Apple Network Server MacOS based ROMs found

[joevt]

I would check the Name Registry using Display Name Registry or DisplayNameRegistry or my DumpNameRegistry to see what is visible to the OS.
https://68kmla.org/bb/threads/silic...ng-easier-way-using-flashrom.7013/post-472471
https://68kmla.org/bb/threads/attem...-the-great-gazelle-pci-hack.45493/post-503468

 

[Mr. Macintosh]

Incredible progress @PappaDF !!! Please share what you can as working together can work at continuing to the main goal! I know you said you were going to post something but I'm very excited. I want to get this info into astra as soon as possible.

 

[Mr. Macintosh]

OHHHHHHHH

Open Firmware ARC Interface Version 3.0 (Jul 12 1996 - 18:46:28)

OFClose(ff8d3740) failed
OFClose(ff8d3640) failed
OFClose(ff8d3500) failed

Booting from 'multi(0)scsi(0)cdrom(0)fdisk(0)\PPC\SETUPLDR'

OFClose(ff8d3000) failed
HJ
OFClose(ff8d2d40) failed

Windows NT Setup

OFClose(ff8d2ac0) failed
OFClose(ff8d2800) failed

Setup is loading files (Windows NT Executive)...

Setup could not determine the type of computer you have, or you have
chosen to manually specify the computer type.

Select the computer type from the following list, or select "Other"
if you have a device support disk provided by your computer manufacturer.

IBM Power Series 6015
IBM Power Series 6020,40,42
IBM Power Series 6050,6070 and RS/6000 Model 7248
IBM RS/6000 Model E20/E30/F30
IBM RS/6000 Model 45M/H45
MOTOROLA PowerStack
MOTOROLA PowerStack2
MOTOROLA Big Bend
Powerized ES, MX, LX, TX (Uniprocessor)
Powerized MX, LX, TX (Multiprocessor)
> Other

ENTER=Select ESC=Cancel F3=Exit

 

[PappaDF]

Incredible progress @PappaDF !!! Please share what you can as working together can work at continuing to the main goal! I know you said you were going to post something but I'm very excited. I want to get this info into astra as soon as possible.

I've now uploaded it here: https://github.com/pappadf/powermac-nt-hal

We can always find a better home for it later on - just want to make sure that you get access to it asap. Note that I didn't really take the time to clean it up or review it properly. There could be a lot of strange Ai-generated things in there.

To follow what exactly my AI agent have done - start with README.md and STORY.md, and continue with docs/* for additional details. Under traces/ you can follow the exact progress in terms of terminal output/screenshots.

It now passes hard drive partitioning and ends with:

 

[Mr. Macintosh]

I think same spot where @PappaDF left off yesterday needing HAL support.

I've now uploaded it here: https://github.com/pappadf/powermac-nt-hal

We can always find a better home for it later on - just want to make sure that you get access to it asap. Note that I didn't really take the time to clean it up or review it properly. There could be a lot of strange Ai-generated things in there.

To follow what exactly my AI agent have done - start with README.md and STORY.md, and continue with docs/* for additional details. Under traces/ you can follow the exact progress in terms of terminal output/screenshots.

It now passes hard drive partitioning and ends with:

View attachment 30215

WOW!!! Thank you @PappaDF

On real hardware with your work and a modified cd I was able to select past multiple screens and have arrived here

 

[PappaDF]

I think same spot where @PappaDF left off yesterday needing HAL support.

WOW!!! Thank you @PappaDF

On real hardware with your work and a modified cd I was able to select past multiple screens and have arrived here

Thanks! Great progress! How far into the boot are you here, i.e. what were the previous screens/output that you passed?

 

[Mr. Macintosh]

Thanks! Great progress! How far into the boot are you here, i.e. what were the previous screens/output that you passed?

For sure! I was able to get past picking the display adapter, and storage controller. I will be able to pick back up in a few hours. I’m telling you we are all so close I can taste it!!

 

[PappaDF]

My latest experiments/progress in a separate branch/PR: https://github.com/pappadf/powermac-nt-hal/pull/1

 

[ClassicHasClass]

I would check the Name Registry using Display Name Registry or DisplayNameRegistry or my DumpNameRegistry to see what is visible to the OS.
https://68kmla.org/bb/threads/silic...ng-easier-way-using-flashrom.7013/post-472471
https://68kmla.org/bb/threads/attem...-the-great-gazelle-pci-hack.45493/post-503468

Oh, I thought I had posted that here before. Here is what the relevant sections look like:

