drbd: add some more explicit drbd_md_sync
authorLars Ellenberg <lars.ellenberg@linbit.com>
Thu, 14 Oct 2010 11:37:40 +0000 (13:37 +0200)
committerPhilipp Reisner <philipp.reisner@linbit.com>
Thu, 14 Oct 2010 17:08:58 +0000 (19:08 +0200)
It sometimes may take a while for the after state change work to be
scheduled, which does drbd_md_sync. At convenient places, we should do
explicit drbd_md_sync to have the new state information on disk as soon
as possible.

Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
drivers/block/drbd/drbd_receiver.c

index 5a4b6dcd48f2d19446602310b991cdbf77a9ee0f..6ec922c623a1f18bc50d1068742970985dd84f8e 100644 (file)
@@ -3804,6 +3804,9 @@ static void drbdd(struct drbd_conf *mdev)
        err_out:
                drbd_force_state(mdev, NS(conn, C_PROTOCOL_ERROR));
        }
+       /* If we leave here, we probably want to update at least the
+        * "Connected" indicator on stable storage. Do so explicitly here. */
+       drbd_md_sync(mdev);
 }
 
 void drbd_flush_workqueue(struct drbd_conf *mdev)
@@ -4685,10 +4688,12 @@ int drbd_asender(struct drbd_thread *thi)
        if (0) {
 reconnect:
                drbd_force_state(mdev, NS(conn, C_NETWORK_FAILURE));
+               drbd_md_sync(mdev);
        }
        if (0) {
 disconnect:
                drbd_force_state(mdev, NS(conn, C_DISCONNECTING));
+               drbd_md_sync(mdev);
        }
        clear_bit(SIGNAL_ASENDER, &mdev->flags);