linux-kernel - Re: [PATCH] x86,sched: Fix sched_smt_power

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]

Date:	Tue, 10 Jan 2012 20:41:08 +0100
From:	Ingo Molnar <mingo@...e.hu>
To:	Arjan van de Ven <arjan@...ux.intel.com>
Cc:	Suresh Siddha <suresh.b.siddha@...el.com>,
	Youquan Song <youquan.song@...el.com>,
	Peter Zijlstra <a.p.zijlstra@...llo.nl>,
	linux-kernel@...r.kernel.org, tglx@...utronix.de, hpa@...or.com,
	akpm@...ux-foundation.org, stable@...r.kernel.org,
	len.brown@...el.com, anhua.xu@...el.com, chaohong.guo@...el.com,
	Youquan Song <youquan.song@...ux.intel.com>
Subject: Re: [PATCH] x86,sched: Fix sched_smt_power_savings totally broken


* Arjan van de Ven <arjan@...ux.intel.com> wrote:

> On 1/10/2012 1:18 AM, Ingo Molnar wrote:
> > 
> > * Suresh Siddha <suresh.b.siddha@...el.com> wrote:
> > 
> >> On Tue, 2012-01-10 at 00:58 -0500, Youquan Song wrote:
> >>> Thanks Peter! Here is the patch. 
> >>
> >> Youquan, As far as I know both the
> >> sched_smt_power_savings/sched_mc_power_savings are broken for atleast an
> >> year.
> > 
> > We want a single knob, sched_power_savings - with the mc_ and 
> > smt_ ones still kept and aliased to sched_power_savings, for 
> > compatibility reasons.
> > 
> > As Peter said, the other reasonable option is to have no knob at 
> > all and restart this code from scratch.
> > 
> > The other thing we should do is to add sane defaults: to turn on 
> > sched_power_savings *AUTOMATICALLY* when a system is obviously 
> > battery driven and turn it off when the system is obviously AC 
> > driven. User-space can still implement policy and override the 
> > kernel's default, but there's absolutely no excuse to not offer 
> > this default ourselves.
> 
> a very good default would be to keep all tasks on one package 
> until half the cores in the package are busy, and then start 
> spreading out.
> 
> I suspect that'll be the 90% case coverage.

Maybe - but there's no reason to connect all the dots within the 
kernel and actually *discover* nd use the very, very likely 
performance preference of the hardware in question.

Like a good resource management system (==kernel) should do.

We can do that with a 99% confidence factor or so - maybe better 
- and leave all the weird cases that the kernel cannot (or 
should not) know about to 'user space policy' knobs.

Thanks,

	Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/