xref: /openbsd-src/sys/netinet6/ip6_input.c (revision 897fc685943471cf985a0fe38ba076ea6fe74fa5)
1 /*	$OpenBSD: ip6_input.c,v 1.214 2018/02/19 08:59:53 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 <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 (ifp->if_type == IFT_CARP &&
249 	    carp_lsdrop(m, AF_INET6, ip6->ip6_src.s6_addr32,
250 	    ip6->ip6_dst.s6_addr32, (ip6->ip6_nxt == IPPROTO_ICMPV6 ? 0 : 1)))
251 		goto bad;
252 #endif
253 	ip6stat_inc(ip6s_nxthist + ip6->ip6_nxt);
254 
255 	/*
256 	 * Check against address spoofing/corruption.
257 	 */
258 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
259 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
260 		/*
261 		 * XXX: "badscope" is not very suitable for a multicast source.
262 		 */
263 		ip6stat_inc(ip6s_badscope);
264 		goto bad;
265 	}
266 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
267 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
268 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
269 		    ip6stat_inc(ip6s_badscope);
270 		    goto bad;
271 	}
272 	/* Drop packets if interface ID portion is already filled. */
273 	if (((IN6_IS_SCOPE_EMBED(&ip6->ip6_src) && ip6->ip6_src.s6_addr16[1]) ||
274 	    (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst) && ip6->ip6_dst.s6_addr16[1])) &&
275 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
276 		ip6stat_inc(ip6s_badscope);
277 		goto bad;
278 	}
279 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
280 	    !(m->m_flags & M_LOOP)) {
281 		/*
282 		 * In this case, the packet should come from the loopback
283 		 * interface.  However, we cannot just check the if_flags,
284 		 * because ip6_mloopback() passes the "actual" interface
285 		 * as the outgoing/incoming interface.
286 		 */
287 		ip6stat_inc(ip6s_badscope);
288 		goto bad;
289 	}
290 
291 	/*
292 	 * The following check is not documented in specs.  A malicious
293 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
294 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
295 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
296 	 *
297 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
298 	 * support IPv4-less kernel compilation, we cannot support SIIT
299 	 * environment at all.  So, it makes more sense for us to reject any
300 	 * malicious packets for non-SIIT environment, than try to do a
301 	 * partial support for SIIT environment.
302 	 */
303 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
304 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
305 		ip6stat_inc(ip6s_badscope);
306 		goto bad;
307 	}
308 
309 	/*
310 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
311 	 *
312 	 * The code forbids automatic tunneling as per RFC4213.
313 	 */
314 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
315 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
316 		ip6stat_inc(ip6s_badscope);
317 		goto bad;
318 	}
319 
320 	/*
321 	 * If the packet has been received on a loopback interface it
322 	 * can be destinated to any local address, not necessarily to
323 	 * an address configured on `ifp'.
324 	 */
325 	if (ifp->if_flags & IFF_LOOPBACK) {
326 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
327 			src_scope = ip6->ip6_src.s6_addr16[1];
328 			ip6->ip6_src.s6_addr16[1] = 0;
329 		}
330 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
331 			dst_scope = ip6->ip6_dst.s6_addr16[1];
332 			ip6->ip6_dst.s6_addr16[1] = 0;
333 		}
334 	}
335 
336 #if NPF > 0
337 	/*
338 	 * Packet filter
339 	 */
340 	odst = ip6->ip6_dst;
341 	if (pf_test(AF_INET6, PF_IN, ifp, mp) != PF_PASS)
342 		goto bad;
343 	m = *mp;
344 	if (m == NULL)
345 		goto bad;
346 
347 	ip6 = mtod(m, struct ip6_hdr *);
348 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
349 #endif
350 
351 	/*
352 	 * Without embedded scope ID we cannot find link-local
353 	 * addresses in the routing table.
354 	 */
355 	if (ifp->if_flags & IFF_LOOPBACK) {
356 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
357 			ip6->ip6_src.s6_addr16[1] = src_scope;
358 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
359 			ip6->ip6_dst.s6_addr16[1] = dst_scope;
360 	} else {
361 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
362 			ip6->ip6_src.s6_addr16[1] = htons(ifp->if_index);
363 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
364 			ip6->ip6_dst.s6_addr16[1] = htons(ifp->if_index);
365 	}
366 
367 	/*
368 	 * Be more secure than RFC5095 and scan for type 0 routing headers.
369 	 * If pf has already scanned the header chain, do not do it twice.
370 	 */
371 	if (!(m->m_pkthdr.pf.flags & PF_TAG_PROCESSED) &&
372 	    ip6_check_rh0hdr(m, offp)) {
373 		ip6stat_inc(ip6s_badoptions);
374 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, *offp);
375 		m = *mp = NULL;
376 		goto bad;
377 	}
378 
379 	if (IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
380 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) {
381 		nxt = ip6_ours(mp, offp, nxt, af);
382 		goto out;
383 	}
384 
385 #if NPF > 0
386 	if (pf_ouraddr(m) == 1) {
387 		nxt = ip6_ours(mp, offp, nxt, af);
388 		goto out;
389 	}
390 #endif
391 
392 	/*
393 	 * Multicast check
394 	 */
395 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
396 		/*
397 		 * Make sure M_MCAST is set.  It should theoretically
398 		 * already be there, but let's play safe because upper
399 		 * layers check for this flag.
400 		 */
401 		m->m_flags |= M_MCAST;
402 
403 		/*
404 		 * See if we belong to the destination multicast group on the
405 		 * arrival interface.
