md/raid5: ensure device failure recorded before write request returns.

When a write to one of the devices of a RAID5/6 fails, the failure is
recorded in the metadata of the other devices so that after a restart
the data on the failed drive wont be trusted even if that drive seems
to be working again (maybe a cable was unplugged).

Similarly when we record a bad-block in response to a write failure,
we must not let the write complete until the bad-block update is safe.

Currently there is no interlock between the write request completing
and the metadata update.  So it is possible that the write will
complete, the app will confirm success in some way, and then the
machine will crash before the metadata update completes.

This is an extremely small hole for a racy to fit in, but it is
theoretically possible and so should be closed.

So:
 - set MD_CHANGE_PENDING when requesting a metadata update for a
   failed device, so we can know with certainty when it completes
 - queue requests that completed when MD_CHANGE_PENDING is set to
   only be processed after the metadata update completes
 - call raid_end_bio_io() on bios in that queue when the time comes.


Signed-off-by: NeilBrown <neilb@suse.com>

2 files changed, 26 insertions, 1 deletions

diff --git a/drivers/md/raid5.c b/drivers/md/raid5.c
index 214dcca0d7f8..4195064460d0 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -2513,6 +2513,7 @@ static void error(struct mddev *mddev, struct md_rdev *rdev)
 	set_bit(Blocked, &rdev->flags);
 	set_bit(Faulty, &rdev->flags);
 	set_bit(MD_CHANGE_DEVS, &mddev->flags);
+	set_bit(MD_CHANGE_PENDING, &mddev->flags);
 	printk(KERN_ALERT
 	       "md/raid:%s: Disk failure on %s, disabling device.\n"
 	       "md/raid:%s: Operation continuing on %d devices.\n",
@@ -4601,7 +4602,15 @@ finish:
 			md_wakeup_thread(conf->mddev->thread);
 	}
 
-	return_io(&s.return_bi);
+	if (!bio_list_empty(&s.return_bi)) {
+		if (test_bit(MD_CHANGE_PENDING, &conf->mddev->flags)) {
+			spin_lock_irq(&conf->device_lock);
+			bio_list_merge(&conf->return_bi, &s.return_bi);
+			spin_unlock_irq(&conf->device_lock);
+			md_wakeup_thread(conf->mddev->thread);
+		} else
+			return_io(&s.return_bi);
+	}
 
 	clear_bit_unlock(STRIPE_ACTIVE, &sh->state);
 }
@@ -5817,6 +5826,18 @@ static void raid5d(struct md_thread *thread)
 
 	md_check_recovery(mddev);
 
+	if (!bio_list_empty(&conf->return_bi) &&
+	    !test_bit(MD_CHANGE_PENDING, &mddev->flags)) {
+		struct bio_list tmp = BIO_EMPTY_LIST;
+		spin_lock_irq(&conf->device_lock);
+		if (!test_bit(MD_CHANGE_PENDING, &mddev->flags)) {
+			bio_list_merge(&tmp, &conf->return_bi);
+			bio_list_init(&conf->return_bi);
+		}
+		spin_unlock_irq(&conf->device_lock);
+		return_io(&tmp);
+	}
+
 	blk_start_plug(&plug);
 	handled = 0;
 	spin_lock_irq(&conf->device_lock);
@@ -6476,6 +6497,7 @@ static struct r5conf *setup_conf(struct mddev *mddev)
 	INIT_LIST_HEAD(&conf->hold_list);
 	INIT_LIST_HEAD(&conf->delayed_list);
 	INIT_LIST_HEAD(&conf->bitmap_list);
+	bio_list_init(&conf->return_bi);
 	init_llist_head(&conf->released_stripes);
 	atomic_set(&conf->active_stripes, 0);
 	atomic_set(&conf->preread_active_stripes, 0);
diff --git a/drivers/md/raid5.h b/drivers/md/raid5.h
index 1de82a6e4c23..828c2925e68f 100644
--- a/drivers/md/raid5.h
+++ b/drivers/md/raid5.h
@@ -476,6 +476,9 @@ struct r5conf {
 	int			skip_copy; /* Don't copy data from bio to stripe cache */
 	struct list_head	*last_hold; /* detect hold_list promotions */
 
+	/* bios to have bi_end_io called after metadata is synced */
+	struct bio_list		return_bi;
+
 	atomic_t		reshape_stripes; /* stripes with pending writes for reshape */
 	/* unfortunately we need two cache names as we temporarily have
 	 * two caches.