GPU passthrough - /dev/dri no such file or directory

Got the above error in a Cli made docker container after updating to latest Electric Eel today. I assume this has moved or similar and is easy to understand. I tried logging into the Jira system to report a bug but it seems to be having a problem today where I can’t log in. So reporting here in the mean time in case anyone else comes across it.

Thanks, please provide the system/hardware details.

Include the Apps and history.

I would normally just upload debug logs, but since I can’t open a jira ticket…

It’s a threadripper 2950x, 128GB RAM, Nvidia p2000, this is custom docker app plex from linuxserver. I assume that’s all you need for that.

Was working find mapping /dev/dri and then after update it isn’t. listing the dev directory there is no longer a dri in there.

As a bit of further troubleshooting, I unset the Nvidia drivers checkbox in the application settings, saved and exited, then reset it. This unfortunately caused the app service to be stopped and not be possible to start it, including from using service docker start. A reboot however did manage to restart the apps, but the /dev/dri device is still missing.

So are you saying that the /dev/dri directory is missing or the Nvidia device you expect to see inside it? The directory should be there as part of the base install, regardless of Nvidia driver status. This is from a 24.10.1 system without an installed gpu device:

% ls /dev/
autofs           dri        input         loop5         ppp     rtc0  sde2      stderr  tty15  tty25  tty35  tty45  tty55  tty8         vcs1    vcsa3   vcsu6
block            fb0        ipmi0         loop6         pps0    sda   sde3      stdin   tty16  tty26  tty36  tty46  tty56  tty9         vcs10   vcsa4   vfio
bsg              fd         kmsg          loop7         pps1    sda1  sg0       stdout  tty17  tty27  tty37  tty47  tty57  ttyS0        vcs2    vcsa5   vga_arbiter
btrfs-control    full       kvm           mapper        psaux   sdb   sg1       tty     tty18  tty28  tty38  tty48  tty58  ttyS1        vcs3    vcsa6   vhost-net
bus              fuse       log           mem           ptmx    sdb1  sg2       tty0    tty19  tty29  tty39  tty49  tty59  ttyS2        vcs4    vcsu    vhost-vsock
char             hidraw0    loop-control  mqueue        ptp0    sdc   sg3       tty1    tty2   tty3   tty4   tty5   tty6   ttyS3        vcs5    vcsu1   watchdog
console          hidraw1    loop0         net           ptp1    sdc1  sg4       tty10   tty20  tty30  tty40  tty50  tty60  uhid         vcs6    vcsu10  watchdog0
core             hpet       loop1         null          pts     sdd   sg5       tty11   tty21  tty31  tty41  tty51  tty61  uinput       vcsa    vcsu2   xconsole
cpu_dma_latency  hugepages  loop2         nvme-fabrics  random  sdd1  shm       tty12   tty22  tty32  tty42  tty52  tty62  urandom      vcsa1   vcsu3   zero
cuse             hwrng      loop3         nvram         rfkill  sde   snapshot  tty13   tty23  tty33  tty43  tty53  tty63  userfaultfd  vcsa10  vcsu4   zfs
disk             initctl    loop4         port          rtc     sde1  snd       tty14   tty24  tty34  tty44  tty54  tty7   vcs          vcsa2   vcsu5
% ls /dev/dri
by-path  card0

I’m also not seeing any issues with our Jira instance at the moment, so you should be able to log in and submit a bug ticket.

Im saying the whole folder is missing. It was there on the previous version before the upgrade, now its not. When I’m back at a computer I’ll upload a listing of the /dev directory. The jira login was just because I cleared all my browser cache etc. when logging in it just comes back to the login page over and over. Tried two browsers safari and Firefox.

Interestingly i’m also missing the dri directory but my nvidia gpu works just fine in jellyfin… wasn’t /dri only used as ram disk for transcoding instead of a permanent folder?

I don’t know, but plex stopped working when it disappeared. As per the instructions here for the linuxserver plex container, /dev/dri is required for hardware acceleration.