/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * xattr.c
 *
 * Copyright (C) 2004, 2008 Oracle.  All rights reserved.
 *
 * CREDITS:
 * Lots of code in this file is copy from linux/fs/ext3/xattr.c.
 * Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
 *
 * 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.
 */

#include <linux/capability.h>
#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/sched.h>
#include <linux/splice.h>
#include <linux/mount.h>
#include <linux/writeback.h>
#include <linux/falloc.h>
#include <linux/sort.h>
#include <linux/init.h>
#include <linux/module.h>
#include <linux/string.h>
#include <linux/security.h>

#include <cluster/masklog.h>

#include "ocfs2.h"
#include "alloc.h"
#include "blockcheck.h"
#include "dlmglue.h"
#include "file.h"
#include "symlink.h"
#include "sysfile.h"
#include "inode.h"
#include "journal.h"
#include "ocfs2_fs.h"
#include "suballoc.h"
#include "uptodate.h"
#include "buffer_head_io.h"
#include "super.h"
#include "xattr.h"
#include "refcounttree.h"
#include "acl.h"
#include "ocfs2_trace.h"

struct ocfs2_xattr_def_value_root {
	struct ocfs2_xattr_value_root	xv;
	struct ocfs2_extent_rec		er;
};

struct ocfs2_xattr_bucket {
	/* The inode these xattrs are associated with */
	struct inode *bu_inode;

	/* The actual buffers that make up the bucket */
	struct buffer_head *bu_bhs[OCFS2_XATTR_MAX_BLOCKS_PER_BUCKET];

	/* How many blocks make up one bucket for this filesystem */
	int bu_blocks;
};

struct ocfs2_xattr_set_ctxt {
	handle_t *handle;
	struct ocfs2_alloc_context *meta_ac;
	struct ocfs2_alloc_context *data_ac;
	struct ocfs2_cached_dealloc_ctxt dealloc;
	int set_abort;
};

#define OCFS2_XATTR_ROOT_SIZE	(sizeof(struct ocfs2_xattr_def_value_root))
#define OCFS2_XATTR_INLINE_SIZE	80
#define OCFS2_XATTR_HEADER_GAP	4
#define OCFS2_XATTR_FREE_IN_IBODY	(OCFS2_MIN_XATTR_INLINE_SIZE \
					 - sizeof(struct ocfs2_xattr_header) \
					 - OCFS2_XATTR_HEADER_GAP)
#define OCFS2_XATTR_FREE_IN_BLOCK(ptr)	((ptr)->i_sb->s_blocksize \
					 - sizeof(struct ocfs2_xattr_block) \
					 - sizeof(struct ocfs2_xattr_header) \
					 - OCFS2_XATTR_HEADER_GAP)

static struct ocfs2_xattr_def_value_root def_xv = {
	.xv.xr_list.l_count = cpu_to_le16(1),
};

const struct xattr_handler *ocfs2_xattr_handlers[] = {
	&ocfs2_xattr_user_handler,
	&ocfs2_xattr_acl_access_handler,
	&ocfs2_xattr_acl_default_handler,
	&ocfs2_xattr_trusted_handler,
	&ocfs2_xattr_security_handler,
	NULL
};

static const struct xattr_handler *ocfs2_xattr_handler_map[OCFS2_XATTR_MAX] = {
	[OCFS2_XATTR_INDEX_USER]	= &ocfs2_xattr_user_handler,
	[OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS]
					= &ocfs2_xattr_acl_access_handler,
	[OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT]
					= &ocfs2_xattr_acl_default_handler,
	[OCFS2_XATTR_INDEX_TRUSTED]	= &ocfs2_xattr_trusted_handler,
	[OCFS2_XATTR_INDEX_SECURITY]	= &ocfs2_xattr_security_handler,
};

struct ocfs2_xattr_info {
	int		xi_name_index;
	const char	*xi_name;
	int		xi_name_len;
	const void	*xi_value;
	size_t		xi_value_len;
};

struct ocfs2_xattr_search {
	struct buffer_head *inode_bh;
	/*
	 * xattr_bh point to the block buffer head which has extended attribute
	 * when extended attribute in inode, xattr_bh is equal to inode_bh.
	 */
	struct buffer_head *xattr_bh;
	struct ocfs2_xattr_header *header;
	struct ocfs2_xattr_bucket *bucket;
	void *base;
	void *end;
	struct ocfs2_xattr_entry *here;
	int not_found;
};

/* Operations on struct ocfs2_xa_entry */
struct ocfs2_xa_loc;
struct ocfs2_xa_loc_operations {
	/*
	 * Journal functions
	 */
	int (*xlo_journal_access)(handle_t *handle, struct ocfs2_xa_loc *loc,
				  int type);
	void (*xlo_journal_dirty)(handle_t *handle, struct ocfs2_xa_loc *loc);

	/*
	 * Return a pointer to the appropriate buffer in loc->xl_storage
	 * at the given offset from loc->xl_header.
	 */
	void *(*xlo_offset_pointer)(struct ocfs2_xa_loc *loc, int offset);

	/* Can we reuse the existing entry for the new value? */
	int (*xlo_can_reuse)(struct ocfs2_xa_loc *loc,
			     struct ocfs2_xattr_info *xi);

	/* How much space is needed for the new value? */
	int (*xlo_check_space)(struct ocfs2_xa_loc *loc,
			       struct ocfs2_xattr_info *xi);

	/*
	 * Return the offset of the first name+value pair.  This is
	 * the start of our downward-filling free space.
	 */
	int (*xlo_get_free_start)(struct ocfs2_xa_loc *loc);

	/*
	 * Remove the name+value at this location.  Do whatever is
	 * appropriate with the remaining name+value pairs.
	 */
	void (*xlo_wipe_namevalue)(struct ocfs2_xa_loc *loc);

	/* Fill xl_entry with a new entry */
	void (*xlo_add_entry)(struct ocfs2_xa_loc *loc, u32 name_hash);

	/* Add name+value storage to an entry */
	void (*xlo_add_namevalue)(struct ocfs2_xa_loc *loc, int size);

	/*
	 * Initialize the value buf's access and bh fields for this entry.
	 * ocfs2_xa_fill_value_buf() will handle the xv pointer.
	 */
	void (*xlo_fill_value_buf)(struct ocfs2_xa_loc *loc,
				   struct ocfs2_xattr_value_buf *vb);
};

/*
 * Describes an xattr entry location.  This is a memory structure
 * tracking the on-disk structure.
 */
struct ocfs2_xa_loc {
	/* This xattr belongs to this inode */
	struct inode *xl_inode;

	/* The ocfs2_xattr_header inside the on-disk storage. Not NULL. */
	struct ocfs2_xattr_header *xl_header;

	/* Bytes from xl_header to the end of the storage */
	int xl_size;

	/*
	 * The ocfs2_xattr_entry this location describes.  If this is
	 * NULL, this location describes the on-disk structure where it
	 * would have been.
	 */
	struct ocfs2_xattr_entry *xl_entry;

	/*
	 * Internal housekeeping
	 */

	/* Buffer(s) containing this entry */
	void *xl_storage;

	/* Operations on the storage backing this location */
	const struct ocfs2_xa_loc_operations *xl_ops;
};

/*
 * Convenience functions to calculate how much space is needed for a
 * given name+value pair
 */
static int namevalue_size(int name_len, uint64_t value_len)
{
	if (value_len > OCFS2_XATTR_INLINE_SIZE)
		return OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE;
	else
		return OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_SIZE(value_len);
}

static int namevalue_size_xi(struct ocfs2_xattr_info *xi)
{
	return namevalue_size(xi->xi_name_len, xi->xi_value_len);
}

static int namevalue_size_xe(struct ocfs2_xattr_entry *xe)
{
	u64 value_len = le64_to_cpu(xe->xe_value_size);

	BUG_ON((value_len > OCFS2_XATTR_INLINE_SIZE) &&
	       ocfs2_xattr_is_local(xe));
	return namevalue_size(xe->xe_name_len, value_len);
}


static int ocfs2_xattr_bucket_get_name_value(struct super_block *sb,
					     struct ocfs2_xattr_header *xh,
					     int index,
					     int *block_off,
					     int *new_offset);

static int ocfs2_xattr_block_find(struct inode *inode,
				  int name_index,
				  const char *name,
				  struct ocfs2_xattr_search *xs);
static int ocfs2_xattr_index_block_find(struct inode *inode,
					struct buffer_head *root_bh,
					int name_index,
					const char *name,
					struct ocfs2_xattr_search *xs);

static int ocfs2_xattr_tree_list_index_block(struct inode *inode,
					struct buffer_head *blk_bh,
					char *buffer,
					size_t buffer_size);

static int ocfs2_xattr_create_index_block(struct inode *inode,
					  struct ocfs2_xattr_search *xs,
					  struct ocfs2_xattr_set_ctxt *ctxt);

static int ocfs2_xattr_set_entry_index_block(struct inode *inode,
					     struct ocfs2_xattr_info *xi,
					     struct ocfs2_xattr_search *xs,
					     struct ocfs2_xattr_set_ctxt *ctxt);

typedef int (xattr_tree_rec_func)(struct inode *inode,
				  struct buffer_head *root_bh,
				  u64 blkno, u32 cpos, u32 len, void *para);
static int ocfs2_iterate_xattr_index_block(struct inode *inode,
					   struct buffer_head *root_bh,
					   xattr_tree_rec_func *rec_func,
					   void *para);
static int ocfs2_delete_xattr_in_bucket(struct inode *inode,
					struct ocfs2_xattr_bucket *bucket,
					void *para);
static int ocfs2_rm_xattr_cluster(struct inode *inode,
				  struct buffer_head *root_bh,
				  u64 blkno,
				  u32 cpos,
				  u32 len,
				  void *para);

static int ocfs2_mv_xattr_buckets(struct inode *inode, handle_t *handle,
				  u64 src_blk, u64 last_blk, u64 to_blk,
				  unsigned int start_bucket,
				  u32 *first_hash);
static int ocfs2_prepare_refcount_xattr(struct inode *inode,
					struct ocfs2_dinode *di,
					struct ocfs2_xattr_info *xi,
					struct ocfs2_xattr_search *xis,
					struct ocfs2_xattr_search *xbs,
					struct ocfs2_refcount_tree **ref_tree,
					int *meta_need,
					int *credits);
static int ocfs2_get_xattr_tree_value_root(struct super_block *sb,
					   struct ocfs2_xattr_bucket *bucket,
					   int offset,
					   struct ocfs2_xattr_value_root **xv,
					   struct buffer_head **bh);

static inline u16 ocfs2_xattr_buckets_per_cluster(struct ocfs2_super *osb)
{
	return (1 << osb->s_clustersize_bits) / OCFS2_XATTR_BUCKET_SIZE;
}

static inline u16 ocfs2_blocks_per_xattr_bucket(struct super_block *sb)
{
	return OCFS2_XATTR_BUCKET_SIZE / (1 << sb->s_blocksize_bits);
}

