linux-kernel - Re: [PATCH v3 3/4] sched: scheduler-driven cpu frequency selection

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

Date:	Mon, 29 Jun 2015 11:55:16 -0500
From:	Felipe Balbi <balbi@...com>
To:	Michael Turquette <mturquette@...libre.com>
CC:	<balbi@...com>, <peterz@...radead.org>, <mingo@...nel.org>,
	<linux-kernel@...r.kernel.org>, <preeti@...ux.vnet.ibm.com>,
	<Morten.Rasmussen@....com>, <riel@...hat.com>, <efault@....de>,
	<nicolas.pitre@...aro.org>, <daniel.lezcano@...aro.org>,
	<dietmar.eggemann@....com>, <vincent.guittot@...aro.org>,
	<amit.kucheria@...aro.org>, <juri.lelli@....com>,
	<rjw@...ysocki.net>, <viresh.kumar@...aro.org>,
	<ashwin.chaugule@...aro.org>, <alex.shi@...aro.org>,
	<linux-pm@...r.kernel.org>, <abelvesa@...il.com>,
	<pebolle@...cali.nl>
Subject: Re: [PATCH v3 3/4] sched: scheduler-driven cpu frequency selection

Hi,

On Mon, Jun 29, 2015 at 09:49:43AM -0700, Michael Turquette wrote:

<snip>

> > > +static int cpufreq_sched_stop(struct cpufreq_policy *policy)
> > > +{
> > > +     struct gov_data *gd = policy->governor_data;
> > > +
> > > +     if (cpufreq_driver_might_sleep()) {
> > 
> > unnecessary curly braces.
> > 
> > > +             kthread_stop(gd->task);
> > 
> 
> Thanks for the review. I'll take into account everything above.
> 
> > should you switch back to some default OPP when this is removed ? Some
> > SoCs can't run at certain OPPs forever (thermal limitations, or whatever
> > else), might be good to switch to something considered safe.
> 
> The above only happens when we unload the module or switch governors,
> and every governor has this characteristic.
> 
> I do not think that open-coding a return to some default opp in every
> governor is a good solution. This sounds like something the cpufreq core
> should take care of.

indeed.

> Also, how do we know which opp is safe?

no idea, that needs to be described somehow.

-- 
balbi

Download attachment "signature.asc" of type "application/pgp-signature" (820 bytes)