xref: /xnu-8019.80.24/bsd/netinet6/ip6_forward.c (revision a325d9c4a84054e40bbe985afedcb50ab80993ea)
1 /*
2  * Copyright (c) 2009-2021 Apple Inc. All rights reserved.
3  *
4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5  *
6  * This file contains Original Code and/or Modifications of Original Code
7  * as defined in and that are subject to the Apple Public Source License
8  * Version 2.0 (the 'License'). You may not use this file except in
9  * compliance with the License. The rights granted to you under the License
10  * may not be used to create, or enable the creation or redistribution of,
11  * unlawful or unlicensed copies of an Apple operating system, or to
12  * circumvent, violate, or enable the circumvention or violation of, any
13  * terms of an Apple operating system software license agreement.
14  *
15  * Please obtain a copy of the License at
16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
17  *
18  * The Original Code and all software distributed under the License are
19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23  * Please see the License for the specific language governing rights and
24  * limitations under the License.
25  *
26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27  */
28 
29 /*	$FreeBSD: src/sys/netinet6/ip6_forward.c,v 1.16 2002/10/16 02:25:05 sam Exp $	*/
30 /*	$KAME: ip6_forward.c,v 1.69 2001/05/17 03:48:30 itojun Exp $	*/
31 
32 /*
33  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
34  * All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. Neither the name of the project nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  */
60 
61 
62 #include <sys/param.h>
63 #include <sys/systm.h>
64 #include <sys/malloc.h>
65 #include <sys/mbuf.h>
66 #include <sys/domain.h>
67 #include <sys/protosw.h>
68 #include <sys/socket.h>
69 #include <sys/errno.h>
70 #include <sys/time.h>
71 #include <sys/kernel.h>
72 #include <sys/syslog.h>
73 
74 #include <net/if.h>
75 #include <net/route.h>
76 
77 #include <netinet/in.h>
78 #include <netinet/in_var.h>
79 #include <netinet/in_systm.h>
80 #include <netinet/ip.h>
81 #include <netinet/ip_var.h>
82 #include <netinet6/in6_var.h>
83 #include <netinet/ip6.h>
84 #include <netinet6/ip6_var.h>
85 #include <netinet/icmp6.h>
86 #include <netinet6/nd6.h>
87 #include <netinet6/scope6_var.h>
88 
89 #include <netinet/in_pcb.h>
90 
91 #if IPSEC
92 #include <netinet6/ipsec.h>
93 #include <netinet6/ipsec6.h>
94 #include <netkey/key.h>
95 extern int ipsec_bypass;
96 #endif /* IPSEC */
97 
98 #include <net/net_osdep.h>
99 
100 #if DUMMYNET
101 #include <netinet/ip_dummynet.h>
102 #endif /* DUMMYNET */
103 
104 #if PF
105 #include <net/pfvar.h>
106 #endif /* PF */
107 
108 /*
109  * Forward a packet.  If some error occurs return the sender
110  * an icmp packet.  Note we can't always generate a meaningful
111  * icmp message because icmp doesn't have a large enough repertoire
112  * of codes and types.
113  *
114  * If not forwarding, just drop the packet.  This could be confusing
115  * if ipforwarding was zero but some routing protocol was advancing
116  * us as a gateway to somewhere.  However, we must let the routing
117  * protocol deal with that.