#define bucket_blkno(_b) ((_b)->bu_bhs[0]->b_blocknr)
#define bucket_block(_b, _n) ((_b)->bu_bhs[(_n)]->b_data)
#define bucket_xh(_b) ((struct ocfs2_xattr_header *)bucket_block((_b), 0))

static struct ocfs2_xattr_bucket *ocfs2_xattr_bucket_new(struct inode *inode)
{
	struct ocfs2_xattr_bucket *bucket;
	int blks = ocfs2_blocks_per_xattr_bucket(inode->i_sb);

	BUG_ON(blks > OCFS2_XATTR_MAX_BLOCKS_PER_BUCKET);

	bucket = kzalloc(sizeof(struct ocfs2_xattr_bucket), GFP_NOFS);
	if (bucket) {
		bucket->bu_inode = inode;
		bucket->bu_blocks = blks;
	}

	return bucket;
}

static void ocfs2_xattr_bucket_relse(struct ocfs2_xattr_bucket *bucket)
{
	int i;

	for (i = 0; i < bucket->bu_blocks; i++) {
		brelse(bucket->bu_bhs[i]);
		bucket->bu_bhs[i] = NULL;
	}
}

static void ocfs2_xattr_bucket_free(struct ocfs2_xattr_bucket *bucket)
{
	if (bucket) {
		ocfs2_xattr_bucket_relse(bucket);
		bucket->bu_inode = NULL;
		kfree(bucket);
	}
}

/*
 * A bucket that has never been written to disk doesn't need to be
 * read.  We just need the buffer_heads.  Don't call this for
 * buckets that are already on disk.  ocfs2_read_xattr_bucket() initializes
 * them fully.
 */
static int ocfs2_init_xattr_bucket(struct ocfs2_xattr_bucket *bucket,
				   u64 xb_blkno)
{
	int i, rc = 0;

	for (i = 0; i < bucket->bu_blocks; i++) {
		bucket->bu_bhs[i] = sb_getblk(bucket->bu_inode->i_sb,
					      xb_blkno + i);
		if (!bucket->bu_bhs[i]) {
			rc = -EIO;
			mlog_errno(rc);
			break;
		}

		if (!ocfs2_buffer_uptodate(INODE_CACHE(bucket->bu_inode),
					   bucket->bu_bhs[i]))
			ocfs2_set_new_buffer_uptodate(INODE_CACHE(bucket->bu_inode),
						      bucket->bu_bhs[i]);
	}

	if (rc)
		ocfs2_xattr_bucket_relse(bucket);
	return rc;
}

/* Read the xattr bucket at xb_blkno */
static int ocfs2_read_xattr_bucket(struct ocfs2_xattr_bucket *bucket,
				   u64 xb_blkno)
{
	int rc;

	rc = ocfs2_read_blocks(INODE_CACHE(bucket->bu_inode), xb_blkno,
			       bucket->bu_blocks, bucket->bu_bhs, 0,
			       NULL);
	if (!rc) {
		spin_lock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock);
		rc = ocfs2_validate_meta_ecc_bhs(bucket->bu_inode->i_sb,
						 bucket->bu_bhs,
						 bucket->bu_blocks,
						 &bucket_xh(bucket)->xh_check);
		spin_unlock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock);
		if (rc)
			mlog_errno(rc);
	}

	if (rc)
		ocfs2_xattr_bucket_relse(bucket);
	return rc;
}

static int ocfs2_xattr_bucket_journal_access(handle_t *handle,
					     struct ocfs2_xattr_bucket *bucket,
					     int type)
{
	int i, rc = 0;

	for (i = 0; i < bucket->bu_blocks; i++) {
		rc = ocfs2_journal_access(handle,
					  INODE_CACHE(bucket->bu_inode),
					  bucket->bu_bhs[i], type);
		if (rc) {
			mlog_errno(rc);
			break;
		}
	}

	return rc;
}

static void ocfs2_xattr_bucket_journal_dirty(handle_t *handle,
					     struct ocfs2_xattr_bucket *bucket)
{
	int i;

	spin_lock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock);
	ocfs2_compute_meta_ecc_bhs(bucket->bu_inode->i_sb,
				   bucket->bu_bhs, bucket->bu_blocks,
				   &bucket_xh(bucket)->xh_check);
	spin_unlock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock);

	for (i = 0; i < bucket->bu_blocks; i++)
		ocfs2_journal_dirty(handle, bucket->bu_bhs[i]);
}

static void ocfs2_xattr_bucket_copy_data(struct ocfs2_xattr_bucket *dest,
					 struct ocfs2_xattr_bucket *src)
{
	int i;
	int blocksize = src->bu_inode->i_sb->s_blocksize;

	BUG_ON(dest->bu_blocks != src->bu_blocks);
	BUG_ON(dest->bu_inode != src->bu_inode);

	for (i = 0; i < src->bu_blocks; i++) {
		memcpy(bucket_block(dest, i), bucket_block(src, i),
		       blocksize);
	}
}

static int ocfs2_validate_xattr_block(struct super_block *sb,
				      struct buffer_head *bh)
{
	int rc;
	struct ocfs2_xattr_block *xb =
		(struct ocfs2_xattr_block *)bh->b_data;

	trace_ocfs2_validate_xattr_block((unsigned long long)bh->b_blocknr);

	BUG_ON(!buffer_uptodate(bh));

	/*
	 * If the ecc fails, we return the error but otherwise
	 * leave the filesystem running.  We know any error is
	 * local to this block.
	 */
	rc = ocfs2_validate_meta_ecc(sb, bh->b_data, &xb->xb_check);
	if (rc)
		return rc;

	/*
	 * Errors after here are fatal
	 */

	if (!OCFS2_IS_VALID_XATTR_BLOCK(xb)) {
		ocfs2_error(sb,
			    "Extended attribute block #%llu has bad "
			    "signature %.*s",
			    (unsigned long long)bh->b_blocknr, 7,
			    xb->xb_signature);
		return -EINVAL;
	}

	if (le64_to_cpu(xb->xb_blkno) != bh->b_blocknr) {
		ocfs2_error(sb,
			    "Extended attribute block #%llu has an "
			    "invalid xb_blkno of %llu",
			    (unsigned long long)bh->b_blocknr,
			    (unsigned long long)le64_to_cpu(xb->xb_blkno));
		return -EINVAL;
	}

	if (le32_to_cpu(xb->xb_fs_generation) != OCFS2_SB(sb)->fs_generation) {
		ocfs2_error(sb,
			    "Extended attribute block #%llu has an invalid "
			    "xb_fs_generation of #%u",
			    (unsigned long long)bh->b_blocknr,
			    le32_to_cpu(xb->xb_fs_generation));
		return -EINVAL;
	}

	return 0;
}

static int ocfs2_read_xattr_block(struct inode *inode, u64 xb_blkno,
				  struct buffer_head **bh)
{
	int rc;
	struct buffer_head *tmp = *bh;

	rc = ocfs2_read_block(INODE_CACHE(inode), xb_blkno, &tmp,
			      ocfs2_validate_xattr_block);

	/* If ocfs2_read_block() got us a new bh, pass it up. */
	if (!rc && !*bh)
		*bh = tmp;

	return rc;
}

static inline const char *ocfs2_xattr_prefix(int name_index)
{
	const struct xattr_handler *handler = NULL;

	if (name_index > 0 && name_index < OCFS2_XATTR_MAX)
		handler = ocfs2_xattr_handler_map[name_index];

	return handler ? handler->prefix : NULL;
}

static u32 ocfs2_xattr_name_hash(struct inode *inode,
				 const char *name,
				 int name_len)
{
	/* Get hash value of uuid from super block */
	u32 hash = OCFS2_SB(inode->i_sb)->uuid_hash;
	int i;

	/* hash extended attribute name */
	for (i = 0; i < name_len; i++) {
		hash = (hash << OCFS2_HASH_SHIFT) ^
		       (hash >> (8*sizeof(hash) - OCFS2_HASH_SHIFT)) ^
		       *name++;
	}

	return hash;
}

static int ocfs2_xattr_entry_real_size(int name_len, size_t value_len)
{
	return namevalue_size(name_len, value_len) +
		sizeof(struct ocfs2_xattr_entry);
}

static int ocfs2_xi_entry_usage(struct ocfs2_xattr_info *xi)
{
	return namevalue_size_xi(xi) +
		sizeof(struct ocfs2_xattr_entry);
}

static int ocfs2_xe_entry_usage(struct ocfs2_xattr_entry *xe)
{
	return namevalue_size_xe(xe) +
		sizeof(struct ocfs2_xattr_entry);
}

int ocfs2_calc_security_init(struct inode *dir,
			     struct ocfs2_security_xattr_info *si,
			     int *want_clusters,
			     int *xattr_credits,
			     struct ocfs2_alloc_context **xattr_ac)
{
	int ret = 0;
	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
	int s_size = ocfs2_xattr_entry_real_size(strlen(si->name),
						 si->value_len);

