]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/netinet6/ip6_input.c
xnu-2050.7.9.tar.gz
[apple/xnu.git] / bsd / netinet6 / ip6_input.c
index e933fa07d28c819404f3759a68548940645f9fca..9fcc992cdab38469d443f5221a55cfc59fc2c950 100644 (file)
@@ -1,4 +1,32 @@
-/*     $KAME: ip6_input.c,v 1.75 2000/03/28 23:11:05 itojun Exp $      */
+/*
+ * Copyright (c) 2003-2012 Apple Inc. All rights reserved.
+ *
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
+ * 
+ * This file contains Original Code and/or Modifications of Original Code
+ * as defined in and that are subject to the Apple Public Source License
+ * Version 2.0 (the 'License'). You may not use this file except in
+ * compliance with the License. The rights granted to you under the License
+ * may not be used to create, or enable the creation or redistribution of,
+ * unlawful or unlicensed copies of an Apple operating system, or to
+ * circumvent, violate, or enable the circumvention or violation of, any
+ * terms of an Apple operating system software license agreement.
+ * 
+ * Please obtain a copy of the License at
+ * http://www.opensource.apple.com/apsl/ and read it before using this file.
+ * 
+ * The Original Code and all software distributed under the License are
+ * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
+ * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
+ * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
+ * Please see the License for the specific language governing rights and
+ * limitations under the License.
+ * 
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
+ */
+/*     $FreeBSD: src/sys/netinet6/ip6_input.c,v 1.11.2.10 2001/07/24 19:10:18 brooks Exp $     */
+/*     $KAME: ip6_input.c,v 1.194 2001/05/27 13:28:35 itojun Exp $     */
 
 /*
  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
  *
  *     @(#)ip_input.c  8.2 (Berkeley) 1/4/94
  */
-#define _IP_VHL
-#ifdef __FreeBSD__
-#include "opt_ip6fw.h"
-#endif
-#if (defined(__FreeBSD__) && __FreeBSD__ >= 3) || defined(__NetBSD__)
-#include "opt_inet.h"
-#ifdef __NetBSD__      /*XXX*/
-#include "opt_ipsec.h"
-#endif
-#endif
+
 
 #include <sys/param.h>
 #include <sys/systm.h>
 #include <sys/time.h>
 #include <sys/kernel.h>
 #include <sys/syslog.h>
-#if !defined(__bsdi__) && !(defined(__FreeBSD__) && __FreeBSD__ < 3)
+#include <sys/sysctl.h>
 #include <sys/proc.h>
-#endif
+#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_types.h>
 #include <net/if_dl.h>
 #include <net/route.h>
-#include <net/netisr.h>
+#include <net/kpi_protocol.h>
+#include <net/ntstat.h>
+#include <net/init.h>
 
 #include <netinet/in.h>
 #include <netinet/in_systm.h>
 #include <netinet/ip6.h>
 #include <netinet6/in6_var.h>
 #include <netinet6/ip6_var.h>
-#if (defined(__FreeBSD__) && __FreeBSD__ >= 3) || defined(__OpenBSD__) || (defined(__bsdi__) && _BSDI_VERSION >= 199802) || defined (__APPLE__)
 #include <netinet/in_pcb.h>
-#endif
-#if defined(__OpenBSD__) || (defined(__bsdi__) && _BSDI_VERSION >= 199802)
-#include <netinet6/in6_pcb.h>
-#endif
 #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 MIP6
-#include <netinet6/mip6.h>
+#if IPSEC
+#include <netinet6/ipsec.h>
+#if INET6
+#include <netinet6/ipsec6.h>
+#endif
+extern int ipsec_bypass;
 #endif
 
-#if IPV6FIREWALL
 #include <netinet6/ip6_fw.h>
-#endif
+
+#if DUMMYNET
+#include <netinet/ip_fw.h>
+#include <netinet/ip_dummynet.h>
+#endif /* DUMMYNET */
+
+#include <netinet/kpi_ipfilter_var.h>
 
 #include <netinet6/ip6protosw.h>
 
 /* we need it for NLOOP. */
-#ifndef __bsdi__
 #include "loop.h"
-#endif
-#include "faith.h"
-#include "gif.h"
-#include "bpfilter.h"
 
 #include <net/net_osdep.h>
 
+#if PF
+#include <net/pfvar.h>
+#endif /* PF */
+
 extern struct domain inet6domain;
 extern struct ip6protosw inet6sw[];
-#ifdef __bsdi__
-#if _BSDI_VERSION < 199802
-extern struct ifnet loif;
-#else
-extern struct ifnet *loifp;
-#endif
-#endif
 
 struct ip6protosw *  ip6_protox[IPPROTO_MAX];
 static int ip6qmaxlen = IFQ_MAXLEN;
-struct in6_ifaddr *in6_ifaddr;
-#if !(defined(__FreeBSD__) && __FreeBSD__ >= 4)
-struct ifqueue ip6intrq;
-#endif
 
-#if defined(__NetBSD__) || defined(__OpenBSD__)
-extern struct ifnet loif[NLOOP];
-#endif
+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 */
 int ip6_sourcecheck_interval;          /* XXX */
-#if defined(__FreeBSD__) && __FreeBSD__ >= 4
 const int int6intrq_present = 1;
-#endif
 
-#if IPV6FIREWALL
+int ip6_ours_check_algorithm;
+
+#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 */
 ip6_fw_chk_t *ip6_fw_chk_ptr;
 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
+int ip6_fw_enable = 1;
 #endif
 
 struct ip6stat ip6stat;
 
-static void ip6_init2 __P((void *));
+#ifdef __APPLE__
+struct ifqueue ip6intrq;
+decl_lck_mtx_data(, ip6_init_mutex);
+decl_lck_mtx_data(, proxy6_lock);
+decl_lck_mtx_data(, dad6_mutex_data);
+decl_lck_mtx_data(, nd6_mutex_data);
+decl_lck_mtx_data(, prefix6_mutex_data);
+decl_lck_mtx_data(, scope6_mutex_data);
+lck_mtx_t              *dad6_mutex = &dad6_mutex_data;
+lck_mtx_t              *nd6_mutex = &nd6_mutex_data;
+lck_mtx_t              *prefix6_mutex = &prefix6_mutex_data;
+lck_mtx_t              *scope6_mutex = &scope6_mutex_data;
+#ifdef ENABLE_ADDRSEL
+decl_lck_mtx_data(, addrsel_mutex_data);
+lck_mtx_t              *addrsel_mutex = &addrsel_mutex_data;
+#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 int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
+static void ip6_init_delayed(void);
+static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *);
+
+static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
 #if PULLDOWN_TEST
-static struct mbuf *ip6_pullexthdr __P((struct mbuf *, size_t, int));
+static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
 #endif
 
-#if NATPT
-extern int             ip6_protocol_tr;
-
-int    natpt_in6       __P((struct mbuf *, struct mbuf **));
-extern void ip_forward __P((struct mbuf *, int));
+#ifdef __APPLE__
+void gifattach(void);
+void stfattach(void);
 #endif
 
-/* Initialize the PF_INET6 domain, and add in the pre-defined protos */
-void
-in6_dinit()
-{      register int i;
-       register struct ip6protosw *pr;
-       register struct domain *dp;
-       static inet6domain_initted = 0;
-       extern int in6_proto_count; 
-
-       if (!inet6domain_initted)
-       {       
-               dp = &inet6domain;
-
-               for (i=0, pr = &inet6sw[0]; i<in6_proto_count; i++, pr++) {
-                       if (net_add_proto(pr, dp))
-                               printf("in6_dinit: warning net_add_proto failed for pr=%x proto #%d\n", pr, i);
-                       
-               }
-               inet6domain_initted = 1;
-       }
-}
+SYSCTL_DECL(_net_inet6_ip6);
 
-#ifdef MIP6
-int (*mip6_new_packet_hook)(struct mbuf *m) = 0;
-int (*mip6_route_optimize_hook)(struct mbuf *m) = 0;
-#endif
+int    ip6_doscopedroute = 1;
+SYSCTL_INT(_net_inet6_ip6, OID_AUTO, scopedroute, CTLFLAG_RD | CTLFLAG_LOCKED,
+     &ip6_doscopedroute, 0, "Enable IPv6 scoped routing");
+
+int    ip6_restrictrecvif = 1;
+SYSCTL_INT(_net_inet6_ip6, OID_AUTO, restrictrecvif,
+    CTLFLAG_RW | CTLFLAG_LOCKED, &ip6_restrictrecvif, 0,
+    "Enable inbound interface restrictions");
+
+/*
+ * On platforms which require strict alignment (currently for anything but
+ * i386 or x86_64), check if the IP header pointer is 32-bit aligned; if not,
+ * copy the contents of the mbuf chain into a new chain, and free the original
+ * one.  Create some head room in the first mbuf of the new chain, in case
+ * it's needed later on.
+ *
+ * RFC 2460 says that IPv6 headers are 64-bit aligned, but network interfaces
+ * mostly align to 32-bit boundaries.  Care should be taken never to use 64-bit
+ * load/store operations on the fields in IPv6 headers.
+ */
+#if defined(__i386__) || defined(__x86_64__)
+#define        IP6_HDR_ALIGNMENT_FIXUP(_m, _ifp, _action) do { } while (0)
+#else /* !__i386__ && !__x86_64__ */
+#define        IP6_HDR_ALIGNMENT_FIXUP(_m, _ifp, _action) do {                 \
+       if (!IP6_HDR_ALIGNED_P(mtod(_m, caddr_t))) {                    \
+               struct mbuf *_n;                                        \
+               struct ifnet *__ifp = (_ifp);                           \
+               atomic_add_64(&(__ifp)->if_alignerrs, 1);               \
+               if (((_m)->m_flags & M_PKTHDR) &&                       \
+                   (_m)->m_pkthdr.header != NULL)                      \
+                       (_m)->m_pkthdr.header = NULL;                   \
+               _n = m_defrag_offset(_m, max_linkhdr, M_NOWAIT);        \
+               if (_n == NULL) {                                       \
+                       ip6stat.ip6s_toosmall++;                        \
+                       m_freem(_m);                                    \
+                       (_m) = NULL;                                    \
+                       _action                                         \
+               } else {                                                \
+                       VERIFY(_n != (_m));                             \
+                       (_m) = _n;                                      \
+               }                                                       \
+       }                                                               \
+} while (0)
+#endif /* !__i386__ && !__x86_64__ */
+
+static void
+ip6_proto_input(
+       __unused protocol_family_t      protocol,
+       mbuf_t                          packet)
+{
+       ip6_input(packet);
+}
 
