delayed-inode.c

/*
 * Copyright (C) 2011 Fujitsu.  All rights reserved.
 * Written by Miao Xie <miaox@cn.fujitsu.com>
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License v2 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/slab.h>
#include "delayed-inode.h"
#include "disk-io.h"
#include "transaction.h"

#define BTRFS_DELAYED_WRITEBACK		400
#define BTRFS_DELAYED_BACKGROUND	100

static struct kmem_cache *delayed_node_cache;

int __init btrfs_delayed_inode_init(void)
{
	delayed_node_cache = kmem_cache_create("delayed_node",
					sizeof(struct btrfs_delayed_node),
					0,
					SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
					NULL);
	if (!delayed_node_cache)
		return -ENOMEM;
	return 0;
}

void btrfs_delayed_inode_exit(void)
{
	if (delayed_node_cache)
		kmem_cache_destroy(delayed_node_cache);
}

static inline void btrfs_init_delayed_node(
				struct btrfs_delayed_node *delayed_node,
				struct btrfs_root *root, u64 inode_id)
{
	delayed_node->root = root;
	delayed_node->inode_id = inode_id;
	atomic_set(&delayed_node->refs, 0);
	delayed_node->count = 0;
	delayed_node->in_list = 0;
	delayed_node->inode_dirty = 0;
	delayed_node->ins_root = RB_ROOT;
	delayed_node->del_root = RB_ROOT;
	mutex_init(&delayed_node->mutex);
	delayed_node->index_cnt = 0;
	INIT_LIST_HEAD(&delayed_node->n_list);
	INIT_LIST_HEAD(&delayed_node->p_list);
	delayed_node->bytes_reserved = 0;
	memset(&delayed_node->inode_item, 0, sizeof(delayed_node->inode_item));
}

static inline int btrfs_is_continuous_delayed_item(
					struct btrfs_delayed_item *item1,
					struct btrfs_delayed_item *item2)
{
	if (item1->key.type == BTRFS_DIR_INDEX_KEY &&
	    item1->key.objectid == item2->key.objectid &&
	    item1->key.type == item2->key.type &&
	    item1->key.offset + 1 == item2->key.offset)
		return 1;
	return 0;
}

static inline struct btrfs_delayed_root *btrfs_get_delayed_root(
							struct btrfs_root *root)
{
	return root->fs_info->delayed_root;
}

static struct btrfs_delayed_node *btrfs_get_delayed_node(struct inode *inode)
{
	struct btrfs_inode *btrfs_inode = BTRFS_I(inode);
	struct btrfs_root *root = btrfs_inode->root;
	u64 ino = btrfs_ino(inode);
	struct btrfs_delayed_node *node;

	node = ACCESS_ONCE(btrfs_inode->delayed_node);
	if (node) {
		atomic_inc(&node->refs);
		return node;
	}

	spin_lock(&root->inode_lock);
	node = radix_tree_lookup(&root->delayed_nodes_tree, ino);
	if (node) {
		if (btrfs_inode->delayed_node) {
			atomic_inc(&node->refs);	/* can be accessed */
			BUG_ON(btrfs_inode->delayed_node != node);
			spin_unlock(&root->inode_lock);
			return node;
		}
		btrfs_inode->delayed_node = node;
		atomic_inc(&node->refs);	/* can be accessed */
		atomic_inc(&node->refs);	/* cached in the inode */
		spin_unlock(&root->inode_lock);
		return node;
	}
	spin_unlock(&root->inode_lock);

	return NULL;
}

/* Will return either the node or PTR_ERR(-ENOMEM) */
static struct btrfs_delayed_node *btrfs_get_or_create_delayed_node(
							struct inode *inode)
{
	struct btrfs_delayed_node *node;
	struct btrfs_inode *btrfs_inode = BTRFS_I(inode);
	struct btrfs_root *root = btrfs_inode->root;
	u64 ino = btrfs_ino(inode);
	int ret;

