correct linkmtu handling.

Obtained from:	KAME

8 files changed, 120 insertions, 107 deletions

diff --git a/sys/netinet6/in6.c b/sys/netinet6/in6.c
index d1c5f8c..83b5913 100644
--- a/sys/netinet6/in6.c
+++ b/sys/netinet6/in6.c
@@ -2288,9 +2288,12 @@ in6_setmaxmtu()
 
 	IFNET_RLOCK();
 	for (ifp = TAILQ_FIRST(&ifnet); ifp; ifp = TAILQ_NEXT(ifp, if_list)) {
+		/* this function can be called during ifnet initialization */
+		if (!ifp->if_afdata[AF_INET6])
+			continue;
 		if ((ifp->if_flags & IFF_LOOPBACK) == 0 &&
-		    ND_IFINFO(ifp)->linkmtu > maxmtu)
-			maxmtu =  ND_IFINFO(ifp)->linkmtu;
+		    IN6_LINKMTU(ifp) > maxmtu)
+			maxmtu = IN6_LINKMTU(ifp);
 	}
 	IFNET_RUNLOCK();
 	if (maxmtu)	     /* update only when maxmtu is positive */
diff --git a/sys/netinet6/in6_rmx.c b/sys/netinet6/in6_rmx.c
index cc1ba310d..2853fe8 100644
--- a/sys/netinet6/in6_rmx.c
+++ b/sys/netinet6/in6_rmx.c
@@ -94,6 +94,7 @@
 #include <netinet6/ip6_var.h>
 
 #include <netinet/icmp6.h>
+#include <netinet6/nd6.h>
 
 #include <netinet/tcp.h>
 #include <netinet/tcp_seq.h>
@@ -149,7 +150,7 @@ in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head,
 
 	if (!rt->rt_rmx.rmx_mtu && !(rt->rt_rmx.rmx_locks & RTV_MTU)
 	    && rt->rt_ifp)
-		rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu;
+		rt->rt_rmx.rmx_mtu = IN6_LINKMTU(rt->rt_ifp);
 
 	ret = rn_addroute(v_arg, n_arg, head, treenodes);
 	if (ret == NULL && rt->rt_flags & RTF_HOST) {
diff --git a/sys/netinet6/ip6_forward.c b/sys/netinet6/ip6_forward.c
index 6414ab5..3cc812e 100644
--- a/sys/netinet6/ip6_forward.c
+++ b/sys/netinet6/ip6_forward.c
@@ -383,7 +383,7 @@ ip6_forward(m, srcrt)
 		return;
 	}
 
-	if (m->m_pkthdr.len > rt->rt_ifp->if_mtu) {
+	if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) {
 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
 		if (mcopy) {
 			u_long mtu;
@@ -393,7 +393,7 @@ ip6_forward(m, srcrt)
 			size_t ipsechdrsiz;
 #endif
 
-			mtu = rt->rt_ifp->if_mtu;
+			mtu = IN6_LINKMTU(rt->rt_ifp);
 #ifdef IPSEC
 			/*
 			 * When we do IPsec tunnel ingress, we need to play
diff --git a/sys/netinet6/ip6_mroute.c b/sys/netinet6/ip6_mroute.c
index eeb9b21..607a2ea 100644
--- a/sys/netinet6/ip6_mroute.c
+++ b/sys/netinet6/ip6_mroute.c
@@ -113,6 +113,7 @@
 
 #include <netinet/ip6.h>
 #include <netinet6/ip6_var.h>
+#include <netinet6/nd6.h>
 #include <netinet6/ip6_mroute.h>
 #include <netinet6/pim6.h>
 #include <netinet6/pim6_var.h>
@@ -1454,6 +1455,7 @@ phyint_send(ip6, mifp, m)
 	static struct route_in6 ro;
 	struct	in6_multi *in6m;
 	struct sockaddr_in6 *dst6;
+	u_long linkmtu;
 