118  *
119  */
120 
121 struct mbuf *
ip6_forward(struct mbuf * m,struct route_in6 * ip6forward_rt,int srcrt)122 ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
123     int srcrt)
124 {
125 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
126 	struct sockaddr_in6 *dst;
127 	struct rtentry *rt;
128 	int error, type = 0, code = 0;
129 	boolean_t proxy = FALSE;
130 	struct mbuf *mcopy = NULL;
131 	struct ifnet *ifp, *rcvifp, *origifp;   /* maybe unnecessary */
132 	u_int32_t inzone, outzone, len = 0, pktcnt = 0;
133 	struct in6_addr src_in6, dst_in6;
134 	uint64_t curtime = net_uptime();
135 #if IPSEC
136 	struct secpolicy *sp = NULL;
137 #endif
138 	unsigned int ifscope = IFSCOPE_NONE;
139 #if PF
140 	struct pf_mtag *pf_mtag;
141 	struct pf_fragment_tag *pf_ftagp, pf_ftag;
142 	boolean_t pf_ftag_valid = FALSE;
143 #endif /* PF */
144 	uint32_t mpktlen = 0;
145 
146 	/*
147 	 * In the prefix proxying case, the route to the proxied node normally
148 	 * gets created by nd6_prproxy_ns_output(), as part of forwarding a
149 	 * NS (NUD/AR) packet to the proxied node.  In the event that such
150 	 * packet did not arrive in time before the correct route gets created,
151 	 * ip6_input() would have performed a rtalloc() which most likely will
152 	 * create the wrong cloned route; this route points back to the same
153 	 * interface as the inbound interface, since the parent non-scoped
154 	 * prefix route points there.  Therefore we check if that is the case
155 	 * and perform the necessary fixup to get the correct route installed.
156 	 */
157 	if (!srcrt && nd6_prproxy &&
158 	    (rt = ip6forward_rt->ro_rt) != NULL && (rt->rt_flags & RTF_PROXY)) {
159 		nd6_proxy_find_fwdroute(m->m_pkthdr.rcvif, ip6forward_rt);
160 		if ((rt = ip6forward_rt->ro_rt) != NULL) {
161 			ifscope = rt->rt_ifp->if_index;
162 		}
163 	}
164 
165 #if PF
166 	pf_mtag = pf_find_mtag(m);
167 	/*
168 	 * save the PF fragmentation metadata as m_copy() removes the
169 	 * mbufs tags from the original mbuf.
170 	 */
171 	pf_ftagp = pf_find_fragment_tag(m);
172 	if (pf_ftagp != NULL) {
173 		ASSERT(pf_mtag->pftag_flags & PF_TAG_REASSEMBLED);
174 		pf_ftag = *pf_ftagp;
175 		pf_ftag_valid = TRUE;
176 		mpktlen = pf_ftag.ft_maxlen;
177 		ASSERT(mpktlen);
178 	}
179 	if (pf_mtag != NULL && pf_mtag->pftag_rtableid != IFSCOPE_NONE) {
180 		ifscope = pf_mtag->pftag_rtableid;
181 	}
182 	pf_mtag = NULL;
183 	pf_ftagp = NULL;
184 	/*
185 	 * If the caller provides a route which is on a different interface
186 	 * than the one specified for scoped forwarding, discard the route
187 	 * and do a lookup below.
188 	 */
189 	if (ifscope != IFSCOPE_NONE && (rt = ip6forward_rt->ro_rt) != NULL) {
190 		RT_LOCK(rt);
191 		if (rt->rt_ifp->if_index != ifscope) {
192 			RT_UNLOCK(rt);
193 			ROUTE_RELEASE(ip6forward_rt);
194 			rt = NULL;
195 		} else {
196 			RT_UNLOCK(rt);
197 		}
198 	}
199 #endif /* PF */
200 
201 #if IPSEC
202 	/*
203 	 * Check AH/ESP integrity.
204 	 */
205 	/*
206 	 * Don't increment ip6s_cantforward because this is the check
207 	 * before forwarding packet actually.
208 	 */
209 	if (ipsec_bypass == 0) {
210 		if (ipsec6_in_reject(m, NULL)) {
211 			IPSEC_STAT_INCREMENT(ipsec6stat.in_polvio);
212 			m_freem(m);
213 			return NULL;
214 		}
215 	}
216 #endif /*IPSEC*/
217 
218 	/*
219 	 * Do not forward packets to multicast destination.
220 	 * Do not forward packets with unspecified source.  It was discussed
221 	 * in July 2000, on ipngwg mailing list.
