xref: /openbsd-src/sys/netinet6/ip6_input.c (revision 25c4e8bd056e974b28f4a0ffd39d76c190a56013)
1 /*	$OpenBSD: ip6_input.c,v 1.248 2022/06/29 22:45:24 bluhm 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_local(struct mbuf **, int *, int, int);
124 int ip6_check_rh0hdr(struct mbuf *, int *);
125 int ip6_hbhchcheck(struct mbuf **, int *, int *);
126 int ip6_hopopts_input(struct mbuf **, int *, u_int32_t *, u_int32_t *);
127 struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
128 int ip6_sysctl_soiikey(void *, size_t *, void *, size_t);
129 
130 static struct mbuf_queue	ip6send_mq;
131 
132 static void ip6_send_dispatch(void *);
133 static struct task ip6send_task =
134 	TASK_INITIALIZER(ip6_send_dispatch, &ip6send_mq);
135 
136 /*
137  * IP6 initialization: fill in IP6 protocol switch table.
138  * All protocols not implemented in kernel go to raw IP6 protocol handler.
139  */
140 void
141 ip6_init(void)
142 {
143 	const struct protosw *pr;
144 	int i;
145 
146 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
147 	if (pr == NULL)
148 		panic("%s", __func__);
149 	for (i = 0; i < IPPROTO_MAX; i++)
150 		ip6_protox[i] = pr - inet6sw;
151 	for (pr = inet6domain.dom_protosw;
152 	    pr < inet6domain.dom_protoswNPROTOSW; pr++)
153 		if (pr->pr_domain->dom_family == PF_INET6 &&
154 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW &&
155 		    pr->pr_protocol < IPPROTO_MAX)
156 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
157 	ip6_randomid_init();
158 	nd6_init();
159 	frag6_init();
160 
161 	mq_init(&ip6send_mq, 64, IPL_SOFTNET);
162 
163 	ip6counters = counters_alloc(ip6s_ncounters);
164 #ifdef MROUTING
165 	rt_timer_queue_init(&ip6_mrouterq, MCAST_EXPIRE_TIMEOUT,
166 	    &mf6c_expire_route);
167 #endif
168 }
169 
170 /*
171  * Enqueue packet for local delivery.  Queuing is used as a boundary
172  * between the network layer (input/forward path) running with shared
173  * NET_RLOCK_IN_SOFTNET() and the transport layer needing it exclusively.
174  */
175 int
176 ip6_ours(struct mbuf **mp, int *offp, int nxt, int af)
177 {
178 	/* We are already in a IPv4/IPv6 local deliver loop. */
179 	if (af != AF_UNSPEC)
180 		return ip6_local(mp, offp, nxt, af);
181 
182 	niq_enqueue(&ip6intrq, *mp);
183 	*mp = NULL;
184 	return IPPROTO_DONE;
185 }
186 
187 /*
188  * Dequeue and process locally delivered packets.
189  */
190 void
191 ip6intr(void)
192 {
193 	struct mbuf *m;
194 	int off, nxt;
195 
196 	while ((m = niq_dequeue(&ip6intrq)) != NULL) {
197 #ifdef DIAGNOSTIC
198 		if ((m->m_flags & M_PKTHDR) == 0)
199 			panic("ip6intr no HDR");
200 #endif
201 		off = 0;
202 		nxt = ip6_local(&m, &off, IPPROTO_IPV6, AF_UNSPEC);
203 		KASSERT(nxt == IPPROTO_DONE);
204 	}
205 }
206 
207 void
208 ipv6_input(struct ifnet *ifp, struct mbuf *m)
209 {
210 	int off, nxt;
211 
212 	off = 0;
213 	nxt = ip6_input_if(&m, &off, IPPROTO_IPV6, AF_UNSPEC, ifp);
214 	KASSERT(nxt == IPPROTO_DONE);
215 }
216 
217 struct mbuf *
218 ipv6_check(struct ifnet *ifp, struct mbuf *m)
219 {
220 	struct ip6_hdr *ip6;
221 
222 	if (m->m_len < sizeof(*ip6)) {
223 		m = m_pullup(m, sizeof(*ip6));
224 		if (m == NULL) {
225 			ip6stat_inc(ip6s_toosmall);
226 			return (NULL);
227 		}
228 	}
229 
230 	ip6 = mtod(m, struct ip6_hdr *);
231 
232 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
233 		ip6stat_inc(ip6s_badvers);
234 		goto bad;
235 	}
236 
237 	/*
238 	 * Check against address spoofing/corruption.
239 	 */
240 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
241 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
242 		/*
243 		 * XXX: "badscope" is not very suitable for a multicast source.
244 		 */
245 		ip6stat_inc(ip6s_badscope);
246 		goto bad;
247 	}
248 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
249 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
250 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
251 		ip6stat_inc(ip6s_badscope);
252 		goto bad;
253 	}
254 	/* Drop packets if interface ID portion is already filled. */
255 	if (((IN6_IS_SCOPE_EMBED(&ip6->ip6_src) && ip6->ip6_src.s6_addr16[1]) ||
256 	    (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst) && ip6->ip6_dst.s6_addr16[1])) &&
257 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
258 		ip6stat_inc(ip6s_badscope);
259 		goto bad;
260 	}
261 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
262 	    !(m->m_flags & M_LOOP)) {
263 		/*
264 		 * In this case, the packet should come from the loopback
265 		 * interface.  However, we cannot just check the if_flags,
266 		 * because ip6_mloopback() passes the "actual" interface
267 		 * as the outgoing/incoming interface.
268 		 */
269 		ip6stat_inc(ip6s_badscope);
270 		goto bad;
271 	}
272 
273 	/*
274 	 * The following check is not documented in specs.  A malicious
275 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
276 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
277 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
278 	 *
279 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
280 	 * support IPv4-less kernel compilation, we cannot support SIIT
281 	 * environment at all.  So, it makes more sense for us to reject any
282 	 * malicious packets for non-SIIT environment, than try to do a
283 	 * partial support for SIIT environment.
284 	 */
285 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
286 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
287 		ip6stat_inc(ip6s_badscope);
288 		goto bad;
289 	}
290 
291 	/*
292 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
293 	 *
294 	 * The code forbids automatic tunneling as per RFC4213.
