]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/netinet6/ip6_input.c
xnu-1699.22.73.tar.gz
[apple/xnu.git] / bsd / netinet6 / ip6_input.c
index 0a8320298a5cb2b408913a452ccb2029be7128d8..ae8fecd689a080106caa62721d80281c2b655e04 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2003-2008 Apple Inc. All rights reserved.
+ * Copyright (c) 2003-2011 Apple Inc. All rights reserved.
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
  * 
 #include <sys/time.h>
 #include <sys/kernel.h>
 #include <sys/syslog.h>
+#include <sys/sysctl.h>
 #include <sys/proc.h>
 #include <sys/kauth.h>
+#include <sys/mcache.h>
+#include <mach/mach_time.h>
+
+#include <pexpert/pexpert.h>
 
 #include <net/if.h>
 #include <net/if_var.h>
 #include <net/if_dl.h>
 #include <net/route.h>
 #include <net/kpi_protocol.h>
+#include <net/ntstat.h>
 
 #include <netinet/in.h>
 #include <netinet/in_systm.h>
 #include <netinet/icmp6.h>
 #include <netinet6/in6_ifattach.h>
 #include <netinet6/nd6.h>
-#include <netinet6/in6_prefix.h>
+#include <netinet6/scope6_var.h>
+#include <mach/sdt.h>
 
 #if IPSEC
 #include <netinet6/ipsec.h>
@@ -159,7 +166,14 @@ extern struct ip6protosw inet6sw[];
 
 struct ip6protosw *  ip6_protox[IPPROTO_MAX];
 static int ip6qmaxlen = IFQ_MAXLEN;
-struct in6_ifaddr *in6_ifaddrs;
+
+static lck_grp_attr_t  *in6_ifaddr_rwlock_grp_attr;
+static lck_grp_t       *in6_ifaddr_rwlock_grp;
+static lck_attr_t      *in6_ifaddr_rwlock_attr;
+decl_lck_rw_data(, in6_ifaddr_rwlock);
+
+/* Protected by in6_ifaddr_rwlock */
+struct in6_ifaddr *in6_ifaddrs = NULL;
 
 int ip6_forward_srcrt;                 /* XXX */
 int ip6_sourcecheck;                   /* XXX */
@@ -168,7 +182,14 @@ const int int6intrq_present = 1;
 
 int ip6_ours_check_algorithm;
 int in6_init2done = 0;
+int in6_init_done = 0;
 
+#define _CASSERT(x)    \
+       switch (0) { case 0: case (x): ; }
+#define IN6_IFSTAT_REQUIRE_ALIGNED_64(f)       \
+       _CASSERT(!(offsetof(struct in6_ifstat, f) % sizeof (uint64_t)))
+#define ICMP6_IFSTAT_REQUIRE_ALIGNED_64(f)     \
+       _CASSERT(!(offsetof(struct icmp6_ifstat, f) % sizeof (uint64_t)))
 
 #if IPFW2
 /* firewall hooks */
@@ -181,17 +202,23 @@ struct ip6stat ip6stat;
 
 #ifdef __APPLE__
 struct ifqueue ip6intrq;
-lck_mtx_t              *ip6_mutex;
+decl_lck_mtx_data(, ip6_init_mutex);
 lck_mtx_t              *dad6_mutex;
 lck_mtx_t              *nd6_mutex;
 lck_mtx_t              *prefix6_mutex;
 lck_mtx_t              *scope6_mutex;
+#ifdef ENABLE_ADDRSEL
+lck_mtx_t              *addrsel_mutex;
+#endif
+decl_lck_rw_data(, in6_ifs_rwlock);
+decl_lck_rw_data(, icmp6_ifs_rwlock);
 lck_attr_t             *ip6_mutex_attr;
 lck_grp_t              *ip6_mutex_grp;
 lck_grp_attr_t         *ip6_mutex_grp_attr;
 extern lck_mtx_t       *inet6_domain_mutex;
 #endif
 extern int loopattach_done;
+extern void addrsel_policy_init(void);
 
 static void ip6_init2(void *);
 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *);
@@ -209,6 +236,11 @@ void stfattach(void);
 
 extern lck_mtx_t *domain_proto_mtx;
 
+SYSCTL_DECL(_net_inet6_ip6);
+
+int    ip6_doscopedroute = 1;
+SYSCTL_INT(_net_inet6_ip6, OID_AUTO, scopedroute, CTLFLAG_RD | CTLFLAG_LOCKED,
+     &ip6_doscopedroute, 0, "Enable IPv6 scoped routing");
 
 static void
 ip6_proto_input(
@@ -229,6 +261,9 @@ ip6_init()
        int i;
        struct timeval tv;
 
+       PE_parse_boot_argn("net.inet6.ip6.scopedroute", &ip6_doscopedroute,
+           sizeof (ip6_doscopedroute));
+
 #if DIAGNOSTIC
        if (sizeof(struct protosw) != sizeof(struct ip6protosw))
                panic("sizeof(protosw) != sizeof(ip6protosw)");
@@ -251,9 +286,6 @@ ip6_init()
        ip6_mutex_grp = lck_grp_alloc_init("ip6", ip6_mutex_grp_attr);
        ip6_mutex_attr = lck_attr_alloc_init();
 
-       if ((ip6_mutex = lck_mtx_alloc_init(ip6_mutex_grp, ip6_mutex_attr)) == NULL) {
-               panic("ip6_init: can't alloc ip6_mutex\n");
-       }
        if ((dad6_mutex = lck_mtx_alloc_init(ip6_mutex_grp, ip6_mutex_attr)) == NULL) {
                panic("ip6_init: can't alloc dad6_mutex\n");
        }
@@ -269,14 +301,90 @@ ip6_init()
                panic("ip6_init: can't alloc scope6_mutex\n");
        }
 
