ip6_forward.c revision 162045
1/* $FreeBSD: head/sys/netinet6/ip6_forward.c 162045 2006-09-05 19:20:42Z jhay $ */ 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 <netinet6/ip6protosw.h> 84 85struct route_in6 ip6_forward_rt; 86 87/* 88 * Forward a packet. If some error occurs return the sender 89 * an icmp packet. Note we can't always generate a meaningful 90 * icmp message because icmp doesn't have a large enough repertoire 91 * of codes and types. 92 * 93 * If not forwarding, just drop the packet. This could be confusing 94 * if ipforwarding was zero but some routing protocol was advancing 95 * us as a gateway to somewhere. However, we must let the routing 96 * protocol deal with that. 97 * 98 */ 99 100void 101ip6_forward(m, srcrt) 102 struct mbuf *m; 103 int srcrt; 104{ 105 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 106 struct sockaddr_in6 *dst = NULL; 107 struct rtentry *rt = NULL; 108 int error, type = 0, code = 0; 109 struct mbuf *mcopy = NULL; 110 struct ifnet *origifp; /* maybe unnecessary */ 111 u_int32_t inzone, outzone; 112 struct in6_addr src_in6, dst_in6; 113#ifdef IPSEC 114 struct secpolicy *sp = NULL; 115 int ipsecrt = 0; 116#endif 117 118 GIANT_REQUIRED; /* XXX bz: ip6_forward_rt */ 119 120#ifdef IPSEC 121 /* 122 * Check AH/ESP integrity. 123 */ 124 /* 125 * Don't increment ip6s_cantforward because this is the check 126 * before forwarding packet actually. 127 */ 128 if (ipsec6_in_reject(m, NULL)) { 129#if !defined(FAST_IPSEC) 130 ipsec6stat.in_polvio++; 131#endif 132 m_freem(m); 133 return; 134 } 135#endif /* IPSEC */ 136 137 /* 138 * Do not forward packets to multicast destination (should be handled 139 * by ip6_mforward(). 140 * Do not forward packets with unspecified source. It was discussed 141 * in July 2000, on the ipngwg mailing list. 142 */ 143 if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 || 144 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 145 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 146 ip6stat.ip6s_cantforward++; 147 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ 148 if (ip6_log_time + ip6_log_interval < time_second) { 149 ip6_log_time = time_second; 150 log(LOG_DEBUG, 151 "cannot forward " 152 "from %s to %s nxt %d received on %s\n", 153 ip6_sprintf(&ip6->ip6_src), 154 ip6_sprintf(&ip6->ip6_dst), 155 ip6->ip6_nxt, 156 if_name(m->m_pkthdr.rcvif)); 157 } 158 m_freem(m); 159 return; 160 } 161 162#ifdef IPSTEALTH 163 if (!ip6stealth) { 164#endif 165 if (ip6->ip6_hlim <= IPV6_HLIMDEC) { 166 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ 167 icmp6_error(m, ICMP6_TIME_EXCEEDED, 168 ICMP6_TIME_EXCEED_TRANSIT, 0); 169 return; 170 } 171 ip6->ip6_hlim -= IPV6_HLIMDEC; 172 173#ifdef IPSTEALTH 174 } 175#endif 176 177 /* 178 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU - 179 * size of IPv6 + ICMPv6 headers) bytes of the packet in case 180 * we need to generate an ICMP6 message to the src. 181 * Thanks to M_EXT, in most cases copy will not occur. 182 * 183 * It is important to save it before IPsec processing as IPsec 184 * processing may modify the mbuf. 185 */ 186 mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN)); 187 188#ifdef IPSEC 189 /* get a security policy for this packet */ 190 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 191 IP_FORWARDING, &error); 192 if (sp == NULL) { 193 ipsec6stat.out_inval++; 194 ip6stat.ip6s_cantforward++; 195 if (mcopy) { 196#if 0 197 /* XXX: what icmp ? */ 198#else 199 m_freem(mcopy); 200#endif 201 } 202 m_freem(m); 203 return; 204 } 205 206 error = 0; 207 208 /* check policy */ 209 switch (sp->policy) { 210 case IPSEC_POLICY_DISCARD: 211 /* 212 * This packet is just discarded. 213 */ 214 ipsec6stat.out_polvio++; 215 ip6stat.ip6s_cantforward++; 216 key_freesp(sp); 217 if (mcopy) { 218#if 0 219 /* XXX: what icmp ? */ 220#else 221 m_freem(mcopy); 222#endif 223 } 224 m_freem(m); 225 return; 226 227 case IPSEC_POLICY_BYPASS: 228 case IPSEC_POLICY_NONE: 229 /* no need to do IPsec. */ 230 key_freesp(sp); 231 goto skip_ipsec; 232 233 case IPSEC_POLICY_IPSEC: 234 if (sp->req == NULL) { 235 /* XXX should be panic ? */ 236 printf("ip6_forward: No IPsec request specified.\n"); 237 ip6stat.ip6s_cantforward++; 238 key_freesp(sp); 239 if (mcopy) { 240#if 0 241 /* XXX: what icmp ? */ 242#else 243 m_freem(mcopy); 244#endif 245 } 246 m_freem(m); 247 return; 248 } 249 /* do IPsec */ 250 break; 251 252 case IPSEC_POLICY_ENTRUST: 253 default: 254 /* should be panic ?? */ 255 printf("ip6_forward: Invalid policy found. %d\n", sp->policy); 256 key_freesp(sp); 257 goto skip_ipsec; 258 } 259 260 { 261 struct ipsecrequest *isr = NULL; 262 struct ipsec_output_state state; 263 264 /* 265 * when the kernel forwards a packet, it is not proper to apply 266 * IPsec transport mode to the packet is not proper. this check 267 * avoid from this. 268 * at present, if there is even a transport mode SA request in the 269 * security policy, the kernel does not apply IPsec to the packet. 270 * this check is not enough because the following case is valid. 271 * ipsec esp/tunnel/xxx-xxx/require esp/transport//require; 272 */ 273 for (isr = sp->req; isr; isr = isr->next) { 274 if (isr->saidx.mode == IPSEC_MODE_ANY) 275 goto doipsectunnel; 276 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) 277 goto doipsectunnel; 278 } 279 280 /* 281 * if there's no need for tunnel mode IPsec, skip. 282 */ 283 if (!isr) 284 goto skip_ipsec; 285 286 doipsectunnel: 287 /* 288 * All the extension headers will become inaccessible 289 * (since they can be encrypted). 290 * Don't panic, we need no more updates to extension headers 291 * on inner IPv6 packet (since they are now encapsulated). 292 * 293 * IPv6 [ESP|AH] IPv6 [extension headers] payload 294 */ 295 bzero(&state, sizeof(state)); 296 state.m = m; 297 state.ro = NULL; /* update at ipsec6_output_tunnel() */ 298 state.dst = NULL; /* update at ipsec6_output_tunnel() */ 299 300 error = ipsec6_output_tunnel(&state, sp, 0); 301 302 m = state.m; 303 key_freesp(sp); 304 305 if (error) { 306 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 307 switch (error) { 308 case EHOSTUNREACH: 309 case ENETUNREACH: 310 case EMSGSIZE: 311 case ENOBUFS: 312 case ENOMEM: 313 break; 314 default: 315 printf("ip6_output (ipsec): error code %d\n", error); 316 /* FALLTHROUGH */ 317 case ENOENT: 318 /* don't show these error codes to the user */ 319 break; 320 } 321 ip6stat.ip6s_cantforward++; 322 if (mcopy) { 323#if 0 324 /* XXX: what icmp ? */ 325#else 326 m_freem(mcopy); 327#endif 328 } 329 m_freem(m); 330 return; 331 } 332 333 if (ip6 != mtod(m, struct ip6_hdr *)) { 334 /* 335 * now tunnel mode headers are added. we are originating 336 * packet instead of forwarding the packet. 337 */ 338 ip6_output(m, NULL, NULL, IPV6_FORWARDING/*XXX*/, NULL, NULL, 339 NULL); 340 goto freecopy; 341 } 342 343 /* adjust pointer */ 344 dst = (struct sockaddr_in6 *)state.dst; 345 rt = state.ro ? state.ro->ro_rt : NULL; 346 if (dst != NULL && rt != NULL) 347 ipsecrt = 1; 348 } 349 skip_ipsec: 350#endif /* IPSEC */ 351 352#ifdef IPSEC 353 if (ipsecrt) 354 goto skip_routing; 355#endif 356 357 dst = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst; 358 if (!srcrt) { 359 /* ip6_forward_rt.ro_dst.sin6_addr is equal to ip6->ip6_dst */ 360 if (ip6_forward_rt.ro_rt == 0 || 361 (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0) { 362 if (ip6_forward_rt.ro_rt) { 363 RTFREE(ip6_forward_rt.ro_rt); 364 ip6_forward_rt.ro_rt = 0; 365 } 366 367 /* this probably fails but give it a try again */ 368 rtalloc((struct route *)&ip6_forward_rt); 369 } 370 371 if (ip6_forward_rt.ro_rt == 0) { 372 ip6stat.ip6s_noroute++; 373 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute); 374 if (mcopy) { 375 icmp6_error(mcopy, ICMP6_DST_UNREACH, 376 ICMP6_DST_UNREACH_NOROUTE, 0); 377 } 378 m_freem(m); 379 return; 380 } 381 } else if ((rt = ip6_forward_rt.ro_rt) == 0 || 382 !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) { 383 if (ip6_forward_rt.ro_rt) { 384 RTFREE(ip6_forward_rt.ro_rt); 385 ip6_forward_rt.ro_rt = 0; 386 } 387 bzero(dst, sizeof(*dst)); 388 dst->sin6_len = sizeof(struct sockaddr_in6); 389 dst->sin6_family = AF_INET6; 390 dst->sin6_addr = ip6->ip6_dst; 391 392 rtalloc((struct route *)&ip6_forward_rt); 393 if (ip6_forward_rt.ro_rt == 0) { 394 ip6stat.ip6s_noroute++; 395 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute); 396 if (mcopy) { 397 icmp6_error(mcopy, ICMP6_DST_UNREACH, 398 ICMP6_DST_UNREACH_NOROUTE, 0); 399 } 400 m_freem(m); 401 return; 402 } 403 } 404 rt = ip6_forward_rt.ro_rt; 405#ifdef IPSEC 406 skip_routing:; 407#endif 408 409 /* 410 * Source scope check: if a packet can't be delivered to its 411 * destination for the reason that the destination is beyond the scope 412 * of the source address, discard the packet and return an icmp6 413 * destination unreachable error with Code 2 (beyond scope of source 414 * address). We use a local copy of ip6_src, since in6_setscope() 415 * will possibly modify its first argument. 416 * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1] 417 */ 418 src_in6 = ip6->ip6_src; 419 if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) { 420 /* XXX: this should not happen */ 421 ip6stat.ip6s_cantforward++; 422 ip6stat.ip6s_badscope++; 423 m_freem(m); 424 return; 425 } 426 if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) { 427 ip6stat.ip6s_cantforward++; 428 ip6stat.ip6s_badscope++; 429 m_freem(m); 430 return; 431 } 432 if (inzone != outzone 433#ifdef IPSEC 434 && !ipsecrt 435#endif 436 ) { 437 ip6stat.ip6s_cantforward++; 438 ip6stat.ip6s_badscope++; 439 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard); 440 441 if (ip6_log_time + ip6_log_interval < time_second) { 442 ip6_log_time = time_second; 443 log(LOG_DEBUG, 444 "cannot forward " 445 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 446 ip6_sprintf(&ip6->ip6_src), 447 ip6_sprintf(&ip6->ip6_dst), 448 ip6->ip6_nxt, 449 if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp)); 450 } 451 if (mcopy) 452 icmp6_error(mcopy, ICMP6_DST_UNREACH, 453 ICMP6_DST_UNREACH_BEYONDSCOPE, 0); 454 m_freem(m); 455 return; 456 } 457 458 /* 459 * Destination scope check: if a packet is going to break the scope 460 * zone of packet's destination address, discard it. This case should 461 * usually be prevented by appropriately-configured routing table, but 462 * we need an explicit check because we may mistakenly forward the 463 * packet to a different zone by (e.g.) a default route. 464 */ 465 dst_in6 = ip6->ip6_dst; 466 if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 || 467 in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 || 468 inzone != outzone) { 469 ip6stat.ip6s_cantforward++; 470 ip6stat.ip6s_badscope++; 471 m_freem(m); 472 return; 473 } 474 475 if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) { 476 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig); 477 if (mcopy) { 478 u_long mtu; 479#ifdef IPSEC 480 struct secpolicy *sp; 481 int ipsecerror; 482 size_t ipsechdrsiz; 483#endif 484 485 mtu = IN6_LINKMTU(rt->rt_ifp); 486#ifdef IPSEC 487 /* 488 * When we do IPsec tunnel ingress, we need to play 489 * with the link value (decrement IPsec header size 490 * from mtu value). The code is much simpler than v4 491 * case, as we have the outgoing interface for 492 * encapsulated packet as "rt->rt_ifp". 493 */ 494 sp = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND, 495 IP_FORWARDING, &ipsecerror); 496 if (sp) { 497 ipsechdrsiz = ipsec6_hdrsiz(mcopy, 498 IPSEC_DIR_OUTBOUND, NULL); 499 if (ipsechdrsiz < mtu) 500 mtu -= ipsechdrsiz; 501 } 502 503 /* 504 * if mtu becomes less than minimum MTU, 505 * tell minimum MTU (and I'll need to fragment it). 506 */ 507 if (mtu < IPV6_MMTU) 508 mtu = IPV6_MMTU; 509#endif 510 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu); 511 } 512 m_freem(m); 513 return; 514 } 515 516 if (rt->rt_flags & RTF_GATEWAY) 517 dst = (struct sockaddr_in6 *)rt->rt_gateway; 518 519 /* 520 * If we are to forward the packet using the same interface 521 * as one we got the packet from, perhaps we should send a redirect 522 * to sender to shortcut a hop. 523 * Only send redirect if source is sending directly to us, 524 * and if packet was not source routed (or has any options). 