xref: /openbsd/sys/netinet6/ip6_input.c (revision 55cc5ba3)
1 /*	$OpenBSD: ip6_input.c,v 1.230 2020/11/16 06:44:39 gnezdo 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 cpumem *ip6counters;
119 
120 uint8_t ip6_soiikey[IP6_SOIIKEY_LEN];
121 
122 int ip6_ours(struct mbuf **, int *, int, int);
123 int ip6_check_rh0hdr(struct mbuf *, int *);
124 int ip6_hbhchcheck(struct mbuf *, int *, int *, int *);
125 int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
126 struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
127 int ip6_sysctl_soiikey(void *, size_t *, void *, size_t);
128 
129 static struct mbuf_queue	ip6send_mq;
130 
131 static void ip6_send_dispatch(void *);
132 static struct task ip6send_task =
133 	TASK_INITIALIZER(ip6_send_dispatch, &ip6send_mq);
134 
135 /*
136  * IP6 initialization: fill in IP6 protocol switch table.
137  * All protocols not implemented in kernel go to raw IP6 protocol handler.
138  */
139 void
140 ip6_init(void)
141 {
142 	const struct protosw *pr;
143 	int i;
144 
145 	pr = pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
146 	if (pr == NULL)
147 		panic("%s", __func__);
148 	for (i = 0; i < IPPROTO_MAX; i++)
149 		ip6_protox[i] = pr - inet6sw;
150 	for (pr = inet6domain.dom_protosw;
151 	    pr < inet6domain.dom_protoswNPROTOSW; pr++)
152 		if (pr->pr_domain->dom_family == PF_INET6 &&
153 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW &&
154 		    pr->pr_protocol < IPPROTO_MAX)
155 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
156 	ip6_randomid_init();
157 	nd6_init();
158 	frag6_init();
159 
160 	mq_init(&ip6send_mq, 64, IPL_SOFTNET);
161 
162 	ip6counters = counters_alloc(ip6s_ncounters);
163 }
164 
165 void
166 ipv6_input(struct ifnet *ifp, struct mbuf *m)
167 {
168 	int off, nxt;
169 
170 	off = 0;
171 	nxt = ip6_input_if(&m, &off, IPPROTO_IPV6, AF_UNSPEC, ifp);
172 	KASSERT(nxt == IPPROTO_DONE);
173 }
174 
175 int
176 ip6_input_if(struct mbuf **mp, int *offp, int nxt, int af, struct ifnet *ifp)
177 {
178 	struct mbuf *m = *mp;
179 	struct ip6_hdr *ip6;
180 	struct sockaddr_in6 sin6;
181 	struct rtentry *rt = NULL;
182 	int ours = 0;
183 	u_int16_t src_scope, dst_scope;
184 #if NPF > 0
185 	struct in6_addr odst;
186 #endif
187 	int srcrt = 0;
188 
189 	KASSERT(*offp == 0);
190 
191 	ip6stat_inc(ip6s_total);
192 
193 	if (m->m_len < sizeof(struct ip6_hdr)) {
194 		if ((m = *mp = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
195 			ip6stat_inc(ip6s_toosmall);
196 			goto bad;
197 		}
198 	}
199 
200 	ip6 = mtod(m, struct ip6_hdr *);
201 
202 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
203 		ip6stat_inc(ip6s_badvers);
204 		goto bad;
205 	}
206 
207 #if NCARP > 0
208 	if (carp_lsdrop(ifp, m, AF_INET6, ip6->ip6_src.s6_addr32,
209 	    ip6->ip6_dst.s6_addr32, (ip6->ip6_nxt == IPPROTO_ICMPV6 ? 0 : 1)))
210 		goto bad;
211 #endif
212 	ip6stat_inc(ip6s_nxthist + ip6->ip6_nxt);
213 
214 	/*
215 	 * Check against address spoofing/corruption.
216 	 */
217 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
218 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
219 		/*
220 		 * XXX: "badscope" is not very suitable for a multicast source.
221 		 */
222 		ip6stat_inc(ip6s_badscope);
223 		goto bad;
224 	}
225 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
226 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
227 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
228 		    ip6stat_inc(ip6s_badscope);
229 		    goto bad;
230 	}
231 	/* Drop packets if interface ID portion is already filled. */
232 	if (((IN6_IS_SCOPE_EMBED(&ip6->ip6_src) && ip6->ip6_src.s6_addr16[1]) ||
233 	    (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst) && ip6->ip6_dst.s6_addr16[1])) &&
234 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
235 		ip6stat_inc(ip6s_badscope);
236 		goto bad;
237 	}
238 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
239 	    !(m->m_flags & M_LOOP)) {
240 		/*
241 		 * In this case, the packet should come from the loopback
242 		 * interface.  However, we cannot just check the if_flags,
243 		 * because ip6_mloopback() passes the "actual" interface
244 		 * as the outgoing/incoming interface.
245 		 */
246 		ip6stat_inc(ip6s_badscope);
247 		goto bad;
248 	}
249 
250 	/*
251 	 * The following check is not documented in specs.  A malicious
252 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
253 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
254 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
255 	 *
256 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
257 	 * support IPv4-less kernel compilation, we cannot support SIIT
258 	 * environment at all.  So, it makes more sense for us to reject any
259 	 * malicious packets for non-SIIT environment, than try to do a
260 	 * partial support for SIIT environment.
261 	 */
262 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
263 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
264 		ip6stat_inc(ip6s_badscope);
265 		goto bad;
266 	}
267 
268 	/*
269 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
270 	 *
271 	 * The code forbids automatic tunneling as per RFC4213.
272 	 */
273 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
274 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
275 		ip6stat_inc(ip6s_badscope);
276 		goto bad;
277 	}
278 
279 	/*
280 	 * If the packet has been received on a loopback interface it
281 	 * can be destinated to any local address, not necessarily to
282 	 * an address configured on `ifp'.
