xref: /openbsd-src/sys/netinet6/ip6_input.c (revision 1ad61ae0a79a724d2d3ec69e69c8e1d1ff6b53a0)
1 /*	$OpenBSD: ip6_input.c,v 1.256 2023/09/16 09:33:27 mpi Exp $	*/
2 /*	$KAME: ip6_input.c,v 1.188 2001/03/29 05:34:31 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 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include "pf.h"
65 #include "carp.h"
66 
67 #include <sys/param.h>
68 #include <sys/systm.h>
69 #include <sys/mbuf.h>
70 #include <sys/domain.h>
71 #include <sys/sysctl.h>
72 #include <sys/protosw.h>
73 #include <sys/socket.h>
74 #include <sys/socketvar.h>
75 #include <sys/errno.h>
76 #include <sys/time.h>
77 #include <sys/timeout.h>
78 #include <sys/kernel.h>
79 #include <sys/syslog.h>
80 #include <sys/task.h>
81 
82 #include <net/if.h>
83 #include <net/if_var.h>
84 #include <net/if_types.h>
85 #include <net/route.h>
86 #include <net/netisr.h>
87 
88 #include <netinet/in.h>
89 
90 #include <netinet/ip.h>
91 
92 #include <netinet/in_pcb.h>
93 #include <netinet/ip_var.h>
94 #include <netinet6/in6_var.h>
95 #include <netinet6/in6_ifattach.h>
96 #include <netinet/ip6.h>
97 #include <netinet6/ip6_var.h>
98 #include <netinet/icmp6.h>
99 #include <netinet6/nd6.h>
100 
101 #include "gif.h"
102 #include "bpfilter.h"
103 
104 #ifdef MROUTING
105 #include <netinet6/ip6_mroute.h>
106 #endif
107 
108 #if NPF > 0
109 #include <net/pfvar.h>
110 #endif
111 
112 #if NCARP > 0
113 #include <netinet/ip_carp.h>
114 #endif
115 
116 struct niqueue ip6intrq = NIQUEUE_INITIALIZER(IPQ_MAXLEN, NETISR_IPV6);
117 
118 struct cpumem *ip6counters;
119 
120 uint8_t ip6_soiikey[IP6_SOIIKEY_LEN];
121 
122 int ip6_ours(struct mbuf **, int *, int, int);
123 int ip6_check_rh0hdr(struct mbuf *, int *);
124 int ip6_hbhchcheck(struct mbuf **, int *, int *);
125 int ip6_hopopts_input(struct mbuf **, int *, u_int32_t *, u_int32_t *);
126 struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
127 int ip6_sysctl_soiikey(void *, size_t *, void *, size_t);
128 
129 static struct mbuf_queue	ip6send_mq;
130 
131 static void ip6_send_dispatch(void *);
132 static struct task ip6send_task =
133 	TASK_INITIALIZER(ip6_send_dispatch, &ip6send_mq);
134 
135 /*
136  * IP6 initialization: fill in IP6 protocol switch table.
137  * All protocols not implemented in kernel go to raw IP6 protocol handler.
138  */
139 void
140 ip6_init(void)
141 {
142 	const struct protosw *pr;
143 	int i;
144 
145 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
146 	if (pr == NULL)
147 		panic("%s", __func__);
148 	for (i = 0; i < IPPROTO_MAX; i++)
149 		ip6_protox[i] = pr - inet6sw;
150 	for (pr = inet6domain.dom_protosw;
151 	    pr < inet6domain.dom_protoswNPROTOSW; pr++)
152 		if (pr->pr_domain->dom_family == PF_INET6 &&
153 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW &&
154 		    pr->pr_protocol < IPPROTO_MAX)
155 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
156 	ip6_randomid_init();
157 	nd6_init();
158 	frag6_init();
159 
160 	mq_init(&ip6send_mq, 64, IPL_SOFTNET);
161 
162 	ip6counters = counters_alloc(ip6s_ncounters);
163 #ifdef MROUTING
164 	rt_timer_queue_init(&ip6_mrouterq, MCAST_EXPIRE_TIMEOUT,
165 	    &mf6c_expire_route);
166 #endif
167 }
168 
169 struct ip6_offnxt {
170 	int	ion_off;
171 	int	ion_nxt;
172 };
173 
174 /*
175  * Enqueue packet for local delivery.  Queuing is used as a boundary
176  * between the network layer (input/forward path) running with
177  * NET_LOCK_SHARED() and the transport layer needing it exclusively.
178  */
179 int
180 ip6_ours(struct mbuf **mp, int *offp, int nxt, int af)
181 {
182 	/* ip6_hbhchcheck() may be run before, then off and nxt are set */
183 	if (*offp == 0) {
184 		nxt = ip6_hbhchcheck(mp, offp, NULL);
185 		if (nxt == IPPROTO_DONE)
186 			return IPPROTO_DONE;
187 	}
188 
189 	/* We are already in a IPv4/IPv6 local deliver loop. */
190 	if (af != AF_UNSPEC)
191 		return nxt;
192 
193 	/* save values for later, use after dequeue */
194 	if (*offp != sizeof(struct ip6_hdr)) {
195 		struct m_tag *mtag;
196 		struct ip6_offnxt *ion;
197 
198 		/* mbuf tags are expensive, but only used for header options */
199 		mtag = m_tag_get(PACKET_TAG_IP6_OFFNXT, sizeof(*ion),
200 		    M_NOWAIT);
201 		if (mtag == NULL) {
202 			ip6stat_inc(ip6s_idropped);
203 			m_freemp(mp);
204 			return IPPROTO_DONE;
205 		}
206 		ion = (struct ip6_offnxt *)(mtag + 1);
207 		ion->ion_off = *offp;
208 		ion->ion_nxt = nxt;
209 
210 		m_tag_prepend(*mp, mtag);
211 	}
212 
213 	niq_enqueue(&ip6intrq, *mp);
214 	*mp = NULL;
215 	return IPPROTO_DONE;
216 }
217 
218 /*
219  * Dequeue and process locally delivered packets.
220  * This is called with exclusive NET_LOCK().
