extent-tree.c

/*
 * Copyright (C) 2007 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 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/sched.h>
#include <linux/pagemap.h>
#include <linux/writeback.h>
#include "hash.h"
#include "crc32c.h"
#include "ctree.h"
#include "disk-io.h"
#include "print-tree.h"
#include "transaction.h"
#include "volumes.h"
#include "locking.h"

#define BLOCK_GROUP_DATA     EXTENT_WRITEBACK
#define BLOCK_GROUP_METADATA EXTENT_UPTODATE
#define BLOCK_GROUP_SYSTEM   EXTENT_NEW

#define BLOCK_GROUP_DIRTY EXTENT_DIRTY

static int finish_current_insert(struct btrfs_trans_handle *trans, struct
				 btrfs_root *extent_root);
static int del_pending_extents(struct btrfs_trans_handle *trans, struct
			       btrfs_root *extent_root);
static struct btrfs_block_group_cache *
__btrfs_find_block_group(struct btrfs_root *root,
			 struct btrfs_block_group_cache *hint,
			 u64 search_start, int data, int owner);

void maybe_lock_mutex(struct btrfs_root *root)
{
	if (root != root->fs_info->extent_root &&
	    root != root->fs_info->chunk_root &&
	    root != root->fs_info->dev_root) {
		mutex_lock(&root->fs_info->alloc_mutex);
	}
}

void maybe_unlock_mutex(struct btrfs_root *root)
{
	if (root != root->fs_info->extent_root &&
	    root != root->fs_info->chunk_root &&
	    root != root->fs_info->dev_root) {
		mutex_unlock(&root->fs_info->alloc_mutex);
	}
}

static int cache_block_group(struct btrfs_root *root,
			     struct btrfs_block_group_cache *block_group)
{
	struct btrfs_path *path;
	int ret;
	struct btrfs_key key;
	struct extent_buffer *leaf;
	struct extent_io_tree *free_space_cache;
	int slot;
	u64 last = 0;
	u64 hole_size;
	u64 first_free;
	int found = 0;

	if (!block_group)
		return 0;

	root = root->fs_info->extent_root;
	free_space_cache = &root->fs_info->free_space_cache;

	if (block_group->cached)
		return 0;

	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

	path->reada = 2;
	/*
	 * we get into deadlocks with paths held by callers of this function.
	 * since the alloc_mutex is protecting things right now, just
	 * skip the locking here
	 */
	path->skip_locking = 1;
	first_free = block_group->key.objectid;
	key.objectid = block_group->key.objectid;
	key.offset = 0;
	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
	if (ret < 0)
		return ret;
	ret = btrfs_previous_item(root, path, 0, BTRFS_EXTENT_ITEM_KEY);
	if (ret < 0)
		return ret;
	if (ret == 0) {
		leaf = path->nodes[0];
		btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
		if (key.objectid + key.offset > first_free)
			first_free = key.objectid + key.offset;
	}
	while(1) {
		leaf = path->nodes[0];
		slot = path->slots[0];
		if (slot >= btrfs_header_nritems(leaf)) {
			ret = btrfs_next_leaf(root, path);
			if (ret < 0)
				goto err;
			if (ret == 0) {
				continue;
			} else {
				break;
			}
		}
		btrfs_item_key_to_cpu(leaf, &key, slot);
		if (key.objectid < block_group->key.objectid) {
			goto next;
		}
		if (key.objectid >= block_group->key.objectid +
		    block_group->key.offset) {
			break;
		}

		if (btrfs_key_type(&key) == BTRFS_EXTENT_ITEM_KEY) {
			if (!found) {
				last = first_free;
				found = 1;
			}
			if (key.objectid > last) {
				hole_size = key.objectid - last;
				set_extent_dirty(free_space_cache, last,
						 last + hole_size - 1,
						 GFP_NOFS);
			}
			last = key.objectid + key.offset;
		}
next:
		path->slots[0]++;
	}

	if (!found)
		last = first_free;
	if (block_group->key.objectid +
	    block_group->key.offset > last) {
		hole_size = block_group->key.objectid +
			block_group->key.offset - last;
		set_extent_dirty(free_space_cache, last,
				 last + hole_size - 1, GFP_NOFS);
	}
	block_group->cached = 1;
err:
	btrfs_free_path(path);
	return 0;
}

