xref: /openbsd/sys/netinet6/ip6_forward.c (revision 91f110e0)
1 /*	$OpenBSD: ip6_forward.c,v 1.64 2014/01/29 00:50:56 dlg Exp $	*/
2 /*	$KAME: ip6_forward.c,v 1.75 2001/06/29 12:42:13 jinmei 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 "pf.h"
34 
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/malloc.h>
38 #include <sys/mbuf.h>
39 #include <sys/domain.h>
40 #include <sys/protosw.h>
41 #include <sys/socket.h>
42 #include <sys/errno.h>
43 #include <sys/time.h>
44 #include <sys/kernel.h>
45 #include <sys/syslog.h>
46 
47 #include <net/if.h>
48 #include <net/if_enc.h>
49 #include <net/route.h>
50 
51 #include <netinet/in.h>
52 #include <netinet/ip_var.h>
53 #include <netinet6/in6_var.h>
54 #include <netinet/ip6.h>
55 #include <netinet6/ip6_var.h>
56 #include <netinet/icmp6.h>
57 #include <netinet6/nd6.h>
58 
59 #if NPF > 0
60 #include <net/pfvar.h>
61 #endif
62 
63 #ifdef IPSEC
64 #include <netinet/ip_ipsp.h>
65 #include <netinet/ip_ah.h>
66 #include <netinet/ip_esp.h>
67 #include <netinet/udp.h>
68 #include <netinet/tcp.h>
69 #include <net/pfkeyv2.h>
70 #endif
71 
72 struct	route_in6 ip6_forward_rt;
73 
74 /*
75  * Forward a packet.  If some error occurs return the sender
76  * an icmp packet.  Note we can't always generate a meaningful
77  * icmp message because icmp doesn't have a large enough repertoire
78  * of codes and types.
79  *
80  * If not forwarding, just drop the packet.  This could be confusing
81  * if ipforwarding was zero but some routing protocol was advancing
82  * us as a gateway to somewhere.  However, we must let the routing
83  * protocol deal with that.
84  *
85  */
86 
87 void
88 ip6_forward(struct mbuf *m, int srcrt)
89 {
90 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
91 	struct sockaddr_in6 *dst;
92 	struct rtentry *rt;
93 	int error = 0, type = 0, code = 0;
94 	struct mbuf *mcopy = NULL;
95 	struct ifnet *origifp;	/* maybe unnecessary */
96 #ifdef IPSEC
97 	u_int8_t sproto = 0;
98 	struct m_tag *mtag;
99 	union sockaddr_union sdst;
100 	struct tdb_ident *tdbi;
101 	u_int32_t sspi;
102 	struct tdb *tdb;
103 #if NPF > 0
104 	struct ifnet *encif;
105 #endif
106 #endif /* IPSEC */
107 	u_int rtableid = 0;
108 	char src6[INET6_ADDRSTRLEN], dst6[INET6_ADDRSTRLEN];
109 
110 	/*
111 	 * Do not forward packets to multicast destination (should be handled
112 	 * by ip6_mforward().
113 	 * Do not forward packets with unspecified source.  It was discussed
114 	 * in July 2000, on ipngwg mailing list.
115 	 */
116 	if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
117 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
118 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
119 		ip6stat.ip6s_cantforward++;
120 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
121 		if (ip6_log_time + ip6_log_interval < time_second) {
122 			ip6_log_time = time_second;
123 			inet_ntop(AF_INET6, &ip6->ip6_src, src6, sizeof(src6));
124 			inet_ntop(AF_INET6, &ip6->ip6_dst, dst6, sizeof(dst6));
125 			log(LOG_DEBUG,
126 			    "cannot forward "
127 			    "from %s to %s nxt %d received on %s\n",
128 			    src6, dst6,
129 			    ip6->ip6_nxt,
130 			    m->m_pkthdr.rcvif->if_xname);
131 		}
132 		m_freem(m);
133 		return;
134 	}
135 
136 	if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
137 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
138 		icmp6_error(m, ICMP6_TIME_EXCEEDED,
139 				ICMP6_TIME_EXCEED_TRANSIT, 0);
140 		return;
141 	}
142 	ip6->ip6_hlim -= IPV6_HLIMDEC;
143 
144 	/*
145 	 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
146 	 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
147 	 * we need to generate an ICMP6 message to the src.
148 	 * Thanks to M_EXT, in most cases copy will not occur.
149 	 *
150 	 * It is important to save it before IPsec processing as IPsec
151 	 * processing may modify the mbuf.
152 	 */
153 	mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
154 
155 #if NPF > 0
156 reroute:
157 #endif
158 
159 #ifdef IPSEC
160 	if (!ipsec_in_use)
161 		goto done_spd;
162 
163 	/*
164 	 * Check if there was an outgoing SA bound to the flow
165 	 * from a transport protocol.
166 	 */
167 
168 	/* Do we have any pending SAs to apply ? */
169 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
170 	if (mtag != NULL) {
171 #ifdef DIAGNOSTIC
172 		if (mtag->m_tag_len != sizeof (struct tdb_ident))
173 			panic("ip6_forward: tag of length %hu (should be %zu",
174 			    mtag->m_tag_len, sizeof (struct tdb_ident));
175 #endif
176 		tdbi = (struct tdb_ident *)(mtag + 1);
177 		tdb = gettdb(tdbi->rdomain, tdbi->spi, &tdbi->dst,
178 		    tdbi->proto);
179 		if (tdb == NULL)
180 			error = -EINVAL;
181 		m_tag_delete(m, mtag);
182 	} else
183 		tdb = ipsp_spd_lookup(m, AF_INET6, sizeof(struct ip6_hdr),
184 		    &error, IPSP_DIRECTION_OUT, NULL, NULL, 0);
185 
186 	if (tdb == NULL) {
187 		if (error == 0) {
188 		        /*
189 			 * No IPsec processing required, we'll just send the
190 			 * packet out.
191 			 */
192 		        sproto = 0;
193 
194 			/* Fall through to routing/multicast handling */
195 		} else {
196 		        /*
197 			 * -EINVAL is used to indicate that the packet should
198 			 * be silently dropped, typically because we've asked
199 			 * key management for an SA.
200 			 */
201 		        if (error == -EINVAL) /* Should silently drop packet */
202 				error = 0;
203 
204 			goto freecopy;
205 		}
206 	} else {
207 		/* Loop detection */
208 		for (mtag = m_tag_first(m); mtag != NULL;
209 		    mtag = m_tag_next(m, mtag)) {
210 			if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
211 			    mtag->m_tag_id !=
212 			    PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
213 				continue;
214 			tdbi = (struct tdb_ident *)(mtag + 1);
215 			if (tdbi->spi == tdb->tdb_spi &&
216 			    tdbi->proto == tdb->tdb_sproto &&
217 			    tdbi->rdomain == tdb->tdb_rdomain &&
218 			    !bcmp(&tdbi->dst, &tdb->tdb_dst,
219 			    sizeof(union sockaddr_union))) {
220 				sproto = 0; /* mark as no-IPsec-needed */
221 				goto done_spd;
222 			}
223 		}
224 
225 	        /* We need to do IPsec */
226 	        bcopy(&tdb->tdb_dst, &sdst, sizeof(sdst));
227 		sspi = tdb->tdb_spi;
228 		sproto = tdb->tdb_sproto;
229 	}
230 
231 	/* Fall through to the routing/multicast handling code */
232  done_spd:
233 #endif /* IPSEC */
234 
235 #if NPF > 0
236 	rtableid = m->m_pkthdr.rdomain;
237 #endif
238 
239 	dst = &ip6_forward_rt.ro_dst;
240 	if (!srcrt) {
241 		/*
242 		 * ip6_forward_rt.ro_dst.sin6_addr is equal to ip6->ip6_dst
243 		 */
244 		if (ip6_forward_rt.ro_rt == 0 ||
245 		    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0 ||
246 		    ip6_forward_rt.ro_tableid != rtableid) {
247 			if (ip6_forward_rt.ro_rt) {
248 				RTFREE(ip6_forward_rt.ro_rt);
249 				ip6_forward_rt.ro_rt = 0;
250 			}
251 			/* this probably fails but give it a try again */
252 			ip6_forward_rt.ro_tableid = rtableid;
253 			rtalloc_mpath((struct route *)&ip6_forward_rt,
254 			    &ip6->ip6_src.s6_addr32[0]);
255 		}
256 
257 		if (ip6_forward_rt.ro_rt == 0) {
258 			ip6stat.ip6s_noroute++;
259 			/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
260 			if (mcopy) {
261 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
262 					    ICMP6_DST_UNREACH_NOROUTE, 0);
263 			}
264 			m_freem(m);
265 			return;
266 		}
267 	} else if (ip6_forward_rt.ro_rt == 0 ||
268 	   (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0 ||
269 	   !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr) ||
270 	   ip6_forward_rt.ro_tableid != rtableid) {
271 		if (ip6_forward_rt.ro_rt) {
272 			RTFREE(ip6_forward_rt.ro_rt);
273 			ip6_forward_rt.ro_rt = 0;
274 		}
275 		bzero(dst, sizeof(*dst));
276 		dst->sin6_len = sizeof(struct sockaddr_in6);
277 		dst->sin6_family = AF_INET6;
278 		dst->sin6_addr = ip6->ip6_dst;
279 		ip6_forward_rt.ro_tableid = rtableid;
280 
281 		rtalloc_mpath((struct route *)&ip6_forward_rt,
282 		    &ip6->ip6_src.s6_addr32[0]);
283 
284 		if (ip6_forward_rt.ro_rt == 0) {
285 			ip6stat.ip6s_noroute++;
286 			/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
287 			if (mcopy) {
288 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
289 					    ICMP6_DST_UNREACH_NOROUTE, 0);
290 			}
291 			m_freem(m);
292 			return;
293 		}
294 	}
295 	rt = ip6_forward_rt.ro_rt;
296 
297 	/*
298 	 * Scope check: if a packet can't be delivered to its destination
299 	 * for the reason that the destination is beyond the scope of the
300 	 * source address, discard the packet and return an icmp6 destination
301 	 * unreachable error with Code 2 (beyond scope of source address).
302 	 * [draft-ietf-ipngwg-icmp-v3-00.txt, Section 3.1]
303 	 */
304 	if (in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_src) !=
305 	    in6_addr2scopeid(rt->rt_ifp, &ip6->ip6_src)) {
306 		ip6stat.ip6s_cantforward++;
307 		ip6stat.ip6s_badscope++;
308 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
309 
310 		if (ip6_log_time + ip6_log_interval < time_second) {
311 			ip6_log_time = time_second;
312 			inet_ntop(AF_INET6, &ip6->ip6_src, src6, sizeof(src6));
313 			inet_ntop(AF_INET6, &ip6->ip6_dst, dst6, sizeof(dst6));
314 			log(LOG_DEBUG,
315 			    "cannot forward "
316 			    "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
317 			    src6, dst6,
318 			    ip6->ip6_nxt,
319 			    m->m_pkthdr.rcvif->if_xname, rt->rt_ifp->if_xname);
320 		}
321 		if (mcopy)
322 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
323 				    ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
324 		m_freem(m);
325 		goto freert;
326 	}
327 
328 #ifdef IPSEC
329 	/*
330 	 * Check if the packet needs encapsulation.
331 	 * ipsp_process_packet will never come back to here.
332 	 * XXX ipsp_process_packet() calls ip6_output(), and there'll be no
333 	 * PMTU notification.  is it okay?
