[v2] block: Consider only dispatched requests for inflight statistic

Commit Message

Gabriel Krisman Bertazi Sept. 2, 2020, 8:19 p.m. UTC

Jens Axboe <[e-mail]> writes:

> We just need to decide if this makes sense or not. I think we should
> apply this for 5.10, with Ming's suggestion of using
> blk_mq_request_started(). Then I guess we'll see what happens...

Hello,

Here is the second version, then.  But, instead of
blk_mq_request_started as suggested on the review, this uses
blk_mq_rq_state to access the state attribute, since we don't want to
include MQ_RQ_COMPLETE.

Also, improved the commit message a bit.

Thanks,

>8

According to Documentation/block/stat.rst, inflight should not include
I/O requests that are in the queue but not yet dispatched to the device,
but blk-mq identifies as inflight any request that has a tag allocated,
which, for queues without elevator, happens at request allocation time
and before it is queued in the ctx (default case in blk_mq_submit_bio).

In addition, current behavior is different for queues with elevator from
queues without it, since for the former the driver tag is allocated at
dispatch time.  A more precise approach would be to only consider
requests with state MQ_RQ_IN_FLIGHT.

This effectively reverts commit 6131837b1de6 ("blk-mq: count allocated
but not started requests in iostats inflight") to consolidate blk-mq
behavior with itself (elevator case) and with original documentation,
but it differs from the behavior used by the legacy path.

Signed-off-by: Gabriel Krisman Bertazi <[e-mail]>
---
Changes Since v1:
  - Use blk_mq_rq_state to fetch req->state
  - Improve commit message

 block/blk-mq.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

Comments

Gabriel Krisman Bertazi Sept. 15, 2020, 4:11 p.m. UTC | #1

Gabriel Krisman Bertazi <[e-mail]> writes:

> Jens Axboe <[e-mail]> writes:
>
>> We just need to decide if this makes sense or not. I think we should
>> apply this for 5.10, with Ming's suggestion of using
>> blk_mq_request_started(). Then I guess we'll see what happens...
>
> Hello,
>
> Here is the second version, then.  But, instead of
> blk_mq_request_started as suggested on the review, this uses
> blk_mq_rq_state to access the state attribute, since we don't want to
> include MQ_RQ_COMPLETE.
>
> Also, improved the commit message a bit.
>

Hi Jens,

Sorry for the ping.  Have you made a decision here?

Thanks,

Patch

diff --git a/block/blk-mq.c b/block/blk-mq.c
index 0015a1892153..bee55f80fb69 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -105,7 +105,7 @@  static bool blk_mq_check_inflight(struct blk_mq_hw_ctx *hctx,
 {
 	struct mq_inflight *mi = priv;
 
-	if (rq->part == mi->part)
+	if (rq->part == mi->part && blk_mq_rq_state(rq) == MQ_RQ_IN_FLIGHT)
 		mi->inflight[rq_data_dir(rq)]++;
 
 	return true;