222 	 */
223 	if ((m->m_flags & (M_BCAST | M_MCAST)) != 0 ||
224 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
225 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
226 		ip6stat.ip6s_cantforward++;
227 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
228 		if (ip6_log_time + ip6_log_interval < curtime) {
229 			ip6_log_time = curtime;
230 			log(LOG_DEBUG,
231 			    "cannot forward "
232 			    "from %s to %s nxt %d received on %s\n",
233 			    ip6_sprintf(&ip6->ip6_src),
234 			    ip6_sprintf(&ip6->ip6_dst),
235 			    ip6->ip6_nxt,
236 			    if_name(m->m_pkthdr.rcvif));
237 		}
238 		m_freem(m);
239 		return NULL;
240 	}
241 
242 	if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
243 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
244 		icmp6_error_flag(m, ICMP6_TIME_EXCEEDED,
245 		    ICMP6_TIME_EXCEED_TRANSIT, 0, 0);
246 		return NULL;
247 	}
248 
249 	/*
250 	 * See if the destination is a proxied address, and if so pretend
251 	 * that it's for us.  This is mostly to handle NUD probes against
252 	 * the proxied addresses.  We filter for ICMPv6 here and will let
253 	 * icmp6_input handle the rest.
254 	 */
255 	if (!srcrt && nd6_prproxy) {
256 		VERIFY(!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst));
257 		proxy = nd6_prproxy_isours(m, ip6, ip6forward_rt, ifscope);
258 		/*
259 		 * Don't update hop limit while proxying; RFC 4389 4.1.
260 		 * Also skip IPsec forwarding path processing as this
261 		 * packet is not to be forwarded.
262 		 */
263 		if (proxy) {
264 			goto skip_ipsec;
265 		}
266 	}
267 
268 	ip6->ip6_hlim -= IPV6_HLIMDEC;
269 
270 	/*
271 	 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
272 	 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
273 	 * we need to generate an ICMP6 message to the src.
274 	 * Thanks to M_EXT, in most cases copy will not occur.
275 	 *
276 	 * It is important to save it before IPsec processing as IPsec
277 	 * processing may modify the mbuf.
278 	 */
279 	mcopy = m_copy(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN));
280 
281 #if IPSEC
282 	if (ipsec_bypass != 0) {
283 		goto skip_ipsec;
284 	}
285 	/* get a security policy for this packet */
286 	sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, IP_FORWARDING,
287 	    &error);
288 	if (sp == NULL) {
289 		IPSEC_STAT_INCREMENT(ipsec6stat.out_inval);
290 		ip6stat.ip6s_cantforward++;
291 		if (mcopy) {
292 #if 0
293 			/* XXX: what icmp ? */
294 #else
295 			m_freem(mcopy);
296 #endif
297 		}
298 		m_freem(m);
299 		return NULL;
300 	}
301 
302 	error = 0;
303 
304 	/* check policy */
305 	switch (sp->policy) {
306 	case IPSEC_POLICY_DISCARD:
307 	case IPSEC_POLICY_GENERATE:
308 		/*
309 		 * This packet is just discarded.