283 	 */
284 	if (ifp->if_flags & IFF_LOOPBACK) {
285 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
286 			src_scope = ip6->ip6_src.s6_addr16[1];
287 			ip6->ip6_src.s6_addr16[1] = 0;
288 		}
289 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
290 			dst_scope = ip6->ip6_dst.s6_addr16[1];
291 			ip6->ip6_dst.s6_addr16[1] = 0;
292 		}
293 	}
294 
295 #if NPF > 0
296 	/*
297 	 * Packet filter
298 	 */
299 	odst = ip6->ip6_dst;
300 	if (pf_test(AF_INET6, PF_IN, ifp, mp) != PF_PASS)
301 		goto bad;
302 	m = *mp;
303 	if (m == NULL)
304 		goto bad;
305 
306 	ip6 = mtod(m, struct ip6_hdr *);
307 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
308 #endif
309 
310 	/*
311 	 * Without embedded scope ID we cannot find link-local
312 	 * addresses in the routing table.
313 	 */
314 	if (ifp->if_flags & IFF_LOOPBACK) {
315 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
316 			ip6->ip6_src.s6_addr16[1] = src_scope;
317 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
318 			ip6->ip6_dst.s6_addr16[1] = dst_scope;
319 	} else {
320 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
321 			ip6->ip6_src.s6_addr16[1] = htons(ifp->if_index);
322 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
323 			ip6->ip6_dst.s6_addr16[1] = htons(ifp->if_index);
324 	}
325 
326 	/*
327 	 * Be more secure than RFC5095 and scan for type 0 routing headers.
328 	 * If pf has already scanned the header chain, do not do it twice.
329 	 */
330 	if (!(m->m_pkthdr.pf.flags & PF_TAG_PROCESSED) &&
331 	    ip6_check_rh0hdr(m, offp)) {
332 		ip6stat_inc(ip6s_badoptions);
333 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, *offp);
334 		m = *mp = NULL;
335 		goto bad;
336 	}
337 
338 #if NPF > 0
339 	if (pf_ouraddr(m) == 1) {
340 		nxt = ip6_ours(mp, offp, nxt, af);
341 		goto out;
342 	}
343 #endif
344 
345 	/*
346 	 * Multicast check
347 	 */
348 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
349 		/*
350 		 * Make sure M_MCAST is set.  It should theoretically
351 		 * already be there, but let's play safe because upper
352 		 * layers check for this flag.
353 		 */
354 		m->m_flags |= M_MCAST;
355 
356 		/*
357 		 * See if we belong to the destination multicast group on the
358 		 * arrival interface.
359 		 */
360 		if (in6_hasmulti(&ip6->ip6_dst, ifp))
361 			ours = 1;
362 
363 #ifdef MROUTING
364 		if (ip6_mforwarding && ip6_mrouter[ifp->if_rdomain]) {
365 			int error;
366 
367 			if (ip6_hbhchcheck(m, offp, &nxt, &ours))
368 				goto out;
369 
370 			ip6 = mtod(m, struct ip6_hdr *);
371 
372 			/*
373 			 * If we are acting as a multicast router, all
374 			 * incoming multicast packets are passed to the
375 			 * kernel-level multicast forwarding function.
376 			 * The packet is returned (relatively) intact; if
377 			 * ip6_mforward() returns a non-zero value, the packet
378 			 * must be discarded, else it may be accepted below.
379 			 */
380 			KERNEL_LOCK();
381 			error = ip6_mforward(ip6, ifp, m);
382 			KERNEL_UNLOCK();
383 			if (error) {
384 				ip6stat_inc(ip6s_cantforward);
385 				goto bad;
386 			}
387 
388 			if (ours) {
389 				if (af == AF_UNSPEC) {
390 					KERNEL_LOCK();
391 					nxt = ip_deliver(mp, offp, nxt,
392 					    AF_INET6);
393 					KERNEL_UNLOCK();
394 				}
395 				goto out;
396 			}
397 			goto bad;
398 		}
399 #endif
400 		if (!ours) {
401 			ip6stat_inc(ip6s_notmember);
402 			if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
403 				ip6stat_inc(ip6s_cantforward);
404 			goto bad;
405 		}
406 		nxt = ip6_ours(mp, offp, nxt, af);
407 		goto out;
408 	}
409 
410 
411 	/*
412 	 *  Unicast check
413 	 */
414 	memset(&sin6, 0, sizeof(struct sockaddr_in6));
415 	sin6.sin6_len = sizeof(struct sockaddr_in6);
416 	sin6.sin6_family = AF_INET6;
417 	sin6.sin6_addr = ip6->ip6_dst;
418 	rt = rtalloc_mpath(sin6tosa(&sin6), &ip6->ip6_src.s6_addr32[0],
419 	    m->m_pkthdr.ph_rtableid);
420 
421 	/*
422 	 * Accept the packet if the route to the destination is marked
423 	 * as local.
424 	 */
425 	if (rtisvalid(rt) && ISSET(rt->rt_flags, RTF_LOCAL)) {
426 		struct in6_ifaddr *ia6 = ifatoia6(rt->rt_ifa);
427 
428 		if (ip6_forwarding == 0 && rt->rt_ifidx != ifp->if_index &&
429 		    !((ifp->if_flags & IFF_LOOPBACK) ||
430 		    (ifp->if_type == IFT_ENC) ||
431 		    (m->m_pkthdr.pf.flags & PF_TAG_TRANSLATE_LOCALHOST))) {
432 			/* received on wrong interface */
433 #if NCARP > 0
434 			struct ifnet *out_if;
435 
436 			/*
437 			 * Virtual IPs on carp interfaces need to be checked
438 			 * also against the parent interface and other carp
439 			 * interfaces sharing the same parent.
440 			 */
441 			out_if = if_get(rt->rt_ifidx);
442 			if (!(out_if && carp_strict_addr_chk(out_if, ifp))) {
443 				ip6stat_inc(ip6s_wrongif);
444 				if_put(out_if);
445 				goto bad;
446 			}
447 			if_put(out_if);
448 #else
449 			ip6stat_inc(ip6s_wrongif);
450 			goto bad;
451 #endif
452 		}
453 		/*
454 		 * packets to a tentative, duplicated, or somehow invalid
455 		 * address must not be accepted.
