xref: /freebsd/sys/netinet6/ip6_input.c (revision c6ec7d31830ab1c80edae95ad5e4b9dba10c47ac)
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 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl);
130 VNET_DEFINE(u_long, in6_ifaddrhmask);
131 
132 static struct netisr_handler ip6_nh = {
133 	.nh_name = "ip6",
134 	.nh_handler = ip6_input,
135 	.nh_proto = NETISR_IPV6,
136 	.nh_policy = NETISR_POLICY_FLOW,
137 };
138 
139 VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch);
140 #define	V_in6_tmpaddrtimer_ch		VNET(in6_tmpaddrtimer_ch)
141 
142 VNET_DEFINE(struct pfil_head, inet6_pfil_hook);
143 
144 VNET_DEFINE(struct ip6stat, ip6stat);
145 
146 struct rwlock in6_ifaddr_lock;
147 RW_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
148 
149 static void ip6_init2(void *);
150 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *);
151 static struct ip6aux *ip6_addaux(struct mbuf *);
152 static struct ip6aux *ip6_findaux(struct mbuf *m);
153 static void ip6_delaux (struct mbuf *);
154 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
155 #ifdef PULLDOWN_TEST
156 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
157 #endif
158 
159 /*
160  * IP6 initialization: fill in IP6 protocol switch table.
161  * All protocols not implemented in kernel go to raw IP6 protocol handler.
162  */
163 void
164 ip6_init(void)
165 {
166 	struct ip6protosw *pr;
167 	int i;
168 
169 	TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
170 	    &V_ip6_auto_linklocal);
171 	TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
172 	TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
173 
174 	TAILQ_INIT(&V_in6_ifaddrhead);
175 	V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR,
176 	    &V_in6_ifaddrhmask);
177 
178 	/* Initialize packet filter hooks. */
179 	V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
180 	V_inet6_pfil_hook.ph_af = AF_INET6;
181 	if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0)
182 		printf("%s: WARNING: unable to register pfil hook, "
183 			"error %d\n", __func__, i);
184 
185 	scope6_init();
186 	addrsel_policy_init();
187 	nd6_init();
188 	frag6_init();
189 
190 #ifdef FLOWTABLE
191 	if (TUNABLE_INT_FETCH("net.inet6.ip6.output_flowtable_size",
192 		&V_ip6_output_flowtable_size)) {
193 		if (V_ip6_output_flowtable_size < 256)
194 			V_ip6_output_flowtable_size = 256;
195 		if (!powerof2(V_ip6_output_flowtable_size)) {
196 			printf("flowtable must be power of 2 size\n");
197 			V_ip6_output_flowtable_size = 2048;
198 		}
199 	} else {
200 		/*
201 		 * round up to the next power of 2
202 		 */
203 		V_ip6_output_flowtable_size = 1 << fls((1024 + maxusers * 64)-1);
204 	}
205 	V_ip6_ft = flowtable_alloc("ipv6", V_ip6_output_flowtable_size, FL_IPV6|FL_PCPU);
206 #endif
207 
208 	V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
209 
210 	/* Skip global initialization stuff for non-default instances. */
211 	if (!IS_DEFAULT_VNET(curvnet))
212 		return;
213 
214 #ifdef DIAGNOSTIC
215 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
216 		panic("sizeof(protosw) != sizeof(ip6protosw)");
217 #endif
218 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
219 	if (pr == NULL)
220 		panic("ip6_init");
221 
222 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
223 	for (i = 0; i < IPPROTO_MAX; i++)
224 		ip6_protox[i] = pr - inet6sw;
225 	/*
226 	 * Cycle through IP protocols and put them into the appropriate place
227 	 * in ip6_protox[].
228 	 */
229 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
230 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
231 		if (pr->pr_domain->dom_family == PF_INET6 &&
232 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
233 			/* Be careful to only index valid IP protocols. */
234 			if (pr->pr_protocol < IPPROTO_MAX)
235 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
236 		}
237 
238 	netisr_register(&ip6_nh);
239 }
240 
241 /*
242  * The protocol to be inserted into ip6_protox[] must be already registered
243  * in inet6sw[], either statically or through pf_proto_register().
244  */
245 int
246 ip6proto_register(short ip6proto)
247 {
248 	struct ip6protosw *pr;
249 
250 	/* Sanity checks. */
251 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
252 		return (EPROTONOSUPPORT);
253 
254 	/*
255 	 * The protocol slot must not be occupied by another protocol
256 	 * already.  An index pointing to IPPROTO_RAW is unused.
257 	 */
258 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
259 	if (pr == NULL)
260 		return (EPFNOSUPPORT);
261 	if (ip6_protox[ip6proto] != pr - inet6sw)	/* IPPROTO_RAW */
262 		return (EEXIST);
263 
264 	/*
265 	 * Find the protocol position in inet6sw[] and set the index.
266 	 */
267 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
268 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) {
269 		if (pr->pr_domain->dom_family == PF_INET6 &&
270 		    pr->pr_protocol && pr->pr_protocol == ip6proto) {
271 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
272 			return (0);
273 		}
274 	}
275 	return (EPROTONOSUPPORT);
276 }
277 
278 int
279 ip6proto_unregister(short ip6proto)
280 {
281 	struct ip6protosw *pr;
282 
283 	/* Sanity checks. */
284 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
285 		return (EPROTONOSUPPORT);
286 
287 	/* Check if the protocol was indeed registered. */
288 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
289 	if (pr == NULL)
290 		return (EPFNOSUPPORT);
291 	if (ip6_protox[ip6proto] == pr - inet6sw)	/* IPPROTO_RAW */
292 		return (ENOENT);
293 
294 	/* Reset the protocol slot to IPPROTO_RAW. */
295 	ip6_protox[ip6proto] = pr - inet6sw;
296 	return (0);
297 }
298 
299 #ifdef VIMAGE
300 void
301 ip6_destroy()
302 {
303 
304 	hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask);
305 	nd6_destroy();
306 	callout_drain(&V_in6_tmpaddrtimer_ch);
307 }
308 #endif
309 
310 static int
311 ip6_init2_vnet(const void *unused __unused)
312 {
313 
314 	/* nd6_timer_init */
315 	callout_init(&V_nd6_timer_ch, 0);
316 	callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
317 
318 	/* timer for regeneranation of temporary addresses randomize ID */
319 	callout_init(&V_in6_tmpaddrtimer_ch, 0);
320 	callout_reset(&V_in6_tmpaddrtimer_ch,
321 		      (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
322 		       V_ip6_temp_regen_advance) * hz,
323 		      in6_tmpaddrtimer, curvnet);
324 
325 	return (0);
326 }
327 
328 static void
329 ip6_init2(void *dummy)
330 {
331 
332 	ip6_init2_vnet(NULL);
333 }
334 
335 /* cheat */
336 /* This must be after route_init(), which is now SI_ORDER_THIRD */
337 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
338 
339 static int
340 ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off,
341     int *nxt, int *ours)
342 {
343 	struct ip6_hdr *ip6;
344 	struct ip6_hbh *hbh;
345 
346 	if (ip6_hopopts_input(plen, rtalert, &m, off)) {
347 #if 0	/*touches NULL pointer*/
348 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
349 #endif
350 		goto out;	/* m have already been freed */
351 	}
352 
353 	/* adjust pointer */
354 	ip6 = mtod(m, struct ip6_hdr *);
355 
356 	/*
357 	 * if the payload length field is 0 and the next header field
358 	 * indicates Hop-by-Hop Options header, then a Jumbo Payload
359 	 * option MUST be included.
