1 /*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: ip6_forward.c,v 1.69 2001/05/17 03:48:30 itojun Exp $ 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_inet.h" 36 #include "opt_inet6.h" 37 #include "opt_ipsec.h" 38 #include "opt_ipstealth.h" 39 40 #include <sys/param.h> 41 #include <sys/systm.h> 42 #include <sys/malloc.h> 43 #include <sys/mbuf.h> 44 #include <sys/domain.h> 45 #include <sys/protosw.h> 46 #include <sys/socket.h> 47 #include <sys/errno.h> 48 #include <sys/time.h> 49 #include <sys/kernel.h> 50 #include <sys/syslog.h> 51 52 #include <net/if.h> 53 #include <net/if_var.h> 54 #include <net/netisr.h> 55 #include <net/route.h> 56 #include <net/pfil.h> 57 58 #include <netinet/in.h> 59 #include <netinet/in_var.h> 60 #include <netinet/in_systm.h> 61 #include <netinet/ip.h> 62 #include <netinet/ip_var.h> 63 #include <netinet6/in6_var.h> 64 #include <netinet/ip6.h> 65 #include <netinet6/ip6_var.h> 66 #include <netinet6/scope6_var.h> 67 #include <netinet/icmp6.h> 68 #include <netinet6/nd6.h> 69 70 #include <netinet/in_pcb.h> 71 72 #ifdef IPSEC 73 #include <netinet6/ip6_ipsec.h> 74 #include <netipsec/ipsec.h> 75 #include <netipsec/ipsec6.h> 76 #include <netipsec/key.h> 77 #endif /* IPSEC */ 78 79 /* 80 * Forward a packet. If some error occurs return the sender 81 * an icmp packet. Note we can't always generate a meaningful 82 * icmp message because icmp doesn't have a large enough repertoire 83 * of codes and types. 84 * 85 * If not forwarding, just drop the packet. This could be confusing 86 * if ipforwarding was zero but some routing protocol was advancing 87 * us as a gateway to somewhere. However, we must let the routing 88 * protocol deal with that. 89 * 90 */ 91 void 92 ip6_forward(struct mbuf *m, int srcrt) 93 { 94 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 95 struct sockaddr_in6 *dst = NULL; 96 struct rtentry *rt = NULL; 97 struct route_in6 rin6; 98 int error, type = 0, code = 0; 99 struct mbuf *mcopy = NULL; 100 struct ifnet *origifp; /* maybe unnecessary */ 101 u_int32_t inzone, outzone; 102 struct in6_addr src_in6, dst_in6, odst; 103 #ifdef IPSEC 104 struct secpolicy *sp = NULL; 105 #endif 106 struct m_tag *fwd_tag; 107 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 108 109 /* 110 * Do not forward packets to multicast destination (should be handled 111 * by ip6_mforward(). 112 * Do not forward packets with unspecified source. It was discussed 113 * in July 2000, on the ipngwg mailing list. 114 */ 115 if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 || 116 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 117 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 118 IP6STAT_INC(ip6s_cantforward); 119 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ 120 if (V_ip6_log_time + V_ip6_log_interval < time_uptime) { 121 V_ip6_log_time = time_uptime; 122 log(LOG_DEBUG, 123 "cannot forward " 124 "from %s to %s nxt %d received on %s\n", 125 ip6_sprintf(ip6bufs, &ip6->ip6_src), 126 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 127 ip6->ip6_nxt, 128 if_name(m->m_pkthdr.rcvif)); 129 } 130 m_freem(m); 131 return; 132 } 133 #ifdef IPSEC 134 /* 135 * Check if this packet has an active SA and needs to be dropped 136 * instead of forwarded. 