bsd/netinet6/ip6_forward.c

   1 /*
   2  * Copyright (c) 2009-2013 Apple Inc. All rights reserved.
   3  *
   4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
   5  *
   6  * This file contains Original Code and/or Modifications of Original Code
   7  * as defined in and that are subject to the Apple Public Source License
   8  * Version 2.0 (the 'License'). You may not use this file except in
   9  * compliance with the License. The rights granted to you under the License
  10  * may not be used to create, or enable the creation or redistribution of,
  11  * unlawful or unlicensed copies of an Apple operating system, or to
  12  * circumvent, violate, or enable the circumvention or violation of, any
  13  * terms of an Apple operating system software license agreement.
  14  *
  15  * Please obtain a copy of the License at
  16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
  17  *
  18  * The Original Code and all software distributed under the License are
  19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  23  * Please see the License for the specific language governing rights and
  24  * limitations under the License.
  25  *
  26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  27  */
  28
  29 /*      $FreeBSD: src/sys/netinet6/ip6_forward.c,v 1.16 2002/10/16 02:25:05 sam Exp $   */
  30 /*      $KAME: ip6_forward.c,v 1.69 2001/05/17 03:48:30 itojun Exp $    */
  31
  32 /*
  33  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
  34  * All rights reserved.
  35  *
  36  * Redistribution and use in source and binary forms, with or without
  37  * modification, are permitted provided that the following conditions
  38  * are met:
  39  * 1. Redistributions of source code must retain the above copyright
  40  *    notice, this list of conditions and the following disclaimer.
  41  * 2. Redistributions in binary form must reproduce the above copyright
  42  *    notice, this list of conditions and the following disclaimer in the
  43  *    documentation and/or other materials provided with the distribution.
  44  * 3. Neither the name of the project nor the names of its contributors
  45  *    may be used to endorse or promote products derived from this software
  46  *    without specific prior written permission.
  47  *
  48  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  58  * SUCH DAMAGE.
  59  */
  60
  61
  62 #include <sys/param.h>
  63 #include <sys/systm.h>
  64 #include <sys/malloc.h>
  65 #include <sys/mbuf.h>
  66 #include <sys/domain.h>
  67 #include <sys/protosw.h>
  68 #include <sys/socket.h>
  69 #include <sys/errno.h>
  70 #include <sys/time.h>
  71 #include <sys/kernel.h>
  72 #include <sys/syslog.h>
  73
  74 #include <net/if.h>
  75 #include <net/route.h>
  76
  77 #include <netinet/in.h>
  78 #include <netinet/in_var.h>
  79 #include <netinet/in_systm.h>
  80 #include <netinet/ip.h>
  81 #include <netinet/ip_var.h>
  82 #include <netinet6/in6_var.h>
  83 #include <netinet/ip6.h>
  84 #include <netinet6/ip6_var.h>
  85 #include <netinet/icmp6.h>
  86 #include <netinet6/nd6.h>
  87 #include <netinet6/scope6_var.h>
  88
  89 #include <netinet/in_pcb.h>
  90
  91 #if IPSEC
  92 #include <netinet6/ipsec.h>
  93 #if INET6
  94 #include <netinet6/ipsec6.h>
  95 #endif
  96 #include <netkey/key.h>
  97 extern int ipsec_bypass;
  98 #endif /* IPSEC */
  99
 100 #include <netinet6/ip6_fw.h>
 101
 102 #include <net/net_osdep.h>
 103
 104 #if PF
 105 #include <net/pfvar.h>
 106 #endif /* PF */
 107
 108 /*
 109  * Forward a packet.  If some error occurs return the sender
 110  * an icmp packet.  Note we can't always generate a meaningful
 111  * icmp message because icmp doesn't have a large enough repertoire
 112  * of codes and types.
 113  *
 114  * If not forwarding, just drop the packet.  This could be confusing
 115  * if ipforwarding was zero but some routing protocol was advancing
 116  * us as a gateway to somewhere.  However, we must let the routing
 117  * protocol deal with that.
 118  *
 119  */
 120
 121 struct mbuf *
 122 ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
 123     int srcrt)
 124 {
 125         struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
 126         struct sockaddr_in6 *dst;
 127         struct rtentry *rt;
 128         int error, type = 0, code = 0;
 129         boolean_t proxy = FALSE;
 130         struct mbuf *mcopy = NULL;
 131         struct ifnet *ifp, *rcvifp, *origifp;   /* maybe unnecessary */
 132         u_int32_t inzone, outzone, len;
 133         struct in6_addr src_in6, dst_in6;
 134         uint64_t curtime = net_uptime();
 135 #if IPSEC
 136         struct secpolicy *sp = NULL;
 137 #endif
 138         unsigned int ifscope = IFSCOPE_NONE;
 139 #if PF
 140         struct pf_mtag *pf_mtag;
 141 #endif /* PF */
 142
 143         /*
 144          * In the prefix proxying case, the route to the proxied node normally
 145          * gets created by nd6_prproxy_ns_output(), as part of forwarding a
 146          * NS (NUD/AR) packet to the proxied node.  In the event that such
 147          * packet did not arrive in time before the correct route gets created,
 148          * ip6_input() would have performed a rtalloc() which most likely will
 149          * create the wrong cloned route; this route points back to the same
 150          * interface as the inbound interface, since the parent non-scoped
 151          * prefix route points there.  Therefore we check if that is the case
 152          * and perform the necessary fixup to get the correct route installed.
 153          */
 154         if (!srcrt && nd6_prproxy &&
 155             (rt = ip6forward_rt->ro_rt) != NULL && (rt->rt_flags & RTF_PROXY)) {
 156                 nd6_proxy_find_fwdroute(m->m_pkthdr.rcvif, ip6forward_rt);
 157                 if ((rt = ip6forward_rt->ro_rt) != NULL)
 158                         ifscope = rt->rt_ifp->if_index;
 159         }
 160
 161 #if PF
 162         pf_mtag = pf_find_mtag(m);
 163         if (pf_mtag != NULL && pf_mtag->pftag_rtableid != IFSCOPE_NONE)
 164                 ifscope = pf_mtag->pftag_rtableid;
 165
 166         /*
 167          * If the caller provides a route which is on a different interface
 168          * than the one specified for scoped forwarding, discard the route
 169          * and do a lookup below.
 170          */
 171         if (ifscope != IFSCOPE_NONE && (rt = ip6forward_rt->ro_rt) != NULL) {
 172                 RT_LOCK(rt);
 173                 if (rt->rt_ifp->if_index != ifscope) {
 174                         RT_UNLOCK(rt);
 175                         ROUTE_RELEASE(ip6forward_rt);
 176                         rt = NULL;
 177                 } else {
 178                         RT_UNLOCK(rt);
 179                 }
 180         }
 181 #endif /* PF */
 182
 183 #if IPSEC
 184         /*
 185          * Check AH/ESP integrity.
 186          */
 187         /*
 188          * Don't increment ip6s_cantforward because this is the check
 189          * before forwarding packet actually.
 190          */
 191         if (ipsec_bypass == 0) {
 192                 if (ipsec6_in_reject(m, NULL)) {
 193                         IPSEC_STAT_INCREMENT(ipsec6stat.in_polvio);
 194                         m_freem(m);
 195                         return (NULL);
 196                 }
 197         }
 198 #endif /*IPSEC*/
 199
 200         /*
 201          * Do not forward packets to multicast destination.
 202          * Do not forward packets with unspecified source.  It was discussed
 203          * in July 2000, on ipngwg mailing list.
 204          */
 205         if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
 206             IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
 207             IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
 208                 ip6stat.ip6s_cantforward++;
 209                 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
 210                 if (ip6_log_time + ip6_log_interval < curtime) {
 211                         ip6_log_time = curtime;
 212                         log(LOG_DEBUG,
 213                             "cannot forward "
 214                             "from %s to %s nxt %d received on %s\n",
 215                             ip6_sprintf(&ip6->ip6_src),
 216                             ip6_sprintf(&ip6->ip6_dst),
 217                             ip6->ip6_nxt,
 218                             if_name(m->m_pkthdr.rcvif));
 219                 }
 220                 m_freem(m);
 221                 return (NULL);
 222         }
 223
 224         if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
 225                 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
 226                 icmp6_error(m, ICMP6_TIME_EXCEEDED,
 227                                 ICMP6_TIME_EXCEED_TRANSIT, 0);
 228                 return (NULL);
 229         }
 230
 231         /*
 232          * See if the destination is a proxied address, and if so pretend
 233          * that it's for us.  This is mostly to handle NUD probes against
 234          * the proxied addresses.  We filter for ICMPv6 here and will let
 235          * icmp6_input handle the rest.
