xref: /openbsd-src/sys/netinet6/ip6_input.c (revision c1a45aed656e7d5627c30c92421893a76f370ccb)
1 /*	$OpenBSD: ip6_input.c,v 1.241 2022/04/20 09:38:26 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 cpumem *ip6counters;
117 
118 uint8_t ip6_soiikey[IP6_SOIIKEY_LEN];
119 
120 int ip6_ours(struct mbuf **, int *, int, int);
121 int ip6_check_rh0hdr(struct mbuf *, int *);
122 int ip6_hbhchcheck(struct mbuf *, int *, int *, int *);
123 int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
124 struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
125 int ip6_sysctl_soiikey(void *, size_t *, void *, size_t);
126 
127 static struct mbuf_queue	ip6send_mq;
128 
129 static void ip6_send_dispatch(void *);
130 static struct task ip6send_task =
131 	TASK_INITIALIZER(ip6_send_dispatch, &ip6send_mq);
132 
133 /*
134  * IP6 initialization: fill in IP6 protocol switch table.
135  * All protocols not implemented in kernel go to raw IP6 protocol handler.
136  */
137 void
138 ip6_init(void)
139 {
140 	const struct protosw *pr;
141 	int i;
142 
143 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
144 	if (pr == NULL)
145 		panic("%s", __func__);
146 	for (i = 0; i < IPPROTO_MAX; i++)
147 		ip6_protox[i] = pr - inet6sw;
148 	for (pr = inet6domain.dom_protosw;
149 	    pr < inet6domain.dom_protoswNPROTOSW; pr++)
150 		if (pr->pr_domain->dom_family == PF_INET6 &&
151 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW &&
152 		    pr->pr_protocol < IPPROTO_MAX)
153 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
154 	ip6_randomid_init();
155 	nd6_init();
156 	frag6_init();
157 
158 	mq_init(&ip6send_mq, 64, IPL_SOFTNET);
159 
160 	ip6counters = counters_alloc(ip6s_ncounters);
161 }
162 
163 void
164 ipv6_input(struct ifnet *ifp, struct mbuf *m)
165 {
166 	int off, nxt;
167 
168 	off = 0;
169 	nxt = ip6_input_if(&m, &off, IPPROTO_IPV6, AF_UNSPEC, ifp);
170 	KASSERT(nxt == IPPROTO_DONE);
171 }
172 
173 struct mbuf *
174 ipv6_check(struct ifnet *ifp, struct mbuf *m)
175 {
176 	struct ip6_hdr *ip6;
177 
178 	if (m->m_len < sizeof(*ip6)) {
179 		m = m_pullup(m, sizeof(*ip6));
180 		if (m == NULL) {
181 			ip6stat_inc(ip6s_toosmall);
182 			return (NULL);
183 		}
184 	}
185 
186 	ip6 = mtod(m, struct ip6_hdr *);
187 
188 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
189 		ip6stat_inc(ip6s_badvers);
190 		goto bad;
191 	}
192 
193 	/*
194 	 * Check against address spoofing/corruption.
195 	 */
196 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
197 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
198 		/*
199 		 * XXX: "badscope" is not very suitable for a multicast source.
200 		 */
201 		ip6stat_inc(ip6s_badscope);
202 		goto bad;
203 	}
204 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
205 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
206 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
207 		ip6stat_inc(ip6s_badscope);
208 		goto bad;
209 	}
210 	/* Drop packets if interface ID portion is already filled. */
211 	if (((IN6_IS_SCOPE_EMBED(&ip6->ip6_src) && ip6->ip6_src.s6_addr16[1]) ||
212 	    (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst) && ip6->ip6_dst.s6_addr16[1])) &&
213 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
214 		ip6stat_inc(ip6s_badscope);
215 		goto bad;
216 	}
217 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
218 	    !(m->m_flags & M_LOOP)) {
219 		/*
220 		 * In this case, the packet should come from the loopback
221 		 * interface.  However, we cannot just check the if_flags,
222 		 * because ip6_mloopback() passes the "actual" interface
223 		 * as the outgoing/incoming interface.
224 		 */
225 		ip6stat_inc(ip6s_badscope);
226 		goto bad;
227 	}
228 
229 	/*
230 	 * The following check is not documented in specs.  A malicious
231 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
232 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
233 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
234 	 *
235 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
236 	 * support IPv4-less kernel compilation, we cannot support SIIT
237 	 * environment at all.  So, it makes more sense for us to reject any
238 	 * malicious packets for non-SIIT environment, than try to do a
239 	 * partial support for SIIT environment.
240 	 */
241 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
242 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
243 		ip6stat_inc(ip6s_badscope);
244 		goto bad;
245 	}
246 
247 	/*
248 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
249 	 *
250 	 * The code forbids automatic tunneling as per RFC4213.
251 	 */
252 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
253 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
254 		ip6stat_inc(ip6s_badscope);
255 		goto bad;
256 	}
257 
258 	return (m);
259 bad:
260 	m_freem(m);
261 	return (NULL);
262 }
263 
264 int
265 ip6_input_if(struct mbuf **mp, int *offp, int nxt, int af, struct ifnet *ifp)
266 {
267 	struct mbuf *m;
268 	struct ip6_hdr *ip6;
269 	struct sockaddr_in6 sin6;
270 	struct rtentry *rt = NULL;
271 	int ours = 0;
272 	u_int16_t src_scope, dst_scope;
273 #if NPF > 0
274 	struct in6_addr odst;
275 #endif
276 	int srcrt = 0;
277 
278 	KASSERT(*offp == 0);
279 
280 	ip6stat_inc(ip6s_total);
281 
282 	m = *mp = ipv6_check(ifp, *mp);
283 	if (m == NULL)
284 		goto bad;
285 
286 	ip6 = mtod(m, struct ip6_hdr *);
287 
288 #if NCARP > 0
289 	if (carp_lsdrop(ifp, m, AF_INET6, ip6->ip6_src.s6_addr32,
290 	    ip6->ip6_dst.s6_addr32, (ip6->ip6_nxt == IPPROTO_ICMPV6 ? 0 : 1)))
291 		goto bad;
292 #endif
293 	ip6stat_inc(ip6s_nxthist + ip6->ip6_nxt);
294 
295 	/*
296 	 * If the packet has been received on a loopback interface it
297 	 * can be destined to any local address, not necessarily to
298 	 * an address configured on `ifp'.
