panic on HP rx1600

[Marcel Moolenaar]

On Jul 20, 2008, at 9:45 AM, Hiroki Sato wrote:

> Hiroki Sato <hrs at freebsd.org> wrote
>  in <20080721.003831.44053407.hrs at allbsd.org>:
>
> hr> Marcel Moolenaar <xcllnt at mac.com> wrote
> hr>   in <CB80C6F8-02CF-4E4F-82A1-2E0E0DFD910D at mac.com>:
> hr>
> hr> xc> If you don't have INVARIANTS, could you add it to your kernel
> hr> xc> configuration. I've seen kernel panics without INVARIANTS
> hr> xc> after GCC 4 was imported. Also, if you're using ULE, then
> hr> xc> please switch to 4BSD.
> hr>
> hr>  Okay, thanks.  I will do and report what is changed.  The  
> kernel was
> hr>  the same configuration as GENERIC.
>
> I confirmed that adding INVARIANTS prevents it from triggering a
> panic.  Hmm, do you know exactly which part of the tree is the cause,
> or should we narrow down this issue?

I haven't had the time to go to the bottom of this. This
is not to say that I think we should narrow it down.

Looking at the context (unlocking, turnstiles), it could
be related to atomic operations and I already fixed a GCC
bug in that respect:
	src/sys/ia64/include/atomic.h, rev 1.14

It would be extreme luck if that would also fix the panic,
but we might as well try.... Hmmm, I'd better MFC it...

-- 
Marcel Moolenaar
xcllnt at mac.com