again:
	node = btrfs_get_delayed_node(inode);
	if (node)
		return node;

	node = kmem_cache_alloc(delayed_node_cache, GFP_NOFS);
	if (!node)
		return ERR_PTR(-ENOMEM);
	btrfs_init_delayed_node(node, root, ino);

	atomic_inc(&node->refs);	/* cached in the btrfs inode */
	atomic_inc(&node->refs);	/* can be accessed */

	ret = radix_tree_preload(GFP_NOFS & ~__GFP_HIGHMEM);
	if (ret) {
		kmem_cache_free(delayed_node_cache, node);
		return ERR_PTR(ret);
	}

	spin_lock(&root->inode_lock);
	ret = radix_tree_insert(&root->delayed_nodes_tree, ino, node);
	if (ret == -EEXIST) {
		kmem_cache_free(delayed_node_cache, node);
		spin_unlock(&root->inode_lock);
		radix_tree_preload_end();
		goto again;
	}
	btrfs_inode->delayed_node = node;
	spin_unlock(&root->inode_lock);
	radix_tree_preload_end();

	return node;
}

/*
 * Call it when holding delayed_node->mutex
 *
 * If mod = 1, add this node into the prepared list.
 */
static void btrfs_queue_delayed_node(struct btrfs_delayed_root *root,
				     struct btrfs_delayed_node *node,
				     int mod)
{
	spin_lock(&root->lock);
	if (node->in_list) {
		if (!list_empty(&node->p_list))
			list_move_tail(&node->p_list, &root->prepare_list);
		else if (mod)
			list_add_tail(&node->p_list, &root->prepare_list);
	} else {
		list_add_tail(&node->n_list, &root->node_list);
		list_add_tail(&node->p_list, &root->prepare_list);
		atomic_inc(&node->refs);	/* inserted into list */
		root->nodes++;
		node->in_list = 1;
	}
	spin_unlock(&root->lock);
}

/* Call it when holding delayed_node->mutex */
static void btrfs_dequeue_delayed_node(struct btrfs_delayed_root *root,
				       struct btrfs_delayed_node *node)
{
	spin_lock(&root->lock);
	if (node->in_list) {
		root->nodes--;
		atomic_dec(&node->refs);	/* not in the list */
		list_del_init(&node->n_list);
		if (!list_empty(&node->p_list))
			list_del_init(&node->p_list);
		node->in_list = 0;
	}
	spin_unlock(&root->lock);
}

struct btrfs_delayed_node *btrfs_first_delayed_node(
			struct btrfs_delayed_root *delayed_root)
{
	struct list_head *p;
	struct btrfs_delayed_node *node = NULL;

	spin_lock(&delayed_root->lock);
	if (list_empty(&delayed_root->node_list))
		goto out;

	p = delayed_root->node_list.next;
	node = list_entry(p, struct btrfs_delayed_node, n_list);
	atomic_inc(&node->refs);
out:
	spin_unlock(&delayed_root->lock);

	return node;
}

struct btrfs_delayed_node *btrfs_next_delayed_node(
						struct btrfs_delayed_node *node)
{
	struct btrfs_delayed_root *delayed_root;
	struct list_head *p;
	struct btrfs_delayed_node *next = NULL;

	delayed_root = node->root->fs_info->delayed_root;
	spin_lock(&delayed_root->lock);
	if (!node->in_list) {	/* not in the list */
		if (list_empty(&delayed_root->node_list))
			goto out;
		p = delayed_root->node_list.next;
	} else if (list_is_last(&node->n_list, &delayed_root->node_list))
		goto out;
	else
		p = node->n_list.next;

	next = list_entry(p, struct btrfs_delayed_node, n_list);
	atomic_inc(&next->refs);
out:
	spin_unlock(&delayed_root->lock);

	return next;
}

static void __btrfs_release_delayed_node(
				struct btrfs_delayed_node *delayed_node,
				int mod)
{
	struct btrfs_delayed_root *delayed_root;

