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("$FreeBSD$"); 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 #include <netinet6/vinet6.h> 69 70 #include <netinet/in_pcb.h> 71 72 #ifdef IPSEC 73 #include <netipsec/ipsec.h> 74 #include <netipsec/ipsec6.h> 75 #include <netipsec/key.h> 76 #endif /* IPSEC */ 77 78 #include <netinet6/ip6protosw.h> 79 80 /* 81 * Forward a packet. If some error occurs return the sender 82 * an icmp packet. Note we can't always generate a meaningful 83 * icmp message because icmp doesn't have a large enough repertoire 84 * of codes and types. 85 * 86 * If not forwarding, just drop the packet. This could be confusing 87 * if ipforwarding was zero but some routing protocol was advancing 88 * us as a gateway to somewhere. However, we must let the routing 89 * protocol deal with that. 90 * 91 */ 92 void 93 ip6_forward(struct mbuf *m, int srcrt) 94 { 95 INIT_VNET_INET6(curvnet); 96 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 97 struct sockaddr_in6 *dst = NULL; 98 struct rtentry *rt = NULL; 99 struct route_in6 rin6; 100 int error, type = 0, code = 0; 101 struct mbuf *mcopy = NULL; 102 struct ifnet *origifp; /* maybe unnecessary */ 103 u_int32_t inzone, outzone; 104 struct in6_addr src_in6, dst_in6; 105 #ifdef IPSEC 106 INIT_VNET_IPSEC(curvnet); 107 struct secpolicy *sp = NULL; 108 int ipsecrt = 0; 109 #endif 110 char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN]; 111 112 #ifdef IPSEC 113 /* 114 * Check AH/ESP integrity. 115 */ 116 /* 117 * Don't increment ip6s_cantforward because this is the check 118 * before forwarding packet actually. 119 */ 120 if (ipsec6_in_reject(m, NULL)) { 121 V_ipsec6stat.in_polvio++; 122 m_freem(m); 123 return; 124 } 125 #endif /* IPSEC */ 126 127 /* 128 * Do not forward packets to multicast destination (should be handled 129 * by ip6_mforward(). 130 * Do not forward packets with unspecified source. It was discussed 131 * in July 2000, on the ipngwg mailing list. 132 */ 133 if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 || 134 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 135 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 136 V_ip6stat.ip6s_cantforward++; 137 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ 138 if (V_ip6_log_time + V_ip6_log_interval < time_second) { 139 V_ip6_log_time = time_second; 140 log(LOG_DEBUG, 141 "cannot forward " 142 "from %s to %s nxt %d received on %s\n", 143 ip6_sprintf(ip6bufs, &ip6->ip6_src), 144 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 145 ip6->ip6_nxt, 146 if_name(m->m_pkthdr.rcvif)); 147 } 148 m_freem(m); 149 return; 150 } 151 152 #ifdef IPSTEALTH 153 if (!V_ip6stealth) { 154 #endif 155 if (ip6->ip6_hlim <= IPV6_HLIMDEC) { 156 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ 157 icmp6_error(m, ICMP6_TIME_EXCEEDED, 158 ICMP6_TIME_EXCEED_TRANSIT, 0); 159 return; 160 } 161 ip6->ip6_hlim -= IPV6_HLIMDEC; 162 163 #ifdef IPSTEALTH 164 } 165 #endif 166 167 /* 168 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU - 169 * size of IPv6 + ICMPv6 headers) bytes of the packet in case 170 * we need to generate an ICMP6 message to the src. 171 * Thanks to M_EXT, in most cases copy will not occur. 172 * 173 * It is important to save it before IPsec processing as IPsec 174 * processing may modify the mbuf. 175 */ 176 mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN)); 177 178 #ifdef IPSEC 179 /* get a security policy for this packet */ 180 sp = ipsec_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 181 IP_FORWARDING, &error); 182 if (sp == NULL) { 183 V_ipsec6stat.out_inval++; 184 V_ip6stat.ip6s_cantforward++; 185 if (mcopy) { 186 #if 0 187 /* XXX: what icmp ? */ 188 #else 189 m_freem(mcopy); 190 #endif 191 } 192 m_freem(m); 193 return; 194 } 195 196 error = 0; 197 198 /* check policy */ 199 switch (sp->policy) { 200 case IPSEC_POLICY_DISCARD: 201 /* 202 * This packet is just discarded. 203 */ 204 V_ipsec6stat.out_polvio++; 205 V_ip6stat.ip6s_cantforward++; 206 KEY_FREESP(&sp); 207 if (mcopy) { 208 #if 0 209 /* XXX: what icmp ? */ 210 #else 211 m_freem(mcopy); 212 #endif 213 } 214 m_freem(m); 215 return; 216 217 case IPSEC_POLICY_BYPASS: 218 case IPSEC_POLICY_NONE: 219 /* no need to do IPsec. */ 220 KEY_FREESP(&sp); 221 goto skip_ipsec; 222 223 case IPSEC_POLICY_IPSEC: 224 if (sp->req == NULL) { 225 /* XXX should be panic ? */ 226 printf("ip6_forward: No IPsec request specified.\n"); 227 V_ip6stat.ip6s_cantforward++; 228 KEY_FREESP(&sp); 229 if (mcopy) { 230 #if 0 231 /* XXX: what icmp ? */ 232 #else 233 m_freem(mcopy); 234 #endif 235 } 236 m_freem(m); 237 return; 238 } 239 /* do IPsec */ 240 break; 241 242 case IPSEC_POLICY_ENTRUST: 243 default: 244 /* should be panic ?? */ 245 printf("ip6_forward: Invalid policy found. %d\n", sp->policy); 246 KEY_FREESP(&sp); 247 goto skip_ipsec; 248 } 249 250 { 251 struct ipsecrequest *isr = NULL; 252 struct ipsec_output_state state; 253 254 /* 255 * when the kernel forwards a packet, it is not proper to apply 256 * IPsec transport mode to the packet is not proper. this check 257 * avoid from this. 258 * at present, if there is even a transport mode SA request in the 259 * security policy, the kernel does not apply IPsec to the packet. 260 * this check is not enough because the following case is valid. 261 * ipsec esp/tunnel/xxx-xxx/require esp/transport//require; 262 */ 263 for (isr = sp->req; isr; isr = isr->next) { 264 if (isr->saidx.mode == IPSEC_MODE_ANY) 265 goto doipsectunnel; 266 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) 267 goto doipsectunnel; 268 } 269 270 /* 271 * if there's no need for tunnel mode IPsec, skip. 272 */ 273 if (!isr) 274 goto skip_ipsec; 275 276 doipsectunnel: 277 /* 278 * All the extension headers will become inaccessible 279 * (since they can be encrypted). 280 * Don't panic, we need no more updates to extension headers 281 * on inner IPv6 packet (since they are now encapsulated). 282 * 283 * IPv6 [ESP|AH] IPv6 [extension headers] payload 284 */ 285 bzero(&state, sizeof(state)); 286 state.m = m; 287 state.ro = NULL; /* update at ipsec6_output_tunnel() */ 288 state.dst = NULL; /* update at ipsec6_output_tunnel() */ 289 290 error = ipsec6_output_tunnel(&state, sp, 0); 291 292 m = state.m; 293 KEY_FREESP(&sp); 294 295 if (error) { 296 /* mbuf is already reclaimed in ipsec6_output_tunnel. */ 297 switch (error) { 298 case EHOSTUNREACH: 299 case ENETUNREACH: 300 case EMSGSIZE: 301 case ENOBUFS: 302 case ENOMEM: 303 break; 304 default: 305 printf("ip6_output (ipsec): error code %d\n", error); 306 /* FALLTHROUGH */ 307 case ENOENT: 308 /* don't show these error codes to the user */ 309 break; 310 } 311 V_ip6stat.ip6s_cantforward++; 312 if (mcopy) { 313 #if 0 314 /* XXX: what icmp ? */ 315 #else 316 m_freem(mcopy); 317 #endif 318 } 319 m_freem(m); 320 return; 321 } else { 322 /* 323 * In the FAST IPSec case we have already 324 * re-injected the packet and it has been freed 325 * by the ipsec_done() function. So, just clean 326 * up after ourselves. 327 */ 328 m = NULL; 329 goto freecopy; 330 } 331 332 if ((m != NULL) && (ip6 != mtod(m, struct ip6_hdr *)) ){ 333 /* 334 * now tunnel mode headers are added. we are originating 335 * packet instead of forwarding the packet. 336 */ 337 ip6_output(m, NULL, NULL, IPV6_FORWARDING/*XXX*/, NULL, NULL, 338 NULL); 339 goto freecopy; 340 } 341 342 /* adjust pointer */ 343 dst = (struct sockaddr_in6 *)state.dst; 344 rt = state.ro ? state.ro->ro_rt : NULL; 345 if (dst != NULL && rt != NULL) 346 ipsecrt = 1; 347 } 348 if (ipsecrt) 349 goto skip_routing; 350 skip_ipsec: 351 #endif 352 353 bzero(&rin6, sizeof(struct route_in6)); 354 dst = (struct sockaddr_in6 *)&rin6.ro_dst; 355 dst->sin6_len = sizeof(struct sockaddr_in6); 356 dst->sin6_family = AF_INET6; 357 dst->sin6_addr = ip6->ip6_dst; 358 359 rin6.ro_rt = rtalloc1((struct sockaddr *)dst, 0, 0); 360 if (rin6.ro_rt != NULL) 361 RT_UNLOCK(rin6.ro_rt); 362 else { 363 V_ip6stat.ip6s_noroute++; 364 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute); 365 if (mcopy) { 366 icmp6_error(mcopy, ICMP6_DST_UNREACH, 367 ICMP6_DST_UNREACH_NOROUTE, 0); 368 } 369 goto bad; 370 } 371 rt = rin6.ro_rt; 372 #ifdef IPSEC 373 skip_routing: 374 #endif 375 376 /* 377 * Source scope check: if a packet can't be delivered to its 378 * destination for the reason that the destination is beyond the scope 379 * of the source address, discard the packet and return an icmp6 380 * destination unreachable error with Code 2 (beyond scope of source 381 * address). We use a local copy of ip6_src, since in6_setscope() 382 * will possibly modify its first argument. 383 * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1] 384 */ 385 src_in6 = ip6->ip6_src; 386 if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) { 387 /* XXX: this should not happen */ 388 V_ip6stat.ip6s_cantforward++; 389 V_ip6stat.ip6s_badscope++; 390 goto bad; 391 } 392 if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) { 393 V_ip6stat.ip6s_cantforward++; 394 V_ip6stat.ip6s_badscope++; 395 goto bad; 396 } 397 if (inzone != outzone 398 #ifdef IPSEC 399 && !ipsecrt 400 #endif 401 ) { 402 V_ip6stat.ip6s_cantforward++; 403 V_ip6stat.ip6s_badscope++; 404 in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard); 405 406 if (V_ip6_log_time + V_ip6_log_interval < time_second) { 407 V_ip6_log_time = time_second; 408 log(LOG_DEBUG, 409 "cannot forward " 410 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 411 ip6_sprintf(ip6bufs, &ip6->ip6_src), 412 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 413 ip6->ip6_nxt, 414 if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp)); 415 } 416 if (mcopy) 417 icmp6_error(mcopy, ICMP6_DST_UNREACH, 418 ICMP6_DST_UNREACH_BEYONDSCOPE, 0); 419 goto bad; 420 } 421 422 /* 423 * Destination scope check: if a packet is going to break the scope 424 * zone of packet's destination address, discard it. This case should 425 * usually be prevented by appropriately-configured routing table, but 426 * we need an explicit check because we may mistakenly forward the 427 * packet to a different zone by (e.g.) a default route. 428 */ 429 dst_in6 = ip6->ip6_dst; 430 if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 || 431 in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 || 432 inzone != outzone) { 433 V_ip6stat.ip6s_cantforward++; 434 V_ip6stat.ip6s_badscope++; 435 goto bad; 436 } 437 438 if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) { 439 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig); 440 if (mcopy) { 441 u_long mtu; 442 #ifdef IPSEC 443 struct secpolicy *sp; 444 int ipsecerror; 445 size_t ipsechdrsiz; 446 #endif /* IPSEC */ 447 448 mtu = IN6_LINKMTU(rt->rt_ifp); 449 #ifdef IPSEC 450 /* 451 * When we do IPsec tunnel ingress, we need to play 452 * with the link value (decrement IPsec header size 453 * from mtu value). The code is much simpler than v4 454 * case, as we have the outgoing interface for 455 * encapsulated packet as "rt->rt_ifp". 456 */ 457 sp = ipsec_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND, 458 IP_FORWARDING, &ipsecerror); 459 if (sp) { 460 ipsechdrsiz = ipsec_hdrsiz(mcopy, 461 IPSEC_DIR_OUTBOUND, NULL); 462 if (ipsechdrsiz < mtu) 463 mtu -= ipsechdrsiz; 464 } 465 466 /* 467 * if mtu becomes less than minimum MTU, 468 * tell minimum MTU (and I'll need to fragment it). 469 */ 470 if (mtu < IPV6_MMTU) 471 mtu = IPV6_MMTU; 472 #endif /* IPSEC */ 473 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu); 474 } 475 goto bad; 476 } 477 478 if (rt->rt_flags & RTF_GATEWAY) 479 dst = (struct sockaddr_in6 *)rt->rt_gateway; 480 481 /* 482 * If we are to forward the packet using the same interface 483 * as one we got the packet from, perhaps we should send a redirect 484 * to sender to shortcut a hop. 485 * Only send redirect if source is sending directly to us, 486 * and if packet was not source routed (or has any options). 