456 		 */
457 		if ((ia6->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED))) {
458 			char src[INET6_ADDRSTRLEN], dst[INET6_ADDRSTRLEN];
459 
460 			inet_ntop(AF_INET6, &ip6->ip6_src, src, sizeof(src));
461 			inet_ntop(AF_INET6, &ip6->ip6_dst, dst, sizeof(dst));
462 			/* address is not ready, so discard the packet. */
463 			nd6log((LOG_INFO,
464 			    "%s: packet to an unready address %s->%s\n",
465 			    __func__, src, dst));
466 
467 			goto bad;
468 		} else {
469 			nxt = ip6_ours(mp, offp, nxt, af);
470 			goto out;
471 		}
472 	}
473 
474 #if NCARP > 0
475 	if (ip6->ip6_nxt == IPPROTO_ICMPV6 &&
476 	    carp_lsdrop(ifp, m, AF_INET6, ip6->ip6_src.s6_addr32,
477 	    ip6->ip6_dst.s6_addr32, 1))
478 		goto bad;
479 #endif
480 	/*
481 	 * Now there is no reason to process the packet if it's not our own
482 	 * and we're not a router.
483 	 */
484 	if (!ip6_forwarding) {
485 		ip6stat_inc(ip6s_cantforward);
486 		goto bad;
487 	}
488 
489 	if (ip6_hbhchcheck(m, offp, &nxt, &ours))
490 		goto out;
491 
492 	if (ours) {
493 		if (af == AF_UNSPEC) {
494 			KERNEL_LOCK();
495 			nxt = ip_deliver(mp, offp, nxt, AF_INET6);
496 			KERNEL_UNLOCK();
497 		}
498 		goto out;
499 	}
500 
501 #ifdef IPSEC
502 	if (ipsec_in_use) {
503 		int rv;
504 
505 		rv = ipsec_forward_check(m, *offp, AF_INET6);
506 		if (rv != 0) {
507 			ip6stat_inc(ip6s_cantforward);
508 			goto bad;
509 		}
510 		/*
511 		 * Fall through, forward packet. Outbound IPsec policy
512 		 * checking will occur in ip6_forward().
513 		 */
514 	}
515 #endif /* IPSEC */
516 
517 	ip6_forward(m, rt, srcrt);
518 	*mp = NULL;
519 	return IPPROTO_DONE;
520  bad:
521 	nxt = IPPROTO_DONE;
522 	m_freemp(mp);
523  out:
524 	rtfree(rt);
525 	return nxt;
526 }
527 
528 int
529 ip6_ours(struct mbuf **mp, int *offp, int nxt, int af)
530 {
531 	if (ip6_hbhchcheck(*mp, offp, &nxt, NULL))
532 		return IPPROTO_DONE;
533 
534 	/* Check wheter we are already in a IPv4/IPv6 local deliver loop. */
535 	if (af == AF_UNSPEC)
536 		nxt = ip_deliver(mp, offp, nxt, AF_INET6);
537 	return nxt;
538 }
539 
540 int
541 ip6_hbhchcheck(struct mbuf *m, int *offp, int *nxtp, int *oursp)
542 {
543 	struct ip6_hdr *ip6;
544 	u_int32_t plen, rtalert = ~0;
545 
546 	ip6 = mtod(m, struct ip6_hdr *);
547 
548 	/*
549 	 * Process Hop-by-Hop options header if it's contained.
550 	 * m may be modified in ip6_hopopts_input().
551 	 * If a JumboPayload option is included, plen will also be modified.
552 	 */
553 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
554 	*offp = sizeof(struct ip6_hdr);
555 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
556 		struct ip6_hbh *hbh;
557 
558 		if (ip6_hopopts_input(&plen, &rtalert, &m, offp)) {
559 			goto bad;	/* m have already been freed */
560 		}
561 
562 		/* adjust pointer */
563 		ip6 = mtod(m, struct ip6_hdr *);
564 
565 		/*
566 		 * if the payload length field is 0 and the next header field
567 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
568 		 * option MUST be included.
569 		 */
570 		if (ip6->ip6_plen == 0 && plen == 0) {
571 			/*
572 			 * Note that if a valid jumbo payload option is
573 			 * contained, ip6_hopopts_input() must set a valid
574 			 * (non-zero) payload length to the variable plen.
575 			 */
576 			ip6stat_inc(ip6s_badoptions);
577 			icmp6_error(m, ICMP6_PARAM_PROB,
578 				    ICMP6_PARAMPROB_HEADER,
579 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
580 			goto bad;
581 		}
582 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
583 			sizeof(struct ip6_hbh));
584 		if (hbh == NULL) {
585 			ip6stat_inc(ip6s_tooshort);
586 			goto bad;
587 		}
588 		*nxtp = hbh->ip6h_nxt;
589 
590 		/*
591 		 * accept the packet if a router alert option is included
592 		 * and we act as an IPv6 router.
593 		 */
594 		if (rtalert != ~0 && ip6_forwarding && oursp != NULL)
595 			*oursp = 1;
596 	} else
597 		*nxtp = ip6->ip6_nxt;
598 
599 	/*
600 	 * Check that the amount of data in the buffers
601 	 * is as at least much as the IPv6 header would have us expect.
602 	 * Trim mbufs if longer than we expect.
603 	 * Drop packet if shorter than we expect.
604 	 */
605 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
606 		ip6stat_inc(ip6s_tooshort);
607 		m_freem(m);
608 		goto bad;
609 	}
610 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
611 		if (m->m_len == m->m_pkthdr.len) {
612 			m->m_len = sizeof(struct ip6_hdr) + plen;
613 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
614 		} else {
615 			m_adj(m,
616 			    sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
617 		}
618 	}
619 
620 	return (0);
621 
622  bad:
623 	*nxtp = IPPROTO_DONE;
624 	return (-1);
625 }
626 
627 /* scan packet for RH0 routing header. Mostly stolen from pf.c:pf_test() */
628 int
629 ip6_check_rh0hdr(struct mbuf *m, int *offp)
630 {
631 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
632 	struct ip6_rthdr rthdr;
633 	struct ip6_ext opt6;
634 	u_int8_t proto = ip6->ip6_nxt;
635 	int done = 0, lim, off, rh_cnt = 0;
636 
637 	off = ((caddr_t)ip6 - m->m_data) + sizeof(struct ip6_hdr);
638 	lim = min(m->m_pkthdr.len, ntohs(ip6->ip6_plen) + sizeof(*ip6));
639 	do {
640 		switch (proto) {
641 		case IPPROTO_ROUTING:
642 			*offp = off;
643 			if (rh_cnt++) {
644 				/* more than one rh header present */
645 				return (1);
646 			}
647 
648 			if (off + sizeof(rthdr) > lim) {
649 				/* packet to short to make sense */
650 				return (1);
651 			}
652 
653 			m_copydata(m, off, sizeof(rthdr), (caddr_t)&rthdr);
654 
655 			if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
656 				*offp += offsetof(struct ip6_rthdr, ip6r_type);
657 				return (1);
658 			}
659 
660 			off += (rthdr.ip6r_len + 1) * 8;
661 			proto = rthdr.ip6r_nxt;
662 			break;
663 		case IPPROTO_AH:
664 		case IPPROTO_HOPOPTS:
665 		case IPPROTO_DSTOPTS:
666 			/* get next header and header length */
667 			if (off + sizeof(opt6) > lim) {
668 				/*
669 				 * Packet to short to make sense, we could
670 				 * reject the packet but as a router we
671 				 * should not do that so forward it.
672 				 */
673 				return (0);
674 			}
675 
676 			m_copydata(m, off, sizeof(opt6), (caddr_t)&opt6);
677 
678 			if (proto == IPPROTO_AH)
679 				off += (opt6.ip6e_len + 2) * 4;
680 			else
681 				off += (opt6.ip6e_len + 1) * 8;
682 			proto = opt6.ip6e_nxt;
683 			break;
684 		case IPPROTO_FRAGMENT:
685 		default:
686 			/* end of header stack */
687 			done = 1;
688 			break;
689 		}
690 	} while (!done);
691 
692 	return (0);
693 }
694 
695 /*
696  * Hop-by-Hop options header processing. If a valid jumbo payload option is
697  * included, the real payload length will be stored in plenp.
698  *
699  * rtalertp - XXX: should be stored in a more smart way
700  */
701 int
702 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, struct mbuf **mp,
703     int *offp)
704 {
705 	struct mbuf *m = *mp;
706 	int off = *offp, hbhlen;
707 	struct ip6_hbh *hbh;
708 
709 	/* validation of the length of the header */
710 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
711 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
712 	if (hbh == NULL) {
713 		ip6stat_inc(ip6s_tooshort);
714 		return -1;
715 	}
716 	hbhlen = (hbh->ip6h_len + 1) << 3;
717 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
718 		hbhlen);
719 	if (hbh == NULL) {
720 		ip6stat_inc(ip6s_tooshort);
721 		return -1;
722 	}
723 	off += hbhlen;
724 	hbhlen -= sizeof(struct ip6_hbh);
725 
726 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
727 				hbhlen, rtalertp, plenp) < 0)
728 		return (-1);
729 
730 	*offp = off;
731 	*mp = m;
732 	return (0);
733 }
734 
735 /*
736  * Search header for all Hop-by-hop options and process each option.
737  * This function is separate from ip6_hopopts_input() in order to
738  * handle a case where the sending node itself process its hop-by-hop
739  * options header. In such a case, the function is called from ip6_output().
740  *
741  * The function assumes that hbh header is located right after the IPv6 header
742  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
743  * opthead + hbhlen is located in continuous memory region.
744  */
745 int
746 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
747     u_int32_t *rtalertp, u_int32_t *plenp)
748 {
749 	struct ip6_hdr *ip6;
750 	int optlen = 0;
751 	u_int8_t *opt = opthead;
752 	u_int16_t rtalert_val;
753 	u_int32_t jumboplen;
754 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
755 
756 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
757 		switch (*opt) {
758 		case IP6OPT_PAD1:
759 			optlen = 1;
760 			break;
761 		case IP6OPT_PADN:
762 			if (hbhlen < IP6OPT_MINLEN) {
763 				ip6stat_inc(ip6s_toosmall);
764 				goto bad;
765 			}
766 			optlen = *(opt + 1) + 2;
767 			break;
768 		case IP6OPT_ROUTER_ALERT:
769 			/* XXX may need check for alignment */
770 			if (hbhlen < IP6OPT_RTALERT_LEN) {
771 				ip6stat_inc(ip6s_toosmall);
772 				goto bad;
773 			}
774 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
775 				/* XXX stat */
776 				icmp6_error(m, ICMP6_PARAM_PROB,
777 				    ICMP6_PARAMPROB_HEADER,
778 				    erroff + opt + 1 - opthead);
779 				return (-1);
780 			}
781 			optlen = IP6OPT_RTALERT_LEN;
782 			memcpy((caddr_t)&rtalert_val, (caddr_t)(opt + 2), 2);
783 			*rtalertp = ntohs(rtalert_val);
784 			break;
785 		case IP6OPT_JUMBO:
786 			/* XXX may need check for alignment */
787 			if (hbhlen < IP6OPT_JUMBO_LEN) {
788 				ip6stat_inc(ip6s_toosmall);
789 				goto bad;
790 			}
791 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
792 				/* XXX stat */
793 				icmp6_error(m, ICMP6_PARAM_PROB,
794 				    ICMP6_PARAMPROB_HEADER,
795 				    erroff + opt + 1 - opthead);
796 				return (-1);
797 			}
798 			optlen = IP6OPT_JUMBO_LEN;
799 
800 			/*
801 			 * IPv6 packets that have non 0 payload length
802 			 * must not contain a jumbo payload option.
803 			 */
804 			ip6 = mtod(m, struct ip6_hdr *);
805 			if (ip6->ip6_plen) {
806 				ip6stat_inc(ip6s_badoptions);
807 				icmp6_error(m, ICMP6_PARAM_PROB,
808 				    ICMP6_PARAMPROB_HEADER,
809 				    erroff + opt - opthead);
810 				return (-1);
811 			}
812 
813 			/*
814 			 * We may see jumbolen in unaligned location, so
815 			 * we'd need to perform memcpy().