360 	 */
361 	if (ip6->ip6_plen == 0 && *plen == 0) {
362 		/*
363 		 * Note that if a valid jumbo payload option is
364 		 * contained, ip6_hopopts_input() must set a valid
365 		 * (non-zero) payload length to the variable plen.
366 		 */
367 		V_ip6stat.ip6s_badoptions++;
368 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
369 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
370 		icmp6_error(m, ICMP6_PARAM_PROB,
371 			    ICMP6_PARAMPROB_HEADER,
372 			    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
373 		goto out;
374 	}
375 #ifndef PULLDOWN_TEST
376 	/* ip6_hopopts_input() ensures that mbuf is contiguous */
377 	hbh = (struct ip6_hbh *)(ip6 + 1);
378 #else
379 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
380 		sizeof(struct ip6_hbh));
381 	if (hbh == NULL) {
382 		V_ip6stat.ip6s_tooshort++;
383 		goto out;
384 	}
385 #endif
386 	*nxt = hbh->ip6h_nxt;
387 
388 	/*
389 	 * If we are acting as a router and the packet contains a
390 	 * router alert option, see if we know the option value.
391 	 * Currently, we only support the option value for MLD, in which
392 	 * case we should pass the packet to the multicast routing
393 	 * daemon.
394 	 */
395 	if (*rtalert != ~0) {
396 		switch (*rtalert) {
397 		case IP6OPT_RTALERT_MLD:
398 			if (V_ip6_forwarding)
399 				*ours = 1;
400 			break;
401 		default:
402 			/*
403 			 * RFC2711 requires unrecognized values must be
404 			 * silently ignored.
405 			 */
406 			break;
407 		}
408 	}
409 
410 	return (0);
411 
412 out:
413 	return (1);
414 }
415 
416 void
417 ip6_input(struct mbuf *m)
418 {
419 	struct ip6_hdr *ip6;
420 	int off = sizeof(struct ip6_hdr), nest;
421 	u_int32_t plen;
422 	u_int32_t rtalert = ~0;
423 	int nxt, ours = 0;
424 	struct ifnet *deliverifp = NULL, *ifp = NULL;
425 	struct in6_addr odst;
426 	struct route_in6 rin6;
427 	int srcrt = 0;
428 	struct llentry *lle = NULL;
429 	struct sockaddr_in6 dst6, *dst;
430 
431 	bzero(&rin6, sizeof(struct route_in6));
432 #ifdef IPSEC
433 	/*
434 	 * should the inner packet be considered authentic?
435 	 * see comment in ah4_input().
436 	 * NB: m cannot be NULL when passed to the input routine
437 	 */
438 
439 	m->m_flags &= ~M_AUTHIPHDR;
440 	m->m_flags &= ~M_AUTHIPDGM;
441 
442 #endif /* IPSEC */
443 
444 	/*
445 	 * make sure we don't have onion peering information into m_tag.
446 	 */
447 	ip6_delaux(m);
448 
449 	if (m->m_flags & M_FASTFWD_OURS) {
450 		/*
451 		 * Firewall changed destination to local.
452 		 */
453 		m->m_flags &= ~M_FASTFWD_OURS;
454 		ours = 1;
455 		deliverifp = m->m_pkthdr.rcvif;
456 		ip6 = mtod(m, struct ip6_hdr *);
457 		goto hbhcheck;
458 	}
459 
460 	/*
461 	 * mbuf statistics
462 	 */
463 	if (m->m_flags & M_EXT) {
464 		if (m->m_next)
465 			V_ip6stat.ip6s_mext2m++;
466 		else
467 			V_ip6stat.ip6s_mext1++;
468 	} else {
469 #define M2MMAX	(sizeof(V_ip6stat.ip6s_m2m)/sizeof(V_ip6stat.ip6s_m2m[0]))
470 		if (m->m_next) {
471 			if (m->m_flags & M_LOOP) {
472 				V_ip6stat.ip6s_m2m[V_loif->if_index]++;
473 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
474 				V_ip6stat.ip6s_m2m[m->m_pkthdr.rcvif->if_index]++;
475 			else
476 				V_ip6stat.ip6s_m2m[0]++;
477 		} else
478 			V_ip6stat.ip6s_m1++;
479 #undef M2MMAX
480 	}
481 
482 	/* drop the packet if IPv6 operation is disabled on the IF */
483 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
484 		m_freem(m);
485 		return;
486 	}
487 
488 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
489 	V_ip6stat.ip6s_total++;
490 
491 #ifndef PULLDOWN_TEST
492 	/*
493 	 * L2 bridge code and some other code can return mbuf chain
494 	 * that does not conform to KAME requirement.  too bad.
495 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
496 	 */
497 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
498 		struct mbuf *n;
499 
500 		MGETHDR(n, M_NOWAIT, MT_HEADER);
501 		if (n)
502 			M_MOVE_PKTHDR(n, m);
503 		if (n && n->m_pkthdr.len > MHLEN) {
504 			MCLGET(n, M_NOWAIT);
505 			if ((n->m_flags & M_EXT) == 0) {
506 				m_freem(n);
507 				n = NULL;
508 			}
509 		}
510 		if (n == NULL) {
511 			m_freem(m);
512 			return;	/* ENOBUFS */
513 		}
514 
515 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
516 		n->m_len = n->m_pkthdr.len;
517 		m_freem(m);
518 		m = n;
519 	}
520 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
521 #endif
522 
523 	if (m->m_len < sizeof(struct ip6_hdr)) {
524 		struct ifnet *inifp;
525 		inifp = m->m_pkthdr.rcvif;
526 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
527 			V_ip6stat.ip6s_toosmall++;
528 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
529 			return;
530 		}
531 	}
532 
533 	ip6 = mtod(m, struct ip6_hdr *);
534 
535 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
536 		V_ip6stat.ip6s_badvers++;
537 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
538 		goto bad;
539 	}
540 
541 	V_ip6stat.ip6s_nxthist[ip6->ip6_nxt]++;
542 
543 	/*
544 	 * Check against address spoofing/corruption.
545 	 */
546 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
547 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
548 		/*
549 		 * XXX: "badscope" is not very suitable for a multicast source.
550 		 */
551 		V_ip6stat.ip6s_badscope++;
552 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
553 		goto bad;
554 	}
555 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
556 	    !(m->m_flags & M_LOOP)) {
557 		/*
558 		 * In this case, the packet should come from the loopback
559 		 * interface.  However, we cannot just check the if_flags,
560 		 * because ip6_mloopback() passes the "actual" interface
561 		 * as the outgoing/incoming interface.