221  */
222 void
223 ip6intr(void)
224 {
225 	struct mbuf *m;
226 
227 	while ((m = niq_dequeue(&ip6intrq)) != NULL) {
228 		struct m_tag *mtag;
229 		int off, nxt;
230 
231 #ifdef DIAGNOSTIC
232 		if ((m->m_flags & M_PKTHDR) == 0)
233 			panic("ip6intr no HDR");
234 #endif
235 		mtag = m_tag_find(m, PACKET_TAG_IP6_OFFNXT, NULL);
236 		if (mtag != NULL) {
237 			struct ip6_offnxt *ion;
238 
239 			ion = (struct ip6_offnxt *)(mtag + 1);
240 			off = ion->ion_off;
241 			nxt = ion->ion_nxt;
242 
243 			m_tag_delete(m, mtag);
244 		} else {
245 			struct ip6_hdr *ip6;
246 
247 			ip6 = mtod(m, struct ip6_hdr *);
248 			off = sizeof(struct ip6_hdr);
249 			nxt = ip6->ip6_nxt;
250 		}
251 		nxt = ip_deliver(&m, &off, nxt, AF_INET6);
252 		KASSERT(nxt == IPPROTO_DONE);
253 	}
254 }
255 
256 void
257 ipv6_input(struct ifnet *ifp, struct mbuf *m)
258 {
259 	int off, nxt;
260 
261 	off = 0;
262 	nxt = ip6_input_if(&m, &off, IPPROTO_IPV6, AF_UNSPEC, ifp);
263 	KASSERT(nxt == IPPROTO_DONE);
264 }
265 
266 struct mbuf *
267 ipv6_check(struct ifnet *ifp, struct mbuf *m)
268 {
269 	struct ip6_hdr *ip6;
270 
271 	if (m->m_len < sizeof(*ip6)) {
272 		m = m_pullup(m, sizeof(*ip6));
273 		if (m == NULL) {
274 			ip6stat_inc(ip6s_toosmall);
275 			return (NULL);
276 		}
277 	}
278 
279 	ip6 = mtod(m, struct ip6_hdr *);
280 
281 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
282 		ip6stat_inc(ip6s_badvers);
283 		goto bad;
284 	}
285 
286 	/*
287 	 * Check against address spoofing/corruption.
288 	 */
289 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
290 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
291 		/*
292 		 * XXX: "badscope" is not very suitable for a multicast source.
293 		 */
294 		ip6stat_inc(ip6s_badscope);
295 		goto bad;
296 	}
297 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
298 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
299 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
300 		ip6stat_inc(ip6s_badscope);
301 		goto bad;
302 	}
303 	/* Drop packets if interface ID portion is already filled. */
304 	if (((IN6_IS_SCOPE_EMBED(&ip6->ip6_src) && ip6->ip6_src.s6_addr16[1]) ||
305 	    (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst) && ip6->ip6_dst.s6_addr16[1])) &&
306 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
307 		ip6stat_inc(ip6s_badscope);
308 		goto bad;
309 	}
310 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
311 	    !(m->m_flags & M_LOOP)) {
312 		/*
313 		 * In this case, the packet should come from the loopback
314 		 * interface.  However, we cannot just check the if_flags,
315 		 * because ip6_mloopback() passes the "actual" interface
316 		 * as the outgoing/incoming interface.
317 		 */
318 		ip6stat_inc(ip6s_badscope);
319 		goto bad;
320 	}
321 
322 	/*
323 	 * The following check is not documented in specs.  A malicious
324 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
325 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
326 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
327 	 *
328 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
329 	 * support IPv4-less kernel compilation, we cannot support SIIT
330 	 * environment at all.  So, it makes more sense for us to reject any
331 	 * malicious packets for non-SIIT environment, than try to do a
332 	 * partial support for SIIT environment.
333 	 */
334 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
335 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
336 		ip6stat_inc(ip6s_badscope);
337 		goto bad;
338 	}
339 
340 	/*
341 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
342 	 *
343 	 * The code forbids automatic tunneling as per RFC4213.
344 	 */
345 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
346 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
347 		ip6stat_inc(ip6s_badscope);
348 		goto bad;
349 	}
350 
351 	return (m);
352 bad:
353 	m_freem(m);
354 	return (NULL);
355 }
356 
357 int
358 ip6_input_if(struct mbuf **mp, int *offp, int nxt, int af, struct ifnet *ifp)
359 {
360 	struct mbuf *m;
361 	struct ip6_hdr *ip6;
362 	struct sockaddr_in6 sin6;
363 	struct rtentry *rt = NULL;
364 	int ours = 0;
365 	u_int16_t src_scope, dst_scope;
366 #if NPF > 0
367 	struct in6_addr odst;
368 #endif
369 	int srcrt = 0;
370 
371 	KASSERT(*offp == 0);
372 
373 	ip6stat_inc(ip6s_total);
374 
375 	m = *mp = ipv6_check(ifp, *mp);
376 	if (m == NULL)
377 		goto bad;
378 
379 	ip6 = mtod(m, struct ip6_hdr *);
380 
381 #if NCARP > 0
382 	if (carp_lsdrop(ifp, m, AF_INET6, ip6->ip6_src.s6_addr32,
383 	    ip6->ip6_dst.s6_addr32, (ip6->ip6_nxt == IPPROTO_ICMPV6 ? 0 : 1)))
384 		goto bad;
385 #endif
386 	ip6stat_inc(ip6s_nxthist + ip6->ip6_nxt);
387 
388 	/*
389 	 * If the packet has been received on a loopback interface it
390 	 * can be destined to any local address, not necessarily to
391 	 * an address configured on `ifp'.
392 	 */
393 	if (ifp->if_flags & IFF_LOOPBACK) {
394 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
395 			src_scope = ip6->ip6_src.s6_addr16[1];
396 			ip6->ip6_src.s6_addr16[1] = 0;
397 		}
398 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
399 			dst_scope = ip6->ip6_dst.s6_addr16[1];
400 			ip6->ip6_dst.s6_addr16[1] = 0;
401 		}
402 	}
403 
404 #if NPF > 0
405 	/*
406 	 * Packet filter
407 	 */
408 	odst = ip6->ip6_dst;
409 	if (pf_test(AF_INET6, PF_IN, ifp, mp) != PF_PASS)
410 		goto bad;
411 	m = *mp;
412 	if (m == NULL)
413 		goto bad;
414 
415 	ip6 = mtod(m, struct ip6_hdr *);
416 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
417 #endif
418 
419 	/*
420 	 * Without embedded scope ID we cannot find link-local
421 	 * addresses in the routing table.
422 	 */
423 	if (ifp->if_flags & IFF_LOOPBACK) {
424 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
425 			ip6->ip6_src.s6_addr16[1] = src_scope;
426 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
427 			ip6->ip6_dst.s6_addr16[1] = dst_scope;
428 	} else {
429 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
430 			ip6->ip6_src.s6_addr16[1] = htons(ifp->if_index);
431 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
432 			ip6->ip6_dst.s6_addr16[1] = htons(ifp->if_index);
433 	}
434 
435 	/*
436 	 * Be more secure than RFC5095 and scan for type 0 routing headers.
437 	 * If pf has already scanned the header chain, do not do it twice.
438 	 */
439 	if (!(m->m_pkthdr.pf.flags & PF_TAG_PROCESSED) &&
440 	    ip6_check_rh0hdr(m, offp)) {
441 		ip6stat_inc(ip6s_badoptions);
442 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, *offp);
443 		m = *mp = NULL;
444 		goto bad;
445 	}
446 
447 #if NPF > 0
448 	if (pf_ouraddr(m) == 1) {
449 		nxt = ip6_ours(mp, offp, nxt, af);
450 		goto out;
451 	}
452 #endif
453 
454 	/*
455 	 * Multicast check
456 	 */
457 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
458 		/*
459 		 * Make sure M_MCAST is set.  It should theoretically
460 		 * already be there, but let's play safe because upper
461 		 * layers check for this flag.
