xref: /xnu-10063.121.3/bsd/netinet6/ip6_forward.c (revision 2c2f96dc2b9a4408a43d3150ae9c105355ca3daa)
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 static void
adjust_scope_and_pktlen(struct mbuf * m,unsigned int * ifscope_p,uint32_t * mpktlen_p)107 adjust_scope_and_pktlen(struct mbuf *m,
108     unsigned int *ifscope_p, uint32_t *mpktlen_p)
109 {
110 	struct pf_mtag *pf_mtag;
111 	struct pf_fragment_tag *pf_ftagp;
112 
113 	pf_mtag = pf_find_mtag(m);
114 	ASSERT(pf_mtag != NULL);
115 	if (pf_mtag->pftag_rtableid != IFSCOPE_NONE) {
116 		*ifscope_p = pf_mtag->pftag_rtableid;
117 	}
118 	pf_ftagp = pf_find_fragment_tag(m);
119 	if (pf_ftagp != NULL) {
120 		ASSERT(pf_mtag->pftag_flags & PF_TAG_REASSEMBLED);
121 		*mpktlen_p = pf_ftagp->ft_maxlen;
122 		ASSERT(*mpktlen_p);
123 	}
124 }
125 
126 #endif /* PF */
127 
128 #include <net/sockaddr_utils.h>
129 
130 /*
131  * Forward a packet.  If some error occurs return the sender
132  * an icmp packet.  Note we can't always generate a meaningful
133  * icmp message because icmp doesn't have a large enough repertoire
134  * of codes and types.
135  *
136  * If not forwarding, just drop the packet.  This could be confusing
137  * if ipforwarding was zero but some routing protocol was advancing
138  * us as a gateway to somewhere.  However, we must let the routing
139  * protocol deal with that.
140  *
141  */
142 
143 struct mbuf *
ip6_forward(struct mbuf * m,struct route_in6 * ip6forward_rt,int srcrt)144 ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
145     int srcrt)
146 {
147 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
148 	struct sockaddr_in6 *dst;
149 	struct rtentry *rt;
150 	int error, type = 0, code = 0;
151 	boolean_t proxy = FALSE;
152 	struct mbuf *mcopy = NULL;
153 	struct ifnet *ifp, *rcvifp, *origifp;   /* maybe unnecessary */
154 	u_int32_t inzone, outzone, len = 0, pktcnt = 0;
155 	struct in6_addr src_in6, dst_in6;
156 	uint64_t curtime = net_uptime();
157 #if IPSEC
158 	struct secpolicy *sp = NULL;
159 #endif
160 	unsigned int ifscope = IFSCOPE_NONE;
161 	uint32_t mpktlen = 0;
162 
163 	/*
164 	 * In the prefix proxying case, the route to the proxied node normally
165 	 * gets created by nd6_prproxy_ns_output(), as part of forwarding a
166 	 * NS (NUD/AR) packet to the proxied node.  In the event that such
167 	 * packet did not arrive in time before the correct route gets created,
168 	 * ip6_input() would have performed a rtalloc() which most likely will
169 	 * create the wrong cloned route; this route points back to the same
170 	 * interface as the inbound interface, since the parent non-scoped
171 	 * prefix route points there.  Therefore we check if that is the case
172 	 * and perform the necessary fixup to get the correct route installed.
173 	 */
174 	if (!srcrt && nd6_prproxy &&
175 	    (rt = ip6forward_rt->ro_rt) != NULL && (rt->rt_flags & RTF_PROXY)) {
176 		nd6_proxy_find_fwdroute(m->m_pkthdr.rcvif, ip6forward_rt);
177 		if ((rt = ip6forward_rt->ro_rt) != NULL) {
178 			ifscope = rt->rt_ifp->if_index;
179 		}
180 	}
181 
182 #if PF
183 	adjust_scope_and_pktlen(m, &ifscope, &mpktlen);
184 
185 	/*
186 	 * If the caller provides a route which is on a different interface
187 	 * than the one specified for scoped forwarding, discard the route
188 	 * and do a lookup below.