295 	 */
296 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
297 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
298 		ip6stat_inc(ip6s_badscope);
299 		goto bad;
300 	}
301 
302 	return (m);
303 bad:
304 	m_freem(m);
305 	return (NULL);
306 }
307 
308 int
309 ip6_input_if(struct mbuf **mp, int *offp, int nxt, int af, struct ifnet *ifp)
310 {
311 	struct mbuf *m;
312 	struct ip6_hdr *ip6;
313 	struct sockaddr_in6 sin6;
314 	struct rtentry *rt = NULL;
315 	int ours = 0;
316 	u_int16_t src_scope, dst_scope;
317 #if NPF > 0
318 	struct in6_addr odst;
319 #endif
320 	int srcrt = 0;
321 
322 	KASSERT(*offp == 0);
323 
324 	ip6stat_inc(ip6s_total);
325 
326 	m = *mp = ipv6_check(ifp, *mp);
327 	if (m == NULL)
328 		goto bad;
329 
330 	ip6 = mtod(m, struct ip6_hdr *);
331 
332 #if NCARP > 0
333 	if (carp_lsdrop(ifp, m, AF_INET6, ip6->ip6_src.s6_addr32,
334 	    ip6->ip6_dst.s6_addr32, (ip6->ip6_nxt == IPPROTO_ICMPV6 ? 0 : 1)))
335 		goto bad;
336 #endif
337 	ip6stat_inc(ip6s_nxthist + ip6->ip6_nxt);
338 
339 	/*
340 	 * If the packet has been received on a loopback interface it
341 	 * can be destined to any local address, not necessarily to
342 	 * an address configured on `ifp'.
343 	 */
344 	if (ifp->if_flags & IFF_LOOPBACK) {
345 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
346 			src_scope = ip6->ip6_src.s6_addr16[1];
347 			ip6->ip6_src.s6_addr16[1] = 0;
348 		}
349 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
350 			dst_scope = ip6->ip6_dst.s6_addr16[1];
351 			ip6->ip6_dst.s6_addr16[1] = 0;
352 		}
353 	}
354 
355 #if NPF > 0
356 	/*
357 	 * Packet filter
358 	 */
359 	odst = ip6->ip6_dst;
360 	if (pf_test(AF_INET6, PF_IN, ifp, mp) != PF_PASS)
361 		goto bad;
362 	m = *mp;
363 	if (m == NULL)
364 		goto bad;
365 
366 	ip6 = mtod(m, struct ip6_hdr *);
367 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
368 #endif
369 
370 	/*
371 	 * Without embedded scope ID we cannot find link-local
372 	 * addresses in the routing table.
373 	 */
374 	if (ifp->if_flags & IFF_LOOPBACK) {
375 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
376 			ip6->ip6_src.s6_addr16[1] = src_scope;
377 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
378 			ip6->ip6_dst.s6_addr16[1] = dst_scope;
379 	} else {
380 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
381 			ip6->ip6_src.s6_addr16[1] = htons(ifp->if_index);
382 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
383 			ip6->ip6_dst.s6_addr16[1] = htons(ifp->if_index);
384 	}
385 
386 	/*
387 	 * Be more secure than RFC5095 and scan for type 0 routing headers.
388 	 * If pf has already scanned the header chain, do not do it twice.
389 	 */
390 	if (!(m->m_pkthdr.pf.flags & PF_TAG_PROCESSED) &&
391 	    ip6_check_rh0hdr(m, offp)) {
392 		ip6stat_inc(ip6s_badoptions);
393 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, *offp);
394 		m = *mp = NULL;
395 		goto bad;
396 	}
397 
398 #if NPF > 0
399 	if (pf_ouraddr(m) == 1) {
400 		nxt = ip6_ours(mp, offp, nxt, af);
401 		goto out;
402 	}
403 #endif
404 
405 	/*
406 	 * Multicast check
407 	 */
408 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
409 		/*
410 		 * Make sure M_MCAST is set.  It should theoretically
411 		 * already be there, but let's play safe because upper
412 		 * layers check for this flag.
413 		 */
414 		m->m_flags |= M_MCAST;
415 
416 		/*
417 		 * See if we belong to the destination multicast group on the
418 		 * arrival interface.
419 		 */
420 		if (in6_hasmulti(&ip6->ip6_dst, ifp))
421 			ours = 1;
422 
423 #ifdef MROUTING
424 		if (ip6_mforwarding && ip6_mrouter[ifp->if_rdomain]) {
425 			int error;
426 
427 			nxt = ip6_hbhchcheck(&m, offp, &ours);
428 			if (nxt == IPPROTO_DONE)
429 				goto out;
430 
431 			ip6 = mtod(m, struct ip6_hdr *);
432 
433 			/*
434 			 * If we are acting as a multicast router, all
435 			 * incoming multicast packets are passed to the
436 			 * kernel-level multicast forwarding function.
437 			 * The packet is returned (relatively) intact; if
438 			 * ip6_mforward() returns a non-zero value, the packet
439 			 * must be discarded, else it may be accepted below.
440 			 */
441 			KERNEL_LOCK();
442 			error = ip6_mforward(ip6, ifp, m);
443 			KERNEL_UNLOCK();
444 			if (error) {
445 				ip6stat_inc(ip6s_cantforward);
446 				goto bad;
447 			}
448 
449 			if (ours) {
450 				if (af == AF_UNSPEC)
451 					nxt = ip_deliver(mp, offp, nxt,
452 					    AF_INET6);
453 				goto out;
454 			}
455 			goto bad;
456 		}
457 #endif
458 		if (!ours) {
459 			ip6stat_inc(ip6s_notmember);
460 			if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
461 				ip6stat_inc(ip6s_cantforward);
462 			goto bad;
463 		}
464 		nxt = ip6_ours(mp, offp, nxt, af);
465 		goto out;
466 	}
467 
468 
469 	/*
470 	 *  Unicast check
471 	 */
472 	memset(&sin6, 0, sizeof(struct sockaddr_in6));
473 	sin6.sin6_len = sizeof(struct sockaddr_in6);
474 	sin6.sin6_family = AF_INET6;
475 	sin6.sin6_addr = ip6->ip6_dst;
476 	rt = rtalloc_mpath(sin6tosa(&sin6), &ip6->ip6_src.s6_addr32[0],
477 	    m->m_pkthdr.ph_rtableid);
478 
479 	/*
480 	 * Accept the packet if the route to the destination is marked
481 	 * as local.
