Re: Strange reboots on 2.6.29-rc6-wl

[Sitsofe Wheeler]

On Fri, Mar 20, 2009 at 11:21:07AM +0100, Nico -telmich- Schottelius wrote:
> Sitsofe Wheeler [Tue, Mar 17, 2009 at 02:57:53PM +0000]:
> > On Tue, Mar 17, 2009 at 08:54:24AM +0100, Nico -telmich- Schottelius wrote:
> > > 
> > > Has anyone a clue on howto debug the build environment or
> > > debugging the kernel before it's getting booted?
> > 
> > If you have a serial console you can probably get output on that.
> 
> There's no native serial port on the machine :-/
> 
> > If you have flipped EMBEDDED to Y then you might also have to enable
> > EARLY_PRINTK. Another trick might be to turn set
> > BOOT_PRINTK_DELAY (Kernel hacking -> Delay each boot printk message by N
> > milliseconds) to something like 2000 and see if you can read/catch the
> > messages before the kernel reboots...
> 
> I've EARLY_PRINTK set, but other than the edd probing message
> I don't see anything.

What about when you set BOOT_PRINTK_DELAY to a value high enough to
cause pauses - do you see more messages? Additionally are booting your
kernel without the quiet option? Have you checked to make sure the
kernel isn't set to reboot on oops/panic etc? Are you using any binary
drivers?

> > Failing that start a git bisection (
> > http://www.kernel.org/doc/local/git-quick.html#bisect or
> > http://www.stardust.webpages.pl/files/handbook/handbook-en-0.3-rc1.pdf )
> > from the last good kernel...
> 
> I really don't think that it's the kernel's fault, because _all_
> new versions I compile instantly reboot.

So even a freshly compiled 2.6.27 and 2.6.28 reboot instantly?! If even
old kernels that worked suddently don't work that hints strongly at
hardware errors or a duff compiler toolchain...

You can also make the kernel spit out decompression error messages by
enabling CONFIG_X86_VERBOSE_BOOTUP . I guess it would also be wise to
make sure CONFIG_DOUBLEFAULT is also set. Are you sure nothing else has
changed though? Have you enabled some kernel option that was not set
before?  Have you changed any BIOS settings recently?

> Last one I tried was v2.6.29-rc8-241-g65c2449 (linux-2.6).
> Isn't there some flag to enable step by step execution (like in gdb)?

I don't think I undrstand what you mean - the kernel isn't a debugger...
If you are running a User Mode Linux kernel or running inside QEMU etc.
then you can do what you describe because your "machine" is a program
but I think the only way to do that with real hardware is to learn how
to use KGDB ( http://kgdb.sourceforge.net ) and have another machine to
hand. I suppose you could also sprinkle beep patches/debug printks about
the place to see just how far it gets before it reboots...

-- 
Sitsofe | http://sucs.org/~sits/
--
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/