linux-kernel - Re: [PATCH v7 2/4] cpufreq: ondemand: call dbs_check

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:	Thu, 31 Jan 2013 09:39:09 +0100
From:	Fabio Baltieri <fabio.baltieri@...aro.org>
To:	Viresh Kumar <viresh.kumar@...aro.org>
Cc:	"Rafael J. Wysocki" <rjw@...k.pl>, cpufreq@...r.kernel.org,
	linux-pm@...r.kernel.org, Linus Walleij <linus.walleij@...aro.org>,
	swarren@...dotorg.org, linaro-dev@...ts.linaro.org,
	Nicolas Pitre <nicolas.pitre@...aro.org>,
	mathieu.poirier@...aro.org, Joseph Lo <josephl@...dia.com>,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH v7 2/4] cpufreq: ondemand: call dbs_check_cpu only when
 necessary

Hello Viresh,

On Wed, Jan 30, 2013 at 10:41:08PM +0530, Viresh Kumar wrote:
> On 30 January 2013 22:16, Fabio Baltieri <fabio.baltieri@...aro.org> wrote:
> > Isn't that how it works now?  The current cpu ktime is not checked
> > against its own, but against the "leader" cpu (dbs_info_local->cdbs.cpu),
> > that's why it's initialized only for the first.
> >
> > Maybe I should have used dbs_info_leader/dbs_info instead of
> > dbs_info_local/dbs_info.
> 
> This routine is called as wq handler. Which will recover dbs_info from work
> using container_of. Which would give dbs_info_local for the cpu j.
> 
> Then we will execute below code.
> 
> +       /* use leader CPU's dbs_info */
> +       dbs_info = &per_cpu(od_cpu_dbs_info, dbs_info_local->cdbs.cpu);
> 
> dbs_info_local->cdbs.cpu was uninitialized for all cpus except policy->cpu.

Ok, now I see the problem: cdbs->cpu is initialized only on the leader
cpu and this is working by coincidence on my system, while
cdbs->time_stamp is initialized only on the leader cpu, and that should
be correct even when cpu hotplugging as that's reinitialized every time.

That's a fix so I'll send a patch just to set ->cpu into the
for_each_cpu cycle.

Thanks,
Fabio

> And so, might be initialized with 0 as its a global variable... But if you
> offline cpu 0 and online it back, then policy->cpu would be 1 and this logic,
> which worked by mistake will fail.
> 
> +       mutex_lock(&dbs_info->cdbs.timer_mutex);
> +
> +       time_now = ktime_get();
> +       delta_us = ktime_us_delta(time_now, dbs_info->cdbs.time_stamp);
> 
> and so as this.
> 
> Correct?
> --
> To unsubscribe from this list: send the line "unsubscribe cpufreq" in
> the body of a message to majordomo@...r.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

-- 
Fabio Baltieri
--
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/