816 			 */
817 			memcpy(&jumboplen, opt + 2, sizeof(jumboplen));
818 			jumboplen = (u_int32_t)htonl(jumboplen);
819 
820 #if 1
821 			/*
822 			 * if there are multiple jumbo payload options,
823 			 * *plenp will be non-zero and the packet will be
824 			 * rejected.
825 			 * the behavior may need some debate in ipngwg -
826 			 * multiple options does not make sense, however,
827 			 * there's no explicit mention in specification.
828 			 */
829 			if (*plenp != 0) {
830 				ip6stat_inc(ip6s_badoptions);
831 				icmp6_error(m, ICMP6_PARAM_PROB,
832 				    ICMP6_PARAMPROB_HEADER,
833 				    erroff + opt + 2 - opthead);
834 				return (-1);
835 			}
836 #endif
837 
838 			/*
839 			 * jumbo payload length must be larger than 65535.
840 			 */
841 			if (jumboplen <= IPV6_MAXPACKET) {
842 				ip6stat_inc(ip6s_badoptions);
843 				icmp6_error(m, ICMP6_PARAM_PROB,
844 				    ICMP6_PARAMPROB_HEADER,
845 				    erroff + opt + 2 - opthead);
846 				return (-1);
847 			}
848 			*plenp = jumboplen;
849 
850 			break;
851 		default:		/* unknown option */
852 			if (hbhlen < IP6OPT_MINLEN) {
853 				ip6stat_inc(ip6s_toosmall);
854 				goto bad;
855 			}
856 			optlen = ip6_unknown_opt(opt, m,
857 			    erroff + opt - opthead);
858 			if (optlen == -1)
859 				return (-1);
860 			optlen += 2;
861 			break;
862 		}
863 	}
864 
865 	return (0);
866 
867   bad:
868 	m_freem(m);
869 	return (-1);
870 }
871 
872 /*
873  * Unknown option processing.
874  * The third argument `off' is the offset from the IPv6 header to the option,
875  * which allows returning an ICMPv6 error even if the IPv6 header and the
876  * option header are not continuous.
877  */
878 int
879 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
880 {
881 	struct ip6_hdr *ip6;
882 
883 	switch (IP6OPT_TYPE(*optp)) {
884 	case IP6OPT_TYPE_SKIP: /* ignore the option */
885 		return ((int)*(optp + 1));
886 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
887 		m_freem(m);
888 		return (-1);
889 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
890 		ip6stat_inc(ip6s_badoptions);
891 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
892 		return (-1);
893 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
894 		ip6stat_inc(ip6s_badoptions);
895 		ip6 = mtod(m, struct ip6_hdr *);
896 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
897 		    (m->m_flags & (M_BCAST|M_MCAST)))
898 			m_freem(m);
899 		else
900 			icmp6_error(m, ICMP6_PARAM_PROB,
901 				    ICMP6_PARAMPROB_OPTION, off);
902 		return (-1);
903 	}
904 
905 	m_freem(m);		/* XXX: NOTREACHED */
906 	return (-1);
907 }
908 
909 /*
910  * Create the "control" list for this pcb.
911  *
912  * The routine will be called from upper layer handlers like udp_input().
913  * Thus the routine assumes that the caller (udp_input) have already
914  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
915  * very first mbuf on the mbuf chain.
916  * We may want to add some infinite loop prevention or sanity checks for safety.
917  * (This applies only when you are using KAME mbuf chain restriction, i.e.
918  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
919  */
920 void
921 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
922 {
923 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
924 
925 	if (in6p->inp_socket->so_options & SO_TIMESTAMP) {
926 		struct timeval tv;
927 
928 		m_microtime(m, &tv);
929 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
930 		    SCM_TIMESTAMP, SOL_SOCKET);
931 		if (*mp)
932 			mp = &(*mp)->m_next;
933 	}
934 
935 	/* RFC 2292 sec. 5 */
936 	if ((in6p->inp_flags & IN6P_PKTINFO) != 0) {
937 		struct in6_pktinfo pi6;
938 		memcpy(&pi6.ipi6_addr, &ip6->ip6_dst, sizeof(struct in6_addr));
939 		if (IN6_IS_SCOPE_EMBED(&pi6.ipi6_addr))
940 			pi6.ipi6_addr.s6_addr16[1] = 0;
941 		pi6.ipi6_ifindex = m ? m->m_pkthdr.ph_ifidx : 0;
942 		*mp = sbcreatecontrol((caddr_t) &pi6,
943 		    sizeof(struct in6_pktinfo),
944 		    IPV6_PKTINFO, IPPROTO_IPV6);
945 		if (*mp)
946 			mp = &(*mp)->m_next;
947 	}
948 
949 	if ((in6p->inp_flags & IN6P_HOPLIMIT) != 0) {
950 		int hlim = ip6->ip6_hlim & 0xff;
951 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
952 		    IPV6_HOPLIMIT, IPPROTO_IPV6);
953 		if (*mp)
954 			mp = &(*mp)->m_next;
955 	}
956 
957 	if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
958 		u_int32_t flowinfo;
959 		int tclass;
960 
961 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
962 		flowinfo >>= 20;
963 
964 		tclass = flowinfo & 0xff;
965 		*mp = sbcreatecontrol((caddr_t)&tclass, sizeof(tclass),
966 		    IPV6_TCLASS, IPPROTO_IPV6);
967 		if (*mp)
968 			mp = &(*mp)->m_next;
969 	}
970 
971 	/*
972 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
973 	 * privilege for the option (see ip6_ctloutput), but it might be too
974 	 * strict, since there might be some hop-by-hop options which can be
975 	 * returned to normal user.
976 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
977 	 */
978 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
979 		/*
980 		 * Check if a hop-by-hop options header is contained in the
981 		 * received packet, and if so, store the options as ancillary
982 		 * data. Note that a hop-by-hop options header must be
983 		 * just after the IPv6 header, which is assured through the
984 		 * IPv6 input processing.
