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 #ifdef SCTP 107 int sw_csum; 108 #endif 109 struct m_tag *fwd_tag; 110 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 111 112 /* 113 * Do not forward packets to multicast destination (should be handled 114 * by ip6_mforward(). 115 * Do not forward packets with unspecified source. It was discussed 116 * in July 2000, on the ipngwg mailing list. 117 */ 118 if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 || 119 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 120 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 121 IP6STAT_INC(ip6s_cantforward); 122 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ 123 if (V_ip6_log_time + V_ip6_log_interval < time_uptime) { 124 V_ip6_log_time = time_uptime; 125 log(LOG_DEBUG, 126 "cannot forward " 127 "from %s to %s nxt %d received on %s\n", 128 ip6_sprintf(ip6bufs, &ip6->ip6_src), 129 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 130 ip6->ip6_nxt, 131 if_name(m->m_pkthdr.rcvif)); 132 } 133 m_freem(m); 134 return; 135 } 136 #ifdef IPSEC 137 /* 138 * Check if this packet has an active SA and needs to be dropped 139 * instead of forwarded. 140 */ 141 if (ip6_ipsec_fwd(m) != 0) { 142 IP6STAT_INC(ip6s_cantforward); 143 m_freem(m); 144 return; 145 } 146 #endif /* IPSEC */ 147 148 #ifdef IPSTEALTH 149 if (!V_ip6stealth) { 150 #endif 151 if (ip6->ip6_hlim <= IPV6_HLIMDEC) { 152 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ 153 icmp6_error(m, ICMP6_TIME_EXCEEDED, 154 ICMP6_TIME_EXCEED_TRANSIT, 0); 155 return; 156 } 157 ip6->ip6_hlim -= IPV6_HLIMDEC; 158 159 #ifdef IPSTEALTH 160 } 161 #endif 162 163 /* 164 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU - 165 * size of IPv6 + ICMPv6 headers) bytes of the packet in case 166 * we need to generate an ICMP6 message to the src. 167 * Thanks to M_EXT, in most cases copy will not occur. 168 * 169 * It is important to save it before IPsec processing as IPsec 170 * processing may modify the mbuf. 171 */ 172 mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN)); 173 174 #ifdef IPSEC 175 /* get a security policy for this packet */ 176 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, &error); 177 if (sp == NULL) { 178 IPSEC6STAT_INC(ips_out_inval); 179 IP6STAT_INC(ip6s_cantforward); 180 if (mcopy) { 181 #if 0 182 /* XXX: what icmp ? */ 183 #else 184 m_freem(mcopy); 185 #endif 186 } 187 m_freem(m); 188 return; 189 } 190 191 error = 0; 192 193 /* check policy */ 194 switch (sp->policy) { 195 case IPSEC_POLICY_DISCARD: 196 /* 197 * This packet is just discarded. 198 */ 199 IPSEC6STAT_INC(ips_out_polvio); 200 IP6STAT_INC(ip6s_cantforward); 201 KEY_FREESP(&sp); 202 if (mcopy) { 203 #if 0 204 /* XXX: what icmp ? */ 205 #else 206 m_freem(mcopy); 207 #endif 208 } 209 m_freem(m); 210 return; 211 212 case IPSEC_POLICY_BYPASS: 213 case IPSEC_POLICY_NONE: 214 /* no need to do IPsec. */ 215 KEY_FREESP(&sp); 216 goto skip_ipsec; 217 218 case IPSEC_POLICY_IPSEC: 219 if (sp->req == NULL) { 220 /* XXX should be panic ? */ 221 printf("ip6_forward: No IPsec request specified.\n"); 222 IP6STAT_INC(ip6s_cantforward); 223 KEY_FREESP(&sp); 224 if (mcopy) { 225 #if 0 226 /* XXX: what icmp ? */ 227 #else 228 m_freem(mcopy); 229 #endif 230 } 231 m_freem(m); 232 return; 233 } 234 /* do IPsec */ 235 break; 236 237 case IPSEC_POLICY_ENTRUST: 238 default: 239 /* should be panic ?? */ 240 printf("ip6_forward: Invalid policy found. %d\n", sp->policy); 241 KEY_FREESP(&sp); 242 goto skip_ipsec; 243 } 244 245 { 246 struct ipsecrequest *isr = NULL; 247 248 /* 249 * when the kernel forwards a packet, it is not proper to apply 250 * IPsec transport mode to the packet. This check avoid from this. 251 * at present, if there is even a transport mode SA request in the 252 * security policy, the kernel does not apply IPsec to the packet. 