blk-mq: return when hctx is stopped in blk_mq_run_work_fn

If a hardware queue is stopped, it should not be run again before
explicitly started. Ignore stopped queues in blk_mq_run_work_fn(),
fixing a regression recently introduced when the START_ON_RUN bit
was removed.

Fixes: 15fe8a90 ("blk-mq: remove blk_mq_delay_queue()")
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Bart Van Assche <bart.vanassche@wdc.com>
Signed-off-by: Jianchao Wang <jianchao.w.wang@oracle.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block/blk-mq.c

@@ -1475,7 +1475,7 @@ static void blk_mq_run_work_fn(struct work_struct *work)
 	 * If we are stopped, don't run the queue.
 	 */
 	if (test_bit(BLK_MQ_S_STOPPED, &hctx->state))
-		clear_bit(BLK_MQ_S_STOPPED, &hctx->state);
+		return;
 
 	__blk_mq_run_hw_queue(hctx);
 }