Revert "md: allow a partially recovered device to be hot-added to an array."
authorNeilBrown <neilb@suse.com>
Sat, 31 Oct 2015 00:00:56 +0000 (11:00 +1100)
committerNeilBrown <neilb@suse.com>
Sat, 31 Oct 2015 00:00:56 +0000 (11:00 +1100)
This reverts commit 7eb418851f3278de67126ea0c427641ab4792c57.

This commit is poorly justified, I can find not discusison in email,
and it clearly causes a problem.

If a device which is being recovered fails and is subsequently
re-added to an array, there could easily have been changes to the
array *before* the point where the recovery was up to.  So the
recovery must start again from the beginning.

If a spare is being recovered and fails, then when it is re-added we
really should do a bitmap-based recovery up to the recovery-offset,
and then a full recovery from there.  Before this reversion, we only
did the "full recovery from there" which is not corect.  After this
reversion with will do a full recovery from the start, which is safer
but not ideal.

It will be left to a future patch to arrange the two different styles
of recovery.

Reported-and-tested-by: Nate Dailey <nate.dailey@stratus.com>
Signed-off-by: NeilBrown <neilb@suse.com>
Cc: stable@vger.kernel.org (3.14+)
Fixes: 7eb418851f32 ("md: allow a partially recovered device to be hot-added to an array.")

drivers/md/md.c

index c702de18207ae76ab56f1235ed5c98a9095ed050..3fe3d04a968ad1ae5e148abbc93cde43be576104 100644 (file)
@@ -8040,8 +8040,7 @@ static int remove_and_add_spares(struct mddev *mddev,
                       !test_bit(Bitmap_sync, &rdev->flags)))
                        continue;
 
-               if (rdev->saved_raid_disk < 0)
-                       rdev->recovery_offset = 0;
+               rdev->recovery_offset = 0;
                if (mddev->pers->
                    hot_add_disk(mddev, rdev) == 0) {
                        if (sysfs_link_rdev(mddev, rdev))