xref: /openbsd-src/sys/netinet6/ip6_input.c (revision d59bb9942320b767f2a19aaa7690c8c6e30b724c)
1 /*	$OpenBSD: ip6_input.c,v 1.180 2017/02/28 09:59:34 mpi Exp $	*/
2 /*	$KAME: ip6_input.c,v 1.188 2001/03/29 05:34:31 itojun Exp $	*/
3 
4 /*
5  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
6  * All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the project nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 /*
34  * Copyright (c) 1982, 1986, 1988, 1993
35  *	The Regents of the University of California.  All rights reserved.
36  *
37  * Redistribution and use in source and binary forms, with or without
38  * modification, are permitted provided that the following conditions
39  * are met:
40  * 1. Redistributions of source code must retain the above copyright
41  *    notice, this list of conditions and the following disclaimer.
42  * 2. Redistributions in binary form must reproduce the above copyright
43  *    notice, this list of conditions and the following disclaimer in the
44  *    documentation and/or other materials provided with the distribution.
45  * 3. Neither the name of the University nor the names of its contributors
46  *    may be used to endorse or promote products derived from this software
47  *    without specific prior written permission.
48  *
49  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
50  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
51  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
52  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
53  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
54  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
55  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
56  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
57  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
58  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
59  * SUCH DAMAGE.
60  *
61  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
62  */
63 
64 #include "pf.h"
65 #include "carp.h"
66 
67 #include <sys/param.h>
68 #include <sys/systm.h>
69 #include <sys/mbuf.h>
70 #include <sys/domain.h>
71 #include <sys/sysctl.h>
72 #include <sys/protosw.h>
73 #include <sys/socket.h>
74 #include <sys/socketvar.h>
75 #include <sys/errno.h>
76 #include <sys/time.h>
77 #include <sys/timeout.h>
78 #include <sys/kernel.h>
79 #include <sys/syslog.h>
80 #include <sys/task.h>
81 
82 #include <net/if.h>
83 #include <net/if_var.h>
84 #include <net/if_types.h>
85 #include <net/route.h>
86 #include <net/netisr.h>
87 
88 #include <netinet/in.h>
89 
90 #include <netinet/ip.h>
91 
92 #include <netinet/in_pcb.h>
93 #include <netinet/ip_var.h>
94 #include <netinet6/in6_var.h>
95 #include <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 cpumem *ip6counters;
121 
122 void ip6_ours(struct mbuf *);
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 
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 protosw *pr;
142 	int i;
143 
144 	pr = 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 = inet6domain.dom_protosw;
150 	    pr < 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 	ip6counters = counters_alloc(ip6s_ncounters);
163 }
164 
165 /*
166  * IP6 input interrupt handling. Just pass the packet to ip6_input.
167  */
168 void
169 ip6intr(void)
170 {
171 	struct mbuf *m;
172 
173 	while ((m = niq_dequeue(&ip6intrq)) != NULL)
174 		ip6_input(m);
175 }
176 
177 void
178 ip6_input(struct mbuf *m)
179 {
180 	struct ifnet *ifp;
181 	struct ip6_hdr *ip6;
182 	struct sockaddr_in6 sin6;
183 	struct rtentry *rt = NULL;
184 	int off, nxt, ours = 0;
185 	u_int16_t src_scope, dst_scope;
186 #if NPF > 0
187 	struct in6_addr odst;
188 #endif
189 	int srcrt = 0;
190 
191 	ifp = if_get(m->m_pkthdr.ph_ifidx);
192 	if (ifp == NULL)
193 		goto bad;
194 
195 	ip6stat_inc(ip6s_total);
196 
197 	if (m->m_len < sizeof(struct ip6_hdr)) {
198 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
199 			ip6stat_inc(ip6s_toosmall);
200 			goto out;
201 		}
202 	}
203 
204 	ip6 = mtod(m, struct ip6_hdr *);
205 
206 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
207 		ip6stat_inc(ip6s_badvers);
208 		goto bad;
209 	}
210 
211 #if NCARP > 0
212 	if (ifp->if_type == IFT_CARP && ip6->ip6_nxt != IPPROTO_ICMPV6 &&
213 	    carp_lsdrop(m, AF_INET6, ip6->ip6_src.s6_addr32,
214 	    ip6->ip6_dst.s6_addr32))
215 		goto bad;
216 #endif
217 	ip6stat_inc(ip6s_nxthist + ip6->ip6_nxt);
218 
219 	/*
220 	 * Check against address spoofing/corruption.
