xref: /openbsd-src/sys/netinet6/ip6_input.c (revision 6a13ef69787db04ae501a22e92fa10865b44fd7c)
1 /*	$OpenBSD: ip6_input.c,v 1.174 2016/12/27 18:45:01 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 <netinet/ip6.h>
96 #include <netinet6/ip6_var.h>
97 #include <netinet/icmp6.h>
98 #include <netinet6/nd6.h>
99 
100 #include <netinet6/ip6protosw.h>
101 
102 #include "gif.h"
103 #include "bpfilter.h"
104 
105 #ifdef MROUTING
106 #include <netinet6/ip6_mroute.h>
107 #endif
108 
109 #if NPF > 0
110 #include <net/pfvar.h>
111 #endif
112 
113 #if NCARP > 0
114 #include <netinet/ip_carp.h>
115 #endif
116 
117 struct in6_ifaddrhead in6_ifaddr;
118 struct niqueue ip6intrq = NIQUEUE_INITIALIZER(IFQ_MAXLEN, NETISR_IPV6);
119 
120 struct ip6stat ip6stat;
121 
122 int ip6_check_rh0hdr(struct mbuf *, int *);
123 
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 
128 static struct mbuf_queue	ip6send_mq;
129 
130 static void ip6_send_dispatch(void *);
131 static struct task ip6send_task =
132 	TASK_INITIALIZER(ip6_send_dispatch, &ip6send_mq);
133 
134 /*
135  * IP6 initialization: fill in IP6 protocol switch table.
136  * All protocols not implemented in kernel go to raw IP6 protocol handler.
137  */
138 void
139 ip6_init(void)
140 {
141 	struct ip6protosw *pr;
142 	int i;
143 
144 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
145 	if (pr == NULL)
146 		panic("ip6_init");
147 	for (i = 0; i < IPPROTO_MAX; i++)
148 		ip6_protox[i] = pr - inet6sw;
149 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
150 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
151 		if (pr->pr_domain->dom_family == PF_INET6 &&
152 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW &&
153 		    pr->pr_protocol < IPPROTO_MAX)
154 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
155 	TAILQ_INIT(&in6_ifaddr);
156 	ip6_randomid_init();
157 	nd6_init();
158 	frag6_init();
159 
160 	mq_init(&ip6send_mq, 64, IPL_SOFTNET);
161 }
162 
163 /*
164  * IP6 input interrupt handling. Just pass the packet to ip6_input.
165  */
166 void
167 ip6intr(void)
168 {
169 	struct mbuf *m;
170 
171 	while ((m = niq_dequeue(&ip6intrq)) != NULL)
172 		ip6_input(m);
173 }
174 
175 void
176 ip6_input(struct mbuf *m)
177 {
178 	struct ifnet *ifp;
179 	struct ip6_hdr *ip6;
180 	struct sockaddr_in6 sin6;
181 	struct rtentry *rt = NULL;
182 	int off, nest;
183 	u_int16_t src_scope, dst_scope;
184 	int nxt, ours = 0;
185 #if NPF > 0
186 	struct in6_addr odst;
187 #endif
188 	int srcrt = 0;
189 
190 	ifp = if_get(m->m_pkthdr.ph_ifidx);
191 	if (ifp == NULL)
192 		goto bad;
193 
194 	if (m->m_flags & M_EXT) {
195 		if (m->m_next)
196 			ip6stat.ip6s_mext2m++;
197 		else
198 			ip6stat.ip6s_mext1++;
199 	} else {
200 		if (m->m_next) {
201 			int ifidx = m->m_pkthdr.ph_ifidx;
202 			if (m->m_flags & M_LOOP)
203 				ifidx = rtable_loindex(m->m_pkthdr.ph_rtableid);
204 			if (ifidx < nitems(ip6stat.ip6s_m2m))
205 				ip6stat.ip6s_m2m[ifidx]++;
206 			else
207 				ip6stat.ip6s_m2m[0]++;
208 		} else
209 			ip6stat.ip6s_m1++;
210 	}
211 
212 	ip6stat.ip6s_total++;
213 
214 	if (m->m_len < sizeof(struct ip6_hdr)) {
215 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
216 			ip6stat.ip6s_toosmall++;
217 			if_put(ifp);
218 			return;
219 		}
220 	}
221 
222 	ip6 = mtod(m, struct ip6_hdr *);
223 
224 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
225 		ip6stat.ip6s_badvers++;
226 		goto bad;
227 	}
228 
229 #if NCARP > 0
230 	if (ifp->if_type == IFT_CARP && ip6->ip6_nxt != IPPROTO_ICMPV6 &&
231 	    carp_lsdrop(m, AF_INET6, ip6->ip6_src.s6_addr32,
232 	    ip6->ip6_dst.s6_addr32))
233 		goto bad;
234 #endif
235 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
236 
237 	/*
238 	 * Check against address spoofing/corruption.
239 	 */
240 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
241 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
242 		/*
243 		 * XXX: "badscope" is not very suitable for a multicast source.
244 		 */
245 		ip6stat.ip6s_badscope++;
246 		goto bad;
247 	}
248 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
249 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
250 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
251 		    ip6stat.ip6s_badscope++;
252 		    goto bad;
253 	}
254 	/* Drop packets if interface ID portion is already filled. */
255 	if (((IN6_IS_SCOPE_EMBED(&ip6->ip6_src) && ip6->ip6_src.s6_addr16[1]) ||
256 	    (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst) && ip6->ip6_dst.s6_addr16[1])) &&
257 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
258 		ip6stat.ip6s_badscope++;
259 		goto bad;
260 	}
261 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
262 	    !(m->m_flags & M_LOOP)) {
263 		/*
264 		 * In this case, the packet should come from the loopback
265 		 * interface.  However, we cannot just check the if_flags,
266 		 * because ip6_mloopback() passes the "actual" interface
267 		 * as the outgoing/incoming interface.