 /*
  * IP6 initialization: fill in IP6 protocol switch table.
@@ -219,114 +303,170 @@ int (*mip6_route_optimize_hook)(struct mbuf *m) = 0;
 void
 ip6_init()
 {
-       register struct protosw *pr;
-       register int i;
+       struct ip6protosw *pr;
+       int i;
        struct timeval tv;
 
-       pr = (struct protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
+       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)");
+#endif
+       pr = (struct ip6protosw *)pffindproto_locked(PF_INET6, IPPROTO_RAW, SOCK_RAW);
        if (pr == 0)
                panic("ip6_init");
        for (i = 0; i < IPPROTO_MAX; i++)
                ip6_protox[i] = pr;
-       for (pr = inet6domain.dom_protosw; pr; pr = pr->pr_next) {
-               if(!((unsigned int)pr->pr_domain)) continue;    /* If uninitialized, skip */
+       for (pr = (struct ip6protosw*)inet6domain.dom_protosw; pr; pr = pr->pr_next) {
+               if(!(pr->pr_domain)) continue;    /* If uninitialized, skip */
                if (pr->pr_domain->dom_family == PF_INET6 &&
                    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
                        ip6_protox[pr->pr_protocol] = pr;
                }
        }
 
+       ip6_mutex_grp_attr  = lck_grp_attr_alloc_init();
+
+       ip6_mutex_grp = lck_grp_alloc_init("ip6", ip6_mutex_grp_attr);
+       ip6_mutex_attr = lck_attr_alloc_init();
+
+       lck_mtx_init(dad6_mutex, ip6_mutex_grp, ip6_mutex_attr);
+       lck_mtx_init(nd6_mutex, ip6_mutex_grp, ip6_mutex_attr);
+       lck_mtx_init(prefix6_mutex, ip6_mutex_grp, ip6_mutex_attr);
+       lck_mtx_init(scope6_mutex, ip6_mutex_grp, ip6_mutex_attr);
+
+#ifdef ENABLE_ADDRSEL
+       lck_mtx_init(addrsel_mutex, ip6_mutex_grp, ip6_mutex_attr);
+#endif
+
+       lck_mtx_init(&proxy6_lock, ip6_mutex_grp, ip6_mutex_attr);
+       lck_mtx_init(&ip6_init_mutex, ip6_mutex_grp, ip6_mutex_attr);
+
+       lck_rw_init(&in6_ifs_rwlock, ip6_mutex_grp, ip6_mutex_attr);
+       lck_rw_init(&icmp6_ifs_rwlock, 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();
-#if IPV6FIREWALL
-       ip6_fw_init();
-#endif
+       icmp6_init();
+       addrsel_policy_init();
        /*
         * in many cases, random() here does NOT return random number
         * as initialization during bootstrap time occur in fixed order.
         */
        microtime(&tv);
        ip6_flow_seq = random() ^ tv.tv_usec;
-       timeout(ip6_init2, (caddr_t)0, 6 * hz);
-}
-
-static void
-ip6_init2(dummy)
-       void *dummy;
-{
-       int ret;
-#if defined(__bsdi__) && _BSDI_VERSION < 199802
-       struct ifnet *loifp = &loif;
-#endif
-#ifdef __APPLE__
-       boolean_t   funnel_state;
-       funnel_state = thread_set_funneled(TRUE);
-#endif
-
-       /* get EUI64 from somewhere */
-       ret = in6_ifattach_getifid(NULL);
+       microtime(&tv);
+       ip6_desync_factor = (random() ^ tv.tv_usec) % MAX_TEMP_DESYNC_FACTOR;
 
        /*
-        * to route local address of p2p link to loopback,
-        * assign loopback address first.
+        * P2P interfaces often route the local address to the loopback
+        * interface. At this point, lo0 hasn't been initialized yet, which
+        * means that we need to delay the IPv6 configuration of lo0.
         */
-       in6_ifattach(&loif[0], IN6_IFT_LOOP, NULL, 0);
+       net_init_add(ip6_init_delayed);
 
-#if MIP6
-       /* Initialize the Mobile IPv6 code */
-       mip6_init();
-#endif
+       domain_proto_mtx_unlock(TRUE);
+       proto_register_input(PF_INET6, ip6_proto_input, NULL, 0);
+       domain_proto_mtx_lock();
+}
 
-#import <gif.h>
-#if NGIF > 0
-      gifattach();
-#endif
-#import <faith.h>
-#if NFAITH > 0
-      faithattach();
-#endif
+static void
+ip6_init_delayed(void)
+{
+       (void) in6_ifattach(lo_ifp, NULL, NULL);
 
+#ifdef __APPLE__
        /* nd6_timer_init */
        timeout(nd6_timer, (caddr_t)0, hz);
-       /* router renumbering prefix list maintenance */
-       timeout(in6_rr_timer, (caddr_t)0, hz);
-#ifdef __APPLE__
-       (void) thread_set_funneled(funnel_state);
-#endif
-}
 
-#if __FreeBSD__
-/* cheat */
-SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip6_init2, NULL);
-#endif
+       /* timer for regeneranation of temporary addresses randomize ID */
+       timeout(in6_tmpaddrtimer, (caddr_t)0,
+               (ip6_temp_preferred_lifetime - ip6_desync_factor -
+                      ip6_temp_regen_advance) * hz);
 
-/*
- * IP6 input interrupt handling. Just pass the packet to ip6_input.
- */
-void
-ip6intr(void)
-{
-       int s;
-       struct mbuf *m;
-
-       for (;;) {
-               s = splimp();
-               IF_DEQUEUE(&ip6intrq, m);
-               splx(s);
-               if (m == 0)
-                       return;
-               ip6_input(m);
-       }
+#if NGIF
+       gifattach();
+#endif
+#if NSTF
+       stfattach();
+#endif
+#endif /* __APPLE__ */
 }
 
-NETISR_SET(NETISR_IPV6, ip6intr);
-
-extern struct  route_in6 ip6_forward_rt;
-
 void
-ip6_input(m)
-       struct mbuf *m;
+ip6_input(struct mbuf *m)
 {
        struct ip6_hdr *ip6;
        int off = sizeof(struct ip6_hdr), nest;
@@ -334,9 +474,59 @@ ip6_input(m)
        u_int32_t rtalert = ~0;
        int nxt = 0, ours = 0;
        struct ifnet *deliverifp = NULL;
-#if defined(__bsdi__) && _BSDI_VERSION < 199802
-       struct ifnet *loifp = &loif;
-#endif
+       ipfilter_t inject_ipfref = 0;
+       int seen;
+       struct in6_ifaddr *ia6 = NULL;
+       struct route_in6 ip6_forward_rt;
+       struct sockaddr_in6 *dst6;
+#if DUMMYNET
+       struct m_tag    *tag;
+       struct ip_fw_args args;
+               
+       bzero(&args, sizeof(struct ip_fw_args));
+#endif /* DUMMYNET */
+
+       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);
+
+       /* Perform IP header alignment fixup, if needed */
+       IP6_HDR_ALIGNMENT_FIXUP(m, m->m_pkthdr.rcvif, return;);
+
+#if DUMMYNET
+       if ((tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
+           KERNEL_TAG_TYPE_DUMMYNET, NULL)) != NULL) {
+               struct dn_pkt_tag       *dn_tag;
+
+               dn_tag = (struct dn_pkt_tag *)(tag+1);
+               
+               args.fwa_pf_rule = dn_tag->dn_pf_rule;
+               
+               m_tag_delete(m, tag);
+       }
+       
+       if (args.fwa_pf_rule) {
+               ip6 = mtod(m, struct ip6_hdr *); /* In case PF got disabled */
+
+               goto check_with_pf;
+       }
+#endif /* DUMMYNET */
+
+       /*
+        * No need to proccess packet twice if we've 
+        * already seen it
+        */
+       inject_ipfref = ipf_get_inject_filter(m);
+       if (inject_ipfref != 0) {
+               ip6 = mtod(m, struct ip6_hdr *);
+               nxt = ip6->ip6_nxt;
+               seen = 0;
+               goto injectit;
+       } else
+               seen = 1;
 
