ip6_forward.c revision 139826
1/* $FreeBSD: head/sys/netinet6/ip6_forward.c 139826 2005-01-07 02:30:35Z imp $ */ 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_ip6fw.h" 34#include "opt_inet.h" 35#include "opt_inet6.h" 36#include "opt_ipsec.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 <netinet/icmp6.h> 63#include <netinet6/nd6.h> 64 65#include <netinet/in_pcb.h> 66 67#ifdef IPSEC 68#include <netinet6/ipsec.h> 69#ifdef INET6 70#include <netinet6/ipsec6.h> 71#endif 72#include <netkey/key.h> 73#endif /* IPSEC */ 74 75#ifdef FAST_IPSEC 76#include <netipsec/ipsec.h> 77#include <netipsec/ipsec6.h> 78#include <netipsec/key.h> 79#define IPSEC 80#endif /* FAST_IPSEC */ 81 82#include <netinet6/ip6_fw.h> 83 84#include <net/net_osdep.h> 85 86#include <netinet6/ip6protosw.h> 87 88struct route_in6 ip6_forward_rt; 89 90/* 91 * Forward a packet. If some error occurs return the sender 92 * an icmp packet. Note we can't always generate a meaningful 93 * icmp message because icmp doesn't have a large enough repertoire 94 * of codes and types. 95 * 96 * If not forwarding, just drop the packet. This could be confusing 97 * if ipforwarding was zero but some routing protocol was advancing 98 * us as a gateway to somewhere. However, we must let the routing 99 * protocol deal with that. 100 * 101 */ 102 103void 104ip6_forward(m, srcrt) 105 struct mbuf *m; 106 int srcrt; 107{ 108 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 109 struct sockaddr_in6 *dst = NULL; 110 struct rtentry *rt = NULL; 111 int error, type = 0, code = 0; 112 struct mbuf *mcopy = NULL; 113 struct ifnet *origifp; /* maybe unnecessary */ 114 u_int32_t srczone, dstzone; 115#ifdef IPSEC 116 struct secpolicy *sp = NULL; 117 int ipsecrt = 0; 118#endif 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 if (ip6->ip6_hlim <= IPV6_HLIMDEC) { 163 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ 164 icmp6_error(m, ICMP6_TIME_EXCEEDED, 165 ICMP6_TIME_EXCEED_TRANSIT, 0); 166 return; 167 } 168 ip6->ip6_hlim -= IPV6_HLIMDEC; 169 170 /* 171 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU - 172 * size of IPv6 + ICMPv6 headers) bytes of the packet in case 173 * we need to generate an ICMP6 message to the src. 174 * Thanks to M_EXT, in most cases copy will not occur. 175 * 176 * It is important to save it before IPsec processing as IPsec 177 * processing may modify the mbuf. 178 */ 179 mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN)); 180 181#ifdef IPSEC 182 /* get a security policy for this packet */ 183 sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 184 IP_FORWARDING, &error); 185 if (sp == NULL) { 186 ipsec6stat.out_inval++; 187 ip6stat.ip6s_cantforward++; 188 if (mcopy) { 189#if 0 190 /* XXX: what icmp ? */ 191#else 192 m_freem(mcopy); 193#endif 194 } 195 m_freem(m); 196 return; 197 } 198 199 error = 0; 200 201 /* check policy */ 202 switch (sp->policy) { 203 case IPSEC_POLICY_DISCARD: 204 /* 205 * This packet is just discarded. 206 */ 207 ipsec6stat.out_polvio++; 208 ip6stat.ip6s_cantforward++; 209 key_freesp(sp); 210 if (mcopy) { 211#if 0 212 /* XXX: what icmp ? */ 213#else 214 m_freem(mcopy); 215#endif 216 } 217 m_freem(m); 218 return; 219 220 case IPSEC_POLICY_BYPASS: 221 case IPSEC_POLICY_NONE: 222 /* no need to do IPsec. */ 223 key_freesp(sp); 224 goto skip_ipsec; 225 226 case IPSEC_POLICY_IPSEC: 227 if (sp->req == NULL) { 228 /* XXX should be panic ? */ 229 printf("ip6_forward: No IPsec request specified.