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 (should be handled
 202          * by ip6_mforward().
 203          * Do not forward packets with unspecified source.  It was discussed
 204          * in July 2000, on ipngwg mailing list.
 205          */
 206         if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
 207             IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
 208             IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
 209                 ip6stat.ip6s_cantforward++;
 210                 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
 211                 if (ip6_log_time + ip6_log_interval < curtime) {
 212                         ip6_log_time = curtime;
 213                         log(LOG_DEBUG,
 214                             "cannot forward "
 215                             "from %s to %s nxt %d received on %s\n",
 216                             ip6_sprintf(&ip6->ip6_src),
 217                             ip6_sprintf(&ip6->ip6_dst),
 218                             ip6->ip6_nxt,
 219                             if_name(m->m_pkthdr.rcvif));
 220                 }
 221                 m_freem(m);
 222                 return (NULL);
 223         }
 224
 225         if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
 226                 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
 227                 icmp6_error(m, ICMP6_TIME_EXCEEDED,
 228                                 ICMP6_TIME_EXCEED_TRANSIT, 0);
 229                 return (NULL);
 230         }
 231
 232         /*
 233          * See if the destination is a proxied address, and if so pretend
 234          * that it's for us.  This is mostly to handle NUD probes against
 235          * the proxied addresses.  We filter for ICMPv6 here and will let
 236          * icmp6_input handle the rest.
 237          */
 238         if (!srcrt && nd6_prproxy) {
 239                 VERIFY(!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst));
 240                 proxy = nd6_prproxy_isours(m, ip6, ip6forward_rt, ifscope);
 241                 /*
 242                  * Don't update hop limit while proxying; RFC 4389 4.1.
 243                  * Also skip IPsec forwarding path processing as this
 244                  * packet is not to be forwarded.
 245                  */
 246                 if (proxy)
 247                         goto skip_ipsec;
 248         }
 249
 250         ip6->ip6_hlim -= IPV6_HLIMDEC;
 251
 252         /*
 253          * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
 254          * size of IPv6 + ICMPv6 headers) bytes of the packet in case
 255          * we need to generate an ICMP6 message to the src.
 256          * Thanks to M_EXT, in most cases copy will not occur.
 257          *
 258          * It is important to save it before IPsec processing as IPsec
 259          * processing may modify the mbuf.
 260          */
 261         mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
 262
 263 #if IPSEC
 264         if (ipsec_bypass != 0)
 265                 goto skip_ipsec;
 266         /* get a security policy for this packet */
 267         sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, IP_FORWARDING,
 268             &error);
 269         if (sp == NULL) {
 270                 IPSEC_STAT_INCREMENT(ipsec6stat.out_inval);
 271                 ip6stat.ip6s_cantforward++;
 272                 if (mcopy) {
 273 #if 0
 274                         /* XXX: what icmp ? */
 275 #else
 276                         m_freem(mcopy);
 277 #endif
 278                 }
 279                 m_freem(m);
 280                 return (NULL);
 281         }
 282
 283         error = 0;
 284
 285         /* check policy */
 286         switch (sp->policy) {
 287         case IPSEC_POLICY_DISCARD:
 288         case IPSEC_POLICY_GENERATE:
 289                 /*
 290                  * This packet is just discarded.
 291                  */
 292                 IPSEC_STAT_INCREMENT(ipsec6stat.out_polvio);
 293                 ip6stat.ip6s_cantforward++;
 294                 key_freesp(sp, KEY_SADB_UNLOCKED);
 295                 if (mcopy) {
 296 #if 0
 297                         /* XXX: what icmp ? */
 298 #else
 299                         m_freem(mcopy);
 300 #endif
 301                 }
 302                 m_freem(m);
 303                 return (NULL);
 304
 305         case IPSEC_POLICY_BYPASS:
 306         case IPSEC_POLICY_NONE:
 307                 /* no need to do IPsec. */
 308                 key_freesp(sp, KEY_SADB_UNLOCKED);
 309                 goto skip_ipsec;
 310
 311         case IPSEC_POLICY_IPSEC:
 312                 if (sp->req == NULL) {
 313                         /* XXX should be panic ? */
 314                         printf("ip6_forward: No IPsec request specified.\n");
 315                         ip6stat.ip6s_cantforward++;
 316                         key_freesp(sp, KEY_SADB_UNLOCKED);
 317                         if (mcopy) {
 318 #if 0
 319                                 /* XXX: what icmp ? */
 320 #else
 321                                 m_freem(mcopy);
 322 #endif
 323                         }
 324                         m_freem(m);
 325                         return (NULL);
 326                 }
 327                 /* do IPsec */
 328                 break;
 329
 330         case IPSEC_POLICY_ENTRUST:
 331         default:
 332                 /* should be panic ?? */
 333                 printf("ip6_forward: Invalid policy found. %d\n", sp->policy);
 334                 key_freesp(sp, KEY_SADB_UNLOCKED);
 335                 goto skip_ipsec;
 336         }
 337
 338     {
 339         struct ipsec_output_state state;
 340
 341         /*
 342          * All the extension headers will become inaccessible
 343          * (since they can be encrypted).
 344          * Don't panic, we need no more updates to extension headers
 345          * on inner IPv6 packet (since they are now encapsulated).
 346          *
 347          * IPv6 [ESP|AH] IPv6 [extension headers] payload
 348          */
 349         bzero(&state, sizeof(state));
 350         state.m = m;
 351         state.dst = NULL;       /* update at ipsec6_output_tunnel() */
 352
 353         error = ipsec6_output_tunnel(&state, sp, 0);
 354         key_freesp(sp, KEY_SADB_UNLOCKED);
 355         if (state.tunneled == 4) {
 356                 ROUTE_RELEASE(&state.ro);
 357                 return (NULL);  /* packet is gone - sent over IPv4 */
 358         }
 359
 360         m = state.m;
 361         ROUTE_RELEASE(&state.ro);
 362
 363         if (error) {
 364                 /* mbuf is already reclaimed in ipsec6_output_tunnel. */
 365                 switch (error) {
 366                 case EHOSTUNREACH:
 367                 case ENETUNREACH:
 368                 case EMSGSIZE:
 369                 case ENOBUFS:
 370                 case ENOMEM:
 371                         break;
 372                 default:
 373                         printf("ip6_output (ipsec): error code %d\n", error);
 374                         /* fall through */
 375                 case ENOENT:
 376                         /* don't show these error codes to the user */
 377                         break;
 378                 }
 379                 ip6stat.ip6s_cantforward++;
 380                 if (mcopy) {
 381 #if 0
 382                         /* XXX: what icmp ? */
 383 #else
 384                         m_freem(mcopy);
 385 #endif
 386                 }
 387                 m_freem(m);
 388                 return (NULL);
 389         }
 390     }
 391 #endif /* IPSEC */
 392     skip_ipsec:
 393
 394         dst = (struct sockaddr_in6 *)&ip6forward_rt->ro_dst;
 395         if ((rt = ip6forward_rt->ro_rt) != NULL) {
 396                 RT_LOCK(rt);
 397                 /* Take an extra ref for ourselves */
 398                 RT_ADDREF_LOCKED(rt);
 399         }
 400
 401         VERIFY(rt == NULL || rt == ip6forward_rt->ro_rt);
 402         if (!srcrt) {
 403                 /*
 404                  * ip6forward_rt->ro_dst.sin6_addr is equal to ip6->ip6_dst
 405                  */
 406                 if (ROUTE_UNUSABLE(ip6forward_rt)) {
 407                         if (rt != NULL) {
 408                                 /* Release extra ref */
 409                                 RT_REMREF_LOCKED(rt);
 410                                 RT_UNLOCK(rt);
 411                         }
 412                         ROUTE_RELEASE(ip6forward_rt);
 413
 414                         /* this probably fails but give it a try again */
 415                         rtalloc_scoped_ign((struct route *)ip6forward_rt,
 416                             RTF_PRCLONING, ifscope);
 417                         if ((rt = ip6forward_rt->ro_rt) != NULL) {
 418                                 RT_LOCK(rt);
 419                                 /* Take an extra ref for ourselves */
 420                                 RT_ADDREF_LOCKED(rt);
 421                         }
 422                 }
 423
 424                 if (rt == NULL) {
 425                         ip6stat.ip6s_noroute++;
 426                         in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
 427                         if (mcopy)
 428                                 icmp6_error(mcopy, ICMP6_DST_UNREACH,
 429                                             ICMP6_DST_UNREACH_NOROUTE, 0);
 430                         m_freem(m);
 431                         return (NULL);
 432                 }
 433                 RT_LOCK_ASSERT_HELD(rt);
 434         } else if (ROUTE_UNUSABLE(ip6forward_rt) ||
 435             !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) {
 436                 if (rt != NULL) {
 437                         /* Release extra ref */
 438                         RT_REMREF_LOCKED(rt);
 439                         RT_UNLOCK(rt);
 440                 }
 441                 ROUTE_RELEASE(ip6forward_rt);
 442
 443                 bzero(dst, sizeof(*dst));
 444                 dst->sin6_len = sizeof(struct sockaddr_in6);
 445                 dst->sin6_family = AF_INET6;
 446                 dst->sin6_addr = ip6->ip6_dst;
 447
 448                 rtalloc_scoped_ign((struct route *)ip6forward_rt,
 449                     RTF_PRCLONING, ifscope);
 450                 if ((rt = ip6forward_rt->ro_rt) == NULL) {
 451                         ip6stat.ip6s_noroute++;
 452                         in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