310 		 */
311 		IPSEC_STAT_INCREMENT(ipsec6stat.out_polvio);
312 		ip6stat.ip6s_cantforward++;
313 		key_freesp(sp, KEY_SADB_UNLOCKED);
314 		if (mcopy) {
315 #if 0
316 			/* XXX: what icmp ? */
317 #else
318 			m_freem(mcopy);
319 #endif
320 		}
321 		m_freem(m);
322 		return NULL;
323 
324 	case IPSEC_POLICY_BYPASS:
325 	case IPSEC_POLICY_NONE:
326 		/* no need to do IPsec. */
327 		key_freesp(sp, KEY_SADB_UNLOCKED);
328 		goto skip_ipsec;
329 
330 	case IPSEC_POLICY_IPSEC:
331 		if (sp->req == NULL) {
332 			/* XXX should be panic ? */
333 			printf("ip6_forward: No IPsec request specified.\n");
334 			ip6stat.ip6s_cantforward++;
335 			key_freesp(sp, KEY_SADB_UNLOCKED);
336 			if (mcopy) {
337 #if 0
338 				/* XXX: what icmp ? */
339 #else
340 				m_freem(mcopy);
341 #endif
342 			}
343 			m_freem(m);
344 			return NULL;
345 		}
346 		/* do IPsec */
347 		break;
348 
349 	case IPSEC_POLICY_ENTRUST:
350 	default:
351 		/* should be panic ?? */
352 		printf("ip6_forward: Invalid policy found. %d\n", sp->policy);
353 		key_freesp(sp, KEY_SADB_UNLOCKED);
354 		goto skip_ipsec;
355 	}
356 
357 	{
358 		struct ipsec_output_state state;
359 
360 		/*
361 		 * All the extension headers will become inaccessible
362 		 * (since they can be encrypted).
363 		 * Don't panic, we need no more updates to extension headers
364 		 * on inner IPv6 packet (since they are now encapsulated).
365 		 *
366 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
367 		 */
368 		bzero(&state, sizeof(state));
369 		state.m = m;
370 		state.dst = NULL; /* update at ipsec6_output_tunnel() */
371 
372 		error = ipsec6_output_tunnel(&state, sp, 0);
373 		key_freesp(sp, KEY_SADB_UNLOCKED);
374 		if (state.tunneled == 4) {
375 			ROUTE_RELEASE(&state.ro);
376 			return NULL; /* packet is gone - sent over IPv4 */
377 		}
378 
379 		m = state.m;
380 		ROUTE_RELEASE(&state.ro);
381 
382 		if (error) {
383 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
384 			switch (error) {
385 			case EHOSTUNREACH:
386 			case ENETUNREACH:
387 			case EMSGSIZE:
388 			case ENOBUFS:
389 			case ENOMEM:
390 				break;
391 			default:
392 				printf("ip6_output (ipsec): error code %d\n", error);
393 				OS_FALLTHROUGH;
394 			case ENOENT:
395 				/* don't show these error codes to the user */
396 				break;
397 			}
398 			ip6stat.ip6s_cantforward++;
399 			if (mcopy) {
400 #if 0
401 				/* XXX: what icmp ? */
402 #else
403 				m_freem(mcopy);
404 #endif
405 			}
406 			m_freem(m);
407 			return NULL;
408 		}
409 	}
410 #endif /* IPSEC */
411 skip_ipsec:
412 
413 	dst = (struct sockaddr_in6 *)&ip6forward_rt->ro_dst;
414 	if ((rt = ip6forward_rt->ro_rt) != NULL) {
415 		RT_LOCK(rt);
416 		/* Take an extra ref for ourselves */
417 		RT_ADDREF_LOCKED(rt);
418 	}
419 
420 	VERIFY(rt == NULL || rt == ip6forward_rt->ro_rt);
421 	if (!srcrt) {
422 		/*
423 		 * ip6forward_rt->ro_dst.sin6_addr is equal to ip6->ip6_dst
424 		 */
425 		if (ROUTE_UNUSABLE(ip6forward_rt)) {
426 			if (rt != NULL) {
427 				/* Release extra ref */
428 				RT_REMREF_LOCKED(rt);
429 				RT_UNLOCK(rt);
430 			}
431 			ROUTE_RELEASE(ip6forward_rt);
432 
433 			/* this probably fails but give it a try again */
434 			rtalloc_scoped_ign((struct route *)ip6forward_rt,
435 			    RTF_PRCLONING, ifscope);
436 			if ((rt = ip6forward_rt->ro_rt) != NULL) {
437 				RT_LOCK(rt);
438 				/* Take an extra ref for ourselves */
439 				RT_ADDREF_LOCKED(rt);
440 			}
441 		}
442 
443 		if (rt == NULL) {
444 			ip6stat.ip6s_noroute++;
