1 /* $NetBSD: ip6_forward.c,v 1.43 2004/07/16 01:12:02 itojun Exp $ */ 2 /* $KAME: ip6_forward.c,v 1.109 2002/09/11 08:10:17 sakane 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 <sys/cdefs.h> 34 __KERNEL_RCSID(0, "$NetBSD: ip6_forward.c,v 1.43 2004/07/16 01:12:02 itojun Exp $"); 35 36 #include "opt_ipsec.h" 37 #include "opt_pfil_hooks.h" 38 39 #include <sys/param.h> 40 #include <sys/systm.h> 41 #include <sys/malloc.h> 42 #include <sys/mbuf.h> 43 #include <sys/domain.h> 44 #include <sys/protosw.h> 45 #include <sys/socket.h> 46 #include <sys/errno.h> 47 #include <sys/time.h> 48 #include <sys/kernel.h> 49 #include <sys/syslog.h> 50 51 #include <net/if.h> 52 #include <net/route.h> 53 54 #include <netinet/in.h> 55 #include <netinet/in_var.h> 56 #include <netinet/ip_var.h> 57 #include <netinet/ip6.h> 58 #include <netinet6/ip6_var.h> 59 #include <netinet/icmp6.h> 60 #include <netinet6/nd6.h> 61 62 #ifdef IPSEC 63 #include <netinet6/ipsec.h> 64 #include <netkey/key.h> 65 #endif /* IPSEC */ 66 67 #ifdef PFIL_HOOKS 68 #include <net/pfil.h> 69 #endif 70 71 #include <net/net_osdep.h> 72 73 struct route_in6 ip6_forward_rt; 74 75 #ifdef PFIL_HOOKS 76 extern struct pfil_head inet6_pfil_hook; /* XXX */ 77 #endif 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 92 void 93 ip6_forward(m, srcrt) 94 struct mbuf *m; 95 int srcrt; 96 { 97 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 98 struct sockaddr_in6 *dst; 99 struct rtentry *rt; 100 int error = 0, type = 0, code = 0; 101 struct mbuf *mcopy = NULL; 102 struct ifnet *origifp; /* maybe unnecessary */ 103 #ifdef IPSEC 104 struct secpolicy *sp = NULL; 105 int ipsecrt = 0; 106 #endif 107 108 #ifdef IPSEC 109 /* 110 * Check AH/ESP integrity. 111 */ 112 /* 113 * Don't increment ip6s_cantforward because this is the check 114 * before forwarding packet actually. 115 */ 116 if (ipsec6_in_reject(m, NULL)) { 117 ipsec6stat.in_polvio++; 118 m_freem(m); 119 return; 120 } 121 #endif /* IPSEC */ 122 123 /* 124 * Do not forward packets to multicast destination (should be handled 125 * by ip6_mforward(). 126 * Do not forward packets with unspecified source. It was discussed 127 * in July 2000, on ipngwg mailing list. 128 */ 129 if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 || 130 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 131 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 132 ip6stat.ip6s_cantforward++; 133 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ 134 if (ip6_log_time + ip6_log_interval < time.tv_sec) { 135 ip6_log_time = time.tv_sec; 136 log(LOG_DEBUG, 137 "cannot forward " 138 "from %s to %s nxt %d received on %s\n", 139 ip6_sprintf(&ip6->ip6_src), 140 ip6_sprintf(&ip6->ip6_dst), 141 ip6->ip6_nxt, 142 if_name(m->m_pkthdr.rcvif)); 143 } 144 m_freem(m); 145 return; 146 } 147 148 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 /* 157 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU - 158 * size of IPv6 + ICMPv6 headers) bytes of the packet in case 159 * we need to generate an ICMP6 message to the src. 160 * Thanks to M_EXT, in most cases copy will not occur. 161 * 162 * It is important to save it before IPsec processing as IPsec 163 * processing may modify the mbuf. 164 */ 165 mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN)); 166 167 #ifdef IPSEC 168 /* get a security policy for this packet */ 169 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 170 IP_FORWARDING, &error); 171 if (sp == NULL) { 172 ipsec6stat.out_inval++; 173 ip6stat.ip6s_cantforward++; 174 if (mcopy) { 175 #if 0 176 /* XXX: what icmp ? */ 177 #else 178 m_freem(mcopy); 179 #endif 180 } 181 m_freem(m); 182 return; 183 } 184 185 error = 0; 186 187 /* check policy */ 188 switch (sp->policy) { 189 case IPSEC_POLICY_DISCARD: 190 /* 191 * This packet is just discarded. 