Ring test issue (intermittent?)
I have a weird problem at boot time with drm and this may be related to this tiling bug : https://bugs.freedesktop.org/attachment.cgi?id=38539
When I reinstalled the opensource rig yesterday for testing the new S3TC stuff, I got these result:
1st boot : Everything OK I can go to the desktop and the 3d acceleration is enabled. But I did not patch the cs checker so I added a return 0; in the good place in the kernel and rebooted.
2rd boot : It's boot, no problem... except when the desktop load, it look like the linked image...
3rd boot : I rebooted on the 1st boot kernel (after recompilling all the mesa stuff... same as 2rd boot
4rd boot : same as 3rd... I finally disabled tiling... and I was able to get to the desktop!! So I proceded to repatch the kernel (cs checker) and reboot...
5rd boot : Everything is working including a S3TC enabled game (EVE-Online) but the tiling was still Disabled and I did not test it yet (I'm at work)
So here what I found while reading the logs :
1st and 5rd boot :
[drm] r600 ring test
[drm] ring test succeeded in 1 usecs
[drm] radeon: ib pool ready.
[drm] ib test succeeded in 0 usecs
On all other boot :
[drm:r600_ring_test] *ERROR* radeon: ring test failed (scratch(0x8504)=0xCAFEDEAD)
radeon 0000:01:00.0: disabling GPU acceleration
So I start to think that the ring test can be an issue for the tiling but it's just a guess. When I got the corruption problem between the first and last boot, it was the whole desktop, always.. not just games windows like in the screenshot. And the second question is why? because the same kernel has been used in the 2rd and 5rd boot and between all these boot, I just rebooted the system, no full shutdown (except one between 3rd and 4rd to test... Can it be an intermittent problem or an initialization bug on my card?
I use :
Kernel git master 2.6.38-rc4
libdrm git master
mesa git master
xf86-video-ati git master
xorg git master
It was working correctly for one week, with my two radeon on a multiseats configuration but now the problem is back and i'm unable to get the GPU accelecation.
It look intermittent. But it's not the videocards because I use two (discrete and onboard) and they have the same problem. Maybe hardware related or something else. Catalyst Linux and Windows work correctly so they probably use some quirks to initialize my card correctly.
I created a entry on the kernel bugzilla : https://bugzilla.kernel.org/show_bug.cgi?id=30102