562 		 */
563 		V_ip6stat.ip6s_badscope++;
564 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
565 		goto bad;
566 	}
567 
568 #ifdef ALTQ
569 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
570 		/* packet is dropped by traffic conditioner */
571 		return;
572 	}
573 #endif
574 	/*
575 	 * The following check is not documented in specs.  A malicious
576 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
577 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
578 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
579 	 *
580 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
581 	 * support IPv4-less kernel compilation, we cannot support SIIT
582 	 * environment at all.  So, it makes more sense for us to reject any
583 	 * malicious packets for non-SIIT environment, than try to do a
584 	 * partial support for SIIT environment.
585 	 */
586 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
587 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
588 		V_ip6stat.ip6s_badscope++;
589 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
590 		goto bad;
591 	}
592 #if 0
593 	/*
594 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
595 	 *
596 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
597 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
598 	 * is revised to forbid relaying case.
599 	 */
600 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
601 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
602 		V_ip6stat.ip6s_badscope++;
603 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
604 		goto bad;
605 	}
606 #endif
607 #ifdef IPSEC
608 	/*
609 	 * Bypass packet filtering for packets previously handled by IPsec.
610 	 */
611 	if (ip6_ipsec_filtertunnel(m))
612 		goto passin;
613 #endif /* IPSEC */
614 
615 	/*
616 	 * Run through list of hooks for input packets.
617 	 *
618 	 * NB: Beware of the destination address changing
619 	 *     (e.g. by NAT rewriting).  When this happens,
620 	 *     tell ip6_forward to do the right thing.
621 	 */
622 	odst = ip6->ip6_dst;
623 
624 	/* Jump over all PFIL processing if hooks are not active. */
625 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
626 		goto passin;
627 
628 	if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
629 	    m->m_pkthdr.rcvif, PFIL_IN, NULL))
630 		return;
631 	if (m == NULL)			/* consumed by filter */
632 		return;
633 	ip6 = mtod(m, struct ip6_hdr *);
634 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
635 
636 	if (m->m_flags & M_FASTFWD_OURS) {
637 		m->m_flags &= ~M_FASTFWD_OURS;
638 		ours = 1;
639 		deliverifp = m->m_pkthdr.rcvif;
640 		goto hbhcheck;
641 	}
642 	if ((m->m_flags & M_IP6_NEXTHOP) &&
643 	    m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
644 		/*
645 		 * Directly ship the packet on.  This allows forwarding
646 		 * packets originally destined to us to some other directly
647 		 * connected host.
648 		 */
649 		ip6_forward(m, 1);
650 		goto out;
651 	}
652 
653 passin:
654 	/*
655 	 * Disambiguate address scope zones (if there is ambiguity).
656 	 * We first make sure that the original source or destination address
657 	 * is not in our internal form for scoped addresses.  Such addresses
658 	 * are not necessarily invalid spec-wise, but we cannot accept them due
659 	 * to the usage conflict.
660 	 * in6_setscope() then also checks and rejects the cases where src or
661 	 * dst are the loopback address and the receiving interface
662 	 * is not loopback.
663 	 */
664 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
665 		V_ip6stat.ip6s_badscope++; /* XXX */
666 		goto bad;
667 	}
668 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
669 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
670 		V_ip6stat.ip6s_badscope++;
671 		goto bad;
672 	}
673 
674 	/*
675 	 * Multicast check. Assume packet is for us to avoid
676 	 * prematurely taking locks.
677 	 */
678 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
679 		ours = 1;
680 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
681 		deliverifp = m->m_pkthdr.rcvif;
682 		goto hbhcheck;
683 	}
684 
685 	/*
686 	 *  Unicast check
687 	 */
688 
689 	bzero(&dst6, sizeof(dst6));
690 	dst6.sin6_family = AF_INET6;
691 	dst6.sin6_len = sizeof(struct sockaddr_in6);
692 	dst6.sin6_addr = ip6->ip6_dst;
693 	ifp = m->m_pkthdr.rcvif;
694 	IF_AFDATA_RLOCK(ifp);
695 	lle = lla_lookup(LLTABLE6(ifp), 0,
696 	     (struct sockaddr *)&dst6);
697 	IF_AFDATA_RUNLOCK(ifp);
698 	if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) {
699 		struct ifaddr *ifa;
700 		struct in6_ifaddr *ia6;
701 		int bad;
702 
703 		bad = 1;
704 #define	sa_equal(a1, a2)						\
705 	(bcmp((a1), (a2), ((a1))->sin6_len) == 0)
706 		IF_ADDR_RLOCK(ifp);
707 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
708 			if (ifa->ifa_addr->sa_family != dst6.sin6_family)
709 				continue;
710 			if (sa_equal(&dst6, ifa->ifa_addr))
711 				break;
712 		}
713 		KASSERT(ifa != NULL, ("%s: ifa not found for lle %p",
714 		    __func__, lle));
715 #undef sa_equal
716 
717 		ia6 = (struct in6_ifaddr *)ifa;
718 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
719 			/* Count the packet in the ip address stats */
720 			ia6->ia_ifa.if_ipackets++;
721 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
722 
723 			/*
724 			 * record address information into m_tag.
725 			 */
726 			(void)ip6_setdstifaddr(m, ia6);
727 
728 			bad = 0;
729 		} else {
730 			char ip6bufs[INET6_ADDRSTRLEN];
731 			char ip6bufd[INET6_ADDRSTRLEN];
732 			/* address is not ready, so discard the packet. */
733 			nd6log((LOG_INFO,
734 			    "ip6_input: packet to an unready address %s->%s\n",
735 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
736 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
737 		}
738 		IF_ADDR_RUNLOCK(ifp);
739 		LLE_RUNLOCK(lle);
740 		if (bad)
741 			goto bad;
742 		else {
743 			ours = 1;
744 			deliverifp = ifp;
745 			goto hbhcheck;
746 		}
747 	}
748 	if (lle != NULL)
749 		LLE_RUNLOCK(lle);
750 
751 	dst = &rin6.ro_dst;
752 	dst->sin6_len = sizeof(struct sockaddr_in6);
753 	dst->sin6_family = AF_INET6;
754 	dst->sin6_addr = ip6->ip6_dst;
755 	rin6.ro_rt = in6_rtalloc1((struct sockaddr *)dst, 0, 0, M_GETFIB(m));
756 	if (rin6.ro_rt)
757 		RT_UNLOCK(rin6.ro_rt);
758 
759 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
760 
761 	/*
762 	 * Accept the packet if the forwarding interface to the destination
763 	 * according to the routing table is the loopback interface,
764 	 * unless the associated route has a gateway.
765 	 * Note that this approach causes to accept a packet if there is a
766 	 * route to the loopback interface for the destination of the packet.
767 	 * But we think it's even useful in some situations, e.g. when using
768 	 * a special daemon which wants to intercept the packet.
769 	 *
770 	 * XXX: some OSes automatically make a cloned route for the destination
771 	 * of an outgoing packet.  If the outgoing interface of the packet
772 	 * is a loopback one, the kernel would consider the packet to be
773 	 * accepted, even if we have no such address assinged on the interface.
774 	 * We check the cloned flag of the route entry to reject such cases,
775 	 * assuming that route entries for our own addresses are not made by
776 	 * cloning (it should be true because in6_addloop explicitly installs
777 	 * the host route).  However, we might have to do an explicit check
778 	 * while it would be less efficient.  Or, should we rather install a
779 	 * reject route for such a case?
780 	 */
781 	if (rin6.ro_rt &&
782 	    (rin6.ro_rt->rt_flags &
783 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
784 #ifdef RTF_WASCLONED
785 	    !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
786 #endif
787 #ifdef RTF_CLONED
788 	    !(rin6.ro_rt->rt_flags & RTF_CLONED) &&
789 #endif
790 #if 0
791 	    /*
792 	     * The check below is redundant since the comparison of
793 	     * the destination and the key of the rtentry has
794 	     * already done through looking up the routing table.
795 	     */
796 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
797 	    &rt6_key(rin6.ro_rt)->sin6_addr)
798 #endif
799 	    rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
800 		int free_ia6 = 0;
801 		struct in6_ifaddr *ia6;
802 
803 		/*
804 		 * found the loopback route to the interface address
805 		 */
806 		if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) {
807 			struct sockaddr_in6 dest6;
808 
809 			bzero(&dest6, sizeof(dest6));
810 			dest6.sin6_family = AF_INET6;
811 			dest6.sin6_len = sizeof(dest6);
812 			dest6.sin6_addr = ip6->ip6_dst;
813 			ia6 = (struct in6_ifaddr *)
814 			    ifa_ifwithaddr((struct sockaddr *)&dest6);
815 			if (ia6 == NULL)
816 				goto bad;
817 			free_ia6 = 1;
818 		}
819 		else
820 			ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
821 
822 		/*
823 		 * record address information into m_tag.
824 		 */
825 		(void)ip6_setdstifaddr(m, ia6);
826 
827 		/*
828 		 * packets to a tentative, duplicated, or somehow invalid
829 		 * address must not be accepted.
830 		 */
831 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
832 			/* this address is ready */
833 			ours = 1;
834 			deliverifp = ia6->ia_ifp;	/* correct? */
835 			/* Count the packet in the ip address stats */
836 			ia6->ia_ifa.if_ipackets++;
837 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
838 			if (ia6 != NULL && free_ia6 != 0)
839 				ifa_free(&ia6->ia_ifa);
840 			goto hbhcheck;
841 		} else {
842 			char ip6bufs[INET6_ADDRSTRLEN];
843 			char ip6bufd[INET6_ADDRSTRLEN];
844 			/* address is not ready, so discard the packet. */
845 			nd6log((LOG_INFO,
846 			    "ip6_input: packet to an unready address %s->%s\n",
847 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
848 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
849 
850 			if (ia6 != NULL && free_ia6 != 0)
851 				ifa_free(&ia6->ia_ifa);
852 			goto bad;
853 		}
854 	}
855 
856 	/*
857 	 * FAITH (Firewall Aided Internet Translator)
858 	 */
859 	if (V_ip6_keepfaith) {
860 		if (rin6.ro_rt && rin6.ro_rt->rt_ifp &&
861 		    rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) {
862 			/* XXX do we need more sanity checks? */
863 			ours = 1;
864 			deliverifp = rin6.ro_rt->rt_ifp; /* faith */
865 			goto hbhcheck;
866 		}
867 	}
868 
869 	/*
870 	 * Now there is no reason to process the packet if it's not our own
871 	 * and we're not a router.
872 	 */
873 	if (!V_ip6_forwarding) {
874 		V_ip6stat.ip6s_cantforward++;
875 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
876 		goto bad;
877 	}
878 
879   hbhcheck:
880 	/*
881 	 * record address information into m_tag, if we don't have one yet.
882 	 * note that we are unable to record it, if the address is not listed
883 	 * as our interface address (e.g. multicast addresses, addresses
884 	 * within FAITH prefixes and such).
885 	 */
886 	if (deliverifp) {
887 		struct in6_ifaddr *ia6;
888 
889  		if ((ia6 = ip6_getdstifaddr(m)) != NULL) {
890 			ifa_free(&ia6->ia_ifa);
891 		} else {
892 			ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
893 			if (ia6) {
894 				if (!ip6_setdstifaddr(m, ia6)) {
895 					/*
896 					 * XXX maybe we should drop the packet here,
897 					 * as we could not provide enough information
898 					 * to the upper layers.
899 					 */
900 				}
901 				ifa_free(&ia6->ia_ifa);
902 			}
903 		}
904 	}
905 
906 	/*
907 	 * Process Hop-by-Hop options header if it's contained.
908 	 * m may be modified in ip6_hopopts_input().
909 	 * If a JumboPayload option is included, plen will also be modified.
910 	 */
911 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
912 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
913 		int error;
914 
915 		error = ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours);
916 		if (error != 0)
917 			goto out;
918 	} else
919 		nxt = ip6->ip6_nxt;
920 
921 	/*
922 	 * Check that the amount of data in the buffers
923 	 * is as at least much as the IPv6 header would have us expect.
924 	 * Trim mbufs if longer than we expect.
925 	 * Drop packet if shorter than we expect.
926 	 */
927 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
928 		V_ip6stat.ip6s_tooshort++;
929 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
930 		goto bad;
931 	}
932 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
933 		if (m->m_len == m->m_pkthdr.len) {
934 			m->m_len = sizeof(struct ip6_hdr) + plen;
935 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
936 		} else
937 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
938 	}
939 
940 	/*
941 	 * Forward if desirable.
942 	 */
943 	if (V_ip6_mrouter &&
944 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
945 		/*
946 		 * If we are acting as a multicast router, all
947 		 * incoming multicast packets are passed to the
948 		 * kernel-level multicast forwarding function.
949 		 * The packet is returned (relatively) intact; if
950 		 * ip6_mforward() returns a non-zero value, the packet
951 		 * must be discarded, else it may be accepted below.
952 		 *
953 		 * XXX TODO: Check hlim and multicast scope here to avoid
954 		 * unnecessarily calling into ip6_mforward().
955 		 */
956 		if (ip6_mforward &&
957 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
958 			IP6STAT_INC(ip6s_cantforward);
959 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
960 			goto bad;
961 		}
962 	} else if (!ours) {
963 		ip6_forward(m, srcrt);
964 		goto out;
965 	}
966 
967 	ip6 = mtod(m, struct ip6_hdr *);
968 
969 	/*
970 	 * Malicious party may be able to use IPv4 mapped addr to confuse
971 	 * tcp/udp stack and bypass security checks (act as if it was from
972 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
973 	 *
974 	 * For SIIT end node behavior, you may want to disable the check.
975 	 * However, you will  become vulnerable to attacks using IPv4 mapped
976 	 * source.
