From: Yu Kuai <yukuai3@huawei.com>
Now that blk-mq also use per_cpu counter to trace inflight as bio-based
device, they can be replaced by part_in_flight() and part_in_flight_rw()
directly.
Noted that there will be a change that inflight will be accounted from
blk_account_io_start() instead of blk_mq_start_request(). This also fix
an inconsistence for rq-based device that if there are rq allocated but
not started, io_ticks will be updated from blk_account_io_start() but not
from part_stat_show() or diskstats_show(). For consequence, for example:
blk_account_io_start -> 0s
-> something is wrong with driver, rq can't be dispatched to driver
-> finially the driver recovers
blk_mq_start_request -> 10s
Then in this case, if user is using "iostat 1", then user will found
that 'util' is 0 from 0s-9s, because diskstats_show() doesn't update
io_ticks, then in 9s-10s user will found 'util' is 1000%.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
---
block/blk-mq.c | 32 --------------------------------
block/blk-mq.h | 5 -----
block/genhd.c | 18 +++---------------
3 files changed, 3 insertions(+), 52 deletions(-)
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 82045f0ab5ba..dfbb4e24f04a 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -87,38 +87,6 @@ struct mq_inflight {
unsigned int inflight[2];
};
-static bool blk_mq_check_inflight(struct request *rq, void *priv)
-{
- struct mq_inflight *mi = priv;
-
- if (rq->part && blk_do_io_stat(rq) &&
- (!mi->part->bd_partno || rq->part == mi->part) &&
- blk_mq_rq_state(rq) == MQ_RQ_IN_FLIGHT)
- mi->inflight[rq_data_dir(rq)]++;
-
- return true;
-}
-
-unsigned int blk_mq_in_flight(struct request_queue *q,
- struct block_device *part)
-{
- struct mq_inflight mi = { .part = part };
-
- blk_mq_queue_tag_busy_iter(q, blk_mq_check_inflight, &mi);
-
- return mi.inflight[0] + mi.inflight[1];
-}
-
-void blk_mq_in_flight_rw(struct request_queue *q, struct block_device *part,
- unsigned int inflight[2])
-{
- struct mq_inflight mi = { .part = part };
-
- blk_mq_queue_tag_busy_iter(q, blk_mq_check_inflight, &mi);
- inflight[0] = mi.inflight[0];
- inflight[1] = mi.inflight[1];
-}
-
void blk_freeze_queue_start(struct request_queue *q)
{
mutex_lock(&q->mq_freeze_lock);
diff --git a/block/blk-mq.h b/block/blk-mq.h
index f75a9ecfebde..aa92e1317f18 100644
--- a/block/blk-mq.h
+++ b/block/blk-mq.h
@@ -236,11 +236,6 @@ static inline bool blk_mq_hw_queue_mapped(struct blk_mq_hw_ctx *hctx)
return hctx->nr_ctx && hctx->tags;
}
-unsigned int blk_mq_in_flight(struct request_queue *q,
- struct block_device *part);
-void blk_mq_in_flight_rw(struct request_queue *q, struct block_device *part,
- unsigned int inflight[2]);
-
static inline void blk_mq_put_dispatch_budget(struct request_queue *q,
int budget_token)
{
diff --git a/block/genhd.c b/block/genhd.c
index 782a42718965..9c64d34b7dc3 100644
--- a/block/genhd.c
+++ b/block/genhd.c
@@ -954,15 +954,10 @@ 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);
@@ -1003,13 +998,9 @@ ssize_t part_inflight_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);
unsigned int inflight[2];
- if (queue_is_mq(q))
- blk_mq_in_flight_rw(q, bdev, inflight);
- else
- part_in_flight_rw(bdev, inflight);
+ part_in_flight_rw(bdev, inflight);
return sprintf(buf, "%8u %8u\n", inflight[0], inflight[1]);
}
@@ -1251,11 +1242,8 @@ 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);
--
2.39.2On Fri, Mar 22 2024 at 11:59P -0400,
Yu Kuai <yukuai1@huaweicloud.com> wrote:
> From: Yu Kuai <yukuai3@huawei.com>
>
> Now that blk-mq also use per_cpu counter to trace inflight as bio-based
> device, they can be replaced by part_in_flight() and part_in_flight_rw()
> directly.