525 * Also, don't send redirect if forwarding using a route 526 * modified by a redirect. 527 */ 528 if (ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && 529#ifdef IPSEC 530 !ipsecrt && 531#endif 532 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) { 533 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) { 534 /* 535 * If the incoming interface is equal to the outgoing 536 * one, and the link attached to the interface is 537 * point-to-point, then it will be highly probable 538 * that a routing loop occurs. Thus, we immediately 539 * drop the packet and send an ICMPv6 error message. 540 * 541 * type/code is based on suggestion by Rich Draves. 542 * not sure if it is the best pick. 543 */ 544 icmp6_error(mcopy, ICMP6_DST_UNREACH, 545 ICMP6_DST_UNREACH_ADDR, 0); 546 m_freem(m); 547 return; 548 } 549 type = ND_REDIRECT; 550 } 551 552 /* 553 * Fake scoped addresses. Note that even link-local source or 554 * destinaion can appear, if the originating node just sends the 555 * packet to us (without address resolution for the destination). 556 * Since both icmp6_error and icmp6_redirect_output fill the embedded 557 * link identifiers, we can do this stuff after making a copy for 558 * returning an error. 559 */ 560 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { 561 /* 562 * See corresponding comments in ip6_output. 563 * XXX: but is it possible that ip6_forward() sends a packet 564 * to a loopback interface? I don't think so, and thus 565 * I bark here. (jinmei@kame.net) 566 * XXX: it is common to route invalid packets to loopback. 567 * also, the codepath will be visited on use of ::1 in 568 * rthdr. (itojun) 569 */ 570#if 1 571 if (0) 572#else 573 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0) 574#endif 575 { 576 printf("ip6_forward: outgoing interface is loopback. " 577 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 578 ip6_sprintf(&ip6->ip6_src), 579 ip6_sprintf(&ip6->ip6_dst), 580 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif), 581 if_name(rt->rt_ifp)); 582 } 583 584 /* we can just use rcvif in forwarding. */ 585 origifp = m->m_pkthdr.rcvif; 586 } 587 else 588 origifp = rt->rt_ifp; 589 /* 590 * clear embedded scope identifiers if necessary. 591 * in6_clearscope will touch the addresses only when necessary. 592 */ 593 in6_clearscope(&ip6->ip6_src); 594 in6_clearscope(&ip6->ip6_dst); 595 596 /* Jump over all PFIL processing if hooks are not active. */ 597 if (!PFIL_HOOKED(&inet6_pfil_hook)) 598 goto pass; 599 600 /* Run through list of hooks for output packets. */ 601 error = pfil_run_hooks(&inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT, NULL); 602 if (error != 0) 603 goto senderr; 604 if (m == NULL) 605 goto freecopy; 606 ip6 = mtod(m, struct ip6_hdr *); 607 608pass: 609 error = nd6_output(rt->rt_ifp, origifp, m, dst, rt); 610 if (error) { 611 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard); 612 ip6stat.ip6s_cantforward++; 613 } else { 614 ip6stat.ip6s_forward++; 615 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward); 616 if (type) 617 ip6stat.ip6s_redirectsent++; 618 else { 619 if (mcopy) 620 goto freecopy; 621 } 622 } 623 624senderr: 625 if (mcopy == NULL) 626 return; 627 switch (error) { 628 case 0: 629 if (type == ND_REDIRECT) { 630 icmp6_redirect_output(mcopy, rt); 631 return; 632 } 633 goto freecopy; 634 635 case EMSGSIZE: 636 /* xxx MTU is constant in PPP? */ 637 goto freecopy; 638 639 case ENOBUFS: 640 /* Tell source to slow down like source quench in IP? */ 641 goto freecopy; 642 643 case ENETUNREACH: /* shouldn't happen, checked above */ 644 case EHOSTUNREACH: 645 case ENETDOWN: 646 case EHOSTDOWN: 647 default: 648 type = ICMP6_DST_UNREACH; 649 code = ICMP6_DST_UNREACH_ADDR; 650 break; 651 } 652 icmp6_error(mcopy, type, code, 0); 653 return; 654 655 freecopy: 656 m_freem(mcopy); 657 return; 658} 659