482 	 */
483 	if (rtisvalid(rt) && ISSET(rt->rt_flags, RTF_LOCAL)) {
484 		struct in6_ifaddr *ia6 = ifatoia6(rt->rt_ifa);
485 
486 		if (ip6_forwarding == 0 && rt->rt_ifidx != ifp->if_index &&
487 		    !((ifp->if_flags & IFF_LOOPBACK) ||
488 		    (ifp->if_type == IFT_ENC) ||
489 		    (m->m_pkthdr.pf.flags & PF_TAG_TRANSLATE_LOCALHOST))) {
490 			/* received on wrong interface */
491 #if NCARP > 0
492 			struct ifnet *out_if;
493 
494 			/*
495 			 * Virtual IPs on carp interfaces need to be checked
496 			 * also against the parent interface and other carp
497 			 * interfaces sharing the same parent.
498 			 */
499 			out_if = if_get(rt->rt_ifidx);
500 			if (!(out_if && carp_strict_addr_chk(out_if, ifp))) {
501 				ip6stat_inc(ip6s_wrongif);
502 				if_put(out_if);
503 				goto bad;
504 			}
505 			if_put(out_if);
506 #else
507 			ip6stat_inc(ip6s_wrongif);
508 			goto bad;
509 #endif
510 		}
511 		/*
512 		 * packets to a tentative, duplicated, or somehow invalid
513 		 * address must not be accepted.
514 		 */
515 		if ((ia6->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED))) {
516 			char src[INET6_ADDRSTRLEN], dst[INET6_ADDRSTRLEN];
517 
518 			inet_ntop(AF_INET6, &ip6->ip6_src, src, sizeof(src));
519 			inet_ntop(AF_INET6, &ip6->ip6_dst, dst, sizeof(dst));
520 			/* address is not ready, so discard the packet. */
521 			nd6log((LOG_INFO,
522 			    "%s: packet to an unready address %s->%s\n",
523 			    __func__, src, dst));
524 
525 			goto bad;
526 		} else {
527 			nxt = ip6_ours(mp, offp, nxt, af);
528 			goto out;
529 		}
530 	}
531 
532 #if NCARP > 0
533 	if (ip6->ip6_nxt == IPPROTO_ICMPV6 &&
534 	    carp_lsdrop(ifp, m, AF_INET6, ip6->ip6_src.s6_addr32,
535 	    ip6->ip6_dst.s6_addr32, 1))
536 		goto bad;
537 #endif
538 	/*
539 	 * Now there is no reason to process the packet if it's not our own
540 	 * and we're not a router.
541 	 */
542 	if (!ip6_forwarding) {
543 		ip6stat_inc(ip6s_cantforward);
544 		goto bad;
545 	}
546 
547 	nxt = ip6_hbhchcheck(&m, offp, &ours);
548 	if (nxt == IPPROTO_DONE)
549 		goto out;
550 
551 	if (ours) {
552 		if (af == AF_UNSPEC)
553 			nxt = ip_deliver(mp, offp, nxt, AF_INET6);
554 		goto out;
555 	}
556 
557 #ifdef IPSEC
558 	if (ipsec_in_use) {
559 		int rv;
560 
561 		rv = ipsec_forward_check(m, *offp, AF_INET6);
562 		if (rv != 0) {
563 			ip6stat_inc(ip6s_cantforward);
564 			goto bad;
565 		}
566 		/*
567 		 * Fall through, forward packet. Outbound IPsec policy
568 		 * checking will occur in ip6_forward().
569 		 */
570 	}
571 #endif /* IPSEC */
572 
573 	ip6_forward(m, rt, srcrt);
574 	*mp = NULL;
575 	return IPPROTO_DONE;
576  bad:
577 	nxt = IPPROTO_DONE;
578 	m_freemp(mp);
579  out:
580 	rtfree(rt);
581 	return nxt;
582 }
583 
584 int
585 ip6_local(struct mbuf **mp, int *offp, int nxt, int af)
586 {
587 	NET_ASSERT_WLOCKED();
588 
589 	nxt = ip6_hbhchcheck(mp, offp, NULL);
590 	if (nxt == IPPROTO_DONE)
591 		return IPPROTO_DONE;
592 
593 	/* Check whether we are already in a IPv4/IPv6 local deliver loop. */
594 	if (af == AF_UNSPEC)
595 		nxt = ip_deliver(mp, offp, nxt, AF_INET6);
596 	return nxt;
597 }
598 
599 /* On error free mbuf and return IPPROTO_DONE. */
600 int
601 ip6_hbhchcheck(struct mbuf **mp, int *offp, int *oursp)
602 {
603 	struct ip6_hdr *ip6;
604 	u_int32_t plen, rtalert = ~0;
605 	int nxt;
606 
607 	ip6 = mtod(*mp, struct ip6_hdr *);
608 
609 	/*
610 	 * Process Hop-by-Hop options header if it's contained.
611 	 * m may be modified in ip6_hopopts_input().
612 	 * If a JumboPayload option is included, plen will also be modified.
613 	 */
614 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
615 	*offp = sizeof(struct ip6_hdr);
616 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
617 		struct ip6_hbh *hbh;
618 
619 		if (ip6_hopopts_input(mp, offp, &plen, &rtalert))
620 			goto bad;	/* m have already been freed */
621 
622 		/* adjust pointer */
623 		ip6 = mtod(*mp, struct ip6_hdr *);
624 
625 		/*
626 		 * if the payload length field is 0 and the next header field
627 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
628 		 * option MUST be included.
629 		 */
630 		if (ip6->ip6_plen == 0 && plen == 0) {
631 			/*
632 			 * Note that if a valid jumbo payload option is
633 			 * contained, ip6_hopopts_input() must set a valid
634 			 * (non-zero) payload length to the variable plen.
635 			 */
636 			ip6stat_inc(ip6s_badoptions);
637 			icmp6_error(*mp, ICMP6_PARAM_PROB,
638 				    ICMP6_PARAMPROB_HEADER,
639 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
640 			goto bad;
641 		}
642 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, *mp,
643 		    sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
644 		if (hbh == NULL) {
645 			ip6stat_inc(ip6s_tooshort);
646 			goto bad;
647 		}
648 		nxt = hbh->ip6h_nxt;
649 
650 		/*
651 		 * accept the packet if a router alert option is included
652 		 * and we act as an IPv6 router.
653 		 */
654 		if (rtalert != ~0 && ip6_forwarding && oursp != NULL)
655 			*oursp = 1;
656 	} else
657 		nxt = ip6->ip6_nxt;
658 
659 	/*
660 	 * Check that the amount of data in the buffers
661 	 * is as at least much as the IPv6 header would have us expect.
662 	 * Trim mbufs if longer than we expect.
663 	 * Drop packet if shorter than we expect.
