Re: [PATCH] userns: Support fuse interacting with multiple user namespaces

[ebiederm-aS9lmoZGLiVWk0Htik3J/w@public.gmane.org (Eric W. Biederman)]

Miklos Szeredi <miklos-sUDqSbJrdHQHWmgEVkV9KA@public.gmane.org> writes:

> ebiederm-aS9lmoZGLiVWk0Htik3J/w@public.gmane.org (Eric W. Biederman) writes:
>
>> Use kuid_t and kgid_t in struct fuse_conn and struct fuse_mount_data.
>>
>> The connection between between a fuse filesystem and a fuse daemon is
>> established when a fuse filesystem is mounted and provided with a file
>> descriptor the fuse daemon created by opening /dev/fuse.
>>
>> For now restrict the communication of uids and gids between the fuse
>> filesystem and the fuse daemon to the initial user namespace.
>
> Why?
>
> I think far more logical would be to limit a single instance of the
> filesystem and the daemon to an arbitrary but *single* namespace.
> I.e. one fuse_conn <-> one user namespace.
>
> Is there a reason to treat the initial namespace specially?

Stepwise making this work.

The initial user namespace is special in that is the only user namespace
you can currently mount a filesystem in.

My goal with the posted patch is to support users on the other side of
the vfs that are in multiple user namespaces.  Which basically means
accepting kuid_t and kgid_t values and converting them to uid_t and
gid_t values where needed.

As for creating a version of fuse that can be mounted in different user
namespaces and can communicate with the fuse daemon in the user
namespace the filesystem was mounted in, I have a preliminary patch for
that.  My goal with that patch is to support unprivileged mounts of fuse
without the help of a suid root mount program.  But that patch is more
involved and not quite ready yet.

I want to get to the point that I can enable fuse and user namespaces at
the same time (what this patch allows) and then put in the effort to
fully take advantage of user namespaces.

Eric