Chris Gilbert <chris@paradox.demon.co.uk> writes:

> On Saturday 02 June 2001 11:27 pm, John Fremlin wrote:
> > Ben Harris <bjh21@netbsd.org> writes:
> 
> > I get the copyright printed fine, then (this time 100% reproduceable)
> > I get
> >
> >         panic: uvm_km_suballoc: unable to allocate space in parent map
> 
> If you're jumping into main yep, it will blow up, pmap requires some
> things to be setup before attempting anything vaguely normal ;)

What like? I finish initarm nicely now and go through locore.S like
the other ports.

The actual problem was that with my memory organisation I didn't have
physical_freestart set to a page boundary, so one crucial variable,
holding the kernel stack address, actually on that page, got wasted by
pmap_bootstrap. Could have been a nasty one :-)

> > which I don't understand at all. It shouldn't be running out of memory
> > and other pmap things have worked.
> 
> You are probably causing a fault to happen, what's the trace for it?

No sign of a fault, and it gets stuck still in the same place.

> > > That's not good.  It'll screw up other users of disassem() (like
> > > DDB).  The abort-fixup code should be printing the address itself.
> >
> > Agreed.
> >
> > > I'll look at making it do this.
> >
> > Cool! But please don't touch cpufunc.c for a moment - I have no CVS
> > write access and pending changes to it :-(
> 
> CVS will happily try to merge stuff, and ben has just submitted the
> changes for this ;)

I thought that Ben was going to change the stuff I had changed: if I
had been him, I should have ripped out the panic strings, that I (me
not Ben) tidied up, and bunged them into a single helper
subroutine. The result would have been an annoying clash :-)

-- 

	http://ape.n3.net