On 2005-06-16 simonb@wasabisystems.com wrote:
> On Mon, Jun 13, 2005 at 01:36:31PM +0000, Andy Ruhl wrote:
>
> > I just got done building 3.0_BETA yesterday for hpcmips. I built a
> > release on an i386 running 2.0.2.
> >
> > The build was failing for various reasons on my i386 and also on my
> > amd64 3.0_BETA machines, but I cvs'd the sources every time and
> > yesterday it finally decided to build completely.
> >
> > Here are the exact messages on the screen:
> >
> > mem_cluster_cnt = 2
> > mem_clusters[0] = {0x0,0x37b000}
> > mem_clusters[1] = {0x37b000,0x1c85000}
> > loading 0x37b000,0xc85000
> > Copyright (c) 1996, 1997, 1999, 2000, 2001, 2002, 2003, 2004, 2005
> >         The NetBSD Foundation, Inc. All rights reserved.
> > Copyright (c) 1982, 1986, 1989, 1991, 1993
> >         The Regents of the University of California. All rights reserved.
> >
> >
> > And that's it. It won't boot any farther. Do I need to pass in some
> > options via pbsdboot.exe, or possibly recompile with some debug
> > options?
>
> sys/arch/hpcmips/hpcmips/machdep.c does:
>
> 	printf(version);
> 	sprintf(cpu_model, "%s (%s)", platid_name(&platid), cpu_name);
> 	printf("%s\n", cpu_model);
>
> You're seeing the version, but not the cpu model, so it looks like the
> platid stuff is failing somehow.

  It is before that, uvm_kmapent_alloc called from uvm_map
from uvm_km_alloc1 in kmeminit.  The line that dies is:
                entry->flags = UVM_MAP_KERNEL;
which is trying to set 0xc0000fd0.  Trying to examine that in DDB just
before it would die gets a TLB miss trap, so I guess something didn't get
set up right.  I can't look any more right now.

Matthew Orgass
darkstar@city-net.com