xref: /openbsd-src/sys/netinet6/ip6_input.c (revision 43003dfe3ad45d1698bed8a37f2b0f5b14f20d4f)
1 /*	$OpenBSD: ip6_input.c,v 1.91 2009/05/18 20:37:13 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/malloc.h>
70 #include <sys/mbuf.h>
71 #include <sys/domain.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/kernel.h>
78 #include <sys/syslog.h>
79 #include <sys/proc.h>
80 
81 #include <net/if.h>
82 #include <net/if_types.h>
83 #include <net/if_dl.h>
84 #include <net/route.h>
85 #include <net/netisr.h>
86 
87 #include <netinet/in.h>
88 #include <netinet/in_systm.h>
89 
90 #ifdef INET
91 #include <netinet/ip.h>
92 #include <netinet/ip_icmp.h>
93 #endif /*INET*/
94 
95 #include <netinet/in_pcb.h>
96 #include <netinet6/in6_var.h>
97 #include <netinet/ip6.h>
98 #include <netinet6/ip6_var.h>
99 #include <netinet/icmp6.h>
100 #include <netinet6/in6_ifattach.h>
101 #include <netinet6/nd6.h>
102 
103 #include <netinet6/ip6protosw.h>
104 
105 #include "faith.h"
106 #include "gif.h"
107 #include "bpfilter.h"
108 
109 #ifdef MROUTING
110 #include <netinet6/ip6_mroute.h>
111 #endif
112 
113 #if NPF > 0
114 #include <net/pfvar.h>
115 #endif
116 
117 #if NCARP > 0
118 #include <netinet/in_var.h>
119 #include <netinet/ip_carp.h>
120 #endif
121 
122 extern struct domain inet6domain;
123 extern struct ip6protosw inet6sw[];
124 
125 u_char ip6_protox[IPPROTO_MAX];
126 static int ip6qmaxlen = IFQ_MAXLEN;
127 struct in6_ifaddr *in6_ifaddr;
128 struct ifqueue ip6intrq;
129 
130 int ip6_forward_srcrt;			/* XXX */
131 int ip6_sourcecheck;			/* XXX */
132 int ip6_sourcecheck_interval;		/* XXX */
133 
134 struct ip6stat ip6stat;
135 
136 static void ip6_init2(void *);
137 int ip6_check_rh0hdr(struct mbuf *);
138 
139 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
140 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
141 
142 /*
143  * IP6 initialization: fill in IP6 protocol switch table.
144  * All protocols not implemented in kernel go to raw IP6 protocol handler.
145  */
146 void
147 ip6_init()
148 {
149 	struct ip6protosw *pr;
150 	int i;
151 
152 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
153 	if (pr == 0)
154 		panic("ip6_init");
155 	for (i = 0; i < IPPROTO_MAX; i++)
156 		ip6_protox[i] = pr - inet6sw;
157 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
158 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
159 		if (pr->pr_domain->dom_family == PF_INET6 &&
160 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW)
161 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
162 	ip6intrq.ifq_maxlen = ip6qmaxlen;
163 	ip6_randomid_init();
164 	nd6_init();
165 	frag6_init();
166 	ip6_init2((void *)0);
167 }
168 
169 static void
170 ip6_init2(void *dummy)
171 {
172 
173 	/* nd6_timer_init */
174 	bzero(&nd6_timer_ch, sizeof(nd6_timer_ch));
175 	timeout_set(&nd6_timer_ch, nd6_timer, NULL);
176 	timeout_add_sec(&nd6_timer_ch, 1);
177 }
178 
179 /*
180  * IP6 input interrupt handling. Just pass the packet to ip6_input.
181  */
182 void
183 ip6intr()
184 {
185 	int s;
186 	struct mbuf *m;
187 
188 	for (;;) {
189 		s = splnet();
190 		IF_DEQUEUE(&ip6intrq, m);
191 		splx(s);
192 		if (m == NULL)
193 			return;
194 		ip6_input(m);
195 	}
196 }
197 
198 extern struct	route_in6 ip6_forward_rt;
199 extern u_int	ip6_forward_rtableid;
200 
201 void
202 ip6_input(struct mbuf *m)
203 {
204 	struct ip6_hdr *ip6;
205 	int off = sizeof(struct ip6_hdr), nest;
206 	u_int32_t plen;
207 	u_int32_t rtalert = ~0;
208 	int nxt, ours = 0;
209 	struct ifnet *deliverifp = NULL;
210 #if NPF > 0
211 	struct in6_addr odst;
212 #endif
213 	int srcrt = 0, isanycast = 0;
214 	u_int rtableid = 0;
215 
216 	/*
217 	 * mbuf statistics by kazu
218 	 */
219 	if (m->m_flags & M_EXT) {
220 		if (m->m_next)
221 			ip6stat.ip6s_mext2m++;
222 		else
223 			ip6stat.ip6s_mext1++;
224 	} else {
225 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
226 		if (m->m_next) {
227 			if (m->m_flags & M_LOOP) {
228 				ip6stat.ip6s_m2m[lo0ifp->if_index]++;	/*XXX*/
229 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
230 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
231 			else
232 				ip6stat.ip6s_m2m[0]++;
233 		} else
234 			ip6stat.ip6s_m1++;
235 #undef M2MMAX
236 	}
237 
238 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
239 	ip6stat.ip6s_total++;
240 
241 	if (m->m_len < sizeof(struct ip6_hdr)) {
242 		struct ifnet *inifp;
243 		inifp = m->m_pkthdr.rcvif;
244 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
245 			ip6stat.ip6s_toosmall++;
246 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
247 			return;
248 		}
249 	}
250 
251 	ip6 = mtod(m, struct ip6_hdr *);
252 
253 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
254 		ip6stat.ip6s_badvers++;
255 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
256 		goto bad;
257 	}
258 
259 #if NCARP > 0
260 	if (m->m_pkthdr.rcvif->if_type == IFT_CARP &&
261 	    ip6->ip6_nxt != IPPROTO_ICMPV6 &&
262 	    carp_lsdrop(m, AF_INET6, ip6->ip6_src.s6_addr32,
263 	    ip6->ip6_dst.s6_addr32))
264 		goto bad;
265 #endif
266 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
267 
268 	/*
269 	 * Check against address spoofing/corruption.
