xref: /freebsd/sys/netinet6/ip6_input.c (revision 9ecd54f24fe9fa373e07c9fd7c052deb2188f545)
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/sdt.h>
80 #include <sys/socket.h>
81 #include <sys/socketvar.h>
82 #include <sys/errno.h>
83 #include <sys/time.h>
84 #include <sys/kernel.h>
85 #include <sys/syslog.h>
86 
87 #include <net/if.h>
88 #include <net/if_var.h>
89 #include <net/if_types.h>
90 #include <net/if_dl.h>
91 #include <net/route.h>
92 #include <net/netisr.h>
93 #include <net/pfil.h>
94 #include <net/vnet.h>
95 
96 #include <netinet/in.h>
97 #include <netinet/in_kdtrace.h>
98 #include <netinet/ip_var.h>
99 #include <netinet/in_systm.h>
100 #include <net/if_llatbl.h>
101 #ifdef INET
102 #include <netinet/ip.h>
103 #include <netinet/ip_icmp.h>
104 #endif /* INET */
105 #include <netinet/ip6.h>
106 #include <netinet6/in6_var.h>
107 #include <netinet6/ip6_var.h>
108 #include <netinet/in_pcb.h>
109 #include <netinet/icmp6.h>
110 #include <netinet6/scope6_var.h>
111 #include <netinet6/in6_ifattach.h>
112 #include <netinet6/nd6.h>
113 
114 #ifdef IPSEC
115 #include <netipsec/ipsec.h>
116 #include <netinet6/ip6_ipsec.h>
117 #include <netipsec/ipsec6.h>
118 #endif /* IPSEC */
119 
120 #include <netinet6/ip6protosw.h>
121 
122 extern struct domain inet6domain;
123 
124 u_char ip6_protox[IPPROTO_MAX];
125 VNET_DEFINE(struct in6_ifaddrhead, in6_ifaddrhead);
126 VNET_DEFINE(struct in6_ifaddrlisthead *, in6_ifaddrhashtbl);
127 VNET_DEFINE(u_long, in6_ifaddrhmask);
128 
129 static struct netisr_handler ip6_nh = {
130 	.nh_name = "ip6",
131 	.nh_handler = ip6_input,
132 	.nh_proto = NETISR_IPV6,
133 	.nh_policy = NETISR_POLICY_FLOW,
134 };
135 
136 VNET_DECLARE(struct callout, in6_tmpaddrtimer_ch);
137 #define	V_in6_tmpaddrtimer_ch		VNET(in6_tmpaddrtimer_ch)
138 
139 VNET_DEFINE(struct pfil_head, inet6_pfil_hook);
140 
141 VNET_PCPUSTAT_DEFINE(struct ip6stat, ip6stat);
142 VNET_PCPUSTAT_SYSINIT(ip6stat);
143 #ifdef VIMAGE
144 VNET_PCPUSTAT_SYSUNINIT(ip6stat);
145 #endif /* VIMAGE */
146 
147 struct rwlock in6_ifaddr_lock;
148 RW_SYSINIT(in6_ifaddr_lock, &in6_ifaddr_lock, "in6_ifaddr_lock");
149 
150 static void ip6_init2(void *);
151 static struct ip6aux *ip6_setdstifaddr(struct mbuf *, struct in6_ifaddr *);
152 static struct ip6aux *ip6_addaux(struct mbuf *);
153 static struct ip6aux *ip6_findaux(struct mbuf *m);
154 static void ip6_delaux (struct mbuf *);
155 static int ip6_hopopts_input(u_int32_t *, u_int32_t *, struct mbuf **, int *);
156 #ifdef PULLDOWN_TEST
157 static struct mbuf *ip6_pullexthdr(struct mbuf *, size_t, int);
158 #endif
159 
160 /*
161  * IP6 initialization: fill in IP6 protocol switch table.
162  * All protocols not implemented in kernel go to raw IP6 protocol handler.
