xref: /openbsd-src/sys/netinet6/ip6_forward.c (revision d13be5d47e4149db2549a9828e244d59dbc43f15)
1 /*	$OpenBSD: ip6_forward.c,v 1.53 2011/07/04 06:54:49 claudio 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/in_var.h>
53 #include <netinet/ip_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 	int s;
104 #if NPF > 0
105 	struct ifnet *encif;
106 #endif
107 #endif /* IPSEC */
108 	u_int rtableid = 0;
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 			log(LOG_DEBUG,
124 			    "cannot forward "
125 			    "from %s to %s nxt %d received on %s\n",
126 			    ip6_sprintf(&ip6->ip6_src),
127 			    ip6_sprintf(&ip6->ip6_dst),
128 			    ip6->ip6_nxt,
129 			    m->m_pkthdr.rcvif->if_xname);
130 		}
131 		m_freem(m);
132 		return;
133 	}
134 
135 	if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
136 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
137 		icmp6_error(m, ICMP6_TIME_EXCEEDED,
138 				ICMP6_TIME_EXCEED_TRANSIT, 0);
139 		return;
140 	}
141 	ip6->ip6_hlim -= IPV6_HLIMDEC;
142 
143 #if NPF > 0
144 reroute:
145 #endif
146 
147 #ifdef IPSEC
148 	if (!ipsec_in_use)
149 		goto done_spd;
150 
151 	s = splnet();
152 
153 	/*
154 	 * Check if there was an outgoing SA bound to the flow
155 	 * from a transport protocol.
156 	 */
157 
158 	/* Do we have any pending SAs to apply ? */
159 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
160 	if (mtag != NULL) {
161 #ifdef DIAGNOSTIC
162 		if (mtag->m_tag_len != sizeof (struct tdb_ident))
163 			panic("ip6_forward: tag of length %d (should be %d",
164 			    mtag->m_tag_len, sizeof (struct tdb_ident));
165 #endif
166 		tdbi = (struct tdb_ident *)(mtag + 1);
167 		tdb = gettdb(tdbi->rdomain, tdbi->spi, &tdbi->dst,
168 		    tdbi->proto);
169 		if (tdb == NULL)
170 			error = -EINVAL;
171 		m_tag_delete(m, mtag);
172 	} else
173 		tdb = ipsp_spd_lookup(m, AF_INET6, sizeof(struct ip6_hdr),
174 		    &error, IPSP_DIRECTION_OUT, NULL, NULL);
175 
176 	if (tdb == NULL) {
177 	        splx(s);
178 
179 		if (error == 0) {
180 		        /*
181 			 * No IPsec processing required, we'll just send the
182 			 * packet out.
183 			 */
184 		        sproto = 0;
185 
186 			/* Fall through to routing/multicast handling */
187 		} else {
188 		        /*
189 			 * -EINVAL is used to indicate that the packet should
190 			 * be silently dropped, typically because we've asked
191 			 * key management for an SA.
192 			 */
193 		        if (error == -EINVAL) /* Should silently drop packet */
194 				error = 0;
195 
196 			goto freecopy;
197 		}
198 	} else {
199 		/* Loop detection */
200 		for (mtag = m_tag_first(m); mtag != NULL;
201 		    mtag = m_tag_next(m, mtag)) {
202 			if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
203 			    mtag->m_tag_id !=
204 			    PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
205 				continue;
206 			tdbi = (struct tdb_ident *)(mtag + 1);
207 			if (tdbi->spi == tdb->tdb_spi &&
208 			    tdbi->proto == tdb->tdb_sproto &&
209 			    tdbi->rdomain == tdb->tdb_rdomain &&
210 			    !bcmp(&tdbi->dst, &tdb->tdb_dst,
211 			    sizeof(union sockaddr_union))) {
212 				splx(s);
213 				sproto = 0; /* mark as no-IPsec-needed */
214 				goto done_spd;
215 			}
216 		}
217 
218 	        /* We need to do IPsec */
219 	        bcopy(&tdb->tdb_dst, &sdst, sizeof(sdst));
220 		sspi = tdb->tdb_spi;
221 		sproto = tdb->tdb_sproto;
222 	        splx(s);
223 	}
224 
225 	/* Fall through to the routing/multicast handling code */
226  done_spd:
227 #endif /* IPSEC */
228 
229 #if NPF > 0
230 	rtableid = m->m_pkthdr.rdomain;
231 #endif
232 
233 	/*
234 	 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
235 	 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
236 	 * we need to generate an ICMP6 message to the src.
