cache->cached = BTRFS_CACHE_NO;
} else {
cache->cached = BTRFS_CACHE_STARTED;
+ cache->has_caching_ctl = 1;
}
}
spin_unlock(&cache->lock);
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);
struct btrfs_block_group_cache *cache)
{
struct rb_node *node;
+
spin_lock(&root->fs_info->block_group_cache_lock);
+
+ /* If our block group was removed, we need a full search. */
+ if (RB_EMPTY_NODE(&cache->cache_node)) {
+ const u64 next_bytenr = cache->key.objectid + cache->key.offset;
+
+ spin_unlock(&root->fs_info->block_group_cache_lock);
+ btrfs_put_block_group(cache);
+ cache = btrfs_lookup_first_block_group(root->fs_info,
+ next_bytenr);
+ return cache;
+ }
node = rb_next(&cache->cache_node);
btrfs_put_block_group(cache);
if (node) {
found->chunk_alloc = 0;
found->flush = 0;
init_waitqueue_head(&found->wait);
+ INIT_LIST_HEAD(&found->ro_bgs);
ret = kobject_init_and_add(&found->kobj, &space_info_ktype,
info->space_info_kobj, "%s",
min_allocable_bytes <= sinfo->total_bytes) {
sinfo->bytes_readonly += num_bytes;
cache->ro = 1;
+ list_add_tail(&cache->ro_list, &sinfo->ro_bgs);
ret = 0;
}
out:
/*
* helper to account the unused space of all the readonly block group in the
- * list. takes mirrors into account.
+ * space_info. takes mirrors into account.
*/
-static u64 __btrfs_get_ro_block_group_free_space(struct list_head *groups_list)
+u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo)
{
struct btrfs_block_group_cache *block_group;
u64 free_bytes = 0;
int factor;
- list_for_each_entry(block_group, groups_list, list) {
+ /* It's df, we don't care if it's racey */
+ if (list_empty(&sinfo->ro_bgs))
+ return 0;
+
+ spin_lock(&sinfo->lock);
+ list_for_each_entry(block_group, &sinfo->ro_bgs, ro_list) {
spin_lock(&block_group->lock);
if (!block_group->ro) {
spin_unlock(&block_group->lock);
}
-
- return free_bytes;
-}
-
-/*
- * helper to account the unused space of all the readonly block group in the
- * space_info. takes mirrors into account.
- */
-u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo)
-{
- int i;
- u64 free_bytes = 0;
-
- spin_lock(&sinfo->lock);
-
- for (i = 0; i < BTRFS_NR_RAID_TYPES; i++)
- if (!list_empty(&sinfo->block_groups[i]))
- free_bytes += __btrfs_get_ro_block_group_free_space(
- &sinfo->block_groups[i]);
-
spin_unlock(&sinfo->lock);
return free_bytes;
cache->bytes_super - btrfs_block_group_used(&cache->item);
sinfo->bytes_readonly -= num_bytes;
cache->ro = 0;
+ list_del_init(&cache->ro_list);
spin_unlock(&cache->lock);
spin_unlock(&sinfo->lock);
}
INIT_LIST_HEAD(&cache->list);
INIT_LIST_HEAD(&cache->cluster_list);
INIT_LIST_HEAD(&cache->bg_list);
+ INIT_LIST_HEAD(&cache->ro_list);
btrfs_init_free_space_ctl(cache);
return cache;
int ret;
int index;
int factor;
+ struct btrfs_caching_control *caching_ctl = NULL;
root = root->fs_info->extent_root;
spin_lock(&root->fs_info->block_group_cache_lock);
rb_erase(&block_group->cache_node,
&root->fs_info->block_group_cache_tree);
+ RB_CLEAR_NODE(&block_group->cache_node);
if (root->fs_info->first_logical_byte == block_group->key.objectid)
root->fs_info->first_logical_byte = (u64)-1;
* are still on the list after taking the semaphore
*/
list_del_init(&block_group->list);
+ list_del_init(&block_group->ro_list);
if (list_empty(&block_group->space_info->block_groups[index])) {
kobj = block_group->space_info->block_group_kobjs[index];
block_group->space_info->block_group_kobjs[index] = NULL;
kobject_put(kobj);
}
+ if (block_group->has_caching_ctl)
+ caching_ctl = get_caching_control(block_group);
if (block_group->cached == BTRFS_CACHE_STARTED)
wait_block_group_cache_done(block_group);
+ if (block_group->has_caching_ctl) {
+ down_write(&root->fs_info->commit_root_sem);
+ if (!caching_ctl) {
+ struct btrfs_caching_control *ctl;
+
+ list_for_each_entry(ctl,
+ &root->fs_info->caching_block_groups, list)
+ if (ctl->block_group == block_group) {
+ caching_ctl = ctl;
+ atomic_inc(&caching_ctl->count);
+ break;
+ }
+ }
+ if (caching_ctl)
+ list_del_init(&caching_ctl->list);
+ up_write(&root->fs_info->commit_root_sem);
+ if (caching_ctl) {
+ /* Once for the caching bgs list and once for us. */
+ put_caching_control(caching_ctl);
+ put_caching_control(caching_ctl);
+ }
+ }
btrfs_remove_free_space_cache(block_group);
*/
start = block_group->key.objectid;
end = start + block_group->key.offset - 1;
- clear_extent_bits(&fs_info->freed_extents[0], start, end,
+ ret = clear_extent_bits(&fs_info->freed_extents[0], start, end,
EXTENT_DIRTY, GFP_NOFS);
- clear_extent_bits(&fs_info->freed_extents[1], start, end,
+ if (ret) {
+ btrfs_set_block_group_rw(root, block_group);
+ goto end_trans;
+ }
+ ret = clear_extent_bits(&fs_info->freed_extents[1], start, end,
EXTENT_DIRTY, GFP_NOFS);
+ if (ret) {
+ btrfs_set_block_group_rw(root, block_group);
+ goto end_trans;
+ }
/* Reset pinned so btrfs_put_block_group doesn't complain */
block_group->pinned = 0;
*/
ret = btrfs_remove_chunk(trans, root,
block_group->key.objectid);
+end_trans:
btrfs_end_transaction(trans, root);
next:
btrfs_put_block_group(block_group);
}
/*
- * btrfs_{start,end}_write() is similar to mnt_{want, drop}_write(),
- * they are used to prevent the some tasks writing data into the page cache
- * by nocow before the subvolume is snapshoted, but flush the data into
- * the disk after the snapshot creation.
+ * btrfs_{start,end}_write_no_snapshoting() are similar to
+ * mnt_{want,drop}_write(), they are used to prevent some tasks from writing
+ * data into the page cache through nocow before the subvolume is snapshoted,
+ * but flush the data into disk after the snapshot creation, or to prevent
+ * operations while snapshoting is ongoing and that cause the snapshot to be
+ * inconsistent (writes followed by expanding truncates for example).
*/
-void btrfs_end_nocow_write(struct btrfs_root *root)
+void btrfs_end_write_no_snapshoting(struct btrfs_root *root)
{
percpu_counter_dec(&root->subv_writers->counter);
/*
wake_up(&root->subv_writers->wait);
}
-int btrfs_start_nocow_write(struct btrfs_root *root)
+int btrfs_start_write_no_snapshoting(struct btrfs_root *root)
{
if (atomic_read(&root->will_be_snapshoted))
return 0;
*/
smp_mb();
if (atomic_read(&root->will_be_snapshoted)) {
- btrfs_end_nocow_write(root);
+ btrfs_end_write_no_snapshoting(root);
return 0;
}
return 1;