+#ifdef ENABLE_ADDRSEL
+       if ((addrsel_mutex = lck_mtx_alloc_init(ip6_mutex_grp, ip6_mutex_attr)) == NULL) {
+               panic("ip6_init: can't alloc addrsel_mutex\n");
+       }
+#endif
+
+       lck_rw_init(&in6_ifs_rwlock, ip6_mutex_grp, ip6_mutex_attr);
+       lck_rw_init(&icmp6_ifs_rwlock, ip6_mutex_grp, ip6_mutex_attr);
+       lck_mtx_init(&ip6_init_mutex, ip6_mutex_grp, ip6_mutex_attr);
 
        inet6domain.dom_flags = DOM_REENTRANT;  
 
        ip6intrq.ifq_maxlen = ip6qmaxlen;
+
+       in6_ifaddr_rwlock_grp_attr = lck_grp_attr_alloc_init();
+       in6_ifaddr_rwlock_grp = lck_grp_alloc_init("in6_ifaddr_rwlock",
+           in6_ifaddr_rwlock_grp_attr);
+       in6_ifaddr_rwlock_attr = lck_attr_alloc_init();
+       lck_rw_init(&in6_ifaddr_rwlock, in6_ifaddr_rwlock_grp,
+           in6_ifaddr_rwlock_attr);
+
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_receive);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_hdrerr);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_toobig);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_noroute);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_addrerr);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_protounknown);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_truncated);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_discard);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_deliver);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_forward);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_request);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_discard);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_fragok);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_fragfail);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_fragcreat);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_reass_reqd);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_reass_ok);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_reass_fail);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_mcast);
+       IN6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_mcast); 
+
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_msg);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_error);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_dstunreach);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_adminprohib);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_timeexceed);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_paramprob);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_pkttoobig);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_echo);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_echoreply);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_routersolicit);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_routeradvert);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_neighborsolicit);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_neighboradvert);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_redirect);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_mldquery);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_mldreport);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_in_mlddone);
+
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_msg);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_error);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_dstunreach);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_adminprohib);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_timeexceed);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_paramprob);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_pkttoobig);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_echo);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_echoreply);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_routersolicit);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_routeradvert);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_neighborsolicit);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_neighboradvert);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_redirect);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_mldquery);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_mldreport);
+       ICMP6_IFSTAT_REQUIRE_ALIGNED_64(ifs6_out_mlddone);
+
        in6_ifaddr_init();
+       ip6_moptions_init();
        nd6_init();
        frag6_init();
        icmp6_init();
+       addrsel_policy_init();
        /*
         * in many cases, random() here does NOT return random number
         * as initialization during bootstrap time occur in fixed order.
@@ -310,9 +418,6 @@ ip6_init2(
        /* nd6_timer_init */
        timeout(nd6_timer, (caddr_t)0, hz);
 
-       /* router renumbering prefix list maintenance */
-       timeout(in6_rr_timer, (caddr_t)0, hz);
-
        /* timer for regeneranation of temporary addresses randomize ID */
        timeout(in6_tmpaddrtimer, (caddr_t)0,
                (ip6_temp_preferred_lifetime - ip6_desync_factor -
@@ -327,42 +432,27 @@ ip6_init2(
 #if NSTF
        stfattach();
 #endif
-#else
-       /* nd6_timer_init */
-
-       callout_init(&nd6_timer_ch);
-       callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
-
-       /* router renumbering prefix list maintenance */
-       callout_init(&in6_rr_timer_ch);
-       callout_reset(&in6_rr_timer_ch, hz, in6_rr_timer, NULL);
-
-       /* timer for regeneranation of temporary addresses randomize ID */
-       callout_reset(&in6_tmpaddrtimer_ch,
-                     (ip6_temp_preferred_lifetime - ip6_desync_factor -
-                      ip6_temp_regen_advance) * hz,
-                     in6_tmpaddrtimer, NULL);
 #endif
-
        in6_init2done = 1;
-}
 
-#if __FreeBSD__
-/* cheat */
-/* This must be after route_init(), which is now SI_ORDER_THIRD */
-SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
-#endif
+       lck_mtx_lock(&ip6_init_mutex);
+       in6_init_done = 1;
+       wakeup(&in6_init_done);
+       lck_mtx_unlock(&ip6_init_mutex);
+}
 
-/*
- * ip6_forward_rt contains the route entry that was recently used during
- * the forwarding of an IPv6 packet and thus acts as a route cache.  Access
- * to this variable is protected by the global lock ip6_mutex.
- */
-static struct route_in6 ip6_forward_rt;
+void
+ip6_fin()
+{
+       lck_mtx_lock(&ip6_init_mutex);
+       while (in6_init_done == 0) {
+               (void) msleep(&in6_init_done, &ip6_init_mutex, 0, "ip6_fin()", NULL);
+       }
+       lck_mtx_unlock(&ip6_init_mutex);
+}
 
 void
-ip6_input(m)
-       struct mbuf *m;
+ip6_input(struct mbuf *m)
 {
        struct ip6_hdr *ip6;
        int off = sizeof(struct ip6_hdr), nest;
@@ -372,6 +462,16 @@ ip6_input(m)
        struct ifnet *deliverifp = NULL;
        ipfilter_t inject_ipfref = 0;
        int seen;
+       struct in6_ifaddr *ia6 = NULL;
+       struct route_in6 ip6_forward_rt;
+       struct sockaddr_in6 *dst6;
+
+       bzero(&ip6_forward_rt, sizeof(ip6_forward_rt));
+
+       /* Check if the packet we received is valid after interface filter
+        * processing
+        */
+       MBUF_INPUT_CHECK(m, m->m_pkthdr.rcvif);
 
        /*
         * No need to proccess packet twice if we've 
@@ -402,7 +502,6 @@ ip6_input(m)
         */
        ip6_delaux(m);
 
-       lck_mtx_lock(ip6_mutex);
        /*
         * mbuf statistics
         */
@@ -425,6 +524,15 @@ ip6_input(m)
 #undef M2MMAX
        }
 
+       /* drop the packet if IPv6 operation is disabled on the IF */
+       lck_rw_lock_shared(nd_if_rwlock);
+       if (m->m_pkthdr.rcvif->if_index < nd_ifinfo_indexlim &&
+           (nd_ifinfo[m->m_pkthdr.rcvif->if_index].flags & ND6_IFF_IFDISABLED)) {
+               lck_rw_done(nd_if_rwlock);
+               goto bad;
+       }
+       lck_rw_done(nd_if_rwlock);
+
        in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
        ip6stat.ip6s_total++;
 
@@ -447,11 +555,8 @@ ip6_input(m)
                                n = NULL;
                        }
                }
-               if (n == NULL) {
-                       m_freem(m);
-                       lck_mtx_unlock(ip6_mutex);
-                       return; /*ENOBUFS*/
-               }
+               if (n == NULL)
+                       goto bad;
 
                m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
                n->m_len = m->m_pkthdr.len;
@@ -459,7 +564,7 @@ ip6_input(m)
                m = n;
        }
        IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr),
-               {lck_mtx_unlock(ip6_mutex); return;}); 
+               {goto done;}); 
 #endif
 
        if (m->m_len < sizeof(struct ip6_hdr)) {
@@ -468,8 +573,7 @@ ip6_input(m)
                if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
                        ip6stat.ip6s_toosmall++;
                        in6_ifstat_inc(inifp, ifs6_in_hdrerr);
-                       lck_mtx_unlock(ip6_mutex);
-                       return;
+                       goto done;
                }
        }
 
