[PATCH] blk-mq: don't lose requests if a stopped queue restarts

[Shaohua Li <shli@fb.com>]

Normally if driver is busy to dispatch a request the logic is like below:
block layer:					driver:
	__blk_mq_run_hw_queue
a.						blk_mq_stop_hw_queue
b.	rq add to ctx->dispatch

later:
1.						blk_mq_start_hw_queue
2.	__blk_mq_run_hw_queue

But it's possible step 1-2 runs between a and b. And since rq isn't in
ctx->dispatch yet, step 2 will not run rq. The rq might get lost if
there are no subsequent requests kick in.

Signed-off-by: Shaohua Li <shli@fb.com>
---
 block/blk-mq.c | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index ade8a2d..e6822a2 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -772,6 +772,7 @@ static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)
 
 	WARN_ON(!cpumask_test_cpu(raw_smp_processor_id(), hctx->cpumask));
 
+again:
 	if (unlikely(test_bit(BLK_MQ_S_STOPPED, &hctx->state)))
 		return;
 
@@ -853,8 +854,16 @@ static void __blk_mq_run_hw_queue(struct blk_mq_hw_ctx *hctx)
 	 */
 	if (!list_empty(&rq_list)) {
 		spin_lock(&hctx->lock);
-		list_splice(&rq_list, &hctx->dispatch);
+		list_splice_init(&rq_list, &hctx->dispatch);
 		spin_unlock(&hctx->lock);
+		/*
+		 * the queue is expected stopped with BLK_MQ_RQ_QUEUE_BUSY, but
+		 * it's possible the queue is stopped and restarted again
+		 * before this. Queue restart will dispatch requests. And since
+		 * requests in rq_list aren't added into hctx->dispatch yet,
+		 * the requests in rq_list might get lost.
+		 **/
+		goto again;
 	}
 }
 
-- 
1.8.1