IBM SCSI w/cache in PS/2 Model 50 troubles

[Maikudou]

I am trying to use a Spock Prime/Anosprime SCSI adapter in Model 50z
It’s Reference disk finds it ok, and it even seems to work, I see ZuluSCSI’s devices properly, I’d 6 is an attempt to use reference disk image (it does not boot form it), and ID1 is an empty 20MB one.

During auto setup it finds it, but when it goes to set up SCSI devices it shows “An error occurred during installation of an initialization program. Select Test the Computer from the main menu to determine the cause of the problem.”


Needless to say testing does not show anything wrong, even scsi test succeed.

When I go to “Set and View SCSI Device Configuration” it shows devices (ZuluSCSI image files):

But any attempt to save configuration ends up with: “An error occurred while saving the SCSI data. Restart the computer and configure the SCSI devices again.”

What could be wrong? Where does it tries to save it and why might fail?

ADFs are from system’s reference disk 1990 dates , if I try to put newer ones 1993 dates, it can’t find the device during auto setup.

It is the same if I remove zuluscsi, it does not matter if built-in hdd controller is installed or removed. The hdd does not work, well, it works, but heads scratch disks with unbearable noice, at least it tries to seek, but I think it’s unrelated.

 

[ardent-blue]

Go for the burn!

AdapterID 0DF9Fh
AdapterName "Integrated Fixed Disk and Controller"
Fixedresources
int 0eh ["e" is 14]

Sharing an IRQ is not bad, but the next part has the DBA-ESDI drive booting first:
Prompt "Primary/Alternate Port Addresses"
Choice "Primary" io 3510h-3517h
Choice "Alternate" io 3518h-351fh

And both choices come before the lowest address of Spock... So booting from Spock with a DBA-ESDI installed may not work.

Next, zuluscsi might not play well with PS/2s. Ryan Alswede could not get a AzureSCSI [based on ZuluSCSI] to work on a 77 Bermuda. Though the Bermuda is an IML system, and the 50z is not, there are possible issues with the file images of HDs. I dunno.

> ADFs are from system’s reference disk 1990 dates , if I try to put newer ones 1993 dates, it can’t find the device during auto setup.

Uh, have you obeyed the demands by the dominatrix known as IBM?

 

[Maikudou]

Thanks! Booting from it is not on the table right now, I just struggle to make config updated. As I said it finds it but can’t “install init” or “save scsi settings”. And thus it’s always 162 two beeps on post. The fixed drive adapter is not installed at all.

Telling from what I can see Zulu scsi devices properly listed and correct size it sort of works? And it is the same without zuluscsi connected at all.

 

[ardent-blue]

ADFs are from system’s reference disk 1990 dates , if I try to put newer ones 1993 dates, it can’t find the device during auto setup.

Took me a bit to wrap my kontuzed brain around this.

First, the 50/50z looks to support the SCSI w/cache, there was an ECA on an early Tribble or Spock and a downlevel SCSI adapter. But the three can is not affected.

There are ways to update SC.EXE to the later '92 version. Try BOPT103.EXE for an update to the 8550 SC.EXE

 

[Maikudou]

Thanks again, this did not help. It also messes with my Kingston KTM-609/16 with new sc.exe. That memory extension only works with original ref disk. I guess this newer scsi card just can’t work in model 50 for some reason.

 

[Maikudou]

I also tried a "Spock", supposedly older card, with the same result. It was different first time: it set it up without errors, and then after restart after I added a sci device to it started to throw the same errors as in pics. And after that it can't set it up even after I disconnect the scsi device from it. "Remembered" something somewhere? Its strange. It only has ROM as far as I understand.
I read a bit more and it seems it SC.exe might want to instal initialization program on drive C:? Maybe that is failure reason, there is no HDD, even no HDD adapter present.
It seems it's not possible to set up a SCSI adapter without working HDD?
Is it known if anyone succeed with using SCSI without hdd on Model 50?
It's a pity, because the system works otherwise, boots from floppy and such.

How hard it would be to set it up "manually", via setting bits on the cmos or whatever it is there? Technical manual says I should not but if it mentions it, it means it is possible?

 

[ardent-blue]

I also tried a "Spock", supposedly older card, with the same result. It was different first time: it set it up without errors, and then after restart after I added a sci device to it started to throw the same errors as in pics. And after that it can't set it up even after I disconnect the scsi device from it. "Remembered" something somewhere? Its strange. It only has ROM as far as I understand.
I read a bit more and it seems it SC.exe might want to instal initialization program on drive C:? Maybe that is failure reason, there is no HDD, even no HDD adapter present.
It seems it's not possible to set up a SCSI adapter without working HDD?
Is it known if anyone succeed with using SCSI without hdd on Model 50?
It's a pity, because the system works otherwise, boots from floppy and such.