@@ -495,10 +599,8 @@ ip6_input(m)
                        m_freem(m);
                        m = NULL;
                }
-               if (!m) {
-                       lck_mtx_unlock(ip6_mutex);
-                       return;
-               }
+               if (!m)
+                       goto done;
        }
 #endif
 
@@ -514,9 +616,14 @@ ip6_input(m)
                in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
                goto bad;
        }
-       if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
-            IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
-           (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
+       if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
+           !(m->m_flags & M_LOOP)) {
+               /*
+                * In this case, the packet should come from the loopback
+                * interface.  However, we cannot just check the if_flags,
+                * because ip6_mloopback() passes the "actual" interface
+                * as the outgoing/incoming interface.
+                */
                ip6stat.ip6s_badscope++;
                in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
                goto bad;
@@ -526,13 +633,13 @@ ip6_input(m)
         * The following check is not documented in specs.  A malicious
         * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
         * and bypass security checks (act as if it was from 127.0.0.1 by using
-        * IPv6 src ::ffff:127.0.0.1).  Be cautious.
+        * IPv6 src ::ffff:127.0.0.1).  Be cautious.
         *
         * This check chokes if we are in an SIIT cloud.  As none of BSDs
         * support IPv4-less kernel compilation, we cannot support SIIT
         * environment at all.  So, it makes more sense for us to reject any
         * malicious packets for non-SIIT environment, than try to do a
-        * partical support for SIIT environment.
+        * partial support for SIIT environment.
         */
        if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
            IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
@@ -556,19 +663,42 @@ ip6_input(m)
        }
 #endif
 
+       /*
+        * Naively assume we can attribute inbound data to the route we would
+        * use to send to this destination. Asymetric routing breaks this
+        * assumption, but it still allows us to account for traffic from
+        * a remote node in the routing table.
+        * this has a very significant performance impact so we bypass
+        * if nstat_collect is disabled. We may also bypass if the
+        * protocol is tcp in the future because tcp will have a route that
+        * we can use to attribute the data to. That does mean we would not
+        * account for forwarded tcp traffic.
+        */
+       if (nstat_collect) {
+               struct rtentry *rte =
+                   ifnet_cached_rtlookup_inet6(m->m_pkthdr.rcvif,
+                   &ip6->ip6_src);
+               if (rte != NULL) {
+                       nstat_route_rx(rte, 1, m->m_pkthdr.len, 0);
+                       rtfree(rte);
+               }
+       }
+
 #if PF
        /* Invoke inbound packet filter */
-       lck_mtx_unlock(ip6_mutex);
-       if (pf_af_hook(m->m_pkthdr.rcvif, NULL, &m, AF_INET6, TRUE) != 0) {
-               if (m != NULL) {
-                       panic("%s: unexpected packet %p\n", __func__, m);
-                       /* NOTREACHED */
+       if (PF_IS_ENABLED) {
+               int error;
+               error = pf_af_hook(m->m_pkthdr.rcvif, NULL, &m, AF_INET6, TRUE);
+               if (error != 0) {
+                       if (m != NULL) {
+                               panic("%s: unexpected packet %p\n", __func__, m);
+                               /* NOTREACHED */
+                       }
+                       /* Already freed by callee */
+                       goto done;
                }
-               /* Already freed by callee */
-               return;
+               ip6 = mtod(m, struct ip6_hdr *);
        }
-       ip6 = mtod(m, struct ip6_hdr *);
-       lck_mtx_lock(ip6_mutex);
 #endif /* PF */
 
        /* drop packets if interface ID portion is already filled */
@@ -592,39 +722,11 @@ ip6_input(m)
                ip6->ip6_dst.s6_addr16[1]
                        = htons(m->m_pkthdr.rcvif->if_index);
 
-#if 0 /* this case seems to be unnecessary. (jinmei, 20010401) */
-       /*
-        * We use rt->rt_ifp to determine if the address is ours or not.
-        * If rt_ifp is lo0, the address is ours.
-        * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
-        * so any address under fe80::%lo0/64 will be mistakenly considered
-        * local.  The special case is supplied to handle the case properly
-        * by actually looking at interface addresses
-        * (using in6ifa_ifpwithaddr).
-        */
-       if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) != 0 &&
-           IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
-           struct in6_ifaddr *ia6;
-               if (!(ia6 = in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst))) {
-                       lck_mtx_unlock(ip6_mutex);
-                       icmp6_error(m, ICMP6_DST_UNREACH,
-                           ICMP6_DST_UNREACH_ADDR, 0);
-                       /* m is already freed */
-                       return;
-               }
-               ifafree(&ia6->ia_ifa);
-
-               ours = 1;
-               deliverifp = m->m_pkthdr.rcvif;
-               goto hbhcheck;
-       }
-#endif
-
        /*
         * Multicast check
         */
        if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
-               struct  in6_multi *in6m = 0;
+               struct  in6_multi *in6m = NULL;
                struct ifnet *ifp = m->m_pkthdr.rcvif;
 
                in6_ifstat_inc(ifp, ifs6_in_mcast);
@@ -632,16 +734,18 @@ ip6_input(m)
                 * See if we belong to the destination multicast group on the
                 * arrival interface.
                 */
-               ifnet_lock_shared(ifp);
-               IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
-               ifnet_lock_done(ifp);
-               if (in6m)
+               in6_multihead_lock_shared();
+               IN6_LOOKUP_MULTI(&ip6->ip6_dst, ifp, in6m);
+               in6_multihead_lock_done();
+               if (in6m != NULL) {
+                       IN6M_REMREF(in6m);
                        ours = 1;
+               }
+               else 
 #if MROUTING
-               else if (!ip6_mrouter) {
-#else
-               else {
+               if (!ip6_mrouter)
 #endif
+               {
                        ip6stat.ip6s_notmember++;
                        ip6stat.ip6s_cantforward++;
                        in6_ifstat_inc(ifp, ifs6_in_discard);
@@ -651,42 +755,18 @@ ip6_input(m)
                goto hbhcheck;
        }
 
-       if (ip6_forward_rt.ro_rt != NULL)
-               RT_LOCK(ip6_forward_rt.ro_rt);
        /*
         *  Unicast check
         */
-       if (ip6_forward_rt.ro_rt != NULL &&
-           (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) &&
-           IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
-           &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr) &&
-           ip6_forward_rt.ro_rt->generation_id == route_generation) {
-               ip6stat.ip6s_forward_cachehit++;
-       } else {
-               struct sockaddr_in6 *dst6;
-
-               if (ip6_forward_rt.ro_rt != NULL) {
-                       /* route is down/stale or destination is different */
-                       ip6stat.ip6s_forward_cachemiss++;
-                       RT_UNLOCK(ip6_forward_rt.ro_rt);
-                       rtfree(ip6_forward_rt.ro_rt);
-                       ip6_forward_rt.ro_rt = NULL;
-               }
-
-               bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
-               dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
-               dst6->sin6_len = sizeof(struct sockaddr_in6);
-               dst6->sin6_family = AF_INET6;
-               dst6->sin6_addr = ip6->ip6_dst;
-#if SCOPEDROUTING
-               ip6_forward_rt.ro_dst.sin6_scope_id =
-                       in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_dst);
-#endif
+       dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
+       dst6->sin6_len = sizeof(struct sockaddr_in6);
+       dst6->sin6_family = AF_INET6;
+       dst6->sin6_addr = ip6->ip6_dst;
 
-               rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
-               if (ip6_forward_rt.ro_rt != NULL)
-                       RT_LOCK(ip6_forward_rt.ro_rt);
-       }
+       rtalloc_scoped_ign((struct route *)&ip6_forward_rt,
+           RTF_PRCLONING, IFSCOPE_NONE);
+       if (ip6_forward_rt.ro_rt != NULL)
+               RT_LOCK(ip6_forward_rt.ro_rt);
 
 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
 
@@ -726,8 +806,7 @@ ip6_input(m)
                                &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
 #endif
            ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
-               struct in6_ifaddr *ia6 =
-                       (struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
+               ia6 = (struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
 
                /*
                 * record address information into m_aux.
@@ -738,31 +817,32 @@ ip6_input(m)
                 * packets to a tentative, duplicated, or somehow invalid
                 * address must not be accepted.
                 */
+               RT_CONVERT_LOCK(ip6_forward_rt.ro_rt);  /* just in case */
+               IFA_LOCK_SPIN(&ia6->ia_ifa);
                if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
+                       IFA_UNLOCK(&ia6->ia_ifa);
                        /* this address is ready */
                        ours = 1;
                        deliverifp = ia6->ia_ifp;       /* correct? */
                        /* Count the packet in the ip address stats */
-#ifndef __APPLE__
 
-                       ia6->ia_ifa.if_ipackets++;
-                       ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
-#endif
                        RT_UNLOCK(ip6_forward_rt.ro_rt);
+                       ia6 = NULL;
                        goto hbhcheck;
-               } else {
-                       RT_UNLOCK(ip6_forward_rt.ro_rt);
-                       /* address is not ready, so discard the packet. */
-                       nd6log((LOG_INFO,
-                           "ip6_input: packet to an unready address %s->%s\n",
-                           ip6_sprintf(&ip6->ip6_src),
-                           ip6_sprintf(&ip6->ip6_dst)));
-                       goto bad;
                }
+               IFA_UNLOCK(&ia6->ia_ifa);
+               RT_UNLOCK(ip6_forward_rt.ro_rt);
+               /* address is not ready, so discard the packet. */
+               nd6log((LOG_INFO,
+                   "ip6_input: packet to an unready address %s->%s\n",
+                   ip6_sprintf(&ip6->ip6_src),
+                   ip6_sprintf(&ip6->ip6_dst)));
+               ia6 = NULL;
+               goto bad;
        }
 
        /*
-        * FAITH(Firewall Aided Internet Translator)
+        * FAITH (Firewall Aided Internet Translator)
         */
 #if defined(NFAITH) && 0 < NFAITH
        if (ip6_keepfaith) {
@@ -796,9 +876,7 @@ ip6_input(m)
         * as our interface address (e.g. multicast addresses, addresses
         * within FAITH prefixes and such).
         */
-       if (deliverifp && !ip6_getdstifaddr(m)) {
-               struct in6_ifaddr *ia6;
-
+       if (deliverifp && (ia6 = ip6_getdstifaddr(m)) == NULL) {
                ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
                if (ia6) {
                        if (!ip6_setdstifaddr(m, ia6)) {
@@ -808,10 +886,16 @@ ip6_input(m)
                                 * to the upper layers.
                                 */
                        }
-                       ifafree(&ia6->ia_ifa);
+                       IFA_REMREF(&ia6->ia_ifa);
+                       ia6 = NULL;
                }
        }
 
+       if (ia6 != NULL) {
+               IFA_REMREF(&ia6->ia_ifa);
+               ia6 = NULL;
+       }
+
        /*
         * Process Hop-by-Hop options header if it's contained.
         * m may be modified in ip6_hopopts_input().
@@ -825,8 +909,7 @@ ip6_input(m)
 #if 0  /*touches NULL pointer*/
                        in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
 #endif
-                       lck_mtx_unlock(ip6_mutex);
-                       return; /* m have already been freed */
+                       goto done;      /* m have already been freed */
                }
 
                /* adjust pointer */
@@ -840,17 +923,16 @@ ip6_input(m)
                if (ip6->ip6_plen == 0 && plen == 0) {
                        /*
                         * Note that if a valid jumbo payload option is
-                        * contained, ip6_hoptops_input() must set a valid
-                        * (non-zero) payload length to the variable plen. 
+                        * contained, ip6_hopopts_input() must set a valid
+                        * (non-zero) payload length to the variable plen.
                         */
                        ip6stat.ip6s_badoptions++;
                        in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
                        in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
-                       lck_mtx_unlock(ip6_mutex);
                        icmp6_error(m, ICMP6_PARAM_PROB,
                                    ICMP6_PARAMPROB_HEADER,
                                    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
-                       return;
+                       goto done;
                }
 #ifndef PULLDOWN_TEST
                /* ip6_hopopts_input() ensures that mbuf is contiguous */
@@ -860,18 +942,31 @@ ip6_input(m)
                        sizeof(struct ip6_hbh));
                if (hbh == NULL) {
                        ip6stat.ip6s_tooshort++;
-                       lck_mtx_unlock(ip6_mutex);
-                       return;
+                       goto done;
                }
 #endif
                nxt = hbh->ip6h_nxt;
 
