Re: [GIT PULL] userns related vfs enhancements for v4.8

[Eric W. Biederman]

MichaÅ Zegan <webczat_200@xxxxxxxxxxxxxx> writes:

> I have one question: what uids are written in the filesystem inodes?
> those that the kernel sees, or those mapped by user namespaces? and so,
> is it true that the filesystem will still require shifting uids,
> directly or indirectly, to be usable from inside user namespace? can't
> uids be mapped dynamically from what is in the inode to kernel uids
> based on the user namespace they are used from? Maybe because the fs can
> be mounted from outside before starting the container?

Feel free to read the code.  As best I can understand your question the
best answer is in reading and understanding the code yourself.

Eric