xref: /freebsd/sys/netinet6/ip6_input.c (revision 055aefb1bcd6c54859c45274c8e03f03b3f5e681)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ip6_input.c,v 1.95 2000/07/02 07:49:37 jinmei 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. All advertising materials mentioning features or use of this software
46  *    must display the following acknowledgement:
47  *	This product includes software developed by the University of
48  *	California, Berkeley and its contributors.
49  * 4. Neither the name of the University nor the names of its contributors
50  *    may be used to endorse or promote products derived from this software
51  *    without specific prior written permission.
52  *
53  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
54  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
55  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
56  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
57  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
58  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
59  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
61  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
62  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
63  * SUCH DAMAGE.
64  *
65  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
66  */
67 
68 #include "opt_ip6fw.h"
69 #include "opt_inet.h"
70 #include "opt_inet6.h"
71 #include "opt_ipsec.h"
72 #include "opt_pfil_hooks.h"
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/mbuf.h>
77 #include <sys/domain.h>
78 #include <sys/protosw.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/errno.h>
82 #include <sys/time.h>
83 #include <sys/kernel.h>
84 #include <sys/syslog.h>
85 #include <sys/proc.h>
86 
87 #include <net/if.h>
88 #include <net/if_types.h>
89 #include <net/if_dl.h>
90 #include <net/route.h>
91 #include <net/netisr.h>
92 #include <net/intrq.h>
93 #ifdef PFIL_HOOKS
94 #include <net/pfil.h>
95 #endif
96 
97 #include <netinet/in.h>
98 #include <netinet/in_systm.h>
99 #ifdef INET
100 #include <netinet/ip.h>
101 #include <netinet/ip_icmp.h>
102 #endif /*INET*/
103 #include <netinet/ip6.h>
104 #include <netinet6/in6_var.h>
105 #include <netinet6/ip6_var.h>
106 #include <netinet/in_pcb.h>
107 #include <netinet/icmp6.h>
108 #include <netinet6/in6_ifattach.h>
109 #include <netinet6/nd6.h>
110 #include <netinet6/in6_prefix.h>
111 
112 #ifdef IPV6FIREWALL
113 #include <netinet6/ip6_fw.h>
114 #endif
115 
116 #include <netinet6/ip6protosw.h>
117 
118 /* we need it for NLOOP. */
119 #include "loop.h"
120 #include "faith.h"
121 #include "gif.h"
122 
123 #include <net/net_osdep.h>
124 
125 extern struct domain inet6domain;
126 extern struct ip6protosw inet6sw[];
127 
128 u_char ip6_protox[IPPROTO_MAX];
129 static int ip6qmaxlen = IFQ_MAXLEN;
130 struct in6_ifaddr *in6_ifaddr;
131 
132 int ip6_forward_srcrt;			/* XXX */
133 int ip6_sourcecheck;			/* XXX */
134 int ip6_sourcecheck_interval;		/* XXX */
135 const int int6intrq_present = 1;
136 
137 #ifdef IPV6FIREWALL
138 /* firewall hooks */
139 ip6_fw_chk_t *ip6_fw_chk_ptr;
140 ip6_fw_ctl_t *ip6_fw_ctl_ptr;
141 #endif
142 
143 struct ip6stat ip6stat;
144 
145 static void ip6_init2 __P((void *));
146 
147 static int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
148 #ifdef PULLDOWN_TEST
149 static struct mbuf *ip6_pullexthdr __P((struct mbuf *, size_t, int));
150 #endif
151 
152 /*
153  * IP6 initialization: fill in IP6 protocol switch table.
154  * All protocols not implemented in kernel go to raw IP6 protocol handler.
155  */
156 void
157 ip6_init()
158 {
159 	register struct ip6protosw *pr;
160 	register int i;
161 	struct timeval tv;
162 
163 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
164 	if (pr == 0)
165 		panic("ip6_init");
166 	for (i = 0; i < IPPROTO_MAX; i++)
167 		ip6_protox[i] = pr - inet6sw;
168 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
169 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
170 		if (pr->pr_domain->dom_family == PF_INET6 &&
171 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
172 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
173 	ip6intrq.ifq_maxlen = ip6qmaxlen;
174 	register_netisr(NETISR_IPV6, ip6intr);
175 	nd6_init();
176 	frag6_init();
177 #ifdef IPV6FIREWALL
178 	ip6_fw_init();
179 #endif
180 	/*
181 	 * in many cases, random() here does NOT return random number
182 	 * as initialization during bootstrap time occur in fixed order.
183 	 */
184 	microtime(&tv);
185 	ip6_flow_seq = random() ^ tv.tv_usec;
186 }
187 
188 static void
189 ip6_init2(dummy)
190 	void *dummy;
191 {
192 
193 	/*
194 	 * to route local address of p2p link to loopback,
195 	 * assign loopback address first.
196 	 */
197 	in6_ifattach(&loif[0], NULL);
198 
199 	/* nd6_timer_init */
200 	timeout(nd6_timer, (caddr_t)0, hz);
201 	/* router renumbering prefix list maintenance */
202 	timeout(in6_rr_timer, (caddr_t)0, hz);
203 }
204 
205 /* cheat */
206 /* This must be after route_init(), which is now SI_ORDER_THIRD */
207 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
208 
209 /*
210  * IP6 input interrupt handling. Just pass the packet to ip6_input.
