blob: e2d7e86b51d1b24059dda46fda085174ac993a3f [file] [log] [blame]
/*
* 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/sched/signal.h>
#include <linux/pagemap.h>
#include <linux/writeback.h>
#include <linux/blkdev.h>
#include <linux/sort.h>
#include <linux/rcupdate.h>
#include <linux/kthread.h>
#include <linux/slab.h>
#include <linux/ratelimit.h>
#include <linux/percpu_counter.h>
#include "hash.h"
#include "tree-log.h"
#include "disk-io.h"
#include "print-tree.h"
#include "volumes.h"
#include "raid56.h"
#include "locking.h"
#include "free-space-cache.h"
#include "free-space-tree.h"
#include "math.h"
#include "sysfs.h"
#include "qgroup.h"
#undef SCRAMBLE_DELAYED_REFS
/*
* control flags for do_chunk_alloc's force field
* CHUNK_ALLOC_NO_FORCE means to only allocate a chunk
* if we really need one.
*
* CHUNK_ALLOC_LIMITED means to only try and allocate one
* if we have very few chunks already allocated. This is
* used as part of the clustering code to help make sure
* we have a good pool of storage to cluster in, without
* filling the FS with empty chunks
*
* CHUNK_ALLOC_FORCE means it must try to allocate one
*
*/
enum {
CHUNK_ALLOC_NO_FORCE = 0,
CHUNK_ALLOC_LIMITED = 1,
CHUNK_ALLOC_FORCE = 2,
};
static int update_block_group(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info, u64 bytenr,
u64 num_bytes, int alloc);
static int __btrfs_free_extent(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_delayed_ref_node *node, u64 parent,
u64 root_objectid, u64 owner_objectid,
u64 owner_offset, int refs_to_drop,
struct btrfs_delayed_extent_op *extra_op);
static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
struct extent_buffer *leaf,
struct btrfs_extent_item *ei);
static int alloc_reserved_file_extent(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
u64 parent, u64 root_objectid,
u64 flags, u64 owner, u64 offset,
struct btrfs_key *ins, int ref_mod);
static int alloc_reserved_tree_block(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
u64 parent, u64 root_objectid,
u64 flags, struct btrfs_disk_key *key,
int level, struct btrfs_key *ins);
static int do_chunk_alloc(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info, u64 flags,
int force);
static int find_next_key(struct btrfs_path *path, int level,
struct btrfs_key *key);
static void dump_space_info(struct btrfs_fs_info *fs_info,
struct btrfs_space_info *info, u64 bytes,
int dump_block_groups);
static int btrfs_add_reserved_bytes(struct btrfs_block_group_cache *cache,
u64 ram_bytes, u64 num_bytes, int delalloc);
static int btrfs_free_reserved_bytes(struct btrfs_block_group_cache *cache,
u64 num_bytes, int delalloc);
static int block_rsv_use_bytes(struct btrfs_block_rsv *block_rsv,
u64 num_bytes);
static int __reserve_metadata_bytes(struct btrfs_fs_info *fs_info,
struct btrfs_space_info *space_info,
u64 orig_bytes,
enum btrfs_reserve_flush_enum flush,
bool system_chunk);
static void space_info_add_new_bytes(struct btrfs_fs_info *fs_info,
struct btrfs_space_info *space_info,
u64 num_bytes);
static void space_info_add_old_bytes(struct btrfs_fs_info *fs_info,
struct btrfs_space_info *space_info,
u64 num_bytes);
static noinline int
block_group_cache_done(struct btrfs_block_group_cache *cache)
{
smp_mb();
return cache->cached == BTRFS_CACHE_FINISHED ||
cache->cached == BTRFS_CACHE_ERROR;
}
static int block_group_bits(struct btrfs_block_group_cache *cache, u64 bits)
{
return (cache->flags & bits) == bits;
}
void btrfs_get_block_group(struct btrfs_block_group_cache *cache)
{
atomic_inc(&cache->count);
}
void btrfs_put_block_group(struct btrfs_block_group_cache *cache)
{
if (atomic_dec_and_test(&cache->count)) {
WARN_ON(cache->pinned > 0);
WARN_ON(cache->reserved > 0);
/*
* If not empty, someone is still holding mutex of
* full_stripe_lock, which can only be released by caller.
* And it will definitely cause use-after-free when caller
* tries to release full stripe lock.
*
* No better way to resolve, but only to warn.
*/
WARN_ON(!RB_EMPTY_ROOT(&cache->full_stripe_locks_root.root));
kfree(cache->free_space_ctl);
kfree(cache);
}
}
/*
* this adds the block group to the fs_info rb tree for the block group
* cache
*/
static int btrfs_add_block_group_cache(struct btrfs_fs_info *info,
struct btrfs_block_group_cache *block_group)
{
struct rb_node **p;
struct rb_node *parent = NULL;
struct btrfs_block_group_cache *cache;
spin_lock(&info->block_group_cache_lock);
p = &info->block_group_cache_tree.rb_node;
while (*p) {
parent = *p;
cache = rb_entry(parent, struct btrfs_block_group_cache,
cache_node);
if (block_group->key.objectid < cache->key.objectid) {
p = &(*p)->rb_left;
} else if (block_group->key.objectid > cache->key.objectid) {
p = &(*p)->rb_right;
} else {
spin_unlock(&info->block_group_cache_lock);
return -EEXIST;
}
}
rb_link_node(&block_group->cache_node, parent, p);
rb_insert_color(&block_group->cache_node,
&info->block_group_cache_tree);
if (info->first_logical_byte > block_group->key.objectid)
info->first_logical_byte = block_group->key.objectid;
spin_unlock(&info->block_group_cache_lock);
return 0;
}
/*
* This will return the block group at or after bytenr if contains is 0, else
* it will return the block group that contains the bytenr
*/
static struct btrfs_block_group_cache *
block_group_cache_tree_search(struct btrfs_fs_info *info, u64 bytenr,
int contains)
{
struct btrfs_block_group_cache *cache, *ret = NULL;
struct rb_node *n;
u64 end, start;
spin_lock(&info->block_group_cache_lock);
n = info->block_group_cache_tree.rb_node;
while (n) {
cache = rb_entry(n, struct btrfs_block_group_cache,
cache_node);
end = cache->key.objectid + cache->key.offset - 1;
start = cache->key.objectid;
if (bytenr < start) {
if (!contains && (!ret || start < ret->key.objectid))
ret = cache;
n = n->rb_left;
} else if (bytenr > start) {
if (contains && bytenr <= end) {
ret = cache;
break;
}
n = n->rb_right;
} else {
ret = cache;
break;
}
}
if (ret) {
btrfs_get_block_group(ret);
if (bytenr == 0 && info->first_logical_byte > ret->key.objectid)
info->first_logical_byte = ret->key.objectid;
}
spin_unlock(&info->block_group_cache_lock);
return ret;
}
static int add_excluded_extent(struct btrfs_fs_info *fs_info,
u64 start, u64 num_bytes)
{
u64 end = start + num_bytes - 1;
set_extent_bits(&fs_info->freed_extents[0],
start, end, EXTENT_UPTODATE);
set_extent_bits(&fs_info->freed_extents[1],
start, end, EXTENT_UPTODATE);
return 0;
}
static void free_excluded_extents(struct btrfs_fs_info *fs_info,
struct btrfs_block_group_cache *cache)
{
u64 start, end;
start = cache->key.objectid;
end = start + cache->key.offset - 1;
clear_extent_bits(&fs_info->freed_extents[0],
start, end, EXTENT_UPTODATE);
clear_extent_bits(&fs_info->freed_extents[1],
start, end, EXTENT_UPTODATE);
}
static int exclude_super_stripes(struct btrfs_fs_info *fs_info,
struct btrfs_block_group_cache *cache)
{
u64 bytenr;
u64 *logical;
int stripe_len;
int i, nr, ret;
if (cache->key.objectid < BTRFS_SUPER_INFO_OFFSET) {
stripe_len = BTRFS_SUPER_INFO_OFFSET - cache->key.objectid;
cache->bytes_super += stripe_len;
ret = add_excluded_extent(fs_info, cache->key.objectid,
stripe_len);
if (ret)
return ret;
}
for (i = 0; i < BTRFS_SUPER_MIRROR_MAX; i++) {
bytenr = btrfs_sb_offset(i);
ret = btrfs_rmap_block(fs_info, cache->key.objectid,
bytenr, 0, &logical, &nr, &stripe_len);
if (ret)
return ret;
while (nr--) {
u64 start, len;
if (logical[nr] > cache->key.objectid +
cache->key.offset)
continue;
if (logical[nr] + stripe_len <= cache->key.objectid)
continue;
start = logical[nr];
if (start < cache->key.objectid) {
start = cache->key.objectid;
len = (logical[nr] + stripe_len) - start;
} else {
len = min_t(u64, stripe_len,
cache->key.objectid +
cache->key.offset - start);
}
cache->bytes_super += len;
ret = add_excluded_extent(fs_info, start, len);
if (ret) {
kfree(logical);
return ret;
}
}
kfree(logical);
}
return 0;
}
static struct btrfs_caching_control *
get_caching_control(struct btrfs_block_group_cache *cache)
{
struct btrfs_caching_control *ctl;
spin_lock(&cache->lock);
if (!cache->caching_ctl) {
spin_unlock(&cache->lock);
return NULL;
}
ctl = cache->caching_ctl;
refcount_inc(&ctl->count);
spin_unlock(&cache->lock);
return ctl;
}
static void put_caching_control(struct btrfs_caching_control *ctl)
{
if (refcount_dec_and_test(&ctl->count))
kfree(ctl);
}
#ifdef CONFIG_BTRFS_DEBUG
static void fragment_free_space(struct btrfs_block_group_cache *block_group)
{
struct btrfs_fs_info *fs_info = block_group->fs_info;
u64 start = block_group->key.objectid;
u64 len = block_group->key.offset;
u64 chunk = block_group->flags & BTRFS_BLOCK_GROUP_METADATA ?
fs_info->nodesize : fs_info->sectorsize;
u64 step = chunk << 1;
while (len > chunk) {
btrfs_remove_free_space(block_group, start, chunk);
start += step;
if (len < step)
len = 0;
else
len -= step;
}
}
#endif
/*
* this is only called by cache_block_group, since we could have freed extents
* we need to check the pinned_extents for any extents that can't be used yet
* since their free space will be released as soon as the transaction commits.
