xref: /netbsd-src/sys/netinet6/ip6_forward.c (revision 4472dbe5e3bd91ef2540bada7a7ca7384627ff9b)
1 /*	$NetBSD: ip6_forward.c,v 1.12 2000/06/03 14:36:36 itojun Exp $	*/
2 /*	$KAME: ip6_forward.c,v 1.37 2000/05/28 12:17:19 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 <sys/param.h>
34 #include <sys/systm.h>
35 #include <sys/malloc.h>
36 #include <sys/mbuf.h>
37 #include <sys/domain.h>
38 #include <sys/protosw.h>
39 #include <sys/socket.h>
40 #include <sys/errno.h>
41 #include <sys/time.h>
42 #include <sys/kernel.h>
43 #include <sys/syslog.h>
44 
45 #include <net/if.h>
46 #include <net/route.h>
47 
48 #include <netinet/in.h>
49 #include <netinet/in_var.h>
50 #include <netinet/ip_var.h>
51 #include <netinet/ip6.h>
52 #include <netinet6/ip6_var.h>
53 #include <netinet/icmp6.h>
54 #include <netinet6/nd6.h>
55 
56 #ifdef IPSEC_IPV6FWD
57 #include <netinet6/ipsec.h>
58 #include <netkey/key.h>
59 #include <netkey/key_debug.h>
60 #endif /* IPSEC_IPV6FWD */
61 
62 #ifdef IPV6FIREWALL
63 #include <netinet6/ip6_fw.h>
64 #endif
65 
66 #include <net/net_osdep.h>
67 
68 struct	route_in6 ip6_forward_rt;
69 
70 /*
71  * Forward a packet.  If some error occurs return the sender
72  * an icmp packet.  Note we can't always generate a meaningful
73  * icmp message because icmp doesn't have a large enough repertoire
74  * of codes and types.
75  *
76  * If not forwarding, just drop the packet.  This could be confusing
77  * if ipforwarding was zero but some routing protocol was advancing
78  * us as a gateway to somewhere.  However, we must let the routing
79  * protocol deal with that.
80  *
81  */
82 
83 void
84 ip6_forward(m, srcrt)
85 	struct mbuf *m;
86 	int srcrt;
87 {
88 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
89 	register struct sockaddr_in6 *dst;
90 	register struct rtentry *rt;
91 	int error, type = 0, code = 0;
92 	struct mbuf *mcopy = NULL;
93 	struct ifnet *origifp;	/* maybe unnecessary */
94 #ifdef IPSEC_IPV6FWD
95 	struct secpolicy *sp = NULL;
96 #endif
97 	long time_second = time.tv_sec;
98 
99 #ifdef IPSEC_IPV6FWD
100 	/*
101 	 * Check AH/ESP integrity.
102 	 */
103 	/*
104 	 * Don't increment ip6s_cantforward because this is the check
105 	 * before forwarding packet actually.
106 	 */
107 	if (ipsec6_in_reject(m, NULL)) {
108 		ipsec6stat.in_polvio++;
109 		m_freem(m);
110 		return;
111 	}
112 #endif /*IPSEC_IPV6FWD*/
113 
114 	if ((m->m_flags & (M_BCAST|M_MCAST)) != 0 ||
115 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
116 		ip6stat.ip6s_cantforward++;
117 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
118 		if (ip6_log_time + ip6_log_interval < time_second) {
119 			ip6_log_time = time_second;
120 			log(LOG_DEBUG,
121 			    "cannot forward "
122 			    "from %s to %s nxt %d received on %s\n",
123 			    ip6_sprintf(&ip6->ip6_src),
124 			    ip6_sprintf(&ip6->ip6_dst),
125 			    ip6->ip6_nxt,
126 			    if_name(m->m_pkthdr.rcvif));
127 		}
128 		m_freem(m);
129 		return;
130 	}
131 
132 	if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
133 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
134 		icmp6_error(m, ICMP6_TIME_EXCEEDED,
135 				ICMP6_TIME_EXCEED_TRANSIT, 0);
136 		return;
137 	}
138 	ip6->ip6_hlim -= IPV6_HLIMDEC;
139 
140 	/*
141 	 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
142 	 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
143 	 * we need to generate an ICMP6 message to the src.
144 	 * Thanks to M_EXT, in most cases copy will not occur.
145 	 *
146 	 * It is important to save it before IPsec processing as IPsec
147 	 * processing may modify the mbuf.