406 		 */
407 		if (in6_hasmulti(&ip6->ip6_dst, ifp))
408 			ours = 1;
409 
410 #ifdef MROUTING
411 		if (ip6_mforwarding && ip6_mrouter[ifp->if_rdomain]) {
412 			int error;
413 
414 			if (ip6_hbhchcheck(m, offp, &nxt, &ours))
415 				goto out;
416 
417 			ip6 = mtod(m, struct ip6_hdr *);
418 
419 			/*
420 			 * If we are acting as a multicast router, all
421 			 * incoming multicast packets are passed to the
422 			 * kernel-level multicast forwarding function.
423 			 * The packet is returned (relatively) intact; if
424 			 * ip6_mforward() returns a non-zero value, the packet
425 			 * must be discarded, else it may be accepted below.
426 			 */
427 			KERNEL_LOCK();
428 			error = ip6_mforward(ip6, ifp, m);
429 			KERNEL_UNLOCK();
430 			if (error) {
431 				ip6stat_inc(ip6s_cantforward);
432 				goto bad;
433 			}
434 
435 			if (ours) {
436 				if (af == AF_UNSPEC) {
437 					KERNEL_LOCK();
438 					nxt = ip_deliver(mp, offp, nxt,
439 					    AF_INET6);
440 					KERNEL_UNLOCK();
441 				}
442 				goto out;
443 			}
444 			goto bad;
445 		}
446 #endif
447 		if (!ours) {
448 			ip6stat_inc(ip6s_notmember);
449 			if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
450 				ip6stat_inc(ip6s_cantforward);
451 			goto bad;
452 		}
453 		nxt = ip6_ours(mp, offp, nxt, af);
454 		goto out;
455 	}
456 
457 
458 	/*
459 	 *  Unicast check
460 	 */
461 	memset(&sin6, 0, sizeof(struct sockaddr_in6));
462 	sin6.sin6_len = sizeof(struct sockaddr_in6);
463 	sin6.sin6_family = AF_INET6;
464 	sin6.sin6_addr = ip6->ip6_dst;
465 	rt = rtalloc_mpath(sin6tosa(&sin6), &ip6->ip6_src.s6_addr32[0],
466 	    m->m_pkthdr.ph_rtableid);
467 
468 	/*
469 	 * Accept the packet if the route to the destination is marked
470 	 * as local.
471 	 */
472 	if (rtisvalid(rt) && ISSET(rt->rt_flags, RTF_LOCAL)) {
473 		struct in6_ifaddr *ia6 = ifatoia6(rt->rt_ifa);
474 		if (ia6->ia6_flags & IN6_IFF_ANYCAST)
475 			m->m_flags |= M_ACAST;
476 		/*
477 		 * packets to a tentative, duplicated, or somehow invalid
478 		 * address must not be accepted.
479 		 */
480 		if ((ia6->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED))) {
481 			char src[INET6_ADDRSTRLEN], dst[INET6_ADDRSTRLEN];
482 
483 			inet_ntop(AF_INET6, &ip6->ip6_src, src, sizeof(src));
484 			inet_ntop(AF_INET6, &ip6->ip6_dst, dst, sizeof(dst));
485 			/* address is not ready, so discard the packet. */
486 			nd6log((LOG_INFO,
487 			    "%s: packet to an unready address %s->%s\n",
488 			    __func__, src, dst));
489 
490 			goto bad;
491 		} else {
492 			nxt = ip6_ours(mp, offp, nxt, af);
493 			goto out;
494 		}
495 	}
496 
497 #if NCARP > 0
498 	if (ifp->if_type == IFT_CARP && ip6->ip6_nxt == IPPROTO_ICMPV6 &&
499 	    carp_lsdrop(m, AF_INET6, ip6->ip6_src.s6_addr32,
500 	    ip6->ip6_dst.s6_addr32, 1))
501 		goto bad;
502 #endif
503 	/*
504 	 * Now there is no reason to process the packet if it's not our own
505 	 * and we're not a router.
506 	 */
507 	if (!ip6_forwarding) {
508 		ip6stat_inc(ip6s_cantforward);
509 		goto bad;
510 	}
511 
512 	if (ip6_hbhchcheck(m, offp, &nxt, &ours))
513 		goto out;
514 
515 	if (ours) {
516 		if (af == AF_UNSPEC) {
517 			KERNEL_LOCK();
518 			nxt = ip_deliver(mp, offp, nxt, AF_INET6);
519 			KERNEL_UNLOCK();
520 		}
521 		goto out;
522 	}
523 
524 #ifdef IPSEC
525 	if (ipsec_in_use) {
526 		int rv;
527 
528 		rv = ipsec_forward_check(m, *offp, AF_INET6);
529 		if (rv != 0) {
530 			ip6stat_inc(ip6s_cantforward);
531 			goto bad;
532 		}
533 		/*
534 		 * Fall through, forward packet. Outbound IPsec policy
535 		 * checking will occur in ip6_forward().