\n"); 230 ip6stat.ip6s_cantforward++; 231 key_freesp(sp); 232 if (mcopy) { 233#if 0 234 /* XXX: what icmp ? */ 235#else 236 m_freem(mcopy); 237#endif 238 } 239 m_freem(m); 240 return; 241 } 242 /* do IPsec */ 243 break; 244 245 case IPSEC_POLICY_ENTRUST: 246 default: 247 /* should be panic ?? */ 248 printf("ip6_forward: Invalid policy found. %d\n", sp->policy); 249 key_freesp(sp); 250 goto skip_ipsec; 251 } 252 253 { 254 struct ipsecrequest *isr = NULL; 255 struct ipsec_output_state state; 256 257 /* 258 * when the kernel forwards a packet, it is not proper to apply 259 * IPsec transport mode to the packet is not proper. this check 260 * avoid from this. 261 * at present, if there is even a transport mode SA request in the 262 * security policy, the kernel does not apply IPsec to the packet. 263 * this check is not enough because the following case is valid. 264 * ipsec esp/tunnel/xxx-xxx/require esp/transport//require; 265 */ 266 for (isr = sp->req; isr; isr = isr->next) { 267 if (isr->saidx.mode == IPSEC_MODE_ANY) 268 goto doipsectunnel; 269 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) 270 goto doipsectunnel; 271 } 272 273 /* 274 * if there's no need for tunnel mode IPsec, skip. 275 */ 276 if (!isr) 277 goto skip_ipsec; 278 279 doipsectunnel: 280 /* 281 * All the extension headers will become inaccessible 282 * (since they can be encrypted). 283 * Don't panic, we need no more updates to extension headers 284 * on inner IPv6 packet (since they are now encapsulated). 285 * 286 * IPv6 [ESP|AH] IPv6 [extension headers] payload 287 */ 288 bzero(&state, sizeof(state)); 289 state.m = m; 290 state.ro = NULL; /* update at ipsec6_output_tunnel() */ 291 state.dst = NULL; /* update at ipsec6_output_tunnel() */ 292 293 error = ipsec6_output_tunnel(&state, sp, 0); 294 295 m = state.m; 296 key_freesp(sp); 297 298 if (error) { 299 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 300 switch (error) { 301 case EHOSTUNREACH: 302 case ENETUNREACH: 303 case EMSGSIZE: 304 case ENOBUFS: 305 case ENOMEM: 306 break; 307 default: 308 printf("ip6_output (ipsec): error code %d\n", error); 309 /* FALLTHROUGH */ 310 case ENOENT: 311 /* don't show these error codes to the user */ 312 break; 313 } 314 ip6stat.ip6s_cantforward++; 315 if (mcopy) { 316#if 0 317 /* XXX: what icmp ? */ 318#else 319 m_freem(mcopy); 320#endif 321 } 322 m_freem(m); 323 return; 324 } 325 326 if (ip6 != mtod(m, struct ip6_hdr *)) { 327 /* 328 * now tunnel mode headers are added. we are originating 329 * packet instead of forwarding the packet. 330 */ 331 ip6_output(m, NULL, NULL, IPV6_FORWARDING/*XXX*/, NULL, NULL, 332 NULL); 333 goto freecopy; 334 } 335 336 /* adjust pointer */ 337 dst = (struct sockaddr_in6 *)state.dst; 338 rt = state.ro ? state.ro->ro_rt : NULL; 339 if (dst != NULL && rt != NULL) 340 ipsecrt = 1; 341 } 342 skip_ipsec: 343#endif /* IPSEC */ 344 345#ifdef IPSEC 346 if (ipsecrt) 347 goto skip_routing; 348#endif 349 350 dst = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst; 351 if (!srcrt) { 352 /* ip6_forward_rt.ro_dst.sin6_addr is equal to ip6->ip6_dst */ 353 if (ip6_forward_rt.ro_rt == 0 || 354 (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0) { 355 if (ip6_forward_rt.ro_rt) { 356 RTFREE(ip6_forward_rt.ro_rt); 357 ip6_forward_rt.ro_rt = 0; 358 } 359 360 /* this probably fails but give it a try again */ 361 rtalloc((struct route *)&ip6_forward_rt); 362 } 363 364 if (ip6_forward_rt.ro_rt == 0) { 365 ip6stat.ip6s_noroute++; 366 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute); 367 if (mcopy) { 368 icmp6_error(mcopy, ICMP6_DST_UNREACH, 369 ICMP6_DST_UNREACH_NOROUTE, 0); 370 } 371 m_freem(m); 372 return; 373 } 374 } else if ((rt = ip6_forward_rt.ro_rt) == 0 || 375 !