Actually, as routing over an interface can work without assigning global addresses at all, it should not be necessary to assign any of them. Indeed, KAME derived IPv6 stacks do not require global IPv6 addresses to be assigned to a tunnel interface (as tested by the author with NetBSD and FreeBSD 3.x). In that case, one needs to explicitly add a route over the tunnel interface to the linklocal address of the remote tunnel end (or run a routing protocol that will figure it out for itself using the well-known multicast address). For example, to add a default route over a tunnel interface to a remote end with link-local address fe80::2, one would add a route like this:

# route add –inet6 default fe80::2%gif0

or

# route add –inet6 default –ifp gif0

Such unnumbered tunnels will typically only make sense when both tunnel endpoint use a KAME derived IPv6 stack as other implementations usually don’t support unnumbered tunnels.

Alternative with FreeBSD

As an alternative to the above way of creating a manual IPv6-in-IPv4 tunnel, FreeBSD also offers the possibility to set up a tunnel by only modifying the contents of /etc/rc.conf file such that the tunnel is automatically created and configured at boot time. Add the following lines to the rc.conf file:

gif_interfaces=“gif0”
gifconfig_gif0=“130.1.2.3 134.5.6.7” \
ipv6_ifconfig_gif0=“2001:DEAD:BEAF:ABBA::1 \
2001:DEAD:BEAF:ABBA::2 prefixlen 128”

Alternative with NetBSD

With NetBSD, one can easily configure a tunnel interface by making a corresponding /etc/ifconfig.<ifn> file, where <ifn> is the name if the interface (i.e. gif0).

For the example above one would create a file /etc/ifconfig.gif0 containing the following lines:

create
tunnel 130.1.2.3 134.5.6.7
inet6 2001:DEAD:BEAF:ABBA::1 2001:DEAD:BEAF:ABBA::2 prefixlen 128