462 		 */
463 		m->m_flags |= M_MCAST;
464 
465 		/*
466 		 * See if we belong to the destination multicast group on the
467 		 * arrival interface.
468 		 */
469 		if (in6_hasmulti(&ip6->ip6_dst, ifp))
470 			ours = 1;
471 
472 #ifdef MROUTING
473 		if (ip6_mforwarding && ip6_mrouter[ifp->if_rdomain]) {
474 			int error;
475 
476 			nxt = ip6_hbhchcheck(&m, offp, &ours);
477 			if (nxt == IPPROTO_DONE)
478 				goto out;
479 
480 			ip6 = mtod(m, struct ip6_hdr *);
481 
482 			/*
483 			 * If we are acting as a multicast router, all
484 			 * incoming multicast packets are passed to the
485 			 * kernel-level multicast forwarding function.
486 			 * The packet is returned (relatively) intact; if
487 			 * ip6_mforward() returns a non-zero value, the packet
488 			 * must be discarded, else it may be accepted below.
489 			 */
490 			KERNEL_LOCK();
491 			error = ip6_mforward(ip6, ifp, m);
492 			KERNEL_UNLOCK();
493 			if (error) {
494 				ip6stat_inc(ip6s_cantforward);
495 				goto bad;
496 			}
497 
498 			if (ours) {
499 				if (af == AF_UNSPEC)
500 					nxt = ip6_ours(mp, offp, nxt, af);
501 				goto out;
502 			}
503 			goto bad;
504 		}
505 #endif
506 		if (!ours) {
507 			ip6stat_inc(ip6s_notmember);
508 			if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
509 				ip6stat_inc(ip6s_cantforward);
510 			goto bad;
511 		}
512 		nxt = ip6_ours(mp, offp, nxt, af);
513 		goto out;
514 	}
515 
516 
517 	/*
518 	 *  Unicast check
519 	 */
520 	memset(&sin6, 0, sizeof(struct sockaddr_in6));
521 	sin6.sin6_len = sizeof(struct sockaddr_in6);
522 	sin6.sin6_family = AF_INET6;
523 	sin6.sin6_addr = ip6->ip6_dst;
524 	rt = rtalloc_mpath(sin6tosa(&sin6), &ip6->ip6_src.s6_addr32[0],
525 	    m->m_pkthdr.ph_rtableid);
526 
527 	/*
528 	 * Accept the packet if the route to the destination is marked
529 	 * as local.
530 	 */
531 	if (rtisvalid(rt) && ISSET(rt->rt_flags, RTF_LOCAL)) {
532 		struct in6_ifaddr *ia6 = ifatoia6(rt->rt_ifa);
533 
534 		if (ip6_forwarding == 0 && rt->rt_ifidx != ifp->if_index &&
535 		    !((ifp->if_flags & IFF_LOOPBACK) ||
536 		    (ifp->if_type == IFT_ENC) ||
537 		    (m->m_pkthdr.pf.flags & PF_TAG_TRANSLATE_LOCALHOST))) {
538 			/* received on wrong interface */
539 #if NCARP > 0
540 			struct ifnet *out_if;
541 
542 			/*
543 			 * Virtual IPs on carp interfaces need to be checked
544 			 * also against the parent interface and other carp
545 			 * interfaces sharing the same parent.
546 			 */
547 			out_if = if_get(rt->rt_ifidx);
548 			if (!(out_if && carp_strict_addr_chk(out_if, ifp))) {
549 				ip6stat_inc(ip6s_wrongif);
550 				if_put(out_if);
551 				goto bad;
552 			}
553 			if_put(out_if);
554 #else
555 			ip6stat_inc(ip6s_wrongif);
556 			goto bad;
557 #endif
558 		}
559 		/*
560 		 * packets to a tentative, duplicated, or somehow invalid
561 		 * address must not be accepted.
562 		 */
563 		if ((ia6->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED))) {
564 			char src[INET6_ADDRSTRLEN], dst[INET6_ADDRSTRLEN];
565 
566 			inet_ntop(AF_INET6, &ip6->ip6_src, src, sizeof(src));
567 			inet_ntop(AF_INET6, &ip6->ip6_dst, dst, sizeof(dst));
568 			/* address is not ready, so discard the packet. */
569 			nd6log((LOG_INFO,
570 			    "%s: packet to an unready address %s->%s\n",
571 			    __func__, src, dst));
572 
573 			goto bad;
574 		} else {
575 			nxt = ip6_ours(mp, offp, nxt, af);
576 			goto out;
577 		}
578 	}
579 
580 #if NCARP > 0
581 	if (ip6->ip6_nxt == IPPROTO_ICMPV6 &&
582 	    carp_lsdrop(ifp, m, AF_INET6, ip6->ip6_src.s6_addr32,
583 	    ip6->ip6_dst.s6_addr32, 1))
584 		goto bad;
585 #endif
586 	/*
587 	 * Now there is no reason to process the packet if it's not our own
588 	 * and we're not a router.
589 	 */
590 	if (!ip6_forwarding) {
591 		ip6stat_inc(ip6s_cantforward);
592 		goto bad;
593 	}
594 
595 	nxt = ip6_hbhchcheck(&m, offp, &ours);
596 	if (nxt == IPPROTO_DONE)
597 		goto out;
598 
599 	if (ours) {
600 		if (af == AF_UNSPEC)
601 			nxt = ip6_ours(mp, offp, nxt, af);
602 		goto out;
603 	}
604 
605 #ifdef IPSEC
606 	if (ipsec_in_use) {
607 		int rv;
608 
609 		rv = ipsec_forward_check(m, *offp, AF_INET6);
610 		if (rv != 0) {
611 			ip6stat_inc(ip6s_cantforward);
612 			goto bad;
613 		}
614 		/*
615 		 * Fall through, forward packet. Outbound IPsec policy
616 		 * checking will occur in ip6_forward().
617 		 */
618 	}
619 #endif /* IPSEC */
620 
621 	ip6_forward(m, rt, srcrt);
622 	*mp = NULL;
623 	return IPPROTO_DONE;
624  bad:
625 	nxt = IPPROTO_DONE;
626 	m_freemp(mp);
627  out:
628 	rtfree(rt);
629 	return nxt;
630 }
631 
632 /* On error free mbuf and return IPPROTO_DONE. */
633 int
634 ip6_hbhchcheck(struct mbuf **mp, int *offp, int *oursp)
635 {
636 	struct ip6_hdr *ip6;
637 	u_int32_t plen, rtalert = ~0;
638 	int nxt;
639 
640 	ip6 = mtod(*mp, struct ip6_hdr *);
641 
642 	/*
643 	 * Process Hop-by-Hop options header if it's contained.
