Re: problems setting freq on dual core t2600, 2.6.17-1.2157_FC5smp

[Thomas Renninger <trenn@suse.de>]

On Sun, 2006-08-06 at 21:43 +0200, Tobias Bocanegra wrote:
> 
> it working! it was the processor_perflib and not the processor_thermal.
> i commented the call to cpufreq_verify_within_limits() and now i can
> set the 2.17ghz. i just wonder how to control the powerstates in
> perflib?

Sorry for reading that so late.
I got two reports about this problem (X60 and HP nx9420).
Both with a Core Duo CPU (T2400).

On the X60 it was hard to reproduce and the phenomenon vanished for the
X60 when changing a performance-management setting in BIOS from
"Powersave" to "Automatic".
https://bugzilla.novell.com/show_bug.cgi?id=177963


On the HP nx9420 (nc6320 and nx7400 should also be affected) I could
identify that _PPC always returned 1.

But the value calculated in _PPC comes from several function invocations
and I gave up as I hoped some HP BIOS guys would have a look at it.
https://bugzilla.novell.com/show_bug.cgi?id=179702

This is no cpufreq, but an ACPI/AML problem.
Maybe someone who owns such a machine could do an ACPI_DEBUG trace when
_PPC is entered. Adding these two lines (ACPI_DEBUG=y must be compiled
into kernel) around the acpi_evaluated_integer() call should limit acpi
debug output to a sane level. Better unload all unnecessary ACPI modules
(thermal/fan/..) before loading acpi-cpufreq/speedstep-centrino:

drivers/acpi/processor_perf.c:

acpi_dbg_level=0xFFFFFFFF;
       status = acpi_evaluate_integer(pr->handle, "_PPC", NULL, &ppc);
acpi_dbg_level=0x3;


It would be nice if we could go on with the novell bug, if people do not
like to open an account there, please open a separate
bugzilla.kernel.org bug and take me into CC.
Please also describe the affected model and attach acpidump output.

Thanks,

    Thomas