/Devices/device-tree/bandit
    name                             bandit                                     
                                62616E64 697400
    ?ESB                             FFFFFFFF                                   
                                "<FF><FF><FF><FF>"
    device_type                      pci                                        
                                70636900
    model                            AAPL,343S1126                              
                                4141504C 2C333433 53313132 3600
    reg                              F2000000 02000000                          
                                "<F2>......."
    #address-cells                   00000003
    #size-cells                      00000002
    clock-frequency                  01FCA055 = 33.333333 MHz
    slot-names                       00006000
                                     SLOT1_PCI0                                                                 534C4F54 315F5043 493000
                                     SLOT2_PCI0                                                                 534C4F54 325F5043 493000
    ranges                           02000000 00000000 F3000000 F3000000 00000000 01000000 01000000 00000000    "........<F3>...<F3>..................."
                                     00000000 F2000000 00000000 00800000 02000000 00000000 80000000 80000000    "....<F2>........<80>..........<80>...<80>..."
                                     00000000 10000000
    bus-range                        00000000 00000000

/Devices/device-tree/bandit
    name                             bandit                                     
                                62616E64 697400
    device_type                      pci                                        
                                70636900
    model                            AAPL,343S1126                              
                                4141504C 2C333433 53313132 3600
    reg                              F4000000 02000000                          
                                "<F4>......."
    #address-cells                   00000003
    #size-cells                      00000002
    clock-frequency                  01FCA055 = 33.333333 MHz
    slot-names                       0001E000
                                     SLOT3_PCI1                                 
                                534C4F54 335F5043 493100
                                     SLOT4_PCI1                                                                 534C4F54 345F5043 493100
                                     SLOT5_PCI1                                                                 534C4F54 355F5043 493100
                                     SLOT6_PCI1                                                                 534C4F54 365F5043 493100
    ranges                           02000000 00000000 F5000000 F5000000 00000000 01000000 01000000 00000000    "........<F5>...<F5>..................."
                                     00000000 F4000000 00000000 00800000 02000000 00000000 90000000 90000000    "....<F4>........<80>..........<90>...<90>..."
                                     00000000 10000000
    bus-range                        00000001 00000001

/Devices/device-tree/bandit/53c825
    name                             53c825                                                                     35336338 323500
    vendor-id                        00001000
    device-id                        00000003
    revision-id                      00000011
    class-code                       00010000
    interrupts                       00000001
    min-grant                        00000008
    max-latency                      00000040                                                                   "...@"
    devsel-speed                     00000001
    AAPL,interrupts                  00000016
    model                            NCR,825A                                                                   4E43522C 38323541 00
    compatible                       pci1000,3                                                                  70636931 3030302C 3300
    device_type                      scsi                                                                       73637369 00
    reg                              00008800 00000000 00000000 00000000 00000000 01008810 00000000 00000000    "..<88>...................<88>........."
                                     00000000 00000100 02008814 00000000 00000000 00000000 00000100 02008818    "..........<88>...................<88>."
                                     00000000 00000000 00000000 00001000 02008830 00000000 00000000 00000000    "..................<88>0............"
                                     00000000
    AAPL,connector                   scsi-int                                                                   73637369 2D696E74 00
    assigned-addresses               82008830 00000000 F3102000 00000000 00000000 82008818 00000000 F3101000    "<82>.<88>0....<F3>. .........<82>.<88>.....<F3><F3>..."
                                     00000000 00001000 82008814 00000000 F3100000 00000000 00000100 81008810    "........<82>.<88>.....<F3>...........<81>.<88>."
                                     00000000 00000400 00000000 00000100
    AAPL,address                     F3102000 F3101000 F3100000 F2000400                                        "<F3>. .<F3>...<F3>...<F2>..."
    driver-ist                       00057380 00000006 00000000 00000000 00000000 00000000                      "..s<80>...................."

