Re: Re: Re: AmigaOS 3.2 kickstart ROM crashing on NetBSD partitions (Was NetBSD/amiga sysinst)

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

On Sat, 25 Mar 2023 at 18:00, Karoly Balogh <charlie%scenergy.dfmk.hu@localhost> wrote:
>
> Hi,
>
> On Sat, 25 Mar 2023, Karoly Balogh wrote:
>
> > OK, so my initial hypothesis was wrong, but now I have proof, that on
> > Kickstart 3.2.x opening console.device fails from the bootblock, in
> > console.c, function consinit(). That OpenDevice("console.device"...)
> > call in particular, returns -1:
> >
> > (...)
> >
> > I let the people who actually have interest in these newer Kickstarts
> > figure it out. :P
>
> So, this worked, Toni Willen (of WinUAE fame) figured out these new
> Kickstarts do not initialize console.device before bootstrap is called,
> and then Thomas Richter (of MMULib, P96, "modern Kickstart", etc fame)
> proposed a workaround which I implemented, and it seems to work.
>
> Patch here:
> https://gist.github.com/chainq/c2ae18e16d6603d9de2f6ea1f6fa88cc
>
> This made the bootblock work for me on KS v47.102:
> https://twitter.com/chainq/status/1639687720929841154
>
> The patch works like this: if opening "console.device" fails, it tries to
> initialize the "console.device" ROM module using FindResident and
> InitResident, and then open the device again. For older Kickstarts, the
> code is practically unchanged, because there the first opening already
> works. For newer Kickstarts, this works around the initialization issue.

Nice work!

I can definitely get this committed to the NetBSD tree, and pulled up
into earlier branches.
As you have everything to hand are you able to confirm it works fine
on 3.1 (and probably 2.05) - or throw up a quick hdf for others to
look at? (My build box is busy churning through hpcmips kernel builds
to try to track down a Workpad z50 issue :)

Thanks again

David