 	/*
 	 * Make a new reference to the packet; make sure that
@@ -1513,7 +1515,8 @@ phyint_send(ip6, mifp, m)
 	 * Put the packet into the sending queue of the outgoing interface
 	 * if it would fit in the MTU of the interface.
 	 */
-	if (mb_copy->m_pkthdr.len <= ifp->if_mtu || ifp->if_mtu < IPV6_MMTU) {
+	linkmtu = IN6_LINKMTU(ifp);
+	if (mb_copy->m_pkthdr.len <= linkmtu || linkmtu < IPV6_MMTU) {
 		dst6->sin6_len = sizeof(struct sockaddr_in6);
 		dst6->sin6_family = AF_INET6;
 		dst6->sin6_addr = ip6->ip6_dst;
@@ -1530,7 +1533,7 @@ phyint_send(ip6, mifp, m)
 #endif
 	} else {
 #ifdef MULTICAST_PMTUD
-		icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
+		icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, linkmtu);
 #else
 #ifdef MRT6DEBUG
 		if (mrt6debug & DEBUG_XMIT)
diff --git a/sys/netinet6/ip6_output.c b/sys/netinet6/ip6_output.c
index 1f0c0cc..12dd282 100644
--- a/sys/netinet6/ip6_output.c
+++ b/sys/netinet6/ip6_output.c
@@ -137,6 +137,8 @@ static int ip6_insertfraghdr __P((struct mbuf *, struct mbuf *, int,
 	struct ip6_frag **));
 static int ip6_insert_jumboopt __P((struct ip6_exthdrs *, u_int32_t));
 static int ip6_splithdr __P((struct mbuf *, struct ip6_exthdrs *));
+static int ip6_getpmtu __P((struct route_in6 *, struct route_in6 *,
+	struct ifnet *, struct in6_addr *, u_long *));
 
 
 /*
@@ -771,51 +773,9 @@ skip_ipsec2:;
 	if (ifpp)
 		*ifpp = ifp;
 
-	/*
-	 * Determine path MTU.
-	 */
-	if (ro_pmtu != ro) {
-		/* The first hop and the final destination may differ. */
-		struct sockaddr_in6 *sin6_fin =
-			(struct sockaddr_in6 *)&ro_pmtu->ro_dst;
-		if (ro_pmtu->ro_rt && ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
-				       !IN6_ARE_ADDR_EQUAL(&sin6_fin->sin6_addr,
-							   &finaldst))) {
-			RTFREE(ro_pmtu->ro_rt);
-			ro_pmtu->ro_rt = (struct rtentry *)0;
-		}
-		if (ro_pmtu->ro_rt == 0) {
-			bzero(sin6_fin, sizeof(*sin6_fin));
-			sin6_fin->sin6_family = AF_INET6;
-			sin6_fin->sin6_len = sizeof(struct sockaddr_in6);
-			sin6_fin->sin6_addr = finaldst;
-
-			rtalloc((struct route *)ro_pmtu);
-		}
-	}
-	if (ro_pmtu->ro_rt != NULL) {
-		u_int32_t ifmtu = ND_IFINFO(ifp)->linkmtu;
-
-		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
-		if (mtu > ifmtu || mtu == 0) {
-			/*
-			 * The MTU on the route is larger than the MTU on
-			 * the interface!  This shouldn't happen, unless the
-			 * MTU of the interface has been changed after the
-			 * interface was brought up.  Change the MTU in the
-			 * route to match the interface MTU (as long as the
-			 * field isn't locked).
-			 *
-			 * if MTU on the route is 0, we need to fix the MTU.
-			 * this case happens with path MTU discovery timeouts.
-			 */
-			 mtu = ifmtu;
-			 if ((ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU) == 0)
-				 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu; /* XXX */
-		}
-	} else {
-		mtu = ND_IFINFO(ifp)->linkmtu;
-	}
+	/* Determine path MTU. */
+	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu)) != 0)
+		goto bad;
 
 	/*
 	 * advanced API (IPV6_USE_MIN_MTU) overrides mtu setting
@@ -1300,6 +1260,69 @@ ip6_insertfraghdr(m0, m, hlen, frghdrp)
 	return (0);
 }
 
+static int
+ip6_getpmtu(ro_pmtu, ro, ifp, dst, mtup)
+	struct route_in6 *ro_pmtu, *ro;
+	struct ifnet *ifp;
+	struct in6_addr *dst;
+	u_long *mtup;
+{
+	u_int32_t mtu = 0;
+	int error = 0;
+
+	/*
+	 * Determine path MTU.
+	 */
+	if (ro_pmtu != ro) {
+		/* The first hop and the final destination may differ. */
+		struct sockaddr_in6 *sa6_dst =
+			(struct sockaddr_in6 *)&ro_pmtu->ro_dst;
+		if (ro_pmtu->ro_rt &&
+		    ((ro->ro_rt->rt_flags & RTF_UP) == 0 ||
+		     !IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))) {
+			RTFREE(ro_pmtu->ro_rt);
+			ro_pmtu->ro_rt = (struct rtentry *)NULL;
+		}
+		if (ro_pmtu->ro_rt == NULL) {
+			bzero(sa6_dst, sizeof(*sa6_dst));
+			sa6_dst->sin6_family = AF_INET6;
+			sa6_dst->sin6_len = sizeof(struct sockaddr_in6);
+			sa6_dst->sin6_addr = *dst;
+
+			rtalloc((struct route *)ro_pmtu);
+		}
+	}
+	if (ro_pmtu->ro_rt) {
+		u_int32_t ifmtu;
+
+		if (ifp == NULL)
+			ifp = ro_pmtu->ro_rt->rt_ifp;
+		ifmtu = IN6_LINKMTU(ifp);
+		mtu = ro_pmtu->ro_rt->rt_rmx.rmx_mtu;
+		if (mtu == 0)
+			mtu = ifmtu;
+		else if (mtu > ifmtu || mtu == 0) {
+			/*
+			 * The MTU on the route is larger than the MTU on
+			 * the interface!  This shouldn't happen, unless the
+			 * MTU of the interface has been changed after the
+			 * interface was brought up.  Change the MTU in the
+			 * route to match the interface MTU (as long as the
+			 * field isn't locked).
+			 */
+			 mtu = ifmtu;
+			 if (!(ro_pmtu->ro_rt->rt_rmx.rmx_locks & RTV_MTU))
+				 ro_pmtu->ro_rt->rt_rmx.rmx_mtu = mtu;
+		}
+	} else if (ifp) {
+		mtu = IN6_LINKMTU(ifp);
+	} else
+		error = EHOSTUNREACH; /* XXX */
+
+	*mtup = mtu;
+	return (error);
+}
+
 /*
  * IP6 socket option processing.
  */
diff --git a/sys/netinet6/nd6.c b/sys/netinet6/nd6.c
index d7a0e6f..7a74cdf 100644
--- a/sys/netinet6/nd6.c
+++ b/sys/netinet6/nd6.c
@@ -51,6 +51,7 @@
 #include <sys/sysctl.h>
 
 #include <net/if.h>
+#include <net/if_arc.h>
 #include <net/if_dl.h>
 #include <net/if_types.h>
 #include <net/if_atm.h>
@@ -149,12 +150,10 @@ nd6_ifattach(ifp)
 
 	nd->initialized = 1;
 
-	nd->linkmtu = ifnet_byindex(ifp->if_index)->if_mtu;
 	nd->chlim = IPV6_DEFHLIM;
 	nd->basereachable = REACHABLE_TIME;
 	nd->reachable = ND_COMPUTE_RTIME(nd->basereachable);
 	nd->retrans = RETRANS_TIMER;
-	nd->receivedra = 0;
 	/*
 	 * Note that the default value of ip6_accept_rtadv is 0, which means
 	 * we won't accept RAs by default even if we set ND6_IFF_ACCEPT_RTADV
@@ -194,21 +193,19 @@ nd6_setmtu0(ifp, ndi)
 	struct ifnet *ifp;
 	struct nd_ifinfo *ndi;
 {
-	u_long oldmaxmtu;
-	u_long oldlinkmtu;
+	u_int32_t omaxmtu;
 
-	oldmaxmtu = ndi->maxmtu;
-	oldlinkmtu = ndi->linkmtu;
+	omaxmtu = ndi->maxmtu;
 
 	switch (ifp->if_type) {
-	case IFT_ARCNET:	/* XXX MTU handling needs more work */
-		ndi->maxmtu = MIN(60480, ifp->if_mtu);
+	case IFT_ARCNET:
+		ndi->maxmtu = MIN(ARC_PHDS_MAXMTU, ifp->if_mtu); /* RFC2497 */
 		break;
 	case IFT_ETHER:
 		ndi->maxmtu = MIN(ETHERMTU, ifp->if_mtu);
 		break;
 	case IFT_FDDI:
-		ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu);
+		ndi->maxmtu = MIN(FDDIIPMTU, ifp->if_mtu); /* RFC2467 */
 		break;
 	case IFT_ATM:
 		ndi->maxmtu = MIN(ATMMTU, ifp->if_mtu);
@@ -229,29 +226,21 @@ nd6_setmtu0(ifp, ndi)
 		break;
 	}
 
-	if (oldmaxmtu != ndi->maxmtu) {
-		/*
-		 * If the ND level MTU is not set yet, or if the maxmtu
-		 * is reset to a smaller value than the ND level MTU,
-		 * also reset the ND level MTU.
-		 */
-		if (ndi->linkmtu == 0 ||
-		    ndi->maxmtu < ndi->linkmtu) {
-			ndi->linkmtu = ndi->maxmtu;
-			/* also adjust in6_maxmtu if necessary. */
-			if (oldlinkmtu == 0) {
-				/*
-				 * XXX: the case analysis is grotty, but
-				 * it is not efficient to call in6_setmaxmtu()
-				 * here when we are during the initialization
-				 * procedure.
-				 */
-				if (in6_maxmtu < ndi->linkmtu)
-					in6_maxmtu = ndi->linkmtu;
-			} else
-				in6_setmaxmtu();
-		}
+	/*
+	 * Decreasing the interface MTU under IPV6 minimum MTU may cause
+	 * undesirable situation.  We thus notify the operator of the change
+	 * explicitly.  The check for omaxmtu is necessary to restrict the
+	 * log to the case of changing the MTU, not initializing it.
+	 */
+	if (omaxmtu >= IPV6_MMTU && ndi->maxmtu < IPV6_MMTU) {
+		log(LOG_NOTICE, "nd6_setmtu0: "
+		    "new link MTU on %s (%lu) is too small for IPv6\n",
+		    if_name(ifp), (unsigned long)ndi->maxmtu);
 	}
+
+	if (ndi->maxmtu > in6_maxmtu)
+		in6_setmaxmtu(); /* check all interfaces just in case */
+
 #undef MIN
 }
 
@@ -1449,7 +1438,7 @@ nd6_ioctl(cmd, data, ifp)
 	case OSIOCGIFINFO_IN6:
 		/* XXX: old ndp(8) assumes a positive value for linkmtu. */
 		bzero(&ndi->ndi, sizeof(ndi->ndi));
-		ndi->ndi.linkmtu = ND_IFINFO(ifp)->linkmtu;
+		ndi->ndi.linkmtu = IN6_LINKMTU(ifp);
 		ndi->ndi.maxmtu = ND_IFINFO(ifp)->maxmtu;
 		ndi->ndi.basereachable = ND_IFINFO(ifp)->basereachable;
 		ndi->ndi.reachable = ND_IFINFO(ifp)->reachable;
@@ -1457,10 +1446,10 @@ nd6_ioctl(cmd, data, ifp)
 		ndi->ndi.flags = ND_IFINFO(ifp)->flags;
 		ndi->ndi.recalctm = ND_IFINFO(ifp)->recalctm;
 		ndi->ndi.chlim = ND_IFINFO(ifp)->chlim;
-		ndi->ndi.receivedra = ND_IFINFO(ifp)->receivedra;
 		break;
 	case SIOCGIFINFO_IN6:
 		ndi->ndi = *ND_IFINFO(ifp);
+		ndi->ndi.linkmtu = IN6_LINKMTU(ifp);
 		break;
 	case SIOCSIFINFO_FLAGS:
 		ND_IFINFO(ifp)->flags = ndi->ndi.flags;
diff --git a/sys/netinet6/nd6.h b/sys/netinet6/nd6.h
index 2ee4a48..5762312 100644
--- a/sys/netinet6/nd6.h
+++ b/sys/netinet6/nd6.h
@@ -79,7 +79,6 @@ struct nd_ifinfo {
 	u_int32_t flags;		/* Flags */
 	int recalctm;			/* BaseReacable re-calculation timer */
 	u_int8_t chlim;			/* CurHopLimit */
-	u_int8_t receivedra;
 	u_int8_t initialized; /* Flag to see the entry is initialized */
 	/* the following 3 members are for privacy extension for addrconf */
 	u_int8_t randomseed0[8]; /* upper 64 bits of MD5 digest */
diff --git a/sys/netinet6/nd6_rtr.c b/sys/netinet6/nd6_rtr.c
index ec65cb7..e5e61dd 100644
--- a/sys/netinet6/nd6_rtr.c
+++ b/sys/netinet6/nd6_rtr.c
@@ -364,38 +364,33 @@ nd6_ra_input(m, off, icmp6len)
 	 * MTU
 	 */
 	if (ndopts.nd_opts_mtu && ndopts.nd_opts_mtu->nd_opt_mtu_len == 1) {
-		u_int32_t mtu;
+		u_long mtu;
+		u_long maxmtu;
 
-		mtu = ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
+		mtu = (u_long)ntohl(ndopts.nd_opts_mtu->nd_opt_mtu_mtu);
 
 		/* lower bound */
 		if (mtu < IPV6_MMTU) {
 			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu option "
-			    "mtu=%d sent from %s, ignoring\n",
+			    "mtu=%lu sent from %s, ignoring\n",
 			    mtu, ip6_sprintf(&ip6->ip6_src)));
 			goto skip;
 		}
 
 		/* upper bound */
-		if (ndi->maxmtu) {
-			if (mtu <= ndi->maxmtu) {
-				int change = (ndi->linkmtu != mtu);
-
-				ndi->linkmtu = mtu;
-				if (change) /* in6_maxmtu may change */
-					in6_setmaxmtu();
-			} else {
-				nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
-				    "mtu=%d sent from %s; "
-				    "exceeds maxmtu %d, ignoring\n",
-				    mtu, ip6_sprintf(&ip6->ip6_src),
-				    ndi->maxmtu));
-			}
+		maxmtu = (ndi->maxmtu && ndi->maxmtu < ifp->if_mtu)
+		    ? ndi->maxmtu : ifp->if_mtu;
+		if (mtu <= maxmtu) {
+			int change = (ndi->linkmtu != mtu);
+
+			ndi->linkmtu = mtu;
+			if (change) /* in6_maxmtu may change */
+				in6_setmaxmtu();
 		} else {
-			nd6log((LOG_INFO, "nd6_ra_input: mtu option "
-			    "mtu=%d sent from %s; maxmtu unknown, "
-			    "ignoring\n",
-			    mtu, ip6_sprintf(&ip6->ip6_src)));
+			nd6log((LOG_INFO, "nd6_ra_input: bogus mtu "
+			    "mtu=%lu sent from %s; "
+			    "exceeds maxmtu %lu, ignoring\n",
+			    mtu, ip6_sprintf(&ip6->ip6_src), maxmtu));
 		}
 	}