linux-kernel - Re: [PATCH v2 1/3] cpufreq: add resolve

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: <20160531184855.GK9864@graphite.smuckle.net>
Date:	Tue, 31 May 2016 11:48:55 -0700
From:	Steve Muckle <steve.muckle@...aro.org>
To:	Viresh Kumar <viresh.kumar@...aro.org>
Cc:	Steve Muckle <steve.muckle@...aro.org>,
	Peter Zijlstra <peterz@...radead.org>,
	Ingo Molnar <mingo@...hat.com>,
	"Rafael J. Wysocki" <rafael@...nel.org>,
	linux-kernel@...r.kernel.org, linux-pm@...r.kernel.org,
	Vincent Guittot <vincent.guittot@...aro.org>,
	Morten Rasmussen <morten.rasmussen@....com>,
	Dietmar Eggemann <dietmar.eggemann@....com>,
	Juri Lelli <Juri.Lelli@....com>,
	Patrick Bellasi <patrick.bellasi@....com>,
	Michael Turquette <mturquette@...libre.com>
Subject: Re: [PATCH v2 1/3] cpufreq: add resolve_freq driver callback

On Tue, May 31, 2016 at 11:00:11AM +0530, Viresh Kumar wrote:
> On 30-05-16, 08:31, Steve Muckle wrote:
> > My goal here was to have the system operate in this case in a manner
> > that is obviously not optimized (running at fmax), so the platform owner
> > realizes that the cpufreq driver doesn't fully support the schedutil
> > governor.
> > 
> > I was originally going to just return an error code but that also means
> > having to check for it which would be nice to avoid if possible on this
> > fast path.
> 
> Okay, I get what you are saying.
> 
> But all we are doing here is to make things fast by not sending IPIs,
> etc. That should *not* lead to a behavior where the frequency stays at
> MAX all the time even if the driver doesn't provide this callback or
> the freq-table.
> 
> If we just return the target_freq in this case instead of UINT_MAX,
> the platform may eventually have some unnecessary IPIs, wakeups, etc,
> but its frequency will still be switched properly.
> 
> Wouldn't that be a better choice ?

I'm still concerned that a platform owner may use this and accept
suboptimal perf/power because they aren't aware their cpufreq driver is
not fully compliant. But I agree it'd be better to have it work as well
as it can. I will make the change.

Maybe a warning message can be added when schedutil initializes if
resolve_freq is not supported.

thanks,
Steve