MFM Emulation on IBM PS/2 Model 50 -- success

[Alison Telford]

Hi all -- not sure if this was already accomplished, but I was forever curious as to whether I could get the emulator to work with any of my microchannel PS/2 machines. I had acquired a model 80 with an mfm controller card and dead drive, and my model 50 came with a (working-ish) 20mb mfm drive and the model 50 mfm controller (both cards are equivalent, see https://www.ardent-tool.com/storage/MFM.html), specifically the "late model" mfm adapter, IBM PN 90X8643.)

ANYWAY, long story short, I was able to eventually get it to work. Using the 90X8643 card and the two cables that came in my model 80 (a 34 pin control cable with two drive connectors and a twist, much like an ibm floppy cable -- though I'm not sure if the twist is in the same location) and a 20 pin data cable, I connected to the emulator using the card edge connectors, set the drive select jumpers to 2 (e.g, DS1, e.g., second position in the jumper block of 1-4) and (CRUCIALLY) removed the resistor pack from RN1. There is a discussion on the ibm cabling on the page link above).

I set up a drive on the mfm with 615 cylinders and 4 heads (20mb) and set the PS2 bios to type 2 (a full listing of the IBM types is here (I think it's pretty accurate): https://www.win.tue.nl/~aeb/linux/hdtypes/hdtypes-3.html

At this point the PS2 hated the drive. So I saved the changes to setup, rebooted back in, and pressed CTRL-A on the main menu to get to the low level format option. I chose that, and the PS2 complained that the drive wasn't ibm. So i pressed f3 and it asked if I wanted to "prepare it" and I said yes. It hated that too, so I pressed F3 AGAIN, and then it asked if I wanted to FACTORY prepare it, so I said yes, and away it went :)

This low-level format takes quite a long time, especially with bigger drives, so make sure you have other things to do.

After that, I was able to restart the machine (I had to actually power it down, not just reboot) and then installed dos 7 after using dos 7 fdisk to create a partition.

I was also alble to get a type 22 (30mb), type 4 (60mb) and a type 9 (112mb) drive working. Sometimes it was a bit gremliny and fdisk would fail with "can't read fixed disk". Playing with the arbitration level setting of the mfm controller card in the ps2 setup and powering off (rather than rebooting) seemed to do it, though I haven't yet 100% established the pattern. I think perhaps there's a bug in the ibm code somewhere, not sure.

ANYWAY, with stacker installed on the 112mb drive, I've got a nice model 50 with an mfm drive and a 250mb or so C partition. Happy! The emulator works a treat.

I also played around a bit with speedstor 6.5 when I was having the gremlins, but though it worked it didn't seem necessary, as the IBM setup low level format was always able to complete successfully. speedstor can be found on minuszerodegrees.net

Best to all, Alison

[Alison Telford]

Hi all -- just a follow up on this -- I did some more playing around on this, seeing if I could get a protected mode OS to work correctly on my model 50, targeting OS/2 version 1.3 and 1.0 (no xtide drivers available for those afaik :)). I've managed to get both versions of OS/2 to install and operate correctly, but I wanted to ask about the "gremliny" behavior I commented on above. What this experimentation showed me was that, with this bios/controller/emulator combo, if anything is done to the disk partitioning or formatting (low level or even a filesystem format), the system as a whole enters a bugged state. For example, if the system is in a working state and I boot up into (eg) DOS from a floppy and run fdisk to partition the disk (which will succeed), then if I reboot the machine, once back to the dos prompt, the C drive is unformattable (format command fails) and fdisk will fail to load.

What seems to clear this up is to power off the pc completely AND reboot (or better power cycle) the beaglebone. (I'm running the emulator on a separate power supply). If I don't fully power off the pc and don't fully reboot or power cycle the beaglebone, the system will not work again, even if I stop and restart the mfm_emu process, or CTRL-ALT-DEL the machine, etc. Only a full emulator restart and system power cycle will return the whole to a working state. Then, (eg if I had run fdisk and created a partition before power cycling) I can continue on and format the disk. I then need to re-powercycle everything, and after that I'm usually good to go -- normal computer resets and mfm_emu process restarts work. This got messy during the installation process a couple of times, as OS/2 has you reboot several times during the install.

So.... any thoughts and/or ideas on what might be getting out of sync would be appreciated! Suggestions as to tests I could run to try to narrow down the issue also welcome! None of this is a show-stopper, and it may boil down to buggy IBM controller logic, but hey.

Cheers! Alison

[David Gesswein]

I haven't come up with likely cause. Very strange that both sides need to
be fully restarted to make it work again. What error does format give?
What does fdisk fails to load mean?

Have you tried any diagnostic programs that may give more information on what
is going wrong? Not sure what options for your machine are available. I
have used spinrite to do testing. Can also use debug to manually call the
disk read call and see what the error returned.

https://forum.vcfed.org/index.php?threads/victor-9k-sirius-1-software.1244209/page-12#post-1361793

There is a lower level way to do IO that may give better info. I would need
to look that up.

To make progress we will need to figure out some way of determining what is
really going wrong.

> --
> You received this message because you are subscribed to the Google Groups "MFM Discuss" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to mfm-discuss...@googlegroups.com.
> To view this discussion on the web visit https://groups.google.com/d/msgid/mfm-discuss/69982344-a7ab-422b-b975-5f66779d97e6n%40googlegroups.com.

