Re: 43p: Install kernel boots, generic not

[David Brownlee <abs%netbsd.org@localhost>]

On Thu, 11 Jul 2019 at 20:18, Ulrich Teichert <lists%ulrich-teichert.de@localhost> wrote:
> > > I found out what the problem was/is: apparently, the firmware seems to
> > > have a limit for the file size it can boot. It's somewhere below 2M.
> > > I was able to boot my 43p with a kernel build by this configuration:
> > > [... kernel config ...]
> >
> > Does it load to the ">> NetBSD/prep BOOT" stage and then fail? If so I
> > wonder if its an issue with the NetBSD bootloader not being able to
> > find a large enough contiguous chunk of memory (which I would imagine
> > should be fixable).
>
> No, in that cases it looked like this:
> -----------------------snip------------------------
> 0 > boot net
> LOAD: Waiting 60 seconds for Spanning Tree
> BOOTP R = 1 BOOTP S = 1
> FILE: generic_com0.fs
> Load Addr=0x600000 Max Size=0x10000000
> Packet Count = 100...FINAL Packet Count = 5181   Final File Size = 2652339 bytes.

Mmmm. It should be possible to netboot a copy of the NetBSD boot
loader and then have it load the kernel (this is what Sun boxes do).
Of course it means you need a prep NetBSD boot loader that can be used
that way (I don't know if that is the case)

Another option might be to build a binary which is a compressed kernel
with a small loader which decompresses and then jumps into the kernel
- which would give a little more headroom, but still eventually fall
into the size limit.

> > Do kernel modules work on port-prep - an easy way to test might be to
> > try 'modload efs' and then 'modstat'. That could be a way to keep the
> > base kernel down while still being able to support everything
>
> Well, the first kernel with the option enabled died on boot-up,
> is there something to watch out for? It was not selected in the GENERIC
> kernel for prep anyway, so I think that's more work for later...

So to confirm - adding

options         MODULAR         # new style module(7) framework
options         MODULAR_DEFAULT_AUTOLOAD

fails to load the kernel (could this also be a size thing?)

David