> > On a similar topic, supporting something like the sun2 VME SCSI controller is
> > also interesting, since it's really SASI, not SCSI.
>
> Actually an alpha version of this driver is pretty much finished.  See
> sys/dev/ic/sunscpal.[ch], sys/dev/vme/sc_vme.c, and
> sys/arch/sun2/dev/sc_mbmem.c.  I use the Multibus version all of the
> time; alpha testers for the VME version are welcome.
>
> This controller is so simple, it doesn't look at the commands you ask
> it to put on the wire - meaning it's not really a SASI-only or SCSI-1
> only controller.  I usually have more modern SCSI-2 disks and tapes
> hooked up and everything works fine.

Well, disconnect/reconnect doesn't work last I checked. When I did the SCSI
rewrite at Sun (aeons ago) I also did unified ncr (for the 5 different 5380
variants that Sun had) and the sc driver as well. I found it indeed did mostly
work with a base disk driver that assumed CCS.

> > That actually gives me an idea. Since the ACB4000 && the MT-02 really aren't
> > even SCSI-1, let alone even CCS, perhaps another sub-bus attachment defined as
> > "sasi" might be in order... naw, it's really too complex. Maybe the best
> > approach here is to add
> >
> > #ifdef	sun2
> >
> > and do quirks. I dunno. Like I said- this is tough. It's more than this,
> > though. There is a base assumption for most of the SCSI code about some
> > minimal response level and sane response. Chances are pretty good that you'd
> > choke over other commands as well,
>
> I think that except for INQUIRY, the board knows the commands that
> drives are supposed to know.
>

If you say it works with the current sd- and that the ACB4000 at the other end
should work, that's good. There are things beginning to creep in (like
Synchronize Cache) that might send such a board into the weeds- you have to
see what they do.

> mhitch's patch is the kind of thing I was thinking about.  This
> weekend I'll try it out.  Maybe it could be extended to fake more of
> an INQUIRY response, by attempting to read a Sun disklabel?  I think
> it would be great for the sun2, and even sun3 and sparc ports, if
> we could get some kind of support in.

I don't think reading a sun disk label is right (for bootstrap reasons).

But it sounds like this hack will work for the ACB4000. The MT-02 might be
another issue (and, no, I no longer have any :-()...

-matt