268 		 */
269 		ip6stat.ip6s_badscope++;
270 		goto bad;
271 	}
272 
273 	/*
274 	 * The following check is not documented in specs.  A malicious
275 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
276 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
277 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
278 	 *
279 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
280 	 * support IPv4-less kernel compilation, we cannot support SIIT
281 	 * environment at all.  So, it makes more sense for us to reject any
282 	 * malicious packets for non-SIIT environment, than try to do a
283 	 * partial support for SIIT environment.
284 	 */
285 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
286 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
287 		ip6stat.ip6s_badscope++;
288 		goto bad;
289 	}
290 
291 	/*
292 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
293 	 *
294 	 * The code forbids automatic tunneling as per RFC4213.
295 	 */
296 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
297 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
298 		ip6stat.ip6s_badscope++;
299 		goto bad;
300 	}
301 
302 	/*
303 	 * If the packet has been received on a loopback interface it
304 	 * can be destinated to any local address, not necessarily to
305 	 * an address configured on `ifp'.
306 	 */
307 	if (ifp->if_flags & IFF_LOOPBACK) {
308 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
309 			src_scope = ip6->ip6_src.s6_addr16[1];
310 			ip6->ip6_src.s6_addr16[1] = 0;
311 		}
312 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
313 			dst_scope = ip6->ip6_dst.s6_addr16[1];
314 			ip6->ip6_dst.s6_addr16[1] = 0;
315 		}
316 	}
317 
318 #if NPF > 0
319         /*
320          * Packet filter
321          */
322 	odst = ip6->ip6_dst;
323 	if (pf_test(AF_INET6, PF_IN, ifp, &m) != PF_PASS)
324 		goto bad;
325 	if (m == NULL)
326 		goto bad;
327 
328 	ip6 = mtod(m, struct ip6_hdr *);
329 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
330 #endif
331 
332 	/*
333 	 * Without embedded scope ID we cannot find link-local
334 	 * addresses in the routing table.
335 	 */
336 	if (ifp->if_flags & IFF_LOOPBACK) {
337 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
338 			ip6->ip6_src.s6_addr16[1] = src_scope;
339 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
340 			ip6->ip6_dst.s6_addr16[1] = dst_scope;
341 	} else {
342 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
343 			ip6->ip6_src.s6_addr16[1] = htons(ifp->if_index);
344 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
345 			ip6->ip6_dst.s6_addr16[1] = htons(ifp->if_index);
346 	}
347 
348 	/*
349 	 * Be more secure than RFC5095 and scan for type 0 routing headers.
350 	 * If pf has already scanned the header chain, do not do it twice.
351 	 */
352 	if (!(m->m_pkthdr.pf.flags & PF_TAG_PROCESSED) &&
353 	    ip6_check_rh0hdr(m, &off)) {
354 		ip6stat.ip6s_badoptions++;
355 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, off);
356 		/* m is already freed */
357 		if_put(ifp);
358 		return;
359 	}
360 
361 	if (IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
362 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) {
363 		ours = 1;
364 		goto hbhcheck;
365 	}
366 
367 #if NPF > 0
368 	if (pf_ouraddr(m) == 1) {
369 		ours = 1;
370 		goto hbhcheck;
371 	}
372 #endif
373 
374 	/*
375 	 * Multicast check
376 	 */
377 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
378 		/*
379 		 * Make sure M_MCAST is set.  It should theoretically
380 		 * already be there, but let's play safe because upper
381 		 * layers check for this flag.
382 		 */
383 		m->m_flags |= M_MCAST;
384 
385 		/*
386 		 * See if we belong to the destination multicast group on the
387 		 * arrival interface.
388 		 */
389 		if (in6_hasmulti(&ip6->ip6_dst, ifp))
390 			ours = 1;
391 
392 #ifdef MROUTING
393 		if (ip6_mforwarding && ip6_mrouter) {
394 			if (ip6_hbhchcheck(m, &off, &nxt, &ours)) {
395 				rtfree(rt);
396 				if_put(ifp);
397 				return;	/* m have already been freed */
398 			}
399 
400 			ip6 = mtod(m, struct ip6_hdr *);
401 
402 			/*
403 			 * If we are acting as a multicast router, all
404 			 * incoming multicast packets are passed to the
405 			 * kernel-level multicast forwarding function.
406 			 * The packet is returned (relatively) intact; if
407 			 * ip6_mforward() returns a non-zero value, the packet
408 			 * must be discarded, else it may be accepted below.
409 			 */
410 		    	if (ip6_mforward(ip6, ifp, m)) {
411 				ip6stat.ip6s_cantforward++;
412 				goto bad;
413 			}
414 
415 			if (!ours)
416 				goto bad;
417 			goto ours;
418 		}
419 #endif
420 		if (!ours) {
421 			ip6stat.ip6s_notmember++;
422 			if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
423 				ip6stat.ip6s_cantforward++;
424 			goto bad;
425 		}
426 		goto hbhcheck;
427 	}
428 
429 
430 	/*
431 	 *  Unicast check
432 	 */
433 	memset(&sin6, 0, sizeof(struct sockaddr_in6));
434 	sin6.sin6_len = sizeof(struct sockaddr_in6);
435 	sin6.sin6_family = AF_INET6;
436 	sin6.sin6_addr = ip6->ip6_dst;
437 	rt = rtalloc_mpath(sin6tosa(&sin6), &ip6->ip6_src.s6_addr32[0],
438 	    m->m_pkthdr.ph_rtableid);
439 
440 	/*
441 	 * Accept the packet if the route to the destination is marked
442 	 * as local.
