criu
criu copied to clipboard
Direction for WAR-ing seccomp restrictions in containers
I have the following usecase:
- Run criu inside a container whose manager (Docker, LXC, etc) apply seccomp filter for the root process
- Seccomp cannot be removed. It's part of the setup.
I think that a direction for this is to allow users to ignore filters at their own demise.
For example: criu ... --ignore-seccomp ... The argument would need to be provided on save and restore.
Problems:
- Application may rely on seccomp rules for correctness.
- Removing seccomp filter on restore can cause security issues.
Benefits:
- criu works in default docker and lxc configurations
I'm looking for the right architectural solution to support my usecase. I'm happy to get my hands dirty once direction is agreed.
The idea sounds good. My main question is, especially about containers created by Podman and Docker. To use CRIU in one of those containers I am used to start it with --privileged
, because it does not work otherwise. Containers started with --privileged
are running without seccomp if I remember it correctly.
Are you using CRIU in containers without --privileged
and the only problem you are seeing is seccomp related?
In particular, I don't know how the container is started since I do not own the infra. My guess is that it's started with strict seccomp applied for security.
Thanks for the direction, patches are on me.
Are you sure that seccomp is the only thing that blocks you? I would expect there are more problems.
A friendly reminder that this issue had no activity for 30 days.