xref: /openbsd-src/sys/netinet6/ip6_input.c (revision d1df930ffab53da22f3324c32bed7ac5709915e6)
1 /*	$OpenBSD: ip6_input.c,v 1.215 2018/05/21 15:52:22 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 <netinet6/ip6protosw.h>
102 
103 #include "gif.h"
104 #include "bpfilter.h"
105 
106 #ifdef MROUTING
107 #include <netinet6/ip6_mroute.h>
108 #endif
109 
110 #if NPF > 0
111 #include <net/pfvar.h>
112 #endif
113 
114 #if NCARP > 0
115 #include <netinet/ip_carp.h>
116 #endif
117 
118 struct niqueue ip6intrq = NIQUEUE_INITIALIZER(IPQ_MAXLEN, NETISR_IPV6);
119 
120 struct cpumem *ip6counters;
121 
122 uint8_t ip6_soiikey[IP6_SOIIKEY_LEN];
123 
124 int ip6_ours(struct mbuf **, int *, int, int);
125 int ip6_local(struct mbuf **, int *, int, int);
126 int ip6_check_rh0hdr(struct mbuf *, int *);
127 int ip6_hbhchcheck(struct mbuf *, int *, int *, int *);
128 int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
129 struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
130 int ip6_sysctl_soiikey(void *, size_t *, void *, size_t);
131 
132 static struct mbuf_queue	ip6send_mq;
133 
134 static void ip6_send_dispatch(void *);
135 static struct task ip6send_task =
136 	TASK_INITIALIZER(ip6_send_dispatch, &ip6send_mq);
137 
138 /*
139  * IP6 initialization: fill in IP6 protocol switch table.
140  * All protocols not implemented in kernel go to raw IP6 protocol handler.
141  */
142 void
143 ip6_init(void)
144 {
145 	const struct protosw *pr;
146 	int i;
147 
148 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
149 	if (pr == NULL)
150 		panic("ip6_init");
151 	for (i = 0; i < IPPROTO_MAX; i++)
152 		ip6_protox[i] = pr - inet6sw;
153 	for (pr = inet6domain.dom_protosw;
154 	    pr < inet6domain.dom_protoswNPROTOSW; pr++)
155 		if (pr->pr_domain->dom_family == PF_INET6 &&
156 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW &&
157 		    pr->pr_protocol < IPPROTO_MAX)
158 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
159 	ip6_randomid_init();
160 	nd6_init();
161 	frag6_init();
162 
163 	mq_init(&ip6send_mq, 64, IPL_SOFTNET);
164 
165 	ip6counters = counters_alloc(ip6s_ncounters);
166 }
167 
168 /*
169  * Enqueue packet for local delivery.  Queuing is used as a boundary
170  * between the network layer (input/forward path) running without
171  * KERNEL_LOCK() and the transport layer still needing it.
172  */
173 int
174 ip6_ours(struct mbuf **mp, int *offp, int nxt, int af)
175 {
176 	/* We are already in a IPv4/IPv6 local deliver loop. */
177 	if (af != AF_UNSPEC)
178 		return ip6_local(mp, offp, nxt, af);
179 
180 	niq_enqueue(&ip6intrq, *mp);
181 	*mp = NULL;
182 	return IPPROTO_DONE;
183 }
184 
185 /*
186  * Dequeue and process locally delivered packets.
187  */
188 void
189 ip6intr(void)
190 {
191 	struct mbuf *m;
192 	int off, nxt;
193 
194 	while ((m = niq_dequeue(&ip6intrq)) != NULL) {
195 #ifdef DIAGNOSTIC
196 		if ((m->m_flags & M_PKTHDR) == 0)
197 			panic("ip6intr no HDR");
198 #endif
199 		off = 0;
200 		nxt = ip6_local(&m, &off, IPPROTO_IPV6, AF_UNSPEC);
201 		KASSERT(nxt == IPPROTO_DONE);
202 	}
203 }
204 
205 void
206 ipv6_input(struct ifnet *ifp, struct mbuf *m)
207 {
208 	int off, nxt;
209 
210 	off = 0;
211 	nxt = ip6_input_if(&m, &off, IPPROTO_IPV6, AF_UNSPEC, ifp);
212 	KASSERT(nxt == IPPROTO_DONE);
213 }
214 
215 int
216 ip6_input_if(struct mbuf **mp, int *offp, int nxt, int af, struct ifnet *ifp)
217 {
218 	struct mbuf *m = *mp;
219 	struct ip6_hdr *ip6;
220 	struct sockaddr_in6 sin6;
221 	struct rtentry *rt = NULL;
222 	int ours = 0;
223 	u_int16_t src_scope, dst_scope;
224 #if NPF > 0
225 	struct in6_addr odst;
226 #endif
227 	int srcrt = 0;
228 
229 	KASSERT(*offp == 0);
230 
231 	ip6stat_inc(ip6s_total);
232 
233 	if (m->m_len < sizeof(struct ip6_hdr)) {
234 		if ((m = *mp = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
235 			ip6stat_inc(ip6s_toosmall);
236 			goto bad;
237 		}
238 	}
239 
240 	ip6 = mtod(m, struct ip6_hdr *);
241 
242 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
243 		ip6stat_inc(ip6s_badvers);
244 		goto bad;
245 	}
246 
247 #if NCARP > 0
248 	if (carp_lsdrop(ifp, m, AF_INET6, ip6->ip6_src.s6_addr32,
249 	    ip6->ip6_dst.s6_addr32, (ip6->ip6_nxt == IPPROTO_ICMPV6 ? 0 : 1)))
250 		goto bad;
251 #endif
252 	ip6stat_inc(ip6s_nxthist + ip6->ip6_nxt);
253 
254 	/*
255 	 * Check against address spoofing/corruption.
256 	 */
257 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
258 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
259 		/*
260 		 * XXX: "badscope" is not very suitable for a multicast source.
