Subject:
Re: kern/122126: [ed] [panic] panic when using WD-8013 with ed(4)
From:
Peter Much <pmc@citylink.dinoex.sub.org>
Date:
Tue, 1 Apr 2008 01:00:32 +0200
To:
Volker Werth <vwe@FreeBSD.org>

Dear Volker!

On Sun, Mar 30, 2008 at 10:54:19PM +0200, Volker Werth wrote:
! The one you've now already filed in the PR ticket should help hunt the
! problem a bit. Unfortunately this seems to be an ithread problem and I'm
! not familiar with threading.

I never heard about this one either.

! Don't worry, I'm doing exactly the same!  ;)  While not doing anything
! stupid, this should be fine (from the security viewpoint). And I was
! already able to cut my electricity bill by using green (EE) hardware.

Yes, that would be the approach if I would use new hardware -
while currently I don't see a reason for it: a Pentium-2 does the 
routing and fileserving and mldonkey just fine, even when busy,
and I think it's power footprint is still below the new chips.
So currently I focus on arranging diskspace: I'll put the main 
filesystems on a 2 1/2" laptop disk and everything else on some 
SCSI drives handled by gstopd. That should give more diskspace,
less power, and - the confidence-inspiring sound of server SCSI 
disks...  :) 

! I would like to keep you busy with one more stupid question (I could do
! it myself if I would have a 6.3 system handy but I don't).
! 
! Can you please do a
! 
! `/usr/bin/addr2line -e /boot/kernel/kernel.symbols 0xc052f98c'
! 
! and drop me the output just to make sure, it points into the same direction?

Problem - the output is surely not what You had in mind:
> > /usr/bin/addr2line: '/boot/kernel/kernel.symbols': No such file

I went searching, but it looks like there was never ever any file
named "kernel.symbols" on my systems. Could it be this is a feature
that came with Release 7.0 ?

So, what I think You have in mind is: 
   add 'makeoptions DEBUG=-g' to the Kernelconfig,
   recompile the Kernel,
   and use addr2line on the kernel.debug file
This gives:

> > # addr2line -e ./kernel.debug 0xc052f98c
> > ../../../kern/kern_intr.c:338

But, now I was not sure if the linker arrangement with the new
makeoptions is still the same. 
So I recreated the panic with the newly compiled kernel. 
And then - really bad surprize! - this kernel now shows the same
behaviour as the GENERIC kernel! It shows multiple "panics" scrolling
off the screen and then freezes. 
So this comes from the 'makeoptions DEBUG=-g'.

And, as a consequence, I would not think this linenumber is reliable.
The funnier aspect is: adding debug info makes it impossible to
dump.:-/ I think I'm missing something here...

! As soon as you'll post me the output of addr2line, I'll check who will
! handle this issue. I would not wonder if it's a completely different
! issue and not just plain threading.

This is getting a bit strange indeed.

rgds,
PMc