644 	 * m may be modified in ip6_hopopts_input().
645 	 * If a JumboPayload option is included, plen will also be modified.
646 	 */
647 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
648 	*offp = sizeof(struct ip6_hdr);
649 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
650 		struct ip6_hbh *hbh;
651 
652 		if (ip6_hopopts_input(mp, offp, &plen, &rtalert))
653 			goto bad;	/* m have already been freed */
654 
655 		/* adjust pointer */
656 		ip6 = mtod(*mp, struct ip6_hdr *);
657 
658 		/*
659 		 * if the payload length field is 0 and the next header field
660 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
661 		 * option MUST be included.
662 		 */
663 		if (ip6->ip6_plen == 0 && plen == 0) {
664 			/*
665 			 * Note that if a valid jumbo payload option is
666 			 * contained, ip6_hopopts_input() must set a valid
667 			 * (non-zero) payload length to the variable plen.
668 			 */
669 			ip6stat_inc(ip6s_badoptions);
670 			icmp6_error(*mp, ICMP6_PARAM_PROB,
671 				    ICMP6_PARAMPROB_HEADER,
672 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
673 			goto bad;
674 		}
675 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, *mp,
676 		    sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
677 		if (hbh == NULL) {
678 			ip6stat_inc(ip6s_tooshort);
679 			goto bad;
680 		}
681 		nxt = hbh->ip6h_nxt;
682 
683 		/*
684 		 * accept the packet if a router alert option is included
685 		 * and we act as an IPv6 router.
686 		 */
687 		if (rtalert != ~0 && ip6_forwarding && oursp != NULL)
688 			*oursp = 1;
689 	} else
690 		nxt = ip6->ip6_nxt;
691 
692 	/*
693 	 * Check that the amount of data in the buffers
694 	 * is as at least much as the IPv6 header would have us expect.
695 	 * Trim mbufs if longer than we expect.
696 	 * Drop packet if shorter than we expect.
697 	 */
698 	if ((*mp)->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
699 		ip6stat_inc(ip6s_tooshort);
700 		m_freemp(mp);
701 		goto bad;
702 	}
703 	if ((*mp)->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
704 		if ((*mp)->m_len == (*mp)->m_pkthdr.len) {
705 			(*mp)->m_len = sizeof(struct ip6_hdr) + plen;
706 			(*mp)->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
707 		} else {
708 			m_adj((*mp), sizeof(struct ip6_hdr) + plen -
709 			    (*mp)->m_pkthdr.len);
710 		}
711 	}
712 
713 	return nxt;
714  bad:
715 	return IPPROTO_DONE;
716 }
717 
718 /* scan packet for RH0 routing header. Mostly stolen from pf.c:pf_test() */
719 int
720 ip6_check_rh0hdr(struct mbuf *m, int *offp)
721 {
722 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
723 	struct ip6_rthdr rthdr;
724 	struct ip6_ext opt6;
725 	u_int8_t proto = ip6->ip6_nxt;
726 	int done = 0, lim, off, rh_cnt = 0;
727 
728 	off = ((caddr_t)ip6 - m->m_data) + sizeof(struct ip6_hdr);
729 	lim = min(m->m_pkthdr.len, ntohs(ip6->ip6_plen) + sizeof(*ip6));
730 	do {
731 		switch (proto) {
732 		case IPPROTO_ROUTING:
733 			if (rh_cnt++) {
734 				/* more than one rh header present */
735 				*offp = off;
736 				return (1);
737 			}
738 
739 			if (off + sizeof(rthdr) > lim) {
740 				/* packet to short to make sense */
741 				*offp = off;
742 				return (1);
743 			}
744 
745 			m_copydata(m, off, sizeof(rthdr), &rthdr);
746 
747 			if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
748 				*offp = off +
749 				    offsetof(struct ip6_rthdr, ip6r_type);
750 				return (1);
751 			}
752 
753 			off += (rthdr.ip6r_len + 1) * 8;
754 			proto = rthdr.ip6r_nxt;
755 			break;
756 		case IPPROTO_AH:
757 		case IPPROTO_HOPOPTS:
758 		case IPPROTO_DSTOPTS:
759 			/* get next header and header length */
760 			if (off + sizeof(opt6) > lim) {
761 				/*
762 				 * Packet to short to make sense, we could
763 				 * reject the packet but as a router we
764 				 * should not do that so forward it.
765 				 */
766 				return (0);
767 			}
768 
769 			m_copydata(m, off, sizeof(opt6), &opt6);
770 
771 			if (proto == IPPROTO_AH)
772 				off += (opt6.ip6e_len + 2) * 4;
773 			else
774 				off += (opt6.ip6e_len + 1) * 8;
775 			proto = opt6.ip6e_nxt;
776 			break;
777 		case IPPROTO_FRAGMENT:
778 		default:
779 			/* end of header stack */
780 			done = 1;
781 			break;
782 		}
783 	} while (!done);
784 
785 	return (0);
786 }
787 
788 /*
789  * Hop-by-Hop options header processing. If a valid jumbo payload option is
790  * included, the real payload length will be stored in plenp.
791  * On error free mbuf and return -1.
792  *
793  * rtalertp - XXX: should be stored in a more smart way
794  */
795 int
796 ip6_hopopts_input(struct mbuf **mp, int *offp, u_int32_t *plenp,
797     u_int32_t *rtalertp)
798 {
799 	int off = *offp, hbhlen;
800 	struct ip6_hbh *hbh;
801 
802 	/* validation of the length of the header */
803 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, *mp,
804 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
805 	if (hbh == NULL) {
806 		ip6stat_inc(ip6s_tooshort);
807 		return -1;
808 	}
809 	hbhlen = (hbh->ip6h_len + 1) << 3;
810 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, *mp, sizeof(struct ip6_hdr),
811 		hbhlen);
812 	if (hbh == NULL) {
813 		ip6stat_inc(ip6s_tooshort);
814 		return -1;
815 	}
816 	off += hbhlen;
817 	hbhlen -= sizeof(struct ip6_hbh);
818 
819 	if (ip6_process_hopopts(mp, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
820 				hbhlen, rtalertp, plenp) < 0)
821 		return (-1);
822 
823 	*offp = off;
824 	return (0);
825 }
826 
827 /*
828  * Search header for all Hop-by-hop options and process each option.
829  * This function is separate from ip6_hopopts_input() in order to
830  * handle a case where the sending node itself process its hop-by-hop
831  * options header. In such a case, the function is called from ip6_output().
832  * On error free mbuf and return -1.
833  *
834  * The function assumes that hbh header is located right after the IPv6 header
835  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
836  * opthead + hbhlen is located in continuous memory region.
