openfirmware entrypoint on i386

Hi,

I've been exploring OpenFirmware on x86, under qemu with coreboot.

I've been trying to run the hello.elf example.  I have trouble
loading it from an ext2 file system, I get an message:

ok report-disk
ok boot /pci/pci-ide@1,1/ide@0/disk@0:\hello.elf
Boot device: /pci/pci-ide@1,1/ide@0/disk@0:\hello.elf  Arguments: root=/dev/ram0 mount_ramdisk=1
Unimplemented package interface procedure
Undefined word encountered 
ok

I also tried using using qemu's -option-rom function
to throw a raw binary into the address space that i
could then place in memory with 'cmove' and then
tried to 'branch' or 'execute' it.
This method probably isn't worth pursuing much,
even thought it did get me the furthest.

The problem there was that the pointer to the
openfirmware interface function wasn't getting
passed in %eax, but then again, I wasn't using 'boot'.

Anyway, tips on this would be appreciated.

Jonathan Kollasch

Commit policy (Was: openfirmware entrypoint on i386)

[Torsten Duwe]

On Monday 28 January 2008, [e-mail] wrote:
> I've been trying to run the hello.elf example.  I have trouble
> loading it from an ext2 file system, I get an message:
>
> ok report-disk
> ok boot /pci/pci-ide@1,1/ide@0/disk@0:\hello.elf
> Boot device: /pci/pci-ide@1,1/ide@0/disk@0:\hello.elf 
> Unimplemented package interface procedure
> Anyway, tips on this would be appreciated.

I've had the same problem, and solved it (look for "... virtual mode" 
2007-10-08 in the archives). However, the commit procedure to OFW is somehow 
unclear, at least to me. Currently, only Mitch commits to the x86 part of 
OFW, and blueswirl hacks on sparc -- any others?

Mitch, do you insist on checking all patches to the x86 part yourself, or do 
you allow commits by others? One Ack or two? Or must we split/branch the 
repository for OFW as coreboot payload?

Torsten

Commit policy (Was: openfirmware entrypoint on i386)

[Blue Swirl]

On 1/28/08, Torsten Duwe  wrote:
> On Monday 28 January 2008, [e-mail] wrote:
> > I've been trying to run the hello.elf example.  I have trouble
> > loading it from an ext2 file system, I get an message:
> >
> > ok report-disk
> > ok boot /pci/pci-ide@1,1/ide@0/disk@0:\hello.elf
> > Boot device: /pci/pci-ide@1,1/ide@0/disk@0:\hello.elf
> > Unimplemented package interface procedure
> > Anyway, tips on this would be appreciated.
>
> I've had the same problem, and solved it (look for "... virtual mode"
> 2007-10-08 in the archives). However, the commit procedure to OFW is somehow
> unclear, at least to me. Currently, only Mitch commits to the x86 part of
> OFW, and blueswirl hacks on sparc -- any others?

More specifically I only hack OpenBIOS/Sparc, not OFW or coreboot.
> Mitch, do you insist on checking all patches to the x86 part yourself, or do
> you allow commits by others? One Ack or two? Or must we split/branch the
> repository for OFW as coreboot payload?

If it's not too much trouble, I'd ask that at least the mailing list
should be split between OpenBIOS/OFW/coreboot.

Commit policy (Was: openfirmware entrypoint on i386)

[Tarl Neustaedter]

> If it's not too much trouble, I'd ask that at least the mailing list
> should be split between OpenBIOS/OFW/coreboot.

Do we really have enough traffic to justify splitting into multiple
mailing lists?