985 		 */
986 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
987 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
988 			struct ip6_hbh *hbh;
989 			int hbhlen = 0;
990 			struct mbuf *ext;
991 
992 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
993 			    ip6->ip6_nxt);
994 			if (ext == NULL) {
995 				ip6stat_inc(ip6s_tooshort);
996 				return;
997 			}
998 			hbh = mtod(ext, struct ip6_hbh *);
999 			hbhlen = (hbh->ip6h_len + 1) << 3;
1000 			if (hbhlen != ext->m_len) {
1001 				m_freem(ext);
1002 				ip6stat_inc(ip6s_tooshort);
1003 				return;
1004 			}
1005 
1006 			/*
1007 			 * XXX: We copy the whole header even if a
1008 			 * jumbo payload option is included, the option which
1009 			 * is to be removed before returning according to
1010 			 * RFC2292.
1011 			 * Note: this constraint is removed in RFC3542.
1012 			 */
1013 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1014 			    IPV6_HOPOPTS,
1015 			    IPPROTO_IPV6);
1016 			if (*mp)
1017 				mp = &(*mp)->m_next;
1018 			m_freem(ext);
1019 		}
1020 	}
1021 
1022 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1023 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1024 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1025 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1026 
1027 		/*
1028 		 * Search for destination options headers or routing
1029 		 * header(s) through the header chain, and stores each
1030 		 * header as ancillary data.
1031 		 * Note that the order of the headers remains in
1032 		 * the chain of ancillary data.
1033 		 */
1034 		while (1) {	/* is explicit loop prevention necessary? */
1035 			struct ip6_ext *ip6e = NULL;
1036 			int elen;
1037 			struct mbuf *ext = NULL;
1038 
1039 			/*
1040 			 * if it is not an extension header, don't try to
1041 			 * pull it from the chain.
1042 			 */
1043 			switch (nxt) {
1044 			case IPPROTO_DSTOPTS:
1045 			case IPPROTO_ROUTING:
1046 			case IPPROTO_HOPOPTS:
1047 			case IPPROTO_AH: /* is it possible? */
1048 				break;
1049 			default:
1050 				goto loopend;
1051 			}
1052 
1053 			ext = ip6_pullexthdr(m, off, nxt);
1054 			if (ext == NULL) {
1055 				ip6stat_inc(ip6s_tooshort);
1056 				return;
1057 			}
1058 			ip6e = mtod(ext, struct ip6_ext *);
1059 			if (nxt == IPPROTO_AH)
1060 				elen = (ip6e->ip6e_len + 2) << 2;
1061 			else
1062 				elen = (ip6e->ip6e_len + 1) << 3;
1063 			if (elen != ext->m_len) {
1064 				m_freem(ext);
1065 				ip6stat_inc(ip6s_tooshort);
1066 				return;
1067 			}
1068 
1069 			switch (nxt) {
1070 			case IPPROTO_DSTOPTS:
1071 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1072 					break;
1073 
1074 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1075 				    IPV6_DSTOPTS,
1076 				    IPPROTO_IPV6);
1077 				if (*mp)
1078 					mp = &(*mp)->m_next;
1079 				break;
1080 
1081 			case IPPROTO_ROUTING:
1082 				if (!(in6p->inp_flags & IN6P_RTHDR))
1083 					break;
1084 
1085 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1086 				    IPV6_RTHDR,
1087 				    IPPROTO_IPV6);
1088 				if (*mp)
1089 					mp = &(*mp)->m_next;
1090 				break;
1091 
1092 			case IPPROTO_HOPOPTS:
1093 			case IPPROTO_AH: /* is it possible? */
1094 				break;
1095 
1096 			default:
1097 				/*
1098 				 * other cases have been filtered in the above.
1099 				 * none will visit this case.  here we supply
1100 				 * the code just in case (nxt overwritten or
1101 				 * other cases).
1102 				 */
1103 				m_freem(ext);
1104 				goto loopend;
1105 
1106 			}
1107 
1108 			/* proceed with the next header. */
1109 			off += elen;
1110 			nxt = ip6e->ip6e_nxt;
1111 			ip6e = NULL;
1112 			m_freem(ext);
1113 			ext = NULL;
1114 		}
1115 loopend:
1116 		;
1117 	}
1118 }
1119 
1120 /*
1121  * pull single extension header from mbuf chain.  returns single mbuf that
1122  * contains the result, or NULL on error.
1123  */
1124 struct mbuf *
1125 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1126 {
1127 	struct ip6_ext ip6e;
1128 	size_t elen;
1129 	struct mbuf *n;
1130 
1131 #ifdef DIAGNOSTIC
1132 	switch (nxt) {
1133 	case IPPROTO_DSTOPTS:
1134 	case IPPROTO_ROUTING:
1135 	case IPPROTO_HOPOPTS:
1136 	case IPPROTO_AH: /* is it possible? */
1137 		break;
1138 	default:
1139 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1140 	}
1141 #endif
1142 
1143 	if (off + sizeof(ip6e) > m->m_pkthdr.len)
1144 		return NULL;
1145 
1146 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1147 	if (nxt == IPPROTO_AH)
1148 		elen = (ip6e.ip6e_len + 2) << 2;
1149 	else
1150 		elen = (ip6e.ip6e_len + 1) << 3;
1151 
1152 	if (off + elen > m->m_pkthdr.len)
1153 		return NULL;
1154 
1155 	MGET(n, M_DONTWAIT, MT_DATA);
1156 	if (n && elen >= MLEN) {
1157 		MCLGET(n, M_DONTWAIT);
1158 		if ((n->m_flags & M_EXT) == 0) {
1159 			m_free(n);
1160 			n = NULL;
1161 		}
1162 	}
1163 	if (!n)
1164 		return NULL;
1165 
1166 	n->m_len = 0;
1167 	if (elen >= m_trailingspace(n)) {
1168 		m_free(n);
1169 		return NULL;
1170 	}
1171 
1172 	m_copydata(m, off, elen, mtod(n, caddr_t));
1173 	n->m_len = elen;
1174 	return n;
1175 }
1176 
1177 /*
1178  * Get offset to the previous header followed by the header
1179  * currently processed.