*/
u64 add_new_free_space(struct btrfs_block_group_cache *block_group,
struct btrfs_fs_info *info, u64 start, u64 end)
{
u64 extent_start, extent_end, size, total_added = 0;
int ret;
while (start < end) {
ret = find_first_extent_bit(info->pinned_extents, start,
&extent_start, &extent_end,
EXTENT_DIRTY | EXTENT_UPTODATE,
NULL);
if (ret)
break;
if (extent_start <= start) {
start = extent_end + 1;
} else if (extent_start > start && extent_start < end) {
size = extent_start - start;
total_added += size;
ret = btrfs_add_free_space(block_group, start,
size);
BUG_ON(ret); /* -ENOMEM or logic error */
start = extent_end + 1;
} else {
break;
}
}
if (start < end) {
size = end - start;
total_added += size;
ret = btrfs_add_free_space(block_group, start, size);
BUG_ON(ret); /* -ENOMEM or logic error */
}
return total_added;
}
static int load_extent_tree_free(struct btrfs_caching_control *caching_ctl)
{
struct btrfs_block_group_cache *block_group = caching_ctl->block_group;
struct btrfs_fs_info *fs_info = block_group->fs_info;
struct btrfs_root *extent_root = fs_info->extent_root;
struct btrfs_path *path;
struct extent_buffer *leaf;
struct btrfs_key key;
u64 total_found = 0;
u64 last = 0;
u32 nritems;
int ret;
bool wakeup = true;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
last = max_t(u64, block_group->key.objectid, BTRFS_SUPER_INFO_OFFSET);
#ifdef CONFIG_BTRFS_DEBUG
/*
* If we're fragmenting we don't want to make anybody think we can
* allocate from this block group until we've had a chance to fragment
* the free space.
*/
if (btrfs_should_fragment_free_space(block_group))
wakeup = false;
#endif
/*
* We don't want to deadlock with somebody trying to allocate a new
* extent for the extent root while also trying to search the extent
* root to add free space. So we skip locking and search the commit
* root, since its read-only
*/
path->skip_locking = 1;
path->search_commit_root = 1;
path->reada = READA_FORWARD;
key.objectid = last;
key.offset = 0;
key.type = BTRFS_EXTENT_ITEM_KEY;
next:
ret = btrfs_search_slot(NULL, extent_root, &key, path, 0, 0);
if (ret < 0)
goto out;
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
while (1) {
if (btrfs_fs_closing(fs_info) > 1) {
last = (u64)-1;
break;
}
if (path->slots[0] < nritems) {
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
} else {
ret = find_next_key(path, 0, &key);
if (ret)
break;
if (need_resched() ||
rwsem_is_contended(&fs_info->commit_root_sem)) {
if (wakeup)
caching_ctl->progress = last;
btrfs_release_path(path);
up_read(&fs_info->commit_root_sem);
mutex_unlock(&caching_ctl->mutex);
cond_resched();
mutex_lock(&caching_ctl->mutex);
down_read(&fs_info->commit_root_sem);
goto next;
}
ret = btrfs_next_leaf(extent_root, path);
if (ret < 0)
goto out;
if (ret)
break;
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
continue;
}
if (key.objectid < last) {
key.objectid = last;
key.offset = 0;
key.type = BTRFS_EXTENT_ITEM_KEY;
if (wakeup)
caching_ctl->progress = last;
btrfs_release_path(path);
goto next;
}
if (key.objectid < block_group->key.objectid) {
path->slots[0]++;
continue;
}
if (key.objectid >= block_group->key.objectid +
block_group->key.offset)
break;
if (key.type == BTRFS_EXTENT_ITEM_KEY ||
key.type == BTRFS_METADATA_ITEM_KEY) {
total_found += add_new_free_space(block_group,
fs_info, last,
key.objectid);
if (key.type == BTRFS_METADATA_ITEM_KEY)
last = key.objectid +
fs_info->nodesize;
else
last = key.objectid + key.offset;
if (total_found > CACHING_CTL_WAKE_UP) {
total_found = 0;
if (wakeup)
wake_up(&caching_ctl->wait);
}
}
path->slots[0]++;
}
ret = 0;
total_found += add_new_free_space(block_group, fs_info, last,
block_group->key.objectid +
block_group->key.offset);
caching_ctl->progress = (u64)-1;
out:
btrfs_free_path(path);
return ret;
}
static noinline void caching_thread(struct btrfs_work *work)
{
struct btrfs_block_group_cache *block_group;
struct btrfs_fs_info *fs_info;
struct btrfs_caching_control *caching_ctl;
struct btrfs_root *extent_root;
int ret;
caching_ctl = container_of(work, struct btrfs_caching_control, work);
block_group = caching_ctl->block_group;
fs_info = block_group->fs_info;
extent_root = fs_info->extent_root;
mutex_lock(&caching_ctl->mutex);
down_read(&fs_info->commit_root_sem);
if (btrfs_fs_compat_ro(fs_info, FREE_SPACE_TREE))
ret = load_free_space_tree(caching_ctl);
else
ret = load_extent_tree_free(caching_ctl);
spin_lock(&block_group->lock);
block_group->caching_ctl = NULL;
block_group->cached = ret ? BTRFS_CACHE_ERROR : BTRFS_CACHE_FINISHED;
spin_unlock(&block_group->lock);
#ifdef CONFIG_BTRFS_DEBUG
if (btrfs_should_fragment_free_space(block_group)) {
u64 bytes_used;
spin_lock(&block_group->space_info->lock);
spin_lock(&block_group->lock);
bytes_used = block_group->key.offset -
btrfs_block_group_used(&block_group->item);
block_group->space_info->bytes_used += bytes_used >> 1;
spin_unlock(&block_group->lock);
spin_unlock(&block_group->space_info->lock);
fragment_free_space(block_group);
}
#endif
caching_ctl->progress = (u64)-1;
up_read(&fs_info->commit_root_sem);
free_excluded_extents(fs_info, block_group);
mutex_unlock(&caching_ctl->mutex);
wake_up(&caching_ctl->wait);
put_caching_control(caching_ctl);
btrfs_put_block_group(block_group);
}
static int cache_block_group(struct btrfs_block_group_cache *cache,
int load_cache_only)
{
DEFINE_WAIT(wait);
struct btrfs_fs_info *fs_info = cache->fs_info;
struct btrfs_caching_control *caching_ctl;
int ret = 0;
caching_ctl = kzalloc(sizeof(*caching_ctl), GFP_NOFS);
if (!caching_ctl)
return -ENOMEM;
INIT_LIST_HEAD(&caching_ctl->list);
mutex_init(&caching_ctl->mutex);
init_waitqueue_head(&caching_ctl->wait);
caching_ctl->block_group = cache;
caching_ctl->progress = cache->key.objectid;
refcount_set(&caching_ctl->count, 1);
btrfs_init_work(&caching_ctl->work, btrfs_cache_helper,
caching_thread, NULL, NULL);
spin_lock(&cache->lock);
/*
* This should be a rare occasion, but this could happen I think in the
* case where one thread starts to load the space cache info, and then
* some other thread starts a transaction commit which tries to do an
* allocation while the other thread is still loading the space cache
* info. The previous loop should have kept us from choosing this block
* group, but if we've moved to the state where we will wait on caching
* block groups we need to first check if we're doing a fast load here,
* so we can wait for it to finish, otherwise we could end up allocating
* from a block group who's cache gets evicted for one reason or
* another.
*/
while (cache->cached == BTRFS_CACHE_FAST) {
struct btrfs_caching_control *ctl;
ctl = cache->caching_ctl;
refcount_inc(&ctl->count);
prepare_to_wait(&ctl->wait, &wait, TASK_UNINTERRUPTIBLE);
spin_unlock(&cache->lock);
schedule();
finish_wait(&ctl->wait, &wait);
put_caching_control(ctl);
spin_lock(&cache->lock);
}
if (cache->cached != BTRFS_CACHE_NO) {
spin_unlock(&cache->lock);
kfree(caching_ctl);
return 0;
}
WARN_ON(cache->caching_ctl);
cache->caching_ctl = caching_ctl;
cache->cached = BTRFS_CACHE_FAST;
spin_unlock(&cache->lock);
if (fs_info->mount_opt & BTRFS_MOUNT_SPACE_CACHE) {
mutex_lock(&caching_ctl->mutex);
ret = load_free_space_cache(fs_info, cache);
spin_lock(&cache->lock);
if (ret == 1) {
cache->caching_ctl = NULL;
cache->cached = BTRFS_CACHE_FINISHED;
cache->last_byte_to_unpin = (u64)-1;
caching_ctl->progress = (u64)-1;
} else {
if (load_cache_only) {
cache->caching_ctl = NULL;
cache->cached = BTRFS_CACHE_NO;
} else {
cache->cached = BTRFS_CACHE_STARTED;
cache->has_caching_ctl = 1;
}
}
spin_unlock(&cache->lock);
#ifdef CONFIG_BTRFS_DEBUG
if (ret == 1 &&
btrfs_should_fragment_free_space(cache)) {
u64 bytes_used;
spin_lock(&cache->space_info->lock);
spin_lock(&cache->lock);
bytes_used = cache->key.offset -
btrfs_block_group_used(&cache->item);
cache->space_info->bytes_used += bytes_used >> 1;
spin_unlock(&cache->lock);
spin_unlock(&cache->space_info->lock);
fragment_free_space(cache);
}
#endif
mutex_unlock(&caching_ctl->mutex);
wake_up(&caching_ctl->wait);
if (ret == 1) {
put_caching_control(caching_ctl);
free_excluded_extents(fs_info, cache);
return 0;
}
} else {
/*
* We're either using the free space tree or no caching at all.
* Set cached to the appropriate value and wakeup any waiters.
*/
spin_lock(&cache->lock);
if (load_cache_only) {
cache->caching_ctl = NULL;
cache->cached = BTRFS_CACHE_NO;
} else {
cache->cached = BTRFS_CACHE_STARTED;
cache->has_caching_ctl = 1;
}
spin_unlock(&cache->lock);
wake_up(&caching_ctl->wait);
}
if (load_cache_only) {
put_caching_control(caching_ctl);
return 0;
}
down_write(&fs_info->commit_root_sem);
refcount_inc(&caching_ctl->count);
list_add_tail(&caching_ctl->list, &fs_info->caching_block_groups);
up_write(&fs_info->commit_root_sem);
btrfs_get_block_group(cache);
btrfs_queue_work(fs_info->caching_workers, &caching_ctl->work);
return ret;
}
/*
* return the block group that starts at or after bytenr
*/
static struct btrfs_block_group_cache *
btrfs_lookup_first_block_group(struct btrfs_fs_info *info, u64 bytenr)
{
return block_group_cache_tree_search(info, bytenr, 0);
}
/*
* return the block group that contains the given bytenr
*/
struct btrfs_block_group_cache *btrfs_lookup_block_group(
struct btrfs_fs_info *info,
u64 bytenr)
{
return block_group_cache_tree_search(info, bytenr, 1);
}
static struct btrfs_space_info *__find_space_info(struct btrfs_fs_info *info,
u64 flags)
{
struct list_head *head = &info->space_info;
struct btrfs_space_info *found;
flags &= BTRFS_BLOCK_GROUP_TYPE_MASK;
rcu_read_lock();
list_for_each_entry_rcu(found, head, list) {
if (found->flags & flags) {
rcu_read_unlock();
return found;
}
}
rcu_read_unlock();
return NULL;
}
static void add_pinned_bytes(struct btrfs_fs_info *fs_info, s64 num_bytes,
u64 owner, u64 root_objectid)
{
struct btrfs_space_info *space_info;
u64 flags;
if (owner < BTRFS_FIRST_FREE_OBJECTID) {
if (root_objectid == BTRFS_CHUNK_TREE_OBJECTID)
flags = BTRFS_BLOCK_GROUP_SYSTEM;
else
flags = BTRFS_BLOCK_GROUP_METADATA;
} else {
flags = BTRFS_BLOCK_GROUP_DATA;
}
space_info = __find_space_info(fs_info, flags);
ASSERT(space_info);
percpu_counter_add(&space_info->total_bytes_pinned, num_bytes);
}
/*
* after adding space to the filesystem, we need to clear the full flags
* on all the space infos.