 #if IPSEC
        /*
@@ -350,7 +540,12 @@ ip6_input(m)
 #endif
 
        /*
-        * mbuf statistics by kazu
+        * make sure we don't have onion peering information into m_aux.
+        */
+       ip6_delaux(m);
+
+       /*
+        * mbuf statistics
         */
        if (m->m_flags & M_EXT) {
                if (m->m_next)
@@ -358,23 +553,64 @@ ip6_input(m)
                else
                        ip6stat.ip6s_mext1++;
        } else {
+#define M2MMAX (sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
                if (m->m_next) {
-                       if (m->m_flags & M_LOOP)
-                               ip6stat.ip6s_m2m[loif[0].if_index]++;   /*XXX*/
-                       else if (m->m_pkthdr.rcvif->if_index <= 31)
+                       if (m->m_flags & M_LOOP) {
+                               ip6stat.ip6s_m2m[ifnet_index(lo_ifp)]++;        /* XXX */
+                       } else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
                                ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
                        else
                                ip6stat.ip6s_m2m[0]++;
                } else
                        ip6stat.ip6s_m1++;
+#undef M2MMAX
+       }
+
+       /*
+        * Drop the packet if IPv6 operation is disabled on the IF;
+        * accessing the flag is done without acquiring nd_ifinfo lock
+        * for performance reasons.
+        */
+       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++;
 
 #ifndef PULLDOWN_TEST
-       /* XXX is the line really necessary? */
-       IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /*nothing*/);
+       /*
+        * L2 bridge code and some other code can return mbuf chain
+        * that does not conform to KAME requirement.  too bad.
+        * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
+        */
+       if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
+               struct mbuf *n;
+
+               MGETHDR(n, M_DONTWAIT, MT_HEADER);      /* MAC-OK */
+               if (n)
+                       M_COPY_PKTHDR(n, m);
+               if (n && m->m_pkthdr.len > MHLEN) {
+                       MCLGET(n, M_DONTWAIT);
+                       if ((n->m_flags & M_EXT) == 0) {
+                               m_freem(n);
+                               n = NULL;
+                       }
+               }
+               if (n == NULL)
+                       goto bad;
+
+               m_copydata(m, 0, m->m_pkthdr.len, mtod(n, caddr_t));
+               n->m_len = m->m_pkthdr.len;
+               m_freem(m);
+               m = n;
+       }
+       IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr),
+               {goto done;}); 
 #endif
 
        if (m->m_len < sizeof(struct ip6_hdr)) {
@@ -383,7 +619,7 @@ ip6_input(m)
                if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
                        ip6stat.ip6s_toosmall++;
                        in6_ifstat_inc(inifp, ifs6_in_hdrerr);
-                       return;
+                       goto done;
                }
        }
 
@@ -397,11 +633,11 @@ ip6_input(m)
 
        ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
 
-#if IPV6FIREWALL
+#if IPFW2
        /*
         * Check with the firewall...
         */
-       if (ip6_fw_chk_ptr) {
+       if (ip6_fw_enable && ip6_fw_chk_ptr) {
                u_short port = 0;
                /* If ipfw says divert, we have to just drop packet */
                /* use port as a dummy argument */
@@ -410,25 +646,53 @@ ip6_input(m)
                        m = NULL;
                }
                if (!m)
-                       return;
+                       goto done;
        }
 #endif
 
        /*
-        * Scope check
+        * Check against address spoofing/corruption.
         */
        if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
            IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
+               /*
+                * XXX: "badscope" is not very suitable for a multicast source.
+                */
+               ip6stat.ip6s_badscope++;
+               in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
+               goto bad;
+       }
+       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;
        }
 
        /*
-        * Don't check IPv4 mapped address here.  SIIT assumes that
-        * routers would forward IPv6 native packets with IPv4 mapped
-        * address normally.
+        * 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.
+        *
+        * 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
+        * partial support for SIIT environment.
         */
+       if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
+           IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
+               ip6stat.ip6s_badscope++;
+               in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
+               goto bad;
+       }
 #if 0
        /*
         * Reject packets with IPv4 compatible addresses (auto tunnel).
@@ -444,79 +708,117 @@ ip6_input(m)
                goto bad;
        }
 #endif
-       if (IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
-           IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) {
-               if (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) {
-                       ours = 1;
-                       deliverifp = m->m_pkthdr.rcvif;
-                       goto hbhcheck;
-               } else {
-                       ip6stat.ip6s_badscope++;
-                       in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
-                       goto bad;
+
+       /*
+        * 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 (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) {
-               if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
-                       ours = 1;
-                       deliverifp = m->m_pkthdr.rcvif;
-                       goto hbhcheck;
+#if DUMMYNET
+check_with_pf:
+#endif
+#if PF
+       /* Invoke inbound packet filter */
+       if (PF_IS_ENABLED) {
+               int error;
+#if DUMMYNET
+               error = pf_af_hook(m->m_pkthdr.rcvif, NULL, &m, AF_INET6, TRUE, &args);
+#else
+               error = pf_af_hook(m->m_pkthdr.rcvif, NULL, &m, AF_INET6, TRUE, NULL);
+#endif
+               if (error != 0 || m == NULL) {
+                       if (m != NULL) {
+                               panic("%s: unexpected packet %p\n", __func__, m);
+                               /* NOTREACHED */
+                       }
+                       /* Already freed by callee */
+                       goto done;
                }
-       } else {
-               if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
-                       ip6->ip6_src.s6_addr16[1]
-                               = htons(m->m_pkthdr.rcvif->if_index);
-               if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
-                       ip6->ip6_dst.s6_addr16[1]
-                               = htons(m->m_pkthdr.rcvif->if_index);
+               ip6 = mtod(m, struct ip6_hdr *);
        }
+#endif /* PF */
+
+       /* drop packets if interface ID portion is already filled */
+       if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
+               if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src) &&
+                   ip6->ip6_src.s6_addr16[1]) {
+                       ip6stat.ip6s_badscope++;
+                       goto bad;
+               }
+               if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst) &&
+                   ip6->ip6_dst.s6_addr16[1]) {
+                       ip6stat.ip6s_badscope++;
+                       goto bad;
+               }
+       }
+
+       if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
+               ip6->ip6_src.s6_addr16[1]
+                       = htons(m->m_pkthdr.rcvif->if_index);
+       if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
+               ip6->ip6_dst.s6_addr16[1]
+                       = htons(m->m_pkthdr.rcvif->if_index);
 
        /*
         * 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(m->m_pkthdr.rcvif, ifs6_in_mcast);
+               in6_ifstat_inc(ifp, ifs6_in_mcast);
                /*
                 * See if we belong to the destination multicast group on the
                 * arrival interface.
                 */
-               IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
-               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 (!ip6_mrouter) {
+               } else if (!nd6_prproxy
+#if MROUTING
+                   && !ip6_mrouter
+#endif
+                   ) {
                        ip6stat.ip6s_notmember++;
                        ip6stat.ip6s_cantforward++;
-                       in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
+                       in6_ifstat_inc(ifp, ifs6_in_discard);
                        goto bad;
                }
-               deliverifp = m->m_pkthdr.rcvif;
+               deliverifp = ifp;
                goto hbhcheck;
        }
 
        /*
         *  Unicast check
         */
-       if (ip6_forward_rt.ro_rt == 0 ||
-           !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
-                               &ip6_forward_rt.ro_dst.sin6_addr)) {
-               if (ip6_forward_rt.ro_rt) {
-                       RTFREE(ip6_forward_rt.ro_rt);
-                       ip6_forward_rt.ro_rt = 0;
-               }
-               bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
-               ip6_forward_rt.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
-               ip6_forward_rt.ro_dst.sin6_family = AF_INET6;
-               ip6_forward_rt.ro_dst.sin6_addr = ip6->ip6_dst;
+       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 __FreeBSD__ || defined(__APPLE__)
-               rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
-#else
-               rtalloc((struct route *)&ip6_forward_rt);
-#endif
-       }
+       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))
 
@@ -528,10 +830,24 @@ ip6_input(m)
         * route to the loopback interface for the destination of the packet.
         * But we think it's even useful in some situations, e.g. when using
         * a special daemon which wants to intercept the packet.
+        *
+        * XXX: some OSes automatically make a cloned route for the destination
+        * of an outgoing packet.  If the outgoing interface of the packet
+        * is a loopback one, the kernel would consider the packet to be
+        * accepted, even if we have no such address assinged on the interface.
+        * We check the cloned flag of the route entry to reject such cases,
+        * assuming that route entries for our own addresses are not made by
+        * cloning (it should be true because in6_addloop explicitly installs
+        * the host route).  However, we might have to do an explicit check
+        * while it would be less efficient.  Or, should we rather install a
+        * reject route for such a case?
         */
-       if (ip6_forward_rt.ro_rt &&
+       if (ip6_forward_rt.ro_rt != NULL &&
            (ip6_forward_rt.ro_rt->rt_flags &
             (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
+#if RTF_WASCLONED
+           !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
+#endif
 #if 0
            /*
             * The check below is redundant since the comparison of
@@ -539,91 +855,46 @@ ip6_input(m)
             * already done through looking up the routing table.
             */
            IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
-                              &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr) &&
+                               &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;
-               /* packet to tentative address must not be received */
-               if (ia6->ia6_flags & IN6_IFF_ANYCAST)
-                       m->m_flags |= M_ANYCAST6;
+               ia6 = (struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
+
+               /*
+                * record address information into m_aux.
+                */
+               (void)ip6_setdstifaddr(m, ia6);
+
+               /*
+                * 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)) {
-                       /* this interface is ready */
+                       IFA_UNLOCK(&ia6->ia_ifa);
+                       /* this address is ready */
                        ours = 1;
                        deliverifp = ia6->ia_ifp;       /* correct? */
-                       goto hbhcheck;
-               } else {
-                       /* this interface is not ready, fall through */
-               }
-       }
+                       /* Count the packet in the ip address stats */
 
-       /*
-        * FAITH(Firewall Aided Internet Translator)
-        */
-#if defined(NFAITH) && 0 < NFAITH
-       if (ip6_keepfaith) {
-               if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
-                && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
-                       /* XXX do we need more sanity checks? */
-                       ours = 1;
-                       deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /*faith*/
+                       RT_UNLOCK(ip6_forward_rt.ro_rt);
+                       ia6 = NULL;
                        goto hbhcheck;
                }
-       }
-#endif
-
-#if NATPT
-       /*
-        * NAT-PT (Network Address Translation - Protocol Translation)
-        */
-       if (ip6_protocol_tr)
-       {
-           struct mbuf *m1 = NULL;
-
-           switch (natpt_in6(m, &m1))
-           {
-             case IPPROTO_IP:                                  goto processpacket;
-             case IPPROTO_IPV4:        ip_forward(m1, 0);      break;
-             case IPPROTO_IPV6:        ip6_forward(m1, 0);     break;
-             case IPPROTO_MAX:                 /* discard this packet  */
-             default:                                          break;
-
-             case IPPROTO_DONE:                /* discard without free */
-               return;
-           }
-
-           if (m != m1)
-               m_freem(m);
-
-           return;
+               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;
        }
 
-  processpacket:
-#endif
-
-#if 0
-    {
-       /*
-        * Last resort: check in6_ifaddr for incoming interface.
-        * The code is here until I update the "goto ours hack" code above
-        * working right.
-        */
-       struct ifaddr *ifa;
-       for (ifa = m->m_pkthdr.rcvif->if_addrlist.tqh_first;
-            ifa;
-            ifa = ifa->ifa_list.tqe_next) {
-               if (ifa->ifa_addr == NULL)
-                       continue;       /* just for safety */
-               if (ifa->ifa_addr->sa_family != AF_INET6)
-                       continue;
-               if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ip6->ip6_dst)) {
-                       ours = 1;
-                       deliverifp = ifa->ifa_ifp;
-                       goto hbhcheck;
-               }
-       }
-    }
-#endif
+       if (ip6_forward_rt.ro_rt != NULL)
+               RT_UNLOCK(ip6_forward_rt.ro_rt);
 
        /*
         * Now there is no reason to process the packet if it's not our own
@@ -636,6 +907,31 @@ ip6_input(m)
        }
 
   hbhcheck:
+       /*
+        * record address information into m_aux, if we don't have one yet.
+        * note that we are unable to record it, if the address is not listed
+        * as our interface address (e.g. multicast addresses, etc.)
+        */
+       if (deliverifp && (ia6 = ip6_getdstifaddr(m)) == NULL) {
+               ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
+               if (ia6) {
+                       if (!ip6_setdstifaddr(m, ia6)) {
+                               /*
+                                * XXX maybe we should drop the packet here,
+                                * as we could not provide enough information
+                                * to the upper layers.
+                                */
+                       }
+                       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().
@@ -649,10 +945,31 @@ ip6_input(m)
 #if 0  /*touches NULL pointer*/
                        in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
 #endif
-                       return; /* m have already been freed */
+                       goto done;      /* m have already been freed */
                }
+
                /* adjust pointer */
                ip6 = mtod(m, struct ip6_hdr *);
+
+               /*
+                * if the payload length field is 0 and the next header field
+                * indicates Hop-by-Hop Options header, then a Jumbo Payload
+                * option MUST be included.
+                */
+               if (ip6->ip6_plen == 0 && plen == 0) {
+                       /*
+                        * Note that if a valid jumbo payload option is
+                        * 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);
+                       icmp6_error(m, ICMP6_PARAM_PROB,
+                                   ICMP6_PARAMPROB_HEADER,
+                                   (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
+                       goto done;
+               }
 #ifndef PULLDOWN_TEST
                /* ip6_hopopts_input() ensures that mbuf is contiguous */
                hbh = (struct ip6_hbh *)(ip6 + 1);
@@ -661,17 +978,31 @@ ip6_input(m)
                        sizeof(struct ip6_hbh));
                if (hbh == NULL) {
                        ip6stat.ip6s_tooshort++;
-                       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;
 
@@ -706,19 +1037,46 @@ ip6_input(m)
                 * ip6_mforward() returns a non-zero value, the packet
                 * must be discarded, else it may be accepted below.
                 */
+#if MROUTING
                if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
                        ip6stat.ip6s_cantforward++;
-                       m_freem(m);
-                       return;
+                       goto bad;
                }
-               if (!ours) {
-                       m_freem(m);
-                       return;
+#endif
+               if (!ours && nd6_prproxy) {
+                       /*
+                        * If this isn't for us, this might be a Neighbor
+                        * Solicitation (dst is solicited-node multicast)
+                        * against an address in one of the proxied prefixes;
+                        * if so, claim the packet and let icmp6_input()
+                        * handle the rest.
+                        */
+                       ours = nd6_prproxy_isours(m, ip6, NULL, IFSCOPE_NONE);
+                       VERIFY(!ours ||
+                           (m->m_pkthdr.aux_flags & MAUXF_PROXY_DST));
                }
+               if (!ours)
+                       goto bad;
        } else if (!ours) {
-               ip6_forward(m, 0);
-               return;
-       }       
+               /*
+                * The unicast forwarding function might return the packet
+                * if we are proxying prefix(es), and if the packet is an
+                * ICMPv6 packet that has failed the zone checks, but is
+                * targetted towards a proxied address (this is optimized by
+                * way of RTF_PROXY test.)  If so, claim the packet as ours
+                * and let icmp6_input() handle the rest.  The packet's hop
+                * limit value is kept intact (it's not decremented).  This
+                * is for supporting Neighbor Unreachability Detection between
+                * proxied nodes on different links (src is link-local, dst
+                * is target address.)
+                */
+               if ((m = ip6_forward(m, &ip6_forward_rt, 0)) == NULL)
+                       goto done;
+               VERIFY(ip6_forward_rt.ro_rt != NULL);
+               VERIFY(m->m_pkthdr.aux_flags & MAUXF_PROXY_DST);
+               deliverifp = ip6_forward_rt.ro_rt->rt_ifp;
+               ours = 1;
+       }
 
        ip6 = mtod(m, struct ip6_hdr *);
 
@@ -741,31 +1099,23 @@ ip6_input(m)
        /*
         * Tell launch routine the next header
         */
-#if defined(__NetBSD__) && defined(IFA_STATS)
-       if (IFA_STATS && deliverifp != NULL) {
-               struct in6_ifaddr *ia6;
-               ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
-               if (ia6)
-                       ia6->ia_ifa.ifa_data.ifad_inbytes += m->m_pkthdr.len;
-       }
-#endif
        ip6stat.ip6s_delivered++;
        in6_ifstat_inc(deliverifp, ifs6_in_deliver);
+
+injectit:
        nest = 0;
 
-#if MIP6
        /*
-        * Mobile IPv6
-        *
-        * Assume that the received packet shall be processed by MIPv6 when
-        * the destination header has been taken care of. Because of this,
-        * some flags have to be reset for later evaluation.
+        * Perform IP header alignment fixup again, if needed.  Note that
+        * we do it once for the outermost protocol, and we assume each
+        * protocol handler wouldn't mess with the alignment afterwards.
         */
-       if (mip6_new_packet_hook)
-               (*mip6_new_packet_hook)(m);
-#endif /* MIP6 */
+       IP6_HDR_ALIGNMENT_FIXUP(m, m->m_pkthdr.rcvif, return;);
 
        while (nxt != IPPROTO_DONE) {
+               struct ipfilter *filter;
+               int (*pr_input)(struct mbuf **, int *, int);
+
                if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
                        ip6stat.ip6s_toomanyhdr++;
                        goto bad;
@@ -775,25 +1125,115 @@ ip6_input(m)
                 * protection against faulty packet - there should be
                 * more sanity checks in header chain processing.
                 */
-               if (m->m_pkthdr.len == 0 || m->m_pkthdr.len < off) {
+               if (m->m_pkthdr.len < off) {
                        ip6stat.ip6s_tooshort++;
                        in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
                        goto bad;
                }
-               
-#if MIP6
-               if ((nxt != IPPROTO_HOPOPTS) && (nxt != IPPROTO_DSTOPTS) &&
-                   (nxt != IPPROTO_ROUTING) && (nxt != IPPROTO_FRAGMENT) &&
-                   (nxt != IPPROTO_ESP) && (nxt != IPPROTO_AH)) {
-                       if (mip6_route_optimize_hook)
-                               (*mip6_route_optimize_hook)(m);
+
+
+#if IPSEC
+               /*
+                * enforce IPsec policy checking if we are seeing last header.
+                * note that we do not visit this with protocols with pcb layer
+                * code - like udp/tcp/raw ip.
+                */
+               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 bad;
+                   }
+               }
+#endif
+
+               /*
+                * Call IP filter
+                */
+               if (!TAILQ_EMPTY(&ipv6_filters)) {
+                       ipf_ref();
+                       TAILQ_FOREACH(filter, &ipv6_filters, ipf_link) {
+                               if (seen == 0) {
+                                       if ((struct ipfilter *)inject_ipfref ==
+                                           filter)
+                                               seen = 1;
+                               } else if (filter->ipf_filter.ipf_input) {
+                                       errno_t result;
+
+                                       result = filter->ipf_filter.ipf_input(
+                                               filter->ipf_filter.cookie,
+                                               (mbuf_t *)&m, off, nxt);
+                                       if (result == EJUSTRETURN) {
+                                               ipf_unref();
+                                               goto done;
+                                       }
+                                       if (result != 0) {
+                                               ipf_unref();
+                                               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);
+                       lck_mtx_unlock(inet6_domain_mutex);
+               } else {
+                       nxt = pr_input(&m, &off, nxt);
                }
-#endif
-               nxt = (*ip6_protox[nxt]->pr_input)(&m, &off, nxt);
        }
+done:
+       if (ip6_forward_rt.ro_rt != NULL)
+               rtfree(ip6_forward_rt.ro_rt);
        return;
  bad:
        m_freem(m);
+       goto done;
+}
+
+/*
+ * set/grab in6_ifaddr correspond to IPv6 destination address.
+ * XXX backward compatibility wrapper
+ */
+static struct ip6aux *
+ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
+{
+       struct ip6aux *n;
+
+       n = ip6_addaux(m);
+       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 */
+}
+
+struct in6_ifaddr *
+ip6_getdstifaddr(m)
+       struct mbuf *m;
+{
+       struct ip6aux *n;
+
+       n = ip6_findaux(m);
+       if (n != NULL) {
+               if (n->ip6a_dstia6 != NULL)
+                       IFA_ADDREF(&n->ip6a_dstia6->ia_ifa);
+               return (n->ip6a_dstia6);
+       }
+       return (NULL);
 }
 
