linux-kernel - Re: [PATCH v2] block/mq-deadline: adjust the timeout period of the per

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 for Android: free password hash cracker in your pocket

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

Message-ID: <db87a85d-e433-4daf-97c7-d5156849db0f@acm.org>
Date: Thu, 9 Oct 2025 09:50:39 -0700
From: Bart Van Assche <bvanassche@....org>
To: chengkaitao <pilgrimtao@...il.com>, axboe@...nel.dk, dlemoal@...nel.org
Cc: linux-block@...r.kernel.org, linux-kernel@...r.kernel.org,
 Chengkaitao <chengkaitao@...inos.cn>
Subject: Re: [PATCH v2] block/mq-deadline: adjust the timeout period of the
 per_prio->dispatch

On 10/9/25 8:52 AM, chengkaitao wrote:
> On the other hand, the Commit (725f22a1477c) merges the effects of
> fifo_expire and prio_aging_expire on the same code behavior, creating
> redundant interactions. To address this, our new patch introduces
> numerical compensation for {dd->fifo_expire[data_dir]} when adding
> requests to dispatch lists. To maintain original logic as much as
> possible while enhancing dispatch list priority, we additionally
> subtract {dd->prio_aging_expire / 2} from the fifo_time, with default
> values, {dd->prio_aging_expire / 2} equals {dd->fifo_expire[DD_WRITE]}.

No assumptions should be made about the relative values of
dd->prio_aging_expire and dd->fifo_expire[DD_WRITE] since these values
can be modified via sysfs.

> diff --git a/block/mq-deadline.c b/block/mq-deadline.c
> index 3e741d33142d..fedc66187150 100644
> --- a/block/mq-deadline.c
> +++ b/block/mq-deadline.c
> @@ -659,7 +659,8 @@ static void dd_insert_request(struct blk_mq_hw_ctx *hctx, struct request *rq,
>   
>   	if (flags & BLK_MQ_INSERT_AT_HEAD) {
>   		list_add(&rq->queuelist, &per_prio->dispatch);
> -		rq->fifo_time = jiffies;
> +		rq->fifo_time = jiffies + dd->fifo_expire[data_dir]
> +				- dd->prio_aging_expire / 2;
>   	} else {
>   		deadline_add_rq_rb(per_prio, rq);

Thanks for having added a detailed patch description. Please remove
"/ 2" from the above patch to make sure that BLK_MQ_INSERT_AT_HEAD
requests are submitted to the block driver before other requests. This
is important if a request is requeued. Additionally, a comment should be
added above the modified line of code that explains the purpose of the
calculation. How about this untested patch?

diff --git a/block/mq-deadline.c b/block/mq-deadline.c
index 3e741d33142d..566646591ddd 100644
--- a/block/mq-deadline.c
+++ b/block/mq-deadline.c
@@ -659,7 +659,14 @@ static void dd_insert_request(struct blk_mq_hw_ctx 
*hctx, struct request *rq,

  	if (flags & BLK_MQ_INSERT_AT_HEAD) {
  		list_add(&rq->queuelist, &per_prio->dispatch);
-		rq->fifo_time = jiffies;
+		/*
+		 * started_after() subtracts dd->fifo_expire[data_dir] from
+		 * rq->fifo_time. Hence, add it here. Subtract
+		 * dd->prio_aging_expire to ensure that AT HEAD requests are
+		 * submitted before higher priority requests.
+		 */
+		rq->fifo_time = jiffies + dd->fifo_expire[data_dir] -
+				dd->prio_aging_expire;
  	} else {
  		deadline_add_rq_rb(per_prio, rq);

Thanks,

Bart.