163  */
164 void
165 ip6_init(void)
166 {
167 	struct ip6protosw *pr;
168 	int i;
169 
170 	TUNABLE_INT_FETCH("net.inet6.ip6.auto_linklocal",
171 	    &V_ip6_auto_linklocal);
172 	TUNABLE_INT_FETCH("net.inet6.ip6.accept_rtadv", &V_ip6_accept_rtadv);
173 	TUNABLE_INT_FETCH("net.inet6.ip6.no_radr", &V_ip6_no_radr);
174 
175 	TAILQ_INIT(&V_in6_ifaddrhead);
176 	V_in6_ifaddrhashtbl = hashinit(IN6ADDR_NHASH, M_IFADDR,
177 	    &V_in6_ifaddrhmask);
178 
179 	/* Initialize packet filter hooks. */
180 	V_inet6_pfil_hook.ph_type = PFIL_TYPE_AF;
181 	V_inet6_pfil_hook.ph_af = AF_INET6;
182 	if ((i = pfil_head_register(&V_inet6_pfil_hook)) != 0)
183 		printf("%s: WARNING: unable to register pfil hook, "
184 			"error %d\n", __func__, i);
185 
186 	scope6_init();
187 	addrsel_policy_init();
188 	nd6_init();
189 	frag6_init();
190 
191 	V_ip6_desync_factor = arc4random() % MAX_TEMP_DESYNC_FACTOR;
192 
193 	/* Skip global initialization stuff for non-default instances. */
194 	if (!IS_DEFAULT_VNET(curvnet))
195 		return;
196 
197 #ifdef DIAGNOSTIC
198 	if (sizeof(struct protosw) != sizeof(struct ip6protosw))
199 		panic("sizeof(protosw) != sizeof(ip6protosw)");
200 #endif
201 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
202 	if (pr == NULL)
203 		panic("ip6_init");
204 
205 	/* Initialize the entire ip6_protox[] array to IPPROTO_RAW. */
206 	for (i = 0; i < IPPROTO_MAX; i++)
207 		ip6_protox[i] = pr - inet6sw;
208 	/*
209 	 * Cycle through IP protocols and put them into the appropriate place
210 	 * in ip6_protox[].
211 	 */
212 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
213 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++)
214 		if (pr->pr_domain->dom_family == PF_INET6 &&
215 		    pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) {
216 			/* Be careful to only index valid IP protocols. */
217 			if (pr->pr_protocol < IPPROTO_MAX)
218 				ip6_protox[pr->pr_protocol] = pr - inet6sw;
219 		}
220 
221 	netisr_register(&ip6_nh);
222 }
223 
224 /*
225  * The protocol to be inserted into ip6_protox[] must be already registered
226  * in inet6sw[], either statically or through pf_proto_register().
227  */
228 int
229 ip6proto_register(short ip6proto)
230 {
231 	struct ip6protosw *pr;
232 
233 	/* Sanity checks. */
234 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
235 		return (EPROTONOSUPPORT);
236 
237 	/*
238 	 * The protocol slot must not be occupied by another protocol
239 	 * already.  An index pointing to IPPROTO_RAW is unused.
240 	 */
241 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
242 	if (pr == NULL)
243 		return (EPFNOSUPPORT);
244 	if (ip6_protox[ip6proto] != pr - inet6sw)	/* IPPROTO_RAW */
245 		return (EEXIST);
246 
247 	/*
248 	 * Find the protocol position in inet6sw[] and set the index.