211  */
212 void
213 ip6intr()
214 {
215 	int s;
216 	struct mbuf *m;
217 
218 	for (;;) {
219 		s = splimp();
220 		IF_DEQUEUE(&ip6intrq, m);
221 		splx(s);
222 		if (m == 0)
223 			return;
224 		ip6_input(m);
225 	}
226 }
227 
228 extern struct	route_in6 ip6_forward_rt;
229 
230 void
231 ip6_input(m)
232 	struct mbuf *m;
233 {
234 	struct ip6_hdr *ip6;
235 	int off = sizeof(struct ip6_hdr), nest;
236 	u_int32_t plen;
237 	u_int32_t rtalert = ~0;
238 	int nxt, ours = 0;
239 	struct ifnet *deliverifp = NULL;
240 #ifdef  PFIL_HOOKS
241 	struct packet_filter_hook *pfh;
242 	struct mbuf *m0;
243 	int rv;
244 #endif  /* PFIL_HOOKS */
245 
246 #ifdef IPSEC
247 	/*
248 	 * should the inner packet be considered authentic?
249 	 * see comment in ah4_input().
250 	 */
251 	if (m) {
252 		m->m_flags &= ~M_AUTHIPHDR;
253 		m->m_flags &= ~M_AUTHIPDGM;
254 	}
255 #endif
256 
257 	/*
258 	 * mbuf statistics by kazu
259 	 */
260 	if (m->m_flags & M_EXT) {
261 		if (m->m_next)
262 			ip6stat.ip6s_mext2m++;
263 		else
264 			ip6stat.ip6s_mext1++;
265 	} else {
266 		if (m->m_next) {
267 			if (m->m_flags & M_LOOP) {
268 				ip6stat.ip6s_m2m[loif[0].if_index]++;	/*XXX*/
269 			} else if (m->m_pkthdr.rcvif->if_index <= 31)
270 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
271 			else
272 				ip6stat.ip6s_m2m[0]++;
273 		} else
274 			ip6stat.ip6s_m1++;
275 	}
276 
277 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
278 	ip6stat.ip6s_total++;
279 
280 #ifndef PULLDOWN_TEST
281 	/* XXX is the line really necessary? */
282 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /*nothing*/);
283 #endif
284 
285 	if (m->m_len < sizeof(struct ip6_hdr)) {
286 		struct ifnet *inifp;
287 		inifp = m->m_pkthdr.rcvif;
288 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == 0) {
289 			ip6stat.ip6s_toosmall++;
290 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
291 			return;
292 		}
293 	}
294 
295 	ip6 = mtod(m, struct ip6_hdr *);
296 
297 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
298 		ip6stat.ip6s_badvers++;
299 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
300 		goto bad;
301 	}
302 
303 #ifdef PFIL_HOOKS
304 	/*
305 	 * Run through list of hooks for input packets.  If there are any
306 	 * filters which require that additional packets in the flow are
307 	 * not fast-forwarded, they must clear the M_CANFASTFWD flag.
308 	 * Note that filters must _never_ set this flag, as another filter
309 	 * in the list may have previously cleared it.
310 	 */
311 	m0 = m;
312 	pfh = pfil_hook_get(PFIL_IN, &inet6sw[ip6_protox[IPPROTO_IPV6]].pr_pfh);
313 	for (; pfh; pfh = pfh->pfil_link.tqe_next)
314 		if (pfh->pfil_func) {
315 			rv = pfh->pfil_func(ip6, sizeof(*ip6),
316 					    m->m_pkthdr.rcvif, 0, &m0);
317 			if (rv)
318 				return;
319 			m = m0;
320 			if (m == NULL)
321 				return;
322 			ip6 = mtod(m, struct ip6_hdr *);
323 		}
324 #endif /* PFIL_HOOKS */
325 
326 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
327 
328 #ifdef IPV6FIREWALL
329 	/*
330 	 * Check with the firewall...
331 	 */
332 	if (ip6_fw_chk_ptr) {
333 		u_short port = 0;
334 		/* If ipfw says divert, we have to just drop packet */
335 		/* use port as a dummy argument */
336 		if ((*ip6_fw_chk_ptr)(&ip6, NULL, &port, &m)) {
337 			m_freem(m);
338 			m = NULL;
339 		}
340 		if (!m)
341 			return;
342 	}
343 #endif
344 
345 	/*
346 	 * Scope check
347 	 */
348 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
349 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
350 		ip6stat.ip6s_badscope++;
351 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
352 		goto bad;
353 	}
354 
355 	/*
356 	 * Don't check IPv4 mapped address here.  SIIT assumes that
357 	 * routers would forward IPv6 native packets with IPv4 mapped
358 	 * address normally.
359 	 */
360 #if 0
361 	/*
362 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
363 	 *
364 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
365 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
366 	 * is revised to forbid relaying case.
