[Discuss] linux cpu governors

[Matthew Gillen]

On 1/12/2016 10:20 AM, Dan Ritter wrote:
> On Tue, Jan 12, 2016 at 10:00:18AM -0500, Matthew Gillen wrote:
>> Recent kernels (not sure exactly when this started) have been driving
>> me crazy.
>> For reference: Machine: Dell Precision with Intel Core i5
>> OS: Fedora 23 (kernel 4.2.8-300.fc23.x86_64)
>>
>> When running on battery, everything is fine when using the default
>> "powersave" governor.  Lately though, if I don't use the 180W power
>> supply that came with the computer (e.g. because I don't want that
>> beast in my travel bag), the kernel seems to be restricting CPU
>> speeds to something outside of what the governor claims to be
>> managing.
>>
>> Here's what I mean:
>> $cat /proc/cpuinfo
>>   processor       : 0
>>   vendor_id       : GenuineIntel
>>   cpu family      : 6
>>   model           : 60
>>   model name      : Intel(R) Core(TM) i5-4340M CPU @ 2.90GHz
>>   cpu MHz         : 365.332
>> ...
>> $ cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver
>>  intel_pstate
>> $cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
>>  powersave
>> $ cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq
>> 3600000
>> $ cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_min_freq
>> 800000
>> $ cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq
>> 367257
>>
> 
> The problem is in the intel_pstate governor.
> 
> https://www.kernel.org/doc/Documentation/cpu-freq/intel-pstate.txt
> 
> ....
> 
> For contemporary Intel processors, the frequency is controlled by the
> processor itself and the P-states exposed to software are related to
> performance levels.  The idea that frequency can be set to a single
> frequency is fiction for Intel Core processors.
> .....

Looks like I may have just had the misfortune of needing my computer to
work at the wrong time: today F23 did a major rebase of their kernel to
4.3.3, and it appears that the problem is gone; i.e. when I plug in to
AC power, regardless of the wattage, the CPU behaves "normally" and
doesn't throttle itself to 367MHz.

FWIW, the output of
 $ cat /sys/devices/system/cpu/intel_pstate/*

doesn't change from the old kernel to the new kernel, or any combination
of kernel and battery vs. AC power.  So it appears to be functionality
that could not be controlled from userspace.  Glad it appears to have
magically disappeared.

Thanks,
Matt