 453                         if (mcopy)
 454                                 icmp6_error(mcopy, ICMP6_DST_UNREACH,
 455                                     ICMP6_DST_UNREACH_NOROUTE, 0);
 456                         m_freem(m);
 457                         return (NULL);
 458                 }
 459                 RT_LOCK(rt);
 460                 /* Take an extra ref for ourselves */
 461                 RT_ADDREF_LOCKED(rt);
 462         }
 463
 464         /*
 465          * Source scope check: if a packet can't be delivered to its
 466          * destination for the reason that the destination is beyond the scope
 467          * of the source address, discard the packet and return an icmp6
 468          * destination unreachable error with Code 2 (beyond scope of source
 469          * address) unless we are proxying (source address is link local
 470          * for NUDs.)  We use a local copy of ip6_src, since in6_setscope()
 471          * will possibly modify its first argument.
 472          * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1]
 473          */
 474         src_in6 = ip6->ip6_src;
 475         if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) {
 476                 /* XXX: this should not happen */
 477                 ip6stat.ip6s_cantforward++;
 478                 ip6stat.ip6s_badscope++;
 479                 m_freem(m);
 480                 return (NULL);
 481         }
 482         if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) {
 483                 ip6stat.ip6s_cantforward++;
 484                 ip6stat.ip6s_badscope++;
 485                 m_freem(m);
 486                 return (NULL);
 487         }
 488
 489         if (inzone != outzone && !proxy) {
 490                 ip6stat.ip6s_cantforward++;
 491                 ip6stat.ip6s_badscope++;
 492                 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
 493
 494                 if (ip6_log_time + ip6_log_interval < curtime) {
 495                         ip6_log_time = curtime;
 496                         log(LOG_DEBUG,
 497                             "cannot forward "
 498                             "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
 499                             ip6_sprintf(&ip6->ip6_src),
 500                             ip6_sprintf(&ip6->ip6_dst),
 501                             ip6->ip6_nxt,
 502                             if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
 503                 }
 504                 /* Release extra ref */
 505                 RT_REMREF_LOCKED(rt);
 506                 RT_UNLOCK(rt);
 507                 if (mcopy) {
 508                         icmp6_error(mcopy, ICMP6_DST_UNREACH,
 509                                     ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
 510                 }
 511                 m_freem(m);
 512                 return (NULL);
 513         }
 514
 515         /*
 516          * Destination scope check: if a packet is going to break the scope
 517          * zone of packet's destination address, discard it.  This case should
 518          * usually be prevented by appropriately-configured routing table, but
 519          * we need an explicit check because we may mistakenly forward the
 520          * packet to a different zone by (e.g.) a default route.
 521          */
 522         dst_in6 = ip6->ip6_dst;
 523         if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 ||
 524             in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 ||
 525             inzone != outzone) {
 526                 ip6stat.ip6s_cantforward++;
 527                 ip6stat.ip6s_badscope++;
 528                 m_freem(m);
 529                 return (NULL);
 530         }
 531
 532         if (m->m_pkthdr.len > rt->rt_ifp->if_mtu) {
 533                 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
 534                 if (mcopy) {
 535                         uint32_t mtu;
 536 #if IPSEC
 537                         struct secpolicy *sp2;
 538                         int ipsecerror;
 539                         size_t ipsechdrsiz;
 540 #endif
 541
 542                         mtu = rt->rt_ifp->if_mtu;
 543 #if IPSEC
 544                         /*
 545                          * When we do IPsec tunnel ingress, we need to play
 546                          * with the link value (decrement IPsec header size
 547                          * from mtu value).  The code is much simpler than v4
 548                          * case, as we have the outgoing interface for
 549                          * encapsulated packet as "rt->rt_ifp".