[Alison Telford]

Hi David -- I posted a reply to this yesterday, but it failed to send (or I failed to hit send lol)

ANYWAY, while doing a bunch of hardware troubleshooting (cables, controller cards, power supplies), I've perhaps stumbled on the use of the --sync (-s) switch on the mfm_emu command that SEEMS??? to be helping with stability (in that I've yet to experience a crash or failure while using it). Does that make any sense to you? I've been running the emu without the power caps charged up but in my first session I'm pretty sure they were (but now I'm wondering if I'm delusional).

Cheers! A

[David Gesswein]

Not sure why --sync should help. It sets the O_DSYNC flag so data is
immediatly written to disk. Not sure how it could help unless your killing
the emulator in a way that preventes it from flushing the last data to disk.
I didn't see crash in your previous description so not sure what you are
referring to by that.

Are you saying that the same sequence where you reboot and disk can't be
accessed works if you use --sync. Is that using it before reboot, after,
or both?

> --
> You received this message because you are subscribed to the Google Groups "MFM Discuss" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to mfm-discuss...@googlegroups.com.

> To view this discussion on the web visit https://groups.google.com/d/msgid/mfm-discuss/c5dcada1-8e2a-4cb2-923c-d022e2c94962n%40googlegroups.com.

[Alison Telford]

Hi David,

I did some thinking on it today -- I'm going to spend some time looking into the possibility of bad hardware (I have a spare controller card and I've made up some more cables) and maybe even try the whole thing in a model 50Z, just to narrow down possibilities. When I say that fdisk fails, it just returns to the command prompt almost immediately with the message "can't read fixed disk". There is never any sign of an error from the emulator (I'm running it interactively, so I can see the output). Sometimes it stops registering I/O completely, but other times it will continue to indicate seeks and reads (eg on a warm reboot I will always get a seek when the machine accesses the disk before the OS boots) even when the machine is failing to operate correctly.

The effects within operating systems vary. Sometimes I'll get a "non-system disk" (eg in DOS on boot). Today in OS/2 1.0 it would sometimes give me a TRAP error (I forgot the record the specifics, I will do so later) just after startup, but other times it would pass that point but then complain about broken drivers in the config file repeatedly until it hung. (This last is interesting, as it seems to indicate it was able to read the config file at least, but then failed to load the drivers indicated by it?)

ANYWAY, I will keep working away on the issue. I agree I need to try and narrow down a specific failure behavior so I can test carefully. I'll give the debug option you mention above a try, and see about getting a copy of spinrite as well.

Cheers and thanks!! Alison

[Alison Telford]

Hi -- I don't know why, but google groups is messing with my replies here, I wrote a longer one but it got deleted.. So I will keep it short. I THINK, right now, that the problem was a weak power supply for the emulator (wall-wart style). No issues since switching to ATX psu. Running without -s switch and so far it's ok. 

What would you recommend as minimum amperage capability?

Best, Alison

[Alison Telford]

Sorry -- my posts above somehow got put out of order in this thread. Sigh.

ANYWAY -- I'm running the machine right now, without the -s option (but also having commented out the --pool 200,0.6 option in /etc/mfm_emu.conf as I'm running without the capacitors charged).  I changed out the power supply from a wall-wart style to a full ATX unit. I'm wondering if I was just operating on a borderline current level with the smaller power supply. Which would make more sense given the flakiness of everything, perhaps.

So -- let's put a pin in the --sync option, I agree, I wasn't sure how that could possibly be affecting things (I had initially thought it was purely the power supply, but then I experienced what I THOUGHT was a hang even with the ATX psu (and no -s switch) but I think now it was just OS/2 doing some VERY SLOW disk verification in the background). I will post back only if I keep experiencing crashes (and if I make any useful deductions about their cause).

Question: what would you recommend as the minimum amperage of a power supply to use?

Cheers! Alison

[d...@pdp8online.com]

For some reason google groups is marking postings as possible spam that don't look anything like spam. They don't give an option to not do spam checking. I also only get one summary each day with pending postings. Couldn't find an option to email right away. Thats why your posting are showing up strangely.

The page for Bealgebone recommends 2A at 5V. To be safe 2.5A to cover my board should be fine. By wall wart are you referring to using powering the beaglebone through the 5V barrel jack? I try to avoid that since I'm not sure if the DC/DC converter on the board can get annoyed with 5V applied to its output. Haven't really seen problems. If you were trying to power through the USB port that normally won't even power up the board when its attached to my board.

Powering from the 12V drive connector 1.25A is fine.

[Alison Telford]

I was using an (indicated) 12V 2A (wall-wart) plug attached to a converted (barrel-jack-to-molex). This psu was included in a PATA-to-USB disk adapter kit from Vantec. The molex end was plugged into the drive connector on the emulator. So, theoretically enough power, but the whole thing is pretty el cheapo and it may well be that it was delivering marginal/crappy power. I'm likely going to make up some additional ps2-to-floppy adapter boards and I ca use those for power take-off, so I don't need the external PSU. At any rate, will keep testing and working away and will let you know if the gremlins re-emerge :)

Cheers, A