Re: [PATCH v2] cpufreq: stats: Add 'load_table' debugfs file to show accumulated data of CPUs
From: Rafael J. Wysocki
Date: Fri Jun 21 2013 - 09:04:25 EST
On Friday, June 21, 2013 01:01:35 PM Chanwoo Choi wrote:
> On 06/21/2013 12:42 AM, Viresh Kumar wrote:
> > On 20 June 2013 16:48, Chanwoo Choi <cw00.choi@xxxxxxxxxxx> wrote:
> >> But,
> >> To show old frequency/new frequency on load_table debugfs file,
> >> governor function(dbs_check_cpu()) pass calculated CPUs load to specific governor(e.g., ondemand)
> >> as below function flow.
> >> dbs_check_cpu() (in cpufreq.c)
> >> -> od_check_cpu() (in cpufreq_ondemand.c)
> >> -> __cpufreq_driver_target() (in cpufreq.c)
> >> -> cpufreq_driver->target(policy)
> >> Also, The __cpufreq_driver_target() is external function which can be called on other file
> >> so I must consider exception case.
> >> If send CPUFREQ_LOADCHECK noti after changed cpu frequency,
> >> I think it is complicated and has quite a little difficulty.
> >> What is your opinion?
> > What you can do is:
> > - create another routine: cpufreq_governor_driver_target()
> > - replace all __cpufreq_driver_target() from ondemand/conservative governors
> > with this one
> > - In cpufreq_governor_driver_target() call __cpufreq_driver_target() and
> > take a note of new freq.
> > Maybe you don't need to check the actual freq that is set (even
> > that would be simple to implement), but what is requested.
> OK, I understand and will try to implement it.
> If possible, I want to know the your opinion about this.
Well, to be honest, I don't like the whole thing. The more changes you need
to make to the common code to support it, the worse.
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/