192 */ 193 ipsec6stat.out_polvio++; 194 ip6stat.ip6s_cantforward++; 195 key_freesp(sp); 196 if (mcopy) { 197 #if 0 198 /* XXX: what icmp ? */ 199 #else 200 m_freem(mcopy); 201 #endif 202 } 203 m_freem(m); 204 return; 205 206 case IPSEC_POLICY_BYPASS: 207 case IPSEC_POLICY_NONE: 208 /* no need to do IPsec. */ 209 key_freesp(sp); 210 goto skip_ipsec; 211 212 case IPSEC_POLICY_IPSEC: 213 if (sp->req == NULL) { 214 /* XXX should be panic ? */ 215 printf("ip6_forward: No IPsec request specified.\n"); 216 ip6stat.ip6s_cantforward++; 217 key_freesp(sp); 218 if (mcopy) { 219 #if 0 220 /* XXX: what icmp ? */ 221 #else 222 m_freem(mcopy); 223 #endif 224 } 225 m_freem(m); 226 return; 227 } 228 /* do IPsec */ 229 break; 230 231 case IPSEC_POLICY_ENTRUST: 232 default: 233 /* should be panic ?? */ 234 printf("ip6_forward: Invalid policy found. %d\n", sp->policy); 235 key_freesp(sp); 236 goto skip_ipsec; 237 } 238 239 { 240 struct ipsecrequest *isr = NULL; 241 struct ipsec_output_state state; 242 243 /* 244 * when the kernel forwards a packet, it is not proper to apply 245 * IPsec transport mode to the packet is not proper. this check 246 * avoid from this. 247 * at present, if there is even a transport mode SA request in the 248 * security policy, the kernel does not apply IPsec to the packet. 249 * this check is not enough because the following case is valid. 250 * ipsec esp/tunnel/xxx-xxx/require esp/transport//require; 251 */ 252 for (isr = sp->req; isr; isr = isr->next) { 253 if (isr->saidx.mode == IPSEC_MODE_ANY) 254 goto doipsectunnel; 255 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) 256 goto doipsectunnel; 257 } 258 259 /* 260 * if there's no need for tunnel mode IPsec, skip. 261 */ 262 if (!isr) 263 goto skip_ipsec; 264 265 doipsectunnel: 266 /* 267 * All the extension headers will become inaccessible 268 * (since they can be encrypted). 269 * Don't panic, we need no more updates to extension headers 270 * on inner IPv6 packet (since they are now encapsulated). 271 * 272 * IPv6 [ESP|AH] IPv6 [extension headers] payload 273 */ 274 bzero(&state, sizeof(state)); 275 state.m = m; 276 state.ro = NULL; /* update at ipsec6_output_tunnel() */ 277 state.dst = NULL; /* update at ipsec6_output_tunnel() */ 278 279 error = ipsec6_output_tunnel(&state, sp, 0); 280 281 m = state.m; 282 key_freesp(sp); 283 284 if (error) { 285 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 286 switch (error) { 287 case EHOSTUNREACH: 288 case ENETUNREACH: 289 case EMSGSIZE: 290 case ENOBUFS: 291 case ENOMEM: 292 break; 293 default: 294 printf("ip6_output (ipsec): error code %d\n", error); 295 /* FALLTHROUGH */ 296 case ENOENT: 297 /* don't show these error codes to the user */ 298 break; 299 } 300 ip6stat.ip6s_cantforward++; 301 if (mcopy) { 302 #if 0 303 /* XXX: what icmp ? */ 304 #else 305 m_freem(mcopy); 306 #endif 307 } 308 m_freem(m); 309 return; 310 } 311 312 if (ip6 != mtod(m, struct ip6_hdr *)) { 313 /* 314 * now tunnel mode headers are added. we are originating 315 * packet instead of forwarding the packet. 316 */ 317 ip6_output(m, NULL, NULL, IPV6_FORWARDING/*XXX*/, NULL, NULL, 318 NULL); 319 goto freecopy; 320 } 321 322 /* adjust pointer */ 323 rt = state.ro ? state.ro->ro_rt : NULL; 324 dst = (struct sockaddr_in6 *)state.dst; 325 if (dst != NULL && rt != NULL) { 326 ipsecrt = 1; 327 goto skip_routing; 328 } 329 } 330 skip_ipsec: 331 #endif /* IPSEC */ 332 333 dst = &ip6_forward_rt.ro_dst; 334 if (!srcrt) { 335 /* 336 * ip6_forward_rt.ro_dst.sin6_addr is equal to ip6->ip6_dst 337 */ 338 if (ip6_forward_rt.ro_rt == 0 || 339 (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0) { 340 if (ip6_forward_rt.ro_rt) { 341 RTFREE(ip6_forward_rt.ro_rt); 342 ip6_forward_rt.ro_rt = 0; 343 } 344 /* this probably fails but give it a try again */ 345 rtalloc((struct route *)&ip6_forward_rt); 346 } 347 348 if (ip6_forward_rt.ro_rt == 0) { 349 ip6stat.ip6s_noroute++; 350 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */ 351 if (mcopy) { 352 icmp6_error(mcopy, ICMP6_DST_UNREACH, 353 ICMP6_DST_UNREACH_NOROUTE, 0); 354 } 355 m_freem(m); 356 return; 357 } 358 } else if ((rt = ip6_forward_rt.ro_rt) == 0 || 359 !