                /*
-                * accept the packet if a router alert option is included
-                * and we act as an IPv6 router.
+                * If we are acting as a router and the packet contains a
+                * router alert option, see if we know the option value.
+                * Currently, we only support the option value for MLD, in which
+                * case we should pass the packet to the multicast routing
+                * daemon.
                 */
-               if (rtalert != ~0 && ip6_forwarding)
-                       ours = 1;
+               if (rtalert != ~0 && ip6_forwarding) {
+                       switch (rtalert) {
+                       case IP6OPT_RTALERT_MLD:
+                               ours = 1;
+                               break;
+                       default:
+                               /*
+                                * RFC2711 requires unrecognized values must be
+                                * silently ignored.
+                                */
+                               break;
+                       }
+               }
        } else
                nxt = ip6->ip6_nxt;
 
@@ -909,20 +1004,14 @@ ip6_input(m)
 #if MROUTING
                if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
                        ip6stat.ip6s_cantforward++;
-                       m_freem(m);
-                       lck_mtx_unlock(ip6_mutex);
-                       return;
+                       goto bad;
                }
 #endif
-               if (!ours) {
-                       m_freem(m);
-                       lck_mtx_unlock(ip6_mutex);
-                       return;
-               }
+               if (!ours)
+                       goto bad;
        } else if (!ours) {
-               ip6_forward(m, &ip6_forward_rt, 0, 1);
-               lck_mtx_unlock(ip6_mutex);
-               return;
+               ip6_forward(m, &ip6_forward_rt, 0);
+               goto done;
        }       
 
        ip6 = mtod(m, struct ip6_hdr *);
@@ -949,17 +1038,16 @@ ip6_input(m)
        ip6stat.ip6s_delivered++;
        in6_ifstat_inc(deliverifp, ifs6_in_deliver);
 
-       lck_mtx_unlock(ip6_mutex);
 injectit:
        nest = 0;
 
        while (nxt != IPPROTO_DONE) {
                struct ipfilter *filter;
-               int (*pr_input)(struct mbuf **, int *);
+               int (*pr_input)(struct mbuf **, int *, int);
 
                if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
                        ip6stat.ip6s_toomanyhdr++;
-                       goto badunlocked;
+                       goto bad;
                }
 
                /*
@@ -969,24 +1057,9 @@ injectit:
                if (m->m_pkthdr.len < off) {
                        ip6stat.ip6s_tooshort++;
                        in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
-                       goto badunlocked;
-               }
-
-#if 0
-               /*
-                * do we need to do it for every header?  yeah, other
-                * functions can play with it (like re-allocate and copy).
-                */
-               mhist = ip6_addaux(m);
-               if (mhist && M_TRAILINGSPACE(mhist) >= sizeof(nxt)) {
-                       hist = mtod(mhist, caddr_t) + mhist->m_len;
-                       bcopy(&nxt, hist, sizeof(nxt));
-                       mhist->m_len += sizeof(nxt);
-               } else {
-                       ip6stat.ip6s_toomanyhdr++;
                        goto bad;
                }
-#endif
+
 
 #if IPSEC
                /*
@@ -997,7 +1070,7 @@ injectit:
                if ((ipsec_bypass == 0) && (ip6_protox[nxt]->pr_flags & PR_LASTHDR) != 0) {
                        if (ipsec6_in_reject(m, NULL)) {
                                IPSEC_STAT_INCREMENT(ipsec6stat.in_polvio);
-                               goto badunlocked;
+                               goto bad;
                    }
                }
 #endif
@@ -1018,36 +1091,40 @@ injectit:
                                                filter->ipf_filter.cookie, (mbuf_t*)&m, off, nxt);
                                        if (result == EJUSTRETURN) {
                                                ipf_unref();
-                                               return;
+                                               goto done;
                                        }
                                        if (result != 0) {
                                                ipf_unref();
-                                               m_freem(m);
-                                               return;
+                                               goto bad;
                                        }
                                }
                        }
                        ipf_unref();
                }
 
+               DTRACE_IP6(receive, struct mbuf *, m, struct inpcb *, NULL,
+                       struct ip6_hdr *, ip6, struct ifnet *, m->m_pkthdr.rcvif,
+                       struct ip *, NULL, struct ip6_hdr *, ip6);
+
                if ((pr_input = ip6_protox[nxt]->pr_input) == NULL) {
                        m_freem(m);
                        m = NULL;
                        nxt = IPPROTO_DONE;
                } else if (!(ip6_protox[nxt]->pr_flags & PR_PROTOLOCK)) {
                        lck_mtx_lock(inet6_domain_mutex);
-                       nxt = pr_input(&m, &off);
+                       nxt = pr_input(&m, &off, nxt);
                        lck_mtx_unlock(inet6_domain_mutex);
                } else {
-                       nxt = pr_input(&m, &off);
+                       nxt = pr_input(&m, &off, nxt);
                }
        }
+done:
+       if (ip6_forward_rt.ro_rt != NULL)
+               rtfree(ip6_forward_rt.ro_rt);
        return;
  bad:
-       lck_mtx_unlock(ip6_mutex);
- badunlocked:
        m_freem(m);
-       return;
+       goto done;
 }
 
 /*
@@ -1060,8 +1137,13 @@ ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
        struct ip6aux *n;
 
        n = ip6_addaux(m);
-       if (n)
+       if (n != NULL) {
+               if (ia6 != NULL)
+                       IFA_ADDREF(&ia6->ia_ifa);
+               if (n->ip6a_dstia6 != NULL)
+                       IFA_REMREF(&n->ip6a_dstia6->ia_ifa);
                n->ip6a_dstia6 = ia6;
+       }
        return (struct ip6aux *)n;      /* NULL if failed to set */
 }
 
@@ -1072,10 +1154,12 @@ ip6_getdstifaddr(m)
        struct ip6aux *n;
 
        n = ip6_findaux(m);
-       if (n)
-               return n->ip6a_dstia6;
-       else
-               return NULL;
+       if (n != NULL) {
+               if (n->ip6a_dstia6 != NULL)
+                       IFA_ADDREF(&n->ip6a_dstia6->ia_ifa);
+               return (n->ip6a_dstia6);
+       }
+       return (NULL);
 }
 
 /*
@@ -1083,11 +1167,8 @@ ip6_getdstifaddr(m)
  * included, the real payload length will be stored in plenp.
  */
 static int
-ip6_hopopts_input(plenp, rtalertp, mp, offp)
-       u_int32_t *plenp;
-       u_int32_t *rtalertp;    /* XXX: should be stored more smart way */
-       struct mbuf **mp;
-       int *offp;
+ip6_hopopts_input(uint32_t *plenp, uint32_t *rtalertp, struct mbuf **mp,
+    int *offp)
 {
        struct mbuf *m = *mp;
        int off = *offp, hbhlen;
@@ -1123,11 +1204,11 @@ ip6_hopopts_input(plenp, rtalertp, mp, offp)
 
        if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
                                hbhlen, rtalertp, plenp) < 0)
-               return(-1);
+               return (-1);
 