*/
void btrfs_clear_space_info_full(struct btrfs_fs_info *info)
{
struct list_head *head = &info->space_info;
struct btrfs_space_info *found;
rcu_read_lock();
list_for_each_entry_rcu(found, head, list)
found->full = 0;
rcu_read_unlock();
}
/* simple helper to search for an existing data extent at a given offset */
int btrfs_lookup_data_extent(struct btrfs_fs_info *fs_info, u64 start, u64 len)
{
int ret;
struct btrfs_key key;
struct btrfs_path *path;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
key.objectid = start;
key.offset = len;
key.type = BTRFS_EXTENT_ITEM_KEY;
ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
btrfs_free_path(path);
return ret;
}
/*
* helper function to lookup reference count and flags of a tree block.
*
* the head node for delayed ref is used to store the sum of all the
* reference count modifications queued up in the rbtree. the head
* node may also store the extent flags to set. This way you can check
* to see what the reference count and extent flags would be if all of
* the delayed refs are not processed.
*/
int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info, u64 bytenr,
u64 offset, int metadata, u64 *refs, u64 *flags)
{
struct btrfs_delayed_ref_head *head;
struct btrfs_delayed_ref_root *delayed_refs;
struct btrfs_path *path;
struct btrfs_extent_item *ei;
struct extent_buffer *leaf;
struct btrfs_key key;
u32 item_size;
u64 num_refs;
u64 extent_flags;
int ret;
/*
* If we don't have skinny metadata, don't bother doing anything
* different
*/
if (metadata && !btrfs_fs_incompat(fs_info, SKINNY_METADATA)) {
offset = fs_info->nodesize;
metadata = 0;
}
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
if (!trans) {
path->skip_locking = 1;
path->search_commit_root = 1;
}
search_again:
key.objectid = bytenr;
key.offset = offset;
if (metadata)
key.type = BTRFS_METADATA_ITEM_KEY;
else
key.type = BTRFS_EXTENT_ITEM_KEY;
ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 0);
if (ret < 0)
goto out_free;
if (ret > 0 && metadata && key.type == BTRFS_METADATA_ITEM_KEY) {
if (path->slots[0]) {
path->slots[0]--;
btrfs_item_key_to_cpu(path->nodes[0], &key,
path->slots[0]);
if (key.objectid == bytenr &&
key.type == BTRFS_EXTENT_ITEM_KEY &&
key.offset == fs_info->nodesize)
ret = 0;
}
}
if (ret == 0) {
leaf = path->nodes[0];
item_size = btrfs_item_size_nr(leaf, path->slots[0]);
if (item_size >= sizeof(*ei)) {
ei = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_item);
num_refs = btrfs_extent_refs(leaf, ei);
extent_flags = btrfs_extent_flags(leaf, ei);
} else {
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
struct btrfs_extent_item_v0 *ei0;
BUG_ON(item_size != sizeof(*ei0));
ei0 = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_item_v0);
num_refs = btrfs_extent_refs_v0(leaf, ei0);
/* FIXME: this isn't correct for data */
extent_flags = BTRFS_BLOCK_FLAG_FULL_BACKREF;
#else
BUG();
#endif
}
BUG_ON(num_refs == 0);
} else {
num_refs = 0;
extent_flags = 0;
ret = 0;
}
if (!trans)
goto out;
delayed_refs = &trans->transaction->delayed_refs;
spin_lock(&delayed_refs->lock);
head = btrfs_find_delayed_ref_head(delayed_refs, bytenr);
if (head) {
if (!mutex_trylock(&head->mutex)) {
refcount_inc(&head->node.refs);
spin_unlock(&delayed_refs->lock);
btrfs_release_path(path);
/*
* Mutex was contended, block until it's released and try
* again
*/
mutex_lock(&head->mutex);
mutex_unlock(&head->mutex);
btrfs_put_delayed_ref(&head->node);
goto search_again;
}
spin_lock(&head->lock);
if (head->extent_op && head->extent_op->update_flags)
extent_flags |= head->extent_op->flags_to_set;
else
BUG_ON(num_refs == 0);
num_refs += head->node.ref_mod;
spin_unlock(&head->lock);
mutex_unlock(&head->mutex);
}
spin_unlock(&delayed_refs->lock);
out:
WARN_ON(num_refs == 0);
if (refs)
*refs = num_refs;
if (flags)
*flags = extent_flags;
out_free:
btrfs_free_path(path);
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.
*
* There are two kinds of back refs. The implicit back refs is optimized
* for pointers in non-shared tree blocks. For a given pointer in a block,
* back refs of this kind provide information about the block's owner tree
* and the pointer's key. These information allow us to find the block by
* b-tree searching. The full back refs is for pointers in tree blocks not
* referenced by their owner trees. The location of tree block is recorded
* in the back refs. Actually the full back refs is generic, and can be
* used in all cases the implicit back refs is used. The major shortcoming
* of the full back refs is its overhead. Every time a tree block gets
* COWed, we have to update back refs entry for all pointers in it.
*
* For a newly allocated tree block, we use implicit back refs for
* pointers in it. This means most tree related operations only involve
* implicit back refs. For a tree block created in old transaction, the
* only way to drop a reference to it is COW it. So we can detect the
* event that tree block loses its owner tree's reference and do the
* back refs conversion.
*
* When a tree block is COWed through a tree, there are four cases:
*
* The reference count of the block is one and the tree is the block's
* owner tree. Nothing to do in this case.
*
* The reference count of the block is one and the tree is not the
* block's owner tree. In this case, full back refs is used for pointers
* in the block. Remove these full back refs, add implicit back refs for
* every pointers in the new block.
*
* The reference count of the block is greater than one and the tree is
* the block's owner tree. In this case, implicit back refs is used for
* pointers in the block. Add full back refs for every pointers in the
* block, increase lower level extents' reference counts. The original
* implicit back refs are entailed to the new block.
*
* The reference count of the block is greater than one and the tree is
* not the block's owner tree. Add implicit back refs for every pointer in
* the new block, increase lower level extents' reference count.
*
* Back Reference Key composing:
*
* The key objectid corresponds to the first byte in the extent,
* The key type is used to differentiate between types of back refs.
* There are different meanings of the key offset for different types
* of back refs.
*
* File extents can be referenced by:
*
* - multiple snapshots, subvolumes, or different generations in one subvol
* - different files inside a single subvolume
* - different offsets inside a file (bookend extents in file.c)
*
* The extent ref structure for the implicit back refs has fields for:
*
* - Objectid of the subvolume root
* - objectid of the file holding the reference
* - original offset in the file
* - how many bookend extents
*
* The key offset for the implicit back refs is hash of the first
* three fields.
*
* The extent ref structure for the full back refs has field for:
*
* - number of pointers in the tree leaf
*
* The key offset for the implicit back refs is the first byte of
* the tree leaf
*
* When a file extent is allocated, The implicit back refs is used.
* the fields are filled in:
*
* (root_key.objectid, inode objectid, offset in file, 1)
*
* When a file extent is removed file truncation, we find the
* corresponding implicit back refs and check the following fields:
*
* (btrfs_header_owner(leaf), inode objectid, offset in file)
*
* Btree extents can be referenced by:
*
* - Different subvolumes
*
* Both the implicit back refs and the full back refs for tree blocks
* only consist of key. The key offset for the implicit back refs is
* objectid of block's owner tree. The key offset for the full back refs
* is the first byte of parent block.
*
* When implicit back refs is used, information about the lowest key and
* level of the tree block are required. These information are stored in
* tree block info structure.
*/
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
static int convert_extent_item_v0(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_path *path,
u64 owner, u32 extra_size)
{
struct btrfs_root *root = fs_info->extent_root;
struct btrfs_extent_item *item;
struct btrfs_extent_item_v0 *ei0;
struct btrfs_extent_ref_v0 *ref0;
struct btrfs_tree_block_info *bi;
struct extent_buffer *leaf;
struct btrfs_key key;
struct btrfs_key found_key;
u32 new_size = sizeof(*item);
u64 refs;
int ret;
leaf = path->nodes[0];
BUG_ON(btrfs_item_size_nr(leaf, path->slots[0]) != sizeof(*ei0));
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
ei0 = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_item_v0);
refs = btrfs_extent_refs_v0(leaf, ei0);
if (owner == (u64)-1) {
while (1) {
if (path->slots[0] >= btrfs_header_nritems(leaf)) {
ret = btrfs_next_leaf(root, path);
if (ret < 0)
return ret;
BUG_ON(ret > 0); /* Corruption */
leaf = path->nodes[0];
}
btrfs_item_key_to_cpu(leaf, &found_key,
path->slots[0]);
BUG_ON(key.objectid != found_key.objectid);
if (found_key.type != BTRFS_EXTENT_REF_V0_KEY) {
path->slots[0]++;
continue;
}
ref0 = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_ref_v0);
owner = btrfs_ref_objectid_v0(leaf, ref0);
break;
}
}
btrfs_release_path(path);
if (owner < BTRFS_FIRST_FREE_OBJECTID)
new_size += sizeof(*bi);
new_size -= sizeof(*ei0);
ret = btrfs_search_slot(trans, root, &key, path,
new_size + extra_size, 1);
if (ret < 0)
return ret;
BUG_ON(ret); /* Corruption */
btrfs_extend_item(fs_info, path, new_size);
leaf = path->nodes[0];
item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
btrfs_set_extent_refs(leaf, item, refs);
/* FIXME: get real generation */
btrfs_set_extent_generation(leaf, item, 0);
if (owner < BTRFS_FIRST_FREE_OBJECTID) {
btrfs_set_extent_flags(leaf, item,
BTRFS_EXTENT_FLAG_TREE_BLOCK |
BTRFS_BLOCK_FLAG_FULL_BACKREF);
bi = (struct btrfs_tree_block_info *)(item + 1);
/* FIXME: get first key of the block */
memzero_extent_buffer(leaf, (unsigned long)bi, sizeof(*bi));
btrfs_set_tree_block_level(leaf, bi, (int)owner);
} else {
btrfs_set_extent_flags(leaf, item, BTRFS_EXTENT_FLAG_DATA);
}
btrfs_mark_buffer_dirty(leaf);
return 0;
}
#endif
/*
* is_data == BTRFS_REF_TYPE_BLOCK, tree block type is required,
* is_data == BTRFS_REF_TYPE_DATA, data type is requried,
* is_data == BTRFS_REF_TYPE_ANY, either type is OK.
