On Sun, Aug 27, 2000 at 01:07:38PM +0200, Anders Magnusson wrote:
> > On Sat, Aug 26, 2000 at 07:31:07PM +0200, Anders Magnusson wrote:
> > > > 
> > > > So, the thing that could have changed is the "wired" attribute.  This
> > > > is a purely software attribute, that communicates "this mapping is not
> > > > forgettable" to the pmap layer.
> > > > 
> > > Yep, I check this too and it is not a wiring change either :-)
> > > The panic shows up only if it is an identical mapping request to the
> > > previous one :-)
> > 
> > Do you see both directions? that is: wired to wired AND unwired to unwired?
> > 
> Eh, parse problems with this sentence :-) But to explain better:
> 
> I have a wired bit in the PTE to keep track of wired pages, so I get the
> wired comparition "for free". So, if a page changes from wired to unwired
> or from unwired to wired I catch this.

Yes. Is there any information about whether the observed panics mentioned are
caused by wired-wired changes or unwired-unwired changes, or both? This might
give some hint to the UVM gurus about what the problem is.

Regards,
	-is


-- 
 * Progress (n.): The process through which Usenet has evolved from
   smart people in front of dumb terminals to dumb people in front of
   smart terminals.  -- obs@burnout.demon.co.uk (obscurity)