270 	 */
271 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
272 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
273 		/*
274 		 * XXX: "badscope" is not very suitable for a multicast source.
275 		 */
276 		ip6stat.ip6s_badscope++;
277 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
278 		goto bad;
279 	}
280 
281 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
282 	    !(m->m_flags & M_LOOP)) {
283 		/*
284 		 * In this case, the packet should come from the loopback
285 		 * interface.  However, we cannot just check the if_flags,
286 		 * because ip6_mloopback() passes the "actual" interface
287 		 * as the outgoing/incoming interface.
288 		 */
289 		ip6stat.ip6s_badscope++;
290 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
291 		goto bad;
292 	}
293 
294 	/*
295 	 * The following check is not documented in specs.  A malicious
296 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
297 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
298 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
299 	 *
300 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
301 	 * support IPv4-less kernel compilation, we cannot support SIIT
302 	 * environment at all.  So, it makes more sense for us to reject any
303 	 * malicious packets for non-SIIT environment, than try to do a
304 	 * partial support for SIIT environment.
305 	 */
306 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
307 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
308 		ip6stat.ip6s_badscope++;
309 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
310 		goto bad;
311 	}
312 #if 0
313 	/*
314 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
315 	 *
316 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
317 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
318 	 * is revised to forbid relaying case.
319 	 */
320 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
321 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
322 		ip6stat.ip6s_badscope++;
323 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
324 		goto bad;
325 	}
326 #endif
327 
328 	if (ip6_check_rh0hdr(m)) {
329 		ip6stat.ip6s_badoptions++;
330 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
331 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
332 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, 0);
333 		/* m is already freed */
334 		return;
335 	}
336 
337 #if NPF > 0
338         /*
339          * Packet filter
340          */
341 	odst = ip6->ip6_dst;
342 	if (pf_test6(PF_IN, m->m_pkthdr.rcvif, &m, NULL) != PF_PASS)
343 		goto bad;
344 	if (m == NULL)
345 		return;
346 
347 	ip6 = mtod(m, struct ip6_hdr *);
348 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
349 #endif
350 
351 	if (IN6_IS_ADDR_LOOPBACK(&ip6->ip6_src) ||
352 	    IN6_IS_ADDR_LOOPBACK(&ip6->ip6_dst)) {
353 		if (m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) {
354 			ours = 1;
355 			deliverifp = m->m_pkthdr.rcvif;
356 			goto hbhcheck;
357 		} else {
358 			ip6stat.ip6s_badscope++;
359 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
360 			goto bad;
361 		}
362 	}
363 
364 	/* drop packets if interface ID portion is already filled */
365 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) == 0) {
366 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src) &&
367 		    ip6->ip6_src.s6_addr16[1]) {
368 			ip6stat.ip6s_badscope++;
369 			goto bad;
370 		}
371 		if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst) &&
372 		    ip6->ip6_dst.s6_addr16[1]) {
373 			ip6stat.ip6s_badscope++;
374 			goto bad;
375 		}
376 	}
377 
378 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_src))
379 		ip6->ip6_src.s6_addr16[1] = htons(m->m_pkthdr.rcvif->if_index);
380 	if (IN6_IS_SCOPE_EMBED(&ip6->ip6_dst))
381 		ip6->ip6_dst.s6_addr16[1] = htons(m->m_pkthdr.rcvif->if_index);
382 
383 	/*
384 	 * We use rt->rt_ifp to determine if the address is ours or not.
385 	 * If rt_ifp is lo0, the address is ours.
386 	 * The problem here is, rt->rt_ifp for fe80::%lo0/64 is set to lo0,
387 	 * so any address under fe80::%lo0/64 will be mistakenly considered
388 	 * local.  The special case is supplied to handle the case properly
389 	 * by actually looking at interface addresses
390 	 * (using in6ifa_ifpwithaddr).
391 	 */
392 	if ((m->m_pkthdr.rcvif->if_flags & IFF_LOOPBACK) != 0 &&
393 	    IN6_IS_ADDR_LINKLOCAL(&ip6->ip6_dst)) {
394 		if (!in6ifa_ifpwithaddr(m->m_pkthdr.rcvif, &ip6->ip6_dst)) {
395 			icmp6_error(m, ICMP6_DST_UNREACH,
396 			    ICMP6_DST_UNREACH_ADDR, 0);
397 			/* m is already freed */
398 			return;
399 		}
400 
401 		ours = 1;
402 		deliverifp = m->m_pkthdr.rcvif;
403 		goto hbhcheck;
404 	}
405 
406 	if (m->m_pkthdr.pf.flags & PF_TAG_DIVERTED) {
407 		ours = 1;
408 		deliverifp = m->m_pkthdr.rcvif;
409 		goto hbhcheck;
410 	}
411 
412 	/*
413 	 * Multicast check
414 	 */
415 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
416 	  	struct	in6_multi *in6m = 0;
417 
418 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
419 		/*
420 		 * See if we belong to the destination multicast group on the
421 		 * arrival interface.
