Re: [PATCH RFC] user-namespaced file capabilities - now with even more magic

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

"Serge E. Hallyn" <serge@hallyn.com> writes:

> On Thu, Dec 08, 2016 at 05:43:09PM +1300, Eric W. Biederman wrote:
>> "Serge E. Hallyn" <serge@hallyn.com> writes:
>> 
>> > Root in a user ns cannot be trusted to write a traditional
>> > security.capability xattr.  If it were allowed to do so, then any
>> > unprivileged user on the host could map his own uid to root in a
>> > namespace, write the xattr, and execute the file with privilege on the
>> > host.
>> >
>> > This patch introduces v3 of the security.capability xattr.  It builds a
>> > vfs_ns_cap_data struct by appending a uid_t rootid to struct
>> > vfs_cap_data.  This is the absolute uid_t (i.e. the uid_t in
>> > init_user_ns) of the root id (uid 0 in a namespace) in whose namespaces
>> > the file capabilities may take effect.
>> >
>> > When a task in a user ns (which is privileged with CAP_SETFCAP toward
>> > that user_ns) asks to write v2 security.capability, the kernel will
>> > transparently rewrite the xattr as a v3 with the appropriate rootid.
>> > Subsequently, any task executing the file which has the noted kuid as
>> > its root uid, or which is in a descendent user_ns of such a user_ns,
>> > will run the file with capabilities.
>> >
>> > If a task writes a v3 security.capability, then it can provide a
>> > uid (valid within its own user namespace, over which it has CAP_SETFCAP)
>> > for the xattr.  The kernel will translate that to the absolute uid, and
>> > write that to disk.  After this, a task in the writer's namespace will
>> > not be able to use those capabilities, but a task in a namespace where
>> > the given uid is root will.
>> >
>> > Only a single security.capability xattr may be written.  A task may
>> > overwrite the existing one so long as it was written by a user mapped
>> > into his own user_ns over which he has CAP_SETFCAP.
>> >
>> > This allows a simple setxattr to work, allows tar/untar to work, and
>> > allows us to tar in one namespace and untar in another while preserving
>> > the capability, without risking leaking privilege into a parent
>> > namespace.
>> 
>> Any chance of a singed-off-by?
>
> Yes, sorry, Stéphane had pointed out that I'd apparently forgotten to do
> -s.  Do you want me to resend the whole shebang, or does
>
> Signed-off-by: Serge Hallyn <serge@hallyn.com>
>
> suffice?  (My previous iterations did have it fwiw so I don't think I could
> legally disavow it now :)

That should be good enough.  I just wanted to make certain it existed
somewhere.

The whole inode->i_op->getxattr reference was also a bit of a problem
as that method was removed in 4.9-rc1 but otherwise things are looking
reasonable.

Thank you,
Eric