xref: /xnu-11215/bsd/netinet6/ip6_forward.c (revision 8d741a5d)
1 /*
2  * Copyright (c) 2009-2024 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 #include <net/droptap.h>
77 
78 #include <netinet/in.h>
79 #include <netinet/in_var.h>
80 #include <netinet/in_systm.h>
81 #include <netinet/ip.h>
82 #include <netinet/ip_var.h>
83 #include <netinet6/in6_var.h>
84 #include <netinet/ip6.h>
85 #include <netinet6/ip6_var.h>
86 #include <netinet/icmp6.h>
87 #include <netinet6/nd6.h>
88 #include <netinet6/scope6_var.h>
89 
90 #include <netinet/in_pcb.h>
91 
92 #if IPSEC
93 #include <netinet6/ipsec.h>
94 #include <netinet6/ipsec6.h>
95 #include <netkey/key.h>
96 extern int ipsec_bypass;
97 #endif /* IPSEC */
98 
99 #include <net/net_osdep.h>
100 
101 #if DUMMYNET
102 #include <netinet/ip_dummynet.h>
103 #endif /* DUMMYNET */
104 
105 #if PF
106 #include <net/pfvar.h>
107 static void
adjust_scope_and_pktlen(struct mbuf * m,unsigned int * ifscope_p,uint32_t * mpktlen_p)108 adjust_scope_and_pktlen(struct mbuf *m,
109     unsigned int *ifscope_p, uint32_t *mpktlen_p)
110 {
111 	struct pf_mtag *__single pf_mtag;
112 	struct pf_fragment_tag *__single pf_ftagp;
113 
114 	pf_mtag = pf_find_mtag(m);
115 	ASSERT(pf_mtag != NULL);
116 	if (pf_mtag->pftag_rtableid != IFSCOPE_NONE) {
117 		*ifscope_p = pf_mtag->pftag_rtableid;
118 	}
119 	pf_ftagp = pf_find_fragment_tag(m);
120 	if (pf_ftagp != NULL) {
121 		ASSERT(pf_mtag->pftag_flags & PF_TAG_REASSEMBLED);
122 		*mpktlen_p = pf_ftagp->ft_maxlen;
123 		ASSERT(*mpktlen_p);
124 	}
125 }
126 
127 #endif /* PF */
128 
129 #include <net/sockaddr_utils.h>
130 
131 /*
132  * Forward a packet.  If some error occurs return the sender
133  * an icmp packet.  Note we can't always generate a meaningful
134  * icmp message because icmp doesn't have a large enough repertoire
135  * of codes and types.
136  *
137  * If not forwarding, just drop the packet.  This could be confusing
138  * if ipforwarding was zero but some routing protocol was advancing
139  * us as a gateway to somewhere.  However, we must let the routing
140  * protocol deal with that.
141  *
142  */
143 
144 struct mbuf *
ip6_forward(struct mbuf * m,struct route_in6 * ip6forward_rt,int srcrt)145 ip6_forward(struct mbuf *m, struct route_in6 *ip6forward_rt,
146     int srcrt)
147 {
148 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
149 	struct sockaddr_in6 *__single dst;
150 	struct rtentry *__single rt;
151 	int error, type = 0, code = 0;
152 	boolean_t proxy = FALSE;
153 	struct mbuf *mcopy = NULL;
154 	struct ifnet *__single ifp, *__single rcvifp, *__single origifp;
155 	u_int32_t inzone, outzone, len = 0, pktcnt = 0;
156 	struct in6_addr src_in6, dst_in6;
157 	uint64_t curtime = net_uptime();
158 #if IPSEC
159 	struct secpolicy *__single sp = NULL;
160 #endif
161 	unsigned int ifscope = IFSCOPE_NONE;
162 	uint32_t mpktlen = 0;
163 
164 	/*
165 	 * In the prefix proxying case, the route to the proxied node normally
166 	 * gets created by nd6_prproxy_ns_output(), as part of forwarding a
167 	 * NS (NUD/AR) packet to the proxied node.  In the event that such
168 	 * packet did not arrive in time before the correct route gets created,
169 	 * ip6_input() would have performed a rtalloc() which most likely will
170 	 * create the wrong cloned route; this route points back to the same
171 	 * interface as the inbound interface, since the parent non-scoped
172 	 * prefix route points there.  Therefore we check if that is the case
173 	 * and perform the necessary fixup to get the correct route installed.
