[PATCH] mq-deadline: make it clear that __dd_dispatch_request() works on all hw queues

[PATCH] mq-deadline: make it clear that __dd_dispatch_request() works on all hw queues

[Jens Axboe]

Don't pass in the hardware queue to __dd_dispatch_request(), since it
leads the reader to believe that we are returning a request for that
specific hardware queue. That's not how mq-deadline works, the state
for determining which request to serve next is shared across all
hardware queues for a device.

Signed-off-by: Jens Axboe <axboe@kernel.dk>

---

diff --git a/block/mq-deadline.c b/block/mq-deadline.c
index d56972e8ebda..c56f211c8440 100644
--- a/block/mq-deadline.c
+++ b/block/mq-deadline.c
@@ -267,9 +267,8 @@ deadline_next_request(struct deadline_data *dd, int data_dir)
  * deadline_dispatch_requests selects the best request according to
  * read/write expire, fifo_batch, etc
  */
-static struct request *__dd_dispatch_request(struct blk_mq_hw_ctx *hctx)
+static struct request *__dd_dispatch_request(struct deadline_data *dd)
 {
-	struct deadline_data *dd = hctx->queue->elevator->elevator_data;
 	struct request *rq, *next_rq;
 	bool reads, writes;
 	int data_dir;
@@ -372,13 +371,19 @@ static struct request *__dd_dispatch_request(struct blk_mq_hw_ctx *hctx)
 	return rq;
 }
 
+/*
+ * One confusing aspect here is that we get called for a specific
+ * hardware queue, but we return a request that may not be for a
+ * different hardware queue. This is because mq-deadline has shared
+ * state for all hardware queues, in terms of sorting, FIFOs, etc.
+ */
 static struct request *dd_dispatch_request(struct blk_mq_hw_ctx *hctx)
 {
 	struct deadline_data *dd = hctx->queue->elevator->elevator_data;
 	struct request *rq;
 
 	spin_lock(&dd->lock);
-	rq = __dd_dispatch_request(hctx);
+	rq = __dd_dispatch_request(dd);
 	spin_unlock(&dd->lock);
 
 	return rq;

-- 
Jens Axboe

Re: [PATCH] mq-deadline: make it clear that __dd_dispatch_request() works on all hw queues

[Omar Sandoval]

On Fri, Jan 05, 2018 at 02:56:38PM -0700, Jens Axboe wrote:
> Don't pass in the hardware queue to __dd_dispatch_request(), since it
> leads the reader to believe that we are returning a request for that
> specific hardware queue. That's not how mq-deadline works, the state
> for determining which request to serve next is shared across all
> hardware queues for a device.

Reviewed-by: Omar Sandoval <osandov@fb.com>

> Signed-off-by: Jens Axboe <axboe@kernel.dk>
> 
> ---
> 
> diff --git a/block/mq-deadline.c b/block/mq-deadline.c
> index d56972e8ebda..c56f211c8440 100644
> --- a/block/mq-deadline.c
> +++ b/block/mq-deadline.c
> @@ -267,9 +267,8 @@ deadline_next_request(struct deadline_data *dd, int data_dir)
>   * deadline_dispatch_requests selects the best request according to
>   * read/write expire, fifo_batch, etc
>   */
> -static struct request *__dd_dispatch_request(struct blk_mq_hw_ctx *hctx)
> +static struct request *__dd_dispatch_request(struct deadline_data *dd)
>  {
> -	struct deadline_data *dd = hctx->queue->elevator->elevator_data;
>  	struct request *rq, *next_rq;
>  	bool reads, writes;
>  	int data_dir;
> @@ -372,13 +371,19 @@ static struct request *__dd_dispatch_request(struct blk_mq_hw_ctx *hctx)
>  	return rq;
>  }
>  
> +/*
> + * One confusing aspect here is that we get called for a specific
> + * hardware queue, but we return a request that may not be for a
> + * different hardware queue. This is because mq-deadline has shared
> + * state for all hardware queues, in terms of sorting, FIFOs, etc.
> + */
>  static struct request *dd_dispatch_request(struct blk_mq_hw_ctx *hctx)
>  {
>  	struct deadline_data *dd = hctx->queue->elevator->elevator_data;
>  	struct request *rq;
>  
>  	spin_lock(&dd->lock);
> -	rq = __dd_dispatch_request(hctx);
> +	rq = __dd_dispatch_request(dd);
>  	spin_unlock(&dd->lock);
>  
>  	return rq;
> 
> -- 
> Jens Axboe
> 

Re: [PATCH] mq-deadline: make it clear that __dd_dispatch_request() works on all hw queues

[Ming Lei]

On Fri, Jan 05, 2018 at 02:56:38PM -0700, Jens Axboe wrote:
> Don't pass in the hardware queue to __dd_dispatch_request(), since it
> leads the reader to believe that we are returning a request for that
> specific hardware queue. That's not how mq-deadline works, the state
> for determining which request to serve next is shared across all
> hardware queues for a device.
> 
> Signed-off-by: Jens Axboe <axboe@kernel.dk>

Reviewed-by: Ming Lei <ming.lei@redhat.com>

-- 
Ming