xref: /freebsd/sys/netinet6/ip6_forward.c (revision b52b9d56d4e96089873a75f9e29062eec19fabba)
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 struct	route_in6 ip6_forward_rt;
84 
85 /*
86  * Forward a packet.  If some error occurs return the sender
87  * an icmp packet.  Note we can't always generate a meaningful
88  * icmp message because icmp doesn't have a large enough repertoire
89  * of codes and types.
90  *
91  * If not forwarding, just drop the packet.  This could be confusing
92  * if ipforwarding was zero but some routing protocol was advancing
93  * us as a gateway to somewhere.  However, we must let the routing
94  * protocol deal with that.
95  *
96  */
97 
98 void
99 ip6_forward(m, srcrt)
100 	struct mbuf *m;
101 	int srcrt;
102 {
103 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
104 	struct sockaddr_in6 *dst;
105 	struct rtentry *rt;
106 	int error, type = 0, code = 0;
107 	struct mbuf *mcopy = NULL;
108 	struct ifnet *origifp;	/* maybe unnecessary */
109 #ifdef PFIL_HOOKS
110 	struct packet_filter_hook *pfh;
111 	struct mbuf *m1;
112 	int rv;
113 #endif /* PFIL_HOOKS */
114 #ifdef IPSEC
115 	struct secpolicy *sp = NULL;
116 #endif
117 
118 #ifdef IPSEC
119 	/*
120 	 * Check AH/ESP integrity.
121 	 */
122 	/*
123 	 * Don't increment ip6s_cantforward because this is the check
124 	 * before forwarding packet actually.
125 	 */
126 	if (ipsec6_in_reject(m, NULL)) {
127 		ipsec6stat.in_polvio++;
128 		m_freem(m);
129 		return;
130 	}
131 #endif /* IPSEC */
132 
133 	/*
134 	 * Do not forward packets to multicast destination (should be handled
135 	 * by ip6_mforward().
136 	 * Do not forward packets with unspecified source.  It was discussed
137 	 * in July 2000, on ipngwg mailing list.
138 	 */
139 	if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
140 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
141 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
142 		ip6stat.ip6s_cantforward++;
143 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
144 		if (ip6_log_time + ip6_log_interval < time_second) {
145 			ip6_log_time = time_second;
146 			log(LOG_DEBUG,
147 			    "cannot forward "
148 			    "from %s to %s nxt %d received on %s\n",
149 			    ip6_sprintf(&ip6->ip6_src),
150 			    ip6_sprintf(&ip6->ip6_dst),
151 			    ip6->ip6_nxt,
152 			    if_name(m->m_pkthdr.rcvif));
153 		}
154 		m_freem(m);
155 		return;
156 	}
157 
158 	if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
159 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
160 		icmp6_error(m, ICMP6_TIME_EXCEEDED,
161 				ICMP6_TIME_EXCEED_TRANSIT, 0);
162 		return;
163 	}
164 	ip6->ip6_hlim -= IPV6_HLIMDEC;
165 
166 	/*
167 	 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
168 	 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
169 	 * we need to generate an ICMP6 message to the src.
170 	 * Thanks to M_EXT, in most cases copy will not occur.
171 	 *
172 	 * It is important to save it before IPsec processing as IPsec
173 	 * processing may modify the mbuf.
174 	 */
175 	mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
176 
177 #ifdef IPSEC
178 	/* get a security policy for this packet */
179 	sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, IP_FORWARDING,
180 	    &error);
181 	if (sp == NULL) {
182 		ipsec6stat.out_inval++;
183 		ip6stat.ip6s_cantforward++;
184 		if (mcopy) {
185 #if 0
186 			/* XXX: what icmp ? */
187 #else
188 			m_freem(mcopy);
189 #endif
190 		}
191 		m_freem(m);
192 		return;
193 	}
194 
195 	error = 0;
196 
197 	/* check policy */
198 	switch (sp->policy) {
199 	case IPSEC_POLICY_DISCARD:
200 		/*
201 		 * This packet is just discarded.
202 		 */
203 		ipsec6stat.out_polvio++;
204 		ip6stat.ip6s_cantforward++;
205 		key_freesp(sp);
206 		if (mcopy) {
207 #if 0
208 			/* XXX: what icmp ? */
209 #else
210 			m_freem(mcopy);
211 #endif
212 		}
213 		m_freem(m);
214 		return;
215 
216 	case IPSEC_POLICY_BYPASS:
217 	case IPSEC_POLICY_NONE:
218 		/* no need to do IPsec. */
219 		key_freesp(sp);
220 		goto skip_ipsec;
221 
222 	case IPSEC_POLICY_IPSEC:
223 		if (sp->req == NULL) {
224 			/* XXX should be panic ? */
225 			printf("ip6_forward: No IPsec request specified.\n");
226 			ip6stat.ip6s_cantforward++;
227 			key_freesp(sp);
228 			if (mcopy) {
229 #if 0
230 				/* XXX: what icmp ? */
231 #else
232 				m_freem(mcopy);
233 #endif
234 			}
235 			m_freem(m);
236 			return;
237 		}
238 		/* do IPsec */
239 		break;
240 
241 	case IPSEC_POLICY_ENTRUST:
242 	default:
243 		/* should be panic ?? */
244 		printf("ip6_forward: Invalid policy found. %d\n", sp->policy);
245 		key_freesp(sp);
246 		goto skip_ipsec;
247 	}
248 
249     {
250 	struct ipsec_output_state state;
251 
252 	/*
253 	 * All the extension headers will become inaccessible
254 	 * (since they can be encrypted).
255 	 * Don't panic, we need no more updates to extension headers
256 	 * on inner IPv6 packet (since they are now encapsulated).
257 	 *
258 	 * IPv6 [ESP|AH] IPv6 [extension headers] payload
259 	 */
260 	bzero(&state, sizeof(state));
261 	state.m = m;
262 	state.ro = NULL;	/* update at ipsec6_output_tunnel() */
263 	state.dst = NULL;	/* update at ipsec6_output_tunnel() */
264 
265 	error = ipsec6_output_tunnel(&state, sp, 0);
266 
267 	m = state.m;
268 	key_freesp(sp);
269 
270 	if (error) {
271 		/* mbuf is already reclaimed in ipsec6_output_tunnel. */
272 		switch (error) {
273 		case EHOSTUNREACH:
274 		case ENETUNREACH:
275 		case EMSGSIZE:
276 		case ENOBUFS:
277 		case ENOMEM:
278 			break;
279 		default:
280 			printf("ip6_output (ipsec): error code %d\n", error);
281 			/* fall through */
282 		case ENOENT:
283 			/* don't show these error codes to the user */
284 			break;
285 		}
286 		ip6stat.ip6s_cantforward++;
287 		if (mcopy) {
288 #if 0
289 			/* XXX: what icmp ? */
290 #else
291 			m_freem(mcopy);
292 #endif
293 		}
294 		m_freem(m);
295 		return;
296 	}
297     }
298     skip_ipsec:
299 #endif /* IPSEC */
300 
301 	dst = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
302 	if (!srcrt) {
303 		/*
304 		 * ip6_forward_rt.ro_dst.sin6_addr is equal to ip6->ip6_dst
305 		 */
306 		if (ip6_forward_rt.ro_rt == 0 ||
307 		    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0) {
308 			if (ip6_forward_rt.ro_rt) {
309 				RTFREE(ip6_forward_rt.ro_rt);
310 				ip6_forward_rt.ro_rt = 0;
311 			}
312 			/* this probably fails but give it a try again */
313 			rtalloc_ign((struct route *)&ip6_forward_rt,
314 				    RTF_PRCLONING);
315 		}
316 
317 		if (ip6_forward_rt.ro_rt == 0) {
318 			ip6stat.ip6s_noroute++;
319 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
320 			if (mcopy) {
321 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
322 					    ICMP6_DST_UNREACH_NOROUTE, 0);
323 			}
324 			m_freem(m);
325 			return;
326 		}
327 	} else if ((rt = ip6_forward_rt.ro_rt) == 0 ||
328 		 !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) {
329 		if (ip6_forward_rt.ro_rt) {
330 			RTFREE(ip6_forward_rt.ro_rt);
331 			ip6_forward_rt.ro_rt = 0;
332 		}
333 		bzero(dst, sizeof(*dst));
334 		dst->sin6_len = sizeof(struct sockaddr_in6);
335 		dst->sin6_family = AF_INET6;
336 		dst->sin6_addr = ip6->ip6_dst;
337 
338   		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
339 		if (ip6_forward_rt.ro_rt == 0) {
340 			ip6stat.ip6s_noroute++;
341 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
342 			if (mcopy) {
343 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
344 					    ICMP6_DST_UNREACH_NOROUTE, 0);
345 			}
346 			m_freem(m);
347 			return;
348 		}
349 	}
350 	rt = ip6_forward_rt.ro_rt;
351 
352 	/*
353 	 * Scope check: if a packet can't be delivered to its destination
354 	 * for the reason that the destination is beyond the scope of the
355 	 * source address, discard the packet and return an icmp6 destination
356 	 * unreachable error with Code 2 (beyond scope of source address).
