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_ipfw.h" 38 #include "opt_ipsec.h" 39 #include "opt_ipstealth.h" 40 41 #include <sys/param.h> 42 #include <sys/systm.h> 43 #include <sys/malloc.h> 44 #include <sys/mbuf.h> 45 #include <sys/domain.h> 46 #include <sys/protosw.h> 47 #include <sys/socket.h> 48 #include <sys/errno.h> 49 #include <sys/time.h> 50 #include <sys/kernel.h> 51 #include <sys/syslog.h> 52 53 #include <net/if.h> 54 #include <net/if_var.h> 55 #include <net/netisr.h> 56 #include <net/route.h> 57 #include <net/pfil.h> 58 59 #include <netinet/in.h> 60 #include <netinet/in_var.h> 61 #include <netinet/in_systm.h> 62 #include <netinet/ip.h> 63 #include <netinet/ip_var.h> 64 #include <netinet6/in6_var.h> 65 #include <netinet/ip6.h> 66 #include <netinet6/ip6_var.h> 67 #include <netinet6/scope6_var.h> 68 #include <netinet/icmp6.h> 69 #include <netinet6/nd6.h> 70 71 #include <netinet/in_pcb.h> 72 73 #ifdef IPSEC 74 #include <netinet6/ip6_ipsec.h> 75 #include <netipsec/ipsec.h> 76 #include <netipsec/ipsec6.h> 77 #include <netipsec/key.h> 78 #endif /* IPSEC */ 79 80 /* 81 * Forward a packet. If some error occurs return the sender 82 * an icmp packet. Note we can't always generate a meaningful 83 * icmp message because icmp doesn't have a large enough repertoire 84 * of codes and types. 85 * 86 * If not forwarding, just drop the packet. This could be confusing 87 * if ipforwarding was zero but some routing protocol was advancing 88 * us as a gateway to somewhere. However, we must let the routing 89 * protocol deal with that. 90 * 91 */ 92 void 93 ip6_forward(struct mbuf *m, int srcrt) 94 { 95 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 96 struct sockaddr_in6 *dst = NULL; 97 struct rtentry *rt = NULL; 98 struct route_in6 rin6; 99 int error, type = 0, code = 0; 100 struct mbuf *mcopy = NULL; 101 struct ifnet *origifp; /* maybe unnecessary */ 102 u_int32_t inzone, outzone; 103 struct in6_addr src_in6, dst_in6, odst; 104 #ifdef IPSEC 105 struct secpolicy *sp = NULL; 106 #endif 107 #ifdef SCTP 108 int sw_csum; 109 #endif 110 struct m_tag *fwd_tag; 111 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 112 113 /* 114 * Do not forward packets to multicast destination (should be handled 115 * by ip6_mforward(). 116 * Do not forward packets with unspecified source. It was discussed 117 * in July 2000, on the ipngwg mailing list. 118 */ 119 if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 || 120 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 121 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 122 IP6STAT_INC(ip6s_cantforward); 123 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ 124 if (V_ip6_log_time + V_ip6_log_interval < time_uptime) { 125 V_ip6_log_time = time_uptime; 126 log(LOG_DEBUG, 127 "cannot forward " 128 "from %s to %s nxt %d received on %s\n", 129 ip6_sprintf(ip6bufs, &ip6->ip6_src), 130 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 131 ip6->ip6_nxt, 132 if_name(m->m_pkthdr.rcvif)); 133 } 134 m_freem(m); 135 return; 136 } 137 #ifdef IPSEC 138 /* 139 * Check if this packet has an active SA and needs to be dropped 140 * instead of forwarded. 141 */ 142 if (ip6_ipsec_fwd(m) != 0) { 143 IP6STAT_INC(ip6s_cantforward); 144 m_freem(m); 145 return; 146 } 147 #endif /* IPSEC */ 148 149 #ifdef IPSTEALTH 150 if (!V_ip6stealth) { 151 #endif 152 if (ip6->ip6_hlim <= IPV6_HLIMDEC) { 153 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ 154 icmp6_error(m, ICMP6_TIME_EXCEEDED, 155 ICMP6_TIME_EXCEED_TRANSIT, 0); 156 return; 157 } 158 ip6->ip6_hlim -= IPV6_HLIMDEC; 159 160 #ifdef IPSTEALTH 161 } 162 #endif 163 164 /* 165 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU - 166 * size of IPv6 + ICMPv6 headers) bytes of the packet in case 167 * we need to generate an ICMP6 message to the src. 168 * Thanks to M_EXT, in most cases copy will not occur. 