261 		 */
262 		ip6stat_inc(ip6s_badscope);
263 		goto bad;
264 	}
265 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
266 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
267 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
268 		    ip6stat_inc(ip6s_badscope);
269 		    goto bad;
270 	}
271 	/* Drop packets if interface ID portion is already filled. */
272 	if (((IN6_IS_SCOPE_EMBED(&ip6->ip6_src) && ip6->ip6_src.s6_addr16[1]) ||
273 	    (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst) && ip6->ip6_dst.s6_addr16[1])) &&
274 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
275 		ip6stat_inc(ip6s_badscope);
276 		goto bad;
277 	}
278 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
279 	    !(m->m_flags & M_LOOP)) {
280 		/*
281 		 * In this case, the packet should come from the loopback
282 		 * interface.  However, we cannot just check the if_flags,
283 		 * because ip6_mloopback() passes the "actual" interface
284 		 * as the outgoing/incoming interface.
285 		 */
286 		ip6stat_inc(ip6s_badscope);
287 		goto bad;
288 	}
289 
290 	/*
291 	 * The following check is not documented in specs.  A malicious
292 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
293 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
294 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
295 	 *
296 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
297 	 * support IPv4-less kernel compilation, we cannot support SIIT
298 	 * environment at all.  So, it makes more sense for us to reject any
299 	 * malicious packets for non-SIIT environment, than try to do a
300 	 * partial support for SIIT environment.
301 	 */
302 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
303 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
304 		ip6stat_inc(ip6s_badscope);
305 		goto bad;
306 	}
307 
308 	/*
309 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
310 	 *
311 	 * The code forbids automatic tunneling as per RFC4213.
312 	 */
313 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
314 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
315 		ip6stat_inc(ip6s_badscope);
316 		goto bad;
317 	}
318 
319 	/*
320 	 * If the packet has been received on a loopback interface it
321 	 * can be destinated to any local address, not necessarily to
322 	 * an address configured on `ifp'.
323 	 */
324 	if (ifp->if_flags & IFF_LOOPBACK) {
325 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
326 			src_scope = ip6->ip6_src.s6_addr16[1];
327 			ip6->ip6_src.s6_addr16[1] = 0;
328 		}
329 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
330 			dst_scope = ip6->ip6_dst.s6_addr16[1];
331 			ip6->ip6_dst.s6_addr16[1] = 0;
332 		}
333 	}
334 
335 #if NPF > 0
336 	/*
337 	 * Packet filter
338 	 */
339 	odst = ip6->ip6_dst;
340 	if (pf_test(AF_INET6, PF_IN, ifp, mp) != PF_PASS)
341 		goto bad;
342 	m = *mp;
343 	if (m == NULL)
344 		goto bad;
345 
346 	ip6 = mtod(m, struct ip6_hdr *);
347 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
348 #endif
349 
350 	/*
351 	 * Without embedded scope ID we cannot find link-local
352 	 * addresses in the routing table.
353 	 */
354 	if (ifp->if_flags & IFF_LOOPBACK) {
355 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
356 			ip6->ip6_src.s6_addr16[1] = src_scope;
357 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
358 			ip6->ip6_dst.s6_addr16[1] = dst_scope;
359 	} else {
360 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
361 			ip6->ip6_src.s6_addr16[1] = htons(ifp->if_index);
362 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
363 			ip6->ip6_dst.s6_addr16[1] = htons(ifp->if_index);
364 	}
365 
366 	/*
367 	 * Be more secure than RFC5095 and scan for type 0 routing headers.
368 	 * If pf has already scanned the header chain, do not do it twice.
369 	 */
370 	if (!(m->m_pkthdr.pf.flags & PF_TAG_PROCESSED) &&
371 	    ip6_check_rh0hdr(m, offp)) {
372 		ip6stat_inc(ip6s_badoptions);
373 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, *offp);
374 		m = *mp = NULL;
375 		goto bad;
376 	}
377 
378 	if (IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
379 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) {
380 		nxt = ip6_ours(mp, offp, nxt, af);
381 		goto out;
382 	}
383 
384 #if NPF > 0
385 	if (pf_ouraddr(m) == 1) {
386 		nxt = ip6_ours(mp, offp, nxt, af);
387 		goto out;
388 	}
389 #endif
390 
391 	/*
392 	 * Multicast check
393 	 */
394 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
395 		/*
396 		 * Make sure M_MCAST is set.  It should theoretically
397 		 * already be there, but let's play safe because upper
398 		 * layers check for this flag.
399 		 */
400 		m->m_flags |= M_MCAST;
401 
402 		/*
403 		 * See if we belong to the destination multicast group on the
404 		 * arrival interface.
405 		 */
406 		if (in6_hasmulti(&ip6->ip6_dst, ifp))
407 			ours = 1;
408 
409 #ifdef MROUTING
410 		if (ip6_mforwarding && ip6_mrouter[ifp->if_rdomain]) {
411 			int error;
412 
413 			if (ip6_hbhchcheck(m, offp, &nxt, &ours))
414 				goto out;
415 
416 			ip6 = mtod(m, struct ip6_hdr *);
417 
418 			/*
419 			 * If we are acting as a multicast router, all
420 			 * incoming multicast packets are passed to the
421 			 * kernel-level multicast forwarding function.
422 			 * The packet is returned (relatively) intact; if
423 			 * ip6_mforward() returns a non-zero value, the packet
424 			 * must be discarded, else it may be accepted below.
425 			 */
426 			KERNEL_LOCK();
427 			error = ip6_mforward(ip6, ifp, m);
428 			KERNEL_UNLOCK();
429 			if (error) {
430 				ip6stat_inc(ip6s_cantforward);
431 				goto bad;
432 			}
433 
434 			if (ours) {
435 				if (af == AF_UNSPEC) {
436 					KERNEL_LOCK();
437 					nxt = ip_deliver(mp, offp, nxt,
438 					    AF_INET6);
439 					KERNEL_UNLOCK();
440 				}
441 				goto out;
442 			}
443 			goto bad;
444 		}
445 #endif
446 		if (!ours) {
447 			ip6stat_inc(ip6s_notmember);
448 			if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
449 				ip6stat_inc(ip6s_cantforward);
450 			goto bad;
451 		}
452 		nxt = ip6_ours(mp, offp, nxt, af);
453 		goto out;
454 	}
455 
456 
457 	/*
458 	 *  Unicast check
459 	 */
460 	memset(&sin6, 0, sizeof(struct sockaddr_in6));
461 	sin6.sin6_len = sizeof(struct sockaddr_in6);
462 	sin6.sin6_family = AF_INET6;
463 	sin6.sin6_addr = ip6->ip6_dst;
464 	rt = rtalloc_mpath(sin6tosa(&sin6), &ip6->ip6_src.s6_addr32[0],
465 	    m->m_pkthdr.ph_rtableid);
466 
467 	/*
468 	 * Accept the packet if the route to the destination is marked
469 	 * as local.
