SOLVED: HW transcoding not working... Cannot load libcuda.so.1

[GuigZer]

Jellyfin NVIDIA Hardware Transcoding Issue on Debian with Quadro P620

Environment

- **OS**: Debian GNU/Linux 12 (bookworm)    
- **Kernel**: 6.1.0-28-amd64
- **GPU**: NVIDIA Quadro P620
- **Driver Version**: 535.183.01
- **Container**: Jellyfin (lscr.io/linuxserver/jellyfin:latest)

Issue Description

I recently installed a Quadro P620 GPU for hardware transcoding in Jellyfin. While the GPU is properly detected both on the host system and inside the Docker container, FFmpeg fails to initialize CUDA for transcoding with an error related to loading the CUDA library.

Attempted Solutions

I have already tried several solutions suggested in these forum posts:
- Verifying NVIDIA runtime is properly installed and configured in Docker
- Adding all necessary GPU capabilities in docker-compose file
- Setting NVIDIA_DRIVER_CAPABILITIES and NVIDIA_VISIBLE_DEVICES to "all"
- Ensuring proper permissions for the Jellyfin user in the container
- Checking if CUDA libraries are present in the container
- Verifying GPU detection using nvidia-smi both on host and in container

None of these common solutions have resolved the issue.

System Configuration

Host System Information

bash
$ uname -a
Linux openmediavault 6.1.0-28-amd64 #1 SMP PREEMPT_DYNAMIC Debian 6.1.119-1 (2024-11-22) x86_64 GNU/Linux

$ cat /etc/os-release
PRETTY_NAME="Debian GNU/Linux 12 (bookworm)"
NAME="Debian GNU/Linux"
VERSION_ID="12"
VERSION="12 (bookworm)"


GPU Detection

bash
$ nvidia-smi
Mon Nov 25 22:00:28 2024
+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 535.183.01            Driver Version: 535.183.01  CUDA Version: N/A      |
|-----------------------------------------+----------------------+----------------------+
| GPU  Name                Persistence-M | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf          Pwr:Usage/Cap |        Memory-Usage | GPU-Util  Compute M. |
|                                        |                      |              MIG M. |
|=========================================+======================+======================|
|  0  Quadro P620                    Off | 00000000:02:00.0 Off |                  N/A |
| 28%  43C    P0              N/A /  N/A |      0MiB /  2048MiB |      0%      Default |
|                                        |                      |                  N/A |
+-----------------------------------------+----------------------+----------------------+


Docker Compose Configuration
yaml
services:
  jellyfin:
    image: lscr.io/linuxserver/jellyfin:latest
    container_name: jellyfin
    environment:
      - PUID=1000
      - PGID=1000
      - TZ=Etc/UTC
      - JELLYFIN_PublishedServerUrl=http://192.168.1.100
      - NVIDIA_DRIVER_CAPABILITIES=all
      - NVIDIA_VISIBLE_DEVICES=all
    volumes:
      - ./config:/config
      - /srv/dev-disk-by-uuid-504655e3-3559-44c8-b61b-9eaed009f4a3/plex/serie:/data/tvshows
      - /srv/dev-disk-by-uuid-504655e3-3559-44c8-b61b-9eaed009f4a3/plex/film:/data/movies
      - ./transcode:/config/transcodes
    ports:
      - 8096:8096
      - 8920:8920
      - 7359:7359/udp
      - 1900:1900/udp
    restart: unless-stopped
    runtime: nvidia
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: ["gpu","compute","video","utility"]


Error Message
When attempting to play a video file, FFmpeg logs show:

[AVHWDeviceContext @ 0x55ab2d7c1d00] Cannot load libcuda.so.1
[AVHWDeviceContext @ 0x55ab2d7c1d00] Could not dynamically load CUDA
Device creation failed: -1.
Failed to set value 'cuda=cu:0' for option 'init_hw_device': Operation not permitted
Error parsing global options: Operation not permitted


Additional Information

- The GPU is properly detected inside the container (verified using nvidia-smi)
- NVIDIA runtime is properly configured in Docker
- FFmpeg is built with NVIDIA support (as shown in the configuration output)


Any assistance would be greatly appreciated.