----- Original Message -----
>
>
> On 23 September 2011 14:41, Dave Anderson < anderson@xxxxxxxxxx >
> wrote:
>
>
>
>
> Why would any user task do that?
>
>
> Generally because it's buggy and has just smashed the stack, which
> dovetails nicely with the question "Why am I running a debugger?"
>
> (I'm not really sure what the right behaviour is here)
>
But a buggy task such as that would only be relevant to the
crash utility *if*:
(1) it were to generate a kernel-mode crash (highly unlikely), or
(2) if it were the active task on a cpu when a kernel crash occurred
on another cpu.
In the second case, it would receive an NMI from the crashing cpu,
which would bring it into the kernel, and the backtrace on that cpu
would start from the NMI stack. Now, in that bizarre case, I'm not
sure whether the transition from the NMI stack back (in this case)
to user space would work as expected. That's why I asked for an
example of a backtrace. But is it even worth caring about? And
if it is, it should probably be addressed in the backtrace code,
and not as the patch did it.
Dave
--
Crash-utility mailing list
Crash-utility@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/crash-utility
