Re: NetBSD-current on Quadra 610

[Martin Husemann <martin%duskware.de@localhost>]

On Sat, Aug 16, 2014 at 02:50:28AM +0900, Izumi Tsutsui wrote:
> If we can specify -fstack-protector per each source
> and we can investicate which source requires -fstack-protector,
> it would help to track the route cause.

Yes, but rebooting this machine takes such a long time I didn't track
it down yet.

> There are also a bit awful comment around rtclock_intr() in locore.s
> and I wonder if such old hack still works for -current MI hardclock...
> (INTERRUPT_SAVEREG and INTERRUPT_RESTOREREG pair also look weird)

Indeed, especially the hard code stack unwinding starting from frame pointer
to get at the clockframe seems ... not very safe.

mlelstv suggested to get rid of it (see attached patch, mostly from him),
however, this does not change the behaviour for my case.

I also tried KMEM_SIZE and KMEM_REDZONE without DIAGNOSTIC, but that did
not trigger (and still runs into the problem).

I'll bite and do the full -fstack-protector per file dance.

Martin