249 	 */
250 	for (pr = (struct ip6protosw *)inet6domain.dom_protosw;
251 	    pr < (struct ip6protosw *)inet6domain.dom_protoswNPROTOSW; pr++) {
252 		if (pr->pr_domain->dom_family == PF_INET6 &&
253 		    pr->pr_protocol && pr->pr_protocol == ip6proto) {
254 			ip6_protox[pr->pr_protocol] = pr - inet6sw;
255 			return (0);
256 		}
257 	}
258 	return (EPROTONOSUPPORT);
259 }
260 
261 int
262 ip6proto_unregister(short ip6proto)
263 {
264 	struct ip6protosw *pr;
265 
266 	/* Sanity checks. */
267 	if (ip6proto <= 0 || ip6proto >= IPPROTO_MAX)
268 		return (EPROTONOSUPPORT);
269 
270 	/* Check if the protocol was indeed registered. */
271 	pr = (struct ip6protosw *)pffindproto(PF_INET6, IPPROTO_RAW, SOCK_RAW);
272 	if (pr == NULL)
273 		return (EPFNOSUPPORT);
274 	if (ip6_protox[ip6proto] == pr - inet6sw)	/* IPPROTO_RAW */
275 		return (ENOENT);
276 
277 	/* Reset the protocol slot to IPPROTO_RAW. */
278 	ip6_protox[ip6proto] = pr - inet6sw;
279 	return (0);
280 }
281 
282 #ifdef VIMAGE
283 void
284 ip6_destroy()
285 {
286 	int i;
287 
288 	if ((i = pfil_head_unregister(&V_inet6_pfil_hook)) != 0)
289 		printf("%s: WARNING: unable to unregister pfil hook, "
290 		    "error %d\n", __func__, i);
291 	hashdestroy(V_in6_ifaddrhashtbl, M_IFADDR, V_in6_ifaddrhmask);
292 	nd6_destroy();
293 	callout_drain(&V_in6_tmpaddrtimer_ch);
294 }
295 #endif
296 
297 static int
298 ip6_init2_vnet(const void *unused __unused)
299 {
300 
301 	/* nd6_timer_init */
302 	callout_init(&V_nd6_timer_ch, 0);
303 	callout_reset(&V_nd6_timer_ch, hz, nd6_timer, curvnet);
304 
305 	/* timer for regeneranation of temporary addresses randomize ID */
306 	callout_init(&V_in6_tmpaddrtimer_ch, 0);
307 	callout_reset(&V_in6_tmpaddrtimer_ch,
308 		      (V_ip6_temp_preferred_lifetime - V_ip6_desync_factor -
309 		       V_ip6_temp_regen_advance) * hz,
310 		      in6_tmpaddrtimer, curvnet);
311 
312 	return (0);
313 }
314 
315 static void
316 ip6_init2(void *dummy)
317 {
318 
319 	ip6_init2_vnet(NULL);
320 }
321 
322 /* cheat */
323 /* This must be after route_init(), which is now SI_ORDER_THIRD */
324 SYSINIT(netinet6init2, SI_SUB_PROTO_DOMAIN, SI_ORDER_MIDDLE, ip6_init2, NULL);
325 
326 static int
327 ip6_input_hbh(struct mbuf *m, uint32_t *plen, uint32_t *rtalert, int *off,
328     int *nxt, int *ours)
329 {
330 	struct ip6_hdr *ip6;
331 	struct ip6_hbh *hbh;
332 
333 	if (ip6_hopopts_input(plen, rtalert, &m, off)) {
334 #if 0	/*touches NULL pointer*/
335 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
336 #endif
337 		goto out;	/* m have already been freed */
338 	}
339 
340 	/* adjust pointer */
341 	ip6 = mtod(m, struct ip6_hdr *);
342 
343 	/*
344 	 * if the payload length field is 0 and the next header field
345 	 * indicates Hop-by-Hop Options header, then a Jumbo Payload
346 	 * option MUST be included.
347 	 */
348 	if (ip6->ip6_plen == 0 && *plen == 0) {
349 		/*
350 		 * Note that if a valid jumbo payload option is
351 		 * contained, ip6_hopopts_input() must set a valid
352 		 * (non-zero) payload length to the variable plen.