357 	 * [draft-ietf-ipngwg-icmp-v3-02.txt, Section 3.1]
358 	 */
359 	if (in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_src) !=
360 	    in6_addr2scopeid(rt->rt_ifp, &ip6->ip6_src)) {
361 		ip6stat.ip6s_cantforward++;
362 		ip6stat.ip6s_badscope++;
363 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
364 
365 		if (ip6_log_time + ip6_log_interval < time_second) {
366 			ip6_log_time = time_second;
367 			log(LOG_DEBUG,
368 			    "cannot forward "
369 			    "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
370 			    ip6_sprintf(&ip6->ip6_src),
371 			    ip6_sprintf(&ip6->ip6_dst),
372 			    ip6->ip6_nxt,
373 			    if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
374 		}
375 		if (mcopy)
376 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
377 				    ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
378 		m_freem(m);
379 		return;
380 	}
381 
382 	if (m->m_pkthdr.len > rt->rt_ifp->if_mtu) {
383 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
384 		if (mcopy) {
385 			u_long mtu;
386 #ifdef IPSEC
387 			struct secpolicy *sp;
388 			int ipsecerror;
389 			size_t ipsechdrsiz;
390 #endif
391 
392 			mtu = rt->rt_ifp->if_mtu;
393 #ifdef IPSEC
394 			/*
395 			 * When we do IPsec tunnel ingress, we need to play
396 			 * with if_mtu value (decrement IPsec header size
397 			 * from mtu value).  The code is much simpler than v4
398 			 * case, as we have the outgoing interface for
399 			 * encapsulated packet as "rt->rt_ifp".
400 			 */
401 			sp = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
402 				IP_FORWARDING, &ipsecerror);
403 			if (sp) {
404 				ipsechdrsiz = ipsec6_hdrsiz(mcopy,
405 					IPSEC_DIR_OUTBOUND, NULL);
406 				if (ipsechdrsiz < mtu)
407 					mtu -= ipsechdrsiz;
408 			}
409 
410 			/*
411 			 * if mtu becomes less than minimum MTU,
412 			 * tell minimum MTU (and I'll need to fragment it).
413 			 */
414 			if (mtu < IPV6_MMTU)
415 				mtu = IPV6_MMTU;
416 #endif
417 			icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
418 		}
419 		m_freem(m);
420 		return;
421  	}
422 
423 	if (rt->rt_flags & RTF_GATEWAY)
424 		dst = (struct sockaddr_in6 *)rt->rt_gateway;
425 
426 	/*
427 	 * If we are to forward the packet using the same interface
428 	 * as one we got the packet from, perhaps we should send a redirect
429 	 * to sender to shortcut a hop.
430 	 * Only send redirect if source is sending directly to us,
431 	 * and if packet was not source routed (or has any options).
432 	 * Also, don't send redirect if forwarding using a route
433 	 * modified by a redirect.
434 	 */
435 	if (rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
436 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0) {
437 		if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) {
438 			/*
439 			 * If the incoming interface is equal to the outgoing
440 			 * one, and the link attached to the interface is
441 			 * point-to-point, then it will be highly probable
442 			 * that a routing loop occurs. Thus, we immediately
443 			 * drop the packet and send an ICMPv6 error message.
444 			 *
445 			 * type/code is based on suggestion by Rich Draves.
446 			 * not sure if it is the best pick.
447 			 */
448 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
449 				    ICMP6_DST_UNREACH_ADDR, 0);
450 			m_freem(m);
451 			return;
452 		}
453 		type = ND_REDIRECT;
454 	}
455 
456 	/*
457 	 * Check with the firewall...