169 * 170 * It is important to save it before IPsec processing as IPsec 171 * processing may modify the mbuf. 172 */ 173 mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN)); 174 175 #ifdef IPSEC 176 /* get a security policy for this packet */ 177 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, &error); 178 if (sp == NULL) { 179 IPSEC6STAT_INC(ips_out_inval); 180 IP6STAT_INC(ip6s_cantforward); 181 if (mcopy) { 182 #if 0 183 /* XXX: what icmp ? */ 184 #else 185 m_freem(mcopy); 186 #endif 187 } 188 m_freem(m); 189 return; 190 } 191 192 error = 0; 193 194 /* check policy */ 195 switch (sp->policy) { 196 case IPSEC_POLICY_DISCARD: 197 /* 198 * This packet is just discarded. 199 */ 200 IPSEC6STAT_INC(ips_out_polvio); 201 IP6STAT_INC(ip6s_cantforward); 202 KEY_FREESP(&sp); 203 if (mcopy) { 204 #if 0 205 /* XXX: what icmp ? */ 206 #else 207 m_freem(mcopy); 208 #endif 209 } 210 m_freem(m); 211 return; 212 213 case IPSEC_POLICY_BYPASS: 214 case IPSEC_POLICY_NONE: 215 /* no need to do IPsec. */ 216 KEY_FREESP(&sp); 217 goto skip_ipsec; 218 219 case IPSEC_POLICY_IPSEC: 220 if (sp->req == NULL) { 221 /* XXX should be panic ? */ 222 printf("ip6_forward: No IPsec request specified.\n"); 223 IP6STAT_INC(ip6s_cantforward); 224 KEY_FREESP(&sp); 225 if (mcopy) { 226 #if 0 227 /* XXX: what icmp ? */ 228 #else 229 m_freem(mcopy); 230 #endif 231 } 232 m_freem(m); 233 return; 234 } 235 /* do IPsec */ 236 break; 237 238 case IPSEC_POLICY_ENTRUST: 239 default: 240 /* should be panic ?? */ 241 printf("ip6_forward: Invalid policy found. %d\n", sp->policy); 242 KEY_FREESP(&sp); 243 goto skip_ipsec; 244 } 245 246 { 247 struct ipsecrequest *isr = NULL; 248 249 /* 250 * when the kernel forwards a packet, it is not proper to apply 251 * IPsec transport mode to the packet is not proper. this check 252 * avoid from this. 253 * at present, if there is even a transport mode SA request in the 254 * security policy, the kernel does not apply IPsec to the packet. 255 * this check is not enough because the following case is valid. 256 * ipsec esp/tunnel/xxx-xxx/require esp/transport//require; 257 */ 258 for (isr = sp->req; isr; isr = isr->next) { 259 if (isr->saidx.mode == IPSEC_MODE_ANY) 260 goto doipsectunnel; 261 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) 262 goto doipsectunnel; 263 } 264 265 /* 266 * if there's no need for tunnel mode IPsec, skip. 267 */ 268 if (!isr) 269 goto skip_ipsec; 270 271 doipsectunnel: 272 /* 273 * All the extension headers will become inaccessible 274 * (since they can be encrypted). 275 * Don't panic, we need no more updates to extension headers 276 * on inner IPv6 packet (since they are now encapsulated). 277 * 278 * IPv6 [ESP|AH] IPv6 [extension headers] payload 279 */ 280 281 /* 282 * If we need to encapsulate the packet, do it here 283 * ipsec6_proces_packet will send the packet using ip6_output 284 */ 285 error = ipsec6_process_packet(m, sp->req); 286 287 KEY_FREESP(&sp); 288 289 if (error == EJUSTRETURN) { 290 /* 291 * We had a SP with a level of 'use' and no SA. We 292 * will just continue to process the packet without 293 * IPsec processing. 294 */ 295 error = 0; 296 goto skip_ipsec; 297 } 298 299 if (error) { 300 /* mbuf is already reclaimed in ipsec6_process_packet. */ 301 switch (error) { 302 case EHOSTUNREACH: 303 case ENETUNREACH: 304 case EMSGSIZE: 305 case ENOBUFS: 306 case ENOMEM: 307 break; 308 default: 309 printf("ip6_output (ipsec): error code %d\n", error); 310 /* FALLTHROUGH */ 311 case ENOENT: 312 /* don't show these error codes to the user */ 313 break; 314 } 315 IP6STAT_INC(ip6s_cantforward); 316 if (mcopy) { 317 #if 0 318 /* XXX: what icmp ? */ 319 #else 320 m_freem(mcopy); 321 #endif 322 } 323 return; 324 } else { 325 /* 326 * In the FAST IPSec case we have already 327 * re-injected the packet and it has been freed 328 * by the ipsec_done() function. So, just clean 329 * up after ourselves. 