977 	 */
978 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
979 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
980 		V_ip6stat.ip6s_badscope++;
981 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
982 		goto bad;
983 	}
984 
985 	/*
986 	 * Tell launch routine the next header
987 	 */
988 	V_ip6stat.ip6s_delivered++;
989 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
990 	nest = 0;
991 
992 	while (nxt != IPPROTO_DONE) {
993 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
994 			V_ip6stat.ip6s_toomanyhdr++;
995 			goto bad;
996 		}
997 
998 		/*
999 		 * protection against faulty packet - there should be
1000 		 * more sanity checks in header chain processing.
1001 		 */
1002 		if (m->m_pkthdr.len < off) {
1003 			V_ip6stat.ip6s_tooshort++;
1004 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
1005 			goto bad;
1006 		}
1007 
1008 #ifdef IPSEC
1009 		/*
1010 		 * enforce IPsec policy checking if we are seeing last header.
1011 		 * note that we do not visit this with protocols with pcb layer
1012 		 * code - like udp/tcp/raw ip.
1013 		 */
1014 		if (ip6_ipsec_input(m, nxt))
1015 			goto bad;
1016 #endif /* IPSEC */
1017 
1018 		/*
1019 		 * Use mbuf flags to propagate Router Alert option to
1020 		 * ICMPv6 layer, as hop-by-hop options have been stripped.
1021 		 */
1022 		if (nxt == IPPROTO_ICMPV6 && rtalert != ~0)
1023 			m->m_flags |= M_RTALERT_MLD;
1024 
1025 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
1026 	}
1027 	goto out;
1028 bad:
1029 	m_freem(m);
1030 out:
1031 	if (rin6.ro_rt)
1032 		RTFREE(rin6.ro_rt);
1033 }
1034 
1035 /*
1036  * set/grab in6_ifaddr correspond to IPv6 destination address.
1037  * XXX backward compatibility wrapper
1038  *
1039  * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag,
1040  * and then bump it when the tag is copied, and release it when the tag is
1041  * freed.  Unfortunately, m_tags don't support deep copies (yet), so instead
1042  * we just bump the ia refcount when we receive it.  This should be fixed.
1043  */
1044 static struct ip6aux *
1045 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
1046 {
1047 	struct ip6aux *ip6a;
1048 
1049 	ip6a = ip6_addaux(m);
1050 	if (ip6a)
1051 		ip6a->ip6a_dstia6 = ia6;
1052 	return ip6a;	/* NULL if failed to set */
1053 }
1054 
1055 struct in6_ifaddr *
1056 ip6_getdstifaddr(struct mbuf *m)
1057 {
1058 	struct ip6aux *ip6a;
1059 	struct in6_ifaddr *ia;
1060 
1061 	ip6a = ip6_findaux(m);
1062 	if (ip6a) {
1063 		ia = ip6a->ip6a_dstia6;
1064 		ifa_ref(&ia->ia_ifa);
1065 		return ia;
1066 	} else
1067 		return NULL;
1068 }
1069 
1070 /*
1071  * Hop-by-Hop options header processing. If a valid jumbo payload option is
1072  * included, the real payload length will be stored in plenp.
1073  *
1074  * rtalertp - XXX: should be stored more smart way
1075  */
1076 static int
1077 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
1078     struct mbuf **mp, int *offp)
1079 {
1080 	struct mbuf *m = *mp;
1081 	int off = *offp, hbhlen;
1082 	struct ip6_hbh *hbh;
1083 	u_int8_t *opt;
1084 
1085 	/* validation of the length of the header */
1086 #ifndef PULLDOWN_TEST
1087 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
1088 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1089 	hbhlen = (hbh->ip6h_len + 1) << 3;
1090 
1091 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
1092 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1093 #else
1094 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1095 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
1096 	if (hbh == NULL) {
1097 		V_ip6stat.ip6s_tooshort++;
1098 		return -1;
1099 	}
1100 	hbhlen = (hbh->ip6h_len + 1) << 3;
1101 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
1102 		hbhlen);
1103 	if (hbh == NULL) {
1104 		V_ip6stat.ip6s_tooshort++;
1105 		return -1;
1106 	}
1107 #endif
1108 	off += hbhlen;
1109 	hbhlen -= sizeof(struct ip6_hbh);
1110 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
1111 
1112 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1113 				hbhlen, rtalertp, plenp) < 0)
1114 		return (-1);
1115 
1116 	*offp = off;
1117 	*mp = m;
1118 	return (0);
1119 }
1120 
1121 /*
1122  * Search header for all Hop-by-hop options and process each option.
1123  * This function is separate from ip6_hopopts_input() in order to
1124  * handle a case where the sending node itself process its hop-by-hop
1125  * options header. In such a case, the function is called from ip6_output().
1126  *
1127  * The function assumes that hbh header is located right after the IPv6 header
1128  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1129  * opthead + hbhlen is located in contiguous memory region.
1130  */
1131 int
1132 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1133     u_int32_t *rtalertp, u_int32_t *plenp)
1134 {
1135 	struct ip6_hdr *ip6;
1136 	int optlen = 0;
1137 	u_int8_t *opt = opthead;
1138 	u_int16_t rtalert_val;
1139 	u_int32_t jumboplen;
1140 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1141 
1142 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1143 		switch (*opt) {
1144 		case IP6OPT_PAD1:
1145 			optlen = 1;
1146 			break;
1147 		case IP6OPT_PADN:
1148 			if (hbhlen < IP6OPT_MINLEN) {
1149 				V_ip6stat.ip6s_toosmall++;
1150 				goto bad;
1151 			}
1152 			optlen = *(opt + 1) + 2;
1153 			break;
1154 		case IP6OPT_ROUTER_ALERT:
1155 			/* XXX may need check for alignment */
1156 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1157 				V_ip6stat.ip6s_toosmall++;
1158 				goto bad;
1159 			}
1160 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1161 				/* XXX stat */
1162 				icmp6_error(m, ICMP6_PARAM_PROB,
1163 				    ICMP6_PARAMPROB_HEADER,
1164 				    erroff + opt + 1 - opthead);
1165 				return (-1);
1166 			}
1167 			optlen = IP6OPT_RTALERT_LEN;
1168 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1169 			*rtalertp = ntohs(rtalert_val);
1170 			break;
1171 		case IP6OPT_JUMBO:
1172 			/* XXX may need check for alignment */
1173 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1174 				V_ip6stat.ip6s_toosmall++;
1175 				goto bad;
1176 			}
1177 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1178 				/* XXX stat */
1179 				icmp6_error(m, ICMP6_PARAM_PROB,
1180 				    ICMP6_PARAMPROB_HEADER,
1181 				    erroff + opt + 1 - opthead);
1182 				return (-1);
1183 			}
1184 			optlen = IP6OPT_JUMBO_LEN;
1185 
1186 			/*
1187 			 * IPv6 packets that have non 0 payload length
1188 			 * must not contain a jumbo payload option.
1189 			 */
1190 			ip6 = mtod(m, struct ip6_hdr *);
1191 			if (ip6->ip6_plen) {
1192 				V_ip6stat.ip6s_badoptions++;
1193 				icmp6_error(m, ICMP6_PARAM_PROB,
1194 				    ICMP6_PARAMPROB_HEADER,
1195 				    erroff + opt - opthead);
1196 				return (-1);
1197 			}
1198 
1199 			/*
1200 			 * We may see jumbolen in unaligned location, so
1201 			 * we'd need to perform bcopy().
1202 			 */
1203 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1204 			jumboplen = (u_int32_t)htonl(jumboplen);
1205 
1206 #if 1
1207 			/*
1208 			 * if there are multiple jumbo payload options,
1209 			 * *plenp will be non-zero and the packet will be
1210 			 * rejected.
1211 			 * the behavior may need some debate in ipngwg -
1212 			 * multiple options does not make sense, however,
1213 			 * there's no explicit mention in specification.
1214 			 */
1215 			if (*plenp != 0) {
1216 				V_ip6stat.ip6s_badoptions++;
1217 				icmp6_error(m, ICMP6_PARAM_PROB,
1218 				    ICMP6_PARAMPROB_HEADER,
1219 				    erroff + opt + 2 - opthead);
1220 				return (-1);
1221 			}
1222 #endif
1223 
1224 			/*
1225 			 * jumbo payload length must be larger than 65535.
1226 			 */
1227 			if (jumboplen <= IPV6_MAXPACKET) {
1228 				V_ip6stat.ip6s_badoptions++;
1229 				icmp6_error(m, ICMP6_PARAM_PROB,
1230 				    ICMP6_PARAMPROB_HEADER,
1231 				    erroff + opt + 2 - opthead);
1232 				return (-1);
1233 			}
1234 			*plenp = jumboplen;
1235 
1236 			break;
1237 		default:		/* unknown option */
1238 			if (hbhlen < IP6OPT_MINLEN) {
1239 				V_ip6stat.ip6s_toosmall++;
1240 				goto bad;
1241 			}
1242 			optlen = ip6_unknown_opt(opt, m,
1243 			    erroff + opt - opthead);
1244 			if (optlen == -1)
1245 				return (-1);
1246 			optlen += 2;
1247 			break;
1248 		}
1249 	}
1250 
1251 	return (0);
1252 
1253   bad:
1254 	m_freem(m);
1255 	return (-1);
1256 }
1257 
1258 /*
1259  * Unknown option processing.
1260  * The third argument `off' is the offset from the IPv6 header to the option,
1261  * which is necessary if the IPv6 header the and option header and IPv6 header
1262  * is not contiguous in order to return an ICMPv6 error.
1263  */
1264 int
1265 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1266 {
1267 	struct ip6_hdr *ip6;
1268 
1269 	switch (IP6OPT_TYPE(*optp)) {
1270 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1271 		return ((int)*(optp + 1));
1272 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1273 		m_freem(m);
1274 		return (-1);
1275 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1276 		V_ip6stat.ip6s_badoptions++;
1277 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1278 		return (-1);
1279 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1280 		V_ip6stat.ip6s_badoptions++;
1281 		ip6 = mtod(m, struct ip6_hdr *);
1282 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1283 		    (m->m_flags & (M_BCAST|M_MCAST)))
1284 			m_freem(m);
1285 		else
1286 			icmp6_error(m, ICMP6_PARAM_PROB,
1287 				    ICMP6_PARAMPROB_OPTION, off);
1288 		return (-1);
1289 	}
1290 
1291 	m_freem(m);		/* XXX: NOTREACHED */
1292 	return (-1);
1293 }
1294 
1295 /*
1296  * Create the "control" list for this pcb.
1297  * These functions will not modify mbuf chain at all.
1298  *
1299  * With KAME mbuf chain restriction:
1300  * The routine will be called from upper layer handlers like tcp6_input().
1301  * Thus the routine assumes that the caller (tcp6_input) have already
1302  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1303  * very first mbuf on the mbuf chain.
1304  *
1305  * ip6_savecontrol_v4 will handle those options that are possible to be
1306  * set on a v4-mapped socket.
1307  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1308  * options and handle the v6-only ones itself.
1309  */
1310 struct mbuf **
1311 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1312     int *v4only)
1313 {
1314 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1315 
1316 #ifdef SO_TIMESTAMP
1317 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1318 		struct timeval tv;
1319 
1320 		microtime(&tv);
1321 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1322 		    SCM_TIMESTAMP, SOL_SOCKET);
1323 		if (*mp)
1324 			mp = &(*mp)->m_next;
1325 	}
1326 #endif
1327 
1328 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1329 	/* RFC 2292 sec. 5 */
1330 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1331 		struct in6_pktinfo pi6;
1332 
1333 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1334 #ifdef INET
1335 			struct ip *ip;
1336 
1337 			ip = mtod(m, struct ip *);
1338 			pi6.ipi6_addr.s6_addr32[0] = 0;
1339 			pi6.ipi6_addr.s6_addr32[1] = 0;
1340 			pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1341 			pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1342 #else
1343 			/* We won't hit this code */
1344 			bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1345 #endif
1346 		} else {
1347 			bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1348 			in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1349 		}
1350 		pi6.ipi6_ifindex =
1351 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1352 
1353 		*mp = sbcreatecontrol((caddr_t) &pi6,
1354 		    sizeof(struct in6_pktinfo),
1355 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1356 		if (*mp)
1357 			mp = &(*mp)->m_next;
1358 	}
1359 
1360 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1361 		int hlim;
1362 
1363 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1364 #ifdef INET
1365 			struct ip *ip;
1366 
1367 			ip = mtod(m, struct ip *);
1368 			hlim = ip->ip_ttl;
1369 #else
1370 			/* We won't hit this code */
1371 			hlim = 0;
1372 #endif
1373 		} else {
1374 			hlim = ip6->ip6_hlim & 0xff;
1375 		}
1376 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1377 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1378 		    IPPROTO_IPV6);
1379 		if (*mp)
1380 			mp = &(*mp)->m_next;
1381 	}
1382 
1383 	if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1384 		int tclass;
1385 
1386 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1387 #ifdef INET
1388 			struct ip *ip;
1389 
1390 			ip = mtod(m, struct ip *);
1391 			tclass = ip->ip_tos;
1392 #else
1393 			/* We won't hit this code */
1394 			tclass = 0;
1395 #endif
1396 		} else {
1397 			u_int32_t flowinfo;
1398 
1399 			flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1400 			flowinfo >>= 20;
1401 			tclass = flowinfo & 0xff;
1402 		}
1403 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int),
1404 		    IPV6_TCLASS, IPPROTO_IPV6);
1405 		if (*mp)
1406 			mp = &(*mp)->m_next;
1407 	}
1408 
1409 	if (v4only != NULL) {
1410 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1411 			*v4only = 1;
1412 		} else {
1413 			*v4only = 0;
1414 		}
1415 	}
1416 
1417 	return (mp);
1418 }
1419 
1420 void
1421 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1422 {
1423 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1424 	int v4only = 0;
1425 
1426 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1427 	if (v4only)
1428 		return;
1429 
1430 	/*
1431 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1432 	 * privilege for the option (see ip6_ctloutput), but it might be too
1433 	 * strict, since there might be some hop-by-hop options which can be
1434 	 * returned to normal user.