443 	 */
444 	if (rtisvalid(rt) && ISSET(rt->rt_flags, RTF_LOCAL)) {
445 		struct in6_ifaddr *ia6 = ifatoia6(rt->rt_ifa);
446 		if (ia6->ia6_flags & IN6_IFF_ANYCAST)
447 			m->m_flags |= M_ACAST;
448 		/*
449 		 * packets to a tentative, duplicated, or somehow invalid
450 		 * address must not be accepted.
451 		 */
452 		if ((ia6->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED))) {
453 			char src[INET6_ADDRSTRLEN], dst[INET6_ADDRSTRLEN];
454 
455 			inet_ntop(AF_INET6, &ip6->ip6_src, src, sizeof(src));
456 			inet_ntop(AF_INET6, &ip6->ip6_dst, dst, sizeof(dst));
457 			/* address is not ready, so discard the packet. */
458 			nd6log((LOG_INFO,
459 			    "ip6_input: packet to an unready address %s->%s\n",
460 			    src, dst));
461 
462 			goto bad;
463 		} else {
464 			/* this address is ready */
465 			ours = 1;
466 			goto hbhcheck;
467 		}
468 	}
469 
470 #if NCARP > 0
471 	if (ifp->if_type == IFT_CARP && ip6->ip6_nxt == IPPROTO_ICMPV6 &&
472 	    carp_lsdrop(m, AF_INET6, ip6->ip6_src.s6_addr32,
473 	    ip6->ip6_dst.s6_addr32))
474 		goto bad;
475 #endif
476 	/*
477 	 * Now there is no reason to process the packet if it's not our own
478 	 * and we're not a router.
479 	 */
480 	if (!ip6_forwarding) {
481 		ip6stat.ip6s_cantforward++;
482 		goto bad;
483 	}
484 
485   hbhcheck:
486 
487 	if (ip6_hbhchcheck(m, &off, &nxt, &ours)) {
488 		rtfree(rt);
489 		if_put(ifp);
490 		return;	/* m have already been freed */
491 	}
492 
493 	/* adjust pointer */
494 	ip6 = mtod(m, struct ip6_hdr *);
495 
496 	/*
497 	 * Forward if desirable.
498 	 */
499 	if (!ours) {
500 		ip6_forward(m, rt, srcrt);
501 		if_put(ifp);
502 		return;
503 	}
504 #ifdef MROUTING
505   ours:
506 #endif
507 	/* pf might have changed things */
508 	in6_proto_cksum_out(m, NULL);
509 
510 	ip6 = mtod(m, struct ip6_hdr *);
511 
512 	/*
513 	 * Tell launch routine the next header
514 	 */
515 	ip6stat.ip6s_delivered++;
516 	nest = 0;
517 
518 	while (nxt != IPPROTO_DONE) {
519 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
520 			ip6stat.ip6s_toomanyhdr++;
521 			goto bad;
522 		}
523 
524 		/*
525 		 * protection against faulty packet - there should be
526 		 * more sanity checks in header chain processing.
527 		 */
528 		if (m->m_pkthdr.len < off) {
529 			ip6stat.ip6s_tooshort++;
530 			goto bad;
531 		}
532 
533 		/* draft-itojun-ipv6-tcp-to-anycast */
534 		if (ISSET(m->m_flags, M_ACAST) && (nxt == IPPROTO_TCP)) {
535 			if (m->m_len >= sizeof(struct ip6_hdr)) {
536 				icmp6_error(m, ICMP6_DST_UNREACH,
537 					ICMP6_DST_UNREACH_ADDR,
538 					offsetof(struct ip6_hdr, ip6_dst));
539 				break;
540 			} else
541 				goto bad;
542 		}
543 
544 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
545 	}
546 	rtfree(rt);
547 	if_put(ifp);
548 	return;
549  bad:
550 	rtfree(rt);
551 	if_put(ifp);
552 	m_freem(m);
553 }
554 
555 int
556 ip6_hbhchcheck(struct mbuf *m, int *offp, int *nxtp, int *oursp)
557 {
558 	struct ip6_hdr *ip6;
559 	u_int32_t plen, rtalert = ~0;
560 
561 	ip6 = mtod(m, struct ip6_hdr *);
562 
563 	/*
564 	 * Process Hop-by-Hop options header if it's contained.
565 	 * m may be modified in ip6_hopopts_input().
566 	 * If a JumboPayload option is included, plen will also be modified.
567 	 */
568 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
569 	*offp = sizeof(struct ip6_hdr);
570 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
571 		struct ip6_hbh *hbh;
572 
573 		if (ip6_hopopts_input(&plen, &rtalert, &m, offp)) {
574 			return (-1);	/* m have already been freed */
575 		}
576 
577 		/* adjust pointer */
578 		ip6 = mtod(m, struct ip6_hdr *);
579 
580 		/*
581 		 * if the payload length field is 0 and the next header field
582 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
583 		 * option MUST be included.
584 		 */
585 		if (ip6->ip6_plen == 0 && plen == 0) {
586 			/*
587 			 * Note that if a valid jumbo payload option is
588 			 * contained, ip6_hopopts_input() must set a valid
589 			 * (non-zero) payload length to the variable plen.