445 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
446 			if (mcopy) {
447 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
448 				    ICMP6_DST_UNREACH_NOROUTE, 0);
449 			}
450 			m_freem(m);
451 			return NULL;
452 		}
453 		RT_LOCK_ASSERT_HELD(rt);
454 	} else if (ROUTE_UNUSABLE(ip6forward_rt) ||
455 	    !in6_are_addr_equal_scoped(&ip6->ip6_dst, &dst->sin6_addr, ip6_input_getdstifscope(m), dst->sin6_scope_id)) {
456 		if (rt != NULL) {
457 			/* Release extra ref */
458 			RT_REMREF_LOCKED(rt);
459 			RT_UNLOCK(rt);
460 		}
461 		ROUTE_RELEASE(ip6forward_rt);
462 
463 		bzero(dst, sizeof(*dst));
464 		dst->sin6_len = sizeof(struct sockaddr_in6);
465 		dst->sin6_family = AF_INET6;
466 		dst->sin6_addr = ip6->ip6_dst;
467 
468 		rtalloc_scoped_ign((struct route *)ip6forward_rt,
469 		    RTF_PRCLONING, ifscope);
470 		if ((rt = ip6forward_rt->ro_rt) == NULL) {
471 			ip6stat.ip6s_noroute++;
472 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
473 			if (mcopy) {
474 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
475 				    ICMP6_DST_UNREACH_NOROUTE, 0);
476 			}
477 			m_freem(m);
478 			return NULL;
479 		}
480 		RT_LOCK(rt);
481 		/* Take an extra ref for ourselves */
482 		RT_ADDREF_LOCKED(rt);
483 	}
484 
485 	/*
486 	 * Source scope check: if a packet can't be delivered to its
487 	 * destination for the reason that the destination is beyond the scope
488 	 * of the source address, discard the packet and return an icmp6
489 	 * destination unreachable error with Code 2 (beyond scope of source
490 	 * address) unless we are proxying (source address is link local
491 	 * for NUDs.)  We use a local copy of ip6_src, since in6_setscope()
492 	 * will possibly modify its first argument.
493 	 * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1]
494 	 */
495 	src_in6 = ip6->ip6_src;
496 	if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) {
497 		RT_REMREF_LOCKED(rt);
498 		RT_UNLOCK(rt);
499 		/* XXX: this should not happen */
500 		ip6stat.ip6s_cantforward++;
501 		ip6stat.ip6s_badscope++;
502 		m_freem(m);
503 		return NULL;
504 	}
505 	if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) {
506 		RT_REMREF_LOCKED(rt);
507 		RT_UNLOCK(rt);
508 		ip6stat.ip6s_cantforward++;
509 		ip6stat.ip6s_badscope++;
510 		m_freem(m);
511 		return NULL;
512 	}
513 
514 	if (inzone != outzone && !proxy) {
515 		ip6stat.ip6s_cantforward++;
516 		ip6stat.ip6s_badscope++;
517 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
518 
519 		if (ip6_log_time + ip6_log_interval < curtime) {
520 			ip6_log_time = curtime;
521 			log(LOG_DEBUG,
522 			    "cannot forward "
523 			    "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
524 			    ip6_sprintf(&ip6->ip6_src),
525 			    ip6_sprintf(&ip6->ip6_dst),
526 			    ip6->ip6_nxt,
527 			    if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
528 		}
529 		/* Release extra ref */
530 		RT_REMREF_LOCKED(rt);
531 		RT_UNLOCK(rt);
532 		if (mcopy) {
533 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
534 			    ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
535 		}
536 		m_freem(m);
537 		return NULL;
538 	}
539 
540 	/*
541 	 * Destination scope check: if a packet is going to break the scope
542 	 * zone of packet's destination address, discard it.  This case should
543 	 * usually be prevented by appropriately-configured routing table, but
544 	 * we need an explicit check because we may mistakenly forward the
545 	 * packet to a different zone by (e.g.) a default route.