330 */ 331 m = NULL; 332 goto freecopy; 333 } 334 } 335 skip_ipsec: 336 #endif 337 again: 338 bzero(&rin6, sizeof(struct route_in6)); 339 dst = (struct sockaddr_in6 *)&rin6.ro_dst; 340 dst->sin6_len = sizeof(struct sockaddr_in6); 341 dst->sin6_family = AF_INET6; 342 dst->sin6_addr = ip6->ip6_dst; 343 again2: 344 rin6.ro_rt = in6_rtalloc1((struct sockaddr *)dst, 0, 0, M_GETFIB(m)); 345 if (rin6.ro_rt != NULL) 346 RT_UNLOCK(rin6.ro_rt); 347 else { 348 IP6STAT_INC(ip6s_noroute); 349 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute); 350 if (mcopy) { 351 icmp6_error(mcopy, ICMP6_DST_UNREACH, 352 ICMP6_DST_UNREACH_NOROUTE, 0); 353 } 354 goto bad; 355 } 356 rt = rin6.ro_rt; 357 358 /* 359 * Source scope check: if a packet can't be delivered to its 360 * destination for the reason that the destination is beyond the scope 361 * of the source address, discard the packet and return an icmp6 362 * destination unreachable error with Code 2 (beyond scope of source 363 * address). We use a local copy of ip6_src, since in6_setscope() 364 * will possibly modify its first argument. 365 * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1] 366 */ 367 src_in6 = ip6->ip6_src; 368 if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) { 369 /* XXX: this should not happen */ 370 IP6STAT_INC(ip6s_cantforward); 371 IP6STAT_INC(ip6s_badscope); 372 goto bad; 373 } 374 if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) { 375 IP6STAT_INC(ip6s_cantforward); 376 IP6STAT_INC(ip6s_badscope); 377 goto bad; 378 } 379 if (inzone != outzone) { 380 IP6STAT_INC(ip6s_cantforward); 381 IP6STAT_INC(ip6s_badscope); 382 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard); 383 384 if (V_ip6_log_time + V_ip6_log_interval < time_uptime) { 385 V_ip6_log_time = time_uptime; 386 log(LOG_DEBUG, 387 "cannot forward " 388 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 389 ip6_sprintf(ip6bufs, &ip6->ip6_src), 390 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 391 ip6->ip6_nxt, 392 if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp)); 393 } 394 if (mcopy) 395 icmp6_error(mcopy, ICMP6_DST_UNREACH, 396 ICMP6_DST_UNREACH_BEYONDSCOPE, 0); 397 goto bad; 398 } 399 400 /* 401 * Destination scope check: if a packet is going to break the scope 402 * zone of packet's destination address, discard it. This case should 403 * usually be prevented by appropriately-configured routing table, but 404 * we need an explicit check because we may mistakenly forward the 405 * packet to a different zone by (e.g.) a default route. 406 */ 407 dst_in6 = ip6->ip6_dst; 408 if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 || 409 in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 || 410 inzone != outzone) { 411 IP6STAT_INC(ip6s_cantforward); 412 IP6STAT_INC(ip6s_badscope); 413 goto bad; 414 } 415 416 if (rt->rt_flags & RTF_GATEWAY) 417 dst = (struct sockaddr_in6 *)rt->rt_gateway; 418 419 /* 420 * If we are to forward the packet using the same interface 421 * as one we got the packet from, perhaps we should send a redirect 422 * to sender to shortcut a hop. 423 * Only send redirect if source is sending directly to us, 424 * and if packet was not source routed (or has any options). 425 * Also, don't send redirect if forwarding using a route 426 * modified by a redirect. 427 */ 428 if (V_ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && 429 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) { 430 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) { 431 /* 432 * If the incoming interface is equal to the outgoing 433 * one, and the link attached to the interface is 434 * point-to-point, then it will be highly probable 435 * that a routing loop occurs. Thus, we immediately 436 * drop the packet and send an ICMPv6 error message. 437 * 438 * type/code is based on suggestion by Rich Draves. 439 * not sure if it is the best pick. 440 */ 441 icmp6_error(mcopy, ICMP6_DST_UNREACH, 442 ICMP6_DST_UNREACH_ADDR, 0); 443 goto bad; 444 } 445 type = ND_REDIRECT; 446 } 447 448 /* 449 * Fake scoped addresses. Note that even link-local source or 450 * destinaion can appear, if the originating node just sends the 451 * packet to us (without address resolution for the destination). 452 * Since both icmp6_error and icmp6_redirect_output fill the embedded 453 * link identifiers, we can do this stuff after making a copy for 454 * returning an error. 455 */ 456 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { 457 /* 458 * See corresponding comments in ip6_output. 