Handle IPV6_PATHMTU option by spliting ip6_getpmtu_ctl() from ip6_getpmtu().

Add ro_mtu field to 'struct route' to be able to pass lookup MTU back to
  the caller.

Currently, ip6_getpmtu() has 2 totally different use cases:
1) control plane (IPV6_PATHMTU req), where we just need to calculate MTU
  and return it, w/o any reusability.
2) Actual ip6_output() data path where we (nearly) always use the provided
  route lookup data. If this data is not 'valid' we need to perform another
  lookup and save the result (which cannot be re-used by ip6_output()).

Given that, handle 1) by calling separate function doing rte lookup itself.
  Resulting MTU is calculated by (newly-added) ip6_calcmtu() used by both
  ip6_getpmtu_ctl() and ip6_getpmtu().
For 2) instead of storing ref'ed rte, store mtu (the only needed data
  from the lookup result) inside newly-added ro_mtu field.
  'struct route' was shrinked by 8(or 4 bytes) in r292978. Grow it again
  by 4 bytes. New ro_mtu field will be used in other places like
  ip/tcp_output (EMSGSIZE handling from output routines).

Reviewed by:	ae

2 files changed, 96 insertions, 32 deletions

diff --git a/sys/netinet6/in6.h b/sys/netinet6/in6.h
index 9bc142a..ae049f0 100644
--- a/sys/netinet6/in6.h
+++ b/sys/netinet6/in6.h
@@ -378,6 +378,8 @@ struct route_in6 {
 	char		*ro_prepend;
 	uint16_t	ro_plen;
 	uint16_t	ro_flags;
+	uint16_t	ro_mtu;	/* saved ro_rt mtu */
+	uint16_t	spare;
 	struct	sockaddr_in6 ro_dst;
 };
 #endif
diff --git a/sys/netinet6/ip6_output.c b/sys/netinet6/ip6_output.c
index af9fdba..16e860b 100644
--- a/sys/netinet6/ip6_output.c
+++ b/sys/netinet6/ip6_output.c
@@ -147,8 +147,11 @@ static int ip6_insertfraghdr(struct mbuf *, struct mbuf *, int,
 	struct ip6_frag **);
 static int ip6_insert_jumboopt(struct ip6_exthdrs *, u_int32_t);
 static int ip6_splithdr(struct mbuf *, struct ip6_exthdrs *);
-static int ip6_getpmtu(struct route_in6 *, struct route_in6 *,
+static int ip6_getpmtu(struct route_in6 *, int,
 	struct ifnet *, struct in6_addr *, u_long *, int *, u_int);
+static int ip6_calcmtu(struct ifnet *, const struct in6_addr *, u_long,
+	u_long *, int *);
+static int ip6_getpmtu_ctl(u_int, struct in6_addr *, u_long *);
 static int copypktopts(struct ip6_pktopts *, struct ip6_pktopts *, int);
 
 
@@ -712,7 +715,7 @@ again:
 		*ifpp = ifp;
 
 	/* Determine path MTU. */
-	if ((error = ip6_getpmtu(ro_pmtu, ro, ifp, &finaldst, &mtu,
+	if ((error = ip6_getpmtu(ro_pmtu, ro != ro_pmtu, ifp, &finaldst, &mtu,
 	    &alwaysfrag, fibnum)) != 0)
 		goto bad;
 
@@ -1045,8 +1048,6 @@ sendorfree:
 done:
 	if (ro == &ip6route)
 		RO_RTFREE(ro);
-	if (ro_pmtu == &ip6route)
-		RO_RTFREE(ro_pmtu);
 	return (error);
 
 freehdrs:
@@ -1215,35 +1216,104 @@ ip6_insertfraghdr(struct mbuf *m0, struct mbuf *m, int hlen,
 	return (0);
 }
 
+/*
+ * Calculates IPv6 path mtu for destination @dst.
+ * Resulting MTU is stored in @mtup.
+ *
+ * Returns 0 on success.
+ */
 static int
-ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
+ip6_getpmtu_ctl(u_int fibnum, struct in6_addr *dst, u_long *mtup)
+{
+	struct route_in6 ro_pmtu;
+	struct ifnet *ifp;
+	struct sockaddr_in6 *sa6_dst;
+	u_long mtu;
+
+	sa6_dst = (struct sockaddr_in6 *)&ro_pmtu.ro_dst;
+	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;
+
+	in6_rtalloc(&ro_pmtu, fibnum);
+
+	if (ro_pmtu.ro_rt == NULL)
+		return (EHOSTUNREACH);
+
+	ifp = ro_pmtu.ro_rt->rt_ifp;
+	mtu = ro_pmtu.ro_rt->rt_mtu;
+	RO_RTFREE(&ro_pmtu);
+
+	return (ip6_calcmtu(ifp, dst, mtu, mtup, NULL));
+}
+
+/*
+ * Calculates IPv6 path MTU for @dst based on transmit @ifp,
+ * and cached data in @ro_pmtu.
+ * MTU from (successful) route lookup is saved (along with dst)
+ * inside @ro_pmtu to avoid subsequent route lookups after packet
+ * filter processing.
+ *
+ * Stores mtu and always-frag value into @mtup and @alwaysfragp.
+ * Returns 0 on success.
+ */
+static int
+ip6_getpmtu(struct route_in6 *ro_pmtu, int do_lookup,
     struct ifnet *ifp, struct in6_addr *dst, u_long *mtup,
     int *alwaysfragp, u_int fibnum)
 {
-	u_int32_t mtu = 0;
-	int alwaysfrag = 0;
-	int error = 0;
+	struct sockaddr_in6 *sa6_dst;
+	u_long 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_pmtu->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) {
+	mtu = 0;
+	if (do_lookup) {
+
+		/*
+		 * Here ro_pmtu has final destination address, while
+		 * ro might represent immediate destination.
+		 * Use ro_pmtu destination since mtu might differ.
+		 */
+		sa6_dst = (struct sockaddr_in6 *)&ro_pmtu->ro_dst;
+		if (!IN6_ARE_ADDR_EQUAL(&sa6_dst->sin6_addr, dst))
+			ro_pmtu->ro_mtu = 0;
+
+		if (ro_pmtu->ro_mtu == 0) {
 			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;
 
 			in6_rtalloc(ro_pmtu, fibnum);
+			if (ro_pmtu->ro_rt) {
+				mtu = ro_pmtu->ro_rt->rt_mtu;
+				RO_RTFREE(ro_pmtu);
+			}
 		}
 	}
-	if (ro_pmtu->ro_rt) {
+
+	if (ro_pmtu->ro_rt)
+		mtu = ro_pmtu->ro_rt->rt_mtu;
+
+	return (ip6_calcmtu(ifp, dst, mtu, mtup, alwaysfragp));
+}
+
+/*
+ * Calculate MTU based on transmit @ifp, route mtu @rt_mtu and
+ * hostcache data for @dst.
+ * Stores mtu and always-frag value into @mtup and @alwaysfragp.
+ *
+ * Returns 0 on success.
+ */
+static int
+ip6_calcmtu(struct ifnet *ifp, const struct in6_addr *dst, u_long rt_mtu,
+    u_long *mtup, int *alwaysfragp)
+{
+	u_long mtu = 0;
+	int alwaysfrag = 0;
+	int error = 0;
+
+	if (rt_mtu > 0) {
 		u_int32_t ifmtu;
 		struct in_conninfo inc;
 
@@ -1251,14 +1321,12 @@ ip6_getpmtu(struct route_in6 *ro_pmtu, struct route_in6 *ro,
 		inc.inc_flags |= INC_ISIPV6;
 		inc.inc6_faddr = *dst;
 
-		if (ifp == NULL)
-			ifp = ro_pmtu->ro_rt->rt_ifp;
 		ifmtu = IN6_LINKMTU(ifp);
 		mtu = tcp_hc_getmtu(&inc);
 		if (mtu)
-			mtu = min(mtu, ro_pmtu->ro_rt->rt_mtu);
+			mtu = min(mtu, rt_mtu);
 		else
-			mtu = ro_pmtu->ro_rt->rt_mtu;
+			mtu = rt_mtu;
 		if (mtu == 0)
 			mtu = ifmtu;
 		else if (mtu < IPV6_MMTU) {
@@ -1936,9 +2004,6 @@ do { \
 			{
 				u_long pmtu = 0;
 				struct ip6_mtuinfo mtuinfo;
-				struct route_in6 sro;
-
-				bzero(&sro, sizeof(sro));
 
 				if (!(so->so_state & SS_ISCONNECTED))
 					return (ENOTCONN);
@@ -1947,11 +2012,8 @@ do { \
 				 * routing, or optional information to specify
 				 * the outgoing interface.
 				 */
-				error = ip6_getpmtu(&sro, NULL, NULL,
-				    &in6p->in6p_faddr, &pmtu, NULL,
-				    so->so_fibnum);
-				if (sro.ro_rt)
-					RTFREE(sro.ro_rt);
+				error = ip6_getpmtu_ctl(so->so_fibnum,
+				    &in6p->in6p_faddr, &pmtu);
 				if (error)
 					break;
 				if (pmtu > IPV6_MAXPACKET)