[PATCH] blk-mq: add timer in blk_mq_start_request

[PATCH] blk-mq: add timer in blk_mq_start_request

[Ming Lei]

This way will become consistent with non-mq case, also
avoid to update rq->deadline twice for mq.

The comment said: "We do this early, to ensure we are on
the right CPU.", but no percpu stuff is used in blk_add_timer(),
so it isn't necessary. Even when inserting from plug list, there
is no such guarantee at all.

Signed-off-by: Ming Lei <tom.leiming@gmail.com>
---
 block/blk-mq.c |   17 +----------------
 1 file changed, 1 insertion(+), 16 deletions(-)

diff --git a/block/blk-mq.c b/block/blk-mq.c
index a6ee74e..bf6e13d 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -406,16 +406,7 @@ static void blk_mq_start_request(struct request *rq, bool last)
 	if (unlikely(blk_bidi_rq(rq)))
 		rq->next_rq->resid_len = blk_rq_bytes(rq->next_rq);
 
-	/*
-	 * Just mark start time and set the started bit. Due to memory
-	 * ordering, we know we'll see the correct deadline as long as
-	 * REQ_ATOMIC_STARTED is seen. Use the default queue timeout,
-	 * unless one has been set in the request.
-	 */
-	if (!rq->timeout)
-		rq->deadline = jiffies + q->rq_timeout;
-	else
-		rq->deadline = jiffies + rq->timeout;
+	blk_add_timer(rq);
 
 	/*
 	 * Mark us as started and clear complete. Complete might have been
@@ -967,11 +958,6 @@ static void __blk_mq_insert_request(struct blk_mq_hw_ctx *hctx,
 		list_add_tail(&rq->queuelist, &ctx->rq_list);
 
 	blk_mq_hctx_mark_pending(hctx, ctx);
-
-	/*
-	 * We do this early, to ensure we are on the right CPU.
-	 */
-	blk_add_timer(rq);
 }
 
 void blk_mq_insert_request(struct request *rq, bool at_head, bool run_queue,
@@ -1216,7 +1202,6 @@ static void blk_mq_make_request(struct request_queue *q, struct bio *bio)
 
 		blk_mq_bio_to_request(rq, bio);
 		blk_mq_start_request(rq, true);
-		blk_add_timer(rq);
 
 		/*
 		 * For OK queue, we are done. For error, kill it. Any other
-- 
1.7.9.5

Re: [PATCH] blk-mq: add timer in blk_mq_start_request

[Jens Axboe]

On 06/09/2014 10:16 AM, Ming Lei wrote:
> This way will become consistent with non-mq case, also
> avoid to update rq->deadline twice for mq.
> 
> The comment said: "We do this early, to ensure we are on
> the right CPU.", but no percpu stuff is used in blk_add_timer(),
> so it isn't necessary. Even when inserting from plug list, there
> is no such guarantee at all.

Thanks, this is a good cleanup. That particular comment predates the
strict mappings of the queues, so it doesn't apply anymore anyway.

-- 
Jens Axboe