253 * this check is not enough because the following case is valid. 254 * ipsec esp/tunnel/xxx-xxx/require esp/transport//require; 255 */ 256 for (isr = sp->req; isr; isr = isr->next) { 257 if (isr->saidx.mode == IPSEC_MODE_ANY) 258 goto doipsectunnel; 259 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) 260 goto doipsectunnel; 261 } 262 263 /* 264 * if there's no need for tunnel mode IPsec, skip. 265 */ 266 if (!isr) 267 goto skip_ipsec; 268 269 doipsectunnel: 270 /* 271 * All the extension headers will become inaccessible 272 * (since they can be encrypted). 273 * Don't panic, we need no more updates to extension headers 274 * on inner IPv6 packet (since they are now encapsulated). 275 * 276 * IPv6 [ESP|AH] IPv6 [extension headers] payload 277 */ 278 279 /* 280 * If we need to encapsulate the packet, do it here 281 * ipsec6_proces_packet will send the packet using ip6_output 282 */ 283 error = ipsec6_process_packet(m, sp->req); 284 /* Release SP if an error occurred */ 285 if (error != 0) 286 KEY_FREESP(&sp); 287 if (error == EJUSTRETURN) { 288 /* 289 * We had a SP with a level of 'use' and no SA. We 290 * will just continue to process the packet without 291 * IPsec processing. 292 */ 293 error = 0; 294 goto skip_ipsec; 295 } 296 297 if (error) { 298 /* mbuf is already reclaimed in ipsec6_process_packet. */ 299 switch (error) { 300 case EHOSTUNREACH: 301 case ENETUNREACH: 302 case EMSGSIZE: 303 case ENOBUFS: 304 case ENOMEM: 305 break; 306 default: 307 printf("ip6_output (ipsec): error code %d\n", error); 308 /* FALLTHROUGH */ 309 case ENOENT: 310 /* don't show these error codes to the user */ 311 break; 312 } 313 IP6STAT_INC(ip6s_cantforward); 314 if (mcopy) { 315 #if 0 316 /* XXX: what icmp ? */ 317 #else 318 m_freem(mcopy); 319 #endif 320 } 321 return; 322 } else { 323 /* 324 * In the FAST IPSec case we have already 325 * re-injected the packet and it has been freed 326 * by the ipsec_done() function. So, just clean 327 * up after ourselves. 328 */ 329 m = NULL; 330 goto freecopy; 331 } 332 } 333 skip_ipsec: 334 #endif 335 again: 336 bzero(&rin6, sizeof(struct route_in6)); 337 dst = (struct sockaddr_in6 *)&rin6.ro_dst; 338 dst->sin6_len = sizeof(struct sockaddr_in6); 339 dst->sin6_family = AF_INET6; 340 dst->sin6_addr = ip6->ip6_dst; 341 again2: 342 rin6.ro_rt = in6_rtalloc1((struct sockaddr *)dst, 0, 0, M_GETFIB(m)); 343 rt = rin6.ro_rt; 344 if (rin6.ro_rt != NULL) 345 RT_UNLOCK(rin6.ro_rt); 346 else { 347 IP6STAT_INC(ip6s_noroute); 348 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute); 349 if (mcopy) { 350 icmp6_error(mcopy, ICMP6_DST_UNREACH, 351 ICMP6_DST_UNREACH_NOROUTE, 0); 352 } 353 goto bad; 354 } 355 356 /* 357 * Source scope check: if a packet can't be delivered to its 358 * destination for the reason that the destination is beyond the scope 359 * of the source address, discard the packet and return an icmp6 360 * destination unreachable error with Code 2 (beyond scope of source 361 * address). We use a local copy of ip6_src, since in6_setscope() 362 * will possibly modify its first argument. 363 * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1] 364 */ 365 src_in6 = ip6->ip6_src; 366 if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) { 367 /* XXX: this should not happen */ 368 IP6STAT_INC(ip6s_cantforward); 369 IP6STAT_INC(ip6s_badscope); 370 goto bad; 371 } 372 if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) { 373 IP6STAT_INC(ip6s_cantforward); 374 IP6STAT_INC(ip6s_badscope); 375 goto bad; 376 } 377 if (inzone != outzone) { 378 IP6STAT_INC(ip6s_cantforward); 379 IP6STAT_INC(ip6s_badscope); 380 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard); 381 382 if (V_ip6_log_time + V_ip6_log_interval < time_uptime) { 383 V_ip6_log_time = time_uptime; 384 log(LOG_DEBUG, 385 "cannot forward " 386 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 387 ip6_sprintf(ip6bufs, &ip6->ip6_src), 388 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 389 ip6->ip6_nxt, 390 if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp)); 391 } 392 if (mcopy) 393 icmp6_error(mcopy, ICMP6_DST_UNREACH, 394 ICMP6_DST_UNREACH_BEYONDSCOPE, 0); 395 goto bad; 396 } 397 398 /* 399 * Destination scope check: if a packet is going to break the scope 400 * zone of packet's destination address, discard it. This case should 401 * usually be prevented by appropriately-configured routing table, but 402 * we need an explicit check because we may mistakenly forward the 403 * packet to a different zone by (e.g.) a default route. 404 */ 405 dst_in6 = ip6->ip6_dst; 406 if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 || 407 in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 || 408 inzone != outzone) { 409 IP6STAT_INC(ip6s_cantforward); 410 IP6STAT_INC(ip6s_badscope); 411 goto bad; 412 } 413 414 if (rt->rt_flags & RTF_GATEWAY) 415 dst = (struct sockaddr_in6 *)rt->rt_gateway; 416 417 /* 418 * If we are to forward the packet using the same interface 419 * as one we got the packet from, perhaps we should send a redirect 420 * to sender to shortcut a hop. 421 * Only send redirect if source is sending directly to us, 422 * and if packet was not source routed (or has any options). 423 * Also, don't send redirect if forwarding using a route 424 * modified by a redirect. 425 */ 426 if (V_ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && 427 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) { 428 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) { 429 /* 430 * If the incoming interface is equal to the outgoing 431 * one, and the link attached to the interface is 432 * point-to-point, then it will be highly probable 433 * that a routing loop occurs. Thus, we immediately 434 * drop the packet and send an ICMPv6 error message. 435 * 436 * type/code is based on suggestion by Rich Draves. 437 * not sure if it is the best pick. 438 */ 439 icmp6_error(mcopy, ICMP6_DST_UNREACH, 440 ICMP6_DST_UNREACH_ADDR, 0); 441 goto bad; 442 } 443 type = ND_REDIRECT; 444 } 445 446 /* 447 * Fake scoped addresses. Note that even link-local source or 448 * destinaion can appear, if the originating node just sends the 449 * packet to us (without address resolution for the destination). 450 * Since both icmp6_error and icmp6_redirect_output fill the embedded 451 * link identifiers, we can do this stuff after making a copy for 452 * returning an error. 453 */ 454 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { 455 /* 456 * See corresponding comments in ip6_output. 457 * XXX: but is it possible that ip6_forward() sends a packet 458 * to a loopback interface? I don't think so, and thus 459 * I bark here. (jinmei@kame.net) 460 * XXX: it is common to route invalid packets to loopback. 461 * also, the codepath will be visited on use of ::1 in 462 * rthdr. (itojun) 463 */ 464 #if 1 465 if (0) 466 #else 467 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0) 468 #endif 469 { 470 printf("ip6_forward: outgoing interface is loopback. " 471 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 472 ip6_sprintf(ip6bufs, &ip6->ip6_src), 473 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 474 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif), 475 if_name(rt->rt_ifp)); 476 } 477 478 /* we can just use rcvif in forwarding. */ 479 origifp = m->m_pkthdr.rcvif; 480 } 481 else 482 origifp = rt->rt_ifp; 483 /* 484 * clear embedded scope identifiers if necessary. 485 * in6_clearscope will touch the addresses only when necessary. 