837  */
838 int
839 ip6_process_hopopts(struct mbuf **mp, u_int8_t *opthead, int hbhlen,
840     u_int32_t *rtalertp, u_int32_t *plenp)
841 {
842 	struct ip6_hdr *ip6;
843 	int optlen = 0;
844 	u_int8_t *opt = opthead;
845 	u_int16_t rtalert_val;
846 	u_int32_t jumboplen;
847 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
848 
849 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
850 		switch (*opt) {
851 		case IP6OPT_PAD1:
852 			optlen = 1;
853 			break;
854 		case IP6OPT_PADN:
855 			if (hbhlen < IP6OPT_MINLEN) {
856 				ip6stat_inc(ip6s_toosmall);
857 				goto bad;
858 			}
859 			optlen = *(opt + 1) + 2;
860 			break;
861 		case IP6OPT_ROUTER_ALERT:
862 			/* XXX may need check for alignment */
863 			if (hbhlen < IP6OPT_RTALERT_LEN) {
864 				ip6stat_inc(ip6s_toosmall);
865 				goto bad;
866 			}
867 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
868 				/* XXX stat */
869 				icmp6_error(*mp, ICMP6_PARAM_PROB,
870 				    ICMP6_PARAMPROB_HEADER,
871 				    erroff + opt + 1 - opthead);
872 				return (-1);
873 			}
874 			optlen = IP6OPT_RTALERT_LEN;
875 			memcpy((caddr_t)&rtalert_val, (caddr_t)(opt + 2), 2);
876 			*rtalertp = ntohs(rtalert_val);
877 			break;
878 		case IP6OPT_JUMBO:
879 			/* XXX may need check for alignment */
880 			if (hbhlen < IP6OPT_JUMBO_LEN) {
881 				ip6stat_inc(ip6s_toosmall);
882 				goto bad;
883 			}
884 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
885 				/* XXX stat */
886 				icmp6_error(*mp, ICMP6_PARAM_PROB,
887 				    ICMP6_PARAMPROB_HEADER,
888 				    erroff + opt + 1 - opthead);
889 				return (-1);
890 			}
891 			optlen = IP6OPT_JUMBO_LEN;
892 
893 			/*
894 			 * IPv6 packets that have non 0 payload length
895 			 * must not contain a jumbo payload option.
896 			 */
897 			ip6 = mtod(*mp, struct ip6_hdr *);
898 			if (ip6->ip6_plen) {
899 				ip6stat_inc(ip6s_badoptions);
900 				icmp6_error(*mp, ICMP6_PARAM_PROB,
901 				    ICMP6_PARAMPROB_HEADER,
902 				    erroff + opt - opthead);
903 				return (-1);
904 			}
905 
906 			/*
907 			 * We may see jumbolen in unaligned location, so
908 			 * we'd need to perform memcpy().
909 			 */
910 			memcpy(&jumboplen, opt + 2, sizeof(jumboplen));
911 			jumboplen = (u_int32_t)htonl(jumboplen);
912 
913 #if 1
914 			/*
915 			 * if there are multiple jumbo payload options,
916 			 * *plenp will be non-zero and the packet will be
917 			 * rejected.
918 			 * the behavior may need some debate in ipngwg -
919 			 * multiple options does not make sense, however,
920 			 * there's no explicit mention in specification.
921 			 */
922 			if (*plenp != 0) {
923 				ip6stat_inc(ip6s_badoptions);
924 				icmp6_error(*mp, ICMP6_PARAM_PROB,
925 				    ICMP6_PARAMPROB_HEADER,
926 				    erroff + opt + 2 - opthead);
927 				return (-1);
928 			}
929 #endif
930 
931 			/*
932 			 * jumbo payload length must be larger than 65535.
933 			 */
934 			if (jumboplen <= IPV6_MAXPACKET) {
935 				ip6stat_inc(ip6s_badoptions);
936 				icmp6_error(*mp, ICMP6_PARAM_PROB,
937 				    ICMP6_PARAMPROB_HEADER,
938 				    erroff + opt + 2 - opthead);
939 				return (-1);
940 			}
941 			*plenp = jumboplen;
942 
943 			break;
944 		default:		/* unknown option */
945 			if (hbhlen < IP6OPT_MINLEN) {
946 				ip6stat_inc(ip6s_toosmall);
947 				goto bad;
948 			}
949 			optlen = ip6_unknown_opt(mp, opt,
950 			    erroff + opt - opthead);
951 			if (optlen == -1)
952 				return (-1);
953 			optlen += 2;
954 			break;
955 		}
956 	}
957 
958 	return (0);
959 
960   bad:
961 	m_freemp(mp);
962 	return (-1);
963 }
964 
965 /*
966  * Unknown option processing.
967  * The third argument `off' is the offset from the IPv6 header to the option,
968  * which allows returning an ICMPv6 error even if the IPv6 header and the
969  * option header are not continuous.
970  * On error free mbuf and return -1.
971  */
972 int
973 ip6_unknown_opt(struct mbuf **mp, u_int8_t *optp, int off)
974 {
975 	struct ip6_hdr *ip6;
976 
977 	switch (IP6OPT_TYPE(*optp)) {
978 	case IP6OPT_TYPE_SKIP: /* ignore the option */
979 		return ((int)*(optp + 1));
980 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
981 		m_freemp(mp);
982 		return (-1);
983 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
984 		ip6stat_inc(ip6s_badoptions);
985 		icmp6_error(*mp, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
986 		return (-1);
987 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
988 		ip6stat_inc(ip6s_badoptions);
989 		ip6 = mtod(*mp, struct ip6_hdr *);
990 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
991 		    ((*mp)->m_flags & (M_BCAST|M_MCAST)))
992 			m_freemp(mp);
993 		else
994 			icmp6_error(*mp, ICMP6_PARAM_PROB,
995 				    ICMP6_PARAMPROB_OPTION, off);
996 		return (-1);
997 	}
998 
999 	m_freemp(mp);		/* XXX: NOTREACHED */
1000 	return (-1);
1001 }
1002 
1003 /*
1004  * Create the "control" list for this pcb.
1005  *
1006  * The routine will be called from upper layer handlers like udp_input().
1007  * Thus the routine assumes that the caller (udp_input) have already
1008  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1009  * very first mbuf on the mbuf chain.
1010  * We may want to add some infinite loop prevention or sanity checks for safety.
1011  * (This applies only when you are using KAME mbuf chain restriction, i.e.
