From: Pavel Begunkov <asml.silence@gmail.com>
To: Hou Tao <houtao1@huawei.com>,
axboe@kernel.dk, linux-block@vger.kernel.org
Cc: osandov@fb.com, ming.lei@redhat.com
Subject: Re: [PATCH 1/2] block: make rq sector size accessible for block stats
Date: Sat, 25 May 2019 16:52:52 +0300 [thread overview]
Message-ID: <48eca4f7-0a26-f11d-c7f5-46f93884787b@gmail.com> (raw)
In-Reply-To: <76a679b4-cd55-8bca-0020-fb320914f491@gmail.com>
[-- Attachment #1.1: Type: text/plain, Size: 5544 bytes --]
I mean this way
https://lore.kernel.org/linux-block/dd30f4d94aa19956ad4500b1177741fd071ec37f.1558791181.git.asml.silence@gmail.com/T/#u
On 25/05/2019 14:46, Pavel Begunkov wrote:
> That's the same bug I posted a patch about 3+ weeks ago. No answer yet,
> however.
> https://www.spinics.net/lists/linux-block/msg40049.html
>
> I think putting time sampling at __blk_mq_end_request() won't give
> sufficient precision for more sophisticated hybrid polling, because path
> __blk_mq_complete_request() -> __blk_mq_end_request() adds a lot of
> overhead (redirect rq to another cpu, dma unmap, etc). That's the case
> for mentioned patchset
> https://www.spinics.net/lists/linux-block/msg40044.html
>
> It works OK for 50-80us requests (e.g. SSD read), but is not good enough
> for 20us and less (e.g. SSD write). I think it would be better to save
> the time in advance, and use it later.
>
> By the way, it seems, that __blk_mq_end_request() is really better place
> to record stats, as not all drivers use __blk_mq_complete_request(),
> even though they don't implement blk_poll().
>
>
> On 21/05/2019 10:59, Hou Tao wrote:
>> Currently rq->data_len will be decreased by partial completion or
>> zeroed by completion, so when blk_stat_add() is invoked, data_len
>> will be zero and there will never be samples in poll_cb because
>> blk_mq_poll_stats_bkt() will return -1 if data_len is zero.
>>
>> We could move blk_stat_add() back to __blk_mq_complete_request(),
>> but that would make the effort of trying to call ktime_get_ns()
>> once in vain. Instead we can reuse throtl_size field, and use
>> it for both block stats and block throttle, and adjust the
>> logic in blk_mq_poll_stats_bkt() accordingly.
>>
>> Fixes: 4bc6339a583c ("block: move blk_stat_add() to __blk_mq_end_request()")
>> Signed-off-by: Hou Tao <houtao1@huawei.com>
>> ---
>> block/blk-mq.c | 11 +++++------
>> block/blk-throttle.c | 3 ++-
>> include/linux/blkdev.h | 15 ++++++++++++---
>> 3 files changed, 19 insertions(+), 10 deletions(-)
>>
>> diff --git a/block/blk-mq.c b/block/blk-mq.c
>> index 08a6248d8536..4d1462172f0f 100644
>> --- a/block/blk-mq.c
>> +++ b/block/blk-mq.c
>> @@ -44,12 +44,12 @@ static void blk_mq_poll_stats_fn(struct blk_stat_callback *cb);
>>
>> static int blk_mq_poll_stats_bkt(const struct request *rq)
>> {
>> - int ddir, bytes, bucket;
>> + int ddir, sectors, bucket;
>>
>> ddir = rq_data_dir(rq);
>> - bytes = blk_rq_bytes(rq);
>> + sectors = blk_rq_stats_sectors(rq);
>>
>> - bucket = ddir + 2*(ilog2(bytes) - 9);
>> + bucket = ddir + 2 * ilog2(sectors);
>>
>> if (bucket < 0)
>> return -1;
>> @@ -329,6 +329,7 @@ static struct request *blk_mq_rq_ctx_init(struct blk_mq_alloc_data *data,
>> else
>> rq->start_time_ns = 0;
>> rq->io_start_time_ns = 0;
>> + rq->stats_sectors = 0;
>> rq->nr_phys_segments = 0;
>> #if defined(CONFIG_BLK_DEV_INTEGRITY)
>> rq->nr_integrity_segments = 0;
