diff --git a/block/blk-mq.c b/block/blk-mq.c
index 68929ba..7e53038 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -680,6 +680,8 @@
 	struct request_queue *q = hctx->queue;
 	struct request *rq;
 	LIST_HEAD(rq_list);
+	LIST_HEAD(driver_list);
+	struct list_head *dptr;
 	int queued;
 
 	WARN_ON(!cpumask_test_cpu(raw_smp_processor_id(), hctx->cpumask));
@@ -706,16 +708,27 @@
 	}
 
 	/*
+	 * Start off with dptr being NULL, so we start the first request
+	 * immediately, even if we have more pending.
+	 */
+	dptr = NULL;
+
+	/*
 	 * Now process all the entries, sending them to the driver.
 	 */
 	queued = 0;
 	while (!list_empty(&rq_list)) {
+		struct blk_mq_queue_data bd;
 		int ret;
 
 		rq = list_first_entry(&rq_list, struct request, queuelist);
 		list_del_init(&rq->queuelist);
 
-		ret = q->mq_ops->queue_rq(hctx, rq, list_empty(&rq_list));
+		bd.rq = rq;
+		bd.list = dptr;
+		bd.last = list_empty(&rq_list);
+
+		ret = q->mq_ops->queue_rq(hctx, &bd);
 		switch (ret) {
 		case BLK_MQ_RQ_QUEUE_OK:
 			queued++;
@@ -734,6 +747,13 @@
 
 		if (ret == BLK_MQ_RQ_QUEUE_BUSY)
 			break;
+
+		/*
+		 * We've done the first request. If we have more than 1
+		 * left in the list, set dptr to defer issue.
+		 */
+		if (!dptr && rq_list.next != rq_list.prev)
+			dptr = &driver_list;
 	}
 
 	if (!queued)
@@ -1153,6 +1173,11 @@
 	}
 
 	if (is_sync) {
+		struct blk_mq_queue_data bd = {
+			.rq = rq,
+			.list = NULL,
+			.last = 1
+		};
 		int ret;
 
 		blk_mq_bio_to_request(rq, bio);
@@ -1162,7 +1187,7 @@
 		 * error (busy), just add it to our list as we previously
 		 * would have done
 		 */
-		ret = q->mq_ops->queue_rq(data.hctx, rq, true);
+		ret = q->mq_ops->queue_rq(data.hctx, &bd);
 		if (ret == BLK_MQ_RQ_QUEUE_OK)
 			goto done;
 		else {
diff --git a/drivers/block/mtip32xx/mtip32xx.c b/drivers/block/mtip32xx/mtip32xx.c
index 1bd5f52..3bd7ca9 100644
--- a/drivers/block/mtip32xx/mtip32xx.c
+++ b/drivers/block/mtip32xx/mtip32xx.c
@@ -3775,9 +3775,10 @@
 	return false;
 }
 
-static int mtip_queue_rq(struct blk_mq_hw_ctx *hctx, struct request *rq,
-		bool last)
+static int mtip_queue_rq(struct blk_mq_hw_ctx *hctx,
+			 const struct blk_mq_queue_data *bd)
 {
+	struct request *rq = bd->rq;
 	int ret;
 
 	if (unlikely(mtip_check_unal_depth(hctx, rq)))
diff --git a/drivers/block/null_blk.c b/drivers/block/null_blk.c
index 2671a3f..8433bc8 100644
--- a/drivers/block/null_blk.c
+++ b/drivers/block/null_blk.c
@@ -313,15 +313,15 @@
 	}
 }
 
-static int null_queue_rq(struct blk_mq_hw_ctx *hctx, struct request *rq,
-		bool last)
+static int null_queue_rq(struct blk_mq_hw_ctx *hctx,
+			 const struct blk_mq_queue_data *bd)
 {
-	struct nullb_cmd *cmd = blk_mq_rq_to_pdu(rq);
+	struct nullb_cmd *cmd = blk_mq_rq_to_pdu(bd->rq);
 
-	cmd->rq = rq;
+	cmd->rq = bd->rq;
 	cmd->nq = hctx->driver_data;
 
-	blk_mq_start_request(rq);
+	blk_mq_start_request(bd->rq);
 
 	null_handle_cmd(cmd);
 	return BLK_MQ_RQ_QUEUE_OK;
diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
index c6a27d5..cecd3f9 100644
--- a/drivers/block/virtio_blk.c
+++ b/drivers/block/virtio_blk.c
@@ -158,10 +158,11 @@
 	spin_unlock_irqrestore(&vblk->vqs[qid].lock, flags);
 }
 
-static int virtio_queue_rq(struct blk_mq_hw_ctx *hctx, struct request *req,
-		bool last)
+static int virtio_queue_rq(struct blk_mq_hw_ctx *hctx,
+			   const struct blk_mq_queue_data *bd)
 {
 	struct virtio_blk *vblk = hctx->queue->queuedata;
+	struct request *req = bd->rq;
 	struct virtblk_req *vbr = blk_mq_rq_to_pdu(req);
 	unsigned long flags;
 	unsigned int num;
@@ -222,7 +223,7 @@
 		return BLK_MQ_RQ_QUEUE_ERROR;
 	}
 
-	if (last && virtqueue_kick_prepare(vblk->vqs[qid].vq))
+	if (bd->last && virtqueue_kick_prepare(vblk->vqs[qid].vq))
 		notify = true;
 	spin_unlock_irqrestore(&vblk->vqs[qid].lock, flags);
 
diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index 9eff8a3..161dcc9 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -1858,9 +1858,10 @@
 	blk_mq_complete_request(cmd->request);
 }
 
-static int scsi_queue_rq(struct blk_mq_hw_ctx *hctx, struct request *req,
-		bool last)
+static int scsi_queue_rq(struct blk_mq_hw_ctx *hctx,
+			 const struct blk_mq_queue_data *bd)
 {
+	struct request *req = bd->rq;
 	struct request_queue *q = req->q;
 	struct scsi_device *sdev = q->queuedata;
 	struct Scsi_Host *shost = sdev->host;
diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index c9be158..be01d7a 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -79,7 +79,13 @@
 	struct list_head	tag_list;
 };
 
-typedef int (queue_rq_fn)(struct blk_mq_hw_ctx *, struct request *, bool);
+struct blk_mq_queue_data {
+	struct request *rq;
+	struct list_head *list;
+	bool last;
+};
+
+typedef int (queue_rq_fn)(struct blk_mq_hw_ctx *, const struct blk_mq_queue_data *);
 typedef struct blk_mq_hw_ctx *(map_queue_fn)(struct request_queue *, const int);
 typedef enum blk_eh_timer_return (timeout_fn)(struct request *, bool);
 typedef int (init_hctx_fn)(struct blk_mq_hw_ctx *, void *, unsigned int);