1 /* $FreeBSD$ */ 2 /* $KAME: ip6_forward.c,v 1.69 2001/05/17 03:48:30 itojun Exp $ */ 3 4 /*- 5 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 6 * All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of the project nor the names of its contributors 17 * may be used to endorse or promote products derived from this software 18 * without specific prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 30 * SUCH DAMAGE. 31 */ 32 33 #include "opt_inet.h" 34 #include "opt_inet6.h" 35 #include "opt_ipsec.h" 36 #include "opt_ipstealth.h" 37 38 #include <sys/param.h> 39 #include <sys/systm.h> 40 #include <sys/malloc.h> 41 #include <sys/mbuf.h> 42 #include <sys/domain.h> 43 #include <sys/protosw.h> 44 #include <sys/socket.h> 45 #include <sys/errno.h> 46 #include <sys/time.h> 47 #include <sys/kernel.h> 48 #include <sys/syslog.h> 49 50 #include <net/if.h> 51 #include <net/route.h> 52 #include <net/pfil.h> 53 54 #include <netinet/in.h> 55 #include <netinet/in_var.h> 56 #include <netinet/in_systm.h> 57 #include <netinet/ip.h> 58 #include <netinet/ip_var.h> 59 #include <netinet6/in6_var.h> 60 #include <netinet/ip6.h> 61 #include <netinet6/ip6_var.h> 62 #include <netinet6/scope6_var.h> 63 #include <netinet/icmp6.h> 64 #include <netinet6/nd6.h> 65 66 #include <netinet/in_pcb.h> 67 68 #ifdef IPSEC 69 #include <netinet6/ipsec.h> 70 #ifdef INET6 71 #include <netinet6/ipsec6.h> 72 #endif 73 #include <netkey/key.h> 74 #endif /* IPSEC */ 75 76 #ifdef FAST_IPSEC 77 #include <netipsec/ipsec.h> 78 #include <netipsec/ipsec6.h> 79 #include <netipsec/key.h> 80 #define IPSEC 81 #endif /* FAST_IPSEC */ 82 83 #include <net/net_osdep.h> 84 85 #include <netinet6/ip6protosw.h> 86 87 struct route_in6 ip6_forward_rt; 88 89 /* 90 * Forward a packet. If some error occurs return the sender 91 * an icmp packet. Note we can't always generate a meaningful 92 * icmp message because icmp doesn't have a large enough repertoire 93 * of codes and types. 94 * 95 * If not forwarding, just drop the packet. This could be confusing 96 * if ipforwarding was zero but some routing protocol was advancing 97 * us as a gateway to somewhere. However, we must let the routing 98 * protocol deal with that. 99 * 100 */ 101 102 void 103 ip6_forward(m, srcrt) 104 struct mbuf *m; 105 int srcrt; 106 { 107 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 108 struct sockaddr_in6 *dst = NULL; 109 struct rtentry *rt = NULL; 110 int error, type = 0, code = 0; 111 struct mbuf *mcopy = NULL; 112 struct ifnet *origifp; /* maybe unnecessary */ 113 u_int32_t inzone, outzone; 114 struct in6_addr src_in6, dst_in6; 115 #ifdef IPSEC 116 struct secpolicy *sp = NULL; 117 int ipsecrt = 0; 118 #endif 119 120 GIANT_REQUIRED; /* XXX bz: ip6_forward_rt */ 121 122 #ifdef IPSEC 123 /* 124 * Check AH/ESP integrity. 125 */ 126 /* 127 * Don't increment ip6s_cantforward because this is the check 128 * before forwarding packet actually. 129 */ 130 if (ipsec6_in_reject(m, NULL)) { 131 #if !defined(FAST_IPSEC) 132 ipsec6stat.in_polvio++; 133 #endif 134 m_freem(m); 135 return; 136 } 137 #endif /* IPSEC */ 138 139 /* 140 * Do not forward packets to multicast destination (should be handled 141 * by ip6_mforward(). 