148 	 */
149 	mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
150 
151 #ifdef IPSEC_IPV6FWD
152 	/* get a security policy for this packet */
153 	sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, 0, &error);
154 	if (sp == NULL) {
155 		ipsec6stat.out_inval++;
156 		ip6stat.ip6s_cantforward++;
157 		if (mcopy) {
158 #if 0
159 			/* XXX: what icmp ? */
160 #else
161 			m_freem(mcopy);
162 #endif
163 		}
164 		m_freem(m);
165 		return;
166 	}
167 
168 	error = 0;
169 
170 	/* check policy */
171 	switch (sp->policy) {
172 	case IPSEC_POLICY_DISCARD:
173 		/*
174 		 * This packet is just discarded.
175 		 */
176 		ipsec6stat.out_polvio++;
177 		ip6stat.ip6s_cantforward++;
178 		key_freesp(sp);
179 		if (mcopy) {
180 #if 0
181 			/* XXX: what icmp ? */
182 #else
183 			m_freem(mcopy);
184 #endif
185 		}
186 		m_freem(m);
187 		return;
188 
189 	case IPSEC_POLICY_BYPASS:
190 	case IPSEC_POLICY_NONE:
191 		/* no need to do IPsec. */
192 		key_freesp(sp);
193 		goto skip_ipsec;
194 
195 	case IPSEC_POLICY_IPSEC:
196 		if (sp->req == NULL) {
197 			/* XXX should be panic ? */
198 			printf("ip6_forward: No IPsec request specified.\n");
199 			ip6stat.ip6s_cantforward++;
200 			key_freesp(sp);
201 			if (mcopy) {
202 #if 0
203 				/* XXX: what icmp ? */
204 #else
205 				m_freem(mcopy);
206 #endif
207 			}
208 			m_freem(m);
209 			return;
210 		}
211 		/* do IPsec */
212 		break;
213 
214 	case IPSEC_POLICY_ENTRUST:
215 	default:
216 		/* should be panic ?? */
217 		printf("ip6_forward: Invalid policy found. %d\n", sp->policy);
218 		key_freesp(sp);
219 		goto skip_ipsec;
220 	}
221 
222     {
223 	struct ipsec_output_state state;
224 
225 	/*
226 	 * All the extension headers will become inaccessible
227 	 * (since they can be encrypted).
228 	 * Don't panic, we need no more updates to extension headers
229 	 * on inner IPv6 packet (since they are now encapsulated).
230 	 *
231 	 * IPv6 [ESP|AH] IPv6 [extension headers] payload
232 	 */
233 	bzero(&state, sizeof(state));
234 	state.m = m;
235 	state.ro = NULL;	/* update at ipsec6_output_tunnel() */
236 	state.dst = NULL;	/* update at ipsec6_output_tunnel() */
237 
238 	error = ipsec6_output_tunnel(&state, sp, 0);
239 
240 	m = state.m;
241 #if 0	/* XXX allocate a route (ro, dst) again later */
242 	ro = (struct route_in6 *)state.ro;
243 	dst = (struct sockaddr_in6 *)state.dst;
244 #endif
245 	key_freesp(sp);
246 
247 	if (error) {
248 		/* mbuf is already reclaimed in ipsec6_output_tunnel. */
249 		switch (error) {
250 		case EHOSTUNREACH:
251 		case ENETUNREACH:
252 		case EMSGSIZE:
253 		case ENOBUFS:
254 		case ENOMEM:
255 			break;
256 		default:
257 			printf("ip6_output (ipsec): error code %d\n", error);
258 			/*fall through*/
259 		case ENOENT:
260 			/* don't show these error codes to the user */
261 			break;
262 		}
263 		ip6stat.ip6s_cantforward++;
264 		if (mcopy) {
265 #if 0
266 			/* XXX: what icmp ? */
267 #else
268 			m_freem(mcopy);
269 #endif
270 		}
271 		m_freem(m);
272 		return;
273 	}
274     }
275     skip_ipsec:
276 #endif /* IPSEC_IPV6FWD */
277 
278 	dst = &ip6_forward_rt.ro_dst;
279 	if (!srcrt) {
280 		/*
281 		 * ip6_forward_rt.ro_dst.sin6_addr is equal to ip6->ip6_dst
282 		 */
283 		if (ip6_forward_rt.ro_rt == 0 ||
284 		    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) == 0) {
285 			if (ip6_forward_rt.ro_rt) {
286 				RTFREE(ip6_forward_rt.ro_rt);
287 				ip6_forward_rt.ro_rt = 0;
288 			}
289 			/* this probably fails but give it a try again */
290 			rtalloc((struct route *)&ip6_forward_rt);
291 		}
292 
293 		if (ip6_forward_rt.ro_rt == 0) {
294 			ip6stat.ip6s_noroute++;
295 			/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
296 			if (mcopy) {