299 	 */
300 	if (ifp->if_flags & IFF_LOOPBACK) {
301 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
302 			src_scope = ip6->ip6_src.s6_addr16[1];
303 			ip6->ip6_src.s6_addr16[1] = 0;
304 		}
305 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
306 			dst_scope = ip6->ip6_dst.s6_addr16[1];
307 			ip6->ip6_dst.s6_addr16[1] = 0;
308 		}
309 	}
310 
311 #if NPF > 0
312 	/*
313 	 * Packet filter
314 	 */
315 	odst = ip6->ip6_dst;
316 	if (pf_test(AF_INET6, PF_IN, ifp, mp) != PF_PASS)
317 		goto bad;
318 	m = *mp;
319 	if (m == NULL)
320 		goto bad;
321 
322 	ip6 = mtod(m, struct ip6_hdr *);
323 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
324 #endif
325 
326 	/*
327 	 * Without embedded scope ID we cannot find link-local
328 	 * addresses in the routing table.
329 	 */
330 	if (ifp->if_flags & IFF_LOOPBACK) {
331 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
332 			ip6->ip6_src.s6_addr16[1] = src_scope;
333 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
334 			ip6->ip6_dst.s6_addr16[1] = dst_scope;
335 	} else {
336 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
337 			ip6->ip6_src.s6_addr16[1] = htons(ifp->if_index);
338 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
339 			ip6->ip6_dst.s6_addr16[1] = htons(ifp->if_index);
340 	}
341 
342 	/*
343 	 * Be more secure than RFC5095 and scan for type 0 routing headers.
344 	 * If pf has already scanned the header chain, do not do it twice.
345 	 */
346 	if (!(m->m_pkthdr.pf.flags & PF_TAG_PROCESSED) &&
347 	    ip6_check_rh0hdr(m, offp)) {
348 		ip6stat_inc(ip6s_badoptions);
349 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, *offp);
350 		m = *mp = NULL;
351 		goto bad;
352 	}
353 
354 #if NPF > 0
355 	if (pf_ouraddr(m) == 1) {
356 		nxt = ip6_ours(mp, offp, nxt, af);
357 		goto out;
358 	}
359 #endif
360 
361 	/*
362 	 * Multicast check
363 	 */
364 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
365 		/*
366 		 * Make sure M_MCAST is set.  It should theoretically
367 		 * already be there, but let's play safe because upper
368 		 * layers check for this flag.
369 		 */
370 		m->m_flags |= M_MCAST;
371 
372 		/*
373 		 * See if we belong to the destination multicast group on the
374 		 * arrival interface.
375 		 */
376 		if (in6_hasmulti(&ip6->ip6_dst, ifp))
377 			ours = 1;
378 
379 #ifdef MROUTING
380 		if (ip6_mforwarding && ip6_mrouter[ifp->if_rdomain]) {
381 			int error;
382 
383 			if (ip6_hbhchcheck(m, offp, &nxt, &ours))
384 				goto out;
385 
386 			ip6 = mtod(m, struct ip6_hdr *);
387 
388 			/*
389 			 * If we are acting as a multicast router, all
390 			 * incoming multicast packets are passed to the
391 			 * kernel-level multicast forwarding function.
392 			 * The packet is returned (relatively) intact; if
393 			 * ip6_mforward() returns a non-zero value, the packet
394 			 * must be discarded, else it may be accepted below.
395 			 */
396 			KERNEL_LOCK();
397 			error = ip6_mforward(ip6, ifp, m);
398 			KERNEL_UNLOCK();
399 			if (error) {
400 				ip6stat_inc(ip6s_cantforward);
401 				goto bad;
402 			}
403 
404 			if (ours) {
405 				if (af == AF_UNSPEC)
406 					nxt = ip_deliver(mp, offp, nxt,
407 					    AF_INET6);
408 				goto out;
409 			}
410 			goto bad;
411 		}
412 #endif
413 		if (!ours) {
414 			ip6stat_inc(ip6s_notmember);
415 			if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
416 				ip6stat_inc(ip6s_cantforward);
417 			goto bad;
418 		}
419 		nxt = ip6_ours(mp, offp, nxt, af);
420 		goto out;
421 	}
422 
423 
424 	/*
425 	 *  Unicast check
426 	 */
427 	memset(&sin6, 0, sizeof(struct sockaddr_in6));
428 	sin6.sin6_len = sizeof(struct sockaddr_in6);
429 	sin6.sin6_family = AF_INET6;
430 	sin6.sin6_addr = ip6->ip6_dst;
431 	rt = rtalloc_mpath(sin6tosa(&sin6), &ip6->ip6_src.s6_addr32[0],
432 	    m->m_pkthdr.ph_rtableid);
433 
434 	/*
435 	 * Accept the packet if the route to the destination is marked
436 	 * as local.
437 	 */
438 	if (rtisvalid(rt) && ISSET(rt->rt_flags, RTF_LOCAL)) {
439 		struct in6_ifaddr *ia6 = ifatoia6(rt->rt_ifa);
440 
441 		if (ip6_forwarding == 0 && rt->rt_ifidx != ifp->if_index &&
442 		    !((ifp->if_flags & IFF_LOOPBACK) ||
443 		    (ifp->if_type == IFT_ENC) ||
444 		    (m->m_pkthdr.pf.flags & PF_TAG_TRANSLATE_LOCALHOST))) {
445 			/* received on wrong interface */
446 #if NCARP > 0
447 			struct ifnet *out_if;
448 
449 			/*
450 			 * Virtual IPs on carp interfaces need to be checked
451 			 * also against the parent interface and other carp
452 			 * interfaces sharing the same parent.
453 			 */
454 			out_if = if_get(rt->rt_ifidx);
455 			if (!(out_if && carp_strict_addr_chk(out_if, ifp))) {
456 				ip6stat_inc(ip6s_wrongif);
457 				if_put(out_if);
458 				goto bad;
459 			}
460 			if_put(out_if);
461 #else
462 			ip6stat_inc(ip6s_wrongif);
463 			goto bad;
464 #endif
465 		}
466 		/*
467 		 * packets to a tentative, duplicated, or somehow invalid
468 		 * address must not be accepted.
