Re: is __insn_barrier() needed in _splraise()

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Wed, May 06, 2009 at 11:51:56AM -0600, Michael L. Hitch wrote:
> On Wed, 6 May 2009, Manuel Bouyer wrote:
> 
> >m68k ports has an inline splhigh (_splraise() really); and I couldn't
> >determine if the assembly in this function is a barrier for
> >the compiler, or if a __insn_barrier() is needed here. Can someone
> >confirms it ?
> 
>   Looking at a disassembly of kern_softint.o for the amiga indicates that 
> it is needed.  The second check of sh_flags after the splhigh() is 
> missing.

thanks. I didn't have a m68k toolchain handy.
Can you take care of it. a __insn_barrier() is needed at the end of
_splraise(), and one is needed at the beggining of splx (before the
instruction actually lowereing the IPL). unfortunably there's
one splx implementation per m68k port ...

> 
>   Note:  the vax also appears to need this, again based on disassembly of 
> kern_softint.o.

thanks, you reached it before me :)
can you take care of splraise/splx on this port too ?

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--