367 	 */
368 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
369 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
370 		ip6stat.ip6s_badscope++;
371 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
372 		goto bad;
373 	}
374 #endif
375 	if (IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
376 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) {
377 		if (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) {
378 			ours = 1;
379 			deliverifp = m->m_pkthdr.rcvif;
380 			goto hbhcheck;
381 		} else {
382 			ip6stat.ip6s_badscope++;
383 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
384 			goto bad;
385 		}
386 	}
387 
388 #ifndef FAKE_LOOPBACK_IF
389 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0)
390 #else
391 	if (1)
392 #endif
393 	{
394 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src))
395 			ip6->ip6_src.s6_addr16[1]
396 				= htons(m->m_pkthdr.rcvif->if_index);
397 		if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst))
398 			ip6->ip6_dst.s6_addr16[1]
399 				= htons(m->m_pkthdr.rcvif->if_index);
400 	}
401 
402 	/*
403 	 * XXX we need this since we do not have "goto ours" hack route
404 	 * for some of our ifaddrs on loopback interface.
405 	 * we should correct it by changing in6_ifattach to install
406 	 * "goto ours" hack route.
407 	 */
408 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) != 0) {
409 		if (IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
410 			ours = 1;
411 			deliverifp = m->m_pkthdr.rcvif;
412 			goto hbhcheck;
413 		}
414 	}
415 
416 	/*
417 	 * Multicast check
418 	 */
419 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
420 	  	struct	in6_multi *in6m = 0;
421 
422 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
423 		/*
424 		 * See if we belong to the destination multicast group on the
425 		 * arrival interface.
426 		 */
427 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
428 		if (in6m)
429 			ours = 1;
430 		else if (!ip6_mrouter) {
431 			ip6stat.ip6s_notmember++;
432 			ip6stat.ip6s_cantforward++;
433 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
434 			goto bad;
435 		}
436 		deliverifp = m->m_pkthdr.rcvif;
437 		goto hbhcheck;
438 	}
439 
440 	/*
441 	 *  Unicast check
442 	 */
443 	if (ip6_forward_rt.ro_rt != NULL &&
444 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
445 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
446 			       &ip6_forward_rt.ro_dst.sin6_addr))
447 		ip6stat.ip6s_forward_cachehit++;
448 	else {
449 		if (ip6_forward_rt.ro_rt) {
450 			/* route is down or destination is different */
451 			ip6stat.ip6s_forward_cachemiss++;
452 			RTFREE(ip6_forward_rt.ro_rt);
453 			ip6_forward_rt.ro_rt = 0;
454 		}
455 
456 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
457 		ip6_forward_rt.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
458 		ip6_forward_rt.ro_dst.sin6_family = AF_INET6;
459 		ip6_forward_rt.ro_dst.sin6_addr = ip6->ip6_dst;
460 #ifdef SCOPEDROUTING
461 		ip6_forward_rt.ro_dst.sin6_scope_id =
462 			in6_addr2scopeid(m->m_pkthdr.rcvif, &ip6->ip6_dst);
463 #endif
464 
465 		rtalloc_ign((struct route *)&ip6_forward_rt, RTF_PRCLONING);
466 	}
467 
468 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
469 
470 	/*
471 	 * Accept the packet if the forwarding interface to the destination
472 	 * according to the routing table is the loopback interface,
473 	 * unless the associated route has a gateway.
474 	 * Note that this approach causes to accept a packet if there is a
475 	 * route to the loopback interface for the destination of the packet.
476 	 * But we think it's even useful in some situations, e.g. when using
477 	 * a special daemon which wants to intercept the packet.
478 	 */
479 	if (ip6_forward_rt.ro_rt &&
480 	    (ip6_forward_rt.ro_rt->rt_flags &
481 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
482 #if 0
483 	    /*
484 	     * The check below is redundant since the comparison of
485 	     * the destination and the key of the rtentry has
486 	     * already done through looking up the routing table.
487 	     */
488 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
489 			       &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr) &&
490 #endif
491 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
492 		struct in6_ifaddr *ia6 =
493 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
494 		if (ia6->ia6_flags & IN6_IFF_ANYCAST)
495 			m->m_flags |= M_ANYCAST6;
496 		/*
497 		 * packets to a tentative, duplicated, or somehow invalid
498 		 * address must not be accepted.
499 		 */
500 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
501 			/* this address is ready */
502 			ours = 1;
503 			deliverifp = ia6->ia_ifp;	/* correct? */
504 			goto hbhcheck;
505 		} else {
506 			/* address is not ready, so discard the packet. */
507 			log(LOG_INFO,
508 			    "ip6_input: packet to an unready address %s->%s",
509 			    ip6_sprintf(&ip6->ip6_src),
510 			    ip6_sprintf(&ip6->ip6_dst));
511 
512 			goto bad;
513 		}
514 	}
515 
516 	/*
517 	 * FAITH(Firewall Aided Internet Translator)
518 	 */
519 #if defined(NFAITH) && 0 < NFAITH
520 	if (ip6_keepfaith) {
521 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
522 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
523 			/* XXX do we need more sanity checks? */
524 			ours = 1;
525 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /*faith*/
526 			goto hbhcheck;
527 		}
528 	}
529 #endif
530 
531 	/*
532 	 * Now there is no reason to process the packet if it's not our own
533 	 * and we're not a router.