470 	 */
471 	if (rtisvalid(rt) && ISSET(rt->rt_flags, RTF_LOCAL)) {
472 		struct in6_ifaddr *ia6 = ifatoia6(rt->rt_ifa);
473 		if (ia6->ia6_flags & IN6_IFF_ANYCAST)
474 			m->m_flags |= M_ACAST;
475 		/*
476 		 * packets to a tentative, duplicated, or somehow invalid
477 		 * address must not be accepted.
478 		 */
479 		if ((ia6->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED))) {
480 			char src[INET6_ADDRSTRLEN], dst[INET6_ADDRSTRLEN];
481 
482 			inet_ntop(AF_INET6, &ip6->ip6_src, src, sizeof(src));
483 			inet_ntop(AF_INET6, &ip6->ip6_dst, dst, sizeof(dst));
484 			/* address is not ready, so discard the packet. */
485 			nd6log((LOG_INFO,
486 			    "%s: packet to an unready address %s->%s\n",
487 			    __func__, src, dst));
488 
489 			goto bad;
490 		} else {
491 			nxt = ip6_ours(mp, offp, nxt, af);
492 			goto out;
493 		}
494 	}
495 
496 #if NCARP > 0
497 	if (ip6->ip6_nxt == IPPROTO_ICMPV6 &&
498 	    carp_lsdrop(ifp, m, AF_INET6, ip6->ip6_src.s6_addr32,
499 	    ip6->ip6_dst.s6_addr32, 1))
500 		goto bad;
501 #endif
502 	/*
503 	 * Now there is no reason to process the packet if it's not our own
504 	 * and we're not a router.
505 	 */
506 	if (!ip6_forwarding) {
507 		ip6stat_inc(ip6s_cantforward);
508 		goto bad;
509 	}
510 
511 	if (ip6_hbhchcheck(m, offp, &nxt, &ours))
512 		goto out;
513 
514 	if (ours) {
515 		if (af == AF_UNSPEC) {
516 			KERNEL_LOCK();
517 			nxt = ip_deliver(mp, offp, nxt, AF_INET6);
518 			KERNEL_UNLOCK();
519 		}
520 		goto out;
521 	}
522 
523 #ifdef IPSEC
524 	if (ipsec_in_use) {
525 		int rv;
526 
527 		rv = ipsec_forward_check(m, *offp, AF_INET6);
528 		if (rv != 0) {
529 			ip6stat_inc(ip6s_cantforward);
530 			goto bad;
531 		}
532 		/*
533 		 * Fall through, forward packet. Outbound IPsec policy
534 		 * checking will occur in ip6_forward().
535 		 */
536 	}
537 #endif /* IPSEC */
538 
539 	ip6_forward(m, rt, srcrt);
540 	*mp = NULL;
541 	return IPPROTO_DONE;
542  bad:
543 	nxt = IPPROTO_DONE;
544 	m_freemp(mp);
545  out:
546 	rtfree(rt);
547 	return nxt;
548 }
549 
550 int
551 ip6_local(struct mbuf **mp, int *offp, int nxt, int af)
552 {
553 	if (ip6_hbhchcheck(*mp, offp, &nxt, NULL))
554 		return IPPROTO_DONE;
555 
556 	/* Check wheter we are already in a IPv4/IPv6 local deliver loop. */
557 	if (af == AF_UNSPEC)
558 		nxt = ip_deliver(mp, offp, nxt, AF_INET6);
559 	return nxt;
560 }
561 
562 int
563 ip6_hbhchcheck(struct mbuf *m, int *offp, int *nxtp, int *oursp)
564 {
565 	struct ip6_hdr *ip6;
566 	u_int32_t plen, rtalert = ~0;
567 
568 	ip6 = mtod(m, struct ip6_hdr *);
569 
570 	/*
571 	 * Process Hop-by-Hop options header if it's contained.
572 	 * m may be modified in ip6_hopopts_input().
573 	 * If a JumboPayload option is included, plen will also be modified.
574 	 */
575 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
576 	*offp = sizeof(struct ip6_hdr);
577 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
578 		struct ip6_hbh *hbh;
579 
580 		if (ip6_hopopts_input(&plen, &rtalert, &m, offp)) {
581 			goto bad;	/* m have already been freed */
582 		}
583 
584 		/* adjust pointer */
585 		ip6 = mtod(m, struct ip6_hdr *);
586 
587 		/*
588 		 * if the payload length field is 0 and the next header field
589 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
590 		 * option MUST be included.
591 		 */
592 		if (ip6->ip6_plen == 0 && plen == 0) {
593 			/*
594 			 * Note that if a valid jumbo payload option is
595 			 * contained, ip6_hopopts_input() must set a valid
596 			 * (non-zero) payload length to the variable plen.
597 			 */
598 			ip6stat_inc(ip6s_badoptions);
599 			icmp6_error(m, ICMP6_PARAM_PROB,
600 				    ICMP6_PARAMPROB_HEADER,
601 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
602 			goto bad;
603 		}
604 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
605 			sizeof(struct ip6_hbh));
606 		if (hbh == NULL) {
607 			ip6stat_inc(ip6s_tooshort);
608 			goto bad;
609 		}
610 		*nxtp = hbh->ip6h_nxt;
611 
612 		/*
613 		 * accept the packet if a router alert option is included
614 		 * and we act as an IPv6 router.
615 		 */
616 		if (rtalert != ~0 && ip6_forwarding && oursp != NULL)
617 			*oursp = 1;
618 	} else
619 		*nxtp = ip6->ip6_nxt;
620 
621 	/*
622 	 * Check that the amount of data in the buffers
623 	 * is as at least much as the IPv6 header would have us expect.
624 	 * Trim mbufs if longer than we expect.
625 	 * Drop packet if shorter than we expect.
626 	 */
627 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
628 		ip6stat_inc(ip6s_tooshort);
629 		m_freem(m);
630 		goto bad;
631 	}
632 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
633 		if (m->m_len == m->m_pkthdr.len) {
634 			m->m_len = sizeof(struct ip6_hdr) + plen;
635 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
636 		} else {
637 			m_adj(m,
638 			    sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
639 		}
640 	}
641 
642 	return (0);
643 
644  bad:
645 	*nxtp = IPPROTO_DONE;
646 	return (-1);
647 }
648 
649 /* scan packet for RH0 routing header. Mostly stolen from pf.c:pf_test() */
650 int
651 ip6_check_rh0hdr(struct mbuf *m, int *offp)
652 {
653 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
654 	struct ip6_rthdr rthdr;
655 	struct ip6_ext opt6;
656 	u_int8_t proto = ip6->ip6_nxt;
657 	int done = 0, lim, off, rh_cnt = 0;
658 
659 	off = ((caddr_t)ip6 - m->m_data) + sizeof(struct ip6_hdr);
660 	lim = min(m->m_pkthdr.len, ntohs(ip6->ip6_plen) + sizeof(*ip6));
661 	do {
662 		switch (proto) {
663 		case IPPROTO_ROUTING:
664 			*offp = off;
665 			if (rh_cnt++) {
666 				/* more than one rh header present */
667 				return (1);
668 			}
669 
670 			if (off + sizeof(rthdr) > lim) {
671 				/* packet to short to make sense */
672 				return (1);
673 			}
674 
675 			m_copydata(m, off, sizeof(rthdr), (caddr_t)&rthdr);
676 
677 			if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
678 				*offp += offsetof(struct ip6_rthdr, ip6r_type);
679 				return (1);
680 			}
681 
682 			off += (rthdr.ip6r_len + 1) * 8;
683 			proto = rthdr.ip6r_nxt;
684 			break;
685 		case IPPROTO_AH:
686 		case IPPROTO_HOPOPTS:
687 		case IPPROTO_DSTOPTS:
688 			/* get next header and header length */
689 			if (off + sizeof(opt6) > lim) {
690 				/*
691 				 * Packet to short to make sense, we could
692 				 * reject the packet but as a router we
693 				 * should not do that so forward it.
694 				 */
695 				return (0);
696 			}
697 
698 			m_copydata(m, off, sizeof(opt6), (caddr_t)&opt6);
699 
700 			if (proto == IPPROTO_AH)
701 				off += (opt6.ip6e_len + 2) * 4;
702 			else
703 				off += (opt6.ip6e_len + 1) * 8;
704 			proto = opt6.ip6e_nxt;
705 			break;
706 		case IPPROTO_FRAGMENT:
707 		default:
708 			/* end of header stack */
709 			done = 1;
710 			break;
711 		}
712 	} while (!done);
713 
714 	return (0);
715 }
716 
717 /*
718  * Hop-by-Hop options header processing. If a valid jumbo payload option is
719  * included, the real payload length will be stored in plenp.
720  *
721  * rtalertp - XXX: should be stored in a more smart way
722  */
723 int
724 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, struct mbuf **mp,
725     int *offp)
726 {
727 	struct mbuf *m = *mp;
728 	int off = *offp, hbhlen;
729 	struct ip6_hbh *hbh;
730 
731 	/* validation of the length of the header */
732 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
733 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
734 	if (hbh == NULL) {
735 		ip6stat_inc(ip6s_tooshort);
736 		return -1;
737 	}
738 	hbhlen = (hbh->ip6h_len + 1) << 3;
739 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
740 		hbhlen);
741 	if (hbh == NULL) {
742 		ip6stat_inc(ip6s_tooshort);
743 		return -1;
744 	}
745 	off += hbhlen;
746 	hbhlen -= sizeof(struct ip6_hbh);
747 
748 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
749 				hbhlen, rtalertp, plenp) < 0)
750 		return (-1);
751 
752 	*offp = off;
753 	*mp = m;
754 	return (0);
755 }
756 
757 /*
758  * Search header for all Hop-by-hop options and process each option.
759  * This function is separate from ip6_hopopts_input() in order to
760  * handle a case where the sending node itself process its hop-by-hop
761  * options header. In such a case, the function is called from ip6_output().
762  *
763  * The function assumes that hbh header is located right after the IPv6 header
764  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
765  * opthead + hbhlen is located in continuous memory region.
766  */
767 int
768 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
769     u_int32_t *rtalertp, u_int32_t *plenp)
770 {
771 	struct ip6_hdr *ip6;
772 	int optlen = 0;
773 	u_int8_t *opt = opthead;
774 	u_int16_t rtalert_val;
775 	u_int32_t jumboplen;
776 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
777 
778 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
779 		switch (*opt) {
780 		case IP6OPT_PAD1:
781 			optlen = 1;
782 			break;
783 		case IP6OPT_PADN:
784 			if (hbhlen < IP6OPT_MINLEN) {
785 				ip6stat_inc(ip6s_toosmall);
786 				goto bad;
787 			}
788 			optlen = *(opt + 1) + 2;
789 			break;
790 		case IP6OPT_ROUTER_ALERT:
791 			/* XXX may need check for alignment */
792 			if (hbhlen < IP6OPT_RTALERT_LEN) {
793 				ip6stat_inc(ip6s_toosmall);
794 				goto bad;
795 			}
796 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
797 				/* XXX stat */
798 				icmp6_error(m, ICMP6_PARAM_PROB,
799 				    ICMP6_PARAMPROB_HEADER,
800 				    erroff + opt + 1 - opthead);
801 				return (-1);
802 			}
803 			optlen = IP6OPT_RTALERT_LEN;
804 			memcpy((caddr_t)&rtalert_val, (caddr_t)(opt + 2), 2);
805 			*rtalertp = ntohs(rtalert_val);
806 			break;
807 		case IP6OPT_JUMBO:
808 			/* XXX may need check for alignment */
809 			if (hbhlen < IP6OPT_JUMBO_LEN) {
810 				ip6stat_inc(ip6s_toosmall);
811 				goto bad;
812 			}
813 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
814 				/* XXX stat */
815 				icmp6_error(m, ICMP6_PARAM_PROB,
816 				    ICMP6_PARAMPROB_HEADER,
817 				    erroff + opt + 1 - opthead);
818 				return (-1);
819 			}
820 			optlen = IP6OPT_JUMBO_LEN;
821 
822 			/*
823 			 * IPv6 packets that have non 0 payload length
824 			 * must not contain a jumbo payload option.
