linux-kernel - Re: [PATCH] cpufreq: Skip all governor-related actions for cpufreq

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:	Fri, 8 Apr 2016 11:14:14 +0530
From:	Viresh Kumar <viresh.kumar@...aro.org>
To:	"Rafael J. Wysocki" <rjw@...ysocki.net>
Cc:	"Rafael J. Wysocki" <rafael@...nel.org>,
	Linux PM list <linux-pm@...r.kernel.org>,
	Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
	Srinivas Pandruvada <srinivas.pandruvada@...ux.intel.com>
Subject: Re: [PATCH] cpufreq: Skip all governor-related actions for
 cpufreq_suspended set

On 08-04-16, 00:05, Rafael J. Wysocki wrote:
> On Thursday, April 07, 2016 05:35:03 PM Viresh Kumar wrote:

> > That's *ugly* and it works by chance, unless I am misreading it
> > completely.
> 
> I'm assuming that what you mean by "ugly" here is "not really straightforward",
> which I agree with,

Yeah.

> but then it is really disappointing to see comments like
> that from you about the code that you helped to write.

I was just trying to say that this isn't how I feel it should be done.
:(

> Moreover, runtime CPU offline *also* doesn't have to run the governor exit/init
> for the same reason why the policy directory doesn't have to be removed on
> CPU offline: it is just pointless to do that.  The governor has been stopped
> already and it won't do anything more.  The only problem here is to prevent
> governor tunable sysfs attributes from triggering actions in that state,
> but that shouldn't be too difficult to arrange for.  If that's done,

Isn't that already guaranteed as userspace should have been frozen by
by the time we reach cpufreq_suspend()?

> cpufreq_suspended can be dropped, modulo changing cpufreq_start_governor()
> to return immediately if the governor has been started already.
> 
> And if something else is needed to protect driver callbacks from being invoked
> outside of the suspend-resume path, a more robust mechanism has to be added
> for that.
> 
> But in the meantime, I'd like to address the fast switch problem first and
> then you're free to clean up things on top of that.  Or I will clean them up
> if I have the time.

Okay..

-- 
viresh