mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH v2 for-6.10/block 0/2] block: fix user-aware inaccurate util
@ 2024-05-09 12:37 Yu Kuai
  2024-05-09 12:37 ` [PATCH for-6.10/block 1/2] block: support to account io_ticks precisely Yu Kuai
                   ` (2 more replies)
  0 siblings, 3 replies; 4+ messages in thread
From: Yu Kuai @ 2024-05-09 12:37 UTC (permalink / raw)
  To: axboe, ming.lei, hch, mpatocka, snitzer
  Cc: linux-block, linux-kernel, yukuai3, yukuai1, yi.zhang, yangerkun

From: Yu Kuai <yukuai3@huawei.com>

Changes in v2:
 - just replace blk_mq_in_flight() with part_in_flight() for
 diskstats_show() and part_stat_show() in patch 2;

Yu Kuai (2):
  block: support to account io_ticks precisely
  block: fix that util can be greater than 100%

 block/blk-core.c  |  9 +++++----
 block/blk-merge.c |  2 ++
 block/blk-mq.c    |  4 ++++
 block/blk.h       |  1 +
 block/genhd.c     | 14 +++-----------
 5 files changed, 15 insertions(+), 15 deletions(-)

-- 
2.39.2


^ permalink raw reply	[flat|nested] 4+ messages in thread

* [PATCH for-6.10/block 1/2] block: support to account io_ticks precisely
  2024-05-09 12:37 [PATCH v2 for-6.10/block 0/2] block: fix user-aware inaccurate util Yu Kuai
@ 2024-05-09 12:37 ` Yu Kuai
  2024-05-09 12:37 ` [PATCH for-6.10/block 2/2] block: fix that util can be greater than 100% Yu Kuai
  2024-05-09 14:02 ` [PATCH v2 for-6.10/block 0/2] block: fix user-aware inaccurate util Jens Axboe
  2 siblings, 0 replies; 4+ messages in thread
From: Yu Kuai @ 2024-05-09 12:37 UTC (permalink / raw)
  To: axboe, ming.lei, hch, mpatocka, snitzer
  Cc: linux-block, linux-kernel, yukuai3, yukuai1, yi.zhang, yangerkun

From: Yu Kuai <yukuai3@huawei.com>

Currently, io_ticks is accounted based on sampling, specifically
update_io_ticks() will always account io_ticks by 1 jiffies from
bdev_start_io_acct()/blk_account_io_start(), and the result can be
inaccurate, for example(HZ is 250):

Test script:
fio -filename=/dev/sda -bs=4k -rw=write -direct=1 -name=test -thinktime=4ms

Test result: util is about 90%, while the disk is really idle.

This behaviour is introduced by commit 5b18b5a73760 ("block: delete
part_round_stats and switch to less precise counting"), however, there
was a key point that is missed that this patch also improve performance
a lot:

Before the commit:
part_round_stats:
  if (part->stamp != now)
   stats |= 1;

  part_in_flight()
  -> there can be lots of task here in 1 jiffies.
  part_round_stats_single()
   __part_stat_add()
  part->stamp = now;

After the commit:
update_io_ticks:
  stamp = part->bd_stamp;
  if (time_after(now, stamp))
   if (try_cmpxchg())
    __part_stat_add()
    -> only one task can reach here in 1 jiffies.

Hence in order to account io_ticks precisely, we only need to know if
there are IO inflight at most once in one jiffies. Noted that for
rq-based device, iterating tags should not be used here because
'tags->lock' is grabbed in blk_mq_find_and_get_req(), hence
part_stat_lock_inc/dec() and part_in_flight() is used to trace inflight.
The additional overhead is quite little:

 - per cpu add/dec for each IO for rq-based device;
 - per cpu sum for each jiffies;

And it's verified by null-blk that there are no performance degration
under heavy IO pressure.

