Re: Bug#374579: cpufreqd: crashing on startup (amd64, Athlon 64 X2)

[Mattia Dongili <malattia@linux.it>]

Hello,

I need some help with this bug, I'm quite convinced it's not a cpufreqd
bug but I'm asking before reassigning (glibc or kernel?).

Below the first and last messages, the BTS contains a little more
discussion.

Thanks in advance,
Mattia

On Tue, Jun 20, 2006 at 02:46:57PM +1000, Vincent Ho wrote:
> Package: cpufreqd
> Version: 2.1.0-1
> Severity: important
> 
> cpufreqd crashes soon after startup.  /var/log/messages contains the
> following:
> 
> Jun 20 14:31:49 grimlock kernel: cpufreqd[14315] general protection rip:2b9833b644b0 rsp:7fffffa74778 error:0
> 
> This happens both with my cpufreqd.conf and if I purge the package and
> reinstall it.
> 
> Kernel is from the linux-image-2.6.16-2-amd64-k8-smp package.
> /proc/acpi exists but not /proc/pmu or /proc/apm.
> 
> For now, I've reverted to cpufreqd 2.0.0-1 which has been running fine.
> 
> -- System Information:
> Debian Release: testing/unstable
>   APT prefers unstable
>   APT policy: (500, 'unstable'), (1, 'experimental')
> Architecture: amd64 (x86_64)
> Shell:  /bin/sh linked to /bin/bash
> Kernel: Linux 2.6.16-2-amd64-k8-smp
> Locale: LANG=en_AU.UTF-8, LC_CTYPE=en_AU.UTF-8 (charmap=UTF-8)
> 
> Versions of packages cpufreqd depends on:
> ii  debconf [debconf-2.0]         1.5.2      Debian configuration management sy
> ii  libc6                         2.3.6-15   GNU C Library: Shared libraries
> ii  libcpufreq0                   002-1      shared library to deal with the cp
> ii  libsensors3                   1:2.10.0-7 library to read temperature/voltag
> ii  lsb-base                      3.1-10     Linux Standard Base 3.1 init scrip
> 
> cpufreqd recommends no packages.
> 
> -- debconf information:
>   cpufreqd/no_pm:
>   cpufreqd/no_procfs_sysfs:
> 
> 

On Wed, Jun 21, 2006 at 07:58:59PM +0200, Mattia Dongili wrote:
> On Wed, Jun 21, 2006 at 04:45:09PM +1000, Vincent Ho wrote:
> > On Tue, Jun 20, 2006 at 06:46:07PM +0200, Mattia Dongili wrote:
> > 
> > > Linus says that maybe
> > >   echo 0 > /proc/sys/kernel/randomize_va_space
> > > can workraround the problem.
> > > 
> > > Can you try it?
> > 
> > This didn't seem to make any difference.  I should add that it doesn't
> > always trigger the kernel message, but in any case it never starts
> > successfully (so presumably segfaults without kernel message).
> > 
> > When tracing the program in gdb, I noticed that the cpufreqd_log()
> > function was opening syslog above (ie. log_opened == 0, so it calls
> > openlog() to do so).  I'm beginning to think it's suspicious that adding
> > -V4 to the arguments makes a difference.  In fact weirdly enough, adding
> > -V 4,5 or 2 helps, but -V3 doesn't.  I presume adding that alters what
> > gets output and thus exactly when syslog is opened.
> 
> it's printing with LOG_ERR there (== 3) so using a verbosity level of
> 3,4,5 doesn't make any difference on the code path...
> Anyway that code (apm plugin and logging code) is exactely the same as
> 2.0.0, this makes the thing even more wierd (on a cpufreqd perspective). 
> 
> Out of curiosity, does moving cpufreqd_apm.{so,la} away from
> /usr/lib/cpufreqd helps? or does it simply crash in a different place?
> 
> > Since it's crashing on the vfprintf() perhaps it's also possible this is
> > a bug in libc?
> 
> Maybe, let's see if people there can help.
> 
> -- 
> mattia
> :wq!
> 
> 

-- 
mattia
:wq!