486 */ 487 in6_clearscope(&ip6->ip6_src); 488 in6_clearscope(&ip6->ip6_dst); 489 490 /* Jump over all PFIL processing if hooks are not active. */ 491 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 492 goto pass; 493 494 odst = ip6->ip6_dst; 495 /* Run through list of hooks for output packets. */ 496 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT, NULL); 497 if (error != 0 || m == NULL) 498 goto freecopy; /* consumed by filter */ 499 ip6 = mtod(m, struct ip6_hdr *); 500 501 /* See if destination IP address was changed by packet filter. */ 502 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) { 503 m->m_flags |= M_SKIP_FIREWALL; 504 /* If destination is now ourself drop to ip6_input(). */ 505 if (in6_localip(&ip6->ip6_dst)) 506 m->m_flags |= M_FASTFWD_OURS; 507 else { 508 RTFREE(rt); 509 goto again; /* Redo the routing table lookup. */ 510 } 511 } 512 513 /* See if local, if yes, send it to netisr. */ 514 if (m->m_flags & M_FASTFWD_OURS) { 515 if (m->m_pkthdr.rcvif == NULL) 516 m->m_pkthdr.rcvif = V_loif; 517 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 518 m->m_pkthdr.csum_flags |= 519 CSUM_DATA_VALID_IPV6 | CSUM_PSEUDO_HDR; 520 m->m_pkthdr.csum_data = 0xffff; 521 } 522 #ifdef SCTP 523 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) 524 m->m_pkthdr.csum_flags |= CSUM_SCTP_VALID; 525 #endif 526 error = netisr_queue(NETISR_IPV6, m); 527 goto out; 528 } 529 /* Or forward to some other address? */ 530 if ((m->m_flags & M_IP6_NEXTHOP) && 531 (fwd_tag = m_tag_find(m, PACKET_TAG_IPFORWARD, NULL)) != NULL) { 532 dst = (struct sockaddr_in6 *)&rin6.ro_dst; 533 bcopy((fwd_tag+1), dst, sizeof(struct sockaddr_in6)); 534 m->m_flags |= M_SKIP_FIREWALL; 535 m->m_flags &= ~M_IP6_NEXTHOP; 536 m_tag_delete(m, fwd_tag); 537 RTFREE(rt); 538 goto again2; 539 } 540 541 pass: 542 /* See if the size was changed by the packet filter. */ 543 if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) { 544 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig); 545 if (mcopy) { 546 u_long mtu; 547 #ifdef IPSEC 548 size_t ipsechdrsiz; 549 #endif /* IPSEC */ 550 551 mtu = IN6_LINKMTU(rt->rt_ifp); 552 #ifdef IPSEC 553 /* 554 * When we do IPsec tunnel ingress, we need to play 555 * with the link value (decrement IPsec header size 556 * from mtu value). The code is much simpler than v4 557 * case, as we have the outgoing interface for 558 * encapsulated packet as "rt->rt_ifp". 559 */ 560 ipsechdrsiz = ipsec_hdrsiz(mcopy, IPSEC_DIR_OUTBOUND, 561 NULL); 562 if (ipsechdrsiz < mtu) 563 mtu -= ipsechdrsiz; 564 /* 565 * if mtu becomes less than minimum MTU, 566 * tell minimum MTU (and I'll need to fragment it). 567 */ 568 if (mtu < IPV6_MMTU) 569 mtu = IPV6_MMTU; 570 #endif /* IPSEC */ 571 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu); 572 } 573 goto bad; 574 } 575 576 error = nd6_output_ifp(rt->rt_ifp, origifp, m, dst, NULL); 577 if (error) { 578 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard); 579 IP6STAT_INC(ip6s_cantforward); 580 } else { 581 IP6STAT_INC(ip6s_forward); 582 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward); 583 if (type) 584 IP6STAT_INC(ip6s_redirectsent); 585 else { 586 if (mcopy) 587 goto freecopy; 588 } 589 } 590 591 if (mcopy == NULL) 592 goto out; 593 switch (error) { 594 case 0: 595 if (type == ND_REDIRECT) { 596 icmp6_redirect_output(mcopy, rt); 597 goto out; 598 } 599 goto freecopy; 600 601 case EMSGSIZE: 602 /* xxx MTU is constant in PPP? */ 603 goto freecopy; 604 605 case ENOBUFS: 606 /* Tell source to slow down like source quench in IP? */ 607 goto freecopy; 608 609 case ENETUNREACH: /* shouldn't happen, checked above */ 610 case EHOSTUNREACH: 611 case ENETDOWN: 612 case EHOSTDOWN: 613 default: 614 type = ICMP6_DST_UNREACH; 615 code = ICMP6_DST_UNREACH_ADDR; 616 break; 617 } 618 icmp6_error(mcopy, type, code, 0); 619 goto out; 620 621 freecopy: 622 m_freem(mcopy); 623 goto out; 624 bad: 625 m_freem(m); 626 out: 627 if (rt != NULL) 628 RTFREE(rt); 629 } 630