Installing 7.1 failure

[Havard Eidnes <he%NetBSD.org@localhost>]

Hi,

I finally got around to replacing the harddrive in my PPC Mac
Mini which is going to run 7.x, and while the installation itself
went relatively smoothly, it appears that either the bootloader
or the GENERIC kernel has problems starting up.

I've a NetBSD-dedicated drive, so partitioned with pdisk, edited
the NetBSD disklabel (restored with -R -r), newfs'ed the file
systems manually, used "existing partition sizes" and had to give
the mount points to the file systems to sysinst etc., and lastly
installed the hfsutils package (from a 6.0 binary) and placed
NetBSD 7.1's ofwboot.xcf on the HFS partition using hfsutils.

The boot process starts but by the looks of it the kernel fails
to run.  What's printed on the console is:

>> NetBSD/macppc OpenFirmware Boot, Revision 1.12
9184964+173112 [385168+379774]=0x9a7c68
 start=0x100000
No ADB support present, assuming USB keyboard
mem region 0 start=0 size=20000000
avail region 0 start=0x12000 size=0xdee000
avail region 1 start=0xe1a000 size=0x1edd6000
trap: kernel ISI by 0xff8472b0 (SRR1 0x40003030), lr: 0x1001e4
panic: trap
Stopped

and that's it.  Now, who's at fault here?  The kernel?  The boot
loader?  The combination?  The "No ADB support" message comes
from the kernel's ofw_consinit.c, so this points towards the
kernel, but see below.

It seems that the GENERIC macppc kernel already has
DDB_COMMANDONENTER="bt", but perhaps this problem happens before
DDB has been initialized?

I was sort of hoping that the ofwboot.xcf problems were now a
thing of the past(?), and the same boot loader (I'm guessing it's
the same) had no problem booting the installation kernel from my
CD-ROM.  And if I read "strings -a ofwboot.xcf" output on my
6.1-running macppc, it also has revision 1.12 of the boot loader.

The weird thing is that the 7.0 GENERIC kernel fails in exactly
the same way, with the exact same addresses in the trap: message.
And ... this machine *did* run 7.0 before the old disk decided to
not cooperate anymore.

I'm sort of out of good ideas for finding the actual source of
this problem by now...  So hints are most welcome.

Regards,

- Håvard