        *offp = off;
        *mp = m;
-       return(0);
+       return (0);
 }
 
 /*
@@ -1167,7 +1248,7 @@ ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
                        }
                        optlen = *(opt + 1) + 2;
                        break;
-               case IP6OPT_RTALERT:
+               case IP6OPT_ROUTER_ALERT:
                        /* XXX may need check for alignment */
                        if (hbhlen < IP6OPT_RTALERT_LEN) {
                                ip6stat.ip6s_toosmall++;
@@ -1175,11 +1256,9 @@ ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
                        }
                        if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
                                /* XXX stat */
-                               lck_mtx_unlock(ip6_mutex);
                                icmp6_error(m, ICMP6_PARAM_PROB,
                                            ICMP6_PARAMPROB_HEADER,
                                            erroff + opt + 1 - opthead);
-                               lck_mtx_lock(ip6_mutex);
                                return(-1);
                        }
                        optlen = IP6OPT_RTALERT_LEN;
@@ -1194,11 +1273,9 @@ ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
                        }
                        if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
                                /* XXX stat */
-                               lck_mtx_unlock(ip6_mutex);
                                icmp6_error(m, ICMP6_PARAM_PROB,
                                            ICMP6_PARAMPROB_HEADER,
                                            erroff + opt + 1 - opthead);
-                               lck_mtx_lock(ip6_mutex);
                                return(-1);
                        }
                        optlen = IP6OPT_JUMBO_LEN;
@@ -1210,11 +1287,9 @@ ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
                        ip6 = mtod(m, struct ip6_hdr *);
                        if (ip6->ip6_plen) {
                                ip6stat.ip6s_badoptions++;
-                               lck_mtx_unlock(ip6_mutex);
                                icmp6_error(m, ICMP6_PARAM_PROB,
                                            ICMP6_PARAMPROB_HEADER,
                                            erroff + opt - opthead);
-                               lck_mtx_lock(ip6_mutex);
                                return(-1);
                        }
 
@@ -1236,11 +1311,9 @@ ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
                         */
                        if (*plenp != 0) {
                                ip6stat.ip6s_badoptions++;
-                               lck_mtx_unlock(ip6_mutex);
                                icmp6_error(m, ICMP6_PARAM_PROB,
                                            ICMP6_PARAMPROB_HEADER,
                                            erroff + opt + 2 - opthead);
-                               lck_mtx_lock(ip6_mutex);
                                return(-1);
                        }
 #endif
@@ -1250,11 +1323,9 @@ ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
                         */
                        if (jumboplen <= IPV6_MAXPACKET) {
                                ip6stat.ip6s_badoptions++;
-                               lck_mtx_unlock(ip6_mutex);
                                icmp6_error(m, ICMP6_PARAM_PROB,
                                            ICMP6_PARAMPROB_HEADER,
                                            erroff + opt + 2 - opthead);
-                               lck_mtx_lock(ip6_mutex);
                                return(-1);
                        }
                        *plenp = jumboplen;
@@ -1266,9 +1337,8 @@ ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
                                goto bad;
                        }
                        optlen = ip6_unknown_opt(opt, m,
-                           erroff + opt - opthead, 1);
+                           erroff + opt - opthead);
                        if (optlen == -1) {
-                               /* ip6_unknown opt unlocked ip6_mutex */
                                return(-1);
                        }
                        optlen += 2;
@@ -1290,11 +1360,7 @@ ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
  * is not continuous in order to return an ICMPv6 error.
  */
 int
-ip6_unknown_opt(optp, m, off, locked)
-       u_int8_t *optp;
-       struct mbuf *m;
-       int off;
-       int locked;
+ip6_unknown_opt(uint8_t *optp, struct mbuf *m, int off)
 {
        struct ip6_hdr *ip6;
 
@@ -1306,11 +1372,7 @@ ip6_unknown_opt(optp, m, off, locked)
                return(-1);
        case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
                ip6stat.ip6s_badoptions++;
-               if (locked)
-                       lck_mtx_unlock(ip6_mutex);
                icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
-               if (locked)
-                       lck_mtx_lock(ip6_mutex);
                return(-1);
        case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
                ip6stat.ip6s_badoptions++;
@@ -1318,109 +1380,147 @@ ip6_unknown_opt(optp, m, off, locked)
                if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
                    (m->m_flags & (M_BCAST|M_MCAST)))
                        m_freem(m);
-               else {
-                       if (locked)
-                               lck_mtx_unlock(ip6_mutex);
+               else
                        icmp6_error(m, ICMP6_PARAM_PROB,
                                    ICMP6_PARAMPROB_OPTION, off);
-                       if (locked)
-                               lck_mtx_lock(ip6_mutex);
-               }
                return(-1);
        }
 
        m_freem(m);             /* XXX: NOTREACHED */
-       return(-1);
+       return (-1);
 }
 
 /*
  * Create the "control" list for this pcb.
- * The function will not modify mbuf chain at all.
+ * These functions will not modify mbuf chain at all.
  *
- * with KAME mbuf chain restriction:
+ * With KAME mbuf chain restriction:
  * The routine will be called from upper layer handlers like tcp6_input().
  * Thus the routine assumes that the caller (tcp6_input) have already
  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
  * very first mbuf on the mbuf chain.
+ *
+ * ip6_savecontrol_v4 will handle those options that are possible to be
+ * set on a v4-mapped socket.
+ * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
+ * options and handle the v6-only ones itself.
  */
-void
-ip6_savecontrol(in6p, mp, ip6, m)
-       struct inpcb *in6p;
-       struct mbuf **mp;
-       struct ip6_hdr *ip6;
-       struct mbuf *m;
+struct mbuf **
+ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
+    int *v4only)
 {
-       int rthdr_exist = 0;
+       struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
 
-#if SO_TIMESTAMP
-       if ((in6p->in6p_socket->so_options & SO_TIMESTAMP) != 0) {
+       if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
                struct timeval tv;
 
                microtime(&tv);
-               *mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
-                                     SCM_TIMESTAMP, SOL_SOCKET);
-               if (*mp) {
-                       mp = &(*mp)->m_next;
-               }
+               mp = sbcreatecontrol_mbuf((caddr_t) &tv, sizeof(tv),
+                   SCM_TIMESTAMP, SOL_SOCKET, mp);
+               if (*mp == NULL) 
+                       return NULL;
        }
-#endif
+        if ((inp->inp_socket->so_options & SO_TIMESTAMP_MONOTONIC) != 0) {
+                uint64_t time;
 