590 			 */
591 			ip6stat.ip6s_badoptions++;
592 			icmp6_error(m, ICMP6_PARAM_PROB,
593 				    ICMP6_PARAMPROB_HEADER,
594 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
595 			return (-1);
596 		}
597 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
598 			sizeof(struct ip6_hbh));
599 		if (hbh == NULL) {
600 			ip6stat.ip6s_tooshort++;
601 			return (-1);
602 		}
603 		*nxtp = hbh->ip6h_nxt;
604 
605 		/*
606 		 * accept the packet if a router alert option is included
607 		 * and we act as an IPv6 router.
608 		 */
609 		if (rtalert != ~0 && ip6_forwarding)
610 			*oursp = 1;
611 	} else
612 		*nxtp = ip6->ip6_nxt;
613 
614 	/*
615 	 * Check that the amount of data in the buffers
616 	 * is as at least much as the IPv6 header would have us expect.
617 	 * Trim mbufs if longer than we expect.
618 	 * Drop packet if shorter than we expect.
619 	 */
620 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
621 		ip6stat.ip6s_tooshort++;
622 		m_freem(m);
623 		return (-1);
624 	}
625 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
626 		if (m->m_len == m->m_pkthdr.len) {
627 			m->m_len = sizeof(struct ip6_hdr) + plen;
628 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
629 		} else {
630 			m_adj(m,
631 			    sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
632 		}
633 	}
634 
635 	return (0);
636 }
637 
638 /* scan packet for RH0 routing header. Mostly stolen from pf.c:pf_test() */
639 int
640 ip6_check_rh0hdr(struct mbuf *m, int *offp)
641 {
642 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
643 	struct ip6_rthdr rthdr;
644 	struct ip6_ext opt6;
645 	u_int8_t proto = ip6->ip6_nxt;
646 	int done = 0, lim, off, rh_cnt = 0;
647 
648 	off = ((caddr_t)ip6 - m->m_data) + sizeof(struct ip6_hdr);
649 	lim = min(m->m_pkthdr.len, ntohs(ip6->ip6_plen) + sizeof(*ip6));
650 	do {
651 		switch (proto) {
652 		case IPPROTO_ROUTING:
653 			*offp = off;
654 			if (rh_cnt++) {
655 				/* more than one rh header present */
656 				return (1);
657 			}
658 
659 			if (off + sizeof(rthdr) > lim) {
660 				/* packet to short to make sense */
661 				return (1);
662 			}
663 
664 			m_copydata(m, off, sizeof(rthdr), (caddr_t)&rthdr);
665 
666 			if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) {
667 				*offp += offsetof(struct ip6_rthdr, ip6r_type);
668 				return (1);
669 			}
670 
671 			off += (rthdr.ip6r_len + 1) * 8;
672 			proto = rthdr.ip6r_nxt;
673 			break;
674 		case IPPROTO_AH:
675 		case IPPROTO_HOPOPTS:
676 		case IPPROTO_DSTOPTS:
677 			/* get next header and header length */
678 			if (off + sizeof(opt6) > lim) {
679 				/*
680 				 * Packet to short to make sense, we could
681 				 * reject the packet but as a router we
682 				 * should not do that so forward it.
683 				 */
684 				return (0);
685 			}
686 
687 			m_copydata(m, off, sizeof(opt6), (caddr_t)&opt6);
688 
689 			if (proto == IPPROTO_AH)
690 				off += (opt6.ip6e_len + 2) * 4;
691 			else
692 				off += (opt6.ip6e_len + 1) * 8;
693 			proto = opt6.ip6e_nxt;
694 			break;
695 		case IPPROTO_FRAGMENT:
696 		default:
697 			/* end of header stack */
698 			done = 1;
699 			break;
700 		}
701 	} while (!done);
702 
703 	return (0);
704 }
705 
706 /*
707  * Hop-by-Hop options header processing. If a valid jumbo payload option is
708  * included, the real payload length will be stored in plenp.
709  *
710  * rtalertp - XXX: should be stored in a more smart way
711  */
712 int
713 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, struct mbuf **mp,
714     int *offp)
715 {
716 	struct mbuf *m = *mp;
717 	int off = *offp, hbhlen;
718 	struct ip6_hbh *hbh;
719 
720 	/* validation of the length of the header */
721 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
722 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
723 	if (hbh == NULL) {
724 		ip6stat.ip6s_tooshort++;
725 		return -1;
726 	}
727 	hbhlen = (hbh->ip6h_len + 1) << 3;
728 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
729 		hbhlen);
730 	if (hbh == NULL) {
731 		ip6stat.ip6s_tooshort++;
732 		return -1;
733 	}
734 	off += hbhlen;
735 	hbhlen -= sizeof(struct ip6_hbh);
736 
737 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
738 				hbhlen, rtalertp, plenp) < 0)
739 		return (-1);
740 
741 	*offp = off;
742 	*mp = m;
743 	return (0);
744 }
745 
746 /*
747  * Search header for all Hop-by-hop options and process each option.
748  * This function is separate from ip6_hopopts_input() in order to
749  * handle a case where the sending node itself process its hop-by-hop
750  * options header. In such a case, the function is called from ip6_output().
751  *
752  * The function assumes that hbh header is located right after the IPv6 header
753  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
754  * opthead + hbhlen is located in continuous memory region.