221 	 */
222 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
223 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
224 		/*
225 		 * XXX: "badscope" is not very suitable for a multicast source.
226 		 */
227 		ip6stat_inc(ip6s_badscope);
228 		goto bad;
229 	}
230 	if ((IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
231 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) &&
232 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
233 		    ip6stat_inc(ip6s_badscope);
234 		    goto bad;
235 	}
236 	/* Drop packets if interface ID portion is already filled. */
237 	if (((IN6_IS_SCOPE_EMBED(&ip6->ip6_src) && ip6->ip6_src.s6_addr16[1]) ||
238 	    (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst) && ip6->ip6_dst.s6_addr16[1])) &&
239 	    (ifp->if_flags & IFF_LOOPBACK) == 0) {
240 		ip6stat_inc(ip6s_badscope);
241 		goto bad;
242 	}
243 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
244 	    !(m->m_flags & M_LOOP)) {
245 		/*
246 		 * In this case, the packet should come from the loopback
247 		 * interface.  However, we cannot just check the if_flags,
248 		 * because ip6_mloopback() passes the "actual" interface
249 		 * as the outgoing/incoming interface.
250 		 */
251 		ip6stat_inc(ip6s_badscope);
252 		goto bad;
253 	}
254 
255 	/*
256 	 * The following check is not documented in specs.  A malicious
257 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
258 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
259 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
260 	 *
261 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
262 	 * support IPv4-less kernel compilation, we cannot support SIIT
263 	 * environment at all.  So, it makes more sense for us to reject any
264 	 * malicious packets for non-SIIT environment, than try to do a
265 	 * partial support for SIIT environment.
266 	 */
267 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
268 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
269 		ip6stat_inc(ip6s_badscope);
270 		goto bad;
271 	}
272 
273 	/*
274 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
275 	 *
276 	 * The code forbids automatic tunneling as per RFC4213.
277 	 */
278 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
279 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
280 		ip6stat_inc(ip6s_badscope);
281 		goto bad;
282 	}
283 
284 	/*
285 	 * If the packet has been received on a loopback interface it
286 	 * can be destinated to any local address, not necessarily to
287 	 * an address configured on `ifp'.
288 	 */
289 	if (ifp->if_flags & IFF_LOOPBACK) {
290 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src)) {
291 			src_scope = ip6->ip6_src.s6_addr16[1];
292 			ip6->ip6_src.s6_addr16[1] = 0;
293 		}
294 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst)) {
295 			dst_scope = ip6->ip6_dst.s6_addr16[1];
296 			ip6->ip6_dst.s6_addr16[1] = 0;
297 		}
298 	}
299 
300 #if NPF > 0
301         /*
302          * Packet filter
303          */
304 	odst = ip6->ip6_dst;
305 	if (pf_test(AF_INET6, PF_IN, ifp, &m) != PF_PASS)
306 		goto bad;
307 	if (m == NULL)
308 		goto bad;
309 
310 	ip6 = mtod(m, struct ip6_hdr *);
311 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
312 #endif
313 
314 	/*
315 	 * Without embedded scope ID we cannot find link-local
316 	 * addresses in the routing table.
317 	 */
318 	if (ifp->if_flags & IFF_LOOPBACK) {
319 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
320 			ip6->ip6_src.s6_addr16[1] = src_scope;
321 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
322 			ip6->ip6_dst.s6_addr16[1] = dst_scope;
323 	} else {
324 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
325 			ip6->ip6_src.s6_addr16[1] = htons(ifp->if_index);
326 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
327 			ip6->ip6_dst.s6_addr16[1] = htons(ifp->if_index);
328 	}
329 
330 	/*
331 	 * Be more secure than RFC5095 and scan for type 0 routing headers.
332 	 * If pf has already scanned the header chain, do not do it twice.
333 	 */
334 	if (!(m->m_pkthdr.pf.flags & PF_TAG_PROCESSED) &&
335 	    ip6_check_rh0hdr(m, &off)) {
336 		ip6stat_inc(ip6s_badoptions);
337 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_HEADER, off);
338 		goto out;
339 	}
340 
341 	if (IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
342 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) {
343 		ip6_ours(m);
344 		goto out;
345 	}
346 
347 #if NPF > 0
348 	if (pf_ouraddr(m) == 1) {
349 		ip6_ours(m);
350 		goto out;
351 	}
352 #endif
353 
354 	/*
355 	 * Multicast check
356 	 */
357 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
358 		/*
359 		 * Make sure M_MCAST is set.  It should theoretically
360 		 * already be there, but let's play safe because upper
361 		 * layers check for this flag.
