diff --git a/block/blk-core.c b/block/blk-core.c
index dba0a83..b1dd4e0 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -390,7 +390,7 @@
 		 * be drained.  Check all the queues and counters.
 		 */
 		if (drain_all) {
-			struct blk_flush_queue *fq = blk_get_flush_queue(q);
+			struct blk_flush_queue *fq = blk_get_flush_queue(q, NULL);
 			drain |= !list_empty(&q->queue_head);
 			for (i = 0; i < 2; i++) {
 				drain |= q->nr_rqs[i];
diff --git a/block/blk-flush.c b/block/blk-flush.c
index 9bc5b4f..004d95e 100644
--- a/block/blk-flush.c
+++ b/block/blk-flush.c
@@ -223,7 +223,7 @@
 	bool queued = false;
 	struct request *rq, *n;
 	unsigned long flags = 0;
-	struct blk_flush_queue *fq = blk_get_flush_queue(q);
+	struct blk_flush_queue *fq = blk_get_flush_queue(q, flush_rq->mq_ctx);
 
 	if (q->mq_ops) {
 		spin_lock_irqsave(&fq->mq_flush_lock, flags);
@@ -319,7 +319,7 @@
 static void flush_data_end_io(struct request *rq, int error)
 {
 	struct request_queue *q = rq->q;
-	struct blk_flush_queue *fq = blk_get_flush_queue(q);
+	struct blk_flush_queue *fq = blk_get_flush_queue(q, NULL);
 
 	/*
 	 * After populating an empty queue, kick it to avoid stall.  Read
@@ -333,11 +333,10 @@
 {
 	struct request_queue *q = rq->q;
 	struct blk_mq_hw_ctx *hctx;
-	struct blk_mq_ctx *ctx;
+	struct blk_mq_ctx *ctx = rq->mq_ctx;
 	unsigned long flags;
-	struct blk_flush_queue *fq = blk_get_flush_queue(q);
+	struct blk_flush_queue *fq = blk_get_flush_queue(q, ctx);
 
-	ctx = rq->mq_ctx;
 	hctx = q->mq_ops->map_queue(q, ctx->cpu);
 
 	/*
@@ -367,7 +366,7 @@
 	struct request_queue *q = rq->q;
 	unsigned int fflags = q->flush_flags;	/* may change, cache */
 	unsigned int policy = blk_flush_policy(fflags, rq);
-	struct blk_flush_queue *fq = blk_get_flush_queue(q);
+	struct blk_flush_queue *fq = blk_get_flush_queue(q, rq->mq_ctx);
 
 	/*
 	 * @policy now records what operations need to be done.  Adjust
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 59ca796..53b6def1 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -518,7 +518,8 @@
 struct request *blk_mq_tag_to_rq(struct blk_mq_tags *tags, unsigned int tag)
 {
 	struct request *rq = tags->rqs[tag];
-	struct blk_flush_queue *fq = blk_get_flush_queue(rq->q);
+	/* mq_ctx of flush rq is always cloned from the corresponding req */
+	struct blk_flush_queue *fq = blk_get_flush_queue(rq->q, rq->mq_ctx);
 
 	if (!is_flush_request(rq, fq, tag))
 		return rq;
diff --git a/block/blk.h b/block/blk.h
index 9eaa6e9..7ecdd85 100644
--- a/block/blk.h
+++ b/block/blk.h
@@ -29,7 +29,7 @@
 extern struct ida blk_queue_ida;
 
 static inline struct blk_flush_queue *blk_get_flush_queue(
-		struct request_queue *q)
+		struct request_queue *q, struct blk_mq_ctx *ctx)
 {
 	return q->fq;
 }
@@ -106,7 +106,7 @@
 static inline struct request *__elv_next_request(struct request_queue *q)
 {
 	struct request *rq;
-	struct blk_flush_queue *fq = blk_get_flush_queue(q);
+	struct blk_flush_queue *fq = blk_get_flush_queue(q, NULL);
 
 	while (1) {
 		if (!list_empty(&q->queue_head)) {