 /*
@@ -801,24 +1241,21 @@ ip6_input(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)
 {
-       register struct mbuf *m = *mp;
+       struct mbuf *m = *mp;
        int off = *offp, hbhlen;
        struct ip6_hbh *hbh;
        u_int8_t *opt;
 
        /* validation of the length of the header */
 #ifndef PULLDOWN_TEST
-       IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
+       IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), return -1);
        hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
        hbhlen = (hbh->ip6h_len + 1) << 3;
 
-       IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
+       IP6_EXTHDR_CHECK(m, off, hbhlen, return -1);
        hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
 #else
        IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
@@ -841,11 +1278,11 @@ 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);
 }
 
 /*
@@ -853,6 +1290,10 @@ ip6_hopopts_input(plenp, rtalertp, mp, offp)
  * This function is separate from ip6_hopopts_input() in order to
  * handle a case where the sending node itself process its hop-by-hop
  * options header. In such a case, the function is called from ip6_output().
+ *
+ * The function assumes that hbh header is located right after the IPv6 header
+ * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
+ * opthead + hbhlen is located in continuous memory region.
  */
 int
 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
@@ -866,103 +1307,122 @@ ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
        int optlen = 0;
        u_int8_t *opt = opthead;
        u_int16_t rtalert_val;
+       u_int32_t jumboplen;
+       const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
 
        for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
-               switch(*opt) {
-                case IP6OPT_PAD1:
-                        optlen = 1;
-                        break;
-                case IP6OPT_PADN:
-                        if (hbhlen < IP6OPT_MINLEN) {
-                                ip6stat.ip6s_toosmall++;
-                                goto bad;
-                        }
-                        optlen = *(opt + 1) + 2;
-                        break;
-                case IP6OPT_RTALERT:
-                        /* XXX may need check for alignment */
-                        if (hbhlen < IP6OPT_RTALERT_LEN) {
-                                ip6stat.ip6s_toosmall++;
-                                goto bad;
-                        }
-                        if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2)
-                                 /* XXX: should we discard the packet? */
-                                log(LOG_ERR, "length of router alert opt is inconsitent(%d)",
-                                    *(opt + 1));
-                        optlen = IP6OPT_RTALERT_LEN;
-                        bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
-                        *rtalertp = ntohs(rtalert_val);
-                        break;
-                case IP6OPT_JUMBO:
-                        /* XXX may need check for alignment */
-                        if (hbhlen < IP6OPT_JUMBO_LEN) {
-                                ip6stat.ip6s_toosmall++;
-                                goto bad;
-                        }
-                        if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2)
-                                 /* XXX: should we discard the packet? */
-                                log(LOG_ERR, "length of jumbopayload opt "
-                                    "is inconsistent(%d)",
-                                    *(opt + 1));
-                        optlen = IP6OPT_JUMBO_LEN;
-
-                        /*
-                         * We can simply cast because of the alignment
-                         * requirement of the jumbo payload option.
-                         */
-#if 0
-                        *plenp = ntohl(*(u_int32_t *)(opt + 2));
-#else
-                        bcopy(opt + 2, plenp, sizeof(*plenp));
-                        *plenp = htonl(*plenp);
+               switch (*opt) {
+               case IP6OPT_PAD1:
+                       optlen = 1;
+                       break;
+               case IP6OPT_PADN:
+                       if (hbhlen < IP6OPT_MINLEN) {
+                               ip6stat.ip6s_toosmall++;
+                               goto bad;
+                       }
+                       optlen = *(opt + 1) + 2;
+                       break;
+               case IP6OPT_ROUTER_ALERT:
+                       /* XXX may need check for alignment */
+                       if (hbhlen < IP6OPT_RTALERT_LEN) {
+                               ip6stat.ip6s_toosmall++;
+                               goto bad;
+                       }
+                       if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
+                               /* XXX stat */
+                               icmp6_error(m, ICMP6_PARAM_PROB,
+                                           ICMP6_PARAMPROB_HEADER,
+                                           erroff + opt + 1 - opthead);
+                               return(-1);
+                       }
+                       optlen = IP6OPT_RTALERT_LEN;
+                       bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
+                       *rtalertp = ntohs(rtalert_val);
+                       break;
+               case IP6OPT_JUMBO:
+                       /* XXX may need check for alignment */
+                       if (hbhlen < IP6OPT_JUMBO_LEN) {
+                               ip6stat.ip6s_toosmall++;
+                               goto bad;
+                       }
+                       if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
+                               /* XXX stat */
+                               icmp6_error(m, ICMP6_PARAM_PROB,
+                                           ICMP6_PARAMPROB_HEADER,
+                                           erroff + opt + 1 - opthead);
+                               return(-1);
+                       }
+                       optlen = IP6OPT_JUMBO_LEN;
+
+                       /*
+                        * IPv6 packets that have non 0 payload length
+                        * must not contain a jumbo payload option.
+                        */
+                       ip6 = mtod(m, struct ip6_hdr *);
+                       if (ip6->ip6_plen) {
+                               ip6stat.ip6s_badoptions++;
+                               icmp6_error(m, ICMP6_PARAM_PROB,
+                                           ICMP6_PARAMPROB_HEADER,
+                                           erroff + opt - opthead);
+                               return(-1);
+                       }
+
+                       /*
+                        * We may see jumbolen in unaligned location, so
+                        * we'd need to perform bcopy().
+                        */
+                       bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
+                       jumboplen = (u_int32_t)htonl(jumboplen);
+
+#if 1
+                       /*
+                        * if there are multiple jumbo payload options,
+                        * *plenp will be non-zero and the packet will be
+                        * rejected.
+                        * the behavior may need some debate in ipngwg -
+                        * multiple options does not make sense, however,
+                        * there's no explicit mention in specification.
+                        */
+                       if (*plenp != 0) {
+                               ip6stat.ip6s_badoptions++;
+                               icmp6_error(m, ICMP6_PARAM_PROB,
+                                           ICMP6_PARAMPROB_HEADER,
+                                           erroff + opt + 2 - opthead);
+                               return(-1);
+                       }
 #endif
-                        if (*plenp <= IPV6_MAXPACKET) {
-                                /*
-                                 * jumbo payload length must be larger
-                                 * than 65535
-                                 */
-                                ip6stat.ip6s_badoptions++;
-                                icmp6_error(m, ICMP6_PARAM_PROB,
-                                            ICMP6_PARAMPROB_HEADER,
-                                            sizeof(struct ip6_hdr) +
-                                            sizeof(struct ip6_hbh) +
-                                            opt + 2 - opthead);
-                                return(-1);
-                        }
-
-                        ip6 = mtod(m, struct ip6_hdr *);
-                        if (ip6->ip6_plen) {
-                                /*
-                                 * IPv6 packets that have non 0 payload length
-                                 * must not contain a jumbo paylod option.
-                                 */
-                                ip6stat.ip6s_badoptions++;
-                                icmp6_error(m, ICMP6_PARAM_PROB,
-                                            ICMP6_PARAMPROB_HEADER,
-                                            sizeof(struct ip6_hdr) +
-                                            sizeof(struct ip6_hbh) +
-                                            opt - opthead);
-                                return(-1);
-                        }
-                        break;
-                default:               /* unknown option */
-                        if (hbhlen < IP6OPT_MINLEN) {
-                                ip6stat.ip6s_toosmall++;
-                                goto bad;
-                        }
-                        if ((optlen = ip6_unknown_opt(opt, m,
-                                                      sizeof(struct ip6_hdr) +
-                                                      sizeof(struct ip6_hbh) +
-                                                      opt - opthead)) == -1)
-                                return(-1);
-                        optlen += 2;
-                        break;
+
+                       /*
+                        * jumbo payload length must be larger than 65535.
+                        */
+                       if (jumboplen <= IPV6_MAXPACKET) {
+                               ip6stat.ip6s_badoptions++;
+                               icmp6_error(m, ICMP6_PARAM_PROB,
+                                           ICMP6_PARAMPROB_HEADER,
+                                           erroff + opt + 2 - opthead);
+                               return(-1);
+                       }
+                       *plenp = jumboplen;
+
+                       break;
+               default:                /* unknown option */
+                       if (hbhlen < IP6OPT_MINLEN) {
+                               ip6stat.ip6s_toosmall++;
+                               goto bad;
+                       }
+                       optlen = ip6_unknown_opt(opt, m,
+                           erroff + opt - opthead);
+                       if (optlen == -1) {
+                               return(-1);
+                       }
+                       optlen += 2;
+                       break;
                }
        }
 
        return(0);
 
-  bad:
+  bad: 
        m_freem(m);
        return(-1);
 }
@@ -974,199 +1434,173 @@ 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)
-       u_int8_t *optp;
-       struct mbuf *m;
-       int off;
+ip6_unknown_opt(uint8_t *optp, struct mbuf *m, int off)
 {
        struct ip6_hdr *ip6;
 
-       switch(IP6OPT_TYPE(*optp)) {
-        case IP6OPT_TYPE_SKIP: /* ignore the option */
-                return((int)*(optp + 1));
-        case IP6OPT_TYPE_DISCARD:      /* silently discard */
-                m_freem(m);
-                return(-1);
-        case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
-                ip6stat.ip6s_badoptions++;
-                icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
-                return(-1);
-        case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
-                ip6stat.ip6s_badoptions++;
-                ip6 = mtod(m, struct ip6_hdr *);
-                if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
-                    (m->m_flags & (M_BCAST|M_MCAST)))
-                        m_freem(m);
-                else
-                        icmp6_error(m, ICMP6_PARAM_PROB,
-                                    ICMP6_PARAMPROB_OPTION, off);
-                return(-1);
+       switch (IP6OPT_TYPE(*optp)) {
+       case IP6OPT_TYPE_SKIP: /* ignore the option */
+               return((int)*(optp + 1));
+       case IP6OPT_TYPE_DISCARD:       /* silently discard */
+               m_freem(m);
+               return(-1);
+       case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
+               ip6stat.ip6s_badoptions++;
+               icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
+               return(-1);
+       case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
+               ip6stat.ip6s_badoptions++;
+               ip6 = mtod(m, struct ip6_hdr *);
+               if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
+                   (m->m_flags & (M_BCAST|M_MCAST)))
+                       m_freem(m);
+               else
+                       icmp6_error(m, ICMP6_PARAM_PROB,
+                                   ICMP6_PARAMPROB_OPTION, off);
+               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, ip6, m, ctl, prevctlp)
-#if defined(__FreeBSD__) && __FreeBSD__ >= 3 || defined(__APPLE__)
-       register struct inpcb *in6p;
-#else
-       register struct in6pcb *in6p;
-#endif
-       register struct ip6_hdr *ip6;
-       register struct mbuf *m;
-       struct ip6_recvpktopts *ctl, **prevctlp;
+struct mbuf **
+ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
+    int *v4only)
 {
-       register struct mbuf **mp;
-       struct cmsghdr *cm = NULL;
-       struct ip6_recvpktopts *prevctl = NULL;
-#if HAVE_NRL_INPCB
-# define in6p_flags    inp_flags
-#endif
-#if defined(__NetBSD__) || (defined(__FreeBSD__) && __FreeBSD__ >= 3) || defined (__APPLE__)
-       struct proc *p = current_proc();        /* XXX */
-#endif
-#ifdef __bsdi__
-# define sbcreatecontrol       so_cmsg
-#endif
-       int privileged = 0;
-
-
-       if (ctl == NULL)        /* validity check */
-               return;
-       bzero(ctl, sizeof(*ctl)); /* XXX is it really OK? */
-       mp = &ctl->head;
-
-       /*
-        * If caller wanted to keep history, allocate space to store the
-        * history at the first time.
-        */
-       if (prevctlp) {
-               if (*prevctlp == NULL) {
-                       MALLOC(prevctl, struct ip6_recvpktopts *,
-                              sizeof(*prevctl), M_IP6OPT, M_NOWAIT);
-                       if (prevctl == NULL) {
-                               printf("ip6_savecontrol: can't allocate "
-                                      " enough space for history\n");
-                               return;
-                       }
-                       bzero(prevctl, sizeof(*prevctl));
-                       *prevctlp = prevctl;
-               }
-               else
-                       prevctl = *prevctlp;
-       }
-
-#if defined(__NetBSD__) || (defined(__FreeBSD__) && __FreeBSD__ == 3)
-       if (p && !suser(p->p_ucred, &p->p_acflag))
-               privileged++;
-#elif defined(__FreeBSD__) && __FreeBSD__ >= 4
-       if (p && !suser(p))
-               privileged++;
-#else
-#if HAVE_NRL_INPCB
-       if ((in6p->inp_socket->so_state & SS_PRIV) != 0)
-               privileged++;
-#else
-       if ((in6p->in6p_socket->so_state & SS_PRIV) != 0)
-               privileged++;
-#endif
-#endif
+       struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
 
-#if SO_TIMESTAMP
-       if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
+       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) {
-                       /* always set regradless of the previous value */
-                       ctl->timestamp = *mp;
-                       mp = &(*mp)->m_next;
-               }
+               mp = sbcreatecontrol_mbuf((caddr_t) &tv, sizeof(tv),
+                   SCM_TIMESTAMP, SOL_SOCKET, mp);
+               if (*mp == NULL) 
+                       return NULL;
+       }
+        if ((inp->inp_socket->so_options & SO_TIMESTAMP_MONOTONIC) != 0) {
+                uint64_t time;
+
+                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_get_traffic_class(m);
+
+               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);
        }
-#endif
 
+#define IS2292(inp, x, y)      (((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
        /* RFC 2292 sec. 5 */
-       if (in6p->in6p_flags & IN6P_PKTINFO) {
-               struct in6_pktinfo pi6, *prevpi = NULL;
+       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;
-               if (prevctl && prevctl->pktinfo) {
-                       cm = mtod(prevctl->pktinfo, struct cmsghdr *);
-                       prevpi = (struct in6_pktinfo *)CMSG_DATA(cm);
-               }
+               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;
+       }
 
-               /*
-                * Make a new option only if this is the first time or if the
-                * option value is chaned from last time.
-                */
-               if (prevpi == NULL || bcmp(prevpi, &pi6, sizeof(pi6))) {
-                       *mp = sbcreatecontrol((caddr_t) &pi6,
-                                             sizeof(struct in6_pktinfo),
-                                             IPV6_PKTINFO,
-                                             IPPROTO_IPV6);
-                       if (*mp) {
-                               ctl->pktinfo = *mp;
-                               mp = &(*mp)->m_next;
-                       }
-               }
+       if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
+               int hlim = ip6->ip6_hlim & 0xff;
+
+               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_HOPLIMIT) {
-               int hlim = ip6->ip6_hlim & 0xff, oldhlim = -1;
+       if (v4only != NULL)
+               *v4only = 0;
+       return (mp);
+}
 
-               if (prevctl && prevctl->hlim) {
-                       cm = mtod(prevctl->hlim, struct cmsghdr *);
-                       oldhlim = (*(int *)CMSG_DATA(cm)) & 0xff;
-               }
+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;
 
-               if (oldhlim < 0 || hlim != oldhlim) {
-                       *mp = sbcreatecontrol((caddr_t) &hlim,
-                                             sizeof(int), IPV6_HOPLIMIT,
-                                             IPPROTO_IPV6);
-                       if (*mp) {
-                               ctl->hlim = *mp;
-                               mp = &(*mp)->m_next;
-                       }
-               }
+       *mp = NULL;
+       np = ip6_savecontrol_v4(in6p, m, mp, &v4only);
+       if (np == NULL)
+               goto no_mbufs;
+
+       mp = np;
+       if (v4only)
+               return(0);
+
+       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. We require super-user privilege
-        * for the option, 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.
+        * 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 also RFC 2292 section 6 (or RFC 3542 section 8).
         */
-       if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
+       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.
                 */
-               struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
+               ip6 = mtod(m, struct ip6_hdr *);
                if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
-                       struct ip6_hbh *hbh, *prevhbh = NULL;
-                       int hbhlen = 0, prevhbhlen = 0;
-#ifdef PULLDOWN_TEST
+                       struct ip6_hbh *hbh;
+                       int hbhlen = 0;
+#if PULLDOWN_TEST
                        struct mbuf *ext;
 #endif
 
@@ -1178,56 +1612,39 @@ ip6_savecontrol(in6p, ip6, m, ctl, prevctlp)
                            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
 
-                       if (prevctl && prevctl->hbh) {
-                               cm = mtod(prevctl->hbh, struct cmsghdr *);
-                               prevhbh = (struct ip6_hbh *)CMSG_DATA(cm);
-                               prevhbhlen = (prevhbh->ip6h_len + 1) << 3;
-                       }
                        /*
-                        * Check if there's difference between the current
-                        * and previous HbH headers.
-                        * XXX: should the next header field be ignored?
+                        * 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
                         */
-                       if (prevhbh == NULL || hbhlen != prevhbhlen ||
-                           bcmp(prevhbh, hbh, hbhlen)) {
-                               /*
-                                * 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.
-                                */
-                               *mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
-                                                     IPV6_HOPOPTS,
-                                                     IPPROTO_IPV6);
-                               if (*mp) {
-                                       ctl->hbh = *mp;
-                                       mp = &(*mp)->m_next;
-                               }
-                       }
-#ifdef PULLDOWN_TEST
+                       mp = sbcreatecontrol_mbuf((caddr_t)hbh, hbhlen,
+                           IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
+                           IPPROTO_IPV6, mp);
+
+#if PULLDOWN_TEST
                        m_freem(ext);
 #endif
+                       if (*mp == NULL) {
+                               goto no_mbufs;
+                       }
                }
        }
 
-       /* IPV6_DSTOPTS and IPV6_RTHDR socket options */
-       if (in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDR)) {
-               struct ip6_hdr *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);
-               int rthdr = 0;  /* flag if we've passed a routing header */
 
                /*
                 * Search for destination options headers or routing
@@ -1239,7 +1656,7 @@ ip6_savecontrol(in6p, ip6, m, ctl, prevctlp)
                while (1) {     /* is explicit loop prevention necessary? */
                        struct ip6_ext *ip6e = NULL;
                        int elen;
-#ifdef PULLDOWN_TEST
+#if PULLDOWN_TEST
                        struct mbuf *ext = NULL;
 #endif
 
