_dinode *fe;
+			status = ocfs2_journal_access(handle, old_dir,
+						      old_dir_bh,
+						      OCFS2_JOURNAL_ACCESS_WRITE);
+			fe = (struct ocfs2_dinode *) old_dir_bh->b_data;
+			fe->i_links_count = cpu_to_le16(old_dir->i_nlink);
+			status = ocfs2_journal_dirty(handle, old_dir_bh);
+		}
+	}
+
+	status = 0;
+bail:
+	if (rename_lock)
+		ocfs2_rename_unlock(osb);
+
+	if (handle)
+		ocfs2_commit_trans(handle);
+
+	if (new_inode)
+		sync_mapping_buffers(old_inode->i_mapping);
+
+	if (new_inode)
+		iput(new_inode);
+	if (newfe_bh)
+		brelse(newfe_bh);
+	if (old_dir_bh)
+		brelse(old_dir_bh);
+	if (new_dir_bh)
+		brelse(new_dir_bh);
+	if (new_de_bh)
+		brelse(new_de_bh);
+	if (old_de_bh)
+		brelse(old_de_bh);
+	if (old_inode_de_bh)
+		brelse(old_inode_de_bh);
+	if (orphan_entry_bh)
+		brelse(orphan_entry_bh);
+	if (insert_entry_bh)
+		brelse(insert_entry_bh);
+
+	mlog_exit(status);
+
+	return status;
+}
+
+/*
+ * we expect i_size = strlen(symname). Copy symname into the file
+ * data, including the null terminator.
+ */
+static int ocfs2_create_symlink_data(struct ocfs2_super *osb,
+				     struct ocfs2_journal_handle *handle,
+				     struct inode *inode,
+				     const char *symname)
+{
+	struct buffer_head **bhs = NULL;
+	const char *c;
+	struct super_block *sb = osb->sb;
+	u64 p_blkno;
+	int p_blocks;
+	int virtual, blocks, status, i, bytes_left;
+
+	bytes_left = i_size_read(inode) + 1;
+	/* we can't trust i_blocks because we're actually going to
+	 * write i_size + 1 bytes. */
+	blocks = (bytes_left + sb->s_blocksize - 1) >> sb->s_blocksize_bits;
+
+	mlog_entry("i_blocks = %lu, i_size = %llu, blocks = %d\n",
+		       inode->i_blocks, i_size_read(inode), blocks);
+
+	/* Sanity check -- make sure we're going to fit. */
+	if (bytes_left >
+	    ocfs2_clusters_to_bytes(sb, OCFS2_I(inode)->ip_clusters)) {
+		status = -EIO;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	bhs = kcalloc(blocks, sizeof(struct buffer_head *), GFP_KERNEL);
+	if (!bhs) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_extent_map_get_blocks(inode, 0, 1, &p_blkno,
+					     &p_blocks);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* links can never be larger than one cluster so we know this
+	 * is all going to be contiguous, but do a sanity check
+	 * anyway. */
+	if ((p_blocks << sb->s_blocksize_bits) < bytes_left) {
+		status = -EIO;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	virtual = 0;
+	while(bytes_left > 0) {
+		c = &symname[virtual * sb->s_blocksize];
+
+		bhs[virtual] = sb_getblk(sb, p_blkno);
+		if (!bhs[virtual]) {
+			status = -ENOMEM;
+			mlog_errno(status);
+			goto bail;
+		}
+		ocfs2_set_new_buffer_uptodate(inode, bhs[virtual]);
+
+		status = ocfs2_journal_access(handle, inode, bhs[virtual],
+					      OCFS2_JOURNAL_ACCESS_CREATE);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		memset(bhs[virtual]->b_data, 0, sb->s_blocksize);
+
+		memcpy(bhs[virtual]->b_data, c,
+		       (bytes_left > sb->s_blocksize) ? sb->s_blocksize :
+		       bytes_left);
+
+		status = ocfs2_journal_dirty(handle, bhs[virtual]);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+
+		virtual++;
+		p_blkno++;
+		bytes_left -= sb->s_blocksize;
+	}
+
+	status = 0;
+bail:
+
+	if (bhs) {
+		for(i = 0; i < blocks; i++)
+			if (bhs[i])
+				brelse(bhs[i]);
+		kfree(bhs);
+	}
+
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_symlink(struct inode *dir,
+			 struct dentry *dentry,
+			 const char *symname)
+{
+	int status, l, credits;
+	u64 newsize;
+	struct ocfs2_super *osb = NULL;
+	struct inode *inode = NULL;
+	struct super_block *sb;
+	struct buffer_head *new_fe_bh = NULL;
+	struct buffer_head *de_bh = NULL;
+	struct buffer_head *parent_fe_bh = NULL;
+	struct ocfs2_dinode *fe = NULL;
+	struct ocfs2_dinode *dirfe;
+	struct ocfs2_journal_handle *handle = NULL;
+	struct ocfs2_alloc_context *inode_ac = NULL;
+	struct ocfs2_alloc_context *data_ac = NULL;
+
+	mlog_entry("(0x%p, 0x%p, symname='%s' actual='%.*s')\n", dir,
+		   dentry, symname, dentry->d_name.len, dentry->d_name.name);
+
+	sb = dir->i_sb;
+	osb = OCFS2_SB(sb);
+
+	l = strlen(symname) + 1;
+
+	credits = ocfs2_calc_symlink_credits(sb);
+
+	handle = ocfs2_alloc_handle(osb);
+	if (handle == NULL) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/* lock the parent directory */
+	status = ocfs2_meta_lock(dir, handle, &parent_fe_bh, 1);
+	if (status < 0) {
+		if (status != -ENOENT)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	dirfe = (struct ocfs2_dinode *) parent_fe_bh->b_data;
+	if (!dirfe->i_links_count) {
+		/* can't make a file in a deleted directory. */
+		status = -ENOENT;
+		goto bail;
+	}
+
+	status = ocfs2_check_dir_for_entry(dir, dentry->d_name.name,
+					   dentry->d_name.len);
+	if (status)
+		goto bail;
+
+	status = ocfs2_prepare_dir_for_insert(osb, dir, parent_fe_bh,
+					      dentry->d_name.name,
+					      dentry->d_name.len, &de_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_reserve_new_inode(osb, handle, &inode_ac);
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	/* don't reserve bitmap space for fast symlinks. */
+	if (l > ocfs2_fast_symlink_chars(sb)) {
+		status = ocfs2_reserve_clusters(osb, handle, 1, &data_ac);
+		if (status < 0) {
+			if (status != -ENOSPC)
+				mlog_errno(status);
+			goto bail;
+		}
+	}
+
+	handle = ocfs2_start_trans(osb, handle, credits);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		handle = NULL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_mknod_locked(osb, dir, dentry,
+				    S_IFLNK | S_IRWXUGO, 0,
+				    &new_fe_bh, parent_fe_bh, handle,
+				    &inode, inode_ac);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	fe = (struct ocfs2_dinode *) new_fe_bh->b_data;
+	inode->i_rdev = 0;
+	newsize = l - 1;
+	if (l > ocfs2_fast_symlink_chars(sb)) {
+		inode->i_op = &ocfs2_symlink_inode_operations;
+		status = ocfs2_do_extend_allocation(osb, inode, 1, new_fe_bh,
+						    handle, data_ac, NULL,
+						    NULL);
+		if (status < 0) {
+			if (status != -ENOSPC && status != -EINTR) {
+				mlog(ML_ERROR, "Failed to extend file to "
+					       "%"MLFu64"\n",
+				     newsize);
+				mlog_errno(status);
+				status = -ENOSPC;
+			}
+			goto bail;
+		}
+		i_size_write(inode, newsize);
+		inode->i_blocks = ocfs2_align_bytes_to_sectors(newsize);
+	} else {
+		inode->i_op = &ocfs2_fast_symlink_inode_operations;
+		memcpy((char *) fe->id2.i_symlink, symname, l);
+		i_size_write(inode, newsize);
+		inode->i_blocks = 0;
+	}
+
+	status = ocfs2_mark_inode_dirty(handle, inode, new_fe_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	if (!ocfs2_inode_is_fast_symlink(inode)) {
+		status = ocfs2_create_symlink_data(osb, handle, inode,
+						   symname);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+	}
+
+	status = ocfs2_add_entry(handle, dentry, inode,
+				 le64_to_cpu(fe->i_blkno), parent_fe_bh,
+				 de_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	insert_inode_hash(inode);
+	dentry->d_op = &ocfs2_dentry_ops;
+	d_instantiate(dentry, inode);
+bail:
+	if (handle)
+		ocfs2_commit_trans(handle);
+	if (new_fe_bh)
+		brelse(new_fe_bh);
+	if (parent_fe_bh)
+		brelse(parent_fe_bh);
+	if (de_bh)
+		brelse(de_bh);
+	if (inode_ac)
+		ocfs2_free_alloc_context(inode_ac);
+	if (data_ac)
+		ocfs2_free_alloc_context(data_ac);
+	if ((status < 0) && inode)
+		iput(inode);
+
+	mlog_exit(status);
+
+	return status;
+}
+
+int ocfs2_check_dir_entry(struct inode * dir,
+			  struct ocfs2_dir_entry * de,
+			  struct buffer_head * bh,
+			  unsigned long offset)
+{
+	const char *error_msg = NULL;
+	const int rlen = le16_to_cpu(de->rec_len);
+
+	if (rlen < OCFS2_DIR_REC_LEN(1))
+		error_msg = "rec_len is smaller than minimal";
+	else if (rlen % 4 != 0)
+		error_msg = "rec_len % 4 != 0";
+	else if (rlen < OCFS2_DIR_REC_LEN(de->name_len))
+		error_msg = "rec_len is too small for name_len";
+	else if (((char *) de - bh->b_data) + rlen > dir->i_sb->s_blocksize)
+		error_msg = "directory entry across blocks";
+
+	if (error_msg != NULL)
+		mlog(ML_ERROR, "bad entry in directory #%"MLFu64": %s - "
+		     "offset=%lu, inode=%"MLFu64", rec_len=%d, name_len=%d\n",
+		     OCFS2_I(dir)->ip_blkno, error_msg, offset,
+		     le64_to_cpu(de->inode), rlen, de->name_len);
+	return error_msg == NULL ? 1 : 0;
+}
+
+/* we don't always have a dentry for what we want to add, so people
+ * like orphan dir can call this instead.
+ *
+ * If you pass me insert_bh, I'll skip the search of the other dir
+ * blocks and put the record in there.
+ */
+static int __ocfs2_add_entry(struct ocfs2_journal_handle *handle,
+			     struct inode *dir,
+			     const char *name, int namelen,
+			     struct inode *inode, u64 blkno,
+			     struct buffer_head *parent_fe_bh,
+			     struct buffer_head *insert_bh)
+{
+	unsigned long offset;
+	unsigned short rec_len;
+	struct ocfs2_dir_entry *de, *de1;
+	struct super_block *sb;
+	int retval, status;
+
+	mlog_entry_void();
+
+	sb = dir->i_sb;
+
+	if (!namelen)
+		return -EINVAL;
+
+	rec_len = OCFS2_DIR_REC_LEN(namelen);
+	offset = 0;
+	de = (struct ocfs2_dir_entry *) insert_bh->b_data;
+	while (1) {
+		BUG_ON((char *)de >= sb->s_blocksize + insert_bh->b_data);
+		/* These checks should've already been passed by the
+		 * prepare function, but I guess we can leave them
+		 * here anyway. */
+		if (!ocfs2_check_dir_entry(dir, de, insert_bh, offset)) {
+			retval = -ENOENT;
+			goto bail;
+		}
+		if (ocfs2_match(namelen, name, de)) {
+			retval = -EEXIST;
+			goto bail;
+		}
+		if (((le64_to_cpu(de->inode) == 0) &&
+		     (le16_to_cpu(de->rec_len) >= rec_len)) ||
+		    (le16_to_cpu(de->rec_len) >=
+		     (OCFS2_DIR_REC_LEN(de->name_len) + rec_len))) {
+			status = ocfs2_journal_access(handle, dir, insert_bh,
+						      OCFS2_JOURNAL_ACCESS_WRITE);
+			/* By now the buffer is marked for journaling */
+			offset += le16_to_cpu(de->rec_len);
+			if (le64_to_cpu(de->inode)) {
+				de1 = (struct ocfs2_dir_entry *)((char *) de +
+					OCFS2_DIR_REC_LEN(de->name_len));
+				de1->rec_len =
+					cpu_to_le16(le16_to_cpu(de->rec_len) -
+					OCFS2_DIR_REC_LEN(de->name_len));
+				de->rec_len = cpu_to_le16(OCFS2_DIR_REC_LEN(de->name_len));
+				de = de1;
+			}
+			de->file_type = OCFS2_FT_UNKNOWN;
+			if (blkno) {
+				de->inode = cpu_to_le64(blkno);
+				ocfs2_set_de_type(de, inode->i_mode);
+			} else
+				de->inode = 0;
+			de->name_len = namelen;
+			memcpy(de->name, name, namelen);
+
+			dir->i_mtime = dir->i_ctime = CURRENT_TIME;
+			dir->i_version++;
+			status = ocfs2_journal_dirty(handle, insert_bh);
+			retval = 0;
+			goto bail;
+		}
+		offset += le16_to_cpu(de->rec_len);
+		de = (struct ocfs2_dir_entry *) ((char *) de + le16_to_cpu(de->rec_len));
+	}
+
+	/* when you think about it, the assert above should prevent us
+	 * from ever getting here. */
+	retval = -ENOSPC;
+bail:
+
+	mlog_exit(retval);
+	return retval;
+}
+
+
+/*
+ * ocfs2_delete_entry deletes a directory entry by merging it with the
+ * previous entry
+ */
+static int ocfs2_delete_entry(struct ocfs2_journal_handle *handle,
+			      struct inode *dir,
+			      struct ocfs2_dir_entry *de_del,
+			      struct buffer_head *bh)
+{
+	struct ocfs2_dir_entry *de, *pde;
+	int i, status = -ENOENT;
+
+	mlog_entry("(0x%p, 0x%p, 0x%p, 0x%p)\n", handle, dir, de_del, bh);
+
+	i = 0;
+	pde = NULL;
+	de = (struct ocfs2_dir_entry *) bh->b_data;
+	while (i < bh->b_size) {
+		if (!ocfs2_check_dir_entry(dir, de, bh, i)) {
+			status = -EIO;
+			mlog_errno(status);
+			goto bail;
+		}
+		if (de == de_del)  {
+			status = ocfs2_journal_access(handle, dir, bh,
+						      OCFS2_JOURNAL_ACCESS_WRITE);
+			if (status < 0) {
+				status = -EIO;
+				mlog_errno(status);
+				goto bail;
+			}
+			if (pde)
+				pde->rec_len =
+					cpu_to_le16(le16_to_cpu(pde->rec_len) +
+						    le16_to_cpu(de->rec_len));
+			else
+				de->inode = 0;
+			dir->i_version++;
+			status = ocfs2_journal_dirty(handle, bh);
+			goto bail;
+		}
+		i += le16_to_cpu(de->rec_len);
+		pde = de;
+		de = (struct ocfs2_dir_entry *)((char *)de + le16_to_cpu(de->rec_len));
+	}
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * Returns 0 if not found, -1 on failure, and 1 on success
+ */
+static int inline ocfs2_search_dirblock(struct buffer_head *bh,
+					struct inode *dir,
+					const char *name, int namelen,
+					unsigned long offset,
+					struct ocfs2_dir_entry **res_dir)
+{
+	struct ocfs2_dir_entry *de;
+	char *dlimit, *de_buf;
+	int de_len;
+	int ret = 0;
+
+	mlog_entry_void();
+
+	de_buf = bh->b_data;
+	dlimit = de_buf + dir->i_sb->s_blocksize;
+
+	while (de_buf < dlimit) {
+		/* this code is executed quadratically often */
+		/* do minimal checking `by hand' */
+
+		de = (struct ocfs2_dir_entry *) de_buf;
+
+		if (de_buf + namelen <= dlimit &&
+		    ocfs2_match(namelen, name, de)) {
+			/* found a match - just to be sure, do a full check */
+			if (!ocfs2_check_dir_entry(dir, de, bh, offset)) {
+				ret = -1;
+				goto bail;
+			}
+			*res_dir = de;
+			ret = 1;
+			goto bail;
+		}
+
+		/* prevent looping on a bad block */
+		de_len = le16_to_cpu(de->rec_len);
+		if (de_len <= 0) {
+			ret = -1;
+			goto bail;
+		}
+
+		de_buf += de_len;
+		offset += de_len;
+	}
+
+bail:
+	mlog_exit(ret);
+	return ret;
+}
+
+struct buffer_head *ocfs2_find_entry(const char *name, int namelen,
+				     struct inode *dir,
+				     struct ocfs2_dir_entry **res_dir)
+{
+	struct super_block *sb;
+	struct buffer_head *bh_use[NAMEI_RA_SIZE];
+	struct buffer_head *bh, *ret = NULL;
+	unsigned long start, block, b;
+	int ra_max = 0;		/* Number of bh's in the readahead
+				   buffer, bh_use[] */
+	int ra_ptr = 0;		/* Current index into readahead
+				   buffer */
+	int num = 0;
+	int nblocks, i, err;
+
+	mlog_entry_void();
+
+	*res_dir = NULL;
+	sb = dir->i_sb;
+
+	nblocks = i_size_read(dir) >> sb->s_blocksize_bits;
+	start = OCFS2_I(dir)->ip_dir_start_lookup;
+	if (start >= nblocks)
+		start = 0;
+	block = start;
+
+restart:
+	do {
+		/*
+		 * We deal with the read-ahead logic here.
+		 */
+		if (ra_ptr >= ra_max) {
+			/* Refill the readahead buffer */
+			ra_ptr = 0;
+			b = block;
+			for (ra_max = 0; ra_max < NAMEI_RA_SIZE; ra_max++) {
+				/*
+				 * Terminate if we reach the end of the
+				 * directory and must wrap, or if our
+				 * search has finished at this block.
+				 */
+				if (b >= nblocks || (num && block == start)) {
+					bh_use[ra_max] = NULL;
+					break;
+				}
+				num++;
+
+				/* XXX: questionable readahead stuff here */
+				bh = ocfs2_bread(dir, b++, &err, 1);
+				bh_use[ra_max] = bh;
+#if 0		// ???
+				if (bh)
+					ll_rw_block(READ, 1, &bh);
+#endif
+			}
+		}
+		if ((bh = bh_use[ra_ptr++]) == NULL)
+			goto next;
+		wait_on_buffer(bh);
+		if (!buffer_uptodate(bh)) {
+			/* read error, skip block & hope for the best */
+			brelse(bh);
+			goto next;
+		}
+		i = ocfs2_search_dirblock(bh, dir, name, namelen,
+					  block << sb->s_blocksize_bits,
+					  res_dir);
+		if (i == 1) {
+			OCFS2_I(dir)->ip_dir_start_lookup = block;
+			ret = bh;
+			goto cleanup_and_exit;
+		} else {
+			brelse(bh);
+			if (i < 0)
+				goto cleanup_and_exit;
+		}
+	next:
+		if (++block >= nblocks)
+			block = 0;
+	} while (block != start);
+
+	/*
+	 * If the directory has grown while we were searching, then
+	 * search the last part of the directory before giving up.
+	 */
+	block = nblocks;
+	nblocks = i_size_read(dir) >> sb->s_blocksize_bits;
+	if (block < nblocks) {
+		start = 0;
+		goto restart;
+	}
+
+cleanup_and_exit:
+	/* Clean up the read-ahead blocks */
+	for (; ra_ptr < ra_max; ra_ptr++)
+		brelse(bh_use[ra_ptr]);
+
+	mlog_exit_ptr(ret);
+	return ret;
+}
+
+static int ocfs2_blkno_stringify(u64 blkno, char *name)
+{
+	int status, namelen;
+
+	mlog_entry_void();
+
+	namelen = snprintf(name, OCFS2_ORPHAN_NAMELEN + 1, "%016"MLFx64,
+			   blkno);
+	if (namelen <= 0) {
+		if (namelen)
+			status = namelen;
+		else
+			status = -EINVAL;
+		mlog_errno(status);
+		goto bail;
+	}
+	if (namelen != OCFS2_ORPHAN_NAMELEN) {
+		status = -EINVAL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	mlog(0, "built filename '%s' for orphan dir (len=%d)\n", name,
+	     namelen);
+
+	status = 0;
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_prepare_orphan_dir(struct ocfs2_super *osb,
+				    struct ocfs2_journal_handle *handle,
+				    struct inode *inode,
+				    char *name,
+				    struct buffer_head **de_bh)
+{
+	struct inode *orphan_dir_inode = NULL;
+	struct buffer_head *orphan_dir_bh = NULL;
+	int status = 0;
+
+	status = ocfs2_blkno_stringify(OCFS2_I(inode)->ip_blkno, name);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	orphan_dir_inode = ocfs2_get_system_file_inode(osb,
+						       ORPHAN_DIR_SYSTEM_INODE,
+						       osb->slot_num);
+	if (!orphan_dir_inode) {
+		status = -ENOENT;
+		mlog_errno(status);
+		goto leave;
+	}
+
+	ocfs2_handle_add_inode(handle, orphan_dir_inode);
+	status = ocfs2_meta_lock(orphan_dir_inode, handle, &orphan_dir_bh, 1);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	status = ocfs2_prepare_dir_for_insert(osb, orphan_dir_inode,
+					      orphan_dir_bh, name,
+					      OCFS2_ORPHAN_NAMELEN, de_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+leave:
+	if (orphan_dir_inode)
+		iput(orphan_dir_inode);
+
+	if (orphan_dir_bh)
+		brelse(orphan_dir_bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_orphan_add(struct ocfs2_super *osb,
+			    struct ocfs2_journal_handle *handle,
+			    struct inode *inode,
+			    struct ocfs2_dinode *fe,
+			    char *name,
+			    struct buffer_head *de_bh)
+{
+	struct inode *orphan_dir_inode = NULL;
+	struct buffer_head *orphan_dir_bh = NULL;
+	int status = 0;
+	struct ocfs2_dinode *orphan_fe;
+
+	mlog_entry("(inode->i_ino = %lu)\n", inode->i_ino);
+
+	orphan_dir_inode = ocfs2_get_system_file_inode(osb,
+						       ORPHAN_DIR_SYSTEM_INODE,
+						       osb->slot_num);
+	if (!orphan_dir_inode) {
+		status = -ENOENT;
+		mlog_errno(status);
+		goto leave;
+	}
+
+	status = ocfs2_read_block(osb,
+				  OCFS2_I(orphan_dir_inode)->ip_blkno,
+				  &orphan_dir_bh, OCFS2_BH_CACHED,
+				  orphan_dir_inode);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	status = ocfs2_journal_access(handle, orphan_dir_inode, orphan_dir_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	/* we're a cluster, and nlink can change on disk from
+	 * underneath us... */
+	orphan_fe = (struct ocfs2_dinode *) orphan_dir_bh->b_data;
+	if (S_ISDIR(inode->i_mode))
+		le16_add_cpu(&orphan_fe->i_links_count, 1);
+	orphan_dir_inode->i_nlink = le16_to_cpu(orphan_fe->i_links_count);
+
+	status = ocfs2_journal_dirty(handle, orphan_dir_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	status = __ocfs2_add_entry(handle, orphan_dir_inode, name,
+				   OCFS2_ORPHAN_NAMELEN, inode,
+				   OCFS2_I(inode)->ip_blkno,
+				   orphan_dir_bh, de_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	le32_add_cpu(&fe->i_flags, OCFS2_ORPHANED_FL);
+
+	/* Record which orphan dir our inode now resides
+	 * in. delete_inode will use this to determine which orphan
+	 * dir to lock. */
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+	OCFS2_I(inode)->ip_orphaned_slot = osb->slot_num;
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+	mlog(0, "Inode %"MLFu64" orphaned in slot %d\n",
+	     OCFS2_I(inode)->ip_blkno, osb->slot_num);
+
+leave:
+	if (orphan_dir_inode)
+		iput(orphan_dir_inode);
+
+	if (orphan_dir_bh)
+		brelse(orphan_dir_bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+/* unlike orphan_add, we expect the orphan dir to already be locked here. */
+int ocfs2_orphan_del(struct ocfs2_super *osb,
+		     struct ocfs2_journal_handle *handle,
+		     struct inode *orphan_dir_inode,
+		     struct inode *inode,
+		     struct buffer_head *orphan_dir_bh)
+{
+	char name[OCFS2_ORPHAN_NAMELEN + 1];
+	struct ocfs2_dinode *orphan_fe;
+	int status = 0;
+	struct buffer_head *target_de_bh = NULL;
+	struct ocfs2_dir_entry *target_de = NULL;
+
+	mlog_entry_void();
+
+	status = ocfs2_blkno_stringify(OCFS2_I(inode)->ip_blkno, name);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	mlog(0, "removing '%s' from orphan dir %"MLFu64" (namelen=%d)\n",
+	     name, OCFS2_I(orphan_dir_inode)->ip_blkno, OCFS2_ORPHAN_NAMELEN);
+
+	/* find it's spot in the orphan directory */
+	target_de_bh = ocfs2_find_entry(name, OCFS2_ORPHAN_NAMELEN,
+					orphan_dir_inode, &target_de);
+	if (!target_de_bh) {
+		status = -ENOENT;
+		mlog_errno(status);
+		goto leave;
+	}
+
+	/* remove it from the orphan directory */
+	status = ocfs2_delete_entry(handle, orphan_dir_inode, target_de,
+				    target_de_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	status = ocfs2_journal_access(handle,orphan_dir_inode,  orphan_dir_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	/* do the i_nlink dance! :) */
+	orphan_fe = (struct ocfs2_dinode *) orphan_dir_bh->b_data;
+	if (S_ISDIR(inode->i_mode))
+		le16_add_cpu(&orphan_fe->i_links_count, -1);
+	orphan_dir_inode->i_nlink = le16_to_cpu(orphan_fe->i_links_count);
+
+	status = ocfs2_journal_dirty(handle, orphan_dir_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+leave:
+	if (target_de_bh)
+		brelse(target_de_bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+struct inode_operations ocfs2_dir_iops = {
+	.create		= ocfs2_create,
+	.lookup		= ocfs2_lookup,
+	.link		= ocfs2_link,
+	.unlink		= ocfs2_unlink,
+	.rmdir		= ocfs2_unlink,
+	.symlink	= ocfs2_symlink,
+	.mkdir		= ocfs2_mkdir,
+	.mknod		= ocfs2_mknod,
+	.rename		= ocfs2_rename,
+	.setattr	= ocfs2_setattr,
+	.getattr	= ocfs2_getattr,
+};
diff --git a/fs/ocfs2/namei.h b/fs/ocfs2/namei.h
new file mode 100644
index 00000000000..deaaa97dbf0
--- /dev/null
+++ b/fs/ocfs2/namei.h
@@ -0,0 +1,58 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * namei.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_NAMEI_H
+#define OCFS2_NAMEI_H
+
+extern struct inode_operations ocfs2_dir_iops;
+
+struct dentry *ocfs2_get_parent(struct dentry *child);
+
+int ocfs2_check_dir_entry (struct inode *dir,
+			   struct ocfs2_dir_entry *de,
+			   struct buffer_head *bh,
+			   unsigned long offset);
+struct buffer_head *ocfs2_find_entry(const char *name,
+				     int namelen,
+				     struct inode *dir,
+				     struct ocfs2_dir_entry **res_dir);
+int ocfs2_orphan_del(struct ocfs2_super *osb,
+		     struct ocfs2_journal_handle *handle,
+		     struct inode *orphan_dir_inode,
+		     struct inode *inode,
+		     struct buffer_head *orphan_dir_bh);
+
+static inline int ocfs2_match(int len,
+			      const char * const name,
+			      struct ocfs2_dir_entry *de)
+{
+	if (len != de->name_len)
+		return 0;
+	if (!de->inode)
+		return 0;
+	return !memcmp(name, de->name, len);
+}
+
+#endif /* OCFS2_NAMEI_H */
diff --git a/fs/ocfs2/ocfs1_fs_compat.h b/fs/ocfs2/ocfs1_fs_compat.h
new file mode 100644
index 00000000000..0b499bccec5
--- /dev/null
+++ b/fs/ocfs2/ocfs1_fs_compat.h
@@ -0,0 +1,109 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs1_fs_compat.h
+ *
+ * OCFS1 volume header definitions.  OCFS2 creates valid but unmountable
+ * OCFS1 volume headers on the first two sectors of an OCFS2 volume.
+ * This allows an OCFS1 volume to see the partition and cleanly fail to
+ * mount it.
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License, version 2,  as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef _OCFS1_FS_COMPAT_H
+#define _OCFS1_FS_COMPAT_H
+
+#define OCFS1_MAX_VOL_SIGNATURE_LEN          128
+#define OCFS1_MAX_MOUNT_POINT_LEN            128
+#define OCFS1_MAX_VOL_ID_LENGTH               16
+#define OCFS1_MAX_VOL_LABEL_LEN               64
+#define OCFS1_MAX_CLUSTER_NAME_LEN            64
+
+#define OCFS1_MAJOR_VERSION              (2)
+#define OCFS1_MINOR_VERSION              (0)
+#define OCFS1_VOLUME_SIGNATURE		 "OracleCFS"
+
+/*
+ * OCFS1 superblock.  Lives at sector 0.
+ */
+struct ocfs1_vol_disk_hdr
+{
+/*00*/	__u32 minor_version;
+	__u32 major_version;
+/*08*/	__u8 signature[OCFS1_MAX_VOL_SIGNATURE_LEN];
+/*88*/	__u8 mount_point[OCFS1_MAX_MOUNT_POINT_LEN];
+/*108*/	__u64 serial_num;
+/*110*/	__u64 device_size;
+	__u64 start_off;
+/*120*/	__u64 bitmap_off;
+	__u64 publ_off;
+/*130*/	__u64 vote_off;
+	__u64 root_bitmap_off;
+/*140*/	__u64 data_start_off;
+	__u64 root_bitmap_size;
+/*150*/	__u64 root_off;
+	__u64 root_size;
+/*160*/	__u64 cluster_size;
+	__u64 num_nodes;
+/*170*/	__u64 num_clusters;
+	__u64 dir_node_size;
+/*180*/	__u64 file_node_size;
+	__u64 internal_off;
+/*190*/	__u64 node_cfg_off;
+	__u64 node_cfg_size;
+/*1A0*/	__u64 new_cfg_off;
+	__u32 prot_bits;
+	__s32 excl_mount;
+/*1B0*/
+};
+
+
+struct ocfs1_disk_lock
+{
+/*00*/	__u32 curr_master;
+	__u8 file_lock;
+	__u8 compat_pad[3];  /* Not in orignal definition.  Used to
+				make the already existing alignment
+				explicit */
+	__u64 last_write_time;
+/*10*/	__u64 last_read_time;
+	__u32 writer_node_num;
+	__u32 reader_node_num;
+/*20*/	__u64 oin_node_map;
+	__u64 dlock_seq_num;
+/*30*/
+};
+
+/*
+ * OCFS1 volume label.  Lives at sector 1.