 236          */
 237         if (!srcrt && nd6_prproxy) {
 238                 VERIFY(!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst));
 239                 proxy = nd6_prproxy_isours(m, ip6, ip6forward_rt, ifscope);
 240                 /*
 241                  * Don't update hop limit while proxying; RFC 4389 4.1.
 242                  * Also skip IPsec forwarding path processing as this
 243                  * packet is not to be forwarded.
 244                  */
 245                 if (proxy)
 246                         goto skip_ipsec;
 247         }
 248
 249         ip6->ip6_hlim -= IPV6_HLIMDEC;
 250
 251         /*
 252          * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
 253          * size of IPv6 + ICMPv6 headers) bytes of the packet in case
 254          * we need to generate an ICMP6 message to the src.
 255          * Thanks to M_EXT, in most cases copy will not occur.
 256          *
 257          * It is important to save it before IPsec processing as IPsec
 258          * processing may modify the mbuf.
 259          */
 260         mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
 261
 262 #if IPSEC
 263         if (ipsec_bypass != 0)
 264                 goto skip_ipsec;
 265         /* get a security policy for this packet */
 266         sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, IP_FORWARDING,
 267             &error);
 268         if (sp == NULL) {
 269                 IPSEC_STAT_INCREMENT(ipsec6stat.out_inval);
 270                 ip6stat.ip6s_cantforward++;
 271                 if (mcopy) {
 272 #if 0
 273                         /* XXX: what icmp ? */
 274 #else
 275                         m_freem(mcopy);
 276 #endif
 277                 }
 278                 m_freem(m);
 279                 return (NULL);
 280         }
 281
 282         error = 0;
 283
 284         /* check policy */
 285         switch (sp->policy) {
 286         case IPSEC_POLICY_DISCARD:
 287         case IPSEC_POLICY_GENERATE:
 288                 /*
 289                  * This packet is just discarded.
 290                  */
 291                 IPSEC_STAT_INCREMENT(ipsec6stat.out_polvio);
 292                 ip6stat.ip6s_cantforward++;
 293                 key_freesp(sp, KEY_SADB_UNLOCKED);
 294                 if (mcopy) {
 295 #if 0
 296                         /* XXX: what icmp ? */
 297 #else
 298                         m_freem(mcopy);
 299 #endif
 300                 }
 301                 m_freem(m);
 302                 return (NULL);
 303
 304         case IPSEC_POLICY_BYPASS:
 305         case IPSEC_POLICY_NONE:
 306                 /* no need to do IPsec. */
 307                 key_freesp(sp, KEY_SADB_UNLOCKED);
 308                 goto skip_ipsec;
 309
 310         case IPSEC_POLICY_IPSEC:
 311                 if (sp->req == NULL) {
 312                         /* XXX should be panic ? */
 313                         printf("ip6_forward: No IPsec request specified.\n");
 314                         ip6stat.ip6s_cantforward++;
 315                         key_freesp(sp, KEY_SADB_UNLOCKED);
 316                         if (mcopy) {
 317 #if 0
 318                                 /* XXX: what icmp ? */
 319 #else
 320                                 m_freem(mcopy);
 321 #endif
 322                         }
 323                         m_freem(m);
 324                         return (NULL);
 325                 }
 326                 /* do IPsec */
 327                 break;
 328
 329         case IPSEC_POLICY_ENTRUST:
 330         default:
 331                 /* should be panic ?? */
 332                 printf("ip6_forward: Invalid policy found. %d\n", sp->policy);
 333                 key_freesp(sp, KEY_SADB_UNLOCKED);
 334                 goto skip_ipsec;
 335         }
 336
 337     {
 338         struct ipsec_output_state state;
 339
 340         /*
 341          * All the extension headers will become inaccessible
 342          * (since they can be encrypted).
 343          * Don't panic, we need no more updates to extension headers
 344          * on inner IPv6 packet (since they are now encapsulated).
 345          *
 346          * IPv6 [ESP|AH] IPv6 [extension headers] payload
 347          */
 348         bzero(&state, sizeof(state));
 349         state.m = m;
 350         state.dst = NULL;       /* update at ipsec6_output_tunnel() */
 351
 352         error = ipsec6_output_tunnel(&state, sp, 0);
 353         key_freesp(sp, KEY_SADB_UNLOCKED);
 354         if (state.tunneled == 4) {
 355                 ROUTE_RELEASE(&state.ro);
 356                 return (NULL);  /* packet is gone - sent over IPv4 */
 357         }
 358
 359         m = state.m;
 360         ROUTE_RELEASE(&state.ro);
 361
 362         if (error) {
 363                 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
 364                 switch (error) {
 365                 case EHOSTUNREACH:
 366                 case ENETUNREACH:
 367                 case EMSGSIZE:
 368                 case ENOBUFS:
 369                 case ENOMEM:
 370                         break;
 371                 default:
 372                         printf("ip6_output (ipsec): error code %d\n", error);
 373                         /* fall through */
 374                 case ENOENT:
 375                         /* don't show these error codes to the user */
 376                         break;
 377                 }
 378                 ip6stat.ip6s_cantforward++;
 379                 if (mcopy) {
 380 #if 0
 381                         /* XXX: what icmp ? */
 382 #else
 383                         m_freem(mcopy);
 384 #endif
 385                 }
 386                 m_freem(m);
 387                 return (NULL);
 388         }
 389     }
 390 #endif /* IPSEC */
 391     skip_ipsec:
 392
 393         dst = (struct sockaddr_in6 *)&ip6forward_rt->ro_dst;
 394         if ((rt = ip6forward_rt->ro_rt) != NULL) {
 395                 RT_LOCK(rt);
 396                 /* Take an extra ref for ourselves */
 397                 RT_ADDREF_LOCKED(rt);
 398         }
 399
 400         VERIFY(rt == NULL || rt == ip6forward_rt->ro_rt);
 401         if (!srcrt) {
 402                 /*
 403                  * ip6forward_rt->ro_dst.sin6_addr is equal to ip6->ip6_dst
 404                  */
 405                 if (ROUTE_UNUSABLE(ip6forward_rt)) {
 406                         if (rt != NULL) {
 407                                 /* Release extra ref */
 408                                 RT_REMREF_LOCKED(rt);
 409                                 RT_UNLOCK(rt);
 410                         }
 411                         ROUTE_RELEASE(ip6forward_rt);
 412
 413                         /* this probably fails but give it a try again */
 414                         rtalloc_scoped_ign((struct route *)ip6forward_rt,
 415                             RTF_PRCLONING, ifscope);
 416                         if ((rt = ip6forward_rt->ro_rt) != NULL) {
 417                                 RT_LOCK(rt);
 418                                 /* Take an extra ref for ourselves */
 419                                 RT_ADDREF_LOCKED(rt);
 420                         }
 421                 }
 422
 423                 if (rt == NULL) {
 424                         ip6stat.ip6s_noroute++;
 425                         in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
 426                         if (mcopy)
 427                                 icmp6_error(mcopy, ICMP6_DST_UNREACH,
 428                                             ICMP6_DST_UNREACH_NOROUTE, 0);
 429                         m_freem(m);
 430                         return (NULL);
 431                 }
 432                 RT_LOCK_ASSERT_HELD(rt);
 433         } else if (ROUTE_UNUSABLE(ip6forward_rt) ||
 434             !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) {
 435                 if (rt != NULL) {
 436                         /* Release extra ref */
 437                         RT_REMREF_LOCKED(rt);
 438                         RT_UNLOCK(rt);
 439                 }
 440                 ROUTE_RELEASE(ip6forward_rt);
 441
 442                 bzero(dst, sizeof(*dst));
 443                 dst->sin6_len = sizeof(struct sockaddr_in6);
 444                 dst->sin6_family = AF_INET6;
 445                 dst->sin6_addr = ip6->ip6_dst;
 446
 447                 rtalloc_scoped_ign((struct route *)ip6forward_rt,
 448                     RTF_PRCLONING, ifscope);
 449                 if ((rt = ip6forward_rt->ro_rt) == NULL) {
 450                         ip6stat.ip6s_noroute++;
 451                         in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
 452                         if (mcopy)
 453                                 icmp6_error(mcopy, ICMP6_DST_UNREACH,
 454                                     ICMP6_DST_UNREACH_NOROUTE, 0);
 455                         m_freem(m);
 456                         return (NULL);
 457                 }
 458                 RT_LOCK(rt);
 459                 /* Take an extra ref for ourselves */
 460                 RT_ADDREF_LOCKED(rt);
 461         }
 462
 463         /*
 464          * Source scope check: if a packet can't be delivered to its
 465          * destination for the reason that the destination is beyond the scope
 466          * of the source address, discard the packet and return an icmp6
 467          * destination unreachable error with Code 2 (beyond scope of source
 468          * address) unless we are proxying (source address is link local
 469          * for NUDs.)  We use a local copy of ip6_src, since in6_setscope()
 470          * will possibly modify its first argument.
