md-cluser: make resync_finish only called after pers->sync_request

It is not reasonable that cluster raid to release resync
lock before the last pers->sync_request has finished.

As the metadata will be changed when node performs resync,
we need to inform other nodes to update metadata, so the
MD_CHANGE_PENDING flag is set before finish resync.

Then metadata_update_finish is move ahead to ensure that
METADATA_UPDATED msg is sent before finish resync, and
metadata_update_start need to be run after "repeat:" label
accordingly.

Reviewed-by: NeilBrown <neilb@suse.com>
Signed-off-by: Guoqing Jiang <gqjiang@suse.com>
Signed-off-by: Shaohua Li <shli@fb.com>

diff --git a/drivers/md/md.c b/drivers/md/md.c
index 4fd7d7757f2d78883b2e174f4075e172e72175a8..dd83a50d892ca4de96d490de0d799da63cb78037 100644 (file)
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -2291,6 +2291,7 @@ void md_update_sb(struct mddev *mddev, int force_change)
                return;
        }
 
+repeat:
        if (mddev_is_clustered(mddev)) {
                if (test_and_clear_bit(MD_CHANGE_DEVS, &mddev->flags))
                        force_change = 1;
@@ -2303,7 +2304,7 @@ void md_update_sb(struct mddev *mddev, int force_change)
                        return;
                }
        }
-repeat:
+
        /* First make sure individual recovery_offsets are correct */
        rdev_for_each(rdev, mddev) {
                if (rdev->raid_disk >= 0 &&
@@ -2430,6 +2431,9 @@ repeat:
        md_super_wait(mddev);
        /* if there was a failure, MD_CHANGE_DEVS was set, and we re-write super */
 
+       if (mddev_is_clustered(mddev) && ret == 0)
+               md_cluster_ops->metadata_update_finish(mddev);
+
        spin_lock(&mddev->lock);
        if (mddev->in_sync != sync_req ||
            test_bit(MD_CHANGE_DEVS, &mddev->flags)) {
@@ -2452,9 +2456,6 @@ repeat:
                clear_bit(BlockedBadBlocks, &rdev->flags);
                wake_up(&rdev->blocked_wait);
        }
-
-       if (mddev_is_clustered(mddev) && ret == 0)
-               md_cluster_ops->metadata_update_finish(mddev);
 }
 EXPORT_SYMBOL(md_update_sb);
 
@@ -7785,7 +7786,6 @@ void md_do_sync(struct md_thread *thread)
        struct md_rdev *rdev;
        char *desc, *action = NULL;
        struct blk_plug plug;
-       bool cluster_resync_finished = false;
        int ret;
 
        /* just incase thread restarts... */
@@ -8103,11 +8103,6 @@ void md_do_sync(struct md_thread *thread)
                mddev->curr_resync_completed = mddev->curr_resync;
                sysfs_notify(&mddev->kobj, NULL, "sync_completed");
        }
-       /* tell personality and other nodes that we are finished */
-       if (mddev_is_clustered(mddev)) {
-               md_cluster_ops->resync_finish(mddev);
-               cluster_resync_finished = true;
-       }
        mddev->pers->sync_request(mddev, max_sectors, &skipped);
 
        if (!test_bit(MD_RECOVERY_CHECK, &mddev->recovery) &&
@@ -8147,9 +8142,15 @@ void md_do_sync(struct md_thread *thread)
        set_bit(MD_CHANGE_DEVS, &mddev->flags);
 
        if (mddev_is_clustered(mddev) &&
-           test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
-           !cluster_resync_finished)
+           ret == 0) {
+               /* set CHANGE_PENDING here since maybe another
+                * update is needed, so other nodes are informed */
+               set_bit(MD_CHANGE_PENDING, &mddev->flags);
+               md_wakeup_thread(mddev->thread);
+               wait_event(mddev->sb_wait,
+                          !test_bit(MD_CHANGE_PENDING, &mddev->flags));
                md_cluster_ops->resync_finish(mddev);
+       }
 
        spin_lock(&mddev->lock);
        if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {