Re: Re: Panic on boot on Sun Blade 6270

[Dan Gora <dan.gora@gmail.com>]

On Mon, Mar 15, 2010 at 6:21 AM, Jan Beulich <JBeulich@novell.com> wrote:
>>After about 5 minutes of sitting there it crashed:
>>
>>BUG: unable to handle kernel NULL pointer dereference at 0000000000000008
>>IP: [<ffffffff8020e9cb>] profile_pc+0x30/0x3d
>
> I would have been tempted to analyze this (given that it may point
> out a general problem with the SLE11 kernel) if this wasn't a kernel
> that you built on your own and if this wasn't a really old one. I
> would guess that it's not only a couple of added printk()-s, but also
> a different .config that you're using. Namely (not having seen the
> disassembly of your kernel's profile_pc()) I could easily imagine
> this oops to be a result of you building with CONFIG_FRAME_POINTER=y.

:)... Yes I did build with CONFIG_FRAME_POINTER=y.  I guess I
considered the panic more incidental... Really the problem was that
the machine locked up on boot.  I didn't realize that
CONFIG_FRAME_POINTER could cause a panic.. I thought that it just
turned on using the frame pointer register so that backtraces would
work.

I tried to install the xen 3.4.1 from the SuSE rpms after this and
also could not get that to work.  That too would panic on boot up.  I
tried to install the xen 4.0.0 rpms, but quickly ran into prerequisite
problems and just gave up and went back to using xen 3.3.1, building
it from the SRPMS (I still cannot get anything to work just building
it from the xen hg repo).

I realize that this Sun box is probably pretty funky (it's a PCI
express module based machine) so it might have some peculiar issues
that would be of interest, but I've got to figure this other issue out
first because I have customers on my back.  If you are interested,
Jan, I can go back and get the backtrace from the 3.4.1 panic, but if
SLES 11 is too old to try and debug I'll just forget about it..

If there's time later and there is more recent stuff that we can try
and test, let me know what it is and I can give it a go.

thanks,
dan