755  */
756 int
757 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
758     u_int32_t *rtalertp, u_int32_t *plenp)
759 {
760 	struct ip6_hdr *ip6;
761 	int optlen = 0;
762 	u_int8_t *opt = opthead;
763 	u_int16_t rtalert_val;
764 	u_int32_t jumboplen;
765 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
766 
767 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
768 		switch (*opt) {
769 		case IP6OPT_PAD1:
770 			optlen = 1;
771 			break;
772 		case IP6OPT_PADN:
773 			if (hbhlen < IP6OPT_MINLEN) {
774 				ip6stat.ip6s_toosmall++;
775 				goto bad;
776 			}
777 			optlen = *(opt + 1) + 2;
778 			break;
779 		case IP6OPT_ROUTER_ALERT:
780 			/* XXX may need check for alignment */
781 			if (hbhlen < IP6OPT_RTALERT_LEN) {
782 				ip6stat.ip6s_toosmall++;
783 				goto bad;
784 			}
785 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
786 				/* XXX stat */
787 				icmp6_error(m, ICMP6_PARAM_PROB,
788 				    ICMP6_PARAMPROB_HEADER,
789 				    erroff + opt + 1 - opthead);
790 				return (-1);
791 			}
792 			optlen = IP6OPT_RTALERT_LEN;
793 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
794 			*rtalertp = ntohs(rtalert_val);
795 			break;
796 		case IP6OPT_JUMBO:
797 			/* XXX may need check for alignment */
798 			if (hbhlen < IP6OPT_JUMBO_LEN) {
799 				ip6stat.ip6s_toosmall++;
800 				goto bad;
801 			}
802 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
803 				/* XXX stat */
804 				icmp6_error(m, ICMP6_PARAM_PROB,
805 				    ICMP6_PARAMPROB_HEADER,
806 				    erroff + opt + 1 - opthead);
807 				return (-1);
808 			}
809 			optlen = IP6OPT_JUMBO_LEN;
810 
811 			/*
812 			 * IPv6 packets that have non 0 payload length
813 			 * must not contain a jumbo payload option.
814 			 */
815 			ip6 = mtod(m, struct ip6_hdr *);
816 			if (ip6->ip6_plen) {
817 				ip6stat.ip6s_badoptions++;
818 				icmp6_error(m, ICMP6_PARAM_PROB,
819 				    ICMP6_PARAMPROB_HEADER,
820 				    erroff + opt - opthead);
821 				return (-1);
822 			}
823 
824 			/*
825 			 * We may see jumbolen in unaligned location, so
826 			 * we'd need to perform bcopy().
827 			 */
828 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
829 			jumboplen = (u_int32_t)htonl(jumboplen);
830 
831 #if 1
832 			/*
833 			 * if there are multiple jumbo payload options,
834 			 * *plenp will be non-zero and the packet will be
835 			 * rejected.
836 			 * the behavior may need some debate in ipngwg -
837 			 * multiple options does not make sense, however,
838 			 * there's no explicit mention in specification.
839 			 */
840 			if (*plenp != 0) {
841 				ip6stat.ip6s_badoptions++;
842 				icmp6_error(m, ICMP6_PARAM_PROB,
843 				    ICMP6_PARAMPROB_HEADER,
844 				    erroff + opt + 2 - opthead);
845 				return (-1);
846 			}
847 #endif
848 
849 			/*
850 			 * jumbo payload length must be larger than 65535.
851 			 */
852 			if (jumboplen <= IPV6_MAXPACKET) {
853 				ip6stat.ip6s_badoptions++;
854 				icmp6_error(m, ICMP6_PARAM_PROB,
855 				    ICMP6_PARAMPROB_HEADER,
856 				    erroff + opt + 2 - opthead);
857 				return (-1);
858 			}
859 			*plenp = jumboplen;
860 
861 			break;
862 		default:		/* unknown option */
863 			if (hbhlen < IP6OPT_MINLEN) {
864 				ip6stat.ip6s_toosmall++;
865 				goto bad;
866 			}
867 			optlen = ip6_unknown_opt(opt, m,
868 			    erroff + opt - opthead);
869 			if (optlen == -1)
870 				return (-1);
871 			optlen += 2;
872 			break;
873 		}
874 	}
875 
876 	return (0);
877 
878   bad:
879 	m_freem(m);
880 	return (-1);
881 }
882 
883 /*
884  * Unknown option processing.
885  * The third argument `off' is the offset from the IPv6 header to the option,
886  * which allows returning an ICMPv6 error even if the IPv6 header and the
887  * option header are not continuous.
888  */
889 int
890 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
891 {
892 	struct ip6_hdr *ip6;
893 
894 	switch (IP6OPT_TYPE(*optp)) {
895 	case IP6OPT_TYPE_SKIP: /* ignore the option */
896 		return ((int)*(optp + 1));
897 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
898 		m_freem(m);
899 		return (-1);
900 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
901 		ip6stat.ip6s_badoptions++;
902 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
903 		return (-1);
904 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
905 		ip6stat.ip6s_badoptions++;
906 		ip6 = mtod(m, struct ip6_hdr *);
907 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
908 		    (m->m_flags & (M_BCAST|M_MCAST)))
909 			m_freem(m);
910 		else
911 			icmp6_error(m, ICMP6_PARAM_PROB,
912 				    ICMP6_PARAMPROB_OPTION, off);
913 		return (-1);
914 	}
915 
916 	m_freem(m);		/* XXX: NOTREACHED */
917 	return (-1);
918 }
919 
920 /*
921  * Create the "control" list for this pcb.
922  *
923  * The routine will be called from upper layer handlers like tcp6_input().
924  * Thus the routine assumes that the caller (tcp6_input) have already
925  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
926  * very first mbuf on the mbuf chain.
927  * We may want to add some infinite loop prevention or sanity checks for safety.
928  * (This applies only when you are using KAME mbuf chain restriction, i.e.
