Btrfs: remove empty block groups automatically

One problem that has plagued us is that a user will use up all of his space with
data, remove a bunch of that data, and then try to create a bunch of small files
and run out of space.  This happens because all the chunks were allocated for
data since the metadata requirements were so low.  But now there's a bunch of
empty data block groups and not enough metadata space to do anything.  This
patch solves this problem by automatically deleting empty block groups.  If we
notice the used count go down to 0 when deleting or on mount notice that a block
group has a used count of 0 then we will queue it to be deleted.

When the cleaner thread runs we will double check to make sure the block group
is still empty and then we will delete it.  This patch has the side effect of no
longer having a bunch of BUG_ON()'s in the chunk delete code, which will be
helpful for both this and relocate.  Thanks,

Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: Chris Mason <clm@fb.com>

1 files changed, 134 insertions, 7 deletions

diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index b30ddb49cfab..28a27d5f02d3 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -5433,6 +5433,20 @@ static int update_block_group(struct btrfs_root *root,
 			spin_unlock(&cache->space_info->lock);
 		} else {
 			old_val -= num_bytes;
+
+			/*
+			 * No longer have used bytes in this block group, queue
+			 * it for deletion.
+			 */
+			if (old_val == 0) {
+				spin_lock(&info->unused_bgs_lock);
+				if (list_empty(&cache->bg_list)) {
+					btrfs_get_block_group(cache);
+					list_add_tail(&cache->bg_list,
+						      &info->unused_bgs);
+				}
+				spin_unlock(&info->unused_bgs_lock);
+			}
 			btrfs_set_block_group_used(&cache->item, old_val);
 			cache->pinned += num_bytes;
 			cache->space_info->bytes_pinned += num_bytes;
@@ -8855,6 +8869,16 @@ int btrfs_free_block_groups(struct btrfs_fs_info *info)
 	}
 	up_write(&info->commit_root_sem);
 
+	spin_lock(&info->unused_bgs_lock);
+	while (!list_empty(&info->unused_bgs)) {
+		block_group = list_first_entry(&info->unused_bgs,
+					       struct btrfs_block_group_cache,
+					       bg_list);
+		list_del_init(&block_group->bg_list);
+		btrfs_put_block_group(block_group);
+	}
+	spin_unlock(&info->unused_bgs_lock);
+
 	spin_lock(&info->block_group_cache_lock);
 	while ((n = rb_last(&info->block_group_cache_tree)) != NULL) {
 		block_group = rb_entry(n, struct btrfs_block_group_cache,
@@ -8989,7 +9013,7 @@ btrfs_create_block_group_cache(struct btrfs_root *root, u64 start, u64 size)
 	init_rwsem(&cache->data_rwsem);
 	INIT_LIST_HEAD(&cache->list);
 	INIT_LIST_HEAD(&cache->cluster_list);
-	INIT_LIST_HEAD(&cache->new_bg_list);
+	INIT_LIST_HEAD(&cache->bg_list);
 	btrfs_init_free_space_ctl(cache);
 
 	return cache;
@@ -9130,8 +9154,18 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 		__link_block_group(space_info, cache);
 
 		set_avail_alloc_bits(root->fs_info, cache->flags);
-		if (btrfs_chunk_readonly(root, cache->key.objectid))
+		if (btrfs_chunk_readonly(root, cache->key.objectid)) {
 			set_block_group_ro(cache, 1);
+		} else if (btrfs_block_group_used(&cache->item) == 0) {
+			spin_lock(&info->unused_bgs_lock);
+			/* Should always be true but just in case. */
+			if (list_empty(&cache->bg_list)) {
+				btrfs_get_block_group(cache);
+				list_add_tail(&cache->bg_list,
+					      &info->unused_bgs);
+			}
+			spin_unlock(&info->unused_bgs_lock);
+		}
 	}
 
 	list_for_each_entry_rcu(space_info, &root->fs_info->space_info, list) {
@@ -9172,10 +9206,8 @@ void btrfs_create_pending_block_groups(struct btrfs_trans_handle *trans,
 	struct btrfs_key key;
 	int ret = 0;
 
