Re: [Crash-utility] xencrash: crash analisys tool for Xen hypervisor
Itsuro ODA wrote:
Hmm, anyway, crash is used by analysts who understand what he do.
I think strict checking is not necessary.
You're right -- I agree -- let's just go with the "xen-syms" check.
> But -- perhaps the simplest way is the better way, and your check
> for the first 8 bytes in the "xen-syms" namelist name is sufficient
> for now.
Yes.
> However, we should also have a "--hyper" command line argument to
> force a hypervisor session in case everything else fails.
But let's add --hyper to "long_options" just in case...
Sorry. x86_64.c is broken now. I will fix.
No problem.
One thing re: x86_64 -- unlike x86, please create your own xen-specific
backtrace function to plug into the machdep->back_trace function pointer.
I'm pretty sure you won't be able to make the current version work
for xen!
> 2. I don't see why it's necessary to bother with the
> additional BT_XEN_HYPER_MODE flag? It seems
> its usage could be replaced with XEN_HYPER_MODE(),
> which you also use in lkcd_x86_trace.c.
>
I want to use XEN_HYPER_MODE() too, but
unfortunatly find_trace() uses local valiable "pc" !!
Ah -- that's right, I remember running into the same thing...
How about doing an #undef XEN_HYPER_MODE in lkcd_x86_trace.h,
and then creating a XEN_HYPER_MODE() function in lkcd_x86_trace.c?
> But on the whole, the patch looks pretty clean,
> and as long as the "binary-file-type-determination" can
> be resolved cleanly, it looking pretty good.
>
> Thanks,
> Dave
>
I will be updating http://people.redhat.com with an updated
version before long, but I will leave 4.0-3.8 around if you
still need it. It can be accessed by:
http://people.redhat.com/anderson/crash-4.0-3.8.tar.gz
Thanks,
Dave