struct btrfs_block_group_cache *btrfs_lookup_first_block_group(struct
						       btrfs_fs_info *info,
							 u64 bytenr)
{
	struct extent_io_tree *block_group_cache;
	struct btrfs_block_group_cache *block_group = NULL;
	u64 ptr;
	u64 start;
	u64 end;
	int ret;

	bytenr = max_t(u64, bytenr,
		       BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE);
	block_group_cache = &info->block_group_cache;
	ret = find_first_extent_bit(block_group_cache,
				    bytenr, &start, &end,
				    BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA |
				    BLOCK_GROUP_SYSTEM);
	if (ret) {
		return NULL;
	}
	ret = get_state_private(block_group_cache, start, &ptr);
	if (ret)
		return NULL;

	block_group = (struct btrfs_block_group_cache *)(unsigned long)ptr;
	return block_group;
}

struct btrfs_block_group_cache *btrfs_lookup_block_group(struct
							 btrfs_fs_info *info,
							 u64 bytenr)
{
	struct extent_io_tree *block_group_cache;
	struct btrfs_block_group_cache *block_group = NULL;
	u64 ptr;
	u64 start;
	u64 end;
	int ret;

	bytenr = max_t(u64, bytenr,
		       BTRFS_SUPER_INFO_OFFSET + BTRFS_SUPER_INFO_SIZE);
	block_group_cache = &info->block_group_cache;
	ret = find_first_extent_bit(block_group_cache,
				    bytenr, &start, &end,
				    BLOCK_GROUP_DATA | BLOCK_GROUP_METADATA |
				    BLOCK_GROUP_SYSTEM);
	if (ret) {
		return NULL;
	}
	ret = get_state_private(block_group_cache, start, &ptr);
	if (ret)
		return NULL;

	block_group = (struct btrfs_block_group_cache *)(unsigned long)ptr;
	if (block_group->key.objectid <= bytenr && bytenr <
	    block_group->key.objectid + block_group->key.offset)
		return block_group;
	return NULL;
}

static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
{
	return (cache->flags & bits) == bits;
}

static int noinline find_search_start(struct btrfs_root *root,
			      struct btrfs_block_group_cache **cache_ret,
			      u64 *start_ret, u64 num, int data)
{
	int ret;
	struct btrfs_block_group_cache *cache = *cache_ret;
	struct extent_io_tree *free_space_cache;
	struct extent_state *state;
	u64 last;
	u64 start = 0;
	u64 cache_miss = 0;
	u64 total_fs_bytes;
	u64 search_start = *start_ret;
	int wrapped = 0;

	WARN_ON(!mutex_is_locked(&root->fs_info->alloc_mutex));
	total_fs_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
	free_space_cache = &root->fs_info->free_space_cache;

	if (!cache)
		goto out;

again:
	ret = cache_block_group(root, cache);
	if (ret) {
		goto out;
	}

	last = max(search_start, cache->key.objectid);
	if (!block_group_bits(cache, data) || cache->ro)
		goto new_group;

	spin_lock_irq(&free_space_cache->lock);
	state = find_first_extent_bit_state(free_space_cache, last, EXTENT_DIRTY);
	while(1) {
		if (!state) {
			if (!cache_miss)
				cache_miss = last;
			spin_unlock_irq(&free_space_cache->lock);
			goto new_group;
		}

		start = max(last, state->start);
		last = state->end + 1;
		if (last - start < num) {
			do {
				state = extent_state_next(state);
			} while(state && !(state->state & EXTENT_DIRTY));
			continue;
		}
		spin_unlock_irq(&free_space_cache->lock);
		if (cache->ro) {
			goto new_group;
		}
		if (start + num > cache->key.objectid + cache->key.offset)
			goto new_group;
		if (!block_group_bits(cache, data)) {
			printk("block group bits don't match %Lu %d\n", cache->flags, data);
		}
		*start_ret = start;
		return 0;
	}
out:
	cache = btrfs_lookup_block_group(root->fs_info, search_start);
	if (!cache) {
		printk("Unable to find block group for %Lu\n", search_start);
		WARN_ON(1);
	}
	return -ENOSPC;

new_group:
	last = cache->key.objectid + cache->key.offset;
wrapped:
	cache = btrfs_lookup_first_block_group(root->fs_info, last);
	if (!cache || cache->key.objectid >= total_fs_bytes) {
no_cache:
		if (!wrapped) {
			wrapped = 1;
			last = search_start;
			goto wrapped;
		}
		goto out;
	}
	if (cache_miss && !cache->cached) {
		cache_block_group(root, cache);
		last = cache_miss;
		cache = btrfs_lookup_first_block_group(root->fs_info, last);
	}
	cache_miss = 0;
	cache = __btrfs_find_block_group(root, cache, last, data, 0);
	if (!cache)
		goto no_cache;
	*cache_ret = cache;
	goto again;
}

