drbd: Allow new IOs while the local disk in in FAILED state

The last bunch of commits prepared the 'detach from tar pit' feature.
With that we can be for long time in disk state FAILED. We need
to accept new IO requests during that time.
Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>

drivers/block/drbd/drbd_int.h

@@ -2310,12 +2310,12 @@ static inline int drbd_state_is_stable(struct drbd_conf *mdev)
 	case D_OUTDATED:
 	case D_CONSISTENT:
 	case D_UP_TO_DATE:
+	case D_FAILED:
 		/* disk state is stable as well. */
 		break;
 
 	/* no new io accepted during tansitional states */
 	case D_ATTACHING:
-	case D_FAILED:
 	case D_NEGOTIATING:
 	case D_UNKNOWN:
 	case D_MASK: