Merge final round of MLD changes from p4:

 ip6_input.c, in6.h:
 * Add netinet6-specific mbuf flag M_RTALERT_MLD, shadowing M_PROTO6.
  * Always set this flag if HBH Router Alert option is present for MLD,
    even when not forwarding.

 icmp6.c:
 * In icmp6_input(), spell m->m_pkthdr.rcvif as ifp to be consistent.
 * Use scope ID for verifying input. Do not apply SSM filters here, no inpcb.
  * Check for M_RTALERT_MLD when validating MLD traffic, as we can't see
    IPv6 hop options outside of ip6_input().

 in6_mcast.c:
 * Use KAME scope/zone ID in in6_multi.
   * Update net.inet6.ip6.mcast.filters implementation to use scope IDs
     for comparisons.
 * Fix scope ID treatment in multicast socket option processing.
   Scope IDs passed in from userland will be ignored as other less
   ambiguous APIs exist for specifying the link.
 * Tighten userland input checks in IPv6 SSM delta and full-state ops.
   * Source filter embedded scope IDs need to be revisited, for now
     just clear them and ignore them on input.
 * Adapt KAME behaviour of looking up the scope ID in the default zone
   for multicast leaves, when the interface is ambiguous.

 mld6.c:
 * Tighten origin checks on MLD traffic as per RFC3810 Section 6.2:
  * ip6_src MAY be the unspecified address for MLDv1 reports.
  * ip6_src MAY have link-local address scope for MLDv1 reports,
    MLDv1 queries, and MLDv2 queries.
  * Perform address field validation *before* accepting queries.
 * Use KAME scope/zone ID in query/report processing.
   * Break const correctness for mld_v1_input_report(), mld_v1_input_query()
     as we temporarily modify the input mbuf chain.
   * Clear the scope ID before handoff to userland MLD daemon.
 * Fix MLDv1 old querier present timer processing.
   With the protocol defaults, hosts should revert to MLDv2 after 260s.
 * Add net.inet6.mld.v1enable sysctl, default to on.

 ifmcstat.c:
 * Use sysctl by default; -K requests kvm(3) if so compiled.

 mld.4:
 * Connect man page to build.

Tested using PCS.

1 files changed, 14 insertions, 2 deletions

diff --git a/sys/netinet6/ip6_input.c b/sys/netinet6/ip6_input.c
index b0cf344..7498860 100644
--- a/sys/netinet6/ip6_input.c
+++ b/sys/netinet6/ip6_input.c
@@ -773,10 +773,11 @@ passin:
 		 * case we should pass the packet to the multicast routing
 		 * daemon.
 		 */
-		if (rtalert != ~0 && V_ip6_forwarding) {
+		if (rtalert != ~0) {
 			switch (rtalert) {
 			case IP6OPT_RTALERT_MLD:
-				ours = 1;
+				if (V_ip6_forwarding)
+					ours = 1;
 				break;
 			default:
 				/*
@@ -820,6 +821,9 @@ passin:
 		 * The packet is returned (relatively) intact; if
 		 * ip6_mforward() returns a non-zero value, the packet
 		 * must be discarded, else it may be accepted below.
+		 *
+		 * XXX TODO: Check hlim and multicast scope here to avoid
+		 * unnecessarily calling into ip6_mforward().
 		 */
 		if (ip6_mforward &&
 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
@@ -882,6 +886,14 @@ passin:
 		if (ip6_ipsec_input(m, nxt))
 			goto bad;
 #endif /* IPSEC */
+
+		/*
+		 * Use mbuf flags to propagate Router Alert option to
+		 * ICMPv6 layer, as hop-by-hop options have been stripped.
+		 */
+		if (nxt == IPPROTO_ICMPV6 && rtalert != ~0)
+			m->m_flags |= M_RTALERT_MLD;
+
 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
 	}
 	goto out;