static u64 div_factor(u64 num, int factor)
{
	if (factor == 10)
		return num;
	num *= factor;
	do_div(num, 10);
	return num;
}

static int block_group_state_bits(u64 flags)
{
	int bits = 0;
	if (flags & BTRFS_BLOCK_GROUP_DATA)
		bits |= BLOCK_GROUP_DATA;
	if (flags & BTRFS_BLOCK_GROUP_METADATA)
		bits |= BLOCK_GROUP_METADATA;
	if (flags & BTRFS_BLOCK_GROUP_SYSTEM)
		bits |= BLOCK_GROUP_SYSTEM;
	return bits;
}

static struct btrfs_block_group_cache *
__btrfs_find_block_group(struct btrfs_root *root,
			 struct btrfs_block_group_cache *hint,
			 u64 search_start, int data, int owner)
{
	struct btrfs_block_group_cache *cache;
	struct extent_io_tree *block_group_cache;
	struct btrfs_block_group_cache *found_group = NULL;
	struct btrfs_fs_info *info = root->fs_info;
	u64 used;
	u64 last = 0;
	u64 start;
	u64 end;
	u64 free_check;
	u64 ptr;
	int bit;
	int ret;
	int full_search = 0;
	int factor = 10;
	int wrapped = 0;

	block_group_cache = &info->block_group_cache;

	if (data & BTRFS_BLOCK_GROUP_METADATA)
		factor = 9;

	bit = block_group_state_bits(data);

	if (search_start) {
		struct btrfs_block_group_cache *shint;
		shint = btrfs_lookup_first_block_group(info, search_start);
		if (shint && block_group_bits(shint, data) && !shint->ro) {
			used = btrfs_block_group_used(&shint->item);
			if (used + shint->pinned <
			    div_factor(shint->key.offset, factor)) {
				return shint;
			}
		}
	}
	if (hint && !hint->ro && block_group_bits(hint, data)) {
		used = btrfs_block_group_used(&hint->item);
		if (used + hint->pinned <
		    div_factor(hint->key.offset, factor)) {
			return hint;
		}
		last = hint->key.objectid + hint->key.offset;
	} else {
		if (hint)
			last = max(hint->key.objectid, search_start);
		else
			last = search_start;
	}
again:
	while(1) {
		ret = find_first_extent_bit(block_group_cache, last,
					    &start, &end, bit);
		if (ret)
			break;

		ret = get_state_private(block_group_cache, start, &ptr);
		if (ret) {
			last = end + 1;
			continue;
		}

		cache = (struct btrfs_block_group_cache *)(unsigned long)ptr;
		last = cache->key.objectid + cache->key.offset;
		used = btrfs_block_group_used(&cache->item);

		if (!cache->ro && block_group_bits(cache, data)) {
			free_check = div_factor(cache->key.offset, factor);
			if (used + cache->pinned < free_check) {
				found_group = cache;
				goto found;
			}
		}
		cond_resched();
	}
	if (!wrapped) {
		last = search_start;
		wrapped = 1;
		goto again;
	}
	if (!full_search && factor < 10) {
		last = search_start;
		full_search = 1;
		factor = 10;
		goto again;
	}
found:
	return found_group;
}

struct btrfs_block_group_cache *btrfs_find_block_group(struct btrfs_root *root,
						 struct btrfs_block_group_cache
						 *hint, u64 search_start,
						 int data, int owner)
{

	struct btrfs_block_group_cache *ret;
	mutex_lock(&root->fs_info->alloc_mutex);
	ret = __btrfs_find_block_group(root, hint, search_start, data, owner);
	mutex_unlock(&root->fs_info->alloc_mutex);
	return ret;
}
static u64 hash_extent_ref(u64 root_objectid, u64 ref_generation,
			   u64 owner, u64 owner_offset)
{
	u32 high_crc = ~(u32)0;
	u32 low_crc = ~(u32)0;
	__le64 lenum;
	lenum = cpu_to_le64(root_objectid);
	high_crc = btrfs_crc32c(high_crc, &lenum, sizeof(lenum));
	lenum = cpu_to_le64(ref_generation);
	low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
	if (owner >= BTRFS_FIRST_FREE_OBJECTID) {
		lenum = cpu_to_le64(owner);
		low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
		lenum = cpu_to_le64(owner_offset);
		low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
	}
	return ((u64)high_crc << 32) | (u64)low_crc;
}

