xref: /openbsd-src/sys/netinet6/ip6_forward.c (revision 43003dfe3ad45d1698bed8a37f2b0f5b14f20d4f)
1 /*	$OpenBSD: ip6_forward.c,v 1.44 2009/10/06 21:21:48 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 u_int	ip6_forward_rtableid;
74 
75 /*
76  * Forward a packet.  If some error occurs return the sender
77  * an icmp packet.  Note we can't always generate a meaningful
78  * icmp message because icmp doesn't have a large enough repertoire
79  * of codes and types.
80  *
81  * If not forwarding, just drop the packet.  This could be confusing
82  * if ipforwarding was zero but some routing protocol was advancing
83  * us as a gateway to somewhere.  However, we must let the routing
84  * protocol deal with that.
85  *
86  */
87 
88 void
89 ip6_forward(struct mbuf *m, int srcrt)
90 {
91 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
92 	struct sockaddr_in6 *dst;
93 	struct rtentry *rt;
94 	int error = 0, type = 0, code = 0;
95 	struct mbuf *mcopy = NULL;
96 	struct ifnet *origifp;	/* maybe unnecessary */
97 #ifdef IPSEC
98 	u_int8_t sproto = 0;
99 	struct m_tag *mtag;
100 	union sockaddr_union sdst;
101 	struct tdb_ident *tdbi;
102 	u_int32_t sspi;
103 	struct tdb *tdb;
104 	int s;
105 #endif /* IPSEC */
106 	u_int rtableid = 0;
107 
108 	/*
109 	 * Do not forward packets to multicast destination (should be handled
110 	 * by ip6_mforward().
111 	 * Do not forward packets with unspecified source.  It was discussed
112 	 * in July 2000, on ipngwg mailing list.
113 	 */
114 	if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
115 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
116 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
117 		ip6stat.ip6s_cantforward++;
118 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
119 		if (ip6_log_time + ip6_log_interval < time_second) {
120 			ip6_log_time = time_second;
121 			log(LOG_DEBUG,
122 			    "cannot forward "
123 			    "from %s to %s nxt %d received on %s\n",
124 			    ip6_sprintf(&ip6->ip6_src),
125 			    ip6_sprintf(&ip6->ip6_dst),
126 			    ip6->ip6_nxt,
127 			    m->m_pkthdr.rcvif->if_xname);
128 		}
129 		m_freem(m);
130 		return;
131 	}
132 
133 	if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
134 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
135 		icmp6_error(m, ICMP6_TIME_EXCEEDED,
136 				ICMP6_TIME_EXCEED_TRANSIT, 0);
137 		return;
138 	}
139 	ip6->ip6_hlim -= IPV6_HLIMDEC;
140 
141 #if NPF > 0
142 reroute:
143 #endif
144 
145 #ifdef IPSEC
146 	if (!ipsec_in_use)
147 		goto done_spd;
148 
149 	s = splnet();
150 
151 	/*
152 	 * Check if there was an outgoing SA bound to the flow
153 	 * from a transport protocol.
154 	 */
155 
156 	/* Do we have any pending SAs to apply ? */
157 	mtag = m_tag_find(m, PACKET_TAG_IPSEC_PENDING_TDB, NULL);
158 	if (mtag != NULL) {
159 #ifdef DIAGNOSTIC
160 		if (mtag->m_tag_len != sizeof (struct tdb_ident))
161 			panic("ip6_forward: tag of length %d (should be %d",
162 			    mtag->m_tag_len, sizeof (struct tdb_ident));
163 #endif
164 		tdbi = (struct tdb_ident *)(mtag + 1);
165 		tdb = gettdb(tdbi->spi, &tdbi->dst, tdbi->proto);
166 		if (tdb == NULL)
167 			error = -EINVAL;
168 		m_tag_delete(m, mtag);
169 	} else
170 		tdb = ipsp_spd_lookup(m, AF_INET6, sizeof(struct ip6_hdr),
171 		    &error, IPSP_DIRECTION_OUT, NULL, NULL);
172 
173 	if (tdb == NULL) {
174 	        splx(s);
175 
176 		if (error == 0) {
177 		        /*
178 			 * No IPsec processing required, we'll just send the
179 			 * packet out.
180 			 */
181 		        sproto = 0;
182 
183 			/* Fall through to routing/multicast handling */
184 		} else {
185 		        /*
186 			 * -EINVAL is used to indicate that the packet should
187 			 * be silently dropped, typically because we've asked
188 			 * key management for an SA.
