[GH-ISSUE #5] ppc barebones drivers #5

New issue

Closed

opened 2026-11-03 22:02:03 -07:00 by gitea-mirror · 2 comments

gitea-mirror commented

2026-11-03 22:02:03 -07:00

Owner

Originally created by @raynorpat on GitHub (Oct 10, 2025).
Original GitHub issue: https://github.com/RhapsodiOS/RhapsodiOS/issues/5

Originally assigned to: @raynorpat on GitHub.

The PPC port has many more drivers than the i386 port, there are still several missing drivers that need to be created and tested

ppc driver list

This compares Mac OS X Server 1.2v3 (kernel 5.6) with our tree, as of 2026-10-18 (branch ppc-qemu). In 1.2v3 a driver is either compiled into mach_kernel (its class is in the kernel and listed in the built-in config table in autoconf_ppc.m) or loaded from /private/Drivers/ppc/*.config/*_reloc by BootX/driverLoader. Many have both forms; the on-disk copy of an in-kernel driver goes unused while the kernel's class matches.

"QEMU" means seen working on our kernel under qemu-system-ppc -M mac99 (vm/ppc-guest.py). On -M g3beige (Gossamer), both stock 5.6 and our kernel get Cuda/ADB and the display up and detect the Heathrow IDE. Our kernel has not yet booted past that point there; see #62.

Built into the 5.6 kernel

5.6 driver	Matches	Our source	Built into our kernel	QEMU
BMacEnet	`bmac bmac+`	`kernel-7/bsd/dev/ppc/drvBMacEnet`	yes	—
UniNEnet (Sawtooth GMAC)	`gmac`	`kernel-7/bsd/dev/ppc/drvUniNEnet`	yes (ppc-qemu)	yes (en0 on sungem)
MaceEnet	`mace`	`kernel-7/.../drvMaceEnet`	yes	—
DEC21x4	`apple21143`, 21x4x PCI IDs	`kernel-7/.../drvDEC21x4` (reconstructed from 5.6)	yes (ppc-qemu)	not tested built in
DECchip21041	`pci1011,14`	`kernel-7/.../drvDECchip21040`	yes	—
Apple96_SCSI	`53c94`	`kernel-7/.../drvApple96_SCSI`	yes	—
AppleMesh_SCSI	`mesh`	`kernel-7/.../drvAppleMesh_SCSI`	yes	—
EIDE (IdeController/AtapiController) + ATADisk	`ide ata ATA pci1095,646`	`kernel-7/.../drvPPCATA`, `drvATADisk`	yes	yes (root disk, CD)
ApplePMU	`via-pmu`	`kernel-7/.../drvPMU`	yes	—
AppleCuda	`via-cuda`	`kernel-7/.../drvCuda`	yes	yes (ADB kbd/mouse)
AppleUSBCMD	USB OHCI	`kernel-7/.../drvUSBCMD`	yes	not tested (USB off)
AppleOHare	`pci106b,7`	`kernel-7/.../drvOHare`	yes	—
Symbios8xx	`apple53C8xx Apple53C875Card`	`kernel-7/.../drvSymbios8xx`	yes	—
AttoScsi	`ATTO,ExpressPCIProLVD ATTO,ExpressPCIProUL2D`	`kernel-7/.../drvAttoScsiController` (reconstructed from Sym8xx and 5.6)	yes (ppc-qemu)	no hardware
AdaptecU2SCSI	`ADPT,2940U2B … ADPT,2940U2W ADPT,2940UW`	`kernel-7/.../drvAdaptecU2SCSI` (OSM and CHIM reconstructed from 5.6)	yes (ppc-qemu)	no hardware
PowerSupply	`redundant-power-supply`	`kernel-7/.../drvPowerSupply` (reconstructed, with `hw.powersupply` sysctl)	yes (ppc-qemu)	no hardware
IOMacRiscPCIBridge / VCI / Grackle / IOPCIBridge	`bandit uni-north` / `chaos` / `grackle` / `pci-bridge`	`driverkit-3/libDriver/ppc`	yes	yes (UniNorth; Grackle on g3beige)
IONDRVFramebuffer	`display`	`driverkit-3/libDriver/ppc`	yes	yes