137 */ 138 if (ip6_ipsec_fwd(m) != 0) { 139 IP6STAT_INC(ip6s_cantforward); 140 m_freem(m); 141 return; 142 } 143 #endif /* IPSEC */ 144 145 #ifdef IPSTEALTH 146 if (!V_ip6stealth) { 147 #endif 148 if (ip6->ip6_hlim <= IPV6_HLIMDEC) { 149 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ 150 icmp6_error(m, ICMP6_TIME_EXCEEDED, 151 ICMP6_TIME_EXCEED_TRANSIT, 0); 152 return; 153 } 154 ip6->ip6_hlim -= IPV6_HLIMDEC; 155 156 #ifdef IPSTEALTH 157 } 158 #endif 159 160 /* 161 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU - 162 * size of IPv6 + ICMPv6 headers) bytes of the packet in case 163 * we need to generate an ICMP6 message to the src. 164 * Thanks to M_EXT, in most cases copy will not occur. 165 * 166 * It is important to save it before IPsec processing as IPsec 167 * processing may modify the mbuf. 168 */ 169 mcopy = m_copym(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN), 170 M_NOWAIT); 171 172 #ifdef IPSEC 173 /* get a security policy for this packet */ 174 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, &error); 175 if (sp == NULL) { 176 IPSEC6STAT_INC(ips_out_inval); 177 IP6STAT_INC(ip6s_cantforward); 178 if (mcopy) { 179 #if 0 180 /* XXX: what icmp ? */ 181 #else 182 m_freem(mcopy); 183 #endif 184 } 185 m_freem(m); 186 return; 187 } 188 189 error = 0; 190 191 /* check policy */ 192 switch (sp->policy) { 193 case IPSEC_POLICY_DISCARD: 194 /* 195 * This packet is just discarded. 196 */ 197 IPSEC6STAT_INC(ips_out_polvio); 198 IP6STAT_INC(ip6s_cantforward); 199 KEY_FREESP(&sp); 200 if (mcopy) { 201 #if 0 202 /* XXX: what icmp ? */ 203 #else 204 m_freem(mcopy); 205 #endif 206 } 207 m_freem(m); 208 return; 209 210 case IPSEC_POLICY_BYPASS: 211 case IPSEC_POLICY_NONE: 212 /* no need to do IPsec. */ 213 KEY_FREESP(&sp); 214 goto skip_ipsec; 215 216 case IPSEC_POLICY_IPSEC: 217 if (sp->req == NULL) { 218 /* XXX should be panic ? */ 219 printf("ip6_forward: No IPsec request specified.\n"); 220 IP6STAT_INC(ip6s_cantforward); 221 KEY_FREESP(&sp); 222 if (mcopy) { 223 #if 0 224 /* XXX: what icmp ? */ 225 #else 226 m_freem(mcopy); 227 #endif 228 } 229 m_freem(m); 230 return; 231 } 232 /* do IPsec */ 233 break; 234 235 case IPSEC_POLICY_ENTRUST: 236 default: 237 /* should be panic ?? */ 238 printf("ip6_forward: Invalid policy found. %d\n", sp->policy); 239 KEY_FREESP(&sp); 240 goto skip_ipsec; 241 } 242 243 { 244 struct ipsecrequest *isr = NULL; 245 246 /* 247 * when the kernel forwards a packet, it is not proper to apply 248 * IPsec transport mode to the packet. This check avoid from this. 249 * at present, if there is even a transport mode SA request in the 250 * security policy, the kernel does not apply IPsec to the packet. 251 * this check is not enough because the following case is valid. 252 * ipsec esp/tunnel/xxx-xxx/require esp/transport//require; 253 */ 254 for (isr = sp->req; isr; isr = isr->next) { 255 if (isr->saidx.mode == IPSEC_MODE_ANY) 256 goto doipsectunnel; 257 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) 258 goto doipsectunnel; 259 } 260 261 /* 262 * if there's no need for tunnel mode IPsec, skip. 263 */ 264 if (!isr) 265 goto skip_ipsec; 266 267 doipsectunnel: 268 /* 269 * All the extension headers will become inaccessible 270 * (since they can be encrypted). 271 * Don't panic, we need no more updates to extension headers 272 * on inner IPv6 packet (since they are now encapsulated). 273 * 274 * IPv6 [ESP|AH] IPv6 [extension headers] payload 275 */ 276 277 /* 278 * If we need to encapsulate the packet, do it here 279 * ipsec6_proces_packet will send the packet using ip6_output 280 */ 281 error = ipsec6_process_packet(m, sp->req); 282 /* Release SP if an error occurred */ 283 if (error != 0) 284 KEY_FREESP(&sp); 285 if (error == EJUSTRETURN) { 286 /* 287 * We had a SP with a level of 'use' and no SA. We 288 * will just continue to process the packet without 289 * IPsec processing. 