189 			 */
190 		        if (error == -EINVAL) /* Should silently drop packet */
191 				error = 0;
192 
193 			goto freecopy;
194 		}
195 	} else {
196 		/* Loop detection */
197 		for (mtag = m_tag_first(m); mtag != NULL;
198 		    mtag = m_tag_next(m, mtag)) {
199 			if (mtag->m_tag_id != PACKET_TAG_IPSEC_OUT_DONE &&
200 			    mtag->m_tag_id !=
201 			    PACKET_TAG_IPSEC_OUT_CRYPTO_NEEDED)
202 				continue;
203 			tdbi = (struct tdb_ident *)(mtag + 1);
204 			if (tdbi->spi == tdb->tdb_spi &&
205 			    tdbi->proto == tdb->tdb_sproto &&
206 			    !bcmp(&tdbi->dst, &tdb->tdb_dst,
207 			    sizeof(union sockaddr_union))) {
208 				splx(s);
209 				sproto = 0; /* mark as no-IPsec-needed */
210 				goto done_spd;
211 			}
212 		}
213 
214 	        /* We need to do IPsec */
215 	        bcopy(&tdb->tdb_dst, &sdst, sizeof(sdst));
216 		sspi = tdb->tdb_spi;
217 		sproto = tdb->tdb_sproto;
218 	        splx(s);
219 	}
220 
221 	/* Fall through to the routing/multicast handling code */
222  done_spd:
223 #endif /* IPSEC */
224 
225 #if NPF > 0
226 	rtableid = m->m_pkthdr.pf.rtableid;
227 #endif
228 
229 	/*
230 	 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
231 	 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
232 	 * we need to generate an ICMP6 message to the src.
233 	 * Thanks to M_EXT, in most cases copy will not occur.
234 	 *
235 	 * It is important to save it before IPsec processing as IPsec
236 	 * processing may modify the mbuf.
237 	 */
238 	mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
239 
240 	dst = &ip6_forward_rt.ro_dst;
241 	if (!srcrt) {
242 		/*
243 		 * ip6_forward_rt.ro_dst.sin6_addr is equal to ip6->ip6_dst
244 		 */
245 		if (ip6_forward_rt.ro_rt == 0 ||
246 		    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0 ||
247 		    ip6_forward_rtableid != rtableid) {
248 			if (ip6_forward_rt.ro_rt) {
249 				RTFREE(ip6_forward_rt.ro_rt);
250 				ip6_forward_rt.ro_rt = 0;
251 			}
252 			/* this probably fails but give it a try again */
253 			rtalloc_mpath((struct route *)&ip6_forward_rt,
254 			    &ip6->ip6_src.s6_addr32[0], rtableid);
255 			ip6_forward_rtableid = rtableid;
256 		}
257 
258 		if (ip6_forward_rt.ro_rt == 0) {
259 			ip6stat.ip6s_noroute++;
260 			/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
261 			if (mcopy) {
262 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
263 					    ICMP6_DST_UNREACH_NOROUTE, 0);
264 			}
265 			m_freem(m);
266 			return;
267 		}
268 	} else if (ip6_forward_rt.ro_rt == 0 ||
269 	   (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0 ||
270 	   !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr) ||
271 	   ip6_forward_rtableid != rtableid) {
272 		if (ip6_forward_rt.ro_rt) {
273 			RTFREE(ip6_forward_rt.ro_rt);
274 			ip6_forward_rt.ro_rt = 0;
275 		}
276 		bzero(dst, sizeof(*dst));
277 		dst->sin6_len = sizeof(struct sockaddr_in6);
278 		dst->sin6_family = AF_INET6;
279 		dst->sin6_addr = ip6->ip6_dst;
280 
281 		rtalloc_mpath((struct route *)&ip6_forward_rt,
282 		    &ip6->ip6_src.s6_addr32[0], rtableid);
283 		ip6_forward_rtableid = rtableid;
284 
285 		if (ip6_forward_rt.ro_rt == 0) {
286 			ip6stat.ip6s_noroute++;
287 			/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
288 			if (mcopy) {
289 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
290 					    ICMP6_DST_UNREACH_NOROUTE, 0);
291 			}
292 			m_freem(m);
293 			return;
294 		}
295 	}
296 	rt = ip6_forward_rt.ro_rt;
297 
298 	/*
299 	 * Scope check: if a packet can't be delivered to its destination
300 	 * for the reason that the destination is beyond the scope of the
301 	 * source address, discard the packet and return an icmp6 destination
302 	 * unreachable error with Code 2 (beyond scope of source address).