362 		 */
363 		m->m_flags |= M_MCAST;
364 
365 		/*
366 		 * See if we belong to the destination multicast group on the
367 		 * arrival interface.
368 		 */
369 		if (in6_hasmulti(&ip6->ip6_dst, ifp))
370 			ours = 1;
371 
372 #ifdef MROUTING
373 		if (ip6_mforwarding && ip6_mrouter) {
374 			if (ip6_hbhchcheck(m, &off, &nxt, &ours))
375 				goto out;
376 
377 			ip6 = mtod(m, struct ip6_hdr *);
378 
379 			/*
380 			 * If we are acting as a multicast router, all
381 			 * incoming multicast packets are passed to the
382 			 * kernel-level multicast forwarding function.
383 			 * The packet is returned (relatively) intact; if
384 			 * ip6_mforward() returns a non-zero value, the packet
385 			 * must be discarded, else it may be accepted below.
386 			 */
387 			KERNEL_LOCK();
388 			if (ip6_mforward(ip6, ifp, m)) {
389 				ip6stat_inc(ip6s_cantforward);
390 				m_freem(m);
391 			} else if (ours) {
392 				ip6_local(m, off, nxt);
393 			} else {
394 				m_freem(m);
395 			}
396 			KERNEL_UNLOCK();
397 			goto out;
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 		ip6_ours(m);
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 		if (ia6->ia6_flags & IN6_IFF_ANYCAST)
428 			m->m_flags |= M_ACAST;
429 		/*
430 		 * packets to a tentative, duplicated, or somehow invalid
431 		 * address must not be accepted.
432 		 */
433 		if ((ia6->ia6_flags & (IN6_IFF_TENTATIVE|IN6_IFF_DUPLICATED))) {
434 			char src[INET6_ADDRSTRLEN], dst[INET6_ADDRSTRLEN];
435 
436 			inet_ntop(AF_INET6, &ip6->ip6_src, src, sizeof(src));
437 			inet_ntop(AF_INET6, &ip6->ip6_dst, dst, sizeof(dst));
438 			/* address is not ready, so discard the packet. */
439 			nd6log((LOG_INFO,
440 			    "ip6_input: packet to an unready address %s->%s\n",
441 			    src, dst));
442 
443 			goto bad;
444 		} else {
445 			ip6_ours(m);
446 			goto out;
447 		}
448 	}
449 
450 #if NCARP > 0
451 	if (ifp->if_type == IFT_CARP && ip6->ip6_nxt == IPPROTO_ICMPV6 &&
452 	    carp_lsdrop(m, AF_INET6, ip6->ip6_src.s6_addr32,
453 	    ip6->ip6_dst.s6_addr32))
454 		goto bad;
455 #endif
456 	/*
457 	 * Now there is no reason to process the packet if it's not our own
458 	 * and we're not a router.
459 	 */
460 	if (!ip6_forwarding) {
461 		ip6stat_inc(ip6s_cantforward);
462 		goto bad;
463 	}
464 
465 	if (ip6_hbhchcheck(m, &off, &nxt, &ours))
466 		goto out;
467 
468 	if (ours) {
469 		KERNEL_LOCK();
470 		ip6_local(m, off, nxt);
471 		KERNEL_UNLOCK();
472 		goto out;
473 	}
474 
475 	ip6_forward(m, rt, srcrt);
476 	if_put(ifp);
477 	return;
478  bad:
479 	m_freem(m);
480  out:
481 	rtfree(rt);
482 	if_put(ifp);
483 }
484 
485 void
486 ip6_ours(struct mbuf *m)
487 {
488 	int off, nxt;
489 
490 	if (ip6_hbhchcheck(m, &off, &nxt, NULL))
491 		return;
492 
493 	ip6_local(m, off, nxt);
494 }
495 
496 void
497 ip6_local(struct mbuf *m, int off, int nxt)
498 {
499 	int nest = 0;
500 
501 	KERNEL_ASSERT_LOCKED();
502 
503 	/* pf might have changed things */
504 	in6_proto_cksum_out(m, NULL);
505 
506 	/*
507 	 * Tell launch routine the next header
508 	 */
509 	ip6stat_inc(ip6s_delivered);
510 
511 	while (nxt != IPPROTO_DONE) {
512 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
513 			ip6stat_inc(ip6s_toomanyhdr);
514 			goto bad;
515 		}
516 
517 		/*
518 		 * protection against faulty packet - there should be
519 		 * more sanity checks in header chain processing.
520 		 */
521 		if (m->m_pkthdr.len < off) {
522 			ip6stat_inc(ip6s_tooshort);
523 			goto bad;
524 		}
525 
526 		/* draft-itojun-ipv6-tcp-to-anycast */
527 		if (ISSET(m->m_flags, M_ACAST) && (nxt == IPPROTO_TCP)) {
528 			if (m->m_len >= sizeof(struct ip6_hdr)) {
529 				icmp6_error(m, ICMP6_DST_UNREACH,
530 					ICMP6_DST_UNREACH_ADDR,
531 					offsetof(struct ip6_hdr, ip6_dst));
532 				break;
533 			} else
534 				goto bad;
535 		}
536 
537 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
538 	}
539 	return;
540  bad:
541 	m_freem(m);
542 }
543 
544 int
545 ip6_hbhchcheck(struct mbuf *m, int *offp, int *nxtp, int *oursp)
546 {
547 	struct ip6_hdr *ip6;
548 	u_int32_t plen, rtalert = ~0;
549 
550 	ip6 = mtod(m, struct ip6_hdr *);
551 
552 	/*
553 	 * Process Hop-by-Hop options header if it's contained.
554 	 * m may be modified in ip6_hopopts_input().
555 	 * If a JumboPayload option is included, plen will also be modified.
556 	 */
557 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
558 	*offp = sizeof(struct ip6_hdr);
559 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
560 		struct ip6_hbh *hbh;
561 
562 		if (ip6_hopopts_input(&plen, &rtalert, &m, offp)) {
563 			return (-1);	/* m have already been freed */
564 		}
565 
566 		/* adjust pointer */
567 		ip6 = mtod(m, struct ip6_hdr *);
568 
569 		/*
570 		 * if the payload length field is 0 and the next header field
571 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
572 		 * option MUST be included.
573 		 */
574 		if (ip6->ip6_plen == 0 && plen == 0) {
575 			/*
576 			 * Note that if a valid jumbo payload option is
577 			 * contained, ip6_hopopts_input() must set a valid
578 			 * (non-zero) payload length to the variable plen.
579 			 */
580 			ip6stat_inc(ip6s_badoptions);
581 			icmp6_error(m, ICMP6_PARAM_PROB,
582 				    ICMP6_PARAMPROB_HEADER,
583 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
584 			return (-1);
585 		}
586 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
587 			sizeof(struct ip6_hbh));
588 		if (hbh == NULL) {
589 			ip6stat_inc(ip6s_tooshort);
590 			return (-1);
591 		}
592 		*nxtp = hbh->ip6h_nxt;
593 
594 		/*
595 		 * accept the packet if a router alert option is included
596 		 * and we act as an IPv6 router.
597 		 */
598 		if (rtalert != ~0 && ip6_forwarding && oursp != NULL)
599 			*oursp = 1;
600 	} else
601 		*nxtp = ip6->ip6_nxt;
602 
603 	/*
604 	 * Check that the amount of data in the buffers
605 	 * is as at least much as the IPv6 header would have us expect.
606 	 * Trim mbufs if longer than we expect.
607 	 * Drop packet if shorter than we expect.
608 	 */
609 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
610 		ip6stat_inc(ip6s_tooshort);
611 		m_freem(m);
612 		return (-1);
613 	}
614 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
615 		if (m->m_len == m->m_pkthdr.len) {
616 			m->m_len = sizeof(struct ip6_hdr) + plen;
617 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
618 		} else {
619 			m_adj(m,
620 			    sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
621 		}
622 	}
623 
624 	return (0);
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 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 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 bcopy().
