linux-kernel - Re: [PATCH -next] blk-throttle: delay the setting of 'BIO

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 18 May 2022 12:07:12 +0800
From:   Ming Lei <ming.lei@...hat.com>
To:     Yu Kuai <yukuai3@...wei.com>
Cc:     tj@...nel.org, axboe@...nel.dk, cgroups@...r.kernel.org,
        linux-block@...r.kernel.org, linux-kernel@...r.kernel.org,
        yi.zhang@...wei.com
Subject: Re: [PATCH -next] blk-throttle: delay the setting of 'BIO_THROTTLED'
 to when throttle is done

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@...alhost ~]# echo "8:0 1024" > /sys/fs/cgroup/blkio/blkio.throttle.write_bps_device
> [root@...alhost ~]# echo $$ > /sys/fs/cgroup/blkio/cgroup.procs
> [root@...alhost ~]# dd if=/dev/zero of=/dev/sda bs=10k count=1 oflag=direct &
> [1] 620
> [root@...alhost ~]# dd if=/dev/zero of=/dev/sda bs=10k count=1 oflag=direct &
> [2] 626
> [root@...alhost ~]# 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@...wei.com>
> ---
>  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.


>  		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.

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

Thanks,
Ming