> Subject: kernel fault in pmap_page_protect4_4c()
> From: Jason Thorpe <thorpej@nas.nasa.gov>
> 
> Well, even after Erik's bug-hunting in pmap_page_protect*(), I still
> seem to randomly lose, although differently :-)
> 
> Came into work this morning to find my SS2 had gotten a kernel fault in
> pmap_page_protect4_4c(), at:
> 
> 4255                    pv_changepte4_4c(pvhead(pa), 0, PG_W);
> 
> Unfortunately, I wasn't able to jot down much more info than this... If
> it happens again, I'll try and get _LOTS_ of info... this is a heads-up,
> mostly..

That seems to mesh with some observations on what I'd seen when trying
to chase the problem.  Much sickness results from calling changepte*
on pages that shouldn't be being messed with.  The ones that I saw prior
to Erik's patch were the pages reserved for copy-on-write, but I may
have been confused.

>From what I can figure, that part of Erik's patch that properly deals with
the end of the lists helps avoid a lot of this.  The part that set pv_va
to zero is probably irrelevant and may in fact be harmful, though I don't
claim to understand why, just that someone told me taking that bit out
fixed a "doesn't boot" kind of problem.

							George