303 	 * [draft-ietf-ipngwg-icmp-v3-00.txt, Section 3.1]
304 	 */
305 	if (in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_src) !=
306 	    in6_addr2scopeid(rt->rt_ifp, &ip6->ip6_src)) {
307 		ip6stat.ip6s_cantforward++;
308 		ip6stat.ip6s_badscope++;
309 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
310 
311 		if (ip6_log_time + ip6_log_interval < time_second) {
312 			ip6_log_time = time_second;
313 			log(LOG_DEBUG,
314 			    "cannot forward "
315 			    "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
316 			    ip6_sprintf(&ip6->ip6_src),
317 			    ip6_sprintf(&ip6->ip6_dst),
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 		s = splnet();
337 
338 #if NPF > 0
339 		if (pf_test6(PF_OUT, &encif[0].sc_if, &m, NULL) != PF_PASS) {
340 			splx(s);
341 			error = EHOSTUNREACH;
342 			m_freem(m);
343 			goto senderr;
344 		}
345 		if (m == NULL) {
346 			splx(s);
347 			goto senderr;
348 		}
349 		ip6 = mtod(m, struct ip6_hdr *);
350 		/*
351 		 * PF_TAG_REROUTE handling or not...
352 		 * Packet is entering IPsec so the routing is
353 		 * already overruled by the IPsec policy.
354 		 * Until now the change was not reconsidered.
355 		 * What's the behaviour?
356 		 */
357 #endif
358 		tdb = gettdb(sspi, &sdst, sproto);
359 		if (tdb == NULL) {
360 			splx(s);
361 			error = EHOSTUNREACH;
362 			m_freem(m);
363 			goto senderr;	/*XXX*/
364 		}
365 
366 		m->m_flags &= ~(M_BCAST | M_MCAST);	/* just in case */
367 
368 		/* Callee frees mbuf */
369 		error = ipsp_process_packet(m, tdb, AF_INET6, 0);
370 		splx(s);
371 		m_freem(mcopy);
372 		goto freert;
373 	}
374 #endif /* IPSEC */
375 
376 	if (m->m_pkthdr.len > IN6_LINKMTU(rt->rt_ifp)) {
377 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
378 		if (mcopy) {
379 			u_long mtu;
380 
381 			mtu = IN6_LINKMTU(rt->rt_ifp);
382 
383 			icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
384 		}
385 		m_freem(m);
386 		goto freert;
387 	}
388 
389 	if (rt->rt_flags & RTF_GATEWAY)
390 		dst = (struct sockaddr_in6 *)rt->rt_gateway;
391 
392 	/*
393 	 * If we are to forward the packet using the same interface
394 	 * as one we got the packet from, perhaps we should send a redirect
395 	 * to sender to shortcut a hop.
396 	 * Only send redirect if source is sending directly to us,
397 	 * and if packet was not source routed (or has any options).
398 	 * Also, don't send redirect if forwarding using a route
399 	 * modified by a redirect.
400 	 */
401 	if (rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt && ip6_sendredirects &&
402 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
403 		if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) &&
404 		    nd6_is_addr_neighbor((struct sockaddr_in6 *)&ip6_forward_rt.ro_dst, rt->rt_ifp)) {
405 			/*
406 			 * If the incoming interface is equal to the outgoing
407 			 * one, the link attached to the interface is
408 			 * point-to-point, and the IPv6 destination is
409 			 * regarded as on-link on the link, then it will be
410 			 * highly probable that the destination address does
411 			 * not exist on the link and that the packet is going
412 			 * to loop.  Thus, we immediately drop the packet and
413 			 * send an ICMPv6 error message.
414 			 * For other routing loops, we dare to let the packet
415 			 * go to the loop, so that a remote diagnosing host
416 			 * can detect the loop by traceroute.
417 			 * type/code is based on suggestion by Rich Draves.
418 			 * not sure if it is the best pick.
419 			 */
420 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
421 				    ICMP6_DST_UNREACH_ADDR, 0);
422 			m_freem(m);
423 			goto freert;
424 		}
425 		type = ND_REDIRECT;
426 	}
427 
428 	/*
429 	 * Fake scoped addresses. Note that even link-local source or
430 	 * destinaion can appear, if the originating node just sends the
431 	 * packet to us (without address resolution for the destination).