534 	 */
535 	if (!ip6_forwarding) {
536 		ip6stat.ip6s_cantforward++;
537 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
538 		goto bad;
539 	}
540 
541   hbhcheck:
542 	/*
543 	 * Process Hop-by-Hop options header if it's contained.
544 	 * m may be modified in ip6_hopopts_input().
545 	 * If a JumboPayload option is included, plen will also be modified.
546 	 */
547 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
548 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
549 		struct ip6_hbh *hbh;
550 
551 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
552 #if 0	/*touches NULL pointer*/
553 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
554 #endif
555 			return;	/* m have already been freed */
556 		}
557 
558 		/* adjust pointer */
559 		ip6 = mtod(m, struct ip6_hdr *);
560 
561 		/*
562 		 * if the payload length field is 0 and the next header field
563 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
564 		 * option MUST be included.
565 		 */
566 		if (ip6->ip6_plen == 0 && plen == 0) {
567 			/*
568 			 * Note that if a valid jumbo payload option is
569 			 * contained, ip6_hoptops_input() must set a valid
570 			 * (non-zero) payload length to the variable plen.
571 			 */
572 			ip6stat.ip6s_badoptions++;
573 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
574 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
575 			icmp6_error(m, ICMP6_PARAM_PROB,
576 				    ICMP6_PARAMPROB_HEADER,
577 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
578 			return;
579 		}
580 #ifndef PULLDOWN_TEST
581 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
582 		hbh = (struct ip6_hbh *)(ip6 + 1);
583 #else
584 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
585 			sizeof(struct ip6_hbh));
586 		if (hbh == NULL) {
587 			ip6stat.ip6s_tooshort++;
588 			return;
589 		}
590 #endif
591 		nxt = hbh->ip6h_nxt;
592 
593 		/*
594 		 * accept the packet if a router alert option is included
595 		 * and we act as an IPv6 router.
596 		 */
597 		if (rtalert != ~0 && ip6_forwarding)
598 			ours = 1;
599 	} else
600 		nxt = ip6->ip6_nxt;
601 
602 	/*
603 	 * Check that the amount of data in the buffers
604 	 * is as at least much as the IPv6 header would have us expect.
605 	 * Trim mbufs if longer than we expect.
606 	 * Drop packet if shorter than we expect.
607 	 */
608 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
609 		ip6stat.ip6s_tooshort++;
610 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
611 		goto bad;
612 	}
613 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
614 		if (m->m_len == m->m_pkthdr.len) {
615 			m->m_len = sizeof(struct ip6_hdr) + plen;
616 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
617 		} else
618 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
619 	}
620 
621 	/*
622 	 * Forward if desirable.
623 	 */
624 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
625 		/*
626 		 * If we are acting as a multicast router, all
627 		 * incoming multicast packets are passed to the
628 		 * kernel-level multicast forwarding function.
629 		 * The packet is returned (relatively) intact; if
630 		 * ip6_mforward() returns a non-zero value, the packet
631 		 * must be discarded, else it may be accepted below.
632 		 */
633 		if (ip6_mrouter && ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
634 			ip6stat.ip6s_cantforward++;
635 			m_freem(m);
636 			return;
637 		}
638 		if (!ours) {
639 			m_freem(m);
640 			return;
641 		}
642 	} else if (!ours) {
643 		ip6_forward(m, 0);
644 		return;
645 	}
646 
647 	ip6 = mtod(m, struct ip6_hdr *);
648 
649 	/*
650 	 * Malicious party may be able to use IPv4 mapped addr to confuse
651 	 * tcp/udp stack and bypass security checks (act as if it was from
652 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
653 	 *
654 	 * For SIIT end node behavior, you may want to disable the check.
655 	 * However, you will  become vulnerable to attacks using IPv4 mapped
656 	 * source.
657 	 */
658 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
659 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
660 		ip6stat.ip6s_badscope++;
661 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
662 		goto bad;
663 	}
664 
665 	/*
666 	 * Tell launch routine the next header
667 	 */
668 	ip6stat.ip6s_delivered++;
669 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
670 	nest = 0;
671 	while (nxt != IPPROTO_DONE) {
672 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
673 			ip6stat.ip6s_toomanyhdr++;
674 			goto bad;
675 		}
676 
677 		/*
678 		 * protection against faulty packet - there should be
679 		 * more sanity checks in header chain processing.
680 		 */
681 		if (m->m_pkthdr.len < off) {
682 			ip6stat.ip6s_tooshort++;
683 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
684 			goto bad;
685 		}
686 
687 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
688 	}
689 	return;
690  bad:
691 	m_freem(m);
692 }
693 
694 /*
695  * Hop-by-Hop options header processing. If a valid jumbo payload option is
696  * included, the real payload length will be stored in plenp.
697  */
698 static int
699 ip6_hopopts_input(plenp, rtalertp, mp, offp)
700 	u_int32_t *plenp;
701 	u_int32_t *rtalertp;	/* XXX: should be stored more smart way */
702 	struct mbuf **mp;
703 	int *offp;
704 {
705 	register struct mbuf *m = *mp;
706 	int off = *offp, hbhlen;
707 	struct ip6_hbh *hbh;
708 	u_int8_t *opt;
709 
710 	/* validation of the length of the header */
711 #ifndef PULLDOWN_TEST
712 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
713 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
714 	hbhlen = (hbh->ip6h_len + 1) << 3;
715 
716 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
717 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
718 #else
719 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
720 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
721 	if (hbh == NULL) {
722 		ip6stat.ip6s_tooshort++;
723 		return -1;
724 	}
725 	hbhlen = (hbh->ip6h_len + 1) << 3;
726 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
727 		hbhlen);
728 	if (hbh == NULL) {
729 		ip6stat.ip6s_tooshort++;
730 		return -1;
731 	}
732 #endif
733 	off += hbhlen;
734 	hbhlen -= sizeof(struct ip6_hbh);
735 	opt = (u_int8_t *)hbh + 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 int
753 ip6_process_hopopts(m, opthead, hbhlen, rtalertp, plenp)
754 	struct mbuf *m;
755 	u_int8_t *opthead;
756 	int hbhlen;
757 	u_int32_t *rtalertp;
758 	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 
766 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
767 		switch(*opt) {
768 		 case IP6OPT_PAD1:
769 			 optlen = 1;
770 			 break;
771 		 case IP6OPT_PADN:
772 			 if (hbhlen < IP6OPT_MINLEN) {
773 				 ip6stat.ip6s_toosmall++;
774 				 goto bad;
775 			 }
776 			 optlen = *(opt + 1) + 2;
777 			 break;
778 		 case IP6OPT_RTALERT:
779 			 /* XXX may need check for alignment */
780 			 if (hbhlen < IP6OPT_RTALERT_LEN) {
781 				 ip6stat.ip6s_toosmall++;
782 				 goto bad;
783 			 }
784 			 if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2)
785 				  /* XXX: should we discard the packet? */
786 				 log(LOG_ERR, "length of router alert opt is inconsitent(%d)",
787 				     *(opt + 1));
788 			 optlen = IP6OPT_RTALERT_LEN;
789 			 bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
790 			 *rtalertp = ntohs(rtalert_val);
791 			 break;
792 		 case IP6OPT_JUMBO:
793 			/* XXX may need check for alignment */
794 			if (hbhlen < IP6OPT_JUMBO_LEN) {
795 				ip6stat.ip6s_toosmall++;
796 				goto bad;
797 			}
798 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2)
799 				 /* XXX: should we discard the packet? */
800 				log(LOG_ERR, "length of jumbopayload opt "
801 				    "is inconsistent(%d)",
802 				    *(opt + 1));
803 			optlen = IP6OPT_JUMBO_LEN;
804 
805 			/*
806 			 * IPv6 packets that have non 0 payload length
807 			 * must not contain a jumbo paylod option.
808 			 */
809 			ip6 = mtod(m, struct ip6_hdr *);
810 			if (ip6->ip6_plen) {
811 				ip6stat.ip6s_badoptions++;
812 				icmp6_error(m, ICMP6_PARAM_PROB,
813 					    ICMP6_PARAMPROB_HEADER,
814 					    sizeof(struct ip6_hdr) +
815 					    sizeof(struct ip6_hbh) +
816 					    opt - opthead);
817 				return(-1);
818 			}
819 
820 			/*
821 			 * We may see jumbolen in unaligned location, so
822 			 * we'd need to perform bcopy().
823 			 */
824 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
825 			jumboplen = (u_int32_t)htonl(jumboplen);
826 
827 #if 1
828 			/*
829 			 * if there are multiple jumbo payload options,
830 			 * *plenp will be non-zero and the packet will be
831 			 * rejected.
832 			 * the behavior may need some debate in ipngwg -
833 			 * multiple options does not make sense, however,
834 			 * there's no explicit mention in specification.
835 			 */
836 			if (*plenp != 0) {
837 				ip6stat.ip6s_badoptions++;
838 				icmp6_error(m, ICMP6_PARAM_PROB,
839 					    ICMP6_PARAMPROB_HEADER,
840 					    sizeof(struct ip6_hdr) +
841 					    sizeof(struct ip6_hbh) +
842 					    opt + 2 - opthead);
843 				return(-1);
844 			}
845 #endif
846 
847 			/*
848 			 * jumbo payload length must be larger than 65535.
849 			 */
850 			if (jumboplen <= IPV6_MAXPACKET) {
851 				ip6stat.ip6s_badoptions++;
852 				icmp6_error(m, ICMP6_PARAM_PROB,
853 					    ICMP6_PARAMPROB_HEADER,
854 					    sizeof(struct ip6_hdr) +
855 					    sizeof(struct ip6_hbh) +
856 					    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 			 if ((optlen = ip6_unknown_opt(opt, m,
868 						       sizeof(struct ip6_hdr) +
869 						       sizeof(struct ip6_hbh) +
870 						       opt - opthead)) == -1)
871 				 return(-1);
872 			 optlen += 2;
873 			 break;
874 		}
875 	}
876 
877 	return(0);
878 
879   bad:
880 	m_freem(m);
881 	return(-1);
882 }
883 
884 /*
885  * Unknown option processing.