929  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
930  */
931 void
932 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
933 {
934 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
935 
936 #ifdef SO_TIMESTAMP
937 	if (in6p->inp_socket->so_options & SO_TIMESTAMP) {
938 		struct timeval tv;
939 
940 		microtime(&tv);
941 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
942 		    SCM_TIMESTAMP, SOL_SOCKET);
943 		if (*mp)
944 			mp = &(*mp)->m_next;
945 	}
946 #endif
947 
948 	/* RFC 2292 sec. 5 */
949 	if ((in6p->inp_flags & IN6P_PKTINFO) != 0) {
950 		struct in6_pktinfo pi6;
951 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
952 		if (IN6_IS_SCOPE_EMBED(&pi6.ipi6_addr))
953 			pi6.ipi6_addr.s6_addr16[1] = 0;
954 		pi6.ipi6_ifindex = m ? m->m_pkthdr.ph_ifidx : 0;
955 		*mp = sbcreatecontrol((caddr_t) &pi6,
956 		    sizeof(struct in6_pktinfo),
957 		    IPV6_PKTINFO, IPPROTO_IPV6);
958 		if (*mp)
959 			mp = &(*mp)->m_next;
960 	}
961 
962 	if ((in6p->inp_flags & IN6P_HOPLIMIT) != 0) {
963 		int hlim = ip6->ip6_hlim & 0xff;
964 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
965 		    IPV6_HOPLIMIT, IPPROTO_IPV6);
966 		if (*mp)
967 			mp = &(*mp)->m_next;
968 	}
969 
970 	if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
971 		u_int32_t flowinfo;
972 		int tclass;
973 
974 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
975 		flowinfo >>= 20;
976 
977 		tclass = flowinfo & 0xff;
978 		*mp = sbcreatecontrol((caddr_t)&tclass, sizeof(tclass),
979 		    IPV6_TCLASS, IPPROTO_IPV6);
980 		if (*mp)
981 			mp = &(*mp)->m_next;
982 	}
983 
984 	/*
985 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
986 	 * privilege for the option (see ip6_ctloutput), but it might be too
987 	 * strict, since there might be some hop-by-hop options which can be
988 	 * returned to normal user.
989 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
990 	 */
991 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
992 		/*
993 		 * Check if a hop-by-hop options header is contained in the
994 		 * received packet, and if so, store the options as ancillary
995 		 * data. Note that a hop-by-hop options header must be
996 		 * just after the IPv6 header, which is assured through the
997 		 * IPv6 input processing.
998 		 */
999 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1000 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1001 			struct ip6_hbh *hbh;
1002 			int hbhlen = 0;
1003 			struct mbuf *ext;
1004 
1005 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1006 			    ip6->ip6_nxt);
1007 			if (ext == NULL) {
1008 				ip6stat.ip6s_tooshort++;
1009 				return;
1010 			}
1011 			hbh = mtod(ext, struct ip6_hbh *);
1012 			hbhlen = (hbh->ip6h_len + 1) << 3;
1013 			if (hbhlen != ext->m_len) {
1014 				m_freem(ext);
1015 				ip6stat.ip6s_tooshort++;
1016 				return;
1017 			}
1018 
1019 			/*
1020 			 * XXX: We copy the whole header even if a
1021 			 * jumbo payload option is included, the option which
1022 			 * is to be removed before returning according to
1023 			 * RFC2292.
1024 			 * Note: this constraint is removed in RFC3542.
1025 			 */
1026 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1027 			    IPV6_HOPOPTS,
1028 			    IPPROTO_IPV6);
1029 			if (*mp)
1030 				mp = &(*mp)->m_next;
1031 			m_freem(ext);
1032 		}
1033 	}
1034 
1035 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1036 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1037 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1038 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1039 
1040 		/*
1041 		 * Search for destination options headers or routing
1042 		 * header(s) through the header chain, and stores each
1043 		 * header as ancillary data.
1044 		 * Note that the order of the headers remains in
1045 		 * the chain of ancillary data.
1046 		 */
1047 		while (1) {	/* is explicit loop prevention necessary? */
1048 			struct ip6_ext *ip6e = NULL;
1049 			int elen;
1050 			struct mbuf *ext = NULL;
1051 
1052 			/*
1053 			 * if it is not an extension header, don't try to
1054 			 * pull it from the chain.
1055 			 */
1056 			switch (nxt) {
1057 			case IPPROTO_DSTOPTS:
1058 			case IPPROTO_ROUTING:
1059 			case IPPROTO_HOPOPTS:
1060 			case IPPROTO_AH: /* is it possible? */
1061 				break;
1062 			default:
1063 				goto loopend;
1064 			}
1065 
1066 			ext = ip6_pullexthdr(m, off, nxt);
1067 			if (ext == NULL) {
1068 				ip6stat.ip6s_tooshort++;
1069 				return;
1070 			}
1071 			ip6e = mtod(ext, struct ip6_ext *);
1072 			if (nxt == IPPROTO_AH)
1073 				elen = (ip6e->ip6e_len + 2) << 2;
1074 			else
1075 				elen = (ip6e->ip6e_len + 1) << 3;
1076 			if (elen != ext->m_len) {
1077 				m_freem(ext);
1078 				ip6stat.ip6s_tooshort++;
1079 				return;
1080 			}
1081 
1082 			switch (nxt) {
1083 			case IPPROTO_DSTOPTS:
1084 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1085 					break;
1086 
1087 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1088 				    IPV6_DSTOPTS,
1089 				    IPPROTO_IPV6);
1090 				if (*mp)
1091 					mp = &(*mp)->m_next;
1092 				break;
1093 
1094 			case IPPROTO_ROUTING:
1095 				if (!(in6p->inp_flags & IN6P_RTHDR))
1096 					break;
1097 
1098 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1099 				    IPV6_RTHDR,
1100 				    IPPROTO_IPV6);
1101 				if (*mp)
1102 					mp = &(*mp)->m_next;
1103 				break;
1104 
1105 			case IPPROTO_HOPOPTS:
1106 			case IPPROTO_AH: /* is it possible? */
1107 				break;
1108 
1109 			default:
1110 				/*
1111 				 * other cases have been filtered in the above.
1112 				 * none will visit this case.  here we supply
1113 				 * the code just in case (nxt overwritten or
1114 				 * other cases).
1115 				 */
1116 				m_freem(ext);
1117 				goto loopend;
1118 
1119 			}
1120 
1121 			/* proceed with the next header. */
1122 			off += elen;
1123 			nxt = ip6e->ip6e_nxt;
1124 			ip6e = NULL;
1125 			m_freem(ext);
1126 			ext = NULL;
1127 		}
1128 loopend:
1129 		;
1130 	}
1131 }
1132 
1133 /*
1134  * pull single extension header from mbuf chain.  returns single mbuf that
1135  * contains the result, or NULL on error.
1136  */
1137 struct mbuf *
1138 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1139 {
1140 	struct ip6_ext ip6e;
1141 	size_t elen;
1142 	struct mbuf *n;
1143 
1144 #ifdef DIAGNOSTIC
1145 	switch (nxt) {
1146 	case IPPROTO_DSTOPTS:
1147 	case IPPROTO_ROUTING:
1148 	case IPPROTO_HOPOPTS:
1149 	case IPPROTO_AH: /* is it possible? */
1150 		break;
1151 	default:
1152 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1153 	}
1154 #endif
1155 
1156 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1157 	if (nxt == IPPROTO_AH)
1158 		elen = (ip6e.ip6e_len + 2) << 2;
1159 	else
1160 		elen = (ip6e.ip6e_len + 1) << 3;
1161 
1162 	MGET(n, M_DONTWAIT, MT_DATA);
1163 	if (n && elen >= MLEN) {
1164 		MCLGET(n, M_DONTWAIT);
1165 		if ((n->m_flags & M_EXT) == 0) {
1166 			m_free(n);
1167 			n = NULL;
1168 		}
1169 	}
1170 	if (!n)
1171 		return NULL;
1172 
1173 	n->m_len = 0;
1174 	if (elen >= M_TRAILINGSPACE(n)) {
1175 		m_free(n);
1176 		return NULL;
1177 	}
1178 
1179 	m_copydata(m, off, elen, mtod(n, caddr_t));
1180 	n->m_len = elen;
1181 	return n;
1182 }
1183 
1184 /*
1185  * Get pointer to the previous header followed by the header
1186  * currently processed.
1187  * XXX: This function supposes that
1188  *	M includes all headers,
1189  *	the next header field and the header length field of each header
1190  *	are valid, and
1191  *	the sum of each header length equals to OFF.
1192  * Because of these assumptions, this function must be called very
1193  * carefully. Moreover, it will not be used in the near future when
1194  * we develop `neater' mechanism to process extension headers.
1195  */
1196 u_int8_t *
1197 ip6_get_prevhdr(struct mbuf *m, int off)
1198 {
1199 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1200 
1201 	if (off == sizeof(struct ip6_hdr))
1202 		return (&ip6->ip6_nxt);
1203 	else {
1204 		int len, nxt;
1205 		struct ip6_ext *ip6e = NULL;
1206 
1207 		nxt = ip6->ip6_nxt;
1208 		len = sizeof(struct ip6_hdr);
1209 		while (len < off) {
1210 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1211 
1212 			switch (nxt) {
1213 			case IPPROTO_FRAGMENT:
1214 				len += sizeof(struct ip6_frag);
1215 				break;
1216 			case IPPROTO_AH:
1217 				len += (ip6e->ip6e_len + 2) << 2;
1218 				break;
1219 			default:
1220 				len += (ip6e->ip6e_len + 1) << 3;
1221 				break;
1222 			}
1223 			nxt = ip6e->ip6e_nxt;
1224 		}
1225 		if (ip6e)
1226 			return (&ip6e->ip6e_nxt);
1227 		else
1228 			return NULL;
1229 	}
1230 }
1231 
1232 /*
1233  * get next header offset.  m will be retained.
1234  */
1235 int
1236 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1237 {
1238 	struct ip6_hdr ip6;
1239 	struct ip6_ext ip6e;
1240 	struct ip6_frag fh;
1241 
1242 	/* just in case */
1243 	if (m == NULL)
1244 		panic("ip6_nexthdr: m == NULL");
1245 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1246 		return -1;
1247 
1248 	switch (proto) {
1249 	case IPPROTO_IPV6:
1250 		if (m->m_pkthdr.len < off + sizeof(ip6))
1251 			return -1;
1252 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1253 		if (nxtp)
1254 			*nxtp = ip6.ip6_nxt;
1255 		off += sizeof(ip6);
1256 		return off;
1257 
1258 	case IPPROTO_FRAGMENT:
1259 		/*
1260 		 * terminate parsing if it is not the first fragment,
1261 		 * it does not make sense to parse through it.
