1 /* $NetBSD: ip6_forward.c,v 1.80 2016/06/28 02:02:56 ozaki-r 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 1.80 2016/06/28 02:02:56 ozaki-r Exp $"); 35 36 #ifdef _KERNEL_OPT 37 #include "opt_gateway.h" 38 #include "opt_ipsec.h" 39 #endif 40 41 #include <sys/param.h> 42 #include <sys/systm.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 52 #include <net/if.h> 53 #include <net/route.h> 54 #include <net/pfil.h> 55 56 #include <netinet/in.h> 57 #include <netinet/in_var.h> 58 #include <netinet/ip_var.h> 59 #include <netinet/ip6.h> 60 #include <netinet6/ip6_var.h> 61 #include <netinet6/ip6_private.h> 62 #include <netinet6/scope6_var.h> 63 #include <netinet/icmp6.h> 64 #include <netinet6/nd6.h> 65 66 #ifdef IPSEC 67 #include <netipsec/ipsec.h> 68 #include <netipsec/ipsec6.h> 69 #include <netipsec/key.h> 70 #include <netipsec/xform.h> 71 #endif /* IPSEC */ 72 73 #include <net/net_osdep.h> 74 75 struct route ip6_forward_rt; 76 77 extern pfil_head_t *inet6_pfil_hook; /* XXX */ 78 79 static void __printflike(4, 5) 80 ip6_cantforward(const struct ip6_hdr *ip6, const struct ifnet *srcifp, 81 const struct ifnet *dstifp, const char *fmt, ...) 82 { 83 char sbuf[INET6_ADDRSTRLEN], dbuf[INET6_ADDRSTRLEN]; 84 char reason[256]; 85 va_list ap; 86 uint64_t *ip6s; 87 88 /* update statistics */ 89 ip6s = IP6_STAT_GETREF(); 90 ip6s[IP6_STAT_CANTFORWARD]++; 91 if (dstifp) 92 ip6s[IP6_STAT_BADSCOPE]++; 93 IP6_STAT_PUTREF(); 94 95 if (dstifp) 96 in6_ifstat_inc(dstifp, ifs6_in_discard); 97 98 if (ip6_log_time + ip6_log_interval >= time_uptime) 99 return; 100 ip6_log_time = time_uptime; 101 102 va_start(ap, fmt); 103 vsnprintf(reason, sizeof(reason), fmt, ap); 104 va_end(ap); 105 106 log(LOG_DEBUG, "Cannot forward from %s@%s to %s@%s nxt %d (%s)\n", 107 IN6_PRINT(sbuf, &ip6->ip6_src), srcifp ? if_name(srcifp) : "?", 108 IN6_PRINT(dbuf, &ip6->ip6_dst), dstifp ? if_name(dstifp) : "?", 109 ip6->ip6_nxt, reason); 110 } 111 112 /* 113 * Forward a packet. If some error occurs return the sender 114 * an icmp packet. Note we can't always generate a meaningful 115 * icmp message because icmp doesn't have a large enough repertoire 116 * of codes and types. 117 * 118 * If not forwarding, just drop the packet. This could be confusing 119 * if ipforwarding was zero but some routing protocol was advancing 120 * us as a gateway to somewhere. However, we must let the routing 121 * protocol deal with that. 122 * 123 */ 124 125 void 126 ip6_forward(struct mbuf *m, int srcrt) 127 { 128 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 129 const struct sockaddr_in6 *dst; 130 struct rtentry *rt; 131 int error = 0, type = 0, code = 0; 132 struct mbuf *mcopy = NULL; 133 struct ifnet *origifp; /* maybe unnecessary */ 134 uint32_t inzone, outzone; 135 struct in6_addr src_in6, dst_in6; 136 struct ifnet *rcvif = NULL; 137 struct psref psref; 138 #ifdef IPSEC 139 int needipsec = 0; 140 struct secpolicy *sp = NULL; 141 #endif 142 143 /* 144 * Clear any in-bound checksum flags for this packet. 145 */ 146 m->m_pkthdr.csum_flags = 0; 147 148 rcvif = m_get_rcvif_psref(m, &psref); 149 if (__predict_false(rcvif == NULL)) 150 goto drop; 151 152 /* 153 * Do not forward packets to multicast destination (should be handled 154 * by ip6_mforward(). 