/Devices/device-tree/bandit/54m30
    name                             54m30                                                                      35346D33 3000
    vendor-id                        00001013
    device-id                        000000A0                                                                   "...<A0>"
    revision-id                      00000022                                                                   "...""
    class-code                       00030000
    interrupts                       00000001
    min-grant                        00000000
    max-latency                      00000000
    devsel-speed                     00000001
    compatible                       pci1013,a0                                                                 70636931 3031332C 613000
    reg                              00007800 00000000 00000000 00000000 00000000 42007810 00000000 00000000    "..x.................B.x........."
                                     00000000 01000000 01007814 00000000 00000000 00000000 FFFF0400 02007830    "..........x.............<FF><FF>....x0"
                                     00000000 00000000 00000000 01000000
    device_type                      display                                                                    64697370 6C617900
    model                            STB,HORIZON+                                                               5354422C 484F5249 5A4F4E2B 00
    AAPL,connector                   monitor                                                                    6D6F6E69 746F7200
    assigned-addresses               C2007810 00000000 82000000 00000000 01000000 82007830 00000000 81000000    "<C2>.x.....<82>...........<82>.x0....<81>..."
                                     00000000 01000000
    AAPL,address                     82000000 81000000                                                          "<82>...<81>..."
    did                              00000020                                                                   "... "
    driver,AAPL,MacOS,PowerPC        4A6F7921 70656666 70777063 00000001 B0E2BF9E 00000000 00000000 00000000    "Joy!peffpwpc....<B0>⿞............"
                                     00030002 00000000 FFFFFFFF 00000000 00007CC4 00007CC4 00007CC4 00000080    "........<FF><FF><FF><FF>......|<C4>..|<C4>..|
<C4>...<80>"
[...]
                                     00000000 00000000 00000000 00000000 00000001 6E647276 7669646F 01000000    "....................ndrvvido...."
    driver-ptr                       009C8BE0                                                                   ".<9C><8B><E0>"
    driver-ref                       FFC9                                                                       "<FF><C9>"
    gprf    mods:00000020            000E000A 00000000

Notice that the 54M30 is not attached to any PCI slot, nor the 53C825. The OS treats the 54M30 as a "built-in display" but that's not good enough for the SCSI controller. The devices I installed in its actual PCI slots do show up, and have slots, e.g.,

/Devices/device-tree/bandit/IMS,tt128mbA
    name                             IMS,tt128mbA                               
                                494D532C 74743132 386D6241 00
    vendor-id                        000010E0                                   
                                "...<E0>"
    device-id                        00009128                                   
                                "..<91>("
    revision-id                      00000001
    class-code                       00030000
    interrupts                       00000001
    min-grant                        00000008
    max-latency                      00000008
    devsel-speed                     00000001
    AAPL,interrupts                  00000017
    AAPL,slot-name                   SLOT1_PCI0                                 
                                534C4F54 315F5043 493000
    fcode-rom-offset                 00000000
    model                            tt128m                                     
                                74743132 386D00
    device_type                      display                                    
                                64697370 6C617900
[...]

The slot enumeration is also different (2.0 uses things like "E1" or "C1" instead of "SLOTx_PCIy").

 

[Mr. Macintosh]

This is where I am currently at:

• New HAL: HALSHINR-0.1-MARKER.
• Both SCSI controllers receive the expected memory addresses and IRQs.
• Interrupts occur on both controllers.
• Final result remains STOP 0x7B — INACCESSIBLE_BOOT_DEVICE, with 0xC0000034.

 

[joevt]

Oh, I thought I had posted that here before. Here is what the relevant sections look like:

Notice that the 54M30 is not attached to any PCI slot, nor the 53C825. The OS treats the 54M30 as a "built-in display" but that's not good enough for the SCSI controller. The devices I installed in its actual PCI slots do show up, and have slots, e.g.,

Right. Apple System Profiler doesn't list every PCI device. Just ones in slots that have slot names.

The slot enumeration is also different (2.0 uses things like "E1" or "C1" instead of "SLOTx_PCIy").

The ANS 2.0 ROM does away with the Apple Network Server slot naming and uses PowerSurge type slot names.

Your problem is that the method used to attach the 54M30 driver to the device does not work for the 53C825 driver. Have you tried looking at TradDriverLoaderLib sample source code?
Search for it at:
https://developer.apple.com/library/archive/navigation/#section=Platforms&topic=macOS
Also search there for MoreIsBetter which contains a more recent TradDriverLoaderLib in the MoreDevices section.
But wait, TradDriverLoaderLib is for 68K drivers. You're trying to load a native PPC driver. In that case, you may need the DriverLoaderLib.

Loading native drivers is discussed in Tech Note 1189 "The Monster Disk Driver Technote".
https://developer.apple.com/library/archive/technotes/tn/tn1189.html
It has a section "Installing and Removing Drivers and Drives" which refers you to other relevant documents and sample source code.
It says that "you can set up your DriverDescription so that the system automatically finds and opens your device driver. "
Have you examined the DriverDescription from the ndrv to make sure it matches the hardware?

The tech note says "A native driver is commonly stored in a file of type 'ndrv'". Is the file type correct? Does the file have a cfrg(0) resource? Is it in the extensions folder?

Maybe you need to use the DriverLoaderLib to load the driver manually.

There's a "SCSI Driver Example" example mentioned by the Tech Note. I think that's a 68K driver though.
Actually, the PCI DDK 3.0 has source code for "an NCR 8250 PCI interface card that uses an NCR 53C825 SCSI interface chip".
You should definitely look at that PCI Driver Development Kit.

