xref: /freebsd/sys/netinet6/ip6_forward.c (revision eacee0ff7ec955b32e09515246bd97b6edcd2b0f)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ip6_forward.c,v 1.69 2001/05/17 03:48:30 itojun 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 "opt_ip6fw.h"
34 #include "opt_inet.h"
35 #include "opt_inet6.h"
36 #include "opt_ipsec.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/malloc.h>
41 #include <sys/mbuf.h>
42 #include <sys/domain.h>
43 #include <sys/protosw.h>
44 #include <sys/socket.h>
45 #include <sys/errno.h>
46 #include <sys/time.h>
47 #include <sys/kernel.h>
48 #include <sys/syslog.h>
49 
50 #include <net/if.h>
51 #include <net/route.h>
52 #ifdef PFIL_HOOKS
53 #include <net/pfil.h>
54 #endif
55 
56 #include <netinet/in.h>
57 #include <netinet/in_var.h>
58 #include <netinet/in_systm.h>
59 #include <netinet/ip.h>
60 #include <netinet/ip_var.h>
61 #include <netinet6/in6_var.h>
62 #include <netinet/ip6.h>
63 #include <netinet6/ip6_var.h>
64 #include <netinet/icmp6.h>
65 #include <netinet6/nd6.h>
66 
67 #include <netinet/in_pcb.h>
68 
69 #ifdef IPSEC
70 #include <netinet6/ipsec.h>
71 #ifdef INET6
72 #include <netinet6/ipsec6.h>
73 #endif
74 #include <netkey/key.h>
75 #endif /* IPSEC */
76 
77 #include <netinet6/ip6_fw.h>
78 
79 #include <net/net_osdep.h>
80 
81 #include <netinet6/ip6protosw.h>
82 
83 extern struct ip6protosw inet6sw[];
84 extern u_char ip6_protox[IPPROTO_MAX];
85 
86 struct	route_in6 ip6_forward_rt;
87 
88 /*
89  * Forward a packet.  If some error occurs return the sender
90  * an icmp packet.  Note we can't always generate a meaningful
91  * icmp message because icmp doesn't have a large enough repertoire
92  * of codes and types.
93  *
94  * If not forwarding, just drop the packet.  This could be confusing
95  * if ipforwarding was zero but some routing protocol was advancing
96  * us as a gateway to somewhere.  However, we must let the routing
97  * protocol deal with that.
98  *
99  */
100 
101 void
102 ip6_forward(m, srcrt)
103 	struct mbuf *m;
104 	int srcrt;
105 {
106 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
107 	struct sockaddr_in6 *dst;
108 	struct rtentry *rt;
109 	int error, type = 0, code = 0;
110 	struct mbuf *mcopy = NULL;
111 	struct ifnet *origifp;	/* maybe unnecessary */
112 #ifdef PFIL_HOOKS
113 	struct packet_filter_hook *pfh;
114 	struct mbuf *m1;
115 	int rv;
116 #endif /* PFIL_HOOKS */
117 #ifdef IPSEC
118 	struct secpolicy *sp = NULL;
119 #endif
120 
121 #ifdef IPSEC
122 	/*
123 	 * Check AH/ESP integrity.
124 	 */
125 	/*
126 	 * Don't increment ip6s_cantforward because this is the check
127 	 * before forwarding packet actually.
128 	 */
129 	if (ipsec6_in_reject(m, NULL)) {
130 		ipsec6stat.in_polvio++;
131 		m_freem(m);
132 		return;
133 	}
134 #endif /*IPSEC*/
135 
136 	/*
137 	 * Do not forward packets to multicast destination (should be handled
138 	 * by ip6_mforward().
139 	 * Do not forward packets with unspecified source.  It was discussed
140 	 * in July 2000, on ipngwg mailing list.
141 	 */
142 	if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
143 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
144 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
145 		ip6stat.ip6s_cantforward++;
146 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
147 		if (ip6_log_time + ip6_log_interval < time_second) {
148 			ip6_log_time = time_second;
149 			log(LOG_DEBUG,
150 			    "cannot forward "
151 			    "from %s to %s nxt %d received on %s\n",
152 			    ip6_sprintf(&ip6->ip6_src),
153 			    ip6_sprintf(&ip6->ip6_dst),
154 			    ip6->ip6_nxt,
155 			    if_name(m->m_pkthdr.rcvif));
156 		}
157 		m_freem(m);
158 		return;
159 	}
160 
161 	if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
162 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
163 		icmp6_error(m, ICMP6_TIME_EXCEEDED,
164 				ICMP6_TIME_EXCEED_TRANSIT, 0);
165 		return;
166 	}
167 	ip6->ip6_hlim -= IPV6_HLIMDEC;
168 
169 	/*
170 	 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
171 	 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
172 	 * we need to generate an ICMP6 message to the src.
