md/raid10: fix deadlock with unaligned read during resync

If the 'bio_split' path in raid10-read is used while
resync/recovery is happening it is possible to deadlock.
Fix this be elevating ->nr_waiting for the duration of both
parts of the split request.

This fixes a bug that has been present since 2.6.22
but has only started manifesting recently for unknown reasons.
It is suitable for and -stable since then.

Reported-by:  Justin Bronder <jsbronder@gentoo.org>
Tested-by:  Justin Bronder <jsbronder@gentoo.org>
Signed-off-by: NeilBrown <neilb@suse.de>
Cc: stable@kernel.org

1 files changed, 18 insertions, 0 deletions

diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index 42e64e4e5e25..d1d689126346 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -825,11 +825,29 @@ static int make_request(mddev_t *mddev, struct bio * bio)
                  */
                 bp = bio_split(bio,
                                chunk_sects - (bio->bi_sector & (chunk_sects - 1)) );
+
+                /* Each of these 'make_request' calls will call 'wait_barrier'.
+                 * If the first succeeds but the second blocks due to the resync
+                 * thread raising the barrier, we will deadlock because the
+                 * IO to the underlying device will be queued in generic_make_request
+                 * and will never complete, so will never reduce nr_pending.
+                 * So increment nr_waiting here so no new raise_barriers will
+                 * succeed, and so the second wait_barrier cannot block.
+                 */
+                spin_lock_irq(&conf->resync_lock);
+                conf->nr_waiting++;
+                spin_unlock_irq(&conf->resync_lock);
+
                 if (make_request(mddev, &bp->bio1))
                         generic_make_request(&bp->bio1);
                 if (make_request(mddev, &bp->bio2))
                         generic_make_request(&bp->bio2);
 
+                spin_lock_irq(&conf->resync_lock);
+                conf->nr_waiting--;
+                wake_up(&conf->wait_barrier);
+                spin_unlock_irq(&conf->resync_lock);
+
                 bio_pair_release(bp);
                 return 0;
         bad_map: