linux-kernel - Re: [tip:sched/core] [sched/fair] d329605287: stress-ng.resched.ops_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: Windows password security audit tool. GUI, reports in PDF.

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

Message-ID: <20240710115932.GZ27299@noisy.programming.kicks-ass.net>
Date: Wed, 10 Jul 2024 13:59:32 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: kernel test robot <oliver.sang@...el.com>
Cc: Tejun Heo <tj@...nel.org>, oe-lkp@...ts.linux.dev, lkp@...el.com,
	linux-kernel@...r.kernel.org, x86@...nel.org, ying.huang@...el.com,
	feng.tang@...el.com, fengwei.yin@...el.com,
	aubrey.li@...ux.intel.com, yu.c.chen@...el.com
Subject: Re: [tip:sched/core] [sched/fair]  d329605287:
 stress-ng.resched.ops_per_sec -24.0% regression

On Wed, Jul 10, 2024 at 12:51:44PM +0800, kernel test robot wrote:
> 
> 
> Hello,
> 
> kernel test robot noticed a -24.0% regression of stress-ng.resched.ops_per_sec on:
> 
> 
> commit: d329605287020c3d1c3b0dadc63d8208e7251382 ("sched/fair: set_load_weight() must also call reweight_task() for SCHED_IDLE tasks")
> https://git.kernel.org/cgit/linux/kernel/git/tip/tip.git sched/core
> 
> testcase: stress-ng
> test machine: 64 threads 2 sockets Intel(R) Xeon(R) Gold 6346 CPU @ 3.10GHz (Ice Lake) with 256G memory
> parameters:
> 
> 	nr_threads: 100%
> 	testtime: 60s
> 	test: resched
> 	cpufreq_governor: performance

Well.... if I read the test source correctly, this seems to test how
fast it can call sched_setscheduler(), rather than test how fast we can
schedule.

And that patch mentioned above makes setting SCHED_IDLE more expensve --
as expensive as SCHED_OTHER and SCHED_BATCH.

I'm thinking this test is rather stupid and doesn't actually measure
anything useful, I don't think I consider sched_setscheduler() a fast
path by any means.

Su yeah, *shrug*.