469 		 */
470 		if ((ia6->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED))) {
471 			char src[INET6_ADDRSTRLEN], dst[INET6_ADDRSTRLEN];
472 
473 			inet_ntop(AF_INET6, &ip6->ip6_src, src, sizeof(src));
474 			inet_ntop(AF_INET6, &ip6->ip6_dst, dst, sizeof(dst));
475 			/* address is not ready, so discard the packet. */
476 			nd6log((LOG_INFO,
477 			    "%s: packet to an unready address %s->%s\n",
478 			    __func__, src, dst));
479 
480 			goto bad;
481 		} else {
482 			nxt = ip6_ours(mp, offp, nxt, af);
483 			goto out;
484 		}
485 	}
486 
487 #if NCARP > 0
488 	if (ip6->ip6_nxt == IPPROTO_ICMPV6 &&
489 	    carp_lsdrop(ifp, m, AF_INET6, ip6->ip6_src.s6_addr32,
490 	    ip6->ip6_dst.s6_addr32, 1))
491 		goto bad;
492 #endif
493 	/*
494 	 * Now there is no reason to process the packet if it's not our own
495 	 * and we're not a router.
496 	 */
497 	if (!ip6_forwarding) {
498 		ip6stat_inc(ip6s_cantforward);
499 		goto bad;
500 	}
501 
502 	if (ip6_hbhchcheck(m, offp, &nxt, &ours))
503 		goto out;
504 
505 	if (ours) {
506 		if (af == AF_UNSPEC)
507 			nxt = ip_deliver(mp, offp, nxt, AF_INET6);
508 		goto out;
509 	}
510 
511 #ifdef IPSEC
512 	if (ipsec_in_use) {
513 		int rv;
514 
515 		rv = ipsec_forward_check(m, *offp, AF_INET6);
516 		if (rv != 0) {
517 			ip6stat_inc(ip6s_cantforward);
518 			goto bad;
519 		}
520 		/*
521 		 * Fall through, forward packet. Outbound IPsec policy
522 		 * checking will occur in ip6_forward().
523 		 */
524 	}
525 #endif /* IPSEC */
526 
527 	ip6_forward(m, rt, srcrt);
528 	*mp = NULL;
529 	return IPPROTO_DONE;
530  bad:
531 	nxt = IPPROTO_DONE;
532 	m_freemp(mp);
533  out:
534 	rtfree(rt);
535 	return nxt;
536 }
537 
538 int
539 ip6_ours(struct mbuf **mp, int *offp, int nxt, int af)
540 {
541 	if (ip6_hbhchcheck(*mp, offp, &nxt, NULL))
542 		return IPPROTO_DONE;
543 
544 	/* Check whether we are already in a IPv4/IPv6 local deliver loop. */
545 	if (af == AF_UNSPEC)
546 		nxt = ip_deliver(mp, offp, nxt, AF_INET6);
547 	return nxt;
548 }
549 
550 int
551 ip6_hbhchcheck(struct mbuf *m, int *offp, int *nxtp, int *oursp)
552 {
553 	struct ip6_hdr *ip6;
554 	u_int32_t plen, rtalert = ~0;
555 
556 	ip6 = mtod(m, struct ip6_hdr *);
557 
558 	/*
559 	 * Process Hop-by-Hop options header if it's contained.
560 	 * m may be modified in ip6_hopopts_input().
561 	 * If a JumboPayload option is included, plen will also be modified.
562 	 */
563 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
564 	*offp = sizeof(struct ip6_hdr);
565 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
566 		struct ip6_hbh *hbh;
567 
568 		if (ip6_hopopts_input(&plen, &rtalert, &m, offp)) {
569 			goto bad;	/* m have already been freed */
570 		}
571 
572 		/* adjust pointer */
573 		ip6 = mtod(m, struct ip6_hdr *);
574 
575 		/*
576 		 * if the payload length field is 0 and the next header field
577 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
578 		 * option MUST be included.
579 		 */
580 		if (ip6->ip6_plen == 0 && plen == 0) {
581 			/*
582 			 * Note that if a valid jumbo payload option is
583 			 * contained, ip6_hopopts_input() must set a valid
584 			 * (non-zero) payload length to the variable plen.
585 			 */
586 			ip6stat_inc(ip6s_badoptions);
587 			icmp6_error(m, ICMP6_PARAM_PROB,
588 				    ICMP6_PARAMPROB_HEADER,
589 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
590 			goto bad;
591 		}
592 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
593 			sizeof(struct ip6_hbh));
594 		if (hbh == NULL) {
595 			ip6stat_inc(ip6s_tooshort);
596 			goto bad;
597 		}
598 		*nxtp = hbh->ip6h_nxt;
599 
600 		/*
601 		 * accept the packet if a router alert option is included
602 		 * and we act as an IPv6 router.
603 		 */
604 		if (rtalert != ~0 && ip6_forwarding && oursp != NULL)
605 			*oursp = 1;
606 	} else
607 		*nxtp = ip6->ip6_nxt;
608 
609 	/*
610 	 * Check that the amount of data in the buffers
611 	 * is as at least much as the IPv6 header would have us expect.
612 	 * Trim mbufs if longer than we expect.
613 	 * Drop packet if shorter than we expect.