1435 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1436 	 */
1437 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1438 		/*
1439 		 * Check if a hop-by-hop options header is contatined in the
1440 		 * received packet, and if so, store the options as ancillary
1441 		 * data. Note that a hop-by-hop options header must be
1442 		 * just after the IPv6 header, which is assured through the
1443 		 * IPv6 input processing.
1444 		 */
1445 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1446 			struct ip6_hbh *hbh;
1447 			int hbhlen = 0;
1448 #ifdef PULLDOWN_TEST
1449 			struct mbuf *ext;
1450 #endif
1451 
1452 #ifndef PULLDOWN_TEST
1453 			hbh = (struct ip6_hbh *)(ip6 + 1);
1454 			hbhlen = (hbh->ip6h_len + 1) << 3;
1455 #else
1456 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1457 			    ip6->ip6_nxt);
1458 			if (ext == NULL) {
1459 				V_ip6stat.ip6s_tooshort++;
1460 				return;
1461 			}
1462 			hbh = mtod(ext, struct ip6_hbh *);
1463 			hbhlen = (hbh->ip6h_len + 1) << 3;
1464 			if (hbhlen != ext->m_len) {
1465 				m_freem(ext);
1466 				V_ip6stat.ip6s_tooshort++;
1467 				return;
1468 			}
1469 #endif
1470 
1471 			/*
1472 			 * XXX: We copy the whole header even if a
1473 			 * jumbo payload option is included, the option which
1474 			 * is to be removed before returning according to
1475 			 * RFC2292.
1476 			 * Note: this constraint is removed in RFC3542
1477 			 */
1478 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1479 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1480 			    IPPROTO_IPV6);
1481 			if (*mp)
1482 				mp = &(*mp)->m_next;
1483 #ifdef PULLDOWN_TEST
1484 			m_freem(ext);
1485 #endif
1486 		}
1487 	}
1488 
1489 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1490 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1491 
1492 		/*
1493 		 * Search for destination options headers or routing
1494 		 * header(s) through the header chain, and stores each
1495 		 * header as ancillary data.
1496 		 * Note that the order of the headers remains in
1497 		 * the chain of ancillary data.
1498 		 */
1499 		while (1) {	/* is explicit loop prevention necessary? */
1500 			struct ip6_ext *ip6e = NULL;
1501 			int elen;
1502 #ifdef PULLDOWN_TEST
1503 			struct mbuf *ext = NULL;
1504 #endif
1505 
1506 			/*
1507 			 * if it is not an extension header, don't try to
1508 			 * pull it from the chain.
1509 			 */
1510 			switch (nxt) {
1511 			case IPPROTO_DSTOPTS:
1512 			case IPPROTO_ROUTING:
1513 			case IPPROTO_HOPOPTS:
1514 			case IPPROTO_AH: /* is it possible? */
1515 				break;
1516 			default:
1517 				goto loopend;
1518 			}
1519 
1520 #ifndef PULLDOWN_TEST
1521 			if (off + sizeof(*ip6e) > m->m_len)
1522 				goto loopend;
1523 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1524 			if (nxt == IPPROTO_AH)
1525 				elen = (ip6e->ip6e_len + 2) << 2;
1526 			else
1527 				elen = (ip6e->ip6e_len + 1) << 3;
1528 			if (off + elen > m->m_len)
1529 				goto loopend;
1530 #else
1531 			ext = ip6_pullexthdr(m, off, nxt);
1532 			if (ext == NULL) {
1533 				V_ip6stat.ip6s_tooshort++;
1534 				return;
1535 			}
1536 			ip6e = mtod(ext, struct ip6_ext *);
1537 			if (nxt == IPPROTO_AH)
1538 				elen = (ip6e->ip6e_len + 2) << 2;
1539 			else
1540 				elen = (ip6e->ip6e_len + 1) << 3;
1541 			if (elen != ext->m_len) {
1542 				m_freem(ext);
1543 				V_ip6stat.ip6s_tooshort++;
1544 				return;
1545 			}
1546 #endif
1547 
1548 			switch (nxt) {
1549 			case IPPROTO_DSTOPTS:
1550 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1551 					break;
1552 
1553 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1554 				    IS2292(in6p,
1555 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1556 				    IPPROTO_IPV6);
1557 				if (*mp)
1558 					mp = &(*mp)->m_next;
1559 				break;
1560 			case IPPROTO_ROUTING:
1561 				if (!(in6p->inp_flags & IN6P_RTHDR))
1562 					break;
1563 
1564 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1565 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1566 				    IPPROTO_IPV6);
1567 				if (*mp)
1568 					mp = &(*mp)->m_next;
1569 				break;
1570 			case IPPROTO_HOPOPTS:
1571 			case IPPROTO_AH: /* is it possible? */
1572 				break;
1573 
1574 			default:
1575 				/*
1576 				 * other cases have been filtered in the above.
1577 				 * none will visit this case.  here we supply
1578 				 * the code just in case (nxt overwritten or
1579 				 * other cases).
1580 				 */
1581 #ifdef PULLDOWN_TEST
1582 				m_freem(ext);
1583 #endif
1584 				goto loopend;
1585 
1586 			}
1587 
1588 			/* proceed with the next header. */
1589 			off += elen;
1590 			nxt = ip6e->ip6e_nxt;
1591 			ip6e = NULL;
1592 #ifdef PULLDOWN_TEST
1593 			m_freem(ext);
1594 			ext = NULL;
1595 #endif
1596 		}
1597 	  loopend:
1598 		;
1599 	}
1600 }
1601 #undef IS2292
1602 
1603 void
1604 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1605 {
1606 	struct socket *so;
1607 	struct mbuf *m_mtu;
1608 	struct ip6_mtuinfo mtuctl;
1609 
1610 	so =  in6p->inp_socket;
1611 
1612 	if (mtu == NULL)
1613 		return;
1614 
1615 #ifdef DIAGNOSTIC
1616 	if (so == NULL)		/* I believe this is impossible */
1617 		panic("ip6_notify_pmtu: socket is NULL");
1618 #endif
1619 
1620 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1621 	mtuctl.ip6m_mtu = *mtu;
1622 	mtuctl.ip6m_addr = *dst;
1623 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1624 		return;
1625 
1626 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1627 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1628 		return;
1629 
1630 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1631 	    == 0) {
1632 		m_freem(m_mtu);
1633 		/* XXX: should count statistics */
1634 	} else
1635 		sorwakeup(so);
1636 
1637 	return;
1638 }
1639 
1640 #ifdef PULLDOWN_TEST
1641 /*
1642  * pull single extension header from mbuf chain.  returns single mbuf that
1643  * contains the result, or NULL on error.