536 		 */
537 	}
538 #endif /* IPSEC */
539 
540 	ip6_forward(m, rt, srcrt);
541 	*mp = NULL;
542 	return IPPROTO_DONE;
543  bad:
544 	nxt = IPPROTO_DONE;
545 	m_freemp(mp);
546  out:
547 	rtfree(rt);
548 	return nxt;
549 }
550 
551 int
552 ip6_local(struct mbuf **mp, int *offp, int nxt, int af)
553 {
554 	if (ip6_hbhchcheck(*mp, offp, &nxt, NULL))
555 		return IPPROTO_DONE;
556 
557 	/* Check wheter we are already in a IPv4/IPv6 local deliver loop. */
558 	if (af == AF_UNSPEC)
559 		nxt = ip_deliver(mp, offp, nxt, AF_INET6);
560 	return nxt;
561 }
562 
563 int
564 ip6_hbhchcheck(struct mbuf *m, int *offp, int *nxtp, int *oursp)
565 {
566 	struct ip6_hdr *ip6;
567 	u_int32_t plen, rtalert = ~0;
568 
569 	ip6 = mtod(m, struct ip6_hdr *);
570 
571 	/*
572 	 * Process Hop-by-Hop options header if it's contained.
573 	 * m may be modified in ip6_hopopts_input().
574 	 * If a JumboPayload option is included, plen will also be modified.
575 	 */
576 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
577 	*offp = sizeof(struct ip6_hdr);
578 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
579 		struct ip6_hbh *hbh;
580 
581 		if (ip6_hopopts_input(&plen, &rtalert, &m, offp)) {
582 			goto bad;	/* m have already been freed */
583 		}
584 
585 		/* adjust pointer */
586 		ip6 = mtod(m, struct ip6_hdr *);
587 
588 		/*
589 		 * if the payload length field is 0 and the next header field
590 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
591 		 * option MUST be included.
592 		 */
593 		if (ip6->ip6_plen == 0 && plen == 0) {
594 			/*
595 			 * Note that if a valid jumbo payload option is
596 			 * contained, ip6_hopopts_input() must set a valid
597 			 * (non-zero) payload length to the variable plen.
598 			 */
599 			ip6stat_inc(ip6s_badoptions);
600 			icmp6_error(m, ICMP6_PARAM_PROB,
601 				    ICMP6_PARAMPROB_HEADER,
602 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
603 			goto bad;
604 		}
605 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
606 			sizeof(struct ip6_hbh));
607 		if (hbh == NULL) {
608 			ip6stat_inc(ip6s_tooshort);
609 			goto bad;
610 		}
611 		*nxtp = hbh->ip6h_nxt;
612 
613 		/*
614 		 * accept the packet if a router alert option is included
615 		 * and we act as an IPv6 router.
616 		 */
617 		if (rtalert != ~0 && ip6_forwarding && oursp != NULL)
618 			*oursp = 1;
619 	} else
620 		*nxtp = ip6->ip6_nxt;
621 
622 	/*
623 	 * Check that the amount of data in the buffers
624 	 * is as at least much as the IPv6 header would have us expect.
625 	 * Trim mbufs if longer than we expect.
626 	 * Drop packet if shorter than we expect.
627 	 */
628 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
629 		ip6stat_inc(ip6s_tooshort);
630 		m_freem(m);
631 		goto bad;
632 	}
633 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
634 		if (m->m_len == m->m_pkthdr.len) {
635 			m->m_len = sizeof(struct ip6_hdr) + plen;
636 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
637 		} else {
638 			m_adj(m,
639 			    sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
640 		}
641 	}
642 
643 	return (0);
644 
645  bad:
646 	*nxtp = IPPROTO_DONE;
647 	return (-1);
648 }
649 
650 /* scan packet for RH0 routing header. Mostly stolen from pf.c:pf_test() */
651 int
652 ip6_check_rh0hdr(struct mbuf *m, int *offp)
653 {
654 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
655 	struct ip6_rthdr rthdr;
656 	struct ip6_ext opt6;
657 	u_int8_t proto = ip6->ip6_nxt;
658 	int done = 0, lim, off, rh_cnt = 0;
659 
660 	off = ((caddr_t)ip6 - m->m_data) + sizeof(struct ip6_hdr);
661 	lim = min(m->m_pkthdr.len, ntohs(ip6->ip6_plen) + sizeof(*ip6));
662 	do {
663 		switch (proto) {
664 		case IPPROTO_ROUTING:
665 			*offp = off;
666 			if (rh_cnt++) {
667 				/* more than one rh header present */
668 				return (1);
669 			}
670 
671 			if (off + sizeof(rthdr) > lim) {
672 				/* packet to short to make sense */
673 				return (1);
674 			}
675 
676 			m_copydata(m, off, sizeof(rthdr), (caddr_t)&rthdr);
677 
678 			if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
679 				*offp += offsetof(struct ip6_rthdr, ip6r_type);
680 				return (1);
681 			}
682 
683 			off += (rthdr.ip6r_len + 1) * 8;
684 			proto = rthdr.ip6r_nxt;
685 			break;
686 		case IPPROTO_AH:
687 		case IPPROTO_HOPOPTS:
688 		case IPPROTO_DSTOPTS:
689 			/* get next header and header length */
690 			if (off + sizeof(opt6) > lim) {
691 				/*
692 				 * Packet to short to make sense, we could
693 				 * reject the packet but as a router we
694 				 * should not do that so forward it.
695 				 */
696 				return (0);
697 			}
698 
699 			m_copydata(m, off, sizeof(opt6), (caddr_t)&opt6);
700 
701 			if (proto == IPPROTO_AH)
702 				off += (opt6.ip6e_len + 2) * 4;
703 			else
704 				off += (opt6.ip6e_len + 1) * 8;
705 			proto = opt6.ip6e_nxt;
706 			break;
707 		case IPPROTO_FRAGMENT:
708 		default:
709 			/* end of header stack */
710 			done = 1;
711 			break;
712 		}
713 	} while (!done);
714 
715 	return (0);
716 }
717 
718 /*
719  * Hop-by-Hop options header processing. If a valid jumbo payload option is
720  * included, the real payload length will be stored in plenp.