142 * Do not forward packets with unspecified source. It was discussed 143 * in July 2000, on the ipngwg mailing list. 144 */ 145 if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 || 146 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 147 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 148 ip6stat.ip6s_cantforward++; 149 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ 150 if (ip6_log_time + ip6_log_interval < time_second) { 151 ip6_log_time = time_second; 152 log(LOG_DEBUG, 153 "cannot forward " 154 "from %s to %s nxt %d received on %s\n", 155 ip6_sprintf(&ip6->ip6_src), 156 ip6_sprintf(&ip6->ip6_dst), 157 ip6->ip6_nxt, 158 if_name(m->m_pkthdr.rcvif)); 159 } 160 m_freem(m); 161 return; 162 } 163 164 #ifdef IPSTEALTH 165 if (!ip6stealth) { 166 #endif 167 if (ip6->ip6_hlim <= IPV6_HLIMDEC) { 168 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ 169 icmp6_error(m, ICMP6_TIME_EXCEEDED, 170 ICMP6_TIME_EXCEED_TRANSIT, 0); 171 return; 172 } 173 ip6->ip6_hlim -= IPV6_HLIMDEC; 174 175 #ifdef IPSTEALTH 176 } 177 #endif 178 179 /* 180 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU - 181 * size of IPv6 + ICMPv6 headers) bytes of the packet in case 182 * we need to generate an ICMP6 message to the src. 183 * Thanks to M_EXT, in most cases copy will not occur. 184 * 185 * It is important to save it before IPsec processing as IPsec 186 * processing may modify the mbuf. 187 */ 188 mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN)); 189 190 #ifdef IPSEC 191 /* get a security policy for this packet */ 192 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 193 IP_FORWARDING, &error); 194 if (sp == NULL) { 195 ipsec6stat.out_inval++; 196 ip6stat.ip6s_cantforward++; 197 if (mcopy) { 198 #if 0 199 /* XXX: what icmp ? */ 200 #else 201 m_freem(mcopy); 202 #endif 203 } 204 m_freem(m); 205 return; 206 } 207 208 error = 0; 209 210 /* check policy */ 211 switch (sp->policy) { 212 case IPSEC_POLICY_DISCARD: 213 /* 214 * This packet is just discarded. 215 */ 216 ipsec6stat.out_polvio++; 217 ip6stat.ip6s_cantforward++; 218 key_freesp(sp); 219 if (mcopy) { 220 #if 0 221 /* XXX: what icmp ? */ 222 #else 223 m_freem(mcopy); 224 #endif 225 } 226 m_freem(m); 227 return; 228 229 case IPSEC_POLICY_BYPASS: 230 case IPSEC_POLICY_NONE: 231 /* no need to do IPsec. */ 232 key_freesp(sp); 233 goto skip_ipsec; 234 235 case IPSEC_POLICY_IPSEC: 236 if (sp->req == NULL) { 237 /* XXX should be panic ? */ 238 printf("ip6_forward: No IPsec request specified.\n"); 239 ip6stat.ip6s_cantforward++; 240 key_freesp(sp); 241 if (mcopy) { 242 #if 0 243 /* XXX: what icmp ? */ 244 #else 245 m_freem(mcopy); 246 #endif 247 } 248 m_freem(m); 249 return; 250 } 251 /* do IPsec */ 252 break; 253 254 case IPSEC_POLICY_ENTRUST: 255 default: 256 /* should be panic ?? */ 257 printf("ip6_forward: Invalid policy found. %d\n", sp->policy); 258 key_freesp(sp); 259 goto skip_ipsec; 260 } 261 262 { 263 struct ipsecrequest *isr = NULL; 264 struct ipsec_output_state state; 265 266 /* 267 * when the kernel forwards a packet, it is not proper to apply 268 * IPsec transport mode to the packet is not proper. this check 269 * avoid from this. 270 * at present, if there is even a transport mode SA request in the 271 * security policy, the kernel does not apply IPsec to the packet. 272 * this check is not enough because the following case is valid. 273 * ipsec esp/tunnel/xxx-xxx/require esp/transport//require; 274 */ 275 for (isr = sp->req; isr; isr = isr->next) { 276 if (isr->saidx.mode == IPSEC_MODE_ANY) 277 goto doipsectunnel; 278 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) 279 goto doipsectunnel; 280 } 281 282 /* 283 * if there's no need for tunnel mode IPsec, skip. 284 */ 285 if (!isr) 286 goto skip_ipsec; 287 288 doipsectunnel: 289 /* 290 * All the extension headers will become inaccessible 291 * (since they can be encrypted). 