614 	 */
615 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
616 		ip6stat_inc(ip6s_tooshort);
617 		m_freem(m);
618 		goto bad;
619 	}
620 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
621 		if (m->m_len == m->m_pkthdr.len) {
622 			m->m_len = sizeof(struct ip6_hdr) + plen;
623 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
624 		} else {
625 			m_adj(m,
626 			    sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
627 		}
628 	}
629 
630 	return (0);
631 
632  bad:
633 	*nxtp = IPPROTO_DONE;
634 	return (-1);
635 }
636 
637 /* scan packet for RH0 routing header. Mostly stolen from pf.c:pf_test() */
638 int
639 ip6_check_rh0hdr(struct mbuf *m, int *offp)
640 {
641 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
642 	struct ip6_rthdr rthdr;
643 	struct ip6_ext opt6;
644 	u_int8_t proto = ip6->ip6_nxt;
645 	int done = 0, lim, off, rh_cnt = 0;
646 
647 	off = ((caddr_t)ip6 - m->m_data) + sizeof(struct ip6_hdr);
648 	lim = min(m->m_pkthdr.len, ntohs(ip6->ip6_plen) + sizeof(*ip6));
649 	do {
650 		switch (proto) {
651 		case IPPROTO_ROUTING:
652 			*offp = off;
653 			if (rh_cnt++) {
654 				/* more than one rh header present */
655 				return (1);
656 			}
657 
658 			if (off + sizeof(rthdr) > lim) {
659 				/* packet to short to make sense */
660 				return (1);
661 			}
662 
663 			m_copydata(m, off, sizeof(rthdr), &rthdr);
664 
665 			if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
666 				*offp += offsetof(struct ip6_rthdr, ip6r_type);
667 				return (1);
668 			}
669 
670 			off += (rthdr.ip6r_len + 1) * 8;
671 			proto = rthdr.ip6r_nxt;
672 			break;
673 		case IPPROTO_AH:
674 		case IPPROTO_HOPOPTS:
675 		case IPPROTO_DSTOPTS:
676 			/* get next header and header length */
677 			if (off + sizeof(opt6) > lim) {
678 				/*
679 				 * Packet to short to make sense, we could
680 				 * reject the packet but as a router we
681 				 * should not do that so forward it.
682 				 */
683 				return (0);
684 			}
685 
686 			m_copydata(m, off, sizeof(opt6), &opt6);
687 
688 			if (proto == IPPROTO_AH)
689 				off += (opt6.ip6e_len + 2) * 4;
690 			else
691 				off += (opt6.ip6e_len + 1) * 8;
692 			proto = opt6.ip6e_nxt;
693 			break;
694 		case IPPROTO_FRAGMENT:
695 		default:
696 			/* end of header stack */
697 			done = 1;
698 			break;
699 		}
700 	} while (!done);
701 
702 	return (0);
703 }
704 
705 /*
706  * Hop-by-Hop options header processing. If a valid jumbo payload option is
707  * included, the real payload length will be stored in plenp.
708  *
709  * rtalertp - XXX: should be stored in a more smart way
710  */
711 int
712 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, struct mbuf **mp,
713     int *offp)
714 {
715 	struct mbuf *m = *mp;
716 	int off = *offp, hbhlen;
717 	struct ip6_hbh *hbh;
718 
719 	/* validation of the length of the header */
720 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
721 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
722 	if (hbh == NULL) {
723 		ip6stat_inc(ip6s_tooshort);
724 		return -1;
725 	}
726 	hbhlen = (hbh->ip6h_len + 1) << 3;
727 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
728 		hbhlen);
729 	if (hbh == NULL) {
730 		ip6stat_inc(ip6s_tooshort);
731 		return -1;
732 	}
733 	off += hbhlen;
734 	hbhlen -= sizeof(struct ip6_hbh);
735 
736 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
737 				hbhlen, rtalertp, plenp) < 0)
738 		return (-1);
739 
740 	*offp = off;
741 	*mp = m;
742 	return (0);
743 }
744 
745 /*
746  * Search header for all Hop-by-hop options and process each option.
747  * This function is separate from ip6_hopopts_input() in order to
748  * handle a case where the sending node itself process its hop-by-hop
749  * options header. In such a case, the function is called from ip6_output().
750  *
751  * The function assumes that hbh header is located right after the IPv6 header
752  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
753  * opthead + hbhlen is located in continuous memory region.
754  */
755 int
756 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
757     u_int32_t *rtalertp, u_int32_t *plenp)
758 {
759 	struct ip6_hdr *ip6;
760 	int optlen = 0;
761 	u_int8_t *opt = opthead;
762 	u_int16_t rtalert_val;
763 	u_int32_t jumboplen;
764 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
765 
766 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
767 		switch (*opt) {
768 		case IP6OPT_PAD1:
769 			optlen = 1;
770 			break;
771 		case IP6OPT_PADN:
772 			if (hbhlen < IP6OPT_MINLEN) {
773 				ip6stat_inc(ip6s_toosmall);
774 				goto bad;
775 			}
776 			optlen = *(opt + 1) + 2;
777 			break;
778 		case IP6OPT_ROUTER_ALERT:
779 			/* XXX may need check for alignment */
780 			if (hbhlen < IP6OPT_RTALERT_LEN) {
781 				ip6stat_inc(ip6s_toosmall);
782 				goto bad;
783 			}
784 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
785 				/* XXX stat */
786 				icmp6_error(m, ICMP6_PARAM_PROB,
787 				    ICMP6_PARAMPROB_HEADER,
788 				    erroff + opt + 1 - opthead);
789 				return (-1);
790 			}
791 			optlen = IP6OPT_RTALERT_LEN;
792 			memcpy((caddr_t)&rtalert_val, (caddr_t)(opt + 2), 2);
793 			*rtalertp = ntohs(rtalert_val);
794 			break;
795 		case IP6OPT_JUMBO:
796 			/* XXX may need check for alignment */
797 			if (hbhlen < IP6OPT_JUMBO_LEN) {
798 				ip6stat_inc(ip6s_toosmall);
799 				goto bad;
800 			}
801 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
802 				/* XXX stat */
803 				icmp6_error(m, ICMP6_PARAM_PROB,
804 				    ICMP6_PARAMPROB_HEADER,
805 				    erroff + opt + 1 - opthead);
806 				return (-1);
807 			}
808 			optlen = IP6OPT_JUMBO_LEN;
809 
810 			/*
811 			 * IPv6 packets that have non 0 payload length
812 			 * must not contain a jumbo payload option.
813 			 */
814 			ip6 = mtod(m, struct ip6_hdr *);
815 			if (ip6->ip6_plen) {
816 				ip6stat_inc(ip6s_badoptions);
817 				icmp6_error(m, ICMP6_PARAM_PROB,
818 				    ICMP6_PARAMPROB_HEADER,
819 				    erroff + opt - opthead);
820 				return (-1);
821 			}
822 
823 			/*
824 			 * We may see jumbolen in unaligned location, so
825 			 * we'd need to perform memcpy().
