Re: Dom0 PAE panic when starting xend

[Christoph Egger <Christoph_Egger%gmx.de@localhost>]

On Tuesday 10 March 2009 19:05:45 Christoph Egger wrote:
> On Tuesday 10 March 2009 18:14:58 Manuel Bouyer wrote:
> > On Tue, Mar 10, 2009 at 05:51:28PM +0100, Christoph Egger wrote:
> > > > I just tried with casting to paddr_t. Unfortunately, this doesn't fix
> > > > the mmu_update failure. xenevtmmap() runs twice before the
> > > > panic happens.
> > >
> > > xen_start_info.store_mfn has the value 0x10088a.
> > >
> > > That's not large enough that casting to paddr_t has an effect.
> >
> > Yes, even after a << PAGE_SHIFT
> >
> > x86_round_page(), x86_trunc_page() and x86_ptob() also need to be changed
> > from (unsigned long) to paddr_t. The other x86_* macros in the same block
> > are safe (only used on virtual addresses).

I just tested this change in local tree. Changing x86_round_page() and 
x86_trunc_page() as said, lets the init process die with SIGSEGV on boot.

So if you are sure, this is the way to go, then we have to fix other bugs, 
too.

> Right. You can also have the overflow problem with a 32bit PAE DomU with
> 4GB RAM. xennet(4) uses x86_btop(), for example.
> Therefore, I think, this fix should be pulled to netbsd-5.

Christoph