721  *
722  * rtalertp - XXX: should be stored in a more smart way
723  */
724 int
725 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, struct mbuf **mp,
726     int *offp)
727 {
728 	struct mbuf *m = *mp;
729 	int off = *offp, hbhlen;
730 	struct ip6_hbh *hbh;
731 
732 	/* validation of the length of the header */
733 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
734 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
735 	if (hbh == NULL) {
736 		ip6stat_inc(ip6s_tooshort);
737 		return -1;
738 	}
739 	hbhlen = (hbh->ip6h_len + 1) << 3;
740 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
741 		hbhlen);
742 	if (hbh == NULL) {
743 		ip6stat_inc(ip6s_tooshort);
744 		return -1;
745 	}
746 	off += hbhlen;
747 	hbhlen -= sizeof(struct ip6_hbh);
748 
749 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
750 				hbhlen, rtalertp, plenp) < 0)
751 		return (-1);
752 
753 	*offp = off;
754 	*mp = m;
755 	return (0);
756 }
757 
758 /*
759  * Search header for all Hop-by-hop options and process each option.
760  * This function is separate from ip6_hopopts_input() in order to
761  * handle a case where the sending node itself process its hop-by-hop
762  * options header. In such a case, the function is called from ip6_output().
763  *
764  * The function assumes that hbh header is located right after the IPv6 header
765  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
766  * opthead + hbhlen is located in continuous memory region.
767  */
768 int
769 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
770     u_int32_t *rtalertp, u_int32_t *plenp)
771 {
772 	struct ip6_hdr *ip6;
773 	int optlen = 0;
774 	u_int8_t *opt = opthead;
775 	u_int16_t rtalert_val;
776 	u_int32_t jumboplen;
777 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
778 
779 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
780 		switch (*opt) {
781 		case IP6OPT_PAD1:
782 			optlen = 1;
783 			break;
784 		case IP6OPT_PADN:
785 			if (hbhlen < IP6OPT_MINLEN) {
786 				ip6stat_inc(ip6s_toosmall);
787 				goto bad;
788 			}
789 			optlen = *(opt + 1) + 2;
790 			break;
791 		case IP6OPT_ROUTER_ALERT:
792 			/* XXX may need check for alignment */
793 			if (hbhlen < IP6OPT_RTALERT_LEN) {
794 				ip6stat_inc(ip6s_toosmall);
795 				goto bad;
796 			}
797 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
798 				/* XXX stat */
799 				icmp6_error(m, ICMP6_PARAM_PROB,
800 				    ICMP6_PARAMPROB_HEADER,
801 				    erroff + opt + 1 - opthead);
802 				return (-1);
803 			}
804 			optlen = IP6OPT_RTALERT_LEN;
805 			memcpy((caddr_t)&rtalert_val, (caddr_t)(opt + 2), 2);
806 			*rtalertp = ntohs(rtalert_val);
807 			break;
808 		case IP6OPT_JUMBO:
809 			/* XXX may need check for alignment */
810 			if (hbhlen < IP6OPT_JUMBO_LEN) {
811 				ip6stat_inc(ip6s_toosmall);
812 				goto bad;
813 			}
814 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
815 				/* XXX stat */
816 				icmp6_error(m, ICMP6_PARAM_PROB,
817 				    ICMP6_PARAMPROB_HEADER,
818 				    erroff + opt + 1 - opthead);
819 				return (-1);
820 			}
821 			optlen = IP6OPT_JUMBO_LEN;
822 
823 			/*
824 			 * IPv6 packets that have non 0 payload length
825 			 * must not contain a jumbo payload option.
826 			 */
827 			ip6 = mtod(m, struct ip6_hdr *);
828 			if (ip6->ip6_plen) {
829 				ip6stat_inc(ip6s_badoptions);
830 				icmp6_error(m, ICMP6_PARAM_PROB,
831 				    ICMP6_PARAMPROB_HEADER,
832 				    erroff + opt - opthead);
833 				return (-1);
834 			}
835 
836 			/*
837 			 * We may see jumbolen in unaligned location, so
838 			 * we'd need to perform memcpy().
839 			 */
840 			memcpy(&jumboplen, opt + 2, sizeof(jumboplen));
841 			jumboplen = (u_int32_t)htonl(jumboplen);
842 
843 #if 1
844 			/*
845 			 * if there are multiple jumbo payload options,
846 			 * *plenp will be non-zero and the packet will be
847 			 * rejected.
848 			 * the behavior may need some debate in ipngwg -
849 			 * multiple options does not make sense, however,
850 			 * there's no explicit mention in specification.
851 			 */
852 			if (*plenp != 0) {
853 				ip6stat_inc(ip6s_badoptions);
854 				icmp6_error(m, ICMP6_PARAM_PROB,
855 				    ICMP6_PARAMPROB_HEADER,
856 				    erroff + opt + 2 - opthead);
857 				return (-1);
858 			}
859 #endif
860 
861 			/*
862 			 * jumbo payload length must be larger than 65535.
863 			 */
864 			if (jumboplen <= IPV6_MAXPACKET) {
865 				ip6stat_inc(ip6s_badoptions);
866 				icmp6_error(m, ICMP6_PARAM_PROB,
867 				    ICMP6_PARAMPROB_HEADER,
868 				    erroff + opt + 2 - opthead);
869 				return (-1);
870 			}
871 			*plenp = jumboplen;
872 
873 			break;
874 		default:		/* unknown option */
875 			if (hbhlen < IP6OPT_MINLEN) {
876 				ip6stat_inc(ip6s_toosmall);
877 				goto bad;
878 			}
879 			optlen = ip6_unknown_opt(opt, m,
880 			    erroff + opt - opthead);
881 			if (optlen == -1)
882 				return (-1);
883 			optlen += 2;
884 			break;
885 		}
886 	}
887 
888 	return (0);
889 
890   bad:
891 	m_freem(m);
892 	return (-1);
893 }
894 
895 /*
896  * Unknown option processing.
897  * The third argument `off' is the offset from the IPv6 header to the option,
898  * which allows returning an ICMPv6 error even if the IPv6 header and the
899  * option header are not continuous.
900  */
901 int
902 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
903 {
904 	struct ip6_hdr *ip6;
905 
906 	switch (IP6OPT_TYPE(*optp)) {
907 	case IP6OPT_TYPE_SKIP: /* ignore the option */
908 		return ((int)*(optp + 1));
909 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
910 		m_freem(m);
911 		return (-1);
912 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
913 		ip6stat_inc(ip6s_badoptions);
914 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
915 		return (-1);
916 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
917 		ip6stat_inc(ip6s_badoptions);
918 		ip6 = mtod(m, struct ip6_hdr *);
919 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
920 		    (m->m_flags & (M_BCAST|M_MCAST)))
921 			m_freem(m);
922 		else
923 			icmp6_error(m, ICMP6_PARAM_PROB,
924 				    ICMP6_PARAMPROB_OPTION, off);
925 		return (-1);
926 	}
927 
928 	m_freem(m);		/* XXX: NOTREACHED */
929 	return (-1);
930 }
931 
932 /*
933  * Create the "control" list for this pcb.
934  *
935  * The routine will be called from upper layer handlers like udp_input().
936  * Thus the routine assumes that the caller (udp_input) have already
937  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
938  * very first mbuf on the mbuf chain.
939  * We may want to add some infinite loop prevention or sanity checks for safety.
940  * (This applies only when you are using KAME mbuf chain restriction, i.e.
941  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
942  */
943 void
944 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
945 {
946 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
947 
948 	if (in6p->inp_socket->so_options & SO_TIMESTAMP) {
949 		struct timeval tv;
950 
951 		microtime(&tv);
952 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
953 		    SCM_TIMESTAMP, SOL_SOCKET);
954 		if (*mp)
955 			mp = &(*mp)->m_next;
956 	}
957 
958 	/* RFC 2292 sec. 5 */
959 	if ((in6p->inp_flags & IN6P_PKTINFO) != 0) {
960 		struct in6_pktinfo pi6;
961 		memcpy(&pi6.ipi6_addr, &ip6->ip6_dst, sizeof(struct in6_addr));
962 		if (IN6_IS_SCOPE_EMBED(&pi6.ipi6_addr))
963 			pi6.ipi6_addr.s6_addr16[1] = 0;
964 		pi6.ipi6_ifindex = m ? m->m_pkthdr.ph_ifidx : 0;
965 		*mp = sbcreatecontrol((caddr_t) &pi6,
966 		    sizeof(struct in6_pktinfo),
967 		    IPV6_PKTINFO, IPPROTO_IPV6);
968 		if (*mp)
969 			mp = &(*mp)->m_next;
970 	}
971 
972 	if ((in6p->inp_flags & IN6P_HOPLIMIT) != 0) {
973 		int hlim = ip6->ip6_hlim & 0xff;
974 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
975 		    IPV6_HOPLIMIT, IPPROTO_IPV6);
976 		if (*mp)
977 			mp = &(*mp)->m_next;
978 	}
979 
980 	if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
981 		u_int32_t flowinfo;
982 		int tclass;
983 
984 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
985 		flowinfo >>= 20;
986 
987 		tclass = flowinfo & 0xff;
988 		*mp = sbcreatecontrol((caddr_t)&tclass, sizeof(tclass),
989 		    IPV6_TCLASS, IPPROTO_IPV6);
990 		if (*mp)
991 			mp = &(*mp)->m_next;
992 	}
993 
994 	/*
995 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
996 	 * privilege for the option (see ip6_ctloutput), but it might be too
997 	 * strict, since there might be some hop-by-hop options which can be
998 	 * returned to normal user.
999 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1000 	 */
1001 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1002 		/*
1003 		 * Check if a hop-by-hop options header is contained in the
1004 		 * received packet, and if so, store the options as ancillary
1005 		 * data. Note that a hop-by-hop options header must be
1006 		 * just after the IPv6 header, which is assured through the
1007 		 * IPv6 input processing.
1008 		 */
1009 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1010 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1011 			struct ip6_hbh *hbh;
1012 			int hbhlen = 0;
1013 			struct mbuf *ext;
1014 
1015 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1016 			    ip6->ip6_nxt);
1017 			if (ext == NULL) {
1018 				ip6stat_inc(ip6s_tooshort);
1019 				return;
1020 			}
1021 			hbh = mtod(ext, struct ip6_hbh *);
1022 			hbhlen = (hbh->ip6h_len + 1) << 3;
1023 			if (hbhlen != ext->m_len) {
1024 				m_freem(ext);
1025 				ip6stat_inc(ip6s_tooshort);
1026 				return;
1027 			}
1028 
1029 			/*
1030 			 * XXX: We copy the whole header even if a
1031 			 * jumbo payload option is included, the option which
1032 			 * is to be removed before returning according to
1033 			 * RFC2292.