static int match_extent_ref(struct extent_buffer *leaf,
			    struct btrfs_extent_ref *disk_ref,
			    struct btrfs_extent_ref *cpu_ref)
{
	int ret;
	int len;

	if (cpu_ref->objectid)
		len = sizeof(*cpu_ref);
	else
		len = 2 * sizeof(u64);
	ret = memcmp_extent_buffer(leaf, cpu_ref, (unsigned long)disk_ref,
				   len);
	return ret == 0;
}

static int noinline lookup_extent_backref(struct btrfs_trans_handle *trans,
					  struct btrfs_root *root,
					  struct btrfs_path *path, u64 bytenr,
					  u64 root_objectid,
					  u64 ref_generation, u64 owner,
					  u64 owner_offset, int del)
{
	u64 hash;
	struct btrfs_key key;
	struct btrfs_key found_key;
	struct btrfs_extent_ref ref;
	struct extent_buffer *leaf;
	struct btrfs_extent_ref *disk_ref;
	int ret;
	int ret2;

	btrfs_set_stack_ref_root(&ref, root_objectid);
	btrfs_set_stack_ref_generation(&ref, ref_generation);
	btrfs_set_stack_ref_objectid(&ref, owner);
	btrfs_set_stack_ref_offset(&ref, owner_offset);

	hash = hash_extent_ref(root_objectid, ref_generation, owner,
			       owner_offset);
	key.offset = hash;
	key.objectid = bytenr;
	key.type = BTRFS_EXTENT_REF_KEY;

	while (1) {
		ret = btrfs_search_slot(trans, root, &key, path,
					del ? -1 : 0, del);
		if (ret < 0)
			goto out;
		leaf = path->nodes[0];
		if (ret != 0) {
			u32 nritems = btrfs_header_nritems(leaf);
			if (path->slots[0] >= nritems) {
				ret2 = btrfs_next_leaf(root, path);
				if (ret2)
					goto out;
				leaf = path->nodes[0];
			}
			btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
			if (found_key.objectid != bytenr ||
			    found_key.type != BTRFS_EXTENT_REF_KEY)
				goto out;
			key.offset = found_key.offset;
			if (del) {
				btrfs_release_path(root, path);
				continue;
			}
		}
		disk_ref = btrfs_item_ptr(path->nodes[0],
					  path->slots[0],
					  struct btrfs_extent_ref);
		if (match_extent_ref(path->nodes[0], disk_ref, &ref)) {
			ret = 0;
			goto out;
		}
		btrfs_item_key_to_cpu(leaf, &found_key, path->slots[0]);
		key.offset = found_key.offset + 1;
		btrfs_release_path(root, path);
	}
out:
	return ret;
}

