Mac OS on ANS - Modding a ROM DIMM

[robin-fo]

So, with all of the wackiness of porting things like Windows NT to PPC Macs, are we going to get modded ROMs for PowerMac 8500/8600/9500/9600 systems that can now run that funky AIX??!

Would be fun to play with!

That would indeed be cool!

 

[gogopuffs]

@Big Ben - would you be willing to document via shaky-cam mobile phone and upload to YouTube or Internet Archive and show the bootup/build process???

Thanks!!!

 

[Big Ben]

I’m afraid there is nothing special about booting MacOS it just boots like any other Power Mac: grey screen, happy mac and so on.
Open Firmware stuff was related to improper configuration.
Front LCD screen is disabled so nothing to see here too.

As far as building a ROM DIMM, I won’t be able to build others until someone design a new pcb board.

 

[Classic_II]

I just wanted to say thank you to Big Ben, John, joevt, stynx, dougg3, Eric & jocketlill.

It is absolutely amazing to see everyone come together to make something amazing happen!!!

If anyone is thinking about making a custom rom board, let me know. I was thinking of reaching out to Joe to see if we could get a blank SIMM created.

 

[joevt]

bitsavers website has PowerSurge Flash ROM pictures and schematics.
http://bitsavers.org/pdf/apple/powerpc/PowerSurge/ROM_Flasher/

 

[winocm]

I forced the boot into Mac OS with setenv boot-command bye

Now the ANS just boot like a normal Mac. To boot AIX I suspect you only have to press Command-Option-O-F and type boot to start AIX.

I tried to install a video card but it wasn't detected, maybe a bad contact or a non-working card as it was untested.

Fun-fact the Cirrus Logic 54m30 chip is limited by... software and hardware (sort-of), because the chip is designed for little-endian operations and the driver lack implementation of endianess convertion.


Source: https://wiki.preterhuman.net/ANS_Hardware_Developer_Notes - page 9 section 2.8

Doc also contains CPU board pinout!

On that note, I ended up making an OpenFirmware driver for the Cirrus Logic GD54{30,34,36,46} cards based on the FirmWorks codebase and some taking apart of the ANS code.

It seemed to work fine when I was emulating the Viper CHRP firmware.

If it's useful for anyone else, here you go.

 

[joevt]

On that note, I ended up making an Open Firmware driver for the Cirrus Logic GD54{30,34,36,46} cards based on the FirmWorks codebase and some taking apart of the ANS code.

The ANS code has one fcode image for the GD5430. I would have just used that to create an fcode image for a PCI Option ROM but some additions/changes would be needed to handle words that are in the ANS ROM but not part of the Open Firmware spec.

Your ROM is currently set to compile for GD5446 (sets the device ID in the fcode image of the PCI Option ROM to 0x00B8) but all the variants (GD54{30,34,36,46}) are handled identically by the code. They don't have differences that matter for the Open Firmware driver?

It seemed to work fine when I was emulating the Viper CHRP firmware.

Do you have more details about this? What emulator? What virtual hardware? What OS or just Open Firmware?

 

[winocm]

The ANS code has one fcode image for the GD5430. I would have just used that to create an fcode image for a PCI Option ROM but some additions/changes would be needed to handle words that are in the ANS ROM but not part of the Open Firmware spec.

Your ROM is currently set to compile for GD5446 (sets the device ID in the fcode image of the PCI Option ROM to 0x00B8) but all the variants (GD54{30,34,36,46}) are handled identically by the code. They don't have differences that matter for the Open Firmware driver?


Do you have more details about this? What emulator? What virtual hardware? What OS or just Open Firmware?

Not really. The GD variants just have different VRAM amounts for the most part. Ideally the ROM variants could be handled better with a better makefile, but I wrote it in my spare time during Winter break.

I hacked together a QEMU that *mostly* boots the CHRP Viper and Cobra ROMs (I couldn't get the timing down for deserializing the weird bitbang DIMM/SIMM presence thing and actually implement the device model for that properly, so I patched 2 words to get past the memory sizing sequence.). Enough to the OF "ok> " prompt anyway.

I tried it with the FirmWorks PReP firmware as well. It "works" on both implementations.