1180  */
1181 int
1182 ip6_get_prevhdr(struct mbuf *m, int off)
1183 {
1184 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1185 
1186 	if (off == sizeof(struct ip6_hdr)) {
1187 		return offsetof(struct ip6_hdr, ip6_nxt);
1188 	} else if (off < sizeof(struct ip6_hdr)) {
1189 		panic("%s: off < sizeof(struct ip6_hdr)", __func__);
1190 	} else {
1191 		int len, nlen, nxt;
1192 		struct ip6_ext ip6e;
1193 
1194 		nxt = ip6->ip6_nxt;
1195 		len = sizeof(struct ip6_hdr);
1196 		nlen = 0;
1197 		while (len < off) {
1198 			m_copydata(m, len, sizeof(ip6e), (caddr_t)&ip6e);
1199 
1200 			switch (nxt) {
1201 			case IPPROTO_FRAGMENT:
1202 				nlen = sizeof(struct ip6_frag);
1203 				break;
1204 			case IPPROTO_AH:
1205 				nlen = (ip6e.ip6e_len + 2) << 2;
1206 				break;
1207 			default:
1208 				nlen = (ip6e.ip6e_len + 1) << 3;
1209 				break;
1210 			}
1211 			len += nlen;
1212 			nxt = ip6e.ip6e_nxt;
1213 		}
1214 
1215 		return (len - nlen);
1216 	}
1217 }
1218 
1219 /*
1220  * get next header offset.  m will be retained.
1221  */
1222 int
1223 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1224 {
1225 	struct ip6_hdr ip6;
1226 	struct ip6_ext ip6e;
1227 	struct ip6_frag fh;
1228 
1229 	/* just in case */
1230 	if (m == NULL)
1231 		panic("%s: m == NULL", __func__);
1232 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1233 		return -1;
1234 
1235 	switch (proto) {
1236 	case IPPROTO_IPV6:
1237 		if (m->m_pkthdr.len < off + sizeof(ip6))
1238 			return -1;
1239 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1240 		if (nxtp)
1241 			*nxtp = ip6.ip6_nxt;
1242 		off += sizeof(ip6);
1243 		return off;
1244 
1245 	case IPPROTO_FRAGMENT:
1246 		/*
1247 		 * terminate parsing if it is not the first fragment,
1248 		 * it does not make sense to parse through it.
1249 		 */
1250 		if (m->m_pkthdr.len < off + sizeof(fh))
1251 			return -1;
1252 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1253 		if ((fh.ip6f_offlg & IP6F_OFF_MASK) != 0)
1254 			return -1;
1255 		if (nxtp)
1256 			*nxtp = fh.ip6f_nxt;
1257 		off += sizeof(struct ip6_frag);
1258 		return off;
1259 
1260 	case IPPROTO_AH:
1261 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1262 			return -1;
1263 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1264 		if (nxtp)
1265 			*nxtp = ip6e.ip6e_nxt;
1266 		off += (ip6e.ip6e_len + 2) << 2;
1267 		if (m->m_pkthdr.len < off)
1268 			return -1;
1269 		return off;
1270 
1271 	case IPPROTO_HOPOPTS:
1272 	case IPPROTO_ROUTING:
1273 	case IPPROTO_DSTOPTS:
1274 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1275 			return -1;
1276 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1277 		if (nxtp)
1278 			*nxtp = ip6e.ip6e_nxt;
1279 		off += (ip6e.ip6e_len + 1) << 3;
1280 		if (m->m_pkthdr.len < off)
1281 			return -1;
1282 		return off;
1283 
1284 	case IPPROTO_NONE:
1285 	case IPPROTO_ESP:
1286 	case IPPROTO_IPCOMP:
1287 		/* give up */
1288 		return -1;
1289 
1290 	default:
1291 		return -1;
1292 	}
1293 
1294 	return -1;
1295 }
1296 
1297 /*
1298  * get offset for the last header in the chain.  m will be kept untainted.
1299  */
1300 int
1301 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1302 {
1303 	int newoff;
1304 	int nxt;
1305 
1306 	if (!nxtp) {
1307 		nxt = -1;
1308 		nxtp = &nxt;
1309 	}
1310 	while (1) {
1311 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1312 		if (newoff < 0)
1313 			return off;
1314 		else if (newoff < off)
1315 			return -1;	/* invalid */
1316 		else if (newoff == off)
1317 			return newoff;
1318 
1319 		off = newoff;
1320 		proto = *nxtp;
1321 	}
1322 }
1323 
1324 /*
1325  * System control for IP6
1326  */
1327 
1328 const u_char inet6ctlerrmap[PRC_NCMDS] = {
1329 	0,		0,		0,		0,
1330 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1331 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1332 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1333 	0,		0,		0,		0,
1334 	ENOPROTOOPT
1335 };
1336 
1337 #ifdef MROUTING
1338 extern int ip6_mrtproto;
1339 #endif
1340 
1341 const struct sysctl_bounded_args ipv6ctl_vars[] = {
1342 	{ IPV6CTL_DAD_PENDING, &ip6_dad_pending, 1, 0 },
1343 #ifdef MROUTING
1344 	{ IPV6CTL_MRTPROTO, &ip6_mrtproto, 1, 0 },
1345 #endif
1346 	{ IPV6CTL_FORWARDING, &ip6_forwarding, 0, 1 },
1347 	{ IPV6CTL_SENDREDIRECTS, &ip6_sendredirects, 0, 1 },
1348 	{ IPV6CTL_DEFHLIM, &ip6_defhlim, 0, 255 },
1349 	{ IPV6CTL_MAXFRAGPACKETS, &ip6_maxfragpackets, 0, 1000 },