1012  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
1013  */
1014 void
1015 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1016 {
1017 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1018 
1019 	if (in6p->inp_socket->so_options & SO_TIMESTAMP) {
1020 		struct timeval tv;
1021 
1022 		m_microtime(m, &tv);
1023 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1024 		    SCM_TIMESTAMP, SOL_SOCKET);
1025 		if (*mp)
1026 			mp = &(*mp)->m_next;
1027 	}
1028 
1029 	/* RFC 2292 sec. 5 */
1030 	if ((in6p->inp_flags & IN6P_PKTINFO) != 0) {
1031 		struct in6_pktinfo pi6;
1032 		memcpy(&pi6.ipi6_addr, &ip6->ip6_dst, sizeof(struct in6_addr));
1033 		if (IN6_IS_SCOPE_EMBED(&pi6.ipi6_addr))
1034 			pi6.ipi6_addr.s6_addr16[1] = 0;
1035 		pi6.ipi6_ifindex = m ? m->m_pkthdr.ph_ifidx : 0;
1036 		*mp = sbcreatecontrol((caddr_t) &pi6,
1037 		    sizeof(struct in6_pktinfo),
1038 		    IPV6_PKTINFO, IPPROTO_IPV6);
1039 		if (*mp)
1040 			mp = &(*mp)->m_next;
1041 	}
1042 
1043 	if ((in6p->inp_flags & IN6P_HOPLIMIT) != 0) {
1044 		int hlim = ip6->ip6_hlim & 0xff;
1045 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1046 		    IPV6_HOPLIMIT, IPPROTO_IPV6);
1047 		if (*mp)
1048 			mp = &(*mp)->m_next;
1049 	}
1050 
1051 	if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
1052 		u_int32_t flowinfo;
1053 		int tclass;
1054 
1055 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1056 		flowinfo >>= 20;
1057 
1058 		tclass = flowinfo & 0xff;
1059 		*mp = sbcreatecontrol((caddr_t)&tclass, sizeof(tclass),
1060 		    IPV6_TCLASS, IPPROTO_IPV6);
1061 		if (*mp)
1062 			mp = &(*mp)->m_next;
1063 	}
1064 
1065 	/*
1066 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1067 	 * privilege for the option (see ip6_ctloutput), but it might be too
1068 	 * strict, since there might be some hop-by-hop options which can be
1069 	 * returned to normal user.
1070 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1071 	 */
1072 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1073 		/*
1074 		 * Check if a hop-by-hop options header is contained in the
1075 		 * received packet, and if so, store the options as ancillary
1076 		 * data. Note that a hop-by-hop options header must be
1077 		 * just after the IPv6 header, which is assured through the
1078 		 * IPv6 input processing.
1079 		 */
1080 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1081 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1082 			struct ip6_hbh *hbh;
1083 			int hbhlen = 0;
1084 			struct mbuf *ext;
1085 
1086 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1087 			    ip6->ip6_nxt);
1088 			if (ext == NULL) {
1089 				ip6stat_inc(ip6s_tooshort);
1090 				return;
1091 			}
1092 			hbh = mtod(ext, struct ip6_hbh *);
1093 			hbhlen = (hbh->ip6h_len + 1) << 3;
1094 			if (hbhlen != ext->m_len) {
1095 				m_freem(ext);
1096 				ip6stat_inc(ip6s_tooshort);
1097 				return;
1098 			}
1099 
1100 			/*
1101 			 * XXX: We copy the whole header even if a
1102 			 * jumbo payload option is included, the option which
1103 			 * is to be removed before returning according to
1104 			 * RFC2292.
1105 			 * Note: this constraint is removed in RFC3542.
1106 			 */
1107 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1108 			    IPV6_HOPOPTS,
1109 			    IPPROTO_IPV6);
1110 			if (*mp)
1111 				mp = &(*mp)->m_next;
1112 			m_freem(ext);
1113 		}
1114 	}
1115 
1116 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1117 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1118 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1119 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1120 
1121 		/*
1122 		 * Search for destination options headers or routing
1123 		 * header(s) through the header chain, and stores each
1124 		 * header as ancillary data.
1125 		 * Note that the order of the headers remains in
1126 		 * the chain of ancillary data.
1127 		 */
1128 		while (1) {	/* is explicit loop prevention necessary? */
1129 			struct ip6_ext *ip6e = NULL;
1130 			int elen;
1131 			struct mbuf *ext = NULL;
1132 
1133 			/*
1134 			 * if it is not an extension header, don't try to
1135 			 * pull it from the chain.
1136 			 */
1137 			switch (nxt) {
1138 			case IPPROTO_DSTOPTS:
1139 			case IPPROTO_ROUTING:
1140 			case IPPROTO_HOPOPTS:
1141 			case IPPROTO_AH: /* is it possible? */
1142 				break;
1143 			default:
1144 				goto loopend;
1145 			}
1146 
1147 			ext = ip6_pullexthdr(m, off, nxt);
1148 			if (ext == NULL) {
1149 				ip6stat_inc(ip6s_tooshort);
1150 				return;
1151 			}
1152 			ip6e = mtod(ext, struct ip6_ext *);
1153 			if (nxt == IPPROTO_AH)
1154 				elen = (ip6e->ip6e_len + 2) << 2;
1155 			else
1156 				elen = (ip6e->ip6e_len + 1) << 3;
1157 			if (elen != ext->m_len) {
1158 				m_freem(ext);
1159 				ip6stat_inc(ip6s_tooshort);
1160 				return;
1161 			}
1162 
1163 			switch (nxt) {
1164 			case IPPROTO_DSTOPTS:
1165 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1166 					break;
1167 
1168 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1169 				    IPV6_DSTOPTS,
1170 				    IPPROTO_IPV6);
1171 				if (*mp)
1172 					mp = &(*mp)->m_next;
1173 				break;
1174 
1175 			case IPPROTO_ROUTING:
1176 				if (!(in6p->inp_flags & IN6P_RTHDR))
1177 					break;
1178 
1179 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1180 				    IPV6_RTHDR,
1181 				    IPPROTO_IPV6);
1182 				if (*mp)
1183 					mp = &(*mp)->m_next;
1184 				break;
1185 
1186 			case IPPROTO_HOPOPTS:
1187 			case IPPROTO_AH: /* is it possible? */
1188 				break;
1189 
1190 			default:
1191 				/*
1192 				 * other cases have been filtered in the above.
1193 				 * none will visit this case.  here we supply
1194 				 * the code just in case (nxt overwritten or
1195 				 * other cases).
1196 				 */
1197 				m_freem(ext);
1198 				goto loopend;
1199 
1200 			}
1201 
1202 			/* proceed with the next header. */
1203 			off += elen;
1204 			nxt = ip6e->ip6e_nxt;
1205 			ip6e = NULL;
1206 			m_freem(ext);
1207 			ext = NULL;
1208 		}
1209 loopend:
1210 		;
1211 	}
1212 }
1213 
1214 /*
1215  * pull single extension header from mbuf chain.  returns single mbuf that
1216  * contains the result, or NULL on error.