	if (!delayed_node)
		return;

	delayed_root = delayed_node->root->fs_info->delayed_root;

	mutex_lock(&delayed_node->mutex);
	if (delayed_node->count)
		btrfs_queue_delayed_node(delayed_root, delayed_node, mod);
	else
		btrfs_dequeue_delayed_node(delayed_root, delayed_node);
	mutex_unlock(&delayed_node->mutex);

	if (atomic_dec_and_test(&delayed_node->refs)) {
		struct btrfs_root *root = delayed_node->root;
		spin_lock(&root->inode_lock);
		if (atomic_read(&delayed_node->refs) == 0) {
			radix_tree_delete(&root->delayed_nodes_tree,
					  delayed_node->inode_id);
			kmem_cache_free(delayed_node_cache, delayed_node);
		}
		spin_unlock(&root->inode_lock);
	}
}

static inline void btrfs_release_delayed_node(struct btrfs_delayed_node *node)
{
	__btrfs_release_delayed_node(node, 0);
}

struct btrfs_delayed_node *btrfs_first_prepared_delayed_node(
					struct btrfs_delayed_root *delayed_root)
{
	struct list_head *p;
	struct btrfs_delayed_node *node = NULL;

	spin_lock(&delayed_root->lock);
	if (list_empty(&delayed_root->prepare_list))
		goto out;

	p = delayed_root->prepare_list.next;
	list_del_init(p);
	node = list_entry(p, struct btrfs_delayed_node, p_list);
	atomic_inc(&node->refs);
out:
	spin_unlock(&delayed_root->lock);

	return node;
}

static inline void btrfs_release_prepared_delayed_node(
					struct btrfs_delayed_node *node)
{
	__btrfs_release_delayed_node(node, 1);
}

struct btrfs_delayed_item *btrfs_alloc_delayed_item(u32 data_len)
{
	struct btrfs_delayed_item *item;
	item = kmalloc(sizeof(*item) + data_len, GFP_NOFS);
	if (item) {
		item->data_len = data_len;
		item->ins_or_del = 0;
		item->bytes_reserved = 0;
		item->delayed_node = NULL;
		atomic_set(&item->refs, 1);
	}
	return item;
}

/*
 * __btrfs_lookup_delayed_item - look up the delayed item by key
 * @delayed_node: pointer to the delayed node
 * @key:	  the key to look up
 * @prev:	  used to store the prev item if the right item isn't found
 * @next:	  used to store the next item if the right item isn't found
 *
 * Note: if we don't find the right item, we will return the prev item and
 * the next item.
 */
static struct btrfs_delayed_item *__btrfs_lookup_delayed_item(
				struct rb_root *root,
				struct btrfs_key *key,
				struct btrfs_delayed_item **prev,
				struct btrfs_delayed_item **next)
{
	struct rb_node *node, *prev_node = NULL;
	struct btrfs_delayed_item *delayed_item = NULL;
	int ret = 0;

	node = root->rb_node;

	while (node) {
		delayed_item = rb_entry(node, struct btrfs_delayed_item,
					rb_node);
		prev_node = node;
		ret = btrfs_comp_cpu_keys(&delayed_item->key, key);
		if (ret < 0)
			node = node->rb_right;
		else if (ret > 0)
			node = node->rb_left;
		else
			return delayed_item;
	}

	if (prev) {
		if (!prev_node)
			*prev = NULL;
		else if (ret < 0)
			*prev = delayed_item;
		else if ((node = rb_prev(prev_node)) != NULL) {
			*prev = rb_entry(node, struct btrfs_delayed_item,
					 rb_node);
		} else
			*prev = NULL;
	}

	if (next) {
		if (!prev_node)
			*next = NULL;
		else if (ret > 0)
			*next = delayed_item;
		else if ((node = rb_next(prev_node)) != NULL) {
			*next = rb_entry(node, struct btrfs_delayed_item,
					 rb_node);
		} else
			*next = NULL;
	}
	return NULL;
}