173 	 * Thanks to M_EXT, in most cases copy will not occur.
174 	 *
175 	 * It is important to save it before IPsec processing as IPsec
176 	 * processing may modify the mbuf.
177 	 */
178 	mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
179 
180 #ifdef IPSEC
181 	/* get a security policy for this packet */
182 	sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, IP_FORWARDING,
183 	    &error);
184 	if (sp == NULL) {
185 		ipsec6stat.out_inval++;
186 		ip6stat.ip6s_cantforward++;
187 		if (mcopy) {
188 #if 0
189 			/* XXX: what icmp ? */
190 #else
191 			m_freem(mcopy);
192 #endif
193 		}
194 		m_freem(m);
195 		return;
196 	}
197 
198 	error = 0;
199 
200 	/* check policy */
201 	switch (sp->policy) {
202 	case IPSEC_POLICY_DISCARD:
203 		/*
204 		 * This packet is just discarded.
205 		 */
206 		ipsec6stat.out_polvio++;
207 		ip6stat.ip6s_cantforward++;
208 		key_freesp(sp);
209 		if (mcopy) {
210 #if 0
211 			/* XXX: what icmp ? */
212 #else
213 			m_freem(mcopy);
214 #endif
215 		}
216 		m_freem(m);
217 		return;
218 
219 	case IPSEC_POLICY_BYPASS:
220 	case IPSEC_POLICY_NONE:
221 		/* no need to do IPsec. */
222 		key_freesp(sp);
223 		goto skip_ipsec;
224 
225 	case IPSEC_POLICY_IPSEC:
226 		if (sp->req == NULL) {
227 			/* XXX should be panic ? */
228 			printf("ip6_forward: No IPsec request specified.\n");
229 			ip6stat.ip6s_cantforward++;
230 			key_freesp(sp);
231 			if (mcopy) {
232 #if 0
233 				/* XXX: what icmp ? */
234 #else
235 				m_freem(mcopy);
236 #endif
237 			}
238 			m_freem(m);
239 			return;
240 		}
241 		/* do IPsec */
242 		break;
243 
244 	case IPSEC_POLICY_ENTRUST:
245 	default:
246 		/* should be panic ?? */
247 		printf("ip6_forward: Invalid policy found. %d\n", sp->policy);
248 		key_freesp(sp);
249 		goto skip_ipsec;
250 	}
251 
252     {
253 	struct ipsec_output_state state;
254 
255 	/*
256 	 * All the extension headers will become inaccessible
257 	 * (since they can be encrypted).
258 	 * Don't panic, we need no more updates to extension headers
259 	 * on inner IPv6 packet (since they are now encapsulated).
260 	 *
261 	 * IPv6 [ESP|AH] IPv6 [extension headers] payload
262 	 */
263 	bzero(&state, sizeof(state));
264 	state.m = m;
265 	state.ro = NULL;	/* update at ipsec6_output_tunnel() */
266 	state.dst = NULL;	/* update at ipsec6_output_tunnel() */
267 
268 	error = ipsec6_output_tunnel(&state, sp, 0);
269 
270 	m = state.m;
271 	key_freesp(sp);
272 
273 	if (error) {
274 		/* mbuf is already reclaimed in ipsec6_output_tunnel. */
275 		switch (error) {
276 		case EHOSTUNREACH:
277 		case ENETUNREACH:
278 		case EMSGSIZE:
279 		case ENOBUFS:
280 		case ENOMEM:
281 			break;
282 		default:
283 			printf("ip6_output (ipsec): error code %d\n", error);
284 			/*fall through*/
285 		case ENOENT:
286 			/* don't show these error codes to the user */
287 			break;
288 		}
289 		ip6stat.ip6s_cantforward++;
290 		if (mcopy) {
291 #if 0
292 			/* XXX: what icmp ? */
293 #else
294 			m_freem(mcopy);
295 #endif
296 		}
297 		m_freem(m);
298 		return;
299 	}
300     }
301     skip_ipsec:
302 #endif /* IPSEC */
303 
304 	dst = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
305 	if (!srcrt) {
306 		/*
307 		 * ip6_forward_rt.ro_dst.sin6_addr is equal to ip6->ip6_dst
308 		 */
309 		if (ip6_forward_rt.ro_rt == 0 ||
310 		    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0) {
311 			if (ip6_forward_rt.ro_rt) {
312 				RTFREE(ip6_forward_rt.ro_rt);
313 				ip6_forward_rt.ro_rt = 0;
314 			}
315 			/* this probably fails but give it a try again */
316 			rtalloc_ign((struct route *)&ip6_forward_rt,
317 				    RTF_PRCLONING);
318 		}
319 
320 		if (ip6_forward_rt.ro_rt == 0) {
321 			ip6stat.ip6s_noroute++;
322 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
323 			if (mcopy) {
324 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
325 					    ICMP6_DST_UNREACH_NOROUTE, 0);
326 			}
327 			m_freem(m);
328 			return;
329 		}
330 	} else if ((rt = ip6_forward_rt.ro_rt) == 0 ||