/*
 * Back reference rules.  Back refs have three main goals:
 *
 * 1) differentiate between all holders of references to an extent so that
 *    when a reference is dropped we can make sure it was a valid reference
 *    before freeing the extent.
 *
 * 2) Provide enough information to quickly find the holders of an extent
 *    if we notice a given block is corrupted or bad.
 *
 * 3) Make it easy to migrate blocks for FS shrinking or storage pool
 *    maintenance.  This is actually the same as #2, but with a slightly
 *    different use case.
 *
 * File extents can be referenced by:
 *
 * - multiple snapshots, subvolumes, or different generations in one subvol
 * - different files inside a single subvolume (in theory, not implemented yet)
 * - different offsets inside a file (bookend extents in file.c)
 *
 * The extent ref structure has fields for:
 *
 * - Objectid of the subvolume root
 * - Generation number of the tree holding the reference
 * - objectid of the file holding the reference
 * - offset in the file corresponding to the key holding the reference
 *
 * When a file extent is allocated the fields are filled in:
 *     (root_key.objectid, trans->transid, inode objectid, offset in file)
 *
 * When a leaf is cow'd new references are added for every file extent found
 * in the leaf.  It looks the same as the create case, but trans->transid
 * will be different when the block is cow'd.
 *
 *     (root_key.objectid, trans->transid, inode objectid, offset in file)
 *
 * When a file extent is removed either during snapshot deletion or file
 * truncation, the corresponding back reference is found
 * by searching for:
 *
 *     (btrfs_header_owner(leaf), btrfs_header_generation(leaf),
 *      inode objectid, offset in file)
 *
 * Btree extents can be referenced by:
 *
 * - Different subvolumes
 * - Different generations of the same subvolume
 *
 * Storing sufficient information for a full reverse mapping of a btree
 * block would require storing the lowest key of the block in the backref,
 * and it would require updating that lowest key either before write out or
 * every time it changed.  Instead, the objectid of the lowest key is stored
 * along with the level of the tree block.  This provides a hint
 * about where in the btree the block can be found.  Searches through the
 * btree only need to look for a pointer to that block, so they stop one
 * level higher than the level recorded in the backref.
 *
 * Some btrees do not do reference counting on their extents.  These
 * include the extent tree and the tree of tree roots.  Backrefs for these
 * trees always have a generation of zero.
 *
 * When a tree block is created, back references are inserted:
 *
 * (root->root_key.objectid, trans->transid or zero, level, lowest_key_objectid)
 *
 * When a tree block is cow'd in a reference counted root,
 * new back references are added for all the blocks it points to.
 * These are of the form (trans->transid will have increased since creation):
 *
 * (root->root_key.objectid, trans->transid, level, lowest_key_objectid)
 *
 * Because the lowest_key_objectid and the level are just hints
 * they are not used when backrefs are deleted.  When a backref is deleted:
 *
 * if backref was for a tree root:
 *     root_objectid = root->root_key.objectid
 * else
 *     root_objectid = btrfs_header_owner(parent)
 *
 * (root_objectid, btrfs_header_generation(parent) or zero, 0, 0)
 *
 * Back Reference Key hashing:
 *
 * Back references have four fields, each 64 bits long.  Unfortunately,
 * This is hashed into a single 64 bit number and placed into the key offset.
 * The key objectid corresponds to the first byte in the extent, and the
 * key type is set to BTRFS_EXTENT_REF_KEY
 */
int btrfs_insert_extent_backref(struct btrfs_trans_handle *trans,
				 struct btrfs_root *root,
				 struct btrfs_path *path, u64 bytenr,
				 u64 root_objectid, u64 ref_generation,
				 u64 owner, u64 owner_offset)
{
	u64 hash;
	struct btrfs_key key;
	struct btrfs_extent_ref ref;
	struct btrfs_extent_ref *disk_ref;
	int ret;

	btrfs_set_stack_ref_root(&ref, root_objectid);
	btrfs_set_stack_ref_generation(&ref, ref_generation);
	btrfs_set_stack_ref_objectid(&ref, owner);
	btrfs_set_stack_ref_offset(&ref, owner_offset);

	hash = hash_extent_ref(root_objectid, ref_generation, owner,
			       owner_offset);
	key.offset = hash;
	key.objectid = bytenr;
	key.type = BTRFS_EXTENT_REF_KEY;

	ret = btrfs_insert_empty_item(trans, root, path, &key, sizeof(ref));
	while (ret == -EEXIST) {
		disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
					  struct btrfs_extent_ref);
		if (match_extent_ref(path->nodes[0], disk_ref, &ref))
			goto out;
		key.offset++;
		btrfs_release_path(root, path);
		ret = btrfs_insert_empty_item(trans, root, path, &key,
					      sizeof(ref));
	}
	if (ret)
		goto out;
	disk_ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
				  struct btrfs_extent_ref);
	write_extent_buffer(path->nodes[0], &ref, (unsigned long)disk_ref,
			    sizeof(ref));
	btrfs_mark_buffer_dirty(path->nodes[0]);
out:
	btrfs_release_path(root, path);
	return ret;
}

static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
				struct btrfs_root *root,
				u64 bytenr, u64 num_bytes,
				u64 root_objectid, u64 ref_generation,
				u64 owner, u64 owner_offset)
{
	struct btrfs_path *path;
	int ret;
	struct btrfs_key key;
	struct extent_buffer *l;
	struct btrfs_extent_item *item;
	u32 refs;