664 	 */
665 	if ((*mp)->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
666 		ip6stat_inc(ip6s_tooshort);
667 		m_freemp(mp);
668 		goto bad;
669 	}
670 	if ((*mp)->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
671 		if ((*mp)->m_len == (*mp)->m_pkthdr.len) {
672 			(*mp)->m_len = sizeof(struct ip6_hdr) + plen;
673 			(*mp)->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
674 		} else {
675 			m_adj((*mp), sizeof(struct ip6_hdr) + plen -
676 			    (*mp)->m_pkthdr.len);
677 		}
678 	}
679 
680 	return nxt;
681  bad:
682 	return IPPROTO_DONE;
683 }
684 
685 /* scan packet for RH0 routing header. Mostly stolen from pf.c:pf_test() */
686 int
687 ip6_check_rh0hdr(struct mbuf *m, int *offp)
688 {
689 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
690 	struct ip6_rthdr rthdr;
691 	struct ip6_ext opt6;
692 	u_int8_t proto = ip6->ip6_nxt;
693 	int done = 0, lim, off, rh_cnt = 0;
694 
695 	off = ((caddr_t)ip6 - m->m_data) + sizeof(struct ip6_hdr);
696 	lim = min(m->m_pkthdr.len, ntohs(ip6->ip6_plen) + sizeof(*ip6));
697 	do {
698 		switch (proto) {
699 		case IPPROTO_ROUTING:
700 			*offp = off;
701 			if (rh_cnt++) {
702 				/* more than one rh header present */
703 				return (1);
704 			}
705 
706 			if (off + sizeof(rthdr) > lim) {
707 				/* packet to short to make sense */
708 				return (1);
709 			}
710 
711 			m_copydata(m, off, sizeof(rthdr), &rthdr);
712 
713 			if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
714 				*offp += offsetof(struct ip6_rthdr, ip6r_type);
715 				return (1);
716 			}
717 
718 			off += (rthdr.ip6r_len + 1) * 8;
719 			proto = rthdr.ip6r_nxt;
720 			break;
721 		case IPPROTO_AH:
722 		case IPPROTO_HOPOPTS:
723 		case IPPROTO_DSTOPTS:
724 			/* get next header and header length */
725 			if (off + sizeof(opt6) > lim) {
726 				/*
727 				 * Packet to short to make sense, we could
728 				 * reject the packet but as a router we
729 				 * should not do that so forward it.
730 				 */
731 				return (0);
732 			}
733 
734 			m_copydata(m, off, sizeof(opt6), &opt6);
735 
736 			if (proto == IPPROTO_AH)
737 				off += (opt6.ip6e_len + 2) * 4;
738 			else
739 				off += (opt6.ip6e_len + 1) * 8;
740 			proto = opt6.ip6e_nxt;
741 			break;
742 		case IPPROTO_FRAGMENT:
743 		default:
744 			/* end of header stack */
745 			done = 1;
746 			break;
747 		}
748 	} while (!done);
749 
750 	return (0);
751 }
752 
753 /*
754  * Hop-by-Hop options header processing. If a valid jumbo payload option is
755  * included, the real payload length will be stored in plenp.
756  * On error free mbuf and return -1.
757  *
758  * rtalertp - XXX: should be stored in a more smart way
759  */
760 int
761 ip6_hopopts_input(struct mbuf **mp, int *offp, u_int32_t *plenp,
762     u_int32_t *rtalertp)
763 {
764 	int off = *offp, hbhlen;
765 	struct ip6_hbh *hbh;
766 
767 	/* validation of the length of the header */
768 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, *mp,
769 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
770 	if (hbh == NULL) {
771 		ip6stat_inc(ip6s_tooshort);
772 		return -1;
773 	}
774 	hbhlen = (hbh->ip6h_len + 1) << 3;
775 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, *mp, sizeof(struct ip6_hdr),
776 		hbhlen);
777 	if (hbh == NULL) {
778 		ip6stat_inc(ip6s_tooshort);
779 		return -1;
780 	}
781 	off += hbhlen;
782 	hbhlen -= sizeof(struct ip6_hbh);
783 
784 	if (ip6_process_hopopts(mp, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
785 				hbhlen, rtalertp, plenp) < 0)
786 		return (-1);
787 
788 	*offp = off;
789 	return (0);
790 }
791 
792 /*
793  * Search header for all Hop-by-hop options and process each option.
794  * This function is separate from ip6_hopopts_input() in order to
795  * handle a case where the sending node itself process its hop-by-hop
796  * options header. In such a case, the function is called from ip6_output().
797  * On error free mbuf and return -1.
798  *
799  * The function assumes that hbh header is located right after the IPv6 header
800  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
801  * opthead + hbhlen is located in continuous memory region.
802  */
803 int
804 ip6_process_hopopts(struct mbuf **mp, u_int8_t *opthead, int hbhlen,
805     u_int32_t *rtalertp, u_int32_t *plenp)
806 {
807 	struct ip6_hdr *ip6;
808 	int optlen = 0;
809 	u_int8_t *opt = opthead;
810 	u_int16_t rtalert_val;
811 	u_int32_t jumboplen;
812 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
813 
814 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
815 		switch (*opt) {
816 		case IP6OPT_PAD1:
817 			optlen = 1;
818 			break;
819 		case IP6OPT_PADN:
820 			if (hbhlen < IP6OPT_MINLEN) {
821 				ip6stat_inc(ip6s_toosmall);
822 				goto bad;
823 			}
824 			optlen = *(opt + 1) + 2;
825 			break;
826 		case IP6OPT_ROUTER_ALERT:
827 			/* XXX may need check for alignment */
828 			if (hbhlen < IP6OPT_RTALERT_LEN) {
829 				ip6stat_inc(ip6s_toosmall);
830 				goto bad;
831 			}
832 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
833 				/* XXX stat */
834 				icmp6_error(*mp, ICMP6_PARAM_PROB,
835 				    ICMP6_PARAMPROB_HEADER,
836 				    erroff + opt + 1 - opthead);
837 				return (-1);
838 			}
839 			optlen = IP6OPT_RTALERT_LEN;
840 			memcpy((caddr_t)&rtalert_val, (caddr_t)(opt + 2), 2);
841 			*rtalertp = ntohs(rtalert_val);
842 			break;
843 		case IP6OPT_JUMBO:
844 			/* XXX may need check for alignment */
845 			if (hbhlen < IP6OPT_JUMBO_LEN) {
846 				ip6stat_inc(ip6s_toosmall);
847 				goto bad;
848 			}
849 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
850 				/* XXX stat */
851 				icmp6_error(*mp, ICMP6_PARAM_PROB,
852 				    ICMP6_PARAMPROB_HEADER,
853 				    erroff + opt + 1 - opthead);
854 				return (-1);
855 			}
856 			optlen = IP6OPT_JUMBO_LEN;
857 
858 			/*
859 			 * IPv6 packets that have non 0 payload length
860 			 * must not contain a jumbo payload option.
