[Crash-utility] crash does not get proper backtrace?

KAMEZAWA Hiroyuki kamezawa.hiroyu at jp.fujitsu.com
Wed Sep 8 23:44:34 UTC 2010


On Tue, 7 Sep 2010 10:02:53 -0400 (EDT)
Dave Anderson <anderson at redhat.com> wrote:

> > If so, what we (fujitsu) has to do is...
> >   1. confirm "virsh dump" saves registers or not.
> >   2. If saved, investigate the format.
> >   3. add crash support. (or write a program of format converter.)
> > 
> > I'm sorry if I don't understand correctly.
>  
> "virsh dump" does save registers, but getting them reliably from the dumpfile,
> and then if you do get them, the values do not necessarily work for
> the crash utlility's purposes.
> 
> You can put some debug code in the cpu_load() function in the crash
> utility's qemu-load.c file.  But unfortunately, that code area functionality
> keeps changing, and I simply take the latest code from the QEMU developers
> and plug it in.
> 
Okay, I'll see it...

> Anyway, when I have tried doing so in the past, I've found register contents
> that were either invalid, or containing eip/esp values that could not be used
> for starting points for backtracing the particular task.
> 

Thank you for informaton. It seems I need some deep work to do so....

Regards,
-Kame




More information about the Crash-utility mailing list