Re: Linux 2.6.17-rc1

[Andrew Morton <akpm@osdl.org>]

Ralf Hildebrandt <Ralf.Hildebrandt@charite.de> wrote:
>

I might not have read your email for two days.  And you removed from Cc the
people who actually have a chance of fixing this.  Please always do
reply-to-all.

> * Andrew Morton <akpm@osdl.org>:
> 
> > In acpi_processor_unregister_performance(), pr->performance is NULL.
> > 
> > Can you add the below?  It should tell us who forgot to register the
> > performance data, as well as working around the crash.
> 
> I added the patch.
> in dmesg I now get:
> 
> > Apr  4 09:54:06 knarzkiste kernel: powernow-k8: Found 1 AMD Athlon 64 / Opteron processors (version 1.60.1)
> > Apr  4 09:54:06 knarzkiste kernel: powernow-k8:    0 : fid 0x0 (800 MHz), vid 0x12 (1100 mV)
> > Apr  4 09:54:06 knarzkiste kernel: powernow-k8:    1 : fid 0x8 (1600 MHz), vid 0x4 (1450 mV)
> > Apr  4 09:54:06 knarzkiste kernel: cpu_init done, current fid 0x8, vid 0x2
> > Apr  4 09:54:06 knarzkiste kernel: powernow-k8: ph2 null fid transition 0x8
> 
> > Apr  4 09:54:25 knarzkiste kernel: powernow-k8: Found 1 AMD Athlon 64 / Opteron processors (version 1.60.1)
> > Apr  4 09:54:25 knarzkiste kernel: powernow-k8:    0 : fid 0x0 (800 MHz), vid 0x12 (1100 mV)
> > Apr  4 09:54:25 knarzkiste kernel: powernow-k8:    1 : fid 0x8 (1600 MHz), vid 0x4 (1450 mV)
> > Apr  4 09:54:25 knarzkiste kernel: cpu_init done, current fid 0x8, vid 0x4
> 
> > Apr  4 09:54:44 knarzkiste kernel: powernow-k8: Found 1 AMD Athlon 64 / Opteron processors (version 1.60.1)
> > Apr  4 09:54:45 knarzkiste kernel: powernow-k8:    0 : fid 0x0 (800 MHz), vid 0x12 (1100 mV)
> > Apr  4 09:54:45 knarzkiste kernel: powernow-k8:    1 : fid 0x8 (1600 MHz), vid 0x4 (1450 mV)
> > Apr  4 09:54:45 knarzkiste kernel: cpu_init done, current fid 0x8, vid 0x4
> 
> So, no more EIPs, but no conclusive messages either!

OK, thanks.  That indicates that we did install a
acpi_processor_performance structure, but something must have later on
zeroed it.

Hopefully the cpufreq guys will be able to reproduce this.

<tries it>

Actually, I cannot even rmmod the thing:

Module                  Size  Used by
p4_clockmod             6980  1 
speedstep_lib           5376  1 p4_clockmod

It looks like either it has a refcounting problem or it has been changed so
that it is deliberately pinned.