[AMDGPU] ring GFX timeout
Issue happens quite frequently. It may happen at any time. Watching a video. Playing a game. Opening an application. Truly anything. System just freezes partially. GPU resets. Returned image is corrupted when in GUI. Lights on GPU go out (reset) and go back on. If i have not switched to TTY the returned image is heavily corrupted thanks to restarted VRAM and corrupted memory as result of the lost power due to restart from the timeout. If im in TTY im able to grab a dmesg which im also attaching.
Im using Wayland and GNOME but happens also on Xorg.
The only workaround i found is reverting to previously installed kernel that doesnt have this issue. For me this is 5.15-rc5. Didnt have 5.16 installed as that one had different issue. 5.18-rc3 seems to partially work as well but results in crash sometimes as well usually with the same error. On 5.15 i can push the GPU in all ways and does not crash or anything.
I do have tiny overclock on the GPU but i ofc did try reverting it to no avail. The same issue occurs.
Hardware description:
- CPU: AMD Ryzen 7 3700X
- GPU: AMD Radeon 5600XT
- System Memory: 16GB 3200Mhz CL16 (Patriot VIPER4)
- Display(s): ASUS VG249Q
- Type of Display Connection: DisplayPort
System information:
- Distro name and Version: Gentoo
- Kernel version: 5.19-rc3
- Custom kernel: Just built from git (linux-git ebuild)
- AMD official driver version: MESA 22.2 straight from GIT. Latest commit today as of now.
How to reproduce the issue:
Power on the system with the affected kernel. Best results are with opening steam and if it doesnt crash there opening some game. I usually open Cyberpunk 2077. 99% chance of crash