*/
int btrfs_get_extent_inline_ref_type(const struct extent_buffer *eb,
struct btrfs_extent_inline_ref *iref,
enum btrfs_inline_ref_type is_data)
{
int type = btrfs_extent_inline_ref_type(eb, iref);
u64 offset = btrfs_extent_inline_ref_offset(eb, iref);
if (type == BTRFS_TREE_BLOCK_REF_KEY ||
type == BTRFS_SHARED_BLOCK_REF_KEY ||
type == BTRFS_SHARED_DATA_REF_KEY ||
type == BTRFS_EXTENT_DATA_REF_KEY) {
if (is_data == BTRFS_REF_TYPE_BLOCK) {
if (type == BTRFS_TREE_BLOCK_REF_KEY)
return type;
if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
ASSERT(eb->fs_info);
/*
* Every shared one has parent tree
* block, which must be aligned to
* nodesize.
*/
if (offset &&
IS_ALIGNED(offset, eb->fs_info->nodesize))
return type;
}
} else if (is_data == BTRFS_REF_TYPE_DATA) {
if (type == BTRFS_EXTENT_DATA_REF_KEY)
return type;
if (type == BTRFS_SHARED_DATA_REF_KEY) {
ASSERT(eb->fs_info);
/*
* Every shared one has parent tree
* block, which must be aligned to
* nodesize.
*/
if (offset &&
IS_ALIGNED(offset, eb->fs_info->nodesize))
return type;
}
} else {
ASSERT(is_data == BTRFS_REF_TYPE_ANY);
return type;
}
}
btrfs_print_leaf((struct extent_buffer *)eb);
btrfs_err(eb->fs_info, "eb %llu invalid extent inline ref type %d",
eb->start, type);
WARN_ON(1);
return BTRFS_REF_TYPE_INVALID;
}
static u64 hash_extent_data_ref(u64 root_objectid, u64 owner, u64 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(owner);
low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
lenum = cpu_to_le64(offset);
low_crc = btrfs_crc32c(low_crc, &lenum, sizeof(lenum));
return ((u64)high_crc << 31) ^ (u64)low_crc;
}
static u64 hash_extent_data_ref_item(struct extent_buffer *leaf,
struct btrfs_extent_data_ref *ref)
{
return hash_extent_data_ref(btrfs_extent_data_ref_root(leaf, ref),
btrfs_extent_data_ref_objectid(leaf, ref),
btrfs_extent_data_ref_offset(leaf, ref));
}
static int match_extent_data_ref(struct extent_buffer *leaf,
struct btrfs_extent_data_ref *ref,
u64 root_objectid, u64 owner, u64 offset)
{
if (btrfs_extent_data_ref_root(leaf, ref) != root_objectid ||
btrfs_extent_data_ref_objectid(leaf, ref) != owner ||
btrfs_extent_data_ref_offset(leaf, ref) != offset)
return 0;
return 1;
}
static noinline int lookup_extent_data_ref(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_path *path,
u64 bytenr, u64 parent,
u64 root_objectid,
u64 owner, u64 offset)
{
struct btrfs_root *root = fs_info->extent_root;
struct btrfs_key key;
struct btrfs_extent_data_ref *ref;
struct extent_buffer *leaf;
u32 nritems;
int ret;
int recow;
int err = -ENOENT;
key.objectid = bytenr;
if (parent) {
key.type = BTRFS_SHARED_DATA_REF_KEY;
key.offset = parent;
} else {
key.type = BTRFS_EXTENT_DATA_REF_KEY;
key.offset = hash_extent_data_ref(root_objectid,
owner, offset);
}
again:
recow = 0;
ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
if (ret < 0) {
err = ret;
goto fail;
}
if (parent) {
if (!ret)
return 0;
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
key.type = BTRFS_EXTENT_REF_V0_KEY;
btrfs_release_path(path);
ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
if (ret < 0) {
err = ret;
goto fail;
}
if (!ret)
return 0;
#endif
goto fail;
}
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
while (1) {
if (path->slots[0] >= nritems) {
ret = btrfs_next_leaf(root, path);
if (ret < 0)
err = ret;
if (ret)
goto fail;
leaf = path->nodes[0];
nritems = btrfs_header_nritems(leaf);
recow = 1;
}
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
if (key.objectid != bytenr ||
key.type != BTRFS_EXTENT_DATA_REF_KEY)
goto fail;
ref = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_data_ref);
if (match_extent_data_ref(leaf, ref, root_objectid,
owner, offset)) {
if (recow) {
btrfs_release_path(path);
goto again;
}
err = 0;
break;
}
path->slots[0]++;
}
fail:
return err;
}
static noinline int insert_extent_data_ref(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_path *path,
u64 bytenr, u64 parent,
u64 root_objectid, u64 owner,
u64 offset, int refs_to_add)
{
struct btrfs_root *root = fs_info->extent_root;
struct btrfs_key key;
struct extent_buffer *leaf;
u32 size;
u32 num_refs;
int ret;
key.objectid = bytenr;
if (parent) {
key.type = BTRFS_SHARED_DATA_REF_KEY;
key.offset = parent;
size = sizeof(struct btrfs_shared_data_ref);
} else {
key.type = BTRFS_EXTENT_DATA_REF_KEY;
key.offset = hash_extent_data_ref(root_objectid,
owner, offset);
size = sizeof(struct btrfs_extent_data_ref);
}
ret = btrfs_insert_empty_item(trans, root, path, &key, size);
if (ret && ret != -EEXIST)
goto fail;
leaf = path->nodes[0];
if (parent) {
struct btrfs_shared_data_ref *ref;
ref = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_shared_data_ref);
if (ret == 0) {
btrfs_set_shared_data_ref_count(leaf, ref, refs_to_add);
} else {
num_refs = btrfs_shared_data_ref_count(leaf, ref);
num_refs += refs_to_add;
btrfs_set_shared_data_ref_count(leaf, ref, num_refs);
}
} else {
struct btrfs_extent_data_ref *ref;
while (ret == -EEXIST) {
ref = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_data_ref);
if (match_extent_data_ref(leaf, ref, root_objectid,
owner, offset))
break;
btrfs_release_path(path);
key.offset++;
ret = btrfs_insert_empty_item(trans, root, path, &key,
size);
if (ret && ret != -EEXIST)
goto fail;
leaf = path->nodes[0];
}
ref = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_data_ref);
if (ret == 0) {
btrfs_set_extent_data_ref_root(leaf, ref,
root_objectid);
btrfs_set_extent_data_ref_objectid(leaf, ref, owner);
btrfs_set_extent_data_ref_offset(leaf, ref, offset);
btrfs_set_extent_data_ref_count(leaf, ref, refs_to_add);
} else {
num_refs = btrfs_extent_data_ref_count(leaf, ref);
num_refs += refs_to_add;
btrfs_set_extent_data_ref_count(leaf, ref, num_refs);
}
}
btrfs_mark_buffer_dirty(leaf);
ret = 0;
fail:
btrfs_release_path(path);
return ret;
}
static noinline int remove_extent_data_ref(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_path *path,
int refs_to_drop, int *last_ref)
{
struct btrfs_key key;
struct btrfs_extent_data_ref *ref1 = NULL;
struct btrfs_shared_data_ref *ref2 = NULL;
struct extent_buffer *leaf;
u32 num_refs = 0;
int ret = 0;
leaf = path->nodes[0];
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
ref1 = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_data_ref);
num_refs = btrfs_extent_data_ref_count(leaf, ref1);
} else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
ref2 = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_shared_data_ref);
num_refs = btrfs_shared_data_ref_count(leaf, ref2);
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
} else if (key.type == BTRFS_EXTENT_REF_V0_KEY) {
struct btrfs_extent_ref_v0 *ref0;
ref0 = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_ref_v0);
num_refs = btrfs_ref_count_v0(leaf, ref0);
#endif
} else {
BUG();
}
BUG_ON(num_refs < refs_to_drop);
num_refs -= refs_to_drop;
if (num_refs == 0) {
ret = btrfs_del_item(trans, fs_info->extent_root, path);
*last_ref = 1;
} else {
if (key.type == BTRFS_EXTENT_DATA_REF_KEY)
btrfs_set_extent_data_ref_count(leaf, ref1, num_refs);
else if (key.type == BTRFS_SHARED_DATA_REF_KEY)
btrfs_set_shared_data_ref_count(leaf, ref2, num_refs);
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
else {
struct btrfs_extent_ref_v0 *ref0;
ref0 = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_ref_v0);
btrfs_set_ref_count_v0(leaf, ref0, num_refs);
}
#endif
btrfs_mark_buffer_dirty(leaf);
}
return ret;
}
static noinline u32 extent_data_ref_count(struct btrfs_path *path,
struct btrfs_extent_inline_ref *iref)
{
struct btrfs_key key;
struct extent_buffer *leaf;
struct btrfs_extent_data_ref *ref1;
struct btrfs_shared_data_ref *ref2;
u32 num_refs = 0;
int type;
leaf = path->nodes[0];
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
if (iref) {
/*
* If type is invalid, we should have bailed out earlier than
* this call.
*/
type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_DATA);
ASSERT(type != BTRFS_REF_TYPE_INVALID);
if (type == BTRFS_EXTENT_DATA_REF_KEY) {
ref1 = (struct btrfs_extent_data_ref *)(&iref->offset);
num_refs = btrfs_extent_data_ref_count(leaf, ref1);
} else {
ref2 = (struct btrfs_shared_data_ref *)(iref + 1);
num_refs = btrfs_shared_data_ref_count(leaf, ref2);
}
} else if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
ref1 = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_data_ref);
num_refs = btrfs_extent_data_ref_count(leaf, ref1);
} else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
ref2 = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_shared_data_ref);
num_refs = btrfs_shared_data_ref_count(leaf, ref2);
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
} else if (key.type == BTRFS_EXTENT_REF_V0_KEY) {
struct btrfs_extent_ref_v0 *ref0;
ref0 = btrfs_item_ptr(leaf, path->slots[0],
struct btrfs_extent_ref_v0);
num_refs = btrfs_ref_count_v0(leaf, ref0);
#endif
} else {
WARN_ON(1);
}
return num_refs;
}
static noinline int lookup_tree_block_ref(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_path *path,
u64 bytenr, u64 parent,
u64 root_objectid)
{
struct btrfs_root *root = fs_info->extent_root;
struct btrfs_key key;
int ret;
key.objectid = bytenr;
if (parent) {
key.type = BTRFS_SHARED_BLOCK_REF_KEY;
key.offset = parent;
} else {
key.type = BTRFS_TREE_BLOCK_REF_KEY;
key.offset = root_objectid;
}
ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
if (ret > 0)
ret = -ENOENT;
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
if (ret == -ENOENT && parent) {
btrfs_release_path(path);
key.type = BTRFS_EXTENT_REF_V0_KEY;
ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
if (ret > 0)
ret = -ENOENT;
}
#endif
return ret;
}
static noinline int insert_tree_block_ref(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_path *path,
u64 bytenr, u64 parent,
u64 root_objectid)
{
struct btrfs_key key;
int ret;
key.objectid = bytenr;
if (parent) {
key.type = BTRFS_SHARED_BLOCK_REF_KEY;
key.offset = parent;
} else {
key.type = BTRFS_TREE_BLOCK_REF_KEY;
key.offset = root_objectid;
}
ret = btrfs_insert_empty_item(trans, fs_info->extent_root,
path, &key, 0);
btrfs_release_path(path);
return ret;
}
static inline int extent_ref_type(u64 parent, u64 owner)
{
int type;
if (owner < BTRFS_FIRST_FREE_OBJECTID) {
if (parent > 0)
type = BTRFS_SHARED_BLOCK_REF_KEY;
else
type = BTRFS_TREE_BLOCK_REF_KEY;
} else {
if (parent > 0)
type = BTRFS_SHARED_DATA_REF_KEY;
else
type = BTRFS_EXTENT_DATA_REF_KEY;
}
return type;
}
static int find_next_key(struct btrfs_path *path, int level,
struct btrfs_key *key)
{
for (; level < BTRFS_MAX_LEVEL; level++) {
if (!path->nodes[level])
break;
if (path->slots[level] + 1 >=
btrfs_header_nritems(path->nodes[level]))
continue;
if (level == 0)
btrfs_item_key_to_cpu(path->nodes[level], key,
path->slots[level] + 1);
else
btrfs_node_key_to_cpu(path->nodes[level], key,
path->slots[level] + 1);
return 0;
}
return 1;
}
/*
* look for inline back ref. if back ref is found, *ref_ret is set
* to the address of inline back ref, and 0 is returned.