825 			 */
826 			ip6 = mtod(m, struct ip6_hdr *);
827 			if (ip6->ip6_plen) {
828 				ip6stat_inc(ip6s_badoptions);
829 				icmp6_error(m, ICMP6_PARAM_PROB,
830 				    ICMP6_PARAMPROB_HEADER,
831 				    erroff + opt - opthead);
832 				return (-1);
833 			}
834 
835 			/*
836 			 * We may see jumbolen in unaligned location, so
837 			 * we'd need to perform memcpy().
838 			 */
839 			memcpy(&jumboplen, opt + 2, sizeof(jumboplen));
840 			jumboplen = (u_int32_t)htonl(jumboplen);
841 
842 #if 1
843 			/*
844 			 * if there are multiple jumbo payload options,
845 			 * *plenp will be non-zero and the packet will be
846 			 * rejected.
847 			 * the behavior may need some debate in ipngwg -
848 			 * multiple options does not make sense, however,
849 			 * there's no explicit mention in specification.
850 			 */
851 			if (*plenp != 0) {
852 				ip6stat_inc(ip6s_badoptions);
853 				icmp6_error(m, ICMP6_PARAM_PROB,
854 				    ICMP6_PARAMPROB_HEADER,
855 				    erroff + opt + 2 - opthead);
856 				return (-1);
857 			}
858 #endif
859 
860 			/*
861 			 * jumbo payload length must be larger than 65535.
862 			 */
863 			if (jumboplen <= IPV6_MAXPACKET) {
864 				ip6stat_inc(ip6s_badoptions);
865 				icmp6_error(m, ICMP6_PARAM_PROB,
866 				    ICMP6_PARAMPROB_HEADER,
867 				    erroff + opt + 2 - opthead);
868 				return (-1);
869 			}
870 			*plenp = jumboplen;
871 
872 			break;
873 		default:		/* unknown option */
874 			if (hbhlen < IP6OPT_MINLEN) {
875 				ip6stat_inc(ip6s_toosmall);
876 				goto bad;
877 			}
878 			optlen = ip6_unknown_opt(opt, m,
879 			    erroff + opt - opthead);
880 			if (optlen == -1)
881 				return (-1);
882 			optlen += 2;
883 			break;
884 		}
885 	}
886 
887 	return (0);
888 
889   bad:
890 	m_freem(m);
891 	return (-1);
892 }
893 
894 /*
895  * Unknown option processing.
896  * The third argument `off' is the offset from the IPv6 header to the option,
897  * which allows returning an ICMPv6 error even if the IPv6 header and the
898  * option header are not continuous.
899  */
900 int
901 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
902 {
903 	struct ip6_hdr *ip6;
904 
905 	switch (IP6OPT_TYPE(*optp)) {
906 	case IP6OPT_TYPE_SKIP: /* ignore the option */
907 		return ((int)*(optp + 1));
908 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
909 		m_freem(m);
910 		return (-1);
911 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
912 		ip6stat_inc(ip6s_badoptions);
913 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
914 		return (-1);
915 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
916 		ip6stat_inc(ip6s_badoptions);
917 		ip6 = mtod(m, struct ip6_hdr *);
918 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
919 		    (m->m_flags & (M_BCAST|M_MCAST)))
920 			m_freem(m);
921 		else
922 			icmp6_error(m, ICMP6_PARAM_PROB,
923 				    ICMP6_PARAMPROB_OPTION, off);
924 		return (-1);
925 	}
926 
927 	m_freem(m);		/* XXX: NOTREACHED */
928 	return (-1);
929 }
930 
931 /*
932  * Create the "control" list for this pcb.
933  *
934  * The routine will be called from upper layer handlers like udp_input().
935  * Thus the routine assumes that the caller (udp_input) have already
936  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
937  * very first mbuf on the mbuf chain.
938  * We may want to add some infinite loop prevention or sanity checks for safety.
939  * (This applies only when you are using KAME mbuf chain restriction, i.e.
940  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
941  */
942 void
943 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
944 {
945 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
946 
947 	if (in6p->inp_socket->so_options & SO_TIMESTAMP) {
948 		struct timeval tv;
949 
950 		microtime(&tv);
951 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
952 		    SCM_TIMESTAMP, SOL_SOCKET);
953 		if (*mp)
954 			mp = &(*mp)->m_next;
955 	}
956 
957 	/* RFC 2292 sec. 5 */
958 	if ((in6p->inp_flags & IN6P_PKTINFO) != 0) {
959 		struct in6_pktinfo pi6;
960 		memcpy(&pi6.ipi6_addr, &ip6->ip6_dst, sizeof(struct in6_addr));
961 		if (IN6_IS_SCOPE_EMBED(&pi6.ipi6_addr))
962 			pi6.ipi6_addr.s6_addr16[1] = 0;
963 		pi6.ipi6_ifindex = m ? m->m_pkthdr.ph_ifidx : 0;
964 		*mp = sbcreatecontrol((caddr_t) &pi6,
965 		    sizeof(struct in6_pktinfo),
966 		    IPV6_PKTINFO, IPPROTO_IPV6);
967 		if (*mp)
968 			mp = &(*mp)->m_next;
969 	}
970 
971 	if ((in6p->inp_flags & IN6P_HOPLIMIT) != 0) {
972 		int hlim = ip6->ip6_hlim & 0xff;
973 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
974 		    IPV6_HOPLIMIT, IPPROTO_IPV6);
975 		if (*mp)
976 			mp = &(*mp)->m_next;
977 	}
978 
979 	if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
980 		u_int32_t flowinfo;
981 		int tclass;
982 
983 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
984 		flowinfo >>= 20;
985 
986 		tclass = flowinfo & 0xff;
987 		*mp = sbcreatecontrol((caddr_t)&tclass, sizeof(tclass),
988 		    IPV6_TCLASS, IPPROTO_IPV6);
989 		if (*mp)
990 			mp = &(*mp)->m_next;
991 	}
992 
993 	/*
994 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
995 	 * privilege for the option (see ip6_ctloutput), but it might be too
996 	 * strict, since there might be some hop-by-hop options which can be
997 	 * returned to normal user.