1034 			 * Note: this constraint is removed in RFC3542.
1035 			 */
1036 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1037 			    IPV6_HOPOPTS,
1038 			    IPPROTO_IPV6);
1039 			if (*mp)
1040 				mp = &(*mp)->m_next;
1041 			m_freem(ext);
1042 		}
1043 	}
1044 
1045 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1046 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1047 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1048 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1049 
1050 		/*
1051 		 * Search for destination options headers or routing
1052 		 * header(s) through the header chain, and stores each
1053 		 * header as ancillary data.
1054 		 * Note that the order of the headers remains in
1055 		 * the chain of ancillary data.
1056 		 */
1057 		while (1) {	/* is explicit loop prevention necessary? */
1058 			struct ip6_ext *ip6e = NULL;
1059 			int elen;
1060 			struct mbuf *ext = NULL;
1061 
1062 			/*
1063 			 * if it is not an extension header, don't try to
1064 			 * pull it from the chain.
1065 			 */
1066 			switch (nxt) {
1067 			case IPPROTO_DSTOPTS:
1068 			case IPPROTO_ROUTING:
1069 			case IPPROTO_HOPOPTS:
1070 			case IPPROTO_AH: /* is it possible? */
1071 				break;
1072 			default:
1073 				goto loopend;
1074 			}
1075 
1076 			ext = ip6_pullexthdr(m, off, nxt);
1077 			if (ext == NULL) {
1078 				ip6stat_inc(ip6s_tooshort);
1079 				return;
1080 			}
1081 			ip6e = mtod(ext, struct ip6_ext *);
1082 			if (nxt == IPPROTO_AH)
1083 				elen = (ip6e->ip6e_len + 2) << 2;
1084 			else
1085 				elen = (ip6e->ip6e_len + 1) << 3;
1086 			if (elen != ext->m_len) {
1087 				m_freem(ext);
1088 				ip6stat_inc(ip6s_tooshort);
1089 				return;
1090 			}
1091 
1092 			switch (nxt) {
1093 			case IPPROTO_DSTOPTS:
1094 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1095 					break;
1096 
1097 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1098 				    IPV6_DSTOPTS,
1099 				    IPPROTO_IPV6);
1100 				if (*mp)
1101 					mp = &(*mp)->m_next;
1102 				break;
1103 
1104 			case IPPROTO_ROUTING:
1105 				if (!(in6p->inp_flags & IN6P_RTHDR))
1106 					break;
1107 
1108 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1109 				    IPV6_RTHDR,
1110 				    IPPROTO_IPV6);
1111 				if (*mp)
1112 					mp = &(*mp)->m_next;
1113 				break;
1114 
1115 			case IPPROTO_HOPOPTS:
1116 			case IPPROTO_AH: /* is it possible? */
1117 				break;
1118 
1119 			default:
1120 				/*
1121 				 * other cases have been filtered in the above.
1122 				 * none will visit this case.  here we supply
1123 				 * the code just in case (nxt overwritten or
1124 				 * other cases).
1125 				 */
1126 				m_freem(ext);
1127 				goto loopend;
1128 
1129 			}
1130 
1131 			/* proceed with the next header. */
1132 			off += elen;
1133 			nxt = ip6e->ip6e_nxt;
1134 			ip6e = NULL;
1135 			m_freem(ext);
1136 			ext = NULL;
1137 		}
1138 loopend:
1139 		;
1140 	}
1141 }
1142 
1143 /*
1144  * pull single extension header from mbuf chain.  returns single mbuf that
1145  * contains the result, or NULL on error.
1146  */
1147 struct mbuf *
1148 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1149 {
1150 	struct ip6_ext ip6e;
1151 	size_t elen;
1152 	struct mbuf *n;
1153 
1154 #ifdef DIAGNOSTIC
1155 	switch (nxt) {
1156 	case IPPROTO_DSTOPTS:
1157 	case IPPROTO_ROUTING:
1158 	case IPPROTO_HOPOPTS:
1159 	case IPPROTO_AH: /* is it possible? */
1160 		break;
1161 	default:
1162 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1163 	}
1164 #endif
1165 
1166 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1167 	if (nxt == IPPROTO_AH)
1168 		elen = (ip6e.ip6e_len + 2) << 2;
1169 	else
1170 		elen = (ip6e.ip6e_len + 1) << 3;
1171 
1172 	MGET(n, M_DONTWAIT, MT_DATA);
1173 	if (n && elen >= MLEN) {
1174 		MCLGET(n, M_DONTWAIT);
1175 		if ((n->m_flags & M_EXT) == 0) {
1176 			m_free(n);
1177 			n = NULL;
1178 		}
1179 	}
1180 	if (!n)
1181 		return NULL;
1182 
1183 	n->m_len = 0;
1184 	if (elen >= M_TRAILINGSPACE(n)) {
1185 		m_free(n);
1186 		return NULL;
1187 	}
1188 
1189 	m_copydata(m, off, elen, mtod(n, caddr_t));
1190 	n->m_len = elen;
1191 	return n;
1192 }
1193 
1194 /*
1195  * Get offset to the previous header followed by the header
1196  * currently processed.