@@ -1271,7 +1688,7 @@ ip6_savecontrol(in6p, ip6, m, ctl, prevctlp)
                        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)
@@ -1281,142 +1698,52 @@ ip6_savecontrol(in6p, ip6, m, ctl, prevctlp)
                        if (elen != ext->m_len) {
                                m_freem(ext);
                                ip6stat.ip6s_tooshort++;
-                               return;
+                               return(0);
                        }
 #endif
 
                        switch (nxt) {
                        case IPPROTO_DSTOPTS:
-                       {
-                               struct ip6_dest *prevdest1 = NULL,
-                                       *prevdest2 = NULL;
-                               int prevdestlen;
-
-                               if ((in6p->in6p_flags &
-                                    (IN6P_DSTOPTS | IN6P_RTHDRDSTOPTS)) == 0)
+                               if (!(in6p->inp_flags & IN6P_DSTOPTS))
                                        break;
 
-                                /*
-                                 * We also require super-user privilege for
-                                 * the option.
-                                 * See the comments on IN6_HOPOPTS.
-                                 */
-                               if (!privileged)
-                                       break;
-
-                                /*
-                                 * Save a dst opt header before a routing
-                                 * header if the user wanted.
-                                 */
-                               if (rthdr == 0 &&
-                                   (in6p->in6p_flags & IN6P_RTHDRDSTOPTS)) {
-                                       if (prevctl && prevctl->dest1) {
-                                               cm = mtod(prevctl->dest1,
-                                                         struct cmsghdr *);
-                                               prevdest1 = (struct ip6_dest *)CMSG_DATA(cm);
-                                               prevdestlen = (prevdest1->ip6d_len + 1) << 3;
-                                       }
-
-                                       /*
-                                        * If this is the 1st dst opt header
-                                        * (that is placed before rthdr)
-                                        * we enconter and this header is
-                                        * not different from the previous one,
-                                        * simply ignore the header.
-                                        */
-                                       if (ctl->dest1 == NULL &&
-                                           (prevdest1 &&
-                                            prevdestlen == elen &&
-                                            bcmp(ip6e, prevdest1, elen) == 0))
-                                               break;
-
-                                       *mp = sbcreatecontrol((caddr_t)ip6e,
-                                                             elen,
-                                                             IPV6_RTHDRDSTOPTS,
-                                                             IPPROTO_IPV6);
-                                       if (ctl->dest1 == NULL)
-                                               ctl->dest1 = *mp;
-                                       if (*mp)
-                                               mp = &(*mp)->m_next;
-                               }
-                               /*
-                                * Save a dst opt header after a routing
-                                * header if the user wanted.
-                                */
-                               if (rthdr &&
-                                   (in6p->in6p_flags & IN6P_DSTOPTS)) {
-                                       if (prevctl && prevctl->dest2) {
-                                               cm = mtod(prevctl->dest2,
-                                                         struct cmsghdr *);
-                                               prevdest2 = (struct ip6_dest *)CMSG_DATA(cm);
-                                               prevdestlen = (prevdest2->ip6d_len + 1) << 3;
-                                       }
-                                       /* see the above comment */
-                                       if (ctl->dest2 == NULL &&
-                                           (prevdest2 &&
-                                            prevdestlen == elen &&
-                                            bcmp(ip6e, prevdest2, elen) == 0))
-                                               break;
-
-                                       *mp = sbcreatecontrol((caddr_t)ip6e,
-                                                             elen,
-                                                             IPV6_DSTOPTS,
-                                                             IPPROTO_IPV6);
-                                       if (ctl->dest2 == NULL)
-                                               ctl->dest2 = *mp;
-
-                                       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:
-                       {
-                               struct ip6_rthdr *prevrth = NULL;
-                               int prevrhlen = 0;
-
-                               rthdr++;
-                               if (!in6p->in6p_flags & IN6P_RTHDR)
+                               if (!in6p->inp_flags & IN6P_RTHDR)
                                        break;
 
-                               if (prevctl && prevctl->rthdr) {
-                                       cm = mtod(prevctl->rthdr,
-                                                 struct cmsghdr *);
-                                       prevrth = (struct ip6_rthdr *)CMSG_DATA(cm);
-                                       prevrhlen =
-                                               (prevrth->ip6r_len + 1) << 3;
+                               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;
                                }
-
-                               /*
-                                * Check if the rthdr should be passed to
-                                * a user. See the comments for dstopt hdr.
-                                */
-                               if (ctl->rthdr == NULL && prevrth &&
-                                   prevrhlen == elen &&
-                                   bcmp(ip6e, prevrth, elen) == 0)
-                                       break;
-
-                               *mp = sbcreatecontrol((caddr_t)ip6e, elen,
-                                                     IPV6_RTHDR,
-                                                     IPPROTO_IPV6);
-                               if (ctl->rthdr == NULL)
-                                       ctl->rthdr = *mp;
-                               if (*mp)
-                                       mp = &(*mp)->m_next;
                                break;
-                       }
                        case IPPROTO_HOPOPTS:
                        case IPPROTO_AH: /* is it possible? */
                                break;
 
                        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).
                                 */
-#ifdef PULLDOWN_TEST
+#if PULLDOWN_TEST
                                m_freem(ext);
 #endif
                                goto loopend;
@@ -1427,23 +1754,60 @@ ip6_savecontrol(in6p, ip6, m, ctl, prevctlp)
                        off += elen;
                        nxt = ip6e->ip6e_nxt;
                        ip6e = NULL;
-#ifdef PULLDOWN_TEST
+#if PULLDOWN_TEST
                        m_freem(ext);
                        ext = NULL;
 #endif
                }
          loopend:
+               ;
        }
+       return(0);
+no_mbufs:
+       ip6stat.ip6s_pktdropcntrl++;
+       /* XXX increment a stat to show the failure */
+       return(ENOBUFS);
+}
+#undef IS2292
 
-#ifdef __bsdi__
-# undef sbcreatecontrol
-#endif
-#ifdef __OpenBSD__
-# undef in6p_flags
+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, TRUE))
+               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;
 }
 
-#ifdef PULLDOWN_TEST
+#if PULLDOWN_TEST
 /*
  * pull single extension header from mbuf chain.  returns single mbuf that
  * contains the result, or NULL on error.
@@ -1458,7 +1822,7 @@ ip6_pullexthdr(m, off, nxt)
        size_t elen;
        struct mbuf *n;
 
-#ifdef DIAGNOSTIC
+#if DIAGNOSTIC
        switch (nxt) {
        case IPPROTO_DSTOPTS:
        case IPPROTO_ROUTING:
@@ -1499,172 +1863,6 @@ ip6_pullexthdr(m, off, nxt)
 }
 #endif
 