-       /* some OSes call this logic with IPv4 packet, for SO_TIMESTAMP */
-       if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION)
-               return;
+                time = mach_absolute_time();
+                mp = sbcreatecontrol_mbuf((caddr_t) &time, sizeof(time),
+                        SCM_TIMESTAMP_MONOTONIC, SOL_SOCKET, mp);
 
+                       if (*mp == NULL) 
+                               return NULL;
+        }
+       if ((inp->inp_socket->so_flags & SOF_RECV_TRAFFIC_CLASS) != 0) {
+               int tc = m->m_pkthdr.prio;
+               
+               mp = sbcreatecontrol_mbuf((caddr_t) &tc, sizeof(tc),
+                       SO_TRAFFIC_CLASS, SOL_SOCKET, mp);
+               if (*mp == NULL) 
+                       return NULL;
+       }
+
+       if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
+               if (v4only != NULL)
+                       *v4only = 1;
+               return (mp);
+       }
+
+#define IS2292(inp, x, y)      (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
        /* RFC 2292 sec. 5 */
-       if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
+       if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
                struct in6_pktinfo pi6;
+
                bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
-               if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
-                       pi6.ipi6_addr.s6_addr16[1] = 0;
-               pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
-                                       ? m->m_pkthdr.rcvif->if_index
-                                       : 0;
-               *mp = sbcreatecontrol((caddr_t) &pi6,
-                       sizeof(struct in6_pktinfo), IPV6_PKTINFO,
-                       IPPROTO_IPV6);
-               if (*mp)
-                       mp = &(*mp)->m_next;
+               in6_clearscope(&pi6.ipi6_addr); /* XXX */
+               pi6.ipi6_ifindex =
+                   (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
+
+               mp = sbcreatecontrol_mbuf((caddr_t) &pi6,
+                   sizeof(struct in6_pktinfo),
+                   IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6, mp);
+               if (*mp == NULL) 
+                       return NULL;
        }
 
-       if ((in6p->in6p_flags & IN6P_HOPLIMIT) != 0) {
+       if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
                int hlim = ip6->ip6_hlim & 0xff;
-               *mp = sbcreatecontrol((caddr_t) &hlim,
-                       sizeof(int), IPV6_HOPLIMIT, IPPROTO_IPV6);
-               if (*mp)
-                       mp = &(*mp)->m_next;
+
+               mp = sbcreatecontrol_mbuf((caddr_t) &hlim, sizeof(int),
+                   IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
+                   IPPROTO_IPV6, mp);
+               if (*mp == NULL) 
+                       return NULL;
        }
 
-        if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
-                u_int32_t flowinfo;
-                int tclass;
+       if (v4only != NULL)
+               *v4only = 0;
+       return (mp);
+}
+
+int
+ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
+{
+       struct mbuf **np;
+       struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
+       int v4only = 0;
+
+       *mp = NULL;
+       np = ip6_savecontrol_v4(in6p, m, mp, &v4only);
+       if (np == NULL)
+               goto no_mbufs;
 
-                flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
-                flowinfo >>= 20;
+       mp = np;
+       if (v4only)
+               return(0);
 
-                tclass = flowinfo & 0xff;
-                *mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
-                    IPV6_TCLASS, IPPROTO_IPV6);
-                if (*mp)
-                        mp = &(*mp)->m_next;
-        }
+       if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
+               u_int32_t flowinfo;
+               int tclass;
+
+               flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
+               flowinfo >>= 20;
+
+               tclass = flowinfo & 0xff;
+               mp = sbcreatecontrol_mbuf((caddr_t) &tclass, sizeof(tclass),
+                   IPV6_TCLASS, IPPROTO_IPV6, mp);
+               if (*mp == NULL) 
+                       goto no_mbufs;
+       }
 
        /*
         * IPV6_HOPOPTS socket option.  Recall that we required super-user
         * privilege for the option (see ip6_ctloutput), but it might be too
         * strict, since there might be some hop-by-hop options which can be
         * returned to normal user.
-        * See RFC 2292 section 6.
+        * See also RFC 2292 section 6 (or RFC 3542 section 8).
         */
-       if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0) {
+       if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
                /*
                 * Check if a hop-by-hop options header is contatined in the
                 * received packet, and if so, store the options as ancillary
                 * data. Note that a hop-by-hop options header must be
-                * just after the IPv6 header, which fact is assured through
-                * the IPv6 input processing.
+                * just after the IPv6 header, which is assured through the
+                * IPv6 input processing.
                 */
                ip6 = mtod(m, struct ip6_hdr *);
                if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
@@ -1438,67 +1538,38 @@ ip6_savecontrol(in6p, mp, ip6, m)
                            ip6->ip6_nxt);
                        if (ext == NULL) {
                                ip6stat.ip6s_tooshort++;
-                               return;
+                               return(0);
                        }
                        hbh = mtod(ext, struct ip6_hbh *);
                        hbhlen = (hbh->ip6h_len + 1) << 3;
                        if (hbhlen != ext->m_len) {
                                m_freem(ext);
                                ip6stat.ip6s_tooshort++;
-                               return;
+                               return(0);
                        }
 #endif
 
                        /*
-                        * XXX: We copy whole the header even if a jumbo
-                        * payload option is included, which option is to
-                        * be removed before returning in the RFC 2292.
-                        * Note: this constraint is removed in 2292bis.
+                        * XXX: We copy the whole header even if a
+                        * jumbo payload option is included, the option which
+                        * is to be removed before returning according to
+                        * RFC2292.
+                        * Note: this constraint is removed in RFC3542
                         */
-                       *mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
-                                             IPV6_HOPOPTS, IPPROTO_IPV6);
-                       if (*mp)
-                               mp = &(*mp)->m_next;
+                       mp = sbcreatecontrol_mbuf((caddr_t)hbh, hbhlen,
+                           IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
+                           IPPROTO_IPV6, mp);
+
 #if PULLDOWN_TEST
                        m_freem(ext);
 #endif
-               }
-       }
-
-       /* IPV6_DSTOPTS and IPV6_RTHDR socket options */
-       if ((in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
-               int proto, off, nxt;
-
-               /*
-                * go through the header chain to see if a routing header is
-                * contained in the packet. We need this information to store
-                * destination options headers (if any) properly.
-                * XXX: performance issue. We should record this info when
-                * processing extension headers in incoming routine.
-                * (todo) use m_aux? 
-                */
-               proto = IPPROTO_IPV6;
-               off = 0;
-               nxt = -1;
-               while (1) {
-                       int newoff;
-
-                       newoff = ip6_nexthdr(m, off, proto, &nxt);
-                       if (newoff < 0)
-                               break;
-                       if (newoff < off) /* invalid, check for safety */
-                               break;
-                       if ((proto = nxt) == IPPROTO_ROUTING) {
-                               rthdr_exist = 1;
-                               break;
+                       if (*mp == NULL) {
+                               goto no_mbufs;
                        }
-                       off = newoff;
                }
        }
 