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) { 360 if (ip6_forward_rt.ro_rt) { 361 RTFREE(ip6_forward_rt.ro_rt); 362 ip6_forward_rt.ro_rt = 0; 363 } 364 bzero(dst, sizeof(*dst)); 365 dst->sin6_len = sizeof(struct sockaddr_in6); 366 dst->sin6_family = AF_INET6; 367 dst->sin6_addr = ip6->ip6_dst; 368 369 rtalloc((struct route *)&ip6_forward_rt); 370 if (ip6_forward_rt.ro_rt == 0) { 371 ip6stat.ip6s_noroute++; 372 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */ 373 if (mcopy) { 374 icmp6_error(mcopy, ICMP6_DST_UNREACH, 375 ICMP6_DST_UNREACH_NOROUTE, 0); 376 } 377 m_freem(m); 378 return; 379 } 380 } 381 rt = ip6_forward_rt.ro_rt; 382 #ifdef IPSEC 383 skip_routing:; 384 #endif /* IPSEC */ 385 386 /* 387 * Scope check: if a packet can't be delivered to its destination 388 * for the reason that the destination is beyond the scope of the 389 * source address, discard the packet and return an icmp6 destination 390 * unreachable error with Code 2 (beyond scope of source address). 391 * [draft-ietf-ipngwg-icmp-v3-00.txt, Section 3.1] 392 */ 393 if (in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_src) != 394 in6_addr2scopeid(rt->rt_ifp, &ip6->ip6_src) 395 #ifdef IPSEC 396 && !ipsecrt 397 #endif 398 ) { 399 ip6stat.ip6s_cantforward++; 400 ip6stat.ip6s_badscope++; 401 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard); 402 403 if (ip6_log_time + ip6_log_interval < time.tv_sec) { 404 ip6_log_time = time.tv_sec; 405 log(LOG_DEBUG, 406 "cannot forward " 407 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 408 ip6_sprintf(&ip6->ip6_src), 409 ip6_sprintf(&ip6->ip6_dst), 410 ip6->ip6_nxt, 411 if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp)); 412 } 413 if (mcopy) 414 icmp6_error(mcopy, ICMP6_DST_UNREACH, 415 ICMP6_DST_UNREACH_BEYONDSCOPE, 0); 416 m_freem(m); 417 return; 418 } 419 420 if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) { 421 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig); 422 if (mcopy) { 423 u_long mtu; 424 #ifdef IPSEC 425 struct secpolicy *sp; 426 int ipsecerror; 427 size_t ipsechdrsiz; 428 #endif 429 430 mtu = IN6_LINKMTU(rt->rt_ifp); 431 #ifdef IPSEC 432 /* 433 * When we do IPsec tunnel ingress, we need to play 434 * with the link value (decrement IPsec header size 435 * from mtu value). The code is much simpler than v4 436 * case, as we have the outgoing interface for 437 * encapsulated packet as "rt->rt_ifp". 438 */ 439 sp = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND, 440 IP_FORWARDING, &ipsecerror); 441 if (sp) { 442 ipsechdrsiz = ipsec6_hdrsiz(mcopy, 443 IPSEC_DIR_OUTBOUND, NULL); 444 if (ipsechdrsiz < mtu) 445 mtu -= ipsechdrsiz; 446 } 447 448 /* 449 * if mtu becomes less than minimum MTU, 450 * tell minimum MTU (and I'll need to fragment it). 451 */ 452 if (mtu < IPV6_MMTU) 453 mtu = IPV6_MMTU; 454 #endif 455 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu); 456 } 457 m_freem(m); 458 return; 459 } 460 461 if (rt->rt_flags & RTF_GATEWAY) 462 dst = (struct sockaddr_in6 *)rt->rt_gateway; 463 464 /* 465 * If we are to forward the packet using the same interface 466 * as one we got the packet from, perhaps we should send a redirect 467 * to sender to shortcut a hop. 468 * Only send redirect if source is sending directly to us, 469 * and if packet was not source routed (or has any options). 470 * Also, don't send redirect if forwarding using a route 471 * modified by a redirect. 