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) { 376 if (ip6_forward_rt.ro_rt) { 377 RTFREE(ip6_forward_rt.ro_rt); 378 ip6_forward_rt.ro_rt = 0; 379 } 380 bzero(dst, sizeof(*dst)); 381 dst->sin6_len = sizeof(struct sockaddr_in6); 382 dst->sin6_family = AF_INET6; 383 dst->sin6_addr = ip6->ip6_dst; 384 385 rtalloc((struct route *)&ip6_forward_rt); 386 if (ip6_forward_rt.ro_rt == 0) { 387 ip6stat.ip6s_noroute++; 388 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute); 389 if (mcopy) { 390 icmp6_error(mcopy, ICMP6_DST_UNREACH, 391 ICMP6_DST_UNREACH_NOROUTE, 0); 392 } 393 m_freem(m); 394 return; 395 } 396 } 397 rt = ip6_forward_rt.ro_rt; 398#ifdef IPSEC 399 skip_routing:; 400#endif 401 402 /* 403 * Scope check: if a packet can't be delivered to its destination 404 * for the reason that the destination is beyond the scope of the 405 * source address, discard the packet and return an icmp6 destination 406 * unreachable error with Code 2 (beyond scope of source address). 407 * [draft-ietf-ipngwg-icmp-v3-02.txt, Section 3.1] 408 */ 409 if (in6_addr2zoneid(m->m_pkthdr.rcvif, &ip6->ip6_src, &srczone) || 410 in6_addr2zoneid(rt->rt_ifp, &ip6->ip6_src, &dstzone)) { 411 /* XXX: this should not happen */ 412 ip6stat.ip6s_cantforward++; 413 ip6stat.ip6s_badscope++; 414 m_freem(m); 415 return; 416 } 417 if (srczone != dstzone 418#ifdef IPSEC 419 && !ipsecrt 420#endif 421 ) { 422 ip6stat.ip6s_cantforward++; 423 ip6stat.ip6s_badscope++; 424 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard); 425 426 if (ip6_log_time + ip6_log_interval < time_second) { 427 ip6_log_time = time_second; 428 log(LOG_DEBUG, 429 "cannot forward " 430 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 431 ip6_sprintf(&ip6->ip6_src), 432 ip6_sprintf(&ip6->ip6_dst), 433 ip6->ip6_nxt, 434 if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp)); 435 } 436 if (mcopy) 437 icmp6_error(mcopy, ICMP6_DST_UNREACH, 438 ICMP6_DST_UNREACH_BEYONDSCOPE, 0); 439 m_freem(m); 440 return; 441 } 442 443 if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) { 444 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig); 445 if (mcopy) { 446 u_long mtu; 447#ifdef IPSEC 448 struct secpolicy *sp; 449 int ipsecerror; 450 size_t ipsechdrsiz; 451#endif 452 453 mtu = IN6_LINKMTU(rt->rt_ifp); 454#ifdef IPSEC 455 /* 456 * When we do IPsec tunnel ingress, we need to play 457 * with the link value (decrement IPsec header size 458 * from mtu value). The code is much simpler than v4 459 * case, as we have the outgoing interface for 460 * encapsulated packet as "rt->rt_ifp". 461 */ 462 sp = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND, 463 IP_FORWARDING, &ipsecerror); 464 if (sp) { 465 ipsechdrsiz = ipsec6_hdrsiz(mcopy, 466 IPSEC_DIR_OUTBOUND, NULL); 467 if (ipsechdrsiz < mtu) 468 mtu -= ipsechdrsiz; 469 } 470 471 /* 472 * if mtu becomes less than minimum MTU, 473 * tell minimum MTU (and I'll need to fragment it). 474 */ 475 if (mtu < IPV6_MMTU) 476 mtu = IPV6_MMTU; 477#endif 478 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu); 479 } 480 m_freem(m); 481 return; 482 } 483 484 if (rt->rt_flags & RTF_GATEWAY) 485 dst = (struct sockaddr_in6 *)rt->rt_gateway; 486 487 /* 488 * If we are to forward the packet using the same interface 489 * as one we got the packet from, perhaps we should send a redirect 490 * to sender to shortcut a hop. 491 * Only send redirect if source is sending directly to us, 492 * and if packet was not source routed (or has any options). 493 * Also, don't send redirect if forwarding using a route 494 * modified by a redirect. 495 */ 496 if (rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && 497#ifdef IPSEC 498 !ipsecrt && 499#endif 500 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) { 501 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) { 502 /* 503 * If the incoming interface is equal to the outgoing 504 * one, and the link attached to the interface is 505 * point-to-point, then it will be highly probable 506 * that a routing loop occurs. Thus, we immediately 507 * drop the packet and send an ICMPv6 error message. 