353 		 */
354 		IP6STAT_INC(ip6s_badoptions);
355 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
356 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
357 		icmp6_error(m, ICMP6_PARAM_PROB,
358 			    ICMP6_PARAMPROB_HEADER,
359 			    (caddr_t)&ip6->ip6_plen - (caddr_t)ip6);
360 		goto out;
361 	}
362 #ifndef PULLDOWN_TEST
363 	/* ip6_hopopts_input() ensures that mbuf is contiguous */
364 	hbh = (struct ip6_hbh *)(ip6 + 1);
365 #else
366 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
367 		sizeof(struct ip6_hbh));
368 	if (hbh == NULL) {
369 		IP6STAT_INC(ip6s_tooshort);
370 		goto out;
371 	}
372 #endif
373 	*nxt = hbh->ip6h_nxt;
374 
375 	/*
376 	 * If we are acting as a router and the packet contains a
377 	 * router alert option, see if we know the option value.
378 	 * Currently, we only support the option value for MLD, in which
379 	 * case we should pass the packet to the multicast routing
380 	 * daemon.
381 	 */
382 	if (*rtalert != ~0) {
383 		switch (*rtalert) {
384 		case IP6OPT_RTALERT_MLD:
385 			if (V_ip6_forwarding)
386 				*ours = 1;
387 			break;
388 		default:
389 			/*
390 			 * RFC2711 requires unrecognized values must be
391 			 * silently ignored.
392 			 */
393 			break;
394 		}
395 	}
396 
397 	return (0);
398 
399 out:
400 	return (1);
401 }
402 
403 void
404 ip6_input(struct mbuf *m)
405 {
406 	struct ip6_hdr *ip6;
407 	int off = sizeof(struct ip6_hdr), nest;
408 	u_int32_t plen;
409 	u_int32_t rtalert = ~0;
410 	int nxt, ours = 0;
411 	struct ifnet *deliverifp = NULL, *ifp = NULL;
412 	struct in6_addr odst;
413 	struct route_in6 rin6;
414 	int srcrt = 0;
415 	struct llentry *lle = NULL;
416 	struct sockaddr_in6 dst6, *dst;
417 
418 	bzero(&rin6, sizeof(struct route_in6));
419 #ifdef IPSEC
420 	/*
421 	 * should the inner packet be considered authentic?
422 	 * see comment in ah4_input().
423 	 * NB: m cannot be NULL when passed to the input routine
424 	 */
425 
426 	m->m_flags &= ~M_AUTHIPHDR;
427 	m->m_flags &= ~M_AUTHIPDGM;
428 
429 #endif /* IPSEC */
430 
431 	/*
432 	 * make sure we don't have onion peering information into m_tag.
433 	 */
434 	ip6_delaux(m);
435 
436 	if (m->m_flags & M_FASTFWD_OURS) {
437 		/*
438 		 * Firewall changed destination to local.
439 		 */
440 		m->m_flags &= ~M_FASTFWD_OURS;
441 		ours = 1;
442 		deliverifp = m->m_pkthdr.rcvif;
443 		ip6 = mtod(m, struct ip6_hdr *);
444 		goto hbhcheck;
445 	}
446 
447 	/*
448 	 * mbuf statistics
449 	 */
450 	if (m->m_flags & M_EXT) {
451 		if (m->m_next)
452 			IP6STAT_INC(ip6s_mext2m);
453 		else
454 			IP6STAT_INC(ip6s_mext1);
455 	} else {
456 		if (m->m_next) {
457 			if (m->m_flags & M_LOOP) {
458 				IP6STAT_INC(ip6s_m2m[V_loif->if_index]);
459 			} else if (m->m_pkthdr.rcvif->if_index < IP6S_M2MMAX)
460 				IP6STAT_INC(
461 				    ip6s_m2m[m->m_pkthdr.rcvif->if_index]);
462 			else
463 				IP6STAT_INC(ip6s_m2m[0]);
464 		} else
465 			IP6STAT_INC(ip6s_m1);
466 	}
467 
468 	/* drop the packet if IPv6 operation is disabled on the IF */
469 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
470 		m_freem(m);
471 		return;
472 	}
473 
474 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
475 	IP6STAT_INC(ip6s_total);
476 
477 #ifndef PULLDOWN_TEST
478 	/*
479 	 * L2 bridge code and some other code can return mbuf chain
480 	 * that does not conform to KAME requirement.  too bad.
