mptcp: netlink: respect v4/v6-only sockets

commit fb00ee4f upstream.

If an MPTCP socket has been created with AF_INET6 and the IPV6_V6ONLY
option has been set, the userspace PM would allow creating subflows
using IPv4 addresses, e.g. mapped in v6.

The kernel side of userspace PM will also accept creating subflows with
local and remote addresses having different families. Depending on the
subflow socket's family, different behaviours are expected:
 - If AF_INET is forced with a v6 address, the kernel will take the last
   byte of the IP and try to connect to that: a new subflow is created
   but to a non expected address.
 - If AF_INET6 is forced with a v4 address, the kernel will try to
   connect to a v4 address (v4-mapped-v6). A -EBADF error from the
   connect() part is then expected.

It is then required to check the given families can be accepted. This is
done by using a new helper for addresses family matching, taking care of
IPv4 vs IPv4-mapped-IPv6 addresses. This helper will be re-used later by
the in-kernel path-manager to use mixed IPv4 and IPv6 addresses.

While at it, a clear error message is now reported if there are some
conflicts with the families that have been passed by the userspace.

Fixes: 702c2f64

 ("mptcp: netlink: allow userspace-driven subflow establishment")
Cc: stable@vger.kernel.org
Reviewed-by: Mat Martineau <mathew.j.martineau@linux.intel.com>
Signed-off-by: Matthieu Baerts <matthieu.baerts@tessares.net>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>