Subject: Re: cpu_fork() initial SPL / NTP stability.
To: Leo Weppelman <leo@wau.mis.ah.nl>
From: Steve Woodford <scw@netbsd.org>
List: tech-kern
Date: 04/11/2001 08:57:58
On Tue, 10 Apr 2001, Leo Weppelman wrote:
> On Tue, Apr 10, 2001 at 08:39:56PM +0200, Ignatios Souvatzis wrote:
> > 
> > Did somebody check the m68k implementation(s) already?
> 
> It looks like the atari implementation does not do this correctly for kernel
> threads since they will never reach the call to 'rei' in proc_trampoline

All the m68k ports, except sun[23] are affected. Looking though the code,
the cleanest way would be to move all the existing MD cpu_fork() functions
into the common m68k codebase and tack the following line onto the end of
the cpu_fork() function:

        pcb->pcb_ps = PSL_LOWIPL;               /* Ensure IPL is 0 */

There are a couple of other functions in each m68k port's vm_machdep.c
which could easily be made common while someone is in there...

Someone with more free time than me could make the changes. ;-)

Cheers, Steve