[915] in SIPB_Linux_Development
Yay! A repeatable problem with Linux AFS... :-/
daemon@ATHENA.MIT.EDU (Erik Nygren)
Thu Jan 26 02:35:05 1995
To: linux-afs-bugs@MIT.EDU
Cc: linux-dev@MIT.EDU
Date: Thu, 26 Jan 1995 02:34:42 -0500
From: Erik Nygren <nygren@MIT.EDU>
Well, I've finally managed to find a repeatable problem
with Linux AFS 1.1.84-1 and Linux 1.1.85.
I can even get Limekiller (a Linux machine in the SIPB office running
1.1.84) to repeat it. To repeat, switch to console mode on a machine
running this combo. (You'll want to be there to see the cool dump.)
Anyways, type:
attach hosts
grep blah /mit/hosts/hosts
The machine will start spewing massive kernel oopses. They'll appear
as columns of pointers into the kernel code. My guess is that
something is oopsing recursively and the columns of numbers is the
call trace.
I've grepped this file with older versions of Linux many times before.
I'm not sure where it stopped working.
Unfortunately, I have had no luck logging anything as things go by too
fast to watch and the kernel probably isn't switching into user mode.
For the same reason I've had no luck running tcpdump (don't
have another machine on this subnet I have easy access to).
--- Erik
For some comic relief, here's an excerpt from the zephyr log on class
sipb when I rlogged into lime to try this there.
(BTW, credit goes to jhyun for discovering the problem on foundation
earlier today.... :-)
From: <nygren>
Could someone switch limekiller into console mode (ctl-alt-f1)
and then watch for a bit...
From: <shabby>
ok. watching.
From: <jhyun>
oh no..you're not...
From: <jhyun>
oh geez..i think you are gonna...
From: <shabby>
that's a lot of numbers.
From: <nygren>
Is it spewing LOTS and LOTS of numbers? :)
From: <shabby>
that'd be a "YES"