174 	 */
175 	if (!srcrt && nd6_prproxy &&
176 	    (rt = ip6forward_rt->ro_rt) != NULL && (rt->rt_flags & RTF_PROXY)) {
177 		nd6_proxy_find_fwdroute(m->m_pkthdr.rcvif, ip6forward_rt);
178 		if ((rt = ip6forward_rt->ro_rt) != NULL) {
179 			ifscope = rt->rt_ifp->if_index;
180 		}
181 	}
182 
183 #if PF
184 	adjust_scope_and_pktlen(m, &ifscope, &mpktlen);
185 
186 	/*
187 	 * If the caller provides a route which is on a different interface
188 	 * than the one specified for scoped forwarding, discard the route
189 	 * and do a lookup below.
190 	 */
191 	if (ifscope != IFSCOPE_NONE && (rt = ip6forward_rt->ro_rt) != NULL) {
192 		RT_LOCK(rt);
193 		if (rt->rt_ifp->if_index != ifscope) {
194 			RT_UNLOCK(rt);
195 			ROUTE_RELEASE(ip6forward_rt);
196 			rt = NULL;
197 		} else {
198 			RT_UNLOCK(rt);
199 		}
200 	}
201 #endif /* PF */
202 
203 #if IPSEC
204 	/*
205 	 * Check AH/ESP integrity.
206 	 */
207 	/*
208 	 * Don't increment ip6s_cantforward because this is the check
209 	 * before forwarding packet actually.
210 	 */
211 	if (ipsec_bypass == 0) {
212 		if (ipsec6_in_reject(m, NULL)) {
213 			IPSEC_STAT_INCREMENT(ipsec6stat.in_polvio);
214 			m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IPSEC_REJECT, NULL, 0);
215 			return NULL;
216 		}
217 	}
218 #endif /*IPSEC*/
219 
220 	/*
221 	 * Do not forward packets to multicast destination.
222 	 * Do not forward packets with unspecified source.  It was discussed
223 	 * in July 2000, on ipngwg mailing list.
224 	 */
225 	if ((m->m_flags & (M_BCAST | M_MCAST)) != 0 ||
226 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
227 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
228 		ip6stat.ip6s_cantforward++;
229 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
230 		if (ip6_log_time + ip6_log_interval < curtime) {
231 			ip6_log_time = curtime;
232 			log(LOG_DEBUG,
233 			    "cannot forward "
234 			    "from %s to %s nxt %d received on %s\n",
235 			    ip6_sprintf(&ip6->ip6_src),
236 			    ip6_sprintf(&ip6->ip6_dst),
237 			    ip6->ip6_nxt,
238 			    if_name(m->m_pkthdr.rcvif));
239 		}
240 		m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_CANNOT_FORWARD, NULL, 0);
241 		return NULL;
242 	}
243 
244 	if (ip6->ip6_hlim <= IPV6_HLIMDEC) {
245 		/* XXX in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard) */
246 		icmp6_error_flag(m, ICMP6_TIME_EXCEEDED,
247 		    ICMP6_TIME_EXCEED_TRANSIT, 0, 0);
248 		return NULL;
249 	}
250 
251 	/*
252 	 * See if the destination is a proxied address, and if so pretend
253 	 * that it's for us.  This is mostly to handle NUD probes against
254 	 * the proxied addresses.  We filter for ICMPv6 here and will let
255 	 * icmp6_input handle the rest.
256 	 */
257 	if (!srcrt && nd6_prproxy) {
258 		VERIFY(!IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst));
259 		proxy = nd6_prproxy_isours(m, ip6, ip6forward_rt, ifscope);
260 		/*
261 		 * Don't update hop limit while proxying; RFC 4389 4.1.