Loadable drivers in `/private/Drivers/ppc`

1.2v3 bundle	Class	Our source	QEMU
IOADBDevice	ADBServer	`drivers-ppc/input/drvIOADBDevice`	—
IOApplePCIBus, IONDRVSupport, IODisplay	(in-kernel classes)	`driverkit-3`	—
drvPPC53c96	Apple96_SCSI	`drivers-ppc/scsi/drvPPC53c96`	—
drvPPCMesh	AppleMesh_SCSI	`drivers-ppc/scsi/drvPPCMesh`	—
drvPPCSym8xx	Sym8xxController	`drivers-ppc/scsi/drvPPCSym8xx`	—
drvPPCATA	IdeDisk, IdeController, AtapiController	`drivers-ppc/ide/drvPPCATA` (kernel ATA sources, disk class renamed to IdeDisk; ppc package builds since `f972a4be7`)	not load-tested; ATA is built into our kernel
drvPPCBMac	BMacEnet	`drivers-ppc/network/drvPPCBMac`	—
drvPPCMace	MaceEnet	`drivers-ppc/network/drvPPCMace`	—
drvPPCDec21040	DECchip21041	`drivers-ppc/network/drvPPCDec21040`	—
DEC21x4Ethernet	DEC21x4	`drivers-ppc/network/drvDEC21x4Ethernet` (stub replaced by the reconstructed driver)	stock binary attaches to tulip; superseded by the built-in DEC21x4 on our kernel
drvPPCGem	GemEnet (Sun GEM card)	`drivers-ppc/network/drvPPCGem` (register encodings match the stock binary since `c3e7c2920`; `_startChip` and the counter clear fixed)	not needed on sungem (UniNEnet)
drvPPCGNic	GNicEnet	`drivers-ppc/network/drvPPCGNic`	—
drvPPCCuda / drvPPCPMU / drvPPCOHare		`drivers-ppc/input/drvPPCCuda`, `input/drvPPCPMU`, `bus/drvPPCOHare`	—
PPCSerialPort	PPCSerialPort	`drivers-ppc/input/drvPPCSerialPort`	registers, "couldn't get interrupts" (#63)
PPCAwacs	PPCAwacs	`drivers-ppc/sound/drvPPCAwacs`	no hardware on mac99
drvPPCBurgundy	PPCBurgundy	`drivers-ppc/sound/drvPPCBurgundy`	no hardware on mac99
(not in 1.2v3)	TAS audio	`drivers-ppc/sound/drvPPCTASAudio`	—
Floppy	FloppyController	`drivers-ppc/ide/drvPPCSwimFloppy`	probe skipped without a swim3; no panic (#63)
SCSITape, SCSIServer, PortServer, BPF		`src/drvSCSITape`, `drvSCSIServer`, `drvPortServer`, `drvBPF`	PortServer/BPF load

Open items

All built-in 5.6 drivers are reconstructed and built in on ppc-qemu (AttoScsi and AdaptecU2SCSI were the last).
Serial interrupts on Sawtooth: #63.
Old World clock calibration and the g3beige boot: #62.

