So is there a workaround? In my case, I'm trying to access an OpenCL gpu
from a userland container. I was assuming that the below might be
enough.
lxc.mount.entry = /dev/dri/card1 dev/dri/card1 none bind,optional,create=file
lxc.mount.entry = /dev/dri/renderD128 dev/dri/renderD128 none
"lxc.cgroup.devices" is meaningless for unprivileged containers as those
can never create those devices anyway, so they'll only ever have access
to whatever devices lxc provides and nothing more. All our own default
configs specifically do not set that cgroup controller for unprivileged
2 matches
Mail list logo