+ */
+struct ocfs1_vol_label
+{
+/*00*/	struct ocfs1_disk_lock disk_lock;
+/*30*/	__u8 label[OCFS1_MAX_VOL_LABEL_LEN];
+/*70*/	__u16 label_len;
+/*72*/	__u8 vol_id[OCFS1_MAX_VOL_ID_LENGTH];
+/*82*/	__u16 vol_id_len;
+/*84*/	__u8 cluster_name[OCFS1_MAX_CLUSTER_NAME_LEN];
+/*A4*/	__u16 cluster_name_len;
+/*A6*/
+};
+
+
+#endif /* _OCFS1_FS_COMPAT_H */
+
diff --git a/fs/ocfs2/ocfs2.h b/fs/ocfs2/ocfs2.h
new file mode 100644
index 00000000000..f468c600cf9
--- /dev/null
+++ b/fs/ocfs2/ocfs2.h
@@ -0,0 +1,464 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs2.h
+ *
+ * Defines macros and structures used in OCFS2
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_H
+#define OCFS2_H
+
+#include <linux/spinlock.h>
+#include <linux/sched.h>
+#include <linux/wait.h>
+#include <linux/list.h>
+#include <linux/rbtree.h>
+#include <linux/workqueue.h>
+#include <linux/kref.h>
+
+#include "cluster/nodemanager.h"
+#include "cluster/heartbeat.h"
+#include "cluster/tcp.h"
+
+#include "dlm/dlmapi.h"
+
+#include "ocfs2_fs.h"
+#include "endian.h"
+#include "ocfs2_lockid.h"
+
+struct ocfs2_extent_map {
+	u32		em_clusters;
+	struct rb_root	em_extents;
+};
+
+/* Most user visible OCFS2 inodes will have very few pieces of
+ * metadata, but larger files (including bitmaps, etc) must be taken
+ * into account when designing an access scheme. We allow a small
+ * amount of inlined blocks to be stored on an array and grow the
+ * structure into a rb tree when necessary. */
+#define OCFS2_INODE_MAX_CACHE_ARRAY 2
+
+struct ocfs2_caching_info {
+	unsigned int		ci_num_cached;
+	union {
+		sector_t	ci_array[OCFS2_INODE_MAX_CACHE_ARRAY];
+		struct rb_root	ci_tree;
+	} ci_cache;
+};
+
+/* this limits us to 256 nodes
+ * if we need more, we can do a kmalloc for the map */
+#define OCFS2_NODE_MAP_MAX_NODES    256
+struct ocfs2_node_map {
+	u16 num_nodes;
+	unsigned long map[BITS_TO_LONGS(OCFS2_NODE_MAP_MAX_NODES)];
+};
+
+enum ocfs2_ast_action {
+	OCFS2_AST_INVALID = 0,
+	OCFS2_AST_ATTACH,
+	OCFS2_AST_CONVERT,
+	OCFS2_AST_DOWNCONVERT,
+};
+
+/* actions for an unlockast function to take. */
+enum ocfs2_unlock_action {
+	OCFS2_UNLOCK_INVALID = 0,
+	OCFS2_UNLOCK_CANCEL_CONVERT,
+	OCFS2_UNLOCK_DROP_LOCK,
+};
+
+/* ocfs2_lock_res->l_flags flags. */
+#define OCFS2_LOCK_ATTACHED      (0x00000001) /* have we initialized
+					       * the lvb */
+#define OCFS2_LOCK_BUSY          (0x00000002) /* we are currently in
+					       * dlm_lock */
+#define OCFS2_LOCK_BLOCKED       (0x00000004) /* blocked waiting to
+					       * downconvert*/
+#define OCFS2_LOCK_LOCAL         (0x00000008) /* newly created inode */
+#define OCFS2_LOCK_NEEDS_REFRESH (0x00000010)
+#define OCFS2_LOCK_REFRESHING    (0x00000020)
+#define OCFS2_LOCK_INITIALIZED   (0x00000040) /* track initialization
+					       * for shutdown paths */
+#define OCFS2_LOCK_FREEING       (0x00000080) /* help dlmglue track
+					       * when to skip queueing
+					       * a lock because it's
+					       * about to be
+					       * dropped. */
+#define OCFS2_LOCK_QUEUED        (0x00000100) /* queued for downconvert */
+
+struct ocfs2_lock_res_ops;
+
+typedef void (*ocfs2_lock_callback)(int status, unsigned long data);
+
+struct ocfs2_lock_res {
+	void                    *l_priv;
+	struct ocfs2_lock_res_ops *l_ops;
+	spinlock_t               l_lock;
+
+	struct list_head         l_blocked_list;
+	struct list_head         l_mask_waiters;
+
+	enum ocfs2_lock_type     l_type;
+	unsigned long		 l_flags;
+	char                     l_name[OCFS2_LOCK_ID_MAX_LEN];
+	int                      l_level;
+	unsigned int             l_ro_holders;
+	unsigned int             l_ex_holders;
+	struct dlm_lockstatus    l_lksb;
+
+	/* used from AST/BAST funcs. */
+	enum ocfs2_ast_action    l_action;
+	enum ocfs2_unlock_action l_unlock_action;
+	int                      l_requested;
+	int                      l_blocking;
+
+	wait_queue_head_t        l_event;
+
+	struct list_head         l_debug_list;
+};
+
+struct ocfs2_dlm_debug {
+	struct kref d_refcnt;
+	struct dentry *d_locking_state;
+	struct list_head d_lockres_tracking;
+};
+
+enum ocfs2_vol_state
+{
+	VOLUME_INIT = 0,
+	VOLUME_MOUNTED,
+	VOLUME_DISMOUNTED,
+	VOLUME_DISABLED
+};
+
+struct ocfs2_alloc_stats
+{
+	atomic_t moves;
+	atomic_t local_data;
+	atomic_t bitmap_data;
+	atomic_t bg_allocs;
+	atomic_t bg_extends;
+};
+
+enum ocfs2_local_alloc_state
+{
+	OCFS2_LA_UNUSED = 0,
+	OCFS2_LA_ENABLED,
+	OCFS2_LA_DISABLED
+};
+
+enum ocfs2_mount_options
+{
+	OCFS2_MOUNT_HB_LOCAL   = 1 << 0, /* Heartbeat started in local mode */
+	OCFS2_MOUNT_BARRIER = 1 << 1,	/* Use block barriers */
+	OCFS2_MOUNT_NOINTR  = 1 << 2,   /* Don't catch signals */
+	OCFS2_MOUNT_ERRORS_PANIC = 1 << 3, /* Panic on errors */
+	OCFS2_MOUNT_DATA_WRITEBACK = 1 << 4, /* No data ordering */
+#ifdef OCFS2_ORACORE_WORKAROUNDS
+	OCFS2_MOUNT_COMPAT_OCFS = 1 << 30, /* ocfs1 compatibility mode */
+#endif
+};
+
+#define OCFS2_OSB_SOFT_RO	0x0001
+#define OCFS2_OSB_HARD_RO	0x0002
+#define OCFS2_OSB_ERROR_FS	0x0004
+
+struct ocfs2_journal;
+struct ocfs2_journal_handle;
+struct ocfs2_super
+{
+	u32 osb_id;		/* id used by the proc interface */
+	struct task_struct *commit_task;
+	struct super_block *sb;
+	struct inode *root_inode;
+	struct inode *sys_root_inode;
+	struct inode *system_inodes[NUM_SYSTEM_INODES];
+
+	struct ocfs2_slot_info *slot_info;
+
+	spinlock_t node_map_lock;
+	struct ocfs2_node_map mounted_map;
+	struct ocfs2_node_map recovery_map;
+	struct ocfs2_node_map umount_map;
+
+	u32 num_clusters;
+	u64 root_blkno;
+	u64 system_dir_blkno;
+	u64 bitmap_blkno;
+	u32 bitmap_cpg;
+	u8 *uuid;
+	char *uuid_str;
+	u8 *vol_label;
+	u64 first_cluster_group_blkno;
+	u32 fs_generation;
+
+	u32 s_feature_compat;
+	u32 s_feature_incompat;
+	u32 s_feature_ro_compat;
+
+	/* Protects s_next_generaion, osb_flags. Could protect more on
+	 * osb as it's very short lived. */
+	spinlock_t osb_lock;
+	u32 s_next_generation;
+	unsigned long osb_flags;
+
+	unsigned long s_mount_opt;
+
+	u16 max_slots;
+	u16 num_nodes;
+	s16 node_num;
+	s16 slot_num;
+	int s_sectsize_bits;
+	int s_clustersize;
+	int s_clustersize_bits;
+	struct proc_dir_entry *proc_sub_dir; /* points to /proc/fs/ocfs2/<maj_min> */
+
+	atomic_t vol_state;
+	struct semaphore recovery_lock;
+	struct task_struct *recovery_thread_task;
+	int disable_recovery;
+	wait_queue_head_t checkpoint_event;
+	atomic_t needs_checkpoint;
+	struct ocfs2_journal *journal;
+
+	enum ocfs2_local_alloc_state local_alloc_state;
+	struct buffer_head *local_alloc_bh;
+
+	/* Next two fields are for local node slot recovery during
+	 * mount. */
+	int dirty;
+	struct ocfs2_dinode *local_alloc_copy;
+
+	struct ocfs2_alloc_stats alloc_stats;
+	char dev_str[20];		/* "major,minor" of the device */
+
+	struct dlm_ctxt *dlm;
+	struct ocfs2_lock_res osb_super_lockres;
+	struct ocfs2_lock_res osb_rename_lockres;
+	struct dlm_eviction_cb osb_eviction_cb;
+	struct ocfs2_dlm_debug *osb_dlm_debug;
+
+	struct dentry *osb_debug_root;
+
+	wait_queue_head_t recovery_event;
+
+	spinlock_t vote_task_lock;
+	struct task_struct *vote_task;
+	wait_queue_head_t vote_event;
+	unsigned long vote_wake_sequence;
+	unsigned long vote_work_sequence;
+
+	struct list_head blocked_lock_list;
+	unsigned long blocked_lock_count;
+
+	struct list_head vote_list;
+	int vote_count;
+
+	u32 net_key;
+	spinlock_t net_response_lock;
+	unsigned int net_response_ids;
+	struct list_head net_response_list;
+
+	struct o2hb_callback_func osb_hb_up;
+	struct o2hb_callback_func osb_hb_down;
+
+	struct list_head	osb_net_handlers;
+
+	wait_queue_head_t		osb_mount_event;
+
+	/* Truncate log info */
+	struct inode			*osb_tl_inode;
+	struct buffer_head		*osb_tl_bh;
+	struct work_struct		osb_truncate_log_wq;
+};
+
+#define OCFS2_SB(sb)	    ((struct ocfs2_super *)(sb)->s_fs_info)
+#define OCFS2_MAX_OSB_ID             65536
+
+static inline int ocfs2_should_order_data(struct inode *inode)
+{
+	if (!S_ISREG(inode->i_mode))
+		return 0;
+	if (OCFS2_SB(inode->i_sb)->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)
+		return 0;
+	return 1;
+}
+
+/* set / clear functions because cluster events can make these happen
+ * in parallel so we want the transitions to be atomic. this also
+ * means that any future flags osb_flags must be protected by spinlock
+ * too! */
+static inline void ocfs2_set_osb_flag(struct ocfs2_super *osb,
+				      unsigned long flag)
+{
+	spin_lock(&osb->osb_lock);
+	osb->osb_flags |= flag;
+	spin_unlock(&osb->osb_lock);
+}
+
+static inline void ocfs2_set_ro_flag(struct ocfs2_super *osb,
+				     int hard)
+{
+	spin_lock(&osb->osb_lock);
+	osb->osb_flags &= ~(OCFS2_OSB_SOFT_RO|OCFS2_OSB_HARD_RO);
+	if (hard)
+		osb->osb_flags |= OCFS2_OSB_HARD_RO;
+	else
+		osb->osb_flags |= OCFS2_OSB_SOFT_RO;
+	spin_unlock(&osb->osb_lock);
+}
+
+static inline int ocfs2_is_hard_readonly(struct ocfs2_super *osb)
+{
+	int ret;
+
+	spin_lock(&osb->osb_lock);
+	ret = osb->osb_flags & OCFS2_OSB_HARD_RO;
+	spin_unlock(&osb->osb_lock);
+
+	return ret;
+}
+
+static inline int ocfs2_is_soft_readonly(struct ocfs2_super *osb)
+{
+	int ret;
+
+	spin_lock(&osb->osb_lock);
+	ret = osb->osb_flags & OCFS2_OSB_SOFT_RO;
+	spin_unlock(&osb->osb_lock);
+
+	return ret;
+}
+
+#define OCFS2_IS_VALID_DINODE(ptr)					\
+	(!strcmp((ptr)->i_signature, OCFS2_INODE_SIGNATURE))
+
+#define OCFS2_RO_ON_INVALID_DINODE(__sb, __di)	do {			\
+	typeof(__di) ____di = (__di);					\
+	ocfs2_error((__sb), 						\
+		"Dinode # %"MLFu64" has bad signature %.*s",		\
+		(____di)->i_blkno, 7,					\
+		(____di)->i_signature);					\
+} while (0);
+
+#define OCFS2_IS_VALID_EXTENT_BLOCK(ptr)				\
+	(!strcmp((ptr)->h_signature, OCFS2_EXTENT_BLOCK_SIGNATURE))
+
+#define OCFS2_RO_ON_INVALID_EXTENT_BLOCK(__sb, __eb)	do {		\
+	typeof(__eb) ____eb = (__eb);					\
+	ocfs2_error((__sb), 						\
+		"Extent Block # %"MLFu64" has bad signature %.*s",	\
+		(____eb)->h_blkno, 7,					\
+		(____eb)->h_signature);					\
+} while (0);
+
+#define OCFS2_IS_VALID_GROUP_DESC(ptr)					\
+	(!strcmp((ptr)->bg_signature, OCFS2_GROUP_DESC_SIGNATURE))
+
+#define OCFS2_RO_ON_INVALID_GROUP_DESC(__sb, __gd)	do {		\
+	typeof(__gd) ____gd = (__gd);					\
+		ocfs2_error((__sb),					\
+		"Group Descriptor # %"MLFu64" has bad signature %.*s",	\
+		(____gd)->bg_blkno, 7,					\
+		(____gd)->bg_signature);				\
+} while (0);
+
+static inline unsigned long ino_from_blkno(struct super_block *sb,
+					   u64 blkno)
+{
+	return (unsigned long)(blkno & (u64)ULONG_MAX);
+}
+
+static inline u64 ocfs2_clusters_to_blocks(struct super_block *sb,
+					   u32 clusters)
+{
+	int c_to_b_bits = OCFS2_SB(sb)->s_clustersize_bits -
+		sb->s_blocksize_bits;
+
+	return (u64)clusters << c_to_b_bits;
+}
+
+static inline u32 ocfs2_blocks_to_clusters(struct super_block *sb,
+					   u64 blocks)
+{
+	int b_to_c_bits = OCFS2_SB(sb)->s_clustersize_bits -
+		sb->s_blocksize_bits;
+
+	return (u32)(blocks >> b_to_c_bits);
+}
+
+static inline unsigned int ocfs2_clusters_for_bytes(struct super_block *sb,
+						    u64 bytes)
+{
+	int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
+	unsigned int clusters;
+
+	bytes += OCFS2_SB(sb)->s_clustersize - 1;
+	/* OCFS2 just cannot have enough clusters to overflow this */
+	clusters = (unsigned int)(bytes >> cl_bits);
+
+	return clusters;
+}
+
+static inline u64 ocfs2_blocks_for_bytes(struct super_block *sb,
+					 u64 bytes)
+{
+	bytes += sb->s_blocksize - 1;
+	return bytes >> sb->s_blocksize_bits;
+}
+
+static inline u64 ocfs2_clusters_to_bytes(struct super_block *sb,
+					  u32 clusters)
+{
+	return (u64)clusters << OCFS2_SB(sb)->s_clustersize_bits;
+}
+
+static inline u64 ocfs2_align_bytes_to_clusters(struct super_block *sb,
+						u64 bytes)
+{
+	int cl_bits = OCFS2_SB(sb)->s_clustersize_bits;
+	unsigned int clusters;
+
+	clusters = ocfs2_clusters_for_bytes(sb, bytes);
+	return (u64)clusters << cl_bits;
+}
+
+static inline u64 ocfs2_align_bytes_to_blocks(struct super_block *sb,
+					      u64 bytes)
+{
+	u64 blocks;
+
+        blocks = ocfs2_blocks_for_bytes(sb, bytes);
+	return blocks << sb->s_blocksize_bits;
+}
+
+static inline unsigned long ocfs2_align_bytes_to_sectors(u64 bytes)
+{
+	return (unsigned long)((bytes + 511) >> 9);
+}
+
+#define ocfs2_set_bit ext2_set_bit
+#define ocfs2_clear_bit ext2_clear_bit
+#define ocfs2_test_bit ext2_test_bit
+#define ocfs2_find_next_zero_bit ext2_find_next_zero_bit
+#endif  /* OCFS2_H */
+
diff --git a/fs/ocfs2/ocfs2_fs.h b/fs/ocfs2/ocfs2_fs.h
new file mode 100644
index 00000000000..dfb8a5bedfc
--- /dev/null
+++ b/fs/ocfs2/ocfs2_fs.h
@@ -0,0 +1,638 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs2_fs.h
+ *
+ * On-disk structures for OCFS2.
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public
+ * License, version 2,  as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef _OCFS2_FS_H
+#define _OCFS2_FS_H
+
+/* Version */
+#define OCFS2_MAJOR_REV_LEVEL		0
+#define OCFS2_MINOR_REV_LEVEL          	90
+
+/*
+ * An OCFS2 volume starts this way:
+ * Sector 0: Valid ocfs1_vol_disk_hdr that cleanly fails to mount OCFS.
+ * Sector 1: Valid ocfs1_vol_label that cleanly fails to mount OCFS.
+ * Block OCFS2_SUPER_BLOCK_BLKNO: OCFS2 superblock.
+ *
+ * All other structures are found from the superblock information.
+ *
+ * OCFS2_SUPER_BLOCK_BLKNO is in blocks, not sectors.  eg, for a
+ * blocksize of 2K, it is 4096 bytes into disk.
+ */
+#define OCFS2_SUPER_BLOCK_BLKNO		2
+
+/*
+ * Cluster size limits. The maximum is kept arbitrarily at 1 MB, and could
+ * grow if needed.
+ */
+#define OCFS2_MIN_CLUSTERSIZE		4096
+#define OCFS2_MAX_CLUSTERSIZE		1048576
+
+/*
+ * Blocks cannot be bigger than clusters, so the maximum blocksize is the
+ * minimum cluster size.
+ */
+#define OCFS2_MIN_BLOCKSIZE		512
+#define OCFS2_MAX_BLOCKSIZE		OCFS2_MIN_CLUSTERSIZE
+
+/* Filesystem magic number */
+#define OCFS2_SUPER_MAGIC		0x7461636f
+
+/* Object signatures */
+#define OCFS2_SUPER_BLOCK_SIGNATURE	"OCFSV2"
+#define OCFS2_INODE_SIGNATURE		"INODE01"
+#define OCFS2_EXTENT_BLOCK_SIGNATURE	"EXBLK01"
+#define OCFS2_GROUP_DESC_SIGNATURE      "GROUP01"
+
+/* Compatibility flags */
+#define OCFS2_HAS_COMPAT_FEATURE(sb,mask)			\
+	( OCFS2_SB(sb)->s_feature_compat & (mask) )
+#define OCFS2_HAS_RO_COMPAT_FEATURE(sb,mask)			\
+	( OCFS2_SB(sb)->s_feature_ro_compat & (mask) )
+#define OCFS2_HAS_INCOMPAT_FEATURE(sb,mask)			\
+	( OCFS2_SB(sb)->s_feature_incompat & (mask) )
+#define OCFS2_SET_COMPAT_FEATURE(sb,mask)			\
+	OCFS2_SB(sb)->s_feature_compat |= (mask)
+#define OCFS2_SET_RO_COMPAT_FEATURE(sb,mask)			\
+	OCFS2_SB(sb)->s_feature_ro_compat |= (mask)
+#define OCFS2_SET_INCOMPAT_FEATURE(sb,mask)			\
+	OCFS2_SB(sb)->s_feature_incompat |= (mask)
+#define OCFS2_CLEAR_COMPAT_FEATURE(sb,mask)			\
+	OCFS2_SB(sb)->s_feature_compat &= ~(mask)
+#define OCFS2_CLEAR_RO_COMPAT_FEATURE(sb,mask)			\
+	OCFS2_SB(sb)->s_feature_ro_compat &= ~(mask)
+#define OCFS2_CLEAR_INCOMPAT_FEATURE(sb,mask)			\
+	OCFS2_SB(sb)->s_feature_incompat &= ~(mask)
+
+#define OCFS2_FEATURE_COMPAT_SUPP	0
+#define OCFS2_FEATURE_INCOMPAT_SUPP	0
+#define OCFS2_FEATURE_RO_COMPAT_SUPP	0
+
+/*
+ * Heartbeat-only devices are missing journals and other files.  The
+ * filesystem driver can't load them, but the library can.  Never put
+ * this in OCFS2_FEATURE_INCOMPAT_SUPP, *ever*.
+ */
+#define OCFS2_FEATURE_INCOMPAT_HEARTBEAT_DEV	0x0002
+
+
+/*
+ * Flags on ocfs2_dinode.i_flags
+ */
+#define OCFS2_VALID_FL		(0x00000001)	/* Inode is valid */
+#define OCFS2_UNUSED2_FL	(0x00000002)
+#define OCFS2_ORPHANED_FL	(0x00000004)	/* On the orphan list */
+#define OCFS2_UNUSED3_FL	(0x00000008)
+/* System inode flags */
+#define OCFS2_SYSTEM_FL		(0x00000010)	/* System inode */
+#define OCFS2_SUPER_BLOCK_FL	(0x00000020)	/* Super block */
+#define OCFS2_LOCAL_ALLOC_FL	(0x00000040)	/* Slot local alloc bitmap */
+#define OCFS2_BITMAP_FL		(0x00000080)	/* Allocation bitmap */
+#define OCFS2_JOURNAL_FL	(0x00000100)	/* Slot local journal */
+#define OCFS2_HEARTBEAT_FL	(0x00000200)	/* Heartbeat area */
+#define OCFS2_CHAIN_FL		(0x00000400)	/* Chain allocator */
+#define OCFS2_DEALLOC_FL	(0x00000800)	/* Truncate log */
+
+/*
+ * Journal Flags (ocfs2_dinode.id1.journal1.i_flags)
+ */
+#define OCFS2_JOURNAL_DIRTY_FL	(0x00000001)	/* Journal needs recovery */
+
+/*
+ * superblock s_state flags
+ */
+#define OCFS2_ERROR_FS		(0x00000001)	/* FS saw errors */
+
+/* Limit of space in ocfs2_dir_entry */
+#define OCFS2_MAX_FILENAME_LEN		255
+
+/* Maximum slots on an ocfs2 file system */
+#define OCFS2_MAX_SLOTS			255
+
+/* Slot map indicator for an empty slot */
+#define OCFS2_INVALID_SLOT		-1
+
+#define OCFS2_VOL_UUID_LEN		16
+#define OCFS2_MAX_VOL_LABEL_LEN		64
+
+/* Journal limits (in bytes) */
+#define OCFS2_MIN_JOURNAL_SIZE		(4 * 1024 * 1024)
+#define OCFS2_MAX_JOURNAL_SIZE		(500 * 1024 * 1024)
+
+struct ocfs2_system_inode_info {
+	char	*si_name;
+	int	si_iflags;
+	int	si_mode;
+};
+
+/* System file index */
+enum {
+	BAD_BLOCK_SYSTEM_INODE = 0,
+	GLOBAL_INODE_ALLOC_SYSTEM_INODE,
+	SLOT_MAP_SYSTEM_INODE,
+#define OCFS2_FIRST_ONLINE_SYSTEM_INODE SLOT_MAP_SYSTEM_INODE
+	HEARTBEAT_SYSTEM_INODE,
+	GLOBAL_BITMAP_SYSTEM_INODE,
+#define OCFS2_LAST_GLOBAL_SYSTEM_INODE GLOBAL_BITMAP_SYSTEM_INODE
+	ORPHAN_DIR_SYSTEM_INODE,
+	EXTENT_ALLOC_SYSTEM_INODE,
+	INODE_ALLOC_SYSTEM_INODE,
+	JOURNAL_SYSTEM_INODE,
+	LOCAL_ALLOC_SYSTEM_INODE,
+	TRUNCATE_LOG_SYSTEM_INODE,
+	NUM_SYSTEM_INODES
+};
+
+static struct ocfs2_system_inode_info ocfs2_system_inodes[NUM_SYSTEM_INODES] = {
+	/* Global system inodes (single copy) */
+	/* The first two are only used from userspace mfks/tunefs */
+	[BAD_BLOCK_SYSTEM_INODE]		= { "bad_blocks", 0, S_IFREG | 0644 },
+	[GLOBAL_INODE_ALLOC_SYSTEM_INODE] 	= { "global_inode_alloc", OCFS2_BITMAP_FL | OCFS2_CHAIN_FL, S_IFREG | 0644 },
+
+	/* These are used by the running filesystem */
+	[SLOT_MAP_SYSTEM_INODE]			= { "slot_map", 0, S_IFREG | 0644 },
+	[HEARTBEAT_SYSTEM_INODE]		= { "heartbeat", OCFS2_HEARTBEAT_FL, S_IFREG | 0644 },
+	[GLOBAL_BITMAP_SYSTEM_INODE]		= { "global_bitmap", 0, S_IFREG | 0644 },
+
+	/* Slot-specific system inodes (one copy per slot) */
+	[ORPHAN_DIR_SYSTEM_INODE]		= { "orphan_dir:%04d", 0, S_IFDIR | 0755 },
+	[EXTENT_ALLOC_SYSTEM_INODE]		= { "extent_alloc:%04d", OCFS2_BITMAP_FL | OCFS2_CHAIN_FL, S_IFREG | 0644 },
+	[INODE_ALLOC_SYSTEM_INODE]		= { "inode_alloc:%04d", OCFS2_BITMAP_FL | OCFS2_CHAIN_FL, S_IFREG | 0644 },
+	[JOURNAL_SYSTEM_INODE]			= { "journal:%04d", OCFS2_JOURNAL_FL, S_IFREG | 0644 },
+	[LOCAL_ALLOC_SYSTEM_INODE]		= { "local_alloc:%04d", OCFS2_BITMAP_FL | OCFS2_LOCAL_ALLOC_FL, S_IFREG | 0644 },
+	[TRUNCATE_LOG_SYSTEM_INODE]		= { "truncate_log:%04d", OCFS2_DEALLOC_FL, S_IFREG | 0644 }
+};
+
+/* Parameter passed from mount.ocfs2 to module */
+#define OCFS2_HB_NONE			"heartbeat=none"
+#define OCFS2_HB_LOCAL			"heartbeat=local"
+
+/*
+ * OCFS2 directory file types.  Only the low 3 bits are used.  The
+ * other bits are reserved for now.
+ */
+#define OCFS2_FT_UNKNOWN	0
+#define OCFS2_FT_REG_FILE	1
+#define OCFS2_FT_DIR		2
+#define OCFS2_FT_CHRDEV		3
+#define OCFS2_FT_BLKDEV		4
+#define OCFS2_FT_FIFO		5
+#define OCFS2_FT_SOCK		6
+#define OCFS2_FT_SYMLINK	7
+
+#define OCFS2_FT_MAX		8
+
+/*
+ * OCFS2_DIR_PAD defines the directory entries boundaries
+ *
+ * NOTE: It must be a multiple of 4
+ */
+#define OCFS2_DIR_PAD			4
+#define OCFS2_DIR_ROUND			(OCFS2_DIR_PAD - 1)
+#define OCFS2_DIR_MEMBER_LEN 		offsetof(struct ocfs2_dir_entry, name)
+#define OCFS2_DIR_REC_LEN(name_len)	(((name_len) + OCFS2_DIR_MEMBER_LEN + \
+                                          OCFS2_DIR_ROUND) & \
+					 ~OCFS2_DIR_ROUND)
+
+#define OCFS2_LINK_MAX		32000
+
+#define S_SHIFT			12
+static unsigned char ocfs2_type_by_mode[S_IFMT >> S_SHIFT] = {
+	[S_IFREG >> S_SHIFT]  = OCFS2_FT_REG_FILE,
+	[S_IFDIR >> S_SHIFT]  = OCFS2_FT_DIR,
+	[S_IFCHR >> S_SHIFT]  = OCFS2_FT_CHRDEV,
+	[S_IFBLK >> S_SHIFT]  = OCFS2_FT_BLKDEV,
+	[S_IFIFO >> S_SHIFT]  = OCFS2_FT_FIFO,
+	[S_IFSOCK >> S_SHIFT] = OCFS2_FT_SOCK,
+	[S_IFLNK >> S_SHIFT]  = OCFS2_FT_SYMLINK,
+};
+
+
+/*
+ * Convenience casts
+ */
+#define OCFS2_RAW_SB(dinode)		(&((dinode)->id2.i_super))
+
+/*
+ * On disk extent record for OCFS2
+ * It describes a range of clusters on disk.
+ */
+struct ocfs2_extent_rec {
+/*00*/	__le32 e_cpos;		/* Offset into the file, in clusters */
+	__le32 e_clusters;	/* Clusters covered by this extent */
+	__le64 e_blkno;		/* Physical disk offset, in blocks */
+/*10*/
+};
+
+struct ocfs2_chain_rec {
+	__le32 c_free;	/* Number of free bits in this chain. */
+	__le32 c_total;	/* Number of total bits in this chain */
+	__le64 c_blkno;	/* Physical disk offset (blocks) of 1st group */
+};
+
+struct ocfs2_truncate_rec {
+	__le32 t_start;		/* 1st cluster in this log */
+	__le32 t_clusters;	/* Number of total clusters covered */
+};
+
+/*
+ * On disk extent list for OCFS2 (node in the tree).  Note that this
+ * is contained inside ocfs2_dinode or ocfs2_extent_block, so the
+ * offsets are relative to ocfs2_dinode.id2.i_list or
+ * ocfs2_extent_block.h_list, respectively.
+ */
+struct ocfs2_extent_list {
+/*00*/	__le16 l_tree_depth;		/* Extent tree depth from this
+					   point.  0 means data extents
+					   hang directly off this
+					   header (a leaf) */
+	__le16 l_count;			/* Number of extent records */
+	__le16 l_next_free_rec;		/* Next unused extent slot */
+	__le16 l_reserved1;
+	__le64 l_reserved2;		/* Pad to
+					   sizeof(ocfs2_extent_rec) */
+/*10*/	struct ocfs2_extent_rec l_recs[0];	/* Extent records */
+};
+
+/*
+ * On disk allocation chain list for OCFS2.  Note that this is
+ * contained inside ocfs2_dinode, so the offsets are relative to
+ * ocfs2_dinode.id2.i_chain.
+ */
+struct ocfs2_chain_list {
+/*00*/	__le16 cl_cpg;			/* Clusters per Block Group */
+	__le16 cl_bpc;			/* Bits per cluster */
+	__le16 cl_count;		/* Total chains in this list */
+	__le16 cl_next_free_rec;	/* Next unused chain slot */
+	__le64 cl_reserved1;
+/*10*/	struct ocfs2_chain_rec cl_recs[0];	/* Chain records */
+};
+
+/*
+ * On disk deallocation log for OCFS2.  Note that this is
+ * contained inside ocfs2_dinode, so the offsets are relative to
+ * ocfs2_dinode.id2.i_dealloc.
+ */
+struct ocfs2_truncate_log {
+/*00*/	__le16 tl_count;		/* Total records in this log */
+	__le16 tl_used;			/* Number of records in use */
+	__le32 tl_reserved1;
+/*08*/	struct ocfs2_truncate_rec tl_recs[0];	/* Truncate records */
+};
+
+/*
+ * On disk extent block (indirect block) for OCFS2
+ */
+struct ocfs2_extent_block
+{
+/*00*/	__u8 h_signature[8];		/* Signature for verification */
+	__le64 h_reserved1;
+/*10*/	__le16 h_suballoc_slot;		/* Slot suballocator this
+					   extent_header belongs to */
+	__le16 h_suballoc_bit;		/* Bit offset in suballocator
+					   block group */
+	__le32 h_fs_generation;		/* Must match super block */
+	__le64 h_blkno;			/* Offset on disk, in blocks */
+/*20*/	__le64 h_reserved3;
+	__le64 h_next_leaf_blk;		/* Offset on disk, in blocks,
+					   of next leaf header pointing
+					   to data */
+/*30*/	struct ocfs2_extent_list h_list;	/* Extent record list */
+/* Actual on-disk size is one block */
+};
+
+/*
+ * On disk superblock for OCFS2
+ * Note that it is contained inside an ocfs2_dinode, so all offsets
+ * are relative to the start of ocfs2_dinode.id2.
+ */
+struct ocfs2_super_block {
+/*00*/	__le16 s_major_rev_level;
+	__le16 s_minor_rev_level;
+	__le16 s_mnt_count;
+	__le16 s_max_mnt_count;
+	__le16 s_state;			/* File system state */
+	__le16 s_errors;			/* Behaviour when detecting errors */
+	__le32 s_checkinterval;		/* Max time between checks */
+/*10*/	__le64 s_lastcheck;		/* Time of last check */
+	__le32 s_creator_os;		/* OS */
+	__le32 s_feature_compat;		/* Compatible feature set */
+/*20*/	__le32 s_feature_incompat;	/* Incompatible feature set */
+	__le32 s_feature_ro_compat;	/* Readonly-compatible feature set */
+	__le64 s_root_blkno;		/* Offset, in blocks, of root directory
+					   dinode */
+/*30*/	__le64 s_system_dir_blkno;	/* Offset, in blocks, of system
+					   directory dinode */
+	__le32 s_blocksize_bits;		/* Blocksize for this fs */
+	__le32 s_clustersize_bits;	/* Clustersize for this fs */
+/*40*/	__le16 s_max_slots;		/* Max number of simultaneous mounts
+					   before tunefs required */
+	__le16 s_reserved1;
+	__le32 s_reserved2;
+	__le64 s_first_cluster_group;	/* Block offset of 1st cluster
+					 * group header */
+/*50*/	__u8  s_label[OCFS2_MAX_VOL_LABEL_LEN];	/* Label for mounting, etc. */
+/*90*/	__u8  s_uuid[OCFS2_VOL_UUID_LEN];	/* 128-bit uuid */
+/*A0*/
+};
+
+/*
+ * Local allocation bitmap for OCFS2 slots
+ * Note that it exists inside an ocfs2_dinode, so all offsets are
+ * relative to the start of ocfs2_dinode.id2.