826 			 */
827 			memcpy(&jumboplen, opt + 2, sizeof(jumboplen));
828 			jumboplen = (u_int32_t)htonl(jumboplen);
829 
830 #if 1
831 			/*
832 			 * if there are multiple jumbo payload options,
833 			 * *plenp will be non-zero and the packet will be
834 			 * rejected.
835 			 * the behavior may need some debate in ipngwg -
836 			 * multiple options does not make sense, however,
837 			 * there's no explicit mention in specification.
838 			 */
839 			if (*plenp != 0) {
840 				ip6stat_inc(ip6s_badoptions);
841 				icmp6_error(m, ICMP6_PARAM_PROB,
842 				    ICMP6_PARAMPROB_HEADER,
843 				    erroff + opt + 2 - opthead);
844 				return (-1);
845 			}
846 #endif
847 
848 			/*
849 			 * jumbo payload length must be larger than 65535.
850 			 */
851 			if (jumboplen <= IPV6_MAXPACKET) {
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 			*plenp = jumboplen;
859 
860 			break;
861 		default:		/* unknown option */
862 			if (hbhlen < IP6OPT_MINLEN) {
863 				ip6stat_inc(ip6s_toosmall);
864 				goto bad;
865 			}
866 			optlen = ip6_unknown_opt(opt, m,
867 			    erroff + opt - opthead);
868 			if (optlen == -1)
869 				return (-1);
870 			optlen += 2;
871 			break;
872 		}
873 	}
874 
875 	return (0);
876 
877   bad:
878 	m_freem(m);
879 	return (-1);
880 }
881 
882 /*
883  * Unknown option processing.
884  * The third argument `off' is the offset from the IPv6 header to the option,
885  * which allows returning an ICMPv6 error even if the IPv6 header and the
886  * option header are not continuous.
887  */
888 int
889 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
890 {
891 	struct ip6_hdr *ip6;
892 
893 	switch (IP6OPT_TYPE(*optp)) {
894 	case IP6OPT_TYPE_SKIP: /* ignore the option */
895 		return ((int)*(optp + 1));
896 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
897 		m_freem(m);
898 		return (-1);
899 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
900 		ip6stat_inc(ip6s_badoptions);
901 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
902 		return (-1);
903 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
904 		ip6stat_inc(ip6s_badoptions);
905 		ip6 = mtod(m, struct ip6_hdr *);
906 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
907 		    (m->m_flags & (M_BCAST|M_MCAST)))
908 			m_freem(m);
909 		else
910 			icmp6_error(m, ICMP6_PARAM_PROB,
911 				    ICMP6_PARAMPROB_OPTION, off);
912 		return (-1);
913 	}
914 
915 	m_freem(m);		/* XXX: NOTREACHED */
916 	return (-1);
917 }
918 
919 /*
920  * Create the "control" list for this pcb.
921  *
922  * The routine will be called from upper layer handlers like udp_input().
923  * Thus the routine assumes that the caller (udp_input) have already
924  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
925  * very first mbuf on the mbuf chain.
926  * We may want to add some infinite loop prevention or sanity checks for safety.
927  * (This applies only when you are using KAME mbuf chain restriction, i.e.
928  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
929  */
930 void
931 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
932 {
933 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
934 
935 	if (in6p->inp_socket->so_options & SO_TIMESTAMP) {
936 		struct timeval tv;
937 
938 		m_microtime(m, &tv);
939 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
940 		    SCM_TIMESTAMP, SOL_SOCKET);
941 		if (*mp)
942 			mp = &(*mp)->m_next;
943 	}
944 
945 	/* RFC 2292 sec. 5 */
946 	if ((in6p->inp_flags & IN6P_PKTINFO) != 0) {
947 		struct in6_pktinfo pi6;
948 		memcpy(&pi6.ipi6_addr, &ip6->ip6_dst, sizeof(struct in6_addr));
949 		if (IN6_IS_SCOPE_EMBED(&pi6.ipi6_addr))
950 			pi6.ipi6_addr.s6_addr16[1] = 0;
951 		pi6.ipi6_ifindex = m ? m->m_pkthdr.ph_ifidx : 0;
952 		*mp = sbcreatecontrol((caddr_t) &pi6,
953 		    sizeof(struct in6_pktinfo),
954 		    IPV6_PKTINFO, IPPROTO_IPV6);
955 		if (*mp)
956 			mp = &(*mp)->m_next;
957 	}
958 
959 	if ((in6p->inp_flags & IN6P_HOPLIMIT) != 0) {
960 		int hlim = ip6->ip6_hlim & 0xff;
961 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
962 		    IPV6_HOPLIMIT, IPPROTO_IPV6);
963 		if (*mp)
964 			mp = &(*mp)->m_next;
965 	}
966 
967 	if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
968 		u_int32_t flowinfo;
969 		int tclass;
970 
971 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
972 		flowinfo >>= 20;
973 
974 		tclass = flowinfo & 0xff;
975 		*mp = sbcreatecontrol((caddr_t)&tclass, sizeof(tclass),
976 		    IPV6_TCLASS, IPPROTO_IPV6);
977 		if (*mp)
978 			mp = &(*mp)->m_next;
979 	}
980 
981 	/*
982 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
983 	 * privilege for the option (see ip6_ctloutput), but it might be too
984 	 * strict, since there might be some hop-by-hop options which can be
985 	 * returned to normal user.
986 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
987 	 */
988 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
989 		/*
990 		 * Check if a hop-by-hop options header is contained in the
991 		 * received packet, and if so, store the options as ancillary
992 		 * data. Note that a hop-by-hop options header must be
993 		 * just after the IPv6 header, which is assured through the
994 		 * IPv6 input processing.
