ip6_forward.c revision 195699
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: head/sys/netinet6/ip6_forward.c 195699 2009-07-14 22:48:30Z rwatson $"); 34 35#include "opt_inet.h" 36#include "opt_inet6.h" 37#include "opt_ipsec.h" 38#include "opt_ipstealth.h" 39 40#include <sys/param.h> 41#include <sys/systm.h> 42#include <sys/malloc.h> 43#include <sys/mbuf.h> 44#include <sys/domain.h> 45#include <sys/protosw.h> 46#include <sys/socket.h> 47#include <sys/errno.h> 48#include <sys/time.h> 49#include <sys/kernel.h> 50#include <sys/syslog.h> 51#include <sys/vimage.h> 52 53#include <net/if.h> 54#include <net/route.h> 55#include <net/pfil.h> 56 57#include <netinet/in.h> 58#include <netinet/in_var.h> 59#include <netinet/in_systm.h> 60#include <netinet/ip.h> 61#include <netinet/ip_var.h> 62#include <netinet6/in6_var.h> 63#include <netinet/ip6.h> 64#include <netinet6/ip6_var.h> 65#include <netinet6/scope6_var.h> 66#include <netinet/icmp6.h> 67#include <netinet6/nd6.h> 68 69#include <netinet/in_pcb.h> 70 71#ifdef IPSEC 72#include <netipsec/ipsec.h> 73#include <netipsec/ipsec6.h> 74#include <netipsec/key.h> 75#endif /* IPSEC */ 76 77#include <netinet6/ip6protosw.h> 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 */ 91void 92ip6_forward(struct mbuf *m, int srcrt) 93{ 94 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 95 struct sockaddr_in6 *dst = NULL; 96 struct rtentry *rt = NULL; 97 struct route_in6 rin6; 98 int error, type = 0, code = 0; 99 struct mbuf *mcopy = NULL; 100 struct ifnet *origifp; /* maybe unnecessary */ 101 u_int32_t inzone, outzone; 102 struct in6_addr src_in6, dst_in6; 103#ifdef IPSEC 104 struct secpolicy *sp = NULL; 105 int ipsecrt = 0; 106#endif 107 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 108 109#ifdef IPSEC 110 /* 111 * Check AH/ESP integrity. 112 */ 113 /* 114 * Don't increment ip6s_cantforward because this is the check 115 * before forwarding packet actually. 116 */ 117 if (ipsec6_in_reject(m, NULL)) { 118 V_ipsec6stat.in_polvio++; 119 m_freem(m); 120 return; 121 } 122#endif /* IPSEC */ 123 124 /* 125 * Do not forward packets to multicast destination (should be handled 126 * by ip6_mforward(). 127 * Do not forward packets with unspecified source. It was discussed 128 * in July 2000, on the ipngwg mailing list. 129 */ 130 if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 || 131 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 132 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 133 V_ip6stat.ip6s_cantforward++; 134 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ 135 if (V_ip6_log_time + V_ip6_log_interval < time_second) { 136 V_ip6_log_time = time_second; 137 log(LOG_DEBUG, 138 "cannot forward " 139 "from %s to %s nxt %d received on %s\n", 140 ip6_sprintf(ip6bufs, &ip6->ip6_src), 141 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 142 ip6->ip6_nxt, 143 if_name(m->m_pkthdr.rcvif)); 144 } 145 m_freem(m); 146 return; 147 } 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, 178 IP_FORWARDING, &error); 179 if (sp == NULL) { 180 V_ipsec6stat.out_inval++; 181 V_ip6stat.ip6s_cantforward++; 182 if (mcopy) { 183#if 0 184 /* XXX: what icmp ? */ 185#else 186 m_freem(mcopy); 187#endif 188 } 189 m_freem(m); 190 return; 191 } 192 193 error = 0; 194 195 /* check policy */ 196 switch (sp->policy) { 197 case IPSEC_POLICY_DISCARD: 198 /* 199 * This packet is just discarded. 200 */ 201 V_ipsec6stat.out_polvio++; 202 V_ip6stat.ip6s_cantforward++; 203 KEY_FREESP(&sp); 204 if (mcopy) { 205#if 0 206 /* XXX: what icmp ? */ 207#else 208 m_freem(mcopy); 209#endif 210 } 211 m_freem(m); 212 return; 213 214 case IPSEC_POLICY_BYPASS: 215 case IPSEC_POLICY_NONE: 216 /* no need to do IPsec. */ 217 KEY_FREESP(&sp); 218 goto skip_ipsec; 219 220 case IPSEC_POLICY_IPSEC: 221 if (sp->req == NULL) { 222 /* XXX should be panic ? */ 223 printf("ip6_forward: No IPsec request specified.