GMA965-X3100 woes

[Lukas Hejtmanek]

On Tue, Aug 21, 2007 at 09:03:09AM +0200, Michel Dänzer wrote:
> Unlikely, hardware access is mostly handled by the client side
> i915tex_dri.so with direct rendering.

So, some other clues. It behaves in the same way with the latest Xserver from
GIT head.

It is the same with XAA or EXA and with or without framebuffercompression
regarding the lockup.

However, EXA is a bit more verbose. I got the following messages from DRM in
kernel:
[   17.346051] [drm:drm_fence_lazy_wait] *ERROR* Fence timeout. GPU lockup or
fence driver was taken down.
[   40.613325] [drm:drm_fence_lazy_wait] *ERROR* Fence timeout. GPU lockup or
fence driver was taken down.
[   40.613578] [drm:drm_bo_expire_fence] *ERROR* Detected GPU lockup or fence
driver was taken down. Evicting buffer.

and I used strace to see what glxgears do last:
19:44:30.499309 ioctl(4, 0xc03064cc, 0xbfe83d1c) = 0
19:44:30.499381 ioctl(4, 0xc03064c9, 0xbfe83c5c) = 0
19:44:30.499443 ioctl(4, 0xc03064c5, 0xbfe83c90) = 0
19:44:30.499503 ioctl(4, 0xc03064c5, 0xbfe83c90) = 0
19:44:30.499564 ioctl(4, 0xc03064c5, 0xbfe83c90) = 0
19:44:30.499628 ioctl(4, 0x40046442, 0xbfe83e40) = 0
19:44:30.499695 gettimeofday({1187718270, 499724}, NULL) = 0
19:44:30.499763 write(3,
"\230\5\4\0\6\0`\0\1\0\22\0,\1,\1\236\4\3\0v\0\0\0\6\0`"..., 36) = 36
19:44:30.499901 ioctl(3, FIONREAD, [0]) = 0
19:44:30.499964 gettimeofday({1187718270, 499992}, {4294967176, 0}) = 0
19:44:30.500037 ioctl(4, 0x4008642a, 0xbfe83cf4) = 0
19:44:30.500105 ioctl(4, 0xc08864d3, 0x80cadec) = 0
19:44:30.500173 ioctl(4, 0x40186443, 0xbfe83cfc) = 0
19:44:30.500235 ioctl(4, 0xc03064cc, 0xbfe83cac) = 0
19:44:30.500330 ioctl(4, 0xc07064cf, 0xbfe8343c) = 0

(fd #4 is connected to /dev/dri/card0, fd #3 is Xserver socket)

-- 
Lukáš Hejtmánek