189 	 */
190 	if (ifscope != IFSCOPE_NONE && (rt = ip6forward_rt->ro_rt) != NULL) {
191 		RT_LOCK(rt);
192 		if (rt->rt_ifp->if_index != ifscope) {
193 			RT_UNLOCK(rt);
194 			ROUTE_RELEASE(ip6forward_rt);
195 			rt = NULL;
196 		} else {
197 			RT_UNLOCK(rt);
198 		}
199 	}
200 #endif /* PF */
201 
202 #if IPSEC
203 	/*
204 	 * Check AH/ESP integrity.
205 	 */
206 	/*
207 	 * Don't increment ip6s_cantforward because this is the check
208 	 * before forwarding packet actually.
209 	 */
210 	if (ipsec_bypass == 0) {
211 		if (ipsec6_in_reject(m, NULL)) {
212 			IPSEC_STAT_INCREMENT(ipsec6stat.in_polvio);
213 			m_freem(m);
214 			return NULL;
215 		}
216 	}
217 #endif /*IPSEC*/
218 
219 	/*
220 	 * Do not forward packets to multicast destination.
221 	 * Do not forward packets with unspecified source.  It was discussed
222 	 * in July 2000, on ipngwg mailing list.
223 	 */
224 	if ((m->m_flags & (M_BCAST | M_MCAST)) != 0 ||
225 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
226 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
227 		ip6stat.ip6s_cantforward++;
228 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
229 		if (ip6_log_time + ip6_log_interval < curtime) {
230 			ip6_log_time = curtime;
231 			log(LOG_DEBUG,
232 			    "cannot forward "
233 			    "from %s to %s nxt %d received on %s\n",
234 			    ip6_sprintf(&ip6->ip6_src),
235 			    ip6_sprintf(&ip6->ip6_dst),
236 			    ip6->ip6_nxt,
237 			    if_name(m->m_pkthdr.rcvif));
238 		}
239 		m_freem(m);
240 		return NULL;
241 	}
242 
243 	if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
244 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
245 		icmp6_error_flag(m, ICMP6_TIME_EXCEEDED,
246 		    ICMP6_TIME_EXCEED_TRANSIT, 0, 0);
247 		return NULL;
248 	}
249 
250 	/*
251 	 * See if the destination is a proxied address, and if so pretend
252 	 * that it's for us.  This is mostly to handle NUD probes against
253 	 * the proxied addresses.  We filter for ICMPv6 here and will let
254 	 * icmp6_input handle the rest.
255 	 */
256 	if (!srcrt && nd6_prproxy) {
257 		VERIFY(!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst));
258 		proxy = nd6_prproxy_isours(m, ip6, ip6forward_rt, ifscope);
259 		/*
260 		 * Don't update hop limit while proxying; RFC 4389 4.1.
261 		 * Also skip IPsec forwarding path processing as this
262 		 * packet is not to be forwarded.
263 		 */
264 		if (proxy) {
265 			goto skip_ipsec;
266 		}
267 	}
268 
269 	ip6->ip6_hlim -= IPV6_HLIMDEC;
270 
271 	/*
272 	 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
273 	 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
274 	 * we need to generate an ICMP6 message to the src.
275 	 * Thanks to M_EXT, in most cases copy will not occur.
276 	 *
277 	 * It is important to save it before IPsec processing as IPsec
278 	 * processing may modify the mbuf.
279 	 */
280 	mcopy = m_copym_mode(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN),
281 	    M_DONTWAIT, NULL, NULL, M_COPYM_COPY_HDR);
282 #if IPSEC
283 	if (ipsec_bypass != 0) {
284 		goto skip_ipsec;
285 	}
286 	/* get a security policy for this packet */
287 	sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, IP_FORWARDING,
288 	    &error);
289 	if (sp == NULL) {
290 		IPSEC_STAT_INCREMENT(ipsec6stat.out_inval);
291 		ip6stat.ip6s_cantforward++;
292 		if (mcopy) {
293 #if 0
294 			/* XXX: what icmp ? */
295 #else
296 			m_freem(mcopy);
297 #endif
298 		}
299 		m_freem(m);
300 		return NULL;
301 	}
302 
303 	error = 0;
304 
305 	/* check policy */
306 	switch (sp->policy) {
307 	case IPSEC_POLICY_DISCARD:
308 	case IPSEC_POLICY_GENERATE:
309 		/*
310 		 * This packet is just discarded.
