xref: /freebsd/sys/netinet6/ip6_input.c (revision 7cd2dcf07629713e5a3d60472cfe4701b705a167)
1 /*-
2  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
3  * All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. Neither the name of the project nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	$KAME: ip6_input.c,v 1.259 2002/01/21 04:58:09 jinmei Exp $
30  */
31 
32 /*-
33  * Copyright (c) 1982, 1986, 1988, 1993
34  *	The Regents of the University of California.  All rights reserved.
35  *
36  * Redistribution and use in source and binary forms, with or without
37  * modification, are permitted provided that the following conditions
38  * are met:
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 4. Neither the name of the University nor the names of its contributors
45  *    may be used to endorse or promote products derived from this software
46  *    without specific prior written permission.
47  *
48  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
58  * SUCH DAMAGE.
59  *
60  *	@(#)ip_input.c	8.2 (Berkeley) 1/4/94
61  */
62 
63 #include <sys/cdefs.h>
64 __FBSDID("$FreeBSD$");
65 
66 #include "opt_inet.h"
67 #include "opt_inet6.h"
68 #include "opt_ipfw.h"
69 #include "opt_ipsec.h"
70 #include "opt_route.h"
71 
72 #include <sys/param.h>
73 #include <sys/systm.h>
74 #include <sys/malloc.h>
75 #include <sys/mbuf.h>
76 #include <sys/proc.h>
77 #include <sys/domain.h>
78 #include <sys/protosw.h>
79 #include <sys/socket.h>
80 #include <sys/socketvar.h>
81 #include <sys/errno.h>
82 #include <sys/time.h>
83 #include <sys/kernel.h>
84 #include <sys/syslog.h>
85 
86 #include <net/if.h>
87 #include <net/if_types.h>
88 #include <net/if_dl.h>
89 #include <net/route.h>
90 #include <net/netisr.h>
91 #include <net/pfil.h>
92 #include <net/vnet.h>
93 
94 #include <netinet/in.h>
95 #include <netinet/ip_var.h>
96 #include <netinet/in_systm.h>
97 #include <net/if_llatbl.h>
98 #ifdef INET
99 #include <netinet/ip.h>
100 #include <netinet/ip_icmp.h>
101 #endif /* INET */
102 #include <netinet/ip6.h>
103 #include <netinet6/in6_var.h>
104 #include <netinet6/ip6_var.h>
105 #include <netinet/in_pcb.h>
106 #include <netinet/icmp6.h>
107 #include <netinet6/scope6_var.h>
108 #include <netinet6/in6_ifattach.h>
109 #include <netinet6/nd6.h>
110 
111 #ifdef IPSEC
112 #include <netipsec/ipsec.h>
113 #include <netinet6/ip6_ipsec.h>
114 #include <netipsec/ipsec6.h>
115 #endif /* IPSEC */
116 
117 #include <netinet6/ip6protosw.h>
118 
119 #ifdef FLOWTABLE
120 #include <net/flowtable.h>
121 VNET_DECLARE(int, ip6_output_flowtable_size);
122 #define	V_ip6_output_flowtable_size	VNET(ip6_output_flowtable_size)
123 #endif
124 
125 extern struct domain inet6domain;
126 
127 u_char ip6_protox[IPPROTO_MAX];
128 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
129 
130 static struct netisr_handler ip6_nh = {
131 	.nh_name = "ip6",
132 	.nh_handler = ip6_input,
133 	.nh_proto = NETISR_IPV6,
134 	.nh_policy = NETISR_POLICY_FLOW,
135 };
136 
137 VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch);
138 #define	V_in6_tmpaddrtimer_ch		VNET(in6_tmpaddrtimer_ch)
139 
140 VNET_DEFINE(struct pfil_head, inet6_pfil_hook);
141 
142 VNET_DEFINE(struct ip6stat, ip6stat);
143 
144 struct rwlock in6_ifaddr_lock;
145 RW_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
146 
147 static void ip6_init2(void *);
148 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *);
149 static struct ip6aux *ip6_addaux(struct mbuf *);
150 static struct ip6aux *ip6_findaux(struct mbuf *m);
151 static void ip6_delaux (struct mbuf *);
152 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
153 #ifdef PULLDOWN_TEST
154 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
155 #endif
156 
157 /*
158  * IP6 initialization: fill in IP6 protocol switch table.
159  * All protocols not implemented in kernel go to raw IP6 protocol handler.
160  */
161 void
162 ip6_init(void)
163 {
164 	struct ip6protosw *pr;
165 	int i;
166 
167 	TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
168 	    &V_ip6_auto_linklocal);
169 	TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
170 	TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
171 
172 	TAILQ_INIT(&V_in6_ifaddrhead);
173 
174 	/* Initialize packet filter hooks. */
175 	V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
176 	V_inet6_pfil_hook.ph_af = AF_INET6;
177 	if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0)
178 		printf("%s: WARNING: unable to register pfil hook, "
179 			"error %d\n", __func__, i);
180 
181 	scope6_init();
182 	addrsel_policy_init();
183 	nd6_init();
184 	frag6_init();
185 
186 #ifdef FLOWTABLE
187 	if (TUNABLE_INT_FETCH("net.inet6.ip6.output_flowtable_size",
188 		&V_ip6_output_flowtable_size)) {
189 		if (V_ip6_output_flowtable_size < 256)
190 			V_ip6_output_flowtable_size = 256;
191 		if (!powerof2(V_ip6_output_flowtable_size)) {
192 			printf("flowtable must be power of 2 size\n");
193 			V_ip6_output_flowtable_size = 2048;
194 		}
195 	} else {
196 		/*
197 		 * round up to the next power of 2
198 		 */
199 		V_ip6_output_flowtable_size = 1 << fls((1024 + maxusers * 64)-1);
200 	}
201 	V_ip6_ft = flowtable_alloc("ipv6", V_ip6_output_flowtable_size, FL_IPV6|FL_PCPU);
202 #endif
203 
204 	V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
205 
206 	/* Skip global initialization stuff for non-default instances. */
207 	if (!IS_DEFAULT_VNET(curvnet))
208 		return;
209 
210 #ifdef DIAGNOSTIC
211 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
212 		panic("sizeof(protosw) != sizeof(ip6protosw)");
213 #endif
214 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
215 	if (pr == NULL)
216 		panic("ip6_init");
217 
218 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
219 	for (i = 0; i < IPPROTO_MAX; i++)
220 		ip6_protox[i] = pr - inet6sw;
221 	/*
222 	 * Cycle through IP protocols and put them into the appropriate place
223 	 * in ip6_protox[].
224 	 */
225 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
226 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
227 		if (pr->pr_domain->dom_family == PF_INET6 &&
228 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
229 			/* Be careful to only index valid IP protocols. */
230 			if (pr->pr_protocol < IPPROTO_MAX)
231 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
232 		}
233 
234 	netisr_register(&ip6_nh);
235 }
236 
237 /*
238  * The protocol to be inserted into ip6_protox[] must be already registered
239  * in inet6sw[], either statically or through pf_proto_register().
240  */
241 int
242 ip6proto_register(short ip6proto)
243 {
244 	struct ip6protosw *pr;
245 
246 	/* Sanity checks. */
247 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
248 		return (EPROTONOSUPPORT);
249 
250 	/*
251 	 * The protocol slot must not be occupied by another protocol
252 	 * already.  An index pointing to IPPROTO_RAW is unused.
253 	 */
254 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
255 	if (pr == NULL)
256 		return (EPFNOSUPPORT);
257 	if (ip6_protox[ip6proto] != pr - inet6sw)	/* IPPROTO_RAW */
258 		return (EEXIST);
259 
260 	/*
261 	 * Find the protocol position in inet6sw[] and set the index.
262 	 */
263 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
264 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) {
265 		if (pr->pr_domain->dom_family == PF_INET6 &&
266 		    pr->pr_protocol && pr->pr_protocol == ip6proto) {
267 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
268 			return (0);
269 		}
270 	}
271 	return (EPROTONOSUPPORT);
272 }
273 
274 int
275 ip6proto_unregister(short ip6proto)
276 {
277 	struct ip6protosw *pr;
278 
279 	/* Sanity checks. */
280 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
281 		return (EPROTONOSUPPORT);
282 
283 	/* Check if the protocol was indeed registered. */
284 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
285 	if (pr == NULL)
286 		return (EPFNOSUPPORT);
287 	if (ip6_protox[ip6proto] == pr - inet6sw)	/* IPPROTO_RAW */
288 		return (ENOENT);
289 
290 	/* Reset the protocol slot to IPPROTO_RAW. */
291 	ip6_protox[ip6proto] = pr - inet6sw;
292 	return (0);
293 }
294 
295 #ifdef VIMAGE
296 void
297 ip6_destroy()
298 {
299 
300 	nd6_destroy();
301 	callout_drain(&V_in6_tmpaddrtimer_ch);
302 }
303 #endif
304 
305 static int
306 ip6_init2_vnet(const void *unused __unused)
307 {
308 
309 	/* nd6_timer_init */
310 	callout_init(&V_nd6_timer_ch, 0);
311 	callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
312 
313 	/* timer for regeneranation of temporary addresses randomize ID */
314 	callout_init(&V_in6_tmpaddrtimer_ch, 0);
315 	callout_reset(&V_in6_tmpaddrtimer_ch,
316 		      (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
317 		       V_ip6_temp_regen_advance) * hz,
318 		      in6_tmpaddrtimer, curvnet);
319 
320 	return (0);
321 }
322 
323 static void
324 ip6_init2(void *dummy)
325 {
326 
327 	ip6_init2_vnet(NULL);
328 }
329 
330 /* cheat */
331 /* This must be after route_init(), which is now SI_ORDER_THIRD */
332 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
333 
334 static int
335 ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off,
336     int *nxt, int *ours)
337 {
338 	struct ip6_hdr *ip6;
339 	struct ip6_hbh *hbh;
340 
341 	if (ip6_hopopts_input(plen, rtalert, &m, off)) {
342 #if 0	/*touches NULL pointer*/
343 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
344 #endif
345 		goto out;	/* m have already been freed */
346 	}
347 
348 	/* adjust pointer */
349 	ip6 = mtod(m, struct ip6_hdr *);
350 
351 	/*
352 	 * if the payload length field is 0 and the next header field
353 	 * indicates Hop-by-Hop Options header, then a Jumbo Payload
354 	 * option MUST be included.
355 	 */
356 	if (ip6->ip6_plen == 0 && *plen == 0) {
357 		/*
358 		 * Note that if a valid jumbo payload option is
359 		 * contained, ip6_hopopts_input() must set a valid
360 		 * (non-zero) payload length to the variable plen.
361 		 */
362 		V_ip6stat.ip6s_badoptions++;
363 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
364 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
365 		icmp6_error(m, ICMP6_PARAM_PROB,
366 			    ICMP6_PARAMPROB_HEADER,
367 			    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
368 		goto out;
369 	}
370 #ifndef PULLDOWN_TEST
371 	/* ip6_hopopts_input() ensures that mbuf is contiguous */
372 	hbh = (struct ip6_hbh *)(ip6 + 1);
373 #else
374 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
375 		sizeof(struct ip6_hbh));
376 	if (hbh == NULL) {
377 		V_ip6stat.ip6s_tooshort++;
378 		goto out;
379 	}
380 #endif
381 	*nxt = hbh->ip6h_nxt;
382 
383 	/*
384 	 * If we are acting as a router and the packet contains a
385 	 * router alert option, see if we know the option value.
386 	 * Currently, we only support the option value for MLD, in which
387 	 * case we should pass the packet to the multicast routing
388 	 * daemon.
389 	 */
390 	if (*rtalert != ~0) {
391 		switch (*rtalert) {
392 		case IP6OPT_RTALERT_MLD:
393 			if (V_ip6_forwarding)
394 				*ours = 1;
395 			break;
396 		default:
397 			/*
398 			 * RFC2711 requires unrecognized values must be
399 			 * silently ignored.
400 			 */
401 			break;
402 		}
403 	}
404 
405 	return (0);
406 
407 out:
408 	return (1);
409 }
410 
411 void
412 ip6_input(struct mbuf *m)
413 {
414 	struct ip6_hdr *ip6;
415 	int off = sizeof(struct ip6_hdr), nest;
416 	u_int32_t plen;
417 	u_int32_t rtalert = ~0;
418 	int nxt, ours = 0;
419 	struct ifnet *deliverifp = NULL, *ifp = NULL;
420 	struct in6_addr odst;
421 	struct route_in6 rin6;
422 	int srcrt = 0;
423 	struct llentry *lle = NULL;
424 	struct sockaddr_in6 dst6, *dst;
425 
426 	bzero(&rin6, sizeof(struct route_in6));
427 #ifdef IPSEC
428 	/*
429 	 * should the inner packet be considered authentic?
430 	 * see comment in ah4_input().
431 	 * NB: m cannot be NULL when passed to the input routine
432 	 */
433 
434 	m->m_flags &= ~M_AUTHIPHDR;
435 	m->m_flags &= ~M_AUTHIPDGM;
436 
437 #endif /* IPSEC */
438 
439 	/*
440 	 * make sure we don't have onion peering information into m_tag.
441 	 */
442 	ip6_delaux(m);
443 
444 	if (m->m_flags & M_FASTFWD_OURS) {
445 		/*
446 		 * Firewall changed destination to local.
447 		 */
448 		m->m_flags &= ~M_FASTFWD_OURS;
449 		ours = 1;
450 		deliverifp = m->m_pkthdr.rcvif;
451 		ip6 = mtod(m, struct ip6_hdr *);
452 		goto hbhcheck;
453 	}
454 
455 	/*
456 	 * mbuf statistics
457 	 */
458 	if (m->m_flags & M_EXT) {
459 		if (m->m_next)
460 			V_ip6stat.ip6s_mext2m++;
461 		else
462 			V_ip6stat.ip6s_mext1++;
463 	} else {
464 #define M2MMAX	(sizeof(V_ip6stat.ip6s_m2m)/sizeof(V_ip6stat.ip6s_m2m[0]))
465 		if (m->m_next) {
466 			if (m->m_flags & M_LOOP) {
467 				V_ip6stat.ip6s_m2m[V_loif->if_index]++;
468 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
469 				V_ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
470 			else
471 				V_ip6stat.ip6s_m2m[0]++;
472 		} else
473 			V_ip6stat.ip6s_m1++;
474 #undef M2MMAX
475 	}
476 
477 	/* drop the packet if IPv6 operation is disabled on the IF */
478 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
479 		m_freem(m);
480 		return;
481 	}
482 
483 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
484 	V_ip6stat.ip6s_total++;
485 
486 #ifndef PULLDOWN_TEST
487 	/*
488 	 * L2 bridge code and some other code can return mbuf chain
489 	 * that does not conform to KAME requirement.  too bad.
490 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
491 	 */
492 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
493 		struct mbuf *n;
494 
495 		MGETHDR(n, M_DONTWAIT, MT_HEADER);
496 		if (n)
497 			M_MOVE_PKTHDR(n, m);
498 		if (n && n->m_pkthdr.len > MHLEN) {
499 			MCLGET(n, M_DONTWAIT);
500 			if ((n->m_flags & M_EXT) == 0) {
501 				m_freem(n);
502 				n = NULL;
503 			}
504 		}
505 		if (n == NULL) {
506 			m_freem(m);
507 			return;	/* ENOBUFS */
508 		}
509 
510 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
511 		n->m_len = n->m_pkthdr.len;
512 		m_freem(m);
513 		m = n;
514 	}
515 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
516 #endif
517 
518 	if (m->m_len < sizeof(struct ip6_hdr)) {
519 		struct ifnet *inifp;
520 		inifp = m->m_pkthdr.rcvif;
521 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
522 			V_ip6stat.ip6s_toosmall++;
523 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
524 			return;
525 		}
526 	}
527 
528 	ip6 = mtod(m, struct ip6_hdr *);
529 
530 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
531 		V_ip6stat.ip6s_badvers++;
532 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
533 		goto bad;
534 	}
535 
536 	V_ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
537 
538 	/*
539 	 * Check against address spoofing/corruption.
540 	 */
541 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
542 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
543 		/*
544 		 * XXX: "badscope" is not very suitable for a multicast source.
545 		 */
546 		V_ip6stat.ip6s_badscope++;
547 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
548 		goto bad;
549 	}
550 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
551 	    !(m->m_flags & M_LOOP)) {
552 		/*
553 		 * In this case, the packet should come from the loopback
554 		 * interface.  However, we cannot just check the if_flags,
555 		 * because ip6_mloopback() passes the "actual" interface
556 		 * as the outgoing/incoming interface.
557 		 */
558 		V_ip6stat.ip6s_badscope++;
559 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
560 		goto bad;
561 	}
562 
563 #ifdef ALTQ
564 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
565 		/* packet is dropped by traffic conditioner */
566 		return;
567 	}
568 #endif
569 	/*
570 	 * The following check is not documented in specs.  A malicious
571 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
572 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
573 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
574 	 *
575 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
576 	 * support IPv4-less kernel compilation, we cannot support SIIT
577 	 * environment at all.  So, it makes more sense for us to reject any
578 	 * malicious packets for non-SIIT environment, than try to do a
579 	 * partial support for SIIT environment.
580 	 */
581 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
582 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
583 		V_ip6stat.ip6s_badscope++;
584 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
585 		goto bad;
586 	}
587 #if 0
588 	/*
589 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
590 	 *
591 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
592 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
593 	 * is revised to forbid relaying case.
594 	 */
595 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
596 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
597 		V_ip6stat.ip6s_badscope++;
598 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
599 		goto bad;
600 	}
601 #endif
602 #ifdef IPSEC
603 	/*
604 	 * Bypass packet filtering for packets previously handled by IPsec.
605 	 */
606 	if (ip6_ipsec_filtertunnel(m))
607 		goto passin;
608 #endif /* IPSEC */
609 
610 	/*
611 	 * Run through list of hooks for input packets.
612 	 *
613 	 * NB: Beware of the destination address changing
614 	 *     (e.g. by NAT rewriting).  When this happens,
615 	 *     tell ip6_forward to do the right thing.
616 	 */
617 	odst = ip6->ip6_dst;
618 
619 	/* Jump over all PFIL processing if hooks are not active. */
620 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
621 		goto passin;
622 
623 	if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
624 	    m->m_pkthdr.rcvif, PFIL_IN, NULL))
625 		return;
626 	if (m == NULL)			/* consumed by filter */
627 		return;
628 	ip6 = mtod(m, struct ip6_hdr *);
629 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
630 
631 	if (m->m_flags & M_FASTFWD_OURS) {
632 		m->m_flags &= ~M_FASTFWD_OURS;
633 		ours = 1;
634 		deliverifp = m->m_pkthdr.rcvif;
635 		goto hbhcheck;
636 	}
637 	if ((m->m_flags & M_IP6_NEXTHOP) &&
638 	    m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
639 		/*
640 		 * Directly ship the packet on.  This allows forwarding
641 		 * packets originally destined to us to some other directly
642 		 * connected host.
643 		 */
644 		ip6_forward(m, 1);
645 		goto out;
646 	}
647 
648 passin:
649 	/*
650 	 * Disambiguate address scope zones (if there is ambiguity).
651 	 * We first make sure that the original source or destination address
652 	 * is not in our internal form for scoped addresses.  Such addresses
653 	 * are not necessarily invalid spec-wise, but we cannot accept them due
654 	 * to the usage conflict.
655 	 * in6_setscope() then also checks and rejects the cases where src or
656 	 * dst are the loopback address and the receiving interface
657 	 * is not loopback.
658 	 */
659 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
660 		V_ip6stat.ip6s_badscope++; /* XXX */
661 		goto bad;
662 	}
663 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
664 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
665 		V_ip6stat.ip6s_badscope++;
666 		goto bad;
667 	}
668 
669 	/*
670 	 * Multicast check. Assume packet is for us to avoid
671 	 * prematurely taking locks.
672 	 */
673 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
674 		ours = 1;
675 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
676 		deliverifp = m->m_pkthdr.rcvif;
677 		goto hbhcheck;
678 	}
679 
680 	/*
681 	 *  Unicast check
682 	 */
683 
684 	bzero(&dst6, sizeof(dst6));
685 	dst6.sin6_family = AF_INET6;
686 	dst6.sin6_len = sizeof(struct sockaddr_in6);
687 	dst6.sin6_addr = ip6->ip6_dst;
688 	ifp = m->m_pkthdr.rcvif;
689 	IF_AFDATA_LOCK(ifp);
690 	lle = lla_lookup(LLTABLE6(ifp), 0,
691 	     (struct sockaddr *)&dst6);
692 	IF_AFDATA_UNLOCK(ifp);
693 	if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) {
694 		struct ifaddr *ifa;
695 		struct in6_ifaddr *ia6;
696 		int bad;
697 
698 		bad = 1;
699 #define	sa_equal(a1, a2)						\
700 	(bcmp((a1), (a2), ((a1))->sin6_len) == 0)
701 		IF_ADDR_RLOCK(ifp);
702 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
703 			if (ifa->ifa_addr->sa_family != dst6.sin6_family)
704 				continue;
705 			if (sa_equal(&dst6, ifa->ifa_addr))
706 				break;
707 		}
708 		KASSERT(ifa != NULL, ("%s: ifa not found for lle %p",
709 		    __func__, lle));
710 #undef sa_equal
711 
712 		ia6 = (struct in6_ifaddr *)ifa;
713 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
714 			/* Count the packet in the ip address stats */
715 			ia6->ia_ifa.if_ipackets++;
716 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
717 
718 			/*
719 			 * record address information into m_tag.
720 			 */
721 			(void)ip6_setdstifaddr(m, ia6);
722 
723 			bad = 0;
724 		} else {
725 			char ip6bufs[INET6_ADDRSTRLEN];
726 			char ip6bufd[INET6_ADDRSTRLEN];
727 			/* address is not ready, so discard the packet. */
728 			nd6log((LOG_INFO,
729 			    "ip6_input: packet to an unready address %s->%s\n",
730 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
731 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
732 		}
733 		IF_ADDR_RUNLOCK(ifp);
734 		LLE_RUNLOCK(lle);
735 		if (bad)
736 			goto bad;
737 		else {
738 			ours = 1;
739 			deliverifp = ifp;
740 			goto hbhcheck;
741 		}
742 	}
743 	if (lle != NULL)
744 		LLE_RUNLOCK(lle);
745 
746 	dst = &rin6.ro_dst;
747 	dst->sin6_len = sizeof(struct sockaddr_in6);
748 	dst->sin6_family = AF_INET6;
749 	dst->sin6_addr = ip6->ip6_dst;
750 	rin6.ro_rt = in6_rtalloc1((struct sockaddr *)dst, 0, 0, M_GETFIB(m));
751 	if (rin6.ro_rt)
752 		RT_UNLOCK(rin6.ro_rt);
753 
754 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
755 
756 	/*
757 	 * Accept the packet if the forwarding interface to the destination
758 	 * according to the routing table is the loopback interface,
759 	 * unless the associated route has a gateway.
760 	 * Note that this approach causes to accept a packet if there is a
761 	 * route to the loopback interface for the destination of the packet.
762 	 * But we think it's even useful in some situations, e.g. when using
763 	 * a special daemon which wants to intercept the packet.
764 	 *
765 	 * XXX: some OSes automatically make a cloned route for the destination
766 	 * of an outgoing packet.  If the outgoing interface of the packet
767 	 * is a loopback one, the kernel would consider the packet to be
768 	 * accepted, even if we have no such address assinged on the interface.
769 	 * We check the cloned flag of the route entry to reject such cases,
770 	 * assuming that route entries for our own addresses are not made by
771 	 * cloning (it should be true because in6_addloop explicitly installs
772 	 * the host route).  However, we might have to do an explicit check
773 	 * while it would be less efficient.  Or, should we rather install a
774 	 * reject route for such a case?
775 	 */
776 	if (rin6.ro_rt &&
777 	    (rin6.ro_rt->rt_flags &
778 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
779 #ifdef RTF_WASCLONED
780 	    !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
781 #endif
782 #ifdef RTF_CLONED
783 	    !(rin6.ro_rt->rt_flags & RTF_CLONED) &&
784 #endif
785 #if 0
786 	    /*
787 	     * The check below is redundant since the comparison of
788 	     * the destination and the key of the rtentry has
789 	     * already done through looking up the routing table.
790 	     */
791 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
792 	    &rt6_key(rin6.ro_rt)->sin6_addr)
793 #endif
794 	    rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
795 		int free_ia6 = 0;
796 		struct in6_ifaddr *ia6;
797 
798 		/*
799 		 * found the loopback route to the interface address
800 		 */
801 		if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) {
802 			struct sockaddr_in6 dest6;
803 
804 			bzero(&dest6, sizeof(dest6));
805 			dest6.sin6_family = AF_INET6;
806 			dest6.sin6_len = sizeof(dest6);
807 			dest6.sin6_addr = ip6->ip6_dst;
808 			ia6 = (struct in6_ifaddr *)
809 			    ifa_ifwithaddr((struct sockaddr *)&dest6);
810 			if (ia6 == NULL)
811 				goto bad;
812 			free_ia6 = 1;
813 		}
814 		else
815 			ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
816 
817 		/*
818 		 * record address information into m_tag.
819 		 */
820 		(void)ip6_setdstifaddr(m, ia6);
821 
822 		/*
823 		 * packets to a tentative, duplicated, or somehow invalid
824 		 * address must not be accepted.
825 		 */
826 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
827 			/* this address is ready */
828 			ours = 1;
829 			deliverifp = ia6->ia_ifp;	/* correct? */
830 			/* Count the packet in the ip address stats */
831 			ia6->ia_ifa.if_ipackets++;
832 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
833 			if (ia6 != NULL && free_ia6 != 0)
834 				ifa_free(&ia6->ia_ifa);
835 			goto hbhcheck;
836 		} else {
837 			char ip6bufs[INET6_ADDRSTRLEN];
838 			char ip6bufd[INET6_ADDRSTRLEN];
839 			/* address is not ready, so discard the packet. */
840 			nd6log((LOG_INFO,
841 			    "ip6_input: packet to an unready address %s->%s\n",
842 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
843 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
844 
845 			if (ia6 != NULL && free_ia6 != 0)
846 				ifa_free(&ia6->ia_ifa);
847 			goto bad;
848 		}
849 	}
850 
851 	/*
852 	 * FAITH (Firewall Aided Internet Translator)
853 	 */
854 	if (V_ip6_keepfaith) {
855 		if (rin6.ro_rt && rin6.ro_rt->rt_ifp &&
856 		    rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) {
857 			/* XXX do we need more sanity checks? */
858 			ours = 1;
859 			deliverifp = rin6.ro_rt->rt_ifp; /* faith */
860 			goto hbhcheck;
861 		}
862 	}
863 
864 	/*
865 	 * Now there is no reason to process the packet if it's not our own
866 	 * and we're not a router.
867 	 */
868 	if (!V_ip6_forwarding) {
869 		V_ip6stat.ip6s_cantforward++;
870 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
871 		goto bad;
872 	}
873 
874   hbhcheck:
875 	/*
876 	 * record address information into m_tag, if we don't have one yet.
877 	 * note that we are unable to record it, if the address is not listed
878 	 * as our interface address (e.g. multicast addresses, addresses
879 	 * within FAITH prefixes and such).
880 	 */
881 	if (deliverifp) {
882 		struct in6_ifaddr *ia6;
883 
884  		if ((ia6 = ip6_getdstifaddr(m)) != NULL) {
885 			ifa_free(&ia6->ia_ifa);
886 		} else {
887 			ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
888 			if (ia6) {
889 				if (!ip6_setdstifaddr(m, ia6)) {
890 					/*
891 					 * XXX maybe we should drop the packet here,
892 					 * as we could not provide enough information
893 					 * to the upper layers.
894 					 */
895 				}
896 				ifa_free(&ia6->ia_ifa);
897 			}
898 		}
899 	}
900 
901 	/*
902 	 * Process Hop-by-Hop options header if it's contained.
903 	 * m may be modified in ip6_hopopts_input().
904 	 * If a JumboPayload option is included, plen will also be modified.
905 	 */
906 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
907 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
908 		int error;
909 
910 		error = ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours);
911 		if (error != 0)
912 			goto out;
913 	} else
914 		nxt = ip6->ip6_nxt;
915 
916 	/*
917 	 * Check that the amount of data in the buffers
918 	 * is as at least much as the IPv6 header would have us expect.
919 	 * Trim mbufs if longer than we expect.
920 	 * Drop packet if shorter than we expect.
921 	 */
922 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
923 		V_ip6stat.ip6s_tooshort++;
924 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
925 		goto bad;
926 	}
927 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
928 		if (m->m_len == m->m_pkthdr.len) {
929 			m->m_len = sizeof(struct ip6_hdr) + plen;
930 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
931 		} else
932 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
933 	}
934 
935 	/*
936 	 * Forward if desirable.
937 	 */
938 	if (V_ip6_mrouter &&
939 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
940 		/*
941 		 * If we are acting as a multicast router, all
942 		 * incoming multicast packets are passed to the
943 		 * kernel-level multicast forwarding function.
944 		 * The packet is returned (relatively) intact; if
945 		 * ip6_mforward() returns a non-zero value, the packet
946 		 * must be discarded, else it may be accepted below.
947 		 *
948 		 * XXX TODO: Check hlim and multicast scope here to avoid
949 		 * unnecessarily calling into ip6_mforward().
950 		 */
951 		if (ip6_mforward &&
952 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
953 			IP6STAT_INC(ip6s_cantforward);
954 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
955 			goto bad;
956 		}
957 	} else if (!ours) {
958 		ip6_forward(m, srcrt);
959 		goto out;
960 	}
961 
962 	ip6 = mtod(m, struct ip6_hdr *);
963 
964 	/*
965 	 * Malicious party may be able to use IPv4 mapped addr to confuse
966 	 * tcp/udp stack and bypass security checks (act as if it was from
967 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
968 	 *
969 	 * For SIIT end node behavior, you may want to disable the check.
970 	 * However, you will  become vulnerable to attacks using IPv4 mapped
971 	 * source.
972 	 */
973 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
974 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
975 		V_ip6stat.ip6s_badscope++;
976 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
977 		goto bad;
978 	}
979 
980 	/*
981 	 * Tell launch routine the next header
982 	 */
983 	V_ip6stat.ip6s_delivered++;
984 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
985 	nest = 0;
986 
987 	while (nxt != IPPROTO_DONE) {
988 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
989 			V_ip6stat.ip6s_toomanyhdr++;
990 			goto bad;
991 		}
992 
993 		/*
994 		 * protection against faulty packet - there should be
995 		 * more sanity checks in header chain processing.
996 		 */
997 		if (m->m_pkthdr.len < off) {
998 			V_ip6stat.ip6s_tooshort++;
999 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
1000 			goto bad;
1001 		}
1002 
1003 #ifdef IPSEC
1004 		/*
1005 		 * enforce IPsec policy checking if we are seeing last header.
1006 		 * note that we do not visit this with protocols with pcb layer
1007 		 * code - like udp/tcp/raw ip.
1008 		 */
1009 		if (ip6_ipsec_input(m, nxt))
1010 			goto bad;
1011 #endif /* IPSEC */
1012 
1013 		/*
1014 		 * Use mbuf flags to propagate Router Alert option to
1015 		 * ICMPv6 layer, as hop-by-hop options have been stripped.
1016 		 */
1017 		if (nxt == IPPROTO_ICMPV6 && rtalert != ~0)
1018 			m->m_flags |= M_RTALERT_MLD;
1019 
1020 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
1021 	}
1022 	goto out;
1023 bad:
1024 	m_freem(m);
1025 out:
1026 	if (rin6.ro_rt)
1027 		RTFREE(rin6.ro_rt);
1028 }
1029 
1030 /*
1031  * set/grab in6_ifaddr correspond to IPv6 destination address.
1032  * XXX backward compatibility wrapper
1033  *
1034  * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag,
1035  * and then bump it when the tag is copied, and release it when the tag is
1036  * freed.  Unfortunately, m_tags don't support deep copies (yet), so instead
1037  * we just bump the ia refcount when we receive it.  This should be fixed.
1038  */
1039 static struct ip6aux *
1040 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
1041 {
1042 	struct ip6aux *ip6a;
1043 
1044 	ip6a = ip6_addaux(m);
1045 	if (ip6a)
1046 		ip6a->ip6a_dstia6 = ia6;
1047 	return ip6a;	/* NULL if failed to set */
1048 }
1049 
1050 struct in6_ifaddr *
1051 ip6_getdstifaddr(struct mbuf *m)
1052 {
1053 	struct ip6aux *ip6a;
1054 	struct in6_ifaddr *ia;
1055 
1056 	ip6a = ip6_findaux(m);
1057 	if (ip6a) {
1058 		ia = ip6a->ip6a_dstia6;
1059 		ifa_ref(&ia->ia_ifa);
1060 		return ia;
1061 	} else
1062 		return NULL;
1063 }
1064 
1065 /*
1066  * Hop-by-Hop options header processing. If a valid jumbo payload option is
1067  * included, the real payload length will be stored in plenp.
1068  *
1069  * rtalertp - XXX: should be stored more smart way
1070  */
1071 static int
1072 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
1073     struct mbuf **mp, int *offp)
1074 {
1075 	struct mbuf *m = *mp;
1076 	int off = *offp, hbhlen;
1077 	struct ip6_hbh *hbh;
1078 	u_int8_t *opt;
1079 
1080 	/* validation of the length of the header */
1081 #ifndef PULLDOWN_TEST
1082 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
1083 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1084 	hbhlen = (hbh->ip6h_len + 1) << 3;
1085 
1086 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
1087 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1088 #else
1089 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1090 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
1091 	if (hbh == NULL) {
1092 		V_ip6stat.ip6s_tooshort++;
1093 		return -1;
1094 	}
1095 	hbhlen = (hbh->ip6h_len + 1) << 3;
1096 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
1097 		hbhlen);
1098 	if (hbh == NULL) {
1099 		V_ip6stat.ip6s_tooshort++;
1100 		return -1;
1101 	}
1102 #endif
1103 	off += hbhlen;
1104 	hbhlen -= sizeof(struct ip6_hbh);
1105 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
1106 
1107 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1108 				hbhlen, rtalertp, plenp) < 0)
1109 		return (-1);
1110 
1111 	*offp = off;
1112 	*mp = m;
1113 	return (0);
1114 }
1115 
1116 /*
1117  * Search header for all Hop-by-hop options and process each option.
1118  * This function is separate from ip6_hopopts_input() in order to
1119  * handle a case where the sending node itself process its hop-by-hop
1120  * options header. In such a case, the function is called from ip6_output().
1121  *
1122  * The function assumes that hbh header is located right after the IPv6 header
1123  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1124  * opthead + hbhlen is located in contiguous memory region.
1125  */
1126 int
1127 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1128     u_int32_t *rtalertp, u_int32_t *plenp)
1129 {
1130 	struct ip6_hdr *ip6;
1131 	int optlen = 0;
1132 	u_int8_t *opt = opthead;
1133 	u_int16_t rtalert_val;
1134 	u_int32_t jumboplen;
1135 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1136 
1137 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1138 		switch (*opt) {
1139 		case IP6OPT_PAD1:
1140 			optlen = 1;
1141 			break;
1142 		case IP6OPT_PADN:
1143 			if (hbhlen < IP6OPT_MINLEN) {
1144 				V_ip6stat.ip6s_toosmall++;
1145 				goto bad;
1146 			}
1147 			optlen = *(opt + 1) + 2;
1148 			break;
1149 		case IP6OPT_ROUTER_ALERT:
1150 			/* XXX may need check for alignment */
1151 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1152 				V_ip6stat.ip6s_toosmall++;
1153 				goto bad;
1154 			}
1155 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1156 				/* XXX stat */
1157 				icmp6_error(m, ICMP6_PARAM_PROB,
1158 				    ICMP6_PARAMPROB_HEADER,
1159 				    erroff + opt + 1 - opthead);
1160 				return (-1);
1161 			}
1162 			optlen = IP6OPT_RTALERT_LEN;
1163 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1164 			*rtalertp = ntohs(rtalert_val);
1165 			break;
1166 		case IP6OPT_JUMBO:
1167 			/* XXX may need check for alignment */
1168 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1169 				V_ip6stat.ip6s_toosmall++;
1170 				goto bad;
1171 			}
1172 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1173 				/* XXX stat */
1174 				icmp6_error(m, ICMP6_PARAM_PROB,
1175 				    ICMP6_PARAMPROB_HEADER,
1176 				    erroff + opt + 1 - opthead);
1177 				return (-1);
1178 			}
1179 			optlen = IP6OPT_JUMBO_LEN;
1180 
1181 			/*
1182 			 * IPv6 packets that have non 0 payload length
1183 			 * must not contain a jumbo payload option.
1184 			 */
1185 			ip6 = mtod(m, struct ip6_hdr *);
1186 			if (ip6->ip6_plen) {
1187 				V_ip6stat.ip6s_badoptions++;
1188 				icmp6_error(m, ICMP6_PARAM_PROB,
1189 				    ICMP6_PARAMPROB_HEADER,
1190 				    erroff + opt - opthead);
1191 				return (-1);
1192 			}
1193 
1194 			/*
1195 			 * We may see jumbolen in unaligned location, so
1196 			 * we'd need to perform bcopy().
1197 			 */
1198 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1199 			jumboplen = (u_int32_t)htonl(jumboplen);
1200 
1201 #if 1
1202 			/*
1203 			 * if there are multiple jumbo payload options,
1204 			 * *plenp will be non-zero and the packet will be
1205 			 * rejected.
1206 			 * the behavior may need some debate in ipngwg -
1207 			 * multiple options does not make sense, however,
1208 			 * there's no explicit mention in specification.
1209 			 */
1210 			if (*plenp != 0) {
1211 				V_ip6stat.ip6s_badoptions++;
1212 				icmp6_error(m, ICMP6_PARAM_PROB,
1213 				    ICMP6_PARAMPROB_HEADER,
1214 				    erroff + opt + 2 - opthead);
1215 				return (-1);
1216 			}
1217 #endif
1218 
1219 			/*
1220 			 * jumbo payload length must be larger than 65535.
1221 			 */
1222 			if (jumboplen <= IPV6_MAXPACKET) {
1223 				V_ip6stat.ip6s_badoptions++;
1224 				icmp6_error(m, ICMP6_PARAM_PROB,
1225 				    ICMP6_PARAMPROB_HEADER,
1226 				    erroff + opt + 2 - opthead);
1227 				return (-1);
1228 			}
1229 			*plenp = jumboplen;
1230 
1231 			break;
1232 		default:		/* unknown option */
1233 			if (hbhlen < IP6OPT_MINLEN) {
1234 				V_ip6stat.ip6s_toosmall++;
1235 				goto bad;
1236 			}
1237 			optlen = ip6_unknown_opt(opt, m,
1238 			    erroff + opt - opthead);
1239 			if (optlen == -1)
1240 				return (-1);
1241 			optlen += 2;
1242 			break;
1243 		}
1244 	}
1245 
1246 	return (0);
1247 
1248   bad:
1249 	m_freem(m);
1250 	return (-1);
1251 }
1252 
1253 /*
1254  * Unknown option processing.
1255  * The third argument `off' is the offset from the IPv6 header to the option,
1256  * which is necessary if the IPv6 header the and option header and IPv6 header
1257  * is not contiguous in order to return an ICMPv6 error.
1258  */
1259 int
1260 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1261 {
1262 	struct ip6_hdr *ip6;
1263 
1264 	switch (IP6OPT_TYPE(*optp)) {
1265 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1266 		return ((int)*(optp + 1));
1267 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1268 		m_freem(m);
1269 		return (-1);
1270 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1271 		V_ip6stat.ip6s_badoptions++;
1272 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1273 		return (-1);
1274 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1275 		V_ip6stat.ip6s_badoptions++;
1276 		ip6 = mtod(m, struct ip6_hdr *);
1277 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1278 		    (m->m_flags & (M_BCAST|M_MCAST)))
1279 			m_freem(m);
1280 		else
1281 			icmp6_error(m, ICMP6_PARAM_PROB,
1282 				    ICMP6_PARAMPROB_OPTION, off);
1283 		return (-1);
1284 	}
1285 
1286 	m_freem(m);		/* XXX: NOTREACHED */
1287 	return (-1);
1288 }
1289 
1290 /*
1291  * Create the "control" list for this pcb.
1292  * These functions will not modify mbuf chain at all.
1293  *
1294  * With KAME mbuf chain restriction:
1295  * The routine will be called from upper layer handlers like tcp6_input().
1296  * Thus the routine assumes that the caller (tcp6_input) have already
1297  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1298  * very first mbuf on the mbuf chain.
1299  *
1300  * ip6_savecontrol_v4 will handle those options that are possible to be
1301  * set on a v4-mapped socket.
1302  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1303  * options and handle the v6-only ones itself.
1304  */
1305 struct mbuf **
1306 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1307     int *v4only)
1308 {
1309 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1310 
1311 #ifdef SO_TIMESTAMP
1312 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1313 		struct timeval tv;
1314 
1315 		microtime(&tv);
1316 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1317 		    SCM_TIMESTAMP, SOL_SOCKET);
1318 		if (*mp)
1319 			mp = &(*mp)->m_next;
1320 	}
1321 #endif
1322 
1323 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1324 	/* RFC 2292 sec. 5 */
1325 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1326 		struct in6_pktinfo pi6;
1327 
1328 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1329 #ifdef INET
1330 			struct ip *ip;
1331 
1332 			ip = mtod(m, struct ip *);
1333 			pi6.ipi6_addr.s6_addr32[0] = 0;
1334 			pi6.ipi6_addr.s6_addr32[1] = 0;
1335 			pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1336 			pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1337 #else
1338 			/* We won't hit this code */
1339 			bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1340 #endif
1341 		} else {
1342 			bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1343 			in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1344 		}
1345 		pi6.ipi6_ifindex =
1346 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1347 
1348 		*mp = sbcreatecontrol((caddr_t) &pi6,
1349 		    sizeof(struct in6_pktinfo),
1350 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1351 		if (*mp)
1352 			mp = &(*mp)->m_next;
1353 	}
1354 
1355 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1356 		int hlim;
1357 
1358 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1359 #ifdef INET
1360 			struct ip *ip;
1361 
1362 			ip = mtod(m, struct ip *);
1363 			hlim = ip->ip_ttl;
1364 #else
1365 			/* We won't hit this code */
1366 			hlim = 0;
1367 #endif
1368 		} else {
1369 			hlim = ip6->ip6_hlim & 0xff;
1370 		}
1371 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1372 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1373 		    IPPROTO_IPV6);
1374 		if (*mp)
1375 			mp = &(*mp)->m_next;
1376 	}
1377 
1378 	if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1379 		int tclass;
1380 
1381 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1382 #ifdef INET
1383 			struct ip *ip;
1384 
1385 			ip = mtod(m, struct ip *);
1386 			tclass = ip->ip_tos;
1387 #else
1388 			/* We won't hit this code */
1389 			tclass = 0;
1390 #endif
1391 		} else {
1392 			u_int32_t flowinfo;
1393 
1394 			flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1395 			flowinfo >>= 20;
1396 			tclass = flowinfo & 0xff;
1397 		}
1398 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int),
1399 		    IPV6_TCLASS, IPPROTO_IPV6);
1400 		if (*mp)
1401 			mp = &(*mp)->m_next;
1402 	}
1403 
1404 	if (v4only != NULL) {
1405 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1406 			*v4only = 1;
1407 		} else {
1408 			*v4only = 0;
1409 		}
1410 	}
1411 
1412 	return (mp);
1413 }
1414 
1415 void
1416 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1417 {
1418 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1419 	int v4only = 0;
1420 
1421 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1422 	if (v4only)
1423 		return;
1424 
1425 	/*
1426 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1427 	 * privilege for the option (see ip6_ctloutput), but it might be too
1428 	 * strict, since there might be some hop-by-hop options which can be
1429 	 * returned to normal user.
1430 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1431 	 */
1432 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1433 		/*
1434 		 * Check if a hop-by-hop options header is contatined in the
1435 		 * received packet, and if so, store the options as ancillary
1436 		 * data. Note that a hop-by-hop options header must be
1437 		 * just after the IPv6 header, which is assured through the
1438 		 * IPv6 input processing.
1439 		 */
1440 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1441 			struct ip6_hbh *hbh;
1442 			int hbhlen = 0;
1443 #ifdef PULLDOWN_TEST
1444 			struct mbuf *ext;
1445 #endif
1446 
1447 #ifndef PULLDOWN_TEST
1448 			hbh = (struct ip6_hbh *)(ip6 + 1);
1449 			hbhlen = (hbh->ip6h_len + 1) << 3;
1450 #else
1451 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1452 			    ip6->ip6_nxt);
1453 			if (ext == NULL) {
1454 				V_ip6stat.ip6s_tooshort++;
1455 				return;
1456 			}
1457 			hbh = mtod(ext, struct ip6_hbh *);
1458 			hbhlen = (hbh->ip6h_len + 1) << 3;
1459 			if (hbhlen != ext->m_len) {
1460 				m_freem(ext);
1461 				V_ip6stat.ip6s_tooshort++;
1462 				return;
1463 			}
1464 #endif
1465 
1466 			/*
1467 			 * XXX: We copy the whole header even if a
1468 			 * jumbo payload option is included, the option which
1469 			 * is to be removed before returning according to
1470 			 * RFC2292.
1471 			 * Note: this constraint is removed in RFC3542
1472 			 */
1473 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1474 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1475 			    IPPROTO_IPV6);
1476 			if (*mp)
1477 				mp = &(*mp)->m_next;
1478 #ifdef PULLDOWN_TEST
1479 			m_freem(ext);
1480 #endif
1481 		}
1482 	}
1483 
1484 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1485 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1486 
1487 		/*
1488 		 * Search for destination options headers or routing
1489 		 * header(s) through the header chain, and stores each
1490 		 * header as ancillary data.
1491 		 * Note that the order of the headers remains in
1492 		 * the chain of ancillary data.
1493 		 */
1494 		while (1) {	/* is explicit loop prevention necessary? */
1495 			struct ip6_ext *ip6e = NULL;
1496 			int elen;
1497 #ifdef PULLDOWN_TEST
1498 			struct mbuf *ext = NULL;
1499 #endif
1500 
1501 			/*
1502 			 * if it is not an extension header, don't try to
1503 			 * pull it from the chain.
1504 			 */
1505 			switch (nxt) {
1506 			case IPPROTO_DSTOPTS:
1507 			case IPPROTO_ROUTING:
1508 			case IPPROTO_HOPOPTS:
1509 			case IPPROTO_AH: /* is it possible? */
1510 				break;
1511 			default:
1512 				goto loopend;
1513 			}
1514 
1515 #ifndef PULLDOWN_TEST
1516 			if (off + sizeof(*ip6e) > m->m_len)
1517 				goto loopend;
1518 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1519 			if (nxt == IPPROTO_AH)
1520 				elen = (ip6e->ip6e_len + 2) << 2;
1521 			else
1522 				elen = (ip6e->ip6e_len + 1) << 3;
1523 			if (off + elen > m->m_len)
1524 				goto loopend;
1525 #else
1526 			ext = ip6_pullexthdr(m, off, nxt);
1527 			if (ext == NULL) {
1528 				V_ip6stat.ip6s_tooshort++;
1529 				return;
1530 			}
1531 			ip6e = mtod(ext, struct ip6_ext *);
1532 			if (nxt == IPPROTO_AH)
1533 				elen = (ip6e->ip6e_len + 2) << 2;
1534 			else
1535 				elen = (ip6e->ip6e_len + 1) << 3;
1536 			if (elen != ext->m_len) {
1537 				m_freem(ext);
1538 				V_ip6stat.ip6s_tooshort++;
1539 				return;
1540 			}
1541 #endif
1542 
1543 			switch (nxt) {
1544 			case IPPROTO_DSTOPTS:
1545 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1546 					break;
1547 
1548 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1549 				    IS2292(in6p,
1550 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1551 				    IPPROTO_IPV6);
1552 				if (*mp)
1553 					mp = &(*mp)->m_next;
1554 				break;
1555 			case IPPROTO_ROUTING:
1556 				if (!(in6p->inp_flags & IN6P_RTHDR))
1557 					break;
1558 
1559 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1560 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1561 				    IPPROTO_IPV6);
1562 				if (*mp)
1563 					mp = &(*mp)->m_next;
1564 				break;
1565 			case IPPROTO_HOPOPTS:
1566 			case IPPROTO_AH: /* is it possible? */
1567 				break;
1568 
1569 			default:
1570 				/*
1571 				 * other cases have been filtered in the above.
1572 				 * none will visit this case.  here we supply
1573 				 * the code just in case (nxt overwritten or
1574 				 * other cases).
1575 				 */
1576 #ifdef PULLDOWN_TEST
1577 				m_freem(ext);
1578 #endif
1579 				goto loopend;
1580 
1581 			}
1582 
1583 			/* proceed with the next header. */
1584 			off += elen;
1585 			nxt = ip6e->ip6e_nxt;
1586 			ip6e = NULL;
1587 #ifdef PULLDOWN_TEST
1588 			m_freem(ext);
1589 			ext = NULL;
1590 #endif
1591 		}
1592 	  loopend:
1593 		;
1594 	}
1595 }
1596 #undef IS2292
1597 
1598 void
1599 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1600 {
1601 	struct socket *so;
1602 	struct mbuf *m_mtu;
1603 	struct ip6_mtuinfo mtuctl;
1604 
1605 	so =  in6p->inp_socket;
1606 
1607 	if (mtu == NULL)
1608 		return;
1609 
1610 #ifdef DIAGNOSTIC
1611 	if (so == NULL)		/* I believe this is impossible */
1612 		panic("ip6_notify_pmtu: socket is NULL");
1613 #endif
1614 
1615 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1616 	mtuctl.ip6m_mtu = *mtu;
1617 	mtuctl.ip6m_addr = *dst;
1618 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1619 		return;
1620 
1621 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1622 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1623 		return;
1624 
1625 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1626 	    == 0) {
1627 		m_freem(m_mtu);
1628 		/* XXX: should count statistics */
1629 	} else
1630 		sorwakeup(so);
1631 
1632 	return;
1633 }
1634 
1635 #ifdef PULLDOWN_TEST
1636 /*
1637  * pull single extension header from mbuf chain.  returns single mbuf that
1638  * contains the result, or NULL on error.
1639  */
1640 static struct mbuf *
1641 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1642 {
1643 	struct ip6_ext ip6e;
1644 	size_t elen;
1645 	struct mbuf *n;
1646 
1647 #ifdef DIAGNOSTIC
1648 	switch (nxt) {
1649 	case IPPROTO_DSTOPTS:
1650 	case IPPROTO_ROUTING:
1651 	case IPPROTO_HOPOPTS:
1652 	case IPPROTO_AH: /* is it possible? */
1653 		break;
1654 	default:
1655 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1656 	}
1657 #endif
1658 
1659 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1660 	if (nxt == IPPROTO_AH)
1661 		elen = (ip6e.ip6e_len + 2) << 2;
1662 	else
1663 		elen = (ip6e.ip6e_len + 1) << 3;
1664 
1665 	MGET(n, M_DONTWAIT, MT_DATA);
1666 	if (n && elen >= MLEN) {
1667 		MCLGET(n, M_DONTWAIT);
1668 		if ((n->m_flags & M_EXT) == 0) {
1669 			m_free(n);
1670 			n = NULL;
1671 		}
1672 	}
1673 	if (!n)
1674 		return NULL;
1675 
1676 	n->m_len = 0;
1677 	if (elen >= M_TRAILINGSPACE(n)) {
1678 		m_free(n);
1679 		return NULL;
1680 	}
1681 
1682 	m_copydata(m, off, elen, mtod(n, caddr_t));
1683 	n->m_len = elen;
1684 	return n;
1685 }
1686 #endif
1687 
1688 /*
1689  * Get pointer to the previous header followed by the header
1690  * currently processed.
1691  * XXX: This function supposes that
1692  *	M includes all headers,
1693  *	the next header field and the header length field of each header
1694  *	are valid, and
1695  *	the sum of each header length equals to OFF.
1696  * Because of these assumptions, this function must be called very
1697  * carefully. Moreover, it will not be used in the near future when
1698  * we develop `neater' mechanism to process extension headers.
1699  */
1700 char *
1701 ip6_get_prevhdr(struct mbuf *m, int off)
1702 {
1703 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1704 
1705 	if (off == sizeof(struct ip6_hdr))
1706 		return (&ip6->ip6_nxt);
1707 	else {
1708 		int len, nxt;
1709 		struct ip6_ext *ip6e = NULL;
1710 
1711 		nxt = ip6->ip6_nxt;
1712 		len = sizeof(struct ip6_hdr);
1713 		while (len < off) {
1714 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1715 
1716 			switch (nxt) {
1717 			case IPPROTO_FRAGMENT:
1718 				len += sizeof(struct ip6_frag);
1719 				break;
1720 			case IPPROTO_AH:
1721 				len += (ip6e->ip6e_len + 2) << 2;
1722 				break;
1723 			default:
1724 				len += (ip6e->ip6e_len + 1) << 3;
1725 				break;
1726 			}
1727 			nxt = ip6e->ip6e_nxt;
1728 		}
1729 		if (ip6e)
1730 			return (&ip6e->ip6e_nxt);
1731 		else
1732 			return NULL;
1733 	}
1734 }
1735 
1736 /*
1737  * get next header offset.  m will be retained.
1738  */
1739 int
1740 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1741 {
1742 	struct ip6_hdr ip6;
1743 	struct ip6_ext ip6e;
1744 	struct ip6_frag fh;
1745 
1746 	/* just in case */
1747 	if (m == NULL)
1748 		panic("ip6_nexthdr: m == NULL");
1749 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1750 		return -1;
1751 
1752 	switch (proto) {
1753 	case IPPROTO_IPV6:
1754 		if (m->m_pkthdr.len < off + sizeof(ip6))
1755 			return -1;
1756 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1757 		if (nxtp)
1758 			*nxtp = ip6.ip6_nxt;
1759 		off += sizeof(ip6);
1760 		return off;
1761 
1762 	case IPPROTO_FRAGMENT:
1763 		/*
1764 		 * terminate parsing if it is not the first fragment,
1765 		 * it does not make sense to parse through it.
1766 		 */
1767 		if (m->m_pkthdr.len < off + sizeof(fh))
1768 			return -1;
1769 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1770 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1771 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1772 			return -1;
1773 		if (nxtp)
1774 			*nxtp = fh.ip6f_nxt;
1775 		off += sizeof(struct ip6_frag);
1776 		return off;
1777 
1778 	case IPPROTO_AH:
1779 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1780 			return -1;
1781 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1782 		if (nxtp)
1783 			*nxtp = ip6e.ip6e_nxt;
1784 		off += (ip6e.ip6e_len + 2) << 2;
1785 		return off;
1786 
1787 	case IPPROTO_HOPOPTS:
1788 	case IPPROTO_ROUTING:
1789 	case IPPROTO_DSTOPTS:
1790 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1791 			return -1;
1792 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1793 		if (nxtp)
1794 			*nxtp = ip6e.ip6e_nxt;
1795 		off += (ip6e.ip6e_len + 1) << 3;
1796 		return off;
1797 
1798 	case IPPROTO_NONE:
1799 	case IPPROTO_ESP:
1800 	case IPPROTO_IPCOMP:
1801 		/* give up */
1802 		return -1;
1803 
1804 	default:
1805 		return -1;
1806 	}
1807 
1808 	return -1;
1809 }
1810 
1811 /*
1812  * get offset for the last header in the chain.  m will be kept untainted.
1813  */
1814 int
1815 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1816 {
1817 	int newoff;
1818 	int nxt;
1819 
1820 	if (!nxtp) {
1821 		nxt = -1;
1822 		nxtp = &nxt;
1823 	}
1824 	while (1) {
1825 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1826 		if (newoff < 0)
1827 			return off;
1828 		else if (newoff < off)
1829 			return -1;	/* invalid */
1830 		else if (newoff == off)
1831 			return newoff;
1832 
1833 		off = newoff;
1834 		proto = *nxtp;
1835 	}
1836 }
1837 
1838 static struct ip6aux *
1839 ip6_addaux(struct mbuf *m)
1840 {
1841 	struct m_tag *mtag;
1842 
1843 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1844 	if (!mtag) {
1845 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1846 		    M_NOWAIT);
1847 		if (mtag) {
1848 			m_tag_prepend(m, mtag);
1849 			bzero(mtag + 1, sizeof(struct ip6aux));
1850 		}
1851 	}
1852 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1853 }
1854 
1855 static struct ip6aux *
1856 ip6_findaux(struct mbuf *m)
1857 {
1858 	struct m_tag *mtag;
1859 
1860 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1861 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1862 }
1863 
1864 static void
1865 ip6_delaux(struct mbuf *m)
1866 {
1867 	struct m_tag *mtag;
1868 
1869 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1870 	if (mtag)
1871 		m_tag_delete(m, mtag);
1872 }
1873 
1874 /*
1875  * System control for IP6
1876  */
1877 
1878 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1879 	0,		0,		0,		0,
1880 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1881 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1882 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1883 	0,		0,		0,		0,
1884 	ENOPROTOOPT
1885 };
1886