Originally created by @raynorpat on GitHub (Oct 10, 2025). Original GitHub issue: https://github.com/RhapsodiOS/RhapsodiOS/issues/5 Originally assigned to: @raynorpat on GitHub. The PPC port has many more drivers than the i386 port, there are still several missing drivers that need to be created and tested ## ppc driver list This compares Mac OS X Server 1.2v3 (kernel 5.6) with our tree, as of 2026-10-18 (branch `ppc-qemu`). In 1.2v3 a driver is either compiled into `mach_kernel` (its class is in the kernel and listed in the built-in config table in `autoconf_ppc.m`) or loaded from `/private/Drivers/ppc/*.config/*_reloc` by BootX/driverLoader. Many have both forms; the on-disk copy of an in-kernel driver goes unused while the kernel's class matches. "QEMU" means seen working on our kernel under `qemu-system-ppc -M mac99` (`vm/ppc-guest.py`). On `-M g3beige` (Gossamer), both stock 5.6 and our kernel get Cuda/ADB and the display up and detect the Heathrow IDE. Our kernel has not yet booted past that point there; see #62. ### Built into the 5.6 kernel | 5.6 driver | Matches | Our source | Built into our kernel | QEMU | |---|---|---|---|---| | BMacEnet | `bmac bmac+` | `kernel-7/bsd/dev/ppc/drvBMacEnet` | yes | — | | UniNEnet (Sawtooth GMAC) | `gmac` | `kernel-7/bsd/dev/ppc/drvUniNEnet` | yes (ppc-qemu) | yes (en0 on sungem) | | MaceEnet | `mace` | `kernel-7/.../drvMaceEnet` | yes | — | | DEC21x4 | `apple21143`, 21x4x PCI IDs | `kernel-7/.../drvDEC21x4` (reconstructed from 5.6) | yes (ppc-qemu) | not tested built in | | DECchip21041 | `pci1011,14` | `kernel-7/.../drvDECchip21040` | yes | — | | Apple96_SCSI | `53c94` | `kernel-7/.../drvApple96_SCSI` | yes | — | | AppleMesh_SCSI | `mesh` | `kernel-7/.../drvAppleMesh_SCSI` | yes | — | | EIDE (IdeController/AtapiController) + ATADisk | `ide ata ATA pci1095,646` | `kernel-7/.../drvPPCATA`, `drvATADisk` | yes | yes (root disk, CD) | | ApplePMU | `via-pmu` | `kernel-7/.../drvPMU` | yes | — | | AppleCuda | `via-cuda` | `kernel-7/.../drvCuda` | yes | yes (ADB kbd/mouse) | | AppleUSBCMD | USB OHCI | `kernel-7/.../drvUSBCMD` | yes | not tested (USB off) | | AppleOHare | `pci106b,7` | `kernel-7/.../drvOHare` | yes | — | | Symbios8xx | `apple53C8xx Apple53C875Card` | `kernel-7/.../drvSymbios8xx` | yes | — | | AttoScsi | `ATTO,ExpressPCIProLVD ATTO,ExpressPCIProUL2D` | `kernel-7/.../drvAttoScsiController` (reconstructed from Sym8xx and 5.6) | yes (ppc-qemu) | no hardware | | AdaptecU2SCSI | `ADPT,2940U2B … ADPT,2940U2W ADPT,2940UW` | `kernel-7/.../drvAdaptecU2SCSI` (OSM and CHIM reconstructed from 5.6) | yes (ppc-qemu) | no hardware | | PowerSupply | `redundant-power-supply` | `kernel-7/.../drvPowerSupply` (reconstructed, with `hw.powersupply` sysctl) | yes (ppc-qemu) | no hardware | | IOMacRiscPCIBridge / VCI / Grackle / IOPCIBridge | `bandit uni-north` / `chaos` / `grackle` / `pci-bridge` | `driverkit-3/libDriver/ppc` | yes | yes (UniNorth; Grackle on g3beige) | | IONDRVFramebuffer | `display` | `driverkit-3/libDriver/ppc` | yes | yes | ### Loadable drivers in `/private/Drivers/ppc` | 1.2v3 bundle | Class | Our source | QEMU | |---|---|---|---| | IOADBDevice | ADBServer | `drivers-ppc/input/drvIOADBDevice` | — | | IOApplePCIBus, IONDRVSupport, IODisplay | (in-kernel classes) | `driverkit-3` | — | | drvPPC53c96 | Apple96_SCSI | `drivers-ppc/scsi/drvPPC53c96` | — | | drvPPCMesh | AppleMesh_SCSI | `drivers-ppc/scsi/drvPPCMesh` | — | | drvPPCSym8xx | Sym8xxController | `drivers-ppc/scsi/drvPPCSym8xx` | — | | drvPPCATA | IdeDisk, IdeController, AtapiController | `drivers-ppc/ide/drvPPCATA` (kernel ATA sources, disk class renamed to IdeDisk; ppc package builds since f972a4be7) | not load-tested; ATA is built into our kernel | | drvPPCBMac | BMacEnet | `drivers-ppc/network/drvPPCBMac` | — | | drvPPCMace | MaceEnet | `drivers-ppc/network/drvPPCMace` | — | | drvPPCDec21040 | DECchip21041 | `drivers-ppc/network/drvPPCDec21040` | — | | DEC21x4Ethernet | DEC21x4 | `drivers-ppc/network/drvDEC21x4Ethernet` (stub replaced by the reconstructed driver) | stock binary attaches to tulip; superseded by the built-in DEC21x4 on our kernel | | drvPPCGem | GemEnet (Sun GEM card) | `drivers-ppc/network/drvPPCGem` (register encodings match the stock binary since c3e7c2920; `_startChip` and the counter clear fixed) | not needed on sungem (UniNEnet) | | drvPPCGNic | GNicEnet | `drivers-ppc/network/drvPPCGNic` | — | | drvPPCCuda / drvPPCPMU / drvPPCOHare | | `drivers-ppc/input/drvPPCCuda`, `input/drvPPCPMU`, `bus/drvPPCOHare` | — | | PPCSerialPort | PPCSerialPort | `drivers-ppc/input/drvPPCSerialPort` | registers, "couldn't get interrupts" (#63) | | PPCAwacs | PPCAwacs | `drivers-ppc/sound/drvPPCAwacs` | no hardware on mac99 | | drvPPCBurgundy | PPCBurgundy | `drivers-ppc/sound/drvPPCBurgundy` | no hardware on mac99 | | (not in 1.2v3) | TAS audio | `drivers-ppc/sound/drvPPCTASAudio` | — | | Floppy | FloppyController | `drivers-ppc/ide/drvPPCSwimFloppy` | probe skipped without a swim3; no panic (#63) | | SCSITape, SCSIServer, PortServer, BPF | | `src/drvSCSITape`, `drvSCSIServer`, `drvPortServer`, `drvBPF` | PortServer/BPF load | ### Open items - All built-in 5.6 drivers are reconstructed and built in on `ppc-qemu` (AttoScsi and AdaptecU2SCSI were the last). - Serial interrupts on Sawtooth: #63. - Old World clock calibration and the g3beige boot: #62.