262 		 * Also skip IPsec forwarding path processing as this
263 		 * packet is not to be forwarded.
264 		 */
265 		if (proxy) {
266 			goto skip_ipsec;
267 		}
268 	}
269 
270 	ip6->ip6_hlim -= IPV6_HLIMDEC;
271 
272 	/*
273 	 * Save at most ICMPV6_PLD_MAXLEN (= the min IPv6 MTU -
274 	 * size of IPv6 + ICMPv6 headers) bytes of the packet in case
275 	 * we need to generate an ICMP6 message to the src.
276 	 * Thanks to M_EXT, in most cases copy will not occur.
277 	 *
278 	 * It is important to save it before IPsec processing as IPsec
279 	 * processing may modify the mbuf.
280 	 */
281 	mcopy = m_copym_mode(m, 0, imin(m->m_pkthdr.len, ICMPV6_PLD_MAXLEN),
282 	    M_DONTWAIT, NULL, NULL, M_COPYM_COPY_HDR);
283 #if IPSEC
284 	if (ipsec_bypass != 0) {
285 		goto skip_ipsec;
286 	}
287 	/* get a security policy for this packet */
288 	sp = ipsec6_getpolicybyaddr(m, IPSEC_DIR_OUTBOUND, IP_FORWARDING,
289 	    &error);
290 	if (sp == NULL) {
291 		IPSEC_STAT_INCREMENT(ipsec6stat.out_inval);
292 		ip6stat.ip6s_cantforward++;
293 		if (mcopy) {
294 #if 0
295 			/* XXX: what icmp ? */
296 #else
297 			m_freem(mcopy);
298 #endif
299 		}
300 		m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_CANNOT_FORWARD, NULL, 0);
301 		return NULL;
302 	}
303 
304 	error = 0;
305 
306 	/* check policy */
307 	switch (sp->policy) {
308 	case IPSEC_POLICY_DISCARD:
309 	case IPSEC_POLICY_GENERATE:
310 		/*
311 		 * This packet is just discarded.
312 		 */
313 		IPSEC_STAT_INCREMENT(ipsec6stat.out_polvio);
314 		ip6stat.ip6s_cantforward++;
315 		key_freesp(sp, KEY_SADB_UNLOCKED);
316 		if (mcopy) {
317 #if 0
318 			/* XXX: what icmp ? */
319 #else
320 			m_freem(mcopy);
321 #endif
322 		}
323 		m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_CANNOT_FORWARD, NULL, 0); // or IPPOLICY ?
324 		return NULL;
325 
326 	case IPSEC_POLICY_BYPASS:
327 	case IPSEC_POLICY_NONE:
328 		/* no need to do IPsec. */
329 		key_freesp(sp, KEY_SADB_UNLOCKED);
330 		goto skip_ipsec;
331 
332 	case IPSEC_POLICY_IPSEC:
333 		if (sp->req == NULL) {
334 			/* XXX should be panic ? */
335 			printf("ip6_forward: No IPsec request specified.\n");
336 			ip6stat.ip6s_cantforward++;
337 			key_freesp(sp, KEY_SADB_UNLOCKED);
338 			if (mcopy) {
339 #if 0
340 				/* XXX: what icmp ? */
341 #else
342 				m_freem(mcopy);
343 #endif
344 			}
345 			m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_CANNOT_FORWARD, NULL, 0); // or IPPOLICY ?
346 			return NULL;
347 		}
348 		/* do IPsec */
349 		break;
350 
351 	case IPSEC_POLICY_ENTRUST:
352 	default:
353 		/* should be panic ?? */
354 		printf("ip6_forward: Invalid policy found. %d\n", sp->policy);
355 		key_freesp(sp, KEY_SADB_UNLOCKED);
356 		goto skip_ipsec;
357 	}
358 
359 	{
360 		struct ipsec_output_state state;
361 
362 		/*
363 		 * All the extension headers will become inaccessible
364 		 * (since they can be encrypted).
365 		 * Don't panic, we need no more updates to extension headers
366 		 * on inner IPv6 packet (since they are now encapsulated).
367 		 *
368 		 * IPv6 [ESP|AH] IPv6 [extension headers] payload
369 		 */
370 		bzero(&state, sizeof(state));
371 		state.m = m;
372 		state.dst = NULL; /* update at ipsec6_output_tunnel() */
373 
374 		error = ipsec6_output_tunnel(&state, sp, 0);
375 		key_freesp(sp, KEY_SADB_UNLOCKED);
376 		if (state.tunneled == 4) {
377 			ROUTE_RELEASE(&state.ro);
378 			return NULL; /* packet is gone - sent over IPv4 */
379 		}
380 
381 		m = state.m;
382 		ROUTE_RELEASE(&state.ro);
383 
384 		if (error) {
385 			/* mbuf is already reclaimed in ipsec6_output_tunnel. */
386 			switch (error) {
387 			case EHOSTUNREACH:
388 			case ENETUNREACH:
389 			case EMSGSIZE:
390 			case ENOBUFS:
391 			case ENOMEM:
392 				break;
393 			default:
394 				printf("ip6_output (ipsec): error code %d\n", error);
395 				OS_FALLTHROUGH;
396 			case ENOENT:
397 				/* don't show these error codes to the user */
398 				break;
399 			}
400 			ip6stat.ip6s_cantforward++;
401 			if (mcopy) {
402 #if 0
403 				/* XXX: what icmp ? */
404 #else
405 				m_freem(mcopy);
406 #endif
407 			}
408 			m_freem(m);
409 			return NULL;
410 		}
411 	}
412 #endif /* IPSEC */
413 skip_ipsec:
414 
415 	dst = SIN6(&ip6forward_rt->ro_dst);
416 	if ((rt = ip6forward_rt->ro_rt) != NULL) {
417 		RT_LOCK(rt);
418 		/* Take an extra ref for ourselves */
419 		RT_ADDREF_LOCKED(rt);
420 	}
421 
422 	VERIFY(rt == NULL || rt == ip6forward_rt->ro_rt);
423 	if (!srcrt) {
424 		/*
425 		 * ip6forward_rt->ro_dst.sin6_addr is equal to ip6->ip6_dst
426 		 */
427 		if (ROUTE_UNUSABLE(ip6forward_rt)) {
428 			if (rt != NULL) {
429 				/* Release extra ref */
430 				RT_REMREF_LOCKED(rt);
431 				RT_UNLOCK(rt);
432 			}
433 			ROUTE_RELEASE(ip6forward_rt);
434 
435 			/* this probably fails but give it a try again */
436 			rtalloc_scoped_ign((struct route *)ip6forward_rt,
437 			    RTF_PRCLONING, ifscope);
438 			if ((rt = ip6forward_rt->ro_rt) != NULL) {
439 				RT_LOCK(rt);
440 				/* Take an extra ref for ourselves */
441 				RT_ADDREF_LOCKED(rt);
442 			}
443 		}
444 
445 		if (rt == NULL) {
446 			ip6stat.ip6s_noroute++;
447 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
448 			if (mcopy) {
449 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
450 				    ICMP6_DST_UNREACH_NOROUTE, 0);
451 			}
452 			m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_NO_ROUTE, NULL, 0);
453 			return NULL;
454 		}
455 		RT_LOCK_ASSERT_HELD(rt);
456 	} else if (ROUTE_UNUSABLE(ip6forward_rt) ||
457 	    !in6_are_addr_equal_scoped(&ip6->ip6_dst, &dst->sin6_addr, ip6_input_getdstifscope(m), dst->sin6_scope_id)) {
458 		if (rt != NULL) {
459 			/* Release extra ref */
460 			RT_REMREF_LOCKED(rt);
461 			RT_UNLOCK(rt);
462 		}
463 		ROUTE_RELEASE(ip6forward_rt);
464 
465 		SOCKADDR_ZERO(dst, sizeof(*dst));
466 		dst->sin6_len = sizeof(struct sockaddr_in6);
467 		dst->sin6_family = AF_INET6;
468 		dst->sin6_addr = ip6->ip6_dst;
469 
470 		rtalloc_scoped_ign((struct route *)ip6forward_rt,
471 		    RTF_PRCLONING, ifscope);
472 		if ((rt = ip6forward_rt->ro_rt) == NULL) {
473 			ip6stat.ip6s_noroute++;
474 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_noroute);
475 			if (mcopy) {
476 				icmp6_error(mcopy, ICMP6_DST_UNREACH,
477 				    ICMP6_DST_UNREACH_NOROUTE, 0);
478 			}
479 			m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP_NO_ROUTE, NULL, 0);
480 			return NULL;
481 		}
482 		RT_LOCK(rt);
483 		/* Take an extra ref for ourselves */
484 		RT_ADDREF_LOCKED(rt);
485 	}
486 
487 	/*
488 	 * Source scope check: if a packet can't be delivered to its
489 	 * destination for the reason that the destination is beyond the scope
490 	 * of the source address, discard the packet and return an icmp6
491 	 * destination unreachable error with Code 2 (beyond scope of source
492 	 * address) unless we are proxying (source address is link local
493 	 * for NUDs.)  We use a local copy of ip6_src, since in6_setscope()
494 	 * will possibly modify its first argument.