*
* if back ref isn't found, *ref_ret is set to the address where it
* should be inserted, and -ENOENT is returned.
*
* if insert is true and there are too many inline back refs, the path
* points to the extent item, and -EAGAIN is returned.
*
* NOTE: inline back refs are ordered in the same way that back ref
* items in the tree are ordered.
*/
static noinline_for_stack
int lookup_inline_extent_backref(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_path *path,
struct btrfs_extent_inline_ref **ref_ret,
u64 bytenr, u64 num_bytes,
u64 parent, u64 root_objectid,
u64 owner, u64 offset, int insert)
{
struct btrfs_root *root = fs_info->extent_root;
struct btrfs_key key;
struct extent_buffer *leaf;
struct btrfs_extent_item *ei;
struct btrfs_extent_inline_ref *iref;
u64 flags;
u64 item_size;
unsigned long ptr;
unsigned long end;
int extra_size;
int type;
int want;
int ret;
int err = 0;
bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
int needed;
key.objectid = bytenr;
key.type = BTRFS_EXTENT_ITEM_KEY;
key.offset = num_bytes;
want = extent_ref_type(parent, owner);
if (insert) {
extra_size = btrfs_extent_inline_ref_size(want);
path->keep_locks = 1;
} else
extra_size = -1;
/*
* Owner is our parent level, so we can just add one to get the level
* for the block we are interested in.
*/
if (skinny_metadata && owner < BTRFS_FIRST_FREE_OBJECTID) {
key.type = BTRFS_METADATA_ITEM_KEY;
key.offset = owner;
}
again:
ret = btrfs_search_slot(trans, root, &key, path, extra_size, 1);
if (ret < 0) {
err = ret;
goto out;
}
/*
* We may be a newly converted file system which still has the old fat
* extent entries for metadata, so try and see if we have one of those.
*/
if (ret > 0 && skinny_metadata) {
skinny_metadata = false;
if (path->slots[0]) {
path->slots[0]--;
btrfs_item_key_to_cpu(path->nodes[0], &key,
path->slots[0]);
if (key.objectid == bytenr &&
key.type == BTRFS_EXTENT_ITEM_KEY &&
key.offset == num_bytes)
ret = 0;
}
if (ret) {
key.objectid = bytenr;
key.type = BTRFS_EXTENT_ITEM_KEY;
key.offset = num_bytes;
btrfs_release_path(path);
goto again;
}
}
if (ret && !insert) {
err = -ENOENT;
goto out;
} else if (WARN_ON(ret)) {
err = -EIO;
goto out;
}
leaf = path->nodes[0];
item_size = btrfs_item_size_nr(leaf, path->slots[0]);
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
if (item_size < sizeof(*ei)) {
if (!insert) {
err = -ENOENT;
goto out;
}
ret = convert_extent_item_v0(trans, fs_info, path, owner,
extra_size);
if (ret < 0) {
err = ret;
goto out;
}
leaf = path->nodes[0];
item_size = btrfs_item_size_nr(leaf, path->slots[0]);
}
#endif
BUG_ON(item_size < sizeof(*ei));
ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
flags = btrfs_extent_flags(leaf, ei);
ptr = (unsigned long)(ei + 1);
end = (unsigned long)ei + item_size;
if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK && !skinny_metadata) {
ptr += sizeof(struct btrfs_tree_block_info);
BUG_ON(ptr > end);
}
if (owner >= BTRFS_FIRST_FREE_OBJECTID)
needed = BTRFS_REF_TYPE_DATA;
else
needed = BTRFS_REF_TYPE_BLOCK;
err = -ENOENT;
while (1) {
if (ptr >= end) {
WARN_ON(ptr > end);
break;
}
iref = (struct btrfs_extent_inline_ref *)ptr;
type = btrfs_get_extent_inline_ref_type(leaf, iref, needed);
if (type == BTRFS_REF_TYPE_INVALID) {
err = -EINVAL;
goto out;
}
if (want < type)
break;
if (want > type) {
ptr += btrfs_extent_inline_ref_size(type);
continue;
}
if (type == BTRFS_EXTENT_DATA_REF_KEY) {
struct btrfs_extent_data_ref *dref;
dref = (struct btrfs_extent_data_ref *)(&iref->offset);
if (match_extent_data_ref(leaf, dref, root_objectid,
owner, offset)) {
err = 0;
break;
}
if (hash_extent_data_ref_item(leaf, dref) <
hash_extent_data_ref(root_objectid, owner, offset))
break;
} else {
u64 ref_offset;
ref_offset = btrfs_extent_inline_ref_offset(leaf, iref);
if (parent > 0) {
if (parent == ref_offset) {
err = 0;
break;
}
if (ref_offset < parent)
break;
} else {
if (root_objectid == ref_offset) {
err = 0;
break;
}
if (ref_offset < root_objectid)
break;
}
}
ptr += btrfs_extent_inline_ref_size(type);
}
if (err == -ENOENT && insert) {
if (item_size + extra_size >=
BTRFS_MAX_EXTENT_ITEM_SIZE(root)) {
err = -EAGAIN;
goto out;
}
/*
* To add new inline back ref, we have to make sure
* there is no corresponding back ref item.
* For simplicity, we just do not add new inline back
* ref if there is any kind of item for this block
*/
if (find_next_key(path, 0, &key) == 0 &&
key.objectid == bytenr &&
key.type < BTRFS_BLOCK_GROUP_ITEM_KEY) {
err = -EAGAIN;
goto out;
}
}
*ref_ret = (struct btrfs_extent_inline_ref *)ptr;
out:
if (insert) {
path->keep_locks = 0;
btrfs_unlock_up_safe(path, 1);
}
return err;
}
/*
* helper to add new inline back ref
*/
static noinline_for_stack
void setup_inline_extent_backref(struct btrfs_fs_info *fs_info,
struct btrfs_path *path,
struct btrfs_extent_inline_ref *iref,
u64 parent, u64 root_objectid,
u64 owner, u64 offset, int refs_to_add,
struct btrfs_delayed_extent_op *extent_op)
{
struct extent_buffer *leaf;
struct btrfs_extent_item *ei;
unsigned long ptr;
unsigned long end;
unsigned long item_offset;
u64 refs;
int size;
int type;
leaf = path->nodes[0];
ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
item_offset = (unsigned long)iref - (unsigned long)ei;
type = extent_ref_type(parent, owner);
size = btrfs_extent_inline_ref_size(type);
btrfs_extend_item(fs_info, path, size);
ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
refs = btrfs_extent_refs(leaf, ei);
refs += refs_to_add;
btrfs_set_extent_refs(leaf, ei, refs);
if (extent_op)
__run_delayed_extent_op(extent_op, leaf, ei);
ptr = (unsigned long)ei + item_offset;
end = (unsigned long)ei + btrfs_item_size_nr(leaf, path->slots[0]);
if (ptr < end - size)
memmove_extent_buffer(leaf, ptr + size, ptr,
end - size - ptr);
iref = (struct btrfs_extent_inline_ref *)ptr;
btrfs_set_extent_inline_ref_type(leaf, iref, type);
if (type == BTRFS_EXTENT_DATA_REF_KEY) {
struct btrfs_extent_data_ref *dref;
dref = (struct btrfs_extent_data_ref *)(&iref->offset);
btrfs_set_extent_data_ref_root(leaf, dref, root_objectid);
btrfs_set_extent_data_ref_objectid(leaf, dref, owner);
btrfs_set_extent_data_ref_offset(leaf, dref, offset);
btrfs_set_extent_data_ref_count(leaf, dref, refs_to_add);
} else if (type == BTRFS_SHARED_DATA_REF_KEY) {
struct btrfs_shared_data_ref *sref;
sref = (struct btrfs_shared_data_ref *)(iref + 1);
btrfs_set_shared_data_ref_count(leaf, sref, refs_to_add);
btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
} else if (type == BTRFS_SHARED_BLOCK_REF_KEY) {
btrfs_set_extent_inline_ref_offset(leaf, iref, parent);
} else {
btrfs_set_extent_inline_ref_offset(leaf, iref, root_objectid);
}
btrfs_mark_buffer_dirty(leaf);
}
static int lookup_extent_backref(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_path *path,
struct btrfs_extent_inline_ref **ref_ret,
u64 bytenr, u64 num_bytes, u64 parent,
u64 root_objectid, u64 owner, u64 offset)
{
int ret;
ret = lookup_inline_extent_backref(trans, fs_info, path, ref_ret,
bytenr, num_bytes, parent,
root_objectid, owner, offset, 0);
if (ret != -ENOENT)
return ret;
btrfs_release_path(path);
*ref_ret = NULL;
if (owner < BTRFS_FIRST_FREE_OBJECTID) {
ret = lookup_tree_block_ref(trans, fs_info, path, bytenr,
parent, root_objectid);
} else {
ret = lookup_extent_data_ref(trans, fs_info, path, bytenr,
parent, root_objectid, owner,
offset);
}
return ret;
}
/*
* helper to update/remove inline back ref
*/
static noinline_for_stack
void update_inline_extent_backref(struct btrfs_fs_info *fs_info,
struct btrfs_path *path,
struct btrfs_extent_inline_ref *iref,
int refs_to_mod,
struct btrfs_delayed_extent_op *extent_op,
int *last_ref)
{
struct extent_buffer *leaf;
struct btrfs_extent_item *ei;
struct btrfs_extent_data_ref *dref = NULL;
struct btrfs_shared_data_ref *sref = NULL;
unsigned long ptr;
unsigned long end;
u32 item_size;
int size;
int type;
u64 refs;
leaf = path->nodes[0];
ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
refs = btrfs_extent_refs(leaf, ei);
WARN_ON(refs_to_mod < 0 && refs + refs_to_mod <= 0);
refs += refs_to_mod;
btrfs_set_extent_refs(leaf, ei, refs);
if (extent_op)
__run_delayed_extent_op(extent_op, leaf, ei);
/*
* If type is invalid, we should have bailed out after
* lookup_inline_extent_backref().