+ */
+struct ocfs2_local_alloc
+{
+/*00*/	__le32 la_bm_off;	/* Starting bit offset in main bitmap */
+	__le16 la_size;		/* Size of included bitmap, in bytes */
+	__le16 la_reserved1;
+	__le64 la_reserved2;
+/*10*/	__u8   la_bitmap[0];
+};
+
+/*
+ * On disk inode for OCFS2
+ */
+struct ocfs2_dinode {
+/*00*/	__u8 i_signature[8];		/* Signature for validation */
+	__le32 i_generation;		/* Generation number */
+	__le16 i_suballoc_slot;		/* Slot suballocator this inode
+					   belongs to */
+	__le16 i_suballoc_bit;		/* Bit offset in suballocator
+					   block group */
+/*10*/	__le32 i_reserved0;
+	__le32 i_clusters;		/* Cluster count */
+	__le32 i_uid;			/* Owner UID */
+	__le32 i_gid;			/* Owning GID */
+/*20*/	__le64 i_size;			/* Size in bytes */
+	__le16 i_mode;			/* File mode */
+	__le16 i_links_count;		/* Links count */
+	__le32 i_flags;			/* File flags */
+/*30*/	__le64 i_atime;			/* Access time */
+	__le64 i_ctime;			/* Creation time */
+/*40*/	__le64 i_mtime;			/* Modification time */
+	__le64 i_dtime;			/* Deletion time */
+/*50*/	__le64 i_blkno;			/* Offset on disk, in blocks */
+	__le64 i_last_eb_blk;		/* Pointer to last extent
+					   block */
+/*60*/	__le32 i_fs_generation;		/* Generation per fs-instance */
+	__le32 i_atime_nsec;
+	__le32 i_ctime_nsec;
+	__le32 i_mtime_nsec;
+/*70*/	__le64 i_reserved1[9];
+/*B8*/	union {
+		__le64 i_pad1;		/* Generic way to refer to this
+					   64bit union */
+		struct {
+			__le64 i_rdev;	/* Device number */
+		} dev1;
+		struct {		/* Info for bitmap system
+					   inodes */
+			__le32 i_used;	/* Bits (ie, clusters) used  */
+			__le32 i_total;	/* Total bits (clusters)
+					   available */
+		} bitmap1;
+		struct {		/* Info for journal system
+					   inodes */
+			__le32 ij_flags;	/* Mounted, version, etc. */
+			__le32 ij_pad;
+		} journal1;
+	} id1;				/* Inode type dependant 1 */
+/*C0*/	union {
+		struct ocfs2_super_block	i_super;
+		struct ocfs2_local_alloc	i_lab;
+		struct ocfs2_chain_list		i_chain;
+		struct ocfs2_extent_list	i_list;
+		struct ocfs2_truncate_log	i_dealloc;
+		__u8               		i_symlink[0];
+	} id2;
+/* Actual on-disk size is one block */
+};
+
+/*
+ * On-disk directory entry structure for OCFS2
+ *
+ * Packed as this structure could be accessed unaligned on 64-bit platforms
+ */
+struct ocfs2_dir_entry {
+/*00*/	__le64   inode;                  /* Inode number */
+	__le16   rec_len;                /* Directory entry length */
+	__u8    name_len;               /* Name length */
+	__u8    file_type;
+/*0C*/	char    name[OCFS2_MAX_FILENAME_LEN];   /* File name */
+/* Actual on-disk length specified by rec_len */
+} __attribute__ ((packed));
+
+/*
+ * On disk allocator group structure for OCFS2
+ */
+struct ocfs2_group_desc
+{
+/*00*/	__u8    bg_signature[8];        /* Signature for validation */
+	__le16   bg_size;                /* Size of included bitmap in
+					   bytes. */
+	__le16   bg_bits;                /* Bits represented by this
+					   group. */
+	__le16	bg_free_bits_count;     /* Free bits count */
+	__le16   bg_chain;               /* What chain I am in. */
+/*10*/	__le32   bg_generation;
+	__le32	bg_reserved1;
+	__le64   bg_next_group;          /* Next group in my list, in
+					   blocks */
+/*20*/	__le64   bg_parent_dinode;       /* dinode which owns me, in
+					   blocks */
+	__le64   bg_blkno;               /* Offset on disk, in blocks */
+/*30*/	__le64   bg_reserved2[2];
+/*40*/	__u8    bg_bitmap[0];
+};
+
+#ifdef __KERNEL__
+static inline int ocfs2_fast_symlink_chars(struct super_block *sb)
+{
+	return  sb->s_blocksize -
+		 offsetof(struct ocfs2_dinode, id2.i_symlink);
+}
+
+static inline int ocfs2_extent_recs_per_inode(struct super_block *sb)
+{
+	int size;
+
+	size = sb->s_blocksize -
+		offsetof(struct ocfs2_dinode, id2.i_list.l_recs);
+
+	return size / sizeof(struct ocfs2_extent_rec);
+}
+
+static inline int ocfs2_chain_recs_per_inode(struct super_block *sb)
+{
+	int size;
+
+	size = sb->s_blocksize -
+		offsetof(struct ocfs2_dinode, id2.i_chain.cl_recs);
+
+	return size / sizeof(struct ocfs2_chain_rec);
+}
+
+static inline u16 ocfs2_extent_recs_per_eb(struct super_block *sb)
+{
+	int size;
+
+	size = sb->s_blocksize -
+		offsetof(struct ocfs2_extent_block, h_list.l_recs);
+
+	return size / sizeof(struct ocfs2_extent_rec);
+}
+
+static inline u16 ocfs2_local_alloc_size(struct super_block *sb)
+{
+	u16 size;
+
+	size = sb->s_blocksize -
+		offsetof(struct ocfs2_dinode, id2.i_lab.la_bitmap);
+
+	return size;
+}
+
+static inline int ocfs2_group_bitmap_size(struct super_block *sb)
+{
+	int size;
+
+	size = sb->s_blocksize -
+		offsetof(struct ocfs2_group_desc, bg_bitmap);
+
+	return size;
+}
+
+static inline int ocfs2_truncate_recs_per_inode(struct super_block *sb)
+{
+	int size;
+
+	size = sb->s_blocksize -
+		offsetof(struct ocfs2_dinode, id2.i_dealloc.tl_recs);
+
+	return size / sizeof(struct ocfs2_truncate_rec);
+}
+#else
+static inline int ocfs2_fast_symlink_chars(int blocksize)
+{
+	return blocksize - offsetof(struct ocfs2_dinode, id2.i_symlink);
+}
+
+static inline int ocfs2_extent_recs_per_inode(int blocksize)
+{
+	int size;
+
+	size = blocksize -
+		offsetof(struct ocfs2_dinode, id2.i_list.l_recs);
+
+	return size / sizeof(struct ocfs2_extent_rec);
+}
+
+static inline int ocfs2_chain_recs_per_inode(int blocksize)
+{
+	int size;
+
+	size = blocksize -
+		offsetof(struct ocfs2_dinode, id2.i_chain.cl_recs);
+
+	return size / sizeof(struct ocfs2_chain_rec);
+}
+
+static inline int ocfs2_extent_recs_per_eb(int blocksize)
+{
+	int size;
+
+	size = blocksize -
+		offsetof(struct ocfs2_extent_block, h_list.l_recs);
+
+	return size / sizeof(struct ocfs2_extent_rec);
+}
+
+static inline int ocfs2_local_alloc_size(int blocksize)
+{
+	int size;
+
+	size = blocksize -
+		offsetof(struct ocfs2_dinode, id2.i_lab.la_bitmap);
+
+	return size;
+}
+
+static inline int ocfs2_group_bitmap_size(int blocksize)
+{
+	int size;
+
+	size = blocksize -
+		offsetof(struct ocfs2_group_desc, bg_bitmap);
+
+	return size;
+}
+
+static inline int ocfs2_truncate_recs_per_inode(int blocksize)
+{
+	int size;
+
+	size = blocksize -
+		offsetof(struct ocfs2_dinode, id2.i_dealloc.tl_recs);
+
+	return size / sizeof(struct ocfs2_truncate_rec);
+}
+#endif  /* __KERNEL__ */
+
+
+static inline int ocfs2_system_inode_is_global(int type)
+{
+	return ((type >= 0) &&
+		(type <= OCFS2_LAST_GLOBAL_SYSTEM_INODE));
+}
+
+static inline int ocfs2_sprintf_system_inode_name(char *buf, int len,
+						  int type, int slot)
+{
+	int chars;
+
+        /*
+         * Global system inodes can only have one copy.  Everything
+         * after OCFS2_LAST_GLOBAL_SYSTEM_INODE in the system inode
+         * list has a copy per slot.
+         */
+	if (type <= OCFS2_LAST_GLOBAL_SYSTEM_INODE)
+		chars = snprintf(buf, len,
+				 ocfs2_system_inodes[type].si_name);
+	else
+		chars = snprintf(buf, len,
+				 ocfs2_system_inodes[type].si_name,
+				 slot);
+
+	return chars;
+}
+
+static inline void ocfs2_set_de_type(struct ocfs2_dir_entry *de,
+				    umode_t mode)
+{
+	de->file_type = ocfs2_type_by_mode[(mode & S_IFMT)>>S_SHIFT];
+}
+
+#endif  /* _OCFS2_FS_H */
+
diff --git a/fs/ocfs2/ocfs2_lockid.h b/fs/ocfs2/ocfs2_lockid.h
new file mode 100644
index 00000000000..7dd9e1e705b
--- /dev/null
+++ b/fs/ocfs2/ocfs2_lockid.h
@@ -0,0 +1,73 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ocfs2_lockid.h
+ *
+ * Defines OCFS2 lockid bits.
+ *
+ * Copyright (C) 2002, 2005 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_LOCKID_H
+#define OCFS2_LOCKID_H
+
+/* lock ids are made up in the following manner:
+ * name[0]     --> type
+ * name[1-6]   --> 6 pad characters, reserved for now
+ * name[7-22]  --> block number, expressed in hex as 16 chars
+ * name[23-30] --> i_generation, expressed in hex 8 chars
+ * name[31]    --> '\0' */
+#define OCFS2_LOCK_ID_MAX_LEN  32
+#define OCFS2_LOCK_ID_PAD "000000"
+
+enum ocfs2_lock_type {
+	OCFS2_LOCK_TYPE_META = 0,
+	OCFS2_LOCK_TYPE_DATA,
+	OCFS2_LOCK_TYPE_SUPER,
+	OCFS2_LOCK_TYPE_RENAME,
+	OCFS2_LOCK_TYPE_RW,
+	OCFS2_NUM_LOCK_TYPES
+};
+
+static inline char ocfs2_lock_type_char(enum ocfs2_lock_type type)
+{
+	char c;
+	switch (type) {
+		case OCFS2_LOCK_TYPE_META:
+			c = 'M';
+			break;
+		case OCFS2_LOCK_TYPE_DATA:
+			c = 'D';
+			break;
+		case OCFS2_LOCK_TYPE_SUPER:
+			c = 'S';
+			break;
+		case OCFS2_LOCK_TYPE_RENAME:
+			c = 'R';
+			break;
+		case OCFS2_LOCK_TYPE_RW:
+			c = 'W';
+			break;
+		default:
+			c = '\0';
+	}
+
+	return c;
+}
+
+#endif  /* OCFS2_LOCKID_H */
diff --git a/fs/ocfs2/slot_map.c b/fs/ocfs2/slot_map.c
new file mode 100644
index 00000000000..871627961d6
--- /dev/null
+++ b/fs/ocfs2/slot_map.c
@@ -0,0 +1,303 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * slot_map.c
+ *
+ *
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/smp_lock.h>
+
+#define MLOG_MASK_PREFIX ML_SUPER
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "dlmglue.h"
+#include "extent_map.h"
+#include "heartbeat.h"
+#include "inode.h"
+#include "slot_map.h"
+#include "super.h"
+#include "sysfile.h"
+
+#include "buffer_head_io.h"
+
+static s16 __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
+				    s16 global);
+static void __ocfs2_fill_slot(struct ocfs2_slot_info *si,
+			      s16 slot_num,
+			      s16 node_num);
+
+/* Use the slot information we've collected to create a map of mounted
+ * nodes. Should be holding an EX on super block. assumes slot info is
+ * up to date. Note that we call this *after* we find a slot, so our
+ * own node should be set in the map too... */
+void ocfs2_populate_mounted_map(struct ocfs2_super *osb)
+{
+	int i;
+	struct ocfs2_slot_info *si = osb->slot_info;
+
+	spin_lock(&si->si_lock);
+
+	for (i = 0; i < si->si_size; i++)
+		if (si->si_global_node_nums[i] != OCFS2_INVALID_SLOT)
+			ocfs2_node_map_set_bit(osb, &osb->mounted_map,
+					      si->si_global_node_nums[i]);
+
+	spin_unlock(&si->si_lock);
+}
+
+/* post the slot information on disk into our slot_info struct. */
+void ocfs2_update_slot_info(struct ocfs2_slot_info *si)
+{
+	int i;
+	__le16 *disk_info;
+
+	/* we don't read the slot block here as ocfs2_super_lock
+	 * should've made sure we have the most recent copy. */
+	spin_lock(&si->si_lock);
+	disk_info = (__le16 *) si->si_bh->b_data;
+
+	for (i = 0; i < si->si_size; i++)
+		si->si_global_node_nums[i] = le16_to_cpu(disk_info[i]);
+
+	spin_unlock(&si->si_lock);
+}
+
+/* post the our slot info stuff into it's destination bh and write it
+ * out. */
+int ocfs2_update_disk_slots(struct ocfs2_super *osb,
+			    struct ocfs2_slot_info *si)
+{
+	int status, i;
+	__le16 *disk_info = (__le16 *) si->si_bh->b_data;
+
+	spin_lock(&si->si_lock);
+	for (i = 0; i < si->si_size; i++)
+		disk_info[i] = cpu_to_le16(si->si_global_node_nums[i]);
+	spin_unlock(&si->si_lock);
+
+	status = ocfs2_write_block(osb, si->si_bh, si->si_inode);
+	if (status < 0)
+		mlog_errno(status);
+
+	return status;
+}
+
+/* try to find global node in the slot info. Returns
+ * OCFS2_INVALID_SLOT if nothing is found. */
+static s16 __ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
+				    s16 global)
+{
+	int i;
+	s16 ret = OCFS2_INVALID_SLOT;
+
+	for(i = 0; i < si->si_num_slots; i++) {
+		if (global == si->si_global_node_nums[i]) {
+			ret = (s16) i;
+			break;
+		}
+	}
+	return ret;
+}
+
+static s16 __ocfs2_find_empty_slot(struct ocfs2_slot_info *si)
+{
+	int i;
+	s16 ret = OCFS2_INVALID_SLOT;
+
+	for(i = 0; i < si->si_num_slots; i++) {
+		if (OCFS2_INVALID_SLOT == si->si_global_node_nums[i]) {
+			ret = (s16) i;
+			break;
+		}
+	}
+	return ret;
+}
+
+s16 ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
+			   s16 global)
+{
+	s16 ret;
+
+	spin_lock(&si->si_lock);
+	ret = __ocfs2_node_num_to_slot(si, global);
+	spin_unlock(&si->si_lock);
+	return ret;
+}
+
+static void __ocfs2_fill_slot(struct ocfs2_slot_info *si,
+			      s16 slot_num,
+			      s16 node_num)
+{
+	BUG_ON(slot_num == OCFS2_INVALID_SLOT);
+	BUG_ON(slot_num >= si->si_num_slots);
+	BUG_ON((node_num != O2NM_INVALID_NODE_NUM) &&
+	       (node_num >= O2NM_MAX_NODES));
+
+	si->si_global_node_nums[slot_num] = node_num;
+}
+
+void ocfs2_clear_slot(struct ocfs2_slot_info *si,
+		      s16 slot_num)
+{
+	spin_lock(&si->si_lock);
+	__ocfs2_fill_slot(si, slot_num, OCFS2_INVALID_SLOT);
+	spin_unlock(&si->si_lock);
+}
+
+int ocfs2_init_slot_info(struct ocfs2_super *osb)
+{
+	int status, i;
+	u64 blkno;
+	struct inode *inode = NULL;
+	struct buffer_head *bh = NULL;
+	struct ocfs2_slot_info *si;
+
+	si = kcalloc(1, sizeof(struct ocfs2_slot_info), GFP_KERNEL);
+	if (!si) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	spin_lock_init(&si->si_lock);
+	si->si_num_slots = osb->max_slots;
+	si->si_size = OCFS2_MAX_SLOTS;
+
+	for(i = 0; i < si->si_num_slots; i++)
+		si->si_global_node_nums[i] = OCFS2_INVALID_SLOT;
+
+	inode = ocfs2_get_system_file_inode(osb, SLOT_MAP_SYSTEM_INODE,
+					    OCFS2_INVALID_SLOT);
+	if (!inode) {
+		status = -EINVAL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_extent_map_get_blocks(inode, 0ULL, 1, &blkno, NULL);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_read_block(osb, blkno, &bh, 0, inode);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	si->si_inode = inode;
+	si->si_bh = bh;
+	osb->slot_info = si;
+bail:
+	if (status < 0 && si)
+		ocfs2_free_slot_info(si);
+
+	return status;
+}
+
+void ocfs2_free_slot_info(struct ocfs2_slot_info *si)
+{
+	if (si->si_inode)
+		iput(si->si_inode);
+	if (si->si_bh)
+		brelse(si->si_bh);
+	kfree(si);
+}
+
+int ocfs2_find_slot(struct ocfs2_super *osb)
+{
+	int status;
+	s16 slot;
+	struct ocfs2_slot_info *si;
+
+	mlog_entry_void();
+
+	si = osb->slot_info;
+
+	ocfs2_update_slot_info(si);
+
+	spin_lock(&si->si_lock);
+	/* search for ourselves first and take the slot if it already
+	 * exists. Perhaps we need to mark this in a variable for our
+	 * own journal recovery? Possibly not, though we certainly
+	 * need to warn to the user */
+	slot = __ocfs2_node_num_to_slot(si, osb->node_num);
+	if (slot == OCFS2_INVALID_SLOT) {
+		/* if no slot yet, then just take 1st available
+		 * one. */
+		slot = __ocfs2_find_empty_slot(si);
+		if (slot == OCFS2_INVALID_SLOT) {
+			spin_unlock(&si->si_lock);
+			mlog(ML_ERROR, "no free slots available!\n");
+			status = -EINVAL;
+			goto bail;
+		}
+	} else
+		mlog(ML_NOTICE, "slot %d is already allocated to this node!\n",
+		     slot);
+
+	__ocfs2_fill_slot(si, slot, osb->node_num);
+	osb->slot_num = slot;
+	spin_unlock(&si->si_lock);
+
+	mlog(ML_NOTICE, "taking node slot %d\n", osb->slot_num);
+
+	status = ocfs2_update_disk_slots(osb, si);
+	if (status < 0)
+		mlog_errno(status);
+
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+void ocfs2_put_slot(struct ocfs2_super *osb)
+{
+	int status;
+	struct ocfs2_slot_info *si = osb->slot_info;
+
+	if (!si)
+		return;
+
+	ocfs2_update_slot_info(si);
+
+	spin_lock(&si->si_lock);
+	__ocfs2_fill_slot(si, osb->slot_num, OCFS2_INVALID_SLOT);
+	osb->slot_num = OCFS2_INVALID_SLOT;
+	spin_unlock(&si->si_lock);
+
+	status = ocfs2_update_disk_slots(osb, si);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+bail:
+	osb->slot_info = NULL;
+	ocfs2_free_slot_info(si);
+}
+
diff --git a/fs/ocfs2/slot_map.h b/fs/ocfs2/slot_map.h
new file mode 100644
index 00000000000..d8c8ceed031
--- /dev/null
+++ b/fs/ocfs2/slot_map.h
@@ -0,0 +1,66 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * slotmap.h
+ *
+ * description here
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+
+#ifndef SLOTMAP_H
+#define SLOTMAP_H
+
+struct ocfs2_slot_info {
+	spinlock_t si_lock;
+
+       	struct inode *si_inode;
+	struct buffer_head *si_bh;
+	unsigned int si_num_slots;
+	unsigned int si_size;
+	s16 si_global_node_nums[OCFS2_MAX_SLOTS];
+};
+
+int ocfs2_init_slot_info(struct ocfs2_super *osb);
+void ocfs2_free_slot_info(struct ocfs2_slot_info *si);
+
+int ocfs2_find_slot(struct ocfs2_super *osb);
+void ocfs2_put_slot(struct ocfs2_super *osb);
+
+void ocfs2_update_slot_info(struct ocfs2_slot_info *si);
+int ocfs2_update_disk_slots(struct ocfs2_super *osb,
+			    struct ocfs2_slot_info *si);
+
+s16 ocfs2_node_num_to_slot(struct ocfs2_slot_info *si,
+			   s16 global);
+void ocfs2_clear_slot(struct ocfs2_slot_info *si,
+		      s16 slot_num);
+
+void ocfs2_populate_mounted_map(struct ocfs2_super *osb);
+
+static inline int ocfs2_is_empty_slot(struct ocfs2_slot_info *si,
+				      int slot_num)
+{
+	BUG_ON(slot_num == OCFS2_INVALID_SLOT);
+	assert_spin_locked(&si->si_lock);
+
+	return si->si_global_node_nums[slot_num] == OCFS2_INVALID_SLOT;
+}
+
+#endif
diff --git a/fs/ocfs2/suballoc.c b/fs/ocfs2/suballoc.c
new file mode 100644
index 00000000000..c46c164aefb
--- /dev/null
+++ b/fs/ocfs2/suballoc.c
@@ -0,0 +1,1651 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * suballoc.c
+ *
+ * metadata alloc and free
+ * Inspired by ext3 block groups.
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+
+#define MLOG_MASK_PREFIX ML_DISK_ALLOC
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "dlmglue.h"
+#include "inode.h"
+#include "journal.h"
+#include "localalloc.h"
+#include "suballoc.h"
+#include "super.h"
+#include "sysfile.h"
+#include "uptodate.h"
+
+#include "buffer_head_io.h"
+
+static inline void ocfs2_debug_bg(struct ocfs2_group_desc *bg);
+static inline void ocfs2_debug_suballoc_inode(struct ocfs2_dinode *fe);
+static inline u16 ocfs2_find_victim_chain(struct ocfs2_chain_list *cl);
+static int ocfs2_block_group_fill(struct ocfs2_journal_handle *handle,
+				  struct inode *alloc_inode,
+				  struct buffer_head *bg_bh,
+				  u64 group_blkno,
+				  u16 my_chain,
+				  struct ocfs2_chain_list *cl);
+static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
+				   struct inode *alloc_inode,
+				   struct buffer_head *bh);
+
+static int ocfs2_reserve_suballoc_bits(struct ocfs2_super *osb,
+				       struct ocfs2_alloc_context *ac);
+
+static int ocfs2_cluster_group_search(struct inode *inode,
+				      struct buffer_head *group_bh,
+				      u32 bits_wanted, u32 min_bits,
+				      u16 *bit_off, u16 *bits_found);
+static int ocfs2_block_group_search(struct inode *inode,
+				    struct buffer_head *group_bh,
+				    u32 bits_wanted, u32 min_bits,
+				    u16 *bit_off, u16 *bits_found);
+static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
+			      u32 bits_wanted,
+			      u32 min_bits,
+			      u16 *bit_off,
+			      unsigned int *num_bits,
+			      u64 *bg_blkno);
+static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
+				     struct ocfs2_alloc_context *ac,
+				     u32 bits_wanted,
+				     u32 min_bits,
+				     u16 *bit_off,
+				     unsigned int *num_bits,
+				     u64 *bg_blkno);
+static int ocfs2_test_bg_bit_allocatable(struct buffer_head *bg_bh,
+					 int nr);
+static int ocfs2_block_group_find_clear_bits(struct ocfs2_super *osb,
+					     struct buffer_head *bg_bh,
+					     unsigned int bits_wanted,
+					     u16 *bit_off,
+					     u16 *bits_found);
+static inline int ocfs2_block_group_set_bits(struct ocfs2_journal_handle *handle,
+					     struct inode *alloc_inode,
+					     struct ocfs2_group_desc *bg,
+					     struct buffer_head *group_bh,
+					     unsigned int bit_off,
+					     unsigned int num_bits);
+static inline int ocfs2_block_group_clear_bits(struct ocfs2_journal_handle *handle,
+					       struct inode *alloc_inode,
+					       struct ocfs2_group_desc *bg,
+					       struct buffer_head *group_bh,
+					       unsigned int bit_off,
+					       unsigned int num_bits);
+
+static int ocfs2_relink_block_group(struct ocfs2_journal_handle *handle,
+				    struct inode *alloc_inode,
+				    struct buffer_head *fe_bh,
+				    struct buffer_head *bg_bh,
+				    struct buffer_head *prev_bg_bh,
+				    u16 chain);
+static inline int ocfs2_block_group_reasonably_empty(struct ocfs2_group_desc *bg,
+						     u32 wanted);
+static int ocfs2_free_suballoc_bits(struct ocfs2_journal_handle *handle,
+				    struct inode *alloc_inode,
+				    struct buffer_head *alloc_bh,
+				    unsigned int start_bit,
+				    u64 bg_blkno,
+				    unsigned int count);
+static inline u64 ocfs2_which_suballoc_group(u64 block,
+					     unsigned int bit);
+static inline u32 ocfs2_desc_bitmap_to_cluster_off(struct inode *inode,
+						   u64 bg_blkno,
+						   u16 bg_bit_off);
+static inline u64 ocfs2_which_cluster_group(struct inode *inode,
+					    u32 cluster);
+static inline void ocfs2_block_to_cluster_group(struct inode *inode,
+						u64 data_blkno,
+						u64 *bg_blkno,
+						u16 *bg_bit_off);
+
+void ocfs2_free_alloc_context(struct ocfs2_alloc_context *ac)
+{
+	if (ac->ac_inode)
+		iput(ac->ac_inode);
+	if (ac->ac_bh)
+		brelse(ac->ac_bh);
+	kfree(ac);
+}
+
+static u32 ocfs2_bits_per_group(struct ocfs2_chain_list *cl)
+{
+	return (u32)le16_to_cpu(cl->cl_cpg) * (u32)le16_to_cpu(cl->cl_bpc);
+}
+
+static int ocfs2_block_group_fill(struct ocfs2_journal_handle *handle,
+				  struct inode *alloc_inode,
+				  struct buffer_head *bg_bh,
+				  u64 group_blkno,
+				  u16 my_chain,
+				  struct ocfs2_chain_list *cl)
+{
+	int status = 0;
+	struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data;
+	struct super_block * sb = alloc_inode->i_sb;
+
+	mlog_entry_void();
+
+	if (((unsigned long long) bg_bh->b_blocknr) != group_blkno) {
+		ocfs2_error(alloc_inode->i_sb, "group block (%"MLFu64") "
+			    "!= b_blocknr (%llu)", group_blkno,
+			    (unsigned long long) bg_bh->b_blocknr);
+		status = -EIO;
+		goto bail;
+	}
+
+	status = ocfs2_journal_access(handle,
+				      alloc_inode,
+				      bg_bh,
+				      OCFS2_JOURNAL_ACCESS_CREATE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	memset(bg, 0, sb->s_blocksize);
+	strcpy(bg->bg_signature, OCFS2_GROUP_DESC_SIGNATURE);
+	bg->bg_generation = cpu_to_le32(OCFS2_SB(sb)->fs_generation);
+	bg->bg_size = cpu_to_le16(ocfs2_group_bitmap_size(sb));
+	bg->bg_bits = cpu_to_le16(ocfs2_bits_per_group(cl));
+	bg->bg_chain = cpu_to_le16(my_chain);
+	bg->bg_next_group = cl->cl_recs[my_chain].c_blkno;
+	bg->bg_parent_dinode = cpu_to_le64(OCFS2_I(alloc_inode)->ip_blkno);
+	bg->bg_blkno = cpu_to_le64(group_blkno);
+	/* set the 1st bit in the bitmap to account for the descriptor block */
+	ocfs2_set_bit(0, (unsigned long *)bg->bg_bitmap);
+	bg->bg_free_bits_count = cpu_to_le16(le16_to_cpu(bg->bg_bits) - 1);
+
+	status = ocfs2_journal_dirty(handle, bg_bh);
+	if (status < 0)
+		mlog_errno(status);
+
+	/* There is no need to zero out or otherwise initialize the
+	 * other blocks in a group - All valid FS metadata in a block
+	 * group stores the superblock fs_generation value at
+	 * allocation time. */
+
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+static inline u16 ocfs2_find_smallest_chain(struct ocfs2_chain_list *cl)
+{
+	u16 curr, best;
+
+	best = curr = 0;
+	while (curr < le16_to_cpu(cl->cl_count)) {
+		if (le32_to_cpu(cl->cl_recs[best].c_total) >
+		    le32_to_cpu(cl->cl_recs[curr].c_total))
+			best = curr;
+		curr++;
+	}
+	return best;
+}
+
+/*
+ * We expect the block group allocator to already be locked.
+ */
+static int ocfs2_block_group_alloc(struct ocfs2_super *osb,
+				   struct inode *alloc_inode,
+				   struct buffer_head *bh)
+{
+	int status, credits;
+	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) bh->b_data;
+	struct ocfs2_chain_list *cl;
+	struct ocfs2_alloc_context *ac = NULL;
+	struct ocfs2_journal_handle *handle = NULL;
+	u32 bit_off, num_bits;
+	u16 alloc_rec;
+	u64 bg_blkno;
+	struct buffer_head *bg_bh = NULL;
+	struct ocfs2_group_desc *bg;
+
+	BUG_ON(ocfs2_is_cluster_bitmap(alloc_inode));
+
+	mlog_entry_void();
+
+	handle = ocfs2_alloc_handle(osb);
+	if (!handle) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	cl = &fe->id2.i_chain;
+	status = ocfs2_reserve_clusters(osb,
+					handle,
+					le16_to_cpu(cl->cl_cpg),
+					&ac);
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	credits = ocfs2_calc_group_alloc_credits(osb->sb,
+						 le16_to_cpu(cl->cl_cpg));
+	handle = ocfs2_start_trans(osb, handle, credits);
+	if (IS_ERR(handle)) {
+		status = PTR_ERR(handle);
+		handle = NULL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_claim_clusters(osb,
+				      handle,
+				      ac,
+				      le16_to_cpu(cl->cl_cpg),
+				      &bit_off,
+				      &num_bits);
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	alloc_rec = ocfs2_find_smallest_chain(cl);
+
+	/* setup the group */
+	bg_blkno = ocfs2_clusters_to_blocks(osb->sb, bit_off);
+	mlog(0, "new descriptor, record %u, at block %"MLFu64"\n",
+	     alloc_rec, bg_blkno);
+
+	bg_bh = sb_getblk(osb->sb, bg_blkno);
+	if (!bg_bh) {
+		status = -EIO;
+		mlog_errno(status);
+		goto bail;
+	}
+	ocfs2_set_new_buffer_uptodate(alloc_inode, bg_bh);
+
+	status = ocfs2_block_group_fill(handle,
+					alloc_inode,
+					bg_bh,
+					bg_blkno,
+					alloc_rec,
+					cl);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	bg = (struct ocfs2_group_desc *) bg_bh->b_data;
+
+	status = ocfs2_journal_access(handle, alloc_inode,
+				      bh, OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	le32_add_cpu(&cl->cl_recs[alloc_rec].c_free,
+		     le16_to_cpu(bg->bg_free_bits_count));
+	le32_add_cpu(&cl->cl_recs[alloc_rec].c_total, le16_to_cpu(bg->bg_bits));
+	cl->cl_recs[alloc_rec].c_blkno  = cpu_to_le64(bg_blkno);
+	if (le16_to_cpu(cl->cl_next_free_rec) < le16_to_cpu(cl->cl_count))
+		le16_add_cpu(&cl->cl_next_free_rec, 1);
+
+	le32_add_cpu(&fe->id1.bitmap1.i_used, le16_to_cpu(bg->bg_bits) -
+					le16_to_cpu(bg->bg_free_bits_count));
+	le32_add_cpu(&fe->id1.bitmap1.i_total, le16_to_cpu(bg->bg_bits));
+	le32_add_cpu(&fe->i_clusters, le16_to_cpu(cl->cl_cpg));
+
+	status = ocfs2_journal_dirty(handle, bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	spin_lock(&OCFS2_I(alloc_inode)->ip_lock);
+	OCFS2_I(alloc_inode)->ip_clusters = le32_to_cpu(fe->i_clusters);
+	fe->i_size = cpu_to_le64(ocfs2_clusters_to_bytes(alloc_inode->i_sb,
+					     le32_to_cpu(fe->i_clusters)));
+	spin_unlock(&OCFS2_I(alloc_inode)->ip_lock);
+	i_size_write(alloc_inode, le64_to_cpu(fe->i_size));
+	alloc_inode->i_blocks =
+		ocfs2_align_bytes_to_sectors(i_size_read(alloc_inode));
+
+	status = 0;
+bail:
+	if (handle)
+		ocfs2_commit_trans(handle);
+
+	if (ac)
+		ocfs2_free_alloc_context(ac);
+
+	if (bg_bh)
+		brelse(bg_bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_reserve_suballoc_bits(struct ocfs2_super *osb,
+				       struct ocfs2_alloc_context *ac)
+{
+	int status;
+	u32 bits_wanted = ac->ac_bits_wanted;
+	struct inode *alloc_inode = ac->ac_inode;
+	struct buffer_head *bh = NULL;
+	struct ocfs2_journal_handle *handle = ac->ac_handle;
+	struct ocfs2_dinode *fe;
+	u32 free_bits;
+
+	mlog_entry_void();
+
+	BUG_ON(handle->flags & OCFS2_HANDLE_STARTED);
+
+	ocfs2_handle_add_inode(handle, alloc_inode);
+	status = ocfs2_meta_lock(alloc_inode, handle, &bh, 1);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	fe = (struct ocfs2_dinode *) bh->b_data;
+	if (!OCFS2_IS_VALID_DINODE(fe)) {
+		OCFS2_RO_ON_INVALID_DINODE(alloc_inode->i_sb, fe);
+		status = -EIO;
+		goto bail;
+	}
+	if (!(fe->i_flags & cpu_to_le32(OCFS2_CHAIN_FL))) {
+		ocfs2_error(alloc_inode->i_sb, "Invalid chain allocator "
+			    "# %"MLFu64, le64_to_cpu(fe->i_blkno));
+		status = -EIO;
+		goto bail;
+	}
+
+	free_bits = le32_to_cpu(fe->id1.bitmap1.i_total) -
+		le32_to_cpu(fe->id1.bitmap1.i_used);
+
+	if (bits_wanted > free_bits) {
+		/* cluster bitmap never grows */
+		if (ocfs2_is_cluster_bitmap(alloc_inode)) {
+			mlog(0, "Disk Full: wanted=%u, free_bits=%u\n",
+			     bits_wanted, free_bits);
+			status = -ENOSPC;
+			goto bail;
+		}
+
+		status = ocfs2_block_group_alloc(osb, alloc_inode, bh);
+		if (status < 0) {
+			if (status != -ENOSPC)
+				mlog_errno(status);
+			goto bail;
+		}
+		atomic_inc(&osb->alloc_stats.bg_extends);
+
+		/* You should never ask for this much metadata */
+		BUG_ON(bits_wanted >
+		       (le32_to_cpu(fe->id1.bitmap1.i_total)
+			- le32_to_cpu(fe->id1.bitmap1.i_used)));
+	}
+
+	get_bh(bh);
+	ac->ac_bh = bh;
+bail:
+	if (bh)
+		brelse(bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+int ocfs2_reserve_new_metadata(struct ocfs2_super *osb,
+			       struct ocfs2_journal_handle *handle,
+			       struct ocfs2_dinode *fe,
+			       struct ocfs2_alloc_context **ac)
+{
+	int status;
+	struct inode *alloc_inode = NULL;
+
+	*ac = kcalloc(1, sizeof(struct ocfs2_alloc_context), GFP_KERNEL);
+	if (!(*ac)) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	(*ac)->ac_bits_wanted = ocfs2_extend_meta_needed(fe);
+	(*ac)->ac_handle = handle;
+	(*ac)->ac_which = OCFS2_AC_USE_META;
+
+#ifndef OCFS2_USE_ALL_METADATA_SUBALLOCATORS
+	alloc_inode = ocfs2_get_system_file_inode(osb,
+						  EXTENT_ALLOC_SYSTEM_INODE,
+						  0);
+#else
+	alloc_inode = ocfs2_get_system_file_inode(osb,
+						  EXTENT_ALLOC_SYSTEM_INODE,
+						  osb->slot_num);
+#endif
+	if (!alloc_inode) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	(*ac)->ac_inode = igrab(alloc_inode);
+	(*ac)->ac_group_search = ocfs2_block_group_search;
+
+	status = ocfs2_reserve_suballoc_bits(osb, (*ac));
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	status = 0;
+bail:
+	if ((status < 0) && *ac) {
+		ocfs2_free_alloc_context(*ac);
+		*ac = NULL;
+	}
+
+	if (alloc_inode)
+		iput(alloc_inode);
+
+	mlog_exit(status);
+	return status;
+}
+
+int ocfs2_reserve_new_inode(struct ocfs2_super *osb,
+			    struct ocfs2_journal_handle *handle,
+			    struct ocfs2_alloc_context **ac)
+{
+	int status;
+	struct inode *alloc_inode = NULL;
+
+	*ac = kcalloc(1, sizeof(struct ocfs2_alloc_context), GFP_KERNEL);
+	if (!(*ac)) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	(*ac)->ac_bits_wanted = 1;
+	(*ac)->ac_handle = handle;
+	(*ac)->ac_which = OCFS2_AC_USE_INODE;
+
+	alloc_inode = ocfs2_get_system_file_inode(osb,
+						  INODE_ALLOC_SYSTEM_INODE,
+						  osb->slot_num);
+	if (!alloc_inode) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	(*ac)->ac_inode = igrab(alloc_inode);
+	(*ac)->ac_group_search = ocfs2_block_group_search;
+
+	status = ocfs2_reserve_suballoc_bits(osb, *ac);
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	status = 0;
+bail:
+	if ((status < 0) && *ac) {
+		ocfs2_free_alloc_context(*ac);
+		*ac = NULL;
+	}
+
+	if (alloc_inode)
+		iput(alloc_inode);
+
+	mlog_exit(status);
+	return status;
+}
+
+/* local alloc code has to do the same thing, so rather than do this
+ * twice.. */
+int ocfs2_reserve_cluster_bitmap_bits(struct ocfs2_super *osb,
+				      struct ocfs2_alloc_context *ac)
+{
+	int status;
+
+	ac->ac_inode = ocfs2_get_system_file_inode(osb,
+						   GLOBAL_BITMAP_SYSTEM_INODE,
+						   OCFS2_INVALID_SLOT);
+	if (!ac->ac_inode) {
+		status = -EINVAL;
+		mlog(ML_ERROR, "Could not get bitmap inode!\n");
+		goto bail;
+	}
+	ac->ac_which = OCFS2_AC_USE_MAIN;
+	ac->ac_group_search = ocfs2_cluster_group_search;
+
+	status = ocfs2_reserve_suballoc_bits(osb, ac);
+	if (status < 0 && status != -ENOSPC)
+		mlog_errno(status);
+bail:
+	return status;
+}
+
+/* Callers don't need to care which bitmap (local alloc or main) to
+ * use so we figure it out for them, but unfortunately this clutters
+ * things a bit. */
+int ocfs2_reserve_clusters(struct ocfs2_super *osb,
+			   struct ocfs2_journal_handle *handle,
+			   u32 bits_wanted,
+			   struct ocfs2_alloc_context **ac)
+{
+	int status;
+
+	mlog_entry_void();
+
+	BUG_ON(!handle);
+
+	*ac = kcalloc(1, sizeof(struct ocfs2_alloc_context), GFP_KERNEL);
+	if (!(*ac)) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	(*ac)->ac_bits_wanted = bits_wanted;
+	(*ac)->ac_handle = handle;
+
+	status = -ENOSPC;
+	if (ocfs2_alloc_should_use_local(osb, bits_wanted)) {
+		status = ocfs2_reserve_local_alloc_bits(osb,
+							handle,
+							bits_wanted,
+							*ac);
+		if ((status < 0) && (status != -ENOSPC)) {
+			mlog_errno(status);
+			goto bail;
+		} else if (status == -ENOSPC) {
+			/* reserve_local_bits will return enospc with
+			 * the local alloc inode still locked, so we
+			 * can change this safely here. */
+			mlog(0, "Disabling local alloc\n");
+			/* We set to OCFS2_LA_DISABLED so that umount
+			 * can clean up what's left of the local
+			 * allocation */
+			osb->local_alloc_state = OCFS2_LA_DISABLED;
+		}
+	}
+
+	if (status == -ENOSPC) {
+		status = ocfs2_reserve_cluster_bitmap_bits(osb, *ac);
+		if (status < 0) {
+			if (status != -ENOSPC)
+				mlog_errno(status);
+			goto bail;
+		}
+	}
+
+	status = 0;
+bail:
+	if ((status < 0) && *ac) {
+		ocfs2_free_alloc_context(*ac);
+		*ac = NULL;
+	}
+
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * More or less lifted from ext3. I'll leave their description below:
+ *
+ * "For ext3 allocations, we must not reuse any blocks which are
+ * allocated in the bitmap buffer's "last committed data" copy.  This
+ * prevents deletes from freeing up the page for reuse until we have
+ * committed the delete transaction.