	/*
	 * The max space of security xattr taken inline is
	 * 256(name) + 80(value) + 16(entry) = 352 bytes,
	 * So reserve one metadata block for it is ok.
	 */
	if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE ||
	    s_size > OCFS2_XATTR_FREE_IN_IBODY) {
		ret = ocfs2_reserve_new_metadata_blocks(osb, 1, xattr_ac);
		if (ret) {
			mlog_errno(ret);
			return ret;
		}
		*xattr_credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS;
	}

	/* reserve clusters for xattr value which will be set in B tree*/
	if (si->value_len > OCFS2_XATTR_INLINE_SIZE) {
		int new_clusters = ocfs2_clusters_for_bytes(dir->i_sb,
							    si->value_len);

		*xattr_credits += ocfs2_clusters_to_blocks(dir->i_sb,
							   new_clusters);
		*want_clusters += new_clusters;
	}
	return ret;
}

int ocfs2_calc_xattr_init(struct inode *dir,
			  struct buffer_head *dir_bh,
			  umode_t mode,
			  struct ocfs2_security_xattr_info *si,
			  int *want_clusters,
			  int *xattr_credits,
			  int *want_meta)
{
	int ret = 0;
	struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
	int s_size = 0, a_size = 0, acl_len = 0, new_clusters;

	if (si->enable)
		s_size = ocfs2_xattr_entry_real_size(strlen(si->name),
						     si->value_len);

	if (osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) {
		acl_len = ocfs2_xattr_get_nolock(dir, dir_bh,
					OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT,
					"", NULL, 0);
		if (acl_len > 0) {
			a_size = ocfs2_xattr_entry_real_size(0, acl_len);
			if (S_ISDIR(mode))
				a_size <<= 1;
		} else if (acl_len != 0 && acl_len != -ENODATA) {
			mlog_errno(ret);
			return ret;
		}
	}

	if (!(s_size + a_size))
		return ret;

	/*
	 * The max space of security xattr taken inline is
	 * 256(name) + 80(value) + 16(entry) = 352 bytes,
	 * The max space of acl xattr taken inline is
	 * 80(value) + 16(entry) * 2(if directory) = 192 bytes,
	 * when blocksize = 512, may reserve one more cluser for
	 * xattr bucket, otherwise reserve one metadata block
	 * for them is ok.
	 * If this is a new directory with inline data,
	 * we choose to reserve the entire inline area for
	 * directory contents and force an external xattr block.
	 */
	if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE ||
	    (S_ISDIR(mode) && ocfs2_supports_inline_data(osb)) ||
	    (s_size + a_size) > OCFS2_XATTR_FREE_IN_IBODY) {
		*want_meta = *want_meta + 1;
		*xattr_credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS;
	}

	if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE &&
	    (s_size + a_size) > OCFS2_XATTR_FREE_IN_BLOCK(dir)) {
		*want_clusters += 1;
		*xattr_credits += ocfs2_blocks_per_xattr_bucket(dir->i_sb);
	}

	/*
	 * reserve credits and clusters for xattrs which has large value
	 * and have to be set outside
	 */
	if (si->enable && si->value_len > OCFS2_XATTR_INLINE_SIZE) {
		new_clusters = ocfs2_clusters_for_bytes(dir->i_sb,
							si->value_len);
		*xattr_credits += ocfs2_clusters_to_blocks(dir->i_sb,
							   new_clusters);
		*want_clusters += new_clusters;
	}
	if (osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL &&
	    acl_len > OCFS2_XATTR_INLINE_SIZE) {
		/* for directory, it has DEFAULT and ACCESS two types of acls */
		new_clusters = (S_ISDIR(mode) ? 2 : 1) *
				ocfs2_clusters_for_bytes(dir->i_sb, acl_len);
		*xattr_credits += ocfs2_clusters_to_blocks(dir->i_sb,
							   new_clusters);
		*want_clusters += new_clusters;
	}

	return ret;
}

static int ocfs2_xattr_extend_allocation(struct inode *inode,
					 u32 clusters_to_add,
					 struct ocfs2_xattr_value_buf *vb,
					 struct ocfs2_xattr_set_ctxt *ctxt)
{
	int status = 0, credits;
	handle_t *handle = ctxt->handle;
	enum ocfs2_alloc_restarted why;
	u32 prev_clusters, logical_start = le32_to_cpu(vb->vb_xv->xr_clusters);
	struct ocfs2_extent_tree et;

	ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb);

	while (clusters_to_add) {
		trace_ocfs2_xattr_extend_allocation(clusters_to_add);

		status = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh,
				       OCFS2_JOURNAL_ACCESS_WRITE);
		if (status < 0) {
			mlog_errno(status);
			break;
		}

		prev_clusters = le32_to_cpu(vb->vb_xv->xr_clusters);
		status = ocfs2_add_clusters_in_btree(handle,
						     &et,
						     &logical_start,
						     clusters_to_add,
						     0,
						     ctxt->data_ac,
						     ctxt->meta_ac,
						     &why);
		if ((status < 0) && (status != -EAGAIN)) {
			if (status != -ENOSPC)
				mlog_errno(status);
			break;
		}

		ocfs2_journal_dirty(handle, vb->vb_bh);

		clusters_to_add -= le32_to_cpu(vb->vb_xv->xr_clusters) -
					 prev_clusters;

		if (why != RESTART_NONE && clusters_to_add) {
			/*
			 * We can only fail in case the alloc file doesn't give
			 * up enough clusters.
			 */
			BUG_ON(why == RESTART_META);

			credits = ocfs2_calc_extend_credits(inode->i_sb,
							    &vb->vb_xv->xr_list,
							    clusters_to_add);
			status = ocfs2_extend_trans(handle, credits);
			if (status < 0) {
				status = -ENOMEM;
				mlog_errno(status);
				break;
			}
		}
	}

	return status;
}

static int __ocfs2_remove_xattr_range(struct inode *inode,
				      struct ocfs2_xattr_value_buf *vb,
				      u32 cpos, u32 phys_cpos, u32 len,
				      unsigned int ext_flags,
				      struct ocfs2_xattr_set_ctxt *ctxt)
{
	int ret;
	u64 phys_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos);
	handle_t *handle = ctxt->handle;
	struct ocfs2_extent_tree et;

	ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb);

	ret = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh,
			    OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	ret = ocfs2_remove_extent(handle, &et, cpos, len, ctxt->meta_ac,
				  &ctxt->dealloc);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	le32_add_cpu(&vb->vb_xv->xr_clusters, -len);
	ocfs2_journal_dirty(handle, vb->vb_bh);

	if (ext_flags & OCFS2_EXT_REFCOUNTED)
		ret = ocfs2_decrease_refcount(inode, handle,
					ocfs2_blocks_to_clusters(inode->i_sb,
								 phys_blkno),
					len, ctxt->meta_ac, &ctxt->dealloc, 1);
	else
		ret = ocfs2_cache_cluster_dealloc(&ctxt->dealloc,
						  phys_blkno, len);
	if (ret)
		mlog_errno(ret);

out:
	return ret;
}

static int ocfs2_xattr_shrink_size(struct inode *inode,
				   u32 old_clusters,
				   u32 new_clusters,
				   struct ocfs2_xattr_value_buf *vb,
				   struct ocfs2_xattr_set_ctxt *ctxt)
{
	int ret = 0;
	unsigned int ext_flags;
	u32 trunc_len, cpos, phys_cpos, alloc_size;
	u64 block;

	if (old_clusters <= new_clusters)
		return 0;

	cpos = new_clusters;
	trunc_len = old_clusters - new_clusters;
	while (trunc_len) {
		ret = ocfs2_xattr_get_clusters(inode, cpos, &phys_cpos,
					       &alloc_size,
					       &vb->vb_xv->xr_list, &ext_flags);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}

		if (alloc_size > trunc_len)
			alloc_size = trunc_len;

		ret = __ocfs2_remove_xattr_range(inode, vb, cpos,
						 phys_cpos, alloc_size,
						 ext_flags, ctxt);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}

		block = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos);
		ocfs2_remove_xattr_clusters_from_cache(INODE_CACHE(inode),
						       block, alloc_size);
		cpos += alloc_size;
		trunc_len -= alloc_size;
	}

out:
	return ret;
}

static int ocfs2_xattr_value_truncate(struct inode *inode,
				      struct ocfs2_xattr_value_buf *vb,
				      int len,
				      struct ocfs2_xattr_set_ctxt *ctxt)
{
	int ret;
	u32 new_clusters = ocfs2_clusters_for_bytes(inode->i_sb, len);
	u32 old_clusters = le32_to_cpu(vb->vb_xv->xr_clusters);

	if (new_clusters == old_clusters)
		return 0;

	if (new_clusters > old_clusters)
		ret = ocfs2_xattr_extend_allocation(inode,
						    new_clusters - old_clusters,
						    vb, ctxt);
	else
		ret = ocfs2_xattr_shrink_size(inode,
					      old_clusters, new_clusters,
					      vb, ctxt);

	return ret;
}

static int ocfs2_xattr_list_entry(char *buffer, size_t size,
				  size_t *result, const char *prefix,
				  const char *name, int name_len)
{
	char *p = buffer + *result;
	int prefix_len = strlen(prefix);
	int total_len = prefix_len + name_len + 1;

	*result += total_len;

