diff --git a/block/blk-core.c b/block/blk-core.c
index fd389a16013ce714a6334288c47a5c34c516b930..35a87c06276efc1499366f54c1f9d4d218b394f3 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -1643,7 +1643,13 @@ void blk_flush_plug(struct blk_plug *plug, bool from_schedule)
 		flush_plug_callbacks(plug, from_schedule);
 	if (!rq_list_empty(plug->mq_list))
 		blk_mq_flush_plug_list(plug, from_schedule);
-	if (unlikely(!from_schedule && plug->cached_rq))
+	/*
+	 * Unconditionally flush out cached requests, even if the unplug
+	 * event came from schedule. Since we know hold references to the
+	 * queue for cached requests, we don't want a blocked task holding
+	 * up a queue freeze/quiesce event.
+	 */
+	if (unlikely(!rq_list_empty(plug->cached_rq)))
 		blk_mq_free_plug_rqs(plug);
 }
 
diff --git a/block/blk-mq.c b/block/blk-mq.c
index c68aa0a332e1c932312d1c3156781362144730d4..5498454c2164a112c4e7b230344409a405fe1b20 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -410,7 +410,10 @@ __blk_mq_alloc_requests_batch(struct blk_mq_alloc_data *data,
 		tag_mask &= ~(1UL << i);
 		rq = blk_mq_rq_ctx_init(data, tags, tag, alloc_time_ns);
 		rq_list_add(data->cached_rq, rq);
+		nr++;
 	}
+	/* caller already holds a reference, add for remainder */
+	percpu_ref_get_many(&data->q->q_usage_counter, nr - 1);
 	data->nr_tags -= nr;
 
 	return rq_list_pop(data->cached_rq);
@@ -630,10 +633,8 @@ void blk_mq_free_plug_rqs(struct blk_plug *plug)
 {
 	struct request *rq;
 
-	while ((rq = rq_list_pop(&plug->cached_rq)) != NULL) {
-		percpu_ref_get(&rq->q->q_usage_counter);
+	while ((rq = rq_list_pop(&plug->cached_rq)) != NULL)
 		blk_mq_free_request(rq);
-	}
 }
 
 static void req_bio_endio(struct request *rq, struct bio *bio,