Re: [PATCH -next] blk-throttle: delay the setting of 'BIO_THROTTLED' to when throttle is done

[yukuai (C)]

在 2022/05/18 12:07, Ming Lei 写道:
> On Tue, May 17, 2022 at 09:49:09PM +0800, Yu Kuai wrote:
> > commit 9f5ede3c01f9 ("block: throttle split bio in case of iops limit")
> > introduce a new problem, for example:
> >
> > [root@localhost ~]# echo "8:0 1024" > /sys/fs/cgroup/blkio/blkio.throttle.write_bps_device
> > [root@localhost ~]# echo $$ > /sys/fs/cgroup/blkio/cgroup.procs
> > [root@localhost ~]# dd if=/dev/zero of=/dev/sda bs=10k count=1 oflag=direct &
> > [1] 620
> > [root@localhost ~]# dd if=/dev/zero of=/dev/sda bs=10k count=1 oflag=direct &
> > [2] 626
> > [root@localhost ~]# 1+0 records in
> > 1+0 records out
> > 10240 bytes (10 kB, 10 KiB) copied, 10.0038 s, 1.0 kB/s1+0 records in
> > 1+0 records out
> >
> > 10240 bytes (10 kB, 10 KiB) copied, 9.23076 s, 1.1 kB/s
> > -> the second bio is issued after 10s instead of 20s.
> >
> > This is because if some bios are already queued, current bio is queued
> > directly and the flag 'BIO_THROTTLED' is set. And later, when former
> > bios are dispatched, this bio will be dispatched without waiting at all,
> > this is due to tg_with_in_bps_limit() will return 0 if the flag is set.
> >
> > Instead of setting the flag when bio starts throttle, delay to when
> > throttle is done to fix the problem.
> >
> > Fixes: 9f5ede3c01f9 ("block: throttle split bio in case of iops limit")
> > Signed-off-by: Yu Kuai <yukuai3@xxxxxxxxxx>
> > ---
> >   block/blk-throttle.c | 10 ++++++----
> >   1 file changed, 6 insertions(+), 4 deletions(-)
> >
> > diff --git a/block/blk-throttle.c b/block/blk-throttle.c
> > index 447e1b8722f7..f952f2d942ff 100644
> > --- a/block/blk-throttle.c
> > +++ b/block/blk-throttle.c
> > @@ -811,7 +811,7 @@ static bool tg_with_in_bps_limit(struct throtl_grp *tg, struct bio *bio,
> >   	unsigned int bio_size = throtl_bio_data_size(bio);
> >   
> >   	/* no need to throttle if this bio's bytes have been accounted */
> > -	if (bps_limit == U64_MAX || bio_flagged(bio, BIO_THROTTLED)) {
> > +	if (bps_limit == U64_MAX) {
>
> This way may double account bio size for re-entered split bio.
Hi, Ming

Yes, you are right, I forgot that...
> >   		if (wait)
> >   			*wait = 0;
> >   		return true;
> > @@ -1226,8 +1226,10 @@ static void blk_throtl_dispatch_work_fn(struct work_struct *work)
> >   
> >   	spin_lock_irq(&q->queue_lock);
> >   	for (rw = READ; rw <= WRITE; rw++)
> > -		while ((bio = throtl_pop_queued(&td_sq->queued[rw], NULL)))
> > +		while ((bio = throtl_pop_queued(&td_sq->queued[rw], NULL))) {
> > +			bio_set_flag(bio, BIO_THROTTLED);
> >   			bio_list_add(&bio_list_on_stack, bio);
> > +		}
> >   	spin_unlock_irq(&q->queue_lock);
> >   
> >   	if (!bio_list_empty(&bio_list_on_stack)) {
> > @@ -2134,7 +2136,8 @@ bool __blk_throtl_bio(struct bio *bio)
> >   			}
> >   			break;
> >   		}
> > -
> > +		/* this bio will be issued directly */
> > +		bio_set_flag(bio, BIO_THROTTLED);
> >   		/* within limits, let's charge and dispatch directly */
> >   		throtl_charge_bio(tg, bio);
>
> Marking BIO_THROTTLED before throtle_charge_bio() causes the bio
> bytes not be charged.
Yes, thanks for spotting this.
> Another simple way is to compensate for previous extra bytes accounting,
> something like the following patch:
>
>
> diff --git a/block/blk-throttle.c b/block/blk-throttle.c
> index 139b2d7a99e2..44773d2ba257 100644
> --- a/block/blk-throttle.c
> +++ b/block/blk-throttle.c
> @@ -810,8 +810,7 @@ static bool tg_with_in_bps_limit(struct throtl_grp *tg, struct bio *bio,
>   	unsigned long jiffy_elapsed, jiffy_wait, jiffy_elapsed_rnd;
>   	unsigned int bio_size = throtl_bio_data_size(bio);
>   
> -	/* no need to throttle if this bio's bytes have been accounted */
> -	if (bps_limit == U64_MAX || bio_flagged(bio, BIO_THROTTLED)) {
> +	if (bps_limit == U64_MAX) {
>   		if (wait)
>   			*wait = 0;
>   		return true;
> @@ -921,10 +920,8 @@ static void throtl_charge_bio(struct throtl_grp *tg, struct bio *bio)
>   	unsigned int bio_size = throtl_bio_data_size(bio);
>   
>   	/* Charge the bio to the group */
> -	if (!bio_flagged(bio, BIO_THROTTLED)) {
> -		tg->bytes_disp[rw] += bio_size;
> -		tg->last_bytes_disp[rw] += bio_size;
> -	}
> +	tg->bytes_disp[rw] += bio_size;
> +	tg->last_bytes_disp[rw] += bio_size;
>   
>   	tg->io_disp[rw]++;
>   	tg->last_io_disp[rw]++;
> @@ -2125,6 +2122,20 @@ bool __blk_throtl_bio(struct bio *bio)
>   		if (sq->nr_queued[rw])
>   			break;
>   
> +		/*
> +		 * re-entered bio has accounted bytes already, so try to
> +		 * compensate previous over-accounting. However, if new
> +		 * slice is started, just forget it
> +		 */
> +		if (bio_flagged(bio, BIO_THROTTLED)) {
> +			unsigned int bio_size = throtl_bio_data_size(bio);
> +
> +			if (tg->bytes_disp[rw] >= bio_size)
> +				tg->bytes_disp[rw] -= bio_size;
> +			if (tg->last_bytes_disp[rw] - bio_size)
> +				tg->last_bytes_disp[rw] -= bio_size;
> +		}
> +
If new slice is not started, then this should fix the problem.
From what I can see, other than tg_conf_updated, new silce can start if
there are still bio throttled:

tg_may_dispatch
 if (!(tg->service_queue.nr_queued[rw]))
  throtl_start_new_slice

Thus I think the change is ok. For the case in tg_conf_updated, I'll
remove the throtl_start_new_slice() to fix a hung problem. I'll add this
patch with this one in next version.

Thanks,
Kuai

>   		/* if above limits, break to queue */
>   		if (!tg_may_dispatch(tg, bio, NULL)) {
>   			tg->last_low_overflow_time[rw] = jiffies;
>
> Thanks,
> Ming
>
> .