	/* we are just looking for how big our buffer needs to be */
	if (!size)
		return 0;

	if (*result > size)
		return -ERANGE;

	memcpy(p, prefix, prefix_len);
	memcpy(p + prefix_len, name, name_len);
	p[prefix_len + name_len] = '\0';

	return 0;
}

static int ocfs2_xattr_list_entries(struct inode *inode,
				    struct ocfs2_xattr_header *header,
				    char *buffer, size_t buffer_size)
{
	size_t result = 0;
	int i, type, ret;
	const char *prefix, *name;

	for (i = 0 ; i < le16_to_cpu(header->xh_count); i++) {
		struct ocfs2_xattr_entry *entry = &header->xh_entries[i];
		type = ocfs2_xattr_get_type(entry);
		prefix = ocfs2_xattr_prefix(type);

		if (prefix) {
			name = (const char *)header +
				le16_to_cpu(entry->xe_name_offset);

			ret = ocfs2_xattr_list_entry(buffer, buffer_size,
						     &result, prefix, name,
						     entry->xe_name_len);
			if (ret)
				return ret;
		}
	}

	return result;
}

int ocfs2_has_inline_xattr_value_outside(struct inode *inode,
					 struct ocfs2_dinode *di)
{
	struct ocfs2_xattr_header *xh;
	int i;

	xh = (struct ocfs2_xattr_header *)
		 ((void *)di + inode->i_sb->s_blocksize -
		 le16_to_cpu(di->i_xattr_inline_size));

	for (i = 0; i < le16_to_cpu(xh->xh_count); i++)
		if (!ocfs2_xattr_is_local(&xh->xh_entries[i]))
			return 1;

	return 0;
}

static int ocfs2_xattr_ibody_list(struct inode *inode,
				  struct ocfs2_dinode *di,
				  char *buffer,
				  size_t buffer_size)
{
	struct ocfs2_xattr_header *header = NULL;
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
	int ret = 0;

	if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL))
		return ret;

	header = (struct ocfs2_xattr_header *)
		 ((void *)di + inode->i_sb->s_blocksize -
		 le16_to_cpu(di->i_xattr_inline_size));

	ret = ocfs2_xattr_list_entries(inode, header, buffer, buffer_size);

	return ret;
}

static int ocfs2_xattr_block_list(struct inode *inode,
				  struct ocfs2_dinode *di,
				  char *buffer,
				  size_t buffer_size)
{
	struct buffer_head *blk_bh = NULL;
	struct ocfs2_xattr_block *xb;
	int ret = 0;

	if (!di->i_xattr_loc)
		return ret;

	ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc),
				     &blk_bh);
	if (ret < 0) {
		mlog_errno(ret);
		return ret;
	}

	xb = (struct ocfs2_xattr_block *)blk_bh->b_data;
	if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) {
		struct ocfs2_xattr_header *header = &xb->xb_attrs.xb_header;
		ret = ocfs2_xattr_list_entries(inode, header,
					       buffer, buffer_size);
	} else
		ret = ocfs2_xattr_tree_list_index_block(inode, blk_bh,
						   buffer, buffer_size);

	brelse(blk_bh);

	return ret;
}

ssize_t ocfs2_listxattr(struct dentry *dentry,
			char *buffer,
			size_t size)
{
	int ret = 0, i_ret = 0, b_ret = 0;
	struct buffer_head *di_bh = NULL;
	struct ocfs2_dinode *di = NULL;
	struct ocfs2_inode_info *oi = OCFS2_I(dentry->d_inode);

	if (!ocfs2_supports_xattr(OCFS2_SB(dentry->d_sb)))
		return -EOPNOTSUPP;

	if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL))
		return ret;

	ret = ocfs2_inode_lock(dentry->d_inode, &di_bh, 0);
	if (ret < 0) {
		mlog_errno(ret);
		return ret;
	}

	di = (struct ocfs2_dinode *)di_bh->b_data;

	down_read(&oi->ip_xattr_sem);
	i_ret = ocfs2_xattr_ibody_list(dentry->d_inode, di, buffer, size);
	if (i_ret < 0)
		b_ret = 0;
	else {
		if (buffer) {
			buffer += i_ret;
			size -= i_ret;
		}
		b_ret = ocfs2_xattr_block_list(dentry->d_inode, di,
					       buffer, size);
		if (b_ret < 0)
			i_ret = 0;
	}
	up_read(&oi->ip_xattr_sem);
	ocfs2_inode_unlock(dentry->d_inode, 0);

	brelse(di_bh);

	return i_ret + b_ret;
}

static int ocfs2_xattr_find_entry(int name_index,
				  const char *name,
				  struct ocfs2_xattr_search *xs)
{
	struct ocfs2_xattr_entry *entry;
	size_t name_len;
	int i, cmp = 1;

	if (name == NULL)
		return -EINVAL;

	name_len = strlen(name);
	entry = xs->here;
	for (i = 0; i < le16_to_cpu(xs->header->xh_count); i++) {
		cmp = name_index - ocfs2_xattr_get_type(entry);
		if (!cmp)
			cmp = name_len - entry->xe_name_len;
		if (!cmp)
			cmp = memcmp(name, (xs->base +
				     le16_to_cpu(entry->xe_name_offset)),
				     name_len);
		if (cmp == 0)
			break;
		entry += 1;
	}
	xs->here = entry;

	return cmp ? -ENODATA : 0;
}

static int ocfs2_xattr_get_value_outside(struct inode *inode,
					 struct ocfs2_xattr_value_root *xv,
					 void *buffer,
					 size_t len)
{
	u32 cpos, p_cluster, num_clusters, bpc, clusters;
	u64 blkno;
	int i, ret = 0;
	size_t cplen, blocksize;
	struct buffer_head *bh = NULL;
	struct ocfs2_extent_list *el;

	el = &xv->xr_list;
	clusters = le32_to_cpu(xv->xr_clusters);
	bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
	blocksize = inode->i_sb->s_blocksize;

	cpos = 0;
	while (cpos < clusters) {
		ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster,
					       &num_clusters, el, NULL);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}

		blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster);
		/* Copy ocfs2_xattr_value */
		for (i = 0; i < num_clusters * bpc; i++, blkno++) {
			ret = ocfs2_read_block(INODE_CACHE(inode), blkno,
					       &bh, NULL);
			if (ret) {
				mlog_errno(ret);
				goto out;
			}

			cplen = len >= blocksize ? blocksize : len;
			memcpy(buffer, bh->b_data, cplen);
			len -= cplen;
			buffer += cplen;

			brelse(bh);
			bh = NULL;
			if (len == 0)
				break;
		}
		cpos += num_clusters;
	}
out:
	return ret;
}

static int ocfs2_xattr_ibody_get(struct inode *inode,
				 int name_index,
				 const char *name,
				 void *buffer,
				 size_t buffer_size,
				 struct ocfs2_xattr_search *xs)
{
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data;
	struct ocfs2_xattr_value_root *xv;
	size_t size;
	int ret = 0;

	if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL))
		return -ENODATA;

	xs->end = (void *)di + inode->i_sb->s_blocksize;
	xs->header = (struct ocfs2_xattr_header *)
			(xs->end - le16_to_cpu(di->i_xattr_inline_size));
	xs->base = (void *)xs->header;
	xs->here = xs->header->xh_entries;

	ret = ocfs2_xattr_find_entry(name_index, name, xs);
	if (ret)
		return ret;
	size = le64_to_cpu(xs->here->xe_value_size);
	if (buffer) {
		if (size > buffer_size)
			return -ERANGE;
		if (ocfs2_xattr_is_local(xs->here)) {
			memcpy(buffer, (void *)xs->base +
			       le16_to_cpu(xs->here->xe_name_offset) +
			       OCFS2_XATTR_SIZE(xs->here->xe_name_len), size);
		} else {
			xv = (struct ocfs2_xattr_value_root *)
				(xs->base + le16_to_cpu(
				 xs->here->xe_name_offset) +
				OCFS2_XATTR_SIZE(xs->here->xe_name_len));
			ret = ocfs2_xattr_get_value_outside(inode, xv,
							    buffer, size);
			if (ret < 0) {
				mlog_errno(ret);
				return ret;
			}
		}
	}

	return size;
}

static int ocfs2_xattr_block_get(struct inode *inode,
				 int name_index,
				 const char *name,
				 void *buffer,
				 size_t buffer_size,
				 struct ocfs2_xattr_search *xs)
{
	struct ocfs2_xattr_block *xb;
	struct ocfs2_xattr_value_root *xv;
	size_t size;
	int ret = -ENODATA, name_offset, name_len, i;
	int uninitialized_var(block_off);

	xs->bucket = ocfs2_xattr_bucket_new(inode);
	if (!xs->bucket) {
		ret = -ENOMEM;
		mlog_errno(ret);
		goto cleanup;
	}

	ret = ocfs2_xattr_block_find(inode, name_index, name, xs);
	if (ret) {
		mlog_errno(ret);
		goto cleanup;
	}

	if (xs->not_found) {
		ret = -ENODATA;
		goto cleanup;
	}

	xb = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data;
	size = le64_to_cpu(xs->here->xe_value_size);
	if (buffer) {
		ret = -ERANGE;
		if (size > buffer_size)
			goto cleanup;

		name_offset = le16_to_cpu(xs->here->xe_name_offset);
		name_len = OCFS2_XATTR_SIZE(xs->here->xe_name_len);
		i = xs->here - xs->header->xh_entries;

		if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) {
			ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb,
								bucket_xh(xs->bucket),
								i,
								&block_off,
								&name_offset);
			xs->base = bucket_block(xs->bucket, block_off);
		}
		if (ocfs2_xattr_is_local(xs->here)) {
			memcpy(buffer, (void *)xs->base +
			       name_offset + name_len, size);
		} else {
			xv = (struct ocfs2_xattr_value_root *)
				(xs->base + name_offset + name_len);
			ret = ocfs2_xattr_get_value_outside(inode, xv,
							    buffer, size);
			if (ret < 0) {
				mlog_errno(ret);
				goto cleanup;
			}
		}
	}
	ret = size;
cleanup:
	ocfs2_xattr_bucket_free(xs->bucket);

	brelse(xs->xattr_bh);
	xs->xattr_bh = NULL;
	return ret;
}

int ocfs2_xattr_get_nolock(struct inode *inode,
			   struct buffer_head *di_bh,
			   int name_index,
			   const char *name,
			   void *buffer,
			   size_t buffer_size)
{
	int ret;
	struct ocfs2_dinode *di = NULL;
	struct ocfs2_inode_info *oi = OCFS2_I(inode);
	struct ocfs2_xattr_search xis = {
		.not_found = -ENODATA,
	};
	struct ocfs2_xattr_search xbs = {
		.not_found = -ENODATA,
	};

	if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb)))
		return -EOPNOTSUPP;

	if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL))
		ret = -ENODATA;

