Skip to contentSkip to conversation

Btrfs: fix race between scrub and block group deletion

PATCH New [email protected] 1 participant 1 message Nov 19, 2015 Btrfs file system

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 Btrfs: fix race between scrub and block group deletion

From: Filipe Manana <[email protected]>

Scrub can race with the cleaner kthread deleting block groups that are unused (and with relocation too) leading to a failure with error -EINVAL that gets returned to user space.

The following diagram illustrates how it happens:

CPU 1 CPU 2

cleaner kthread
btrfs_delete_unused_bgs()

gets block group X from
fs_info->unused_bgs

sets block group to RO

btrfs_remove_chunk(bg X)

deletes device extents

scrub_enumerate_chunks()

searches device tree using its commit root

finds device extent for block group X

gets block group X from the tree fs_info->block_group_cache_tree (via btrfs_lookup_block_group())

sets bg X to RO (again)

btrfs_remove_block_group(bg X)

deletes block group from
fs_info->block_group_cache_tree

removes extent map from
fs_info->mapping_tree

scrub_chunk(offset X)

searches fs_info->mapping_tree for extent map starting at offset X

--> doesn't find any such
extent map
--> returns -EINVAL and scrub
errors out to userspace with -EINVAL

Fix this by dealing with an extent map lookup failure as an indicator of block group deletion.
Issue reproduced with fstest btrfs/071.

---
 fs/btrfs/scrub.c | 20 ++++++++++++++++----
 1 file changed, 16 insertions(+), 4 deletions(-)
fs/btrfs/scrub.c +16 -4
index 2907a77..28fbd83 100644
@@ -3432,7 +3432,9 @@ out:
 static noinline_for_stack int scrub_chunk(struct scrub_ctx *sctx,
 					  struct btrfs_device *scrub_dev,
 					  u64 chunk_offset, u64 length,
-					  u64 dev_offset, int is_dev_replace)
+					  u64 dev_offset,
+					  struct btrfs_block_group_cache *cache,
+					  int is_dev_replace)
 {
 	struct btrfs_mapping_tree *map_tree =
 		&sctx->dev_root->fs_info->mapping_tree;
@@ -3445,8 +3447,18 @@ static noinline_for_stack int scrub_chunk(struct scrub_ctx *sctx,
 	em = lookup_extent_mapping(&map_tree->map_tree, chunk_offset, 1);
 	read_unlock(&map_tree->map_tree.lock);
 
-	if (!em)
-		return -EINVAL;
+	if (!em) {
+		/*
+		 * Might have been an unused block group deleted by the cleaner
+		 * kthread or relocation.
+		 */
+		spin_lock(&cache->lock);
+		if (!cache->removed)
+			ret = -EINVAL;
+		spin_unlock(&cache->lock);
+
+		return ret;
+	}
 
 	map = (struct map_lookup *)em->bdev;
 	if (em->start != chunk_offset)
@@ -3577,7 +3589,7 @@ int scrub_enumerate_chunks(struct scrub_ctx *sctx,
 		dev_replace->cursor_left = found_key.offset;
 		dev_replace->item_needs_writeback = 1;
 		ret = scrub_chunk(sctx, scrub_dev, chunk_offset, length,
-				  found_key.offset, is_dev_replace);
+				  found_key.offset, cache, is_dev_replace);
 
 		/*
 		 * flush, submit all pending read and write bios, afterwards
Signature · 1 line
2.1.3