*/
type = btrfs_get_extent_inline_ref_type(leaf, iref, BTRFS_REF_TYPE_ANY);
ASSERT(type != BTRFS_REF_TYPE_INVALID);
if (type == BTRFS_EXTENT_DATA_REF_KEY) {
dref = (struct btrfs_extent_data_ref *)(&iref->offset);
refs = btrfs_extent_data_ref_count(leaf, dref);
} else if (type == BTRFS_SHARED_DATA_REF_KEY) {
sref = (struct btrfs_shared_data_ref *)(iref + 1);
refs = btrfs_shared_data_ref_count(leaf, sref);
} else {
refs = 1;
BUG_ON(refs_to_mod != -1);
}
BUG_ON(refs_to_mod < 0 && refs < -refs_to_mod);
refs += refs_to_mod;
if (refs > 0) {
if (type == BTRFS_EXTENT_DATA_REF_KEY)
btrfs_set_extent_data_ref_count(leaf, dref, refs);
else
btrfs_set_shared_data_ref_count(leaf, sref, refs);
} else {
*last_ref = 1;
size = btrfs_extent_inline_ref_size(type);
item_size = btrfs_item_size_nr(leaf, path->slots[0]);
ptr = (unsigned long)iref;
end = (unsigned long)ei + item_size;
if (ptr + size < end)
memmove_extent_buffer(leaf, ptr, ptr + size,
end - ptr - size);
item_size -= size;
btrfs_truncate_item(fs_info, path, item_size, 1);
}
btrfs_mark_buffer_dirty(leaf);
}
static noinline_for_stack
int insert_inline_extent_backref(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_path *path,
u64 bytenr, u64 num_bytes, u64 parent,
u64 root_objectid, u64 owner,
u64 offset, int refs_to_add,
struct btrfs_delayed_extent_op *extent_op)
{
struct btrfs_extent_inline_ref *iref;
int ret;
ret = lookup_inline_extent_backref(trans, fs_info, path, &iref,
bytenr, num_bytes, parent,
root_objectid, owner, offset, 1);
if (ret == 0) {
BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID);
update_inline_extent_backref(fs_info, path, iref,
refs_to_add, extent_op, NULL);
} else if (ret == -ENOENT) {
setup_inline_extent_backref(fs_info, path, iref, parent,
root_objectid, owner, offset,
refs_to_add, extent_op);
ret = 0;
}
return ret;
}
static int insert_extent_backref(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_path *path,
u64 bytenr, u64 parent, u64 root_objectid,
u64 owner, u64 offset, int refs_to_add)
{
int ret;
if (owner < BTRFS_FIRST_FREE_OBJECTID) {
BUG_ON(refs_to_add != 1);
ret = insert_tree_block_ref(trans, fs_info, path, bytenr,
parent, root_objectid);
} else {
ret = insert_extent_data_ref(trans, fs_info, path, bytenr,
parent, root_objectid,
owner, offset, refs_to_add);
}
return ret;
}
static int remove_extent_backref(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_path *path,
struct btrfs_extent_inline_ref *iref,
int refs_to_drop, int is_data, int *last_ref)
{
int ret = 0;
BUG_ON(!is_data && refs_to_drop != 1);
if (iref) {
update_inline_extent_backref(fs_info, path, iref,
-refs_to_drop, NULL, last_ref);
} else if (is_data) {
ret = remove_extent_data_ref(trans, fs_info, path, refs_to_drop,
last_ref);
} else {
*last_ref = 1;
ret = btrfs_del_item(trans, fs_info->extent_root, path);
}
return ret;
}
#define in_range(b, first, len) ((b) >= (first) && (b) < (first) + (len))
static int btrfs_issue_discard(struct block_device *bdev, u64 start, u64 len,
u64 *discarded_bytes)
{
int j, ret = 0;
u64 bytes_left, end;
u64 aligned_start = ALIGN(start, 1 << 9);
if (WARN_ON(start != aligned_start)) {
len -= aligned_start - start;
len = round_down(len, 1 << 9);
start = aligned_start;
}
*discarded_bytes = 0;
if (!len)
return 0;
end = start + len;
bytes_left = len;
/* Skip any superblocks on this device. */
for (j = 0; j < BTRFS_SUPER_MIRROR_MAX; j++) {
u64 sb_start = btrfs_sb_offset(j);
u64 sb_end = sb_start + BTRFS_SUPER_INFO_SIZE;
u64 size = sb_start - start;
if (!in_range(sb_start, start, bytes_left) &&
!in_range(sb_end, start, bytes_left) &&
!in_range(start, sb_start, BTRFS_SUPER_INFO_SIZE))
continue;
/*
* Superblock spans beginning of range. Adjust start and
* try again.
*/
if (sb_start <= start) {
start += sb_end - start;
if (start > end) {
bytes_left = 0;
break;
}
bytes_left = end - start;
continue;
}
if (size) {
ret = blkdev_issue_discard(bdev, start >> 9, size >> 9,
GFP_NOFS, 0);
if (!ret)
*discarded_bytes += size;
else if (ret != -EOPNOTSUPP)
return ret;
}
start = sb_end;
if (start > end) {
bytes_left = 0;
break;
}
bytes_left = end - start;
}
if (bytes_left) {
ret = blkdev_issue_discard(bdev, start >> 9, bytes_left >> 9,
GFP_NOFS, 0);
if (!ret)
*discarded_bytes += bytes_left;
}
return ret;
}
int btrfs_discard_extent(struct btrfs_fs_info *fs_info, u64 bytenr,
u64 num_bytes, u64 *actual_bytes)
{
int ret;
u64 discarded_bytes = 0;
struct btrfs_bio *bbio = NULL;
/*
* Avoid races with device replace and make sure our bbio has devices
* associated to its stripes that don't go away while we are discarding.
*/
btrfs_bio_counter_inc_blocked(fs_info);
/* Tell the block device(s) that the sectors can be discarded */
ret = btrfs_map_block(fs_info, BTRFS_MAP_DISCARD, bytenr, &num_bytes,
&bbio, 0);
/* Error condition is -ENOMEM */
if (!ret) {
struct btrfs_bio_stripe *stripe = bbio->stripes;
int i;
for (i = 0; i < bbio->num_stripes; i++, stripe++) {
u64 bytes;
if (!stripe->dev->can_discard)
continue;
ret = btrfs_issue_discard(stripe->dev->bdev,
stripe->physical,
stripe->length,
&bytes);
if (!ret)
discarded_bytes += bytes;
else if (ret != -EOPNOTSUPP)
break; /* Logic errors or -ENOMEM, or -EIO but I don't know how that could happen JDM */
/*
* Just in case we get back EOPNOTSUPP for some reason,
* just ignore the return value so we don't screw up
* people calling discard_extent.
*/
ret = 0;
}
btrfs_put_bbio(bbio);
}
btrfs_bio_counter_dec(fs_info);
if (actual_bytes)
*actual_bytes = discarded_bytes;
if (ret == -EOPNOTSUPP)
ret = 0;
return ret;
}
/* Can return -ENOMEM */
int btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
u64 bytenr, u64 num_bytes, u64 parent,
u64 root_objectid, u64 owner, u64 offset)
{
int old_ref_mod, new_ref_mod;
int ret;
BUG_ON(owner < BTRFS_FIRST_FREE_OBJECTID &&
root_objectid == BTRFS_TREE_LOG_OBJECTID);
if (owner < BTRFS_FIRST_FREE_OBJECTID) {
ret = btrfs_add_delayed_tree_ref(fs_info, trans, bytenr,
num_bytes, parent,
root_objectid, (int)owner,
BTRFS_ADD_DELAYED_REF, NULL,
&old_ref_mod, &new_ref_mod);
} else {
ret = btrfs_add_delayed_data_ref(fs_info, trans, bytenr,
num_bytes, parent,
root_objectid, owner, offset,
0, BTRFS_ADD_DELAYED_REF,
&old_ref_mod, &new_ref_mod);
}
if (ret == 0 && old_ref_mod < 0 && new_ref_mod >= 0)
add_pinned_bytes(fs_info, -num_bytes, owner, root_objectid);
return ret;
}
static int __btrfs_inc_extent_ref(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_delayed_ref_node *node,
u64 parent, u64 root_objectid,
u64 owner, u64 offset, int refs_to_add,
struct btrfs_delayed_extent_op *extent_op)
{
struct btrfs_path *path;
struct extent_buffer *leaf;
struct btrfs_extent_item *item;
struct btrfs_key key;
u64 bytenr = node->bytenr;
u64 num_bytes = node->num_bytes;
u64 refs;
int ret;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
path->reada = READA_FORWARD;
path->leave_spinning = 1;
/* this will setup the path even if it fails to insert the back ref */
ret = insert_inline_extent_backref(trans, fs_info, path, bytenr,
num_bytes, parent, root_objectid,
owner, offset,
refs_to_add, extent_op);
if ((ret < 0 && ret != -EAGAIN) || !ret)
goto out;
/*
* Ok we had -EAGAIN which means we didn't have space to insert and
* inline extent ref, so just update the reference count and add a
* normal backref.
