OS9 Hard Disc Image

[BeebMaster]

Tonight I made an image of my NEC D3126 ST506 drive, which came with the Xebec S1410 controller, containing OS9 v2.3. It was relatively painless, which I was glad about, as I'd been saving it to the end of my hard drive backups which I started in September 2019 and was dreading it a bit really. But everything worked very nicely and the drive seems to be in very good condition.

ADFS didn't like it much, but ADT's DEX let me look at the first sector:

According to this, which was the best thing I could find after nearly 12 seconds of searching, it looked like it could be an OS9 boot sector. The first three bytes are the number of sectors, and it's stored in human readable left-to-right format, instead of back to front like we are used to reading multi-byte hex numbers, &13380 being the exact number you get from the disc geometry of 615x4x32.

So I set my ADFS Imager to work and it did it with one disc error the first time and no errors the second time. The only thing is that ADFS decided that the end of the disc was at sector &4C80, which is about the default setting for MFM bridge boards for a 5MB drive, so I will have to learn more about the Xebec controller command set to feed it the right geometry.

I don't think it matters, as there isn't any data after about two-thirds in to the 5MB image so probably the rest of the disc is blank. So here it is, and now I've preserved it, what can I do with it?

[Pernod]

Thanks, looks like I have a new toy to play with

Are you sure about the geometry, a quick search suggests a D3126 has geometry 615x4x17 and sector size 512.

[BeebMaster]

17 sectors per track at 512 bytes per sector generally translates to 32 or 33 sectors per track (depending on controller and sector skew) at 256 bytes per sector, which is what ADFS requires and I don't think I've seen any other (historical) BBC Micro hard disc implementation which uses anything other than 256 byte sectors. That matches the sector count in the first 3 bytes of sector 0, and scanning through the image it looks like where data would cross a sector boundary, it's all there (ie. text follows on with no obvious gaps which you would get if I was only capturing 256 of every 512 bytes).

What do I do next I wonder? Is this for use with for example the 6809 Pi Tube second processor? I tried it, but it doesn't immediately try to boot from a disc, like the Z80 and 80186 second processors do. Was there an OS9 filing system ROM for the Beeb?

[Pernod]

What do I do next I wonder? Is this for use with for example the 6809 Pi Tube second processor? I tried it, but it doesn't immediately try to boot from a disc, like the Z80 and 80186 second processors do. Was there an OS9 filing system ROM for the Beeb?

I'll be trying to use it with the Cumana 68008 2nd processor, see http://chrisacorns.computinghistory.org ... essor.html, which I've emulated in MAME.

[BeebMaster]

Mmmm so it plugs straight into the Beeb motherboard, displacing the 6502, has its own floppy and (SCSI/SASI) hard drive ports and doesn't use the Tube!

[BeebMaster]

I've spent the last couple of days working through the Xebec manual and writing little utilities to test out many of the controller's op codes. I've realised that there was a controller reset between the time I used the "Initialise Characteristics" command (sort of primitive mode select) and starting up my ADFS imager, so the geometry was lost simply because I pressed BREAK so I could go into the Native ARM co-processor.

I've now found that I can access all sectors up to the actual geometry of 615x4x32 (&13380) if I don't reset the controller after issuing it with the right geometry. Its behaviour in this state is slightly odd when reaching the end of disc, as instead of generating a volume overflow error (&21) like other ST506 controllers do, and indeed the Xebec was doing when it was in its default 153x4x32 geometry, it retries repeatedly before giving a sector address mark not found error (&12), and then trying to continue. So I can't use that as an end-of-disc test as it can be caused by multiple reasons.

But by just imaging 0-&1337F, I've been able to make a complete image of the whole disc. Unfortunately, there's nothing in the part that I couldn't see before, it's just 15MB of blank sectors, but I couldn't know that until I tried of course. Here it is for completeness.

[RobC]

This is great - thanks for taking the effort to preserve it.

I'll be trying to use it with the Cumana 68008 2nd processor

Looks like that it what it's from - there appears to be a 6502 Beeb paged ROM image starting at 0x19700 in the latest dump that says:
"68008 on the BBC Version 2.2 (C) Snowtop Computers 1989" and "OS9 - OS-9/68000 from Cumana ... 68008 Upgrade Board for the BBC from Cumana".