422 		 */
423 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
424 		if (in6m)
425 			ours = 1;
426 #ifdef MROUTING
427 		else if (!ip6_mforwarding || !ip6_mrouter)
428 #else
429 		else
430 #endif
431 		{
432 			ip6stat.ip6s_notmember++;
433 			if (!IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
434 				ip6stat.ip6s_cantforward++;
435 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
436 			goto bad;
437 		}
438 		deliverifp = m->m_pkthdr.rcvif;
439 		goto hbhcheck;
440 	}
441 
442 #if NPF > 0
443 	rtableid = m->m_pkthdr.pf.rtableid;
444 #endif
445 
446 	/*
447 	 *  Unicast check
448 	 */
449 	if (ip6_forward_rt.ro_rt != NULL &&
450 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
451 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
452 			       &ip6_forward_rt.ro_dst.sin6_addr) &&
453 	    rtableid == ip6_forward_rtableid)
454 		ip6stat.ip6s_forward_cachehit++;
455 	else {
456 		if (ip6_forward_rt.ro_rt) {
457 			/* route is down or destination is different */
458 			ip6stat.ip6s_forward_cachemiss++;
459 			RTFREE(ip6_forward_rt.ro_rt);
460 			ip6_forward_rt.ro_rt = 0;
461 		}
462 
463 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
464 		ip6_forward_rt.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
465 		ip6_forward_rt.ro_dst.sin6_family = AF_INET6;
466 		ip6_forward_rt.ro_dst.sin6_addr = ip6->ip6_dst;
467 		ip6_forward_rtableid = rtableid;
468 
469 		rtalloc_mpath((struct route *)&ip6_forward_rt,
470 		    &ip6->ip6_src.s6_addr32[0], rtableid);
471 	}
472 
473 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
474 
475 	/*
476 	 * Accept the packet if the forwarding interface to the destination
477 	 * according to the routing table is the loopback interface,
478 	 * unless the associated route has a gateway.
479 	 * Note that this approach causes to accept a packet if there is a
480 	 * route to the loopback interface for the destination of the packet.
481 	 * But we think it's even useful in some situations, e.g. when using
482 	 * a special daemon which wants to intercept the packet.
483 	 */
484 	if (ip6_forward_rt.ro_rt &&
485 	    (ip6_forward_rt.ro_rt->rt_flags &
486 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
487 #if 0
488 	    /*
489 	     * The check below is redundant since the comparison of
490 	     * the destination and the key of the rtentry has
491 	     * already done through looking up the routing table.
492 	     */
493 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
494 	    &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr) &&
495 #endif
496 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
497 		struct in6_ifaddr *ia6 =
498 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
499 		if (ia6->ia6_flags & IN6_IFF_ANYCAST)
500 			isanycast = 1;
501 		/*
502 		 * packets to a tentative, duplicated, or somehow invalid
503 		 * address must not be accepted.
504 		 */
505 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
506 			/* this address is ready */
507 			ours = 1;
508 			deliverifp = ia6->ia_ifp;	/* correct? */
509 			goto hbhcheck;
510 		} else {
511 			/* address is not ready, so discard the packet. */
512 			nd6log((LOG_INFO,
513 			    "ip6_input: packet to an unready address %s->%s\n",
514 			    ip6_sprintf(&ip6->ip6_src),
515 			    ip6_sprintf(&ip6->ip6_dst)));
516 
517 			goto bad;
518 		}
519 	}
520 
521 	/*
522 	 * FAITH (Firewall Aided Internet Translator)
523 	 */
524 #if defined(NFAITH) && 0 < NFAITH
525 	if (ip6_keepfaith) {
526 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
527 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
528 			/* XXX do we need more sanity checks? */
529 			ours = 1;
530 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /*faith*/
531 			goto hbhcheck;
532 		}
533 	}
534 #endif
535 
536 #if 0
537     {
538 	/*
539 	 * Last resort: check in6_ifaddr for incoming interface.
540 	 * The code is here until I update the "goto ours hack" code above
541 	 * working right.
542 	 */
543 	struct ifaddr *ifa;
544 	TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrlist, ifa_list) {
545 		if (ifa->ifa_addr == NULL)
546 			continue;	/* just for safety */
547 		if (ifa->ifa_addr->sa_family != AF_INET6)
548 			continue;
549 		if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa), &ip6->ip6_dst)) {
550 			ours = 1;
551 			deliverifp = ifa->ifa_ifp;
552 			goto hbhcheck;
553 		}
554 	}
555     }
556 #endif
557 
558 #if NCARP > 0
559 	if (m->m_pkthdr.rcvif->if_type == IFT_CARP &&
560 	    ip6->ip6_nxt == IPPROTO_ICMPV6 &&
561 	    carp_lsdrop(m, AF_INET6, ip6->ip6_src.s6_addr32,
562 	    ip6->ip6_dst.s6_addr32))
563 		goto bad;
564 #endif
565 	/*
566 	 * Now there is no reason to process the packet if it's not our own
567 	 * and we're not a router.
568 	 */
569 	if (!ip6_forwarding) {
570 		ip6stat.ip6s_cantforward++;
571 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
572 		goto bad;
573 	}
574 
575   hbhcheck:
576 	/*
577 	 * Process Hop-by-Hop options header if it's contained.
578 	 * m may be modified in ip6_hopopts_input().
579 	 * If a JumboPayload option is included, plen will also be modified.
580 	 */
581 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
582 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
583 		struct ip6_hbh *hbh;
584 
585 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
586 #if 0	/*touches NULL pointer*/
587 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
588 #endif
589 			return;	/* m have already been freed */
590 		}
591 
592 		/* adjust pointer */
593 		ip6 = mtod(m, struct ip6_hdr *);
594 
595 		/*
596 		 * if the payload length field is 0 and the next header field
597 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
598 		 * option MUST be included.
599 		 */
600 		if (ip6->ip6_plen == 0 && plen == 0) {
601 			/*
602 			 * Note that if a valid jumbo payload option is
603 			 * contained, ip6_hoptops_input() must set a valid
604 			 * (non-zero) payload length to the variable plen.