432 	 * Since both icmp6_error and icmp6_redirect_output fill the embedded
433 	 * link identifiers, we can do this stuff after making a copy for
434 	 * returning an error.
435 	 */
436 	if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
437 		/*
438 		 * See corresponding comments in ip6_output.
439 		 * XXX: but is it possible that ip6_forward() sends a packet
440 		 *      to a loopback interface? I don't think so, and thus
441 		 *      I bark here. (jinmei@kame.net)
442 		 * XXX: it is common to route invalid packets to loopback.
443 		 *	also, the codepath will be visited on use of ::1 in
444 		 *	rthdr. (itojun)
445 		 */
446 #if 1
447 		if (0)
448 #else
449 		if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
450 #endif
451 		{
452 			printf("ip6_forward: outgoing interface is loopback. "
453 			       "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
454 			       ip6_sprintf(&ip6->ip6_src),
455 			       ip6_sprintf(&ip6->ip6_dst),
456 			       ip6->ip6_nxt, m->m_pkthdr.rcvif->if_xname,
457 			       rt->rt_ifp->if_xname);
458 		}
459 
460 		/* we can just use rcvif in forwarding. */
461 		origifp = m->m_pkthdr.rcvif;
462 	}
463 	else
464 		origifp = rt->rt_ifp;
465 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
466 		ip6->ip6_src.s6_addr16[1] = 0;
467 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
468 		ip6->ip6_dst.s6_addr16[1] = 0;
469 
470 #if NPF > 0
471 	if (pf_test6(PF_OUT, rt->rt_ifp, &m, NULL) != PF_PASS) {
472 		m_freem(m);
473 		goto senderr;
474 	}
475 	if (m == NULL)
476 		goto senderr;
477 
478 	ip6 = mtod(m, struct ip6_hdr *);
479 	if ((m->m_pkthdr.pf.flags & (PF_TAG_REROUTE | PF_TAG_GENERATED)) ==
480 	    (PF_TAG_REROUTE | PF_TAG_GENERATED)) {
481 		/* already rerun the route lookup, go on */
482 		m->m_pkthdr.pf.flags &= ~(PF_TAG_GENERATED | PF_TAG_REROUTE);
483 	} else if (m->m_pkthdr.pf.flags & PF_TAG_REROUTE) {
484 		/* tag as generated to skip over pf_test on rerun */
485 		m->m_pkthdr.pf.flags |= PF_TAG_GENERATED;
486 		srcrt = 1;
487 		goto reroute;
488 	}
489 #endif
490 
491 	error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
492 	if (error) {
493 		in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
494 		ip6stat.ip6s_cantforward++;
495 	} else {
496 		ip6stat.ip6s_forward++;
497 		in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
498 		if (type)
499 			ip6stat.ip6s_redirectsent++;
500 		else {
501 			if (mcopy)
502 				goto freecopy;
503 		}
504 	}
505 
506 #if NPF > 0 || defined(IPSEC)
507 senderr:
508 #endif
509 	if (mcopy == NULL)
510 		goto freert;
511 	switch (error) {
512 	case 0:
513 		if (type == ND_REDIRECT) {
514 			icmp6_redirect_output(mcopy, rt);
515 			goto freert;
516 		}
517 		goto freecopy;
518 
519 	case EMSGSIZE:
520 		/* xxx MTU is constant in PPP? */
521 		goto freecopy;
522 
523 	case ENOBUFS:
524 		/* Tell source to slow down like source quench in IP? */
525 		goto freecopy;
526 
527 	case ENETUNREACH:	/* shouldn't happen, checked above */
528 	case EHOSTUNREACH:
529 	case ENETDOWN:
530 	case EHOSTDOWN:
531 	default:
532 		type = ICMP6_DST_UNREACH;
533 		code = ICMP6_DST_UNREACH_ADDR;
534 		break;
535 	}
536 	icmp6_error(mcopy, type, code, 0);
537 	goto freert;
538 
539  freecopy:
540 	m_freem(mcopy);
541  freert:
542 #ifndef SMALL_KERNEL
543 	if (ip6_multipath && ip6_forward_rt.ro_rt &&
544 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_MPATH)) {
545 		RTFREE(ip6_forward_rt.ro_rt);
546 		ip6_forward_rt.ro_rt = 0;
547 	}
548 #endif
549 	return;
550 }
551