]> pilppa.com Git - linux-2.6-omap-h63xx.git/commitdiff
Make sure all changes to md/sync_action are notified.
authorNeil Brown <neilb@notabene.brown>
Fri, 27 Jun 2008 22:31:41 +0000 (08:31 +1000)
committerNeil Brown <neilb@notabene.brown>
Fri, 27 Jun 2008 22:31:41 +0000 (08:31 +1000)
When the 'resync' thread starts or stops, when we explicitly
set sync_action, or when we determine that there is definitely nothing
to do, we notify sync_action.

To stop "sync_action" from occasionally showing the wrong value,
we introduce a new flags - MD_RECOVERY_RECOVER - to say that a
recovery is probably needed or happening, and we make sure
that we set MD_RECOVERY_RUNNING before clearing MD_RECOVERY_NEEDED.

Signed-off-by: Neil Brown <neilb@suse.de>
Documentation/md.txt
drivers/md/md.c
include/linux/raid/md_k.h

index dca97ba4944a9d3fe445b23f0d6940e21b2f2f38..c05bfb55659edd3aed5fa1976f4452c1135b67dc 100644 (file)
@@ -386,6 +386,12 @@ also have
        'check' and 'repair' will start the appropriate process
            providing the current state is 'idle'.
 
+      This file responds to select/poll.  Any important change in the value
+      triggers a poll event.  Sometimes the value will briefly be
+      "recover" if a recovery seems to be needed, but cannot be
+      achieved. In that case, the transition to "recover" isn't
+      notified, but the transition away is.
+
    mismatch_count
       When performing 'check' and 'repair', and possibly when
       performing 'resync', md will count the number of errors that are
index 5b9d4fe4e6e44409b17db186764a5ec0c8e282cb..c26dcad8a3ac98730dfea7d9590eb8fa516cbdc9 100644 (file)
@@ -169,7 +169,6 @@ void md_new_event(mddev_t *mddev)
 {
        atomic_inc(&md_event_count);
        wake_up(&md_event_waiters);
-       sysfs_notify(&mddev->kobj, NULL, "sync_action");
 }
 EXPORT_SYMBOL_GPL(md_new_event);
 
@@ -2936,7 +2935,7 @@ action_show(mddev_t *mddev, char *page)
                                type = "check";
                        else
                                type = "repair";
-               } else
+               } else if (test_bit(MD_RECOVERY_RECOVER, &mddev->recovery))
                        type = "recover";
        }
        return sprintf(page, "%s\n", type);
@@ -2958,9 +2957,12 @@ action_store(mddev_t *mddev, const char *page, size_t len)
        } else if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) ||
                   test_bit(MD_RECOVERY_NEEDED, &mddev->recovery))
                return -EBUSY;
-       else if (cmd_match(page, "resync") || cmd_match(page, "recover"))
+       else if (cmd_match(page, "resync"))
+               set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
+       else if (cmd_match(page, "recover")) {
+               set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
                set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
-       else if (cmd_match(page, "reshape")) {
+       else if (cmd_match(page, "reshape")) {
                int err;
                if (mddev->pers->start_reshape == NULL)
                        return -EINVAL;
@@ -2977,6 +2979,7 @@ action_store(mddev_t *mddev, const char *page, size_t len)
        }
        set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
        md_wakeup_thread(mddev->thread);
+       sysfs_notify(&mddev->kobj, NULL, "sync_action");
        return len;
 }
 