	xis.inode_bh = xbs.inode_bh = di_bh;
	di = (struct ocfs2_dinode *)di_bh->b_data;

	ret = ocfs2_xattr_ibody_get(inode, name_index, name, buffer,
				    buffer_size, &xis);
	if (ret == -ENODATA && di->i_xattr_loc)
		ret = ocfs2_xattr_block_get(inode, name_index, name, buffer,
					    buffer_size, &xbs);

	return ret;
}

/* ocfs2_xattr_get()
 *
 * Copy an extended attribute into the buffer provided.
 * Buffer is NULL to compute the size of buffer required.
 */
static int ocfs2_xattr_get(struct inode *inode,
			   int name_index,
			   const char *name,
			   void *buffer,
			   size_t buffer_size)
{
	int ret;
	struct buffer_head *di_bh = NULL;

	ret = ocfs2_inode_lock(inode, &di_bh, 0);
	if (ret < 0) {
		mlog_errno(ret);
		return ret;
	}
	down_read(&OCFS2_I(inode)->ip_xattr_sem);
	ret = ocfs2_xattr_get_nolock(inode, di_bh, name_index,
				     name, buffer, buffer_size);
	up_read(&OCFS2_I(inode)->ip_xattr_sem);

	ocfs2_inode_unlock(inode, 0);

	brelse(di_bh);

	return ret;
}

static int __ocfs2_xattr_set_value_outside(struct inode *inode,
					   handle_t *handle,
					   struct ocfs2_xattr_value_buf *vb,
					   const void *value,
					   int value_len)
{
	int ret = 0, i, cp_len;
	u16 blocksize = inode->i_sb->s_blocksize;
	u32 p_cluster, num_clusters;
	u32 cpos = 0, bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
	u32 clusters = ocfs2_clusters_for_bytes(inode->i_sb, value_len);
	u64 blkno;
	struct buffer_head *bh = NULL;
	unsigned int ext_flags;
	struct ocfs2_xattr_value_root *xv = vb->vb_xv;

	BUG_ON(clusters > le32_to_cpu(xv->xr_clusters));

	while (cpos < clusters) {
		ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster,
					       &num_clusters, &xv->xr_list,
					       &ext_flags);
		if (ret) {
			mlog_errno(ret);
			goto out;
		}

		BUG_ON(ext_flags & OCFS2_EXT_REFCOUNTED);

		blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster);

		for (i = 0; i < num_clusters * bpc; i++, blkno++) {
			ret = ocfs2_read_block(INODE_CACHE(inode), blkno,
					       &bh, NULL);
			if (ret) {
				mlog_errno(ret);
				goto out;
			}

			ret = ocfs2_journal_access(handle,
						   INODE_CACHE(inode),
						   bh,
						   OCFS2_JOURNAL_ACCESS_WRITE);
			if (ret < 0) {
				mlog_errno(ret);
				goto out;
			}

			cp_len = value_len > blocksize ? blocksize : value_len;
			memcpy(bh->b_data, value, cp_len);
			value_len -= cp_len;
			value += cp_len;
			if (cp_len < blocksize)
				memset(bh->b_data + cp_len, 0,
				       blocksize - cp_len);

			ocfs2_journal_dirty(handle, bh);
			brelse(bh);
			bh = NULL;

			/*
			 * XXX: do we need to empty all the following
			 * blocks in this cluster?
			 */
			if (!value_len)
				break;
		}
		cpos += num_clusters;
	}
out:
	brelse(bh);

	return ret;
}

static int ocfs2_xa_check_space_helper(int needed_space, int free_start,
				       int num_entries)
{
	int free_space;

	if (!needed_space)
		return 0;

	free_space = free_start -
		sizeof(struct ocfs2_xattr_header) -
		(num_entries * sizeof(struct ocfs2_xattr_entry)) -
		OCFS2_XATTR_HEADER_GAP;
	if (free_space < 0)
		return -EIO;
	if (free_space < needed_space)
		return -ENOSPC;

	return 0;
}

static int ocfs2_xa_journal_access(handle_t *handle, struct ocfs2_xa_loc *loc,
				   int type)
{
	return loc->xl_ops->xlo_journal_access(handle, loc, type);
}

static void ocfs2_xa_journal_dirty(handle_t *handle, struct ocfs2_xa_loc *loc)
{
	loc->xl_ops->xlo_journal_dirty(handle, loc);
}

/* Give a pointer into the storage for the given offset */
static void *ocfs2_xa_offset_pointer(struct ocfs2_xa_loc *loc, int offset)
{
	BUG_ON(offset >= loc->xl_size);
	return loc->xl_ops->xlo_offset_pointer(loc, offset);
}

/*
 * Wipe the name+value pair and allow the storage to reclaim it.  This
 * must be followed by either removal of the entry or a call to
 * ocfs2_xa_add_namevalue().
 */
static void ocfs2_xa_wipe_namevalue(struct ocfs2_xa_loc *loc)
{
	loc->xl_ops->xlo_wipe_namevalue(loc);
}

/*
 * Find lowest offset to a name+value pair.  This is the start of our
 * downward-growing free space.
 */
static int ocfs2_xa_get_free_start(struct ocfs2_xa_loc *loc)
{
	return loc->xl_ops->xlo_get_free_start(loc);
}

/* Can we reuse loc->xl_entry for xi? */
static int ocfs2_xa_can_reuse_entry(struct ocfs2_xa_loc *loc,
				    struct ocfs2_xattr_info *xi)
{
	return loc->xl_ops->xlo_can_reuse(loc, xi);
}

/* How much free space is needed to set the new value */
static int ocfs2_xa_check_space(struct ocfs2_xa_loc *loc,
				struct ocfs2_xattr_info *xi)
{
	return loc->xl_ops->xlo_check_space(loc, xi);
}

static void ocfs2_xa_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash)
{
	loc->xl_ops->xlo_add_entry(loc, name_hash);
	loc->xl_entry->xe_name_hash = cpu_to_le32(name_hash);
	/*
	 * We can't leave the new entry's xe_name_offset at zero or
	 * add_namevalue() will go nuts.  We set it to the size of our
	 * storage so that it can never be less than any other entry.
	 */
	loc->xl_entry->xe_name_offset = cpu_to_le16(loc->xl_size);
}

static void ocfs2_xa_add_namevalue(struct ocfs2_xa_loc *loc,
				   struct ocfs2_xattr_info *xi)
{
	int size = namevalue_size_xi(xi);
	int nameval_offset;
	char *nameval_buf;

	loc->xl_ops->xlo_add_namevalue(loc, size);
	loc->xl_entry->xe_value_size = cpu_to_le64(xi->xi_value_len);
	loc->xl_entry->xe_name_len = xi->xi_name_len;
	ocfs2_xattr_set_type(loc->xl_entry, xi->xi_name_index);
	ocfs2_xattr_set_local(loc->xl_entry,
			      xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE);

	nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset);
	nameval_buf = ocfs2_xa_offset_pointer(loc, nameval_offset);
	memset(nameval_buf, 0, size);
	memcpy(nameval_buf, xi->xi_name, xi->xi_name_len);
}

static void ocfs2_xa_fill_value_buf(struct ocfs2_xa_loc *loc,
				    struct ocfs2_xattr_value_buf *vb)
{
	int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset);
	int name_size = OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len);

	/* Value bufs are for value trees */
	BUG_ON(ocfs2_xattr_is_local(loc->xl_entry));
	BUG_ON(namevalue_size_xe(loc->xl_entry) !=
	       (name_size + OCFS2_XATTR_ROOT_SIZE));

	loc->xl_ops->xlo_fill_value_buf(loc, vb);
	vb->vb_xv =
		(struct ocfs2_xattr_value_root *)ocfs2_xa_offset_pointer(loc,
							nameval_offset +
							name_size);
}

static int ocfs2_xa_block_journal_access(handle_t *handle,
					 struct ocfs2_xa_loc *loc, int type)
{
	struct buffer_head *bh = loc->xl_storage;
	ocfs2_journal_access_func access;

	if (loc->xl_size == (bh->b_size -
			     offsetof(struct ocfs2_xattr_block,
				      xb_attrs.xb_header)))
		access = ocfs2_journal_access_xb;
	else
		access = ocfs2_journal_access_di;
	return access(handle, INODE_CACHE(loc->xl_inode), bh, type);
}

static void ocfs2_xa_block_journal_dirty(handle_t *handle,
					 struct ocfs2_xa_loc *loc)
{
	struct buffer_head *bh = loc->xl_storage;

	ocfs2_journal_dirty(handle, bh);
}

static void *ocfs2_xa_block_offset_pointer(struct ocfs2_xa_loc *loc,
					   int offset)
{
	return (char *)loc->xl_header + offset;
}

static int ocfs2_xa_block_can_reuse(struct ocfs2_xa_loc *loc,
				    struct ocfs2_xattr_info *xi)
{
	/*
	 * Block storage is strict.  If the sizes aren't exact, we will
	 * remove the old one and reinsert the new.
	 */
	return namevalue_size_xe(loc->xl_entry) ==
		namevalue_size_xi(xi);
}

static int ocfs2_xa_block_get_free_start(struct ocfs2_xa_loc *loc)
{
	struct ocfs2_xattr_header *xh = loc->xl_header;
	int i, count = le16_to_cpu(xh->xh_count);
	int offset, free_start = loc->xl_size;

	for (i = 0; i < count; i++) {
		offset = le16_to_cpu(xh->xh_entries[i].xe_name_offset);
		if (offset < free_start)
			free_start = offset;
	}

	return free_start;
}

static int ocfs2_xa_block_check_space(struct ocfs2_xa_loc *loc,
				      struct ocfs2_xattr_info *xi)
{
	int count = le16_to_cpu(loc->xl_header->xh_count);
	int free_start = ocfs2_xa_get_free_start(loc);
	int needed_space = ocfs2_xi_entry_usage(xi);