-/*
- * Merge new IPv6 received options to previous ones.
- * If a new option is not given, just re-link the option chain.
- * If an old option exists but a corresponding new one doesn't, just
- * keep the ole option.
- * If a new option exists but a corresponding old one doesn't, just
- * copy the new option.
- * If both new and old options exist, free old one and overwrite the option
- * with the new option.
- * Otherwise, do nothing for the option.
- * XXX: in any case, options that don't follow the recommend order and
- *      number of extension headers (RFC 2460 Section 4.1) are simply ignored.
- * XXX: We assume that each option is stored in a single mbuf.
- */
-#define CLEAN_RECVOPT(old, type) \
-do {                                                           \
-       if ((old)->type && (old)->type->m_next) {               \
-               (old)->type->m_next = NULL;                     \
-       }                                                       \
-} while (0)
-#define MERGE_RECVOPT(new, old, type) if ((new)->type) {\
-               if ((old)->type)\
-                       m_free((old)->type);\
-               (old)->type = m_copy((new)->type, 0, (new)->type->m_len);\
-               if (((old)->type) && ((old)->type->m_next)) {\
-                       m_freem((old)->type);\
-                       old->type = NULL;\
-               }\
-       }
-#define LINK_RECVOPTS(opt, type, p) if ((opt)->type) {\
-               *(p) = (opt)->type;\
-               (p) = &(opt)->type->m_next;\
-       }
-
-static void dump_inputopts __P((char *, struct ip6_recvpktopts *));
-static void
-dump_inputopts(str, p)
-       char *str;
-       struct ip6_recvpktopts *p;
-{
-#if 1
-       return;
-#else
-#define PRINT1(p, name) \
-do { \
-       if (p->name) { \
-               printf(" %s: %p", #name, (p)->name); \
-               if (p->name->m_next) \
-                       printf("[%p]", (p)->name->m_next); \
-       } \
-} while (0)
-
-       printf("%s p=%p head=%p", str, p, p->head);
-       PRINT1(p, hlim);
-       PRINT1(p, pktinfo);
-       PRINT1(p, hbh);
-       PRINT1(p, dest1);
-       PRINT1(p, dest2);
-       PRINT1(p, rthdr);
-       printf("\n");
-#undef PRINT1
-#endif
-}
-
-void
-ip6_update_recvpcbopt(old, new)
-       struct ip6_recvpktopts *new, *old;
-{
-       struct mbuf **mp;
-
-       if (old == NULL) {
-               printf("ip6_update_recvpcbopt: invalid arguments\n");
-               return;
-       }
-
-       dump_inputopts("old before", old);
-       if (new)
-               dump_inputopts("new before", new);
-
-#if 0
-       /*
-        * cleanup m->m_next linkage. note that we do it in reverse order
-        * to prevent possible memory leakage.
-        */
-       old->head = NULL;
-       CLEAN_RECVOPT(old, rthdr);
-       CLEAN_RECVOPT(old, dest2);
-       CLEAN_RECVOPT(old, dest1);
-       CLEAN_RECVOPT(old, hbh);
-       CLEAN_RECVOPT(old, pktinfo);
-       CLEAN_RECVOPT(old, hlim);
-#endif
-
-       if (new) {
-               MERGE_RECVOPT(new, old, hlim);
-               MERGE_RECVOPT(new, old, pktinfo);
-               MERGE_RECVOPT(new, old, hbh);
-               MERGE_RECVOPT(new, old, dest1);
-               MERGE_RECVOPT(new, old, dest2);
-               MERGE_RECVOPT(new, old, rthdr);
-       }
-
-       dump_inputopts("old middle", old);
-       if (new)
-               dump_inputopts("new middle", new);
-
-       /* link options */
-       mp = &old->head;
-       LINK_RECVOPTS(old, hlim, mp);
-       LINK_RECVOPTS(old, pktinfo, mp);
-       LINK_RECVOPTS(old, hbh, mp);
-       LINK_RECVOPTS(old, dest1, mp);
-       LINK_RECVOPTS(old, dest2, mp);
-       LINK_RECVOPTS(old, rthdr, mp);
-       *mp = NULL;
-
-       dump_inputopts("old after", old);
-       if (new)
-               dump_inputopts("new after", new);
-}
-
-#undef MERGE_RECVOPT
-#undef LINK_RECVOPTS
-
-void
-ip6_reset_rcvopt(opts, optname)
-       struct ip6_recvpktopts *opts;
-       int optname;
-{
-       if (opts == NULL)
-               return;
-
-       switch(optname) {
-       case IPV6_RECVPKTINFO:
-               if (opts->pktinfo) m_free(opts->pktinfo);
-               opts->pktinfo = NULL;
-               break;
-       case IPV6_RECVHOPLIMIT:
-               if (opts->hlim) m_free(opts->hlim);
-               opts->hlim = NULL;
-               break;
-       case IPV6_RECVHOPOPTS:
-               if (opts->hbh) m_free(opts->hbh);
-               opts->hbh = NULL;
-               break;
-       case IPV6_RECVRTHDRDSTOPTS:
-               if (opts->dest1) m_free(opts->dest1);
-               opts->dest1 = NULL;
-               break;
-       case IPV6_RECVDSTOPTS:
-               if (opts->dest2) m_free(opts->dest2);
-               opts->dest2 = NULL;
-               break;
-       case IPV6_RECVRTHDR:
-               if (opts->rthdr) m_free(opts->rthdr);
-               opts->rthdr = NULL;
-               break;
-       default:
-               printf("ip6_reset_rcvopt: invalid option name (%d)\n",
-                      optname);
-               return;
-       }
-
-       ip6_update_recvpcbopt(opts, NULL); /* re-link the option chain */
-}
-
 /*
  * Get pointer to the previous header followed by the header
  * currently processed.
@@ -1685,7 +1883,7 @@ ip6_get_prevhdr(m, off)
        struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
 
        if (off == sizeof(struct ip6_hdr))
-               return(&ip6->ip6_nxt);
+               return((char *) &ip6->ip6_nxt);
        else {
                int len, nxt;
                struct ip6_ext *ip6e = NULL;
@@ -1695,7 +1893,7 @@ ip6_get_prevhdr(m, off)
                while (len < off) {
                        ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
 
-                       switch(nxt) {
+                       switch (nxt) {
                        case IPPROTO_FRAGMENT:
                                len += sizeof(struct ip6_frag);
                                break;
@@ -1709,7 +1907,7 @@ ip6_get_prevhdr(m, off)
                        nxt = ip6e->ip6e_nxt;
                }
                if (ip6e)
-                       return(&ip6e->ip6e_nxt);
+                       return((char *) &ip6e->ip6e_nxt);
                else
                        return NULL;
        }
@@ -1753,7 +1951,8 @@ ip6_nexthdr(m, off, proto, nxtp)
                if (m->m_pkthdr.len < off + sizeof(fh))
                        return -1;
                m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
-               if ((ntohs(fh.ip6f_offlg) & IP6F_OFF_MASK) != 0)
+               /* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
+               if (fh.ip6f_offlg & IP6F_OFF_MASK)
                        return -1;
                if (nxtp)
                        *nxtp = fh.ip6f_nxt;
@@ -1824,6 +2023,74 @@ ip6_lasthdr(m, off, proto, nxtp)
        }
 }
 
+struct ip6aux *
+ip6_addaux(struct mbuf *m)
+{
+       struct m_tag            *tag;
+
+       /* Check if one is already allocated */
+       tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
+           KERNEL_TAG_TYPE_INET6, NULL);
+       if (tag == NULL) {
+               /* Allocate a tag */
+               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) {
+                       m_tag_prepend(m, tag);
+               }
+       }
+
+       return (tag ? (struct ip6aux *)(tag + 1) : NULL);
+}
+
+struct ip6aux *
+ip6_findaux(struct mbuf *m)
+{
+       struct m_tag    *tag;
+
+       tag = m_tag_locate(m, KERNEL_MODULE_TAG_ID,
+           KERNEL_TAG_TYPE_INET6, NULL);
+
+       return (tag ? (struct ip6aux *)(tag + 1) : NULL);
+}
+
+void
+ip6_delaux(struct mbuf *m)
+{
+       struct m_tag    *tag;
+
+       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
  */
@@ -1836,102 +2103,3 @@ u_char  inet6ctlerrmap[PRC_NCMDS] = {
        0,              0,              0,              0,
        ENOPROTOOPT
 };
-
-#if defined(__NetBSD__) || defined(__OpenBSD__)
-#include <vm/vm.h>
-#include <sys/sysctl.h>
-
-int
-ip6_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
-       int *name;
-       u_int namelen;
-       void *oldp;
-       size_t *oldlenp;
-       void *newp;
-       size_t newlen;
-{
-       /* All sysctl names at this level are terminal. */
-       if (namelen != 1)
-               return ENOTDIR;
-
-       switch (name[0]) {
-
-       case IPV6CTL_FORWARDING:
-               return sysctl_int(oldp, oldlenp, newp, newlen,
-                                 &ip6_forwarding);
-       case IPV6CTL_SENDREDIRECTS:
-               return sysctl_int(oldp, oldlenp, newp, newlen,
-                               &ip6_sendredirects);
-       case IPV6CTL_DEFHLIM:
-               return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_defhlim);
-       case IPV6CTL_MAXFRAGPACKETS:
-               return sysctl_int(oldp, oldlenp, newp, newlen,
-                               &ip6_maxfragpackets);
-       case IPV6CTL_ACCEPT_RTADV:
-               return sysctl_int(oldp, oldlenp, newp, newlen,
-                               &ip6_accept_rtadv);
-       case IPV6CTL_KEEPFAITH:
-               return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_keepfaith);
-       case IPV6CTL_LOG_INTERVAL:
-               return sysctl_int(oldp, oldlenp, newp, newlen,
-                               &ip6_log_interval);
-       case IPV6CTL_HDRNESTLIMIT:
-               return sysctl_int(oldp, oldlenp, newp, newlen,
-                               &ip6_hdrnestlimit);
-       case IPV6CTL_DAD_COUNT:
-               return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_dad_count);
-       case IPV6CTL_AUTO_FLOWLABEL:
-               return sysctl_int(oldp, oldlenp, newp, newlen,
-                               &ip6_auto_flowlabel);
-       case IPV6CTL_DEFMCASTHLIM:
-               return sysctl_int(oldp, oldlenp, newp, newlen,
-                               &ip6_defmcasthlim);
-       case IPV6CTL_GIF_HLIM:
-               return sysctl_int(oldp, oldlenp, newp, newlen,
-                               &ip6_gif_hlim);
-       case IPV6CTL_KAME_VERSION:
-               return sysctl_rdstring(oldp, oldlenp, newp, __KAME_VERSION);
-       case IPV6CTL_USE_DEPRECATED:
-               return sysctl_int(oldp, oldlenp, newp, newlen,
-                               &ip6_use_deprecated);
-       case IPV6CTL_RR_PRUNE:
-               return sysctl_int(oldp, oldlenp, newp, newlen, &ip6_rr_prune);
-#if defined(__NetBSD__) && !defined(INET6_BINDV6ONLY)
-       case IPV6CTL_BINDV6ONLY:
-               return sysctl_int(oldp, oldlenp, newp, newlen,
-                               &ip6_bindv6only);
-#endif
-       default:
-               return EOPNOTSUPP;
-       }
-       /* NOTREACHED */
-}
-#endif /* __NetBSD__ || __OpenBSD__ */
-
-#ifdef __bsdi__
-int *ip6_sysvars[] = IPV6CTL_VARS;
-
-int
-ip6_sysctl(name, namelen, oldp, oldlenp, newp, newlen)
-       int     *name;
-       u_int   namelen;
-       void    *oldp;
-       size_t  *oldlenp;
-       void    *newp;
-       size_t  newlen;
-{
-       if (name[0] >= IPV6CTL_MAXID)
-               return (EOPNOTSUPP);
-
-       switch (name[0]) {
-       case IPV6CTL_STATS:
-               return sysctl_rdtrunc(oldp, oldlenp, newp, &ip6stat,
-                   sizeof(ip6stat));
-       case IPV6CTL_KAME_VERSION:
-               return sysctl_rdstring(oldp, oldlenp, newp, __KAME_VERSION);
-       default:
-               return (sysctl_int_arr(ip6_sysvars, name, namelen,
-                   oldp, oldlenp, newp, newlen));
-       }
-}
-#endif /* __bsdi__ */