 471          * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1]
 472          */
 473         src_in6 = ip6->ip6_src;
 474         if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) {
 475                 /* XXX: this should not happen */
 476                 ip6stat.ip6s_cantforward++;
 477                 ip6stat.ip6s_badscope++;
 478                 m_freem(m);
 479                 return (NULL);
 480         }
 481         if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) {
 482                 ip6stat.ip6s_cantforward++;
 483                 ip6stat.ip6s_badscope++;
 484                 m_freem(m);
 485                 return (NULL);
 486         }
 487
 488         if (inzone != outzone && !proxy) {
 489                 ip6stat.ip6s_cantforward++;
 490                 ip6stat.ip6s_badscope++;
 491                 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
 492
 493                 if (ip6_log_time + ip6_log_interval < curtime) {
 494                         ip6_log_time = curtime;
 495                         log(LOG_DEBUG,
 496                             "cannot forward "
 497                             "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
 498                             ip6_sprintf(&ip6->ip6_src),
 499                             ip6_sprintf(&ip6->ip6_dst),
 500                             ip6->ip6_nxt,
 501                             if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
 502                 }
 503                 /* Release extra ref */
 504                 RT_REMREF_LOCKED(rt);
 505                 RT_UNLOCK(rt);
 506                 if (mcopy) {
 507                         icmp6_error(mcopy, ICMP6_DST_UNREACH,
 508                                     ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
 509                 }
 510                 m_freem(m);
 511                 return (NULL);
 512         }
 513
 514         /*
 515          * Destination scope check: if a packet is going to break the scope
 516          * zone of packet's destination address, discard it.  This case should
 517          * usually be prevented by appropriately-configured routing table, but
 518          * we need an explicit check because we may mistakenly forward the
 519          * packet to a different zone by (e.g.) a default route.
 520          */
 521         dst_in6 = ip6->ip6_dst;
 522         if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 ||
 523             in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 ||
 524             inzone != outzone) {
 525                 ip6stat.ip6s_cantforward++;
 526                 ip6stat.ip6s_badscope++;
 527                 m_freem(m);
 528                 return (NULL);
 529         }
 530
 531         if (m->m_pkthdr.len > rt->rt_ifp->if_mtu) {
 532                 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
 533                 if (mcopy) {
 534                         uint32_t mtu;
 535 #if IPSEC
 536                         struct secpolicy *sp2;
 537                         int ipsecerror;
 538                         size_t ipsechdrsiz;
 539 #endif
 540
 541                         mtu = rt->rt_ifp->if_mtu;
 542 #if IPSEC
 543                         /*
 544                          * When we do IPsec tunnel ingress, we need to play
 545                          * with the link value (decrement IPsec header size
 546                          * from mtu value).  The code is much simpler than v4
 547                          * case, as we have the outgoing interface for
 548                          * encapsulated packet as "rt->rt_ifp".
 549                          */
 550                         sp2 = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
 551                                 IP_FORWARDING, &ipsecerror);
 552                         if (sp2) {
 553                                 ipsechdrsiz = ipsec6_hdrsiz(mcopy,
 554                                         IPSEC_DIR_OUTBOUND, NULL);
 555                                 if (ipsechdrsiz < mtu)
 556                                         mtu -= ipsechdrsiz;
 557                                 key_freesp(sp2, KEY_SADB_UNLOCKED);
 558                         }
 559                         /*
 560                          * if mtu becomes less than minimum MTU,
 561                          * tell minimum MTU (and I'll need to fragment it).
