Re: novice question -- live kernel debugging?

Richard B. Johnson (root@chaos.analogic.com)
Fri, 5 Dec 1997 17:55:32 -0500 (EST)


On Fri, 5 Dec 1997, Mikhail Sukhar wrote:

> Hi everybody!
>
> Could somebody direct me to the texts explaining a bit more
> about kernel debugging. I can always use printfs and I understand
> how to debug a crash dump, i.e. to find from symbol map and EIP
> the place using gdb the offending lines of code. However I can't
> find any FAQa/HOWTOs on how to debug live kernel, or at least to
> force it to crash. Could something like /dev/kmem be used?

A user is not supposed to be able to crash the kernel. However, there
are some hooks put in so a kernel-programmer can do a few things, some
of which could crash the kernel. Loading an invalid module is one of
the things that can crash the kernel.

You can add a new kernel function-code that will do whatever you want
after the trap to kernel-mode occurs.

Of course, if you just want to randomly mess things up....

become root:
execute cp /dev/random /dev/port

I'd suggest you unmount your hard-disk(s) first...

kill -TERM -1
sync
kill -KILL -1
umount -a

... Then play ....

Cheers,
Dick Johnson

Richard B. Johnson
Project Engineer
Analogic Corporation
Penguin : Linux version 2.1.70 on an i586 machine (66.15 BogoMips).
Warning : It's hard to remain at the trailing edge of technology.