 550                          */
 551                         sp2 = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
 552                                 IP_FORWARDING, &ipsecerror);
 553                         if (sp2) {
 554                                 ipsechdrsiz = ipsec6_hdrsiz(mcopy,
 555                                         IPSEC_DIR_OUTBOUND, NULL);
 556                                 if (ipsechdrsiz < mtu)
 557                                         mtu -= ipsechdrsiz;
 558                                 key_freesp(sp2, KEY_SADB_UNLOCKED);
 559                         }
 560                         /*
 561                          * if mtu becomes less than minimum MTU,
 562                          * tell minimum MTU (and I'll need to fragment it).
 563                          */
 564                         if (mtu < IPV6_MMTU)
 565                                 mtu = IPV6_MMTU;
 566 #endif
 567                         /* Release extra ref */
 568                         RT_REMREF_LOCKED(rt);
 569                         RT_UNLOCK(rt);
 570                         icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
 571                 } else {
 572                         /* Release extra ref */
 573                         RT_REMREF_LOCKED(rt);
 574                         RT_UNLOCK(rt);
 575                 }
 576                 m_freem(m);
 577                 return (NULL);
 578         }
 579
 580         if (rt->rt_flags & RTF_GATEWAY)
 581                 dst = (struct sockaddr_in6 *)(void *)rt->rt_gateway;
 582
 583         /*
 584          * If we are to forward the packet using the same interface
 585          * as one we got the packet from, perhaps we should send a redirect
 586          * to sender to shortcut a hop.
 587          * Only send redirect if source is sending directly to us,
 588          * and if packet was not source routed (or has any options).
 589          * Also, don't send redirect if forwarding using a route
 590          * modified by a redirect.
 591          */
 592         if (!proxy &&
 593             ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
 594             (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
 595                 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) {
 596                         /*
 597                          * If the incoming interface is equal to the outgoing
 598                          * one, and the link attached to the interface is
 599                          * point-to-point, then it will be highly probable
 600                          * that a routing loop occurs. Thus, we immediately
 601                          * drop the packet and send an ICMPv6 error message.
 602                          *
 603                          * type/code is based on suggestion by Rich Draves.
 604                          * not sure if it is the best pick.
 605                          */
 606                         RT_REMREF_LOCKED(rt);   /* Release extra ref */
 607                         RT_UNLOCK(rt);
 608                         icmp6_error(mcopy, ICMP6_DST_UNREACH,
 609                                     ICMP6_DST_UNREACH_ADDR, 0);
 610                         m_freem(m);
 611                         return (NULL);
 612                 }
 613                 type = ND_REDIRECT;
 614         }
 615
 616 #if IPFW2
 617         /*
 618          * Check with the firewall...
 619          */
 620         if (ip6_fw_enable && ip6_fw_chk_ptr) {
 621                 u_short port = 0;
 622                 ifp = rt->rt_ifp;
 623                 /* Drop the lock but retain the extra ref */
 624                 RT_UNLOCK(rt);
 625                 /* If ipfw says divert, we have to just drop packet */
 626                 if (ip6_fw_chk_ptr(&ip6, ifp, &port, &m)) {
 627                         m_freem(m);
 628                         goto freecopy;
 629                 }
 630                 if (!m) {
 631                         goto freecopy;
 632                 }
 633                 /* We still have the extra ref on rt */
 634                 RT_LOCK(rt);
 635         }
 636 #endif
 637
 638         /*
 639          * Fake scoped addresses. Note that even link-local source or
 640          * destinaion can appear, if the originating node just sends the
 641          * packet to us (without address resolution for the destination).
 642          * Since both icmp6_error and icmp6_redirect_output fill the embedded
 643          * link identifiers, we can do this stuff after making a copy for
 644          * returning an error.
 645          */
 646         if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
 647                 /*
 648                  * See corresponding comments in ip6_output.
 649                  * XXX: but is it possible that ip6_forward() sends a packet
 650                  *      to a loopback interface? I don't think so, and thus
 651                  *      I bark here. (jinmei@kame.net)
 652                  * XXX: it is common to route invalid packets to loopback.
 653                  *      also, the codepath will be visited on use of ::1 in
 654                  *      rthdr. (itojun)
 655                  */
 656 #if 1
 657                 if (0)
 658 #else
 659                 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
 660 #endif
 661                 {
 662                         printf("ip6_forward: outgoing interface is loopback. "
 663                                 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
 664                                 ip6_sprintf(&ip6->ip6_src),
 665                                 ip6_sprintf(&ip6->ip6_dst),
 666                                 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
 667                                 if_name(rt->rt_ifp));
 668                 }
 669
 670                 /* we can just use rcvif in forwarding. */
 671                 origifp = rcvifp = m->m_pkthdr.rcvif;
 672         } else if (nd6_prproxy) {
 673                 /*
 674                  * In the prefix proxying case, we need to inform nd6_output()
 675                  * about the inbound interface, so that any subsequent NS
 676                  * packets generated by nd6_prproxy_ns_output() will not be
 677                  * sent back to that same interface.