290 */ 291 error = 0; 292 goto skip_ipsec; 293 } 294 295 if (error) { 296 /* mbuf is already reclaimed in ipsec6_process_packet. */ 297 switch (error) { 298 case EHOSTUNREACH: 299 case ENETUNREACH: 300 case EMSGSIZE: 301 case ENOBUFS: 302 case ENOMEM: 303 break; 304 default: 305 printf("ip6_output (ipsec): error code %d\n", error); 306 /* FALLTHROUGH */ 307 case ENOENT: 308 /* don't show these error codes to the user */ 309 break; 310 } 311 IP6STAT_INC(ip6s_cantforward); 312 if (mcopy) { 313 #if 0 314 /* XXX: what icmp ? */ 315 #else 316 m_freem(mcopy); 317 #endif 318 } 319 return; 320 } else { 321 /* 322 * In the FAST IPSec case we have already 323 * re-injected the packet and it has been freed 324 * by the ipsec_done() function. So, just clean 325 * up after ourselves. 326 */ 327 m = NULL; 328 goto freecopy; 329 } 330 } 331 skip_ipsec: 332 #endif 333 again: 334 bzero(&rin6, sizeof(struct route_in6)); 335 dst = (struct sockaddr_in6 *)&rin6.ro_dst; 336 dst->sin6_len = sizeof(struct sockaddr_in6); 337 dst->sin6_family = AF_INET6; 338 dst->sin6_addr = ip6->ip6_dst; 339 again2: 340 rin6.ro_rt = in6_rtalloc1((struct sockaddr *)dst, 0, 0, M_GETFIB(m)); 341 rt = rin6.ro_rt; 342 if (rin6.ro_rt != NULL) 343 RT_UNLOCK(rin6.ro_rt); 344 else { 345 IP6STAT_INC(ip6s_noroute); 346 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute); 347 if (mcopy) { 348 icmp6_error(mcopy, ICMP6_DST_UNREACH, 349 ICMP6_DST_UNREACH_NOROUTE, 0); 350 } 351 goto bad; 352 } 353 354 /* 355 * Source scope check: if a packet can't be delivered to its 356 * destination for the reason that the destination is beyond the scope 357 * of the source address, discard the packet and return an icmp6 358 * destination unreachable error with Code 2 (beyond scope of source 359 * address). We use a local copy of ip6_src, since in6_setscope() 360 * will possibly modify its first argument. 361 * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1] 362 */ 363 src_in6 = ip6->ip6_src; 364 if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) { 365 /* XXX: this should not happen */ 366 IP6STAT_INC(ip6s_cantforward); 367 IP6STAT_INC(ip6s_badscope); 368 goto bad; 369 } 370 if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) { 371 IP6STAT_INC(ip6s_cantforward); 372 IP6STAT_INC(ip6s_badscope); 373 goto bad; 374 } 375 if (inzone != outzone) { 376 IP6STAT_INC(ip6s_cantforward); 377 IP6STAT_INC(ip6s_badscope); 378 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard); 379 380 if (V_ip6_log_time + V_ip6_log_interval < time_uptime) { 381 V_ip6_log_time = time_uptime; 382 log(LOG_DEBUG, 383 "cannot forward " 384 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 385 ip6_sprintf(ip6bufs, &ip6->ip6_src), 386 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 387 ip6->ip6_nxt, 388 if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp)); 389 } 390 if (mcopy) 391 icmp6_error(mcopy, ICMP6_DST_UNREACH, 392 ICMP6_DST_UNREACH_BEYONDSCOPE, 0); 393 goto bad; 394 } 395 396 /* 397 * Destination scope check: if a packet is going to break the scope 398 * zone of packet's destination address, discard it. This case should 399 * usually be prevented by appropriately-configured routing table, but 400 * we need an explicit check because we may mistakenly forward the 401 * packet to a different zone by (e.g.) a default route. 