On Friday 14 August 2009 11:29:39 Oliver Maurhart wrote:
> Hi *,
>
> > > Oh. That's interesting. This may also explain the sudden drops in CPU
> > > frequency while working (power plugged in and online serving ...).
> > > Maybe my
> > > BIOS is broken ... I'm owning a Dell Latitude D810 and the BIOS.
> >
> > I remember Dells limiting via _PPC call. They needed special handling
> > because they also reduced the freq on their own instead of leaving
> > this to one instance, the OS.
> >
> > > Ok. I'm going to grab the current kernel-sources and compile along
> > > with your patch to check my hw_limit data.
>
> So, I got myself the linux-2.6.31-rc6 from kernel.org, compiled it and
> installed it. I also went on to get any BIOS updates from DELL and flashed
> anything on the laptop, which wasn't fast enough to run and hide.
>
> Things differ a bit, since I managed to get 1,87 GHz ... but just for a few
> seconds. I'm currently stuck at 1.33 GHz.
>
> Though my systems tells me I could run at 2.13 GHz, I'm not able to switch to
> this speed.
>
> Anyway ... I don't get a hw_limit file in the /sys/*/cpufreq/. I looked at
> make menuconfig in the kernel, but missed to see if I had to turn on any extra
> switches. Do I miss somthing, or is this the wrong kernel version?
The patch is not mainline yet, you have to search the cpufreq list
(I pointed to the subject in my last mail) and you have to apply it manually.
Sorry for not mentioning.
> # uname -a
> Linux demandred 2.6.31-rc6 #1 SMP PREEMPT Fri Aug 14 10:18:33 CEST 2009 i686
> Intel(R) Pentium(R) M processor 2.13GHz GenuineIntel GNU/Linux
> # find /sys/devices/system/cpu/ -iname "*hw*limit*"
>
> BTW: where are the max frequencies for the several governors stored? How do
> they compute?
It's not governor specific.
scaling_max_freq can be restricted through:
- the user: echo xy >scaling_max_freq
- through the generic thermal driver
- through the BIOS via _PPC
- theoretically through other kernel layers, but the thermal driver
and ACPI processor driver (notified when BIOS restricts freq) are
the only ones I know which possibly could restrict max_freq
Finding the real culprit if things are restricted through ACPI BIOS can
be cumbersome. First steps would be:
Disassemble acpi tables:
acpidump >acpidump
acpixtract -a acpidump
iasl -d [SD]SDT*.dat
grep _PPC *.dsl
This will show you where the _PPC functions in your ACPI tables are
and looking at them might give a pointer why and where the BIOS
restricts your frequency.
But better open a bugzilla.kernel.org bug if you are sure that BIOS limits
your freq and attach dmesg and acpidump.
The quick workaround would be processor.ignore_ppc=1
Thomas
--
To unsubscribe from this list: send the line "unsubscribe cpufreq" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