487 * Also, don't send redirect if forwarding using a route 488 * modified by a redirect. 489 */ 490 if (V_ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && 491 #ifdef IPSEC 492 !ipsecrt && 493 #endif /* IPSEC */ 494 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) { 495 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) { 496 /* 497 * If the incoming interface is equal to the outgoing 498 * one, and the link attached to the interface is 499 * point-to-point, then it will be highly probable 500 * that a routing loop occurs. Thus, we immediately 501 * drop the packet and send an ICMPv6 error message. 502 * 503 * type/code is based on suggestion by Rich Draves. 504 * not sure if it is the best pick. 505 */ 506 icmp6_error(mcopy, ICMP6_DST_UNREACH, 507 ICMP6_DST_UNREACH_ADDR, 0); 508 goto bad; 509 } 510 type = ND_REDIRECT; 511 } 512 513 /* 514 * Fake scoped addresses. Note that even link-local source or 515 * destinaion can appear, if the originating node just sends the 516 * packet to us (without address resolution for the destination). 517 * Since both icmp6_error and icmp6_redirect_output fill the embedded 518 * link identifiers, we can do this stuff after making a copy for 519 * returning an error. 520 */ 521 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { 522 /* 523 * See corresponding comments in ip6_output. 524 * XXX: but is it possible that ip6_forward() sends a packet 525 * to a loopback interface? I don't think so, and thus 526 * I bark here. (jinmei@kame.net) 527 * XXX: it is common to route invalid packets to loopback. 528 * also, the codepath will be visited on use of ::1 in 529 * rthdr. (itojun) 530 */ 531 #if 1 532 if (0) 533 #else 534 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0) 535 #endif 536 { 537 printf("ip6_forward: outgoing interface is loopback. " 538 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 539 ip6_sprintf(ip6bufs, &ip6->ip6_src), 540 ip6_sprintf(ip6bufd, &ip6->ip6_dst), 541 ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif), 542 if_name(rt->rt_ifp)); 543 } 544 545 /* we can just use rcvif in forwarding. */ 546 origifp = m->m_pkthdr.rcvif; 547 } 548 else 549 origifp = rt->rt_ifp; 550 /* 551 * clear embedded scope identifiers if necessary. 552 * in6_clearscope will touch the addresses only when necessary. 553 */ 554 in6_clearscope(&ip6->ip6_src); 555 in6_clearscope(&ip6->ip6_dst); 556 557 /* Jump over all PFIL processing if hooks are not active. */ 558 if (!PFIL_HOOKED(&inet6_pfil_hook)) 559 goto pass; 560 561 /* Run through list of hooks for output packets. */ 562 error = pfil_run_hooks(&inet6_pfil_hook, &m, rt->rt_ifp, PFIL_OUT, NULL); 563 if (error != 0) 564 goto senderr; 565 if (m == NULL) 566 goto freecopy; 567 ip6 = mtod(m, struct ip6_hdr *); 568 569 pass: 570 error = nd6_output(rt->rt_ifp, origifp, m, dst, rt); 571 if (error) { 572 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard); 573 V_ip6stat.ip6s_cantforward++; 574 } else { 575 V_ip6stat.ip6s_forward++; 576 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward); 577 if (type) 578 V_ip6stat.ip6s_redirectsent++; 579 else { 580 if (mcopy) 581 goto freecopy; 582 } 583 } 584 585 senderr: 586 if (mcopy == NULL) 587 goto out; 588 switch (error) { 589 case 0: 590 if (type == ND_REDIRECT) { 591 icmp6_redirect_output(mcopy, rt); 592 goto out; 593 } 594 goto freecopy; 595 596 case EMSGSIZE: 597 /* xxx MTU is constant in PPP? */ 598 goto freecopy; 599 600 case ENOBUFS: 601 /* Tell source to slow down like source quench in IP? */ 602 goto freecopy; 603 604 case ENETUNREACH: /* shouldn't happen, checked above */ 605 case EHOSTUNREACH: 606 case ENETDOWN: 607 case EHOSTDOWN: 608 default: 609 type = ICMP6_DST_UNREACH; 610 code = ICMP6_DST_UNREACH_ADDR; 611 break; 612 } 613 icmp6_error(mcopy, type, code, 0); 614 goto out; 615 616 freecopy: 617 m_freem(mcopy); 618 goto out; 619 bad: 620 m_freem(m); 621 out: 622 if (rt != NULL 623 #ifdef IPSEC 624 && !ipsecrt 625 #endif 626 ) 627 RTFREE(rt); 628 } 629