Hardware Transcoding with Intel UHD Graphics 605 on Proxmox Privileged LXC

[designingbeing]

Jellyfin is giving the error "Playback failed due to a fatal player error" whenever I enable hardware transcoding.

I've followed the Jellyfin tutorial for Proxmox LXCs as best as I can and found a lot of threads on the subject but I'm not having success:

FFmpeg logs show 

Code:

No VA display found for any default device. 
Device creation failed: -22

The only thing that stands out to me is that

Code:

getent group render

gives different values on LXC=106, Proxmox Host=104. I'm not sure if this is even an issue, or what to do about it.


[*]Jellyfin 10.9.9
[*]Debian 12 running as a Privileged LXC container on Proxmox 8.2.4
[*]Installed via curl https://repo.jellyfin.org/install-debuntu.sh | sudo bash
[*]J5005 processor which iirc is Gemini Lake, based on Kaby Lake with Intel UHD Graphics 605

---

Output from commands run on LXC Container:
vainfo
  vainfo - LXC.txt (Size: 1.32 KB / Downloads: 68)
getent group video, getent group render
  getent group video render - LXC.txt (Size: 89 bytes / Downloads: 52)
groups jellyfin
  groups jellyfin - LXC.txt (Size: 51 bytes / Downloads: 71)
ls -la /dev/dri
  ls -la devdri - LXC.txt (Size: 178 bytes / Downloads: 73)
usr/lib/jellyfin-ffmpeg/ffmpeg -v verbose -init_hw_device vaapi=vadev/dri/renderD128 -init_hw_device opencl@va
  usrlibjellyfin-ffmpegffmpeg - LXC .txt (Size: 2.33 KB / Downloads: 87)

---

Output from commands run on Proxmox Host:
getent group video, getent group render
  getent group video render - PROXMOX.txt (Size: 73 bytes / Downloads: 52)
ls -la /dev/dri 
  ls -la devdri - PROXMOX.txt (Size: 299 bytes / Downloads: 64)

---

LXC Configuration file:

  LXC.conf.txt (Size: 674 bytes / Downloads: 83)

---

FFmpeg logs:

  FFmpeg.Transcode-2024-08-24_22-34-40_ebf5e9fffb88af435b038fe26661932d_6a2c85e4.log.txt (Size: 12.61 KB / Downloads: 71)
  FFmpeg.Transcode-2024-08-24_22-34-42_ebf5e9fffb88af435b038fe26661932d_b511f466.log.txt (Size: 12.58 KB / Downloads: 77)
  FFmpeg.Transcode-2024-08-24_22-34-43_ebf5e9fffb88af435b038fe26661932d_60cec405.log.txt (Size: 12.58 KB / Downloads: 88)
  FFmpeg.Transcode-2024-08-24_22-34-43_ebf5e9fffb88af435b038fe26661932d_74996ef6.log.txt (Size: 12.61 KB / Downloads: 82)
  FFmpeg.Transcode-2024-08-24_22-34-46_ebf5e9fffb88af435b038fe26661932d_32634eca.log.txt (Size: 12.58 KB / Downloads: 67)
  FFmpeg.Transcode-2024-08-24_22-34-47_ebf5e9fffb88af435b038fe26661932d_3e01173c.log.txt (Size: 12.58 KB / Downloads: 61)
  FFmpeg.Transcode-2024-08-24_22-34-47_ebf5e9fffb88af435b038fe26661932d_ff05bcb8.log.txt (Size: 12.61 KB / Downloads: 66)

---

Jellyfin log 1/2 https://pastebin.com/bnq8306r
Jellyfin log 2/2 https://pastebin.com/H6N0am7u

[TheDreadPirate]

Did you pass the GPU into the LXC?

https://jellyfin.org/docs/general/admini...on-proxmox

[Fate]

My mounts look a bit different in lxc conf:

Code:

lxc.cgroup2.devices.allow: c 226:0 rwm
lxc.cgroup2.devices.allow: c 226:128 rwm
lxc.cgroup2.devices.allow: c 29:0 rwm
lxc.mount.entry: /dev/fb0 dev/fb0 none bind,optional,create=file
lxc.mount.entry: /dev/dri dev/dri none bind,optional,create=dir
lxc.mount.entry: /dev/dri/renderD128 dev/dri/renderD128 none bind,optional,create=file

Was your container always privileged? I think if you have changed from unprivileged to privileged you have to reset the user/group ids. (probably easier to reinstall the container)

I would recommend tteck scripts to create your container if you are unsure.

Which intel drivers did you install on the proxmox and on the lxc side?