>> @@ -678,9 +679,7 @@ void blk_mq_start_request(struct request *rq)
>>
>> if (test_bit(QUEUE_FLAG_STATS, &q->queue_flags)) {
>> rq->io_start_time_ns = ktime_get_ns();
>> -#ifdef CONFIG_BLK_DEV_THROTTLING_LOW
>> - rq->throtl_size = blk_rq_sectors(rq);
>> -#endif
>> + rq->stats_sectors = blk_rq_sectors(rq);
>> rq->rq_flags |= RQF_STATS;
>> rq_qos_issue(q, rq);
>> }
>> diff --git a/block/blk-throttle.c b/block/blk-throttle.c
>> index 1b97a73d2fb1..88459a4ac704 100644
>> --- a/block/blk-throttle.c
>> +++ b/block/blk-throttle.c
>> @@ -2249,7 +2249,8 @@ void blk_throtl_stat_add(struct request *rq, u64 time_ns)
>> struct request_queue *q = rq->q;
>> struct throtl_data *td = q->td;
>>
>> - throtl_track_latency(td, rq->throtl_size, req_op(rq), time_ns >> 10);
>> + throtl_track_latency(td, blk_rq_stats_sectors(rq), req_op(rq),
>> + time_ns >> 10);
>> }
>>
>> void blk_throtl_bio_endio(struct bio *bio)
>> diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
>> index 1aafeb923e7b..68a0841d3554 100644
>> --- a/include/linux/blkdev.h
>> +++ b/include/linux/blkdev.h
>> @@ -202,9 +202,12 @@ struct request {
>> #ifdef CONFIG_BLK_WBT
>> unsigned short wbt_flags;
>> #endif
>> -#ifdef CONFIG_BLK_DEV_THROTTLING_LOW
>> - unsigned short throtl_size;
>> -#endif
>> + /*
>> + * rq sectors used for blk stats. It has the same value
>> + * with blk_rq_sectors(rq), except that it never be zeroed
>> + * by completion.
>> + */
>> + unsigned short stats_sectors;
>>
>> /*
>> * Number of scatter-gather DMA addr+len pairs after
>> @@ -892,6 +895,7 @@ static inline struct request_queue *bdev_get_queue(struct block_device *bdev)
>> * blk_rq_err_bytes() : bytes left till the next error boundary
>> * blk_rq_sectors() : sectors left in the entire request
>> * blk_rq_cur_sectors() : sectors left in the current segment
>> + * blk_rq_stats_sectors() : sectors of the entire request used for stats
>> */
>> static inline sector_t blk_rq_pos(const struct request *rq)
>> {
>> @@ -920,6 +924,11 @@ static inline unsigned int blk_rq_cur_sectors(const struct request *rq)
>> return blk_rq_cur_bytes(rq) >> SECTOR_SHIFT;
>> }
>>
>> +static inline unsigned int blk_rq_stats_sectors(const struct request *rq)
>> +{
>> + return rq->stats_sectors;
>> +}
>> +
>> #ifdef CONFIG_BLK_DEV_ZONED
>> static inline unsigned int blk_rq_zone_no(struct request *rq)
>> {
>>
>
--
Yours sincerely,
Pavel Begunkov
[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]
next prev parent reply other threads:[~2019-05-25 13:53 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-05-21 7:59 [PATCH 0/2] fixes for block stats Hou Tao
2019-05-21 7:59 ` [PATCH 1/2] block: make rq sector size accessible " Hou Tao
2019-05-25 11:46 ` Pavel Begunkov
2019-05-25 13:52 ` Pavel Begunkov [this message]
2019-05-27 11:39 ` Hou Tao
2019-05-28 18:25 ` Pavel Begunkov
2019-09-15 16:52 ` Pavel Begunkov
2019-05-21 7:59 ` [PATCH 2/2] block: also check RQF_STATS in blk_mq_need_time_stamp() Hou Tao
2019-05-25 5:12 ` [PATCH 0/2] fixes for block stats Hou Tao
2019-09-15 22:02 ` Jens Axboe
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=48eca4f7-0a26-f11d-c7f5-46f93884787b@gmail.com \
--to=asml.silence@gmail.com \
--cc=axboe@kernel.dk \
--cc=houtao1@huawei.com \
--cc=linux-block@vger.kernel.org \
--cc=ming.lei@redhat.com \
--cc=osandov@fb.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).