402 */ 403 dst_in6 = ip6->ip6_dst; 404 if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 || 405 in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 || 406 inzone != outzone) { 407 IP6STAT_INC(ip6s_cantforward); 408 IP6STAT_INC(ip6s_badscope); 409 goto bad; 410 } 411 412 if (rt->rt_flags & RTF_GATEWAY) 413 dst = (struct sockaddr_in6 *)rt->rt_gateway; 414 415 /* 416 * If we are to forward the packet using the same interface 417 * as one we got the packet from, perhaps we should send a redirect 418 * to sender to shortcut a hop. 419 * Only send redirect if source is sending directly to us, 420 * and if packet was not source routed (or has any options). 421 * Also, don't send redirect if forwarding using a route 422 * modified by a redirect. 423 */ 424 if (V_ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && 425 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) { 426 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) { 427 /* 428 * If the incoming interface is equal to the outgoing 429 * one, and the link attached to the interface is 430 * point-to-point, then it will be highly probable 431 * that a routing loop occurs. Thus, we immediately 432 * drop the packet and send an ICMPv6 error message. 433 * 434 * type/code is based on suggestion by Rich Draves. 435 * not sure if it is the best pick. 436 */ 437 icmp6_error(mcopy, ICMP6_DST_UNREACH, 438 ICMP6_DST_UNREACH_ADDR, 0); 439 goto bad; 440 } 441 type = ND_REDIRECT; 442 } 443 444 /* 445 * Fake scoped addresses. Note that even link-local source or 446 * destinaion can appear, if the originating node just sends the 447 * packet to us (without address resolution for the destination). 448 * Since both icmp6_error and icmp6_redirect_output fill the embedded 449 * link identifiers, we can do this stuff after making a copy for 450 * returning an error. 451 */ 452 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { 453 /* 454 * See corresponding comments in ip6_output. 455 * XXX: but is it possible that ip6_forward() sends a packet 456 * to a loopback interface? I don't think so, and thus 457 * I bark here. (jinmei@kame.net) 458 * XXX: it is common to route invalid packets to loopback. 459 * also, the codepath will be visited on use of ::1 in 460 * rthdr. (itojun) 461 */ 462 #if 1 463 if (0) 464 #else 465 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0) 466 #endif 467 { 468 printf("ip6_forward: outgoing interface is loopback. " 469 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 470 ip6_sprintf(ip6bufs, &ip6->ip6_src), 471 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 472 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif), 473 if_name(rt->rt_ifp)); 474 } 475 476 /* we can just use rcvif in forwarding. */ 477 origifp = m->m_pkthdr.rcvif; 478 } 479 else 480 origifp = rt->rt_ifp; 481 /* 482 * clear embedded scope identifiers if necessary. 483 * in6_clearscope will touch the addresses only when necessary. 