If there is a possibility of aliased mappings with different colors
from a process and a kernel, then it's possible that process writes
(and it goes to memory because cache is write-through) but kernel can
access stalled cache line (which is not coherent) if it copies it.
This somewhat explains why data corruption happens only over nfs.

Extra dcache_flush_page can trash cache contents (it flushes process' 
page not kernel's, so it's not correct stalled cache line btw) and so
occasionally invalidates it.

I have no proof for this schema, it's working scenario.

	Andrey

On Sun, Aug 25, 2002 at 04:34:08PM -0700, Chuck Silvers wrote:
> actually, on second thought, I was wrong.  the change below really
> shouldn't be necessary.  pmap_clear_modify() needs to make sure there
> aren't any dirty cache lines in a VAC, but the VAC on the ultrasparcs
> is write-through, so there can't be any dirty cache lines in the VAC.
> the real problem is something else.
> 
> -Chuck
> 
> 
> On Thu, Aug 22, 2002 at 10:23:16AM -0700, Andrey Petrov wrote:
> > On Thu, Aug 22, 2002 at 11:47:53AM +0200, Hubert Feyrer wrote:
> > >
> > > Using the latest snapshot from releng.netbsd.org, I get lots of "signal
> > > 11"s during compilation. I'v replaced the machine's RAM (4*256 -> 4*64),
> > > but that didn't make any differences.
> > > 
> > 
> > I'm not sure if that hme or not but the following patch fixed
> > 'signal 11' during NFS build for me. I have some theory about
> > aliased shared pages, but it's just guess with no prove at this 
> > point.
> > 
> > 	Andrey
> > 
> > Index: pmap.c
> > ===================================================================
> > RCS file: /cvsroot/syssrc/sys/arch/sparc64/sparc64/pmap.c,v
> > retrieving revision 1.125
> > diff -u -p -r1.125 pmap.c
> > --- pmap.c      2002/08/12 12:04:31     1.125
> > +++ pmap.c      2002/08/22 17:14:52
> > @@ -2805,6 +2805,7 @@ pmap_clear_modify(pg)
> >                                 changed |= 1;
> >                         pv->pv_va &= ~(PV_MOD);
> >                         simple_unlock(&pv->pv_pmap->pm_lock);
> > +                       dcache_flush_page(pa);
> >                 }
> >         splx(s);
> >         pv_check();