Fixes: 5b18b5a73760 ("block: delete part_round_stats and switch to less precise counting")
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
---
 block/blk-core.c  | 9 +++++----
 block/blk-merge.c | 2 ++
 block/blk-mq.c    | 4 ++++
 block/blk.h       | 1 +
 block/genhd.c     | 2 +-
 5 files changed, 13 insertions(+), 5 deletions(-)

diff --git a/block/blk-core.c b/block/blk-core.c
index 47400a4fe851..9ead80d6c6f0 100644
--- a/block/blk-core.c
+++ b/block/blk-core.c
@@ -978,10 +978,11 @@ void update_io_ticks(struct block_device *part, unsigned long now, bool end)
 	unsigned long stamp;
 again:
 	stamp = READ_ONCE(part->bd_stamp);
-	if (unlikely(time_after(now, stamp))) {
-		if (likely(try_cmpxchg(&part->bd_stamp, &stamp, now)))
-			__part_stat_add(part, io_ticks, end ? now - stamp : 1);
-	}
+	if (unlikely(time_after(now, stamp)) &&
+	    likely(try_cmpxchg(&part->bd_stamp, &stamp, now)) &&
+	    (end || part_in_flight(part)))
+		__part_stat_add(part, io_ticks, now - stamp);
+
 	if (part->bd_partno) {
 		part = bdev_whole(part);
 		goto again;
diff --git a/block/blk-merge.c b/block/blk-merge.c
index f64115d72f3d..8534c35e0497 100644
--- a/block/blk-merge.c
+++ b/block/blk-merge.c
@@ -780,6 +780,8 @@ static void blk_account_io_merge_request(struct request *req)
 	if (blk_do_io_stat(req)) {
 		part_stat_lock();
 		part_stat_inc(req->part, merges[op_stat_group(req_op(req))]);
+		part_stat_local_dec(req->part,
+				    in_flight[op_is_write(req_op(req))]);
 		part_stat_unlock();
 	}
 }
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 9f677ea85a52..8e01e4b32e10 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -996,6 +996,8 @@ static inline void blk_account_io_done(struct request *req, u64 now)
 		update_io_ticks(req->part, jiffies, true);
 		part_stat_inc(req->part, ios[sgrp]);
 		part_stat_add(req->part, nsecs[sgrp], now - req->start_time_ns);
+		part_stat_local_dec(req->part,
+				    in_flight[op_is_write(req_op(req))]);
 		part_stat_unlock();
 	}
 }
@@ -1018,6 +1020,8 @@ static inline void blk_account_io_start(struct request *req)
 
 		part_stat_lock();
 		update_io_ticks(req->part, jiffies, false);
+		part_stat_local_inc(req->part,
+				    in_flight[op_is_write(req_op(req))]);
 		part_stat_unlock();
 	}
 }
diff --git a/block/blk.h b/block/blk.h
index 0e46c5d30d5a..6e94c10af798 100644
--- a/block/blk.h
+++ b/block/blk.h
@@ -366,6 +366,7 @@ static inline bool blk_do_io_stat(struct request *rq)
 }
 
 void update_io_ticks(struct block_device *part, unsigned long now, bool end);
+unsigned int part_in_flight(struct block_device *part);
 
 static inline void req_set_nomerge(struct request_queue *q, struct request *req)
 {
diff --git a/block/genhd.c b/block/genhd.c
index dec2ee338fb4..8f1163d2d171 100644
--- a/block/genhd.c
+++ b/block/genhd.c
@@ -118,7 +118,7 @@ static void part_stat_read_all(struct block_device *part,
 	}
 }
 
-static unsigned int part_in_flight(struct block_device *part)
+unsigned int part_in_flight(struct block_device *part)
 {
 	unsigned int inflight = 0;
 	int cpu;
-- 
2.39.2


^ permalink raw reply	[flat|nested] 4+ messages in thread

* [PATCH for-6.10/block 2/2] block: fix that util can be greater than 100%
  2024-05-09 12:37 [PATCH v2 for-6.10/block 0/2] block: fix user-aware inaccurate util Yu Kuai
  2024-05-09 12:37 ` [PATCH for-6.10/block 1/2] block: support to account io_ticks precisely Yu Kuai
@ 2024-05-09 12:37 ` Yu Kuai
  2024-05-09 14:02 ` [PATCH v2 for-6.10/block 0/2] block: fix user-aware inaccurate util Jens Axboe
  2 siblings, 0 replies; 4+ messages in thread
From: Yu Kuai @ 2024-05-09 12:37 UTC (permalink / raw)
  To: axboe, ming.lei, hch, mpatocka, snitzer
  Cc: linux-block, linux-kernel, yukuai3, yukuai1, yi.zhang, yangerkun

From: Yu Kuai <yukuai3@huawei.com>

util means the percentage that disk has IO, and theoretically it should
not be greater than 100%. However, there is a gap for rq-based disk:

io_ticks will be updated when rq is allocated, however, before such rq
dispatch to driver, it will not be account as inflight from
blk_mq_start_request() hence diskstats_show()/part_stat_show() will not
update io_ticks. For example:

1) at t0, issue a new IO, rq is allocated, and blk_account_io_start()
update io_ticks;

2) something is wrong with drivers, and the rq can't be dispatched;

3) at t0 + 10s, drivers recovers and rq is dispatched and done, io_ticks
is updated;

Then if user is using "iostat 1" to monitor "util", between t0 - t0+9s,
util will be zero, and between t0+9s - t0+10s, util will be 1000%.