-	list_for_each_entry_safe(block_group, tmp, &trans->new_bgs,
-				 new_bg_list) {
-		list_del_init(&block_group->new_bg_list);
-
+	list_for_each_entry_safe(block_group, tmp, &trans->new_bgs, bg_list) {
+		list_del_init(&block_group->bg_list);
 		if (ret)
 			continue;
 
@@ -9261,7 +9293,7 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans,
 
 	__link_block_group(cache->space_info, cache);
 
-	list_add_tail(&cache->new_bg_list, &trans->new_bgs);
+	list_add_tail(&cache->bg_list, &trans->new_bgs);
 
 	set_avail_alloc_bits(extent_root->fs_info, type);
 
@@ -9430,6 +9462,101 @@ out:
 	return ret;
 }
 
+/*
+ * Process the unused_bgs list and remove any that don't have any allocated
+ * space inside of them.
+ */
+void btrfs_delete_unused_bgs(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_block_group_cache *block_group;
+	struct btrfs_space_info *space_info;
+	struct btrfs_root *root = fs_info->extent_root;
+	struct btrfs_trans_handle *trans;
+	int ret = 0;
+
+	if (!fs_info->open)
+		return;
+
+	spin_lock(&fs_info->unused_bgs_lock);
+	while (!list_empty(&fs_info->unused_bgs)) {
+		u64 start, end;
+
+		block_group = list_first_entry(&fs_info->unused_bgs,
+					       struct btrfs_block_group_cache,
+					       bg_list);
+		space_info = block_group->space_info;
+		list_del_init(&block_group->bg_list);
+		if (ret || btrfs_mixed_space_info(space_info)) {
+			btrfs_put_block_group(block_group);
+			continue;
+		}
+		spin_unlock(&fs_info->unused_bgs_lock);
+
+		/* Don't want to race with allocators so take the groups_sem */
+		down_write(&space_info->groups_sem);
+		spin_lock(&block_group->lock);
+		if (block_group->reserved ||
+		    btrfs_block_group_used(&block_group->item) ||
+		    block_group->ro) {
+			/*
+			 * We want to bail if we made new allocations or have
+			 * outstanding allocations in this block group.  We do
+			 * the ro check in case balance is currently acting on
+			 * this block group.
+			 */
+			spin_unlock(&block_group->lock);
+			up_write(&space_info->groups_sem);
+			goto next;
+		}
+		spin_unlock(&block_group->lock);
+
+		/* We don't want to force the issue, only flip if it's ok. */
+		ret = set_block_group_ro(block_group, 0);
+		up_write(&space_info->groups_sem);
+		if (ret < 0) {
+			ret = 0;
+			goto next;
+		}
+
+		/*
+		 * Want to do this before we do anything else so we can recover
+		 * properly if we fail to join the transaction.
+		 */
+		trans = btrfs_join_transaction(root);
+		if (IS_ERR(trans)) {
+			btrfs_set_block_group_rw(root, block_group);
+			ret = PTR_ERR(trans);
+			goto next;
+		}
+
+		/*
+		 * We could have pending pinned extents for this block group,
+		 * just delete them, we don't care about them anymore.
+		 */
+		start = block_group->key.objectid;
+		end = start + block_group->key.offset - 1;
+		clear_extent_bits(&fs_info->freed_extents[0], start, end,
+				  EXTENT_DIRTY, GFP_NOFS);
+		clear_extent_bits(&fs_info->freed_extents[1], start, end,
+				  EXTENT_DIRTY, GFP_NOFS);
+
+		/* Reset pinned so btrfs_put_block_group doesn't complain */
+		block_group->pinned = 0;
+
+		/*
+		 * Btrfs_remove_chunk will abort the transaction if things go
+		 * horribly wrong.
+		 */
+		ret = btrfs_remove_chunk(trans, root,
+					 block_group->key.objectid);
+		btrfs_end_transaction(trans, root);
+next:
+		btrfs_put_block_group(block_group);
+		spin_lock(&fs_info->unused_bgs_lock);
+	}
+	spin_unlock(&fs_info->unused_bgs_lock);
+}
+
 int btrfs_init_space_info(struct btrfs_fs_info *fs_info)
 {
 	struct btrfs_space_info *space_info;