459 * XXX: but is it possible that ip6_forward() sends a packet 460 * to a loopback interface? I don't think so, and thus 461 * I bark here. (jinmei@kame.net) 462 * XXX: it is common to route invalid packets to loopback. 463 * also, the codepath will be visited on use of ::1 in 464 * rthdr. (itojun) 465 */ 466 #if 1 467 if (0) 468 #else 469 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0) 470 #endif 471 { 472 printf("ip6_forward: outgoing interface is loopback. " 473 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 474 ip6_sprintf(ip6bufs, &ip6->ip6_src), 475 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 476 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif), 477 if_name(rt->rt_ifp)); 478 } 479 480 /* we can just use rcvif in forwarding. */ 481 origifp = m->m_pkthdr.rcvif; 482 } 483 else 484 origifp = rt->rt_ifp; 485 /* 486 * clear embedded scope identifiers if necessary. 487 * in6_clearscope will touch the addresses only when necessary. 488 */ 489 in6_clearscope(&ip6->ip6_src); 490 in6_clearscope(&ip6->ip6_dst); 491 492 /* Jump over all PFIL processing if hooks are not active. */ 493 if (!PFIL_HOOKED(&V_inet6_pfil_hook)) 494 goto pass; 495 496 odst = ip6->ip6_dst; 497 /* Run through list of hooks for output packets. */ 498 error = pfil_run_hooks(&V_inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT, NULL); 499 if (error != 0 || m == NULL) 500 goto freecopy; /* consumed by filter */ 501 ip6 = mtod(m, struct ip6_hdr *); 502 503 /* See if destination IP address was changed by packet filter. */ 504 if (!IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst)) { 505 m->m_flags |= M_SKIP_FIREWALL; 506 /* If destination is now ourself drop to ip6_input(). */ 507 if (in6_localip(&ip6->ip6_dst)) 508 m->m_flags |= M_FASTFWD_OURS; 509 else 510 goto again; /* Redo the routing table lookup. */ 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 goto again2; 538 } 539 540 pass: 541 /* See if the size was changed by the packet filter. */ 542 if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) { 543 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig); 544 if (mcopy) { 545 u_long mtu; 546 #ifdef IPSEC 547 size_t ipsechdrsiz; 548 #endif /* IPSEC */ 549 550 mtu = IN6_LINKMTU(rt->rt_ifp); 551 #ifdef IPSEC 552 /* 553 * When we do IPsec tunnel ingress, we need to play 554 * with the link value (decrement IPsec header size 555 * from mtu value). The code is much simpler than v4 556 * case, as we have the outgoing interface for 557 * encapsulated packet as "rt->rt_ifp". 558 */ 559 ipsechdrsiz = ipsec_hdrsiz(mcopy, IPSEC_DIR_OUTBOUND, 560 NULL); 561 if (ipsechdrsiz < mtu) 562 mtu -= ipsechdrsiz; 563 /* 564 * if mtu becomes less than minimum MTU, 565 * tell minimum MTU (and I'll need to fragment it). 566 */ 567 if (mtu < IPV6_MMTU) 568 mtu = IPV6_MMTU; 569 #endif /* IPSEC */ 570 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu); 571 } 572 goto bad; 573 } 574 575 error = nd6_output(rt->rt_ifp, origifp, m, dst, rt); 576 if (error) { 577 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard); 578 IP6STAT_INC(ip6s_cantforward); 579 } else { 580 IP6STAT_INC(ip6s_forward); 581 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward); 582 if (type) 583 IP6STAT_INC(ip6s_redirectsent); 584 else { 585 if (mcopy) 586 goto freecopy; 587 } 588 } 589 590 if (mcopy == NULL) 591 goto out; 592 switch (error) { 593 case 0: 594 if (type == ND_REDIRECT) { 595 icmp6_redirect_output(mcopy, rt); 596 goto out; 597 } 598 goto freecopy; 599 600 case EMSGSIZE: 601 /* xxx MTU is constant in PPP? */ 602 goto freecopy; 603 604 case ENOBUFS: 605 /* Tell source to slow down like source quench in IP? */ 606 goto freecopy; 607 608 case ENETUNREACH: /* shouldn't happen, checked above */ 609 case EHOSTUNREACH: 610 case ENETDOWN: 611 case EHOSTDOWN: 612 default: 613 type = ICMP6_DST_UNREACH; 614 code = ICMP6_DST_UNREACH_ADDR; 615 break; 616 } 617 icmp6_error(mcopy, type, code, 0); 618 goto out; 619 620 freecopy: 621 m_freem(mcopy); 622 goto out; 623 bad: 624 m_freem(m); 625 out: 626 if (rt != NULL) 627 RTFREE(rt); 628 } 629