Bug#382129: Is now a good time to revisit installer not booting on OldWorld PowerMac beige G3?
On Sat, Sep 16, 2006 at 03:35:17PM -0400, Rick Thomas wrote:
> On Sep 16, 2006, at 3:05 PM, Sven Luther wrote:
> >On Sat, Sep 16, 2006 at 09:50:03AM +0200, Frans Pop wrote:
> >>On Saturday 16 September 2006 09:09, Rick Thomas wrote:
> >>>Is there someone out there who will work with me to get a kernel
> >>>boots to run debian-installer on my beige G3 PowerMac (OldWorld)
> >>I'd like to see that happen.
> >>>If not, I think it may be time to withdraw OldWorld PowerMacs from
> >>>the list of hardware supported by the Debian Installer.
> >>Note that as I understand there are also problems booting the kernel
> >>outside the installer (i.e. just upgrading an installed system), this
> >>does not seem to be a Debian Installer issue, but rather a powerpc
> >>or kernel configuration issue.
> >Its actually a problem with the ramdisk generators and/or
> >bootloaders, not the
> >kernel proper.
> Actually, from the symptoms I'm seeing, it may not be getting far
> enough to encounter ramdisk problems. (Not to say there are no such
> ramdisk problems, just that what I'm seeing is not diagnostic for
> them.) I gave fairly detailed notes in the previous bug report,
> Bug#382129. Let me know if you want more information.
Did you try 2.6.18-rc6 from : http://kernel-archive.buildserver.net/debian-kernel/trunk
Which contains a workaround for the bootx bug ?
> If Sven is right, that leaves bootloaders.
Indeed, maybe you could try the miboot-on-a-disk solution Aurelien proposed
> However, I'm not completely convinced yet that it's not a kernel
> configuration problem. My evidence for this is that the miboot
> floppy with a 2.6.16 kernel on it will boot and read it's ramdisk
> floppy. From this I conclude that it is *possible* to build a 2.6.16
> kernel (and by inference, a 2.6.17 kernel) that will boot an OldWorld
> machine. If so, the only question is whether such a configuration is
> acceptable as a general-purpose kernel for PowerMac machines,
> including NewWorld?
The problem is that for some obscure reason, the ramdisk is not read, and thus
the modules placed there by the ramdisk generator not loaded. This is the bug,
and building-in this module will pave the way to go back to the bloated 2.4
kernels we used to have, with world+dog builtin.