334 	 */
335 	if (sproto != 0) {
336 		tdb = gettdb(rtable_l2(m->m_pkthdr.rdomain),
337 		    sspi, &sdst, sproto);
338 		if (tdb == NULL) {
339 			error = EHOSTUNREACH;
340 			m_freem(m);
341 			goto senderr;	/*XXX*/
342 		}
343 
344 #if NPF > 0
345 		if ((encif = enc_getif(tdb->tdb_rdomain,
346 		    tdb->tdb_tap)) == NULL ||
347 		    pf_test(AF_INET6, PF_FWD, encif, &m, NULL) != PF_PASS) {
348 			error = EHOSTUNREACH;
349 			m_freem(m);
350 			goto senderr;
351 		}
352 		if (m == NULL)
353 			goto senderr;
354 		ip6 = mtod(m, struct ip6_hdr *);
355 		/*
356 		 * PF_TAG_REROUTE handling or not...
357 		 * Packet is entering IPsec so the routing is
358 		 * already overruled by the IPsec policy.
359 		 * Until now the change was not reconsidered.
360 		 * What's the behaviour?
361 		 */
362 #endif
363 		in6_proto_cksum_out(m, encif);
364 
365 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
366 
367 		/* Callee frees mbuf */
368 		error = ipsp_process_packet(m, tdb, AF_INET6, 0);
369 		m_freem(mcopy);
370 		goto freert;
371 	}
372 #endif /* IPSEC */
373 
374 	if (rt->rt_flags & RTF_GATEWAY)
375 		dst = satosin6(rt->rt_gateway);
376 
377 	/*
378 	 * If we are to forward the packet using the same interface
379 	 * as one we got the packet from, perhaps we should send a redirect
380 	 * to sender to shortcut a hop.
381 	 * Only send redirect if source is sending directly to us,
382 	 * and if packet was not source routed (or has any options).
383 	 * Also, don't send redirect if forwarding using a route
384 	 * modified by a redirect.
385 	 */
386 	if (rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && ip6_sendredirects &&
387 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
388 		if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) &&
389 		    nd6_is_addr_neighbor(&ip6_forward_rt.ro_dst, rt->rt_ifp)) {
390 			/*
391 			 * If the incoming interface is equal to the outgoing
392 			 * one, the link attached to the interface is
393 			 * point-to-point, and the IPv6 destination is
394 			 * regarded as on-link on the link, then it will be
395 			 * highly probable that the destination address does
396 			 * not exist on the link and that the packet is going
397 			 * to loop.  Thus, we immediately drop the packet and
398 			 * send an ICMPv6 error message.
399 			 * For other routing loops, we dare to let the packet
400 			 * go to the loop, so that a remote diagnosing host
401 			 * can detect the loop by traceroute.
402 			 * type/code is based on suggestion by Rich Draves.
403 			 * not sure if it is the best pick.
404 			 */
405 			if (mcopy)
406 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
407 				    ICMP6_DST_UNREACH_ADDR, 0);
408 			m_freem(m);
409 			goto freert;
410 		}
411 		type = ND_REDIRECT;
412 	}
413 
414 	/*
415 	 * Fake scoped addresses. Note that even link-local source or
416 	 * destinaion can appear, if the originating node just sends the
417 	 * packet to us (without address resolution for the destination).
418 	 * Since both icmp6_error and icmp6_redirect_output fill the embedded
419 	 * link identifiers, we can do this stuff after making a copy for
420 	 * returning an error.
421 	 */
422 	if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
423 		/*
424 		 * See corresponding comments in ip6_output.
425 		 * XXX: but is it possible that ip6_forward() sends a packet
426 		 *      to a loopback interface? I don't think so, and thus
427 		 *      I bark here. (jinmei@kame.net)
428 		 * XXX: it is common to route invalid packets to loopback.
