Games won't launch after opening DT 5.4

[xplatinum]

Is there an existing issue for this?

• I checked and did not find my issue in the already reported ones

Describe the bug

Since installing 5.4 DT has prevented me from launching games if it was previously launched during the active session.

Steps to reproduce

1. Launch and close Darktable
2. Launch a game (e.g., Arc Raiders, Path of Exile 2)
3. Game crashes (vlc::Queue:submit: ErrorDeviceLost)
4. Reboot
5. Launch the game again
6. Game works fine
7. Open and close Darktable
8. Launch the game again
9. Game crashes (vlc::Queue:submit: ErrorDeviceLost)

The issue only happens after DT has been launched. It's like it keeps a hold of something in the backend.

Expected behavior

Darktable shouldn't make games crash following it's launch

Logfile | Screenshot | Screencast

Dec 31 14:18:42 arch kernel: amdgpu 0000:03:00.0: amdgpu: ring comp_1.1.0 timeout, signaled seq=7816, emitted seq=7818
Dec 31 14:18:42 arch kernel: amdgpu 0000:03:00.0: amdgpu:  Process Main pid 7828 thread Main pid 7828
Dec 31 14:18:42 arch kernel: amdgpu 0000:03:00.0: amdgpu: Starting comp_1.1.0 ring reset
Dec 31 14:18:42 arch kernel: amdgpu 0000:03:00.0: amdgpu: [gfxhub] page fault (src_id:0 ring:32 vmid:0 pasid:0)
Dec 31 14:18:42 arch kernel: amdgpu 0000:03:00.0: amdgpu:   in page starting at address 0x0000000000000000 from client 10
Dec 31 14:18:42 arch kernel: amdgpu 0000:03:00.0: amdgpu: GCVM_L2_PROTECTION_FAULT_STATUS:0x00040A40
Dec 31 14:18:42 arch kernel: amdgpu 0000:03:00.0: amdgpu:          Faulty UTCL2 client ID: CPC (0x5)
Dec 31 14:18:42 arch kernel: amdgpu 0000:03:00.0: amdgpu:          MORE_FAULTS: 0x0
Dec 31 14:18:42 arch kernel: amdgpu 0000:03:00.0: amdgpu:          WALKER_ERROR: 0x0
Dec 31 14:18:42 arch kernel: amdgpu 0000:03:00.0: amdgpu:          PERMISSION_FAULTS: 0x4
Dec 31 14:18:42 arch kernel: amdgpu 0000:03:00.0: amdgpu:          MAPPING_ERROR: 0x0
Dec 31 14:18:42 arch kernel: amdgpu 0000:03:00.0: amdgpu:          RW: 0x1
Dec 31 14:18:42 arch kernel: amdgpu 0000:03:00.0: amdgpu: Ring comp_1.1.0 reset succeeded

Commit

No response

Where did you obtain darktable from?

darktable.org / GitHub release

darktable version

2:5.4.0-1

What OS are you using?

Linux

What is the version of your OS?

Arch

Describe your system

kernel Linux 6.18.2-arch2-1
de KDE Plasma 6.5.4
vulkan 1.4.328 - radv [Mesa 25.3.2-arch1.1]

Operating System: Arch Linux
KDE Plasma Version: 6.5.4
KDE Frameworks Version: 6.21.0
Qt Version: 6.10.1
Kernel Version: 6.18.2-arch2-1 (64-bit)
Graphics Platform: Wayland
Processors: 20 × 12th Gen Intel® Core™ i7-12700K
Memory: 32 GiB of RAM (31.1 GiB usable)
Graphics Processor: AMD Radeon RX 9070 XT
Manufacturer: ASUS

Are you using OpenCL GPU in darktable?

Yes

If yes, what is the GPU card and driver?

AMD Radeon RX 9070 XT - vulkan 1.4.328 - radv [Mesa 25.3.2-arch1.1]

Please provide additional context if applicable. You can attach files too, but might need to rename to .txt or .zip

No response

[ralfbrown]

Do the crashes happen with OpenCL disabled? (run darktable with the --disable-opencl flag[*]) The error code from VLC would point in that direction.

[*] You can also disable OpenCL in darktable's preferences, but if you use that option, you'll need to reboot before testing, since darktable will have started with OpenCL and thus probably trigger the crash.

[xplatinum]

Do the crashes happen with OpenCL disabled? (run darktable with the --disable-opencl flag[*]) The error code from VLC would point in that direction.

[*] You can also disable OpenCL in darktable's preferences, but if you use that option, you'll need to reboot before testing, since darktable will have started with OpenCL and thus probably trigger the crash.

No issue when running DT with OpenCL turned off. I found that the issue is also present in 5.2. Older ROCm version exhibited the same problem. So it's probably not that either.

It might not be something with DT afterall. I opened the issue on amdgpu too.

[gi-man]

Can you provide a log from dt using -d perf and -d opencl after processing one image in the darkroom? Also, when you close dt, can you look to see if there is dt process running?

[xplatinum]

Can you provide a log from dt using -d perf and -d opencl after processing one image in the darkroom? Also, when you close dt, can you look to see if there is dt process running?

Here's the log. There are no more processes once the app is closed.

darktableDebug.txt

[gi-man]

I was looking for opencl errors and there are none.

[TurboGit]

@jenshannoschwalm : This is puzzling, you may want to have a look at it.

[jenshannoschwalm]

When using #20059

@xplatinum would you be able to self-compile and check?
@MStraeten @zisoft obviously this must be tested on apple harware too :-)

[zisoft]

Well, it compiles and runs fine on my Intel Mac :)

How can I test this?

[jenshannoschwalm]

Basically it's just using dt with opencl enabled and -d opencl and watch for error messages when closing :-) There should be none .

[xplatinum]

Tested with linux-lts (6.12.x): issue does not reproduce.

This appears to be a kernel AMDGPU regression rather than a darktable-specific issue. Disabling OpenCL or using linux-lts avoids the problem.

[jenshannoschwalm]

So another of those crazy ROCm problems .

[gi-man]

So another of those crazy ROCm problems .

I still like the idea of the PR. It clears the CL at the end of the pipe execution or when closing dt?