 678                  */
 679                 origifp = rcvifp = m->m_pkthdr.rcvif;
 680         } else {
 681                 rcvifp = m->m_pkthdr.rcvif;
 682                 origifp = rt->rt_ifp;
 683         }
 684         /*
 685          * clear embedded scope identifiers if necessary.
 686          * in6_clearscope will touch the addresses only when necessary.
 687          */
 688         in6_clearscope(&ip6->ip6_src);
 689         in6_clearscope(&ip6->ip6_dst);
 690
 691         ifp = rt->rt_ifp;
 692         /* Drop the lock but retain the extra ref */
 693         RT_UNLOCK(rt);
 694
 695         /*
 696          * If this is to be processed locally, let ip6_input have it.
 697          */
 698         if (proxy) {
 699                 VERIFY(m->m_pkthdr.pkt_flags & PKTF_PROXY_DST);
 700                 /* Release extra ref */
 701                 RT_REMREF(rt);
 702                 if (mcopy != NULL)
 703                         m_freem(mcopy);
 704                 return (m);
 705         }
 706
 707 #if PF
 708         /* Invoke outbound packet filter */
 709         error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, NULL);
 710
 711         if (error != 0 || m == NULL) {
 712                 if (m != NULL) {
 713                         panic("%s: unexpected packet %p\n", __func__, m);
 714                         /* NOTREACHED */
 715                 }
 716                 /* Already freed by callee */
 717                 goto senderr;
 718         }
 719         ip6 = mtod(m, struct ip6_hdr *);
 720 #endif /* PF */
 721
 722         /* Mark this packet as being forwarded from another interface */
 723         m->m_pkthdr.pkt_flags |= PKTF_FORWARDED;
 724         len = m_pktlen(m);
 725
 726         error = nd6_output(ifp, origifp, m, dst, rt, NULL);
 727         if (error) {
 728                 in6_ifstat_inc(ifp, ifs6_out_discard);
 729                 ip6stat.ip6s_cantforward++;
 730         } else {
 731                 /*
 732                  * Increment stats on the source interface; the ones
 733                  * for destination interface has been taken care of
 734                  * during output above by virtue of PKTF_FORWARDED.
 735                  */
 736                 rcvifp->if_fpackets++;
 737                 rcvifp->if_fbytes += len;
 738
 739                 ip6stat.ip6s_forward++;
 740                 in6_ifstat_inc(ifp, ifs6_out_forward);
 741                 if (type)
 742                         ip6stat.ip6s_redirectsent++;
 743                 else {
 744                         if (mcopy) {
 745                                 goto freecopy;
 746                         }
 747                 }
 748         }
 749 #if PF
 750 senderr:
 751 #endif /* PF */
 752         if (mcopy == NULL) {
 753                 /* Release extra ref */
 754                 RT_REMREF(rt);
 755                 return (NULL);
 756         }
 757         switch (error) {
 758         case 0:
 759 #if 1
 760                 if (type == ND_REDIRECT) {
 761                         icmp6_redirect_output(mcopy, rt);
 762                         /* Release extra ref */
 763                         RT_REMREF(rt);
 764                         return (NULL);
 765                 }
 766 #endif
 767                 goto freecopy;
 768
 769         case EMSGSIZE:
 770                 /* xxx MTU is constant in PPP? */
 771                 goto freecopy;
 772
 773         case ENOBUFS:
 774                 /* Tell source to slow down like source quench in IP? */
 775                 goto freecopy;
 776
 777         case ENETUNREACH:       /* shouldn't happen, checked above */
 778         case EHOSTUNREACH:
 779         case ENETDOWN:
 780         case EHOSTDOWN:
 781         default:
 782                 type = ICMP6_DST_UNREACH;
 783                 code = ICMP6_DST_UNREACH_ADDR;
 784                 break;
 785         }
 786         icmp6_error(mcopy, type, code, 0);
 787         /* Release extra ref */
 788         RT_REMREF(rt);
 789         return (NULL);
 790
 791  freecopy:
 792         m_freem(mcopy);
 793         /* Release extra ref */
 794         RT_REMREF(rt);
 795         return (NULL);
 796 }