diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c
index d6af9d431dc631989dce6709971d1fc042e89dc1..dfd81cab57888b33a39b7b8b8e7d8d1c3bae0506 100644
--- a/block/blk-mq-tag.c
+++ b/block/blk-mq-tag.c
@@ -39,16 +39,20 @@ void __blk_mq_tag_busy(struct blk_mq_hw_ctx *hctx)
 {
 	unsigned int users;
 
+	/*
+	 * calling test_bit() prior to test_and_set_bit() is intentional,
+	 * it avoids dirtying the cacheline if the queue is already active.
+	 */
 	if (blk_mq_is_shared_tags(hctx->flags)) {
 		struct request_queue *q = hctx->queue;
 
-		if (test_bit(QUEUE_FLAG_HCTX_ACTIVE, &q->queue_flags))
+		if (test_bit(QUEUE_FLAG_HCTX_ACTIVE, &q->queue_flags) ||
+		    test_and_set_bit(QUEUE_FLAG_HCTX_ACTIVE, &q->queue_flags))
 			return;
-		set_bit(QUEUE_FLAG_HCTX_ACTIVE, &q->queue_flags);
 	} else {
-		if (test_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
+		if (test_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state) ||
+		    test_and_set_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state))
 			return;
-		set_bit(BLK_MQ_S_TAG_ACTIVE, &hctx->state);
 	}
 
 	users = atomic_inc_return(&hctx->tags->active_queues);