linux-kernel - Re: [PATCH] cpufreq: governors: Remove duplicate check of target freq in supported range

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: <521CE317.20908@semaphore.gr>
Date:	Tue, 27 Aug 2013 20:34:15 +0300
From:	Stratos Karafotis <stratosk@...aphore.gr>
To:	Viresh Kumar <viresh.kumar@...aro.org>
CC:	"Rafael J. Wysocki" <rjw@...k.pl>,
	"cpufreq@...r.kernel.org" <cpufreq@...r.kernel.org>,
	"linux-pm@...r.kernel.org" <linux-pm@...r.kernel.org>,
	LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] cpufreq: governors: Remove duplicate check of target
 freq in supported range

On 08/27/2013 07:07 PM, Viresh Kumar wrote:
> On 27 August 2013 21:16, Stratos Karafotis <stratosk@...aphore.gr> wrote:
>> I think we should keep these checks because:
>>
>> 1) They shorten the execution code (there is no unnecessary call of
>> __cpufreq_driver_target)
>
> I don't really count this one.. This is how code is present everywhere in
> kernel.. These checks are present in routines and callers don't need to
> take care of them..

I mean that if we will get rid of the code you mentioned, we will have
an extra call to function __cpufreq_driver_target in some cases.

>> 2) In case my patch will be accepted, we need them to avoid continuously
>> increase of dbs_info->requested_freq.With my patch the requested_freq
>> can temporarily overcome policy->min and policy->max. __cpufreq_driver_target
>> will select the correct frequency (within policy->min and policy->max).
>> Then, dbs_cpufreq_notifier will adjust requested_freq.
>
> Sorry, I couldn't understand what you meant here :(
>

I'm sorry. Let me try to explain this better.

With my patch, dbs_info->requested_freq will not be capped within
policy->min and policy->max in cs_check_cpu.
So, temporarily it may have a value greater than policy->max
or lower that policy->min.
When we call __cpufreq_driver_target, the correct frequency will be 
selected because __cpufreq_driver_target takes care to adjust the
target frequency within policy range.
But, eventually, dbs_cpufreq_notifier will adjust dbs_info->requested
within policy range, if needed.

If we remove
	if (dbs_info->requested_freq == policy->max)
		return;
and
	if (policy->cur == policy->min)
		return;

request_freq will keep increasing or decreasing in each iteration and
finally will overflow or underflow.

Consider, for example, that in a CPU with policy->max = 1000MHz
the current frequency is 950MHz. With a constant load above
up_threshold, the requested_freq in first iteration will be 1000MHz
and __cpufreq_driver_target will select 1000MHz freq.

In second iteration, requested_freq will be 1050MHz, and 
__cpufreq_driver_target will select 1000MHz. dbs_cpufreq_notifier
will adjust requested_freq back to 1000MHz.

In next iterations, dbs_cpufreq_notifier will not be called, so we
need the above check (dbs_info->requested_freq == policy->max) to
prevent requested_freq to grow arbitrary.

I hope my explanation was better now. :)

Thanks,
Stratos
--
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/