481 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
482 	 */
483 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
484 		struct mbuf *n;
485 
486 		if (m->m_pkthdr.len > MHLEN)
487 			n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
488 		else
489 			n = m_gethdr(M_NOWAIT, MT_DATA);
490 		if (n == NULL) {
491 			m_freem(m);
492 			return;	/* ENOBUFS */
493 		}
494 
495 		m_move_pkthdr(n, m);
496 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
497 		n->m_len = n->m_pkthdr.len;
498 		m_freem(m);
499 		m = n;
500 	}
501 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
502 #endif
503 
504 	if (m->m_len < sizeof(struct ip6_hdr)) {
505 		struct ifnet *inifp;
506 		inifp = m->m_pkthdr.rcvif;
507 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
508 			IP6STAT_INC(ip6s_toosmall);
509 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
510 			return;
511 		}
512 	}
513 
514 	ip6 = mtod(m, struct ip6_hdr *);
515 
516 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
517 		IP6STAT_INC(ip6s_badvers);
518 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
519 		goto bad;
520 	}
521 
522 	IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]);
523 
524 	IP_PROBE(receive, NULL, NULL, ip6, m->m_pkthdr.rcvif, NULL, ip6);
525 
526 	/*
527 	 * Check against address spoofing/corruption.
528 	 */
529 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
530 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
531 		/*
532 		 * XXX: "badscope" is not very suitable for a multicast source.
533 		 */
534 		IP6STAT_INC(ip6s_badscope);
535 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
536 		goto bad;
537 	}
538 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
539 	    !(m->m_flags & M_LOOP)) {
540 		/*
541 		 * In this case, the packet should come from the loopback
542 		 * interface.  However, we cannot just check the if_flags,
543 		 * because ip6_mloopback() passes the "actual" interface
544 		 * as the outgoing/incoming interface.
545 		 */
546 		IP6STAT_INC(ip6s_badscope);
547 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
548 		goto bad;
549 	}
550 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) &&
551 	    IPV6_ADDR_MC_SCOPE(&ip6->ip6_dst) == 0) {
552 		/*
553 		 * RFC4291 2.7:
554 		 * Nodes must not originate a packet to a multicast address
555 		 * whose scop field contains the reserved value 0; if such
556 		 * a packet is received, it must be silently dropped.
557 		 */
558 		IP6STAT_INC(ip6s_badscope);
559 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
560 		goto bad;
561 	}
562 #ifdef ALTQ
563 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
564 		/* packet is dropped by traffic conditioner */
565 		return;
566 	}
567 #endif
568 	/*
569 	 * The following check is not documented in specs.  A malicious
570 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
571 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
572 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
573 	 *
574 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
575 	 * support IPv4-less kernel compilation, we cannot support SIIT
576 	 * environment at all.  So, it makes more sense for us to reject any
577 	 * malicious packets for non-SIIT environment, than try to do a
578 	 * partial support for SIIT environment.
579 	 */
580 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
581 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
582 		IP6STAT_INC(ip6s_badscope);
583 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
584 		goto bad;
585 	}
586 #if 0
587 	/*
588 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
589 	 *
590 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
591 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
592 	 * is revised to forbid relaying case.
593 	 */
594 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
595 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
596 		IP6STAT_INC(ip6s_badscope);
597 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
598 		goto bad;
599 	}
600 #endif
601 #ifdef IPSEC
602 	/*
603 	 * Bypass packet filtering for packets previously handled by IPsec.
604 	 */
605 	if (ip6_ipsec_filtertunnel(m))
606 		goto passin;
607 #endif /* IPSEC */
608 
609 	/*
610 	 * Run through list of hooks for input packets.
