radeon driver: checkpoint

[Benjamin Herrenschmidt]

Ok, I've commited a few more fixes today, and I think the biggest issues
should be ironed out by now.

This is a call for people with radeons to test the current HEAD or
ati-1-0-branch (they should really be equivalent except that the later
builds on 7.0 while the former has the EXA changes for the server from
CVS).

I'll try to go through bugzillas later today or tomorrow, but if you
reported a problem and it's gone, feel free to beat me and update the
entry (CC me if I wasn't already too please).

Note: One of the latest fix is a bit dodgy with those magic delay's. I
think I understand now why they are necessary and how to make them not
be (basically to make CRTC*_OFFSET to stick right away instead of
waiting for a blanking) but I also think that the current stuff should
work fine and is enough for the stable branch. I might rework that in
HEAD later.

Cheers,
Ben.