237 	 * Thanks to M_EXT, in most cases copy will not occur.
238 	 *
239 	 * It is important to save it before IPsec processing as IPsec
240 	 * processing may modify the mbuf.
241 	 */
242 	mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
243 
244 	dst = &ip6_forward_rt.ro_dst;
245 	if (!srcrt) {
246 		/*
247 		 * ip6_forward_rt.ro_dst.sin6_addr is equal to ip6->ip6_dst
248 		 */
249 		if (ip6_forward_rt.ro_rt == 0 ||
250 		    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0 ||
251 		    ip6_forward_rt.ro_tableid != rtableid) {
252 			if (ip6_forward_rt.ro_rt) {
253 				RTFREE(ip6_forward_rt.ro_rt);
254 				ip6_forward_rt.ro_rt = 0;
255 			}
256 			/* this probably fails but give it a try again */
257 			ip6_forward_rt.ro_tableid = rtableid;
258 			rtalloc_mpath((struct route *)&ip6_forward_rt,
259 			    &ip6->ip6_src.s6_addr32[0]);
260 		}
261 
262 		if (ip6_forward_rt.ro_rt == 0) {
263 			ip6stat.ip6s_noroute++;
264 			/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
265 			if (mcopy) {
266 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
267 					    ICMP6_DST_UNREACH_NOROUTE, 0);
268 			}
269 			m_freem(m);
270 			return;
271 		}
272 	} else if (ip6_forward_rt.ro_rt == 0 ||
273 	   (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0 ||
274 	   !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr) ||
275 	   ip6_forward_rt.ro_tableid != rtableid) {
276 		if (ip6_forward_rt.ro_rt) {
277 			RTFREE(ip6_forward_rt.ro_rt);
278 			ip6_forward_rt.ro_rt = 0;
279 		}
280 		bzero(dst, sizeof(*dst));
281 		dst->sin6_len = sizeof(struct sockaddr_in6);
282 		dst->sin6_family = AF_INET6;
283 		dst->sin6_addr = ip6->ip6_dst;
284 		ip6_forward_rt.ro_tableid = rtableid;
285 
286 		rtalloc_mpath((struct route *)&ip6_forward_rt,
287 		    &ip6->ip6_src.s6_addr32[0]);
288 
289 		if (ip6_forward_rt.ro_rt == 0) {
290 			ip6stat.ip6s_noroute++;
291 			/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
292 			if (mcopy) {
293 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
294 					    ICMP6_DST_UNREACH_NOROUTE, 0);
295 			}
296 			m_freem(m);
297 			return;
298 		}
299 	}
300 	rt = ip6_forward_rt.ro_rt;
301 
302 	/*
303 	 * Scope check: if a packet can't be delivered to its destination
304 	 * for the reason that the destination is beyond the scope of the
305 	 * source address, discard the packet and return an icmp6 destination
306 	 * unreachable error with Code 2 (beyond scope of source address).
307 	 * [draft-ietf-ipngwg-icmp-v3-00.txt, Section 3.1]
308 	 */
309 	if (in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_src) !=
310 	    in6_addr2scopeid(rt->rt_ifp, &ip6->ip6_src)) {
311 		ip6stat.ip6s_cantforward++;
312 		ip6stat.ip6s_badscope++;
313 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
314 
315 		if (ip6_log_time + ip6_log_interval < time_second) {
316 			ip6_log_time = time_second;
317 			log(LOG_DEBUG,
318 			    "cannot forward "
319 			    "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
320 			    ip6_sprintf(&ip6->ip6_src),
321 			    ip6_sprintf(&ip6->ip6_dst),
322 			    ip6->ip6_nxt,
323 			    m->m_pkthdr.rcvif->if_xname, rt->rt_ifp->if_xname);
324 		}
325 		if (mcopy)
326 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
327 				    ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
328 		m_freem(m);
329 		goto freert;
330 	}
331 
332 #ifdef IPSEC
333 	/*
334 	 * Check if the packet needs encapsulation.