605 			 */
606 			ip6stat.ip6s_badoptions++;
607 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
608 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
609 			icmp6_error(m, ICMP6_PARAM_PROB,
610 				    ICMP6_PARAMPROB_HEADER,
611 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
612 			return;
613 		}
614 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
615 			sizeof(struct ip6_hbh));
616 		if (hbh == NULL) {
617 			ip6stat.ip6s_tooshort++;
618 			return;
619 		}
620 		nxt = hbh->ip6h_nxt;
621 
622 		/*
623 		 * accept the packet if a router alert option is included
624 		 * and we act as an IPv6 router.
625 		 */
626 		if (rtalert != ~0 && ip6_forwarding)
627 			ours = 1;
628 	} else
629 		nxt = ip6->ip6_nxt;
630 
631 	/*
632 	 * Check that the amount of data in the buffers
633 	 * is as at least much as the IPv6 header would have us expect.
634 	 * Trim mbufs if longer than we expect.
635 	 * Drop packet if shorter than we expect.
636 	 */
637 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
638 		ip6stat.ip6s_tooshort++;
639 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
640 		goto bad;
641 	}
642 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
643 		if (m->m_len == m->m_pkthdr.len) {
644 			m->m_len = sizeof(struct ip6_hdr) + plen;
645 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
646 		} else
647 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
648 	}
649 
650 	/*
651 	 * Forward if desirable.
652 	 */
653 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
654 		/*
655 		 * If we are acting as a multicast router, all
656 		 * incoming multicast packets are passed to the
657 		 * kernel-level multicast forwarding function.
658 		 * The packet is returned (relatively) intact; if
659 		 * ip6_mforward() returns a non-zero value, the packet
660 		 * must be discarded, else it may be accepted below.
661 		 */
662 #ifdef MROUTING
663 		if (ip6_mforwarding && ip6_mrouter &&
664 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
665 			ip6stat.ip6s_cantforward++;
666 			m_freem(m);
667 			return;
668 		}
669 #endif
670 		if (!ours) {
671 			m_freem(m);
672 			return;
673 		}
674 	} else if (!ours) {
675 		ip6_forward(m, srcrt);
676 		return;
677 	}
678 
679 	ip6 = mtod(m, struct ip6_hdr *);
680 
681 	/*
682 	 * Malicious party may be able to use IPv4 mapped addr to confuse
683 	 * tcp/udp stack and bypass security checks (act as if it was from
684 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
685 	 *
686 	 * For SIIT end node behavior, you may want to disable the check.
687 	 * However, you will  become vulnerable to attacks using IPv4 mapped
688 	 * source.
689 	 */
690 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
691 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
692 		ip6stat.ip6s_badscope++;
693 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
694 		goto bad;
695 	}
696 
697 	/*
698 	 * Tell launch routine the next header
699 	 */
700 	ip6stat.ip6s_delivered++;
701 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
702 	nest = 0;
703 
704 	while (nxt != IPPROTO_DONE) {
705 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
706 			ip6stat.ip6s_toomanyhdr++;
707 			goto bad;
708 		}
709 
710 		/*
711 		 * protection against faulty packet - there should be
712 		 * more sanity checks in header chain processing.
713 		 */
714 		if (m->m_pkthdr.len < off) {
715 			ip6stat.ip6s_tooshort++;
716 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
717 			goto bad;
718 		}
719 
720 		/* draft-itojun-ipv6-tcp-to-anycast */
721 		if (isanycast && nxt == IPPROTO_TCP) {
722 			if (m->m_len >= sizeof(struct ip6_hdr)) {
723 				ip6 = mtod(m, struct ip6_hdr *);
724 				icmp6_error(m, ICMP6_DST_UNREACH,
725 					ICMP6_DST_UNREACH_ADDR,
726 					(caddr_t)&ip6->ip6_dst - (caddr_t)ip6);
727 				break;
728 			} else
729 				goto bad;
730 		}
731 
732 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
733 	}
734 	return;
735  bad:
736 	m_freem(m);
737 }
738 
739 /* scan packet for RH0 routing header. Mostly stolen from pf.c:pf_test6() */
740 int
741 ip6_check_rh0hdr(struct mbuf *m)
742 {
743 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
744 	struct ip6_rthdr rthdr;
745 	struct ip6_ext opt6;
746 	u_int8_t proto = ip6->ip6_nxt;
747 	int done = 0, lim, off, rh_cnt = 0;
748 
749 	off = ((caddr_t)ip6 - m->m_data) + sizeof(struct ip6_hdr);
750 	lim = min(m->m_pkthdr.len, ntohs(ip6->ip6_plen) + sizeof(*ip6));
751 	do {
752 		switch (proto) {
753 		case IPPROTO_ROUTING:
754 			if (rh_cnt++) {
755 				/* more then one rh header present */
756 				return (1);
757 			}
758 
759 			if (off + sizeof(rthdr) > lim) {
760 				/* packet to short to make sense */
761 				return (1);
762 			}
763 
764 			m_copydata(m, off, sizeof(rthdr), (caddr_t)&rthdr);
765 
766 			if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0)
767 				return (1);
768 
769 			off += (rthdr.ip6r_len + 1) * 8;
770 			proto = rthdr.ip6r_nxt;
771 			break;
772 		case IPPROTO_AH:
773 		case IPPROTO_HOPOPTS:
774 		case IPPROTO_DSTOPTS:
775 			/* get next header and header length */
776 			if (off + sizeof(opt6) > lim) {
777 				/*
778 				 * Packet to short to make sense, we could
779 				 * reject the packet but as a router we
780 				 * should not do that so forward it.
781 				 */
782 				return (0);
783 			}
784 
785 			m_copydata(m, off, sizeof(opt6), (caddr_t)&opt6);
786 
787 			if (proto == IPPROTO_AH)
788 				off += (opt6.ip6e_len + 2) * 4;
789 			else
790 				off += (opt6.ip6e_len + 1) * 8;
791 			proto = opt6.ip6e_nxt;
792 			break;
793 		case IPPROTO_FRAGMENT:
794 		default:
795 			/* end of header stack */
796 			done = 1;
797 			break;
798 		}
799 	} while (!done);
800 
801 	return (0);
802 }
803 
804 /*
805  * Hop-by-Hop options header processing. If a valid jumbo payload option is
806  * included, the real payload length will be stored in plenp.