611 	 *
612 	 * NB: Beware of the destination address changing
613 	 *     (e.g. by NAT rewriting).  When this happens,
614 	 *     tell ip6_forward to do the right thing.
615 	 */
616 	odst = ip6->ip6_dst;
617 
618 	/* Jump over all PFIL processing if hooks are not active. */
619 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
620 		goto passin;
621 
622 	if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
623 	    m->m_pkthdr.rcvif, PFIL_IN, NULL))
624 		return;
625 	if (m == NULL)			/* consumed by filter */
626 		return;
627 	ip6 = mtod(m, struct ip6_hdr *);
628 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
629 
630 	if (m->m_flags & M_FASTFWD_OURS) {
631 		m->m_flags &= ~M_FASTFWD_OURS;
632 		ours = 1;
633 		deliverifp = m->m_pkthdr.rcvif;
634 		goto hbhcheck;
635 	}
636 	if ((m->m_flags & M_IP6_NEXTHOP) &&
637 	    m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
638 		/*
639 		 * Directly ship the packet on.  This allows forwarding
640 		 * packets originally destined to us to some other directly
641 		 * connected host.
642 		 */
643 		ip6_forward(m, 1);
644 		goto out;
645 	}
646 
647 passin:
648 	/*
649 	 * Disambiguate address scope zones (if there is ambiguity).
650 	 * We first make sure that the original source or destination address
651 	 * is not in our internal form for scoped addresses.  Such addresses
652 	 * are not necessarily invalid spec-wise, but we cannot accept them due
653 	 * to the usage conflict.
654 	 * in6_setscope() then also checks and rejects the cases where src or
655 	 * dst are the loopback address and the receiving interface
656 	 * is not loopback.
657 	 */
658 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
659 		IP6STAT_INC(ip6s_badscope); /* XXX */
660 		goto bad;
661 	}
662 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
663 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
664 		IP6STAT_INC(ip6s_badscope);
665 		goto bad;
666 	}
667 
668 	/*
669 	 * Multicast check. Assume packet is for us to avoid
670 	 * prematurely taking locks.
671 	 */
672 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
673 		ours = 1;
674 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
675 		deliverifp = m->m_pkthdr.rcvif;
676 		goto hbhcheck;
677 	}
678 
679 	/*
680 	 *  Unicast check
681 	 */
682 
683 	bzero(&dst6, sizeof(dst6));
684 	dst6.sin6_family = AF_INET6;
685 	dst6.sin6_len = sizeof(struct sockaddr_in6);
686 	dst6.sin6_addr = ip6->ip6_dst;
687 	ifp = m->m_pkthdr.rcvif;
688 	IF_AFDATA_RLOCK(ifp);
689 	lle = lla_lookup(LLTABLE6(ifp), 0,
690 	     (struct sockaddr *)&dst6);
691 	IF_AFDATA_RUNLOCK(ifp);
692 	if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) {
693 		struct ifaddr *ifa;
694 		struct in6_ifaddr *ia6;
695 		int bad;
696 
697 		bad = 1;
698 #define	sa_equal(a1, a2)						\
699 	(bcmp((a1), (a2), ((a1))->sin6_len) == 0)
700 		IF_ADDR_RLOCK(ifp);
701 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
702 			if (ifa->ifa_addr->sa_family != dst6.sin6_family)
703 				continue;
704 			if (sa_equal(&dst6, ifa->ifa_addr))
705 				break;
706 		}
707 		KASSERT(ifa != NULL, ("%s: ifa not found for lle %p",
708 		    __func__, lle));
709 #undef sa_equal
710 
711 		ia6 = (struct in6_ifaddr *)ifa;
712 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
713 			/* Count the packet in the ip address stats */
714 			counter_u64_add(ia6->ia_ifa.ifa_ipackets, 1);
715 			counter_u64_add(ia6->ia_ifa.ifa_ibytes,
716 			    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 			counter_u64_add(ia6->ia_ifa.ifa_ipackets, 1);
832 			counter_u64_add(ia6->ia_ifa.ifa_ibytes,
833 			    m->m_pkthdr.len);
