Debian Bookworm - XWayland page fault loop + [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx_0.0.0 timeout

Random freeze. I was doing nothing special. I left ffmpeg working and when I came back the graphics were frozen but the system continued working.

Hardware description:

• CPU: Ryzen 7 3700 X

• GPU: 06:00.0 VGA compatible controller [0300]: Advanced Micro Devices, Inc. [AMD/ATI] Navi 23 [Radeon RX 6600/6600 XT/6600M] [1002:73ff] (rev c7)

• System Memory: 16GB

• Display(s): LG 20M35

• Type of Display Connection: D-Sub + HDMI Adaptor

• Distro name and Version: Debian Bookworm (completely updated)

• Kernel version: Linux debian 6.1.0-13-amd64 #1 (closed) SMP PREEMPT_DYNAMIC Debian 6.1.55-1 (2023-09-29) x86_64 GNU/Linux

• AMD official driver version: Mesa 22.3.6

systemctl shows a loop of:

dic 06 15:01:35 debian kernel: amdgpu 0000:06:00.0: amdgpu: [gfxhub] page fault (src_id:0 ring:24 vmid:1 pasid:32771, for process Xwayland pid 2185 thread Xwayland:cs0 pid 2196)
dic 06 15:01:35 debian kernel: amdgpu 0000:06:00.0: amdgpu:   in page starting at address 0x000080034691b000 from client 0x1b (UTCL2)
dic 06 15:01:35 debian kernel: amdgpu 0000:06:00.0: amdgpu: GCVM_L2_PROTECTION_FAULT_STATUS:0x00000000
dic 06 15:01:35 debian kernel: amdgpu 0000:06:00.0: amdgpu:          Faulty UTCL2 client ID: CB/DB (0x0)
dic 06 15:01:35 debian kernel: amdgpu 0000:06:00.0: amdgpu:          MORE_FAULTS: 0x0
dic 06 15:01:35 debian kernel: amdgpu 0000:06:00.0: amdgpu:          WALKER_ERROR: 0x0
dic 06 15:01:35 debian kernel: amdgpu 0000:06:00.0: amdgpu:          PERMISSION_FAULTS: 0x0
dic 06 15:01:35 debian kernel: amdgpu 0000:06:00.0: amdgpu:          MAPPING_ERROR: 0x0
dic 06 15:01:35 debian kernel: amdgpu 0000:06:00.0: amdgpu:          RW: 0x0

that ends with [drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring gfx_0.0.0 timeout, but soft recovered

Attached file has more information.

Screenshots/video files

gpu_error.log

bug 6000 dGPU Series

changed the description

added 6000 dGPU Series label

changed the description

Same error occurs on Manjaro XWayland and kwin_wayland [gfxhub] page fault loop, ending with the same amdgpu_job_timedout error... Followed by an opengl error... look at the attached log

amd_error.log

Once more completelly random crash. This time, looking at JDowloader flatpak version (previous crashes happened using wayland native or xwayland using programs). Speeds measurement charts looked as corrupted graphics with random colored big square blocks (there is not any capture because it was impossible to take one).

Linux manjaro 6.6.5-1-MANJARO #1 (closed) SMP PREEMPT_DYNAMIC Fri Dec 8 12:53:00 UTC 2023 x86_64 GNU/Linux

Same error keeps happening...

Here is the error log: amd-error-otromas.log

If there is any way in wich I can help say it to me, please, I use my computer almost 24/7 and it is a shit being saving files each 30 seconds because I do not know when the next graphics freeze will be.

It is only a graphics freeze+corruption (random colored square blocks appear in some areas). If it happens when using a terminal I am able (whithout seeying anything) to kill some processes (killall program_name) and restart the computer (shutdown -r now).

¡Hi!

I have fresh news. First, I do not know this issue could be related to #2943

The news are that today another crash happened (in Manjaro), but I had nothing better to do than wait and see if something happens. If the log grew as much as to fill all the root filesystem I could log in in Debian and remove it, so there was not any problem. But... 10 minutes after the freeze the session closed and I was sent to SDDM to start a new one.

Here is the systemctl log, just before the session was closed (around 11:24:34) there are some stack traces of failing processes, and before that I guess that there are the common error loops.

amd-error-otromas-reinicio10min.log