linux-kernel - Re: [PATCH RESEND] sched/fair: Only update stats for allowed CPUs when looking for dst group

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]

Message-ID: <20251014113731.GO4067720@noisy.programming.kicks-ass.net>
Date: Tue, 14 Oct 2025 13:37:31 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Adam Li <adamli@...amperecomputing.com>
Cc: mingo@...hat.com, juri.lelli@...hat.com, vincent.guittot@...aro.org,
	dietmar.eggemann@....com, rostedt@...dmis.org, bsegall@...gle.com,
	mgorman@...e.de, vschneid@...hat.com, cl@...ux.com,
	linux-kernel@...r.kernel.org, patches@...erecomputing.com,
	shkaushik@...erecomputing.com
Subject: Re: [PATCH RESEND] sched/fair: Only update stats for allowed CPUs
 when looking for dst group

On Sat, Oct 11, 2025 at 06:43:22AM +0000, Adam Li wrote:
> Load imbalance is observed when the workload frequently forks new threads.
> Due to CPU affinity, the workload can run on CPU 0-7 in the first
> group, and only on CPU 8-11 in the second group. CPU 12-15 are always idle.
> 
> { 0 1 2 3 4 5 6 7 } {8 9 10 11 12 13 14 15}
>   * * * * * * * *    * * *  *
> 
> When looking for dst group for newly forked threads, in many times
> update_sg_wakeup_stats() reports the second group has more idle CPUs
> than the first group. The scheduler thinks the second group is less
> busy. Then it selects least busy CPUs among CPU 8-11. Therefore CPU 8-11
> can be crowded with newly forked threads, at the same time CPU 0-7
> can be idle.
> 
> A task may not use all the CPUs in a schedule group due to CPU affinity.
> Only update schedule group statistics for allowed CPUs.
> 
> Signed-off-by: Adam Li <adamli@...amperecomputing.com>
> ---
> Resending this patch from the patchset:
> https://lore.kernel.org/lkml/20250717062036.432243-2-adamli@os.amperecomputing.com/
> 

Right, lets start with this then ;-)

No need to do the cpumask_and() thing, that's just more changes vs
update_sg_lb_stats().