Re: (Fwd) Re: cbm 9060/9090 scsi command set

From: Ethan Dicks <ethan.dicks_at_gmail.com>
Date: Thu, 14 Nov 2013 11:12:11 -0500
Message-ID: <CAALmimkTgfhoBt_vdr4_6dwFJ6mY=hWwbBdRxjjR32G6bMJMjQ@mail.gmail.com>
On Thu, Nov 14, 2013 at 10:10 AM,  <Ruud@baltissen.org> wrote:
> Hallo Didier,
>
>> does anybody has an idea about the command set really issued by
>> the commodore controller to the scsi hard disk ?
>
> I'm afraid I have some disapointments for you.
> - There isn't a SCSI hard disk on board, it is a MFM one.
> - Between the Commodore controller and this MFM hard disk sits a
> SASI - MFM converter.

Yes.  One could theoretically plug in something different into the
50-pin SASI connector on the "DOS board" in place of the MFM drive
and SASI-MFM bridge, but I don't recall for certain if anyone has
tried it.

> It is full of unknown ICs. OK, unknown to me,

ISTR there are docs on the SASI-MFM board but for most of us,
it's a black box (I don't _think_ there are docs handy that list all
the SASI commands it recognizes).

> - The Commodore controller has a SASI interface, not a SCSI one. A
> German friend of mine said that SASI was the predecessor of SCSI

Yes.  It's "Shugart Associates System Interface" and with a few minor
changes became SCSI-1.

> so
> one should be able to attach a SCSI hard disk. So far he hasn't
> succeeded.

Embedded SCSI drives came a bit after SASI was gone.  Perhaps
a very old 20MB drive (ST225N?) _might_ work.  One issue to
overcome is that the 9060/9090 uses 256-byte sectors and those
sectors get written by the low-level format routines in the drive.
That specifically is something that changed with different ROM
revisions (I have a tech note that describes the LLF going from
approx 45 min to 2 hrs when you upgrade the ROMs).  I haven't
disassembled the ROMs to see what they are doing there, but
it would be a place to start.

Having a SCSI analyzer in the chain would be *extremely*
useful to watch/document the interaction.  I do not have such
a thing.

> On the positive site:
> - I made a disassembly of the ROM of Disk Controller, freely
> available.
> - This board does NOT use the 6530, the ROM is an EPROM in this
> case.

Yes.

> a second solution could be to replace the EPROM by one with my own
> content; not having to use this weird SASI protocol, it would
> simplify the interfacing.

I think it might be possible to tweak the ROM code to properly
support SCSI command packets, but it would probably take a
lot more tweaking to support partial sector writes to a drive that
was formatted to 512 bytes per sector.  The DOS 3.0 filesystem
maxes out around 16MB anyway (if you were to rewrite the code
that expects a ~300-cylinder drive with 32 sectors per track and
either 4 or 6 heads) so there's no real loss in giving up half of
each sector.  Even my smallest embedded SCSI drive is 50MB,
so I couldn't address every block.

I still have my D9090 drives, so I do enjoy discussing ways to
replace the HDAs.  Mine work, but like all drives, they won't
work forever.  25 years ago, I paid $100 to have one completely
torn down and refurbed (it had a bad track zero sensor which is
inside the sealed HDA, so since I was already paying the clean
room fee, I had them replace the platters too).  That drive has
less than 1000 hours on it after repair.  It may last a good long
while, but not forever.

I see two approaches to the same destination - put a SCSI bus
analyzer on a working drive and watch the transactions for
format, reads, and writes, and disassemble the DOS Board
ROMs and look for every command packet that the board
sends to the SASI-MFM board.  It's probably only about half
a dozen kinds.  Once we know what the SASI commands
are, that's a long way to sticking a replacement on the bus.

-ethan

       Message was sent through the cbm-hackers mailing list
Received on 2013-11-14 18:00:06

Archive generated by hypermail 2.2.0.