diff --git a/block/blk-exec.c b/block/blk-exec.c
index d6cd501c0d34..f58ffcad8280 100644
--- a/block/blk-exec.c
+++ b/block/blk-exec.c
@@ -59,7 +59,10 @@ void blk_execute_rq_nowait(struct gendisk *bd_disk, struct request *rq,
 	 * don't check dying flag for MQ because the request won't
 	 * be reused after dying flag is set
 	 */
-	blk_mq_sched_insert_request(rq, at_head, true, false);
+	blk_mq_sched_insert_request(rq, at_head, true,
+				    rq->mq_hctx->flags & BLK_MQ_F_BLOCKING &&
+				    rq->cmd_flags & REQ_NOWAIT);
+
 }
 EXPORT_SYMBOL_GPL(blk_execute_rq_nowait);
 
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 474ec2fdceb8..9da923d49cb2 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -1561,11 +1561,9 @@ static void __blk_mq_delay_run_hw_queue(struct blk_mq_hw_ctx *hctx, bool async,
 	if (unlikely(blk_mq_hctx_stopped(hctx)))
 		return;
 
-	if (!async && !(hctx->flags & BLK_MQ_F_BLOCKING)) {
-		if (cpumask_test_cpu(raw_smp_processor_id(), hctx->cpumask)) {
-			__blk_mq_run_hw_queue(hctx);
-			return;
-		}
+	if (!async && cpumask_test_cpu(raw_smp_processor_id(), hctx->cpumask)) {
+		__blk_mq_run_hw_queue(hctx);
+		return;
 	}
 
 	kblockd_mod_delayed_work_on(blk_mq_hctx_next_cpu(hctx), &hctx->run_work,
@@ -1768,7 +1766,7 @@ void blk_mq_start_hw_queue(struct blk_mq_hw_ctx *hctx)
 {
 	clear_bit(BLK_MQ_S_STOPPED, &hctx->state);
 
-	blk_mq_run_hw_queue(hctx, false);
+	blk_mq_run_hw_queue(hctx, hctx->flags & BLK_MQ_F_BLOCKING);
 }
 EXPORT_SYMBOL(blk_mq_start_hw_queue);
 
@@ -1798,7 +1796,8 @@ void blk_mq_start_stopped_hw_queues(struct request_queue *q, bool async)
 	int i;
 
 	queue_for_each_hw_ctx(q, hctx, i)
-		blk_mq_start_stopped_hw_queue(hctx, async);
+		blk_mq_start_stopped_hw_queue(hctx, async ||
+					(hctx->flags & BLK_MQ_F_BLOCKING));
 }
 EXPORT_SYMBOL(blk_mq_start_stopped_hw_queues);
 
diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index 77f379649f64..8d7bb439855f 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -328,7 +328,8 @@ static void scsi_single_lun_run(struct scsi_device *current_sdev)
 	 * but in most cases, we will be first. Ideally, each LU on the
 	 * target would get some limited time or requests on the target.
 	 */
-	blk_mq_run_hw_queues(current_sdev->request_queue, false);
+	blk_mq_run_hw_queues(current_sdev->request_queue,
+			     shost->queuecommand_may_block);
 
 	spin_lock_irqsave(shost->host_lock, flags);
 	if (starget->starget_sdev_user)