886  * The third argument `off' is the offset from the IPv6 header to the option,
887  * which is necessary if the IPv6 header the and option header and IPv6 header
888  * is not continuous in order to return an ICMPv6 error.
889  */
890 int
891 ip6_unknown_opt(optp, m, off)
892 	u_int8_t *optp;
893 	struct mbuf *m;
894 	int off;
895 {
896 	struct ip6_hdr *ip6;
897 
898 	switch(IP6OPT_TYPE(*optp)) {
899 	 case IP6OPT_TYPE_SKIP: /* ignore the option */
900 		 return((int)*(optp + 1));
901 	 case IP6OPT_TYPE_DISCARD:	/* silently discard */
902 		 m_freem(m);
903 		 return(-1);
904 	 case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
905 		 ip6stat.ip6s_badoptions++;
906 		 icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
907 		 return(-1);
908 	 case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
909 		 ip6stat.ip6s_badoptions++;
910 		 ip6 = mtod(m, struct ip6_hdr *);
911 		 if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
912 		     (m->m_flags & (M_BCAST|M_MCAST)))
913 			 m_freem(m);
914 		 else
915 			 icmp6_error(m, ICMP6_PARAM_PROB,
916 				     ICMP6_PARAMPROB_OPTION, off);
917 		 return(-1);
918 	}
919 
920 	m_freem(m);		/* XXX: NOTREACHED */
921 	return(-1);
922 }
923 
924 /*
925  * Create the "control" list for this pcb.
926  *
927  * The routine will be called from upper layer handlers like tcp6_input().
928  * Thus the routine assumes that the caller (tcp6_input) have already
929  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
930  * very first mbuf on the mbuf chain.
931  * We may want to add some infinite loop prevention or sanity checks for safety.
932  * (This applies only when you are using KAME mbuf chain restriction, i.e.
933  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
934  */
935 void
936 ip6_savecontrol(in6p, mp, ip6, m)
937 	register struct in6pcb *in6p;
938 	register struct mbuf **mp;
939 	register struct ip6_hdr *ip6;
940 	register struct mbuf *m;
941 {
942 	struct proc *p = curproc;	/* XXX */
943 	int privileged;
944 
945 	privileged = 0;
946 	if (p && !suser(p))
947 		privileged++;
948 
949 	if (in6p->in6p_socket->so_options & SO_TIMESTAMP) {
950 		struct timeval tv;
951 
952 		microtime(&tv);
953 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
954 			SCM_TIMESTAMP, SOL_SOCKET);
955 		if (*mp)
956 			mp = &(*mp)->m_next;
957 	}
958 
959 #ifdef noyet
960 	/* options were tossed above */
961 	if (in6p->in6p_flags & IN6P_RECVOPTS)
962 		/* broken */
963 	/* ip6_srcroute doesn't do what we want here, need to fix */
964 	if (in6p->in6p_flags & IPV6P_RECVRETOPTS)
965 		/* broken */
966 #endif
967 
968 	/* RFC 2292 sec. 5 */
969 	if (in6p->in6p_flags & IN6P_PKTINFO) {
970 		struct in6_pktinfo pi6;
971 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
972 		if (IN6_IS_SCOPE_LINKLOCAL(&pi6.ipi6_addr))
973 			pi6.ipi6_addr.s6_addr16[1] = 0;
974 		pi6.ipi6_ifindex = (m && m->m_pkthdr.rcvif)
975 					? m->m_pkthdr.rcvif->if_index
976 					: 0;
977 		*mp = sbcreatecontrol((caddr_t) &pi6,
978 			sizeof(struct in6_pktinfo), IPV6_PKTINFO,
979 			IPPROTO_IPV6);
980 		if (*mp)
981 			mp = &(*mp)->m_next;
982 	}
983 	if (in6p->in6p_flags & IN6P_HOPLIMIT) {
984 		int hlim = ip6->ip6_hlim & 0xff;
985 		*mp = sbcreatecontrol((caddr_t) &hlim,
986 			sizeof(int), IPV6_HOPLIMIT, IPPROTO_IPV6);
987 		if (*mp)
988 			mp = &(*mp)->m_next;
989 	}
990 	/* IN6P_NEXTHOP - for outgoing packet only */
991 
992 	/*
993 	 * IPV6_HOPOPTS socket option. We require super-user privilege
994 	 * for the option, but it might be too strict, since there might
995 	 * be some hop-by-hop options which can be returned to normal user.
996 	 * See RFC 2292 section 6.
997 	 */
998 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
999 		/*
1000 		 * Check if a hop-by-hop options header is contatined in the
1001 		 * received packet, and if so, store the options as ancillary
1002 		 * data. Note that a hop-by-hop options header must be
1003 		 * just after the IPv6 header, which fact is assured through
1004 		 * the IPv6 input processing.