311 		 */
312 		IPSEC_STAT_INCREMENT(ipsec6stat.out_polvio);
313 		ip6stat.ip6s_cantforward++;
314 		key_freesp(sp, KEY_SADB_UNLOCKED);
315 		if (mcopy) {
316 #if 0
317 			/* XXX: what icmp ? */
318 #else
319 			m_freem(mcopy);
320 #endif
321 		}
322 		m_freem(m);
323 		return NULL;
324 
325 	case IPSEC_POLICY_BYPASS:
326 	case IPSEC_POLICY_NONE:
327 		/* no need to do IPsec. */
328 		key_freesp(sp, KEY_SADB_UNLOCKED);
329 		goto skip_ipsec;
330 
331 	case IPSEC_POLICY_IPSEC:
332 		if (sp->req == NULL) {
333 			/* XXX should be panic ? */
334 			printf("ip6_forward: No IPsec request specified.\n");
335 			ip6stat.ip6s_cantforward++;
336 			key_freesp(sp, KEY_SADB_UNLOCKED);
337 			if (mcopy) {
338 #if 0
339 				/* XXX: what icmp ? */
340 #else
341 				m_freem(mcopy);
342 #endif
343 			}
344 			m_freem(m);
345 			return NULL;
346 		}
347 		/* do IPsec */
348 		break;
349 
350 	case IPSEC_POLICY_ENTRUST:
351 	default:
352 		/* should be panic ?? */
353 		printf("ip6_forward: Invalid policy found. %d\n", sp->policy);
354 		key_freesp(sp, KEY_SADB_UNLOCKED);
355 		goto skip_ipsec;
356 	}
357 
358 	{
359 		struct ipsec_output_state state;
360 
361 		/*
362 		 * All the extension headers will become inaccessible
363 		 * (since they can be encrypted).
364 		 * Don't panic, we need no more updates to extension headers
365 		 * on inner IPv6 packet (since they are now encapsulated).
366 		 *
367 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
368 		 */
369 		bzero(&state, sizeof(state));
370 		state.m = m;
371 		state.dst = NULL; /* update at ipsec6_output_tunnel() */
372 
373 		error = ipsec6_output_tunnel(&state, sp, 0);
374 		key_freesp(sp, KEY_SADB_UNLOCKED);
375 		if (state.tunneled == 4) {
376 			ROUTE_RELEASE(&state.ro);
377 			return NULL; /* packet is gone - sent over IPv4 */
378 		}
379 
380 		m = state.m;
381 		ROUTE_RELEASE(&state.ro);
382 
383 		if (error) {
384 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
385 			switch (error) {
386 			case EHOSTUNREACH:
387 			case ENETUNREACH:
388 			case EMSGSIZE:
389 			case ENOBUFS:
390 			case ENOMEM:
391 				break;
392 			default:
393 				printf("ip6_output (ipsec): error code %d\n", error);
394 				OS_FALLTHROUGH;
395 			case ENOENT:
396 				/* don't show these error codes to the user */
397 				break;
398 			}
399 			ip6stat.ip6s_cantforward++;
400 			if (mcopy) {
401 #if 0
402 				/* XXX: what icmp ? */
403 #else
404 				m_freem(mcopy);
405 #endif
406 			}
407 			m_freem(m);
408 			return NULL;
409 		}
410 	}
411 #endif /* IPSEC */
412 skip_ipsec:
413 
414 	dst = SIN6(&ip6forward_rt->ro_dst);