546 	 */
547 	dst_in6 = ip6->ip6_dst;
548 	if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 ||
549 	    in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 ||
550 	    inzone != outzone) {
551 		RT_REMREF_LOCKED(rt);
552 		RT_UNLOCK(rt);
553 		ip6stat.ip6s_cantforward++;
554 		ip6stat.ip6s_badscope++;
555 		m_freem(m);
556 		return NULL;
557 	}
558 
559 	if (mpktlen == 0) {
560 		mpktlen = m->m_pkthdr.len;
561 	}
562 
563 	if (mpktlen > rt->rt_ifp->if_mtu) {
564 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
565 		if (mcopy) {
566 			uint32_t mtu;
567 #if IPSEC
568 			struct secpolicy *sp2;
569 			int ipsecerror;
570 			size_t ipsechdrsiz;
571 #endif
572 
573 			mtu = rt->rt_ifp->if_mtu;
574 #if IPSEC
575 			/*
576 			 * When we do IPsec tunnel ingress, we need to play
577 			 * with the link value (decrement IPsec header size
578 			 * from mtu value).  The code is much simpler than v4
579 			 * case, as we have the outgoing interface for
580 			 * encapsulated packet as "rt->rt_ifp".
581 			 */
582 			sp2 = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
583 			    IP_FORWARDING, &ipsecerror);
584 			if (sp2) {
585 				ipsechdrsiz = ipsec6_hdrsiz(mcopy,
586 				    IPSEC_DIR_OUTBOUND, NULL);
587 				if (ipsechdrsiz < mtu) {
588 					mtu -= ipsechdrsiz;
589 				}
590 				key_freesp(sp2, KEY_SADB_UNLOCKED);
591 			}
592 			/*
593 			 * if mtu becomes less than minimum MTU,
594 			 * tell minimum MTU (and I'll need to fragment it).
595 			 */
596 			if (mtu < IPV6_MMTU) {
597 				mtu = IPV6_MMTU;
598 			}
599 #endif
600 			/* Release extra ref */
601 			RT_REMREF_LOCKED(rt);
602 			RT_UNLOCK(rt);
603 			icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
604 		} else {
605 			/* Release extra ref */
606 			RT_REMREF_LOCKED(rt);
607 			RT_UNLOCK(rt);
608 		}
609 		m_freem(m);
610 		return NULL;
611 	}
612 
613 	if (rt->rt_flags & RTF_GATEWAY) {
614 		dst = (struct sockaddr_in6 *)(void *)rt->rt_gateway;
615 	}
616 
617 	/*
618 	 * If we are to forward the packet using the same interface
619 	 * as one we got the packet from, perhaps we should send a redirect
620 	 * to sender to shortcut a hop.
621 	 * Only send redirect if source is sending directly to us,
622 	 * and if packet was not source routed (or has any options).
623 	 * Also, don't send redirect if forwarding using a route
624 	 * modified by a redirect.
625 	 */
626 	if (!proxy &&
627 	    ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
628 	    (rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) == 0) {
629 		if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) {
630 			/*
631 			 * If the incoming interface is equal to the outgoing
632 			 * one, and the link attached to the interface is
633 			 * point-to-point, then it will be highly probable
634 			 * that a routing loop occurs. Thus, we immediately
635 			 * drop the packet and send an ICMPv6 error message.
636 			 *
637 			 * type/code is based on suggestion by Rich Draves.
638 			 * not sure if it is the best pick.
639 			 */
640 			RT_REMREF_LOCKED(rt);   /* Release extra ref */
641 			RT_UNLOCK(rt);
642 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
643 			    ICMP6_DST_UNREACH_ADDR, 0);
644 			m_freem(m);
645 			return NULL;
646 		}
647 		type = ND_REDIRECT;
648 	}
649 	/*
650 	 * Fake scoped addresses. Note that even link-local source or
651 	 * destinaion can appear, if the originating node just sends the
652 	 * packet to us (without address resolution for the destination).