struct btrfs_delayed_item *__btrfs_lookup_delayed_insertion_item(
					struct btrfs_delayed_node *delayed_node,
					struct btrfs_key *key)
{
	struct btrfs_delayed_item *item;

	item = __btrfs_lookup_delayed_item(&delayed_node->ins_root, key,
					   NULL, NULL);
	return item;
}

struct btrfs_delayed_item *__btrfs_lookup_delayed_deletion_item(
					struct btrfs_delayed_node *delayed_node,
					struct btrfs_key *key)
{
	struct btrfs_delayed_item *item;

	item = __btrfs_lookup_delayed_item(&delayed_node->del_root, key,
					   NULL, NULL);
	return item;
}

struct btrfs_delayed_item *__btrfs_search_delayed_insertion_item(
					struct btrfs_delayed_node *delayed_node,
					struct btrfs_key *key)
{
	struct btrfs_delayed_item *item, *next;

	item = __btrfs_lookup_delayed_item(&delayed_node->ins_root, key,
					   NULL, &next);
	if (!item)
		item = next;

	return item;
}

struct btrfs_delayed_item *__btrfs_search_delayed_deletion_item(
					struct btrfs_delayed_node *delayed_node,
					struct btrfs_key *key)
{
	struct btrfs_delayed_item *item, *next;

	item = __btrfs_lookup_delayed_item(&delayed_node->del_root, key,
					   NULL, &next);
	if (!item)
		item = next;

	return item;
}

static int __btrfs_add_delayed_item(struct btrfs_delayed_node *delayed_node,
				    struct btrfs_delayed_item *ins,
				    int action)
{
	struct rb_node **p, *node;
	struct rb_node *parent_node = NULL;
	struct rb_root *root;
	struct btrfs_delayed_item *item;
	int cmp;

	if (action == BTRFS_DELAYED_INSERTION_ITEM)
		root = &delayed_node->ins_root;
	else if (action == BTRFS_DELAYED_DELETION_ITEM)
		root = &delayed_node->del_root;
	else
		BUG();
	p = &root->rb_node;
	node = &ins->rb_node;

	while (*p) {
		parent_node = *p;
		item = rb_entry(parent_node, struct btrfs_delayed_item,
				 rb_node);

		cmp = btrfs_comp_cpu_keys(&item->key, &ins->key);
		if (cmp < 0)
			p = &(*p)->rb_right;
		else if (cmp > 0)
			p = &(*p)->rb_left;
		else
			return -EEXIST;
	}

	rb_link_node(node, parent_node, p);
	rb_insert_color(node, root);
	ins->delayed_node = delayed_node;
	ins->ins_or_del = action;

	if (ins->key.type == BTRFS_DIR_INDEX_KEY &&
	    action == BTRFS_DELAYED_INSERTION_ITEM &&
	    ins->key.offset >= delayed_node->index_cnt)
			delayed_node->index_cnt = ins->key.offset + 1;

	delayed_node->count++;
	atomic_inc(&delayed_node->root->fs_info->delayed_root->items);
	return 0;
}

static int __btrfs_add_delayed_insertion_item(struct btrfs_delayed_node *node,
					      struct btrfs_delayed_item *item)
{
	return __btrfs_add_delayed_item(node, item,
					BTRFS_DELAYED_INSERTION_ITEM);
}

static int __btrfs_add_delayed_deletion_item(struct btrfs_delayed_node *node,
					     struct btrfs_delayed_item *item)
{
	return __btrfs_add_delayed_item(node, item,
					BTRFS_DELAYED_DELETION_ITEM);
}

static void __btrfs_remove_delayed_item(struct btrfs_delayed_item *delayed_item)
{
	struct rb_root *root;
	struct btrfs_delayed_root *delayed_root;

	delayed_root = delayed_item->delayed_node->root->fs_info->delayed_root;

