The Linux Kernel Mailing List
 help / color / mirror / Atom feed
From: Tao Cui <cui.tao@linux.dev>
To: Tang Yizhou <yizhou.tang@shopee.com>,
	axboe@kernel.dk, guzebing1612@gmail.com
Cc: cui.tao@linux.dev, linux-block@vger.kernel.org,
	linux-kernel@vger.kernel.org, yukuai@fygo.io
Subject: Re: [PATCH v2] blk-wbt: Always change enable_state to MANUAL when setting latency
Date: Fri, 21 Aug 2026 10:49:45 +0800	[thread overview]
Message-ID: <f17cab7f-ee67-4d9c-95f0-01c28b4a520b@linux.dev> (raw)
In-Reply-To: <20260820161514.371181-1-yizhou.tang@shopee.com>



在 2026/8/21 00:15, Tang Yizhou 写道:
> From: Tang Yizhou <yizhou.tang@shopee.com>
> 
> Commit 1e56f30a73f3 ("block: Make WBT latency writes honor enable
> state") compares only the boolean enabled state in the no-op check.
> Writing the current latency value while WBT is in WBT_STATE_ON_DEFAULT
> is therefore still skipped, and enable_state is not changed to
> WBT_STATE_ON_MANUAL. A subsequent elevator switch to BFQ then disables
> WBT through wbt_disable_default(), silently discarding the explicit
> sysfs setting:
> 
>   # enable_state = WBT_STATE_ON_DEFAULT, min_lat_nsec = 2000000
>   echo 2000 > /sys/block/nullb0/queue/wbt_lat_usec  # skipped as a no-op
>   echo bfq > /sys/block/nullb0/queue/scheduler
>   cat /sys/block/nullb0/queue/wbt_lat_usec          # 0, WBT disabled
> 
> Skip the update only if the stored latency matches and enable_state
> already equals the exact target state: WBT_STATE_ON_MANUAL for a
> non-zero value, WBT_STATE_OFF_MANUAL for zero.
> 
> A write that passes the check then reaches wbt_set_min_lat(), which
> performs the actual transition to the MANUAL state.
> 
> Rename the helper to wbt_lat_changed() to improve readability.
> 
> Fixes: 1e56f30a73f3 ("block: Make WBT latency writes honor enable state")
> Reviewed-by: Yu Kuai <yukuai@fygo.io>
> Reviewed-by: Guzebing <guzebing1612@gmail.com>
> Signed-off-by: Tang Yizhou <yizhou.tang@shopee.com>
> ---
> v2: Pick Yuai and Guzebing's Reviewed-by tag.
>  block/blk-wbt.c | 14 +++++++++++---
>  1 file changed, 11 insertions(+), 3 deletions(-)
> 
> diff --git a/block/blk-wbt.c b/block/blk-wbt.c
> index 953d400fd013..4f12a064286e 100644
> --- a/block/blk-wbt.c
> +++ b/block/blk-wbt.c
> @@ -813,7 +813,14 @@ static void wbt_queue_depth_changed(struct rq_qos *rqos)
>  	wbt_update_limits(RQWB(rqos));
>  }
>  
> -static bool wbt_set_lat_changed(struct request_queue *q, u64 val)
> +/*
> + * Return true if writing @val would change the WBT state:
> + * 1) @val differs from the stored min_lat_nsec, or
> + * 2) @val matches, but enable_state is not the corresponding manual state
> + *    (WBT_STATE_ON_MANUAL for non-zero @val, WBT_STATE_OFF_MANUAL for 0),
> + *    so the write still has to update enable_state.
> + */
> +static bool wbt_lat_changed(struct request_queue *q, u64 val)
>  {
>  	struct rq_qos *rqos = wbt_rq_qos(q);
>  	struct rq_wb *rwb;
> @@ -825,7 +832,8 @@ static bool wbt_set_lat_changed(struct request_queue *q, u64 val)
>  	if (rwb->min_lat_nsec != val)
>  		return true;
>  
> -	return rwb_enabled(rwb) != !!val;
> +	return rwb->enable_state !=
> +	       (val ? WBT_STATE_ON_MANUAL : WBT_STATE_OFF_MANUAL);
>  }
>  
>  static void wbt_exit(struct rq_qos *rqos)
> @@ -1021,7 +1029,7 @@ int wbt_set_lat(struct gendisk *disk, s64 val)
>  		val *= 1000ULL;
>  
>  	mutex_lock(&disk->rqos_state_mutex);
> -	if (!wbt_set_lat_changed(q, val)) {
> +	if (!wbt_lat_changed(q, val)) {
>  		mutex_unlock(&disk->rqos_state_mutex);
>  		goto out;
>  	}

Reviewed-by: Tao Cui <cuitao@kylinos.cn>



      reply	other threads:[~2026-08-21  2:49 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-08-20 16:15 [PATCH v2] blk-wbt: Always change enable_state to MANUAL when setting latency Tang Yizhou
2026-08-21  2:49 ` Tao Cui [this message]

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=f17cab7f-ee67-4d9c-95f0-01c28b4a520b@linux.dev \
    --to=cui.tao@linux.dev \
    --cc=axboe@kernel.dk \
    --cc=guzebing1612@gmail.com \
    --cc=linux-block@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=yizhou.tang@shopee.com \
    --cc=yukuai@fygo.io \
    /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