495 	 * [draft-ietf-ipngwg-icmp-v3-04.txt, Section 3.1]
496 	 */
497 	src_in6 = ip6->ip6_src;
498 	if (in6_setscope(&src_in6, rt->rt_ifp, &outzone)) {
499 		RT_REMREF_LOCKED(rt);
500 		RT_UNLOCK(rt);
501 		/* XXX: this should not happen */
502 		ip6stat.ip6s_cantforward++;
503 		ip6stat.ip6s_badscope++;
504 		m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
505 		return NULL;
506 	}
507 	if (in6_setscope(&src_in6, m->m_pkthdr.rcvif, &inzone)) {
508 		RT_REMREF_LOCKED(rt);
509 		RT_UNLOCK(rt);
510 		ip6stat.ip6s_cantforward++;
511 		ip6stat.ip6s_badscope++;
512 		m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
513 		return NULL;
514 	}
515 
516 	if (inzone != outzone && !proxy) {
517 		ip6stat.ip6s_cantforward++;
518 		ip6stat.ip6s_badscope++;
519 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_discard);
520 
521 		if (ip6_log_time + ip6_log_interval < curtime) {
522 			ip6_log_time = curtime;
523 			log(LOG_DEBUG,
524 			    "cannot forward "
525 			    "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
526 			    ip6_sprintf(&ip6->ip6_src),
527 			    ip6_sprintf(&ip6->ip6_dst),
528 			    ip6->ip6_nxt,
529 			    if_name(m->m_pkthdr.rcvif), if_name(rt->rt_ifp));
530 		}
531 		/* Release extra ref */
532 		RT_REMREF_LOCKED(rt);
533 		RT_UNLOCK(rt);
534 		if (mcopy) {
535 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
536 			    ICMP6_DST_UNREACH_BEYONDSCOPE, 0);
537 		}
538 		m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
539 		return NULL;
540 	}
541 
542 	/*
543 	 * Destination scope check: if a packet is going to break the scope
544 	 * zone of packet's destination address, discard it.  This case should
545 	 * usually be prevented by appropriately-configured routing table, but
546 	 * we need an explicit check because we may mistakenly forward the
547 	 * packet to a different zone by (e.g.) a default route.
