Re: NetBSD 8.0 hang on RS/6000 43p-150 (7043-150)

Chris Hanson <[email protected]> Sat, 5 Jan 2019 14:16:19 -0800
Newsgroups gmane.os.netbsd.ports.ofppc
Message-ID <[email protected]>
On Jan 5, 2019, at 3:30 AM, Frank Wille <[email protected]> wrote:
>=20
> Chris Hanson wrote:
>=20
>> On 03.01.19 20:23:44 you wrote:
>>=20
>> I realize (now...) that the 7043-150 isn't in the supported systems
>> list but I figured I'd give it a try.
>=20
> I have no RS/6000 hardware myself and I don't remember the reasons why som=
e
> models were not supported. The hardware doesn't look special at all.

My initial assumption was that nobody who did kernel hacking had one and thu=
s had tried it. I think that=E2=80=99s ultimately been borne out; reading th=
e list archives, several other people have tried to boot the 7043-150 and ru=
n into similar issues.

>> With an 8.0 release CD, the
>> kernel starts but hangs in OpenPIC (transcribed, please forgive any
>> errors):
>>=20
>>  OpenPIC Version 1.1: Supports 4 CPUs and 16 interrupt sources.
>=20
> Interrupts will not be established at OpenPIC initialization, so I guess i=
t
> hangs while starting to scan the OFW PCI bus.

That seems consistent with what people have previously talked about on the l=
ist.

>> I can put images on floppy and attempt to boot them if anyone wants to
>> poke at it. I could also set up a system for building ofppc if people
>> just want to propose tweaks for me to try either atop the netbsd-8
>> branch or current.
>=20
> You should do the latter, if you're interested. You only need a working
> NetBSD system for kernel compilation. Then download the 8.0 source, build
> the cross compilation tools, build the kernel. It's really easy to do. Ask=

> here if you need help.

Thanks, I started this a couple nights ago and have a kernel built from the n=
etbsd-8 branch. :) (I=E2=80=99ve helped figure out some issues with hp300 on=
 some systems earlier.)

> Then you should be able to do some first debugging yourself. Don't expect
> that others do that for you, when they don't own the hardware. :)
> I would start looking in: src/sys/arch/ofppc/pci/ofwpci.c

Oh, I don=E2=80=99t expect folks to debug for me. I was just hoping that if t=
here=E2=80=99s anything obvious people would be able to point it out so I co=
uld try it. For example, starting to look in that file. I don=E2=80=99t know=
 my way around the NetBSD codebase well. :)

> For a start insert a printf() before and after init_ofppc_interrupt() to s=
ee
> whether the OpenPIC initialization returned. If that worked move the
> printf() further down in the function, until you can locate the region wer=
e
> it hangs. For example pci_configure_bus() would be my next guess. Finally
> config_found_ia() starts the auto-configuration of all PCI devices on the
> ofwpci bus.
>=20
> Good luck!

Thanks! I=E2=80=99ll put in some logging.

A couple more questions:

1. How do I actually create the floppy image to boot? Do I just put a kernel=
 named =E2=80=9Cnetbsd=E2=80=9D and ofwboot on a FAT12-format floppy and exp=
ect it to work? I don=E2=80=99t have this system in a convenient place to co=
nnect to a network at the moment, though that=E2=80=99s not impossible.

2. What arguments should I be passing at boot, and how? It looks like I shou=
ld be using e.g. =E2=80=9Cboot -v=E2=80=9D at the Open Firmware =E2=80=9Cok=E2=
=80=9D prompt, but I don=E2=80=99t recall seeing any difference in the boot m=
essages when doing so. Are boot arguments supposed to be listed at kernel st=
art?

Thanks for the assistance!

  =E2=80=94 Chris