md: make error_handler functions more uniform and correct.
authorNeilBrown <neilb@suse.de>
Wed, 11 May 2011 04:38:44 +0000 (14:38 +1000)
committerNeilBrown <neilb@suse.de>
Wed, 11 May 2011 04:38:44 +0000 (14:38 +1000)
- there is no need to test_bit Faulty, as that was already done in
  md_error which is the only caller of these functions.
- MD_CHANGE_DEVS should be set *after* faulty is set to ensure
  metadata is updated correctly.
- spinlock should be held while updating ->degraded.

Signed-off-by: NeilBrown <neilb@suse.de>
drivers/md/multipath.c
drivers/md/raid5.c

index 0254712..3535c23 100644 (file)
@@ -186,6 +186,7 @@ static int multipath_congested(void *data, int bits)
 static void multipath_error (mddev_t *mddev, mdk_rdev_t *rdev)
 {
        multipath_conf_t *conf = mddev->private;
+       char b[BDEVNAME_SIZE];
 
        if (conf->raid_disks - mddev->degraded <= 1) {
                /*
@@ -194,26 +195,27 @@ static void multipath_error (mddev_t *mddev, mdk_rdev_t *rdev)
                 * which has just failed.
                 */
                printk(KERN_ALERT 
-                       "multipath: only one IO path left and IO error.\n");
+                      "multipath: only one IO path left and IO error.\n");
                /* leave it active... it's all we have */
-       } else {
-               /*
-                * Mark disk as unusable
-                */
-               if (!test_bit(Faulty, &rdev->flags)) {
-                       char b[BDEVNAME_SIZE];
-                       clear_bit(In_sync, &rdev->flags);
-                       set_bit(Faulty, &rdev->flags);
-                       set_bit(MD_CHANGE_DEVS, &mddev->flags);
-                       mddev->degraded++;
-                       printk(KERN_ALERT "multipath: IO failure on %s,"
-                               " disabling IO path.\n"
-                               "multipath: Operation continuing"
-                               " on %d IO paths.\n",
-                               bdevname (rdev->bdev,b),
-                               conf->raid_disks - mddev->degraded);
-               }
+               return;
+       }
+       /*
+        * Mark disk as unusable
+        */
+       if (test_and_clear_bit(In_sync, &rdev->flags)) {
+               unsigned long flags;
+               spin_lock_irqsave(&conf->device_lock, flags);
+               mddev->degraded++;
+               spin_unlock_irqrestore(&conf->device_lock, flags);
        }
+       set_bit(Faulty, &rdev->flags);
+       set_bit(MD_CHANGE_DEVS, &mddev->flags);
+       printk(KERN_ALERT "multipath: IO failure on %s,"
+              " disabling IO path.\n"
+              "multipath: Operation continuing"
+              " on %d IO paths.\n",
+              bdevname(rdev->bdev, b),
+              conf->raid_disks - mddev->degraded);
 }
 
 static void print_multipath_conf (multipath_conf_t *conf)
@@ -273,9 +275,11 @@ static int multipath_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
                                                           PAGE_CACHE_SIZE - 1);
                        }
 
+                       spin_lock_irq(&conf->device_lock);
                        mddev->degraded--;
                        rdev->raid_disk = path;
                        set_bit(In_sync, &rdev->flags);
+                       spin_unlock_irq(&conf->device_lock);
                        rcu_assign_pointer(p->rdev, rdev);
                        err = 0;
                        md_integrity_add_rdev(rdev, mddev);
index 49bf5f8..51af7f3 100644 (file)
@@ -1700,27 +1700,25 @@ static void error(mddev_t *mddev, mdk_rdev_t *rdev)
        raid5_conf_t *conf = mddev->private;
        pr_debug("raid456: error called\n");
 
-       if (!test_bit(Faulty, &rdev->flags)) {
-               set_bit(MD_CHANGE_DEVS, &mddev->flags);
-               if (test_and_clear_bit(In_sync, &rdev->flags)) {
-                       unsigned long flags;
-                       spin_lock_irqsave(&conf->device_lock, flags);
-                       mddev->degraded++;
-                       spin_unlock_irqrestore(&conf->device_lock, flags);
-                       /*
-                        * if recovery was running, make sure it aborts.
-                        */
-                       set_bit(MD_RECOVERY_INTR, &mddev->recovery);
-               }
-               set_bit(Faulty, &rdev->flags);
-               printk(KERN_ALERT
-                      "md/raid:%s: Disk failure on %s, disabling device.\n"
-                      "md/raid:%s: Operation continuing on %d devices.\n",
-                      mdname(mddev),
-                      bdevname(rdev->bdev, b),
-                      mdname(mddev),
-                      conf->raid_disks - mddev->degraded);
+       if (test_and_clear_bit(In_sync, &rdev->flags)) {
+               unsigned long flags;
+               spin_lock_irqsave(&conf->device_lock, flags);
+               mddev->degraded++;
+               spin_unlock_irqrestore(&conf->device_lock, flags);
+               /*
+                * if recovery was running, make sure it aborts.
+                */
+               set_bit(MD_RECOVERY_INTR, &mddev->recovery);
        }
+       set_bit(Faulty, &rdev->flags);
+       set_bit(MD_CHANGE_DEVS, &mddev->flags);
+       printk(KERN_ALERT
+              "md/raid:%s: Disk failure on %s, disabling device.\n"
+              "md/raid:%s: Operation continuing on %d devices.\n",
+              mdname(mddev),
+              bdevname(rdev->bdev, b),
+              mdname(mddev),
+              conf->raid_disks - mddev->degraded);
 }
 
 /*