@@ -3682,6 +3685,7 @@ static int do_md_run(mddev_t * mddev)
        mddev->changed = 1;
        md_new_event(mddev);
        sysfs_notify(&mddev->kobj, NULL, "array_state");
+       sysfs_notify(&mddev->kobj, NULL, "sync_action");
        kobject_uevent(&mddev->gendisk->dev.kobj, KOBJ_CHANGE);
        return 0;
 }
@@ -4252,6 +4256,8 @@ static int add_new_disk(mddev_t * mddev, mdu_disk_info_t *info)
                        export_rdev(rdev);
 
                md_update_sb(mddev, 1);
+               if (mddev->degraded)
+                       set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
                set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
                md_wakeup_thread(mddev->thread);
                return err;
@@ -5105,6 +5111,8 @@ void md_error(mddev_t *mddev, mdk_rdev_t *rdev)
        if (!mddev->pers->error_handler)
                return;
        mddev->pers->error_handler(mddev,rdev);
+       if (mddev->degraded)
+               set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
        set_bit(MD_RECOVERY_INTR, &mddev->recovery);
        set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
        md_wakeup_thread(mddev->thread);
@@ -6055,13 +6063,18 @@ void md_check_recovery(mddev_t *mddev)
                        mddev->recovery = 0;
                        /* flag recovery needed just to double check */
                        set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
+                       sysfs_notify(&mddev->kobj, NULL, "sync_action");
                        md_new_event(mddev);
                        goto unlock;
                }
+               /* Set RUNNING before clearing NEEDED to avoid
+                * any transients in the value of "sync_action".
+                */
+               set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
+               clear_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
                /* Clear some bits that don't mean anything, but
                 * might be left set
                 */
-               clear_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
                clear_bit(MD_RECOVERY_INTR, &mddev->recovery);
                clear_bit(MD_RECOVERY_DONE, &mddev->recovery);
 
@@ -6079,17 +6092,19 @@ void md_check_recovery(mddev_t *mddev)
                                /* Cannot proceed */
                                goto unlock;
                        set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
+                       clear_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
                } else if ((spares = remove_and_add_spares(mddev))) {
                        clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
                        clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
+                       set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
                } else if (mddev->recovery_cp < MaxSector) {
                        set_bit(MD_RECOVERY_SYNC, &mddev->recovery);
+                       clear_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
                } else if (!test_bit(MD_RECOVERY_SYNC, &mddev->recovery))
                        /* nothing to be done ... */
                        goto unlock;
 
                if (mddev->pers->sync_request) {
-                       set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
                        if (spares && mddev->bitmap && ! mddev->bitmap->file) {
                                /* We are adding a device or devices to an array
                                 * which has the bitmap stored on all devices.
@@ -6108,9 +6123,16 @@ void md_check_recovery(mddev_t *mddev)
                                mddev->recovery = 0;
                        } else
                                md_wakeup_thread(mddev->sync_thread);
+                       sysfs_notify(&mddev->kobj, NULL, "sync_action");
                        md_new_event(mddev);
                }
        unlock:
+               if (!mddev->sync_thread) {
+                       clear_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
+                       if (test_and_clear_bit(MD_RECOVERY_RECOVER,
+                                              &mddev->recovery))
+                               sysfs_notify(&mddev->kobj, NULL, "sync_action");
+               }
                mddev_unlock(mddev);
        }
 }
index 780e0613e6d5bbe0f552e7051b28444fc856b8d3..62aa9c9a6ddc16ae4cf1fc45dddabe323fe881fe 100644 (file)
@@ -188,6 +188,7 @@ struct mddev_s
         * NEEDED:   we might need to start a resync/recover
         * RUNNING:  a thread is running, or about to be started
         * SYNC:     actually doing a resync, not a recovery
+        * RECOVER:  doing recovery, or need to try it.
         * INTR:     resync needs to be aborted for some reason
         * DONE:     thread is done and is waiting to be reaped
         * REQUEST:  user-space has requested a sync (used with SYNC)
@@ -198,6 +199,7 @@ struct mddev_s
         */
 #define        MD_RECOVERY_RUNNING     0
 #define        MD_RECOVERY_SYNC        1
+#define        MD_RECOVERY_RECOVER     2
 #define        MD_RECOVERY_INTR        3
 #define        MD_RECOVERY_DONE        4
 #define        MD_RECOVERY_NEEDED      5