998 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
999 	 */
1000 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1001 		/*
1002 		 * Check if a hop-by-hop options header is contained in the
1003 		 * received packet, and if so, store the options as ancillary
1004 		 * data. Note that a hop-by-hop options header must be
1005 		 * just after the IPv6 header, which is assured through the
1006 		 * IPv6 input processing.
1007 		 */
1008 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1009 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1010 			struct ip6_hbh *hbh;
1011 			int hbhlen = 0;
1012 			struct mbuf *ext;
1013 
1014 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1015 			    ip6->ip6_nxt);
1016 			if (ext == NULL) {
1017 				ip6stat_inc(ip6s_tooshort);
1018 				return;
1019 			}
1020 			hbh = mtod(ext, struct ip6_hbh *);
1021 			hbhlen = (hbh->ip6h_len + 1) << 3;
1022 			if (hbhlen != ext->m_len) {
1023 				m_freem(ext);
1024 				ip6stat_inc(ip6s_tooshort);
1025 				return;
1026 			}
1027 
1028 			/*
1029 			 * XXX: We copy the whole header even if a
1030 			 * jumbo payload option is included, the option which
1031 			 * is to be removed before returning according to
1032 			 * RFC2292.
1033 			 * Note: this constraint is removed in RFC3542.
1034 			 */
1035 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1036 			    IPV6_HOPOPTS,
1037 			    IPPROTO_IPV6);
1038 			if (*mp)
1039 				mp = &(*mp)->m_next;
1040 			m_freem(ext);
1041 		}
1042 	}
1043 
1044 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1045 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1046 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1047 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1048 
1049 		/*
1050 		 * Search for destination options headers or routing
1051 		 * header(s) through the header chain, and stores each
1052 		 * header as ancillary data.
1053 		 * Note that the order of the headers remains in
1054 		 * the chain of ancillary data.
1055 		 */
1056 		while (1) {	/* is explicit loop prevention necessary? */
1057 			struct ip6_ext *ip6e = NULL;
1058 			int elen;
1059 			struct mbuf *ext = NULL;
1060 
1061 			/*
1062 			 * if it is not an extension header, don't try to
1063 			 * pull it from the chain.
1064 			 */
1065 			switch (nxt) {
1066 			case IPPROTO_DSTOPTS:
1067 			case IPPROTO_ROUTING:
1068 			case IPPROTO_HOPOPTS:
1069 			case IPPROTO_AH: /* is it possible? */
1070 				break;
1071 			default:
1072 				goto loopend;
1073 			}
1074 
1075 			ext = ip6_pullexthdr(m, off, nxt);
1076 			if (ext == NULL) {
1077 				ip6stat_inc(ip6s_tooshort);
1078 				return;
1079 			}
1080 			ip6e = mtod(ext, struct ip6_ext *);
1081 			if (nxt == IPPROTO_AH)
1082 				elen = (ip6e->ip6e_len + 2) << 2;
1083 			else
1084 				elen = (ip6e->ip6e_len + 1) << 3;
1085 			if (elen != ext->m_len) {
1086 				m_freem(ext);
1087 				ip6stat_inc(ip6s_tooshort);
1088 				return;
1089 			}
1090 
1091 			switch (nxt) {
1092 			case IPPROTO_DSTOPTS:
1093 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1094 					break;
1095 
1096 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1097 				    IPV6_DSTOPTS,
1098 				    IPPROTO_IPV6);
1099 				if (*mp)
1100 					mp = &(*mp)->m_next;
1101 				break;
1102 
1103 			case IPPROTO_ROUTING:
1104 				if (!(in6p->inp_flags & IN6P_RTHDR))
1105 					break;
1106 
1107 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1108 				    IPV6_RTHDR,
1109 				    IPPROTO_IPV6);
1110 				if (*mp)
1111 					mp = &(*mp)->m_next;
1112 				break;
1113 
1114 			case IPPROTO_HOPOPTS:
1115 			case IPPROTO_AH: /* is it possible? */
1116 				break;
1117 
1118 			default:
1119 				/*
1120 				 * other cases have been filtered in the above.
1121 				 * none will visit this case.  here we supply
1122 				 * the code just in case (nxt overwritten or
1123 				 * other cases).
1124 				 */
1125 				m_freem(ext);
1126 				goto loopend;
1127 
1128 			}
1129 
1130 			/* proceed with the next header. */
1131 			off += elen;
1132 			nxt = ip6e->ip6e_nxt;
1133 			ip6e = NULL;
1134 			m_freem(ext);
1135 			ext = NULL;
1136 		}
1137 loopend:
1138 		;
1139 	}
1140 }
1141 
1142 /*
1143  * pull single extension header from mbuf chain.  returns single mbuf that
1144  * contains the result, or NULL on error.