292 * Don't panic, we need no more updates to extension headers 293 * on inner IPv6 packet (since they are now encapsulated). 294 * 295 * IPv6 [ESP|AH] IPv6 [extension headers] payload 296 */ 297 bzero(&state, sizeof(state)); 298 state.m = m; 299 state.ro = NULL; /* update at ipsec6_output_tunnel() */ 300 state.dst = NULL; /* update at ipsec6_output_tunnel() */ 301 302 error = ipsec6_output_tunnel(&state, sp, 0); 303 304 m = state.m; 305 key_freesp(sp); 306 307 if (error) { 308 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 309 switch (error) { 310 case EHOSTUNREACH: 311 case ENETUNREACH: 312 case EMSGSIZE: 313 case ENOBUFS: 314 case ENOMEM: 315 break; 316 default: 317 printf("ip6_output (ipsec): error code %d\n", error); 318 /* FALLTHROUGH */ 319 case ENOENT: 320 /* don't show these error codes to the user */ 321 break; 322 } 323 ip6stat.ip6s_cantforward++; 324 if (mcopy) { 325 #if 0 326 /* XXX: what icmp ? */ 327 #else 328 m_freem(mcopy); 329 #endif 330 } 331 m_freem(m); 332 return; 333 } 334 335 if (ip6 != mtod(m, struct ip6_hdr *)) { 336 /* 337 * now tunnel mode headers are added. we are originating 338 * packet instead of forwarding the packet. 339 */ 340 ip6_output(m, NULL, NULL, IPV6_FORWARDING/*XXX*/, NULL, NULL, 341 NULL); 342 goto freecopy; 343 } 344 345 /* adjust pointer */ 346 dst = (struct sockaddr_in6 *)state.dst; 347 rt = state.ro ? state.ro->ro_rt : NULL; 348 if (dst != NULL && rt != NULL) 349 ipsecrt = 1; 350 } 351 skip_ipsec: 352 #endif /* IPSEC */ 353 354 #ifdef IPSEC 355 if (ipsecrt) 356 goto skip_routing; 357 #endif 358 359 dst = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst; 360 if (!srcrt) { 361 /* ip6_forward_rt.ro_dst.sin6_addr is equal to ip6->ip6_dst */ 362 if (ip6_forward_rt.ro_rt == 0 || 363 (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0) { 364 if (ip6_forward_rt.ro_rt) { 365 RTFREE(ip6_forward_rt.ro_rt); 366 ip6_forward_rt.ro_rt = 0; 367 } 368 369 /* this probably fails but give it a try again */ 370 rtalloc((struct route *)&ip6_forward_rt); 371 } 372 373 if (ip6_forward_rt.ro_rt == 0) { 374 ip6stat.ip6s_noroute++; 375 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute); 376 if (mcopy) { 377 icmp6_error(mcopy, ICMP6_DST_UNREACH, 378 ICMP6_DST_UNREACH_NOROUTE, 0); 379 } 380 m_freem(m); 381 return; 382 } 383 } else if ((rt = ip6_forward_rt.ro_rt) == 0 || 384 !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) { 385 if (ip6_forward_rt.ro_rt) { 386 RTFREE(ip6_forward_rt.ro_rt); 387 ip6_forward_rt.ro_rt = 0; 388 } 389 bzero(dst, sizeof(*dst)); 390 dst->sin6_len = sizeof(struct sockaddr_in6); 391 dst->sin6_family = AF_INET6; 392 dst->sin6_addr = ip6->ip6_dst; 393 394 rtalloc((struct route *)&ip6_forward_rt); 395 if (ip6_forward_rt.ro_rt == 0) { 396 ip6stat.ip6s_noroute++; 397 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute); 398 if (mcopy) { 399 icmp6_error(mcopy, ICMP6_DST_UNREACH, 400 ICMP6_DST_UNREACH_NOROUTE, 0); 401 } 402 m_freem(m); 403 return; 404 } 405 } 406 rt = ip6_forward_rt.ro_rt; 407 #ifdef IPSEC 408 skip_routing:; 409 #endif 410 411 /* 412 * Source scope check: if a packet can't be delivered to its 413 * destination for the reason that the destination is beyond the scope 414 * of the source address, discard the packet and return an icmp6 415 * destination unreachable error with Code 2 (beyond scope of source 416 * address). We use a local copy of ip6_src, since in6_setscope() 417 * will possibly modify its first argument. 418 * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1] 419 */ 420 src_in6 = ip6->ip6_src; 421 if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) { 422 /* XXX: this should not happen */ 423 ip6stat.ip6s_cantforward++; 424 ip6stat.ip6s_badscope++; 425 m_freem(m); 426 return; 427 } 428 if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) { 429 ip6stat.ip6s_cantforward++; 430 ip6stat.ip6s_badscope++; 431 m_freem(m); 432 return; 433 } 434 if (inzone != outzone 435 #ifdef IPSEC 436 && !ipsecrt 437 #endif 438 ) { 439 ip6stat.ip6s_cantforward++; 440 ip6stat.ip6s_badscope++; 441 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard); 442 443 if (ip6_log_time + ip6_log_interval < time_second) { 444 ip6_log_time = time_second; 445 log(LOG_DEBUG, 446 "cannot forward " 447 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 448 ip6_sprintf(&ip6->ip6_src), 449 ip6_sprintf(&ip6->ip6_dst), 450 ip6->ip6_nxt, 451 if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp)); 452 } 453 if (mcopy) 454 icmp6_error(mcopy, ICMP6_DST_UNREACH, 455 ICMP6_DST_UNREACH_BEYONDSCOPE, 0); 456 m_freem(m); 457 return; 458 } 459 460 /* 461 * Destination scope check: if a packet is going to break the scope 462 * zone of packet's destination address, discard it. This case should 463 * usually be prevented by appropriately-configured routing table, but 464 * we need an explicit check because we may mistakenly forward the 465 * packet to a different zone by (e.g.) a default route. 466 */ 467 dst_in6 = ip6->ip6_dst; 468 if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 || 469 in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 || 470 inzone != outzone) { 471 ip6stat.ip6s_cantforward++; 472 ip6stat.ip6s_badscope++; 473 m_freem(m); 474 return; 475 } 476 477 if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) { 478 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig); 479 if (mcopy) { 480 u_long mtu; 481 #ifdef IPSEC 482 struct secpolicy *sp; 483 int ipsecerror; 484 size_t ipsechdrsiz; 485 #endif 486 487 mtu = IN6_LINKMTU(rt->rt_ifp); 488 #ifdef IPSEC 489 /* 490 * When we do IPsec tunnel ingress, we need to play 491 * with the link value (decrement IPsec header size 492 * from mtu value). The code is much simpler than v4 493 * case, as we have the outgoing interface for 494 * encapsulated packet as "rt->rt_ifp". 495 */ 496 sp = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND, 497 IP_FORWARDING, &ipsecerror); 498 if (sp) { 499 ipsechdrsiz = ipsec6_hdrsiz(mcopy, 500 IPSEC_DIR_OUTBOUND, NULL); 501 if (ipsechdrsiz < mtu) 502 mtu -= ipsechdrsiz; 503 } 504 505 /* 506 * if mtu becomes less than minimum MTU, 507 * tell minimum MTU (and I'll need to fragment it). 508 */ 509 if (mtu < IPV6_MMTU) 510 mtu = IPV6_MMTU; 511 #endif 512 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu); 513 } 514 m_freem(m); 515 return; 516 } 517 518 if (rt->rt_flags & RTF_GATEWAY) 519 dst = (struct sockaddr_in6 *)rt->rt_gateway; 520 521 /* 522 * If we are to forward the packet using the same interface 523 * as one we got the packet from, perhaps we should send a redirect 524 * to sender to shortcut a hop. 525 * Only send redirect if source is sending directly to us, 526 * and if packet was not source routed (or has any options). 527 * Also, don't send redirect if forwarding using a route 528 * modified by a redirect. 