gitea-mirror

2026-11-03 22:02:03 -07:00

closed this issue
added the
bug

help wanted
labels

gitea-mirror commented

2026-11-03 22:02:04 -07:00

Author

Owner

@raynorpat commented on GitHub (Oct 15, 2025):

I've got most of the driver frameworks complete for the ppc specific items

@raynorpat commented on GitHub (Oct 15, 2025): I've got most of the driver frameworks complete for the ppc specific items

gitea-mirror commented

2026-11-03 22:02:06 -07:00

Author

Owner

@raynorpat commented on GitHub (Oct 23, 2026):

The last two built-in 5.6 drivers are reconstructed and built into the ppc kernel on ppc-qemu.

AttoScsi: rebuilt from Sym8xx plus the 5.6 differences (new interrupt queueing, 895/896 clocking, LVD sense, period tables). The ppc kernel links with it.
AdaptecU2SCSI: all 476 bodies are recovered from 5.6, the OSM/controller plus all 392 CHIM functions with firmware and tables. The ppc kernel links with it, and all 444 reference driver symbols are present. Firmware and tables match 5.6 byte for byte. The body comparison against 5.6 found no behavioral differences beyond labelled divergences, and 2781 frozen routine cases pass on the linked kernel. The config entry now matches 5.6 and adds ADPT,2940U2W ADPT,2940UW.
Labelled fixes to 5.6 bugs:
- AttoScsi: MinSEPeriod unset, the LVD flag cleared instead of set, and the async SDTR message given the sync period.
- AdaptecU2SCSI: a NULL write in ProbeTarget, the parity-retry double completion in NormalPostRoutine, and an IOB leak in threadExecuteRequest:.

Neither driver has been run on hardware or under QEMU, since mac99 has no matching SCSI card. Serial interrupts (#63) and the g3beige boot (#62) stay open in their own issues.

@raynorpat commented on GitHub (Oct 23, 2026): The last two built-in 5.6 drivers are reconstructed and built into the ppc kernel on `ppc-qemu`. - **AttoScsi:** rebuilt from Sym8xx plus the 5.6 differences (new interrupt queueing, 895/896 clocking, LVD sense, period tables). The ppc kernel links with it. - **AdaptecU2SCSI:** all 476 bodies are recovered from 5.6, the OSM/controller plus all 392 CHIM functions with firmware and tables. The ppc kernel links with it, and all 444 reference driver symbols are present. Firmware and tables match 5.6 byte for byte. The body comparison against 5.6 found no behavioral differences beyond labelled divergences, and 2781 frozen routine cases pass on the linked kernel. The config entry now matches 5.6 and adds `ADPT,2940U2W ADPT,2940UW`. - **Labelled fixes to 5.6 bugs:** - AttoScsi: `MinSEPeriod` unset, the LVD flag cleared instead of set, and the async SDTR message given the sync period. - AdaptecU2SCSI: a NULL write in `ProbeTarget`, the parity-retry double completion in `NormalPostRoutine`, and an IOB leak in `threadExecuteRequest:`. Neither driver has been run on hardware or under QEMU, since mac99 has no matching SCSI card. Serial interrupts (#63) and the g3beige boot (#62) stay open in their own issues.