\n"); 224 V_ip6stat.ip6s_cantforward++; 225 KEY_FREESP(&sp); 226 if (mcopy) { 227#if 0 228 /* XXX: what icmp ? */ 229#else 230 m_freem(mcopy); 231#endif 232 } 233 m_freem(m); 234 return; 235 } 236 /* do IPsec */ 237 break; 238 239 case IPSEC_POLICY_ENTRUST: 240 default: 241 /* should be panic ?? */ 242 printf("ip6_forward: Invalid policy found. %d\n", sp->policy); 243 KEY_FREESP(&sp); 244 goto skip_ipsec; 245 } 246 247 { 248 struct ipsecrequest *isr = NULL; 249 struct ipsec_output_state state; 250 251 /* 252 * when the kernel forwards a packet, it is not proper to apply 253 * IPsec transport mode to the packet is not proper. this check 254 * avoid from this. 255 * at present, if there is even a transport mode SA request in the 256 * security policy, the kernel does not apply IPsec to the packet. 257 * this check is not enough because the following case is valid. 258 * ipsec esp/tunnel/xxx-xxx/require esp/transport//require; 259 */ 260 for (isr = sp->req; isr; isr = isr->next) { 261 if (isr->saidx.mode == IPSEC_MODE_ANY) 262 goto doipsectunnel; 263 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) 264 goto doipsectunnel; 265 } 266 267 /* 268 * if there's no need for tunnel mode IPsec, skip. 269 */ 270 if (!isr) 271 goto skip_ipsec; 272 273 doipsectunnel: 274 /* 275 * All the extension headers will become inaccessible 276 * (since they can be encrypted). 277 * Don't panic, we need no more updates to extension headers 278 * on inner IPv6 packet (since they are now encapsulated). 279 * 280 * IPv6 [ESP|AH] IPv6 [extension headers] payload 281 */ 282 bzero(&state, sizeof(state)); 283 state.m = m; 284 state.ro = NULL; /* update at ipsec6_output_tunnel() */ 285 state.dst = NULL; /* update at ipsec6_output_tunnel() */ 286 287 error = ipsec6_output_tunnel(&state, sp, 0); 288 289 m = state.m; 290 KEY_FREESP(&sp); 291 292 if (error) { 293 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 294 switch (error) { 295 case EHOSTUNREACH: 296 case ENETUNREACH: 297 case EMSGSIZE: 298 case ENOBUFS: 299 case ENOMEM: 300 break; 301 default: 302 printf("ip6_output (ipsec): error code %d\n", error); 303 /* FALLTHROUGH */ 304 case ENOENT: 305 /* don't show these error codes to the user */ 306 break; 307 } 308 V_ip6stat.ip6s_cantforward++; 309 if (mcopy) { 310#if 0 311 /* XXX: what icmp ? */ 312#else 313 m_freem(mcopy); 314#endif 315 } 316 m_freem(m); 317 return; 318 } else { 319 /* 320 * In the FAST IPSec case we have already 321 * re-injected the packet and it has been freed 322 * by the ipsec_done() function. So, just clean 323 * up after ourselves. 324 */ 325 m = NULL; 326 goto freecopy; 327 } 328 329 if ((m != NULL) && (ip6 != mtod(m, struct ip6_hdr *)) ){ 330 /* 331 * now tunnel mode headers are added. we are originating 332 * packet instead of forwarding the packet. 333 */ 334 ip6_output(m, NULL, NULL, IPV6_FORWARDING/*XXX*/, NULL, NULL, 335 NULL); 336 goto freecopy; 337 } 338 339 /* adjust pointer */ 340 dst = (struct sockaddr_in6 *)state.dst; 341 rt = state.ro ? state.ro->ro_rt : NULL; 342 if (dst != NULL && rt != NULL) 343 ipsecrt = 1; 344 } 345 if (ipsecrt) 346 goto skip_routing; 347skip_ipsec: 348#endif 349 350 bzero(&rin6, sizeof(struct route_in6)); 351 dst = (struct sockaddr_in6 *)&rin6.ro_dst; 352 dst->sin6_len = sizeof(struct sockaddr_in6); 353 dst->sin6_family = AF_INET6; 354 dst->sin6_addr = ip6->ip6_dst; 355 356 rin6.ro_rt = rtalloc1((struct sockaddr *)dst, 0, 0); 357 if (rin6.ro_rt != NULL) 358 RT_UNLOCK(rin6.ro_rt); 359 else { 360 V_ip6stat.ip6s_noroute++; 361 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute); 362 if (mcopy) { 363 icmp6_error(mcopy, ICMP6_DST_UNREACH, 364 ICMP6_DST_UNREACH_NOROUTE, 0); 365 } 366 goto bad; 367 } 368 rt = rin6.ro_rt; 369#ifdef IPSEC 370skip_routing: 371#endif 372 373 /* 374 * Source scope check: if a packet can't be delivered to its 375 * destination for the reason that the destination is beyond the scope 376 * of the source address, discard the packet and return an icmp6 377 * destination unreachable error with Code 2 (beyond scope of source 378 * address). We use a local copy of ip6_src, since in6_setscope() 379 * will possibly modify its first argument. 