335 	 * ipsp_process_packet will never come back to here.
336 	 * XXX ipsp_process_packet() calls ip6_output(), and there'll be no
337 	 * PMTU notification.  is it okay?
338 	 */
339 	if (sproto != 0) {
340 		s = splnet();
341 
342 		tdb = gettdb(rtable_l2(m->m_pkthdr.rdomain),
343 		    sspi, &sdst, sproto);
344 		if (tdb == NULL) {
345 			splx(s);
346 			error = EHOSTUNREACH;
347 			m_freem(m);
348 			goto senderr;	/*XXX*/
349 		}
350 
351 #if NPF > 0
352 		if ((encif = enc_getif(tdb->tdb_rdomain,
353 		    tdb->tdb_tap)) == NULL ||
354 		    pf_test(AF_INET6, PF_FWD, encif, &m, NULL) != PF_PASS) {
355 			splx(s);
356 			error = EHOSTUNREACH;
357 			m_freem(m);
358 			goto senderr;
359 		}
360 		if (m == NULL) {
361 			splx(s);
362 			goto senderr;
363 		}
364 		ip6 = mtod(m, struct ip6_hdr *);
365 		/*
366 		 * PF_TAG_REROUTE handling or not...
367 		 * Packet is entering IPsec so the routing is
368 		 * already overruled by the IPsec policy.
369 		 * Until now the change was not reconsidered.
370 		 * What's the behaviour?
371 		 */
372 #endif
373 
374 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
375 
376 		/* Callee frees mbuf */
377 		error = ipsp_process_packet(m, tdb, AF_INET6, 0);
378 		splx(s);
379 		m_freem(mcopy);
380 		goto freert;
381 	}
382 #endif /* IPSEC */
383 
384 	if (rt->rt_flags & RTF_GATEWAY)
385 		dst = (struct sockaddr_in6 *)rt->rt_gateway;
386 
387 	/*
388 	 * If we are to forward the packet using the same interface
389 	 * as one we got the packet from, perhaps we should send a redirect
390 	 * to sender to shortcut a hop.
391 	 * Only send redirect if source is sending directly to us,
392 	 * and if packet was not source routed (or has any options).
393 	 * Also, don't send redirect if forwarding using a route
394 	 * modified by a redirect.
395 	 */
396 	if (rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && ip6_sendredirects &&
397 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
398 		if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) &&
399 		    nd6_is_addr_neighbor((struct sockaddr_in6 *)&ip6_forward_rt.ro_dst, rt->rt_ifp)) {
400 			/*
401 			 * If the incoming interface is equal to the outgoing
402 			 * one, the link attached to the interface is
403 			 * point-to-point, and the IPv6 destination is
404 			 * regarded as on-link on the link, then it will be
405 			 * highly probable that the destination address does
406 			 * not exist on the link and that the packet is going
407 			 * to loop.  Thus, we immediately drop the packet and
408 			 * send an ICMPv6 error message.
409 			 * For other routing loops, we dare to let the packet
410 			 * go to the loop, so that a remote diagnosing host
411 			 * can detect the loop by traceroute.
412 			 * type/code is based on suggestion by Rich Draves.
413 			 * not sure if it is the best pick.
414 			 */
415 			if (mcopy)
416 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
417 				    ICMP6_DST_UNREACH_ADDR, 0);
418 			m_freem(m);
419 			goto freert;
420 		}
421 		type = ND_REDIRECT;
422 	}
423 
424 	/*
425 	 * Fake scoped addresses. Note that even link-local source or
426 	 * destinaion can appear, if the originating node just sends the
427 	 * packet to us (without address resolution for the destination).
428 	 * Since both icmp6_error and icmp6_redirect_output fill the embedded
429 	 * link identifiers, we can do this stuff after making a copy for
430 	 * returning an error.
431 	 */
432 	if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
433 		/*
434 		 * See corresponding comments in ip6_output.
435 		 * XXX: but is it possible that ip6_forward() sends a packet
436 		 *      to a loopback interface? I don't think so, and thus
437 		 *      I bark here. (jinmei@kame.net)
438 		 * XXX: it is common to route invalid packets to loopback.