	WARN_ON(num_bytes < root->sectorsize);
	path = btrfs_alloc_path();
	if (!path)
		return -ENOMEM;

	path->reada = 1;
	key.objectid = bytenr;
	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
	key.offset = num_bytes;
	ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
				0, 1);
	if (ret < 0)
		return ret;
	if (ret != 0) {
		BUG();
	}
	BUG_ON(ret != 0);
	l = path->nodes[0];
	item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
	refs = btrfs_extent_refs(l, item);
	btrfs_set_extent_refs(l, item, refs + 1);
	btrfs_mark_buffer_dirty(path->nodes[0]);

	btrfs_release_path(root->fs_info->extent_root, path);

	path->reada = 1;
	ret = btrfs_insert_extent_backref(trans, root->fs_info->extent_root,
					  path, bytenr, root_objectid,
					  ref_generation, owner, owner_offset);
	BUG_ON(ret);
	finish_current_insert(trans, root->fs_info->extent_root);
	del_pending_extents(trans, root->fs_info->extent_root);

	btrfs_free_path(path);
	return 0;
}

int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
				struct btrfs_root *root,
				u64 bytenr, u64 num_bytes,
				u64 root_objectid, u64 ref_generation,
				u64 owner, u64 owner_offset)
{
	int ret;

	mutex_lock(&root->fs_info->alloc_mutex);
	ret = __btrfs_inc_extent_ref(trans, root, bytenr, num_bytes,
				     root_objectid, ref_generation,
				     owner, owner_offset);
	mutex_unlock(&root->fs_info->alloc_mutex);
	return ret;
}

int btrfs_extent_post_op(struct btrfs_trans_handle *trans,
			 struct btrfs_root *root)
{
	finish_current_insert(trans, root->fs_info->extent_root);
	del_pending_extents(trans, root->fs_info->extent_root);
	return 0;
}

static int lookup_extent_ref(struct btrfs_trans_handle *trans,
			     struct btrfs_root *root, u64 bytenr,
			     u64 num_bytes, u32 *refs)
{
	struct btrfs_path *path;
	int ret;
	struct btrfs_key key;
	struct extent_buffer *l;
	struct btrfs_extent_item *item;

	WARN_ON(num_bytes < root->sectorsize);
	path = btrfs_alloc_path();
	path->reada = 1;
	key.objectid = bytenr;
	key.offset = num_bytes;
	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
	ret = btrfs_search_slot(trans, root->fs_info->extent_root, &key, path,
				0, 0);
	if (ret < 0)
		goto out;
	if (ret != 0) {
		btrfs_print_leaf(root, path->nodes[0]);
		printk("failed to find block number %Lu\n", bytenr);
		BUG();
	}
	l = path->nodes[0];
	item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
	*refs = btrfs_extent_refs(l, item);
out:
	btrfs_free_path(path);
	return 0;
}

u32 btrfs_count_snapshots_in_path(struct btrfs_root *root,
				  struct btrfs_path *count_path,
				  u64 expected_owner,
				  u64 first_extent)
{
	struct btrfs_root *extent_root = root->fs_info->extent_root;
	struct btrfs_path *path;
	u64 bytenr;
	u64 found_objectid;
	u64 found_owner;
	u64 root_objectid = root->root_key.objectid;
	u32 total_count = 0;
	u32 extent_refs;
	u32 cur_count;
	u32 nritems;
	int ret;
	struct btrfs_key key;
	struct btrfs_key found_key;
	struct extent_buffer *l;
	struct btrfs_extent_item *item;
	struct btrfs_extent_ref *ref_item;
	int level = -1;

	/* FIXME, needs locking */
	BUG();

	mutex_lock(&root->fs_info->alloc_mutex);
	path = btrfs_alloc_path();
again:
	if (level == -1)
		bytenr = first_extent;
	else
		bytenr = count_path->nodes[level]->start;

	cur_count = 0;
	key.objectid = bytenr;
	key.offset = 0;

	btrfs_set_key_type(&key, BTRFS_EXTENT_ITEM_KEY);
	ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
	if (ret < 0)
		goto out;
	BUG_ON(ret == 0);

	l = path->nodes[0];
	btrfs_item_key_to_cpu(l, &found_key, path->slots[0]);

	if (found_key.objectid != bytenr ||
	    found_key.type != BTRFS_EXTENT_ITEM_KEY) {
		goto out;
	}

	item = btrfs_item_ptr(l, path->slots[0], struct btrfs_extent_item);
	extent_refs = btrfs_extent_refs(l, item);
	while (1) {
		l = path->nodes[0];
		nritems = btrfs_header_nritems(l);
		if (path->slots[0] >= nritems) {
			ret = btrfs_next_leaf(extent_root, path);
			if (ret == 0)
				continue;
			break;
		}
		btrfs_item_key_to_cpu(l, &found_key, path->slots[0]);
		if (found_key.objectid != bytenr)
			break;