1005 		 */
1006 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1007 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1008 			struct ip6_hbh *hbh;
1009 			int hbhlen;
1010 
1011 #ifndef PULLDOWN_TEST
1012 			hbh = (struct ip6_hbh *)(ip6 + 1);
1013 			hbhlen = (hbh->ip6h_len + 1) << 3;
1014 #else
1015 			IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1016 				sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
1017 			if (hbh == NULL) {
1018 				ip6stat.ip6s_tooshort++;
1019 				return;
1020 			}
1021 			hbhlen = (hbh->ip6h_len + 1) << 3;
1022 			IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1023 				sizeof(struct ip6_hdr), hbhlen);
1024 			if (hbh == NULL) {
1025 				ip6stat.ip6s_tooshort++;
1026 				return;
1027 			}
1028 #endif
1029 
1030 			/*
1031 			 * XXX: We copy whole the header even if a jumbo
1032 			 * payload option is included, which option is to
1033 			 * be removed before returning in the RFC 2292.
1034 			 * But it's too painful operation...
1035 			 */
1036 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1037 					      IPV6_HOPOPTS, IPPROTO_IPV6);
1038 			if (*mp)
1039 				mp = &(*mp)->m_next;
1040 		}
1041 	}
1042 
1043 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1044 	if (in6p->in6p_flags & (IN6P_DSTOPTS | IN6P_RTHDR)) {
1045 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1046 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);;
1047 
1048 		/*
1049 		 * Search for destination options headers or routing
1050 		 * header(s) through the header chain, and stores each
1051 		 * header as ancillary data.
1052 		 * Note that the order of the headers remains in
1053 		 * the chain of ancillary data.
1054 		 */
1055 		while(1) {	/* is explicit loop prevention necessary? */
1056 			struct ip6_ext *ip6e;
1057 			int elen;
1058 
1059 #ifndef PULLDOWN_TEST
1060 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1061 			if (nxt == IPPROTO_AH)
1062 				elen = (ip6e->ip6e_len + 2) << 2;
1063 			else
1064 				elen = (ip6e->ip6e_len + 1) << 3;
1065 #else
1066 			IP6_EXTHDR_GET(ip6e, struct ip6_ext *, m, off,
1067 				sizeof(struct ip6_ext));
1068 			if (ip6e == NULL) {
1069 				ip6stat.ip6s_tooshort++;
1070 				return;
1071 			}
1072 			if (nxt == IPPROTO_AH)
1073 				elen = (ip6e->ip6e_len + 2) << 2;
1074 			else
1075 				elen = (ip6e->ip6e_len + 1) << 3;
1076 			IP6_EXTHDR_GET(ip6e, struct ip6_ext *, m, off, elen);
1077 			if (ip6e == NULL) {
1078 				ip6stat.ip6s_tooshort++;
1079 				return;
1080 			}
1081 #endif
1082 
1083 			switch(nxt) {
1084 		         case IPPROTO_DSTOPTS:
1085 				 if (!in6p->in6p_flags & IN6P_DSTOPTS)
1086 					 break;
1087 
1088 				 /*
1089 				  * We also require super-user privilege for
1090 				  * the option.
1091 				  * See the comments on IN6_HOPOPTS.
1092 				  */
1093 				 if (!privileged)
1094 					 break;
1095 
1096 				 *mp = sbcreatecontrol((caddr_t)ip6e, elen,
1097 						       IPV6_DSTOPTS,
1098 						       IPPROTO_IPV6);
1099 				 if (*mp)
1100 					 mp = &(*mp)->m_next;
1101 				 break;
1102 
1103 			 case IPPROTO_ROUTING:
1104 				 if (!in6p->in6p_flags & IN6P_RTHDR)
1105 					 break;
1106 
1107 				 *mp = sbcreatecontrol((caddr_t)ip6e, elen,
1108 						       IPV6_RTHDR,
1109 						       IPPROTO_IPV6);
1110 				 if (*mp)
1111 					 mp = &(*mp)->m_next;
1112 				 break;
1113 
1114 			 case IPPROTO_UDP:
1115 			 case IPPROTO_TCP:
1116 			 case IPPROTO_ICMPV6:
1117 			 default:
1118 				 /*
1119 				  * stop search if we encounter an upper
1120 				  * layer protocol headers.
1121 				  */
1122 				 goto loopend;
1123 
1124 			 case IPPROTO_HOPOPTS:
1125 			 case IPPROTO_AH: /* is it possible? */
1126 				 break;
1127 			}
1128 
1129 			/* proceed with the next header. */
1130 			off += elen;
1131 			nxt = ip6e->ip6e_nxt;
1132 		}
1133 	  loopend:
1134 	}
1135 	if ((in6p->in6p_flags & IN6P_HOPOPTS) && privileged) {
1136 		/* to be done */
1137 	}
1138 	if ((in6p->in6p_flags & IN6P_DSTOPTS) && privileged) {
1139 		/* to be done */
1140 	}
1141 	/* IN6P_RTHDR - to be done */
1142 
1143 }
1144 
1145 /*
1146  * Get pointer to the previous header followed by the header
1147  * currently processed.
1148  * XXX: This function supposes that
1149  *	M includes all headers,
1150  *	the next header field and the header length field of each header
1151  *	are valid, and
1152  *	the sum of each header length equals to OFF.
1153  * Because of these assumptions, this function must be called very
1154  * carefully. Moreover, it will not be used in the near future when
1155  * we develop `neater' mechanism to process extension headers.
1156  */
1157 char *
1158 ip6_get_prevhdr(m, off)
1159 	struct mbuf *m;
1160 	int off;
1161 {
1162 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1163 
1164 	if (off == sizeof(struct ip6_hdr))
1165 		return(&ip6->ip6_nxt);
1166 	else {
1167 		int len, nxt;
1168 		struct ip6_ext *ip6e = NULL;
1169 
1170 		nxt = ip6->ip6_nxt;
1171 		len = sizeof(struct ip6_hdr);
1172 		while (len < off) {
1173 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1174 
1175 			switch(nxt) {
1176 			case IPPROTO_FRAGMENT:
1177 				len += sizeof(struct ip6_frag);
1178 				break;
1179 			case IPPROTO_AH:
1180 				len += (ip6e->ip6e_len + 2) << 2;
1181 				break;
1182 			default:
1183 				len += (ip6e->ip6e_len + 1) << 3;
1184 				break;
1185 			}
1186 			nxt = ip6e->ip6e_nxt;
1187 		}
1188 		if (ip6e)
1189 			return(&ip6e->ip6e_nxt);
1190 		else
1191 			return NULL;
1192 	}
1193 }
1194 
1195 /*
1196  * get next header offset.  m will be retained.
1197  */
1198 int
1199 ip6_nexthdr(m, off, proto, nxtp)
1200 	struct mbuf *m;
1201 	int off;
1202 	int proto;
1203 	int *nxtp;
1204 {
1205 	struct ip6_hdr ip6;
1206 	struct ip6_ext ip6e;
1207 	struct ip6_frag fh;
1208 
1209 	/* just in case */
1210 	if (m == NULL)
1211 		panic("ip6_nexthdr: m == NULL");
1212 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1213 		return -1;
1214 
1215 	switch (proto) {
1216 	case IPPROTO_IPV6:
1217 		if (m->m_pkthdr.len < off + sizeof(ip6))
1218 			return -1;
1219 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1220 		if (nxtp)
1221 			*nxtp = ip6.ip6_nxt;
1222 		off += sizeof(ip6);
1223 		return off;
1224 
1225 	case IPPROTO_FRAGMENT:
1226 		/*
1227 		 * terminate parsing if it is not the first fragment,
1228 		 * it does not make sense to parse through it.
1229 		 */
1230 		if (m->m_pkthdr.len < off + sizeof(fh))
1231 			return -1;
1232 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1233 		if ((ntohs(fh.ip6f_offlg) & IP6F_OFF_MASK) != 0)
1234 			return -1;
1235 		if (nxtp)
1236 			*nxtp = fh.ip6f_nxt;
1237 		off += sizeof(struct ip6_frag);
1238 		return off;
1239 
1240 	case IPPROTO_AH:
1241 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1242 			return -1;
1243 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1244 		if (nxtp)
1245 			*nxtp = ip6e.ip6e_nxt;
1246 		off += (ip6e.ip6e_len + 2) << 2;
1247 		return off;
1248 
1249 	case IPPROTO_HOPOPTS:
1250 	case IPPROTO_ROUTING:
1251 	case IPPROTO_DSTOPTS:
1252 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1253 			return -1;
1254 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1255 		if (nxtp)
1256 			*nxtp = ip6e.ip6e_nxt;
1257 		off += (ip6e.ip6e_len + 1) << 3;
1258 		return off;
1259 
1260 	case IPPROTO_NONE:
1261 	case IPPROTO_ESP:
1262 	case IPPROTO_IPCOMP:
1263 		/* give up */
1264 		return -1;
1265 
1266 	default:
1267 		return -1;
1268 	}
1269 
1270 	return -1;
1271 }
1272 
1273 /*
1274  * get offset for the last header in the chain.  m will be kept untainted.
1275  */
1276 int
1277 ip6_lasthdr(m, off, proto, nxtp)
1278 	struct mbuf *m;
1279 	int off;
1280 	int proto;
1281 	int *nxtp;
1282 {
1283 	int newoff;
1284 	int nxt;
1285 
1286 	if (!nxtp) {
1287 		nxt = -1;
1288 		nxtp = &nxt;
1289 	}
1290 	while (1) {
1291 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1292 		if (newoff < 0)
1293 			return off;
1294 		else if (newoff < off)
1295 			return -1;	/* invalid */
1296 		else if (newoff == off)
1297 			return newoff;
1298 
1299 		off = newoff;
1300 		proto = *nxtp;
1301 	}
1302 }
1303 
1304 /*
1305  * System control for IP6
1306  */
1307 
1308 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1309 	0,		0,		0,		0,
1310 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1311 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1312 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1313 	0,		0,		0,		0,
1314 	ENOPROTOOPT
1315 };
1316