-       if ((in6p->in6p_flags &
-            (IN6P_RTHDR | IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) != 0) {
-               ip6 = mtod(m, struct ip6_hdr *);
+       if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
                int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
 
                /*
@@ -1543,7 +1614,7 @@ ip6_savecontrol(in6p, mp, ip6, m)
                        ext = ip6_pullexthdr(m, off, nxt);
                        if (ext == NULL) {
                                ip6stat.ip6s_tooshort++;
-                               return;
+                               return(0);
                        }
                        ip6e = mtod(ext, struct ip6_ext *);
                        if (nxt == IPPROTO_AH)
@@ -1553,30 +1624,39 @@ ip6_savecontrol(in6p, mp, ip6, m)
                        if (elen != ext->m_len) {
                                m_freem(ext);
                                ip6stat.ip6s_tooshort++;
-                               return;
+                               return(0);
                        }
 #endif
 
                        switch (nxt) {
                        case IPPROTO_DSTOPTS:
-                               if ((in6p->in6p_flags & IN6P_DSTOPTS) == 0)
+                               if (!(in6p->inp_flags & IN6P_DSTOPTS))
                                        break;
 
-                               *mp = sbcreatecontrol((caddr_t)ip6e, elen,
-                                                     IPV6_DSTOPTS,
-                                                     IPPROTO_IPV6);
-                               if (*mp)
-                                       mp = &(*mp)->m_next;
+                               mp = sbcreatecontrol_mbuf((caddr_t)ip6e, elen,
+                                   IS2292(in6p,
+                                       IPV6_2292DSTOPTS, IPV6_DSTOPTS),
+                                   IPPROTO_IPV6, mp);
+                                       if (*mp == NULL) {
+#if PULLDOWN_TEST
+                                       m_freem(ext);
+#endif
+                                       goto no_mbufs;
+                               }
                                break;
                        case IPPROTO_ROUTING:
-                               if (!in6p->in6p_flags & IN6P_RTHDR)
+                               if (!in6p->inp_flags & IN6P_RTHDR)
                                        break;
 
-                               *mp = sbcreatecontrol((caddr_t)ip6e, elen,
-                                                     IPV6_RTHDR,
-                                                     IPPROTO_IPV6);
-                               if (*mp)
-                                       mp = &(*mp)->m_next;
+                               mp = sbcreatecontrol_mbuf((caddr_t)ip6e, elen,
+                                   IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
+                                   IPPROTO_IPV6, mp);
+                               if (*mp == NULL) {
+#if PULLDOWN_TEST
+                                       m_freem(ext);
+#endif
+                                       goto no_mbufs;
+                               }
                                break;
                        case IPPROTO_HOPOPTS:
                        case IPPROTO_AH: /* is it possible? */
@@ -1584,7 +1664,7 @@ ip6_savecontrol(in6p, mp, ip6, m)
 
                        default:
                                /*
-                                * other cases have been filtered in the above.
+                                * other cases have been filtered in the above.
                                 * none will visit this case.  here we supply
                                 * the code just in case (nxt overwritten or
                                 * other cases).
@@ -1608,7 +1688,49 @@ ip6_savecontrol(in6p, mp, ip6, m)
          loopend:
                ;
        }
+       return(0);
+no_mbufs:
+       ip6stat.ip6s_pktdropcntrl++;
+       /* XXX increment a stat to show the failure */
+       return(ENOBUFS);
+}
+#undef IS2292
+
+void
+ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
+{
+       struct socket *so;
+       struct mbuf *m_mtu;
+       struct ip6_mtuinfo mtuctl;
+
+       so =  in6p->inp_socket;
 
+       if (mtu == NULL)
+               return;
+
+#ifdef DIAGNOSTIC
+       if (so == NULL)         /* I believe this is impossible */
+               panic("ip6_notify_pmtu: socket is NULL");
+#endif
+
+       bzero(&mtuctl, sizeof(mtuctl)); /* zero-clear for safety */
+       mtuctl.ip6m_mtu = *mtu;
+       mtuctl.ip6m_addr = *dst;
+       if (sa6_recoverscope(&mtuctl.ip6m_addr))
+               return;
+
+       if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
+           IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
+               return;
+
+       if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu, NULL)
+           == 0) {
+               m_freem(m_mtu);
+               /* XXX: should count statistics */
+       } else
+               sorwakeup(so);
+
+       return;
 }
 
 #if PULLDOWN_TEST
@@ -1837,8 +1959,8 @@ ip6_addaux(
        tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_INET6, NULL);
        if (tag == NULL) {
                /* Allocate a tag */
-               tag = m_tag_alloc(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_INET6,
-                   sizeof (struct ip6aux), M_DONTWAIT);
+               tag = m_tag_create(KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_INET6,
+                   sizeof (struct ip6aux), M_DONTWAIT, m);
 
                /* Attach it to the mbuf */
                if (tag) {
@@ -1855,7 +1977,7 @@ ip6_findaux(
 {
        struct m_tag    *tag;
        
-       tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_ENCAP, NULL);
+       tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_INET6, NULL);
        
        return tag ? (struct ip6aux*)(tag + 1) : NULL;
 }
@@ -1866,12 +1988,35 @@ ip6_delaux(
 {
        struct m_tag    *tag;
 
-       tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_ENCAP, NULL);
+       tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID, KERNEL_TAG_TYPE_INET6, NULL);
        if (tag) {
                m_tag_delete(m, tag);
        }
 }
 
+/*
+ * Called by m_tag_free().
+ */
+void
+ip6_destroyaux(struct ip6aux *n)
+{
+       if (n->ip6a_dstia6 != NULL) {
+               IFA_REMREF(&n->ip6a_dstia6->ia_ifa);
+               n->ip6a_dstia6 = NULL;
+       }
+}
+
+/*
+ * Called by m_tag_copy()
+ */
+void
+ip6_copyaux(struct ip6aux *src, struct ip6aux *dst)
+{
+       bcopy(src, dst, sizeof (*dst));
+       if (dst->ip6a_dstia6 != NULL)
+               IFA_ADDREF(&dst->ip6a_dstia6->ia_ifa);
+}
+
 /*
  * System control for IP6
  */