1217  */
1218 struct mbuf *
1219 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1220 {
1221 	struct ip6_ext ip6e;
1222 	size_t elen;
1223 	struct mbuf *n;
1224 
1225 #ifdef DIAGNOSTIC
1226 	switch (nxt) {
1227 	case IPPROTO_DSTOPTS:
1228 	case IPPROTO_ROUTING:
1229 	case IPPROTO_HOPOPTS:
1230 	case IPPROTO_AH: /* is it possible? */
1231 		break;
1232 	default:
1233 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1234 	}
1235 #endif
1236 
1237 	if (off + sizeof(ip6e) > m->m_pkthdr.len)
1238 		return NULL;
1239 
1240 	m_copydata(m, off, sizeof(ip6e), &ip6e);
1241 	if (nxt == IPPROTO_AH)
1242 		elen = (ip6e.ip6e_len + 2) << 2;
1243 	else
1244 		elen = (ip6e.ip6e_len + 1) << 3;
1245 
1246 	if (off + elen > m->m_pkthdr.len)
1247 		return NULL;
1248 
1249 	MGET(n, M_DONTWAIT, MT_DATA);
1250 	if (n && elen >= MLEN) {
1251 		MCLGET(n, M_DONTWAIT);
1252 		if ((n->m_flags & M_EXT) == 0) {
1253 			m_free(n);
1254 			n = NULL;
1255 		}
1256 	}
1257 	if (n == NULL) {
1258 		ip6stat_inc(ip6s_idropped);
1259 		return NULL;
1260 	}
1261 
1262 	n->m_len = 0;
1263 	if (elen >= m_trailingspace(n)) {
1264 		m_free(n);
1265 		return NULL;
1266 	}
1267 
1268 	m_copydata(m, off, elen, mtod(n, caddr_t));
1269 	n->m_len = elen;
1270 	return n;
1271 }
1272 
1273 /*
1274  * Get offset to the previous header followed by the header
1275  * currently processed.
1276  */
1277 int
1278 ip6_get_prevhdr(struct mbuf *m, int off)
1279 {
1280 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1281 
1282 	if (off == sizeof(struct ip6_hdr)) {
1283 		return offsetof(struct ip6_hdr, ip6_nxt);
1284 	} else if (off < sizeof(struct ip6_hdr)) {
1285 		panic("%s: off < sizeof(struct ip6_hdr)", __func__);
1286 	} else {
1287 		int len, nlen, nxt;
1288 		struct ip6_ext ip6e;
1289 
1290 		nxt = ip6->ip6_nxt;
1291 		len = sizeof(struct ip6_hdr);
1292 		nlen = 0;
1293 		while (len < off) {
1294 			m_copydata(m, len, sizeof(ip6e), &ip6e);
1295 
1296 			switch (nxt) {
1297 			case IPPROTO_FRAGMENT:
1298 				nlen = sizeof(struct ip6_frag);
1299 				break;
1300 			case IPPROTO_AH:
1301 				nlen = (ip6e.ip6e_len + 2) << 2;
1302 				break;
1303 			default:
1304 				nlen = (ip6e.ip6e_len + 1) << 3;
1305 				break;
1306 			}
1307 			len += nlen;
1308 			nxt = ip6e.ip6e_nxt;
1309 		}
1310 
1311 		return (len - nlen);
1312 	}
1313 }
1314 
1315 /*
1316  * get next header offset.  m will be retained.
1317  */
1318 int
1319 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1320 {
1321 	struct ip6_hdr ip6;
1322 	struct ip6_ext ip6e;
1323 	struct ip6_frag fh;
1324 
1325 	/* just in case */
1326 	if (m == NULL)
1327 		panic("%s: m == NULL", __func__);
1328 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1329 		return -1;
1330 
1331 	switch (proto) {
1332 	case IPPROTO_IPV6:
1333 		if (m->m_pkthdr.len < off + sizeof(ip6))
1334 			return -1;
1335 		m_copydata(m, off, sizeof(ip6), &ip6);
1336 		if (nxtp)
1337 			*nxtp = ip6.ip6_nxt;
1338 		off += sizeof(ip6);
1339 		return off;
1340 
1341 	case IPPROTO_FRAGMENT:
1342 		/*
1343 		 * terminate parsing if it is not the first fragment,
1344 		 * it does not make sense to parse through it.
1345 		 */
1346 		if (m->m_pkthdr.len < off + sizeof(fh))
1347 			return -1;
1348 		m_copydata(m, off, sizeof(fh), &fh);
1349 		if ((fh.ip6f_offlg & IP6F_OFF_MASK) != 0)
1350 			return -1;
1351 		if (nxtp)
1352 			*nxtp = fh.ip6f_nxt;
1353 		off += sizeof(struct ip6_frag);
1354 		return off;
1355 
1356 	case IPPROTO_AH:
1357 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1358 			return -1;
1359 		m_copydata(m, off, sizeof(ip6e), &ip6e);
1360 		if (nxtp)
1361 			*nxtp = ip6e.ip6e_nxt;
1362 		off += (ip6e.ip6e_len + 2) << 2;
1363 		if (m->m_pkthdr.len < off)
1364 			return -1;
1365 		return off;
1366 
1367 	case IPPROTO_HOPOPTS:
1368 	case IPPROTO_ROUTING:
1369 	case IPPROTO_DSTOPTS:
1370 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1371 			return -1;
1372 		m_copydata(m, off, sizeof(ip6e), &ip6e);
1373 		if (nxtp)
1374 			*nxtp = ip6e.ip6e_nxt;
1375 		off += (ip6e.ip6e_len + 1) << 3;
1376 		if (m->m_pkthdr.len < off)
1377 			return -1;
1378 		return off;
1379 
1380 	case IPPROTO_NONE:
1381 	case IPPROTO_ESP:
1382 	case IPPROTO_IPCOMP:
1383 		/* give up */
1384 		return -1;
1385 
1386 	default:
1387 		return -1;
1388 	}
1389 
1390 	return -1;
1391 }
1392 
1393 /*
1394  * get offset for the last header in the chain.  m will be kept untainted.