 562                          */
 563                         if (mtu < IPV6_MMTU)
 564                                 mtu = IPV6_MMTU;
 565 #endif
 566                         /* Release extra ref */
 567                         RT_REMREF_LOCKED(rt);
 568                         RT_UNLOCK(rt);
 569                         icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
 570                 } else {
 571                         /* Release extra ref */
 572                         RT_REMREF_LOCKED(rt);
 573                         RT_UNLOCK(rt);
 574                 }
 575                 m_freem(m);
 576                 return (NULL);
 577         }
 578
 579         if (rt->rt_flags & RTF_GATEWAY)
 580                 dst = (struct sockaddr_in6 *)(void *)rt->rt_gateway;
 581
 582         /*
 583          * If we are to forward the packet using the same interface
 584          * as one we got the packet from, perhaps we should send a redirect
 585          * to sender to shortcut a hop.
 586          * Only send redirect if source is sending directly to us,
 587          * and if packet was not source routed (or has any options).
 588          * Also, don't send redirect if forwarding using a route
 589          * modified by a redirect.
 590          */
 591         if (!proxy &&
 592             ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
 593             (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
 594                 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) {
 595                         /*
 596                          * If the incoming interface is equal to the outgoing
 597                          * one, and the link attached to the interface is
 598                          * point-to-point, then it will be highly probable
 599                          * that a routing loop occurs. Thus, we immediately
 600                          * drop the packet and send an ICMPv6 error message.
 601                          *
 602                          * type/code is based on suggestion by Rich Draves.
 603                          * not sure if it is the best pick.
 604                          */
 605                         RT_REMREF_LOCKED(rt);   /* Release extra ref */
 606                         RT_UNLOCK(rt);
 607                         icmp6_error(mcopy, ICMP6_DST_UNREACH,
 608                                     ICMP6_DST_UNREACH_ADDR, 0);
 609                         m_freem(m);
 610                         return (NULL);
 611                 }
 612                 type = ND_REDIRECT;
 613         }
 614
 615 #if IPFW2
 616         /*
 617          * Check with the firewall...
 618          */
 619         if (ip6_fw_enable && ip6_fw_chk_ptr) {
 620                 u_short port = 0;
 621                 ifp = rt->rt_ifp;
 622                 /* Drop the lock but retain the extra ref */
 623                 RT_UNLOCK(rt);
 624                 /* If ipfw says divert, we have to just drop packet */
 625                 if (ip6_fw_chk_ptr(&ip6, ifp, &port, &m)) {
 626                         m_freem(m);
 627                         goto freecopy;
 628                 }
 629                 if (!m) {
 630                         goto freecopy;
 631                 }
 632                 /* We still have the extra ref on rt */
 633                 RT_LOCK(rt);
 634         }
 635 #endif
 636
 637         /*
 638          * Fake scoped addresses. Note that even link-local source or
 639          * destinaion can appear, if the originating node just sends the
 640          * packet to us (without address resolution for the destination).
 641          * Since both icmp6_error and icmp6_redirect_output fill the embedded
 642          * link identifiers, we can do this stuff after making a copy for
 643          * returning an error.
 644          */
 645         if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
 646                 /*
 647                  * See corresponding comments in ip6_output.
 648                  * XXX: but is it possible that ip6_forward() sends a packet
 649                  *      to a loopback interface? I don't think so, and thus
 650                  *      I bark here. (jinmei@kame.net)
 651                  * XXX: it is common to route invalid packets to loopback.
 652                  *      also, the codepath will be visited on use of ::1 in
 653                  *      rthdr. (itojun)
 654                  */
 655 #if 1
 656                 if ((0))
 657 #else
 658                 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
 659 #endif
 660                 {
 661                         printf("ip6_forward: outgoing interface is loopback. "
 662                                 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
 663                                 ip6_sprintf(&ip6->ip6_src),
 664                                 ip6_sprintf(&ip6->ip6_dst),
 665                                 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
 666                                 if_name(rt->rt_ifp));
 667                 }
 668
 669                 /* we can just use rcvif in forwarding. */
 670                 origifp = rcvifp = m->m_pkthdr.rcvif;
 671         } else if (nd6_prproxy) {
 672                 /*
 673                  * In the prefix proxying case, we need to inform nd6_output()
 674                  * about the inbound interface, so that any subsequent NS
 675                  * packets generated by nd6_prproxy_ns_output() will not be
 676                  * sent back to that same interface.