297 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
298 					    ICMP6_DST_UNREACH_NOROUTE, 0);
299 			}
300 			m_freem(m);
301 			return;
302 		}
303 	} else if ((rt = ip6_forward_rt.ro_rt) == 0 ||
304 		 !IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst, &dst->sin6_addr)) {
305 		if (ip6_forward_rt.ro_rt) {
306 			RTFREE(ip6_forward_rt.ro_rt);
307 			ip6_forward_rt.ro_rt = 0;
308 		}
309 		bzero(dst, sizeof(*dst));
310 		dst->sin6_len = sizeof(struct sockaddr_in6);
311 		dst->sin6_family = AF_INET6;
312 		dst->sin6_addr = ip6->ip6_dst;
313 
314 		rtalloc((struct route *)&ip6_forward_rt);
315 		if (ip6_forward_rt.ro_rt == 0) {
316 			ip6stat.ip6s_noroute++;
317 			/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_noroute) */
318 			if (mcopy) {
319 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
320 					    ICMP6_DST_UNREACH_NOROUTE, 0);
321 			}
322 			m_freem(m);
323 			return;
324 		}
325 	}
326 	rt = ip6_forward_rt.ro_rt;
327 
328 	/*
329 	 * Scope check: if a packet can't be delivered to its destination
330 	 * for the reason that the destination is beyond the scope of the
331 	 * source address, discard the packet and return an icmp6 destination
332 	 * unreachable error with Code 2 (beyond scope of source address).
333 	 * [draft-ietf-ipngwg-icmp-v3-00.txt, Section 3.1]
334 	 */
335 	if (in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_src) !=
336 	    in6_addr2scopeid(rt->rt_ifp, &ip6->ip6_src)) {
337 		ip6stat.ip6s_cantforward++;
338 		ip6stat.ip6s_badscope++;
339 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
340 
341 		if (ip6_log_time + ip6_log_interval < time_second) {
342 			ip6_log_time = time_second;
343 			log(LOG_DEBUG,
344 			    "cannot forward "
345 			    "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
346 			    ip6_sprintf(&ip6->ip6_src),
347 			    ip6_sprintf(&ip6->ip6_dst),
348 			    ip6->ip6_nxt,
349 			    if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
350 		}
351 		if (mcopy)
352 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
353 				    ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
354 		m_freem(m);
355 		return;
356 	}
357 
358 	if (m->m_pkthdr.len > rt->rt_ifp->if_mtu) {
359 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
360 		if (mcopy) {
361 			u_long mtu;
362 #ifdef IPSEC_IPV6FWD
363 			struct secpolicy *sp;
364 			int ipsecerror;
365 			size_t ipsechdrsiz;
366 #endif
367 
368 			mtu = rt->rt_ifp->if_mtu;
369 #ifdef IPSEC_IPV6FWD
370 			/*
371 			 * When we do IPsec tunnel ingress, we need to play
372 			 * with if_mtu value (decrement IPsec header size
373 			 * from mtu value).  The code is much simpler than v4
374 			 * case, as we have the outgoing interface for
375 			 * encapsulated packet as "rt->rt_ifp".
376 			 */
377 			sp = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
378 				IP_FORWARDING, &ipsecerror);
379 			if (sp) {
380 				ipsechdrsiz = ipsec6_hdrsiz(mcopy,
381 					IPSEC_DIR_OUTBOUND, NULL);
382 				if (ipsechdrsiz < mtu)
383 					mtu -= ipsechdrsiz;
384 			}
385 
386 			/*
387 			 * if mtu becomes less than minimum MTU,
388 			 * tell minimum MTU (and I'll need to fragment it).
389 			 */
390 			if (mtu < IPV6_MMTU)
391 				mtu = IPV6_MMTU;
392 #endif
393 			icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
394 		}
395 		m_freem(m);
396 		return;
397  	}
398 
399 	if (rt->rt_flags & RTF_GATEWAY)
400 		dst = (struct sockaddr_in6 *)rt->rt_gateway;
401 
402 	/*
403 	 * If we are to forward the packet using the same interface
404 	 * as one we got the packet from, perhaps we should send a redirect
405 	 * to sender to shortcut a hop.
406 	 * Only send redirect if source is sending directly to us,
407 	 * and if packet was not source routed (or has any options).