	BUG_ON(!delayed_root);
	BUG_ON(delayed_item->ins_or_del != BTRFS_DELAYED_DELETION_ITEM &&
	       delayed_item->ins_or_del != BTRFS_DELAYED_INSERTION_ITEM);

	if (delayed_item->ins_or_del == BTRFS_DELAYED_INSERTION_ITEM)
		root = &delayed_item->delayed_node->ins_root;
	else
		root = &delayed_item->delayed_node->del_root;

	rb_erase(&delayed_item->rb_node, root);
	delayed_item->delayed_node->count--;
	if (atomic_dec_return(&delayed_root->items) <
	    BTRFS_DELAYED_BACKGROUND &&
	    waitqueue_active(&delayed_root->wait))
		wake_up(&delayed_root->wait);
}

static void btrfs_release_delayed_item(struct btrfs_delayed_item *item)
{
	if (item) {
		__btrfs_remove_delayed_item(item);
		if (atomic_dec_and_test(&item->refs))
			kfree(item);
	}
}

struct btrfs_delayed_item *__btrfs_first_delayed_insertion_item(
					struct btrfs_delayed_node *delayed_node)
{
	struct rb_node *p;
	struct btrfs_delayed_item *item = NULL;

	p = rb_first(&delayed_node->ins_root);
	if (p)
		item = rb_entry(p, struct btrfs_delayed_item, rb_node);

	return item;
}

struct btrfs_delayed_item *__btrfs_first_delayed_deletion_item(
					struct btrfs_delayed_node *delayed_node)
{
	struct rb_node *p;
	struct btrfs_delayed_item *item = NULL;

	p = rb_first(&delayed_node->del_root);
	if (p)
		item = rb_entry(p, struct btrfs_delayed_item, rb_node);

	return item;
}

struct btrfs_delayed_item *__btrfs_next_delayed_item(
						struct btrfs_delayed_item *item)
{
	struct rb_node *p;
	struct btrfs_delayed_item *next = NULL;

	p = rb_next(&item->rb_node);
	if (p)
		next = rb_entry(p, struct btrfs_delayed_item, rb_node);

	return next;
}

static inline struct btrfs_root *btrfs_get_fs_root(struct btrfs_root *root,
						   u64 root_id)
{
	struct btrfs_key root_key;

	if (root->objectid == root_id)
		return root;

	root_key.objectid = root_id;
	root_key.type = BTRFS_ROOT_ITEM_KEY;
	root_key.offset = (u64)-1;
	return btrfs_read_fs_root_no_name(root->fs_info, &root_key);
}

static int btrfs_delayed_item_reserve_metadata(struct btrfs_trans_handle *trans,
					       struct btrfs_root *root,
					       struct btrfs_delayed_item *item)
{
	struct btrfs_block_rsv *src_rsv;
	struct btrfs_block_rsv *dst_rsv;
	u64 num_bytes;
	int ret;

	if (!trans->bytes_reserved)
		return 0;

	src_rsv = trans->block_rsv;
	dst_rsv = &root->fs_info->delayed_block_rsv;

	num_bytes = btrfs_calc_trans_metadata_size(root, 1);
	ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
	if (!ret) {
		trace_btrfs_space_reservation(root->fs_info, "delayed_item",
					      item->key.objectid,
					      num_bytes, 1);
		item->bytes_reserved = num_bytes;
	}

	return ret;
}

static void btrfs_delayed_item_release_metadata(struct btrfs_root *root,
						struct btrfs_delayed_item *item)
{
	struct btrfs_block_rsv *rsv;

	if (!item->bytes_reserved)
		return;

	rsv = &root->fs_info->delayed_block_rsv;
	trace_btrfs_space_reservation(root->fs_info, "delayed_item",
				      item->key.objectid, item->bytes_reserved,
				      0);
	btrfs_block_rsv_release(root, rsv,
				item->bytes_reserved);
}

static int btrfs_delayed_inode_reserve_metadata(
					struct btrfs_trans_handle *trans,
					struct btrfs_root *root,
					struct inode *inode,
					struct btrfs_delayed_node *node)
{
	struct btrfs_block_rsv *src_rsv;
	struct btrfs_block_rsv *dst_rsv;
	u64 num_bytes;
	int ret;
	bool release = false;

	src_rsv = trans->block_rsv;
	dst_rsv = &root->fs_info->delayed_block_rsv;

	num_bytes = btrfs_calc_trans_metadata_size(root, 1);