Fix this problem by updating io_ticks from diskstats_show() and
part_stat_show() if there are rq allocated.

Signed-off-by: Yu Kuai <yukuai3@huawei.com>
---
 block/genhd.c | 12 ++----------
 1 file changed, 2 insertions(+), 10 deletions(-)

diff --git a/block/genhd.c b/block/genhd.c
index 8f1163d2d171..7f39fbe60753 100644
--- a/block/genhd.c
+++ b/block/genhd.c
@@ -951,15 +951,10 @@ ssize_t part_stat_show(struct device *dev,
 		       struct device_attribute *attr, char *buf)
 {
 	struct block_device *bdev = dev_to_bdev(dev);
-	struct request_queue *q = bdev_get_queue(bdev);
 	struct disk_stats stat;
 	unsigned int inflight;
 
-	if (queue_is_mq(q))
-		inflight = blk_mq_in_flight(q, bdev);
-	else
-		inflight = part_in_flight(bdev);
-
+	inflight = part_in_flight(bdev);
 	if (inflight) {
 		part_stat_lock();
 		update_io_ticks(bdev, jiffies, true);
@@ -1256,11 +1251,8 @@ static int diskstats_show(struct seq_file *seqf, void *v)
 	xa_for_each(&gp->part_tbl, idx, hd) {
 		if (bdev_is_partition(hd) && !bdev_nr_sectors(hd))
 			continue;
-		if (queue_is_mq(gp->queue))
-			inflight = blk_mq_in_flight(gp->queue, hd);
-		else
-			inflight = part_in_flight(hd);
 
+		inflight = part_in_flight(hd);
 		if (inflight) {
 			part_stat_lock();
 			update_io_ticks(hd, jiffies, true);
-- 
2.39.2


^ permalink raw reply	[flat|nested] 4+ messages in thread

* Re: [PATCH v2 for-6.10/block 0/2] block: fix user-aware inaccurate util
  2024-05-09 12:37 [PATCH v2 for-6.10/block 0/2] block: fix user-aware inaccurate util Yu Kuai
  2024-05-09 12:37 ` [PATCH for-6.10/block 1/2] block: support to account io_ticks precisely Yu Kuai
  2024-05-09 12:37 ` [PATCH for-6.10/block 2/2] block: fix that util can be greater than 100% Yu Kuai
@ 2024-05-09 14:02 ` Jens Axboe
  2 siblings, 0 replies; 4+ messages in thread
From: Jens Axboe @ 2024-05-09 14:02 UTC (permalink / raw)
  To: ming.lei, hch, mpatocka, snitzer, Yu Kuai
  Cc: linux-block, linux-kernel, yukuai3, yi.zhang, yangerkun


On Thu, 09 May 2024 20:37:15 +0800, Yu Kuai wrote:
> Changes in v2:
>  - just replace blk_mq_in_flight() with part_in_flight() for
>  diskstats_show() and part_stat_show() in patch 2;
> 
> Yu Kuai (2):
>   block: support to account io_ticks precisely
>   block: fix that util can be greater than 100%
> 
> [...]

Applied, thanks!

[1/2] block: support to account io_ticks precisely
      commit: 99dc422335d8b2bd4d105797241d3e715bae90e9
[2/2] block: fix that util can be greater than 100%
      commit: 7be835694daebbb4adffbc461519081aa0cf28e1

Best regards,
-- 
Jens Axboe




^ permalink raw reply	[flat|nested] 4+ messages in thread

end of thread, other threads:[~2024-05-09 14:02 UTC | newest]

Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2024-05-09 12:37 [PATCH v2 for-6.10/block 0/2] block: fix user-aware inaccurate util Yu Kuai
2024-05-09 12:37 ` [PATCH for-6.10/block 1/2] block: support to account io_ticks precisely Yu Kuai
2024-05-09 12:37 ` [PATCH for-6.10/block 2/2] block: fix that util can be greater than 100% Yu Kuai
2024-05-09 14:02 ` [PATCH v2 for-6.10/block 0/2] block: fix user-aware inaccurate util Jens Axboe

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®