*/
leaf = path->nodes[0];
btrfs_item_key_to_cpu(leaf, &key, path->slots[0]);
item = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
refs = btrfs_extent_refs(leaf, item);
btrfs_set_extent_refs(leaf, item, refs + refs_to_add);
if (extent_op)
__run_delayed_extent_op(extent_op, leaf, item);
btrfs_mark_buffer_dirty(leaf);
btrfs_release_path(path);
path->reada = READA_FORWARD;
path->leave_spinning = 1;
/* now insert the actual backref */
ret = insert_extent_backref(trans, fs_info, path, bytenr, parent,
root_objectid, owner, offset, refs_to_add);
if (ret)
btrfs_abort_transaction(trans, ret);
out:
btrfs_free_path(path);
return ret;
}
static int run_delayed_data_ref(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_delayed_ref_node *node,
struct btrfs_delayed_extent_op *extent_op,
int insert_reserved)
{
int ret = 0;
struct btrfs_delayed_data_ref *ref;
struct btrfs_key ins;
u64 parent = 0;
u64 ref_root = 0;
u64 flags = 0;
ins.objectid = node->bytenr;
ins.offset = node->num_bytes;
ins.type = BTRFS_EXTENT_ITEM_KEY;
ref = btrfs_delayed_node_to_data_ref(node);
trace_run_delayed_data_ref(fs_info, node, ref, node->action);
if (node->type == BTRFS_SHARED_DATA_REF_KEY)
parent = ref->parent;
ref_root = ref->root;
if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
if (extent_op)
flags |= extent_op->flags_to_set;
ret = alloc_reserved_file_extent(trans, fs_info,
parent, ref_root, flags,
ref->objectid, ref->offset,
&ins, node->ref_mod);
} else if (node->action == BTRFS_ADD_DELAYED_REF) {
ret = __btrfs_inc_extent_ref(trans, fs_info, node, parent,
ref_root, ref->objectid,
ref->offset, node->ref_mod,
extent_op);
} else if (node->action == BTRFS_DROP_DELAYED_REF) {
ret = __btrfs_free_extent(trans, fs_info, node, parent,
ref_root, ref->objectid,
ref->offset, node->ref_mod,
extent_op);
} else {
BUG();
}
return ret;
}
static void __run_delayed_extent_op(struct btrfs_delayed_extent_op *extent_op,
struct extent_buffer *leaf,
struct btrfs_extent_item *ei)
{
u64 flags = btrfs_extent_flags(leaf, ei);
if (extent_op->update_flags) {
flags |= extent_op->flags_to_set;
btrfs_set_extent_flags(leaf, ei, flags);
}
if (extent_op->update_key) {
struct btrfs_tree_block_info *bi;
BUG_ON(!(flags & BTRFS_EXTENT_FLAG_TREE_BLOCK));
bi = (struct btrfs_tree_block_info *)(ei + 1);
btrfs_set_tree_block_key(leaf, bi, &extent_op->key);
}
}
static int run_delayed_extent_op(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_delayed_ref_node *node,
struct btrfs_delayed_extent_op *extent_op)
{
struct btrfs_key key;
struct btrfs_path *path;
struct btrfs_extent_item *ei;
struct extent_buffer *leaf;
u32 item_size;
int ret;
int err = 0;
int metadata = !extent_op->is_data;
if (trans->aborted)
return 0;
if (metadata && !btrfs_fs_incompat(fs_info, SKINNY_METADATA))
metadata = 0;
path = btrfs_alloc_path();
if (!path)
return -ENOMEM;
key.objectid = node->bytenr;
if (metadata) {
key.type = BTRFS_METADATA_ITEM_KEY;
key.offset = extent_op->level;
} else {
key.type = BTRFS_EXTENT_ITEM_KEY;
key.offset = node->num_bytes;
}
again:
path->reada = READA_FORWARD;
path->leave_spinning = 1;
ret = btrfs_search_slot(trans, fs_info->extent_root, &key, path, 0, 1);
if (ret < 0) {
err = ret;
goto out;
}
if (ret > 0) {
if (metadata) {
if (path->slots[0] > 0) {
path->slots[0]--;
btrfs_item_key_to_cpu(path->nodes[0], &key,
path->slots[0]);
if (key.objectid == node->bytenr &&
key.type == BTRFS_EXTENT_ITEM_KEY &&
key.offset == node->num_bytes)
ret = 0;
}
if (ret > 0) {
btrfs_release_path(path);
metadata = 0;
key.objectid = node->bytenr;
key.offset = node->num_bytes;
key.type = BTRFS_EXTENT_ITEM_KEY;
goto again;
}
} else {
err = -EIO;
goto out;
}
}
leaf = path->nodes[0];
item_size = btrfs_item_size_nr(leaf, path->slots[0]);
#ifdef BTRFS_COMPAT_EXTENT_TREE_V0
if (item_size < sizeof(*ei)) {
ret = convert_extent_item_v0(trans, fs_info, path, (u64)-1, 0);
if (ret < 0) {
err = ret;
goto out;
}
leaf = path->nodes[0];
item_size = btrfs_item_size_nr(leaf, path->slots[0]);
}
#endif
BUG_ON(item_size < sizeof(*ei));
ei = btrfs_item_ptr(leaf, path->slots[0], struct btrfs_extent_item);
__run_delayed_extent_op(extent_op, leaf, ei);
btrfs_mark_buffer_dirty(leaf);
out:
btrfs_free_path(path);
return err;
}
static int run_delayed_tree_ref(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_delayed_ref_node *node,
struct btrfs_delayed_extent_op *extent_op,
int insert_reserved)
{
int ret = 0;
struct btrfs_delayed_tree_ref *ref;
struct btrfs_key ins;
u64 parent = 0;
u64 ref_root = 0;
bool skinny_metadata = btrfs_fs_incompat(fs_info, SKINNY_METADATA);
ref = btrfs_delayed_node_to_tree_ref(node);
trace_run_delayed_tree_ref(fs_info, node, ref, node->action);
if (node->type == BTRFS_SHARED_BLOCK_REF_KEY)
parent = ref->parent;
ref_root = ref->root;
ins.objectid = node->bytenr;
if (skinny_metadata) {
ins.offset = ref->level;
ins.type = BTRFS_METADATA_ITEM_KEY;
} else {
ins.offset = node->num_bytes;
ins.type = BTRFS_EXTENT_ITEM_KEY;
}
if (node->ref_mod != 1) {
btrfs_err(fs_info,
"btree block(%llu) has %d references rather than 1: action %d ref_root %llu parent %llu",
node->bytenr, node->ref_mod, node->action, ref_root,
parent);
return -EIO;
}
if (node->action == BTRFS_ADD_DELAYED_REF && insert_reserved) {
BUG_ON(!extent_op || !extent_op->update_flags);
ret = alloc_reserved_tree_block(trans, fs_info,
parent, ref_root,
extent_op->flags_to_set,
&extent_op->key,
ref->level, &ins);
} else if (node->action == BTRFS_ADD_DELAYED_REF) {
ret = __btrfs_inc_extent_ref(trans, fs_info, node,
parent, ref_root,
ref->level, 0, 1,
extent_op);
} else if (node->action == BTRFS_DROP_DELAYED_REF) {
ret = __btrfs_free_extent(trans, fs_info, node,
parent, ref_root,
ref->level, 0, 1, extent_op);
} else {
BUG();
}
return ret;
}
/* helper function to actually process a single delayed ref entry */
static int run_one_delayed_ref(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
struct btrfs_delayed_ref_node *node,
struct btrfs_delayed_extent_op *extent_op,
int insert_reserved)
{
int ret = 0;
if (trans->aborted) {
if (insert_reserved)
btrfs_pin_extent(fs_info, node->bytenr,
node->num_bytes, 1);
return 0;
}
if (btrfs_delayed_ref_is_head(node)) {
struct btrfs_delayed_ref_head *head;
/*
* we've hit the end of the chain and we were supposed
* to insert this extent into the tree. But, it got
* deleted before we ever needed to insert it, so all
* we have to do is clean up the accounting
*/
BUG_ON(extent_op);
head = btrfs_delayed_node_to_head(node);
trace_run_delayed_ref_head(fs_info, node, head, node->action);
if (head->total_ref_mod < 0) {
struct btrfs_block_group_cache *cache;
cache = btrfs_lookup_block_group(fs_info, node->bytenr);
ASSERT(cache);
percpu_counter_add(&cache->space_info->total_bytes_pinned,
-node->num_bytes);
btrfs_put_block_group(cache);
}
if (insert_reserved) {
btrfs_pin_extent(fs_info, node->bytenr,
node->num_bytes, 1);
if (head->is_data) {
ret = btrfs_del_csums(trans, fs_info,
node->bytenr,
node->num_bytes);
}
}
/* Also free its reserved qgroup space */
btrfs_qgroup_free_delayed_ref(fs_info, head->qgroup_ref_root,
head->qgroup_reserved);
return ret;
}
if (node->type == BTRFS_TREE_BLOCK_REF_KEY ||
node->type == BTRFS_SHARED_BLOCK_REF_KEY)
ret = run_delayed_tree_ref(trans, fs_info, node, extent_op,
insert_reserved);
else if (node->type == BTRFS_EXTENT_DATA_REF_KEY ||
node->type == BTRFS_SHARED_DATA_REF_KEY)
ret = run_delayed_data_ref(trans, fs_info, node, extent_op,
insert_reserved);
else
BUG();
return ret;
}
static inline struct btrfs_delayed_ref_node *
select_delayed_ref(struct btrfs_delayed_ref_head *head)
{
struct btrfs_delayed_ref_node *ref;
if (list_empty(&head->ref_list))
return NULL;
/*
* Select a delayed ref of type BTRFS_ADD_DELAYED_REF first.
* This is to prevent a ref count from going down to zero, which deletes
* the extent item from the extent tree, when there still are references
* to add, which would fail because they would not find the extent item.
*/
if (!list_empty(&head->ref_add_list))
return list_first_entry(&head->ref_add_list,
struct btrfs_delayed_ref_node, add_list);
ref = list_first_entry(&head->ref_list, struct btrfs_delayed_ref_node,
list);
ASSERT(list_empty(&ref->add_list));
return ref;
}
/*
* Returns 0 on success or if called with an already aborted transaction.
* Returns -ENOMEM or -EIO on failure and will abort the transaction.
*/
static noinline int __btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info,
unsigned long nr)
{
struct btrfs_delayed_ref_root *delayed_refs;
struct btrfs_delayed_ref_node *ref;
struct btrfs_delayed_ref_head *locked_ref = NULL;
struct btrfs_delayed_extent_op *extent_op;
ktime_t start = ktime_get();
int ret;
unsigned long count = 0;
unsigned long actual_count = 0;
int must_insert_reserved = 0;
delayed_refs = &trans->transaction->delayed_refs;
while (1) {
if (!locked_ref) {
if (count >= nr)
break;
spin_lock(&delayed_refs->lock);
locked_ref = btrfs_select_ref_head(trans);
if (!locked_ref) {
spin_unlock(&delayed_refs->lock);
break;
}
/* grab the lock that says we are going to process
* all the refs for this head */
ret = btrfs_delayed_ref_lock(trans, locked_ref);
spin_unlock(&delayed_refs->lock);
/*
* we may have dropped the spin lock to get the head
* mutex lock, and that might have given someone else
* time to free the head. If that's true, it has been
* removed from our list and we can move on.
*/
if (ret == -EAGAIN) {
locked_ref = NULL;
count++;
continue;
}
}
/*
* We need to try and merge add/drops of the same ref since we
* can run into issues with relocate dropping the implicit ref
* and then it being added back again before the drop can
* finish. If we merged anything we need to re-loop so we can
* get a good ref.
* Or we can get node references of the same type that weren't
* merged when created due to bumps in the tree mod seq, and
* we need to merge them to prevent adding an inline extent
* backref before dropping it (triggering a BUG_ON at
* insert_inline_extent_backref()).
*/
spin_lock(&locked_ref->lock);
btrfs_merge_delayed_refs(trans, fs_info, delayed_refs,
locked_ref);
/*
* locked_ref is the head node, so we have to go one
* node back for any delayed ref updates
*/
ref = select_delayed_ref(locked_ref);
if (ref && ref->seq &&
btrfs_check_delayed_seq(fs_info, delayed_refs, ref->seq)) {
spin_unlock(&locked_ref->lock);
spin_lock(&delayed_refs->lock);
locked_ref->processing = 0;
delayed_refs->num_heads_ready++;
spin_unlock(&delayed_refs->lock);
btrfs_delayed_ref_unlock(locked_ref);
locked_ref = NULL;
cond_resched();
count++;
continue;
}
/*
* record the must insert reserved flag before we
* drop the spin lock.
*/
must_insert_reserved = locked_ref->must_insert_reserved;
locked_ref->must_insert_reserved = 0;
extent_op = locked_ref->extent_op;
locked_ref->extent_op = NULL;
if (!ref) {
/* All delayed refs have been processed, Go ahead
* and send the head node to run_one_delayed_ref,
* so that any accounting fixes can happen
*/
ref = &locked_ref->node;
if (extent_op && must_insert_reserved) {
btrfs_free_delayed_extent_op(extent_op);
extent_op = NULL;
}
if (extent_op) {
spin_unlock(&locked_ref->lock);
ret = run_delayed_extent_op(trans, fs_info,
ref, extent_op);
btrfs_free_delayed_extent_op(extent_op);
if (ret) {
/*
* Need to reset must_insert_reserved if
* there was an error so the abort stuff
* can cleanup the reserved space
* properly.