548 	 */
549 	dst_in6 = ip6->ip6_dst;
550 	if (in6_setscope(&dst_in6, m->m_pkthdr.rcvif, &inzone) != 0 ||
551 	    in6_setscope(&dst_in6, rt->rt_ifp, &outzone) != 0 ||
552 	    inzone != outzone) {
553 		RT_REMREF_LOCKED(rt);
554 		RT_UNLOCK(rt);
555 		ip6stat.ip6s_cantforward++;
556 		ip6stat.ip6s_badscope++;
557 		m_drop(m, DROPTAP_FLAG_DIR_IN | DROPTAP_FLAG_L2_MISSING, DROP_REASON_IP6_BAD_SCOPE, NULL, 0);
558 		return NULL;
559 	}
560 
561 	if (mpktlen == 0) {
562 		mpktlen = m->m_pkthdr.len;
563 	}
564 
565 	if (mpktlen > rt->rt_ifp->if_mtu) {
566 		in6_ifstat_inc(rt->rt_ifp, ifs6_in_toobig);
567 		if (mcopy) {
568 			uint32_t mtu;
569 #if IPSEC
570 			struct secpolicy *__single sp2;
571 			int ipsecerror;
572 			size_t ipsechdrsiz;
573 #endif
574 
575 			mtu = rt->rt_ifp->if_mtu;
576 #if IPSEC
577 			/*
578 			 * When we do IPsec tunnel ingress, we need to play
579 			 * with the link value (decrement IPsec header size
580 			 * from mtu value).  The code is much simpler than v4
581 			 * case, as we have the outgoing interface for
582 			 * encapsulated packet as "rt->rt_ifp".
583 			 */
584 			sp2 = ipsec6_getpolicybyaddr(mcopy, IPSEC_DIR_OUTBOUND,
585 			    IP_FORWARDING, &ipsecerror);
586 			if (sp2) {
587 				ipsechdrsiz = ipsec6_hdrsiz(mcopy,
588 				    IPSEC_DIR_OUTBOUND, NULL);
589 				if (ipsechdrsiz < mtu) {
590 					mtu -= ipsechdrsiz;
591 				}
592 				key_freesp(sp2, KEY_SADB_UNLOCKED);
593 			}
594 			/*
595 			 * if mtu becomes less than minimum MTU,
596 			 * tell minimum MTU (and I'll need to fragment it).
597 			 */
598 			if (mtu < IPV6_MMTU) {
599 				mtu = IPV6_MMTU;
600 			}
601 #endif
602 			/* Release extra ref */
603 			RT_REMREF_LOCKED(rt);
604 			RT_UNLOCK(rt);
605 			icmp6_error(mcopy, ICMP6_PACKET_TOO_BIG, 0, mtu);
606 		} else {
607 			/* Release extra ref */
608 			RT_REMREF_LOCKED(rt);
609 			RT_UNLOCK(rt);
610 		}
611 		m_freem(m);
612 		return NULL;
613 	}
614 
615 	if (rt->rt_flags & RTF_GATEWAY) {
616 		dst = SIN6(rt->rt_gateway);
617 	}
618 
619 	/*
620 	 * If we are to forward the packet using the same interface
621 	 * as one we got the packet from, perhaps we should send a redirect
622 	 * to sender to shortcut a hop.
623 	 * Only send redirect if source is sending directly to us,
624 	 * and if packet was not source routed (or has any options).
625 	 * Also, don't send redirect if forwarding using a route
626 	 * modified by a redirect.
627 	 */
628 	if (!proxy &&
629 	    ip6_sendredirects && rt->rt_ifp == m->m_pkthdr.rcvif && !srcrt &&
630 	    (rt->rt_flags & (RTF_DYNAMIC | RTF_MODIFIED)) == 0) {
631 		if ((rt->rt_ifp->if_flags & IFF_POINTOPOINT) != 0) {
632 			/*
633 			 * If the incoming interface is equal to the outgoing
634 			 * one, and the link attached to the interface is
635 			 * point-to-point, then it will be highly probable
636 			 * that a routing loop occurs. Thus, we immediately
637 			 * drop the packet and send an ICMPv6 error message.
638 			 *
639 			 * type/code is based on suggestion by Rich Draves.
640 			 * not sure if it is the best pick.
641 			 */
642 			RT_REMREF_LOCKED(rt);   /* Release extra ref */
643 			RT_UNLOCK(rt);
644 			icmp6_error(mcopy, ICMP6_DST_UNREACH,
645 			    ICMP6_DST_UNREACH_ADDR, 0);
646 			m_freem(m);
647 			return NULL;
648 		}
649 		type = ND_REDIRECT;
650 	}
651 	/*
652 	 * Fake scoped addresses. Note that even link-local source or
653 	 * destinaion can appear, if the originating node just sends the
654 	 * packet to us (without address resolution for the destination).
