tech-x11 archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

re: HEAD on intel graphics



> Another confirmation - My EEE 4G pc also now works apparently perfectly
> with 7.99.9. I still have a problem with a Radeon RV350 - which now never
> crashes, but once Xorg is started in any usual way (i.e. /etc/rc.d/xdm
> onestart or startx) gets me only a black screen with the expected processes
> running; killing them all does not restore the console, which stays black
> until a reboot. On the other hand if I start only Xorg, I get a black
> screen with functional mouse cursor, the process can be killed and the
> console is restored. Go figure...

this sounds like what i see with an old laptop with a radeon
9000M chipset.  i've been trying to track it down but so far
i'm not sure what is up.

what i see is that code in radeon_fence_wait() returns EINTR
always (possibly due to the mouse IO timer?).  this, in turn,
causes X to never move on from this point.

i suspect the reason it sort of works without any clients is
that X doesn't attempt to draw anything except the background,
and it doesn't end up calling anything that wants to wait.
as soon as you start a client that wants to draw something
X will hang.  timing can also get the mouse setup before the
clients start up, and you're left with a working mouse pointer.

i hacked around the EINTR problem by having the code in the
kernel (radeon_fence.c) always use the NOINTR wait path, which
stops it catching the SIGALRM and returning, and while that
sort of works better, it still doesn't work properly.  there
are only windows with borders, no contents.


on my PCI radeon 9250, the closest card i have to this, it
mostly works except for the red/blue swapping others have
mentioned, which makes everything look Strange.

both of these work fine with old drm, but that requires
having the old radeon_drv.so (Xorg driver) that does not
force KMS/DRI2 on you.


.mrg.


Home | Main Index | Thread Index | Old Index