	/*
	 * Block storage will reclaim the original entry before inserting
	 * the new value, so we only need the difference.  If the new
	 * entry is smaller than the old one, we don't need anything.
	 */
	if (loc->xl_entry) {
		/* Don't need space if we're reusing! */
		if (ocfs2_xa_can_reuse_entry(loc, xi))
			needed_space = 0;
		else
			needed_space -= ocfs2_xe_entry_usage(loc->xl_entry);
	}
	if (needed_space < 0)
		needed_space = 0;
	return ocfs2_xa_check_space_helper(needed_space, free_start, count);
}

/*
 * Block storage for xattrs keeps the name+value pairs compacted.  When
 * we remove one, we have to shift any that preceded it towards the end.
 */
static void ocfs2_xa_block_wipe_namevalue(struct ocfs2_xa_loc *loc)
{
	int i, offset;
	int namevalue_offset, first_namevalue_offset, namevalue_size;
	struct ocfs2_xattr_entry *entry = loc->xl_entry;
	struct ocfs2_xattr_header *xh = loc->xl_header;
	int count = le16_to_cpu(xh->xh_count);

	namevalue_offset = le16_to_cpu(entry->xe_name_offset);
	namevalue_size = namevalue_size_xe(entry);
	first_namevalue_offset = ocfs2_xa_get_free_start(loc);

	/* Shift the name+value pairs */
	memmove((char *)xh + first_namevalue_offset + namevalue_size,
		(char *)xh + first_namevalue_offset,
		namevalue_offset - first_namevalue_offset);
	memset((char *)xh + first_namevalue_offset, 0, namevalue_size);

	/* Now tell xh->xh_entries about it */
	for (i = 0; i < count; i++) {
		offset = le16_to_cpu(xh->xh_entries[i].xe_name_offset);
		if (offset <= namevalue_offset)
			le16_add_cpu(&xh->xh_entries[i].xe_name_offset,
				     namevalue_size);
	}

	/*
	 * Note that we don't update xh_free_start or xh_name_value_len
	 * because they're not used in block-stored xattrs.
	 */
}

static void ocfs2_xa_block_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash)
{
	int count = le16_to_cpu(loc->xl_header->xh_count);
	loc->xl_entry = &(loc->xl_header->xh_entries[count]);
	le16_add_cpu(&loc->xl_header->xh_count, 1);
	memset(loc->xl_entry, 0, sizeof(struct ocfs2_xattr_entry));
}

static void ocfs2_xa_block_add_namevalue(struct ocfs2_xa_loc *loc, int size)
{
	int free_start = ocfs2_xa_get_free_start(loc);

	loc->xl_entry->xe_name_offset = cpu_to_le16(free_start - size);
}

static void ocfs2_xa_block_fill_value_buf(struct ocfs2_xa_loc *loc,
					  struct ocfs2_xattr_value_buf *vb)
{
	struct buffer_head *bh = loc->xl_storage;

	if (loc->xl_size == (bh->b_size -
			     offsetof(struct ocfs2_xattr_block,
				      xb_attrs.xb_header)))
		vb->vb_access = ocfs2_journal_access_xb;
	else
		vb->vb_access = ocfs2_journal_access_di;
	vb->vb_bh = bh;
}

/*
 * Operations for xattrs stored in blocks.  This includes inline inode
 * storage and unindexed ocfs2_xattr_blocks.
 */
static const struct ocfs2_xa_loc_operations ocfs2_xa_block_loc_ops = {
	.xlo_journal_access	= ocfs2_xa_block_journal_access,
	.xlo_journal_dirty	= ocfs2_xa_block_journal_dirty,
	.xlo_offset_pointer	= ocfs2_xa_block_offset_pointer,
	.xlo_check_space	= ocfs2_xa_block_check_space,
	.xlo_can_reuse		= ocfs2_xa_block_can_reuse,
	.xlo_get_free_start	= ocfs2_xa_block_get_free_start,
	.xlo_wipe_namevalue	= ocfs2_xa_block_wipe_namevalue,
	.xlo_add_entry		= ocfs2_xa_block_add_entry,
	.xlo_add_namevalue	= ocfs2_xa_block_add_namevalue,
	.xlo_fill_value_buf	= ocfs2_xa_block_fill_value_buf,
};

static int ocfs2_xa_bucket_journal_access(handle_t *handle,
					  struct ocfs2_xa_loc *loc, int type)
{
	struct ocfs2_xattr_bucket *bucket = loc->xl_storage;

	return ocfs2_xattr_bucket_journal_access(handle, bucket, type);
}

static void ocfs2_xa_bucket_journal_dirty(handle_t *handle,
					  struct ocfs2_xa_loc *loc)
{
	struct ocfs2_xattr_bucket *bucket = loc->xl_storage;

	ocfs2_xattr_bucket_journal_dirty(handle, bucket);
}

static void *ocfs2_xa_bucket_offset_pointer(struct ocfs2_xa_loc *loc,
					    int offset)
{
	struct ocfs2_xattr_bucket *bucket = loc->xl_storage;
	int block, block_offset;

	/* The header is at the front of the bucket */
	block = offset >> loc->xl_inode->i_sb->s_blocksize_bits;
	block_offset = offset % loc->xl_inode->i_sb->s_blocksize;

	return bucket_block(bucket, block) + block_offset;
}

static int ocfs2_xa_bucket_can_reuse(struct ocfs2_xa_loc *loc,
				     struct ocfs2_xattr_info *xi)
{
	return namevalue_size_xe(loc->xl_entry) >=
		namevalue_size_xi(xi);
}

static int ocfs2_xa_bucket_get_free_start(struct ocfs2_xa_loc *loc)
{
	struct ocfs2_xattr_bucket *bucket = loc->xl_storage;
	return le16_to_cpu(bucket_xh(bucket)->xh_free_start);
}

static int ocfs2_bucket_align_free_start(struct super_block *sb,
					 int free_start, int size)
{
	/*
	 * We need to make sure that the name+value pair fits within
	 * one block.
	 */
	if (((free_start - size) >> sb->s_blocksize_bits) !=
	    ((free_start - 1) >> sb->s_blocksize_bits))
		free_start -= free_start % sb->s_blocksize;

	return free_start;
}

static int ocfs2_xa_bucket_check_space(struct ocfs2_xa_loc *loc,
				       struct ocfs2_xattr_info *xi)
{
	int rc;
	int count = le16_to_cpu(loc->xl_header->xh_count);
	int free_start = ocfs2_xa_get_free_start(loc);
	int needed_space = ocfs2_xi_entry_usage(xi);
	int size = namevalue_size_xi(xi);
	struct super_block *sb = loc->xl_inode->i_sb;

	/*
	 * Bucket storage does not reclaim name+value pairs it cannot
	 * reuse.  They live as holes until the bucket fills, and then
	 * the bucket is defragmented.  However, the bucket can reclaim
	 * the ocfs2_xattr_entry.
	 */
	if (loc->xl_entry) {
		/* Don't need space if we're reusing! */
		if (ocfs2_xa_can_reuse_entry(loc, xi))
			needed_space = 0;
		else
			needed_space -= sizeof(struct ocfs2_xattr_entry);
	}
	BUG_ON(needed_space < 0);

	if (free_start < size) {
		if (needed_space)
			return -ENOSPC;
	} else {
		/*
		 * First we check if it would fit in the first place.
		 * Below, we align the free start to a block.  This may
		 * slide us below the minimum gap.  By checking unaligned
		 * first, we avoid that error.
		 */
		rc = ocfs2_xa_check_space_helper(needed_space, free_start,
						 count);
		if (rc)
			return rc;
		free_start = ocfs2_bucket_align_free_start(sb, free_start,
							   size);
	}
	return ocfs2_xa_check_space_helper(needed_space, free_start, count);
}

static void ocfs2_xa_bucket_wipe_namevalue(struct ocfs2_xa_loc *loc)
{
	le16_add_cpu(&loc->xl_header->xh_name_value_len,
		     -namevalue_size_xe(loc->xl_entry));
}

static void ocfs2_xa_bucket_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash)
{
	struct ocfs2_xattr_header *xh = loc->xl_header;
	int count = le16_to_cpu(xh->xh_count);
	int low = 0, high = count - 1, tmp;
	struct ocfs2_xattr_entry *tmp_xe;

	/*
	 * We keep buckets sorted by name_hash, so we need to find
	 * our insert place.
	 */
	while (low <= high && count) {
		tmp = (low + high) / 2;
		tmp_xe = &xh->xh_entries[tmp];

		if (name_hash > le32_to_cpu(tmp_xe->xe_name_hash))
			low = tmp + 1;
		else if (name_hash < le32_to_cpu(tmp_xe->xe_name_hash))
			high = tmp - 1;
		else {
			low = tmp;
			break;
		}
	}

	if (low != count)
		memmove(&xh->xh_entries[low + 1],
			&xh->xh_entries[low],
			((count - low) * sizeof(struct ocfs2_xattr_entry)));

	le16_add_cpu(&xh->xh_count, 1);
	loc->xl_entry = &xh->xh_entries[low];
	memset(loc->xl_entry, 0, sizeof(struct ocfs2_xattr_entry));
}

static void ocfs2_xa_bucket_add_namevalue(struct ocfs2_xa_loc *loc, int size)
{
	int free_start = ocfs2_xa_get_free_start(loc);
	struct ocfs2_xattr_header *xh = loc->xl_header;
	struct super_block *sb = loc->xl_inode->i_sb;
	int nameval_offset;

	free_start = ocfs2_bucket_align_free_start(sb, free_start, size);
	nameval_offset = free_start - size;
	loc->xl_entry->xe_name_offset = cpu_to_le16(nameval_offset);
	xh->xh_free_start = cpu_to_le16(nameval_offset);
	le16_add_cpu(&xh->xh_name_value_len, size);

}

static void ocfs2_xa_bucket_fill_value_buf(struct ocfs2_xa_loc *loc,
					   struct ocfs2_xattr_value_buf *vb)
{
	struct ocfs2_xattr_bucket *bucket = loc->xl_storage;
	struct super_block *sb = loc->xl_inode->i_sb;
	int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset);
	int size = namevalue_size_xe(loc->xl_entry);
	int block_offset = nameval_offset >> sb->s_blocksize_bits;