408 	 * Also, don't send redirect if forwarding using a route
409 	 * modified by a redirect.
410 	 */
411 	if (rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
412 	    (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0)
413 		type = ND_REDIRECT;
414 
415 #ifdef IPV6FIREWALL
416 	/*
417 	 * Check with the firewall...
418 	 */
419 	if (ip6_fw_chk_ptr) {
420 		u_short port = 0;
421 		/* If ipfw says divert, we have to just drop packet */
422 		if ((*ip6_fw_chk_ptr)(&ip6, rt->rt_ifp, &port, &m)) {
423 			m_freem(m);
424 			goto freecopy;
425 		}
426 		if (!m)
427 			goto freecopy;
428 	}
429 #endif
430 
431 	/*
432 	 * Fake scoped addresses. Note that even link-local source or
433 	 * destinaion can appear, if the originating node just sends the
434 	 * packet to us (without address resolution for the destination).
435 	 * Since both icmp6_error and icmp6_redirect_output fill the embedded
436 	 * link identifiers, we can do this stuff after make a copy for
437 	 * returning error.
438 	 */
439 	if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
440 		/*
441 		 * See corresponding comments in ip6_output.
442 		 * XXX: but is it possible that ip6_forward() sends a packet
443 		 *      to a loopback interface? I don't think so, and thus
444 		 *      I bark here. (jinmei@kame.net)
445 		 * XXX: it is common to route invalid packets to loopback.
446 		 *	(itojun)
447 		 */
448 
449 		if ((rt->rt_flags & (RTF_BLACKHOLE|RTF_REJECT)) == 0) {
450 			printf("ip6_forward: outgoing interface is loopback. "
451 			       "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
452 			       ip6_sprintf(&ip6->ip6_src),
453 			       ip6_sprintf(&ip6->ip6_dst),
454 			       ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
455 			       if_name(rt->rt_ifp));
456 		}
457 
458 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
459 			origifp = ifindex2ifnet[ntohs(ip6->ip6_src.s6_addr16[1])];
460 		else if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
461 			origifp = ifindex2ifnet[ntohs(ip6->ip6_dst.s6_addr16[1])];
462 		else
463 			origifp = rt->rt_ifp;
464 	}
465 	else
466 		origifp = rt->rt_ifp;
467 #ifndef FAKE_LOOPBACK_IF
468 	if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) == 0)
469 #else
470 	if (1)
471 #endif
472 	{
473 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
474 			ip6->ip6_src.s6_addr16[1] = 0;
475 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
476 			ip6->ip6_dst.s6_addr16[1] = 0;
477 	}
478 
479 #ifdef OLDIP6OUTPUT
480 	error = (*rt->rt_ifp->if_output)(rt->rt_ifp, m,
481 					 (struct sockaddr *)dst,
482 					 ip6_forward_rt.ro_rt);
483 #else
484 	error = nd6_output(rt->rt_ifp, origifp, m, dst, rt);
485 #endif
486 	if (error) {
487 		in6_ifstat_inc(rt->rt_ifp, ifs6_out_discard);
488 		ip6stat.ip6s_cantforward++;
489 	} else {
490 		ip6stat.ip6s_forward++;
491 		in6_ifstat_inc(rt->rt_ifp, ifs6_out_forward);
492 		if (type)
493 			ip6stat.ip6s_redirectsent++;
494 		else {
495 			if (mcopy)
496 				goto freecopy;
497 		}
498 	}
499 	if (mcopy == NULL)
500 		return;
501 
502 	switch (error) {
503 	case 0:
504 #if 1
505 		if (type == ND_REDIRECT) {
506 			icmp6_redirect_output(mcopy, rt);
507 			return;
508 		}
509 #endif
510 		goto freecopy;
511 
512 	case EMSGSIZE:
513 		/* xxx MTU is constant in PPP? */
514 		goto freecopy;
515 
516 	case ENOBUFS:
517 		/* Tell source to slow down like source quench in IP? */
518 		goto freecopy;
519 
520 	case ENETUNREACH:	/* shouldn't happen, checked above */
521 	case EHOSTUNREACH:
522 	case ENETDOWN:
523 	case EHOSTDOWN:
524 	default:
525 		type = ICMP6_DST_UNREACH;
526 		code = ICMP6_DST_UNREACH_ADDR;
527 		break;
528 	}
529 	icmp6_error(mcopy, type, code, 0);
530 	return;
531 
532  freecopy:
533 	m_freem(mcopy);
534 	return;
535 }
536