Repository navigation
[Bug][v0.11.2][WSL2] Licensed LTX worker rejects WSL NVIDIA CUDA driver; Windows version works #556
Description
Describe the bug
The Windows version of Jasna v0.11.2 successfully completes actual LTX restoration on this PC and GPU.
Under Ubuntu 24.04 on WSL2, license activation and LTX Trial mode work. However, actual licensed LTX restoration fails after detection and planning, when the protected worker starts.
The rejected module appears to be the standard NVIDIA CUDA driver provided through WSL.
Relevant log (local username and video filename anonymized):
INFO Using VALI NVDEC decoder for /home/user/sample.mkv
INFO LTX plan: 114387 frames, 64 scene cuts, 373 tracks, 1016 windows
ERROR Failed to process sample.mkv: LTX worker could not start: ProtectionError: LTX executable mapping policy was not enforced: another program loaded this module into the worker (often antivirus, overlay or preload software): /usr/lib/wsl/drivers/nv_dispig.inf_amd64_cc526ccc3287f5cd/libcuda.so.1.1 (core.py:177 <- core.py:138 <- ltx_worker_security.py:96)
To Reproduce
- Launch the official Linux NVIDIA build from the Ubuntu terminal using
./jasna. - Activate the LTX license.
- Select LTX, Standard model, and turn Trial mode OFF.
- Add a local video and start processing.
- Detection and planning complete, then the protected LTX worker fails to start.
Jasna and the input video are stored inside the Linux filesystem under /home/user/, not on a Windows-mounted or network drive.
Desktop
- Host OS: Windows 11 Pro, build 26200
- Linux: Ubuntu 24.04 on WSL2
- WSL kernel: 6.18.54.1-microsoft-standard-WSL2
- Jasna: v0.11.2, official Linux NVIDIA release
- Interface: GUI
- GPU: NVIDIA GeForce RTX 4070 Ti, 12 GB VRAM
- NVIDIA Windows driver: 610.88
- CPU: AMD Ryzen 5 5600X
- System RAM: 32 GB; WSL memory limit approximately 16 GB
- TVAI: not used
Additional information
No LD_PRELOAD or LD_LIBRARY_PATH was found in the running Jasna process environment. No kernel OOM-killer record was found.
Is licensed LTX restoration supported under WSL2? Is there a supported configuration change or an upcoming fix for this case?
I can provide additional diagnostic information if needed.
Activity
gensukeman-del commentedon Oct 19, 2026
Additional test results:
This does not appear to be limited to long videos.
berlingots commentedon Oct 19, 2026
On version Ubuntu 26.04.1 LTS (Resolute Raccoon) experiencing the same issue getting LTX Standard to work.
INFO:jasna.ltx.sampler:LTX quantized linears: Triton kernels
ERROR:jasna.protection.ltx_worker:LTX inference worker failed
torch.AcceleratorError: CUDA error: unknown error
Search for
cudaErrorUnknown' in https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__TYPES.html for more information. CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1 Compile withTORCH_USE_CUDA_DSA` to enable device-side assertions.Ran Jasna as sudo and made sure permissions was not an issue. Everything else works fine including Windows. Also made sure PyTorch and CUDA were all up to date.
Using latest version of Jasna v0.11.2
Nvidia 4090, 24GB
AMD Ryzen 9 9950X
Sytem RAM: 64 GB
gensukeman-del commentedon Oct 19, 2026
@berlingots Thanks for sharing your report! You mentioned Ubuntu 26.04.1 — are you running it under WSL2, or booting Ubuntu directly on your PC?
In my case, LTX trial mode and actual BasicVSR++ processing work under WSL2, but licensed LTX fails at worker startup with a protection-policy error involving the WSL CUDA driver (
libcuda.so.1.1). Windows LTX works on the same PC.Your CUDA “unknown error” looks different from mine, so I’m not sure whether the two failures share a cause. Knowing whether you’re using WSL2 would help clarify that. Thanks!