415 	if ((rt = ip6forward_rt->ro_rt) != NULL) {
416 		RT_LOCK(rt);
417 		/* Take an extra ref for ourselves */
418 		RT_ADDREF_LOCKED(rt);
419 	}
420 
421 	VERIFY(rt == NULL || rt == ip6forward_rt->ro_rt);
422 	if (!srcrt) {
423 		/*
424 		 * ip6forward_rt->ro_dst.sin6_addr is equal to ip6->ip6_dst
425 		 */
426 		if (ROUTE_UNUSABLE(ip6forward_rt)) {
427 			if (rt != NULL) {
428 				/* Release extra ref */
429 				RT_REMREF_LOCKED(rt);
430 				RT_UNLOCK(rt);
431 			}
432 			ROUTE_RELEASE(ip6forward_rt);
433 
434 			/* this probably fails but give it a try again */
435 			rtalloc_scoped_ign((struct route *)ip6forward_rt,
436 			    RTF_PRCLONING, ifscope);
437 			if ((rt = ip6forward_rt->ro_rt) != NULL) {
438 				RT_LOCK(rt);
439 				/* Take an extra ref for ourselves */
440 				RT_ADDREF_LOCKED(rt);
441 			}
442 		}
443 
444 		if (rt == NULL) {
445 			ip6stat.ip6s_noroute++;
446 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
447 			if (mcopy) {
448 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
449 				    ICMP6_DST_UNREACH_NOROUTE, 0);
450 			}
451 			m_freem(m);
452 			return NULL;
453 		}
454 		RT_LOCK_ASSERT_HELD(rt);
455 	} else if (ROUTE_UNUSABLE(ip6forward_rt) ||
456 	    !in6_are_addr_equal_scoped(&ip6->ip6_dst, &dst->sin6_addr, ip6_input_getdstifscope(m), dst->sin6_scope_id)) {
457 		if (rt != NULL) {
458 			/* Release extra ref */
459 			RT_REMREF_LOCKED(rt);
460 			RT_UNLOCK(rt);
461 		}
462 		ROUTE_RELEASE(ip6forward_rt);
463 
464 		SOCKADDR_ZERO(dst, sizeof(*dst));
465 		dst->sin6_len = sizeof(struct sockaddr_in6);
466 		dst->sin6_family = AF_INET6;
467 		dst->sin6_addr = ip6->ip6_dst;
468 
469 		rtalloc_scoped_ign((struct route *)ip6forward_rt,
470 		    RTF_PRCLONING, ifscope);
471 		if ((rt = ip6forward_rt->ro_rt) == NULL) {
472 			ip6stat.ip6s_noroute++;
473 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
474 			if (mcopy) {
475 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
476 				    ICMP6_DST_UNREACH_NOROUTE, 0);
477 			}
478 			m_freem(m);
479 			return NULL;
480 		}
481 		RT_LOCK(rt);
482 		/* Take an extra ref for ourselves */
483 		RT_ADDREF_LOCKED(rt);
484 	}
485 
486 	/*
487 	 * Source scope check: if a packet can't be delivered to its
488 	 * destination for the reason that the destination is beyond the scope
489 	 * of the source address, discard the packet and return an icmp6
490 	 * destination unreachable error with Code 2 (beyond scope of source
491 	 * address) unless we are proxying (source address is link local
492 	 * for NUDs.)  We use a local copy of ip6_src, since in6_setscope()
493 	 * will possibly modify its first argument.