1350 	{ IPV6CTL_LOG_INTERVAL, &ip6_log_interval, 0, INT_MAX },
1351 	{ IPV6CTL_HDRNESTLIMIT, &ip6_hdrnestlimit, 0, 100 },
1352 	{ IPV6CTL_DAD_COUNT, &ip6_dad_count, 0, 10 },
1353 	{ IPV6CTL_AUTO_FLOWLABEL, &ip6_auto_flowlabel, 0, 1 },
1354 	{ IPV6CTL_DEFMCASTHLIM, &ip6_defmcasthlim, 0, 255 },
1355 	{ IPV6CTL_USE_DEPRECATED, &ip6_use_deprecated, 0, 1 },
1356 	{ IPV6CTL_MAXFRAGS, &ip6_maxfrags, 0, 1000 },
1357 	{ IPV6CTL_MFORWARDING, &ip6_mforwarding, 0, 1 },
1358 	{ IPV6CTL_MULTIPATH, &ip6_multipath, 0, 1 },
1359 	{ IPV6CTL_MCAST_PMTU, &ip6_mcast_pmtu, 0, 1 },
1360 	{ IPV6CTL_NEIGHBORGCTHRESH, &ip6_neighborgcthresh, -1, 5 * 2048 },
1361 	{ IPV6CTL_MAXDYNROUTES, &ip6_maxdynroutes, -1, 5 * 4096 },
1362 };
1363 
1364 int
1365 ip6_sysctl_ip6stat(void *oldp, size_t *oldlenp, void *newp)
1366 {
1367 	struct ip6stat *ip6stat;
1368 	int ret;
1369 
1370 	CTASSERT(sizeof(*ip6stat) == (ip6s_ncounters * sizeof(uint64_t)));
1371 
1372 	ip6stat = malloc(sizeof(*ip6stat), M_TEMP, M_WAITOK);
1373 	counters_read(ip6counters, (uint64_t *)ip6stat, ip6s_ncounters);
1374 	ret = sysctl_rdstruct(oldp, oldlenp, newp,
1375 	    ip6stat, sizeof(*ip6stat));
1376 	free(ip6stat, M_TEMP, sizeof(*ip6stat));
1377 
1378 	return (ret);
1379 }
1380 
1381 int
1382 ip6_sysctl_soiikey(void *oldp, size_t *oldlenp, void *newp, size_t newlen)
1383 {
1384 	uint8_t oldkey[IP6_SOIIKEY_LEN];
1385 	int error;
1386 
1387 	error = suser(curproc);
1388 	if (error != 0)
1389 		return (error);
1390 
1391 	memcpy(oldkey, ip6_soiikey, sizeof(oldkey));
1392 
1393 	error = sysctl_struct(oldp, oldlenp, newp, newlen, ip6_soiikey,
1394 	    sizeof(ip6_soiikey));
1395 
1396 	return (error);
1397 }
1398 
1399 int
1400 ip6_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
1401     void *newp, size_t newlen)
1402 {
1403 #ifdef MROUTING
1404 	extern struct mrt6stat mrt6stat;
1405 #endif
1406 	int error;
1407 
1408 	/* Almost all sysctl names at this level are terminal. */
1409 	if (namelen != 1 && name[0] != IPV6CTL_IFQUEUE)
1410 		return (ENOTDIR);
1411 
1412 	switch (name[0]) {
1413 	case IPV6CTL_STATS:
1414 		return (ip6_sysctl_ip6stat(oldp, oldlenp, newp));
1415 #ifdef MROUTING
1416 	case IPV6CTL_MRTSTATS:
1417 		if (newp != NULL)
1418 			return (EPERM);
1419 		NET_LOCK();
1420 		error = sysctl_struct(oldp, oldlenp, newp, newlen,
1421 		    &mrt6stat, sizeof(mrt6stat));
1422 		NET_UNLOCK();
1423 		return (error);
1424 	case IPV6CTL_MRTMIF:
1425 		if (newp)
1426 			return (EPERM);
1427 		NET_LOCK();
1428 		error = mrt6_sysctl_mif(oldp, oldlenp);
1429 		NET_UNLOCK();
1430 		return (error);
1431 	case IPV6CTL_MRTMFC:
1432 		if (newp)
1433 			return (EPERM);
1434 		NET_LOCK();
1435 		error = mrt6_sysctl_mfc(oldp, oldlenp);
1436 		NET_UNLOCK();
1437 		return (error);
1438 #else
1439 	case IPV6CTL_MRTSTATS:
1440 	case IPV6CTL_MRTPROTO:
1441 	case IPV6CTL_MRTMIF:
1442 	case IPV6CTL_MRTMFC:
1443 		return (EOPNOTSUPP);
1444 #endif
1445 	case IPV6CTL_MTUDISCTIMEOUT:
1446 		NET_LOCK();
1447 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1448 		    &ip6_mtudisc_timeout);
1449 		if (icmp6_mtudisc_timeout_q != NULL)
1450 			rt_timer_queue_change(icmp6_mtudisc_timeout_q,
1451 			    ip6_mtudisc_timeout);
1452 		NET_UNLOCK();
1453 		return (error);
1454 	case IPV6CTL_IFQUEUE:
1455 		return (EOPNOTSUPP);
1456 	case IPV6CTL_SOIIKEY:
1457 		return (ip6_sysctl_soiikey(oldp, oldlenp, newp, newlen));
1458 	default:
1459 		NET_LOCK();
1460 		error = sysctl_bounded_arr(ipv6ctl_vars, nitems(ipv6ctl_vars),
1461 		    name, namelen, oldp, oldlenp, newp, newlen);
1462 		NET_UNLOCK();
1463 		return (error);
1464 	}
1465 	/* NOTREACHED */
1466 }
1467 
1468 void
1469 ip6_send_dispatch(void *xmq)
1470 {
1471 	struct mbuf_queue *mq = xmq;
1472 	struct mbuf *m;
1473 	struct mbuf_list ml;
1474 
1475 	mq_delist(mq, &ml);
1476 	if (ml_empty(&ml))
1477 		return;
1478 
1479 	NET_LOCK();
1480 	while ((m = ml_dequeue(&ml)) != NULL) {
1481 		/*
1482 		 * To avoid a "too big" situation at an intermediate router and
1483 		 * the path MTU discovery process, specify the IPV6_MINMTU
1484 		 * flag.  Note that only echo and node information replies are
1485 		 * affected, since the length of ICMP6 errors is limited to the
1486 		 * minimum MTU.
1487 		 */
1488 		ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL);
1489 	}
1490 	NET_UNLOCK();
1491 }
1492 
1493 void
1494 ip6_send(struct mbuf *m)
1495 {
1496 	mq_enqueue(&ip6send_mq, m);
1497 	task_add(net_tq(0), &ip6send_task);
1498 }
1499