On Sat, 11 Apr 2020 03:53:41 +0000 [email protected] wrote: > Similar problem also occur for amdgpu from 6.6 release to 6.7 beta, > but with a different error message: > [drm] *ERROR* ring gfx timeout, signalled seq=10088, emitted seq=10090
My amd64 with amdgpu RAVEN graphics sometimes gets this error. I got the error a few times with supertuxkart, so I don't need Minecraft. > The problem only seems to happen when I'm playing on a server where > there are a lot of activities. On an empty world I haven't > encountered this problem before. I suspect the amount of activities > lead to a huge amount of OpenGL calls and that somehow screw up the > system. supertuxkart has more trouble when I enable all the graphical effects and put the max 19 AI karts in a race. I suspect that something goes wrong when supertuxkart sends lots of data to the gpu. I have also seen glitches in supertuxkart: objects may turn invisible (like a kart with no wheels, or a set of 4 wheels with no kart) or cast out huge, opaque black shadows. This might be a problem with sending shaders to the gpu. If I open 3 supertuxkart windows at the same time, then the 3rd window has fewer glitches. In xfce, my windows may blink or temporarily disappear. I use xfwm4's compositor, which draws shadows around my windows. This might be a problem sending pixels to the gpu. If the pixels got lost, the gpu might draw an entirely transparent and invisible window. The blinking is more frequent with the bsd.sp kernel than with my usual bsd.mp. When the X server hangs, I can still ssh(1) into the system, and may find some Xorg or supertuxkart process stuck waiting at a dma fence in the kernel. "ring gfx timeout" happens if the kernel's drm scheduler got stuck at a dma fence. I have not found the cause of this problem. --George
