diff --git a/block/blk-core.c b/block/blk-core.c
index d709c0e3a2ac..aebe676225e6 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -2342,7 +2342,12 @@ blk_status_t blk_insert_cloned_request(struct request_queue *q, struct request *
 	if (q->mq_ops) {
 		if (blk_queue_io_stat(q))
 			blk_account_io_start(rq, true);
-		blk_mq_sched_insert_request(rq, false, true, false, false);
+		/*
+		 * Since we have a scheduler attached on the top device,
+		 * bypass a potential scheduler on the bottom device for
+		 * insert.
+		 */
+		blk_mq_request_bypass_insert(rq);
 		return BLK_STS_OK;
 	}
 
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 3f18cff80050..98a18609755e 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -1401,6 +1401,22 @@ void __blk_mq_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq,
 	blk_mq_hctx_mark_pending(hctx, ctx);
 }
 
+/*
+ * Should only be used carefully, when the caller knows we want to
+ * bypass a potential IO scheduler on the target device.
+ */
+void blk_mq_request_bypass_insert(struct request *rq)
+{
+	struct blk_mq_ctx *ctx = rq->mq_ctx;
+	struct blk_mq_hw_ctx *hctx = blk_mq_map_queue(rq->q, ctx->cpu);
+
+	spin_lock(&hctx->lock);
+	list_add_tail(&rq->queuelist, &hctx->dispatch);
+	spin_unlock(&hctx->lock);
+
+	blk_mq_run_hw_queue(hctx, false);
+}
+
 void blk_mq_insert_requests(struct blk_mq_hw_ctx *hctx, struct blk_mq_ctx *ctx,
 			    struct list_head *list)
 
diff --git a/block/blk-mq.h b/block/blk-mq.h
index 98252b79b80b..ef15b3414da5 100644
--- a/block/blk-mq.h
+++ b/block/blk-mq.h
@@ -54,6 +54,7 @@ int blk_mq_alloc_rqs(struct blk_mq_tag_set *set, struct blk_mq_tags *tags,
  */
 void __blk_mq_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq,
 				bool at_head);
+void blk_mq_request_bypass_insert(struct request *rq);
 void blk_mq_insert_requests(struct blk_mq_hw_ctx *hctx, struct blk_mq_ctx *ctx,
 				struct list_head *list);