Re: PR# 5087. It won't die.

[Ignatios Souvatzis <is%jocelyn.rhein.de@localhost>]

On Fri, Sep 25, 1998 at 07:33:34PM -0700, David Hopper wrote:
> Add me to the list of folks bitten by this one.
> 
> What confuses me is how chopps' GENERIC install kernel runs just fine.
> It's only when I config, build and boot my own kernel that the console
> 'submarines'.  It's such an odd problem that I thought it was me,
> something buried in iteconfig or grfconfig, until I found the PR,
> generated for 1.3, at
> http://www.NetBSD.org/cgi-bin/query-pr-single.pl?number=5087 .
> 
> In spite of the suggestion that it was fixed in 1.3.1, this is a 1.3.2
> kernel generated from scratch, after a fresh install from the official
> CD (thanks Herb, et.al.!).  This PR should reopen, but I really need
> some feedback on this one.  Why does the GENERIC kernel work?  Did
> chopps work from a slightly different source tree than what ended up in
> the 1.3.2 sources?

He might answer himself, but as far as I know, not.

I suggest you add

options "SYSPTSIZE=4"

to your kernel, which is no solution, just a workaround for _some people_.
The basic problem is that the old HP "pmap" module which the AMiga pmap
module is based upon can't handle big kernel address spaces, specifically
not dynamically.

> 
> Config:  A3K, WarpEngine (40/40, 32M), Picasso II (2 megs, 45ns), 50
> megs RAM total (NetBSD only sees the Warp Engine's 32 megs, total).
> dmesg is otherwise identical to the PR, except that I've  tried
> (desperately) compiling without grfcc support at all, only grfcl.

Just to make sure: did you make clean before make'ing the new kernel after
changing options?

Regards,
        -is