331 		 !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) {
332 		if (ip6_forward_rt.ro_rt) {
333 			RTFREE(ip6_forward_rt.ro_rt);
334 			ip6_forward_rt.ro_rt = 0;
335 		}
336 		bzero(dst, sizeof(*dst));
337 		dst->sin6_len = sizeof(struct sockaddr_in6);
338 		dst->sin6_family = AF_INET6;
339 		dst->sin6_addr = ip6->ip6_dst;
340 
341   		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
342 		if (ip6_forward_rt.ro_rt == 0) {
343 			ip6stat.ip6s_noroute++;
344 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
345 			if (mcopy) {
346 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
347 					    ICMP6_DST_UNREACH_NOROUTE, 0);
348 			}
349 			m_freem(m);
350 			return;
351 		}
352 	}
353 	rt = ip6_forward_rt.ro_rt;
354 
355 	/*
356 	 * Scope check: if a packet can't be delivered to its destination
357 	 * for the reason that the destination is beyond the scope of the
358 	 * source address, discard the packet and return an icmp6 destination
359 	 * unreachable error with Code 2 (beyond scope of source address).
360 	 * [draft-ietf-ipngwg-icmp-v3-00.txt, Section 3.1]
361 	 */
362 	if (in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_src) !=
363 	    in6_addr2scopeid(rt->rt_ifp, &ip6->ip6_src)) {
364 		ip6stat.ip6s_cantforward++;
365 		ip6stat.ip6s_badscope++;
366 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
367 
368 		if (ip6_log_time + ip6_log_interval < time_second) {
369 			ip6_log_time = time_second;
370 			log(LOG_DEBUG,
371 			    "cannot forward "
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,
376 			    if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
377 		}
378 		if (mcopy)
379 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
380 				    ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
381 		m_freem(m);
382 		return;
383 	}
384 
385 	if (m->m_pkthdr.len > rt->rt_ifp->if_mtu) {
386 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
387 		if (mcopy) {
388 			u_long mtu;
389 #ifdef IPSEC
390 			struct secpolicy *sp;
391 			int ipsecerror;
392 			size_t ipsechdrsiz;
393 #endif
394 
395 			mtu = rt->rt_ifp->if_mtu;
396 #ifdef IPSEC
397 			/*
398 			 * When we do IPsec tunnel ingress, we need to play
399 			 * with if_mtu value (decrement IPsec header size
400 			 * from mtu value).  The code is much simpler than v4
401 			 * case, as we have the outgoing interface for
402 			 * encapsulated packet as "rt->rt_ifp".
403 			 */
404 			sp = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
405 				IP_FORWARDING, &ipsecerror);
406 			if (sp) {
407 				ipsechdrsiz = ipsec6_hdrsiz(mcopy,
408 					IPSEC_DIR_OUTBOUND, NULL);
409 				if (ipsechdrsiz < mtu)
410 					mtu -= ipsechdrsiz;
411 			}
412 
413 			/*
414 			 * if mtu becomes less than minimum MTU,
415 			 * tell minimum MTU (and I'll need to fragment it).
416 			 */
417 			if (mtu < IPV6_MMTU)
418 				mtu = IPV6_MMTU;
419 #endif
420 			icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
421 		}
422 		m_freem(m);
423 		return;
424  	}
425 
426 	if (rt->rt_flags & RTF_GATEWAY)
427 		dst = (struct sockaddr_in6 *)rt->rt_gateway;
428 
429 	/*
430 	 * If we are to forward the packet using the same interface
431 	 * as one we got the packet from, perhaps we should send a redirect
432 	 * to sender to shortcut a hop.
433 	 * Only send redirect if source is sending directly to us,
434 	 * and if packet was not source routed (or has any options).
435 	 * Also, don't send redirect if forwarding using a route
436 	 * modified by a redirect.
437 	 */
438 	if (rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
439 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
440 		if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) {
441 			/*
442 			 * If the incoming interface is equal to the outgoing
443 			 * one, and the link attached to the interface is
444 			 * point-to-point, then it will be highly probable
445 			 * that a routing loop occurs. Thus, we immediately
446 			 * drop the packet and send an ICMPv6 error message.
