btrfs: defrag: don't try to defrag extents which are under writeback

Once we start writeback (have called btrfs_run_delalloc_range()), we
allocate an extent, create an extent map point to that extent, with a
generation of (u64)-1, created the ordered extent and then clear the
DELALLOC bit from the range in the inode's io tree.

Such extent map can pass the first call of defrag_collect_targets(), as
its generation is (u64)-1, meets any possible minimal generation check.
And the range will not have DELALLOC bit, also passing the DELALLOC bit
check.

It will only be re-checked in the second call of
defrag_collect_targets(), which will wait for writeback.

But at that stage we have already spent our time waiting for some IO we
may or may not want to defrag.

Let's reject such extents early so we won't waste our time.

CC: stable@vger.kernel.org # 5.16
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

diff --git a/fs/btrfs/ioctl.c b/fs/btrfs/ioctl.c
index b51c8b7..9013656 100644 (file)
--- a/fs/btrfs/ioctl.c
+++ b/fs/btrfs/ioctl.c
@@ -1210,6 +1210,10 @@ static int defrag_collect_targets(struct btrfs_inode *inode,
                if (em->generation < newer_than)
                        goto next;
 
+               /* This em is under writeback, no need to defrag */
+               if (em->generation == (u64)-1)
+                       goto next;
+
                /*
                 * Our start offset might be in the middle of an existing extent
                 * map, so take that into account.