How hard it would be to set it up "manually", via setting bits on the cmos or whatever it is there? Technical manual says I should not but if it mentions it, it means it is possible?

Hot d*mn, wimmen and children down!

After adding the SCSI device, does it throw ANY boot-time errors on the screen?

Initialization usually is for memory on-adapter that the system can use. Examples, any memory expansion adapter or a MCMaster. The 512K on Spock is managed by the on-board 80C188. I suppose it is possible for the system master to access it, but why?

I looked at your set-up, the 50z is not IML, so what ID the SCSI HD is set to is less important, -BUT- you have a 1MB image on ID6, and IBM actually followed the ANSI standard, so it boots from ID6, and your 20MB partition on ID1 would not be bootable. I think. You are blaspheming the Machine Spirit. If anyone has successfully booted on an IBM PS/2 with a file-based SCSI emulator, I'd like to know.

I have never booted a 50, or a 50z, so your efforts are illuminating.

setting bits on the cmos

Gung ho! But during boot, the system compares the bits in the CMOS (or NVRAM) against the values stored on the adapter, and if they don't match, that triggers a configuration error. I think.

Nobody has tried it before.

 

[Maikudou]

Thank you for trying too help, appreciate it!
Let me go step by step what I see with comments what I think of might happening, they are in no way have to be true:

1. Working bare PS/2 Model 50z (8550-031) system with 80387 installed. No MCA adapters, no fixed disk riser/adapter, working floppy. Fresh battery. No boot errors, time is set, post memtest — 01024 KB OK. Boots both from ref disk and from a random dos bootable. Runs Wolfenstein3D an Karateka.
2. Install a Spock SCSI adapter. ANO6451018, rev01, pn 84F8015, one oscillator. No scsi devices attached to it.
3. Boot error 165. Makes sense: new adapter added.
4. Boot from ref disk. Don't update configuration just yet, got to set configuration
5. Got an error but still am able to get to "Set configuration" menu:
   
6. Under "View configuration":
   
7. OK, it says, "Select Test your computer" to see who is to blame,
8. Reboot, now got 162 not 165, keep in mind not configuration was changed yet. Go it advanced diagnostics with a CTRL-A, System checkout -> Run Tests one Time -> 1 Cached SCSI I/O Adapter(s)
9. Card reset test ends with an error
   
10. Looks like it does not like to be unterminated. OK, install scsi2sd V5.2 with auto-termination and an empty 30MB partition set up. Now all tests succeed, except on final step "Doing Device Reassignments" we have a warning:
    
11. OK. "BIOS ROMs being disabled". By whom? It asks to run configuration. OK let's do that.
12. Reboot again, 162, etc. Now let's run Automatic configuration with scsi2sd still connected.
13. Got the same during "Automatic configuration being run".
    
    
14. Then it says "Configuration complete". Enter restarts the system.
15. Still 162
16. Maybe without scsi2sd? Run automatic configuration again without scsi2sd.
17. Still the same "error during installation of initialization program".
18. Maybe if I attach a period correct SCSI HDD from 1989 Mac which I know works as a fact?
19. Nope, still the same. Even if I see it in "Set and View SCSI Device Configuration" (I can see scs2sd and ZuluSCSI volumes too if I connect them):
    

To conclude: the adapter sort of works. It can communicate with attached devices, but "BIOS disabled"?
And I can't set it up so PC boots without 162 with any SCSI device attached or with no devices attached.

Boot from it is a distant task for now. For now how to make configuration set up?

 

[mark0x01]

It has been a very long time since I tangled with MCA, but from this info I don't see it listing it's rom address - ADPItem 1 ROM Address Range

It could be a version mismatch between the configuration tool and the definition file? this shows there are two different ones.

 

[ardent-blue]

Arrghhh. SCSI error codes are different in POST versus advanced diagnostics.

00967014 8210

0096 SCSI w/cache, PUN of 7 [adapter is ID7], LUN of 0 [all IBM PS/2 SCSI use just one LUN], Slot is 1, Capacity is "4" and that is unknown to me.

Unit Reference Code [UFC] is "8" and you are not going anywhere fast...

"Note: If the Unit Reference Code is other than a '9', the following SENSE KEY and ADDITIONAL SENSE CODES (Charts 1 and 2) have DIFFERENT meanings and SHOULD NOT be used."

