Re: Kernel hang with latest snapshot

[David Boreham <david_list at boreham dot org>]

David Boreham wrote:

> Li Guanglei wrote:
>
> > Sorry. It should be:
> >   stap -e "probe addevent.netdev.receive {}" -D ASCII_TRACE
> Also crashes.
>
> So, I've been doing some debugging the old fashioned way.
> I commented out code until the crashing didn't happen, then
> uncommented code until it came back. Using that technique
> I've isolated the crash to this function:
>
> write_events_desc()

BTW I previously determined that the crash is NOT caused
by events that occur when a probe is hit. I did this by copying
the code for the iosyscall.open.entry probe and then commenting
out the underlying C logging payload. I still see the crashes.
So it's something in the initialization phase that's causing
the crash. This theory tends to be supported by the fact
that any LKET probe will crash my systems, even syscalls
that I'm pretty sure are never called.