This mentions SCSI SIM:
https://developer.apple.com/library/archive/qa/dv/dv22.html

I wonder if you can add the driver to the SCSI controller while in Open Firmware? Create Open Firmware commands to create the driver,AAPL,MacOS,PowerPC property and download it via serial port using the dl command. Then see if the SCSI controller is bootable or accessible in Mac OS.

 

[ClassicHasClass]

The SCSI ndrv Extension is configured exactly the same as the 54M30 one, with the right Type and a valid cfrg resource, yes.

I'll look into the PCI DDK. Perhaps it will be simpler to just generate a driver based on that, and tweak it for the 1.1.20.1 ROM's PCI slot enumeration.

 

[PappaDF]

This is where I am currently at:

• New HAL: HALSHINR-0.1-MARKER.
• Both SCSI controllers receive the expected memory addresses and IRQs.
• Interrupts occur on both controllers.
• Final result remains STOP 0x7B — INACCESSIBLE_BOOT_DEVICE, with 0xC0000034.

I'm just the middle-man here as most of my work was carried out by ai, but nevertheless:

This sounds exactly like what my agent hit around wall 18-22 in https://github.com/pappadf/powermac-nt-hal/blob/main/STORY.md:

"Both controllers get correct memory addresses and IRQs, and interrupts occur, yet the boot device is inaccessible" sounds like you have passed what it called wall 18-21, and that you are now at wall 22 in its list. I'm of course guessing here, but if it's correct, the fix isn't a driver at all - it's four bytes in VENEER.EXE.

What wall 22 is. Before NT will name the boot device, it checks that it's still the same disc the loader booted from. It reads one known sector - 2048 bytes at offset 0x8000, where every ISO 9660 disc keeps its volume header - checksums it, and compares that against the checksum the loader recorded earlier. If the two don't agree, \ArcName\multi(0)scsi(0)cdrom(0)fdisk(0) never gets created, the boot device has no name, and you get OBJECT_NAME_NOT_FOUND.

The catch is that the veneer never read that sector. VrOpen appends ":" + partition digits to the Open Firmware path, and OF answers that on a CD by opening the disc's root directory as a file instead of the raw device - so the read came from inside a 428-byte directory listing. Look for a 2048-byte read at 0x8000 coming back short (428 on the copy my agent used) - on a raw device it could only ever return 2048.

How to confirm this on real hardware, no debugger needed. My agent claims that veneer has an undocumented debug mask, and it's settable from Open Firmware. Pass it on the boot line:

> boot <your-boot-device> -vrdebug 0x1200

0x1000 traces VrRead/VrSeek, 0x0200 traces VrOpen; output lands on ttya, which you'll already have on an ANS. Look for a 2048-byte read at 0x8000 coming back as 428. That's conclusive either way, and costs one boot. (0x2000 additionally prints the argv handed to the loader, which is handy later.)

The fix. Turn the branch at veneer image VA 0x54748 into a nop - file offset 0x4948 in VENEER.EXE, currently 0x4086003C (bne cr1,+0x3c), write 0x60000000. A straight four-byte patch on the copy in \PPC.

Important caveat: that patch is CD-only. It later turned out to make an installed disk unbootable - partition(N) then opens the whole device, offset 0 is the MBR, every filesystem recogniser rejects it, and OSLOADER dies with "Could not access disk partition tables". So take it back out for the disk boot.

You may also need wall 16. The veneer's convert_SCSI_device maps NCR,53C810 -> NCRC810, but the ANS reports NCR,825A, which falls through to UNKNOWN SCSI and Setup's mass-storage screen shows <none>. Eleven bytes at image VA 0x5F420 (file offset 0xF620); Setup then announces Symbios Logic C810 PCI SCSI Host Adapter.

All these offsets are from the OEM 000-48303 CD's VENEER.EXE. If yours differs, note that the veneer ships with its COFF symbol table intact (1512 symbols), so you can locate them by name instead

 

[PappaDF]

All my recent experiments are on the pappadf-experiments branch, not main: https://github.com/pappadf/powermac-nt-hal/tree/pappadf-experiments

Where it's got to on the emulator: text-mode Setup runs end to end, the installed system boots from disk (veneer -> OSLOADER -> registry hive -> boot drivers -> NTOSKRNL + the HAL), NT completes I/O initialisation with both 53C825As assigned and taking real interrupts, the hard disk's partition table read through this HAL and turned into C: and D:, and the CD-ROM enumerated - and GUI-mode Setup now runs in graphics mode on the Cirrus 54M30 at 640×480:

 

[PappaDF]

I pushed new content to https://github.com/pappadf/powermac-nt-hal/wiki (as before, indeed based on AI tools).