807  *
808  * rtalertp - XXX: should be stored in a more smart way
809  */
810 static int
811 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp, struct mbuf **mp,
812 	int *offp)
813 {
814 	struct mbuf *m = *mp;
815 	int off = *offp, hbhlen;
816 	struct ip6_hbh *hbh;
817 
818 	/* validation of the length of the header */
819 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
820 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
821 	if (hbh == NULL) {
822 		ip6stat.ip6s_tooshort++;
823 		return -1;
824 	}
825 	hbhlen = (hbh->ip6h_len + 1) << 3;
826 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
827 		hbhlen);
828 	if (hbh == NULL) {
829 		ip6stat.ip6s_tooshort++;
830 		return -1;
831 	}
832 	off += hbhlen;
833 	hbhlen -= sizeof(struct ip6_hbh);
834 
835 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
836 				hbhlen, rtalertp, plenp) < 0)
837 		return (-1);
838 
839 	*offp = off;
840 	*mp = m;
841 	return (0);
842 }
843 
844 /*
845  * Search header for all Hop-by-hop options and process each option.
846  * This function is separate from ip6_hopopts_input() in order to
847  * handle a case where the sending node itself process its hop-by-hop
848  * options header. In such a case, the function is called from ip6_output().
849  *
850  * The function assumes that hbh header is located right after the IPv6 header
851  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
852  * opthead + hbhlen is located in continuous memory region.
853  */
854 int
855 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
856 	u_int32_t *rtalertp, u_int32_t *plenp)
857 {
858 	struct ip6_hdr *ip6;
859 	int optlen = 0;
860 	u_int8_t *opt = opthead;
861 	u_int16_t rtalert_val;
862 	u_int32_t jumboplen;
863 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
864 
865 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
866 		switch (*opt) {
867 		case IP6OPT_PAD1:
868 			optlen = 1;
869 			break;
870 		case IP6OPT_PADN:
871 			if (hbhlen < IP6OPT_MINLEN) {
872 				ip6stat.ip6s_toosmall++;
873 				goto bad;
874 			}
875 			optlen = *(opt + 1) + 2;
876 			break;
877 		case IP6OPT_ROUTER_ALERT:
878 			/* XXX may need check for alignment */
879 			if (hbhlen < IP6OPT_RTALERT_LEN) {
880 				ip6stat.ip6s_toosmall++;
881 				goto bad;
882 			}
883 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
884 				/* XXX stat */
885 				icmp6_error(m, ICMP6_PARAM_PROB,
886 				    ICMP6_PARAMPROB_HEADER,
887 				    erroff + opt + 1 - opthead);
888 				return (-1);
889 			}
890 			optlen = IP6OPT_RTALERT_LEN;
891 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
892 			*rtalertp = ntohs(rtalert_val);
893 			break;
894 		case IP6OPT_JUMBO:
895 			/* XXX may need check for alignment */
896 			if (hbhlen < IP6OPT_JUMBO_LEN) {
897 				ip6stat.ip6s_toosmall++;
898 				goto bad;
899 			}
900 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
901 				/* XXX stat */
902 				icmp6_error(m, ICMP6_PARAM_PROB,
903 				    ICMP6_PARAMPROB_HEADER,
904 				    erroff + opt + 1 - opthead);
905 				return (-1);
906 			}
907 			optlen = IP6OPT_JUMBO_LEN;
908 
909 			/*
910 			 * IPv6 packets that have non 0 payload length
911 			 * must not contain a jumbo payload option.
912 			 */
913 			ip6 = mtod(m, struct ip6_hdr *);
914 			if (ip6->ip6_plen) {
915 				ip6stat.ip6s_badoptions++;
916 				icmp6_error(m, ICMP6_PARAM_PROB,
917 				    ICMP6_PARAMPROB_HEADER,
918 				    erroff + opt - opthead);
919 				return (-1);
920 			}
921 
922 			/*
923 			 * We may see jumbolen in unaligned location, so
924 			 * we'd need to perform bcopy().
925 			 */
926 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
927 			jumboplen = (u_int32_t)htonl(jumboplen);
928 
929 #if 1
930 			/*
931 			 * if there are multiple jumbo payload options,
932 			 * *plenp will be non-zero and the packet will be
933 			 * rejected.
934 			 * the behavior may need some debate in ipngwg -
935 			 * multiple options does not make sense, however,
936 			 * there's no explicit mention in specification.
937 			 */
938 			if (*plenp != 0) {
939 				ip6stat.ip6s_badoptions++;
940 				icmp6_error(m, ICMP6_PARAM_PROB,
941 				    ICMP6_PARAMPROB_HEADER,
942 				    erroff + opt + 2 - opthead);
943 				return (-1);
944 			}
945 #endif
946 
947 			/*
948 			 * jumbo payload length must be larger than 65535.
949 			 */
950 			if (jumboplen <= IPV6_MAXPACKET) {
951 				ip6stat.ip6s_badoptions++;
952 				icmp6_error(m, ICMP6_PARAM_PROB,
953 				    ICMP6_PARAMPROB_HEADER,
954 				    erroff + opt + 2 - opthead);
955 				return (-1);
956 			}
957 			*plenp = jumboplen;
958 
959 			break;
960 		default:		/* unknown option */
961 			if (hbhlen < IP6OPT_MINLEN) {
962 				ip6stat.ip6s_toosmall++;
963 				goto bad;
964 			}
965 			optlen = ip6_unknown_opt(opt, m,
966 			    erroff + opt - opthead);
967 			if (optlen == -1)
968 				return (-1);
969 			optlen += 2;
970 			break;
971 		}
972 	}
973 
974 	return (0);
975 
976   bad:
977 	m_freem(m);
978 	return (-1);
979 }
980 
981 /*
982  * Unknown option processing.
