Re: [PATCH -mm 5/7] add user namespace

[ebiederm@xmission.com (Eric W. Biederman)]

"Serge E. Hallyn" <serue@us.ibm.com> writes:

> Quoting Eric W. Biederman (ebiederm@xmission.com):
>> "Serge E. Hallyn" <serue@us.ibm.com> writes:
>> >> No.  The uids in a filesystem are interpreted in some user namespace
>> >> context.  We can discover that context at the first mount of the
>> >> filesystem.  Assuming the uids on a filesystem are the same set
>> >> of uids your process is using is just wrong.
>> >
>> > But, when I insert a usb keychain disk into my laptop, that fs assumes
>> > the uids on it's fs are the same as uids on my laptop...
>> 
>> I agree that setting the fs_user_namespace at mount time is fine.
>> However when we use a mount that a process in another user namespace
>> we need to not assume the uids are the same.
>> 
>> Do you see the difference?
>
> Aaah - so you don't want to store this on the fs.  So this is actually
> like what I had mentioned many many emails ago?

Quite possibly.  I'm not certain where you go the idea I was thinking
of storing this on the fs.  I think you must have been thinking of
the Linux-Vserver implementation.

>> Actually I was thinking something as mundane as a mapping table.  This
>> uid in this namespace equals that uid in that other namespace.
>
> I see.
>
> That's also what I was imagining earlier, but it seems crass somehow.
> I'd almost prefer to just tag a mount with a user namespace implicitly,
> and only allow the mounter to say 'do' or 'don't' allow this to be read
> by users in another namespace.  Then in the 'don't' case, user joe
> [1000] can't read files belonging to user jack [1000] in another
> namespace.  It's stricter, but clean.
>
> But whether we do mapping tables or simple isolation, I do still like
> the idea of pursuing the use of the keystore for global uids.

Yes.  I guess my thinking is that the mapping effort and keys are an
enhancement after we get the basic user namespace working, to overcome
the limitations.

Eric