Re: bootable mac68k iso (Was: Tools mkhybrid(8) for mac68k/macppc ISO9660/HFS install ISO images

[Stan Johnson <userm57%yahoo.com@localhost>]

Hello,

On 6/3/24 9:27 AM, David Brownlee wrote:
> On Sun, 2 Jun 2024 at 20:21, Izumi Tsutsui <tsutsui%ceres.dti.ne.jp@localhost> wrote:
>>
>> abs@ wrote:
>>
>>>> As you wrote another message, mac68k one should work
>>>> if the previous external mkisofs version has worked.
>>>
>>> OK, so releng have confirmed the changes to include the mac68k ISO
>>> image are being deployed to the autobuild cluster, so bootable mac68k
>>> ISO images should start showing up on https://nycdn.netbsd.org/ soon
> 
> Bootable NetBSD-HEAD mac68k iso images should now be available on
> https://nycdn.netbsd.org/ - direct download should be
> 
> https://nycdn.netbsd.org/pub/NetBSD-daily/HEAD/latest/images/NetBSD-10.99.10-mac68k.iso
> 
> Now calling for people with a mac68k and CD-writer to give it a test :)
> 

I tested NetBSD-10.99.10-mac68k.iso using RAM-based video (RBV; i.e.
built-in video) on a Mac IIci with a BlueSCSI card and 65 MiB memory
(there is 1 MiB in Bank A and 64 MiB in Bank B); 65 MiB seems to be the
ideal memory configuration for both NetBSD and Linux when using IIci RBV.

This is the MD5 checksum of the file I downloaded:
MD5 (NetBSD-10.99.10-mac68k.iso) = 3ad958871f46b4f7b8fcf3f4596b17ec

The SBC installation kernel booted and installed a basic system with no
errors and only a few (mostly minor) issues:

1) I noticed that the installation only lists the first four available
SCSI disks. This would not be a problem in most situations, though with
BlueSCSI I have five targets defined specifically for this sort of testing.

Recommendation: Include all of the SCSI disks in the list of available
drives.

2) I didn't use a real SCSI CD-ROM drive, though the iso would have fit.
The BlueSCSI simulated CD-ROM is faster at around 75-80 KiB/s, compared
to about 20-40 KiB/s for a real Apple SCSI CD-ROM drive.

Recommendation: Include a DVD-sized iso that has all of the software,
including source code. There are still some working SCSI DVD-ROM drives,
though most users will be more likely to be using SCSI2SD or BlueSCSI
cards at this point.

3) There is still an issue on the IIci with RAM-based video (RBV, i.e.
built-in video). The issue is described in NetBSD Problem Report #58164:
https://gnats.netbsd.org/cgi-bin/query-pr-single.pl?number=58164

When a Toby frame buffer video card is installed and RBV is being used,
the kernel panics as described in the problem report. If a Toby card is
installed and is being used, then there is this different panic (seen
only on the serial console):

[   1.0000000] panic: Cannot work with the current memory mappings.
[   1.0000000] cpu0: Begin traceback...
[   1.0000000] ?(100000,385938,363fc4,1efca4,2c596a) at 0
[   1.0000000] cpu0: End traceback...
Stopped in pid 0.0 (system) at  2ea08:  unlk    a6
db>

The memory issue is related to the Nubus video card issue. It's possible
that the problem is only seen on systems that have RBV and Nubus video,
so the IIci and IIsi, but those are two of the most common mac68k
systems. And there's the added complication for the IIci in being able
to have up to 64 MiB in its Bank A. The IIsi has 1 MiB soldered-in for
its Bank A. The most common configuration today for a IIci system that
still works is Nubus video and 64 MiB in both Banks A and B, though both
NetBSD and Linux will ignore all of Bank A memory if RBV is being used,
and NetBSD 10 will crash if RBV is being used and a Nubus video card is
installed.

Nubus unaccelerated video worked on the IIci in NetBSD 8.2, and it
stopped working sometime between 8.2 and 9.3. I don't know how to do the
Linux equivalent of a "git bisect" in NetBSD to identify which commit
may have triggered the regression.

-Stan Johnson