Would be great to see this up and running.

[Pernod]

Looks like that it what it's from - there appears to be a 6502 Beeb paged ROM image starting at 0x19700 in the latest dump that says:
"68008 on the BBC Version 2.2 (C) Snowtop Computers 1989" and "OS9 - OS-9/68000 from Cumana ... 68008 Upgrade Board for the BBC from Cumana".

Would be great to see this up and running.

That's interesting, haven't actually played with it yet but the only known dumped ROM that I'm using says: "68008 on the BBC Version 1.1 (C) Vivaway 1985", which is a 16K ROM containing about 5K code.

I have had the board emulation running, booting from floppy images, but it's very sensitive to interrupts and needs some attention.

[RobC]

I ripped out that ROM image and fired it up in BeebEm. It gives the expected help message and implements *OS9 but then, unsurprisingly, hangs as I guess it waits for the co-pro board to respond.

[BeebMaster]

I extracted it too, using an OSWORD sector read operation on the actual disc. I've no way of telling how long the file is meant to be as I can't interpret the OS9 file system, so I took 16K. It's very likely to be much shorter, as there are some blank sectors (&6C fill pattern) in there.

Master 128 doesn't like it! It just says This is not a language when trying *OS9 or *FX142. However it starts on a BBC B but doesn't get anywhere. I'll have to try the Xebec apparatus connected to a BBC B to see if it is actually waiting for the disc (hopefully) rather than the second processor.

[RobC]

Just tried it in BeebEm again.

I used the debugger to break while it was hanging on the *OS9 call and it was stuck in this loop:

Code: Select all

812E LDA &2400
8131 BEQ &812E

From somewhere (probably on here ), I have a file called OSROM9.ASM which claims to be the assembler source code listing for the "DMA PROCESSOR FOR OS9/68008 (6502 VERO CPU CARD)". It has some lines that say:

Code: Select all

STA $2400          ; Set $2400 ($00000400) to $00 Used to send commands
LDA #$40           ; to BBC B OS IE  PRINT GET

So, it looks like the Cumana board uses a window at 0x2400 to send messages to the host and the ROM is polling this address waiting for commands from the board...

[Pernod]

The Cumana board has full access to the BBC RAM and IO, and the board maps a WD2797 and two 6821 PIA's into FRED, one of which acts as a SASI interface.

All the ROM does is copies a bootstrap into RAM and passes control to the 68008, which proceeds to boot a floppy. I have a floppy image labelled OS9BOOTHD, so wouldn't expect anyone to have any luck without the actual board in place.

[BeebMaster]

Shame, I would really have liked to get it going. I wonder if the person who sold me the drive had the Cumana board as well.

[RobC]

Shame, I would really have liked to get it going. I wonder if the person who sold me the drive had the Cumana board as well.

There are some schematics here but I don't know how complete they are.

It does seem like an interesting board - I have a Casper 68000 co-pro but there's not a lot of software for it.

I wonder how much work would be involved in porting OS/9 to a 68000 TUBE co-pro?

[BeebMaster]

I think Chris Whytehead must be the only person who has one. Interestingly, the box the drive came in originated at Cumana, I've only just made the connection despite having purchased it 15 years ago this coming Saturday!

[RobC]

I think Chris Whytehead must be the only person who has one. Interestingly, the box the drive came in originated at Cumana, I've only just made the connection despite having purchased it 15 years ago this coming Saturday!

I guess it must be at the Centre for Computing History now then?

[Pernod]

I guess it must be at the Centre for Computing History now then?

Yep, they seem to have three of them http://www.computinghistory.org.uk/page ... Match=true

[BeebMaster]

I photographed it this afternoon and put it away, so that brings the Winchester Disc archiving to an end. Now to start on the floppies...

[jgharston]

Master 128 doesn't like it! It just says This is not a language when trying *OS9 or *FX142. However it starts on a BBC B but doesn't get anywhere. I'll have to try the Xebec apparatus connected to a BBC B to see if it is actually waiting for the disc (hopefully) rather than the second processor.

It probably doesn't have the 'language' bit set in the ROM type. The Master is stricter than the Beeb on entering language ROMs, and refuses to enter one if the language bit is not set. The Beeb will enter whatever ROM you tell it to.