155 * Do not forward packets with unspecified source. It was discussed 156 * in July 2000, on ipngwg mailing list. 157 */ 158 if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 || 159 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) || 160 IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) { 161 ip6_cantforward(ip6, rcvif, NULL, 162 ((m->m_flags & (M_BCAST|M_MCAST)) != 0) ? "bcast/mcast" : 163 IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ? "mcast/dst" : 164 "unspec/src"); 165 goto drop; 166 } 167 168 if (ip6->ip6_hlim <= IPV6_HLIMDEC) { 169 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */ 170 icmp6_error(m, ICMP6_TIME_EXCEEDED, 171 ICMP6_TIME_EXCEED_TRANSIT, 0); 172 goto out; 173 } 174 ip6->ip6_hlim -= IPV6_HLIMDEC; 175 176 /* 177 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU - 178 * size of IPv6 + ICMPv6 headers) bytes of the packet in case 179 * we need to generate an ICMP6 message to the src. 180 * Thanks to M_EXT, in most cases copy will not occur. 181 * 182 * It is important to save it before IPsec processing as IPsec 183 * processing may modify the mbuf. 184 */ 185 mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN)); 186 187 #ifdef IPSEC 188 if (ipsec_used) { 189 /* Check the security policy (SP) for the packet */ 190 191 sp = ipsec6_check_policy(m, NULL, 0, &needipsec, &error); 192 if (error != 0) { 193 /* 194 * Hack: -EINVAL is used to signal that a packet 195 * should be silently discarded. This is typically 196 * because we asked key management for an SA and 197 * it was delayed (e.g. kicked up to IKE). 198 */ 199 if (error == -EINVAL) 200 error = 0; 201 goto freecopy; 202 } 203 } 204 #endif /* IPSEC */ 205 206 if (srcrt) { 207 union { 208 struct sockaddr dst; 209 struct sockaddr_in6 dst6; 210 } u; 211 212 sockaddr_in6_init(&u.dst6, &ip6->ip6_dst, 0, 0, 0); 213 if ((rt = rtcache_lookup(&ip6_forward_rt, &u.dst)) == NULL) { 214 IP6_STATINC(IP6_STAT_NOROUTE); 215 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */ 216 if (mcopy) { 217 icmp6_error(mcopy, ICMP6_DST_UNREACH, 218 ICMP6_DST_UNREACH_NOROUTE, 0); 219 } 220 goto drop; 221 } 222 } else if ((rt = rtcache_validate(&ip6_forward_rt)) == NULL && 223 (rt = rtcache_update(&ip6_forward_rt, 1)) == NULL) { 224 /* 225 * rtcache_getdst(ip6_forward_rt)->sin6_addr was equal to 226 * ip6->ip6_dst 227 */ 228 IP6_STATINC(IP6_STAT_NOROUTE); 229 /* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */ 230 if (mcopy) { 231 icmp6_error(mcopy, ICMP6_DST_UNREACH, 232 ICMP6_DST_UNREACH_NOROUTE, 0); 233 } 234 goto drop; 235 } 236 dst = satocsin6(rtcache_getdst(&ip6_forward_rt)); 237 238 /* 239 * Source scope check: if a packet can't be delivered to its 240 * destination for the reason that the destination is beyond the scope 241 * of the source address, discard the packet and return an icmp6 242 * destination unreachable error with Code 2 (beyond scope of source 243 * address). We use a local copy of ip6_src, since in6_setscope() 244 * will possibly modify its first argument. 