816 			 */
817 			bcopy(opt + 2, &jumboplen, 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 #ifdef SO_TIMESTAMP
926 	if (in6p->inp_socket->so_options & SO_TIMESTAMP) {
927 		struct timeval tv;
928 
929 		microtime(&tv);
930 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
931 		    SCM_TIMESTAMP, SOL_SOCKET);
932 		if (*mp)
933 			mp = &(*mp)->m_next;
934 	}
935 #endif
936 
937 	/* RFC 2292 sec. 5 */
938 	if ((in6p->inp_flags & IN6P_PKTINFO) != 0) {
939 		struct in6_pktinfo pi6;
940 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
941 		if (IN6_IS_SCOPE_EMBED(&pi6.ipi6_addr))
942 			pi6.ipi6_addr.s6_addr16[1] = 0;
943 		pi6.ipi6_ifindex = m ? m->m_pkthdr.ph_ifidx : 0;
944 		*mp = sbcreatecontrol((caddr_t) &pi6,
945 		    sizeof(struct in6_pktinfo),
946 		    IPV6_PKTINFO, IPPROTO_IPV6);
947 		if (*mp)
948 			mp = &(*mp)->m_next;
949 	}
950 
951 	if ((in6p->inp_flags & IN6P_HOPLIMIT) != 0) {
952 		int hlim = ip6->ip6_hlim & 0xff;
953 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
954 		    IPV6_HOPLIMIT, IPPROTO_IPV6);
955 		if (*mp)
956 			mp = &(*mp)->m_next;
957 	}
958 
959 	if ((in6p->inp_flags & IN6P_TCLASS) != 0) {
960 		u_int32_t flowinfo;
961 		int tclass;
962 
963 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
964 		flowinfo >>= 20;
965 
966 		tclass = flowinfo & 0xff;
967 		*mp = sbcreatecontrol((caddr_t)&tclass, sizeof(tclass),
968 		    IPV6_TCLASS, IPPROTO_IPV6);
969 		if (*mp)
970 			mp = &(*mp)->m_next;
971 	}
972 
973 	/*
974 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
975 	 * privilege for the option (see ip6_ctloutput), but it might be too
976 	 * strict, since there might be some hop-by-hop options which can be
977 	 * returned to normal user.
978 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
979 	 */
980 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
981 		/*
982 		 * Check if a hop-by-hop options header is contained in the
983 		 * received packet, and if so, store the options as ancillary
984 		 * data. Note that a hop-by-hop options header must be
985 		 * just after the IPv6 header, which is assured through the
986 		 * IPv6 input processing.
987 		 */
988 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
989 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
990 			struct ip6_hbh *hbh;
991 			int hbhlen = 0;
992 			struct mbuf *ext;
993 
994 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
995 			    ip6->ip6_nxt);
996 			if (ext == NULL) {
997 				ip6stat_inc(ip6s_tooshort);
998 				return;
999 			}
1000 			hbh = mtod(ext, struct ip6_hbh *);
1001 			hbhlen = (hbh->ip6h_len + 1) << 3;
1002 			if (hbhlen != ext->m_len) {
1003 				m_freem(ext);
1004 				ip6stat_inc(ip6s_tooshort);
1005 				return;
1006 			}
1007 
1008 			/*
1009 			 * XXX: We copy the whole header even if a
1010 			 * jumbo payload option is included, the option which
1011 			 * is to be removed before returning according to
1012 			 * RFC2292.
1013 			 * Note: this constraint is removed in RFC3542.
1014 			 */
1015 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1016 			    IPV6_HOPOPTS,
1017 			    IPPROTO_IPV6);
1018 			if (*mp)
1019 				mp = &(*mp)->m_next;
1020 			m_freem(ext);
1021 		}
1022 	}
1023 
1024 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1025 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1026 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1027 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1028 
1029 		/*
1030 		 * Search for destination options headers or routing
1031 		 * header(s) through the header chain, and stores each
1032 		 * header as ancillary data.
1033 		 * Note that the order of the headers remains in
1034 		 * the chain of ancillary data.
1035 		 */
1036 		while (1) {	/* is explicit loop prevention necessary? */
1037 			struct ip6_ext *ip6e = NULL;
1038 			int elen;
1039 			struct mbuf *ext = NULL;
1040 
1041 			/*
1042 			 * if it is not an extension header, don't try to
1043 			 * pull it from the chain.
