md-cluster: send BITMAP_NEEDS_SYNC message if reshaping is interrupted

We need to continue the reshaping if it was interrupted in
original node. So original node should call resync_bitmap
in case reshaping is aborted.

Then BITMAP_NEEDS_SYNC message is broadcasted to other nodes,
node which continues the reshaping should restart reshape from
mddev->reshape_position instead of from the first beginning.
Reviewed-by: NeilBrown <neilb@suse.com>
Signed-off-by: Guoqing Jiang <gqjiang@suse.com>
Signed-off-by: Shaohua Li <shli@fb.com>

drivers/md/md-cluster.c

@@ -333,6 +333,12 @@ static void recover_bitmaps(struct md_thread *thread)
 			}
 		spin_unlock_irq(&cinfo->suspend_lock);
 
+		/* Kick off a reshape if needed */
+		if (test_bit(MD_RESYNCING_REMOTE, &mddev->recovery) &&
+		    test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
+		    mddev->reshape_position != MaxSector)
+			md_wakeup_thread(mddev->sync_thread);
+
 		if (hi > 0) {
 			if (lo < mddev->recovery_cp)
 				mddev->recovery_cp = lo;
@@ -1020,10 +1026,17 @@ static int leave(struct mddev *mddev)
 	if (!cinfo)
 		return 0;
 
-	/* BITMAP_NEEDS_SYNC message should be sent when node
+	/*
+	 * BITMAP_NEEDS_SYNC message should be sent when node
 	 * is leaving the cluster with dirty bitmap, also we
-	 * can only deliver it when dlm connection is available */
-	if (cinfo->slot_number > 0 && mddev->recovery_cp != MaxSector)
+	 * can only deliver it when dlm connection is available.
+	 *
+	 * Also, we should send BITMAP_NEEDS_SYNC message in
+	 * case reshaping is interrupted.
+	 */
+	if ((cinfo->slot_number > 0 && mddev->recovery_cp != MaxSector) ||
+	    (mddev->reshape_position != MaxSector &&
+	     test_bit(MD_CLOSING, &mddev->flags)))
 		resync_bitmap(mddev);
 
 	set_bit(MD_CLUSTER_HOLDING_MUTEX_FOR_RECVD, &cinfo->state);

drivers/md/md.c

@@ -8370,9 +8370,17 @@ void md_do_sync(struct md_thread *thread)
 		else if (!mddev->bitmap)
 			j = mddev->recovery_cp;
 
-	} else if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
+	} else if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) {
 		max_sectors = mddev->resync_max_sectors;
-	else {
+		/*
+		 * If the original node aborts reshaping then we continue the
+		 * reshaping, so set j again to avoid restart reshape from the
+		 * first beginning
+		 */
+		if (mddev_is_clustered(mddev) &&
+		    mddev->reshape_position != MaxSector)
+			j = mddev->reshape_position;
+	} else {
 		/* recovery follows the physical size of devices */
 		max_sectors = mddev->dev_sectors;
 		j = MaxSector;