245 * [draft-ietf-ipngwg-icmp-v3-07, Section 3.1] 246 */ 247 src_in6 = ip6->ip6_src; 248 inzone = outzone = ~0; 249 if (in6_setscope(&src_in6, rt->rt_ifp, &outzone) != 0 || 250 in6_setscope(&src_in6, rcvif, &inzone) != 0 || 251 inzone != outzone) { 252 ip6_cantforward(ip6, rcvif, rt->rt_ifp, 253 "src[%s] inzone %d outzone %d", 254 in6_getscopename(&ip6->ip6_src), inzone, outzone); 255 if (mcopy) 256 icmp6_error(mcopy, ICMP6_DST_UNREACH, 257 ICMP6_DST_UNREACH_BEYONDSCOPE, 0); 258 goto drop; 259 } 260 261 #ifdef IPSEC 262 /* 263 * If we need to encapsulate the packet, do it here 264 * ipsec6_proces_packet will send the packet using ip6_output 265 */ 266 if (needipsec) { 267 int s = splsoftnet(); 268 error = ipsec6_process_packet(m, sp->req); 269 splx(s); 270 if (mcopy) 271 goto freecopy; 272 } 273 #endif 274 275 /* 276 * Destination scope check: if a packet is going to break the scope 277 * zone of packet's destination address, discard it. This case should 278 * usually be prevented by appropriately-configured routing table, but 279 * we need an explicit check because we may mistakenly forward the 280 * packet to a different zone by (e.g.) a default route. 281 */ 282 dst_in6 = ip6->ip6_dst; 283 inzone = outzone = ~0; 284 if (in6_setscope(&dst_in6, rcvif, &inzone) != 0 || 285 in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 || 286 inzone != outzone) { 287 ip6_cantforward(ip6, rcvif, rt->rt_ifp, 288 "dst[%s] inzone %d outzone %d", 289 in6_getscopename(&ip6->ip6_dst), inzone, outzone); 290 if (mcopy) 291 icmp6_error(mcopy, ICMP6_DST_UNREACH, 292 ICMP6_DST_UNREACH_BEYONDSCOPE, 0); 293 goto drop; 294 } 295 296 if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) { 297 in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig); 298 if (mcopy) { 299 u_long mtu; 300 301 mtu = IN6_LINKMTU(rt->rt_ifp); 302 icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu); 303 } 304 goto drop; 305 } 306 307 if (rt->rt_flags & RTF_GATEWAY) 308 dst = (struct sockaddr_in6 *)rt->rt_gateway; 309 310 /* 311 * If we are to forward the packet using the same interface 312 * as one we got the packet from, perhaps we should send a redirect 313 * to sender to shortcut a hop. 314 * Only send redirect if source is sending directly to us, 315 * and if packet was not source routed (or has any options). 316 * Also, don't send redirect if forwarding using a route 317 * modified by a redirect. 318 */ 319 if (rt->rt_ifp == rcvif && !srcrt && ip6_sendredirects && 320 (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) { 321 if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) && 322 nd6_is_addr_neighbor( 323 satocsin6(rtcache_getdst(&ip6_forward_rt)), 324 rt->rt_ifp)) { 325 /* 326 * If the incoming interface is equal to the outgoing 327 * one, the link attached to the interface is 328 * point-to-point, and the IPv6 destination is 329 * regarded as on-link on the link, then it will be 330 * highly probable that the destination address does 331 * not exist on the link and that the packet is going 332 * to loop. Thus, we immediately drop the packet and 333 * send an ICMPv6 error message. 334 * For other routing loops, we dare to let the packet 335 * go to the loop, so that a remote diagnosing host 336 * can detect the loop by traceroute. 337 * type/code is based on suggestion by Rich Draves. 