653 	 * Since both icmp6_error and icmp6_redirect_output fill the embedded
654 	 * link identifiers, we can do this stuff after making a copy for
655 	 * returning an error.
656 	 */
657 	if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
658 		/*
659 		 * See corresponding comments in ip6_output.
660 		 * XXX: but is it possible that ip6_forward() sends a packet
661 		 *      to a loopback interface? I don't think so, and thus
662 		 *      I bark here. ([email protected])
663 		 * XXX: it is common to route invalid packets to loopback.
664 		 *	also, the codepath will be visited on use of ::1 in
665 		 *	rthdr. (itojun)
666 		 */
667 #if 1
668 		if ((0))
669 #else
670 		if ((rt->rt_flags & (RTF_BLACKHOLE | RTF_REJECT)) == 0)
671 #endif
672 		{
673 			printf("ip6_forward: outgoing interface is loopback. "
674 			    "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
675 			    ip6_sprintf(&ip6->ip6_src),
676 			    ip6_sprintf(&ip6->ip6_dst),
677 			    ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
678 			    if_name(rt->rt_ifp));
679 		}
680 
681 		/* we can just use rcvif in forwarding. */
682 		origifp = rcvifp = m->m_pkthdr.rcvif;
683 	} else if (nd6_prproxy) {
684 		/*
685 		 * In the prefix proxying case, we need to inform nd6_output()
686 		 * about the inbound interface, so that any subsequent NS
687 		 * packets generated by nd6_prproxy_ns_output() will not be
688 		 * sent back to that same interface.
689 		 */
690 		origifp = rcvifp = m->m_pkthdr.rcvif;
691 	} else {
692 		rcvifp = m->m_pkthdr.rcvif;
693 		origifp = rt->rt_ifp;
694 	}
695 	/*
696 	 * clear embedded scope identifiers if necessary.
697 	 * in6_clearscope will touch the addresses only when necessary.
698 	 */
699 	in6_clearscope(&ip6->ip6_src);
700 	in6_clearscope(&ip6->ip6_dst);
701 
702 	ifp = rt->rt_ifp;
703 	/* Drop the lock but retain the extra ref */
704 	RT_UNLOCK(rt);
705 
706 	/*
707 	 * If this is to be processed locally, let ip6_input have it.
708 	 */
709 	if (proxy) {
710 		VERIFY(m->m_pkthdr.pkt_flags & PKTF_PROXY_DST);
711 		/* Release extra ref */
712 		RT_REMREF(rt);
713 		if (mcopy != NULL) {
714 			m_freem(mcopy);
715 		}
716 		return m;
717 	}
718 
719 	/* Mark this packet as being forwarded from another interface */
720 	m->m_pkthdr.pkt_flags |= PKTF_FORWARDED;
721 
722 #if PF
723 	if (PF_IS_ENABLED) {
724 		/*
725 		 * PF refragments any packet which it reassembled due to scrub
726 		 * rules, in which case it will set the PF_TAG_REFRAGMENTED
727 		 * flag in PF mbuf tag.
728 		 */
729 		if (pf_ftag_valid) {
730 			pf_copy_fragment_tag(m, &pf_ftag, M_DONTWAIT);
731 		}
732 #if DUMMYNET
733 		struct ip_fw_args args;
734 		bzero(&args, sizeof(args));
735 
736 		args.fwa_oif = ifp;
737 		args.fwa_oflags = 0;
738 		args.fwa_ro6 = ip6forward_rt;
739 		args.fwa_ro6_pmtu = ip6forward_rt;
740 		args.fwa_mtu = rt->rt_ifp->if_mtu;
741 		args.fwa_dst6 = dst;
742 		args.fwa_origifp = origifp;
743 		/* Invoke outbound packet filter */
744 		error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, &args);
745 #else /* !DUMMYNET */
746 		error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, NULL);
747 #endif /* !DUMMYNET */
748 		if (error != 0 || m == NULL) {
749 			if (m != NULL) {
750 				panic("%s: unexpected packet %p", __func__, m);
751 				/* NOTREACHED */
752 			}
753 			/* Already freed by callee */
754 			goto senderr;
755 		}
756 
757 		pf_mtag = pf_find_mtag(m);
758 		/*
759 		 * refragmented packets from PF.