380 * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1] 381 */ 382 src_in6 = ip6->ip6_src; 383 if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) { 384 /* XXX: this should not happen */ 385 V_ip6stat.ip6s_cantforward++; 386 V_ip6stat.ip6s_badscope++; 387 goto bad; 388 } 389 if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) { 390 V_ip6stat.ip6s_cantforward++; 391 V_ip6stat.ip6s_badscope++; 392 goto bad; 393 } 394 if (inzone != outzone 395#ifdef IPSEC 396 && !ipsecrt 397#endif 398 ) { 399 V_ip6stat.ip6s_cantforward++; 400 V_ip6stat.ip6s_badscope++; 401 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard); 402 403 if (V_ip6_log_time + V_ip6_log_interval < time_second) { 404 V_ip6_log_time = time_second; 405 log(LOG_DEBUG, 406 "cannot forward " 407 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 408 ip6_sprintf(ip6bufs, &ip6->ip6_src), 409 ip6_sprintf(ip6bufd, &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 goto bad; 417 } 418 419 /* 420 * Destination scope check: if a packet is going to break the scope 421 * zone of packet's destination address, discard it. This case should 422 * usually be prevented by appropriately-configured routing table, but 423 * we need an explicit check because we may mistakenly forward the 424 * packet to a different zone by (e.g.) a default route. 425 */ 426 dst_in6 = ip6->ip6_dst; 427 if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 || 428 in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 || 429 inzone != outzone) { 430 V_ip6stat.ip6s_cantforward++; 431 V_ip6stat.ip6s_badscope++; 432 goto bad; 433 } 434 435 if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) { 436 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig); 437 if (mcopy) { 438 u_long mtu; 439#ifdef IPSEC 440 struct secpolicy *sp; 441 int ipsecerror; 442 size_t ipsechdrsiz; 443#endif /* IPSEC */ 444 445 mtu = IN6_LINKMTU(rt->rt_ifp); 446#ifdef IPSEC 447 /* 448 * When we do IPsec tunnel ingress, we need to play 449 * with the link value (decrement IPsec header size 450 * from mtu value). The code is much simpler than v4 451 * case, as we have the outgoing interface for 452 * encapsulated packet as "rt->rt_ifp". 453 */ 454 sp = ipsec_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND, 455 IP_FORWARDING, &ipsecerror); 456 if (sp) { 457 ipsechdrsiz = ipsec_hdrsiz(mcopy, 458 IPSEC_DIR_OUTBOUND, NULL); 459 if (ipsechdrsiz < mtu) 460 mtu -= ipsechdrsiz; 461 } 462 463 /* 464 * if mtu becomes less than minimum MTU, 465 * tell minimum MTU (and I'll need to fragment it). 466 */ 467 if (mtu < IPV6_MMTU) 468 mtu = IPV6_MMTU; 469#endif /* IPSEC */ 470 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu); 471 } 472 goto bad; 473 } 474 475 if (rt->rt_flags & RTF_GATEWAY) 476 dst = (struct sockaddr_in6 *)rt->rt_gateway; 477 478 /* 479 * If we are to forward the packet using the same interface 480 * as one we got the packet from, perhaps we should send a redirect 481 * to sender to shortcut a hop. 482 * Only send redirect if source is sending directly to us, 483 * and if packet was not source routed (or has any options). 484 * Also, don't send redirect if forwarding using a route 485 * modified by a redirect. 