Re: BUG: kernel crash report, odd...

[Randy Dunlap]

On Tue, 10 Jul 2007 00:29:54 +0200 Jesper Juhl wrote:

> Forgot a few things, see below...
> 
> On 10/07/07, Jesper Juhl <jesper.juhl@xxxxxxxxx> wrote:
> > On 10/07/07, Charles Shannon Hendrix <shannon@xxxxxxxxxxxxx> wrote:
> > >
> > > A system I was using a few minutes ago dumped this to the syslog:
> > >
> > > Jul  9 17:50:38 daydream kernel: [76022.613000] BUG: unable to handle
> > > kernel NUL
> > > L pointer dereference at virtual address 00000010
> > > Jul  9 17:50:38 daydream kernel: [76022.613000]  printing eip:
> > > Jul  9 17:50:38 daydream kernel: [76022.613000] c01ace66
> > > Jul  9 17:50:38 daydream kernel: [76022.613000] *pde = 00000000
> > > Jul  9 17:50:38 daydream kernel: [76022.613000] Oops: 0000 [#1]
> > >
> > >
> > > There was no other information either before or after.
> > >
> > > What I saw was my KDE desktop stopped responding to events.
> > >
> > > Machine didn't respond to power switch.
> > >
> > > Unfortunately, that's all I know.
> > >
> > > In the past when I had bugs like this, I got a fairly extensive kernel
> > > error message, but this time the above is the only thing I saw.
> > >
> > It does look a little short... probably the rest just didn't make it to disk...
> >
> > > I have not had kernel bugs like this until I started running 2.6.20 on a
> > > new Kubuntu install.
> > >
> > > Same hardware, but Slackware and my own custom 2.6.19 kernel never
> > > crashed.
> > >
> > Different kernel source version, different configuration, probably
> > different compiler, Slackware doesn't patch the vanilla kernel - I
> > believe (k)ubuntu does.   All in all that equals a significantly
> > different kernel binary you have been running in those two cases.
> >
> A different userspace could also be the cause. In theory, userspace
> should never be able to crash the kernel, but in real life bugs happen
> and sometimes userspace manages to cause a kernel crash. So, a
> different userspace environment in your new distribution could in
> theory expose a bug that didn't surface with your previous
> distribution.
> 
> > > Is there anything I can do, in case this happens again, to try and
> > > capture more information?
> > >
> > 0. Make sure the kernels loglevel is set so that you get all messages.

i.e., boot with "ignore_loglevel"

> 0.5. Perhaps increase the size of the kernels log buffer (the
> LOG_BUF_SHIFT config option)

or boot with "log_buf_len=n[KMG]" (power of 2 required)

> > 1. Make sure syslog is setup to capture all kernel messages, both
> > errors, warnings, notices, debug messages etc.
> >
> > 2. If you can, check output of 'dmesg' instead of just what made it to syslog.
> >
> > 3. If you have a second PC, setup serial console or netconsole to
> > capture kernel messages on that second box. This can sometimes capture
> > messages that don't make it to disk.
> >
> > 4. Build a custom kernel with some (or all) of the debug options under
> > the Kernel Hacking menu enabled.
> >
> > 5. Even when the system hangs you may in some cases be able to get
> > some info (or just sync the disks and do a reboot) via magic sysrq.

---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/