983  * The third argument `off' is the offset from the IPv6 header to the option,
984  * which is necessary if the IPv6 header the and option header and IPv6 header
985  * is not continuous in order to return an ICMPv6 error.
986  */
987 int
988 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
989 {
990 	struct ip6_hdr *ip6;
991 
992 	switch (IP6OPT_TYPE(*optp)) {
993 	case IP6OPT_TYPE_SKIP: /* ignore the option */
994 		return ((int)*(optp + 1));
995 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
996 		m_freem(m);
997 		return (-1);
998 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
999 		ip6stat.ip6s_badoptions++;
1000 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1001 		return (-1);
1002 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1003 		ip6stat.ip6s_badoptions++;
1004 		ip6 = mtod(m, struct ip6_hdr *);
1005 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1006 		    (m->m_flags & (M_BCAST|M_MCAST)))
1007 			m_freem(m);
1008 		else
1009 			icmp6_error(m, ICMP6_PARAM_PROB,
1010 				    ICMP6_PARAMPROB_OPTION, off);
1011 		return (-1);
1012 	}
1013 
1014 	m_freem(m);		/* XXX: NOTREACHED */
1015 	return (-1);
1016 }
1017 
1018 /*
1019  * Create the "control" list for this pcb.
1020  *
1021  * The routine will be called from upper layer handlers like tcp6_input().
1022  * Thus the routine assumes that the caller (tcp6_input) have already
1023  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1024  * very first mbuf on the mbuf chain.
1025  * We may want to add some infinite loop prevention or sanity checks for safety.
1026  * (This applies only when you are using KAME mbuf chain restriction, i.e.
1027  * you are using IP6_EXTHDR_CHECK() not m_pulldown())
1028  */
1029 void
1030 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1031 {
1032 #define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1033 # define in6p_flags	inp_flags
1034 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1035 
1036 #ifdef SO_TIMESTAMP
1037 	if (in6p->inp_socket->so_options & SO_TIMESTAMP) {
1038 		struct timeval tv;
1039 
1040 		microtime(&tv);
1041 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1042 		    SCM_TIMESTAMP, SOL_SOCKET);
1043 		if (*mp)
1044 			mp = &(*mp)->m_next;
1045 	}
1046 #endif
1047 
1048 	/* RFC 2292 sec. 5 */
1049 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1050 		struct in6_pktinfo pi6;
1051 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1052 		if (IN6_IS_SCOPE_EMBED(&pi6.ipi6_addr))
1053 			pi6.ipi6_addr.s6_addr16[1] = 0;
1054 		pi6.ipi6_ifindex =
1055 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1056 		*mp = sbcreatecontrol((caddr_t) &pi6,
1057 		    sizeof(struct in6_pktinfo),
1058 		    IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1059 		if (*mp)
1060 			mp = &(*mp)->m_next;
1061 	}
1062 
1063 	if ((in6p->in6p_flags & IN6P_HOPLIMIT) != 0) {
1064 		int hlim = ip6->ip6_hlim & 0xff;
1065 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1066 		    IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1067 		if (*mp)
1068 			mp = &(*mp)->m_next;
1069 	}
1070 
1071 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1072 		u_int32_t flowinfo;
1073 		int tclass;
1074 
1075 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1076 		flowinfo >>= 20;
1077 
1078 		tclass = flowinfo & 0xff;
1079 		*mp = sbcreatecontrol((caddr_t)&tclass, sizeof(tclass),
1080 		    IPV6_TCLASS, IPPROTO_IPV6);
1081 		if (*mp)
1082 			mp = &(*mp)->m_next;
1083 	}
1084 
1085 	/*
1086 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1087 	 * privilege for the option (see ip6_ctloutput), but it might be too
1088 	 * strict, since there might be some hop-by-hop options which can be
1089 	 * returned to normal user.
1090 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1091 	 */
1092 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0) {
1093 		/*
1094 		 * Check if a hop-by-hop options header is contatined in the
1095 		 * received packet, and if so, store the options as ancillary
1096 		 * data. Note that a hop-by-hop options header must be
1097 		 * just after the IPv6 header, which is assured through the
1098 		 * IPv6 input processing.
1099 		 */
1100 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1101 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1102 			struct ip6_hbh *hbh;
1103 			int hbhlen = 0;
1104 			struct mbuf *ext;
1105 
1106 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1107 			    ip6->ip6_nxt);
1108 			if (ext == NULL) {
1109 				ip6stat.ip6s_tooshort++;
1110 				return;
1111 			}
1112 			hbh = mtod(ext, struct ip6_hbh *);
1113 			hbhlen = (hbh->ip6h_len + 1) << 3;
1114 			if (hbhlen != ext->m_len) {
1115 				m_freem(ext);
1116 				ip6stat.ip6s_tooshort++;
1117 				return;
1118 			}
1119 
1120 			/*
1121 			 * XXX: We copy the whole header even if a
1122 			 * jumbo payload option is included, the option which
1123 			 * is to be removed before returning according to
1124 			 * RFC2292.
1125 			 * Note: this constraint is removed in RFC3542.
1126 			 */
1127 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1128 			    IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1129 			    IPPROTO_IPV6);
1130 			if (*mp)
1131 				mp = &(*mp)->m_next;
1132 			m_freem(ext);
1133 		}
1134 	}
1135 
1136 	/* IPV6_DSTOPTS and IPV6_RTHDR socket options */
1137 	if ((in6p->in6p_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1138 		struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1139 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1140 
1141 		/*
1142 		 * Search for destination options headers or routing
1143 		 * header(s) through the header chain, and stores each
1144 		 * header as ancillary data.
1145 		 * Note that the order of the headers remains in
1146 		 * the chain of ancillary data.