	/* Values are not allowed to straddle block boundaries */
	BUG_ON(block_offset !=
	       ((nameval_offset + size - 1) >> sb->s_blocksize_bits));
	/* We expect the bucket to be filled in */
	BUG_ON(!bucket->bu_bhs[block_offset]);

	vb->vb_access = ocfs2_journal_access;
	vb->vb_bh = bucket->bu_bhs[block_offset];
}

/* Operations for xattrs stored in buckets. */
static const struct ocfs2_xa_loc_operations ocfs2_xa_bucket_loc_ops = {
	.xlo_journal_access	= ocfs2_xa_bucket_journal_access,
	.xlo_journal_dirty	= ocfs2_xa_bucket_journal_dirty,
	.xlo_offset_pointer	= ocfs2_xa_bucket_offset_pointer,
	.xlo_check_space	= ocfs2_xa_bucket_check_space,
	.xlo_can_reuse		= ocfs2_xa_bucket_can_reuse,
	.xlo_get_free_start	= ocfs2_xa_bucket_get_free_start,
	.xlo_wipe_namevalue	= ocfs2_xa_bucket_wipe_namevalue,
	.xlo_add_entry		= ocfs2_xa_bucket_add_entry,
	.xlo_add_namevalue	= ocfs2_xa_bucket_add_namevalue,
	.xlo_fill_value_buf	= ocfs2_xa_bucket_fill_value_buf,
};

static unsigned int ocfs2_xa_value_clusters(struct ocfs2_xa_loc *loc)
{
	struct ocfs2_xattr_value_buf vb;

	if (ocfs2_xattr_is_local(loc->xl_entry))
		return 0;

	ocfs2_xa_fill_value_buf(loc, &vb);
	return le32_to_cpu(vb.vb_xv->xr_clusters);
}

static int ocfs2_xa_value_truncate(struct ocfs2_xa_loc *loc, u64 bytes,
				   struct ocfs2_xattr_set_ctxt *ctxt)
{
	int trunc_rc, access_rc;
	struct ocfs2_xattr_value_buf vb;

	ocfs2_xa_fill_value_buf(loc, &vb);
	trunc_rc = ocfs2_xattr_value_truncate(loc->xl_inode, &vb, bytes,
					      ctxt);

	/*
	 * The caller of ocfs2_xa_value_truncate() has already called
	 * ocfs2_xa_journal_access on the loc.  However, The truncate code
	 * calls ocfs2_extend_trans().  This may commit the previous
	 * transaction and open a new one.  If this is a bucket, truncate
	 * could leave only vb->vb_bh set up for journaling.  Meanwhile,
	 * the caller is expecting to dirty the entire bucket.  So we must
	 * reset the journal work.  We do this even if truncate has failed,
	 * as it could have failed after committing the extend.
	 */
	access_rc = ocfs2_xa_journal_access(ctxt->handle, loc,
					    OCFS2_JOURNAL_ACCESS_WRITE);

	/* Errors in truncate take precedence */
	return trunc_rc ? trunc_rc : access_rc;
}

static void ocfs2_xa_remove_entry(struct ocfs2_xa_loc *loc)
{
	int index, count;
	struct ocfs2_xattr_header *xh = loc->xl_header;
	struct ocfs2_xattr_entry *entry = loc->xl_entry;

	ocfs2_xa_wipe_namevalue(loc);
	loc->xl_entry = NULL;

	le16_add_cpu(&xh->xh_count, -1);
	count = le16_to_cpu(xh->xh_count);

	/*
	 * Only zero out the entry if there are more remaining.  This is
	 * important for an empty bucket, as it keeps track of the
	 * bucket's hash value.  It doesn't hurt empty block storage.
	 */
	if (count) {
		index = ((char *)entry - (char *)&xh->xh_entries) /
			sizeof(struct ocfs2_xattr_entry);
		memmove(&xh->xh_entries[index], &xh->xh_entries[index + 1],
			(count - index) * sizeof(struct ocfs2_xattr_entry));
		memset(&xh->xh_entries[count], 0,
		       sizeof(struct ocfs2_xattr_entry));
	}
}

/*
 * If we have a problem adjusting the size of an external value during
 * ocfs2_xa_prepare_entry() or ocfs2_xa_remove(), we may have an xattr
 * in an intermediate state.  For example, the value may be partially
 * truncated.
 *
 * If the value tree hasn't changed, the extend/truncate went nowhere.
 * We have nothing to do.  The caller can treat it as a straight error.
 *
 * If the value tree got partially truncated, we now have a corrupted
 * extended attribute.  We're going to wipe its entry and leak the
 * clusters.  Better to leak some storage than leave a corrupt entry.
 *
 * If the value tree grew, it obviously didn't grow enough for the
 * new entry.  We're not going to try and reclaim those clusters either.
 * If there was already an external value there (orig_clusters != 0),
 * the new clusters are attached safely and we can just leave the old
 * value in place.  If there was no external value there, we remove
 * the entry.
 *
 * This way, the xattr block we store in the journal will be consistent.
 * If the size change broke because of the journal, no changes will hit
 * disk anyway.
 */
static void ocfs2_xa_cleanup_value_truncate(struct ocfs2_xa_loc *loc,
					    const char *what,
					    unsigned int orig_clusters)
{
	unsigned int new_clusters = ocfs2_xa_value_clusters(loc);
	char *nameval_buf = ocfs2_xa_offset_pointer(loc,
				le16_to_cpu(loc->xl_entry->xe_name_offset));

	if (new_clusters < orig_clusters) {
		mlog(ML_ERROR,
		     "Partial truncate while %s xattr %.*s.  Leaking "
		     "%u clusters and removing the entry\n",
		     what, loc->xl_entry->xe_name_len, nameval_buf,
		     orig_clusters - new_clusters);
		ocfs2_xa_remove_entry(loc);
	} else if (!orig_clusters) {
		mlog(ML_ERROR,
		     "Unable to allocate an external value for xattr "
		     "%.*s safely.  Leaking %u clusters and removing the "
		     "entry\n",
		     loc->xl_entry->xe_name_len, nameval_buf,
		     new_clusters - orig_clusters);
		ocfs2_xa_remove_entry(loc);
	} else if (new_clusters > orig_clusters)
		mlog(ML_ERROR,
		     "Unable to grow xattr %.*s safely.  %u new clusters "
		     "have been added, but the value will not be "
		     "modified\n",
		     loc->xl_entry->xe_name_len, nameval_buf,
		     new_clusters - orig_clusters);
}

static int ocfs2_xa_remove(struct ocfs2_xa_loc *loc,
			   struct ocfs2_xattr_set_ctxt *ctxt)
{
	int rc = 0;
	unsigned int orig_clusters;

	if (!ocfs2_xattr_is_local(loc->xl_entry)) {
		orig_clusters = ocfs2_xa_value_clusters(loc);
		rc = ocfs2_xa_value_truncate(loc, 0, ctxt);
		if (rc) {
			mlog_errno(rc);
			/*
			 * Since this is remove, we can return 0 if
			 * ocfs2_xa_cleanup_value_truncate() is going to
			 * wipe the entry anyway.  So we check the
			 * cluster count as well.
			 */
			if (orig_clusters != ocfs2_xa_value_clusters(loc))
				rc = 0;
			ocfs2_xa_cleanup_value_truncate(loc, "removing",
							orig_clusters);
			if (rc)
				goto out;
		}
	}

	ocfs2_xa_remove_entry(loc);

out:
	return rc;
}

static void ocfs2_xa_install_value_root(struct ocfs2_xa_loc *loc)
{
	int name_size = OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len);
	char *nameval_buf;

	nameval_buf = ocfs2_xa_offset_pointer(loc,
				le16_to_cpu(loc->xl_entry->xe_name_offset));
	memcpy(nameval_buf + name_size, &def_xv, OCFS2_XATTR_ROOT_SIZE);
}

/*
 * Take an existing entry and make it ready for the new value.  This
 * won't allocate space, but it may free space.  It should be ready for
 * ocfs2_xa_prepare_entry() to finish the work.
 */
static int ocfs2_xa_reuse_entry(struct ocfs2_xa_loc *loc,
				struct ocfs2_xattr_info *xi,
				struct ocfs2_xattr_set_ctxt *ctxt)
{
	int rc = 0;
	int name_size = OCFS2_XATTR_SIZE(xi->xi_name_len);
	unsigned int orig_clusters;
	char *nameval_buf;
	int xe_local = ocfs2_xattr_is_local(loc->xl_entry);
	int xi_local = xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE;

	BUG_ON(OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len) !=
	       name_size);

	nameval_buf = ocfs2_xa_offset_pointer(loc,
				le16_to_cpu(loc->xl_entry->xe_name_offset));
	if (xe_local) {
		memset(nameval_buf + name_size, 0,
		       namevalue_size_xe(loc->xl_entry) - name_size);
		if (!xi_local)
			ocfs2_xa_install_value_root(loc);
	} else {
		orig_clusters = ocfs2_xa_value_clusters(loc);
		if (xi_local) {
			rc = ocfs2_xa_value_truncate(loc, 0, ctxt);
			if (rc < 0)
				mlog_errno(rc);
			else
				memset(nameval_buf + name_size, 0,
				       namevalue_size_xe(loc->xl_entry) -
				       name_size);
		} else if (le64_to_cpu(loc->xl_entry->xe_value_size) >
			   xi->xi_value_len) {
			rc = ocfs2_xa_value_truncate(loc, xi->xi_value_len,
						     ctxt);
			if (rc < 0)
				mlog_errno(rc);
		}

		if (rc) {
			ocfs2_xa_cleanup_value_truncate(loc, "reusing",
							orig_clusters);
			goto out;
		}
	}

	loc->xl_entry->xe_value_size = cpu_to_le64(xi->xi_value_len);
	ocfs2_xattr_set_local(loc->xl_entry, xi_local);

out:
	return rc;
}

/*
 * Prepares loc->xl_entry to receive the new xattr.  This includes
 * properly setting up the name+value pair region.  If loc->xl_entry
 * already exists, it will take care of modifying it appropriately.
 *
 * Note that this modifies the data.  You did journal_access already,
 * right?
 */
static int ocfs2_xa_prepare_entry(struct ocfs2_xa_loc *loc,
				  struct ocfs2_xattr_info *xi,
				  u32 name_hash,
				  struct ocfs2_xattr_set_ctxt *ctxt)
{
	int rc = 0;
	unsigned int orig_clusters;
	__le64 orig_value_size = 0;

	rc = ocfs2_xa_check_space(loc, xi);
	if (rc)
		goto out;

	if (loc->xl_entry) {
		if (ocfs2_xa_can_reuse_entry(loc, xi)) {
			orig_value_size = loc->xl_entry->xe_value_size;
			rc = ocfs2_xa_reuse_entry(loc, xi, ctxt);
			if (rc)
				goto out;
			goto alloc_value;
		}

		if (!ocfs2_xattr_is_local(loc->xl_entry)) {
			orig_clusters = ocfs2_xa_value_clusters(loc);
			rc = ocfs2_xa_value_truncate(loc, 0, ctxt);
			if (rc) {
				mlog_errno(rc);
				ocfs2_xa_cleanup_value_truncate(loc,
								"overwriting",
								orig_clusters);
				goto out;
			}
		}
		ocfs2_xa_wipe_namevalue(loc);
	} else
		ocfs2_xa_add_entry(loc, name_hash);