Please reference the commit that enabled this, e.g.:
With commit XXXXX ("commit subject") blk-mq was updated to use per_cpu
counters to track inflight IO same as bio-based devices, so replace
blk_mq_in_flight* with part_in_flight() and part_in_flight_rw()
accordingly.
(I'm not seeing the commit in question, but I only took a quick look).
Mike
Hi,
在 2024/03/24 0:05, Mike Snitzer 写道:
> On Fri, Mar 22 2024 at 11:59P -0400,
> Yu Kuai <yukuai1@huaweicloud.com> wrote:
>
>> From: Yu Kuai <yukuai3@huawei.com>
>>
>> Now that blk-mq also use per_cpu counter to trace inflight as bio-based
>> device, they can be replaced by part_in_flight() and part_in_flight_rw()
>> directly.
>
> Please reference the commit that enabled this, e.g.:
>
> With commit XXXXX ("commit subject") blk-mq was updated to use per_cpu
> counters to track inflight IO same as bio-based devices, so replace
> blk_mq_in_flight* with part_in_flight() and part_in_flight_rw()
> accordingly.
Patch 1 in this set do this, so there is no commit xxx yet.
Thanks,
Kuai
>
> (I'm not seeing the commit in question, but I only took a quick look).
>
> Mike
> .
>
On Sat, Mar 23 2024 at 10:11P -0400,
Yu Kuai <yukuai1@huaweicloud.com> wrote:
> Hi,
>
> 在 2024/03/24 0:05, Mike Snitzer 写道:
> > On Fri, Mar 22 2024 at 11:59P -0400,
> > Yu Kuai <yukuai1@huaweicloud.com> wrote:
> >
> > > From: Yu Kuai <yukuai3@huawei.com>
> > >
> > > Now that blk-mq also use per_cpu counter to trace inflight as bio-based
> > > device, they can be replaced by part_in_flight() and part_in_flight_rw()
> > > directly.
> >
> > Please reference the commit that enabled this, e.g.:
> >
> > With commit XXXXX ("commit subject") blk-mq was updated to use per_cpu
> > counters to track inflight IO same as bio-based devices, so replace
> > blk_mq_in_flight* with part_in_flight() and part_in_flight_rw()
> > accordingly.
>
> Patch 1 in this set do this, so there is no commit xxx yet.
>
> Thanks,
> Kuai
Would've helped if I looked at 1/2, but please say:
With the previous commit blk-mq was updated to use per_cpu ...
Hi,
在 2024/03/25 5:57, Mike Snitzer 写道:
> On Sat, Mar 23 2024 at 10:11P -0400,
> Yu Kuai <yukuai1@huaweicloud.com> wrote:
>
>> Hi,
>>
>> 在 2024/03/24 0:05, Mike Snitzer 写道:
>>> On Fri, Mar 22 2024 at 11:59P -0400,
>>> Yu Kuai <yukuai1@huaweicloud.com> wrote:
>>>
>>>> From: Yu Kuai <yukuai3@huawei.com>
>>>>
>>>> Now that blk-mq also use per_cpu counter to trace inflight as bio-based
>>>> device, they can be replaced by part_in_flight() and part_in_flight_rw()
>>>> directly.
>>>
>>> Please reference the commit that enabled this, e.g.:
>>>
>>> With commit XXXXX ("commit subject") blk-mq was updated to use per_cpu
>>> counters to track inflight IO same as bio-based devices, so replace
>>> blk_mq_in_flight* with part_in_flight() and part_in_flight_rw()
>>> accordingly.
>>
>> Patch 1 in this set do this, so there is no commit xxx yet.
>>
>> Thanks,
>> Kuai
>
> Would've helped if I looked at 1/2, but please say:
>
> With the previous commit blk-mq was updated to use per_cpu ...
OK, will pay attention to it later. :)
Thanks,
Kuai
> .
>
© 2016 - 2026 Red Hat, Inc.