linux-kernel - Re: Re: [PATCH 2/2] sched/fair: Fix premature check of WAKEUP

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] [day] [month] [year] [list]

Message-ID: <b37f144b-5d99-4b86-a074-e50ffbca619c@bytedance.com>
Date: Wed, 26 Feb 2025 15:34:47 +0800
From: Abel Wu <wuyun.abel@...edance.com>
To: Vincent Guittot <vincent.guittot@...aro.org>,
 Tianchen Ding <dtcccc@...ux.alibaba.com>
Cc: Madadi Vineeth Reddy <vineethr@...ux.ibm.com>,
 Peter Zijlstra <peterz@...radead.org>, Ingo Molnar <mingo@...hat.com>,
 Juri Lelli <juri.lelli@...hat.com>,
 Dietmar Eggemann <dietmar.eggemann@....com>,
 Steven Rostedt <rostedt@...dmis.org>, Ben Segall <bsegall@...gle.com>,
 Mel Gorman <mgorman@...e.de>, Valentin Schneider <vschneid@...hat.com>,
 Josh Don <joshdon@...gle.com>,
 "open list:SCHEDULER" <linux-kernel@...r.kernel.org>
Subject: Re: Re: [PATCH 2/2] sched/fair: Fix premature check of
 WAKEUP_PREEMPTION

On 2/26/25 1:15 AM, Vincent Guittot wrote:
> On Tue, 25 Feb 2025 at 07:29, Abel Wu <wuyun.abel@...edance.com> wrote:
>>
>> On 2/24/25 9:47 PM, Vincent Guittot wrote:
>>>
>>> Or we should just remove it. I'm curious to know who used it during
>>> the last couple of years ? Having in mind that lazy preemption adds
>>
>> TBH I have never used this feature. But since Phil mentioned a case
>> in debugging DELAY_DEQUEUE, I think we'd better keep it, what do you
>> think?
> 
> Yes. And we need to figure out how to deal with the below as well

Hi Vincent, Tianchen,

I'm not sure this is the right way to do to let SCHED_IDLE be promoted
to the full NEED_RESCHED, as LAZY has relaxed responsiveness of normal
tasks to TICK_NSEC/2 in avg and workloads using fair policies should
adjust their expectations on it. And I would also recommend scheduling
policies playing with each other inside the scope of policy, while the
preemption model is another scope. Tying the two scopes together might
make things complicate, although I can imagine that certain workloads
or scenarios will benefit from it.

Best Regards,
	Abel

> 
>>
>>> another level as check_preempt_wakeup_fair()  uses it so sched-idle
>>> tasks might not always be immediately preempted anyway.
>>
>> Right, thanks for mention that.
>>
>>>
>>>
>>>>
>>>> Thanks,
>>>>           Abel
>>>>
>>