861 			 */
862 			ip6 = mtod(*mp, struct ip6_hdr *);
863 			if (ip6->ip6_plen) {
864 				ip6stat_inc(ip6s_badoptions);
865 				icmp6_error(*mp, ICMP6_PARAM_PROB,
866 				    ICMP6_PARAMPROB_HEADER,
867 				    erroff + opt - opthead);
868 				return (-1);
869 			}
870 
871 			/*
872 			 * We may see jumbolen in unaligned location, so
873 			 * we'd need to perform memcpy().
874 			 */
875 			memcpy(&jumboplen, opt + 2, sizeof(jumboplen));
876 			jumboplen = (u_int32_t)htonl(jumboplen);
877 
878 #if 1
879 			/*
880 			 * if there are multiple jumbo payload options,
881 			 * *plenp will be non-zero and the packet will be
882 			 * rejected.
883 			 * the behavior may need some debate in ipngwg -
884 			 * multiple options does not make sense, however,
885 			 * there's no explicit mention in specification.
886 			 */
887 			if (*plenp != 0) {
888 				ip6stat_inc(ip6s_badoptions);
889 				icmp6_error(*mp, ICMP6_PARAM_PROB,
890 				    ICMP6_PARAMPROB_HEADER,
891 				    erroff + opt + 2 - opthead);
892 				return (-1);
893 			}
894 #endif
895 
896 			/*
897 			 * jumbo payload length must be larger than 65535.
898 			 */
899 			if (jumboplen <= IPV6_MAXPACKET) {
900 				ip6stat_inc(ip6s_badoptions);
901 				icmp6_error(*mp, ICMP6_PARAM_PROB,
902 				    ICMP6_PARAMPROB_HEADER,
903 				    erroff + opt + 2 - opthead);
904 				return (-1);
905 			}
906 			*plenp = jumboplen;
907 
908 			break;
909 		default:		/* unknown option */
910 			if (hbhlen < IP6OPT_MINLEN) {
911 				ip6stat_inc(ip6s_toosmall);
912 				goto bad;
913 			}
914 			optlen = ip6_unknown_opt(mp, opt,
915 			    erroff + opt - opthead);
916 			if (optlen == -1)
917 				return (-1);
918 			optlen += 2;
919 			break;
920 		}
921 	}
922 
923 	return (0);
924 
925   bad:
926 	m_freemp(mp);
927 	return (-1);
928 }
929 
930 /*
931  * Unknown option processing.
932  * The third argument `off' is the offset from the IPv6 header to the option,
933  * which allows returning an ICMPv6 error even if the IPv6 header and the
934  * option header are not continuous.
935  * On error free mbuf and return -1.
936  */
937 int
938 ip6_unknown_opt(struct mbuf **mp, u_int8_t *optp, int off)
939 {
940 	struct ip6_hdr *ip6;
941 
942 	switch (IP6OPT_TYPE(*optp)) {
943 	case IP6OPT_TYPE_SKIP: /* ignore the option */
944 		return ((int)*(optp + 1));
945 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
946 		m_freemp(mp);
947 		return (-1);
948 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
949 		ip6stat_inc(ip6s_badoptions);
950 		icmp6_error(*mp, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
951 		return (-1);
952 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
953 		ip6stat_inc(ip6s_badoptions);
954 		ip6 = mtod(*mp, struct ip6_hdr *);
955 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
956 		    ((*mp)->m_flags & (M_BCAST|M_MCAST)))
957 			m_freemp(mp);
958 		else
959 			icmp6_error(*mp, ICMP6_PARAM_PROB,
960 				    ICMP6_PARAMPROB_OPTION, off);
961 		return (-1);
962 	}
963 
964 	m_freemp(mp);		/* XXX: NOTREACHED */
965 	return (-1);
966 }
967 
968 /*
969  * Create the "control" list for this pcb.
970  *
971  * The routine will be called from upper layer handlers like udp_input().
972  * Thus the routine assumes that the caller (udp_input) have already
973  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
974  * very first mbuf on the mbuf chain.
975  * We may want to add some infinite loop prevention or sanity checks for safety.
976  * (This applies only when you are using KAME mbuf chain restriction, i.e.
977  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
978  */
979 void
980 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
981 {
982 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
983 
984 	if (in6p->inp_socket->so_options & SO_TIMESTAMP) {
985 		struct timeval tv;
986 
987 		m_microtime(m, &tv);
988 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
989 		    SCM_TIMESTAMP, SOL_SOCKET);
990 		if (*mp)
991 			mp = &(*mp)->m_next;
992 	}
993 
994 	/* RFC 2292 sec. 5 */
995 	if ((in6p->inp_flags & IN6P_PKTINFO) != 0) {
996 		struct in6_pktinfo pi6;
997 		memcpy(&pi6.ipi6_addr, &ip6->ip6_dst, sizeof(struct in6_addr));
998 		if (IN6_IS_SCOPE_EMBED(&pi6.ipi6_addr))
999 			pi6.ipi6_addr.s6_addr16[1] = 0;
1000 		pi6.ipi6_ifindex = m ? m->m_pkthdr.ph_ifidx : 0;
1001 		*mp = sbcreatecontrol((caddr_t) &pi6,
1002 		    sizeof(struct in6_pktinfo),
1003 		    IPV6_PKTINFO, IPPROTO_IPV6);
1004 		if (*mp)
1005 			mp = &(*mp)->m_next;
1006 	}
1007 
1008 	if ((in6p->inp_flags & IN6P_HOPLIMIT) != 0) {
1009 		int hlim = ip6->ip6_hlim & 0xff;
1010 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1011 		    IPV6_HOPLIMIT, IPPROTO_IPV6);
1012 		if (*mp)
1013 			mp = &(*mp)->m_next;
1014 	}
1015 
1016 	if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
1017 		u_int32_t flowinfo;
1018 		int tclass;
1019 
1020 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1021 		flowinfo >>= 20;
1022 
1023 		tclass = flowinfo & 0xff;
1024 		*mp = sbcreatecontrol((caddr_t)&tclass, sizeof(tclass),
1025 		    IPV6_TCLASS, IPPROTO_IPV6);
1026 		if (*mp)
1027 			mp = &(*mp)->m_next;
1028 	}
1029 
1030 	/*
1031 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1032 	 * privilege for the option (see ip6_ctloutput), but it might be too
1033 	 * strict, since there might be some hop-by-hop options which can be
1034 	 * returned to normal user.