1395  */
1396 int
1397 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1398 {
1399 	int newoff;
1400 	int nxt;
1401 
1402 	if (!nxtp) {
1403 		nxt = -1;
1404 		nxtp = &nxt;
1405 	}
1406 	while (1) {
1407 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1408 		if (newoff < 0)
1409 			return off;
1410 		else if (newoff < off)
1411 			return -1;	/* invalid */
1412 		else if (newoff == off)
1413 			return newoff;
1414 
1415 		off = newoff;
1416 		proto = *nxtp;
1417 	}
1418 }
1419 
1420 /*
1421  * System control for IP6
1422  */
1423 
1424 const u_char inet6ctlerrmap[PRC_NCMDS] = {
1425 	0,		0,		0,		0,
1426 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1427 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1428 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1429 	0,		0,		0,		0,
1430 	ENOPROTOOPT
1431 };
1432 
1433 #ifdef MROUTING
1434 extern int ip6_mrtproto;
1435 #endif
1436 
1437 const struct sysctl_bounded_args ipv6ctl_vars[] = {
1438 	{ IPV6CTL_DAD_PENDING, &ip6_dad_pending, SYSCTL_INT_READONLY },
1439 #ifdef MROUTING
1440 	{ IPV6CTL_MRTPROTO, &ip6_mrtproto, SYSCTL_INT_READONLY },
1441 #endif
1442 	{ IPV6CTL_FORWARDING, &ip6_forwarding, 0, 1 },
1443 	{ IPV6CTL_SENDREDIRECTS, &ip6_sendredirects, 0, 1 },
1444 	{ IPV6CTL_DEFHLIM, &ip6_defhlim, 0, 255 },
1445 	{ IPV6CTL_MAXFRAGPACKETS, &ip6_maxfragpackets, 0, 1000 },
1446 	{ IPV6CTL_LOG_INTERVAL, &ip6_log_interval, 0, INT_MAX },
1447 	{ IPV6CTL_HDRNESTLIMIT, &ip6_hdrnestlimit, 0, 100 },
1448 	{ IPV6CTL_DAD_COUNT, &ip6_dad_count, 0, 10 },
1449 	{ IPV6CTL_AUTO_FLOWLABEL, &ip6_auto_flowlabel, 0, 1 },
1450 	{ IPV6CTL_DEFMCASTHLIM, &ip6_defmcasthlim, 0, 255 },
1451 	{ IPV6CTL_USE_DEPRECATED, &ip6_use_deprecated, 0, 1 },
1452 	{ IPV6CTL_MAXFRAGS, &ip6_maxfrags, 0, 1000 },
1453 	{ IPV6CTL_MFORWARDING, &ip6_mforwarding, 0, 1 },
1454 	{ IPV6CTL_MULTIPATH, &ip6_multipath, 0, 1 },
1455 	{ IPV6CTL_MCAST_PMTU, &ip6_mcast_pmtu, 0, 1 },
1456 	{ IPV6CTL_NEIGHBORGCTHRESH, &ip6_neighborgcthresh, -1, 5 * 2048 },
1457 	{ IPV6CTL_MAXDYNROUTES, &ip6_maxdynroutes, -1, 5 * 4096 },
1458 };
1459 
1460 int
1461 ip6_sysctl_ip6stat(void *oldp, size_t *oldlenp, void *newp)
1462 {
1463 	struct ip6stat *ip6stat;
1464 	int ret;
1465 
1466 	CTASSERT(sizeof(*ip6stat) == (ip6s_ncounters * sizeof(uint64_t)));
1467 
1468 	ip6stat = malloc(sizeof(*ip6stat), M_TEMP, M_WAITOK);
1469 	counters_read(ip6counters, (uint64_t *)ip6stat, ip6s_ncounters, NULL);
1470 	ret = sysctl_rdstruct(oldp, oldlenp, newp,
1471 	    ip6stat, sizeof(*ip6stat));
1472 	free(ip6stat, M_TEMP, sizeof(*ip6stat));
1473 
1474 	return (ret);
1475 }
1476 
1477 int
1478 ip6_sysctl_soiikey(void *oldp, size_t *oldlenp, void *newp, size_t newlen)
1479 {
1480 	uint8_t oldkey[IP6_SOIIKEY_LEN];
1481 	int error;
1482 
1483 	error = suser(curproc);
1484 	if (error != 0)
1485 		return (error);
1486 
1487 	memcpy(oldkey, ip6_soiikey, sizeof(oldkey));
1488 
1489 	error = sysctl_struct(oldp, oldlenp, newp, newlen, ip6_soiikey,
1490 	    sizeof(ip6_soiikey));
1491 
1492 	return (error);
1493 }
1494 
1495 int
1496 ip6_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
1497     void *newp, size_t newlen)
1498 {
1499 #ifdef MROUTING
1500 	extern struct mrt6stat mrt6stat;
1501 #endif
1502 	int error;
1503 
1504 	/* Almost all sysctl names at this level are terminal. */
1505 	if (namelen != 1 && name[0] != IPV6CTL_IFQUEUE)
1506 		return (ENOTDIR);
1507 
1508 	switch (name[0]) {
1509 	case IPV6CTL_STATS:
1510 		return (ip6_sysctl_ip6stat(oldp, oldlenp, newp));
1511 #ifdef MROUTING
1512 	case IPV6CTL_MRTSTATS:
1513 		if (newp != NULL)
1514 			return (EPERM);
1515 		NET_LOCK();
1516 		error = sysctl_struct(oldp, oldlenp, newp, newlen,
1517 		    &mrt6stat, sizeof(mrt6stat));
1518 		NET_UNLOCK();
1519 		return (error);
1520 	case IPV6CTL_MRTMIF:
1521 		if (newp)
1522 			return (EPERM);
1523 		NET_LOCK();
1524 		error = mrt6_sysctl_mif(oldp, oldlenp);
1525 		NET_UNLOCK();
1526 		return (error);
1527 	case IPV6CTL_MRTMFC:
1528 		if (newp)
1529 			return (EPERM);
1530 		NET_LOCK();
1531 		error = mrt6_sysctl_mfc(oldp, oldlenp);
1532 		NET_UNLOCK();
1533 		return (error);
1534 #else
1535 	case IPV6CTL_MRTSTATS:
1536 	case IPV6CTL_MRTPROTO:
1537 	case IPV6CTL_MRTMIF:
1538 	case IPV6CTL_MRTMFC:
1539 		return (EOPNOTSUPP);
1540 #endif
1541 	case IPV6CTL_MTUDISCTIMEOUT:
1542 		NET_LOCK();
1543 		error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
1544 		    &ip6_mtudisc_timeout, 0, INT_MAX);
1545 		rt_timer_queue_change(&icmp6_mtudisc_timeout_q,
1546 		    ip6_mtudisc_timeout);
1547 		NET_UNLOCK();
1548 		return (error);
1549 	case IPV6CTL_IFQUEUE:
1550 		return (sysctl_niq(name + 1, namelen - 1,
1551 		    oldp, oldlenp, newp, newlen, &ip6intrq));
1552 	case IPV6CTL_SOIIKEY:
1553 		return (ip6_sysctl_soiikey(oldp, oldlenp, newp, newlen));
1554 	default:
1555 		NET_LOCK();
1556 		error = sysctl_bounded_arr(ipv6ctl_vars, nitems(ipv6ctl_vars),
1557 		    name, namelen, oldp, oldlenp, newp, newlen);
1558 		NET_UNLOCK();
1559 		return (error);
1560 	}
1561 	/* NOTREACHED */
1562 }
1563 
1564 void
1565 ip6_send_dispatch(void *xmq)
1566 {
1567 	struct mbuf_queue *mq = xmq;
1568 	struct mbuf *m;
1569 	struct mbuf_list ml;
1570 
1571 	mq_delist(mq, &ml);
1572 	if (ml_empty(&ml))
1573 		return;
1574 
1575 	NET_LOCK_SHARED();
1576 	while ((m = ml_dequeue(&ml)) != NULL) {
1577 		ip6_output(m, NULL, NULL, 0, NULL, NULL);
1578 	}
1579 	NET_UNLOCK_SHARED();
1580 }
1581 
1582 void
1583 ip6_send(struct mbuf *m)
1584 {
1585 	mq_enqueue(&ip6send_mq, m);
1586 	task_add(net_tq(0), &ip6send_task);
1587 }
1588