Aymeric Vincent wrote:

> Basically, in the case where FIRMWORKSBUGS is used and the kernel is
> started with the MMU enabled, the openfirmware entry point is faked to
> point to a trampoline in arch/powerpc/powerpc/ofwreal.S. The role of
> this trampoline is to save and restore everything MMU-related before
> and after calls to the openfirmware.
> 
> The first thing is to patch ofwreal.S to disable the MMU while
> restoring the MMU registers. This is really needed and I will commit
> this soon without doubt.

Great! Thanks a lot. I patched this part far too often myself.


> All this leads IMHO to the fact that we should discuss a modular way
> to tell (based on the detection of the platform) which registers
> should be saved/restored before/after calling to the openfirmware and
> getting rid of the FIRMWORKSBUGS #define whose name doesn't reflect
> reality: we have to save and restore stuff on all our powerpc
> platforms, so let's not have code that does this in two places.
> But I'm not familiar with many of the powerpc platforms we support, so
> I would prefer if others could say what they think about this.

Same for me. I only have Efikas and Pegasos here.
I would guess macppc is the most important architecture without
FIRMWORKSBUGS.


BTW, some days ago I did an NFS-boot on my Efika as well, but, IIRC, there
were problems creating and populating the /dev directory.

I'm currently working on another port, but I would be happy to help once you
get over this point.

Would it make sense to create a powerpc/52xx directory, as with ibm4xx?
There is no 5200-support at the moment, AFAICS.

-- 
    _  Frank Wille (frank@phoenix.owl.de)
 _ //  http://sun.hasenbraten.de/~frank/
 \X/   Phx @ #AmigaGer