429 		 *	also, the codepath will be visited on use of ::1 in
430 		 *	rthdr. (itojun)
431 		 */
432 #if 1
433 		if (0)
434 #else
435 		if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
436 #endif
437 		{
438 			inet_ntop(AF_INET6, &ip6->ip6_src, src6, sizeof(src6));
439 			inet_ntop(AF_INET6, &ip6->ip6_dst, dst6, sizeof(dst6));
440 			printf("ip6_forward: outgoing interface is loopback. "
441 			       "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
442 			       src6, dst6,
443 			       ip6->ip6_nxt, m->m_pkthdr.rcvif->if_xname,
444 			       rt->rt_ifp->if_xname);
445 		}
446 
447 		/* we can just use rcvif in forwarding. */
448 		origifp = m->m_pkthdr.rcvif;
449 	}
450 	else
451 		origifp = rt->rt_ifp;
452 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
453 		ip6->ip6_src.s6_addr16[1] = 0;
454 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
455 		ip6->ip6_dst.s6_addr16[1] = 0;
456 
457 #if NPF > 0
458 	if (pf_test(AF_INET6, PF_FWD, rt->rt_ifp, &m, NULL) != PF_PASS) {
459 		m_freem(m);
460 		goto senderr;
461 	}
462 	if (m == NULL)
463 		goto senderr;
464 	ip6 = mtod(m, struct ip6_hdr *);
465 	if ((m->m_pkthdr.pf.flags & (PF_TAG_REROUTE | PF_TAG_GENERATED)) ==
466 	    (PF_TAG_REROUTE | PF_TAG_GENERATED)) {
467 		/* already rerun the route lookup, go on */
468 		m->m_pkthdr.pf.flags &= ~(PF_TAG_GENERATED | PF_TAG_REROUTE);
469 	} else if (m->m_pkthdr.pf.flags & PF_TAG_REROUTE) {
470 		/* tag as generated to skip over pf_test on rerun */
471 		m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
472 		srcrt = 1;
473 		goto reroute;
474 	}
475 #endif
476 	in6_proto_cksum_out(m, rt->rt_ifp);
477 
478 	/* Check the size after pf_test to give pf a chance to refragment. */
479 	if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) {
480 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
481 		if (mcopy) {
482 			u_long mtu;
483 
484 			mtu = IN6_LINKMTU(rt->rt_ifp);
485 
486 			icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
487 		}
488 		m_freem(m);
489 		goto freert;
490 	}
491 
492 	error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
493 	if (error) {
494 		in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
495 		ip6stat.ip6s_cantforward++;
496 	} else {
497 		ip6stat.ip6s_forward++;
498 		in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
499 		if (type)
500 			ip6stat.ip6s_redirectsent++;
501 		else {
502 			if (mcopy)
503 				goto freecopy;
504 		}
505 	}
506 
507 #if NPF > 0 || defined(IPSEC)
508 senderr:
509 #endif
510 	if (mcopy == NULL)
511 		goto freert;
512 	switch (error) {
513 	case 0:
514 		if (type == ND_REDIRECT) {
515 			icmp6_redirect_output(mcopy, rt);
516 			goto freert;
517 		}
518 		goto freecopy;
519 
520 	case EMSGSIZE:
521 		/* xxx MTU is constant in PPP? */
522 		goto freecopy;
523 
524 	case ENOBUFS:
525 		/* Tell source to slow down like source quench in IP? */
526 		goto freecopy;
527 
528 	case ENETUNREACH:	/* shouldn't happen, checked above */
529 	case EHOSTUNREACH:
530 	case ENETDOWN:
531 	case EHOSTDOWN:
532 	default:
533 		type = ICMP6_DST_UNREACH;
534 		code = ICMP6_DST_UNREACH_ADDR;
535 		break;
536 	}
537 	icmp6_error(mcopy, type, code, 0);
538 	goto freert;
539 
540  freecopy:
541 	m_freem(mcopy);
542  freert:
543 #ifndef SMALL_KERNEL
544 	if (ip6_multipath && ip6_forward_rt.ro_rt &&
545 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_MPATH)) {
546 		RTFREE(ip6_forward_rt.ro_rt);
547 		ip6_forward_rt.ro_rt = 0;
548 	}
549 #endif
550 	return;
551 }
552