1035 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1036 	 */
1037 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1038 		/*
1039 		 * Check if a hop-by-hop options header is contained in the
1040 		 * received packet, and if so, store the options as ancillary
1041 		 * data. Note that a hop-by-hop options header must be
1042 		 * just after the IPv6 header, which is assured through the
1043 		 * IPv6 input processing.
1044 		 */
1045 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1046 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1047 			struct ip6_hbh *hbh;
1048 			int hbhlen = 0;
1049 			struct mbuf *ext;
1050 
1051 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1052 			    ip6->ip6_nxt);
1053 			if (ext == NULL) {
1054 				ip6stat_inc(ip6s_tooshort);
1055 				return;
1056 			}
1057 			hbh = mtod(ext, struct ip6_hbh *);
1058 			hbhlen = (hbh->ip6h_len + 1) << 3;
1059 			if (hbhlen != ext->m_len) {
1060 				m_freem(ext);
1061 				ip6stat_inc(ip6s_tooshort);
1062 				return;
1063 			}
1064 
1065 			/*
1066 			 * XXX: We copy the whole header even if a
1067 			 * jumbo payload option is included, the option which
1068 			 * is to be removed before returning according to
1069 			 * RFC2292.
1070 			 * Note: this constraint is removed in RFC3542.
1071 			 */
1072 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1073 			    IPV6_HOPOPTS,
1074 			    IPPROTO_IPV6);
1075 			if (*mp)
1076 				mp = &(*mp)->m_next;
1077 			m_freem(ext);
1078 		}
1079 	}
1080 
1081 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1082 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1083 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1084 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1085 
1086 		/*
1087 		 * Search for destination options headers or routing
1088 		 * header(s) through the header chain, and stores each
1089 		 * header as ancillary data.
1090 		 * Note that the order of the headers remains in
1091 		 * the chain of ancillary data.
1092 		 */
1093 		while (1) {	/* is explicit loop prevention necessary? */
1094 			struct ip6_ext *ip6e = NULL;
1095 			int elen;
1096 			struct mbuf *ext = NULL;
1097 
1098 			/*
1099 			 * if it is not an extension header, don't try to
1100 			 * pull it from the chain.
1101 			 */
1102 			switch (nxt) {
1103 			case IPPROTO_DSTOPTS:
1104 			case IPPROTO_ROUTING:
1105 			case IPPROTO_HOPOPTS:
1106 			case IPPROTO_AH: /* is it possible? */
1107 				break;
1108 			default:
1109 				goto loopend;
1110 			}
1111 
1112 			ext = ip6_pullexthdr(m, off, nxt);
1113 			if (ext == NULL) {
1114 				ip6stat_inc(ip6s_tooshort);
1115 				return;
1116 			}
1117 			ip6e = mtod(ext, struct ip6_ext *);
1118 			if (nxt == IPPROTO_AH)
1119 				elen = (ip6e->ip6e_len + 2) << 2;
1120 			else
1121 				elen = (ip6e->ip6e_len + 1) << 3;
1122 			if (elen != ext->m_len) {
1123 				m_freem(ext);
1124 				ip6stat_inc(ip6s_tooshort);
1125 				return;
1126 			}
1127 
1128 			switch (nxt) {
1129 			case IPPROTO_DSTOPTS:
1130 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1131 					break;
1132 
1133 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1134 				    IPV6_DSTOPTS,
1135 				    IPPROTO_IPV6);
1136 				if (*mp)
1137 					mp = &(*mp)->m_next;
1138 				break;
1139 
1140 			case IPPROTO_ROUTING:
1141 				if (!(in6p->inp_flags & IN6P_RTHDR))
1142 					break;
1143 
1144 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1145 				    IPV6_RTHDR,
1146 				    IPPROTO_IPV6);
1147 				if (*mp)
1148 					mp = &(*mp)->m_next;
1149 				break;
1150 
1151 			case IPPROTO_HOPOPTS:
1152 			case IPPROTO_AH: /* is it possible? */
1153 				break;
1154 
1155 			default:
1156 				/*
1157 				 * other cases have been filtered in the above.
1158 				 * none will visit this case.  here we supply
1159 				 * the code just in case (nxt overwritten or
1160 				 * other cases).
1161 				 */
1162 				m_freem(ext);
1163 				goto loopend;
1164 
1165 			}
1166 
1167 			/* proceed with the next header. */
1168 			off += elen;
1169 			nxt = ip6e->ip6e_nxt;
1170 			ip6e = NULL;
1171 			m_freem(ext);
1172 			ext = NULL;
1173 		}
1174 loopend:
1175 		;
1176 	}
1177 }
1178 
1179 /*
1180  * pull single extension header from mbuf chain.  returns single mbuf that
1181  * contains the result, or NULL on error.
1182  */
1183 struct mbuf *
1184 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1185 {
1186 	struct ip6_ext ip6e;
1187 	size_t elen;
1188 	struct mbuf *n;
1189 
1190 #ifdef DIAGNOSTIC
1191 	switch (nxt) {
1192 	case IPPROTO_DSTOPTS:
1193 	case IPPROTO_ROUTING:
1194 	case IPPROTO_HOPOPTS:
1195 	case IPPROTO_AH: /* is it possible? */
1196 		break;
1197 	default:
1198 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1199 	}
1200 #endif
1201 
1202 	if (off + sizeof(ip6e) > m->m_pkthdr.len)
1203 		return NULL;
1204 
1205 	m_copydata(m, off, sizeof(ip6e), &ip6e);
1206 	if (nxt == IPPROTO_AH)
1207 		elen = (ip6e.ip6e_len + 2) << 2;
1208 	else
1209 		elen = (ip6e.ip6e_len + 1) << 3;
1210 
1211 	if (off + elen > m->m_pkthdr.len)
1212 		return NULL;
1213 
1214 	MGET(n, M_DONTWAIT, MT_DATA);
1215 	if (n && elen >= MLEN) {
1216 		MCLGET(n, M_DONTWAIT);
1217 		if ((n->m_flags & M_EXT) == 0) {
1218 			m_free(n);
1219 			n = NULL;
1220 		}
1221 	}
1222 	if (!n)
1223 		return NULL;
1224 
1225 	n->m_len = 0;
1226 	if (elen >= m_trailingspace(n)) {
1227 		m_free(n);
1228 		return NULL;
1229 	}
1230 
1231 	m_copydata(m, off, elen, mtod(n, caddr_t));
1232 	n->m_len = elen;
1233 	return n;
1234 }
1235 
1236 /*
1237  * Get offset to the previous header followed by the header
1238  * currently processed.