834 			if (free_ia6)
835 				ifa_free(&ia6->ia_ifa);
836 			goto hbhcheck;
837 		} else {
838 			char ip6bufs[INET6_ADDRSTRLEN];
839 			char ip6bufd[INET6_ADDRSTRLEN];
840 			/* address is not ready, so discard the packet. */
841 			nd6log((LOG_INFO,
842 			    "ip6_input: packet to an unready address %s->%s\n",
843 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
844 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
845 
846 			if (free_ia6)
847 				ifa_free(&ia6->ia_ifa);
848 			goto bad;
849 		}
850 	}
851 
852 	/*
853 	 * FAITH (Firewall Aided Internet Translator)
854 	 */
855 	if (V_ip6_keepfaith) {
856 		if (rin6.ro_rt && rin6.ro_rt->rt_ifp &&
857 		    rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) {
858 			/* XXX do we need more sanity checks? */
859 			ours = 1;
860 			deliverifp = rin6.ro_rt->rt_ifp; /* faith */
861 			goto hbhcheck;
862 		}
863 	}
864 
865 	/*
866 	 * Now there is no reason to process the packet if it's not our own
867 	 * and we're not a router.
868 	 */
869 	if (!V_ip6_forwarding) {
870 		IP6STAT_INC(ip6s_cantforward);
871 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
872 		goto bad;
873 	}
874 
875   hbhcheck:
876 	/*
877 	 * record address information into m_tag, if we don't have one yet.
878 	 * note that we are unable to record it, if the address is not listed
879 	 * as our interface address (e.g. multicast addresses, addresses
880 	 * within FAITH prefixes and such).
881 	 */
882 	if (deliverifp) {
883 		struct in6_ifaddr *ia6;
884 
885  		if ((ia6 = ip6_getdstifaddr(m)) != NULL) {
886 			ifa_free(&ia6->ia_ifa);
887 		} else {
888 			ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
889 			if (ia6) {
890 				if (!ip6_setdstifaddr(m, ia6)) {
891 					/*
892 					 * XXX maybe we should drop the packet here,
893 					 * as we could not provide enough information
894 					 * to the upper layers.
895 					 */
896 				}
897 				ifa_free(&ia6->ia_ifa);
898 			}
899 		}
900 	}
901 
902 	/*
903 	 * Process Hop-by-Hop options header if it's contained.
904 	 * m may be modified in ip6_hopopts_input().
905 	 * If a JumboPayload option is included, plen will also be modified.
906 	 */
907 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
908 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
909 		int error;
910 
911 		error = ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours);
912 		if (error != 0)
913 			goto out;
914 	} else
915 		nxt = ip6->ip6_nxt;
916 
917 	/*
918 	 * Check that the amount of data in the buffers
919 	 * is as at least much as the IPv6 header would have us expect.
920 	 * Trim mbufs if longer than we expect.
921 	 * Drop packet if shorter than we expect.
922 	 */
923 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
924 		IP6STAT_INC(ip6s_tooshort);
925 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
926 		goto bad;
927 	}
928 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
929 		if (m->m_len == m->m_pkthdr.len) {
930 			m->m_len = sizeof(struct ip6_hdr) + plen;
931 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
932 		} else
933 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
934 	}
935 
936 	/*
937 	 * Forward if desirable.
938 	 */
939 	if (V_ip6_mrouter &&
940 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
941 		/*
942 		 * If we are acting as a multicast router, all
943 		 * incoming multicast packets are passed to the
944 		 * kernel-level multicast forwarding function.
945 		 * The packet is returned (relatively) intact; if
946 		 * ip6_mforward() returns a non-zero value, the packet
947 		 * must be discarded, else it may be accepted below.