1197  */
1198 int
1199 ip6_get_prevhdr(struct mbuf *m, int off)
1200 {
1201 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1202 
1203 	if (off == sizeof(struct ip6_hdr)) {
1204 		return offsetof(struct ip6_hdr, ip6_nxt);
1205 	} else if (off < sizeof(struct ip6_hdr)) {
1206 		panic("%s: off < sizeof(struct ip6_hdr)", __func__);
1207 	} else {
1208 		int len, nlen, nxt;
1209 		struct ip6_ext ip6e;
1210 
1211 		nxt = ip6->ip6_nxt;
1212 		len = sizeof(struct ip6_hdr);
1213 		nlen = 0;
1214 		while (len < off) {
1215 			m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e);
1216 
1217 			switch (nxt) {
1218 			case IPPROTO_FRAGMENT:
1219 				nlen = sizeof(struct ip6_frag);
1220 				break;
1221 			case IPPROTO_AH:
1222 				nlen = (ip6e.ip6e_len + 2) << 2;
1223 				break;
1224 			default:
1225 				nlen = (ip6e.ip6e_len + 1) << 3;
1226 				break;
1227 			}
1228 			len += nlen;
1229 			nxt = ip6e.ip6e_nxt;
1230 		}
1231 
1232 		return (len - nlen);
1233 	}
1234 }
1235 
1236 /*
1237  * get next header offset.  m will be retained.
1238  */
1239 int
1240 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1241 {
1242 	struct ip6_hdr ip6;
1243 	struct ip6_ext ip6e;
1244 	struct ip6_frag fh;
1245 
1246 	/* just in case */
1247 	if (m == NULL)
1248 		panic("ip6_nexthdr: m == NULL");
1249 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1250 		return -1;
1251 
1252 	switch (proto) {
1253 	case IPPROTO_IPV6:
1254 		if (m->m_pkthdr.len < off + sizeof(ip6))
1255 			return -1;
1256 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1257 		if (nxtp)
1258 			*nxtp = ip6.ip6_nxt;
1259 		off += sizeof(ip6);
1260 		return off;
1261 
1262 	case IPPROTO_FRAGMENT:
1263 		/*
1264 		 * terminate parsing if it is not the first fragment,
1265 		 * it does not make sense to parse through it.
1266 		 */
1267 		if (m->m_pkthdr.len < off + sizeof(fh))
1268 			return -1;
1269 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1270 		if ((fh.ip6f_offlg & IP6F_OFF_MASK) != 0)
1271 			return -1;
1272 		if (nxtp)
1273 			*nxtp = fh.ip6f_nxt;
1274 		off += sizeof(struct ip6_frag);
1275 		return off;
1276 
1277 	case IPPROTO_AH:
1278 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1279 			return -1;
1280 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1281 		if (nxtp)
1282 			*nxtp = ip6e.ip6e_nxt;
1283 		off += (ip6e.ip6e_len + 2) << 2;
1284 		if (m->m_pkthdr.len < off)
1285 			return -1;
1286 		return off;
1287 
1288 	case IPPROTO_HOPOPTS:
1289 	case IPPROTO_ROUTING:
1290 	case IPPROTO_DSTOPTS:
1291 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1292 			return -1;
1293 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1294 		if (nxtp)
1295 			*nxtp = ip6e.ip6e_nxt;
1296 		off += (ip6e.ip6e_len + 1) << 3;
1297 		if (m->m_pkthdr.len < off)
1298 			return -1;
1299 		return off;
1300 
1301 	case IPPROTO_NONE:
1302 	case IPPROTO_ESP:
1303 	case IPPROTO_IPCOMP:
1304 		/* give up */
1305 		return -1;
1306 
1307 	default:
1308 		return -1;
1309 	}
1310 
1311 	return -1;
1312 }
1313 
1314 /*
1315  * get offset for the last header in the chain.  m will be kept untainted.
1316  */
1317 int
1318 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1319 {
1320 	int newoff;
1321 	int nxt;
1322 
1323 	if (!nxtp) {
1324 		nxt = -1;
1325 		nxtp = &nxt;
1326 	}
1327 	while (1) {
1328 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1329 		if (newoff < 0)
1330 			return off;
1331 		else if (newoff < off)
1332 			return -1;	/* invalid */
1333 		else if (newoff == off)
1334 			return newoff;
1335 
1336 		off = newoff;
1337 		proto = *nxtp;
1338 	}
1339 }
1340 
1341 /*
1342  * System control for IP6
1343  */
1344 
1345 const u_char inet6ctlerrmap[PRC_NCMDS] = {
1346 	0,		0,		0,		0,
1347 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1348 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1349 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1350 	0,		0,		0,		0,
1351 	ENOPROTOOPT
1352 };
1353 
1354 int *ipv6ctl_vars[IPV6CTL_MAXID] = IPV6CTL_VARS;
1355 
1356 int
1357 ip6_sysctl_ip6stat(void *oldp, size_t *oldlenp, void *newp)
1358 {
1359 	struct ip6stat *ip6stat;
1360 	int ret;
1361 
1362 	CTASSERT(sizeof(*ip6stat) == (ip6s_ncounters * sizeof(uint64_t)));
