How do I conserve the userns UID/GID mappings when archiving an LXC guest?

Question

I'd like to archive an existing LXC container that has been configured to run as unprivileged LXC container (see this question).

How can I conserve all the file system meta-data that is used to store the mapped UID and GID for file/folder ownership?

NB: I know that the mapping itself happens on the host, but what I mean is that inside the userns there are a number of UIDs and GIDs which all map on the host to an unprivileged user, but which in the guest still resolve to different UIDs and GIDs. So whatever that magic sauce is that keeps these things connected at the file system level, I'd like to conserve it in an archive (tar or 7z or similar).

Answer 1

I have investigated that topic since I asked my question and it turns out that the ranges of sub-GIDs/sub-UIDs are indeed used as file-group/-owner for the files. There is no additional metadata (unless you'd be using SELinux or the like) which is relevant to this.

Also keep in mind that ACLs may have to be modified, if you make use of ACLs.