[Crash-utility] Re: Problem with using crash 4.0-2.21 on ppc

Dave Anderson anderson at redhat.com
Fri Feb 24 14:38:11 UTC 2006


Haren Myneni wrote:

>
> > Is ff807a50 typically a legitimate user-space stack address
> > in ppc64 user VM?  You could probably run the address
> > through IN_TASK_VMA(), and if it is a valid user-space
> > stack address, just indicate that the process was running
> > in user-space.
>
> ff807a50 is in user space. Yes, the kernel address on PPC64 starts at
> c000000000000000. Not only we should print an message says that "running
> is user space", but also to display traces for other active traces. It
> is a bug too. I will send the fix ASAP.
>

Good point -- just returning after the user-space message gets printed
will prevent the premature stoppage of the whole command.

>
> Sure, if you have some other fixes or on other archs. Otherwise, can we
> wait for early next week. I am wondering what is causing for Rachita's
> issue. Is it related to the same paca.dataoffset patch? just want to
> make sure.
>

That's fine -- I'll just wait until next week.

>
> BTW, I ran the crash tool on RHEL4 vmcore (not the recent RHEL4 update
> version) to see whether I am breaking backward compatibility. Small fix.
> I somehow overlooked. Sorry. Probably, that might be the reason I saved
> one RHEL4 vmcore and the corresponding vmlinux.debug.
>
> ----------------------------------------------------------------------------------------------------------------------
> --- crash-4.0-2.21/ppc64.c.orig 2006-02-23 15:55:03.000000000 -0800
> +++ crash-4.0-2.21/ppc64.c      2006-02-23 16:36:53.000000000 -0800
> @@ -1538,7 +1538,7 @@ ppc64_get_dumpfile_stack_frame(struct bt
>         /*
>          * For the kdump vmcore, Use SP and IP values that are saved in ptregs.
>          */
> -       if (pc->flags && KDUMP)
> +       if (pc->flags & KDUMP)
>                 return ppc64_kdump_stack_frame(bt_in, nip, ksp);
>
>          bt = &bt_local;

Good catch -- I've got this one.

Thanks,
  Dave





More information about the Crash-utility mailing list