529 */ 530 if (rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && 531 #ifdef IPSEC 532 !ipsecrt && 533 #endif 534 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) { 535 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) { 536 /* 537 * If the incoming interface is equal to the outgoing 538 * one, and the link attached to the interface is 539 * point-to-point, then it will be highly probable 540 * that a routing loop occurs. Thus, we immediately 541 * drop the packet and send an ICMPv6 error message. 542 * 543 * type/code is based on suggestion by Rich Draves. 544 * not sure if it is the best pick. 545 */ 546 icmp6_error(mcopy, ICMP6_DST_UNREACH, 547 ICMP6_DST_UNREACH_ADDR, 0); 548 m_freem(m); 549 return; 550 } 551 type = ND_REDIRECT; 552 } 553 554 /* 555 * Fake scoped addresses. Note that even link-local source or 556 * destinaion can appear, if the originating node just sends the 557 * packet to us (without address resolution for the destination). 558 * Since both icmp6_error and icmp6_redirect_output fill the embedded 559 * link identifiers, we can do this stuff after making a copy for 560 * returning an error. 561 */ 562 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { 563 /* 564 * See corresponding comments in ip6_output. 565 * XXX: but is it possible that ip6_forward() sends a packet 566 * to a loopback interface? I don't think so, and thus 567 * I bark here. (jinmei@kame.net) 568 * XXX: it is common to route invalid packets to loopback. 569 * also, the codepath will be visited on use of ::1 in 570 * rthdr. (itojun) 571 */ 572 #if 1 573 if (0) 574 #else 575 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0) 576 #endif 577 { 578 printf("ip6_forward: outgoing interface is loopback. " 579 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 580 ip6_sprintf(&ip6->ip6_src), 581 ip6_sprintf(&ip6->ip6_dst), 582 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif), 583 if_name(rt->rt_ifp)); 584 } 585 586 /* we can just use rcvif in forwarding. */ 587 origifp = m->m_pkthdr.rcvif; 588 } 589 else 590 origifp = rt->rt_ifp; 591 /* 592 * clear embedded scope identifiers if necessary. 593 * in6_clearscope will touch the addresses only when necessary. 594 */ 595 in6_clearscope(&ip6->ip6_src); 596 in6_clearscope(&ip6->ip6_dst); 597 598 /* Jump over all PFIL processing if hooks are not active. */ 599 if (!PFIL_HOOKED(&inet6_pfil_hook)) 600 goto pass; 601 602 /* Run through list of hooks for output packets. */ 603 error = pfil_run_hooks(&inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT, NULL); 604 if (error != 0) 605 goto senderr; 606 if (m == NULL) 607 goto freecopy; 608 ip6 = mtod(m, struct ip6_hdr *); 609 610 pass: 611 error = nd6_output(rt->rt_ifp, origifp, m, dst, rt); 612 if (error) { 613 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard); 614 ip6stat.ip6s_cantforward++; 615 } else { 616 ip6stat.ip6s_forward++; 617 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward); 618 if (type) 619 ip6stat.ip6s_redirectsent++; 620 else { 621 if (mcopy) 622 goto freecopy; 623 } 624 } 625 626 senderr: 627 if (mcopy == NULL) 628 return; 629 switch (error) { 630 case 0: 631 if (type == ND_REDIRECT) { 632 icmp6_redirect_output(mcopy, rt); 633 return; 634 } 635 goto freecopy; 636 637 case EMSGSIZE: 638 /* xxx MTU is constant in PPP? */ 639 goto freecopy; 640 641 case ENOBUFS: 642 /* Tell source to slow down like source quench in IP? */ 643 goto freecopy; 644 645 case ENETUNREACH: /* shouldn't happen, checked above */ 646 case EHOSTUNREACH: 647 case ENETDOWN: 648 case EHOSTDOWN: 649 default: 650 type = ICMP6_DST_UNREACH; 651 code = ICMP6_DST_UNREACH_ADDR; 652 break; 653 } 654 icmp6_error(mcopy, type, code, 0); 655 return; 656 657 freecopy: 658 m_freem(mcopy); 659 return; 660 } 661