	/*
	 * btrfs_dirty_inode will update the inode under btrfs_join_transaction
	 * which doesn't reserve space for speed.  This is a problem since we
	 * still need to reserve space for this update, so try to reserve the
	 * space.
	 *
	 * Now if src_rsv == delalloc_block_rsv we'll let it just steal since
	 * we're accounted for.
	 */
	if (!src_rsv || (!trans->bytes_reserved &&
	    src_rsv != &root->fs_info->delalloc_block_rsv)) {
		ret = btrfs_block_rsv_add_noflush(root, dst_rsv, num_bytes);
		/*
		 * Since we're under a transaction reserve_metadata_bytes could
		 * try to commit the transaction which will make it return
		 * EAGAIN to make us stop the transaction we have, so return
		 * ENOSPC instead so that btrfs_dirty_inode knows what to do.
		 */
		if (ret == -EAGAIN)
			ret = -ENOSPC;
		if (!ret) {
			node->bytes_reserved = num_bytes;
			trace_btrfs_space_reservation(root->fs_info,
						      "delayed_inode",
						      btrfs_ino(inode),
						      num_bytes, 1);
		}
		return ret;
	} else if (src_rsv == &root->fs_info->delalloc_block_rsv) {
		spin_lock(&BTRFS_I(inode)->lock);
		if (test_and_clear_bit(BTRFS_INODE_DELALLOC_META_RESERVED,
				       &BTRFS_I(inode)->runtime_flags)) {
			spin_unlock(&BTRFS_I(inode)->lock);
			release = true;
			goto migrate;
		}
		spin_unlock(&BTRFS_I(inode)->lock);

		/* Ok we didn't have space pre-reserved.  This shouldn't happen
		 * too often but it can happen if we do delalloc to an existing
		 * inode which gets dirtied because of the time update, and then
		 * isn't touched again until after the transaction commits and
		 * then we try to write out the data.  First try to be nice and
		 * reserve something strictly for us.  If not be a pain and try
		 * to steal from the delalloc block rsv.
		 */
		ret = btrfs_block_rsv_add_noflush(root, dst_rsv, num_bytes);
		if (!ret)
			goto out;

		ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
		if (!ret)
			goto out;

		/*
		 * Ok this is a problem, let's just steal from the global rsv
		 * since this really shouldn't happen that often.
		 */
		WARN_ON(1);
		ret = btrfs_block_rsv_migrate(&root->fs_info->global_block_rsv,
					      dst_rsv, num_bytes);
		goto out;
	}

migrate:
	ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);

out:
	/*
	 * Migrate only takes a reservation, it doesn't touch the size of the
	 * block_rsv.  This is to simplify people who don't normally have things
	 * migrated from their block rsv.  If they go to release their
	 * reservation, that will decrease the size as well, so if migrate
	 * reduced size we'd end up with a negative size.  But for the
	 * delalloc_meta_reserved stuff we will only know to drop 1 reservation,
	 * but we could in fact do this reserve/migrate dance several times
	 * between the time we did the original reservation and we'd clean it
	 * up.  So to take care of this, release the space for the meta
	 * reservation here.  I think it may be time for a documentation page on
	 * how block rsvs. work.
	 */
	if (!ret) {
		trace_btrfs_space_reservation(root->fs_info, "delayed_inode",
					      btrfs_ino(inode), num_bytes, 1);
		node->bytes_reserved = num_bytes;
	}

