KMS

[James Cloos]

>>>>> "Michel" == Michel Dänzer <michel at daenzer.net> writes:

Michel> I'm not seeing these problems, though I have a rather different
Michel> setup (Mobility RV350 with 128M of VRAM).

Michel> AGP can be unreliable, maybe try if passing radeon.test=1 on
Michel> loading the radeon DRM gives any interesting results.

Disabling AGP has no benefit.

But inspired by one of the bugz threads I tried enabling the COMPOSITE
extension and running xcompmgr -a.  That worked.  All of the anomalies
stopped occurring.

There are still two issues, though.  XV does not work; it looks like the
window is pointed at some semi-random block of ram.  The displayed stream
does not appear to have any similarity to what the client is sending.

And running glxinfo causes X to quit and leaves the display blank and
unusable.  AFAICT the keyboard is also left unusable; only the ACPI
power button works for me (acpid is configured to run »telinit 0« when
it is hit); the resulting shutdown is clean.  The only thing in the logs
when that occurs is evdev's report that it is quiting.

It looks like the code paths which are used when there is no composite
manager have not been debugged yet.

I ended up back to non-KMS for now, but plan on returning to kms when I
next compile the kernel.  It does look like I'll have to Disable glx,
though.

-JImC
-- 
James Cloos <cloos at jhcloos.com>         OpenPGP: 1024D/ED7DAEA6