releases/2.6.36.2/md-raid1-really-fix-recovery-looping-when-single-good-device-fails.patch

   1 From 8f9e0ee38f75d4740daa9e42c8af628d33d19a02 Mon Sep 17 00:00:00 2001
   2 From: NeilBrown <neilb@suse.de>
   3 Date: Wed, 24 Nov 2010 16:39:46 +1100
   4 Subject: md/raid1: really fix recovery looping when single good device fails.
   5 MIME-Version: 1.0
   6 Content-Type: text/plain; charset=UTF-8
   7 Content-Transfer-Encoding: 8bit
   8
   9 From: NeilBrown <neilb@suse.de>
  10
  11 commit 8f9e0ee38f75d4740daa9e42c8af628d33d19a02 upstream.
  12
  13 Commit 4044ba58dd15cb01797c4fd034f39ef4a75f7cc3 supposedly fixed a
  14 problem where if a raid1 with just one good device gets a read-error
  15 during recovery, the recovery would abort and immediately restart in
  16 an infinite loop.
  17
  18 However it depended on raid1_remove_disk removing the spare device
  19 from the array.  But that does not happen in this case.  So add a test
  20 so that in the 'recovery_disabled' case, the device will be removed.
  21
  22 This suitable for any kernel since 2.6.29 which is when
  23 recovery_disabled was introduced.
  24
  25 Reported-by: Sebastian Färber <faerber@gmail.com>
  26 Signed-off-by: NeilBrown <neilb@suse.de>
  27 Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
  28
  29 ---
  30  drivers/md/raid1.c |    1 +
  31  1 file changed, 1 insertion(+)
  32
  33 --- a/drivers/md/raid1.c
  34 +++ b/drivers/md/raid1.c
  35 @@ -1210,6 +1210,7 @@ static int raid1_remove_disk(mddev_t *md
  36                  * is not possible.
  37                  */
  38                 if (!test_bit(Faulty, &rdev->flags) &&
  39 +                   !mddev->recovery_disabled &&
  40                     mddev->degraded < conf->raid_disks) {
  41                         err = -EBUSY;
  42                         goto abort;