1262 		 */
1263 		if (m->m_pkthdr.len < off + sizeof(fh))
1264 			return -1;
1265 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1266 		if ((fh.ip6f_offlg & IP6F_OFF_MASK) != 0)
1267 			return -1;
1268 		if (nxtp)
1269 			*nxtp = fh.ip6f_nxt;
1270 		off += sizeof(struct ip6_frag);
1271 		return off;
1272 
1273 	case IPPROTO_AH:
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 + 2) << 2;
1280 		if (m->m_pkthdr.len < off)
1281 			return -1;
1282 		return off;
1283 
1284 	case IPPROTO_HOPOPTS:
1285 	case IPPROTO_ROUTING:
1286 	case IPPROTO_DSTOPTS:
1287 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1288 			return -1;
1289 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1290 		if (nxtp)
1291 			*nxtp = ip6e.ip6e_nxt;
1292 		off += (ip6e.ip6e_len + 1) << 3;
1293 		if (m->m_pkthdr.len < off)
1294 			return -1;
1295 		return off;
1296 
1297 	case IPPROTO_NONE:
1298 	case IPPROTO_ESP:
1299 	case IPPROTO_IPCOMP:
1300 		/* give up */
1301 		return -1;
1302 
1303 	default:
1304 		return -1;
1305 	}
1306 
1307 	return -1;
1308 }
1309 
1310 /*
1311  * get offset for the last header in the chain.  m will be kept untainted.
1312  */
1313 int
1314 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1315 {
1316 	int newoff;
1317 	int nxt;
1318 
1319 	if (!nxtp) {
1320 		nxt = -1;
1321 		nxtp = &nxt;
1322 	}
1323 	while (1) {
1324 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1325 		if (newoff < 0)
1326 			return off;
1327 		else if (newoff < off)
1328 			return -1;	/* invalid */
1329 		else if (newoff == off)
1330 			return newoff;
1331 
1332 		off = newoff;
1333 		proto = *nxtp;
1334 	}
1335 }
1336 
1337 /*
1338  * System control for IP6
1339  */
1340 
1341 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1342 	0,		0,		0,		0,
1343 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1344 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1345 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1346 	0,		0,		0,		0,
1347 	ENOPROTOOPT
1348 };
1349 
1350 int *ipv6ctl_vars[IPV6CTL_MAXID] = IPV6CTL_VARS;
1351 
1352 int
1353 ip6_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
1354     void *newp, size_t newlen)
1355 {
1356 #ifdef MROUTING
1357 	extern int ip6_mrtproto;
1358 	extern struct mrt6stat mrt6stat;
1359 #endif
1360 	int error;
1361 
1362 	NET_ASSERT_LOCKED();
1363 
1364 	/* Almost all sysctl names at this level are terminal. */
1365 	if (namelen != 1 && name[0] != IPV6CTL_IFQUEUE)
1366 		return (ENOTDIR);
1367 
1368 	switch (name[0]) {
1369 	case IPV6CTL_DAD_PENDING:
1370 		return sysctl_rdint(oldp, oldlenp, newp, ip6_dad_pending);
1371 	case IPV6CTL_STATS:
1372 		if (newp != NULL)
1373 			return (EPERM);
1374 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
1375 		    &ip6stat, sizeof(ip6stat)));
1376 #ifdef MROUTING
1377 	case IPV6CTL_MRTSTATS:
1378 		if (newp != NULL)
1379 			return (EPERM);
1380 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
1381 		    &mrt6stat, sizeof(mrt6stat)));
1382 	case IPV6CTL_MRTPROTO:
1383 		return sysctl_rdint(oldp, oldlenp, newp, ip6_mrtproto);
1384 	case IPV6CTL_MRTMIF:
1385 		if (newp)
1386 			return (EPERM);
1387 		return mrt6_sysctl_mif(oldp, oldlenp);
1388 	case IPV6CTL_MRTMFC:
1389 		if (newp)
1390 			return (EPERM);
1391 		return mrt6_sysctl_mfc(oldp, oldlenp);
1392 #else
1393 	case IPV6CTL_MRTSTATS:
1394 	case IPV6CTL_MRTPROTO:
1395 	case IPV6CTL_MRTMIF:
1396 	case IPV6CTL_MRTMFC:
1397 		return (EOPNOTSUPP);
1398 #endif
1399 	case IPV6CTL_MTUDISCTIMEOUT:
1400 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1401 		   &ip6_mtudisc_timeout);
1402 		if (icmp6_mtudisc_timeout_q != NULL)
1403 			rt_timer_queue_change(icmp6_mtudisc_timeout_q,
1404 					      ip6_mtudisc_timeout);
1405 		return (error);
1406 	case IPV6CTL_IFQUEUE:
1407 		return (sysctl_niq(name + 1, namelen - 1,
1408 		    oldp, oldlenp, newp, newlen, &ip6intrq));
1409 	default:
1410 		if (name[0] < IPV6CTL_MAXID)
1411 			return (sysctl_int_arr(ipv6ctl_vars, name, namelen,
1412 			    oldp, oldlenp, newp, newlen));
1413 		return (EOPNOTSUPP);
1414 	}
1415 	/* NOTREACHED */
1416 }
1417 
1418 void
1419 ip6_send_dispatch(void *xmq)
1420 {
1421 	struct mbuf_queue *mq = xmq;
1422 	struct mbuf *m;
1423 	struct mbuf_list ml;
1424 	int s;
1425 
1426 	mq_delist(mq, &ml);
1427 	if (ml_empty(&ml))
1428 		return;
1429 
1430 	KERNEL_LOCK();
1431 	NET_LOCK(s);
1432 	while ((m = ml_dequeue(&ml)) != NULL) {
1433 		ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL);
1434 	}
1435 	NET_UNLOCK(s);
1436 	KERNEL_UNLOCK();
1437 }
1438 
1439 void
1440 ip6_send(struct mbuf *m)
1441 {
1442 	mq_enqueue(&ip6send_mq, m);
1443 	task_add(softnettq, &ip6send_task);
1444 }
1445