]> git.proxmox.com Git - mirror_ubuntu-kernels.git/commitdiff
md/raid5: Fix a bug about judging if the operation is syncing or replacing
authormajianpeng <majianpeng@gmail.com>
Sun, 1 Apr 2012 15:16:59 +0000 (01:16 +1000)
committerNeilBrown <neilb@suse.de>
Tue, 3 Apr 2012 05:37:38 +0000 (15:37 +1000)
When create a raid5 using assume-clean and echo check or repair to
sync_action.Then component disks did not operated IO but the raid
check/resync faster than normal.
Because the judgement in function analyse_stripe():
if (do_recovery ||
    sh->sector >= conf->mddev->recovery_cp)
s->syncing = 1;
else
s->replacing = 1;
When check or repair,the recovery_cp == MaxSectore,so syncing equal zero
not one.

This bug was introduced by commit 9a3e1101b827
    md/raid5:  detect and handle replacements during recovery.
so this patch is suitable for 3.3-stable.

Cc: stable@vger.kernel.org
Signed-off-by: majianpeng <majianpeng@gmail.com>
Signed-off-by: NeilBrown <neilb@suse.de>
drivers/md/raid5.c

index 9799be80bf3143fef6420e1c8f456fa46fb786f1..f351422938e05f0718d9d71be8ff1b86621b948c 100644 (file)
@@ -3279,12 +3279,14 @@ static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
                /* If there is a failed device being replaced,
                 *     we must be recovering.
                 * else if we are after recovery_cp, we must be syncing
+                * else if MD_RECOVERY_REQUESTED is set, we also are syncing.
                 * else we can only be replacing
                 * sync and recovery both need to read all devices, and so
                 * use the same flag.
                 */
                if (do_recovery ||
-                   sh->sector >= conf->mddev->recovery_cp)
+                   sh->sector >= conf->mddev->recovery_cp ||
+                   test_bit(MD_RECOVERY_REQUESTED, &(conf->mddev->recovery)))
                        s->syncing = 1;
                else
                        s->replacing = 1;