655 	 * Since both icmp6_error and icmp6_redirect_output fill the embedded
656 	 * link identifiers, we can do this stuff after making a copy for
657 	 * returning an error.
658 	 */
659 	if ((rt->rt_ifp->if_flags & IFF_LOOPBACK) != 0) {
660 		/*
661 		 * See corresponding comments in ip6_output.
662 		 * XXX: but is it possible that ip6_forward() sends a packet
663 		 *      to a loopback interface? I don't think so, and thus
664 		 *      I bark here. ([email protected])
665 		 * XXX: it is common to route invalid packets to loopback.
666 		 *	also, the codepath will be visited on use of ::1 in
667 		 *	rthdr. (itojun)
668 		 */
669 #if 1
670 		if ((0))
671 #else
672 		if ((rt->rt_flags & (RTF_BLACKHOLE | RTF_REJECT)) == 0)
673 #endif
674 		{
675 			printf("ip6_forward: outgoing interface is loopback. "
676 			    "src %s, dst %s, nxt %d, rcvif %s, outif %s\n",
677 			    ip6_sprintf(&ip6->ip6_src),
678 			    ip6_sprintf(&ip6->ip6_dst),
679 			    ip6->ip6_nxt, if_name(m->m_pkthdr.rcvif),
680 			    if_name(rt->rt_ifp));
681 		}
682 
683 		/* we can just use rcvif in forwarding. */
684 		origifp = rcvifp = m->m_pkthdr.rcvif;
685 	} else if (nd6_prproxy) {
686 		/*
687 		 * In the prefix proxying case, we need to inform nd6_output()
688 		 * about the inbound interface, so that any subsequent NS
689 		 * packets generated by nd6_prproxy_ns_output() will not be
690 		 * sent back to that same interface.
691 		 */
692 		origifp = rcvifp = m->m_pkthdr.rcvif;
693 	} else {
694 		rcvifp = m->m_pkthdr.rcvif;
695 		origifp = rt->rt_ifp;
696 	}
697 	/*
698 	 * clear embedded scope identifiers if necessary.
699 	 * in6_clearscope will touch the addresses only when necessary.
700 	 */
701 	in6_clearscope(&ip6->ip6_src);
702 	in6_clearscope(&ip6->ip6_dst);
703 
704 	ifp = rt->rt_ifp;
705 	/* Drop the lock but retain the extra ref */
706 	RT_UNLOCK(rt);
707 
708 	/*
709 	 * If this is to be processed locally, let ip6_input have it.
710 	 */
711 	if (proxy) {
712 		VERIFY(m->m_pkthdr.pkt_flags & PKTF_PROXY_DST);
713 		/* Release extra ref */
714 		RT_REMREF(rt);
715 		if (mcopy != NULL) {
716 			m_freem(mcopy);
717 		}
718 		return m;
719 	}
720 
721 	/* Mark this packet as being forwarded from another interface */
722 	m->m_pkthdr.pkt_flags |= PKTF_FORWARDED;
723 
724 #if PF
725 	if (PF_IS_ENABLED) {
726 		/*
727 		 * PF refragments any packet which it reassembled due to scrub
728 		 * rules, in which case it will set the PF_TAG_REFRAGMENTED
729 		 * flag in PF mbuf tag.
730 		 */
731 #if DUMMYNET
732 		struct ip_fw_args args;
733 		struct pf_mtag *__single pf_mtag;
734 
735 		bzero(&args, sizeof(args));
736 
737 		args.fwa_oif = ifp;
738 		args.fwa_oflags = 0;
739 		args.fwa_ro6 = ip6forward_rt;
740 		args.fwa_ro6_pmtu = ip6forward_rt;
741 		args.fwa_mtu = rt->rt_ifp->if_mtu;
742 		args.fwa_dst6 = dst;
743 		args.fwa_origifp = origifp;
744 		/* Invoke outbound packet filter */
745 		error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, &args);
746 #else /* !DUMMYNET */
747 		error = pf_af_hook(ifp, NULL, &m, AF_INET6, FALSE, NULL);
748 #endif /* !DUMMYNET */
749 		if (error != 0 || m == NULL) {
750 			if (m != NULL) {
751 				panic("%s: unexpected packet %p", __func__, m);
752 				/* NOTREACHED */
753 			}
754 			/* Already freed by callee */
755 			goto senderr;
756 		}
757 
758 		pf_mtag = pf_find_mtag(m);
759 		/*
760 		 * refragmented packets from PF.