1239  */
1240 int
1241 ip6_get_prevhdr(struct mbuf *m, int off)
1242 {
1243 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1244 
1245 	if (off == sizeof(struct ip6_hdr)) {
1246 		return offsetof(struct ip6_hdr, ip6_nxt);
1247 	} else if (off < sizeof(struct ip6_hdr)) {
1248 		panic("%s: off < sizeof(struct ip6_hdr)", __func__);
1249 	} else {
1250 		int len, nlen, nxt;
1251 		struct ip6_ext ip6e;
1252 
1253 		nxt = ip6->ip6_nxt;
1254 		len = sizeof(struct ip6_hdr);
1255 		nlen = 0;
1256 		while (len < off) {
1257 			m_copydata(m, len, sizeof(ip6e), &ip6e);
1258 
1259 			switch (nxt) {
1260 			case IPPROTO_FRAGMENT:
1261 				nlen = sizeof(struct ip6_frag);
1262 				break;
1263 			case IPPROTO_AH:
1264 				nlen = (ip6e.ip6e_len + 2) << 2;
1265 				break;
1266 			default:
1267 				nlen = (ip6e.ip6e_len + 1) << 3;
1268 				break;
1269 			}
1270 			len += nlen;
1271 			nxt = ip6e.ip6e_nxt;
1272 		}
1273 
1274 		return (len - nlen);
1275 	}
1276 }
1277 
1278 /*
1279  * get next header offset.  m will be retained.
1280  */
1281 int
1282 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1283 {
1284 	struct ip6_hdr ip6;
1285 	struct ip6_ext ip6e;
1286 	struct ip6_frag fh;
1287 
1288 	/* just in case */
1289 	if (m == NULL)
1290 		panic("%s: m == NULL", __func__);
1291 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1292 		return -1;
1293 
1294 	switch (proto) {
1295 	case IPPROTO_IPV6:
1296 		if (m->m_pkthdr.len < off + sizeof(ip6))
1297 			return -1;
1298 		m_copydata(m, off, sizeof(ip6), &ip6);
1299 		if (nxtp)
1300 			*nxtp = ip6.ip6_nxt;
1301 		off += sizeof(ip6);
1302 		return off;
1303 
1304 	case IPPROTO_FRAGMENT:
1305 		/*
1306 		 * terminate parsing if it is not the first fragment,
1307 		 * it does not make sense to parse through it.
1308 		 */
1309 		if (m->m_pkthdr.len < off + sizeof(fh))
1310 			return -1;
1311 		m_copydata(m, off, sizeof(fh), &fh);
1312 		if ((fh.ip6f_offlg & IP6F_OFF_MASK) != 0)
1313 			return -1;
1314 		if (nxtp)
1315 			*nxtp = fh.ip6f_nxt;
1316 		off += sizeof(struct ip6_frag);
1317 		return off;
1318 
1319 	case IPPROTO_AH:
1320 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1321 			return -1;
1322 		m_copydata(m, off, sizeof(ip6e), &ip6e);
1323 		if (nxtp)
1324 			*nxtp = ip6e.ip6e_nxt;
1325 		off += (ip6e.ip6e_len + 2) << 2;
1326 		if (m->m_pkthdr.len < off)
1327 			return -1;
1328 		return off;
1329 
1330 	case IPPROTO_HOPOPTS:
1331 	case IPPROTO_ROUTING:
1332 	case IPPROTO_DSTOPTS:
1333 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1334 			return -1;
1335 		m_copydata(m, off, sizeof(ip6e), &ip6e);
1336 		if (nxtp)
1337 			*nxtp = ip6e.ip6e_nxt;
1338 		off += (ip6e.ip6e_len + 1) << 3;
1339 		if (m->m_pkthdr.len < off)
1340 			return -1;
1341 		return off;
1342 
1343 	case IPPROTO_NONE:
1344 	case IPPROTO_ESP:
1345 	case IPPROTO_IPCOMP:
1346 		/* give up */
1347 		return -1;
1348 
1349 	default:
1350 		return -1;
1351 	}
1352 
1353 	return -1;
1354 }
1355 
1356 /*
1357  * get offset for the last header in the chain.  m will be kept untainted.
