Btrfs: fix racy system chunk allocation when setting block group ro

If while setting a block group read-only we end up allocating a system
chunk, through check_system_chunk(), we were not doing it while holding
the chunk mutex which is a problem if a concurrent chunk allocation is
happening, through do_chunk_alloc(), as it means both block groups can
end up using the same logical addresses and physical regions in the
device(s). So make sure we hold the chunk mutex.

Cc: stable@vger.kernel.org  # 4.0+
Fixes: 2f0810880f08 ("btrfs: delete chunk allocation attemp when
                      setting block group ro")

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Chris Mason <clm@fb.com>

2 files changed, 3 insertions, 0 deletions

diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index 7effed6f2fa6..45e3f086790b 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -8842,7 +8842,9 @@ again:
 out:
         if (cache->flags & BTRFS_BLOCK_GROUP_SYSTEM) {
                 alloc_flags = update_block_group_flags(root, cache->flags);
+                lock_chunks(root->fs_info->chunk_root);
                 check_system_chunk(trans, root, alloc_flags);
+                unlock_chunks(root->fs_info->chunk_root);
         }
         mutex_unlock(&root->fs_info->ro_block_group_mutex);
 
diff --git a/fs/btrfs/volumes.c b/fs/btrfs/volumes.c
index 96aebf3bcd5b..174f5e1e00ab 100644
--- a/fs/btrfs/volumes.c
+++ b/fs/btrfs/volumes.c
@@ -4625,6 +4625,7 @@ int btrfs_alloc_chunk(struct btrfs_trans_handle *trans,
 {
         u64 chunk_offset;
 
+        ASSERT(mutex_is_locked(&extent_root->fs_info->chunk_mutex));
         chunk_offset = find_next_chunk(extent_root->fs_info);
         return __btrfs_alloc_chunk(trans, extent_root, chunk_offset, type);
 }