1363 
1364 	ip6stat = malloc(sizeof(*ip6stat), M_TEMP, M_WAITOK);
1365 	counters_read(ip6counters, (uint64_t *)ip6stat, ip6s_ncounters);
1366 	ret = sysctl_rdstruct(oldp, oldlenp, newp,
1367 	    ip6stat, sizeof(*ip6stat));
1368 	free(ip6stat, M_TEMP, sizeof(*ip6stat));
1369 
1370 	return (ret);
1371 }
1372 
1373 int
1374 ip6_sysctl_soiikey(void *oldp, size_t *oldlenp, void *newp, size_t newlen)
1375 {
1376 	struct ifnet *ifp;
1377 	uint8_t oldkey[IP6_SOIIKEY_LEN];
1378 	int error;
1379 
1380 	error = suser(curproc);
1381 	if (error != 0)
1382 		return (error);
1383 
1384 	memcpy(oldkey, ip6_soiikey, sizeof(oldkey));
1385 
1386 	error = sysctl_struct(oldp, oldlenp, newp, newlen, ip6_soiikey,
1387 	    sizeof(ip6_soiikey));
1388 
1389 	if (!error && memcmp(ip6_soiikey, oldkey, sizeof(oldkey)) != 0) {
1390 		TAILQ_FOREACH(ifp, &ifnet, if_list) {
1391 			if (ifp->if_flags & IFF_LOOPBACK)
1392 				continue;
1393 			NET_LOCK();
1394 			in6_soiiupdate(ifp);
1395 			NET_UNLOCK();
1396 		}
1397 	}
1398 
1399 	return (error);
1400 }
1401 
1402 int
1403 ip6_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
1404     void *newp, size_t newlen)
1405 {
1406 #ifdef MROUTING
1407 	extern int ip6_mrtproto;
1408 	extern struct mrt6stat mrt6stat;
1409 #endif
1410 	int error;
1411 
1412 	/* Almost all sysctl names at this level are terminal. */
1413 	if (namelen != 1 && name[0] != IPV6CTL_IFQUEUE)
1414 		return (ENOTDIR);
1415 
1416 	switch (name[0]) {
1417 	case IPV6CTL_DAD_PENDING:
1418 		return sysctl_rdint(oldp, oldlenp, newp, ip6_dad_pending);
1419 	case IPV6CTL_STATS:
1420 		return (ip6_sysctl_ip6stat(oldp, oldlenp, newp));
1421 #ifdef MROUTING
1422 	case IPV6CTL_MRTSTATS:
1423 		if (newp != NULL)
1424 			return (EPERM);
1425 		NET_LOCK();
1426 		error = sysctl_struct(oldp, oldlenp, newp, newlen,
1427 		    &mrt6stat, sizeof(mrt6stat));
1428 		NET_UNLOCK();
1429 		return (error);
1430 	case IPV6CTL_MRTPROTO:
1431 		return sysctl_rdint(oldp, oldlenp, newp, ip6_mrtproto);
1432 	case IPV6CTL_MRTMIF:
1433 		if (newp)
1434 			return (EPERM);
1435 		NET_LOCK();
1436 		error = mrt6_sysctl_mif(oldp, oldlenp);
1437 		NET_UNLOCK();
1438 		return (error);
1439 	case IPV6CTL_MRTMFC:
1440 		if (newp)
1441 			return (EPERM);
1442 		NET_LOCK();
1443 		error = mrt6_sysctl_mfc(oldp, oldlenp);
1444 		NET_UNLOCK();
1445 		return (error);
1446 #else
1447 	case IPV6CTL_MRTSTATS:
1448 	case IPV6CTL_MRTPROTO:
1449 	case IPV6CTL_MRTMIF:
1450 	case IPV6CTL_MRTMFC:
1451 		return (EOPNOTSUPP);
1452 #endif
1453 	case IPV6CTL_MTUDISCTIMEOUT:
1454 		NET_LOCK();
1455 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1456 		    &ip6_mtudisc_timeout);
1457 		if (icmp6_mtudisc_timeout_q != NULL)
1458 			rt_timer_queue_change(icmp6_mtudisc_timeout_q,
1459 			    ip6_mtudisc_timeout);
1460 		NET_UNLOCK();
1461 		return (error);
1462 	case IPV6CTL_IFQUEUE:
1463 		return (sysctl_niq(name + 1, namelen - 1,
1464 		    oldp, oldlenp, newp, newlen, &ip6intrq));
1465 	case IPV6CTL_SOIIKEY:
1466 		return (ip6_sysctl_soiikey(oldp, oldlenp, newp, newlen));
1467 	default:
1468 		if (name[0] < IPV6CTL_MAXID) {
1469 			NET_LOCK();
1470 			error = sysctl_int_arr(ipv6ctl_vars, name, namelen,
1471 			    oldp, oldlenp, newp, newlen);
1472 			NET_UNLOCK();
1473 			return (error);
1474 		}
1475 		return (EOPNOTSUPP);
1476 	}
1477 	/* NOTREACHED */
1478 }
1479 
1480 void
1481 ip6_send_dispatch(void *xmq)
1482 {
1483 	struct mbuf_queue *mq = xmq;
1484 	struct mbuf *m;
1485 	struct mbuf_list ml;
1486 
1487 	mq_delist(mq, &ml);
1488 	if (ml_empty(&ml))
1489 		return;
1490 
1491 	NET_RLOCK();
1492 	while ((m = ml_dequeue(&ml)) != NULL) {
1493 		/*
1494 		 * To avoid a "too big" situation at an intermediate router and
1495 		 * the path MTU discovery process, specify the IPV6_MINMTU
1496 		 * flag.  Note that only echo and node information replies are
1497 		 * affected, since the length of ICMP6 errors is limited to the
1498 		 * minimum MTU.
1499 		 */
1500 		ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL);
1501 	}
1502 	NET_RUNLOCK();
1503 }
1504 
1505 void
1506 ip6_send(struct mbuf *m)
1507 {
1508 	mq_enqueue(&ip6send_mq, m);
1509 	task_add(net_tq(0), &ip6send_task);
1510 }
1511