1147 		 */
1148 		while (1) {	/* is explicit loop prevention necessary? */
1149 			struct ip6_ext *ip6e = NULL;
1150 			int elen;
1151 			struct mbuf *ext = NULL;
1152 
1153 			/*
1154 			 * if it is not an extension header, don't try to
1155 			 * pull it from the chain.
1156 			 */
1157 			switch (nxt) {
1158 			case IPPROTO_DSTOPTS:
1159 			case IPPROTO_ROUTING:
1160 			case IPPROTO_HOPOPTS:
1161 			case IPPROTO_AH: /* is it possible? */
1162 				break;
1163 			default:
1164 				goto loopend;
1165 			}
1166 
1167 			ext = ip6_pullexthdr(m, off, nxt);
1168 			if (ext == NULL) {
1169 				ip6stat.ip6s_tooshort++;
1170 				return;
1171 			}
1172 			ip6e = mtod(ext, struct ip6_ext *);
1173 			if (nxt == IPPROTO_AH)
1174 				elen = (ip6e->ip6e_len + 2) << 2;
1175 			else
1176 				elen = (ip6e->ip6e_len + 1) << 3;
1177 			if (elen != ext->m_len) {
1178 				m_freem(ext);
1179 				ip6stat.ip6s_tooshort++;
1180 				return;
1181 			}
1182 
1183 			switch (nxt) {
1184 			case IPPROTO_DSTOPTS:
1185 				if (!(in6p->in6p_flags & IN6P_DSTOPTS))
1186 					break;
1187 
1188 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1189 				    IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1190 				    IPPROTO_IPV6);
1191 				if (*mp)
1192 					mp = &(*mp)->m_next;
1193 				break;
1194 
1195 			case IPPROTO_ROUTING:
1196 				if (!(in6p->in6p_flags & IN6P_RTHDR))
1197 					break;
1198 
1199 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1200 				    IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1201 				    IPPROTO_IPV6);
1202 				if (*mp)
1203 					mp = &(*mp)->m_next;
1204 				break;
1205 
1206 			case IPPROTO_HOPOPTS:
1207 			case IPPROTO_AH: /* is it possible? */
1208 				break;
1209 
1210 			default:
1211 				/*
1212 				 * other cases have been filtered in the above.
1213 				 * none will visit this case.  here we supply
1214 				 * the code just in case (nxt overwritten or
1215 				 * other cases).
1216 				 */
1217 				m_freem(ext);
1218 				goto loopend;
1219 
1220 			}
1221 
1222 			/* proceed with the next header. */
1223 			off += elen;
1224 			nxt = ip6e->ip6e_nxt;
1225 			ip6e = NULL;
1226 			m_freem(ext);
1227 			ext = NULL;
1228 		}
1229 	  loopend:
1230 	  	;
1231 	}
1232 # undef in6p_flags
1233 }
1234 
1235 /*
1236  * pull single extension header from mbuf chain.  returns single mbuf that
1237  * contains the result, or NULL on error.
1238  */
1239 static struct mbuf *
1240 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1241 {
1242 	struct ip6_ext ip6e;
1243 	size_t elen;
1244 	struct mbuf *n;
1245 
1246 #ifdef DIAGNOSTIC
1247 	switch (nxt) {
1248 	case IPPROTO_DSTOPTS:
1249 	case IPPROTO_ROUTING:
1250 	case IPPROTO_HOPOPTS:
1251 	case IPPROTO_AH: /* is it possible? */
1252 		break;
1253 	default:
1254 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1255 	}
1256 #endif
1257 
1258 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1259 	if (nxt == IPPROTO_AH)
1260 		elen = (ip6e.ip6e_len + 2) << 2;
1261 	else
1262 		elen = (ip6e.ip6e_len + 1) << 3;
1263 
1264 	MGET(n, M_DONTWAIT, MT_DATA);
1265 	if (n && elen >= MLEN) {
1266 		MCLGET(n, M_DONTWAIT);
1267 		if ((n->m_flags & M_EXT) == 0) {
1268 			m_free(n);
1269 			n = NULL;
1270 		}
1271 	}
1272 	if (!n)
1273 		return NULL;
1274 
1275 	n->m_len = 0;
1276 	if (elen >= M_TRAILINGSPACE(n)) {
1277 		m_free(n);
1278 		return NULL;
1279 	}
1280 
1281 	m_copydata(m, off, elen, mtod(n, caddr_t));
1282 	n->m_len = elen;
1283 	return n;
1284 }
1285 
1286 /*
1287  * Get pointer to the previous header followed by the header
1288  * currently processed.
1289  * XXX: This function supposes that
1290  *	M includes all headers,
1291  *	the next header field and the header length field of each header
1292  *	are valid, and
1293  *	the sum of each header length equals to OFF.
1294  * Because of these assumptions, this function must be called very
1295  * carefully. Moreover, it will not be used in the near future when
1296  * we develop `neater' mechanism to process extension headers.
1297  */
1298 u_int8_t *
1299 ip6_get_prevhdr(struct mbuf *m, int off)
1300 {
1301 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1302 
1303 	if (off == sizeof(struct ip6_hdr))
1304 		return (&ip6->ip6_nxt);
1305 	else {
1306 		int len, nxt;
1307 		struct ip6_ext *ip6e = NULL;
1308 
1309 		nxt = ip6->ip6_nxt;
1310 		len = sizeof(struct ip6_hdr);
1311 		while (len < off) {
1312 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1313 
1314 			switch (nxt) {
1315 			case IPPROTO_FRAGMENT:
1316 				len += sizeof(struct ip6_frag);
1317 				break;
1318 			case IPPROTO_AH:
1319 				len += (ip6e->ip6e_len + 2) << 2;
1320 				break;
1321 			default:
1322 				len += (ip6e->ip6e_len + 1) << 3;
1323 				break;
1324 			}
1325 			nxt = ip6e->ip6e_nxt;
1326 		}
1327 		if (ip6e)
1328 			return (&ip6e->ip6e_nxt);
1329 		else
1330 			return NULL;
1331 	}
1332 }
1333 
1334 /*
1335  * get next header offset.  m will be retained.