472 */ 473 if (rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && ip6_sendredirects && 474 #ifdef IPSEC 475 !ipsecrt && 476 #endif 477 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) { 478 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) && 479 nd6_is_addr_neighbor((struct sockaddr_in6 *)&ip6_forward_rt.ro_dst, rt->rt_ifp)) { 480 /* 481 * If the incoming interface is equal to the outgoing 482 * one, the link attached to the interface is 483 * point-to-point, and the IPv6 destination is 484 * regarded as on-link on the link, then it will be 485 * highly probable that the destination address does 486 * not exist on the link and that the packet is going 487 * to loop. Thus, we immediately drop the packet and 488 * send an ICMPv6 error message. 489 * For other routing loops, we dare to let the packet 490 * go to the loop, so that a remote diagnosing host 491 * can detect the loop by traceroute. 492 * type/code is based on suggestion by Rich Draves. 493 * not sure if it is the best pick. 494 */ 495 icmp6_error(mcopy, ICMP6_DST_UNREACH, 496 ICMP6_DST_UNREACH_ADDR, 0); 497 m_freem(m); 498 return; 499 } 500 type = ND_REDIRECT; 501 } 502 503 /* 504 * Fake scoped addresses. Note that even link-local source or 505 * destinaion can appear, if the originating node just sends the 506 * packet to us (without address resolution for the destination). 507 * Since both icmp6_error and icmp6_redirect_output fill the embedded 508 * link identifiers, we can do this stuff after making a copy for 509 * returning an error. 510 */ 511 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { 512 /* 513 * See corresponding comments in ip6_output. 514 * XXX: but is it possible that ip6_forward() sends a packet 515 * to a loopback interface? I don't think so, and thus 516 * I bark here. (jinmei@kame.net) 517 * XXX: it is common to route invalid packets to loopback. 518 * also, the codepath will be visited on use of ::1 in 519 * rthdr. (itojun) 520 */ 521 #if 1 522 if (0) 523 #else 524 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0) 525 #endif 526 { 527 printf("ip6_forward: outgoing interface is loopback. " 528 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 529 ip6_sprintf(&ip6->ip6_src), 530 ip6_sprintf(&ip6->ip6_dst), 531 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif), 532 if_name(rt->rt_ifp)); 533 } 534 535 /* we can just use rcvif in forwarding. */ 536 origifp = m->m_pkthdr.rcvif; 537 } 538 else 539 origifp = rt->rt_ifp; 540 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) 541 ip6->ip6_src.s6_addr16[1] = 0; 542 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) 543 ip6->ip6_dst.s6_addr16[1] = 0; 544 545 #ifdef PFIL_HOOKS 546 /* 547 * Run through list of hooks for output packets. 548 */ 549 if ((error = pfil_run_hooks(&inet6_pfil_hook, &m, rt->rt_ifp, 550 PFIL_OUT)) != 0) 551 goto senderr; 552 if (m == NULL) 553 goto freecopy; 554 ip6 = mtod(m, struct ip6_hdr *); 555 #endif /* PFIL_HOOKS */ 556 557 error = nd6_output(rt->rt_ifp, origifp, m, dst, rt); 558 if (error) { 559 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard); 560 ip6stat.ip6s_cantforward++; 561 } else { 562 ip6stat.ip6s_forward++; 563 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward); 564 if (type) 565 ip6stat.ip6s_redirectsent++; 566 else { 567 if (mcopy) 568 goto freecopy; 569 } 570 } 571 572 #ifdef PFIL_HOOKS 573 senderr: 574 #endif 575 if (mcopy == NULL) 576 return; 577 switch (error) { 578 case 0: 579 if (type == ND_REDIRECT) { 580 icmp6_redirect_output(mcopy, rt); 581 return; 582 } 583 goto freecopy; 584 585 case EMSGSIZE: 586 /* xxx MTU is constant in PPP? */ 587 goto freecopy; 588 589 case ENOBUFS: 590 /* Tell source to slow down like source quench in IP? */ 591 goto freecopy; 592 593 case ENETUNREACH: /* shouldn't happen, checked above */ 594 case EHOSTUNREACH: 595 case ENETDOWN: 596 case EHOSTDOWN: 597 default: 598 type = ICMP6_DST_UNREACH; 599 code = ICMP6_DST_UNREACH_ADDR; 600 break; 601 } 602 icmp6_error(mcopy, type, code, 0); 603 return; 604 605 freecopy: 606 m_freem(mcopy); 607 return; 608 } 609