494 	 * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1]
495 	 */
496 	src_in6 = ip6->ip6_src;
497 	if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) {
498 		RT_REMREF_LOCKED(rt);
499 		RT_UNLOCK(rt);
500 		/* XXX: this should not happen */
501 		ip6stat.ip6s_cantforward++;
502 		ip6stat.ip6s_badscope++;
503 		m_freem(m);
504 		return NULL;
505 	}
506 	if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) {
507 		RT_REMREF_LOCKED(rt);
508 		RT_UNLOCK(rt);
509 		ip6stat.ip6s_cantforward++;
510 		ip6stat.ip6s_badscope++;
511 		m_freem(m);
512 		return NULL;
513 	}
514 
515 	if (inzone != outzone && !proxy) {
516 		ip6stat.ip6s_cantforward++;
517 		ip6stat.ip6s_badscope++;
518 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
519 
520 		if (ip6_log_time + ip6_log_interval < curtime) {
521 			ip6_log_time = curtime;
522 			log(LOG_DEBUG,
523 			    "cannot forward "
524 			    "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
525 			    ip6_sprintf(&ip6->ip6_src),
526 			    ip6_sprintf(&ip6->ip6_dst),
527 			    ip6->ip6_nxt,
528 			    if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
529 		}
530 		/* Release extra ref */
531 		RT_REMREF_LOCKED(rt);
532 		RT_UNLOCK(rt);
533 		if (mcopy) {
534 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
535 			    ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
536 		}
537 		m_freem(m);
538 		return NULL;
539 	}
540 
541 	/*
542 	 * Destination scope check: if a packet is going to break the scope
543 	 * zone of packet's destination address, discard it.  This case should
544 	 * usually be prevented by appropriately-configured routing table, but
545 	 * we need an explicit check because we may mistakenly forward the
546 	 * packet to a different zone by (e.g.) a default route.
547 	 */
548 	dst_in6 = ip6->ip6_dst;
549 	if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 ||
550 	    in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 ||
551 	    inzone != outzone) {
552 		RT_REMREF_LOCKED(rt);
553 		RT_UNLOCK(rt);
554 		ip6stat.ip6s_cantforward++;
555 		ip6stat.ip6s_badscope++;
556 		m_freem(m);
557 		return NULL;
558 	}
559 
560 	if (mpktlen == 0) {
561 		mpktlen = m->m_pkthdr.len;
562 	}
563 
564 	if (mpktlen > rt->rt_ifp->if_mtu) {
565 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
566 		if (mcopy) {
567 			uint32_t mtu;
568 #if IPSEC
569 			struct secpolicy *sp2;
570 			int ipsecerror;
571 			size_t ipsechdrsiz;
572 #endif
573 
574 			mtu = rt->rt_ifp->if_mtu;
575 #if IPSEC
576 			/*
577 			 * When we do IPsec tunnel ingress, we need to play
578 			 * with the link value (decrement IPsec header size
579 			 * from mtu value).  The code is much simpler than v4
580 			 * case, as we have the outgoing interface for
581 			 * encapsulated packet as "rt->rt_ifp".
582 			 */
583 			sp2 = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
584 			    IP_FORWARDING, &ipsecerror);
585 			if (sp2) {
586 				ipsechdrsiz = ipsec6_hdrsiz(mcopy,
587 				    IPSEC_DIR_OUTBOUND, NULL);
588 				if (ipsechdrsiz < mtu) {
589 					mtu -= ipsechdrsiz;
590 				}
591 				key_freesp(sp2, KEY_SADB_UNLOCKED);
592 			}
593 			/*
594 			 * if mtu becomes less than minimum MTU,
595 			 * tell minimum MTU (and I'll need to fragment it).
596 			 */
597 			if (mtu < IPV6_MMTU) {
598 				mtu = IPV6_MMTU;
599 			}
600 #endif
601 			/* Release extra ref */
602 			RT_REMREF_LOCKED(rt);
603 			RT_UNLOCK(rt);
604 			icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
605 		} else {
606 			/* Release extra ref */
607 			RT_REMREF_LOCKED(rt);
608 			RT_UNLOCK(rt);
609 		}
610 		m_freem(m);
611 		return NULL;
612 	}
613 
614 	if (rt->rt_flags & RTF_GATEWAY) {
615 		dst = SIN6(rt->rt_gateway);
616 	}
617 
618 	/*
619 	 * If we are to forward the packet using the same interface
620 	 * as one we got the packet from, perhaps we should send a redirect
621 	 * to sender to shortcut a hop.
622 	 * Only send redirect if source is sending directly to us,
623 	 * and if packet was not source routed (or has any options).
624 	 * Also, don't send redirect if forwarding using a route
625 	 * modified by a redirect.
