diff '--color=auto' -urpN a/block/blk-mq.c b/block/blk-mq.c
--- a/block/blk-mq.c	2025-02-21 14:01:47.000000000 +0100
+++ b/block/blk-mq.c	2025-02-21 15:52:51.848041852 +0100
@@ -1418,7 +1418,7 @@ void blk_execute_rq_nowait(struct reques
 	}
 
 	blk_mq_insert_request(rq, at_head ? BLK_MQ_INSERT_AT_HEAD : 0);
-	blk_mq_run_hw_queue(hctx, hctx->flags & BLK_MQ_F_BLOCKING);
+	blk_mq_run_hw_queue(hctx, false);
 }
 EXPORT_SYMBOL_GPL(blk_execute_rq_nowait);
 
@@ -2322,8 +2322,6 @@ void blk_mq_run_hw_queue(struct blk_mq_h
 	 */
 	WARN_ON_ONCE(!async && in_interrupt());
 
-	might_sleep_if(!async && hctx->flags & BLK_MQ_F_BLOCKING);
-
 	need_run = blk_mq_hw_queue_need_run(hctx);
 	if (!need_run) {
 		unsigned long flags;
@@ -2342,7 +2340,8 @@ void blk_mq_run_hw_queue(struct blk_mq_h
 			return;
 	}
 
-	if (async || !cpumask_test_cpu(raw_smp_processor_id(), hctx->cpumask)) {
+	if (async || (hctx->flags & BLK_MQ_F_BLOCKING) ||
+	    !cpumask_test_cpu(raw_smp_processor_id(), hctx->cpumask)) {
 		blk_mq_delay_run_hw_queue(hctx, 0);
 		return;
 	}
@@ -2477,7 +2476,7 @@ void blk_mq_start_hw_queue(struct blk_mq
 {
 	clear_bit(BLK_MQ_S_STOPPED, &hctx->state);
 
-	blk_mq_run_hw_queue(hctx, hctx->flags & BLK_MQ_F_BLOCKING);
+	blk_mq_run_hw_queue(hctx, false);
 }
 EXPORT_SYMBOL(blk_mq_start_hw_queue);
 
@@ -2513,8 +2512,7 @@ void blk_mq_start_stopped_hw_queues(stru
 	unsigned long i;
 
 	queue_for_each_hw_ctx(q, hctx, i)
-		blk_mq_start_stopped_hw_queue(hctx, async ||
-					(hctx->flags & BLK_MQ_F_BLOCKING));
+		blk_mq_start_stopped_hw_queue(hctx, async);
 }
 EXPORT_SYMBOL(blk_mq_start_stopped_hw_queues);
 
@@ -2572,8 +2570,6 @@ static void blk_mq_insert_requests(struc
 	list_for_each_entry(rq, list, queuelist) {
 		BUG_ON(rq->mq_ctx != ctx);
 		trace_block_rq_insert(rq);
-		if (rq->cmd_flags & REQ_NOWAIT)
-			run_queue_async = true;
 	}
 
 	spin_lock(&ctx->lock);
@@ -2739,7 +2735,7 @@ static void blk_mq_try_issue_directly(st
 
 	if ((rq->rq_flags & RQF_USE_SCHED) || !blk_mq_get_budget_and_tag(rq)) {
 		blk_mq_insert_request(rq, 0);
-		blk_mq_run_hw_queue(hctx, rq->cmd_flags & REQ_NOWAIT);
+		blk_mq_run_hw_queue(hctx, false);
 		return;
 	}
 
diff '--color=auto' -urpN a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
--- a/drivers/scsi/scsi_lib.c	2025-02-21 14:01:47.000000000 +0100
+++ b/drivers/scsi/scsi_lib.c	2025-02-21 15:53:54.654044691 +0100
@@ -429,8 +429,7 @@ static void scsi_single_lun_run(struct s
 	 * 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,
-			     shost->queuecommand_may_block);
+	blk_mq_run_hw_queues(current_sdev->request_queue, false);
 
 	spin_lock_irqsave(shost->host_lock, flags);
 	if (!starget->starget_sdev_user)