My group has been setting up dynamic tunnels in mrouted over wireless
interfaces, and found what looks like a bug in
sys/netinet/ip_mroute.c:vif_input().

The protocol is checked against AF_INET (2), when it seems like it
should be checked against ENCAP_PROTO (IPPROTO_IPIP, which is 4).
As a result incoming packets on vifs are dropped.

	if (!vifp || proto != AF_INET) {
		m_freem(m);
		mrtstat.mrts_bad_tunnel++;
		return;
	}

This code was last touched on 08-May-01 in revision 1.54, which
changed to use the new encap matching framework.

Christine Jones of BBN found this, and when changed to ENCAP_PROTO
packets are received on vifs.

The code seems "clearly wrong by inspection", but given that it's been
4 years I wonder if we're somehow confused.  Has mrouted with vifs
worked for anyone since May 2001?  Should I just change AF_INET to
ENCAP_PROTO?