Re: [Crash-utility] [PATCH] Use registers from LKCD header to getstarting ESP register
* Troy Heber <troy.heber(a)hp.com> [2007-05-02 17:45]:
On 05/02/07 11:06, Dave Anderson wrote:
> And adding a bunch of new LKCD files at this point in its life-time
> is not all that appealing if it can be avoided.
I agree.
I can also implement this in one file ...
> > It fixed a problem here where I head a crash dump from a
customer
> > using LKCD that showed a totally wrong backtrace in crash but lcrash
> > worked fine.
I'm curious to know more about the unwind bug and if there is another
approach that can be taken to resolve it. Unwinding on IA-64 has been
That was no IA64 machine. It was x86_64 and 2.6.5. (In fact, the patch
doesn't retrieve the registers on IA64 for the stack pointer.)
pretty solid in the past, so I'm curious to know what the
underlying
problem is. I haven't worked with any 2.6 kernels (> 2.6.10) kernels
and LKCD. Bernhard, since you've already debugged this, can you think
of a way to enhance get_lkcd_regs() to catch the eip as an alternative
solution? In other words, why did it miss finding the eip in your
dump?
Why do we need to guess the registers if we really *have* it in the
dump? What's wrong with the general idea to fetch the dump header
and to use that information?
Thanks,
Bernhard
--
SUSE LINUX Products GmbH Tel. +49 (911) 74053-0
Maxfeldstr. 5 GF: Markus Rex
90409 Nürnberg, Germany HRB 16746 (AG Nürnberg)
OpenPGP DDAF6454: F61F 34CC 09CA FB82 C9F6 BA4B 8865 3696 DDAF 6454
Attachments:
- attachment.sig (application/pgp-signature — 189 bytes)