338 * not sure if it is the best pick. 339 */ 340 icmp6_error(mcopy, ICMP6_DST_UNREACH, 341 ICMP6_DST_UNREACH_ADDR, 0); 342 goto drop; 343 } 344 type = ND_REDIRECT; 345 } 346 347 /* 348 * Fake scoped addresses. Note that even link-local source or 349 * destinaion can appear, if the originating node just sends the 350 * packet to us (without address resolution for the destination). 351 * Since both icmp6_error and icmp6_redirect_output fill the embedded 352 * link identifiers, we can do this stuff after making a copy for 353 * returning an error. 354 */ 355 if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) { 356 /* 357 * See corresponding comments in ip6_output. 358 * XXX: but is it possible that ip6_forward() sends a packet 359 * to a loopback interface? I don't think so, and thus 360 * I bark here. (jinmei@kame.net) 361 * XXX: it is common to route invalid packets to loopback. 362 * also, the codepath will be visited on use of ::1 in 363 * rthdr. (itojun) 364 */ 365 #if 1 366 if (0) 367 #else 368 if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0) 369 #endif 370 { 371 printf("ip6_forward: outgoing interface is loopback. " 372 "src %s, dst %s, nxt %d, rcvif %s, outif %s\n", 373 ip6_sprintf(&ip6->ip6_src), 374 ip6_sprintf(&ip6->ip6_dst), 375 ip6->ip6_nxt, if_name(rcvif), 376 if_name(rt->rt_ifp)); 377 } 378 379 /* we can just use rcvif in forwarding. */ 380 origifp = rcvif; 381 } 382 else 383 origifp = rt->rt_ifp; 384 /* 385 * clear embedded scope identifiers if necessary. 386 * in6_clearscope will touch the addresses only when necessary. 387 */ 388 in6_clearscope(&ip6->ip6_src); 389 in6_clearscope(&ip6->ip6_dst); 390 391 /* 392 * Run through list of hooks for output packets. 393 */ 394 if ((error = pfil_run_hooks(inet6_pfil_hook, &m, rt->rt_ifp, 395 PFIL_OUT)) != 0) 396 goto senderr; 397 if (m == NULL) 398 goto freecopy; 399 ip6 = mtod(m, struct ip6_hdr *); 400 401 error = nd6_output(rt->rt_ifp, origifp, m, dst, rt); 402 if (error) { 403 in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard); 404 IP6_STATINC(IP6_STAT_CANTFORWARD); 405 } else { 406 IP6_STATINC(IP6_STAT_FORWARD); 407 in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward); 408 if (type) 409 IP6_STATINC(IP6_STAT_REDIRECTSENT); 410 else { 411 #ifdef GATEWAY 412 if (m->m_flags & M_CANFASTFWD) 413 ip6flow_create(&ip6_forward_rt, m); 414 #endif 415 if (mcopy) 416 goto freecopy; 417 } 418 } 419 420 senderr: 421 if (mcopy == NULL) 422 goto out; 423 switch (error) { 424 case 0: 425 if (type == ND_REDIRECT) { 426 icmp6_redirect_output(mcopy, rt); 427 goto out; 428 } 429 goto freecopy; 430 431 case EMSGSIZE: 432 /* xxx MTU is constant in PPP? */ 433 goto freecopy; 434 435 case ENOBUFS: 436 /* Tell source to slow down like source quench in IP? */ 437 goto freecopy; 438 439 case ENETUNREACH: /* shouldn't happen, checked above */ 440 case EHOSTUNREACH: 441 case ENETDOWN: 442 case EHOSTDOWN: 443 default: 444 type = ICMP6_DST_UNREACH; 445 code = ICMP6_DST_UNREACH_ADDR; 446 break; 447 } 448 icmp6_error(mcopy, type, code, 0); 449 goto out; 450 451 freecopy: 452 m_freem(mcopy); 453 goto out; 454 drop: 455 m_freem(m); 456 out: 457 if (rcvif != NULL) 458 m_put_rcvif_psref(rcvif, &psref); 459 return; 460 } 461