439 		 *	also, the codepath will be visited on use of ::1 in
440 		 *	rthdr. (itojun)
441 		 */
442 #if 1
443 		if (0)
444 #else
445 		if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
446 #endif
447 		{
448 			printf("ip6_forward: outgoing interface is loopback. "
449 			       "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
450 			       ip6_sprintf(&ip6->ip6_src),
451 			       ip6_sprintf(&ip6->ip6_dst),
452 			       ip6->ip6_nxt, m->m_pkthdr.rcvif->if_xname,
453 			       rt->rt_ifp->if_xname);
454 		}
455 
456 		/* we can just use rcvif in forwarding. */
457 		origifp = m->m_pkthdr.rcvif;
458 	}
459 	else
460 		origifp = rt->rt_ifp;
461 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
462 		ip6->ip6_src.s6_addr16[1] = 0;
463 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
464 		ip6->ip6_dst.s6_addr16[1] = 0;
465 
466 #if NPF > 0
467 	if (pf_test(AF_INET6, PF_FWD, rt->rt_ifp, &m, NULL) != PF_PASS) {
468 		m_freem(m);
469 		goto senderr;
470 	}
471 	if (m == NULL)
472 		goto senderr;
473 
474 	ip6 = mtod(m, struct ip6_hdr *);
475 	if ((m->m_pkthdr.pf.flags & (PF_TAG_REROUTE | PF_TAG_GENERATED)) ==
476 	    (PF_TAG_REROUTE | PF_TAG_GENERATED)) {
477 		/* already rerun the route lookup, go on */
478 		m->m_pkthdr.pf.flags &= ~(PF_TAG_GENERATED | PF_TAG_REROUTE);
479 	} else if (m->m_pkthdr.pf.flags & PF_TAG_REROUTE) {
480 		/* tag as generated to skip over pf_test on rerun */
481 		m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
482 		srcrt = 1;
483 		goto reroute;
484 	}
485 #endif
486 
487 	/* Check the size after pf_test to give pf a chance to refragment. */
488 	if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) {
489 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
490 		if (mcopy) {
491 			u_long mtu;
492 
493 			mtu = IN6_LINKMTU(rt->rt_ifp);
494 
495 			icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
496 		}
497 		m_freem(m);
498 		goto freert;
499 	}
500 
501 	error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
502 	if (error) {
503 		in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
504 		ip6stat.ip6s_cantforward++;
505 	} else {
506 		ip6stat.ip6s_forward++;
507 		in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
508 		if (type)
509 			ip6stat.ip6s_redirectsent++;
510 		else {
511 			if (mcopy)
512 				goto freecopy;
513 		}
514 	}
515 
516 #if NPF > 0 || defined(IPSEC)
517 senderr:
518 #endif
519 	if (mcopy == NULL)
520 		goto freert;
521 	switch (error) {
522 	case 0:
523 		if (type == ND_REDIRECT) {
524 			icmp6_redirect_output(mcopy, rt);
525 			goto freert;
526 		}
527 		goto freecopy;
528 
529 	case EMSGSIZE:
530 		/* xxx MTU is constant in PPP? */
531 		goto freecopy;
532 
533 	case ENOBUFS:
534 		/* Tell source to slow down like source quench in IP? */
535 		goto freecopy;
536 
537 	case ENETUNREACH:	/* shouldn't happen, checked above */
538 	case EHOSTUNREACH:
539 	case ENETDOWN:
540 	case EHOSTDOWN:
541 	default:
542 		type = ICMP6_DST_UNREACH;
543 		code = ICMP6_DST_UNREACH_ADDR;
544 		break;
545 	}
546 	icmp6_error(mcopy, type, code, 0);
547 	goto freert;
548 
549  freecopy:
550 	m_freem(mcopy);
551  freert:
552 #ifndef SMALL_KERNEL
553 	if (ip6_multipath && ip6_forward_rt.ro_rt &&
554 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_MPATH)) {
555 		RTFREE(ip6_forward_rt.ro_rt);
556 		ip6_forward_rt.ro_rt = 0;
557 	}
558 #endif
559 	return;
560 }
561