I just took a quick look on Antonello's laptop...
After resume in poll-mode, it does indeed seem to be staying at the top
speed. PowerTOP reflects ~ 85% of CPU time being spent in C1 vs. C0,
which seems to be true both before and after the resume. Prior to the
resume, poll-mode CPUPM works as expected.
In the default (event) mode on this system, I'm seeing the issue tracked
by:
6818514 Event based CPU power management can sometimes auto-tune too
conservatively
..but once that tuning is fixed the system power manages the CPUs as
expected (even after resume).
-Eric
Mark Haywood wrote:
> Antonello Cruz wrote:
>> Mark Haywood wrote:
>>> Antonello Cruz wrote:
>>>> Aubrey Li wrote:
>>>>> On Tue, Apr 28, 2009 at 3:05 AM, Henrik Johansson
>>>>> <henrikj at henkis.net> wrote:
>>>>>>
>>>>>> My power management works fine when I first boot my Lenovo T61,
>>>>>> but after
>>>>>> putting it into suspend mode it will always run on the highest
>>>>>> frequency.
>>>> I see the same behavior with a Dell XPS 1330M
>>>> I've notice that if I disconnect the power adapter and then
>>>> reconnect, it goes to a more same behavior.
>>> So that I can do some investigation ... I assume you are running
>>> CPUPM in poll-mode?
>> That's right, I've attached my power.conf file
>>
>>
>> Since I cannot type (apparently) here is what happens:
>> If I suspend/resume, after resume powertop report max P-state
>> (2001Mhz(turbo) 100%) and it seems the system is stuck at that cpu
>> frequency. This is confirmed with
>> kstat -m cpu_info -s current_clock_Hz
>>
>> If I unplug the power supply, the cpu frequency drops to 800MHz. When
>> I plug the power supply back in, the frequency goes up to 2001Mhz and
>> gradually goes down to 800MHz again. Frequencies in all steps are
>> confirmed with
>> kstat -m cpu_info -s current_clock_Hz
>
> Folks really ought to use powertop instead of the current_clock_Hz
> kstat now. Unfortunately, I think current_clock_Hz will be a confusing
> kstat now that PAD has integrated. You might still see that you are
> not leaving the 2001MHz P-state, but I would verify with
> /usr/bin/powertop.
>
> Thanks,
> Mark
>
>>
>>
>> Antonello
>> ------------------------------------------------------------------------
>>
>> _______________________________________________
>> pm-discuss mailing list
>> pm-discuss at opensolaris.org
>> http://mail.opensolaris.org/mailman/listinfo/pm-discuss
>
> _______________________________________________
> pm-discuss mailing list
> pm-discuss at opensolaris.org
> http://mail.opensolaris.org/mailman/listinfo/pm-discuss