761 		 */
762 		if ((pf_mtag->pftag_flags & PF_TAG_REFRAGMENTED) != 0) {
763 			struct mbuf *__single t;
764 
765 			pf_mtag->pftag_flags &= ~PF_TAG_REFRAGMENTED;
766 			/* for statistics */
767 			t = m;
768 			while (t != NULL) {
769 				pktcnt++;
770 				len += m_pktlen(t);
771 				t = t->m_nextpkt;
772 			}
773 
774 			/*
775 			 * nd6_output() frees packetchain in both success and
776 			 * failure cases.
777 			 */
778 			error = nd6_output(ifp, origifp, m, dst, rt, NULL);
779 			m = NULL;
780 			goto sent;
781 		}
782 		/*
783 		 * We do not use ip6 header again in the code below,
784 		 * however still adding the bit here so that any new
785 		 * code in future doesn't end up working with the
786 		 * wrong pointer
787 		 */
788 		ip6 = mtod(m, struct ip6_hdr *);
789 	}
790 #endif /* PF */
791 
792 	len = m_pktlen(m);
793 	pktcnt = 1;
794 	error = nd6_output(ifp, origifp, m, dst, rt, NULL);
795 sent:
796 	if (error) {
797 		in6_ifstat_add(ifp, ifs6_out_discard, pktcnt);
798 		ip6stat.ip6s_cantforward += pktcnt;
799 	} else {
800 		/*
801 		 * Increment stats on the source interface; the ones
802 		 * for destination interface has been taken care of
803 		 * during output above by virtue of PKTF_FORWARDED.
804 		 */
805 		rcvifp->if_fpackets += pktcnt;
806 		rcvifp->if_fbytes += len;
807 
808 		ip6stat.ip6s_forward += pktcnt;
809 		in6_ifstat_add(ifp, ifs6_out_forward, pktcnt);
810 		if (type) {
811 			ip6stat.ip6s_redirectsent++;
812 		} else {
813 			if (mcopy) {
814 				goto freecopy;
815 			}
816 		}
817 	}
818 #if PF
819 senderr:
820 #endif /* PF */
821 	if (mcopy == NULL) {
822 		/* Release extra ref */
823 		RT_REMREF(rt);
824 		return NULL;
825 	}
826 	switch (error) {
827 	case 0:
828 #if 1
829 		if (type == ND_REDIRECT) {
830 			icmp6_redirect_output(mcopy, rt);
831 			/* Release extra ref */
832 			RT_REMREF(rt);
833 			return NULL;
834 		}
835 #endif
836 		goto freecopy;
837 
838 	case EMSGSIZE:
839 		/* xxx MTU is constant in PPP? */
840 		goto freecopy;
841 
842 	case ENOBUFS:
843 		/* Tell source to slow down like source quench in IP? */
844 		goto freecopy;
845 
846 	case ENETUNREACH:       /* shouldn't happen, checked above */
847 	case EHOSTUNREACH:
848 	case ENETDOWN:
849 	case EHOSTDOWN:
850 	default:
851 		type = ICMP6_DST_UNREACH;
852 		code = ICMP6_DST_UNREACH_ADDR;
853 		break;
854 	}
855 	icmp6_error(mcopy, type, code, 0);
856 	/* Release extra ref */
857 	RT_REMREF(rt);
858 	return NULL;
859 
860 freecopy:
861 	m_freem(mcopy);
862 	/* Release extra ref */
863 	RT_REMREF(rt);
864 	return NULL;
865 }
866