I have never seen any IBM SCSI where the BIOS is disabled. I actually had to re-read the ADP page. ADPItem 1 ROM Address Range is auto-generated by the Cnnnn.ADF Arrgghhh... this is not [sl]easy...
Potential ROM addresses from C8EFF.ADF are: C0000h-C7FFFh C4000h-CBFFFh C8000h-CFFFFh CC000h-D3FFFh D0000h-D7FFFh D4000h-DBFFFh D8000h-DFFFFh

But if you removed the DBA-ESDI drive, there is no competing HD device loading it's BIOS into RAM.

Initialization Programs use Innn.ADF, not Cnnnn.ADF

Spock SCSI adapter. ANO6451018, rev01, pn 84F8015, one oscillator. No scsi devices attached to it.

ECA 032 Possible Replacement Required
The old adapter w/cache can be identified by PN 64F1333 on module U47 (just below the MMKs). If you have this adapter, replace it.
Ed. I saw something about potential data loss concerning this ECA. Not sure- my U47 is the adapter BIOS.

Arrgghhh... maybe an ECA? Um, looks to be the oldest firmware you can get. This is a later firmware for the one can. I am not sure if the ECA was for the firmware -OR- if it was hardware.

 

[Maikudou]

8EFF does not seem to have initialization program at all… no I8EFF file and no InitProg in ADF.

Perhaps both my Spock’s, usual and Prime broken somehow. What’s the odds if that?
Btw, using newer adf the one with ADPItem makes 50’s sc.exe confused, it throws error reading it.

I’ll continue to use different combinations of adfs/ref discs and scsi emulator configs in case anything helps.

For now it is strange: 8EFF is supposed to be supported by model 50, it’s refdisk has the adfs…

 

[Maikudou]

I think it's cursed, McIDE does not work properly too

 

[whaka]

just for info, i got exact same error with both old spock (one can) and spock prime (three cans) in a 55sx.
i know the spock is fine, as it perfectly work with a zuluscsi in a model 70.
maybe those cards were designed to be used in a real 32 bit machine bus, despite the connector is keyed to be 16 bit compatible.
i mean, it might work in a 16 bit slot, but only in a real 32 bit machine with at least a 386DX.
i'll try it in the 16 bit slot in the 70 to see what happen, and tell you.

 

[whaka]

so i tested the spock prime in the 16 bit slot in the 70, and sure it work.
seems the prime doesn't like the zuluscsi V1.1 (even with the latest firmware), but work fine with the RP2040.

i don't know what to think about the 50/55sx, as ADF are present in the reference disk, i guess it's supposed to work.
but maybe they can only work in the 60/65?

 

[Maikudou]

Thanks for checking!
What's an RP2040?

 

[whaka]

one of the most recent zuluscsi, based on raspberry pi RP2040 microcontroller.

 

[Maikudou]

Thanks, I think I have this one (Rev 2023b), does not work in either Spocks on 50z

 

[whaka]

yep, to be sure RP2040 have a red PCB, older have a blue one.

in my situation, on the model 70, the zuluscsi rev1.1 work flawless with the older spock but not with the spock prime.
it is see, partitions are visible, but it fail to boot on, and sometime spit a long error number after ram counting at power on.
and running advanced test on "hard drive" lead to inquiry error, and show "?????????" for drives models.

 

[Maikudou]

Thanks, mine is green ("Compact Homebrew"), but I never was able to get to the point of actually using it, I can't set up the SCSI card (which seem to see the card on zulu without problems during the half-failed setup steps)
McIDE does not work either, it sets up OK, but after I enable it's BIOS, it shows the BIOS OK (XTIDE?), but it stops booting from FDD (to short stepper jerks and then nothing), and of course does not boot from CF-card either with nothing happening: it just waits for something, no non-system disk error or anything. I have to remove it to get to ref disk setup utility again.
Perhaps XTIDE version McIDE has is not fully compatible with 286-based PS/2s? Authors of McIDE said they don't know what could be wrong, except EPROM re-seating, which I did.

 

[whaka]

i don't know the compact hombrew, i only take the bigger with full size SD.

i can't speak for the McIDE, i only know it by name.
but on the XUB site, there's now a bios called "ps2", i didn't find any advice on what it is exaclty, but maybe it's for the McIDE?
worth a try i guess

if you want to stay with scsi, a known good card for 286/386sx system is the patriot. i saw very recently some for sale on ebay, but it's in UK.
it doesn't have all the bells and whistle of spock, and not initially made to be used as a main scsi controller, but sure it work in these machines.
edit : and i can tell it work flawless with the zuluscsi rev1.1, i didn't try the RP2040 yet.