995 		 */
996 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
997 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
998 			struct ip6_hbh *hbh;
999 			int hbhlen = 0;
1000 			struct mbuf *ext;
1001 
1002 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1003 			    ip6->ip6_nxt);
1004 			if (ext == NULL) {
1005 				ip6stat_inc(ip6s_tooshort);
1006 				return;
1007 			}
1008 			hbh = mtod(ext, struct ip6_hbh *);
1009 			hbhlen = (hbh->ip6h_len + 1) << 3;
1010 			if (hbhlen != ext->m_len) {
1011 				m_freem(ext);
1012 				ip6stat_inc(ip6s_tooshort);
1013 				return;
1014 			}
1015 
1016 			/*
1017 			 * XXX: We copy the whole header even if a
1018 			 * jumbo payload option is included, the option which
1019 			 * is to be removed before returning according to
1020 			 * RFC2292.
1021 			 * Note: this constraint is removed in RFC3542.
1022 			 */
1023 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1024 			    IPV6_HOPOPTS,
1025 			    IPPROTO_IPV6);
1026 			if (*mp)
1027 				mp = &(*mp)->m_next;
1028 			m_freem(ext);
1029 		}
1030 	}
1031 
1032 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1033 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1034 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1035 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1036 
1037 		/*
1038 		 * Search for destination options headers or routing
1039 		 * header(s) through the header chain, and stores each
1040 		 * header as ancillary data.
1041 		 * Note that the order of the headers remains in
1042 		 * the chain of ancillary data.
1043 		 */
1044 		while (1) {	/* is explicit loop prevention necessary? */
1045 			struct ip6_ext *ip6e = NULL;
1046 			int elen;
1047 			struct mbuf *ext = NULL;
1048 
1049 			/*
1050 			 * if it is not an extension header, don't try to
1051 			 * pull it from the chain.
1052 			 */
1053 			switch (nxt) {
1054 			case IPPROTO_DSTOPTS:
1055 			case IPPROTO_ROUTING:
1056 			case IPPROTO_HOPOPTS:
1057 			case IPPROTO_AH: /* is it possible? */
1058 				break;
1059 			default:
1060 				goto loopend;
1061 			}
1062 
1063 			ext = ip6_pullexthdr(m, off, nxt);
1064 			if (ext == NULL) {
1065 				ip6stat_inc(ip6s_tooshort);
1066 				return;
1067 			}
1068 			ip6e = mtod(ext, struct ip6_ext *);
1069 			if (nxt == IPPROTO_AH)
1070 				elen = (ip6e->ip6e_len + 2) << 2;
1071 			else
1072 				elen = (ip6e->ip6e_len + 1) << 3;
1073 			if (elen != ext->m_len) {
1074 				m_freem(ext);
1075 				ip6stat_inc(ip6s_tooshort);
1076 				return;
1077 			}
1078 
1079 			switch (nxt) {
1080 			case IPPROTO_DSTOPTS:
1081 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1082 					break;
1083 
1084 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1085 				    IPV6_DSTOPTS,
1086 				    IPPROTO_IPV6);
1087 				if (*mp)
1088 					mp = &(*mp)->m_next;
1089 				break;
1090 
1091 			case IPPROTO_ROUTING:
1092 				if (!(in6p->inp_flags & IN6P_RTHDR))
1093 					break;
1094 
1095 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1096 				    IPV6_RTHDR,
1097 				    IPPROTO_IPV6);
1098 				if (*mp)
1099 					mp = &(*mp)->m_next;
1100 				break;
1101 
1102 			case IPPROTO_HOPOPTS:
1103 			case IPPROTO_AH: /* is it possible? */
1104 				break;
1105 
1106 			default:
1107 				/*
1108 				 * other cases have been filtered in the above.
1109 				 * none will visit this case.  here we supply
1110 				 * the code just in case (nxt overwritten or
1111 				 * other cases).
1112 				 */
1113 				m_freem(ext);
1114 				goto loopend;
1115 
1116 			}
1117 
1118 			/* proceed with the next header. */
1119 			off += elen;
1120 			nxt = ip6e->ip6e_nxt;
1121 			ip6e = NULL;
1122 			m_freem(ext);
1123 			ext = NULL;
1124 		}
1125 loopend:
1126 		;
1127 	}
1128 }
1129 
1130 /*
1131  * pull single extension header from mbuf chain.  returns single mbuf that
1132  * contains the result, or NULL on error.
1133  */
1134 struct mbuf *
1135 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1136 {
1137 	struct ip6_ext ip6e;
1138 	size_t elen;
1139 	struct mbuf *n;
1140 
1141 #ifdef DIAGNOSTIC
1142 	switch (nxt) {
1143 	case IPPROTO_DSTOPTS:
1144 	case IPPROTO_ROUTING:
1145 	case IPPROTO_HOPOPTS:
1146 	case IPPROTO_AH: /* is it possible? */
1147 		break;
1148 	default:
1149 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1150 	}
1151 #endif
1152 
1153 	if (off + sizeof(ip6e) > m->m_pkthdr.len)
1154 		return NULL;
1155 
1156 	m_copydata(m, off, sizeof(ip6e), &ip6e);
1157 	if (nxt == IPPROTO_AH)
1158 		elen = (ip6e.ip6e_len + 2) << 2;
1159 	else
1160 		elen = (ip6e.ip6e_len + 1) << 3;
1161 
1162 	if (off + elen > m->m_pkthdr.len)
1163 		return NULL;
1164 
1165 	MGET(n, M_DONTWAIT, MT_DATA);
1166 	if (n && elen >= MLEN) {
1167 		MCLGET(n, M_DONTWAIT);
1168 		if ((n->m_flags & M_EXT) == 0) {
1169 			m_free(n);
1170 			n = NULL;
1171 		}
1172 	}
1173 	if (!n)
1174 		return NULL;
1175 
1176 	n->m_len = 0;
1177 	if (elen >= m_trailingspace(n)) {
1178 		m_free(n);
1179 		return NULL;
1180 	}
1181 
1182 	m_copydata(m, off, elen, mtod(n, caddr_t));
1183 	n->m_len = elen;
1184 	return n;
1185 }
1186 
1187 /*
1188  * Get offset to the previous header followed by the header
1189  * currently processed.