	/*
	 * If we get here, we have a blank entry.  Fill it.  We grow our
	 * name+value pair back from the end.
	 */
	ocfs2_xa_add_namevalue(loc, xi);
	if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE)
		ocfs2_xa_install_value_root(loc);

alloc_value:
	if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) {
		orig_clusters = ocfs2_xa_value_clusters(loc);
		rc = ocfs2_xa_value_truncate(loc, xi->xi_value_len, ctxt);
		if (rc < 0) {
			ctxt->set_abort = 1;
			ocfs2_xa_cleanup_value_truncate(loc, "growing",
							orig_clusters);
			/*
			 * If we were growing an existing value,
			 * ocfs2_xa_cleanup_value_truncate() won't remove
			 * the entry. We need to restore the original value
			 * size.
			 */
			if (loc->xl_entry) {
				BUG_ON(!orig_value_size);
				loc->xl_entry->xe_value_size = orig_value_size;
			}
			mlog_errno(rc);
		}
	}

out:
	return rc;
}

/*
 * Store the value portion of the name+value pair.  This will skip
 * values that are stored externally.  Their tree roots were set up
 * by ocfs2_xa_prepare_entry().
 */
static int ocfs2_xa_store_value(struct ocfs2_xa_loc *loc,
				struct ocfs2_xattr_info *xi,
				struct ocfs2_xattr_set_ctxt *ctxt)
{
	int rc = 0;
	int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset);
	int name_size = OCFS2_XATTR_SIZE(xi->xi_name_len);
	char *nameval_buf;
	struct ocfs2_xattr_value_buf vb;

	nameval_buf = ocfs2_xa_offset_pointer(loc, nameval_offset);
	if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) {
		ocfs2_xa_fill_value_buf(loc, &vb);
		rc = __ocfs2_xattr_set_value_outside(loc->xl_inode,
						     ctxt->handle, &vb,
						     xi->xi_value,
						     xi->xi_value_len);
	} else
		memcpy(nameval_buf + name_size, xi->xi_value, xi->xi_value_len);

	return rc;
}

static int ocfs2_xa_set(struct ocfs2_xa_loc *loc,
			struct ocfs2_xattr_info *xi,
			struct ocfs2_xattr_set_ctxt *ctxt)
{
	int ret;
	u32 name_hash = ocfs2_xattr_name_hash(loc->xl_inode, xi->xi_name,
					      xi->xi_name_len);

	ret = ocfs2_xa_journal_access(ctxt->handle, loc,
				      OCFS2_JOURNAL_ACCESS_WRITE);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	/*
	 * From here on out, everything is going to modify the buffer a
	 * little.  Errors are going to leave the xattr header in a
	 * sane state.  Thus, even with errors we dirty the sucker.
	 */

	/* Don't worry, we are never called with !xi_value and !xl_entry */
	if (!xi->xi_value) {
		ret = ocfs2_xa_remove(loc, ctxt);
		goto out_dirty;
	}

	ret = ocfs2_xa_prepare_entry(loc, xi, name_hash, ctxt);
	if (ret) {
		if (ret != -ENOSPC)
			mlog_errno(ret);
		goto out_dirty;
	}

	ret = ocfs2_xa_store_value(loc, xi, ctxt);
	if (ret)
		mlog_errno(ret);

out_dirty:
	ocfs2_xa_journal_dirty(ctxt->handle, loc);

out:
	return ret;
}

static void ocfs2_init_dinode_xa_loc(struct ocfs2_xa_loc *loc,
				     struct inode *inode,
				     struct buffer_head *bh,
				     struct ocfs2_xattr_entry *entry)
{
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;

	BUG_ON(!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_XATTR_FL));

	loc->xl_inode = inode;
	loc->xl_ops = &ocfs2_xa_block_loc_ops;
	loc->xl_storage = bh;
	loc->xl_entry = entry;
	loc->xl_size = le16_to_cpu(di->i_xattr_inline_size);
	loc->xl_header =
		(struct ocfs2_xattr_header *)(bh->b_data + bh->b_size -
					      loc->xl_size);
}

static void ocfs2_init_xattr_block_xa_loc(struct ocfs2_xa_loc *loc,
					  struct inode *inode,
					  struct buffer_head *bh,
					  struct ocfs2_xattr_entry *entry)
{
	struct ocfs2_xattr_block *xb =
		(struct ocfs2_xattr_block *)bh->b_data;

	BUG_ON(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED);

	loc->xl_inode = inode;
	loc->xl_ops = &ocfs2_xa_block_loc_ops;
	loc->xl_storage = bh;
	loc->xl_header = &(xb->xb_attrs.xb_header);
	loc->xl_entry = entry;
	loc->xl_size = bh->b_size - offsetof(struct ocfs2_xattr_block,
					     xb_attrs.xb_header);
}

static void ocfs2_init_xattr_bucket_xa_loc(struct ocfs2_xa_loc *loc,
					   struct ocfs2_xattr_bucket *bucket,
					   struct ocfs2_xattr_entry *entry)
{
	loc->xl_inode = bucket->bu_inode;
	loc->xl_ops = &ocfs2_xa_bucket_loc_ops;
	loc->xl_storage = bucket;
	loc->xl_header = bucket_xh(bucket);
	loc->xl_entry = entry;
	loc->xl_size = OCFS2_XATTR_BUCKET_SIZE;
}

/*
 * In xattr remove, if it is stored outside and refcounted, we may have
 * the chance to split the refcount tree. So need the allocators.
 */
static int ocfs2_lock_xattr_remove_allocators(struct inode *inode,
					struct ocfs2_xattr_value_root *xv,
					struct ocfs2_caching_info *ref_ci,
					struct buffer_head *ref_root_bh,
					struct ocfs2_alloc_context **meta_ac,
					int *ref_credits)
{
	int ret, meta_add = 0;
	u32 p_cluster, num_clusters;
	unsigned int ext_flags;

	*ref_credits = 0;
	ret = ocfs2_xattr_get_clusters(inode, 0, &p_cluster,
				       &num_clusters,
				       &xv->xr_list,
				       &ext_flags);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	if (!(ext_flags & OCFS2_EXT_REFCOUNTED))
		goto out;

	ret = ocfs2_refcounted_xattr_delete_need(inode, ref_ci,
						 ref_root_bh, xv,
						 &meta_add, ref_credits);
	if (ret) {
		mlog_errno(ret);
		goto out;
	}

	ret = ocfs2_reserve_new_metadata_blocks(OCFS2_SB(inode->i_sb),
						meta_add, meta_ac);
	if (ret)
		mlog_errno(ret);

out:
	return ret;
}

static int ocfs2_remove_value_outside(struct inode*inode,
				      struct ocfs2_xattr_value_buf *vb,
				      struct ocfs2_xattr_header *header,
				      struct ocfs2_caching_info *ref_ci,
				      struct buffer_head *ref_root_bh)
{
	int ret = 0, i, ref_credits;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, };
	void *val;

	ocfs2_init_dealloc_ctxt(&ctxt.dealloc);

	for (i = 0; i < le16_to_cpu(header->xh_count); i++) {
		struct ocfs2_xattr_entry *entry = &header->xh_entries[i];

		if (ocfs2_xattr_is_local(entry))
			continue;

		val = (void *)header +
			le16_to_cpu(entry->xe_name_offset);
		vb->vb_xv = (struct ocfs2_xattr_value_root *)
			(val + OCFS2_XATTR_SIZE(entry->xe_name_len));

		ret = ocfs2_lock_xattr_remove_allocators(inode, vb->vb_xv,
							 ref_ci, ref_root_bh,
							 &ctxt.meta_ac,
							 &ref_credits);

		ctxt.handle = ocfs2_start_trans(osb, ref_credits +
					ocfs2_remove_extent_credits(osb->sb));
		if (IS_ERR(ctxt.handle)) {
			ret = PTR_ERR(ctxt.handle);
			mlog_errno(ret);
			break;
		}

		ret = ocfs2_xattr_value_truncate(inode, vb, 0, &ctxt);

		ocfs2_commit_trans(osb, ctxt.handle);
		if (ctxt.meta_ac) {
			ocfs2_free_alloc_context(ctxt.meta_ac);
			ctxt.meta_ac = NULL;
		}

		if (ret < 0) {
			mlog_errno(ret);
			break;
		}

	}

	if (ctxt.meta_ac)
		ocfs2_free_alloc_context(ctxt.meta_ac);
	ocfs2_schedule_truncate_log_flush(osb, 1);
	ocfs2_run_deallocs(osb, &ctxt.dealloc);
	return ret;
}

static int ocfs2_xattr_ibody_remove(struct inode *inode,
				    struct buffer_head *di_bh,
				    str