[Xorg-driver-geode] source of -AMD freeze narrowed down to two items

[Martin-Éric Racine]

Hello everyone,

After much discussion with Debian/Ubuntu maintainers, one thing transpired:

Bugs experienced on hardware booting off a General Software BIOS might
be a red herring, because the -AMD code works fine on other BIOSes and
when compiled against older X server core.

The real cause seems to be that since X server 1.3, the default
behavior under the hood has changed: whereas we used to employ vm86
until 1.2, we now employ x86emu since 1.3. This is necessary to, among
other things, make a one-size-fits-all server core compiled as x86_32
binary to work on x86_64 kernels.

Julien Cristau thus hinted that tracing xf86emu and finding out what
makes it fail on General Software might be a more productive approach
than trying to to nit pick our -AMD driver code. He suggests that
x86emu simply managed to expose bugs in GSW that make it crash rather
than recover gracefully.

The end result will probably involve two actions:

1. fix x86emu to not make it fail on GSW.
2. fix GSW to recover rather than fail, when some fishy BIOS probing
takes place.

The first point involves people on this list, while the second one
involves General software engineers.

The next step will involve finding a way to integrate the VSA code to
the LinuxBIOS target for GX/LX, to help hardware vendors who prefer LB
firmware get a solid performance.

Cheers!
-- 
Martin-Éric Racine
http://q-funk.iki.fi