1190  */
1191 int
1192 ip6_get_prevhdr(struct mbuf *m, int off)
1193 {
1194 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1195 
1196 	if (off == sizeof(struct ip6_hdr)) {
1197 		return offsetof(struct ip6_hdr, ip6_nxt);
1198 	} else if (off < sizeof(struct ip6_hdr)) {
1199 		panic("%s: off < sizeof(struct ip6_hdr)", __func__);
1200 	} else {
1201 		int len, nlen, nxt;
1202 		struct ip6_ext ip6e;
1203 
1204 		nxt = ip6->ip6_nxt;
1205 		len = sizeof(struct ip6_hdr);
1206 		nlen = 0;
1207 		while (len < off) {
1208 			m_copydata(m, len, sizeof(ip6e), &ip6e);
1209 
1210 			switch (nxt) {
1211 			case IPPROTO_FRAGMENT:
1212 				nlen = sizeof(struct ip6_frag);
1213 				break;
1214 			case IPPROTO_AH:
1215 				nlen = (ip6e.ip6e_len + 2) << 2;
1216 				break;
1217 			default:
1218 				nlen = (ip6e.ip6e_len + 1) << 3;
1219 				break;
1220 			}
1221 			len += nlen;
1222 			nxt = ip6e.ip6e_nxt;
1223 		}
1224 
1225 		return (len - nlen);
1226 	}
1227 }
1228 
1229 /*
1230  * get next header offset.  m will be retained.
1231  */
1232 int
1233 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1234 {
1235 	struct ip6_hdr ip6;
1236 	struct ip6_ext ip6e;
1237 	struct ip6_frag fh;
1238 
1239 	/* just in case */
1240 	if (m == NULL)
1241 		panic("%s: m == NULL", __func__);
1242 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1243 		return -1;
1244 
1245 	switch (proto) {
1246 	case IPPROTO_IPV6:
1247 		if (m->m_pkthdr.len < off + sizeof(ip6))
1248 			return -1;
1249 		m_copydata(m, off, sizeof(ip6), &ip6);
1250 		if (nxtp)
1251 			*nxtp = ip6.ip6_nxt;
1252 		off += sizeof(ip6);
1253 		return off;
1254 
1255 	case IPPROTO_FRAGMENT:
1256 		/*
1257 		 * terminate parsing if it is not the first fragment,
1258 		 * it does not make sense to parse through it.
1259 		 */
1260 		if (m->m_pkthdr.len < off + sizeof(fh))
1261 			return -1;
1262 		m_copydata(m, off, sizeof(fh), &fh);
1263 		if ((fh.ip6f_offlg & IP6F_OFF_MASK) != 0)
1264 			return -1;
1265 		if (nxtp)
1266 			*nxtp = fh.ip6f_nxt;
1267 		off += sizeof(struct ip6_frag);
1268 		return off;
1269 
1270 	case IPPROTO_AH:
1271 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1272 			return -1;
1273 		m_copydata(m, off, sizeof(ip6e), &ip6e);
1274 		if (nxtp)
1275 			*nxtp = ip6e.ip6e_nxt;
1276 		off += (ip6e.ip6e_len + 2) << 2;
1277 		if (m->m_pkthdr.len < off)
1278 			return -1;
1279 		return off;
1280 
1281 	case IPPROTO_HOPOPTS:
1282 	case IPPROTO_ROUTING:
1283 	case IPPROTO_DSTOPTS:
1284 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1285 			return -1;
1286 		m_copydata(m, off, sizeof(ip6e), &ip6e);
1287 		if (nxtp)
1288 			*nxtp = ip6e.ip6e_nxt;
1289 		off += (ip6e.ip6e_len + 1) << 3;
1290 		if (m->m_pkthdr.len < off)
1291 			return -1;
1292 		return off;
1293 
1294 	case IPPROTO_NONE:
1295 	case IPPROTO_ESP:
1296 	case IPPROTO_IPCOMP:
1297 		/* give up */
1298 		return -1;
1299 
1300 	default:
1301 		return -1;
1302 	}
1303 
1304 	return -1;
1305 }
1306 
1307 /*
1308  * get offset for the last header in the chain.  m will be kept untainted.
1309  */
1310 int
1311 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1312 {
1313 	int newoff;
1314 	int nxt;
1315 
1316 	if (!nxtp) {
1317 		nxt = -1;
1318 		nxtp = &nxt;
1319 	}
1320 	while (1) {
1321 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1322 		if (newoff < 0)
1323 			return off;
1324 		else if (newoff < off)
1325 			return -1;	/* invalid */
1326 		else if (newoff == off)
1327 			return newoff;
1328 
1329 		off = newoff;
1330 		proto = *nxtp;
1331 	}
1332 }
1333 
1334 /*
1335  * System control for IP6
1336  */
1337 
1338 const u_char inet6ctlerrmap[PRC_NCMDS] = {
1339 	0,		0,		0,		0,
1340 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1341 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1342 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1343 	0,		0,		0,		0,
1344 	ENOPROTOOPT
1345 };
1346 
1347 #ifdef MROUTING
1348 extern int ip6_mrtproto;
1349 #endif
1350 
1351 const struct sysctl_bounded_args ipv6ctl_vars[] = {
1352 	{ IPV6CTL_DAD_PENDING, &ip6_dad_pending, SYSCTL_INT_READONLY },
1353 #ifdef MROUTING
1354 	{ IPV6CTL_MRTPROTO, &ip6_mrtproto, SYSCTL_INT_READONLY },
1355 #endif
1356 	{ IPV6CTL_FORWARDING, &ip6_forwarding, 0, 1 },