	if (release) {
		trace_btrfs_space_reservation(root->fs_info, "delalloc",
					      btrfs_ino(inode), num_bytes, 0);
		btrfs_block_rsv_release(root, src_rsv, num_bytes);
	}

	return ret;
}

static void btrfs_delayed_inode_release_metadata(struct btrfs_root *root,
						struct btrfs_delayed_node *node)
{
	struct btrfs_block_rsv *rsv;

	if (!node->bytes_reserved)
		return;

	rsv = &root->fs_info->delayed_block_rsv;
	trace_btrfs_space_reservation(root->fs_info, "delayed_inode",
				      node->inode_id, node->bytes_reserved, 0);
	btrfs_block_rsv_release(root, rsv,
				node->bytes_reserved);
	node->bytes_reserved = 0;
}

/*
 * This helper will insert some continuous items into the same leaf according
 * to the free space of the leaf.
 */
static int btrfs_batch_insert_items(struct btrfs_trans_handle *trans,
				struct btrfs_root *root,
				struct btrfs_path *path,
				struct btrfs_delayed_item *item)
{
	struct btrfs_delayed_item *curr, *next;
	int free_space;
	int total_data_size = 0, total_size = 0;
	struct extent_buffer *leaf;
	char *data_ptr;
	struct btrfs_key *keys;
	u32 *data_size;
	struct list_head head;
	int slot;
	int nitems;
	int i;
	int ret = 0;

	BUG_ON(!path->nodes[0]);

	leaf = path->nodes[0];
	free_space = btrfs_leaf_free_space(root, leaf);
	INIT_LIST_HEAD(&head);

	next = item;
	nitems = 0;

	/*
	 * count the number of the continuous items that we can insert in batch
	 */
	while (total_size + next->data_len + sizeof(struct btrfs_item) <=
	       free_space) {
		total_data_size += next->data_len;
		total_size += next->data_len + sizeof(struct btrfs_item);
		list_add_tail(&next->tree_list, &head);
		nitems++;

		curr = next;
		next = __btrfs_next_delayed_item(curr);
		if (!next)
			break;

		if (!btrfs_is_continuous_delayed_item(curr, next))
			break;
	}

	if (!nitems) {
		ret = 0;
		goto out;
	}

	/*
	 * we need allocate some memory space, but it might cause the task
	 * to sleep, so we set all locked nodes in the path to blocking locks
	 * first.
	 */
	btrfs_set_path_blocking(path);

	keys = kmalloc(sizeof(struct btrfs_key) * nitems, GFP_NOFS);
	if (!keys) {
		ret = -ENOMEM;
		goto out;
	}

	data_size = kmalloc(sizeof(u32) * nitems, GFP_NOFS);
	if (!data_size) {
		ret = -ENOMEM;
		goto error;
	}

	/* get keys of all the delayed items */
	i = 0;
	list_for_each_entry(next, &head, tree_list) {
		keys[i] = next->key;
		data_size[i] = next->data_len;
		i++;
	}

	/* reset all the locked nodes in the patch to spinning locks. */
	btrfs_clear_path_blocking(path, NULL, 0);

	/* insert the keys of the items */
	setup_items_for_insert(trans, root, path, keys, data_size,
			       total_data_size, total_size, nitems);

	/* insert the dir index items */
	slot = path->slots[0];
	list_for_each_entry_safe(curr, next, &head, tree_list) {
		data_ptr = btrfs_item_ptr(leaf, slot, char);
		write_extent_buffer(leaf, &curr->data,
				    (unsigned long)data_ptr,
				    curr->data_len);
		slot++;

		btrfs_delayed_item_release_metadata(root, curr);

		list_del(&curr->tree_list);
		btrfs_release_delayed_item(curr);
	}

error:
	kfree(data_size);
	kfree(keys);
out:
	return ret;
}

/*
 * This helper can just do simple insertion that needn't extend item for new
 * data, such as directory name index insertion, inode insertion.
 */
static int btrfs_insert_delayed_item(struct btrfs_trans_handle *trans,
				     struct btrfs_root *root,
				     struct btrfs_path *path,
				     struct btrfs_delayed_item *delayed_item)
{
	struct extent_buffer *leaf;
	struct btrfs_item *item;
	char *ptr;
	int ret;