*/
if (must_insert_reserved)
locked_ref->must_insert_reserved = 1;
spin_lock(&delayed_refs->lock);
locked_ref->processing = 0;
delayed_refs->num_heads_ready++;
spin_unlock(&delayed_refs->lock);
btrfs_debug(fs_info,
"run_delayed_extent_op returned %d",
ret);
btrfs_delayed_ref_unlock(locked_ref);
return ret;
}
continue;
}
/*
* Need to drop our head ref lock and re-acquire the
* delayed ref lock and then re-check to make sure
* nobody got added.
*/
spin_unlock(&locked_ref->lock);
spin_lock(&delayed_refs->lock);
spin_lock(&locked_ref->lock);
if (!list_empty(&locked_ref->ref_list) ||
locked_ref->extent_op) {
spin_unlock(&locked_ref->lock);
spin_unlock(&delayed_refs->lock);
continue;
}
ref->in_tree = 0;
delayed_refs->num_heads--;
rb_erase(&locked_ref->href_node,
&delayed_refs->href_root);
spin_unlock(&delayed_refs->lock);
} else {
actual_count++;
ref->in_tree = 0;
list_del(&ref->list);
if (!list_empty(&ref->add_list))
list_del(&ref->add_list);
}
atomic_dec(&delayed_refs->num_entries);
if (!btrfs_delayed_ref_is_head(ref)) {
/*
* when we play the delayed ref, also correct the
* ref_mod on head
*/
switch (ref->action) {
case BTRFS_ADD_DELAYED_REF:
case BTRFS_ADD_DELAYED_EXTENT:
locked_ref->node.ref_mod -= ref->ref_mod;
break;
case BTRFS_DROP_DELAYED_REF:
locked_ref->node.ref_mod += ref->ref_mod;
break;
default:
WARN_ON(1);
}
}
spin_unlock(&locked_ref->lock);
ret = run_one_delayed_ref(trans, fs_info, ref, extent_op,
must_insert_reserved);
btrfs_free_delayed_extent_op(extent_op);
if (ret) {
spin_lock(&delayed_refs->lock);
locked_ref->processing = 0;
delayed_refs->num_heads_ready++;
spin_unlock(&delayed_refs->lock);
btrfs_delayed_ref_unlock(locked_ref);
btrfs_put_delayed_ref(ref);
btrfs_debug(fs_info, "run_one_delayed_ref returned %d",
ret);
return ret;
}
/*
* If this node is a head, that means all the refs in this head
* have been dealt with, and we will pick the next head to deal
* with, so we must unlock the head and drop it from the cluster
* list before we release it.
*/
if (btrfs_delayed_ref_is_head(ref)) {
if (locked_ref->is_data &&
locked_ref->total_ref_mod < 0) {
spin_lock(&delayed_refs->lock);
delayed_refs->pending_csums -= ref->num_bytes;
spin_unlock(&delayed_refs->lock);
}
btrfs_delayed_ref_unlock(locked_ref);
locked_ref = NULL;
}
btrfs_put_delayed_ref(ref);
count++;
cond_resched();
}
/*
* We don't want to include ref heads since we can have empty ref heads
* and those will drastically skew our runtime down since we just do
* accounting, no actual extent tree updates.
*/
if (actual_count > 0) {
u64 runtime = ktime_to_ns(ktime_sub(ktime_get(), start));
u64 avg;
/*
* We weigh the current average higher than our current runtime
* to avoid large swings in the average.
*/
spin_lock(&delayed_refs->lock);
avg = fs_info->avg_delayed_ref_runtime * 3 + runtime;
fs_info->avg_delayed_ref_runtime = avg >> 2; /* div by 4 */
spin_unlock(&delayed_refs->lock);
}
return 0;
}
#ifdef SCRAMBLE_DELAYED_REFS
/*
* Normally delayed refs get processed in ascending bytenr order. This
* correlates in most cases to the order added. To expose dependencies on this
* order, we start to process the tree in the middle instead of the beginning
*/
static u64 find_middle(struct rb_root *root)
{
struct rb_node *n = root->rb_node;
struct btrfs_delayed_ref_node *entry;
int alt = 1;
u64 middle;
u64 first = 0, last = 0;
n = rb_first(root);
if (n) {
entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
first = entry->bytenr;
}
n = rb_last(root);
if (n) {
entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
last = entry->bytenr;
}
n = root->rb_node;
while (n) {
entry = rb_entry(n, struct btrfs_delayed_ref_node, rb_node);
WARN_ON(!entry->in_tree);
middle = entry->bytenr;
if (alt)
n = n->rb_left;
else
n = n->rb_right;
alt = 1 - alt;
}
return middle;
}
#endif
static inline u64 heads_to_leaves(struct btrfs_fs_info *fs_info, u64 heads)
{
u64 num_bytes;
num_bytes = heads * (sizeof(struct btrfs_extent_item) +
sizeof(struct btrfs_extent_inline_ref));
if (!btrfs_fs_incompat(fs_info, SKINNY_METADATA))
num_bytes += heads * sizeof(struct btrfs_tree_block_info);
/*
* We don't ever fill up leaves all the way so multiply by 2 just to be
* closer to what we're really going to want to use.
*/
return div_u64(num_bytes, BTRFS_LEAF_DATA_SIZE(fs_info));
}
/*
* Takes the number of bytes to be csumm'ed and figures out how many leaves it
* would require to store the csums for that many bytes.
*/
u64 btrfs_csum_bytes_to_leaves(struct btrfs_fs_info *fs_info, u64 csum_bytes)
{
u64 csum_size;
u64 num_csums_per_leaf;
u64 num_csums;
csum_size = BTRFS_MAX_ITEM_SIZE(fs_info);
num_csums_per_leaf = div64_u64(csum_size,
(u64)btrfs_super_csum_size(fs_info->super_copy));
num_csums = div64_u64(csum_bytes, fs_info->sectorsize);
num_csums += num_csums_per_leaf - 1;
num_csums = div64_u64(num_csums, num_csums_per_leaf);
return num_csums;
}
int btrfs_check_space_for_delayed_refs(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info)
{
struct btrfs_block_rsv *global_rsv;
u64 num_heads = trans->transaction->delayed_refs.num_heads_ready;
u64 csum_bytes = trans->transaction->delayed_refs.pending_csums;
u64 num_dirty_bgs = trans->transaction->num_dirty_bgs;
u64 num_bytes, num_dirty_bgs_bytes;
int ret = 0;
num_bytes = btrfs_calc_trans_metadata_size(fs_info, 1);
num_heads = heads_to_leaves(fs_info, num_heads);
if (num_heads > 1)
num_bytes += (num_heads - 1) * fs_info->nodesize;
num_bytes <<= 1;
num_bytes += btrfs_csum_bytes_to_leaves(fs_info, csum_bytes) *
fs_info->nodesize;
num_dirty_bgs_bytes = btrfs_calc_trans_metadata_size(fs_info,
num_dirty_bgs);
global_rsv = &fs_info->global_block_rsv;
/*
* If we can't allocate any more chunks lets make sure we have _lots_ of
* wiggle room since running delayed refs can create more delayed refs.
*/
if (global_rsv->space_info->full) {
num_dirty_bgs_bytes <<= 1;
num_bytes <<= 1;
}
spin_lock(&global_rsv->lock);
if (global_rsv->reserved <= num_bytes + num_dirty_bgs_bytes)
ret = 1;
spin_unlock(&global_rsv->lock);
return ret;
}
int btrfs_should_throttle_delayed_refs(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info)
{
u64 num_entries =
atomic_read(&trans->transaction->delayed_refs.num_entries);
u64 avg_runtime;
u64 val;
smp_mb();
avg_runtime = fs_info->avg_delayed_ref_runtime;
val = num_entries * avg_runtime;
if (val >= NSEC_PER_SEC)
return 1;
if (val >= NSEC_PER_SEC / 2)
return 2;
return btrfs_check_space_for_delayed_refs(trans, fs_info);
}
struct async_delayed_refs {
struct btrfs_root *root;
u64 transid;
int count;
int error;
int sync;
struct completion wait;
struct btrfs_work work;
};
static inline struct async_delayed_refs *
to_async_delayed_refs(struct btrfs_work *work)
{
return container_of(work, struct async_delayed_refs, work);
}
static void delayed_ref_async_start(struct btrfs_work *work)
{
struct async_delayed_refs *async = to_async_delayed_refs(work);
struct btrfs_trans_handle *trans;
struct btrfs_fs_info *fs_info = async->root->fs_info;
int ret;
/* if the commit is already started, we don't need to wait here */
if (btrfs_transaction_blocked(fs_info))
goto done;
trans = btrfs_join_transaction(async->root);
if (IS_ERR(trans)) {
async->error = PTR_ERR(trans);
goto done;
}
/*
* trans->sync means that when we call end_transaction, we won't
* wait on delayed refs
*/
trans->sync = true;
/* Don't bother flushing if we got into a different transaction */
if (trans->transid > async->transid)
goto end;
ret = btrfs_run_delayed_refs(trans, fs_info, async->count);
if (ret)
async->error = ret;
end:
ret = btrfs_end_transaction(trans);
if (ret && !async->error)
async->error = ret;
done:
if (async->sync)
complete(&async->wait);
else
kfree(async);
}
int btrfs_async_run_delayed_refs(struct btrfs_fs_info *fs_info,
unsigned long count, u64 transid, int wait)
{
struct async_delayed_refs *async;
int ret;
async = kmalloc(sizeof(*async), GFP_NOFS);
if (!async)
return -ENOMEM;
async->root = fs_info->tree_root;
async->count = count;
async->error = 0;
async->transid = transid;
if (wait)
async->sync = 1;
else
async->sync = 0;
init_completion(&async->wait);
btrfs_init_work(&async->work, btrfs_extent_refs_helper,
delayed_ref_async_start, NULL, NULL);
btrfs_queue_work(fs_info->extent_workers, &async->work);
if (wait) {
wait_for_completion(&async->wait);
ret = async->error;
kfree(async);
return ret;
}
return 0;
}
/*
* this starts processing the delayed reference count updates and
* extent insertions we have queued up so far. count can be
* 0, which means to process everything in the tree at the start
* of the run (but not newly added entries), or it can be some target
* number you'd like to process.
*
* Returns 0 on success or if called with an aborted transaction
* Returns <0 on error and aborts the transaction
*/
int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info, unsigned long count)
{
struct rb_node *node;
struct btrfs_delayed_ref_root *delayed_refs;
struct btrfs_delayed_ref_head *head;
int ret;
int run_all = count == (unsigned long)-1;
bool can_flush_pending_bgs = trans->can_flush_pending_bgs;
/* We'll clean this up in btrfs_cleanup_transaction */
if (trans->aborted)
return 0;
if (test_bit(BTRFS_FS_CREATING_FREE_SPACE_TREE, &fs_info->flags))
return 0;
delayed_refs = &trans->transaction->delayed_refs;
if (count ==