486 */ 487 if (V_ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && 488#ifdef IPSEC 489 !ipsecrt && 490#endif /* IPSEC */ 491 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) { 492 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) { 493 /* 494 * If the incoming interface is equal to the outgoing 495 * one, and the link attached to the interface is 496 * point-to-point, then it will be highly probable 497 * that a routing loop occurs. Thus, we immediately 498 * drop the packet and send an ICMPv6 error message. 499 * 500 * type/code is based on suggestion by Rich Draves. 501 * not sure if it is the best pick. 502 */ 503 icmp6_error(mcopy, ICMP6_DST_UNREACH, 504 ICMP6_DST_UNREACH_ADDR, 0); 505 goto bad; 506 } 507 type = ND_REDIRECT; 508 } 509 510 /* 511 * Fake scoped addresses. Note that even link-local source or 512 * destinaion can appear, if the originating node just sends the 513 * packet to us (without address resolution for the destination). 514 * Since both icmp6_error and icmp6_redirect_output fill the embedded 515 * link identifiers, we can do this stuff after making a copy for 516 * returning an error. 517 */ 518 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { 519 /* 520 * See corresponding comments in ip6_output. 521 * XXX: but is it possible that ip6_forward() sends a packet 522 * to a loopback interface? I don't think so, and thus 523 * I bark here. (jinmei@kame.net) 524 * XXX: it is common to route invalid packets to loopback. 525 * also, the codepath will be visited on use of ::1 in 526 * rthdr. (itojun) 527 */ 528#if 1 529 if (0) 530#else 531 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0) 532#endif 533 { 534 printf("ip6_forward: outgoing interface is loopback. " 535 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 536 ip6_sprintf(ip6bufs, &ip6->ip6_src), 537 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 538 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif), 539 if_name(rt->rt_ifp)); 540 } 541 542 /* we can just use rcvif in forwarding. */ 543 origifp = m->m_pkthdr.rcvif; 544 } 545 else 546 origifp = rt->rt_ifp; 547 /* 548 * clear embedded scope identifiers if necessary. 549 * in6_clearscope will touch the addresses only when necessary. 550 */ 551 in6_clearscope(&ip6->ip6_src); 552 in6_clearscope(&ip6->ip6_dst); 553 554 /* Jump over all PFIL processing if hooks are not active. */ 555 if (!PFIL_HOOKED(&inet6_pfil_hook)) 556 goto pass; 557 558 /* Run through list of hooks for output packets. */ 559 error = pfil_run_hooks(&inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT, NULL); 560 if (error != 0) 561 goto senderr; 562 if (m == NULL) 563 goto freecopy; 564 ip6 = mtod(m, struct ip6_hdr *); 565 566pass: 567 error = nd6_output(rt->rt_ifp, origifp, m, dst, rt); 568 if (error) { 569 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard); 570 V_ip6stat.ip6s_cantforward++; 571 } else { 572 V_ip6stat.ip6s_forward++; 573 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward); 574 if (type) 575 V_ip6stat.ip6s_redirectsent++; 576 else { 577 if (mcopy) 578 goto freecopy; 579 } 580 } 581 582senderr: 583 if (mcopy == NULL) 584 goto out; 585 switch (error) { 586 case 0: 587 if (type == ND_REDIRECT) { 588 icmp6_redirect_output(mcopy, rt); 589 goto out; 590 } 591 goto freecopy; 592 593 case EMSGSIZE: 594 /* xxx MTU is constant in PPP? */ 595 goto freecopy; 596 597 case ENOBUFS: 598 /* Tell source to slow down like source quench in IP? */ 599 goto freecopy; 600 601 case ENETUNREACH: /* shouldn't happen, checked above */ 602 case EHOSTUNREACH: 603 case ENETDOWN: 604 case EHOSTDOWN: 605 default: 606 type = ICMP6_DST_UNREACH; 607 code = ICMP6_DST_UNREACH_ADDR; 608 break; 609 } 610 icmp6_error(mcopy, type, code, 0); 611 goto out; 612 613 freecopy: 614 m_freem(mcopy); 615 goto out; 616bad: 617 m_freem(m); 618out: 619 if (rt != NULL 620#ifdef IPSEC 621 && !ipsecrt 622#endif 623 ) 624 RTFREE(rt); 625} 626