		if (found_key.type != BTRFS_EXTENT_REF_KEY) {
			path->slots[0]++;
			continue;
		}

		cur_count++;
		ref_item = btrfs_item_ptr(l, path->slots[0],
					  struct btrfs_extent_ref);
		found_objectid = btrfs_ref_root(l, ref_item);

		if (found_objectid != root_objectid) {
			total_count = 2;
			goto out;
		}
		if (level == -1) {
			found_owner = btrfs_ref_objectid(l, ref_item);
			if (found_owner != expected_owner) {
				total_count = 2;
				goto out;
			}
			/*
			 * nasty.  we don't count a reference held by
			 * the running transaction.  This allows nodatacow
			 * to avoid cow most of the time
			 */
			if (found_owner >= BTRFS_FIRST_FREE_OBJECTID &&
			    btrfs_ref_generation(l, ref_item) ==
			    root->fs_info->generation) {
				extent_refs--;
			}
		}
		total_count = 1;
		path->slots[0]++;
	}
	/*
	 * if there is more than one reference against a data extent,
	 * we have to assume the other ref is another snapshot
	 */
	if (level == -1 && extent_refs > 1) {
		total_count = 2;
		goto out;
	}
	if (cur_count == 0) {
		total_count = 0;
		goto out;
	}
	if (level >= 0 && root->node == count_path->nodes[level])
		goto out;
	level++;
	btrfs_release_path(root, path);
	goto again;

out:
	btrfs_free_path(path);
	mutex_unlock(&root->fs_info->alloc_mutex);
	return total_count;
}

int btrfs_inc_ref(struct btrfs_trans_handle *trans, struct btrfs_root *root,
		  struct extent_buffer *buf)
{
	u64 bytenr;
	u32 nritems;
	struct btrfs_key key;
	struct btrfs_file_extent_item *fi;
	int i;
	int level;
	int ret;
	int faili;

	if (!root->ref_cows)
		return 0;

	level = btrfs_header_level(buf);
	nritems = btrfs_header_nritems(buf);
	for (i = 0; i < nritems; i++) {
		cond_resched();
		if (level == 0) {
			u64 disk_bytenr;
			btrfs_item_key_to_cpu(buf, &key, i);
			if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
				continue;
			fi = btrfs_item_ptr(buf, i,
					    struct btrfs_file_extent_item);
			if (btrfs_file_extent_type(buf, fi) ==
			    BTRFS_FILE_EXTENT_INLINE)
				continue;
			disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
			if (disk_bytenr == 0)
				continue;

			mutex_lock(&root->fs_info->alloc_mutex);
			ret = __btrfs_inc_extent_ref(trans, root, disk_bytenr,
				    btrfs_file_extent_disk_num_bytes(buf, fi),
				    root->root_key.objectid, trans->transid,
				    key.objectid, key.offset);
			mutex_unlock(&root->fs_info->alloc_mutex);
			if (ret) {
				faili = i;
				WARN_ON(1);
				goto fail;
			}
		} else {
			bytenr = btrfs_node_blockptr(buf, i);
			btrfs_node_key_to_cpu(buf, &key, i);

			mutex_lock(&root->fs_info->alloc_mutex);
			ret = __btrfs_inc_extent_ref(trans, root, bytenr,
					   btrfs_level_size(root, level - 1),
					   root->root_key.objectid,
					   trans->transid,
					   level - 1, key.objectid);
			mutex_unlock(&root->fs_info->alloc_mutex);
			if (ret) {
				faili = i;
				WARN_ON(1);
				goto fail;
			}
		}
	}
	return 0;
fail:
	WARN_ON(1);
#if 0
	for (i =0; i < faili; i++) {
		if (level == 0) {
			u64 disk_bytenr;
			btrfs_item_key_to_cpu(buf, &key, i);
			if (btrfs_key_type(&key) != BTRFS_EXTENT_DATA_KEY)
				continue;
			fi = btrfs_item_ptr(buf, i,
					    struct btrfs_file_extent_item);
			if (btrfs_file_extent_type(buf, fi) ==
			    BTRFS_FILE_EXTENT_INLINE)
				continue;
			disk_bytenr = btrfs_file_extent_disk_bytenr(buf, fi);
			if (disk_bytenr == 0)