+ *
+ * If we didn't do this, then deleting something and reallocating it as
+ * data would allow the old block to be overwritten before the
+ * transaction committed (because we force data to disk before commit).
+ * This would lead to corruption if we crashed between overwriting the
+ * data and committing the delete.
+ *
+ * @@@ We may want to make this allocation behaviour conditional on
+ * data-writes at some point, and disable it for metadata allocations or
+ * sync-data inodes."
+ *
+ * Note: OCFS2 already does this differently for metadata vs data
+ * allocations, as those bitmaps are seperate and undo access is never
+ * called on a metadata group descriptor.
+ */
+static int ocfs2_test_bg_bit_allocatable(struct buffer_head *bg_bh,
+					 int nr)
+{
+	struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data;
+
+	if (ocfs2_test_bit(nr, (unsigned long *)bg->bg_bitmap))
+		return 0;
+	if (!buffer_jbd(bg_bh) || !bh2jh(bg_bh)->b_committed_data)
+		return 1;
+
+	bg = (struct ocfs2_group_desc *) bh2jh(bg_bh)->b_committed_data;
+	return !ocfs2_test_bit(nr, (unsigned long *)bg->bg_bitmap);
+}
+
+static int ocfs2_block_group_find_clear_bits(struct ocfs2_super *osb,
+					     struct buffer_head *bg_bh,
+					     unsigned int bits_wanted,
+					     u16 *bit_off,
+					     u16 *bits_found)
+{
+	void *bitmap;
+	u16 best_offset, best_size;
+	int offset, start, found, status = 0;
+	struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data;
+
+	if (!OCFS2_IS_VALID_GROUP_DESC(bg)) {
+		OCFS2_RO_ON_INVALID_GROUP_DESC(osb->sb, bg);
+		return -EIO;
+	}
+
+	found = start = best_offset = best_size = 0;
+	bitmap = bg->bg_bitmap;
+
+	while((offset = ocfs2_find_next_zero_bit(bitmap,
+						 le16_to_cpu(bg->bg_bits),
+						 start)) != -1) {
+		if (offset == le16_to_cpu(bg->bg_bits))
+			break;
+
+		if (!ocfs2_test_bg_bit_allocatable(bg_bh, offset)) {
+			/* We found a zero, but we can't use it as it
+			 * hasn't been put to disk yet! */
+			found = 0;
+			start = offset + 1;
+		} else if (offset == start) {
+			/* we found a zero */
+			found++;
+			/* move start to the next bit to test */
+			start++;
+		} else {
+			/* got a zero after some ones */
+			found = 1;
+			start = offset + 1;
+		}
+		if (found > best_size) {
+			best_size = found;
+			best_offset = start - found;
+		}
+		/* we got everything we needed */
+		if (found == bits_wanted) {
+			/* mlog(0, "Found it all!\n"); */
+			break;
+		}
+	}
+
+	/* XXX: I think the first clause is equivalent to the second
+	 * 	- jlbec */
+	if (found == bits_wanted) {
+		*bit_off = start - found;
+		*bits_found = found;
+	} else if (best_size) {
+		*bit_off = best_offset;
+		*bits_found = best_size;
+	} else {
+		status = -ENOSPC;
+		/* No error log here -- see the comment above
+		 * ocfs2_test_bg_bit_allocatable */
+	}
+
+	return status;
+}
+
+static inline int ocfs2_block_group_set_bits(struct ocfs2_journal_handle *handle,
+					     struct inode *alloc_inode,
+					     struct ocfs2_group_desc *bg,
+					     struct buffer_head *group_bh,
+					     unsigned int bit_off,
+					     unsigned int num_bits)
+{
+	int status;
+	void *bitmap = bg->bg_bitmap;
+	int journal_type = OCFS2_JOURNAL_ACCESS_WRITE;
+
+	mlog_entry_void();
+
+	if (!OCFS2_IS_VALID_GROUP_DESC(bg)) {
+		OCFS2_RO_ON_INVALID_GROUP_DESC(alloc_inode->i_sb, bg);
+		status = -EIO;
+		goto bail;
+	}
+	BUG_ON(le16_to_cpu(bg->bg_free_bits_count) < num_bits);
+
+	mlog(0, "block_group_set_bits: off = %u, num = %u\n", bit_off,
+	     num_bits);
+
+	if (ocfs2_is_cluster_bitmap(alloc_inode))
+		journal_type = OCFS2_JOURNAL_ACCESS_UNDO;
+
+	status = ocfs2_journal_access(handle,
+				      alloc_inode,
+				      group_bh,
+				      journal_type);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	le16_add_cpu(&bg->bg_free_bits_count, -num_bits);
+
+	while(num_bits--)
+		ocfs2_set_bit(bit_off++, bitmap);
+
+	status = ocfs2_journal_dirty(handle,
+				     group_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+/* find the one with the most empty bits */
+static inline u16 ocfs2_find_victim_chain(struct ocfs2_chain_list *cl)
+{
+	u16 curr, best;
+
+	BUG_ON(!cl->cl_next_free_rec);
+
+	best = curr = 0;
+	while (curr < le16_to_cpu(cl->cl_next_free_rec)) {
+		if (le32_to_cpu(cl->cl_recs[curr].c_free) >
+		    le32_to_cpu(cl->cl_recs[best].c_free))
+			best = curr;
+		curr++;
+	}
+
+	BUG_ON(best >= le16_to_cpu(cl->cl_next_free_rec));
+	return best;
+}
+
+static int ocfs2_relink_block_group(struct ocfs2_journal_handle *handle,
+				    struct inode *alloc_inode,
+				    struct buffer_head *fe_bh,
+				    struct buffer_head *bg_bh,
+				    struct buffer_head *prev_bg_bh,
+				    u16 chain)
+{
+	int status;
+	/* there is a really tiny chance the journal calls could fail,
+	 * but we wouldn't want inconsistent blocks in *any* case. */
+	u64 fe_ptr, bg_ptr, prev_bg_ptr;
+	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) fe_bh->b_data;
+	struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data;
+	struct ocfs2_group_desc *prev_bg = (struct ocfs2_group_desc *) prev_bg_bh->b_data;
+
+	if (!OCFS2_IS_VALID_DINODE(fe)) {
+		OCFS2_RO_ON_INVALID_DINODE(alloc_inode->i_sb, fe);
+		status = -EIO;
+		goto out;
+	}
+	if (!OCFS2_IS_VALID_GROUP_DESC(bg)) {
+		OCFS2_RO_ON_INVALID_GROUP_DESC(alloc_inode->i_sb, bg);
+		status = -EIO;
+		goto out;
+	}
+	if (!OCFS2_IS_VALID_GROUP_DESC(prev_bg)) {
+		OCFS2_RO_ON_INVALID_GROUP_DESC(alloc_inode->i_sb, prev_bg);
+		status = -EIO;
+		goto out;
+	}
+
+	mlog(0, "In suballoc %"MLFu64", chain %u, move group %"MLFu64" to "
+	     "top, prev = %"MLFu64"\n",
+	     fe->i_blkno, chain, bg->bg_blkno, prev_bg->bg_blkno);
+
+	fe_ptr = le64_to_cpu(fe->id2.i_chain.cl_recs[chain].c_blkno);
+	bg_ptr = le64_to_cpu(bg->bg_next_group);
+	prev_bg_ptr = le64_to_cpu(prev_bg->bg_next_group);
+
+	status = ocfs2_journal_access(handle, alloc_inode, prev_bg_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto out_rollback;
+	}
+
+	prev_bg->bg_next_group = bg->bg_next_group;
+
+	status = ocfs2_journal_dirty(handle, prev_bg_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto out_rollback;
+	}
+
+	status = ocfs2_journal_access(handle, alloc_inode, bg_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto out_rollback;
+	}
+
+	bg->bg_next_group = fe->id2.i_chain.cl_recs[chain].c_blkno;
+
+	status = ocfs2_journal_dirty(handle, bg_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto out_rollback;
+	}
+
+	status = ocfs2_journal_access(handle, alloc_inode, fe_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto out_rollback;
+	}
+
+	fe->id2.i_chain.cl_recs[chain].c_blkno = bg->bg_blkno;
+
+	status = ocfs2_journal_dirty(handle, fe_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto out_rollback;
+	}
+
+	status = 0;
+out_rollback:
+	if (status < 0) {
+		fe->id2.i_chain.cl_recs[chain].c_blkno = cpu_to_le64(fe_ptr);
+		bg->bg_next_group = cpu_to_le64(bg_ptr);
+		prev_bg->bg_next_group = cpu_to_le64(prev_bg_ptr);
+	}
+out:
+	mlog_exit(status);
+	return status;
+}
+
+static inline int ocfs2_block_group_reasonably_empty(struct ocfs2_group_desc *bg,
+						     u32 wanted)
+{
+	return le16_to_cpu(bg->bg_free_bits_count) > wanted;
+}
+
+/* return 0 on success, -ENOSPC to keep searching and any other < 0
+ * value on error. */
+static int ocfs2_cluster_group_search(struct inode *inode,
+				      struct buffer_head *group_bh,
+				      u32 bits_wanted, u32 min_bits,
+				      u16 *bit_off, u16 *bits_found)
+{
+	int search = -ENOSPC;
+	int ret;
+	struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) group_bh->b_data;
+	u16 tmp_off, tmp_found;
+
+	BUG_ON(!ocfs2_is_cluster_bitmap(inode));
+
+	if (bg->bg_free_bits_count) {
+		ret = ocfs2_block_group_find_clear_bits(OCFS2_SB(inode->i_sb),
+							group_bh, bits_wanted,
+							&tmp_off, &tmp_found);
+		if (ret)
+			return ret;
+
+		/* ocfs2_block_group_find_clear_bits() might
+		 * return success, but we still want to return
+		 * -ENOSPC unless it found the minimum number
+		 * of bits. */
+		if (min_bits <= tmp_found) {
+			*bit_off = tmp_off;
+			*bits_found = tmp_found;
+			search = 0; /* success */
+		}
+	}
+
+	return search;
+}
+
+static int ocfs2_block_group_search(struct inode *inode,
+				    struct buffer_head *group_bh,
+				    u32 bits_wanted, u32 min_bits,
+				    u16 *bit_off, u16 *bits_found)
+{
+	int ret = -ENOSPC;
+	struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) group_bh->b_data;
+
+	BUG_ON(min_bits != 1);
+	BUG_ON(ocfs2_is_cluster_bitmap(inode));
+
+	if (bg->bg_free_bits_count)
+		ret = ocfs2_block_group_find_clear_bits(OCFS2_SB(inode->i_sb),
+							group_bh, bits_wanted,
+							bit_off, bits_found);
+
+	return ret;
+}
+
+static int ocfs2_search_chain(struct ocfs2_alloc_context *ac,
+			      u32 bits_wanted,
+			      u32 min_bits,
+			      u16 *bit_off,
+			      unsigned int *num_bits,
+			      u64 *bg_blkno)
+{
+	int status;
+	u16 chain, tmp_bits;
+	u32 tmp_used;
+	u64 next_group;
+	struct ocfs2_journal_handle *handle = ac->ac_handle;
+	struct inode *alloc_inode = ac->ac_inode;
+	struct buffer_head *group_bh = NULL;
+	struct buffer_head *prev_group_bh = NULL;
+	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) ac->ac_bh->b_data;
+	struct ocfs2_chain_list *cl = (struct ocfs2_chain_list *) &fe->id2.i_chain;
+	struct ocfs2_group_desc *bg;
+
+	chain = ac->ac_chain;
+	mlog(0, "trying to alloc %u bits from chain %u, inode %"MLFu64"\n",
+	     bits_wanted, chain, OCFS2_I(alloc_inode)->ip_blkno);
+
+	status = ocfs2_read_block(OCFS2_SB(alloc_inode->i_sb),
+				  le64_to_cpu(cl->cl_recs[chain].c_blkno),
+				  &group_bh, OCFS2_BH_CACHED, alloc_inode);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+	bg = (struct ocfs2_group_desc *) group_bh->b_data;
+	if (!OCFS2_IS_VALID_GROUP_DESC(bg)) {
+		OCFS2_RO_ON_INVALID_GROUP_DESC(alloc_inode->i_sb, bg);
+		status = -EIO;
+		goto bail;
+	}
+
+	status = -ENOSPC;
+	/* for now, the chain search is a bit simplistic. We just use
+	 * the 1st group with any empty bits. */
+	while ((status = ac->ac_group_search(alloc_inode, group_bh,
+					     bits_wanted, min_bits, bit_off,
+					     &tmp_bits)) == -ENOSPC) {
+		if (!bg->bg_next_group)
+			break;
+
+		if (prev_group_bh) {
+			brelse(prev_group_bh);
+			prev_group_bh = NULL;
+		}
+		next_group = le64_to_cpu(bg->bg_next_group);
+		prev_group_bh = group_bh;
+		group_bh = NULL;
+		status = ocfs2_read_block(OCFS2_SB(alloc_inode->i_sb),
+					  next_group, &group_bh,
+					  OCFS2_BH_CACHED, alloc_inode);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+		bg = (struct ocfs2_group_desc *) group_bh->b_data;
+		if (!OCFS2_IS_VALID_GROUP_DESC(bg)) {
+			OCFS2_RO_ON_INVALID_GROUP_DESC(alloc_inode->i_sb, bg);
+			status = -EIO;
+			goto bail;
+		}
+	}
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	mlog(0, "alloc succeeds: we give %u bits from block group %"MLFu64"\n",
+	     tmp_bits, bg->bg_blkno);
+
+	*num_bits = tmp_bits;
+
+	BUG_ON(*num_bits == 0);
+
+	/*
+	 * Keep track of previous block descriptor read. When
+	 * we find a target, if we have read more than X
+	 * number of descriptors, and the target is reasonably
+	 * empty, relink him to top of his chain.
+	 *
+	 * We've read 0 extra blocks and only send one more to
+	 * the transaction, yet the next guy to search has a
+	 * much easier time.
+	 *
+	 * Do this *after* figuring out how many bits we're taking out
+	 * of our target group.
+	 */
+	if (ac->ac_allow_chain_relink &&
+	    (prev_group_bh) &&
+	    (ocfs2_block_group_reasonably_empty(bg, *num_bits))) {
+		status = ocfs2_relink_block_group(handle, alloc_inode,
+						  ac->ac_bh, group_bh,
+						  prev_group_bh, chain);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
+	}
+
+	/* Ok, claim our bits now: set the info on dinode, chainlist
+	 * and then the group */
+	status = ocfs2_journal_access(handle,
+				      alloc_inode,
+				      ac->ac_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	tmp_used = le32_to_cpu(fe->id1.bitmap1.i_used);
+	fe->id1.bitmap1.i_used = cpu_to_le32(*num_bits + tmp_used);
+	le32_add_cpu(&cl->cl_recs[chain].c_free, -(*num_bits));
+
+	status = ocfs2_journal_dirty(handle,
+				     ac->ac_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_block_group_set_bits(handle,
+					    alloc_inode,
+					    bg,
+					    group_bh,
+					    *bit_off,
+					    *num_bits);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	mlog(0, "Allocated %u bits from suballocator %"MLFu64"\n",
+	     *num_bits, fe->i_blkno);
+
+	*bg_blkno = le64_to_cpu(bg->bg_blkno);
+bail:
+	if (group_bh)
+		brelse(group_bh);
+	if (prev_group_bh)
+		brelse(prev_group_bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+/* will give out up to bits_wanted contiguous bits. */
+static int ocfs2_claim_suballoc_bits(struct ocfs2_super *osb,
+				     struct ocfs2_alloc_context *ac,
+				     u32 bits_wanted,
+				     u32 min_bits,
+				     u16 *bit_off,
+				     unsigned int *num_bits,
+				     u64 *bg_blkno)
+{
+	int status;
+	u16 victim, i;
+	struct ocfs2_chain_list *cl;
+	struct ocfs2_dinode *fe;
+
+	mlog_entry_void();
+
+	BUG_ON(ac->ac_bits_given >= ac->ac_bits_wanted);
+	BUG_ON(bits_wanted > (ac->ac_bits_wanted - ac->ac_bits_given));
+	BUG_ON(!ac->ac_bh);
+
+	fe = (struct ocfs2_dinode *) ac->ac_bh->b_data;
+	if (!OCFS2_IS_VALID_DINODE(fe)) {
+		OCFS2_RO_ON_INVALID_DINODE(osb->sb, fe);
+		status = -EIO;
+		goto bail;
+	}
+	if (le32_to_cpu(fe->id1.bitmap1.i_used) >=
+	    le32_to_cpu(fe->id1.bitmap1.i_total)) {
+		ocfs2_error(osb->sb, "Chain allocator dinode %"MLFu64" has %u"
+			    "used bits but only %u total.",
+			    le64_to_cpu(fe->i_blkno),
+			    le32_to_cpu(fe->id1.bitmap1.i_used),
+			    le32_to_cpu(fe->id1.bitmap1.i_total));
+		status = -EIO;
+		goto bail;
+	}
+
+	cl = (struct ocfs2_chain_list *) &fe->id2.i_chain;
+
+	victim = ocfs2_find_victim_chain(cl);
+	ac->ac_chain = victim;
+	ac->ac_allow_chain_relink = 1;
+
+	status = ocfs2_search_chain(ac, bits_wanted, min_bits, bit_off,
+				    num_bits, bg_blkno);
+	if (!status)
+		goto bail;
+	if (status < 0 && status != -ENOSPC) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	mlog(0, "Search of victim chain %u came up with nothing, "
+	     "trying all chains now.\n", victim);
+
+	/* If we didn't pick a good victim, then just default to
+	 * searching each chain in order. Don't allow chain relinking
+	 * because we only calculate enough journal credits for one
+	 * relink per alloc. */
+	ac->ac_allow_chain_relink = 0;
+	for (i = 0; i < le16_to_cpu(cl->cl_next_free_rec); i ++) {
+		if (i == victim)
+			continue;
+		if (!cl->cl_recs[i].c_free)
+			continue;
+
+		ac->ac_chain = i;
+		status = ocfs2_search_chain(ac, bits_wanted, min_bits,
+					    bit_off, num_bits,
+					    bg_blkno);
+		if (!status)
+			break;
+		if (status < 0 && status != -ENOSPC) {
+			mlog_errno(status);
+			goto bail;
+		}
+	}
+bail:
+
+	mlog_exit(status);
+	return status;
+}
+
+int ocfs2_claim_metadata(struct ocfs2_super *osb,
+			 struct ocfs2_journal_handle *handle,
+			 struct ocfs2_alloc_context *ac,
+			 u32 bits_wanted,
+			 u16 *suballoc_bit_start,
+			 unsigned int *num_bits,
+			 u64 *blkno_start)
+{
+	int status;
+	u64 bg_blkno;
+
+	BUG_ON(!ac);
+	BUG_ON(ac->ac_bits_wanted < (ac->ac_bits_given + bits_wanted));
+	BUG_ON(ac->ac_which != OCFS2_AC_USE_META);
+	BUG_ON(ac->ac_handle != handle);
+
+	status = ocfs2_claim_suballoc_bits(osb,
+					   ac,
+					   bits_wanted,
+					   1,
+					   suballoc_bit_start,
+					   num_bits,
+					   &bg_blkno);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+	atomic_inc(&osb->alloc_stats.bg_allocs);
+
+	*blkno_start = bg_blkno + (u64) *suballoc_bit_start;
+	ac->ac_bits_given += (*num_bits);
+	status = 0;
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+int ocfs2_claim_new_inode(struct ocfs2_super *osb,
+			  struct ocfs2_journal_handle *handle,
+			  struct ocfs2_alloc_context *ac,
+			  u16 *suballoc_bit,
+			  u64 *fe_blkno)
+{
+	int status;
+	unsigned int num_bits;
+	u64 bg_blkno;
+
+	mlog_entry_void();
+
+	BUG_ON(!ac);
+	BUG_ON(ac->ac_bits_given != 0);
+	BUG_ON(ac->ac_bits_wanted != 1);
+	BUG_ON(ac->ac_which != OCFS2_AC_USE_INODE);
+	BUG_ON(ac->ac_handle != handle);
+
+	status = ocfs2_claim_suballoc_bits(osb,
+					   ac,
+					   1,
+					   1,
+					   suballoc_bit,
+					   &num_bits,
+					   &bg_blkno);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+	atomic_inc(&osb->alloc_stats.bg_allocs);
+
+	BUG_ON(num_bits != 1);
+
+	*fe_blkno = bg_blkno + (u64) (*suballoc_bit);
+	ac->ac_bits_given++;
+	status = 0;
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+/* translate a group desc. blkno and it's bitmap offset into
+ * disk cluster offset. */
+static inline u32 ocfs2_desc_bitmap_to_cluster_off(struct inode *inode,
+						   u64 bg_blkno,
+						   u16 bg_bit_off)
+{
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	u32 cluster = 0;
+
+	BUG_ON(!ocfs2_is_cluster_bitmap(inode));
+
+	if (bg_blkno != osb->first_cluster_group_blkno)
+		cluster = ocfs2_blocks_to_clusters(inode->i_sb, bg_blkno);
+	cluster += (u32) bg_bit_off;
+	return cluster;
+}
+
+/* given a cluster offset, calculate which block group it belongs to
+ * and return that block offset. */
+static inline u64 ocfs2_which_cluster_group(struct inode *inode,
+					    u32 cluster)
+{
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	u32 group_no;
+
+	BUG_ON(!ocfs2_is_cluster_bitmap(inode));
+
+	group_no = cluster / osb->bitmap_cpg;
+	if (!group_no)
+		return osb->first_cluster_group_blkno;
+	return ocfs2_clusters_to_blocks(inode->i_sb,
+					group_no * osb->bitmap_cpg);
+}
+
+/* given the block number of a cluster start, calculate which cluster
+ * group and descriptor bitmap offset that corresponds to. */
+static inline void ocfs2_block_to_cluster_group(struct inode *inode,
+						u64 data_blkno,
+						u64 *bg_blkno,
+						u16 *bg_bit_off)
+{
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	u32 data_cluster = ocfs2_blocks_to_clusters(osb->sb, data_blkno);
+
+	BUG_ON(!ocfs2_is_cluster_bitmap(inode));
+
+	*bg_blkno = ocfs2_which_cluster_group(inode,
+					      data_cluster);
+
+	if (*bg_blkno == osb->first_cluster_group_blkno)
+		*bg_bit_off = (u16) data_cluster;
+	else
+		*bg_bit_off = (u16) ocfs2_blocks_to_clusters(osb->sb,
+							     data_blkno - *bg_blkno);
+}
+
+/*
+ * min_bits - minimum contiguous chunk from this total allocation we
+ * can handle. set to what we asked for originally for a full
+ * contig. allocation, set to '1' to indicate we can deal with extents
+ * of any size.
+ */
+int ocfs2_claim_clusters(struct ocfs2_super *osb,
+			 struct ocfs2_journal_handle *handle,
+			 struct ocfs2_alloc_context *ac,
+			 u32 min_clusters,
+			 u32 *cluster_start,
+			 u32 *num_clusters)
+{
+	int status;
+	unsigned int bits_wanted = ac->ac_bits_wanted - ac->ac_bits_given;
+	u64 bg_blkno;
+	u16 bg_bit_off;
+
+	mlog_entry_void();
+
+	BUG_ON(!ac);
+	BUG_ON(ac->ac_bits_given >= ac->ac_bits_wanted);
+
+	BUG_ON(ac->ac_which != OCFS2_AC_USE_LOCAL
+	       && ac->ac_which != OCFS2_AC_USE_MAIN);
+	BUG_ON(ac->ac_handle != handle);
+
+	if (ac->ac_which == OCFS2_AC_USE_LOCAL) {
+		status = ocfs2_claim_local_alloc_bits(osb,
+						      handle,
+						      ac,
+						      bits_wanted,
+						      cluster_start,
+						      num_clusters);
+		if (!status)
+			atomic_inc(&osb->alloc_stats.local_data);
+	} else {
+		if (min_clusters > (osb->bitmap_cpg - 1)) {
+			/* The only paths asking for contiguousness
+			 * should know about this already. */
+			mlog(ML_ERROR, "minimum allocation requested exceeds "
+				       "group bitmap size!");
+			status = -ENOSPC;
+			goto bail;
+		}
+		/* clamp the current request down to a realistic size. */
+		if (bits_wanted > (osb->bitmap_cpg - 1))
+			bits_wanted = osb->bitmap_cpg - 1;
+
+		status = ocfs2_claim_suballoc_bits(osb,
+						   ac,
+						   bits_wanted,
+						   min_clusters,
+						   &bg_bit_off,
+						   num_clusters,
+						   &bg_blkno);
+		if (!status) {
+			*cluster_start =
+				ocfs2_desc_bitmap_to_cluster_off(ac->ac_inode,
+								 bg_blkno,
+								 bg_bit_off);
+			atomic_inc(&osb->alloc_stats.bitmap_data);
+		}
+	}
+	if (status < 0) {
+		if (status != -ENOSPC)
+			mlog_errno(status);
+		goto bail;
+	}
+
+	ac->ac_bits_given += *num_clusters;
+
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+static inline int ocfs2_block_group_clear_bits(struct ocfs2_journal_handle *handle,
+					       struct inode *alloc_inode,
+					       struct ocfs2_group_desc *bg,
+					       struct buffer_head *group_bh,
+					       unsigned int bit_off,
+					       unsigned int num_bits)
+{
+	int status;
+	unsigned int tmp;
+	int journal_type = OCFS2_JOURNAL_ACCESS_WRITE;
+	struct ocfs2_group_desc *undo_bg = NULL;
+
+	mlog_entry_void();
+
+	if (!OCFS2_IS_VALID_GROUP_DESC(bg)) {
+		OCFS2_RO_ON_INVALID_GROUP_DESC(alloc_inode->i_sb, bg);
+		status = -EIO;
+		goto bail;
+	}
+
+	mlog(0, "off = %u, num = %u\n", bit_off, num_bits);
+
+	if (ocfs2_is_cluster_bitmap(alloc_inode))
+		journal_type = OCFS2_JOURNAL_ACCESS_UNDO;
+
+	status = ocfs2_journal_access(handle, alloc_inode, group_bh,
+				      journal_type);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	if (ocfs2_is_cluster_bitmap(alloc_inode))
+		undo_bg = (struct ocfs2_group_desc *) bh2jh(group_bh)->b_committed_data;
+
+	tmp = num_bits;
+	while(tmp--) {
+		ocfs2_clear_bit((bit_off + tmp),
+				(unsigned long *) bg->bg_bitmap);
+		if (ocfs2_is_cluster_bitmap(alloc_inode))
+			ocfs2_set_bit(bit_off + tmp,
+				      (unsigned long *) undo_bg->bg_bitmap);
+	}
+	le16_add_cpu(&bg->bg_free_bits_count, num_bits);
+
+	status = ocfs2_journal_dirty(handle, group_bh);
+	if (status < 0)
+		mlog_errno(status);
+bail:
+	return status;
+}
+
+/*
+ * expects the suballoc inode to already be locked.
+ */
+static int ocfs2_free_suballoc_bits(struct ocfs2_journal_handle *handle,
+				    struct inode *alloc_inode,
+				    struct buffer_head *alloc_bh,
+				    unsigned int start_bit,
+				    u64 bg_blkno,
+				    unsigned int count)
+{
+	int status = 0;
+	u32 tmp_used;
+	struct ocfs2_super *osb = OCFS2_SB(alloc_inode->i_sb);
+	struct ocfs2_dinode *fe = (struct ocfs2_dinode *) alloc_bh->b_data;
+	struct ocfs2_chain_list *cl = &fe->id2.i_chain;
+	struct buffer_head *group_bh = NULL;
+	struct ocfs2_group_desc *group;
+
+	mlog_entry_void();
+
+	if (!OCFS2_IS_VALID_DINODE(fe)) {
+		OCFS2_RO_ON_INVALID_DINODE(alloc_inode->i_sb, fe);
+		status = -EIO;
+		goto bail;
+	}
+	BUG_ON((count + start_bit) > ocfs2_bits_per_group(cl));
+
+	mlog(0, "suballocator %"MLFu64": freeing %u bits from group %"MLFu64
+	     ", starting at %u\n",
+	     OCFS2_I(alloc_inode)->ip_blkno, count, bg_blkno,
+	     start_bit);
+
+	status = ocfs2_read_block(osb, bg_blkno, &group_bh, OCFS2_BH_CACHED,
+				  alloc_inode);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	group = (struct ocfs2_group_desc *) group_bh->b_data;
+	if (!OCFS2_IS_VALID_GROUP_DESC(group)) {
+		OCFS2_RO_ON_INVALID_GROUP_DESC(alloc_inode->i_sb, group);
+		status = -EIO;
+		goto bail;
+	}
+	BUG_ON((count + start_bit) > le16_to_cpu(group->bg_bits));
+
+	status = ocfs2_block_group_clear_bits(handle, alloc_inode,
+					      group, group_bh,
+					      start_bit, count);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = ocfs2_journal_access(handle, alloc_inode, alloc_bh,
+				      OCFS2_JOURNAL_ACCESS_WRITE);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	le32_add_cpu(&cl->cl_recs[le16_to_cpu(group->bg_chain)].c_free,
+		     count);
+	tmp_used = le32_to_cpu(fe->id1.bitmap1.i_used);
+	fe->id1.bitmap1.i_used = cpu_to_le32(tmp_used - count);
+
+	status = ocfs2_journal_dirty(handle, alloc_bh);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+bail:
+	if (group_bh)
+		brelse(group_bh);
+
+	mlog_exit(status);
+	return status;
+}
+
+static inline u64 ocfs2_which_suballoc_group(u64 block, unsigned int bit)
+{
+	u64 group = block - (u64) bit;
+
+	return group;
+}
+
+int ocfs2_free_dinode(struct ocfs2_journal_handle *handle,
+		      struct inode *inode_alloc_inode,
+		      struct buffer_head *inode_alloc_bh,
+		      struct ocfs2_dinode *di)
+{
+	u64 blk = le64_to_cpu(di->i_blkno);
+	u16 bit = le16_to_cpu(di->i_suballoc_bit);
+	u64 bg_blkno = ocfs2_which_suballoc_group(blk, bit);
+
+	return ocfs2_free_suballoc_bits(handle, inode_alloc_inode,
+					inode_alloc_bh, bit, bg_blkno, 1);
+}
+
+int ocfs2_free_extent_block(struct ocfs2_journal_handle *handle,
+			    struct inode *eb_alloc_inode,
+			    struct buffer_head *eb_alloc_bh,
+			    struct ocfs2_extent_block *eb)
+{
+	u64 blk = le64_to_cpu(eb->h_blkno);
+	u16 bit = le16_to_cpu(eb->h_suballoc_bit);
+	u64 bg_blkno = ocfs2_which_suballoc_group(blk, bit);
+
+	return ocfs2_free_suballoc_bits(handle, eb_alloc_inode, eb_alloc_bh,
+					bit, bg_blkno, 1);
+}
+
+int ocfs2_free_clusters(struct ocfs2_journal_handle *handle,
+		       struct inode *bitmap_inode,
+		       struct buffer_head *bitmap_bh,
+		       u64 start_blk,
+		       unsigned int num_clusters)
+{
+	int status;
+	u16 bg_start_bit;
+	u64 bg_blkno;
+	struct ocfs2_dinode *fe;
+
+	/* You can't ever have a contiguous set of clusters
+	 * bigger than a block group bitmap so we never have to worry
+	 * about looping on them. */
+
+	mlog_entry_void();
+
+	/* This is expensive. We can safely remove once this stuff has
+	 * gotten tested really well. */
+	BUG_ON(start_blk != ocfs2_clusters_to_blocks(bitmap_inode->i_sb, ocfs2_blocks_to_clusters(bitmap_inode->i_sb, start_blk)));
+
+	fe = (struct ocfs2_dinode *) bitmap_bh->b_data;
+
+	ocfs2_block_to_cluster_group(bitmap_inode, start_blk, &bg_blkno,
+				     &bg_start_bit);
+
+	mlog(0, "want to free %u clusters starting at block %"MLFu64"\n",
+	     num_clusters, start_blk);
+	mlog(0, "bg_blkno = %"MLFu64", bg_start_bit = %u\n",
+	     bg_blkno, bg_start_bit);
+
+	status = ocfs2_free_suballoc_bits(handle, bitmap_inode, bitmap_bh,
+					  bg_start_bit, bg_blkno,
+					  num_clusters);
+	if (status < 0)
+		mlog_errno(status);
+
+	mlog_exit(status);
+	return status;
+}
+
+static inline void ocfs2_debug_bg(struct ocfs2_group_desc *bg)
+{
+	printk("Block Group:\n");
+	printk("bg_signature:       %s\n", bg->bg_signature);
+	printk("bg_size:            %u\n", bg->bg_size);
+	printk("bg_bits:            %u\n", bg->bg_bits);
+	printk("bg_free_bits_count: %u\n", bg->bg_free_bits_count);
+	printk("bg_chain:           %u\n", bg->bg_chain);
+	printk("bg_generation:      %u\n", le32_to_cpu(bg->bg_generation));
+	printk("bg_next_group:      %"MLFu64"\n", bg->bg_next_group);
+	printk("bg_parent_dinode:   %"MLFu64"\n", bg->bg_parent_dinode);
+	printk("bg_blkno:           %"MLFu64"\n", bg->bg_blkno);
+}
+
+static inline void ocfs2_debug_suballoc_inode(struct ocfs2_dinode *fe)
+{
+	int i;
+
+	printk("Suballoc Inode %"MLFu64":\n", fe->i_blkno);
+	printk("i_signature:                  %s\n", fe->i_signature);
+	printk("i_size:                       %"MLFu64"\n", fe->i_size);
+	printk("i_clusters:                   %u\n", fe->i_clusters);
+	printk("i_generation:                 %u\n",
+	       le32_to_cpu(fe->i_generation));
+	printk("id1.bitmap1.i_used:           %u\n",
+	       le32_to_cpu(fe->id1.bitmap1.i_used));
+	printk("id1.bitmap1.i_total:          %u\n",
+	       le32_to_cpu(fe->id1.bitmap1.i_total));
+	printk("id2.i_chain.cl_cpg:           %u\n", fe->id2.i_chain.cl_cpg);
+	printk("id2.i_chain.cl_bpc:           %u\n", fe->id2.i_chain.cl_bpc);
+	printk("id2.i_chain.cl_count:         %u\n", fe->id2.i_chain.cl_count);
+	printk("id2.i_chain.cl_next_free_rec: %u\n",
+	       fe->id2.i_chain.cl_next_free_rec);
+	for(i = 0; i < fe->id2.i_chain.cl_next_free_rec; i++) {
+		printk("fe->id2.i_chain.cl_recs[%d].c_free:  %u\n", i,
+		       fe->id2.i_chain.cl_recs[i].c_free);
+		printk("fe->id2.i_chain.cl_recs[%d].c_total: %u\n", i,
+		       fe->id2.i_chain.cl_recs[i].c_total);
+		printk("fe->id2.i_chain.cl_recs[%d].c_blkno: %"MLFu64"\n", i,
+		       fe->id2.i_chain.cl_recs[i].c_blkno);
+	}
+}
diff --git a/fs/ocfs2/suballoc.h b/fs/ocfs2/suballoc.h
new file mode 100644
index 00000000000..a76c82a7cea
--- /dev/null
+++ b/fs/ocfs2/suballoc.h
@@ -0,0 +1,132 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * suballoc.h
+ *
+ * Defines sub allocator api
+ *
+ * Copyright (C) 2003, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef _CHAINALLOC_H_
+#define _CHAINALLOC_H_
+
+typedef int (group_search_t)(struct inode *,
+			     struct buffer_head *,
+			     u32,
+			     u32,
+			     u16 *,
+			     u16 *);
+
+struct ocfs2_alloc_context {
+	struct inode *ac_inode;    /* which bitmap are we allocating from? */
+	struct buffer_head *ac_bh; /* file entry bh */
+	u32    ac_bits_wanted;
+	u32    ac_bits_given;
+#define OCFS2_AC_USE_LOCAL 1
+#define OCFS2_AC_USE_MAIN  2
+#define OCFS2_AC_USE_INODE 3
+#define OCFS2_AC_USE_META  4
+	u32    ac_which;
+	struct ocfs2_journal_handle *ac_handle;
+
+	/* these are used by the chain search */
+	u16    ac_chain;
+	int    ac_allow_chain_relink;
+	group_search_t *ac_group_search;
+};
+
+void ocfs2_free_alloc_context(struct ocfs2_alloc_context *ac);
+static inline int ocfs2_alloc_context_bits_left(struct ocfs2_alloc_context *ac)
+{
+	return ac->ac_bits_wanted - ac->ac_bits_given;
+}
+
+int ocfs2_reserve_new_metadata(struct ocfs2_super *osb,
+			       struct ocfs2_journal_handle *handle,
+			       struct ocfs2_dinode *fe,
+			       struct ocfs2_alloc_context **ac);
+int ocfs2_reserve_new_inode(struct ocfs2_super *osb,
+			    struct ocfs2_journal_handle *handle,
+			    struct ocfs2_alloc_context **ac);
+int ocfs2_reserve_clusters(struct ocfs2_super *osb,
+			   struct ocfs2_journal_handle *handle,
+			   u32 bits_wanted,
+			   struct ocfs2_alloc_context **ac);
+
+int ocfs2_claim_metadata(struct ocfs2_super *osb,
+			 struct ocfs2_journal_handle *handle,
+			 struct ocfs2_alloc_context *ac,
+			 u32 bits_wanted,
+			 u16 *suballoc_bit_start,
+			 u32 *num_bits,
+			 u64 *blkno_start);
+int ocfs2_claim_new_inode(struct ocfs2_super *osb,
+			  struct ocfs2_journal_handle *handle,
+			  struct ocfs2_alloc_context *ac,
+			  u16 *suballoc_bit,
+			  u64 *fe_blkno);
+int ocfs2_claim_clusters(struct ocfs2_super *osb,
+			 struct ocfs2_journal_handle *handle,
+			 struct ocfs2_alloc_context *ac,
+			 u32 min_clusters,
+			 u32 *cluster_start,
+			 u32 *num_clusters);
+
+int ocfs2_free_dinode(struct ocfs2_journal_handle *handle,
+		      struct inode *inode_alloc_inode,
+		      struct buffer_head *inode_alloc_bh,
+		      struct ocfs2_dinode *di);
+int ocfs2_free_extent_block(struct ocfs2_journal_handle *handle,
+			    struct inode *eb_alloc_inode,
+			    struct buffer_head *eb_alloc_bh,
+			    struct ocfs2_extent_block *eb);
+int ocfs2_free_clusters(struct ocfs2_journal_handle *handle,
+			struct inode *bitmap_inode,
+			struct buffer_head *bitmap_bh,
+			u64 start_blk,
+			unsigned int num_clusters);
+
+static inline u32 ocfs2_cluster_from_desc(struct ocfs2_super *osb,
+					  u64 bg_blkno)
+{
+	/* This should work for all block group descriptors as only
+	 * the 1st group descriptor of the cluster bitmap is
+	 * different. */
+
+	if (bg_blkno == osb->first_cluster_group_blkno)
+		return 0;
+
+	/* the rest of the block groups are located at the beginning
+	 * of their 1st cluster, so a direct translation just
+	 * works. */
+	return ocfs2_blocks_to_clusters(osb->sb, bg_blkno);
+}
+
+static inline int ocfs2_is_cluster_bitmap(struct inode *inode)
+{
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	return osb->bitmap_blkno == OCFS2_I(inode)->ip_blkno;
+}
+
+/* This is for local alloc ONLY. Others should use the task-specific
+ * apis above. */
+int ocfs2_reserve_cluster_bitmap_bits(struct ocfs2_super *osb,
+				      struct ocfs2_alloc_context *ac);
+
+#endif /* _CHAINALLOC_H_ */
diff --git a/fs/ocfs2/super.c b/fs/ocfs2/super.c
new file mode 100644
index 00000000000..48bf7f0ce54
--- /dev/null
+++ b/fs/ocfs2/super.c
@@ -0,0 +1,1733 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * super.c
+ *
+ * load/unload driver, mount/dismount volumes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/utsname.h>
+#include <linux/init.h>
+#include <linux/random.h>
+#include <linux/statfs.h>
+#include <linux/moduleparam.h>
+#include <linux/blkdev.h>
+#include <linux/socket.h>
+#include <linux/inet.h>
+#include <linux/parser.h>
+#include <linux/crc32.h>
+#include <linux/debugfs.h>
+
+#include <cluster/nodemanager.h>
+
+#define MLOG_MASK_PREFIX ML_SUPER
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+/* this should be the only file to include a version 1 header */
+#include "ocfs1_fs_compat.h"
+
+#include "alloc.h"
+#include "dlmglue.h"
+#include "export.h"
+#include "extent_map.h"
+#include "heartbeat.h"
+#include "inode.h"
+#include "journal.h"
+#include "localalloc.h"
+#include "namei.h"
+#include "slot_map.h"
+#include "super.h"
+#include "sysfile.h"
+#include "uptodate.h"
+#include "ver.h"
+#include "vote.h"
+
+#include "buffer_head_io.h"
+
+/*
+ * Globals
+ */
+static spinlock_t ocfs2_globals_lock = SPIN_LOCK_UNLOCKED;
+
+static u32 osb_id;             /* Keeps track of next available OSB Id */
+
+static kmem_cache_t *ocfs2_inode_cachep = NULL;
+
+kmem_cache_t *ocfs2_lock_cache = NULL;
+
+/* OCFS2 needs to schedule several differnt types of work which
+ * require cluster locking, disk I/O, recovery waits, etc. Since these
+ * types of work tend to be heavy we avoid using the kernel events
+ * workqueue and schedule on our own. */
+struct workqueue_struct *ocfs2_wq = NULL;
+
+static struct dentry *ocfs2_debugfs_root = NULL;
+
+MODULE_AUTHOR("Oracle");
+MODULE_LICENSE("GPL");
+
+static int ocfs2_parse_options(struct super_block *sb, char *options,
+			       unsigned long *mount_opt, int is_remount);
+static void ocfs2_put_super(struct super_block *sb);
+static int ocfs2_mount_volume(struct super_block *sb);
+static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
+static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
+static int ocfs2_initialize_mem_caches(void);
+static void ocfs2_free_mem_caches(void);
+static void ocfs2_delete_osb(struct ocfs2_super *osb);
+
+static int ocfs2_statfs(struct super_block *sb, struct kstatfs *buf);
+
+static int ocfs2_sync_fs(struct super_block *sb, int wait);
+
+static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
+static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
+static int ocfs2_release_system_inodes(struct ocfs2_super *osb);
+static int ocfs2_fill_local_node_info(struct ocfs2_super *osb);
+static int ocfs2_check_volume(struct ocfs2_super *osb);
+static int ocfs2_verify_volume(struct ocfs2_dinode *di,
+			       struct buffer_head *bh,
+			       u32 sectsize);
+static int ocfs2_initialize_super(struct super_block *sb,
+				  struct buffer_head *bh,
+				  int sector_size);
+static int ocfs2_get_sector(struct super_block *sb,
+			    struct buffer_head **bh,
+			    int block,
+			    int sect_size);
+static void ocfs2_write_super(struct super_block *sb);
+static struct inode *ocfs2_alloc_inode(struct super_block *sb);
+static void ocfs2_destroy_inode(struct inode *inode);
+
+static unsigned long long ocfs2_max_file_offset(unsigned int blockshift);
+
+static struct super_operations ocfs2_sops = {
+	.statfs		= ocfs2_statfs,
+	.alloc_inode	= ocfs2_alloc_inode,
+	.destroy_inode	= ocfs2_destroy_inode,
+	.drop_inode	= ocfs2_drop_inode,
+	.clear_inode	= ocfs2_clear_inode,
+	.delete_inode	= ocfs2_delete_inode,
+	.sync_fs	= ocfs2_sync_fs,
+	.write_super	= ocfs2_write_super,
+	.put_super	= ocfs2_put_super,
+	.remount_fs	= ocfs2_remount,
+};
+
+enum {
+	Opt_barrier,
+	Opt_err_panic,
+	Opt_err_ro,
+	Opt_intr,
+	Opt_nointr,
+	Opt_hb_none,
+	Opt_hb_local,
+	Opt_data_ordered,
+	Opt_data_writeback,
+	Opt_err,
+};
+
+static match_table_t tokens = {
+	{Opt_barrier, "barrier=%u"},
+	{Opt_err_panic, "errors=panic"},
+	{Opt_err_ro, "errors=remount-ro"},
+	{Opt_intr, "intr"},
+	{Opt_nointr, "nointr"},
+	{Opt_hb_none, OCFS2_HB_NONE},
+	{Opt_hb_local, OCFS2_HB_LOCAL},
+	{Opt_data_ordered, "data=ordered"},
+	{Opt_data_writeback, "data=writeback"},
+	{Opt_err, NULL}
+};
+
+/*
+ * write_super and sync_fs ripped right out of ext3.
+ */
+static void ocfs2_write_super(struct super_block *sb)
+{
+	if (down_trylock(&sb->s_lock) == 0)
+		BUG();
+	sb->s_dirt = 0;
+}
+
+static int ocfs2_sync_fs(struct super_block *sb, int wait)
+{
+	int status = 0;
+	tid_t target;
+	struct ocfs2_super *osb = OCFS2_SB(sb);
+
+	sb->s_dirt = 0;
+
+	if (ocfs2_is_hard_readonly(osb))
+		return -EROFS;
+
+	if (wait) {
+		status = ocfs2_flush_truncate_log(osb);
+		if (status < 0)
+			mlog_errno(status);
+	} else {
+		ocfs2_schedule_truncate_log_flush(osb, 0);
+	}
+
+	if (journal_start_commit(OCFS2_SB(sb)->journal->j_journal, &target)) {
+		if (wait)
+			log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
+					target);
+	}
+	return 0;
+}
+
+static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
+{
+	struct inode *new = NULL;
+	int status = 0;
+	int i;
+
+	mlog_entry_void();
+
+	new = ocfs2_iget(osb, osb->root_blkno);
+	if (IS_ERR(new)) {
+		status = PTR_ERR(new);
+		mlog_errno(status);
+		goto bail;
+	}
+	osb->root_inode = new;
+
+	new = ocfs2_iget(osb, osb->system_dir_blkno);
+	if (IS_ERR(new)) {
+		status = PTR_ERR(new);
+		mlog_errno(status);
+		goto bail;
+	}
+	osb->sys_root_inode = new;
+
+	for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
+	     i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
+		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
+		if (!new) {
+			ocfs2_release_system_inodes(osb);
+			status = -EINVAL;
+			mlog_errno(status);
+			/* FIXME: Should ERROR_RO_FS */
+			mlog(ML_ERROR, "Unable to load system inode %d, "
+			     "possibly corrupt fs?", i);
+			goto bail;
+		}
+		// the array now has one ref, so drop this one
+		iput(new);
+	}
+
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
+{
+	struct inode *new = NULL;
+	int status = 0;
+	int i;
+
+	mlog_entry_void();
+
+	for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
+	     i < NUM_SYSTEM_INODES;
+	     i++) {
+		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
+		if (!new) {
+			ocfs2_release_system_inodes(osb);
+			status = -EINVAL;
+			mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
+			     status, i, osb->slot_num);
+			goto bail;
+		}
+		/* the array now has one ref, so drop this one */
+		iput(new);
+	}
+
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_release_system_inodes(struct ocfs2_super *osb)
+{
+	int status = 0, i;
+	struct inode *inode;
+
+	mlog_entry_void();
+
+	for (i = 0; i < NUM_SYSTEM_INODES; i++) {
+		inode = osb->system_inodes[i];
+		if (inode) {
+			iput(inode);
+			osb->system_inodes[i] = NULL;
+		}
+	}
+
+	inode = osb->sys_root_inode;
+	if (inode) {
+		iput(inode);
+		osb->sys_root_inode = NULL;
+	}
+
+	inode = osb->root_inode;
+	if (inode) {
+		iput(inode);
+		osb->root_inode = NULL;
+	}
+
+	mlog_exit(status);
+	return status;
+}
+
+/* We're allocating fs objects, use GFP_NOFS */
+static struct inode *ocfs2_alloc_inode(struct super_block *sb)
+{
+	struct ocfs2_inode_info *oi;
+
+	oi = kmem_cache_alloc(ocfs2_inode_cachep, SLAB_NOFS);
+	if (!oi)
+		return NULL;
+
+	return &oi->vfs_inode;
+}
+
+static void ocfs2_destroy_inode(struct inode *inode)
+{
+	kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
+}
+
+/* From xfs_super.c:xfs_max_file_offset
+ * Copyright (c) 2000-2004 Silicon Graphics, Inc.
+ */
+static unsigned long long ocfs2_max_file_offset(unsigned int blockshift)
+{
+	unsigned int pagefactor = 1;
+	unsigned int bitshift = BITS_PER_LONG - 1;
+
+	/* Figure out maximum filesize, on Linux this can depend on
+	 * the filesystem blocksize (on 32 bit platforms).
+	 * __block_prepare_write does this in an [unsigned] long...
+	 *      page->index << (PAGE_CACHE_SHIFT - bbits)
+	 * So, for page sized blocks (4K on 32 bit platforms),
+	 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
+	 *      (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
+	 * but for smaller blocksizes it is less (bbits = log2 bsize).
+	 * Note1: get_block_t takes a long (implicit cast from above)
+	 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
+	 * can optionally convert the [unsigned] long from above into
+	 * an [unsigned] long long.
+	 */
+
+#if BITS_PER_LONG == 32
+# if defined(CONFIG_LBD)
+	BUG_ON(sizeof(sector_t) != 8);
+	pagefactor = PAGE_CACHE_SIZE;
+	bitshift = BITS_PER_LONG;
+# else
+	pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
+# endif
+#endif
+
+	return (((unsigned long long)pagefactor) << bitshift) - 1;
+}
+
+static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
+{
+	int incompat_features;
+	int ret = 0;
+	unsigned long parsed_options;
+	struct ocfs2_super *osb = OCFS2_SB(sb);
+
+	if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
+	    (parsed_options & OCFS2_MOUNT_HB_LOCAL)) {
+		ret = -EINVAL;
+		mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
+		goto out;
+	}
+
+	if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
+	    (parsed_options & OCFS2_MOUNT_DATA_WRITEBACK)) {
+		ret = -EINVAL;
+		mlog(ML_ERROR, "Cannot change data mode on remount\n");
+		goto out;
+	}
+
+	/* We're going to/from readonly mode. */
+	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
+		/* Lock here so the check of HARD_RO and the potential
+		 * setting of SOFT_RO is atomic. */
+		spin_lock(&osb->osb_lock);
+		if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
+			mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
+			ret = -EROFS;
+			goto unlock_osb;
+		}
+
+		if (*flags & MS_RDONLY) {
+			mlog(0, "Going to ro mode.\n");
+			sb->s_flags |= MS_RDONLY;
+			osb->osb_flags |= OCFS2_OSB_SOFT_RO;
+		} else {
+			mlog(0, "Making ro filesystem writeable.\n");
+
+			if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
+				mlog(ML_ERROR, "Cannot remount RDWR "
+				     "filesystem due to previous errors.\n");
+				ret = -EROFS;
+				goto unlock_osb;
+			}
+			incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
+			if (incompat_features) {
+				mlog(ML_ERROR, "Cannot remount RDWR because "
+				     "of unsupported optional features "
+				     "(%x).\n", incompat_features);
+				ret = -EINVAL;
+				goto unlock_osb;
+			}
+			sb->s_flags &= ~MS_RDONLY;
+			osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
+		}
+unlock_osb:
+		spin_unlock(&osb->osb_lock);
+	}
+
+	if (!ret) {
+		if (!ocfs2_is_hard_readonly(osb))
+			ocfs2_set_journal_params(osb);
+
+		/* Only save off the new mount options in case of a successful
+		 * remount. */
+		osb->s_mount_opt = parsed_options;
+	}
+out:
+	return ret;
+}
+
+static int ocfs2_sb_probe(struct super_block *sb,
+			  struct buffer_head **bh,
+			  int *sector_size)
+{
+	int status = 0, tmpstat;
+	struct ocfs1_vol_disk_hdr *hdr;
+	struct ocfs2_dinode *di;
+	int blksize;
+
+	*bh = NULL;
+
+	/* may be > 512 */
+	*sector_size = bdev_hardsect_size(sb->s_bdev);
+	if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
+		mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
+		     *sector_size, OCFS2_MAX_BLOCKSIZE);
+		status = -EINVAL;
+		goto bail;
+	}
+
+	/* Can this really happen? */
+	if (*sector_size < OCFS2_MIN_BLOCKSIZE)
+		*sector_size = OCFS2_MIN_BLOCKSIZE;
+
+	/* check block zero for old format */
+	status = ocfs2_get_sector(sb, bh, 0, *sector_size);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+	hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
+	if (hdr->major_version == OCFS1_MAJOR_VERSION) {
+		mlog(ML_ERROR, "incompatible version: %u.%u\n",
+		     hdr->major_version, hdr->minor_version);
+		status = -EINVAL;
+	}
+	if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
+		   strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
+		mlog(ML_ERROR, "incompatible volume signature: %8s\n",
+		     hdr->signature);
+		status = -EINVAL;
+	}
+	brelse(*bh);
+	*bh = NULL;
+	if (status < 0) {
+		mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
+		     "upgraded before mounting with ocfs v2\n");
+		goto bail;
+	}
+
+	/*
+	 * Now check at magic offset for 512, 1024, 2048, 4096
+	 * blocksizes.  4096 is the maximum blocksize because it is
+	 * the minimum clustersize.
+	 */
+	status = -EINVAL;
+	for (blksize = *sector_size;
+	     blksize <= OCFS2_MAX_BLOCKSIZE;
+	     blksize <<= 1) {
+		tmpstat = ocfs2_get_sector(sb, bh,
+					   OCFS2_SUPER_BLOCK_BLKNO,
+					   blksize);
+		if (tmpstat < 0) {
+			status = tmpstat;
+			mlog_errno(status);
+			goto bail;
+		}
+		di = (struct ocfs2_dinode *) (*bh)->b_data;
+		status = ocfs2_verify_volume(di, *bh, blksize);
+		if (status >= 0)
+			goto bail;
+		brelse(*bh);
+		*bh = NULL;
+		if (status != -EAGAIN)
+			break;
+	}
+
+bail:
+	return status;
+}
+
+static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
+{
+	struct dentry *root;
+	int status, sector_size;
+	unsigned long parsed_opt;
+	struct inode *inode = NULL;
+	struct ocfs2_super *osb = NULL;
+	struct buffer_head *bh = NULL;
+
+	mlog_entry("%p, %p, %i", sb, data, silent);
+
+	/* for now we only have one cluster/node, make sure we see it
+	 * in the heartbeat universe */
+	if (!o2hb_check_local_node_heartbeating()) {
+		status = -EINVAL;
+		goto read_super_error;
+	}
+
+	/* probe for superblock */
+	status = ocfs2_sb_probe(sb, &bh, &sector_size);
+	if (status < 0) {
+		mlog(ML_ERROR, "superblock probe failed!\n");
+		goto read_super_error;
+	}
+
+	status = ocfs2_initialize_super(sb, bh, sector_size);
+	osb = OCFS2_SB(sb);
+	if (status < 0) {
+		mlog_errno(status);
+		goto read_super_error;
+	}
+	brelse(bh);
+	bh = NULL;
+
+	if (!ocfs2_parse_options(sb, data, &parsed_opt, 0)) {
+		status = -EINVAL;
+		goto read_super_error;
+	}
+	osb->s_mount_opt = parsed_opt;
+
+	sb->s_magic = OCFS2_SUPER_MAGIC;
+
+	/* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
+	 * heartbeat=none */
+	if (bdev_read_only(sb->s_bdev)) {
+		if (!(sb->s_flags & MS_RDONLY)) {
+			status = -EACCES;
+			mlog(ML_ERROR, "Readonly device detected but readonly "
+			     "mount was not specified.\n");
+			goto read_super_error;
+		}
+
+		/* You should not be able to start a local heartbeat
+		 * on a readonly device. */
+		if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
+			status = -EROFS;
+			mlog(ML_ERROR, "Local heartbeat specified on readonly "
+			     "device.\n");
+			goto read_super_error;
+		}
+
+		status = ocfs2_check_journals_nolocks(osb);
+		if (status < 0) {
+			if (status == -EROFS)
+				mlog(ML_ERROR, "Recovery required on readonly "
+				     "file system, but write access is "
+				     "unavailable.\n");
+			else
+				mlog_errno(status);			
+			goto read_super_error;
+		}
+
+		ocfs2_set_ro_flag(osb, 1);
+
+		printk(KERN_NOTICE "Readonly device detected. No cluster "
+		       "services will be utilized for this mount. Recovery "
+		       "will be skipped.\n");
+	}
+
+	if (!ocfs2_is_hard_readonly(osb)) {
+		/* If this isn't a hard readonly mount, then we need
+		 * to make sure that heartbeat is in a valid state,
+		 * and that we mark ourselves soft readonly is -oro
+		 * was specified. */
+		if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
+			mlog(ML_ERROR, "No heartbeat for device (%s)\n",
+			     sb->s_id);
+			status = -EINVAL;
+			goto read_super_error;
+		}
+
+		if (sb->s_flags & MS_RDONLY)
+			ocfs2_set_ro_flag(osb, 0);
+	}
+
+	osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
+						 ocfs2_debugfs_root);
+	if (!osb->osb_debug_root) {
+		status = -EINVAL;
+		mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
+		goto read_super_error;
+	}
+
+	status = ocfs2_mount_volume(sb);
+	if (osb->root_inode)
+		inode = igrab(osb->root_inode);
+
+	if (status < 0)
+		goto read_super_error;
+
+	if (!inode) {
+		status = -EIO;
+		mlog_errno(status);
+		goto read_super_error;
+	}
+
+	root = d_alloc_root(inode);
+	if (!root) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto read_super_error;
+	}
+
+	sb->s_root = root;
+
+	ocfs2_complete_mount_recovery(osb);
+
+	printk("ocfs2: Mounting device (%u,%u) on (node %d, slot %d) with %s "
+	       "data mode.\n",
+	       MAJOR(sb->s_dev), MINOR(sb->s_dev), osb->node_num,
+	       osb->slot_num,
+	       osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
+	       "ordered");
+
+	atomic_set(&osb->vol_state, VOLUME_MOUNTED);
+	wake_up(&osb->osb_mount_event);
+
+	mlog_exit(status);
+	return status;
+
+read_super_error:
+	if (bh != NULL)
+		brelse(bh);
+
+	if (inode)
+		iput(inode);
+
+	if (osb) {
+		atomic_set(&osb->vol_state, VOLUME_DISABLED);
+		wake_up(&osb->osb_mount_event);
+		ocfs2_dismount_volume(sb, 1);
+	}
+
+	mlog_exit(status);
+	return status;
+}
+
+static struct super_block *ocfs2_get_sb(struct file_system_type *fs_type,
+					int flags,
+					const char *dev_name,
+					void *data)
+{
+	return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
+}
+
+static struct file_system_type ocfs2_fs_type = {
+	.owner          = THIS_MODULE,
+	.name           = "ocfs2",
+	.get_sb         = ocfs2_get_sb, /* is this called when we mount
+					* the fs? */
+	.kill_sb        = kill_block_super, /* set to the generic one
+					     * right now, but do we
+					     * need to change that? */
+	.fs_flags       = FS_REQUIRES_DEV,
+	.next           = NULL
+};
+
+static int ocfs2_parse_options(struct super_block *sb,
+			       char *options,
+			       unsigned long *mount_opt,
+			       int is_remount)
+{
+	int status;
+	char *p;
+
+	mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
+		   options ? options : "(none)");
+
+	*mount_opt = 0;
+
+	if (!options) {
+		status = 1;
+		goto bail;
+	}
+
+	while ((p = strsep(&options, ",")) != NULL) {
+		int token, option;
+		substring_t args[MAX_OPT_ARGS];
+
+		if (!*p)
+			continue;
+
+		token = match_token(p, tokens, args);
+		switch (token) {
+		case Opt_hb_local:
+			*mount_opt |= OCFS2_MOUNT_HB_LOCAL;
+			break;
+		case Opt_hb_none:
+			*mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
+			break;
+		case Opt_barrier:
+			if (match_int(&args[0], &option)) {
+				status = 0;
+				goto bail;
+			}
+			if (option)
+				*mount_opt |= OCFS2_MOUNT_BARRIER;
+			else
+				*mount_opt &= ~OCFS2_MOUNT_BARRIER;
+			break;
+		case Opt_intr:
+			*mount_opt &= ~OCFS2_MOUNT_NOINTR;
+			break;
+		case Opt_nointr:
+			*mount_opt |= OCFS2_MOUNT_NOINTR;
+			break;
+		case Opt_err_panic:
+			*mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
+			break;
+		case Opt_err_ro:
+			*mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
+			break;
+		case Opt_data_ordered:
+			*mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
+			break;
+		case Opt_data_writeback:
+			*mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
+			break;
+		default:
+			mlog(ML_ERROR,
+			     "Unrecognized mount option \"%s\" "
+			     "or missing value\n", p);
+			status = 0;
+			goto bail;
+		}
+	}
+
+	status = 1;
+
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+static int __init ocfs2_init(void)
+{
+	int status;
+
+	mlog_entry_void();
+
+	ocfs2_print_version();
+
+	if (init_ocfs2_extent_maps())
+		return -ENOMEM;
+
+	status = init_ocfs2_uptodate_cache();
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	status = ocfs2_initialize_mem_caches();
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
+	if (!ocfs2_wq) {
+		status = -ENOMEM;
+		goto leave;
+	}
+
+	spin_lock(&ocfs2_globals_lock);
+	osb_id = 0;
+	spin_unlock(&ocfs2_globals_lock);
+
+	ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
+	if (!ocfs2_debugfs_root) {
+		status = -EFAULT;
+		mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
+	}
+
+leave:
+	if (status < 0) {
+		ocfs2_free_mem_caches();
+		exit_ocfs2_uptodate_cache();
+		exit_ocfs2_extent_maps();
+	}
+
+	mlog_exit(status);
+
+	if (status >= 0) {
+		return register_filesystem(&ocfs2_fs_type);
+	} else
+		return -1;
+}
+
+static void __exit ocfs2_exit(void)
+{
+	mlog_entry_void();
+
+	if (ocfs2_wq) {
+		flush_workqueue(ocfs2_wq);
+		destroy_workqueue(ocfs2_wq);
+	}
+
+	debugfs_remove(ocfs2_debugfs_root);
+
+	ocfs2_free_mem_caches();
+
+	unregister_filesystem(&ocfs2_fs_type);
+
+	exit_ocfs2_extent_maps();
+
+	exit_ocfs2_uptodate_cache();
+
+	mlog_exit_void();
+}
+
+static void ocfs2_put_super(struct super_block *sb)
+{
+	mlog_entry("(0x%p)\n", sb);
+
+	ocfs2_sync_blockdev(sb);
+	ocfs2_dismount_volume(sb, 0);
+
+	mlog_exit_void();
+}
+
+static int ocfs2_statfs(struct super_block *sb, struct kstatfs *buf)
+{
+	struct ocfs2_super *osb;
+	u32 numbits, freebits;
+	int status;
+	struct ocfs2_dinode *bm_lock;
+	struct buffer_head *bh = NULL;
+	struct inode *inode = NULL;
+
+	mlog_entry("(%p, %p)\n", sb, buf);
+
+	osb = OCFS2_SB(sb);
+
+	inode = ocfs2_get_system_file_inode(osb,
+					    GLOBAL_BITMAP_SYSTEM_INODE,
+					    OCFS2_INVALID_SLOT);
+	if (!inode) {
+		mlog(ML_ERROR, "failed to get bitmap inode\n");
+		status = -EIO;
+		goto bail;
+	}
+
+	status = ocfs2_meta_lock(inode, NULL, &bh, 0);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	bm_lock = (struct ocfs2_dinode *) bh->b_data;
+
+	numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
+	freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
+
+	buf->f_type = OCFS2_SUPER_MAGIC;
+	buf->f_bsize = sb->s_blocksize;
+	buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
+	buf->f_blocks = ((sector_t) numbits) *
+			(osb->s_clustersize >> osb->sb->s_blocksize_bits);
+	buf->f_bfree = ((sector_t) freebits) *
+		       (osb->s_clustersize >> osb->sb->s_blocksize_bits);
+	buf->f_bavail = buf->f_bfree;
+	buf->f_files = numbits;
+	buf->f_ffree = freebits;
+
+	brelse(bh);
+
+	ocfs2_meta_unlock(inode, 0);
+	status = 0;
+bail:
+	if (inode)
+		iput(inode);
+
+	mlog_exit(status);
+
+	return status;
+}
+
+static void ocfs2_inode_init_once(void *data,
+				  kmem_cache_t *cachep,
+				  unsigned long flags)
+{
+	struct ocfs2_inode_info *oi = data;
+
+	if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
+	    SLAB_CTOR_CONSTRUCTOR) {
+		oi->ip_flags = 0;
+		oi->ip_open_count = 0;
+		spin_lock_init(&oi->ip_lock);
+		ocfs2_extent_map_init(&oi->vfs_inode);
+		INIT_LIST_HEAD(&oi->ip_handle_list);
+		INIT_LIST_HEAD(&oi->ip_io_markers);
+		oi->ip_handle = NULL;
+		oi->ip_created_trans = 0;
+		oi->ip_last_trans = 0;
+		oi->ip_dir_start_lookup = 0;
+
+		init_rwsem(&oi->ip_alloc_sem);
+		init_MUTEX(&(oi->ip_io_sem));
+
+		oi->ip_blkno = 0ULL;
+		oi->ip_clusters = 0;
+
+		ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
+		ocfs2_lock_res_init_once(&oi->ip_meta_lockres);
+		ocfs2_lock_res_init_once(&oi->ip_data_lockres);
+
+		ocfs2_metadata_cache_init(&oi->vfs_inode);
+
+		inode_init_once(&oi->vfs_inode);
+	}
+}
+
+static int ocfs2_initialize_mem_caches(void)
+{
+	ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
+					       sizeof(struct ocfs2_inode_info),
+					       0, SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT,
+					       ocfs2_inode_init_once, NULL);
+	if (!ocfs2_inode_cachep)
+		return -ENOMEM;
+
+	ocfs2_lock_cache = kmem_cache_create("ocfs2_lock",
+					     sizeof(struct ocfs2_journal_lock),
+					     0,
+					     SLAB_NO_REAP|SLAB_HWCACHE_ALIGN,
+					     NULL, NULL);
+	if (!ocfs2_lock_cache)
+		return -ENOMEM;
+
+	return 0;
+}
+
+static void ocfs2_free_mem_caches(void)
+{
+	if (ocfs2_inode_cachep)
+		kmem_cache_destroy(ocfs2_inode_cachep);
+	if (ocfs2_lock_cache)
+		kmem_cache_destroy(ocfs2_lock_cache);
+
+	ocfs2_inode_cachep = NULL;
+	ocfs2_lock_cache = NULL;
+}
+
+static int ocfs2_get_sector(struct super_block *sb,
+			    struct buffer_head **bh,
+			    int block,
+			    int sect_size)
+{
+	if (!sb_set_blocksize(sb, sect_size)) {
+		mlog(ML_ERROR, "unable to set blocksize\n");
+		return -EIO;
+	}
+
+	*bh = sb_getblk(sb, block);
+	if (!*bh) {
+		mlog_errno(-EIO);
+		return -EIO;
+	}
+	lock_buffer(*bh);
+	if (!buffer_dirty(*bh))
+		clear_buffer_uptodate(*bh);
+	unlock_buffer(*bh);
+	ll_rw_block(READ, 1, bh);
+	wait_on_buffer(*bh);
+	return 0;
+}
+
+/* ocfs2 1.0 only allows one cluster and node identity per kernel image. */
+static int ocfs2_fill_local_node_info(struct ocfs2_super *osb)
+{
+	int status;
+
+	/* XXX hold a ref on the node while mounte?  easy enough, if
+	 * desirable. */
+	osb->node_num = o2nm_this_node();
+	if (osb->node_num == O2NM_MAX_NODES) {
+		mlog(ML_ERROR, "could not find this host's node number\n");
+		status = -ENOENT;
+		goto bail;
+	}
+
+	mlog(ML_NOTICE, "I am node %d\n", osb->node_num);
+
+	status = 0;
+bail:
+	return status;
+}
+
+static int ocfs2_mount_volume(struct super_block *sb)
+{
+	int status = 0;
+	int unlock_super = 0;
+	struct ocfs2_super *osb = OCFS2_SB(sb);
+
+	mlog_entry_void();
+
+	if (ocfs2_is_hard_readonly(osb))
+		goto leave;
+
+	status = ocfs2_fill_local_node_info(osb);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	status = ocfs2_register_hb_callbacks(osb);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	status = ocfs2_dlm_init(osb);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	/* requires vote_thread to be running. */
+	status = ocfs2_register_net_handlers(osb);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	status = ocfs2_super_lock(osb, 1);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+	unlock_super = 1;
+
+	/* This will load up the node map and add ourselves to it. */
+	status = ocfs2_find_slot(osb);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	ocfs2_populate_mounted_map(osb);
+
+	/* load all node-local system inodes */
+	status = ocfs2_init_local_system_inodes(osb);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	status = ocfs2_check_volume(osb);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	status = ocfs2_truncate_log_init(osb);
+	if (status < 0) {
+		mlog_errno(status);
+		goto leave;
+	}
+
+	/* This should be sent *after* we recovered our journal as it
+	 * will cause other nodes to unmark us as needing
+	 * recovery. However, we need to send it *before* dropping the
+	 * super block lock as otherwise their recovery threads might
+	 * try to clean us up while we're live! */
+	status = ocfs2_request_mount_vote(osb);
+	if (status < 0)
+		mlog_errno(status);
+
+leave:
+	if (unlock_super)
+		ocfs2_super_unlock(osb, 1);
+
+	mlog_exit(status);
+	return status;
+}
+
+/* we can't grab the goofy sem lock from inside wait_event, so we use
+ * memory barriers to make sure that we'll see the null task before
+ * being woken up */
+static int ocfs2_recovery_thread_running(struct ocfs2_super *osb)
+{
+	mb();
+	return osb->recovery_thread_task != NULL;
+}
+
+static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
+{
+	int tmp;
+	struct ocfs2_super *osb = NULL;
+
+	mlog_entry("(0x%p)\n", sb);
+
+	BUG_ON(!sb);
+	osb = OCFS2_SB(sb);
+	BUG_ON(!osb);
+
+	ocfs2_shutdown_local_alloc(osb);
+
+	ocfs2_truncate_log_shutdown(osb);
+
+	/* disable any new recovery threads and wait for any currently
+	 * running ones to exit. Do this before setting the vol_state. */
+	down(&osb->recovery_lock);
+	osb->disable_recovery = 1;
+	up(&osb->recovery_lock);
+	wait_event(osb->recovery_event, !ocfs2_recovery_thread_running(osb));
+
+	/* At this point, we know that no more recovery threads can be
+	 * launched, so wait for any recovery completion work to
+	 * complete. */
+	flush_workqueue(ocfs2_wq);
+
+	ocfs2_journal_shutdown(osb);
+
+	ocfs2_sync_blockdev(sb);
+
+	/* No dlm means we've failed during mount, so skip all the
+	 * steps which depended on that to complete. */
+	if (osb->dlm) {
+		tmp = ocfs2_super_lock(osb, 1);
+		if (tmp < 0) {
+			mlog_errno(tmp);
+			return;
+		}
+
+		tmp = ocfs2_request_umount_vote(osb);
+		if (tmp < 0)
+			mlog_errno(tmp);
+
+		if (osb->slot_num != OCFS2_INVALID_SLOT)
+			ocfs2_put_slot(osb);
+
+		ocfs2_super_unlock(osb, 1);
+	}
+
+	ocfs2_release_system_inodes(osb);
+
+	if (osb->dlm) {
+		ocfs2_unregister_net_handlers(osb);
+
+		ocfs2_dlm_shutdown(osb);
+	}
+
+	ocfs2_clear_hb_callbacks(osb);
+
+	debugfs_remove(osb->osb_debug_root);
+
+	if (!mnt_err)
+		ocfs2_stop_heartbeat(osb);
+
+	atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
+
+	printk("ocfs2: Unmounting device (%u,%u) on (node %d)\n",
+	       MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev), osb->node_num);
+
+	ocfs2_delete_osb(osb);
+	kfree(osb);
+	sb->s_dev = 0;
+	sb->s_fs_info = NULL;
+}
+
+static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
+				unsigned uuid_bytes)
+{
+	int i, ret;
+	char *ptr;
+
+	BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
+
+	osb->uuid_str = kcalloc(1, OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
+	if (osb->uuid_str == NULL)
+		return -ENOMEM;
+
+	memcpy(osb->uuid, uuid, OCFS2_VOL_UUID_LEN);
+
+	for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
+		/* print with null */
+		ret = snprintf(ptr, 3, "%02X", uuid[i]);
+		if (ret != 2) /* drop super cleans up */
+			return -EINVAL;
+		/* then only advance past the last char */
+		ptr += 2;
+	}
+
+	return 0;
+}
+
+static int ocfs2_initialize_super(struct super_block *sb,
+				  struct buffer_head *bh,
+				  int sector_size)
+{
+	int status = 0;
+	int i;
+	struct ocfs2_dinode *di = NULL;
+	struct inode *inode = NULL;
+	struct buffer_head *bitmap_bh = NULL;
+	struct ocfs2_journal *journal;
+	__le32 uuid_net_key;
+	struct ocfs2_super *osb;
+
+	mlog_entry_void();
+
+	osb = kcalloc(1, sizeof(struct ocfs2_super), GFP_KERNEL);
+	if (!osb) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	sb->s_fs_info = osb;
+	sb->s_op = &ocfs2_sops;
+	sb->s_export_op = &ocfs2_export_ops;
+	sb->s_flags |= MS_NOATIME;
+	/* this is needed to support O_LARGEFILE */
+	sb->s_maxbytes = ocfs2_max_file_offset(sb->s_blocksize_bits);
+
+	osb->sb = sb;
+	/* Save off for ocfs2_rw_direct */
+	osb->s_sectsize_bits = blksize_bits(sector_size);
+	if (!osb->s_sectsize_bits)
+		BUG();
+
+	osb->net_response_ids = 0;
+	spin_lock_init(&osb->net_response_lock);
+	INIT_LIST_HEAD(&osb->net_response_list);
+
+	INIT_LIST_HEAD(&osb->osb_net_handlers);
+	init_waitqueue_head(&osb->recovery_event);
+	spin_lock_init(&osb->vote_task_lock);
+	init_waitqueue_head(&osb->vote_event);
+	osb->vote_work_sequence = 0;
+	osb->vote_wake_sequence = 0;
+	INIT_LIST_HEAD(&osb->blocked_lock_list);
+	osb->blocked_lock_count = 0;
+	INIT_LIST_HEAD(&osb->vote_list);
+	spin_lock_init(&osb->osb_lock);
+
+	atomic_set(&osb->alloc_stats.moves, 0);
+	atomic_set(&osb->alloc_stats.local_data, 0);
+	atomic_set(&osb->alloc_stats.bitmap_data, 0);
+	atomic_set(&osb->alloc_stats.bg_allocs, 0);
+	atomic_set(&osb->alloc_stats.bg_extends, 0);
+
+	ocfs2_init_node_maps(osb);
+
+	snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
+		 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
+
+	init_MUTEX(&osb->recovery_lock);
+
+	osb->disable_recovery = 0;
+	osb->recovery_thread_task = NULL;
+
+	init_waitqueue_head(&osb->checkpoint_event);
+	atomic_set(&osb->needs_checkpoint, 0);
+
+	osb->node_num = O2NM_INVALID_NODE_NUM;
+	osb->slot_num = OCFS2_INVALID_SLOT;
+
+	osb->local_alloc_state = OCFS2_LA_UNUSED;
+	osb->local_alloc_bh = NULL;
+
+	ocfs2_setup_hb_callbacks(osb);
+
+	init_waitqueue_head(&osb->osb_mount_event);
+
+	osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
+	if (!osb->vol_label) {
+		mlog(ML_ERROR, "unable to alloc vol label\n");
+		status = -ENOMEM;
+		goto bail;
+	}
+
+	osb->uuid = kmalloc(OCFS2_VOL_UUID_LEN, GFP_KERNEL);
+	if (!osb->uuid) {
+		mlog(ML_ERROR, "unable to alloc uuid\n");
+		status = -ENOMEM;
+		goto bail;
+	}
+
+	di = (struct ocfs2_dinode *)bh->b_data;
+
+	osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
+	if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
+		mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
+		     osb->max_slots);
+		status = -EINVAL;
+		goto bail;
+	}
+	mlog(ML_NOTICE, "max_slots for this device: %u\n", osb->max_slots);
+
+	osb->s_feature_compat =
+		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
+	osb->s_feature_ro_compat =
+		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
+	osb->s_feature_incompat =
+		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
+
+	if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
+		mlog(ML_ERROR, "couldn't mount because of unsupported "
+		     "optional features (%x).\n", i);
+		status = -EINVAL;
+		goto bail;
+	}
+	if (!(osb->sb->s_flags & MS_RDONLY) &&
+	    (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
+		mlog(ML_ERROR, "couldn't mount RDWR because of "
+		     "unsupported optional features (%x).\n", i);
+		status = -EINVAL;
+		goto bail;
+	}
+
+	get_random_bytes(&osb->s_next_generation, sizeof(u32));
+
+	/* FIXME
+	 * This should be done in ocfs2_journal_init(), but unknown
+	 * ordering issues will cause the filesystem to crash.
+	 * If anyone wants to figure out what part of the code
+	 * refers to osb->journal before ocfs2_journal_init() is run,
+	 * be my guest.
+	 */
+	/* initialize our journal structure */
+
+	journal = kcalloc(1, sizeof(struct ocfs2_journal), GFP_KERNEL);
+	if (!journal) {
+		mlog(ML_ERROR, "unable to alloc journal\n");
+		status = -ENOMEM;
+		goto bail;
+	}
+	osb->journal = journal;
+	journal->j_osb = osb;
+
+	atomic_set(&journal->j_num_trans, 0);
+	init_rwsem(&journal->j_trans_barrier);
+	init_waitqueue_head(&journal->j_checkpointed);
+	spin_lock_init(&journal->j_lock);
+	journal->j_trans_id = (unsigned long) 1;
+	INIT_LIST_HEAD(&journal->j_la_cleanups);
+	INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery, osb);
+	journal->j_state = OCFS2_JOURNAL_FREE;
+
+	/* get some pseudo constants for clustersize bits */
+	osb->s_clustersize_bits =
+		le32_to_cpu(di->id2.i_super.s_clustersize_bits);
+	osb->s_clustersize = 1 << osb->s_clustersize_bits;
+	mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
+
+	if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
+	    osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
+		mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
+		     osb->s_clustersize);
+		status = -EINVAL;
+		goto bail;
+	}
+
+	if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
+	    > (u32)~0UL) {
+		mlog(ML_ERROR, "Volume might try to write to blocks beyond "
+		     "what jbd can address in 32 bits.\n");
+		status = -EINVAL;
+		goto bail;
+	}
+
+	if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
+				 sizeof(di->id2.i_super.s_uuid))) {
+		mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
+		status = -ENOMEM;
+		goto bail;
+	}
+
+	memcpy(&uuid_net_key, &osb->uuid[i], sizeof(osb->net_key));
+	osb->net_key = le32_to_cpu(uuid_net_key);
+
+	strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
+	osb->vol_label[63] = '\0';
+	osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
+	osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
+	osb->first_cluster_group_blkno =
+		le64_to_cpu(di->id2.i_super.s_first_cluster_group);
+	osb->fs_generation = le32_to_cpu(di->i_fs_generation);
+	mlog(0, "vol_label: %s\n", osb->vol_label);
+	mlog(0, "uuid: %s\n", osb->uuid_str);
+	mlog(0, "root_blkno=%"MLFu64", system_dir_blkno=%"MLFu64"\n",
+	     osb->root_blkno, osb->system_dir_blkno);
+
+	osb->osb_dlm_debug = ocfs2_new_dlm_debug();
+	if (!osb->osb_dlm_debug) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	atomic_set(&osb->vol_state, VOLUME_INIT);
+
+	/* load root, system_dir, and all global system inodes */
+	status = ocfs2_init_global_system_inodes(osb);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/*
+	 * global bitmap
+	 */
+	inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
+					    OCFS2_INVALID_SLOT);
+	if (!inode) {
+		status = -EINVAL;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
+
+	status = ocfs2_read_block(osb, osb->bitmap_blkno, &bitmap_bh, 0,
+				  inode);
+	iput(inode);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	di = (struct ocfs2_dinode *) bitmap_bh->b_data;
+	osb->bitmap_cpg = le16_to_cpu(di->id2.i_chain.cl_cpg);
+	osb->num_clusters = le32_to_cpu(di->id1.bitmap1.i_total);
+	brelse(bitmap_bh);
+	mlog(0, "cluster bitmap inode: %"MLFu64", clusters per group: %u\n",
+	     osb->bitmap_blkno, osb->bitmap_cpg);
+
+	status = ocfs2_init_slot_info(osb);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	/*  Link this osb onto the global linked list of all osb structures. */
+	/*  The Global Link List is mainted for the whole driver . */
+	spin_lock(&ocfs2_globals_lock);
+	osb->osb_id = osb_id;
+	if (osb_id < OCFS2_MAX_OSB_ID)
+		osb_id++;
+	else {
+		mlog(ML_ERROR, "Too many volumes mounted\n");
+		status = -ENOMEM;
+	}
+	spin_unlock(&ocfs2_globals_lock);
+
+bail:
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * will return: -EAGAIN if it is ok to keep searching for superblocks
+ *              -EINVAL if there is a bad superblock
+ *              0 on success
+ */
+static int ocfs2_verify_volume(struct ocfs2_dinode *di,
+			       struct buffer_head *bh,
+			       u32 blksz)
+{
+	int status = -EAGAIN;
+
+	mlog_entry_void();
+
+	if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
+		   strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
+		status = -EINVAL;
+		if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
+			mlog(ML_ERROR, "found superblock with incorrect block "
+			     "size: found %u, should be %u\n",
+			     1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
+			       blksz);
+		} else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
+			   OCFS2_MAJOR_REV_LEVEL ||
+			   le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
+			   OCFS2_MINOR_REV_LEVEL) {
+			mlog(ML_ERROR, "found superblock with bad version: "
+			     "found %u.%u, should be %u.%u\n",
+			     le16_to_cpu(di->id2.i_super.s_major_rev_level),
+			     le16_to_cpu(di->id2.i_super.s_minor_rev_level),
+			     OCFS2_MAJOR_REV_LEVEL,
+			     OCFS2_MINOR_REV_LEVEL);
+		} else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
+			mlog(ML_ERROR, "bad block number on superblock: "
+			     "found %"MLFu64", should be %llu\n",
+			     di->i_blkno, (unsigned long long)bh->b_blocknr);
+		} else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
+			    le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
+			mlog(ML_ERROR, "bad cluster size found: %u\n",
+			     1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
+		} else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
+			mlog(ML_ERROR, "bad root_blkno: 0\n");
+		} else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
+			mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
+		} else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
+			mlog(ML_ERROR,
+			     "Superblock slots found greater than file system "
+			     "maximum: found %u, max %u\n",
+			     le16_to_cpu(di->id2.i_super.s_max_slots),
+			     OCFS2_MAX_SLOTS);
+		} else {
+			/* found it! */
+			status = 0;
+		}
+	}
+
+	mlog_exit(status);
+	return status;
+}
+
+static int ocfs2_check_volume(struct ocfs2_super *osb)
+{
+	int status = 0;
+	int dirty;
+	struct ocfs2_dinode *local_alloc = NULL; /* only used if we
+						  * recover
+						  * ourselves. */
+
+	mlog_entry_void();
+
+	/* Init our journal object. */
+	status = ocfs2_journal_init(osb->journal, &dirty);
+	if (status < 0) {
+		mlog(ML_ERROR, "Could not initialize journal!\n");
+		goto finally;
+	}
+
+	/* If the journal was unmounted cleanly then we don't want to
+	 * recover anything. Otherwise, journal_load will do that
+	 * dirty work for us :) */
+	if (!dirty) {
+		status = ocfs2_journal_wipe(osb->journal, 0);
+		if (status < 0) {
+			mlog_errno(status);
+			goto finally;
+		}
+	} else {
+		mlog(ML_NOTICE, "File system was not unmounted cleanly, "
+		     "recovering volume.\n");
+	}
+
+	/* will play back anything left in the journal. */
+	ocfs2_journal_load(osb->journal);
+
+	if (dirty) {
+		/* recover my local alloc if we didn't unmount cleanly. */
+		status = ocfs2_begin_local_alloc_recovery(osb,
+							  osb->slot_num,
+							  &local_alloc);
+		if (status < 0) {
+			mlog_errno(status);
+			goto finally;
+		}
+		/* we complete the recovery process after we've marked
+		 * ourselves as mounted. */
+	}
+
+	mlog(0, "Journal loaded.\n");
+
+	status = ocfs2_load_local_alloc(osb);
+	if (status < 0) {
+		mlog_errno(status);
+		goto finally;
+	}
+
+	if (dirty) {
+		/* Recovery will be completed after we've mounted the
+		 * rest of the volume. */
+		osb->dirty = 1;
+		osb->local_alloc_copy = local_alloc;
+		local_alloc = NULL;
+	}
+
+	/* go through each journal, trylock it and if you get the
+	 * lock, and it's marked as dirty, set the bit in the recover
+	 * map and launch a recovery thread for it. */
+	status = ocfs2_mark_dead_nodes(osb);
+	if (status < 0)
+		mlog_errno(status);
+
+finally:
+	if (local_alloc)
+		kfree(local_alloc);
+
+	mlog_exit(status);
+	return status;
+}
+
+/*
+ * The routine gets called from dismount or close whenever a dismount on
+ * volume is requested and the osb open count becomes 1.
+ * It will remove the osb from the global list and also free up all the
+ * initialized resources and fileobject.
+ */
+static void ocfs2_delete_osb(struct ocfs2_super *osb)
+{
+	mlog_entry_void();
+
+	/* This function assumes that the caller has the main osb resource */
+
+	if (osb->slot_info)
+		ocfs2_free_slot_info(osb->slot_info);
+
+	/* FIXME
+	 * This belongs in journal shutdown, but because we have to
+	 * allocate osb->journal at the start of ocfs2_initalize_osb(),
+	 * we free it here.
+	 */
+	kfree(osb->journal);
+	if (osb->local_alloc_copy)
+		kfree(osb->local_alloc_copy);
+	kfree(osb->uuid_str);
+	ocfs2_put_dlm_debug(osb->osb_dlm_debug);
+	memset(osb, 0, sizeof(struct ocfs2_super));
+
+	mlog_exit_void();
+}
+
+/* Put OCFS2 into a readonly state, or (if the user specifies it),
+ * panic(). We do not support continue-on-error operation. */
+static void ocfs2_handle_error(struct super_block *sb)
+{
+	struct ocfs2_super *osb = OCFS2_SB(sb);
+
+	if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
+		panic("OCFS2: (device %s): panic forced after error\n",
+		      sb->s_id);
+
+	ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
+
+	if (sb->s_flags & MS_RDONLY &&
+	    (ocfs2_is_soft_readonly(osb) ||
+	     ocfs2_is_hard_readonly(osb)))
+		return;
+
+	printk(KERN_CRIT "File system is now read-only due to the potential "
+	       "of on-disk corruption. Please run fsck.ocfs2 once the file "
+	       "system is unmounted.\n");
+	sb->s_flags |= MS_RDONLY;
+	ocfs2_set_ro_flag(osb, 0);
+}
+
+static char error_buf[1024];
+
+void __ocfs2_error(struct super_block *sb,
+		   const char *function,
+		   const char *fmt, ...)
+{
+	va_list args;
+
+	va_start(args, fmt);
+	vsprintf(error_buf, fmt, args);
+	va_end(args);
+
+	/* Not using mlog here because we want to show the actual
+	 * function the error came from. */
+	printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
+	       sb->s_id, function, error_buf);
+
+	ocfs2_handle_error(sb);
+}
+
+/* Handle critical errors. This is intentionally more drastic than
+ * ocfs2_handle_error, so we only use for things like journal errors,
+ * etc. */
+void __ocfs2_abort(struct super_block* sb,
+		   const char *function,
+		   const char *fmt, ...)
+{
+	va_list args;
+
+	va_start(args, fmt);
+	vsprintf(error_buf, fmt, args);
+	va_end(args);
+
+	printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
+	       sb->s_id, function, error_buf);
+
+	/* We don't have the cluster support yet to go straight to
+	 * hard readonly in here. Until then, we want to keep
+	 * ocfs2_abort() so that we can at least mark critical
+	 * errors.
+	 *
+	 * TODO: This should abort the journal and alert other nodes
+	 * that our slot needs recovery. */
+
+	/* Force a panic(). This stinks, but it's better than letting
+	 * things continue without having a proper hard readonly
+	 * here. */
+	OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
+	ocfs2_handle_error(sb);
+}
+
+module_init(ocfs2_init);
+module_exit(ocfs2_exit);
diff --git a/fs/ocfs2/super.h b/fs/ocfs2/super.h
new file mode 100644
index 00000000000..c564177dfbd
--- /dev/null
+++ b/fs/ocfs2/super.h
@@ -0,0 +1,44 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * super.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_SUPER_H
+#define OCFS2_SUPER_H
+
+extern struct workqueue_struct *ocfs2_wq;
+
+int ocfs2_publish_get_mount_state(struct ocfs2_super *osb,
+				  int node_num);
+
+void __ocfs2_error(struct super_block *sb,
+		   const char *function,
+		   const char *fmt, ...);
+#define ocfs2_error(sb, fmt, args...) __ocfs2_error(sb, __PRETTY_FUNCTION__, fmt, ##args)
+
+void __ocfs2_abort(struct super_block *sb,
+		   const char *function,
+		   const char *fmt, ...);
+#define ocfs2_abort(sb, fmt, args...) __ocfs2_abort(sb, __PRETTY_FUNCTION__, fmt, ##args)
+
+#endif /* OCFS2_SUPER_H */
diff --git a/fs/ocfs2/symlink.c b/fs/ocfs2/symlink.c
new file mode 100644
index 00000000000..f6986bd79e7
--- /dev/null
+++ b/fs/ocfs2/symlink.c
@@ -0,0 +1,180 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ *  linux/cluster/ssi/cfs/symlink.c
+ *
+ *	This program 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 2 of
+ *	the License, or (at your option) any later version.
+ *
+ *	This program is distributed in the hope that it will be useful,
+ *	but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *	MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE
+ *	or NON INFRINGEMENT.  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, write to the Free Software
+ * 	Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ *
+ *	Questions/Comments/Bugfixes to ssic-linux-devel@lists.sourceforge.net
+ *
+ *  Copyright (C) 1992  Rick Sladkey
+ *
+ *  Optimization changes Copyright (C) 1994 Florian La Roche
+ *
+ *  Jun 7 1999, cache symlink lookups in the page cache.  -DaveM
+ *
+ *  Portions Copyright (C) 2001 Compaq Computer Corporation
+ *
+ *  ocfs2 symlink handling code.
+ *
+ *  Copyright (C) 2004, 2005 Oracle.
+ *
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/pagemap.h>
+#include <linux/utsname.h>
+
+#define MLOG_MASK_PREFIX ML_NAMEI
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "file.h"
+#include "inode.h"
+#include "journal.h"
+#include "symlink.h"
+
+#include "buffer_head_io.h"
+
+static char *ocfs2_page_getlink(struct dentry * dentry,
+				struct page **ppage);
+static char *ocfs2_fast_symlink_getlink(struct inode *inode,
+					struct buffer_head **bh);
+
+/* get the link contents into pagecache */
+static char *ocfs2_page_getlink(struct dentry * dentry,
+				struct page **ppage)
+{
+	struct page * page;
+	struct address_space *mapping = dentry->d_inode->i_mapping;
+	page = read_cache_page(mapping, 0,
+			       (filler_t *)mapping->a_ops->readpage, NULL);
+	if (IS_ERR(page))
+		goto sync_fail;
+	wait_on_page_locked(page);
+	if (!PageUptodate(page))
+		goto async_fail;
+	*ppage = page;
+	return kmap(page);
+
+async_fail:
+	page_cache_release(page);
+	return ERR_PTR(-EIO);
+
+sync_fail:
+	return (char*)page;
+}
+
+static char *ocfs2_fast_symlink_getlink(struct inode *inode,
+					struct buffer_head **bh)
+{
+	int status;
+	char *link = NULL;
+	struct ocfs2_dinode *fe;
+
+	mlog_entry_void();
+
+	status = ocfs2_read_block(OCFS2_SB(inode->i_sb),
+				  OCFS2_I(inode)->ip_blkno,
+				  bh,
+				  OCFS2_BH_CACHED,
+				  inode);
+	if (status < 0) {
+		mlog_errno(status);
+		link = ERR_PTR(status);
+		goto bail;
+	}
+
+	fe = (struct ocfs2_dinode *) (*bh)->b_data;
+	link = (char *) fe->id2.i_symlink;
+bail:
+	mlog_exit(status);
+
+	return link;
+}
+
+static int ocfs2_readlink(struct dentry *dentry,
+			  char __user *buffer,
+			  int buflen)
+{
+	int ret;
+	char *link;
+	struct buffer_head *bh = NULL;
+	struct inode *inode = dentry->d_inode;
+
+	mlog_entry_void();
+
+	link = ocfs2_fast_symlink_getlink(inode, &bh);
+	if (IS_ERR(link)) {
+		ret = PTR_ERR(link);
+		goto out;
+	}
+
+	ret = vfs_readlink(dentry, buffer, buflen, link);
+
+	brelse(bh);
+out:
+	mlog_exit(ret);
+	return ret;
+}
+
+static void *ocfs2_follow_link(struct dentry *dentry,
+			       struct nameidata *nd)
+{
+	int status;
+	char *link;
+	struct inode *inode = dentry->d_inode;
+	struct page *page = NULL;
+	struct buffer_head *bh = NULL;
+	
+	if (ocfs2_inode_is_fast_symlink(inode))
+		link = ocfs2_fast_symlink_getlink(inode, &bh);
+	else
+		link = ocfs2_page_getlink(dentry, &page);
+	if (IS_ERR(link)) {
+		status = PTR_ERR(link);
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = vfs_follow_link(nd, link);
+	if (status)
+		mlog_errno(status);
+bail:
+	if (page) {
+		kunmap(page);
+		page_cache_release(page);
+	}
+	if (bh)
+		brelse(bh);
+
+	return ERR_PTR(status);
+}
+
+struct inode_operations ocfs2_symlink_inode_operations = {
+	.readlink	= page_readlink,
+	.follow_link	= ocfs2_follow_link,
+	.getattr	= ocfs2_getattr,
+};
+struct inode_operations ocfs2_fast_symlink_inode_operations = {
+	.readlink	= ocfs2_readlink,
+	.follow_link	= ocfs2_follow_link,
+	.getattr	= ocfs2_getattr,
+};
diff --git a/fs/ocfs2/symlink.h b/fs/ocfs2/symlink.h
new file mode 100644
index 00000000000..1ea9e4d9e9e
--- /dev/null
+++ b/fs/ocfs2/symlink.h
@@ -0,0 +1,42 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * symlink.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_SYMLINK_H
+#define OCFS2_SYMLINK_H
+
+extern struct inode_operations ocfs2_symlink_inode_operations;
+extern struct inode_operations ocfs2_fast_symlink_inode_operations;
+
+/*
+ * Test whether an inode is a fast symlink.
+ */
+static inline int ocfs2_inode_is_fast_symlink(struct inode *inode)
+{
+	return (S_ISLNK(inode->i_mode) &&
+		inode->i_blocks == 0);
+}
+
+
+#endif /* OCFS2_SYMLINK_H */
diff --git a/fs/ocfs2/sysfile.c b/fs/ocfs2/sysfile.c
new file mode 100644
index 00000000000..600a8bc5b54
--- /dev/null
+++ b/fs/ocfs2/sysfile.c
@@ -0,0 +1,131 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * sysfile.c
+ *
+ * Initialize, read, write, etc. system files.
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+
+#include "ocfs2.h"
+
+#define MLOG_MASK_PREFIX ML_INODE
+#include <cluster/masklog.h>
+
+#include "alloc.h"
+#include "dir.h"
+#include "inode.h"
+#include "journal.h"
+#include "sysfile.h"
+
+#include "buffer_head_io.h"
+
+static struct inode * _ocfs2_get_system_file_inode(struct ocfs2_super *osb,
+						   int type,
+						   u32 slot);
+
+static inline int is_global_system_inode(int type);
+static inline int is_in_system_inode_array(struct ocfs2_super *osb,
+					   int type,
+					   u32 slot);
+
+static inline int is_global_system_inode(int type)
+{
+	return type >= OCFS2_FIRST_ONLINE_SYSTEM_INODE &&
+		type <= OCFS2_LAST_GLOBAL_SYSTEM_INODE;
+}
+
+static inline int is_in_system_inode_array(struct ocfs2_super *osb,
+					   int type,
+					   u32 slot)
+{
+	return slot == osb->slot_num || is_global_system_inode(type);
+}
+
+struct inode *ocfs2_get_system_file_inode(struct ocfs2_super *osb,
+					  int type,
+					  u32 slot)
+{
+	struct inode *inode = NULL;
+	struct inode **arr = NULL;
+
+	/* avoid the lookup if cached in local system file array */
+	if (is_in_system_inode_array(osb, type, slot))
+		arr = &(osb->system_inodes[type]);
+
+	if (arr && ((inode = *arr) != NULL)) {
+		/* get a ref in addition to the array ref */
+		inode = igrab(inode);
+		if (!inode)
+			BUG();
+
+		return inode;
+	}
+
+	/* this gets one ref thru iget */
+	inode = _ocfs2_get_system_file_inode(osb, type, slot);
+
+	/* add one more if putting into array for first time */
+	if (arr && inode) {
+		*arr = igrab(inode);
+		if (!*arr)
+			BUG();
+	}
+	return inode;
+}
+
+static struct inode * _ocfs2_get_system_file_inode(struct ocfs2_super *osb,
+						   int type,
+						   u32 slot)
+{
+	char namebuf[40];
+	struct inode *inode = NULL;
+	u64 blkno;
+	struct buffer_head *dirent_bh = NULL;
+	struct ocfs2_dir_entry *de = NULL;
+	int status = 0;
+
+	ocfs2_sprintf_system_inode_name(namebuf,
+					sizeof(namebuf),
+					type, slot);
+
+	status = ocfs2_find_files_on_disk(namebuf, strlen(namebuf),
+					  &blkno, osb->sys_root_inode,
+					  &dirent_bh, &de);
+	if (status < 0) {
+		goto bail;
+	}
+
+	inode = ocfs2_iget(osb, blkno);
+	if (IS_ERR(inode)) {
+		mlog_errno(PTR_ERR(inode));
+		inode = NULL;
+		goto bail;
+	}
+bail:
+	if (dirent_bh)
+		brelse(dirent_bh);
+	return inode;
+}
+
diff --git a/fs/ocfs2/sysfile.h b/fs/ocfs2/sysfile.h
new file mode 100644
index 00000000000..cc9ea661ffc
--- /dev/null
+++ b/fs/ocfs2/sysfile.h
@@ -0,0 +1,33 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * sysfile.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_SYSFILE_H
+#define OCFS2_SYSFILE_H
+
+struct inode * ocfs2_get_system_file_inode(struct ocfs2_super *osb,
+					   int type,
+					   u32 slot);
+
+#endif /* OCFS2_SYSFILE_H */
diff --git a/fs/ocfs2/uptodate.c b/fs/ocfs2/uptodate.c
new file mode 100644
index 00000000000..3a0458fd3e1
--- /dev/null
+++ b/fs/ocfs2/uptodate.c
@@ -0,0 +1,544 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * uptodate.c
+ *
+ * Tracking the up-to-date-ness of a local buffer_head with respect to
+ * the cluster.
+ *
+ * Copyright (C) 2002, 2004, 2005 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ *
+ * Standard buffer head caching flags (uptodate, etc) are insufficient
+ * in a clustered environment - a buffer may be marked up to date on
+ * our local node but could have been modified by another cluster
+ * member. As a result an additional (and performant) caching scheme
+ * is required. A further requirement is that we consume as little
+ * memory as possible - we never pin buffer_head structures in order
+ * to cache them.
+ *
+ * We track the existence of up to date buffers on the inodes which
+ * are associated with them. Because we don't want to pin
+ * buffer_heads, this is only a (strong) hint and several other checks
+ * are made in the I/O path to ensure that we don't use a stale or
+ * invalid buffer without going to disk:
+ *	- buffer_jbd is used liberally - if a bh is in the journal on
+ *	  this node then it *must* be up to date.
+ *	- the standard buffer_uptodate() macro is used to detect buffers
+ *	  which may be invalid (even if we have an up to date tracking
+ * 	  item for them)
+ *
+ * For a full understanding of how this code works together, one
+ * should read the callers in dlmglue.c, the I/O functions in
+ * buffer_head_io.c and ocfs2_journal_access in journal.c
+ */
+
+#include <linux/fs.h>
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/buffer_head.h>
+#include <linux/rbtree.h>
+#include <linux/jbd.h>
+
+#define MLOG_MASK_PREFIX ML_UPTODATE
+
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "inode.h"
+#include "uptodate.h"
+
+struct ocfs2_meta_cache_item {
+	struct rb_node	c_node;
+	sector_t	c_block;
+};
+
+static kmem_cache_t *ocfs2_uptodate_cachep = NULL;
+
+void ocfs2_metadata_cache_init(struct inode *inode)
+{
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+	struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
+
+	oi->ip_flags |= OCFS2_INODE_CACHE_INLINE;
+	ci->ci_num_cached = 0;
+}
+
+/* No lock taken here as 'root' is not expected to be visible to other
+ * processes. */
+static unsigned int ocfs2_purge_copied_metadata_tree(struct rb_root *root)
+{
+	unsigned int purged = 0;
+	struct rb_node *node;
+	struct ocfs2_meta_cache_item *item;
+
+	while ((node = rb_last(root)) != NULL) {
+		item = rb_entry(node, struct ocfs2_meta_cache_item, c_node);
+
+		mlog(0, "Purge item %llu\n",
+		     (unsigned long long) item->c_block);
+
+		rb_erase(&item->c_node, root);
+		kmem_cache_free(ocfs2_uptodate_cachep, item);
+
+		purged++;
+	}
+	return purged;
+}
+
+/* Called from locking and called from ocfs2_clear_inode. Dump the
+ * cache for a given inode.
+ *
+ * This function is a few more lines longer than necessary due to some
+ * accounting done here, but I think it's worth tracking down those
+ * bugs sooner -- Mark */
+void ocfs2_metadata_cache_purge(struct inode *inode)
+{
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+	unsigned int tree, to_purge, purged;
+	struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
+	struct rb_root root = RB_ROOT;
+
+	spin_lock(&oi->ip_lock);
+	tree = !(oi->ip_flags & OCFS2_INODE_CACHE_INLINE);
+	to_purge = ci->ci_num_cached;
+
+	mlog(0, "Purge %u %s items from Inode %"MLFu64"\n", to_purge,
+	     tree ? "array" : "tree", oi->ip_blkno);
+
+	/* If we're a tree, save off the root so that we can safely
+	 * initialize the cache. We do the work to free tree members
+	 * without the spinlock. */
+	if (tree)
+		root = ci->ci_cache.ci_tree;
+
+	ocfs2_metadata_cache_init(inode);
+	spin_unlock(&oi->ip_lock);
+
+	purged = ocfs2_purge_copied_metadata_tree(&root);
+	/* If possible, track the number wiped so that we can more
+	 * easily detect counting errors. Unfortunately, this is only
+	 * meaningful for trees. */
+	if (tree && purged != to_purge)
+		mlog(ML_ERROR, "Inode %"MLFu64", count = %u, purged = %u\n",
+		     oi->ip_blkno, to_purge, purged);
+}
+
+/* Returns the index in the cache array, -1 if not found.
+ * Requires ip_lock. */
+static int ocfs2_search_cache_array(struct ocfs2_caching_info *ci,
+				    sector_t item)
+{
+	int i;
+
+	for (i = 0; i < ci->ci_num_cached; i++) {
+		if (item == ci->ci_cache.ci_array[i])
+			return i;
+	}
+
+	return -1;
+}
+
+/* Returns the cache item if found, otherwise NULL.
+ * Requires ip_lock. */
+static struct ocfs2_meta_cache_item *
+ocfs2_search_cache_tree(struct ocfs2_caching_info *ci,
+			sector_t block)
+{
+	struct rb_node * n = ci->ci_cache.ci_tree.rb_node;
+	struct ocfs2_meta_cache_item *item = NULL;
+
+	while (n) {
+		item = rb_entry(n, struct ocfs2_meta_cache_item, c_node);
+
+		if (block < item->c_block)
+			n = n->rb_left;
+		else if (block > item->c_block)
+			n = n->rb_right;
+		else
+			return item;
+	}
+
+	return NULL;
+}
+
+static int ocfs2_buffer_cached(struct ocfs2_inode_info *oi,
+			       struct buffer_head *bh)
+{
+	int index = -1;
+	struct ocfs2_meta_cache_item *item = NULL;
+
+	spin_lock(&oi->ip_lock);
+
+	mlog(0, "Inode %"MLFu64", query block %llu (inline = %u)\n",
+	     oi->ip_blkno, (unsigned long long) bh->b_blocknr,
+	     !!(oi->ip_flags & OCFS2_INODE_CACHE_INLINE));
+
+	if (oi->ip_flags & OCFS2_INODE_CACHE_INLINE)
+		index = ocfs2_search_cache_array(&oi->ip_metadata_cache,
+						 bh->b_blocknr);
+	else
+		item = ocfs2_search_cache_tree(&oi->ip_metadata_cache,
+					       bh->b_blocknr);
+
+	spin_unlock(&oi->ip_lock);
+
+	mlog(0, "index = %d, item = %p\n", index, item);
+
+	return (index != -1) || (item != NULL);
+}
+
+/* Warning: even if it returns true, this does *not* guarantee that
+ * the block is stored in our inode metadata cache. */
+int ocfs2_buffer_uptodate(struct inode *inode,
+			  struct buffer_head *bh)
+{
+	/* Doesn't matter if the bh is in our cache or not -- if it's
+	 * not marked uptodate then we know it can't have correct
+	 * data. */
+	if (!buffer_uptodate(bh))
+		return 0;
+
+	/* OCFS2 does not allow multiple nodes to be changing the same
+	 * block at the same time. */
+	if (buffer_jbd(bh))
+		return 1;
+
+	/* Ok, locally the buffer is marked as up to date, now search
+	 * our cache to see if we can trust that. */
+	return ocfs2_buffer_cached(OCFS2_I(inode), bh);
+}
+
+/* Requires ip_lock */
+static void ocfs2_append_cache_array(struct ocfs2_caching_info *ci,
+				     sector_t block)
+{
+	BUG_ON(ci->ci_num_cached >= OCFS2_INODE_MAX_CACHE_ARRAY);
+
+	mlog(0, "block %llu takes position %u\n", (unsigned long long) block,
+	     ci->ci_num_cached);
+
+	ci->ci_cache.ci_array[ci->ci_num_cached] = block;
+	ci->ci_num_cached++;
+}
+
+/* By now the caller should have checked that the item does *not*
+ * exist in the tree.
+ * Requires ip_lock. */
+static void __ocfs2_insert_cache_tree(struct ocfs2_caching_info *ci,
+				      struct ocfs2_meta_cache_item *new)
+{
+	sector_t block = new->c_block;
+	struct rb_node *parent = NULL;
+	struct rb_node **p = &ci->ci_cache.ci_tree.rb_node;
+	struct ocfs2_meta_cache_item *tmp;
+
+	mlog(0, "Insert block %llu num = %u\n", (unsigned long long) block,
+	     ci->ci_num_cached);
+
+	while(*p) {
+		parent = *p;
+
+		tmp = rb_entry(parent, struct ocfs2_meta_cache_item, c_node);
+
+		if (block < tmp->c_block)
+			p = &(*p)->rb_left;
+		else if (block > tmp->c_block)
+			p = &(*p)->rb_right;
+		else {
+			/* This should never happen! */
+			mlog(ML_ERROR, "Duplicate block %llu cached!\n",
+			     (unsigned long long) block);
+			BUG();
+		}
+	}
+
+	rb_link_node(&new->c_node, parent, p);
+	rb_insert_color(&new->c_node, &ci->ci_cache.ci_tree);
+	ci->ci_num_cached++;
+}
+
+static inline int ocfs2_insert_can_use_array(struct ocfs2_inode_info *oi,
+					     struct ocfs2_caching_info *ci)
+{
+	assert_spin_locked(&oi->ip_lock);
+
+	return (oi->ip_flags & OCFS2_INODE_CACHE_INLINE) &&
+		(ci->ci_num_cached < OCFS2_INODE_MAX_CACHE_ARRAY);
+}
+
+/* tree should be exactly OCFS2_INODE_MAX_CACHE_ARRAY wide. NULL the
+ * pointers in tree after we use them - this allows caller to detect
+ * when to free in case of error. */
+static void ocfs2_expand_cache(struct ocfs2_inode_info *oi,
+			       struct ocfs2_meta_cache_item **tree)
+{
+	int i;
+	struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
+
+	mlog_bug_on_msg(ci->ci_num_cached != OCFS2_INODE_MAX_CACHE_ARRAY,
+			"Inode %"MLFu64", num cached = %u, should be %u\n",
+			oi->ip_blkno, ci->ci_num_cached,
+			OCFS2_INODE_MAX_CACHE_ARRAY);
+	mlog_bug_on_msg(!(oi->ip_flags & OCFS2_INODE_CACHE_INLINE),
+			"Inode %"MLFu64" not marked as inline anymore!\n",
+			oi->ip_blkno);
+	assert_spin_locked(&oi->ip_lock);
+
+	/* Be careful to initialize the tree members *first* because
+	 * once the ci_tree is used, the array is junk... */
+	for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++)
+		tree[i]->c_block = ci->ci_cache.ci_array[i];
+
+	oi->ip_flags &= ~OCFS2_INODE_CACHE_INLINE;
+	ci->ci_cache.ci_tree = RB_ROOT;
+	/* this will be set again by __ocfs2_insert_cache_tree */
+	ci->ci_num_cached = 0;
+
+	for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++) {
+		__ocfs2_insert_cache_tree(ci, tree[i]);
+		tree[i] = NULL;
+	}
+
+	mlog(0, "Expanded %"MLFu64" to a tree cache: flags 0x%x, num = %u\n",
+	     oi->ip_blkno, oi->ip_flags, ci->ci_num_cached);
+}
+
+/* Slow path function - memory allocation is necessary. See the
+ * comment above ocfs2_set_buffer_uptodate for more information. */
+static void __ocfs2_set_buffer_uptodate(struct ocfs2_inode_info *oi,
+					sector_t block,
+					int expand_tree)
+{
+	int i;
+	struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
+	struct ocfs2_meta_cache_item *new = NULL;
+	struct ocfs2_meta_cache_item *tree[OCFS2_INODE_MAX_CACHE_ARRAY] =
+		{ NULL, };
+
+	mlog(0, "Inode %"MLFu64", block %llu, expand = %d\n",
+	     oi->ip_blkno, (unsigned long long) block, expand_tree);
+
+	new = kmem_cache_alloc(ocfs2_uptodate_cachep, GFP_KERNEL);
+	if (!new) {
+		mlog_errno(-ENOMEM);
+		return;
+	}
+	new->c_block = block;
+
+	if (expand_tree) {
+		/* Do *not* allocate an array here - the removal code
+		 * has no way of tracking that. */
+		for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++) {
+			tree[i] = kmem_cache_alloc(ocfs2_uptodate_cachep,
+						   GFP_KERNEL);
+			if (!tree[i]) {
+				mlog_errno(-ENOMEM);
+				goto out_free;
+			}
+
+			/* These are initialized in ocfs2_expand_cache! */
+		}
+	}
+
+	spin_lock(&oi->ip_lock);
+	if (ocfs2_insert_can_use_array(oi, ci)) {
+		mlog(0, "Someone cleared the tree underneath us\n");
+		/* Ok, items were removed from the cache in between
+		 * locks. Detect this and revert back to the fast path */
+		ocfs2_append_cache_array(ci, block);
+		spin_unlock(&oi->ip_lock);
+		goto out_free;
+	}
+
+	if (expand_tree)
+		ocfs2_expand_cache(oi, tree);
+
+	__ocfs2_insert_cache_tree(ci, new);
+	spin_unlock(&oi->ip_lock);
+
+	new = NULL;
+out_free:
+	if (new)
+		kmem_cache_free(ocfs2_uptodate_cachep, new);
+
+	/* If these were used, then ocfs2_expand_cache re-set them to
+	 * NULL for us. */
+	if (tree[0]) {
+		for(i = 0; i < OCFS2_INODE_MAX_CACHE_ARRAY; i++)
+			if (tree[i])
+				kmem_cache_free(ocfs2_uptodate_cachep,
+						tree[i]);
+	}
+}
+
+/* Item insertion is guarded by ip_io_sem, so the insertion path takes
+ * advantage of this by not rechecking for a duplicate insert during
+ * the slow case. Additionally, if the cache needs to be bumped up to
+ * a tree, the code will not recheck after acquiring the lock --
+ * multiple paths cannot be expanding to a tree at the same time.
+ *
+ * The slow path takes into account that items can be removed
+ * (including the whole tree wiped and reset) when this process it out
+ * allocating memory. In those cases, it reverts back to the fast
+ * path.
+ *
+ * Note that this function may actually fail to insert the block if
+ * memory cannot be allocated. This is not fatal however (but may
+ * result in a performance penalty) */
+void ocfs2_set_buffer_uptodate(struct inode *inode,
+			       struct buffer_head *bh)
+{
+	int expand;
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+	struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
+
+	/* The block may very well exist in our cache already, so avoid
+	 * doing any more work in that case. */
+	if (ocfs2_buffer_cached(oi, bh))
+		return;
+
+	mlog(0, "Inode %"MLFu64", inserting block %llu\n", oi->ip_blkno,
+	     (unsigned long long) bh->b_blocknr);
+
+	/* No need to recheck under spinlock - insertion is guarded by
+	 * ip_io_sem */
+	spin_lock(&oi->ip_lock);
+	if (ocfs2_insert_can_use_array(oi, ci)) {
+		/* Fast case - it's an array and there's a free
+		 * spot. */
+		ocfs2_append_cache_array(ci, bh->b_blocknr);
+		spin_unlock(&oi->ip_lock);
+		return;
+	}
+
+	expand = 0;
+	if (oi->ip_flags & OCFS2_INODE_CACHE_INLINE) {
+		/* We need to bump things up to a tree. */
+		expand = 1;
+	}
+	spin_unlock(&oi->ip_lock);
+
+	__ocfs2_set_buffer_uptodate(oi, bh->b_blocknr, expand);
+}
+
+/* Called against a newly allocated buffer. Most likely nobody should
+ * be able to read this sort of metadata while it's still being
+ * allocated, but this is careful to take ip_io_sem anyway. */
+void ocfs2_set_new_buffer_uptodate(struct inode *inode,
+				   struct buffer_head *bh)
+{
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+
+	/* This should definitely *not* exist in our cache */
+	BUG_ON(ocfs2_buffer_cached(oi, bh));
+
+	set_buffer_uptodate(bh);
+
+	down(&oi->ip_io_sem);
+	ocfs2_set_buffer_uptodate(inode, bh);
+	up(&oi->ip_io_sem);
+}
+
+/* Requires ip_lock. */
+static void ocfs2_remove_metadata_array(struct ocfs2_caching_info *ci,
+					int index)
+{
+	sector_t *array = ci->ci_cache.ci_array;
+	int bytes;
+
+	BUG_ON(index < 0 || index >= OCFS2_INODE_MAX_CACHE_ARRAY);
+	BUG_ON(index >= ci->ci_num_cached);
+	BUG_ON(!ci->ci_num_cached);
+
+	mlog(0, "remove index %d (num_cached = %u\n", index,
+	     ci->ci_num_cached);
+
+	ci->ci_num_cached--;
+
+	/* don't need to copy if the array is now empty, or if we
+	 * removed at the tail */
+	if (ci->ci_num_cached && index < ci->ci_num_cached) {
+		bytes = sizeof(sector_t) * (ci->ci_num_cached - index);
+		memmove(&array[index], &array[index + 1], bytes);
+	}
+}
+
+/* Requires ip_lock. */
+static void ocfs2_remove_metadata_tree(struct ocfs2_caching_info *ci,
+				       struct ocfs2_meta_cache_item *item)
+{
+	mlog(0, "remove block %llu from tree\n",
+	     (unsigned long long) item->c_block);
+
+	rb_erase(&item->c_node, &ci->ci_cache.ci_tree);
+	ci->ci_num_cached--;
+}
+
+/* Called when we remove a chunk of metadata from an inode. We don't
+ * bother reverting things to an inlined array in the case of a remove
+ * which moves us back under the limit. */
+void ocfs2_remove_from_cache(struct inode *inode,
+			     struct buffer_head *bh)
+{
+	int index;
+	sector_t block = bh->b_blocknr;
+	struct ocfs2_meta_cache_item *item = NULL;
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+	struct ocfs2_caching_info *ci = &oi->ip_metadata_cache;
+
+	spin_lock(&oi->ip_lock);
+	mlog(0, "Inode %"MLFu64", remove %llu, items = %u, array = %u\n",
+	     oi->ip_blkno, (unsigned long long) block, ci->ci_num_cached,
+	     oi->ip_flags & OCFS2_INODE_CACHE_INLINE);
+
+	if (oi->ip_flags & OCFS2_INODE_CACHE_INLINE) {
+		index = ocfs2_search_cache_array(ci, block);
+		if (index != -1)
+			ocfs2_remove_metadata_array(ci, index);
+	} else {
+		item = ocfs2_search_cache_tree(ci, block);
+		if (item)
+			ocfs2_remove_metadata_tree(ci, item);
+	}
+	spin_unlock(&oi->ip_lock);
+
+	if (item)
+		kmem_cache_free(ocfs2_uptodate_cachep, item);
+}
+
+int __init init_ocfs2_uptodate_cache(void)
+{
+	ocfs2_uptodate_cachep = kmem_cache_create("ocfs2_uptodate",
+				  sizeof(struct ocfs2_meta_cache_item),
+				  0, SLAB_HWCACHE_ALIGN, NULL, NULL);
+	if (!ocfs2_uptodate_cachep)
+		return -ENOMEM;
+
+	mlog(0, "%u inlined cache items per inode.\n",
+	     OCFS2_INODE_MAX_CACHE_ARRAY);
+
+	return 0;
+}
+
+void __exit exit_ocfs2_uptodate_cache(void)
+{
+	if (ocfs2_uptodate_cachep)
+		kmem_cache_destroy(ocfs2_uptodate_cachep);
+}
diff --git a/fs/ocfs2/uptodate.h b/fs/ocfs2/uptodate.h
new file mode 100644
index 00000000000..e5aacdf4eab
--- /dev/null
+++ b/fs/ocfs2/uptodate.h
@@ -0,0 +1,44 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * uptodate.h
+ *
+ * Cluster uptodate tracking
+ *
+ * Copyright (C) 2002, 2004, 2005 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_UPTODATE_H
+#define OCFS2_UPTODATE_H
+
+int __init init_ocfs2_uptodate_cache(void);
+void __exit exit_ocfs2_uptodate_cache(void);
+
+void ocfs2_metadata_cache_init(struct inode *inode);
+void ocfs2_metadata_cache_purge(struct inode *inode);
+
+int ocfs2_buffer_uptodate(struct inode *inode,
+			  struct buffer_head *bh);
+void ocfs2_set_buffer_uptodate(struct inode *inode,
+			       struct buffer_head *bh);
+void ocfs2_set_new_buffer_uptodate(struct inode *inode,
+				   struct buffer_head *bh);
+void ocfs2_remove_from_cache(struct inode *inode,
+			     struct buffer_head *bh);
+
+#endif /* OCFS2_UPTODATE_H */
diff --git a/fs/ocfs2/ver.c b/fs/ocfs2/ver.c
new file mode 100644
index 00000000000..5405ce121c9
--- /dev/null
+++ b/fs/ocfs2/ver.c
@@ -0,0 +1,43 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ver.c
+ *
+ * version string
+ *
+ * Copyright (C) 2002, 2005 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/module.h>
+#include <linux/string.h>
+#include <linux/kernel.h>
+
+#include "ver.h"
+
+#define OCFS2_BUILD_VERSION "1.3.3"
+
+#define VERSION_STR "OCFS2 " OCFS2_BUILD_VERSION
+
+void ocfs2_print_version(void)
+{
+	printk(KERN_INFO "%s\n", VERSION_STR);
+}
+
+MODULE_DESCRIPTION(VERSION_STR);
+
+MODULE_VERSION(OCFS2_BUILD_VERSION);
diff --git a/fs/ocfs2/ver.h b/fs/ocfs2/ver.h
new file mode 100644
index 00000000000..d7395cb91d2
--- /dev/null
+++ b/fs/ocfs2/ver.h
@@ -0,0 +1,31 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * ver.h
+ *
+ * Function prototypes
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#ifndef OCFS2_VER_H
+#define OCFS2_VER_H
+
+void ocfs2_print_version(void);
+
+#endif /* OCFS2_VER_H */
diff --git a/fs/ocfs2/vote.c b/fs/ocfs2/vote.c
new file mode 100644
index 00000000000..021978e0576
--- /dev/null
+++ b/fs/ocfs2/vote.c
@@ -0,0 +1,1202 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * vote.c
+ *
+ * description here
+ *
+ * Copyright (C) 2003, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+#include <linux/types.h>
+#include <linux/slab.h>
+#include <linux/highmem.h>
+#include <linux/smp_lock.h>
+#include <linux/kthread.h>
+
+#include <cluster/heartbeat.h>
+#include <cluster/nodemanager.h>
+#include <cluster/tcp.h>
+
+#include <dlm/dlmapi.h>
+
+#define MLOG_MASK_PREFIX ML_VOTE
+#include <cluster/masklog.h>
+
+#include "ocfs2.h"
+
+#include "alloc.h"
+#include "dlmglue.h"
+#include "extent_map.h"
+#include "heartbeat.h"
+#include "inode.h"
+#include "journal.h"
+#include "slot_map.h"
+#include "vote.h"
+
+#include "buffer_head_io.h"
+
+#define OCFS2_MESSAGE_TYPE_VOTE     (0x1)
+#define OCFS2_MESSAGE_TYPE_RESPONSE (0x2)
+struct ocfs2_msg_hdr
+{
+	__be32 h_response_id; /* used to lookup message handle on sending
+			    * node. */
+	__be32 h_request;
+	__be64 h_blkno;
+	__be32 h_generation;
+	__be32 h_node_num;    /* node sending this particular message. */
+};
+
+/* OCFS2_MAX_FILENAME_LEN is 255 characters, but we want to align this
+ * for the network. */
+#define OCFS2_VOTE_FILENAME_LEN 256
+struct ocfs2_vote_msg
+{
+	struct ocfs2_msg_hdr v_hdr;
+	union {
+		__be32 v_generic1;
+		__be32 v_orphaned_slot;	/* Used during delete votes */
+		__be32 v_nlink;		/* Used during unlink votes */
+	} md1;				/* Message type dependant 1 */
+	__be32 v_unlink_namelen;
+	__be64 v_unlink_parent;
+	u8  v_unlink_dirent[OCFS2_VOTE_FILENAME_LEN];
+};
+
+/* Responses are given these values to maintain backwards
+ * compatibility with older ocfs2 versions */
+#define OCFS2_RESPONSE_OK		(0)
+#define OCFS2_RESPONSE_BUSY		(-16)
+#define OCFS2_RESPONSE_BAD_MSG		(-22)
+
+struct ocfs2_response_msg
+{
+	struct ocfs2_msg_hdr r_hdr;
+	__be32 r_response;
+	__be32 r_orphaned_slot;
+};
+
+struct ocfs2_vote_work {
+	struct list_head   w_list;
+	struct ocfs2_vote_msg w_msg;
+};
+
+enum ocfs2_vote_request {
+	OCFS2_VOTE_REQ_INVALID = 0,
+	OCFS2_VOTE_REQ_DELETE,
+	OCFS2_VOTE_REQ_UNLINK,
+	OCFS2_VOTE_REQ_RENAME,
+	OCFS2_VOTE_REQ_MOUNT,
+	OCFS2_VOTE_REQ_UMOUNT,
+	OCFS2_VOTE_REQ_LAST
+};
+
+static inline int ocfs2_is_valid_vote_request(int request)
+{
+	return OCFS2_VOTE_REQ_INVALID < request &&
+		request < OCFS2_VOTE_REQ_LAST;
+}
+
+typedef void (*ocfs2_net_response_callback)(void *priv,
+					    struct ocfs2_response_msg *resp);
+struct ocfs2_net_response_cb {
+	ocfs2_net_response_callback	rc_cb;
+	void				*rc_priv;
+};
+
+struct ocfs2_net_wait_ctxt {
+	struct list_head        n_list;
+	u32                     n_response_id;
+	wait_queue_head_t       n_event;
+	struct ocfs2_node_map   n_node_map;
+	int                     n_response; /* an agreggate response. 0 if
+					     * all nodes are go, < 0 on any
+					     * negative response from any
+					     * node or network error. */
+	struct ocfs2_net_response_cb *n_callback;
+};
+
+static void ocfs2_process_mount_request(struct ocfs2_super *osb,
+					unsigned int node_num)
+{
+	mlog(0, "MOUNT vote from node %u\n", node_num);
+	/* The other node only sends us this message when he has an EX
+	 * on the superblock, so our recovery threads (if having been
+	 * launched) are waiting on it.*/
+	ocfs2_recovery_map_clear(osb, node_num);
+	ocfs2_node_map_set_bit(osb, &osb->mounted_map, node_num);
+
+	/* We clear the umount map here because a node may have been
+	 * previously mounted, safely unmounted but never stopped
+	 * heartbeating - in which case we'd have a stale entry. */
+	ocfs2_node_map_clear_bit(osb, &osb->umount_map, node_num);
+}
+
+static void ocfs2_process_umount_request(struct ocfs2_super *osb,
+					 unsigned int node_num)
+{
+	mlog(0, "UMOUNT vote from node %u\n", node_num);
+	ocfs2_node_map_clear_bit(osb, &osb->mounted_map, node_num);
+	ocfs2_node_map_set_bit(osb, &osb->umount_map, node_num);
+}
+
+void ocfs2_mark_inode_remotely_deleted(struct inode *inode)
+{
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+
+	assert_spin_locked(&oi->ip_lock);
+	/* We set the SKIP_DELETE flag on the inode so we don't try to
+	 * delete it in delete_inode ourselves, thus avoiding
+	 * unecessary lock pinging. If the other node failed to wipe
+	 * the inode as a result of a crash, then recovery will pick
+	 * up the slack. */
+	oi->ip_flags |= OCFS2_INODE_DELETED|OCFS2_INODE_SKIP_DELETE;
+}
+
+static int ocfs2_process_delete_request(struct inode *inode,
+					int *orphaned_slot)
+{
+	int response = OCFS2_RESPONSE_BUSY;
+
+	mlog(0, "DELETE vote on inode %lu, read lnk_cnt = %u, slot = %d\n",
+	     inode->i_ino, inode->i_nlink, *orphaned_slot);
+
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+
+	/* Whatever our vote response is, we want to make sure that
+	 * the orphaned slot is recorded properly on this node *and*
+	 * on the requesting node. Technically, if the requesting node
+	 * did not know which slot the inode is orphaned in but we
+	 * respond with BUSY he doesn't actually need the orphaned
+	 * slot, but it doesn't hurt to do it here anyway. */
+	if ((*orphaned_slot) != OCFS2_INVALID_SLOT) {
+		mlog_bug_on_msg(OCFS2_I(inode)->ip_orphaned_slot !=
+				OCFS2_INVALID_SLOT &&
+				OCFS2_I(inode)->ip_orphaned_slot !=
+				(*orphaned_slot),
+				"Inode %"MLFu64": This node thinks it's "
+				"orphaned in slot %d, messaged it's in %d\n",
+				OCFS2_I(inode)->ip_blkno,
+				OCFS2_I(inode)->ip_orphaned_slot,
+				*orphaned_slot);
+
+		mlog(0, "Setting orphaned slot for inode %"MLFu64" to %d\n",
+		     OCFS2_I(inode)->ip_blkno, *orphaned_slot);
+
+		OCFS2_I(inode)->ip_orphaned_slot = *orphaned_slot;
+	} else {
+		mlog(0, "Sending back orphaned slot %d for inode %"MLFu64"\n",
+		     OCFS2_I(inode)->ip_orphaned_slot,
+		     OCFS2_I(inode)->ip_blkno);
+
+		*orphaned_slot = OCFS2_I(inode)->ip_orphaned_slot;
+	}
+
+	/* vote no if the file is still open. */
+	if (OCFS2_I(inode)->ip_open_count) {
+		mlog(0, "open count = %u\n",
+		     OCFS2_I(inode)->ip_open_count);
+		spin_unlock(&OCFS2_I(inode)->ip_lock);
+		goto done;
+	}
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+	/* directories are a bit ugly... What if someone is sitting in
+	 * it? We want to make sure the inode is removed completely as
+	 * a result of the iput in process_vote. */
+	if (S_ISDIR(inode->i_mode) && (atomic_read(&inode->i_count) != 1)) {
+		mlog(0, "i_count = %u\n", atomic_read(&inode->i_count));
+		goto done;
+	}
+
+	if (filemap_fdatawrite(inode->i_mapping)) {
+		mlog(ML_ERROR, "Could not sync inode %"MLFu64" for delete!\n",
+		     OCFS2_I(inode)->ip_blkno);
+		goto done;
+	}
+	sync_mapping_buffers(inode->i_mapping);
+	truncate_inode_pages(inode->i_mapping, 0);
+	ocfs2_extent_map_trunc(inode, 0);
+
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+	/* double check open count - someone might have raced this
+	 * thread into ocfs2_file_open while we were writing out
+	 * data. If we're to allow a wipe of this inode now, we *must*
+	 * hold the spinlock until we've marked it. */
+	if (OCFS2_I(inode)->ip_open_count) {
+		mlog(0, "Raced to wipe! open count = %u\n",
+		     OCFS2_I(inode)->ip_open_count);
+		spin_unlock(&OCFS2_I(inode)->ip_lock);
+		goto done;
+	}
+
+	/* Mark the inode as being wiped from disk. */
+	ocfs2_mark_inode_remotely_deleted(inode);
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+	/* Not sure this is necessary anymore. */
+	d_prune_aliases(inode);
+
+	/* If we get here, then we're voting 'yes', so commit the
+	 * delete on our side. */
+	response = OCFS2_RESPONSE_OK;
+done:
+	return response;
+}
+
+static int ocfs2_match_dentry(struct dentry *dentry,
+			      u64 parent_blkno,
+			      unsigned int namelen,
+			      const char *name)
+{
+	struct inode *parent;
+
+	if (!dentry->d_parent) {
+		mlog(0, "Detached from parent.\n");
+		return 0;
+	}
+
+	parent = dentry->d_parent->d_inode;
+	/* Negative parent dentry? */
+	if (!parent)
+		return 0;
+
+	/* Name is in a different directory. */
+	if (OCFS2_I(parent)->ip_blkno != parent_blkno)
+		return 0;
+
+	if (dentry->d_name.len != namelen)
+		return 0;
+
+	/* comparison above guarantees this is safe. */
+	if (memcmp(dentry->d_name.name, name, namelen))
+		return 0;
+
+	return 1;
+}
+
+static void ocfs2_process_dentry_request(struct inode *inode,
+					 int rename,
+					 unsigned int new_nlink,
+					 u64 parent_blkno,
+					 unsigned int namelen,
+					 const char *name)
+{
+	struct dentry *dentry = NULL;
+	struct list_head *p;
+	struct ocfs2_inode_info *oi = OCFS2_I(inode);
+
+	mlog(0, "parent %"MLFu64", namelen = %u, name = %.*s\n", parent_blkno,
+	     namelen, namelen, name);
+
+	spin_lock(&dcache_lock);
+
+	/* Another node is removing this name from the system. It is
+	 * up to us to find the corresponding dentry and if it exists,
+	 * unhash it from the dcache. */
+	list_for_each(p, &inode->i_dentry) {
+		dentry = list_entry(p, struct dentry, d_alias);
+
+		if (ocfs2_match_dentry(dentry, parent_blkno, namelen, name)) {
+			mlog(0, "dentry found: %.*s\n",
+			     dentry->d_name.len, dentry->d_name.name);
+
+			dget_locked(dentry);
+			break;
+		}
+
+		dentry = NULL;
+	}
+
+	spin_unlock(&dcache_lock);
+
+	if (dentry) {
+		d_delete(dentry);
+		dput(dentry);
+	}
+
+	/* rename votes don't send link counts */
+	if (!rename) {
+		mlog(0, "new_nlink = %u\n", new_nlink);
+
+		/* We don't have the proper locks here to directly
+		 * change i_nlink and besides, the vote is sent
+		 * *before* the operation so it may have failed on the
+		 * other node. This passes a hint to ocfs2_drop_inode
+		 * to force ocfs2_delete_inode, who will take the
+		 * proper cluster locks to sort things out. */
+		if (new_nlink == 0) {
+			spin_lock(&oi->ip_lock);
+			oi->ip_flags |= OCFS2_INODE_MAYBE_ORPHANED;
+			spin_unlock(&OCFS2_I(inode)->ip_lock);
+		}
+	}
+}
+
+static void ocfs2_process_vote(struct ocfs2_super *osb,
+			       struct ocfs2_vote_msg *msg)
+{
+	int net_status, vote_response;
+	int orphaned_slot = 0;
+	int rename = 0;
+	unsigned int node_num, generation, new_nlink, namelen;
+	u64 blkno, parent_blkno;
+	enum ocfs2_vote_request request;
+	struct inode *inode = NULL;
+	struct ocfs2_msg_hdr *hdr = &msg->v_hdr;
+	struct ocfs2_response_msg response;
+
+	/* decode the network mumbo jumbo into local variables. */
+	request = be32_to_cpu(hdr->h_request);
+	blkno = be64_to_cpu(hdr->h_blkno);
+	generation = be32_to_cpu(hdr->h_generation);
+	node_num = be32_to_cpu(hdr->h_node_num);
+	if (request == OCFS2_VOTE_REQ_DELETE)
+		orphaned_slot = be32_to_cpu(msg->md1.v_orphaned_slot);
+
+	mlog(0, "processing vote: request = %u, blkno = %"MLFu64", "
+	     "generation = %u, node_num = %u, priv1 = %u\n", request,
+	     blkno, generation, node_num, be32_to_cpu(msg->md1.v_generic1));
+
+	if (!ocfs2_is_valid_vote_request(request)) {
+		mlog(ML_ERROR, "Invalid vote request %d from node %u\n",
+		     request, node_num);
+		vote_response = OCFS2_RESPONSE_BAD_MSG;
+		goto respond;
+	}
+
+	vote_response = OCFS2_RESPONSE_OK;
+
+	switch (request) {
+	case OCFS2_VOTE_REQ_UMOUNT:
+		ocfs2_process_umount_request(osb, node_num);
+		goto respond;
+	case OCFS2_VOTE_REQ_MOUNT:
+		ocfs2_process_mount_request(osb, node_num);
+		goto respond;
+	default:
+		/* avoids a gcc warning */
+		break;
+	}
+
+	/* We cannot process the remaining message types before we're
+	 * fully mounted. It's perfectly safe however to send a 'yes'
+	 * response as we can't possibly have any of the state they're
+	 * asking us to modify yet. */
+	if (atomic_read(&osb->vol_state) == VOLUME_INIT)
+		goto respond;
+
+	/* If we get here, then the request is against an inode. */
+	inode = ocfs2_ilookup_for_vote(osb, blkno,
+				       request == OCFS2_VOTE_REQ_DELETE);
+
+	/* Not finding the inode is perfectly valid - it means we're
+	 * not interested in what the other node is about to do to it
+	 * so in those cases we automatically respond with an
+	 * affirmative. Cluster locking ensures that we won't race
+	 * interest in the inode with this vote request. */
+	if (!inode)
+		goto respond;
+
+	/* Check generation values. It's possible for us to get a
+	 * request against a stale inode. If so then we proceed as if
+	 * we had not found an inode in the first place. */
+	if (inode->i_generation != generation) {
+		mlog(0, "generation passed %u != inode generation = %u, "
+		     "ip_flags = %x, ip_blkno = %"MLFu64", msg %"MLFu64", "
+		     "i_count = %u, message type = %u\n",
+		     generation, inode->i_generation, OCFS2_I(inode)->ip_flags,
+		     OCFS2_I(inode)->ip_blkno, blkno,
+		     atomic_read(&inode->i_count), request);
+		iput(inode);
+		inode = NULL;
+		goto respond;
+	}
+
+	switch (request) {
+	case OCFS2_VOTE_REQ_DELETE:
+		vote_response = ocfs2_process_delete_request(inode,
+							     &orphaned_slot);
+		break;
+	case OCFS2_VOTE_REQ_RENAME:
+		rename = 1;
+		/* fall through */
+	case OCFS2_VOTE_REQ_UNLINK:
+		parent_blkno = be64_to_cpu(msg->v_unlink_parent);
+		namelen = be32_to_cpu(msg->v_unlink_namelen);
+		/* new_nlink will be ignored in case of a rename vote */
+		new_nlink = be32_to_cpu(msg->md1.v_nlink);
+		ocfs2_process_dentry_request(inode, rename, new_nlink,
+					     parent_blkno, namelen,
+					     msg->v_unlink_dirent);
+		break;
+	default:
+		mlog(ML_ERROR, "node %u, invalid request: %u\n",
+		     node_num, request);
+		vote_response = OCFS2_RESPONSE_BAD_MSG;
+	}
+
+respond:
+	/* Response struture is small so we just put it on the stack
+	 * and stuff it inline. */
+	memset(&response, 0, sizeof(struct ocfs2_response_msg));
+	response.r_hdr.h_response_id = hdr->h_response_id;
+	response.r_hdr.h_blkno = hdr->h_blkno;
+	response.r_hdr.h_generation = hdr->h_generation;
+	response.r_hdr.h_node_num = cpu_to_be32(osb->node_num);
+	response.r_response = cpu_to_be32(vote_response);
+	response.r_orphaned_slot = cpu_to_be32(orphaned_slot);
+
+	net_status = o2net_send_message(OCFS2_MESSAGE_TYPE_RESPONSE,
+					osb->net_key,
+					&response,
+					sizeof(struct ocfs2_response_msg),
+					node_num,
+					NULL);
+	/* We still want to error print for ENOPROTOOPT here. The
+	 * sending node shouldn't have unregistered his net handler
+	 * without sending an unmount vote 1st */
+	if (net_status < 0
+	    && net_status != -ETIMEDOUT
+	    && net_status != -ENOTCONN)
+		mlog(ML_ERROR, "message to node %u fails with error %d!\n",
+		     node_num, net_status);
+
+	if (inode)
+		iput(inode);
+}
+
+static void ocfs2_vote_thread_do_work(struct ocfs2_super *osb)
+{
+	unsigned long processed;
+	struct ocfs2_lock_res *lockres;
+	struct ocfs2_vote_work *work;
+
+	mlog_entry_void();
+
+	spin_lock(&osb->vote_task_lock);
+	/* grab this early so we know to try again if a state change and
+	 * wake happens part-way through our work  */
+	osb->vote_work_sequence = osb->vote_wake_sequence;
+
+	processed = osb->blocked_lock_count;
+	while (processed) {
+		BUG_ON(list_empty(&osb->blocked_lock_list));
+
+		lockres = list_entry(osb->blocked_lock_list.next,
+				     struct ocfs2_lock_res, l_blocked_list);
+		list_del_init(&lockres->l_blocked_list);
+		osb->blocked_lock_count--;
+		spin_unlock(&osb->vote_task_lock);
+
+		BUG_ON(!processed);
+		processed--;
+
+		ocfs2_process_blocked_lock(osb, lockres);
+
+		spin_lock(&osb->vote_task_lock);
+	}
+
+	while (osb->vote_count) {
+		BUG_ON(list_empty(&osb->vote_list));
+		work = list_entry(osb->vote_list.next,
+				  struct ocfs2_vote_work, w_list);
+		list_del(&work->w_list);
+		osb->vote_count--;
+		spin_unlock(&osb->vote_task_lock);
+
+		ocfs2_process_vote(osb, &work->w_msg);
+		kfree(work);
+
+		spin_lock(&osb->vote_task_lock);
+	}
+	spin_unlock(&osb->vote_task_lock);
+
+	mlog_exit_void();
+}
+
+static int ocfs2_vote_thread_lists_empty(struct ocfs2_super *osb)
+{
+	int empty = 0;
+
+	spin_lock(&osb->vote_task_lock);
+	if (list_empty(&osb->blocked_lock_list) &&
+	    list_empty(&osb->vote_list))
+		empty = 1;
+
+	spin_unlock(&osb->vote_task_lock);
+	return empty;
+}
+
+static int ocfs2_vote_thread_should_wake(struct ocfs2_super *osb)
+{
+	int should_wake = 0;
+
+	spin_lock(&osb->vote_task_lock);
+	if (osb->vote_work_sequence != osb->vote_wake_sequence)
+		should_wake = 1;
+	spin_unlock(&osb->vote_task_lock);
+
+	return should_wake;
+}
+
+int ocfs2_vote_thread(void *arg)
+{
+	int status = 0;
+	struct ocfs2_super *osb = arg;
+
+	/* only quit once we've been asked to stop and there is no more
+	 * work available */
+	while (!(kthread_should_stop() &&
+		 ocfs2_vote_thread_lists_empty(osb))) {
+
+		wait_event_interruptible(osb->vote_event,
+					 ocfs2_vote_thread_should_wake(osb) ||
+					 kthread_should_stop());
+
+		mlog(0, "vote_thread: awoken\n");
+
+		ocfs2_vote_thread_do_work(osb);
+	}
+
+	osb->vote_task = NULL;
+	return status;
+}
+
+static struct ocfs2_net_wait_ctxt *ocfs2_new_net_wait_ctxt(unsigned int response_id)
+{
+	struct ocfs2_net_wait_ctxt *w;
+
+	w = kcalloc(1, sizeof(*w), GFP_KERNEL);
+	if (!w) {
+		mlog_errno(-ENOMEM);
+		goto bail;
+	}
+
+	INIT_LIST_HEAD(&w->n_list);
+	init_waitqueue_head(&w->n_event);
+	ocfs2_node_map_init(&w->n_node_map);
+	w->n_response_id = response_id;
+	w->n_callback = NULL;
+bail:
+	return w;
+}
+
+static unsigned int ocfs2_new_response_id(struct ocfs2_super *osb)
+{
+	unsigned int ret;
+
+	spin_lock(&osb->net_response_lock);
+	ret = ++osb->net_response_ids;
+	spin_unlock(&osb->net_response_lock);
+
+	return ret;
+}
+
+static void ocfs2_dequeue_net_wait_ctxt(struct ocfs2_super *osb,
+					struct ocfs2_net_wait_ctxt *w)
+{
+	spin_lock(&osb->net_response_lock);
+	list_del(&w->n_list);
+	spin_unlock(&osb->net_response_lock);
+}
+
+static void ocfs2_queue_net_wait_ctxt(struct ocfs2_super *osb,
+				      struct ocfs2_net_wait_ctxt *w)
+{
+	spin_lock(&osb->net_response_lock);
+	list_add_tail(&w->n_list,
+		      &osb->net_response_list);
+	spin_unlock(&osb->net_response_lock);
+}
+
+static void __ocfs2_mark_node_responded(struct ocfs2_super *osb,
+					struct ocfs2_net_wait_ctxt *w,
+					int node_num)
+{
+	assert_spin_locked(&osb->net_response_lock);
+
+	ocfs2_node_map_clear_bit(osb, &w->n_node_map, node_num);
+	if (ocfs2_node_map_is_empty(osb, &w->n_node_map))
+		wake_up(&w->n_event);
+}
+
+/* Intended to be called from the node down callback, we fake remove
+ * the node from all our response contexts */
+void ocfs2_remove_node_from_vote_queues(struct ocfs2_super *osb,
+					int node_num)
+{
+	struct list_head *p;
+	struct ocfs2_net_wait_ctxt *w = NULL;
+
+	spin_lock(&osb->net_response_lock);
+
+	list_for_each(p, &osb->net_response_list) {
+		w = list_entry(p, struct ocfs2_net_wait_ctxt, n_list);
+
+		__ocfs2_mark_node_responded(osb, w, node_num);
+	}
+
+	spin_unlock(&osb->net_response_lock);
+}
+
+static int ocfs2_broadcast_vote(struct ocfs2_super *osb,
+				struct ocfs2_vote_msg *request,
+				unsigned int response_id,
+				int *response,
+				struct ocfs2_net_response_cb *callback)
+{
+	int status, i, remote_err;
+	struct ocfs2_net_wait_ctxt *w = NULL;
+	int dequeued = 0;
+
+	mlog_entry_void();
+
+	w = ocfs2_new_net_wait_ctxt(response_id);
+	if (!w) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+	w->n_callback = callback;
+
+	/* we're pretty much ready to go at this point, and this fills
+	 * in n_response which we need anyway... */
+	ocfs2_queue_net_wait_ctxt(osb, w);
+
+	i = ocfs2_node_map_iterate(osb, &osb->mounted_map, 0);
+
+	while (i != O2NM_INVALID_NODE_NUM) {
+		if (i != osb->node_num) {
+			mlog(0, "trying to send request to node %i\n", i);
+			ocfs2_node_map_set_bit(osb, &w->n_node_map, i);
+
+			remote_err = 0;
+			status = o2net_send_message(OCFS2_MESSAGE_TYPE_VOTE,
+						    osb->net_key,
+						    request,
+						    sizeof(*request),
+						    i,
+						    &remote_err);
+			if (status == -ETIMEDOUT) {
+				mlog(0, "remote node %d timed out!\n", i);
+				status = -EAGAIN;
+				goto bail;
+			}
+			if (remote_err < 0) {
+				status = remote_err;
+				mlog(0, "remote error %d on node %d!\n",
+				     remote_err, i);
+				mlog_errno(status);
+				goto bail;
+			}
+			if (status < 0) {
+				mlog_errno(status);
+				goto bail;
+			}
+		}
+		i++;
+		i = ocfs2_node_map_iterate(osb, &osb->mounted_map, i);
+		mlog(0, "next is %d, i am %d\n", i, osb->node_num);
+	}
+	mlog(0, "done sending, now waiting on responses...\n");
+
+	wait_event(w->n_event, ocfs2_node_map_is_empty(osb, &w->n_node_map));
+
+	ocfs2_dequeue_net_wait_ctxt(osb, w);
+	dequeued = 1;
+
+	*response = w->n_response;
+	status = 0;
+bail:
+	if (w) {
+		if (!dequeued)
+			ocfs2_dequeue_net_wait_ctxt(osb, w);
+		kfree(w);
+	}
+
+	mlog_exit(status);
+	return status;
+}
+
+static struct ocfs2_vote_msg * ocfs2_new_vote_request(struct ocfs2_super *osb,
+						      u64 blkno,
+						      unsigned int generation,
+						      enum ocfs2_vote_request type,
+						      u32 priv)
+{
+	struct ocfs2_vote_msg *request;
+	struct ocfs2_msg_hdr *hdr;
+
+	BUG_ON(!ocfs2_is_valid_vote_request(type));
+
+	request = kcalloc(1, sizeof(*request), GFP_KERNEL);
+	if (!request) {
+		mlog_errno(-ENOMEM);
+	} else {
+		hdr = &request->v_hdr;
+		hdr->h_node_num = cpu_to_be32(osb->node_num);
+		hdr->h_request = cpu_to_be32(type);
+		hdr->h_blkno = cpu_to_be64(blkno);
+		hdr->h_generation = cpu_to_be32(generation);
+
+		request->md1.v_generic1 = cpu_to_be32(priv);
+	}
+
+	return request;
+}
+
+/* Complete the buildup of a new vote request and process the
+ * broadcast return value. */
+static int ocfs2_do_request_vote(struct ocfs2_super *osb,
+				 struct ocfs2_vote_msg *request,
+				 struct ocfs2_net_response_cb *callback)
+{
+	int status, response;
+	unsigned int response_id;
+	struct ocfs2_msg_hdr *hdr;
+
+	response_id = ocfs2_new_response_id(osb);
+
+	hdr = &request->v_hdr;
+	hdr->h_response_id = cpu_to_be32(response_id);
+
+	status = ocfs2_broadcast_vote(osb, request, response_id, &response,
+				      callback);
+	if (status < 0) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = response;
+bail:
+
+	return status;
+}
+
+static int ocfs2_request_vote(struct inode *inode,
+			      struct ocfs2_vote_msg *request,
+			      struct ocfs2_net_response_cb *callback)
+{
+	int status;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+
+	if (ocfs2_inode_is_new(inode))
+		return 0;
+
+	status = -EAGAIN;
+	while (status == -EAGAIN) {
+		if (!(osb->s_mount_opt & OCFS2_MOUNT_NOINTR) &&
+		    signal_pending(current))
+			return -ERESTARTSYS;
+
+		status = ocfs2_super_lock(osb, 0);
+		if (status < 0) {
+			mlog_errno(status);
+			break;
+		}
+
+		status = 0;
+		if (!ocfs2_node_map_is_only(osb, &osb->mounted_map,
+					   osb->node_num))
+			status = ocfs2_do_request_vote(osb, request, callback);
+
+		ocfs2_super_unlock(osb, 0);
+	}
+	return status;
+}
+
+static void ocfs2_delete_response_cb(void *priv,
+				     struct ocfs2_response_msg *resp)
+{
+	int orphaned_slot, node;
+	struct inode *inode = priv;
+
+	orphaned_slot = be32_to_cpu(resp->r_orphaned_slot);
+	node = be32_to_cpu(resp->r_hdr.h_node_num);
+	mlog(0, "node %d tells us that inode %"MLFu64" is orphaned in slot "
+	     "%d\n", node, OCFS2_I(inode)->ip_blkno, orphaned_slot);
+
+	/* The other node may not actually know which slot the inode
+	 * is orphaned in. */
+	if (orphaned_slot == OCFS2_INVALID_SLOT)
+		return;
+
+	/* Ok, the responding node knows which slot this inode is
+	 * orphaned in. We verify that the information is correct and
+	 * then record this in the inode. ocfs2_delete_inode will use
+	 * this information to determine which lock to take. */
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+	mlog_bug_on_msg(OCFS2_I(inode)->ip_orphaned_slot != orphaned_slot &&
+			OCFS2_I(inode)->ip_orphaned_slot
+			!= OCFS2_INVALID_SLOT, "Inode %"MLFu64": Node %d "
+			"says it's orphaned in slot %d, we think it's in %d\n",
+			OCFS2_I(inode)->ip_blkno,
+			be32_to_cpu(resp->r_hdr.h_node_num),
+			orphaned_slot, OCFS2_I(inode)->ip_orphaned_slot);
+
+	OCFS2_I(inode)->ip_orphaned_slot = orphaned_slot;
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+}
+
+int ocfs2_request_delete_vote(struct inode *inode)
+{
+	int orphaned_slot, status;
+	struct ocfs2_net_response_cb delete_cb;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	struct ocfs2_vote_msg *request;
+
+	spin_lock(&OCFS2_I(inode)->ip_lock);
+	orphaned_slot = OCFS2_I(inode)->ip_orphaned_slot;
+	spin_unlock(&OCFS2_I(inode)->ip_lock);
+
+	delete_cb.rc_cb = ocfs2_delete_response_cb;
+	delete_cb.rc_priv = inode;
+
+	mlog(0, "Inode %"MLFu64", we start thinking orphaned slot is %d\n",
+	     OCFS2_I(inode)->ip_blkno, orphaned_slot);
+
+	status = -ENOMEM;
+	request = ocfs2_new_vote_request(osb, OCFS2_I(inode)->ip_blkno,
+					 inode->i_generation,
+					 OCFS2_VOTE_REQ_DELETE, orphaned_slot);
+	if (request) {
+		status = ocfs2_request_vote(inode, request, &delete_cb);
+
+		kfree(request);
+	}
+
+	return status;
+}
+
+static void ocfs2_setup_unlink_vote(struct ocfs2_vote_msg *request,
+				    struct dentry *dentry)
+{
+	struct inode *parent = dentry->d_parent->d_inode;
+
+	/* We need some values which will uniquely identify a dentry
+	 * on the other nodes so that they can find it and run
+	 * d_delete against it. Parent directory block and full name
+	 * should suffice. */
+
+	mlog(0, "unlink/rename request: parent: %"MLFu64" name: %.*s\n",
+	     OCFS2_I(parent)->ip_blkno, dentry->d_name.len,
+	     dentry->d_name.name);
+
+	request->v_unlink_parent = cpu_to_be64(OCFS2_I(parent)->ip_blkno);
+	request->v_unlink_namelen = cpu_to_be32(dentry->d_name.len);
+	memcpy(request->v_unlink_dirent, dentry->d_name.name,
+	       dentry->d_name.len);
+}
+
+int ocfs2_request_unlink_vote(struct inode *inode,
+			      struct dentry *dentry,
+			      unsigned int nlink)
+{
+	int status;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	struct ocfs2_vote_msg *request;
+
+	if (dentry->d_name.len > OCFS2_VOTE_FILENAME_LEN)
+		return -ENAMETOOLONG;
+
+	status = -ENOMEM;
+	request = ocfs2_new_vote_request(osb, OCFS2_I(inode)->ip_blkno,
+					 inode->i_generation,
+					 OCFS2_VOTE_REQ_UNLINK, nlink);
+	if (request) {
+		ocfs2_setup_unlink_vote(request, dentry);
+
+		status = ocfs2_request_vote(inode, request, NULL);
+
+		kfree(request);
+	}
+	return status;
+}
+
+int ocfs2_request_rename_vote(struct inode *inode,
+			      struct dentry *dentry)
+{
+	int status;
+	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
+	struct ocfs2_vote_msg *request;
+
+	if (dentry->d_name.len > OCFS2_VOTE_FILENAME_LEN)
+		return -ENAMETOOLONG;
+
+	status = -ENOMEM;
+	request = ocfs2_new_vote_request(osb, OCFS2_I(inode)->ip_blkno,
+					 inode->i_generation,
+					 OCFS2_VOTE_REQ_RENAME, 0);
+	if (request) {
+		ocfs2_setup_unlink_vote(request, dentry);
+
+		status = ocfs2_request_vote(inode, request, NULL);
+
+		kfree(request);
+	}
+	return status;
+}
+
+int ocfs2_request_mount_vote(struct ocfs2_super *osb)
+{
+	int status;
+	struct ocfs2_vote_msg *request = NULL;
+
+	request = ocfs2_new_vote_request(osb, 0ULL, 0,
+					 OCFS2_VOTE_REQ_MOUNT, 0);
+	if (!request) {
+		status = -ENOMEM;
+		goto bail;
+	}
+
+	status = -EAGAIN;
+	while (status == -EAGAIN) {
+		if (!(osb->s_mount_opt & OCFS2_MOUNT_NOINTR) &&
+		    signal_pending(current)) {
+			status = -ERESTARTSYS;
+			goto bail;
+		}
+
+		if (ocfs2_node_map_is_only(osb, &osb->mounted_map,
+					   osb->node_num)) {
+			status = 0;
+			goto bail;
+		}
+
+		status = ocfs2_do_request_vote(osb, request, NULL);
+	}
+
+bail:
+	if (request)
+		kfree(request);
+
+	return status;
+}
+
+int ocfs2_request_umount_vote(struct ocfs2_super *osb)
+{
+	int status;
+	struct ocfs2_vote_msg *request = NULL;
+
+	request = ocfs2_new_vote_request(osb, 0ULL, 0,
+					 OCFS2_VOTE_REQ_UMOUNT, 0);
+	if (!request) {
+		status = -ENOMEM;
+		goto bail;
+	}
+
+	status = -EAGAIN;
+	while (status == -EAGAIN) {
+		/* Do not check signals on this vote... We really want
+		 * this one to go all the way through. */
+
+		if (ocfs2_node_map_is_only(osb, &osb->mounted_map,
+					   osb->node_num)) {
+			status = 0;
+			goto bail;
+		}
+
+		status = ocfs2_do_request_vote(osb, request, NULL);
+	}
+
+bail:
+	if (request)
+		kfree(request);
+
+	return status;
+}
+
+/* TODO: This should eventually be a hash table! */
+static struct ocfs2_net_wait_ctxt * __ocfs2_find_net_wait_ctxt(struct ocfs2_super *osb,
+							       u32 response_id)
+{
+	struct list_head *p;
+	struct ocfs2_net_wait_ctxt *w = NULL;
+
+	list_for_each(p, &osb->net_response_list) {
+		w = list_entry(p, struct ocfs2_net_wait_ctxt, n_list);
+		if (response_id == w->n_response_id)
+			break;
+		w = NULL;
+	}
+
+	return w;
+}
+
+/* Translate response codes into local node errno values */
+static inline int ocfs2_translate_response(int response)
+{
+	int ret;
+
+	switch (response) {
+	case OCFS2_RESPONSE_OK:
+		ret = 0;
+		break;
+
+	case OCFS2_RESPONSE_BUSY:
+		ret = -EBUSY;
+		break;
+
+	default:
+		ret = -EINVAL;
+	}
+
+	return ret;
+}
+
+static int ocfs2_handle_response_message(struct o2net_msg *msg,
+					 u32 len,
+					 void *data)
+{
+	unsigned int response_id, node_num;
+	int response_status;
+	struct ocfs2_super *osb = data;
+	struct ocfs2_response_msg *resp;
+	struct ocfs2_net_wait_ctxt * w;
+	struct ocfs2_net_response_cb *resp_cb;
+
+	resp = (struct ocfs2_response_msg *) msg->buf;
+
+	response_id = be32_to_cpu(resp->r_hdr.h_response_id);
+	node_num = be32_to_cpu(resp->r_hdr.h_node_num);
+	response_status = 
+		ocfs2_translate_response(be32_to_cpu(resp->r_response));
+
+	mlog(0, "received response message:\n");
+	mlog(0, "h_response_id = %u\n", response_id);
+	mlog(0, "h_request = %u\n", be32_to_cpu(resp->r_hdr.h_request));
+	mlog(0, "h_blkno = %"MLFu64"\n", be64_to_cpu(resp->r_hdr.h_blkno));
+	mlog(0, "h_generation = %u\n", be32_to_cpu(resp->r_hdr.h_generation));
+	mlog(0, "h_node_num = %u\n", node_num);
+	mlog(0, "r_response = %d\n", response_status);
+
+	spin_lock(&osb->net_response_lock);
+	w = __ocfs2_find_net_wait_ctxt(osb, response_id);
+	if (!w) {
+		mlog(0, "request not found!\n");
+		goto bail;
+	}
+	resp_cb = w->n_callback;
+
+	if (response_status && (!w->n_response)) {
+		/* we only really need one negative response so don't
+		 * set it twice. */
+		w->n_response = response_status;
+	}
+
+	if (resp_cb) {
+		spin_unlock(&osb->net_response_lock);
+
+		resp_cb->rc_cb(resp_cb->rc_priv, resp);
+
+		spin_lock(&osb->net_response_lock);
+	}
+
+	__ocfs2_mark_node_responded(osb, w, node_num);
+bail:
+	spin_unlock(&osb->net_response_lock);
+
+	return 0;
+}
+
+static int ocfs2_handle_vote_message(struct o2net_msg *msg,
+				     u32 len,
+				     void *data)
+{
+	int status;
+	struct ocfs2_super *osb = data;
+	struct ocfs2_vote_work *work;
+
+	work = kmalloc(sizeof(struct ocfs2_vote_work), GFP_KERNEL);
+	if (!work) {
+		status = -ENOMEM;
+		mlog_errno(status);
+		goto bail;
+	}
+
+	INIT_LIST_HEAD(&work->w_list);
+	memcpy(&work->w_msg, msg->buf, sizeof(struct ocfs2_vote_msg));
+
+	mlog(0, "scheduling vote request:\n");
+	mlog(0, "h_response_id = %u\n",
+	     be32_to_cpu(work->w_msg.v_hdr.h_response_id));
+	mlog(0, "h_request = %u\n", be32_to_cpu(work->w_msg.v_hdr.h_request));
+	mlog(0, "h_blkno = %"MLFu64"\n",
+	     be64_to_cpu(work->w_msg.v_hdr.h_blkno));
+	mlog(0, "h_generation = %u\n",
+	     be32_to_cpu(work->w_msg.v_hdr.h_generation));
+	mlog(0, "h_node_num = %u\n",
+	     be32_to_cpu(work->w_msg.v_hdr.h_node_num));
+	mlog(0, "v_generic1 = %u\n", be32_to_cpu(work->w_msg.md1.v_generic1));
+
+	spin_lock(&osb->vote_task_lock);
+	list_add_tail(&work->w_list, &osb->vote_list);
+	osb->vote_count++;
+	spin_unlock(&osb->vote_task_lock);
+
+	ocfs2_kick_vote_thread(osb);
+
+	status = 0;
+bail:
+	return status;
+}
+
+void ocfs2_unregister_net_handlers(struct ocfs2_super *osb)
+{
+	if (!osb->net_key)
+		return;
+
+	o2net_unregister_handler_list(&osb->osb_net_handlers);
+
+	if (!list_empty(&osb->net_response_list))
+		mlog(ML_ERROR, "net response list not empty!\n");
+
+	osb->net_key = 0;
+}
+
+int ocfs2_register_net_handlers(struct ocfs2_super *osb)
+{
+	int status = 0;
+
+	status = o2net_register_handler(OCFS2_MESSAGE_TYPE_RESPONSE,
+					osb->net_key,
+					sizeof(struct ocfs2_response_msg),
+					ocfs2_handle_response_message,
+					osb, &osb->osb_net_handlers);
+	if (status) {
+		mlog_errno(status);
+		goto bail;
+	}
+
+	status = o2net_register_handler(OCFS2_MESSAGE_TYPE_VOTE,
+					osb->net_key,
+					sizeof(struct ocfs2_vote_msg),
+					ocfs2_handle_vote_message,
+					osb, &osb->osb_net_handlers);
+	if (status) {
+		mlog_errno(status);
+		goto bail;
+	}
+bail:
+	if (status < 0)
+		ocfs2_unregister_net_handlers(osb);
+
+	return status;
+}
diff --git a/fs/ocfs2/vote.h b/fs/ocfs2/vote.h
new file mode 100644
index 00000000000..9cce6070346
--- /dev/null
+++ b/fs/ocfs2/vote.h
@@ -0,0 +1,56 @@
+/* -*- mode: c; c-basic-offset: 8; -*-
+ * vim: noexpandtab sw=8 ts=8 sts=0:
+ *
+ * vote.h
+ *
+ * description here
+ *
+ * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
+ *
+ * This program 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 2 of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that 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, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+
+
+#ifndef VOTE_H
+#define VOTE_H
+
+int ocfs2_vote_thread(void *arg);
+static inline void ocfs2_kick_vote_thread(struct ocfs2_super *osb)
+{
+	spin_lock(&osb->vote_task_lock);
+	/* make sure the voting thread gets a swipe at whatever changes
+	 * the caller may have made to the voting state */
+	osb->vote_wake_sequence++;
+	spin_unlock(&osb->vote_task_lock);
+	wake_up(&osb->vote_event);
+}
+
+int ocfs2_request_delete_vote(struct inode *inode);
+int ocfs2_request_unlink_vote(struct inode *inode,
+			      struct dentry *dentry,
+			      unsigned int nlink);
+int ocfs2_request_rename_vote(struct inode *inode,
+			      struct dentry *dentry);
+int ocfs2_request_mount_vote(struct ocfs2_super *osb);
+int ocfs2_request_umount_vote(struct ocfs2_super *osb);
+int ocfs2_register_net_handlers(struct ocfs2_super *osb);
+void ocfs2_unregister_net_handlers(struct ocfs2_super *osb);
+
+void ocfs2_mark_inode_remotely_deleted(struct inode *inode);
+
+void ocfs2_remove_node_from_vote_queues(struct ocfs2_super *osb,
+					int node_num);
+#endif