	ret = btrfs_insert_empty_item(trans, root, path, &delayed_item->key,
				      delayed_item->data_len);
	if (ret < 0 && ret != -EEXIST)
		return ret;

	leaf = path->nodes[0];

	item = btrfs_item_nr(leaf, path->slots[0]);
	ptr = btrfs_item_ptr(leaf, path->slots[0], char);

	write_extent_buffer(leaf, delayed_item->data, (unsigned long)ptr,
			    delayed_item->data_len);
	btrfs_mark_buffer_dirty(leaf);

	btrfs_delayed_item_release_metadata(root, delayed_item);
	return 0;
}

/*
 * we insert an item first, then if there are some continuous items, we try
 * to insert those items into the same leaf.
 */
static int btrfs_insert_delayed_items(struct btrfs_trans_handle *trans,
				      struct btrfs_path *path,
				      struct btrfs_root *root,
				      struct btrfs_delayed_node *node)
{
	struct btrfs_delayed_item *curr, *prev;
	int ret = 0;

do_again:
	mutex_lock(&node->mutex);
	curr = __btrfs_first_delayed_insertion_item(node);
	if (!curr)
		goto insert_end;

	ret = btrfs_insert_delayed_item(trans, root, path, curr);
	if (ret < 0) {
		btrfs_release_path(path);
		goto insert_end;
	}

	prev = curr;
	curr = __btrfs_next_delayed_item(prev);
	if (curr && btrfs_is_continuous_delayed_item(prev, curr)) {
		/* insert the continuous items into the same leaf */
		path->slots[0]++;
		btrfs_batch_insert_items(trans, root, path, curr);
	}
	btrfs_release_delayed_item(prev);
	btrfs_mark_buffer_dirty(path->nodes[0]);

	btrfs_release_path(path);
	mutex_unlock(&node->mutex);
	goto do_again;

insert_end:
	mutex_unlock(&node->mutex);
	return ret;
}

static int btrfs_batch_delete_items(struct btrfs_trans_handle *trans,
				    struct btrfs_root *root,
				    struct btrfs_path *path,
				    struct btrfs_delayed_item *item)
{
	struct btrfs_delayed_item *curr, *next;
	struct extent_buffer *leaf;
	struct btrfs_key key;
	struct list_head head;
	int nitems, i, last_item;
	int ret = 0;

	BUG_ON(!path->nodes[0]);

	leaf = path->nodes[0];

	i = path->slots[0];
	last_item = btrfs_header_nritems(leaf) - 1;
	if (i > last_item)
		return -ENOENT;	/* FIXME: Is errno suitable? */

	next = item;
	INIT_LIST_HEAD(&head);
	btrfs_item_key_to_cpu(leaf, &key, i);
	nitems = 0;
	/*
	 * count the number of the dir index items that we can delete in batch
	 */
	while (btrfs_comp_cpu_keys(&next->key, &key) == 0) {
		list_add_tail(&next->tree_list, &head);
		nitems++;

		curr = next;
		next = __btrfs_next_delayed_item(curr);
		if (!next)
			break;

		if (!btrfs_is_continuous_delayed_item(curr, next))
			break;

		i++;
		if (i > last_item)
			break;
		btrfs_item_key_to_cpu(leaf, &key, i);
	}

	if (!nitems)
		return 0;

	ret = btrfs_del_items(trans, root, path, path->slots[0], nitems);
	if (ret)
		goto out;

	list_for_each_entry_safe(curr, next, &head, tree_list) {
		btrfs_delayed_item_release_metadata(root, curr);
		list_del(&curr->tree_list);
		btrfs_release_delayed_item(curr);
	}

out: