xref: /freebsd/sys/netinet6/ip6_input.c (revision b28624fde638caadd4a89f50c9b7e7da0f98c4d2)
1 /*	$FreeBSD$	*/
2 /*	$KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 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  * 4. 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 "opt_inet.h"
65 #include "opt_inet6.h"
66 #include "opt_ipsec.h"
67 
68 #include <sys/param.h>
69 #include <sys/systm.h>
70 #include <sys/malloc.h>
71 #include <sys/mbuf.h>
72 #include <sys/proc.h>
73 #include <sys/domain.h>
74 #include <sys/protosw.h>
75 #include <sys/socket.h>
76 #include <sys/socketvar.h>
77 #include <sys/errno.h>
78 #include <sys/time.h>
79 #include <sys/kernel.h>
80 #include <sys/syslog.h>
81 
82 #include <net/if.h>
83 #include <net/if_types.h>
84 #include <net/if_dl.h>
85 #include <net/route.h>
86 #include <net/netisr.h>
87 #include <net/pfil.h>
88 
89 #include <netinet/in.h>
90 #include <netinet/in_systm.h>
91 #ifdef INET
92 #include <netinet/ip.h>
93 #include <netinet/ip_icmp.h>
94 #endif /* INET */
95 #include <netinet/ip6.h>
96 #include <netinet6/in6_var.h>
97 #include <netinet6/ip6_var.h>
98 #include <netinet/in_pcb.h>
99 #include <netinet/icmp6.h>
100 #include <netinet6/scope6_var.h>
101 #include <netinet6/in6_ifattach.h>
102 #include <netinet6/nd6.h>
103 
104 #ifdef IPSEC
105 #include <netipsec/ipsec.h>
106 #include <netinet6/ip6_ipsec.h>
107 #include <netipsec/ipsec6.h>
108 #endif /* IPSEC */
109 
110 #include <netinet6/ip6protosw.h>
111 
112 extern struct domain inet6domain;
113 
114 u_char ip6_protox[IPPROTO_MAX];
115 static struct ifqueue ip6intrq;
116 static int ip6qmaxlen = IFQ_MAXLEN;
117 struct in6_ifaddr *in6_ifaddr;
118 
119 extern struct callout in6_tmpaddrtimer_ch;
120 
121 int ip6_forward_srcrt;			/* XXX */
122 int ip6_sourcecheck;			/* XXX */
123 int ip6_sourcecheck_interval;		/* XXX */
124 
125 int ip6_ours_check_algorithm;
126 
127 struct pfil_head inet6_pfil_hook;
128 
129 struct ip6stat ip6stat;
130 
131 static void ip6_init2 __P((void *));
132 static struct ip6aux *ip6_setdstifaddr __P((struct mbuf *, struct in6_ifaddr *));
133 static int ip6_hopopts_input __P((u_int32_t *, u_int32_t *, struct mbuf **, int *));
134 #ifdef PULLDOWN_TEST
135 static struct mbuf *ip6_pullexthdr __P((struct mbuf *, size_t, int));
136 #endif
137 
138 /*
139  * IP6 initialization: fill in IP6 protocol switch table.
140  * All protocols not implemented in kernel go to raw IP6 protocol handler.
141  */
142 void
143 ip6_init(void)
144 {
145 	struct ip6protosw *pr;
146 	int i;
147 
148 #ifdef DIAGNOSTIC
149 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
150 		panic("sizeof(protosw) != sizeof(ip6protosw)");
151 #endif
152 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
153 	if (pr == 0)
154 		panic("ip6_init");
155 
156 	/* Initialize the entire ip_protox[] array to IPPROTO_RAW. */
157 	for (i = 0; i < IPPROTO_MAX; i++)
158 		ip6_protox[i] = pr - inet6sw;
159 	/*
160 	 * Cycle through IP protocols and put them into the appropriate place
161 	 * in ip6_protox[].
162 	 */
163 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
164 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
165 		if (pr->pr_domain->dom_family == PF_INET6 &&
166 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
167 			/* Be careful to only index valid IP protocols. */
168 			if (pr->pr_protocol < IPPROTO_MAX)
169 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
170 		}
171 
172 	/* Initialize packet filter hooks. */
173 	inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
174 	inet6_pfil_hook.ph_af = AF_INET6;
175 	if ((i = pfil_head_register(&inet6_pfil_hook)) != 0)
176 		printf("%s: WARNING: unable to register pfil hook, "
177 			"error %d\n", __func__, i);
178 
179 	ip6intrq.ifq_maxlen = ip6qmaxlen;
180 	mtx_init(&ip6intrq.ifq_mtx, "ip6_inq", NULL, MTX_DEF);
181 	netisr_register(NETISR_IPV6, ip6_input, &ip6intrq, 0);
182 	scope6_init();
183 	addrsel_policy_init();
184 	nd6_init();
185 	frag6_init();
186 	ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
187 }
188 
189 static void
190 ip6_init2(void *dummy)
191 {
192 
193 	/* nd6_timer_init */
194 	callout_init(&nd6_timer_ch, 0);
195 	callout_reset(&nd6_timer_ch, hz, nd6_timer, NULL);
196 
197 	/* timer for regeneranation of temporary addresses randomize ID */
198 	callout_init(&in6_tmpaddrtimer_ch, 0);
199 	callout_reset(&in6_tmpaddrtimer_ch,
200 		      (ip6_temp_preferred_lifetime - ip6_desync_factor -
201 		       ip6_temp_regen_advance) * hz,
202 		      in6_tmpaddrtimer, NULL);
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 extern struct	route_in6 ip6_forward_rt;
210 
211 void
212 ip6_input(struct mbuf *m)
213 {
214 	struct ip6_hdr *ip6;
215 	int off = sizeof(struct ip6_hdr), nest;
216 	u_int32_t plen;
217 	u_int32_t rtalert = ~0;
218 	int nxt, ours = 0;
219 	struct ifnet *deliverifp = NULL;
220 	struct in6_addr odst;
221 	int srcrt = 0;
222 
223 	GIANT_REQUIRED;			/* XXX for now */
224 
225 #ifdef IPSEC
226 	/*
227 	 * should the inner packet be considered authentic?
228 	 * see comment in ah4_input().
229 	 * NB: m cannot be NULL when passed to the input routine
230 	 */
231 
232 	m->m_flags &= ~M_AUTHIPHDR;
233 	m->m_flags &= ~M_AUTHIPDGM;
234 
235 #endif /* IPSEC */
236 
237 	/*
238 	 * make sure we don't have onion peering information into m_tag.
239 	 */
240 	ip6_delaux(m);
241 
242 	/*
243 	 * mbuf statistics
244 	 */
245 	if (m->m_flags & M_EXT) {
246 		if (m->m_next)
247 			ip6stat.ip6s_mext2m++;
248 		else
249 			ip6stat.ip6s_mext1++;
250 	} else {
251 #define M2MMAX	(sizeof(ip6stat.ip6s_m2m)/sizeof(ip6stat.ip6s_m2m[0]))
252 		if (m->m_next) {
253 			if (m->m_flags & M_LOOP) {
254 				ip6stat.ip6s_m2m[loif[0].if_index]++; /* XXX */
255 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
256 				ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
257 			else
258 				ip6stat.ip6s_m2m[0]++;
259 		} else
260 			ip6stat.ip6s_m1++;
261 #undef M2MMAX
262 	}
263 
264 	/* drop the packet if IPv6 operation is disabled on the IF */
265 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
266 		m_freem(m);
267 		return;
268 	}
269 
270 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
271 	ip6stat.ip6s_total++;
272 
273 #ifndef PULLDOWN_TEST
274 	/*
275 	 * L2 bridge code and some other code can return mbuf chain
276 	 * that does not conform to KAME requirement.  too bad.
277 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
278 	 */
279 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
280 		struct mbuf *n;
281 
282 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
283 		if (n)
284 			M_MOVE_PKTHDR(n, m);
285 		if (n && n->m_pkthdr.len > MHLEN) {
286 			MCLGET(n, M_DONTWAIT);
287 			if ((n->m_flags & M_EXT) == 0) {
288 				m_freem(n);
289 				n = NULL;
290 			}
291 		}
292 		if (n == NULL) {
293 			m_freem(m);
294 			return;	/* ENOBUFS */
295 		}
296 
297 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
298 		n->m_len = n->m_pkthdr.len;
299 		m_freem(m);
300 		m = n;
301 	}
302 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
303 #endif
304 
305 	if (m->m_len < sizeof(struct ip6_hdr)) {
306 		struct ifnet *inifp;
307 		inifp = m->m_pkthdr.rcvif;
308 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
309 			ip6stat.ip6s_toosmall++;
310 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
311 			return;
312 		}
313 	}
314 
315 	ip6 = mtod(m, struct ip6_hdr *);
316 
317 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
318 		ip6stat.ip6s_badvers++;
319 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
320 		goto bad;
321 	}
322 
323 	ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
324 
325 	/*
326 	 * Check against address spoofing/corruption.
327 	 */
328 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
329 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
330 		/*
331 		 * XXX: "badscope" is not very suitable for a multicast source.
332 		 */
333 		ip6stat.ip6s_badscope++;
334 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
335 		goto bad;
336 	}
337 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
338 	    !(m->m_flags & M_LOOP)) {
339 		/*
340 		 * In this case, the packet should come from the loopback
341 		 * interface.  However, we cannot just check the if_flags,
342 		 * because ip6_mloopback() passes the "actual" interface
343 		 * as the outgoing/incoming interface.
344 		 */
345 		ip6stat.ip6s_badscope++;
346 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
347 		goto bad;
348 	}
349 
350 #ifdef ALTQ
351 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
352 		/* packet is dropped by traffic conditioner */
353 		return;
354 	}
355 #endif
356 	/*
357 	 * The following check is not documented in specs.  A malicious
358 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
359 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
360 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
361 	 *
362 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
363 	 * support IPv4-less kernel compilation, we cannot support SIIT
364 	 * environment at all.  So, it makes more sense for us to reject any
365 	 * malicious packets for non-SIIT environment, than try to do a
366 	 * partial support for SIIT environment.
367 	 */
368 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
369 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
370 		ip6stat.ip6s_badscope++;
371 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
372 		goto bad;
373 	}
374 #if 0
375 	/*
376 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
377 	 *
378 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
379 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
380 	 * is revised to forbid relaying case.
381 	 */
382 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
383 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
384 		ip6stat.ip6s_badscope++;
385 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
386 		goto bad;
387 	}
388 #endif
389 
390 	/*
391 	 * Run through list of hooks for input packets.
392 	 *
393 	 * NB: Beware of the destination address changing
394 	 *     (e.g. by NAT rewriting).  When this happens,
395 	 *     tell ip6_forward to do the right thing.
396 	 */
397 	odst = ip6->ip6_dst;
398 
399 	/* Jump over all PFIL processing if hooks are not active. */
400 	if (!PFIL_HOOKED(&inet6_pfil_hook))
401 		goto passin;
402 
403 	if (pfil_run_hooks(&inet6_pfil_hook, &m, m->m_pkthdr.rcvif, PFIL_IN, NULL))
404 		return;
405 	if (m == NULL)			/* consumed by filter */
406 		return;
407 	ip6 = mtod(m, struct ip6_hdr *);
408 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
409 
410 passin:
411 	/*
412 	 * Disambiguate address scope zones (if there is ambiguity).
413 	 * We first make sure that the original source or destination address
414 	 * is not in our internal form for scoped addresses.  Such addresses
415 	 * are not necessarily invalid spec-wise, but we cannot accept them due
416 	 * to the usage conflict.
417 	 * in6_setscope() then also checks and rejects the cases where src or
418 	 * dst are the loopback address and the receiving interface
419 	 * is not loopback.
420 	 */
421 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
422 		ip6stat.ip6s_badscope++; /* XXX */
423 		goto bad;
424 	}
425 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
426 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
427 		ip6stat.ip6s_badscope++;
428 		goto bad;
429 	}
430 
431 	/*
432 	 * Multicast check
433 	 */
434 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
435 		struct in6_multi *in6m = 0;
436 
437 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
438 		/*
439 		 * See if we belong to the destination multicast group on the
440 		 * arrival interface.
441 		 */
442 		IN6_LOOKUP_MULTI(ip6->ip6_dst, m->m_pkthdr.rcvif, in6m);
443 		if (in6m)
444 			ours = 1;
445 		else if (!ip6_mrouter) {
446 			ip6stat.ip6s_notmember++;
447 			ip6stat.ip6s_cantforward++;
448 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
449 			goto bad;
450 		}
451 		deliverifp = m->m_pkthdr.rcvif;
452 		goto hbhcheck;
453 	}
454 
455 	/*
456 	 *  Unicast check
457 	 */
458 	if (ip6_forward_rt.ro_rt != NULL &&
459 	    (ip6_forward_rt.ro_rt->rt_flags & RTF_UP) != 0 &&
460 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
461 	    &((struct sockaddr_in6 *)(&ip6_forward_rt.ro_dst))->sin6_addr))
462 		ip6stat.ip6s_forward_cachehit++;
463 	else {
464 		struct sockaddr_in6 *dst6;
465 
466 		if (ip6_forward_rt.ro_rt) {
467 			/* route is down or destination is different */
468 			ip6stat.ip6s_forward_cachemiss++;
469 			RTFREE(ip6_forward_rt.ro_rt);
470 			ip6_forward_rt.ro_rt = 0;
471 		}
472 
473 		bzero(&ip6_forward_rt.ro_dst, sizeof(struct sockaddr_in6));
474 		dst6 = (struct sockaddr_in6 *)&ip6_forward_rt.ro_dst;
475 		dst6->sin6_len = sizeof(struct sockaddr_in6);
476 		dst6->sin6_family = AF_INET6;
477 		dst6->sin6_addr = ip6->ip6_dst;
478 
479 		rtalloc((struct route *)&ip6_forward_rt);
480 	}
481 
482 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
483 
484 	/*
485 	 * Accept the packet if the forwarding interface to the destination
486 	 * according to the routing table is the loopback interface,
487 	 * unless the associated route has a gateway.
488 	 * Note that this approach causes to accept a packet if there is a
489 	 * route to the loopback interface for the destination of the packet.
490 	 * But we think it's even useful in some situations, e.g. when using
491 	 * a special daemon which wants to intercept the packet.
492 	 *
493 	 * XXX: some OSes automatically make a cloned route for the destination
494 	 * of an outgoing packet.  If the outgoing interface of the packet
495 	 * is a loopback one, the kernel would consider the packet to be
496 	 * accepted, even if we have no such address assinged on the interface.
497 	 * We check the cloned flag of the route entry to reject such cases,
498 	 * assuming that route entries for our own addresses are not made by
499 	 * cloning (it should be true because in6_addloop explicitly installs
500 	 * the host route).  However, we might have to do an explicit check
501 	 * while it would be less efficient.  Or, should we rather install a
502 	 * reject route for such a case?
503 	 */
504 	if (ip6_forward_rt.ro_rt &&
505 	    (ip6_forward_rt.ro_rt->rt_flags &
506 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
507 #ifdef RTF_WASCLONED
508 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_WASCLONED) &&
509 #endif
510 #ifdef RTF_CLONED
511 	    !(ip6_forward_rt.ro_rt->rt_flags & RTF_CLONED) &&
512 #endif
513 #if 0
514 	    /*
515 	     * The check below is redundant since the comparison of
516 	     * the destination and the key of the rtentry has
517 	     * already done through looking up the routing table.
518 	     */
519 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
520 	    &rt6_key(ip6_forward_rt.ro_rt)->sin6_addr)
521 #endif
522 	    ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_LOOP) {
523 		struct in6_ifaddr *ia6 =
524 			(struct in6_ifaddr *)ip6_forward_rt.ro_rt->rt_ifa;
525 
526 		/*
527 		 * record address information into m_tag.
528 		 */
529 		(void)ip6_setdstifaddr(m, ia6);
530 
531 		/*
532 		 * packets to a tentative, duplicated, or somehow invalid
533 		 * address must not be accepted.
534 		 */
535 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
536 			/* this address is ready */
537 			ours = 1;
538 			deliverifp = ia6->ia_ifp;	/* correct? */
539 			/* Count the packet in the ip address stats */
540 			ia6->ia_ifa.if_ipackets++;
541 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
542 			goto hbhcheck;
543 		} else {
544 			char ip6bufs[INET6_ADDRSTRLEN];
545 			char ip6bufd[INET6_ADDRSTRLEN];
546 			/* address is not ready, so discard the packet. */
547 			nd6log((LOG_INFO,
548 			    "ip6_input: packet to an unready address %s->%s\n",
549 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
550 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
551 
552 			goto bad;
553 		}
554 	}
555 
556 	/*
557 	 * FAITH (Firewall Aided Internet Translator)
558 	 */
559 	if (ip6_keepfaith) {
560 		if (ip6_forward_rt.ro_rt && ip6_forward_rt.ro_rt->rt_ifp
561 		 && ip6_forward_rt.ro_rt->rt_ifp->if_type == IFT_FAITH) {
562 			/* XXX do we need more sanity checks? */
563 			ours = 1;
564 			deliverifp = ip6_forward_rt.ro_rt->rt_ifp; /* faith */
565 			goto hbhcheck;
566 		}
567 	}
568 
569 	/*
570 	 * Now there is no reason to process the packet if it's not our own
571 	 * and we're not a router.
572 	 */
573 	if (!ip6_forwarding) {
574 		ip6stat.ip6s_cantforward++;
575 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
576 		goto bad;
577 	}
578 
579   hbhcheck:
580 	/*
581 	 * record address information into m_tag, if we don't have one yet.
582 	 * note that we are unable to record it, if the address is not listed
583 	 * as our interface address (e.g. multicast addresses, addresses
584 	 * within FAITH prefixes and such).
585 	 */
586 	if (deliverifp && !ip6_getdstifaddr(m)) {
587 		struct in6_ifaddr *ia6;
588 
589 		ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
590 		if (ia6) {
591 			if (!ip6_setdstifaddr(m, ia6)) {
592 				/*
593 				 * XXX maybe we should drop the packet here,
594 				 * as we could not provide enough information
595 				 * to the upper layers.
596 				 */
597 			}
598 		}
599 	}
600 
601 	/*
602 	 * Process Hop-by-Hop options header if it's contained.
603 	 * m may be modified in ip6_hopopts_input().
604 	 * If a JumboPayload option is included, plen will also be modified.
605 	 */
606 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
607 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
608 		struct ip6_hbh *hbh;
609 
610 		if (ip6_hopopts_input(&plen, &rtalert, &m, &off)) {
611 #if 0	/*touches NULL pointer*/
612 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
613 #endif
614 			return;	/* m have already been freed */
615 		}
616 
617 		/* adjust pointer */
618 		ip6 = mtod(m, struct ip6_hdr *);
619 
620 		/*
621 		 * if the payload length field is 0 and the next header field
622 		 * indicates Hop-by-Hop Options header, then a Jumbo Payload
623 		 * option MUST be included.
624 		 */
625 		if (ip6->ip6_plen == 0 && plen == 0) {
626 			/*
627 			 * Note that if a valid jumbo payload option is
628 			 * contained, ip6_hopopts_input() must set a valid
629 			 * (non-zero) payload length to the variable plen.
630 			 */
631 			ip6stat.ip6s_badoptions++;
632 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
633 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
634 			icmp6_error(m, ICMP6_PARAM_PROB,
635 				    ICMP6_PARAMPROB_HEADER,
636 				    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
637 			return;
638 		}
639 #ifndef PULLDOWN_TEST
640 		/* ip6_hopopts_input() ensures that mbuf is contiguous */
641 		hbh = (struct ip6_hbh *)(ip6 + 1);
642 #else
643 		IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
644 			sizeof(struct ip6_hbh));
645 		if (hbh == NULL) {
646 			ip6stat.ip6s_tooshort++;
647 			return;
648 		}
649 #endif
650 		nxt = hbh->ip6h_nxt;
651 
652 		/*
653 		 * If we are acting as a router and the packet contains a
654 		 * router alert option, see if we know the option value.
655 		 * Currently, we only support the option value for MLD, in which
656 		 * case we should pass the packet to the multicast routing
657 		 * daemon.
658 		 */
659 		if (rtalert != ~0 && ip6_forwarding) {
660 			switch (rtalert) {
661 			case IP6OPT_RTALERT_MLD:
662 				ours = 1;
663 				break;
664 			default:
665 				/*
666 				 * RFC2711 requires unrecognized values must be
667 				 * silently ignored.
668 				 */
669 				break;
670 			}
671 		}
672 	} else
673 		nxt = ip6->ip6_nxt;
674 
675 	/*
676 	 * Check that the amount of data in the buffers
677 	 * is as at least much as the IPv6 header would have us expect.
678 	 * Trim mbufs if longer than we expect.
679 	 * Drop packet if shorter than we expect.
680 	 */
681 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
682 		ip6stat.ip6s_tooshort++;
683 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
684 		goto bad;
685 	}
686 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
687 		if (m->m_len == m->m_pkthdr.len) {
688 			m->m_len = sizeof(struct ip6_hdr) + plen;
689 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
690 		} else
691 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
692 	}
693 
694 	/*
695 	 * Forward if desirable.
696 	 */
697 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
698 		/*
699 		 * If we are acting as a multicast router, all
700 		 * incoming multicast packets are passed to the
701 		 * kernel-level multicast forwarding function.
702 		 * The packet is returned (relatively) intact; if
703 		 * ip6_mforward() returns a non-zero value, the packet
704 		 * must be discarded, else it may be accepted below.
705 		 */
706 		if (ip6_mrouter && ip6_mforward &&
707 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
708 			ip6stat.ip6s_cantforward++;
709 			m_freem(m);
710 			return;
711 		}
712 		if (!ours) {
713 			m_freem(m);
714 			return;
715 		}
716 	} else if (!ours) {
717 		ip6_forward(m, srcrt);
718 		return;
719 	}
720 
721 	ip6 = mtod(m, struct ip6_hdr *);
722 
723 	/*
724 	 * Malicious party may be able to use IPv4 mapped addr to confuse
725 	 * tcp/udp stack and bypass security checks (act as if it was from
726 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
727 	 *
728 	 * For SIIT end node behavior, you may want to disable the check.
729 	 * However, you will  become vulnerable to attacks using IPv4 mapped
730 	 * source.
731 	 */
732 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
733 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
734 		ip6stat.ip6s_badscope++;
735 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
736 		goto bad;
737 	}
738 
739 	/*
740 	 * Tell launch routine the next header
741 	 */
742 	ip6stat.ip6s_delivered++;
743 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
744 	nest = 0;
745 
746 	while (nxt != IPPROTO_DONE) {
747 		if (ip6_hdrnestlimit && (++nest > ip6_hdrnestlimit)) {
748 			ip6stat.ip6s_toomanyhdr++;
749 			goto bad;
750 		}
751 
752 		/*
753 		 * protection against faulty packet - there should be
754 		 * more sanity checks in header chain processing.
755 		 */
756 		if (m->m_pkthdr.len < off) {
757 			ip6stat.ip6s_tooshort++;
758 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
759 			goto bad;
760 		}
761 
762 #ifdef IPSEC
763 		/*
764 		 * enforce IPsec policy checking if we are seeing last header.
765 		 * note that we do not visit this with protocols with pcb layer
766 		 * code - like udp/tcp/raw ip.
767 		 */
768 		if (ip6_ipsec_input(m, nxt))
769 			goto bad;
770 #endif /* IPSEC */
771 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
772 	}
773 	return;
774  bad:
775 	m_freem(m);
776 }
777 
778 /*
779  * set/grab in6_ifaddr correspond to IPv6 destination address.
780  * XXX backward compatibility wrapper
781  */
782 static struct ip6aux *
783 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
784 {
785 	struct ip6aux *ip6a;
786 
787 	ip6a = ip6_addaux(m);
788 	if (ip6a)
789 		ip6a->ip6a_dstia6 = ia6;
790 	return ip6a;	/* NULL if failed to set */
791 }
792 
793 struct in6_ifaddr *
794 ip6_getdstifaddr(struct mbuf *m)
795 {
796 	struct ip6aux *ip6a;
797 
798 	ip6a = ip6_findaux(m);
799 	if (ip6a)
800 		return ip6a->ip6a_dstia6;
801 	else
802 		return NULL;
803 }
804 
805 /*
806  * Hop-by-Hop options header processing. If a valid jumbo payload option is
807  * included, the real payload length will be stored in plenp.
808  *
809  * rtalertp - XXX: should be stored more smart way
810  */
811 static int
812 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
813     struct mbuf **mp, int *offp)
814 {
815 	struct mbuf *m = *mp;
816 	int off = *offp, hbhlen;
817 	struct ip6_hbh *hbh;
818 	u_int8_t *opt;
819 
820 	/* validation of the length of the header */
821 #ifndef PULLDOWN_TEST
822 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
823 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
824 	hbhlen = (hbh->ip6h_len + 1) << 3;
825 
826 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
827 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
828 #else
829 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
830 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
831 	if (hbh == NULL) {
832 		ip6stat.ip6s_tooshort++;
833 		return -1;
834 	}
835 	hbhlen = (hbh->ip6h_len + 1) << 3;
836 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
837 		hbhlen);
838 	if (hbh == NULL) {
839 		ip6stat.ip6s_tooshort++;
840 		return -1;
841 	}
842 #endif
843 	off += hbhlen;
844 	hbhlen -= sizeof(struct ip6_hbh);
845 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
846 
847 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
848 				hbhlen, rtalertp, plenp) < 0)
849 		return (-1);
850 
851 	*offp = off;
852 	*mp = m;
853 	return (0);
854 }
855 
856 /*
857  * Search header for all Hop-by-hop options and process each option.
858  * This function is separate from ip6_hopopts_input() in order to
859  * handle a case where the sending node itself process its hop-by-hop
860  * options header. In such a case, the function is called from ip6_output().
861  *
862  * The function assumes that hbh header is located right after the IPv6 header
863  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
864  * opthead + hbhlen is located in continuous memory region.
865  */
866 int
867 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
868     u_int32_t *rtalertp, u_int32_t *plenp)
869 {
870 	struct ip6_hdr *ip6;
871 	int optlen = 0;
872 	u_int8_t *opt = opthead;
873 	u_int16_t rtalert_val;
874 	u_int32_t jumboplen;
875 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
876 
877 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
878 		switch (*opt) {
879 		case IP6OPT_PAD1:
880 			optlen = 1;
881 			break;
882 		case IP6OPT_PADN:
883 			if (hbhlen < IP6OPT_MINLEN) {
884 				ip6stat.ip6s_toosmall++;
885 				goto bad;
886 			}
887 			optlen = *(opt + 1) + 2;
888 			break;
889 		case IP6OPT_ROUTER_ALERT:
890 			/* XXX may need check for alignment */
891 			if (hbhlen < IP6OPT_RTALERT_LEN) {
892 				ip6stat.ip6s_toosmall++;
893 				goto bad;
894 			}
895 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
896 				/* XXX stat */
897 				icmp6_error(m, ICMP6_PARAM_PROB,
898 				    ICMP6_PARAMPROB_HEADER,
899 				    erroff + opt + 1 - opthead);
900 				return (-1);
901 			}
902 			optlen = IP6OPT_RTALERT_LEN;
903 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
904 			*rtalertp = ntohs(rtalert_val);
905 			break;
906 		case IP6OPT_JUMBO:
907 			/* XXX may need check for alignment */
908 			if (hbhlen < IP6OPT_JUMBO_LEN) {
909 				ip6stat.ip6s_toosmall++;
910 				goto bad;
911 			}
912 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
913 				/* XXX stat */
914 				icmp6_error(m, ICMP6_PARAM_PROB,
915 				    ICMP6_PARAMPROB_HEADER,
916 				    erroff + opt + 1 - opthead);
917 				return (-1);
918 			}
919 			optlen = IP6OPT_JUMBO_LEN;
920 
921 			/*
922 			 * IPv6 packets that have non 0 payload length
923 			 * must not contain a jumbo payload option.
924 			 */
925 			ip6 = mtod(m, struct ip6_hdr *);
926 			if (ip6->ip6_plen) {
927 				ip6stat.ip6s_badoptions++;
928 				icmp6_error(m, ICMP6_PARAM_PROB,
929 				    ICMP6_PARAMPROB_HEADER,
930 				    erroff + opt - opthead);
931 				return (-1);
932 			}
933 
934 			/*
935 			 * We may see jumbolen in unaligned location, so
936 			 * we'd need to perform bcopy().
937 			 */
938 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
939 			jumboplen = (u_int32_t)htonl(jumboplen);
940 
941 #if 1
942 			/*
943 			 * if there are multiple jumbo payload options,
944 			 * *plenp will be non-zero and the packet will be
945 			 * rejected.
946 			 * the behavior may need some debate in ipngwg -
947 			 * multiple options does not make sense, however,
948 			 * there's no explicit mention in specification.
949 			 */
950 			if (*plenp != 0) {
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 #endif
958 
959 			/*
960 			 * jumbo payload length must be larger than 65535.
961 			 */
962 			if (jumboplen <= IPV6_MAXPACKET) {
963 				ip6stat.ip6s_badoptions++;
964 				icmp6_error(m, ICMP6_PARAM_PROB,
965 				    ICMP6_PARAMPROB_HEADER,
966 				    erroff + opt + 2 - opthead);
967 				return (-1);
968 			}
969 			*plenp = jumboplen;
970 
971 			break;
972 		default:		/* unknown option */
973 			if (hbhlen < IP6OPT_MINLEN) {
974 				ip6stat.ip6s_toosmall++;
975 				goto bad;
976 			}
977 			optlen = ip6_unknown_opt(opt, m,
978 			    erroff + opt - opthead);
979 			if (optlen == -1)
980 				return (-1);
981 			optlen += 2;
982 			break;
983 		}
984 	}
985 
986 	return (0);
987 
988   bad:
989 	m_freem(m);
990 	return (-1);
991 }
992 
993 /*
994  * Unknown option processing.
995  * The third argument `off' is the offset from the IPv6 header to the option,
996  * which is necessary if the IPv6 header the and option header and IPv6 header
997  * is not continuous in order to return an ICMPv6 error.
998  */
999 int
1000 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1001 {
1002 	struct ip6_hdr *ip6;
1003 
1004 	switch (IP6OPT_TYPE(*optp)) {
1005 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1006 		return ((int)*(optp + 1));
1007 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1008 		m_freem(m);
1009 		return (-1);
1010 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1011 		ip6stat.ip6s_badoptions++;
1012 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1013 		return (-1);
1014 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1015 		ip6stat.ip6s_badoptions++;
1016 		ip6 = mtod(m, struct ip6_hdr *);
1017 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1018 		    (m->m_flags & (M_BCAST|M_MCAST)))
1019 			m_freem(m);
1020 		else
1021 			icmp6_error(m, ICMP6_PARAM_PROB,
1022 				    ICMP6_PARAMPROB_OPTION, off);
1023 		return (-1);
1024 	}
1025 
1026 	m_freem(m);		/* XXX: NOTREACHED */
1027 	return (-1);
1028 }
1029 
1030 /*
1031  * Create the "control" list for this pcb.
1032  * The function will not modify mbuf chain at all.
1033  *
1034  * with KAME mbuf chain restriction:
1035  * The routine will be called from upper layer handlers like tcp6_input().
1036  * Thus the routine assumes that the caller (tcp6_input) have already
1037  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1038  * very first mbuf on the mbuf chain.
1039  */
1040 void
1041 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1042 {
1043 #define IS2292(x, y)	((in6p->in6p_flags & IN6P_RFC2292) ? (x) : (y))
1044 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1045 
1046 #ifdef SO_TIMESTAMP
1047 	if ((in6p->in6p_socket->so_options & SO_TIMESTAMP) != 0) {
1048 		struct timeval tv;
1049 
1050 		microtime(&tv);
1051 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1052 		    SCM_TIMESTAMP, SOL_SOCKET);
1053 		if (*mp)
1054 			mp = &(*mp)->m_next;
1055 	}
1056 #endif
1057 
1058 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION)
1059 		return;
1060 
1061 	/* RFC 2292 sec. 5 */
1062 	if ((in6p->in6p_flags & IN6P_PKTINFO) != 0) {
1063 		struct in6_pktinfo pi6;
1064 
1065 		bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1066 		in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1067 		pi6.ipi6_ifindex =
1068 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1069 
1070 		*mp = sbcreatecontrol((caddr_t) &pi6,
1071 		    sizeof(struct in6_pktinfo),
1072 		    IS2292(IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1073 		if (*mp)
1074 			mp = &(*mp)->m_next;
1075 	}
1076 
1077 	if ((in6p->in6p_flags & IN6P_HOPLIMIT) != 0) {
1078 		int hlim = ip6->ip6_hlim & 0xff;
1079 
1080 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1081 		    IS2292(IPV6_2292HOPLIMIT, IPV6_HOPLIMIT), IPPROTO_IPV6);
1082 		if (*mp)
1083 			mp = &(*mp)->m_next;
1084 	}
1085 
1086 	if ((in6p->in6p_flags & IN6P_TCLASS) != 0) {
1087 		u_int32_t flowinfo;
1088 		int tclass;
1089 
1090 		flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1091 		flowinfo >>= 20;
1092 
1093 		tclass = flowinfo & 0xff;
1094 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(tclass),
1095 		    IPV6_TCLASS, IPPROTO_IPV6);
1096 		if (*mp)
1097 			mp = &(*mp)->m_next;
1098 	}
1099 
1100 	/*
1101 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1102 	 * privilege for the option (see ip6_ctloutput), but it might be too
1103 	 * strict, since there might be some hop-by-hop options which can be
1104 	 * returned to normal user.
1105 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1106 	 */
1107 	if ((in6p->in6p_flags & IN6P_HOPOPTS) != 0) {
1108 		/*
1109 		 * Check if a hop-by-hop options header is contatined in the
1110 		 * received packet, and if so, store the options as ancillary
1111 		 * data. Note that a hop-by-hop options header must be
1112 		 * just after the IPv6 header, which is assured through the
1113 		 * IPv6 input processing.
1114 		 */
1115 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1116 			struct ip6_hbh *hbh;
1117 			int hbhlen = 0;
1118 #ifdef PULLDOWN_TEST
1119 			struct mbuf *ext;
1120 #endif
1121 
1122 #ifndef PULLDOWN_TEST
1123 			hbh = (struct ip6_hbh *)(ip6 + 1);
1124 			hbhlen = (hbh->ip6h_len + 1) << 3;
1125 #else
1126 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1127 			    ip6->ip6_nxt);
1128 			if (ext == NULL) {
1129 				ip6stat.ip6s_tooshort++;
1130 				return;
1131 			}
1132 			hbh = mtod(ext, struct ip6_hbh *);
1133 			hbhlen = (hbh->ip6h_len + 1) << 3;
1134 			if (hbhlen != ext->m_len) {
1135 				m_freem(ext);
1136 				ip6stat.ip6s_tooshort++;
1137 				return;
1138 			}
1139 #endif
1140 
1141 			/*
1142 			 * XXX: We copy the whole header even if a
1143 			 * jumbo payload option is included, the option which
1144 			 * is to be removed before returning according to
1145 			 * RFC2292.
1146 			 * Note: this constraint is removed in RFC3542
1147 			 */
1148 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1149 			    IS2292(IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1150 			    IPPROTO_IPV6);
1151 			if (*mp)
1152 				mp = &(*mp)->m_next;
1153 #ifdef PULLDOWN_TEST
1154 			m_freem(ext);
1155 #endif
1156 		}
1157 	}
1158 
1159 	if ((in6p->in6p_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1160 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1161 
1162 		/*
1163 		 * Search for destination options headers or routing
1164 		 * header(s) through the header chain, and stores each
1165 		 * header as ancillary data.
1166 		 * Note that the order of the headers remains in
1167 		 * the chain of ancillary data.
1168 		 */
1169 		while (1) {	/* is explicit loop prevention necessary? */
1170 			struct ip6_ext *ip6e = NULL;
1171 			int elen;
1172 #ifdef PULLDOWN_TEST
1173 			struct mbuf *ext = NULL;
1174 #endif
1175 
1176 			/*
1177 			 * if it is not an extension header, don't try to
1178 			 * pull it from the chain.
1179 			 */
1180 			switch (nxt) {
1181 			case IPPROTO_DSTOPTS:
1182 			case IPPROTO_ROUTING:
1183 			case IPPROTO_HOPOPTS:
1184 			case IPPROTO_AH: /* is it possible? */
1185 				break;
1186 			default:
1187 				goto loopend;
1188 			}
1189 
1190 #ifndef PULLDOWN_TEST
1191 			if (off + sizeof(*ip6e) > m->m_len)
1192 				goto loopend;
1193 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1194 			if (nxt == IPPROTO_AH)
1195 				elen = (ip6e->ip6e_len + 2) << 2;
1196 			else
1197 				elen = (ip6e->ip6e_len + 1) << 3;
1198 			if (off + elen > m->m_len)
1199 				goto loopend;
1200 #else
1201 			ext = ip6_pullexthdr(m, off, nxt);
1202 			if (ext == NULL) {
1203 				ip6stat.ip6s_tooshort++;
1204 				return;
1205 			}
1206 			ip6e = mtod(ext, struct ip6_ext *);
1207 			if (nxt == IPPROTO_AH)
1208 				elen = (ip6e->ip6e_len + 2) << 2;
1209 			else
1210 				elen = (ip6e->ip6e_len + 1) << 3;
1211 			if (elen != ext->m_len) {
1212 				m_freem(ext);
1213 				ip6stat.ip6s_tooshort++;
1214 				return;
1215 			}
1216 #endif
1217 
1218 			switch (nxt) {
1219 			case IPPROTO_DSTOPTS:
1220 				if (!(in6p->in6p_flags & IN6P_DSTOPTS))
1221 					break;
1222 
1223 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1224 				    IS2292(IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1225 				    IPPROTO_IPV6);
1226 				if (*mp)
1227 					mp = &(*mp)->m_next;
1228 				break;
1229 			case IPPROTO_ROUTING:
1230 				if (!in6p->in6p_flags & IN6P_RTHDR)
1231 					break;
1232 
1233 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1234 				    IS2292(IPV6_2292RTHDR, IPV6_RTHDR),
1235 				    IPPROTO_IPV6);
1236 				if (*mp)
1237 					mp = &(*mp)->m_next;
1238 				break;
1239 			case IPPROTO_HOPOPTS:
1240 			case IPPROTO_AH: /* is it possible? */
1241 				break;
1242 
1243 			default:
1244 				/*
1245 				 * other cases have been filtered in the above.
1246 				 * none will visit this case.  here we supply
1247 				 * the code just in case (nxt overwritten or
1248 				 * other cases).
1249 				 */
1250 #ifdef PULLDOWN_TEST
1251 				m_freem(ext);
1252 #endif
1253 				goto loopend;
1254 
1255 			}
1256 
1257 			/* proceed with the next header. */
1258 			off += elen;
1259 			nxt = ip6e->ip6e_nxt;
1260 			ip6e = NULL;
1261 #ifdef PULLDOWN_TEST
1262 			m_freem(ext);
1263 			ext = NULL;
1264 #endif
1265 		}
1266 	  loopend:
1267 		;
1268 	}
1269 
1270 #undef IS2292
1271 }
1272 
1273 void
1274 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1275 {
1276 	struct socket *so;
1277 	struct mbuf *m_mtu;
1278 	struct ip6_mtuinfo mtuctl;
1279 
1280 	so =  in6p->inp_socket;
1281 
1282 	if (mtu == NULL)
1283 		return;
1284 
1285 #ifdef DIAGNOSTIC
1286 	if (so == NULL)		/* I believe this is impossible */
1287 		panic("ip6_notify_pmtu: socket is NULL");
1288 #endif
1289 
1290 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1291 	mtuctl.ip6m_mtu = *mtu;
1292 	mtuctl.ip6m_addr = *dst;
1293 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1294 		return;
1295 
1296 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1297 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1298 		return;
1299 
1300 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1301 	    == 0) {
1302 		m_freem(m_mtu);
1303 		/* XXX: should count statistics */
1304 	} else
1305 		sorwakeup(so);
1306 
1307 	return;
1308 }
1309 
1310 #ifdef PULLDOWN_TEST
1311 /*
1312  * pull single extension header from mbuf chain.  returns single mbuf that
1313  * contains the result, or NULL on error.
1314  */
1315 static struct mbuf *
1316 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1317 {
1318 	struct ip6_ext ip6e;
1319 	size_t elen;
1320 	struct mbuf *n;
1321 
1322 #ifdef DIAGNOSTIC
1323 	switch (nxt) {
1324 	case IPPROTO_DSTOPTS:
1325 	case IPPROTO_ROUTING:
1326 	case IPPROTO_HOPOPTS:
1327 	case IPPROTO_AH: /* is it possible? */
1328 		break;
1329 	default:
1330 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1331 	}
1332 #endif
1333 
1334 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1335 	if (nxt == IPPROTO_AH)
1336 		elen = (ip6e.ip6e_len + 2) << 2;
1337 	else
1338 		elen = (ip6e.ip6e_len + 1) << 3;
1339 
1340 	MGET(n, M_DONTWAIT, MT_DATA);
1341 	if (n && elen >= MLEN) {
1342 		MCLGET(n, M_DONTWAIT);
1343 		if ((n->m_flags & M_EXT) == 0) {
1344 			m_free(n);
1345 			n = NULL;
1346 		}
1347 	}
1348 	if (!n)
1349 		return NULL;
1350 
1351 	n->m_len = 0;
1352 	if (elen >= M_TRAILINGSPACE(n)) {
1353 		m_free(n);
1354 		return NULL;
1355 	}
1356 
1357 	m_copydata(m, off, elen, mtod(n, caddr_t));
1358 	n->m_len = elen;
1359 	return n;
1360 }
1361 #endif
1362 
1363 /*
1364  * Get pointer to the previous header followed by the header
1365  * currently processed.
1366  * XXX: This function supposes that
1367  *	M includes all headers,
1368  *	the next header field and the header length field of each header
1369  *	are valid, and
1370  *	the sum of each header length equals to OFF.
1371  * Because of these assumptions, this function must be called very
1372  * carefully. Moreover, it will not be used in the near future when
1373  * we develop `neater' mechanism to process extension headers.
1374  */
1375 char *
1376 ip6_get_prevhdr(struct mbuf *m, int off)
1377 {
1378 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1379 
1380 	if (off == sizeof(struct ip6_hdr))
1381 		return (&ip6->ip6_nxt);
1382 	else {
1383 		int len, nxt;
1384 		struct ip6_ext *ip6e = NULL;
1385 
1386 		nxt = ip6->ip6_nxt;
1387 		len = sizeof(struct ip6_hdr);
1388 		while (len < off) {
1389 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1390 
1391 			switch (nxt) {
1392 			case IPPROTO_FRAGMENT:
1393 				len += sizeof(struct ip6_frag);
1394 				break;
1395 			case IPPROTO_AH:
1396 				len += (ip6e->ip6e_len + 2) << 2;
1397 				break;
1398 			default:
1399 				len += (ip6e->ip6e_len + 1) << 3;
1400 				break;
1401 			}
1402 			nxt = ip6e->ip6e_nxt;
1403 		}
1404 		if (ip6e)
1405 			return (&ip6e->ip6e_nxt);
1406 		else
1407 			return NULL;
1408 	}
1409 }
1410 
1411 /*
1412  * get next header offset.  m will be retained.
1413  */
1414 int
1415 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1416 {
1417 	struct ip6_hdr ip6;
1418 	struct ip6_ext ip6e;
1419 	struct ip6_frag fh;
1420 
1421 	/* just in case */
1422 	if (m == NULL)
1423 		panic("ip6_nexthdr: m == NULL");
1424 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1425 		return -1;
1426 
1427 	switch (proto) {
1428 	case IPPROTO_IPV6:
1429 		if (m->m_pkthdr.len < off + sizeof(ip6))
1430 			return -1;
1431 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1432 		if (nxtp)
1433 			*nxtp = ip6.ip6_nxt;
1434 		off += sizeof(ip6);
1435 		return off;
1436 
1437 	case IPPROTO_FRAGMENT:
1438 		/*
1439 		 * terminate parsing if it is not the first fragment,
1440 		 * it does not make sense to parse through it.
1441 		 */
1442 		if (m->m_pkthdr.len < off + sizeof(fh))
1443 			return -1;
1444 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1445 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1446 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1447 			return -1;
1448 		if (nxtp)
1449 			*nxtp = fh.ip6f_nxt;
1450 		off += sizeof(struct ip6_frag);
1451 		return off;
1452 
1453 	case IPPROTO_AH:
1454 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1455 			return -1;
1456 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1457 		if (nxtp)
1458 			*nxtp = ip6e.ip6e_nxt;
1459 		off += (ip6e.ip6e_len + 2) << 2;
1460 		return off;
1461 
1462 	case IPPROTO_HOPOPTS:
1463 	case IPPROTO_ROUTING:
1464 	case IPPROTO_DSTOPTS:
1465 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1466 			return -1;
1467 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1468 		if (nxtp)
1469 			*nxtp = ip6e.ip6e_nxt;
1470 		off += (ip6e.ip6e_len + 1) << 3;
1471 		return off;
1472 
1473 	case IPPROTO_NONE:
1474 	case IPPROTO_ESP:
1475 	case IPPROTO_IPCOMP:
1476 		/* give up */
1477 		return -1;
1478 
1479 	default:
1480 		return -1;
1481 	}
1482 
1483 	return -1;
1484 }
1485 
1486 /*
1487  * get offset for the last header in the chain.  m will be kept untainted.
1488  */
1489 int
1490 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1491 {
1492 	int newoff;
1493 	int nxt;
1494 
1495 	if (!nxtp) {
1496 		nxt = -1;
1497 		nxtp = &nxt;
1498 	}
1499 	while (1) {
1500 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1501 		if (newoff < 0)
1502 			return off;
1503 		else if (newoff < off)
1504 			return -1;	/* invalid */
1505 		else if (newoff == off)
1506 			return newoff;
1507 
1508 		off = newoff;
1509 		proto = *nxtp;
1510 	}
1511 }
1512 
1513 struct ip6aux *
1514 ip6_addaux(struct mbuf *m)
1515 {
1516 	struct m_tag *mtag;
1517 
1518 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1519 	if (!mtag) {
1520 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1521 		    M_NOWAIT);
1522 		if (mtag) {
1523 			m_tag_prepend(m, mtag);
1524 			bzero(mtag + 1, sizeof(struct ip6aux));
1525 		}
1526 	}
1527 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1528 }
1529 
1530 struct ip6aux *
1531 ip6_findaux(struct mbuf *m)
1532 {
1533 	struct m_tag *mtag;
1534 
1535 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1536 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1537 }
1538 
1539 void
1540 ip6_delaux(struct mbuf *m)
1541 {
1542 	struct m_tag *mtag;
1543 
1544 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1545 	if (mtag)
1546 		m_tag_delete(m, mtag);
1547 }
1548 
1549 /*
1550  * System control for IP6
1551  */
1552 
1553 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1554 	0,		0,		0,		0,
1555 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1556 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1557 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1558 	0,		0,		0,		0,
1559 	ENOPROTOOPT
1560 };
1561