458 	 */
459 	if (ip6_fw_enable && ip6_fw_chk_ptr) {
460 		u_short port = 0;
461 		/* If ipfw says divert, we have to just drop packet */
462 		if ((*ip6_fw_chk_ptr)(&ip6, rt->rt_ifp, &port, &m)) {
463 			m_freem(m);
464 			goto freecopy;
465 		}
466 		if (!m)
467 			goto freecopy;
468 	}
469 
470 	/*
471 	 * Fake scoped addresses. Note that even link-local source or
472 	 * destinaion can appear, if the originating node just sends the
473 	 * packet to us (without address resolution for the destination).
474 	 * Since both icmp6_error and icmp6_redirect_output fill the embedded
475 	 * link identifiers, we can do this stuff after making a copy for
476 	 * returning an error.
477 	 */
478 	if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
479 		/*
480 		 * See corresponding comments in ip6_output.
481 		 * XXX: but is it possible that ip6_forward() sends a packet
482 		 *      to a loopback interface? I don't think so, and thus
483 		 *      I bark here. (jinmei@kame.net)
484 		 * XXX: it is common to route invalid packets to loopback.
485 		 *	also, the codepath will be visited on use of ::1 in
486 		 *	rthdr. (itojun)
487 		 */
488 #if 1
489 		if (0)
490 #else
491 		if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0)
492 #endif
493 		{
494 			printf("ip6_forward: outgoing interface is loopback. "
495 				"src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
496 				ip6_sprintf(&ip6->ip6_src),
497 				ip6_sprintf(&ip6->ip6_dst),
498 				ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
499 				if_name(rt->rt_ifp));
500 		}
501 
502 		/* we can just use rcvif in forwarding. */
503 		origifp = m->m_pkthdr.rcvif;
504 	}
505 	else
506 		origifp = rt->rt_ifp;
507 #ifndef SCOPEDROUTING
508 	/*
509 	 * clear embedded scope identifiers if necessary.
510 	 * in6_clearscope will touch the addresses only when necessary.
511 	 */
512 	in6_clearscope(&ip6->ip6_src);
513 	in6_clearscope(&ip6->ip6_dst);
514 #endif
515 
516 #ifdef PFIL_HOOKS
517 	/*
518 	 * Run through list of hooks for output packets.
519 	 */
520 	m1 = m;
521 	pfh = pfil_hook_get(PFIL_OUT, &inet6sw[ip6_protox[IPPROTO_IPV6]].pr_pfh);
522 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
523 		if (pfh->pfil_func) {
524 			rv = pfh->pfil_func(ip6, sizeof(*ip6),
525 					    rt->rt_ifp, 1, &m1);
526 			if (rv) {
527 				error = EHOSTUNREACH;
528 				goto freecopy;
529 			}
530 			m = m1;
531 			if (m == NULL)
532 				goto freecopy;
533 			ip6 = mtod(m, struct ip6_hdr *);
534 		}
535 #endif /* PFIL_HOOKS */
536 
537 	error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
538 	if (error) {
539 		in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
540 		ip6stat.ip6s_cantforward++;
541 	} else {
542 		ip6stat.ip6s_forward++;
543 		in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
544 		if (type)
545 			ip6stat.ip6s_redirectsent++;
546 		else {
547 			if (mcopy)
548 				goto freecopy;
549 		}
550 	}
551 	if (mcopy == NULL)
552 		return;
553 	switch (error) {
554 	case 0:
555 #if 1
556 		if (type == ND_REDIRECT) {
557 			icmp6_redirect_output(mcopy, rt);
558 			return;
559 		}
560 #endif
561 		goto freecopy;
562 
563 	case EMSGSIZE:
564 		/* xxx MTU is constant in PPP? */
565 		goto freecopy;
566 
567 	case ENOBUFS:
568 		/* Tell source to slow down like source quench in IP? */
569 		goto freecopy;
570 
571 	case ENETUNREACH:	/* shouldn't happen, checked above */
572 	case EHOSTUNREACH:
573 	case ENETDOWN:
574 	case EHOSTDOWN:
575 	default:
576 		type = ICMP6_DST_UNREACH;
577 		code = ICMP6_DST_UNREACH_ADDR;
578 		break;
579 	}
580 	icmp6_error(mcopy, type, code, 0);
581 	return;
582 
583  freecopy:
584 	m_freem(mcopy);
585 	return;
586 }
587