md: Fix skipping recovery for read-only arrays.
authorLukasz Dorau <lukasz.dorau@intel.com>
Thu, 24 Oct 2013 01:55:17 +0000 (12:55 +1100)
committerNeilBrown <neilb@suse.de>
Thu, 24 Oct 2013 01:55:17 +0000 (12:55 +1100)
Since:
        commit 7ceb17e87bde79d285a8b988cfed9eaeebe60b86
        md: Allow devices to be re-added to a read-only array.

spares are activated on a read-only array. In case of raid1 and raid10
personalities it causes that not-in-sync devices are marked in-sync
without checking if recovery has been finished.

If a read-only array is degraded and one of its devices is not in-sync
(because the array has been only partially recovered) recovery will be skipped.

This patch adds checking if recovery has been finished before marking a device
in-sync for raid1 and raid10 personalities. In case of raid5 personality
such condition is already present (at raid5.c:6029).

Bug was introduced in 3.10 and causes data corruption.

Cc: stable@vger.kernel.org
Signed-off-by: Pawel Baldysiak <pawel.baldysiak@intel.com>
Signed-off-by: Lukasz Dorau <lukasz.dorau@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>
drivers/md/raid1.c
drivers/md/raid10.c

index d60412c7f9952a350e85002f6578bd66ee416a44..aacf6bf352d87978a15633b05a0d51d3579ce16d 100644 (file)
@@ -1479,6 +1479,7 @@ static int raid1_spare_active(struct mddev *mddev)
                        }
                }
                if (rdev
+                   && rdev->recovery_offset == MaxSector
                    && !test_bit(Faulty, &rdev->flags)
                    && !test_and_set_bit(In_sync, &rdev->flags)) {
                        count++;
index df7b0a06b0ea2d820aec6ad5a752a13bb64e13e8..73dc8a377522e19af92c9bed26519a479f36dbea 100644 (file)
@@ -1782,6 +1782,7 @@ static int raid10_spare_active(struct mddev *mddev)
                        }
                        sysfs_notify_dirent_safe(tmp->replacement->sysfs_state);
                } else if (tmp->rdev
+                          && tmp->rdev->recovery_offset == MaxSector
                           && !test_bit(Faulty, &tmp->rdev->flags)
                           && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
                        count++;