[IBM portable PC]

I think Chris Whytehead must be the only person who has one.

Not quite… about 7 years ago I purchased a Cumana 68008 board on ebay for about US$70. The owner had modified it to work without a beeb, via a serial port. I removed his interesting modifications and restored it to working order inside my beeb. I also made PDF of manuals available, along with dumps of a few ROMs which it came with, and also images of all of the OS/9 floppy disks.

The board runs, although OS/9 68K does hang occasionally, I suspect due to overheating issues inside the beeb. It’s a large board and this does impact airflow, at some point I am going to try adding a couple of small fans.

What I do not have is a hard drive to attach to the SASI port. To use this image, the simplest approach might be to try and source the same type of drive it came from? I could try one of the many SASI/SCSI emulators, however, that will introduce more complexity when I would simply be trying to get the image to boot at this stage?

[BeebMaster]

Looking again at the pictures of the Cumana 68008 on Chris's Acorns, as it's a 50-pin connector, but described as SASI rather than SCSI, I would expect that it needs a controller board between the Cumana board and the drive, like the Adaptec ACB4000 series which were in common use with BBC Micro ST506 drives. My OS9 drive had a Xebec controller, which performs the same function but operates differently. I don't think it would be possible just to attach an early 50-pin SCSI drive to that connector, but who knows? The installation guide doesn't describe in any detail what the "hard disc unit" entails.

Probably SCSI2SD could be made to work, that's the only one I have any experience of using.

[Pernod]

I also made PDF of manuals available, along with dumps of a few ROMs which it came with, and also images of all of the OS/9 floppy disks.

Check your PM's, I sent you one in 2019 requesting photos of the floppy labels.

[Prime]

At a quick glance, it looks like the OS9 tools that the CoCo/Dragon community use can access this image.

For example the root dir :

Code: Select all

$ os9 dir OS9v2

                           Directory of OS9v2,
OS9Boot         OS9DEV          CMDS            C               SYS
DEFS            LIB             startup         init.ramdisk    greeting
2.2FILES        ASM             OldBoot

Toolshed is available here :

https://sourceforge.net/projects/toolshed/

Once compiled os9 on it's own will list the available commands.
Paths in the file are specified as filename,path/path e.g. to list the files in SOURCE within C you would use :

os9 dir OS9v2,C/SOURCE

To show the contents of the READ_ME file in that directory :

os9 list OS9v2,C/SOURCE/READ_ME

To dump the file in hex/ascii
os9 dump OS9v2,C/SOURCE/READ_ME

To copy the file from the os9 image to your native os :
os9 copy OS9v2,C/SOURCE/READ_ME readme

Cheers.

Phill.

[BeebMaster]

Oo, is this from my image I posted? If it is, how exciting! If not, it's still exciting.

[Prime]

Oo, is this from my image I posted? If it is, how exciting! If not, it's still exciting.

Yep from your image, I've poked it a bit and it looks good, I've accessed a couple of files and the contents seem reasonable.

Cheers.

Phill.

[Prime]

Found this :

Code: Select all

$ os9 list OS9v2,greeting

*************************************************************************
*************************************************************************
****                                                                 ****
****                 WELCOME TO OS-9/68K Version 2.3                 ****
****                                                                 ****
****                      for the BBC Upgrade                        ****
****                                                                 ****
****             Implemented by Snowtop Computers Limited.           ****
****                                                                 ****
*************************************************************************
*************************************************************************

May be of help.....

Cheers.

Phill.

[BeebMaster]

That's good news then, if we can figure out a way to use SCSI2SD or similar with the Cumana board then this image might be of use.

[IBM portable PC]

That's good news then, if we can figure out a way to use SCSI2SD or similar with the Cumana board then this image might be of use.

I have a couple of Xebec controllers in storage, what model did the drive come with and was there a firmware version number on an EPROM etc?

I am also considering buying a Decromancer MFM emulator https://decromancer.ca/mfm-emulator/ although not clear yet how your image would be loaded onto the SD card etc…

Overall, given there are schematics and I can provide high resolution photos of my board, I wonder if someone would be inclined to recreate the PCB? I suspect most of the chips are still around e.g. there are plenty of cheap 68008’s on US ebay currently.