1044 			 */
1045 			switch (nxt) {
1046 			case IPPROTO_DSTOPTS:
1047 			case IPPROTO_ROUTING:
1048 			case IPPROTO_HOPOPTS:
1049 			case IPPROTO_AH: /* is it possible? */
1050 				break;
1051 			default:
1052 				goto loopend;
1053 			}
1054 
1055 			ext = ip6_pullexthdr(m, off, nxt);
1056 			if (ext == NULL) {
1057 				ip6stat_inc(ip6s_tooshort);
1058 				return;
1059 			}
1060 			ip6e = mtod(ext, struct ip6_ext *);
1061 			if (nxt == IPPROTO_AH)
1062 				elen = (ip6e->ip6e_len + 2) << 2;
1063 			else
1064 				elen = (ip6e->ip6e_len + 1) << 3;
1065 			if (elen != ext->m_len) {
1066 				m_freem(ext);
1067 				ip6stat_inc(ip6s_tooshort);
1068 				return;
1069 			}
1070 
1071 			switch (nxt) {
1072 			case IPPROTO_DSTOPTS:
1073 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1074 					break;
1075 
1076 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1077 				    IPV6_DSTOPTS,
1078 				    IPPROTO_IPV6);
1079 				if (*mp)
1080 					mp = &(*mp)->m_next;
1081 				break;
1082 
1083 			case IPPROTO_ROUTING:
1084 				if (!(in6p->inp_flags & IN6P_RTHDR))
1085 					break;
1086 
1087 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1088 				    IPV6_RTHDR,
1089 				    IPPROTO_IPV6);
1090 				if (*mp)
1091 					mp = &(*mp)->m_next;
1092 				break;
1093 
1094 			case IPPROTO_HOPOPTS:
1095 			case IPPROTO_AH: /* is it possible? */
1096 				break;
1097 
1098 			default:
1099 				/*
1100 				 * other cases have been filtered in the above.
1101 				 * none will visit this case.  here we supply
1102 				 * the code just in case (nxt overwritten or
1103 				 * other cases).
1104 				 */
1105 				m_freem(ext);
1106 				goto loopend;
1107 
1108 			}
1109 
1110 			/* proceed with the next header. */
1111 			off += elen;
1112 			nxt = ip6e->ip6e_nxt;
1113 			ip6e = NULL;
1114 			m_freem(ext);
1115 			ext = NULL;
1116 		}
1117 loopend:
1118 		;
1119 	}
1120 }
1121 
1122 /*
1123  * pull single extension header from mbuf chain.  returns single mbuf that
1124  * contains the result, or NULL on error.
1125  */
1126 struct mbuf *
1127 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1128 {
1129 	struct ip6_ext ip6e;
1130 	size_t elen;
1131 	struct mbuf *n;
1132 
1133 #ifdef DIAGNOSTIC
1134 	switch (nxt) {
1135 	case IPPROTO_DSTOPTS:
1136 	case IPPROTO_ROUTING:
1137 	case IPPROTO_HOPOPTS:
1138 	case IPPROTO_AH: /* is it possible? */
1139 		break;
1140 	default:
1141 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1142 	}
1143 #endif
1144 
1145 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1146 	if (nxt == IPPROTO_AH)
1147 		elen = (ip6e.ip6e_len + 2) << 2;
1148 	else
1149 		elen = (ip6e.ip6e_len + 1) << 3;
1150 
1151 	MGET(n, M_DONTWAIT, MT_DATA);
1152 	if (n && elen >= MLEN) {
1153 		MCLGET(n, M_DONTWAIT);
1154 		if ((n->m_flags & M_EXT) == 0) {
1155 			m_free(n);
1156 			n = NULL;
1157 		}
1158 	}
1159 	if (!n)
1160 		return NULL;
1161 
1162 	n->m_len = 0;
1163 	if (elen >= M_TRAILINGSPACE(n)) {
1164 		m_free(n);
1165 		return NULL;
1166 	}
1167 
1168 	m_copydata(m, off, elen, mtod(n, caddr_t));
1169 	n->m_len = elen;
1170 	return n;
1171 }
1172 
1173 /*
1174  * Get pointer to the previous header followed by the header
1175  * currently processed.
1176  * XXX: This function supposes that
1177  *	M includes all headers,
1178  *	the next header field and the header length field of each header
1179  *	are valid, and
1180  *	the sum of each header length equals to OFF.
1181  * Because of these assumptions, this function must be called very
1182  * carefully. Moreover, it will not be used in the near future when
1183  * we develop `neater' mechanism to process extension headers.