1357 	{ IPV6CTL_SENDREDIRECTS, &ip6_sendredirects, 0, 1 },
1358 	{ IPV6CTL_DEFHLIM, &ip6_defhlim, 0, 255 },
1359 	{ IPV6CTL_MAXFRAGPACKETS, &ip6_maxfragpackets, 0, 1000 },
1360 	{ IPV6CTL_LOG_INTERVAL, &ip6_log_interval, 0, INT_MAX },
1361 	{ IPV6CTL_HDRNESTLIMIT, &ip6_hdrnestlimit, 0, 100 },
1362 	{ IPV6CTL_DAD_COUNT, &ip6_dad_count, 0, 10 },
1363 	{ IPV6CTL_AUTO_FLOWLABEL, &ip6_auto_flowlabel, 0, 1 },
1364 	{ IPV6CTL_DEFMCASTHLIM, &ip6_defmcasthlim, 0, 255 },
1365 	{ IPV6CTL_USE_DEPRECATED, &ip6_use_deprecated, 0, 1 },
1366 	{ IPV6CTL_MAXFRAGS, &ip6_maxfrags, 0, 1000 },
1367 	{ IPV6CTL_MFORWARDING, &ip6_mforwarding, 0, 1 },
1368 	{ IPV6CTL_MULTIPATH, &ip6_multipath, 0, 1 },
1369 	{ IPV6CTL_MCAST_PMTU, &ip6_mcast_pmtu, 0, 1 },
1370 	{ IPV6CTL_NEIGHBORGCTHRESH, &ip6_neighborgcthresh, -1, 5 * 2048 },
1371 	{ IPV6CTL_MAXDYNROUTES, &ip6_maxdynroutes, -1, 5 * 4096 },
1372 };
1373 
1374 int
1375 ip6_sysctl_ip6stat(void *oldp, size_t *oldlenp, void *newp)
1376 {
1377 	struct ip6stat *ip6stat;
1378 	int ret;
1379 
1380 	CTASSERT(sizeof(*ip6stat) == (ip6s_ncounters * sizeof(uint64_t)));
1381 
1382 	ip6stat = malloc(sizeof(*ip6stat), M_TEMP, M_WAITOK);
1383 	counters_read(ip6counters, (uint64_t *)ip6stat, ip6s_ncounters);
1384 	ret = sysctl_rdstruct(oldp, oldlenp, newp,
1385 	    ip6stat, sizeof(*ip6stat));
1386 	free(ip6stat, M_TEMP, sizeof(*ip6stat));
1387 
1388 	return (ret);
1389 }
1390 
1391 int
1392 ip6_sysctl_soiikey(void *oldp, size_t *oldlenp, void *newp, size_t newlen)
1393 {
1394 	uint8_t oldkey[IP6_SOIIKEY_LEN];
1395 	int error;
1396 
1397 	error = suser(curproc);
1398 	if (error != 0)
1399 		return (error);
1400 
1401 	memcpy(oldkey, ip6_soiikey, sizeof(oldkey));
1402 
1403 	error = sysctl_struct(oldp, oldlenp, newp, newlen, ip6_soiikey,
1404 	    sizeof(ip6_soiikey));
1405 
1406 	return (error);
1407 }
1408 
1409 int
1410 ip6_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
1411     void *newp, size_t newlen)
1412 {
1413 #ifdef MROUTING
1414 	extern struct mrt6stat mrt6stat;
1415 #endif
1416 	int error;
1417 
1418 	/* Almost all sysctl names at this level are terminal. */
1419 	if (namelen != 1 && name[0] != IPV6CTL_IFQUEUE)
1420 		return (ENOTDIR);
1421 
1422 	switch (name[0]) {
1423 	case IPV6CTL_STATS:
1424 		return (ip6_sysctl_ip6stat(oldp, oldlenp, newp));
1425 #ifdef MROUTING
1426 	case IPV6CTL_MRTSTATS:
1427 		if (newp != NULL)
1428 			return (EPERM);
1429 		NET_LOCK();
1430 		error = sysctl_struct(oldp, oldlenp, newp, newlen,
1431 		    &mrt6stat, sizeof(mrt6stat));
1432 		NET_UNLOCK();
1433 		return (error);
1434 	case IPV6CTL_MRTMIF:
1435 		if (newp)
1436 			return (EPERM);
1437 		NET_LOCK();
1438 		error = mrt6_sysctl_mif(oldp, oldlenp);
1439 		NET_UNLOCK();
1440 		return (error);
1441 	case IPV6CTL_MRTMFC:
1442 		if (newp)
1443 			return (EPERM);
1444 		NET_LOCK();
1445 		error = mrt6_sysctl_mfc(oldp, oldlenp);
1446 		NET_UNLOCK();
1447 		return (error);
1448 #else
1449 	case IPV6CTL_MRTSTATS:
1450 	case IPV6CTL_MRTPROTO:
1451 	case IPV6CTL_MRTMIF:
1452 	case IPV6CTL_MRTMFC:
1453 		return (EOPNOTSUPP);
1454 #endif
1455 	case IPV6CTL_MTUDISCTIMEOUT:
1456 		NET_LOCK();
1457 		error = sysctl_int_bounded(oldp, oldlenp, newp, newlen,
1458 		    &ip6_mtudisc_timeout, 0, INT_MAX);
1459 		rt_timer_queue_change(icmp6_mtudisc_timeout_q,
1460 		    ip6_mtudisc_timeout);
1461 		NET_UNLOCK();
1462 		return (error);
1463 	case IPV6CTL_IFQUEUE:
1464 		return (EOPNOTSUPP);
1465 	case IPV6CTL_SOIIKEY:
1466 		return (ip6_sysctl_soiikey(oldp, oldlenp, newp, newlen));
1467 	default:
1468 		NET_LOCK();
1469 		error = sysctl_bounded_arr(ipv6ctl_vars, nitems(ipv6ctl_vars),
1470 		    name, namelen, oldp, oldlenp, newp, newlen);
1471 		NET_UNLOCK();
1472 		return (error);
1473 	}
1474 	/* NOTREACHED */
1475 }
1476 
1477 void
1478 ip6_send_dispatch(void *xmq)
1479 {
1480 	struct mbuf_queue *mq = xmq;
1481 	struct mbuf *m;
1482 	struct mbuf_list ml;
1483 
1484 	mq_delist(mq, &ml);
1485 	if (ml_empty(&ml))
1486 		return;
1487 
1488 	NET_LOCK();
1489 	while ((m = ml_dequeue(&ml)) != NULL) {
1490 		ip6_output(m, NULL, NULL, 0, NULL, NULL);
1491 	}
1492 	NET_UNLOCK();
1493 }
1494 
1495 void
1496 ip6_send(struct mbuf *m)
1497 {
1498 	mq_enqueue(&ip6send_mq, m);
1499 	task_add(net_tq(0), &ip6send_task);
1500 }
1501