626 	 */
627 	if (!proxy &&
628 	    ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
629 	    (rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) == 0) {
630 		if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) {
631 			/*
632 			 * If the incoming interface is equal to the outgoing
633 			 * one, and the link attached to the interface is
634 			 * point-to-point, then it will be highly probable
635 			 * that a routing loop occurs. Thus, we immediately
636 			 * drop the packet and send an ICMPv6 error message.
637 			 *
638 			 * type/code is based on suggestion by Rich Draves.
639 			 * not sure if it is the best pick.
640 			 */
641 			RT_REMREF_LOCKED(rt);   /* Release extra ref */
642 			RT_UNLOCK(rt);
643 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
644 			    ICMP6_DST_UNREACH_ADDR, 0);
645 			m_freem(m);
646 			return NULL;
647 		}
648 		type = ND_REDIRECT;
649 	}
650 	/*
651 	 * Fake scoped addresses. Note that even link-local source or
652 	 * destinaion can appear, if the originating node just sends the
653 	 * packet to us (without address resolution for the destination).
654 	 * Since both icmp6_error and icmp6_redirect_output fill the embedded
655 	 * link identifiers, we can do this stuff after making a copy for
656 	 * returning an error.
657 	 */
658 	if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
659 		/*
660 		 * See corresponding comments in ip6_output.
661 		 * XXX: but is it possible that ip6_forward() sends a packet
662 		 *      to a loopback interface? I don't think so, and thus
663 		 *      I bark here. ([email protected])
664 		 * XXX: it is common to route invalid packets to loopback.
665 		 *	also, the codepath will be visited on use of ::1 in
666 		 *	rthdr. (itojun)
667 		 */
668 #if 1
669 		if ((0))
670 #else
671 		if ((rt->rt_flags & (RTF_BLACKHOLE | RTF_REJECT)) == 0)
672 #endif
673 		{
674 			printf("ip6_forward: outgoing interface is loopback. "
675 			    "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
676 			    ip6_sprintf(&ip6->ip6_src),
677 			    ip6_sprintf(&ip6->ip6_dst),
678 			    ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
679 			    if_name(rt->rt_ifp));
680 		}
681 
682 		/* we can just use rcvif in forwarding. */
683 		origifp = rcvifp = m->m_pkthdr.rcvif;
684 	} else if (nd6_prproxy) {
685 		/*
686 		 * In the prefix proxying case, we need to inform nd6_output()
687 		 * about the inbound interface, so that any subsequent NS
688 		 * packets generated by nd6_prproxy_ns_output() will not be
689 		 * sent back to that same interface.
690 		 */
691 		origifp = rcvifp = m->m_pkthdr.rcvif;
692 	} else {
693 		rcvifp = m->m_pkthdr.rcvif;
694 		origifp = rt->rt_ifp;
695 	}
696 	/*
697 	 * clear embedded scope identifiers if necessary.
698 	 * in6_clearscope will touch the addresses only when necessary.
699 	 */
700 	in6_clearscope(&ip6->ip6_src);
701 	in6_clearscope(&ip6->ip6_dst);
702 
703 	ifp = rt->rt_ifp;
704 	/* Drop the lock but retain the extra ref */
705 	RT_UNLOCK(rt);
706 
707 	/*
708 	 * If this is to be processed locally, let ip6_input have it.
709 	 */
710 	if (proxy) {
711 		VERIFY(m->m_pkthdr.pkt_flags & PKTF_PROXY_DST);
712 		/* Release extra ref */
713 		RT_REMREF(rt);
714 		if (mcopy != NULL) {
715 			m_freem(mcopy);
716 		}
717 		return m;
718 	}
719 
720 	/* Mark this packet as being forwarded from another interface */
721 	m->m_pkthdr.pkt_flags |= PKTF_FORWARDED;
722 
723 #if PF
724 	if (PF_IS_ENABLED) {
725 		/*
726 		 * PF refragments any packet which it reassembled due to scrub
727 		 * rules, in which case it will set the PF_TAG_REFRAGMENTED
728 		 * flag in PF mbuf tag.
729 		 */
730 #if DUMMYNET
731 		struct ip_fw_args args;
732 		struct pf_mtag *pf_mtag;
733 
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