1184  */
1185 u_int8_t *
1186 ip6_get_prevhdr(struct mbuf *m, int off)
1187 {
1188 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1189 
1190 	if (off == sizeof(struct ip6_hdr))
1191 		return (&ip6->ip6_nxt);
1192 	else {
1193 		int len, nxt;
1194 		struct ip6_ext *ip6e = NULL;
1195 
1196 		nxt = ip6->ip6_nxt;
1197 		len = sizeof(struct ip6_hdr);
1198 		while (len < off) {
1199 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1200 
1201 			switch (nxt) {
1202 			case IPPROTO_FRAGMENT:
1203 				len += sizeof(struct ip6_frag);
1204 				break;
1205 			case IPPROTO_AH:
1206 				len += (ip6e->ip6e_len + 2) << 2;
1207 				break;
1208 			default:
1209 				len += (ip6e->ip6e_len + 1) << 3;
1210 				break;
1211 			}
1212 			nxt = ip6e->ip6e_nxt;
1213 		}
1214 		if (ip6e)
1215 			return (&ip6e->ip6e_nxt);
1216 		else
1217 			return NULL;
1218 	}
1219 }
1220 
1221 /*
1222  * get next header offset.  m will be retained.
1223  */
1224 int
1225 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1226 {
1227 	struct ip6_hdr ip6;
1228 	struct ip6_ext ip6e;
1229 	struct ip6_frag fh;
1230 
1231 	/* just in case */
1232 	if (m == NULL)
1233 		panic("ip6_nexthdr: m == NULL");
1234 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1235 		return -1;
1236 
1237 	switch (proto) {
1238 	case IPPROTO_IPV6:
1239 		if (m->m_pkthdr.len < off + sizeof(ip6))
1240 			return -1;
1241 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1242 		if (nxtp)
1243 			*nxtp = ip6.ip6_nxt;
1244 		off += sizeof(ip6);
1245 		return off;
1246 
1247 	case IPPROTO_FRAGMENT:
1248 		/*
1249 		 * terminate parsing if it is not the first fragment,
1250 		 * it does not make sense to parse through it.
1251 		 */
1252 		if (m->m_pkthdr.len < off + sizeof(fh))
1253 			return -1;
1254 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1255 		if ((fh.ip6f_offlg & IP6F_OFF_MASK) != 0)
1256 			return -1;
1257 		if (nxtp)
1258 			*nxtp = fh.ip6f_nxt;
1259 		off += sizeof(struct ip6_frag);
1260 		return off;
1261 
1262 	case IPPROTO_AH:
1263 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1264 			return -1;
1265 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1266 		if (nxtp)
1267 			*nxtp = ip6e.ip6e_nxt;
1268 		off += (ip6e.ip6e_len + 2) << 2;
1269 		if (m->m_pkthdr.len < off)
1270 			return -1;
1271 		return off;
1272 
1273 	case IPPROTO_HOPOPTS:
1274 	case IPPROTO_ROUTING:
1275 	case IPPROTO_DSTOPTS:
1276 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1277 			return -1;
1278 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1279 		if (nxtp)
1280 			*nxtp = ip6e.ip6e_nxt;
1281 		off += (ip6e.ip6e_len + 1) << 3;
1282 		if (m->m_pkthdr.len < off)
1283 			return -1;
1284 		return off;
1285 
1286 	case IPPROTO_NONE:
1287 	case IPPROTO_ESP:
1288 	case IPPROTO_IPCOMP:
1289 		/* give up */
1290 		return -1;
1291 
1292 	default:
1293 		return -1;
1294 	}
1295 
1296 	return -1;
1297 }
1298 
1299 /*
1300  * get offset for the last header in the chain.  m will be kept untainted.
1301  */
1302 int
1303 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1304 {
1305 	int newoff;
1306 	int nxt;
1307 
1308 	if (!nxtp) {
1309 		nxt = -1;
1310 		nxtp = &nxt;
1311 	}
1312 	while (1) {
1313 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1314 		if (newoff < 0)
1315 			return off;
1316 		else if (newoff < off)
1317 			return -1;	/* invalid */
1318 		else if (newoff == off)
1319 			return newoff;
1320 
1321 		off = newoff;
1322 		proto = *nxtp;
1323 	}
1324 }
1325 
1326 /*
1327  * System control for IP6
1328  */
1329 
1330 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1331 	0,		0,		0,		0,
1332 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1333 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1334 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1335 	0,		0,		0,		0,
1336 	ENOPROTOOPT
1337 };
1338 
1339 int *ipv6ctl_vars[IPV6CTL_MAXID] = IPV6CTL_VARS;
1340 
1341 int
1342 ip6_sysctl_ip6stat(void *oldp, size_t *oldlenp, void *newp)
1343 {
1344 	struct ip6stat *ip6stat;
1345 	int ret;
1346 
1347 	CTASSERT(sizeof(*ip6stat) == (ip6s_ncounters * sizeof(uint64_t)));
1348 
1349 	ip6stat = malloc(sizeof(*ip6stat), M_TEMP, M_WAITOK);
1350 	counters_read(ip6counters, (uint64_t *)ip6stat, ip6s_ncounters);
1351 	ret = sysctl_rdstruct(oldp, oldlenp, newp,
1352 	    ip6stat, sizeof(*ip6stat));
1353 	free(ip6stat, M_TEMP, sizeof(*ip6stat));
1354 
1355 	return (ret);
1356 }
1357 
1358 int
1359 ip6_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
1360     void *newp, size_t newlen)
1361 {
1362 #ifdef MROUTING
1363 	extern int ip6_mrtproto;
1364 	extern struct mrt6stat mrt6stat;
1365 #endif
1366 	int error;
1367 
1368 	NET_ASSERT_LOCKED();
1369 
1370 	/* Almost all sysctl names at this level are terminal. */
1371 	if (namelen != 1 && name[0] != IPV6CTL_IFQUEUE)
1372 		return (ENOTDIR);
1373 
1374 	switch (name[0]) {
1375 	case IPV6CTL_DAD_PENDING:
1376 		return sysctl_rdint(oldp, oldlenp, newp, ip6_dad_pending);
1377 	case IPV6CTL_STATS:
1378 		return (ip6_sysctl_ip6stat(oldp, oldlenp, newp));
1379 #ifdef MROUTING
1380 	case IPV6CTL_MRTSTATS:
1381 		if (newp != NULL)
1382 			return (EPERM);
1383 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
1384 		    &mrt6stat, sizeof(mrt6stat)));
1385 	case IPV6CTL_MRTPROTO:
1386 		return sysctl_rdint(oldp, oldlenp, newp, ip6_mrtproto);
1387 	case IPV6CTL_MRTMIF:
1388 		if (newp)
1389 			return (EPERM);
1390 		return mrt6_sysctl_mif(oldp, oldlenp);
1391 	case IPV6CTL_MRTMFC:
1392 		if (newp)
1393 			return (EPERM);
1394 		return mrt6_sysctl_mfc(oldp, oldlenp);
1395 #else
1396 	case IPV6CTL_MRTSTATS:
1397 	case IPV6CTL_MRTPROTO:
1398 	case IPV6CTL_MRTMIF:
1399 	case IPV6CTL_MRTMFC:
1400 		return (EOPNOTSUPP);
1401 #endif
1402 	case IPV6CTL_MTUDISCTIMEOUT:
1403 		error = sysctl_int(oldp, oldlenp, newp, newlen,
1404 		   &ip6_mtudisc_timeout);
1405 		if (icmp6_mtudisc_timeout_q != NULL)
1406 			rt_timer_queue_change(icmp6_mtudisc_timeout_q,
1407 					      ip6_mtudisc_timeout);
1408 		return (error);
1409 	case IPV6CTL_IFQUEUE:
1410 		return (sysctl_niq(name + 1, namelen - 1,
1411 		    oldp, oldlenp, newp, newlen, &ip6intrq));
1412 	default:
1413 		if (name[0] < IPV6CTL_MAXID)
1414 			return (sysctl_int_arr(ipv6ctl_vars, name, namelen,
1415 			    oldp, oldlenp, newp, newlen));
1416 		return (EOPNOTSUPP);
1417 	}
1418 	/* NOTREACHED */
1419 }
1420 
1421 void
1422 ip6_send_dispatch(void *xmq)
1423 {
1424 	struct mbuf_queue *mq = xmq;
1425 	struct mbuf *m;
1426 	struct mbuf_list ml;
1427 	int s;
1428 
1429 	mq_delist(mq, &ml);
1430 	if (ml_empty(&ml))
1431 		return;
1432 
1433 	KERNEL_LOCK();
1434 	NET_LOCK(s);
1435 	while ((m = ml_dequeue(&ml)) != NULL) {
1436 		ip6_output(m, NULL, NULL, IPV6_MINMTU, NULL, NULL);
1437 	}
1438 	NET_UNLOCK(s);
1439 	KERNEL_UNLOCK();
1440 }
1441 
1442 void
1443 ip6_send(struct mbuf *m)
1444 {
1445 	mq_enqueue(&ip6send_mq, m);
1446 	task_add(softnettq, &ip6send_task);
1447 }
1448