Re: prom-libc status (was Re: putting yaboot and the bootinfoin one file)

["Benjamin Herrenschmidt" <benh@kernel.crashing.org>]

>for the purposes of prom-libc testing and support the machine needs to
>not require apple patches.  the 7200 and 8500 i believe also currently
>work without patching with quik, though they appear to be less
>reliable then the 9600.  

7200 maybe, but 7500, 7500 and 8500 have the same OF that sorta
works with MESH, I know I use quik without patches ;) Actually,
there are a few small OF patches floating around for those (and
nor originating from Apple) that fix a small timeout problem
with MESH (iirc) and fix the control driver in OF, though I don't
use those on the 8500 I have here.

I still think we should support installing the Apple patches even
if they are actually distributed as a separate package for licencing
reasons.

Also, Apple licence isn't _that_ bad, we can perfectly distribute
modified (improved ?) versions of their patches.

I don't know if you have the faintest clue of the work it would be
to redo those patches... You need very intimage knowledge of each
of these OF implementations, their breakage, the actual HW details
of each machine involved, etc...

>the idea for initial testing is starting with as solid a oldworld OF
>base as possible, so we can better distinguish bugs in libc from flaws
>in OF.  i think your 8500 will be useful for initial testing, the
>wallstreet maybe later, but i believe it has pretty extensivly broken
>firmware like the beige g3.  hopefully Segher can help write Free
>patches to support it though (as with the beige g3 if we are
>optimistic). 

The 8500 has the same general flaws as the 8600 (that is broken
memory allocation in OF mostly, for oldworld, we really want to
just use OF map() method to map in physical RAM and do our own
stuffs in there).

Most of the brokenness of other OFs isn't really the in core OF,
but mostly broken drivers. 

Ben.