1644  */
1645 static struct mbuf *
1646 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1647 {
1648 	struct ip6_ext ip6e;
1649 	size_t elen;
1650 	struct mbuf *n;
1651 
1652 #ifdef DIAGNOSTIC
1653 	switch (nxt) {
1654 	case IPPROTO_DSTOPTS:
1655 	case IPPROTO_ROUTING:
1656 	case IPPROTO_HOPOPTS:
1657 	case IPPROTO_AH: /* is it possible? */
1658 		break;
1659 	default:
1660 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1661 	}
1662 #endif
1663 
1664 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1665 	if (nxt == IPPROTO_AH)
1666 		elen = (ip6e.ip6e_len + 2) << 2;
1667 	else
1668 		elen = (ip6e.ip6e_len + 1) << 3;
1669 
1670 	MGET(n, M_NOWAIT, MT_DATA);
1671 	if (n && elen >= MLEN) {
1672 		MCLGET(n, M_NOWAIT);
1673 		if ((n->m_flags & M_EXT) == 0) {
1674 			m_free(n);
1675 			n = NULL;
1676 		}
1677 	}
1678 	if (!n)
1679 		return NULL;
1680 
1681 	n->m_len = 0;
1682 	if (elen >= M_TRAILINGSPACE(n)) {
1683 		m_free(n);
1684 		return NULL;
1685 	}
1686 
1687 	m_copydata(m, off, elen, mtod(n, caddr_t));
1688 	n->m_len = elen;
1689 	return n;
1690 }
1691 #endif
1692 
1693 /*
1694  * Get pointer to the previous header followed by the header
1695  * currently processed.
1696  * XXX: This function supposes that
1697  *	M includes all headers,
1698  *	the next header field and the header length field of each header
1699  *	are valid, and
1700  *	the sum of each header length equals to OFF.
1701  * Because of these assumptions, this function must be called very
1702  * carefully. Moreover, it will not be used in the near future when
1703  * we develop `neater' mechanism to process extension headers.
1704  */
1705 char *
1706 ip6_get_prevhdr(struct mbuf *m, int off)
1707 {
1708 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1709 
1710 	if (off == sizeof(struct ip6_hdr))
1711 		return (&ip6->ip6_nxt);
1712 	else {
1713 		int len, nxt;
1714 		struct ip6_ext *ip6e = NULL;
1715 
1716 		nxt = ip6->ip6_nxt;
1717 		len = sizeof(struct ip6_hdr);
1718 		while (len < off) {
1719 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1720 
1721 			switch (nxt) {
1722 			case IPPROTO_FRAGMENT:
1723 				len += sizeof(struct ip6_frag);
1724 				break;
1725 			case IPPROTO_AH:
1726 				len += (ip6e->ip6e_len + 2) << 2;
1727 				break;
1728 			default:
1729 				len += (ip6e->ip6e_len + 1) << 3;
1730 				break;
1731 			}
1732 			nxt = ip6e->ip6e_nxt;
1733 		}
1734 		if (ip6e)
1735 			return (&ip6e->ip6e_nxt);
1736 		else
1737 			return NULL;
1738 	}
1739 }
1740 
1741 /*
1742  * get next header offset.  m will be retained.
1743  */
1744 int
1745 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1746 {
1747 	struct ip6_hdr ip6;
1748 	struct ip6_ext ip6e;
1749 	struct ip6_frag fh;
1750 
1751 	/* just in case */
1752 	if (m == NULL)
1753 		panic("ip6_nexthdr: m == NULL");
1754 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1755 		return -1;
1756 
1757 	switch (proto) {
1758 	case IPPROTO_IPV6:
1759 		if (m->m_pkthdr.len < off + sizeof(ip6))
1760 			return -1;
1761 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1762 		if (nxtp)
1763 			*nxtp = ip6.ip6_nxt;
1764 		off += sizeof(ip6);
1765 		return off;
1766 
1767 	case IPPROTO_FRAGMENT:
1768 		/*
1769 		 * terminate parsing if it is not the first fragment,
1770 		 * it does not make sense to parse through it.
1771 		 */
1772 		if (m->m_pkthdr.len < off + sizeof(fh))
1773 			return -1;
1774 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1775 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1776 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1777 			return -1;
1778 		if (nxtp)
1779 			*nxtp = fh.ip6f_nxt;
1780 		off += sizeof(struct ip6_frag);
1781 		return off;
1782 
1783 	case IPPROTO_AH:
1784 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1785 			return -1;
1786 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1787 		if (nxtp)
1788 			*nxtp = ip6e.ip6e_nxt;
1789 		off += (ip6e.ip6e_len + 2) << 2;
1790 		return off;
1791 
1792 	case IPPROTO_HOPOPTS:
1793 	case IPPROTO_ROUTING:
1794 	case IPPROTO_DSTOPTS:
1795 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1796 			return -1;
1797 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1798 		if (nxtp)
1799 			*nxtp = ip6e.ip6e_nxt;
1800 		off += (ip6e.ip6e_len + 1) << 3;
1801 		return off;
1802 
1803 	case IPPROTO_NONE:
1804 	case IPPROTO_ESP:
1805 	case IPPROTO_IPCOMP:
1806 		/* give up */
1807 		return -1;
1808 
1809 	default:
1810 		return -1;
1811 	}
1812 
1813 	return -1;
1814 }
1815 
1816 /*
1817  * get offset for the last header in the chain.  m will be kept untainted.
1818  */
1819 int
1820 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1821 {
1822 	int newoff;
1823 	int nxt;
1824 
1825 	if (!nxtp) {
1826 		nxt = -1;
1827 		nxtp = &nxt;
1828 	}
1829 	while (1) {
1830 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1831 		if (newoff < 0)
1832 			return off;
1833 		else if (newoff < off)
1834 			return -1;	/* invalid */
1835 		else if (newoff == off)
1836 			return newoff;
1837 
1838 		off = newoff;
1839 		proto = *nxtp;
1840 	}
1841 }
1842 
1843 static struct ip6aux *
1844 ip6_addaux(struct mbuf *m)
1845 {
1846 	struct m_tag *mtag;
1847 
1848 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1849 	if (!mtag) {
1850 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1851 		    M_NOWAIT);
1852 		if (mtag) {
1853 			m_tag_prepend(m, mtag);
1854 			bzero(mtag + 1, sizeof(struct ip6aux));
1855 		}
1856 	}
1857 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1858 }
1859 
1860 static struct ip6aux *
1861 ip6_findaux(struct mbuf *m)
1862 {
1863 	struct m_tag *mtag;
1864 
1865 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1866 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1867 }
1868 
1869 static void
1870 ip6_delaux(struct mbuf *m)
1871 {
1872 	struct m_tag *mtag;
1873 
1874 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1875 	if (mtag)
1876 		m_tag_delete(m, mtag);
1877 }
1878 
1879 /*
1880  * System control for IP6
1881  */
1882 
1883 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1884 	0,		0,		0,		0,
1885 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1886 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1887 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1888 	0,		0,		0,		0,
1889 	ENOPROTOOPT
1890 };
1891