760 		 */
761 		if ((pf_mtag->pftag_flags & PF_TAG_REFRAGMENTED) != 0) {
762 			struct mbuf *t;
763 
764 			pf_mtag->pftag_flags &= ~PF_TAG_REFRAGMENTED;
765 			/* for statistics */
766 			t = m;
767 			while (t != NULL) {
768 				pktcnt++;
769 				len += m_pktlen(t);
770 				t = t->m_nextpkt;
771 			}
772 
773 			/*
774 			 * nd6_output() frees packetchain in both success and
775 			 * failure cases.
776 			 */
777 			error = nd6_output(ifp, origifp, m, dst, rt, NULL);
778 			m = NULL;
779 			goto sent;
780 		}
781 		/*
782 		 * We do not use ip6 header again in the code below,
783 		 * however still adding the bit here so that any new
784 		 * code in future doesn't end up working with the
785 		 * wrong pointer
786 		 */
787 		ip6 = mtod(m, struct ip6_hdr *);
788 	}
789 #endif /* PF */
790 
791 	len = m_pktlen(m);
792 	pktcnt = 1;
793 	error = nd6_output(ifp, origifp, m, dst, rt, NULL);
794 sent:
795 	if (error) {
796 		in6_ifstat_add(ifp, ifs6_out_discard, pktcnt);
797 		ip6stat.ip6s_cantforward += pktcnt;
798 	} else {
799 		/*
800 		 * Increment stats on the source interface; the ones
801 		 * for destination interface has been taken care of
802 		 * during output above by virtue of PKTF_FORWARDED.
803 		 */
804 		rcvifp->if_fpackets += pktcnt;
805 		rcvifp->if_fbytes += len;
806 
807 		ip6stat.ip6s_forward += pktcnt;
808 		in6_ifstat_add(ifp, ifs6_out_forward, pktcnt);
809 		if (type) {
810 			ip6stat.ip6s_redirectsent++;
811 		} else {
812 			if (mcopy) {
813 				goto freecopy;
814 			}
815 		}
816 	}
817 #if PF
818 senderr:
819 #endif /* PF */
820 	if (mcopy == NULL) {
821 		/* Release extra ref */
822 		RT_REMREF(rt);
823 		return NULL;
824 	}
825 	switch (error) {
826 	case 0:
827 #if 1
828 		if (type == ND_REDIRECT) {
829 			icmp6_redirect_output(mcopy, rt);
830 			/* Release extra ref */
831 			RT_REMREF(rt);
832 			return NULL;
833 		}
834 #endif
835 		goto freecopy;
836 
837 	case EMSGSIZE:
838 		/* xxx MTU is constant in PPP? */
839 		goto freecopy;
840 
841 	case ENOBUFS:
842 		/* Tell source to slow down like source quench in IP? */
843 		goto freecopy;
844 
845 	case ENETUNREACH:       /* shouldn't happen, checked above */
846 	case EHOSTUNREACH:
847 	case ENETDOWN:
848 	case EHOSTDOWN:
849 	default:
850 		type = ICMP6_DST_UNREACH;
851 		code = ICMP6_DST_UNREACH_ADDR;
852 		break;
853 	}
854 	icmp6_error(mcopy, type, code, 0);
855 	/* Release extra ref */
856 	RT_REMREF(rt);
857 	return NULL;
858 
859 freecopy:
860 	m_freem(mcopy);
861 	/* Release extra ref */
862 	RT_REMREF(rt);
863 	return NULL;
864 }
865