 677                  */
 678                 origifp = rcvifp = m->m_pkthdr.rcvif;
 679         } else {
 680                 rcvifp = m->m_pkthdr.rcvif;
 681                 origifp = rt->rt_ifp;
 682         }
 683         /*
 684          * clear embedded scope identifiers if necessary.
 685          * in6_clearscope will touch the addresses only when necessary.
 686          */
 687         in6_clearscope(&ip6->ip6_src);
 688         in6_clearscope(&ip6->ip6_dst);
 689
 690         ifp = rt->rt_ifp;
 691         /* Drop the lock but retain the extra ref */
 692         RT_UNLOCK(rt);
 693
 694         /*
 695          * If this is to be processed locally, let ip6_input have it.
 696          */
 697         if (proxy) {
 698                 VERIFY(m->m_pkthdr.pkt_flags & PKTF_PROXY_DST);
 699                 /* Release extra ref */
 700                 RT_REMREF(rt);
 701                 if (mcopy != NULL)
 702                         m_freem(mcopy);
 703                 return (m);
 704         }
 705
 706 #if PF
 707         /* Invoke outbound packet filter */
 708         error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, NULL);
 709
 710         if (error != 0 || m == NULL) {
 711                 if (m != NULL) {
 712                         panic("%s: unexpected packet %p\n", __func__, m);
 713                         /* NOTREACHED */
 714                 }
 715                 /* Already freed by callee */
 716                 goto senderr;
 717         }
 718         ip6 = mtod(m, struct ip6_hdr *);
 719 #endif /* PF */
 720
 721         /* Mark this packet as being forwarded from another interface */
 722         m->m_pkthdr.pkt_flags |= PKTF_FORWARDED;
 723         len = m_pktlen(m);
 724
 725         error = nd6_output(ifp, origifp, m, dst, rt, NULL);
 726         if (error) {
 727                 in6_ifstat_inc(ifp, ifs6_out_discard);
 728                 ip6stat.ip6s_cantforward++;
 729         } else {
 730                 /*
 731                  * Increment stats on the source interface; the ones
 732                  * for destination interface has been taken care of
 733                  * during output above by virtue of PKTF_FORWARDED.
 734                  */
 735                 rcvifp->if_fpackets++;
 736                 rcvifp->if_fbytes += len;
 737
 738                 ip6stat.ip6s_forward++;
 739                 in6_ifstat_inc(ifp, ifs6_out_forward);
 740                 if (type)
 741                         ip6stat.ip6s_redirectsent++;
 742                 else {
 743                         if (mcopy) {
 744                                 goto freecopy;
 745                         }
 746                 }
 747         }
 748 #if PF
 749 senderr:
 750 #endif /* PF */
 751         if (mcopy == NULL) {
 752                 /* Release extra ref */
 753                 RT_REMREF(rt);
 754                 return (NULL);
 755         }
 756         switch (error) {
 757         case 0:
 758 #if 1
 759                 if (type == ND_REDIRECT) {
 760                         icmp6_redirect_output(mcopy, rt);
 761                         /* Release extra ref */
 762                         RT_REMREF(rt);
 763                         return (NULL);
 764                 }
 765 #endif
 766                 goto freecopy;
 767
 768         case EMSGSIZE:
 769                 /* xxx MTU is constant in PPP? */
 770                 goto freecopy;
 771
 772         case ENOBUFS:
 773                 /* Tell source to slow down like source quench in IP? */
 774                 goto freecopy;
 775
 776         case ENETUNREACH:       /* shouldn't happen, checked above */
 777         case EHOSTUNREACH:
 778         case ENETDOWN:
 779         case EHOSTDOWN:
 780         default:
 781                 type = ICMP6_DST_UNREACH;
 782                 code = ICMP6_DST_UNREACH_ADDR;
 783                 break;
 784         }
 785         icmp6_error(mcopy, type, code, 0);
 786         /* Release extra ref */
 787         RT_REMREF(rt);
 788         return (NULL);
 789
 790  freecopy:
 791         m_freem(mcopy);
 792         /* Release extra ref */
 793         RT_REMREF(rt);
 794         return (NULL);
 795 }