On Mon, Feb 16, 2009 at 1:49 AM, Chris Ball cjb@laptop.org wrote:
Hi,
Well, I was right about olpc_has_dcon() failing, and wrong about the reason why. This upstream commit breaks model detection, and reverting it gets the FB working again:
A proper fix is headed upstream now:
http://git.kernel.org/?p=linux/kernel/git/mingo/linux-2.6-x86.git;a=commit;h...
Next up is X, which is failing because it doesn't think a DCON is present, because /sys/platform/devices/dcon doesn't exist, because olpc_dcon isn't loaded, because it isn't upstream. I'll work on a better detection mechanism to put in the geode driver -- Jordan, any suggestions?
There were a few problems here:
- The DCON detection problem above. I've solved this by just changing
the test to point to a driver that's actually upstream:
http://cgit.freedesktop.org/xorg/driver/xf86-video-geode/commit/?id=990ff710...
The current HEAD was segfaulting, also now solved in the driver.
I eventually worked out that "modprobe msr" is needed before starting
the driver on Rawhide -- if you don't, you get a screen full of white snow. We'll need to get this logic ("if on an xo, do modprobe msr before X starts") into Fedora proper somehow. Any ideas?
Is there anyway to detect and load this using HAL rules?
I'll release a new version of the X driver into Rawhide within a day or so, and the kernel patch above should make its way into Rawhide via Linus at a similar pace. Once both of these are applied, I'm happy to report that I can get to a functional GNOME desktop, using Sebastian's olpc-gnome kickstart file.
Is this on the net somewhere?
Peter