1358  */
1359 int
1360 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1361 {
1362 	int newoff;
1363 	int nxt;
1364 
1365 	if (!nxtp) {
1366 		nxt = -1;
1367 		nxtp = &nxt;
1368 	}
1369 	while (1) {
1370 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1371 		if (newoff < 0)
1372 			return off;
1373 		else if (newoff < off)
1374 			return -1;	/* invalid */
1375 		else if (newoff == off)
1376 			return newoff;
1377 
1378 		off = newoff;
1379 		proto = *nxtp;
1380 	}
1381 }
1382 
1383 /*
1384  * System control for IP6
1385  */
1386 
1387 const u_char inet6ctlerrmap[PRC_NCMDS] = {
1388 	0,		0,		0,		0,
1389 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1390 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1391 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1392 	0,		0,		0,		0,
1393 	ENOPROTOOPT
1394 };
1395 
1396 #ifdef MROUTING
1397 extern int ip6_mrtproto;
1398 #endif
1399 
1400 const struct sysctl_bounded_args ipv6ctl_vars[] = {
1401 	{ IPV6CTL_DAD_PENDING, &ip6_dad_pending, SYSCTL_INT_READONLY },
1402 #ifdef MROUTING
1403 	{ IPV6CTL_MRTPROTO, &ip6_mrtproto, SYSCTL_INT_READONLY },
1404 #endif
1405 	{ IPV6CTL_FORWARDING, &ip6_forwarding, 0, 1 },
1406 	{ IPV6CTL_SENDREDIRECTS, &ip6_sendredirects, 0, 1 },
1407 	{ IPV6CTL_DEFHLIM, &ip6_defhlim, 0, 255 },
1408 	{ IPV6CTL_MAXFRAGPACKETS, &ip6_maxfragpackets, 0, 1000 },
1409 	{ IPV6CTL_LOG_INTERVAL, &ip6_log_interval, 0, INT_MAX },
1410 	{ IPV6CTL_HDRNESTLIMIT, &ip6_hdrnestlimit, 0, 100 },
1411 	{ IPV6CTL_DAD_COUNT, &ip6_dad_count, 0, 10 },
1412 	{ IPV6CTL_AUTO_FLOWLABEL, &ip6_auto_flowlabel, 0, 1 },
1413 	{ IPV6CTL_DEFMCASTHLIM, &ip6_defmcasthlim, 0, 255 },
1414 	{ IPV6CTL_USE_DEPRECATED, &ip6_use_deprecated, 0, 1 },
1415 	{ IPV6CTL_MAXFRAGS, &ip6_maxfrags, 0, 1000 },
1416 	{ IPV6CTL_MFORWARDING, &ip6_mforwarding, 0, 1 },
1417 	{ IPV6CTL_MULTIPATH, &ip6_multipath, 0, 1 },
1418 	{ IPV6CTL_MCAST_PMTU, &ip6_mcast_pmtu, 0, 1 },
1419 	{ IPV6CTL_NEIGHBORGCTHRESH, &ip6_neighborgcthresh, -1, 5 * 2048 },
1420 	{ IPV6CTL_MAXDYNROUTES, &ip6_maxdynroutes, -1, 5 * 4096 },
1421 };
1422 
1423 int
1424 ip6_sysctl_ip6stat(void *oldp, size_t *oldlenp, void *newp)
1425 {
1426 	struct ip6stat *ip6stat;
1427 	int ret;
1428 
1429 	CTASSERT(sizeof(*ip6stat) == (ip6s_ncounters * sizeof(uint64_t)));
1430 
1431 	ip6stat = malloc(sizeof(*ip6stat), M_TEMP, M_WAITOK);
1432 	counters_read(ip6counters, (uint64_t *)ip6stat, ip6s_ncounters);
1433 	ret = sysctl_rdstruct(oldp, oldlenp, newp,
1434 	    ip6stat, sizeof(*ip6stat));
1435 	free(ip6stat, M_TEMP, sizeof(*ip6stat));
1436 
1437 	return (ret);
1438 }
1439 
1440 int
1441 ip6_sysctl_soiikey(void *oldp, size_t *oldlenp, void *newp, size_t newlen)
1442 {
1443 	uint8_t oldkey[IP6_SOIIKEY_LEN];
1444 	int error;
1445 
1446 	error = suser(curproc);
1447 	if (error != 0)
1448 		return (error);
1449 
1450 	memcpy(oldkey, ip6_soiikey, sizeof(oldkey));
1451 
1452 	error = sysctl_struct(oldp, oldlenp, newp, newlen, ip6_soiikey,
1453 	    sizeof(ip6_soiikey));
1454 
1455 	return (error);
1456 }
1457 
1458 int
1459 ip6_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
1460     void *newp, size_t newlen)
1461 {
1462 #ifdef MROUTING
1463 	extern struct mrt6stat mrt6stat;
1464 #endif
1465 	int error;
1466 
1467 	/* Almost all sysctl names at this level are terminal. */
1468 	if (namelen != 1 && name[0] != IPV6CTL_IFQUEUE)
1469 		return (ENOTDIR);
1470 
1471 	switch (name[0]) {
1472 	case IPV6CTL_STATS:
1473 		return (ip6_sysctl_ip6stat(oldp, oldlenp, newp));
1474 #ifdef MROUTING
1475 	case IPV6CTL_MRTSTATS:
1476 		if (newp != NULL)
1477 			return (EPERM);
1478 		NET_LOCK();
1479 		error = sysctl_struct(oldp, oldlenp, newp, newlen,
1480 		    &mrt6stat, sizeof(mrt6stat));
1481 		NET_UNLOCK();
1482 		return (error);
1483 	case IPV6CTL_MRTMIF:
1484 		if (newp)
1485 			return (EPERM);
1486 		NET_LOCK();
1487 		error = mrt6_sysctl_mif(oldp, oldlenp);
1488 		NET_UNLOCK();
1489 		return (error);
1490 	case IPV6CTL_MRTMFC:
1491 		if (newp)
1492 			return (EPERM);
1493 		NET_LOCK();
1494 		error = mrt6_sysctl_mfc(oldp, oldlenp);
1495 		NET_UNLOCK();
1496 		return (error);
1497 #else
1498 	case IPV6CTL_MRTSTATS:
1499 	case IPV6CTL_MRTPROTO:
1500 	case IPV6CTL_MRTMIF:
1501 	case IPV6CTL_MRTMFC:
1502 		return (EOPNOTSUPP);
1503 #endif
1504 	case IPV6CTL_MTUDISCTIMEOUT:
1505 		NET_LOCK();
1506 		error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
1507 		    &ip6_mtudisc_timeout, 0, INT_MAX);
1508 		rt_timer_queue_change(&icmp6_mtudisc_timeout_q,
1509 		    ip6_mtudisc_timeout);
1510 		NET_UNLOCK();
1511 		return (error);
1512 	case IPV6CTL_IFQUEUE:
1513 		return (sysctl_niq(name + 1, namelen - 1,
1514 		    oldp, oldlenp, newp, newlen, &ip6intrq));
1515 	case IPV6CTL_SOIIKEY:
1516 		return (ip6_sysctl_soiikey(oldp, oldlenp, newp, newlen));
1517 	default:
1518 		NET_LOCK();
1519 		error = sysctl_bounded_arr(ipv6ctl_vars, nitems(ipv6ctl_vars),
1520 		    name, namelen, oldp, oldlenp, newp, newlen);
1521 		NET_UNLOCK();
1522 		return (error);
1523 	}
1524 	/* NOTREACHED */
1525 }
1526 
1527 void
1528 ip6_send_dispatch(void *xmq)
1529 {
1530 	struct mbuf_queue *mq = xmq;
1531 	struct mbuf *m;
1532 	struct mbuf_list ml;
1533 
1534 	mq_delist(mq, &ml);
1535 	if (ml_empty(&ml))
1536 		return;
1537 
1538 	NET_LOCK();
1539 	while ((m = ml_dequeue(&ml)) != NULL) {
1540 		ip6_output(m, NULL, NULL, 0, NULL, NULL);
1541 	}
1542 	NET_UNLOCK();
1543 }
1544 
1545 void
1546 ip6_send(struct mbuf *m)
1547 {
1548 	mq_enqueue(&ip6send_mq, m);
1549 	task_add(net_tq(0), &ip6send_task);
1550 }
1551