484 */ 485 in6_clearscope(&ip6->ip6_src); 486 in6_clearscope(&ip6->ip6_dst); 487 488 /* Jump over all PFIL processing if hooks are not active. */ 489 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 490 goto pass; 491 492 odst = ip6->ip6_dst; 493 /* Run through list of hooks for output packets. */ 494 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT, NULL); 495 if (error != 0 || m == NULL) 496 goto freecopy; /* consumed by filter */ 497 ip6 = mtod(m, struct ip6_hdr *); 498 499 /* See if destination IP address was changed by packet filter. */ 500 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) { 501 m->m_flags |= M_SKIP_FIREWALL; 502 /* If destination is now ourself drop to ip6_input(). */ 503 if (in6_localip(&ip6->ip6_dst)) 504 m->m_flags |= M_FASTFWD_OURS; 505 else { 506 RTFREE(rt); 507 goto again; /* Redo the routing table lookup. */ 508 } 509 } 510 511 /* See if local, if yes, send it to netisr. */ 512 if (m->m_flags & M_FASTFWD_OURS) { 513 if (m->m_pkthdr.rcvif == NULL) 514 m->m_pkthdr.rcvif = V_loif; 515 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 516 m->m_pkthdr.csum_flags |= 517 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; 518 m->m_pkthdr.csum_data = 0xffff; 519 } 520 #ifdef SCTP 521 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) 522 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 523 #endif 524 error = netisr_queue(NETISR_IPV6, m); 525 goto out; 526 } 527 /* Or forward to some other address? */ 528 if ((m->m_flags & M_IP6_NEXTHOP) && 529 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { 530 dst = (struct sockaddr_in6 *)&rin6.ro_dst; 531 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in6)); 532 m->m_flags |= M_SKIP_FIREWALL; 533 m->m_flags &= ~M_IP6_NEXTHOP; 534 m_tag_delete(m, fwd_tag); 535 RTFREE(rt); 536 goto again2; 537 } 538 539 pass: 540 /* See if the size was changed by the packet filter. */ 541 if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) { 542 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig); 543 if (mcopy) { 544 u_long mtu; 545 #ifdef IPSEC 546 size_t ipsechdrsiz; 547 #endif /* IPSEC */ 548 549 mtu = IN6_LINKMTU(rt->rt_ifp); 550 #ifdef IPSEC 551 /* 552 * When we do IPsec tunnel ingress, we need to play 553 * with the link value (decrement IPsec header size 554 * from mtu value). The code is much simpler than v4 555 * case, as we have the outgoing interface for 556 * encapsulated packet as "rt->rt_ifp". 557 */ 558 ipsechdrsiz = ipsec_hdrsiz(mcopy, IPSEC_DIR_OUTBOUND, 559 NULL); 560 if (ipsechdrsiz < mtu) 561 mtu -= ipsechdrsiz; 562 /* 563 * if mtu becomes less than minimum MTU, 564 * tell minimum MTU (and I'll need to fragment it). 565 */ 566 if (mtu < IPV6_MMTU) 567 mtu = IPV6_MMTU; 568 #endif /* IPSEC */ 569 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu); 570 } 571 goto bad; 572 } 573 574 error = nd6_output_ifp(rt->rt_ifp, origifp, m, dst, NULL); 575 if (error) { 576 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard); 577 IP6STAT_INC(ip6s_cantforward); 578 } else { 579 IP6STAT_INC(ip6s_forward); 580 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward); 581 if (type) 582 IP6STAT_INC(ip6s_redirectsent); 583 else { 584 if (mcopy) 585 goto freecopy; 586 } 587 } 588 589 if (mcopy == NULL) 590 goto out; 591 switch (error) { 592 case 0: 593 if (type == ND_REDIRECT) { 594 icmp6_redirect_output(mcopy, rt); 595 goto out; 596 } 597 goto freecopy; 598 599 case EMSGSIZE: 600 /* xxx MTU is constant in PPP? */ 601 goto freecopy; 602 603 case ENOBUFS: 604 /* Tell source to slow down like source quench in IP? */ 605 goto freecopy; 606 607 case ENETUNREACH: /* shouldn't happen, checked above */ 608 case EHOSTUNREACH: 609 case ENETDOWN: 610 case EHOSTDOWN: 611 default: 612 type = ICMP6_DST_UNREACH; 613 code = ICMP6_DST_UNREACH_ADDR; 614 break; 615 } 616 icmp6_error(mcopy, type, code, 0); 617 goto out; 618 619 freecopy: 620 m_freem(mcopy); 621 goto out; 622 bad: 623 m_freem(m); 624 out: 625 if (rt != NULL) 626 RTFREE(rt); 627 } 628