1145  */
1146 struct mbuf *
1147 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1148 {
1149 	struct ip6_ext ip6e;
1150 	size_t elen;
1151 	struct mbuf *n;
1152 
1153 #ifdef DIAGNOSTIC
1154 	switch (nxt) {
1155 	case IPPROTO_DSTOPTS:
1156 	case IPPROTO_ROUTING:
1157 	case IPPROTO_HOPOPTS:
1158 	case IPPROTO_AH: /* is it possible? */
1159 		break;
1160 	default:
1161 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1162 	}
1163 #endif
1164 
1165 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1166 	if (nxt == IPPROTO_AH)
1167 		elen = (ip6e.ip6e_len + 2) << 2;
1168 	else
1169 		elen = (ip6e.ip6e_len + 1) << 3;
1170 
1171 	MGET(n, M_DONTWAIT, MT_DATA);
1172 	if (n && elen >= MLEN) {
1173 		MCLGET(n, M_DONTWAIT);
1174 		if ((n->m_flags & M_EXT) == 0) {
1175 			m_free(n);
1176 			n = NULL;
1177 		}
1178 	}
1179 	if (!n)
1180 		return NULL;
1181 
1182 	n->m_len = 0;
1183 	if (elen >= M_TRAILINGSPACE(n)) {
1184 		m_free(n);
1185 		return NULL;
1186 	}
1187 
1188 	m_copydata(m, off, elen, mtod(n, caddr_t));
1189 	n->m_len = elen;
1190 	return n;
1191 }
1192 
1193 /*
1194  * Get offset to the previous header followed by the header
1195  * currently processed.
1196  */
1197 int
1198 ip6_get_prevhdr(struct mbuf *m, int off)
1199 {
1200 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1201 
1202 	if (off == sizeof(struct ip6_hdr)) {
1203 		return offsetof(struct ip6_hdr, ip6_nxt);
1204 	} else if (off < sizeof(struct ip6_hdr)) {
1205 		panic("%s: off < sizeof(struct ip6_hdr)", __func__);
1206 	} else {
1207 		int len, nlen, nxt;
1208 		struct ip6_ext ip6e;
1209 
1210 		nxt = ip6->ip6_nxt;
1211 		len = sizeof(struct ip6_hdr);
1212 		nlen = 0;
1213 		while (len < off) {
1214 			m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e);
1215 
1216 			switch (nxt) {
1217 			case IPPROTO_FRAGMENT:
1218 				nlen = sizeof(struct ip6_frag);
1219 				break;
1220 			case IPPROTO_AH:
1221 				nlen = (ip6e.ip6e_len + 2) << 2;
1222 				break;
1223 			default:
1224 				nlen = (ip6e.ip6e_len + 1) << 3;
1225 				break;
1226 			}
1227 			len += nlen;
1228 			nxt = ip6e.ip6e_nxt;
1229 		}
1230 
1231 		return (len - nlen);
1232 	}
1233 }
1234 
1235 /*
1236  * get next header offset.  m will be retained.
1237  */
1238 int
1239 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1240 {
1241 	struct ip6_hdr ip6;
1242 	struct ip6_ext ip6e;
1243 	struct ip6_frag fh;
1244 
1245 	/* just in case */
1246 	if (m == NULL)
1247 		panic("ip6_nexthdr: m == NULL");
1248 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1249 		return -1;
1250 
1251 	switch (proto) {
1252 	case IPPROTO_IPV6:
1253 		if (m->m_pkthdr.len < off + sizeof(ip6))
1254 			return -1;
1255 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1256 		if (nxtp)
1257 			*nxtp = ip6.ip6_nxt;
1258 		off += sizeof(ip6);
1259 		return off;
1260 
1261 	case IPPROTO_FRAGMENT:
1262 		/*
1263 		 * terminate parsing if it is not the first fragment,
1264 		 * it does not make sense to parse through it.
1265 		 */
1266 		if (m->m_pkthdr.len < off + sizeof(fh))
1267 			return -1;
1268 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1269 		if ((fh.ip6f_offlg & IP6F_OFF_MASK) != 0)
1270 			return -1;
1271 		if (nxtp)
1272 			*nxtp = fh.ip6f_nxt;
1273 		off += sizeof(struct ip6_frag);
1274 		return off;
1275 
1276 	case IPPROTO_AH:
1277 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1278 			return -1;
1279 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1280 		if (nxtp)
1281 			*nxtp = ip6e.ip6e_nxt;
1282 		off += (ip6e.ip6e_len + 2) << 2;
1283 		if (m->m_pkthdr.len < off)
1284 			return -1;
1285 		return off;
1286 
1287 	case IPPROTO_HOPOPTS:
1288 	case IPPROTO_ROUTING:
1289 	case IPPROTO_DSTOPTS:
1290 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1291 			return -1;
1292 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1293 		if (nxtp)
1294 			*nxtp = ip6e.ip6e_nxt;
1295 		off += (ip6e.ip6e_len + 1) << 3;
1296 		if (m->m_pkthdr.len < off)
1297 			return -1;
1298 		return off;
1299 
1300 	case IPPROTO_NONE:
1301 	case IPPROTO_ESP:
1302 	case IPPROTO_IPCOMP:
1303 		/* give up */
1304 		return -1;
1305 
1306 	default:
1307 		return -1;
1308 	}
1309 
1310 	return -1;
1311 }
1312 
1313 /*
1314  * get offset for the last header in the chain.  m will be kept untainted.