508 * 509 * type/code is based on suggestion by Rich Draves. 510 * not sure if it is the best pick. 511 */ 512 icmp6_error(mcopy, ICMP6_DST_UNREACH, 513 ICMP6_DST_UNREACH_ADDR, 0); 514 m_freem(m); 515 return; 516 } 517 type = ND_REDIRECT; 518 } 519 520 /* 521 * Check with the firewall... 522 */ 523 if (ip6_fw_enable && ip6_fw_chk_ptr) { 524 u_short port = 0; 525 /* If ipfw says divert, we have to just drop packet */ 526 if ((*ip6_fw_chk_ptr)(&ip6, rt->rt_ifp, &port, &m)) { 527 m_freem(m); 528 goto freecopy; 529 } 530 if (!m) 531 goto freecopy; 532 } 533 534 /* 535 * Fake scoped addresses. Note that even link-local source or 536 * destinaion can appear, if the originating node just sends the 537 * packet to us (without address resolution for the destination). 538 * Since both icmp6_error and icmp6_redirect_output fill the embedded 539 * link identifiers, we can do this stuff after making a copy for 540 * returning an error. 541 */ 542 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { 543 /* 544 * See corresponding comments in ip6_output. 545 * XXX: but is it possible that ip6_forward() sends a packet 546 * to a loopback interface? I don't think so, and thus 547 * I bark here. (jinmei@kame.net) 548 * XXX: it is common to route invalid packets to loopback. 549 * also, the codepath will be visited on use of ::1 in 550 * rthdr. (itojun) 551 */ 552#if 1 553 if (0) 554#else 555 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0) 556#endif 557 { 558 printf("ip6_forward: outgoing interface is loopback. " 559 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 560 ip6_sprintf(&ip6->ip6_src), 561 ip6_sprintf(&ip6->ip6_dst), 562 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif), 563 if_name(rt->rt_ifp)); 564 } 565 566 /* we can just use rcvif in forwarding. */ 567 origifp = m->m_pkthdr.rcvif; 568 } 569 else 570 origifp = rt->rt_ifp; 571 /* 572 * clear embedded scope identifiers if necessary. 573 * in6_clearscope will touch the addresses only when necessary. 574 */ 575 in6_clearscope(&ip6->ip6_src); 576 in6_clearscope(&ip6->ip6_dst); 577 578 /* Jump over all PFIL processing if hooks are not active. */ 579 if (inet6_pfil_hook.ph_busy_count == -1) 580 goto pass; 581 582 /* Run through list of hooks for output packets. */ 583 error = pfil_run_hooks(&inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT, NULL); 584 if (error != 0) 585 goto senderr; 586 if (m == NULL) 587 goto freecopy; 588 ip6 = mtod(m, struct ip6_hdr *); 589 590pass: 591 error = nd6_output(rt->rt_ifp, origifp, m, dst, rt); 592 if (error) { 593 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard); 594 ip6stat.ip6s_cantforward++; 595 } else { 596 ip6stat.ip6s_forward++; 597 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward); 598 if (type) 599 ip6stat.ip6s_redirectsent++; 600 else { 601 if (mcopy) 602 goto freecopy; 603 } 604 } 605 606senderr: 607 if (mcopy == NULL) 608 return; 609 switch (error) { 610 case 0: 611 if (type == ND_REDIRECT) { 612 icmp6_redirect_output(mcopy, rt); 613 return; 614 } 615 goto freecopy; 616 617 case EMSGSIZE: 618 /* xxx MTU is constant in PPP? */ 619 goto freecopy; 620 621 case ENOBUFS: 622 /* Tell source to slow down like source quench in IP? */ 623 goto freecopy; 624 625 case ENETUNREACH: /* shouldn't happen, checked above */ 626 case EHOSTUNREACH: 627 case ENETDOWN: 628 case EHOSTDOWN: 629 default: 630 type = ICMP6_DST_UNREACH; 631 code = ICMP6_DST_UNREACH_ADDR; 632 break; 633 } 634 icmp6_error(mcopy, type, code, 0); 635 return; 636 637 freecopy: 638 m_freem(mcopy); 639 return; 640} 641