[designingbeing]

Thanks for the reply

Did you pass the GPU into the LXC?

https://jellyfin.org/docs/general/admini...on-proxmox

Yeah, I'm pretty sure I did at least. I added the following lines to my LXC.conf:

Code:

lxc.cgroup2.devices.allow: c 226:0 rwm
lxc.cgroup2.devices.allow: c 226:128 rwm
lxc.hook.pre-start: sh -c "chown 0:104 /dev/dri/renderD128"

which contains variables (226:0, 226:128 ) based on the output of "ls -la /dev/dri" run on the proxmox host

Code:

total 0
drwxr-xr-x  3 root root        100 Aug 25 04:32 .
drwxr-xr-x 20 root root      4780 Aug 25 04:50 ..
drwxr-xr-x  2 root root        80 Aug 25 04:32 by-path
crw-rw----  1 root video  226,  0 Aug 25 04:32 card0
crw-rw----  1 root render 226, 128 Aug 25 04:32 renderD128

I've tried changing the 104 in "chown 0:104 /dev/dri/renderD128" to 106 as well, to cover the render getent group in both Proxmox host as well as LXC guest.

[Fate]

Thanks for the reply

Did you pass the GPU into the LXC?

https://jellyfin.org/docs/general/admini...on-proxmox

Yeah, I'm pretty sure I did at least. I added the following lines to my LXC.conf:

Code:

lxc.cgroup2.devices.allow: c 226:0 rwm
lxc.cgroup2.devices.allow: c 226:128 rwm
lxc.hook.pre-start: sh -c "chown 0:104 /dev/dri/renderD128"

which contains variables (226:0, 226:128 ) based on the output of "ls -la /dev/dri" run on the proxmox host

Code:

total 0
drwxr-xr-x  3 root root        100 Aug 25 04:32 .
drwxr-xr-x 20 root root      4780 Aug 25 04:50 ..
drwxr-xr-x  2 root root        80 Aug 25 04:32 by-path
crw-rw----  1 root video  226,  0 Aug 25 04:32 card0
crw-rw----  1 root render 226, 128 Aug 25 04:32 renderD128

I've tried changing the 104 in "chown 0:104 /dev/dri/renderD128" to 106 as well, to cover the render getent group in both Proxmox host as well as LXC guest.

On my proxmox:

Code:

root@pvx:/dev/dri# ls -lh
total 0
drwxr-xr-x 2 root root        80 Jun  5 00:03 by-path
crw-rw---- 1 root video  226,  0 Jun  5 00:03 card0
crw-rw---- 1 root render 226, 128 Jun  5 00:03 renderD128

on my lxc:

Code:

[root@kuroko dri]# ls -lh
total 0
drwxr-xr-x 2 root root        80 Jun  5 00:03 by-path
crw-rw---- 1 root video  226,  0 Jun  5 00:03 card0
crw-rw---- 1 root render 226, 128 Jun  5 00:03 renderD128

Looking at your log earlier it seems to mismatch your proxmox side, so I'm pretty sure this is your issue.

Could you spawn a test container with tteck scrit and compare how the script mounts your devices?

Just to make sure can oyu post
dmesg | grep i915 on proxmox side?

[designingbeing]

Thanks for the reply

My mounts look a bit different in lxc conf:

Code:

lxc.cgroup2.devices.allow: c 226:0 rwm
lxc.cgroup2.devices.allow: c 226:128 rwm
lxc.cgroup2.devices.allow: c 29:0 rwm
lxc.mount.entry: /dev/fb0 dev/fb0 none bind,optional,create=file
lxc.mount.entry: /dev/dri dev/dri none bind,optional,create=dir
lxc.mount.entry: /dev/dri/renderD128 dev/dri/renderD128 none bind,optional,create=file

Was your container always privileged? I think if you have changed from unprivileged to privileged you have to reset the user/group ids. (probably easier to reinstall the container)

I would recommend tteck scripts to create your container if you are unsure.

Which intel drivers did you install on the proxmox and on the lxc side?

- My container was always privileged

- I think the drivers might have been the issue.

On Proxmox host I could see:

Code:

i965-va-driver-shaders/stable,now 2.4.1-1 amd64 [installed]
intel-media-va-driver-non-free/stable,now 23.1.1+ds1-1 amd64 [installed]

On the LXC guest I could see:

Code:

i965-va-driver/stable,now 2.4.1+dfsg1-1 amd64 [installed,automatic]
intel-media-va-driver/stable,now 23.1.1+dfsg1-1 amd64 [installed,automatic]

On the LXC guest I then ran:

Code:

apt install software-properties-common -y
add-apt-repository -y non-free
apt install intel-media-va-driver-non-free -y

and on the proxmox host, updated my LXC.conf with the value from "getent group render | cut -d":" -f3" as run on the LXC guest

Code:

lxc.hook.pre-start: sh -c "chown 0:{LXC getent output} /dev/dri/renderD128"

then rebooted the LXC.

After this I forced a transcode and for the first time saw activity in "intel_gpu_top" as run on the Proxmox host


I'm cautiously optimistic that this has resolved the issue. Thank you for your input.

[Fate]

Awesome! Glad I could help!