1315  */
1316 int
1317 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1318 {
1319 	int newoff;
1320 	int nxt;
1321 
1322 	if (!nxtp) {
1323 		nxt = -1;
1324 		nxtp = &nxt;
1325 	}
1326 	while (1) {
1327 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1328 		if (newoff < 0)
1329 			return off;
1330 		else if (newoff < off)
1331 			return -1;	/* invalid */
1332 		else if (newoff == off)
1333 			return newoff;
1334 
1335 		off = newoff;
1336 		proto = *nxtp;
1337 	}
1338 }
1339 
1340 /*
1341  * System control for IP6
1342  */
1343 
1344 const u_char inet6ctlerrmap[PRC_NCMDS] = {
1345 	0,		0,		0,		0,
1346 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1347 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1348 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1349 	0,		0,		0,		0,
1350 	ENOPROTOOPT
1351 };
1352 
1353 int *ipv6ctl_vars[IPV6CTL_MAXID] = IPV6CTL_VARS;
1354 
1355 int
1356 ip6_sysctl_ip6stat(void *oldp, size_t *oldlenp, void *newp)
1357 {
1358 	struct ip6stat *ip6stat;
1359 	int ret;
1360 
1361 	CTASSERT(sizeof(*ip6stat) == (ip6s_ncounters * sizeof(uint64_t)));
1362 
1363 	ip6stat = malloc(sizeof(*ip6stat), M_TEMP, M_WAITOK);
1364 	counters_read(ip6counters, (uint64_t *)ip6stat, ip6s_ncounters);
1365 	ret = sysctl_rdstruct(oldp, oldlenp, newp,
1366 	    ip6stat, sizeof(*ip6stat));
1367 	free(ip6stat, M_TEMP, sizeof(*ip6stat));
1368 
1369 	return (ret);
1370 }
1371 
1372 int
1373 ip6_sysctl_soiikey(void *oldp, size_t *oldlenp, void *newp, size_t newlen)
1374 {
1375 	struct ifnet *ifp;
1376 	uint8_t oldkey[IP6_SOIIKEY_LEN];
1377 	int error;
1378 
1379 	error = suser(curproc);
1380 	if (error != 0)
1381 		return (error);
1382 
1383 	memcpy(oldkey, ip6_soiikey, sizeof(oldkey));
1384 
1385 	error = sysctl_struct(oldp, oldlenp, newp, newlen, ip6_soiikey,
1386 	    sizeof(ip6_soiikey));
1387 
1388 	if (!error && memcmp(ip6_soiikey, oldkey, sizeof(oldkey)) != 0) {
1389 		TAILQ_FOREACH(ifp, &ifnet, if_list) {
1390 			if (ifp->if_flags & IFF_LOOPBACK)
1391 				continue;
1392 			NET_LOCK();
1393 			in6_soiiupdate(ifp);
1394 			NET_UNLOCK();
1395 		}
1396 	}
1397 
1398 	return (error);
1399 }
1400 
1401 int
1402 ip6_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
1403     void *newp, size_t newlen)
1404 {
1405 #ifdef MROUTING
1406 	extern int ip6_mrtproto;
1407 	extern struct mrt6stat mrt6stat;
1408 #endif
1409 	int error;
1410 
1411 	/* Almost all sysctl names at this level are terminal. */
1412 	if (namelen != 1 && name[0] != IPV6CTL_IFQUEUE)
1413 		return (ENOTDIR);
1414 
1415 	switch (name[0]) {
1416 	case IPV6CTL_DAD_PENDING:
1417 		return sysctl_rdint(oldp, oldlenp, newp, ip6_dad_pending);
1418 	case IPV6CTL_STATS:
1419 		return (ip6_sysctl_ip6stat(oldp, oldlenp, newp));
1420 #ifdef MROUTING
1421 	case IPV6CTL_MRTSTATS:
1422 		if (newp != NULL)
1423 			return (EPERM);
1424 		NET_LOCK();
1425 		error = sysctl_struct(oldp, oldlenp, newp, newlen,
1426 		    &mrt6stat, sizeof(mrt6stat));
1427 		NET_UNLOCK();
1428 		return (error);
1429 	case IPV6CTL_MRTPROTO:
1430 		return sysctl_rdint(oldp, oldlenp, newp, ip6_mrtproto);
1431 	case IPV6CTL_MRTMIF:
1432 		if (newp)
1433 			return (EPERM);
1434 		NET_LOCK();
1435 		error = mrt6_sysctl_mif(oldp, oldlenp);
1436 		NET_UNLOCK();
1437 		return (error);
1438 	case IPV6CTL_MRTMFC:
1439 		if (newp)
1440 			return (EPERM);
1441 		NET_LOCK();
1442 		error = mrt6_sysctl_mfc(oldp, oldlenp);
1443 		NET_UNLOCK();
1444 		return (error);
1445 #else
1446 	case IPV6CTL_MRTSTATS:
1447 	case IPV6CTL_MRTPROTO:
1448 	case IPV6CTL_MRTMIF:
1449 	case IPV6CTL_MRTMFC:
1450 		return (EOPNOTSUPP);
1451 #endif
1452 	case IPV6CTL_MTUDISCTIMEOUT:
1453 		NET_LOCK();
1454 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1455 		    &ip6_mtudisc_timeout);
1456 		if (icmp6_mtudisc_timeout_q != NULL)
1457 			rt_timer_queue_change(icmp6_mtudisc_timeout_q,
1458 			    ip6_mtudisc_timeout);
1459 		NET_UNLOCK();
1460 		return (error);
1461 	case IPV6CTL_IFQUEUE:
1462 		return (sysctl_niq(name + 1, namelen - 1,
1463 		    oldp, oldlenp, newp, newlen, &ip6intrq));
1464 	case IPV6CTL_SOIIKEY:
1465 		return (ip6_sysctl_soiikey(oldp, oldlenp, newp, newlen));
1466 	default:
1467 		if (name[0] < IPV6CTL_MAXID) {
1468 			NET_LOCK();
1469 			error = sysctl_int_arr(ipv6ctl_vars, name, namelen,
1470 			    oldp, oldlenp, newp, newlen);
1471 			NET_UNLOCK();
1472 			return (error);
1473 		}
1474 		return (EOPNOTSUPP);
1475 	}
1476 	/* NOTREACHED */
1477 }
1478 
1479 void
1480 ip6_send_dispatch(void *xmq)
1481 {
1482 	struct mbuf_queue *mq = xmq;
1483 	struct mbuf *m;
1484 	struct mbuf_list ml;
1485 
1486 	mq_delist(mq, &ml);
1487 	if (ml_empty(&ml))
1488 		return;
1489 
1490 	NET_RLOCK();
1491 	while ((m = ml_dequeue(&ml)) != NULL) {
1492 		/*
1493 		 * To avoid a "too big" situation at an intermediate router and
1494 		 * the path MTU discovery process, specify the IPV6_MINMTU
1495 		 * flag.  Note that only echo and node information replies are
1496 		 * affected, since the length of ICMP6 errors is limited to the
1497 		 * minimum MTU.
1498 		 */
1499 		ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL);
1500 	}
1501 	NET_RUNLOCK();
1502 }
1503 
1504 void
1505 ip6_send(struct mbuf *m)
1506 {
1507 	mq_enqueue(&ip6send_mq, m);
1508 	task_add(net_tq(0), &ip6send_task);
1509 }
1510