1336  */
1337 int
1338 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1339 {
1340 	struct ip6_hdr ip6;
1341 	struct ip6_ext ip6e;
1342 	struct ip6_frag fh;
1343 
1344 	/* just in case */
1345 	if (m == NULL)
1346 		panic("ip6_nexthdr: m == NULL");
1347 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1348 		return -1;
1349 
1350 	switch (proto) {
1351 	case IPPROTO_IPV6:
1352 		if (m->m_pkthdr.len < off + sizeof(ip6))
1353 			return -1;
1354 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1355 		if (nxtp)
1356 			*nxtp = ip6.ip6_nxt;
1357 		off += sizeof(ip6);
1358 		return off;
1359 
1360 	case IPPROTO_FRAGMENT:
1361 		/*
1362 		 * terminate parsing if it is not the first fragment,
1363 		 * it does not make sense to parse through it.
1364 		 */
1365 		if (m->m_pkthdr.len < off + sizeof(fh))
1366 			return -1;
1367 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1368 		if ((fh.ip6f_offlg & IP6F_OFF_MASK) != 0)
1369 			return -1;
1370 		if (nxtp)
1371 			*nxtp = fh.ip6f_nxt;
1372 		off += sizeof(struct ip6_frag);
1373 		return off;
1374 
1375 	case IPPROTO_AH:
1376 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1377 			return -1;
1378 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1379 		if (nxtp)
1380 			*nxtp = ip6e.ip6e_nxt;
1381 		off += (ip6e.ip6e_len + 2) << 2;
1382 		if (m->m_pkthdr.len < off)
1383 			return -1;
1384 		return off;
1385 
1386 	case IPPROTO_HOPOPTS:
1387 	case IPPROTO_ROUTING:
1388 	case IPPROTO_DSTOPTS:
1389 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1390 			return -1;
1391 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1392 		if (nxtp)
1393 			*nxtp = ip6e.ip6e_nxt;
1394 		off += (ip6e.ip6e_len + 1) << 3;
1395 		if (m->m_pkthdr.len < off)
1396 			return -1;
1397 		return off;
1398 
1399 	case IPPROTO_NONE:
1400 	case IPPROTO_ESP:
1401 	case IPPROTO_IPCOMP:
1402 		/* give up */
1403 		return -1;
1404 
1405 	default:
1406 		return -1;
1407 	}
1408 
1409 	return -1;
1410 }
1411 
1412 /*
1413  * get offset for the last header in the chain.  m will be kept untainted.
1414  */
1415 int
1416 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1417 {
1418 	int newoff;
1419 	int nxt;
1420 
1421 	if (!nxtp) {
1422 		nxt = -1;
1423 		nxtp = &nxt;
1424 	}
1425 	while (1) {
1426 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1427 		if (newoff < 0)
1428 			return off;
1429 		else if (newoff < off)
1430 			return -1;	/* invalid */
1431 		else if (newoff == off)
1432 			return newoff;
1433 
1434 		off = newoff;
1435 		proto = *nxtp;
1436 	}
1437 }
1438 
1439 /*
1440  * System control for IP6
1441  */
1442 
1443 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1444 	0,		0,		0,		0,
1445 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1446 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1447 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1448 	0,		0,		0,		0,
1449 	ENOPROTOOPT
1450 };
1451 
1452 #include <uvm/uvm_extern.h>
1453 #include <sys/sysctl.h>
1454 
1455 int *ipv6ctl_vars[IPV6CTL_MAXID] = IPV6CTL_VARS;
1456 
1457 int
1458 ip6_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp,
1459 	void *newp, size_t newlen)
1460 {
1461 #ifdef MROUTING
1462 	extern int ip6_mrtproto;
1463 	extern struct mrt6stat mrt6stat;
1464 #endif
1465 
1466 	/* All sysctl names at this level are terminal. */
1467 	if (namelen != 1)
1468 		return ENOTDIR;
1469 
1470 	switch (name[0]) {
1471 	case IPV6CTL_KAME_VERSION:
1472 		return sysctl_rdstring(oldp, oldlenp, newp, __KAME_VERSION);
1473 	case IPV6CTL_V6ONLY:
1474 		return sysctl_rdint(oldp, oldlenp, newp, ip6_v6only);
1475 	case IPV6CTL_DAD_PENDING:
1476 		return sysctl_rdint(oldp, oldlenp, newp, ip6_dad_pending);
1477 	case IPV6CTL_STATS:
1478 		if (newp != NULL)
1479 			return (EPERM);
1480 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
1481 		    &ip6stat, sizeof(ip6stat)));
1482 	case IPV6CTL_MRTSTATS:
1483 #ifdef MROUTING
1484 		if (newp != NULL)
1485 			return (EPERM);
1486 		return (sysctl_struct(oldp, oldlenp, newp, newlen,
1487 		    &mrt6stat, sizeof(mrt6stat)));
1488 #else
1489 		return (EOPNOTSUPP);
1490 #endif
1491 	case IPV6CTL_MRTPROTO:
1492 #ifdef MROUTING
1493 		return sysctl_rdint(oldp, oldlenp, newp, ip6_mrtproto);
1494 #else
1495 		return (EOPNOTSUPP);
1496 #endif
1497 	default:
1498 		if (name[0] < IPV6CTL_MAXID)
1499 			return (sysctl_int_arr(ipv6ctl_vars, name, namelen,
1500 			    oldp, oldlenp, newp, newlen));
1501 		return (EOPNOTSUPP);
1502 	}
1503 	/* NOTREACHED */
1504 }
1505