447 			 *
448 			 * type/code is based on suggestion by Rich Draves.
449 			 * not sure if it is the best pick.
450 			 */
451 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
452 				    ICMP6_DST_UNREACH_ADDR, 0);
453 			m_freem(m);
454 			return;
455 		}
456 		type = ND_REDIRECT;
457 	}
458 
459 	/*
460 	 * Check with the firewall...
461 	 */
462 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
463 		u_short port = 0;
464 		/* If ipfw says divert, we have to just drop packet */
465 		if ((*ip6_fw_chk_ptr)(&ip6, rt->rt_ifp, &port, &m)) {
466 			m_freem(m);
467 			goto freecopy;
468 		}
469 		if (!m)
470 			goto freecopy;
471 	}
472 
473 	/*
474 	 * Fake scoped addresses. Note that even link-local source or
475 	 * destinaion can appear, if the originating node just sends the
476 	 * packet to us (without address resolution for the destination).
477 	 * Since both icmp6_error and icmp6_redirect_output fill the embedded
478 	 * link identifiers, we can do this stuff after making a copy for
479 	 * returning an error.
480 	 */
481 	if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
482 		/*
483 		 * See corresponding comments in ip6_output.
484 		 * XXX: but is it possible that ip6_forward() sends a packet
485 		 *      to a loopback interface? I don't think so, and thus
486 		 *      I bark here. (jinmei@kame.net)
487 		 * XXX: it is common to route invalid packets to loopback.
488 		 *	also, the codepath will be visited on use of ::1 in
489 		 *	rthdr. (itojun)
490 		 */
491 #if 1
492 		if (0)
493 #else
494 		if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
495 #endif
496 		{
497 			printf("ip6_forward: outgoing interface is loopback. "
498 				"src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
499 				ip6_sprintf(&ip6->ip6_src),
500 				ip6_sprintf(&ip6->ip6_dst),
501 				ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
502 				if_name(rt->rt_ifp));
503 		}
504 
505 		/* we can just use rcvif in forwarding. */
506 		origifp = m->m_pkthdr.rcvif;
507 	}
508 	else
509 		origifp = rt->rt_ifp;
510 #ifndef SCOPEDROUTING
511 	/*
512 	 * clear embedded scope identifiers if necessary.
513 	 * in6_clearscope will touch the addresses only when necessary.
514 	 */
515 	in6_clearscope(&ip6->ip6_src);
516 	in6_clearscope(&ip6->ip6_dst);
517 #endif
518 
519 #ifdef PFIL_HOOKS
520 	/*
521 	 * Run through list of hooks for output packets.
522 	 */
523 	m1 = m;
524 	pfh = pfil_hook_get(PFIL_OUT, &inet6sw[ip6_protox[IPPROTO_IPV6]].pr_pfh);
525 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
526 		if (pfh->pfil_func) {
527 			rv = pfh->pfil_func(ip6, sizeof(*ip6),
528 					    rt->rt_ifp, 1, &m1);
529 			if (rv) {
530 				error = EHOSTUNREACH;
531 				goto freecopy;
532 			}
533 			m = m1;
534 			if (m == NULL)
535 				goto freecopy;
536 			ip6 = mtod(m, struct ip6_hdr *);
537 		}
538 #endif /* PFIL_HOOKS */
539 
540 	error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
541 	if (error) {
542 		in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
543 		ip6stat.ip6s_cantforward++;
544 	} else {
545 		ip6stat.ip6s_forward++;
546 		in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
547 		if (type)
548 			ip6stat.ip6s_redirectsent++;
549 		else {
550 			if (mcopy)
551 				goto freecopy;
552 		}
553 	}
554 	if (mcopy == NULL)
555 		return;
556 
557 	switch (error) {
558 	case 0:
559 #if 1
560 		if (type == ND_REDIRECT) {
561 			icmp6_redirect_output(mcopy, rt);
562 			return;
563 		}
564 #endif
565 		goto freecopy;
566 
567 	case EMSGSIZE:
568 		/* xxx MTU is constant in PPP? */
569 		goto freecopy;
570 
571 	case ENOBUFS:
572 		/* Tell source to slow down like source quench in IP? */
573 		goto freecopy;
574 
575 	case ENETUNREACH:	/* shouldn't happen, checked above */
576 	case EHOSTUNREACH:
577 	case ENETDOWN:
578 	case EHOSTDOWN:
579 	default:
580 		type = ICMP6_DST_UNREACH;
581 		code = ICMP6_DST_UNREACH_ADDR;
582 		break;
583 	}
584 	icmp6_error(mcopy, type, code, 0);
585 	return;
586 
587  freecopy:
588 	m_freem(mcopy);
589 	return;
590 }
591