948 		 *
949 		 * XXX TODO: Check hlim and multicast scope here to avoid
950 		 * unnecessarily calling into ip6_mforward().
951 		 */
952 		if (ip6_mforward &&
953 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
954 			IP6STAT_INC(ip6s_cantforward);
955 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
956 			goto bad;
957 		}
958 	} else if (!ours) {
959 		ip6_forward(m, srcrt);
960 		goto out;
961 	}
962 
963 	ip6 = mtod(m, struct ip6_hdr *);
964 
965 	/*
966 	 * Malicious party may be able to use IPv4 mapped addr to confuse
967 	 * tcp/udp stack and bypass security checks (act as if it was from
968 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
969 	 *
970 	 * For SIIT end node behavior, you may want to disable the check.
971 	 * However, you will  become vulnerable to attacks using IPv4 mapped
972 	 * source.
973 	 */
974 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
975 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
976 		IP6STAT_INC(ip6s_badscope);
977 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
978 		goto bad;
979 	}
980 
981 	/*
982 	 * Tell launch routine the next header
983 	 */
984 	IP6STAT_INC(ip6s_delivered);
985 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
986 	nest = 0;
987 
988 	while (nxt != IPPROTO_DONE) {
989 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
990 			IP6STAT_INC(ip6s_toomanyhdr);
991 			goto bad;
992 		}
993 
994 		/*
995 		 * protection against faulty packet - there should be
996 		 * more sanity checks in header chain processing.
997 		 */
998 		if (m->m_pkthdr.len < off) {
999 			IP6STAT_INC(ip6s_tooshort);
1000 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
1001 			goto bad;
1002 		}
1003 
1004 #ifdef IPSEC
1005 		/*
1006 		 * enforce IPsec policy checking if we are seeing last header.
1007 		 * note that we do not visit this with protocols with pcb layer
1008 		 * code - like udp/tcp/raw ip.
1009 		 */
1010 		if (ip6_ipsec_input(m, nxt))
1011 			goto bad;
1012 #endif /* IPSEC */
1013 
1014 		/*
1015 		 * Use mbuf flags to propagate Router Alert option to
1016 		 * ICMPv6 layer, as hop-by-hop options have been stripped.
1017 		 */
1018 		if (nxt == IPPROTO_ICMPV6 && rtalert != ~0)
1019 			m->m_flags |= M_RTALERT_MLD;
1020 
1021 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
1022 	}
1023 	goto out;
1024 bad:
1025 	m_freem(m);
1026 out:
1027 	if (rin6.ro_rt)
1028 		RTFREE(rin6.ro_rt);
1029 }
1030 
1031 /*
1032  * set/grab in6_ifaddr correspond to IPv6 destination address.
1033  * XXX backward compatibility wrapper
1034  *
1035  * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag,
1036  * and then bump it when the tag is copied, and release it when the tag is
1037  * freed.  Unfortunately, m_tags don't support deep copies (yet), so instead
1038  * we just bump the ia refcount when we receive it.  This should be fixed.
1039  */
1040 static struct ip6aux *
1041 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
1042 {
1043 	struct ip6aux *ip6a;
1044 
1045 	ip6a = ip6_addaux(m);
1046 	if (ip6a)
1047 		ip6a->ip6a_dstia6 = ia6;
1048 	return ip6a;	/* NULL if failed to set */
1049 }
1050 
1051 struct in6_ifaddr *
1052 ip6_getdstifaddr(struct mbuf *m)
1053 {
1054 	struct ip6aux *ip6a;
1055 	struct in6_ifaddr *ia;
1056 
1057 	ip6a = ip6_findaux(m);
1058 	if (ip6a) {
1059 		ia = ip6a->ip6a_dstia6;
1060 		ifa_ref(&ia->ia_ifa);
1061 		return ia;
1062 	} else
1063 		return NULL;
1064 }
1065 
1066 /*
1067  * Hop-by-Hop options header processing. If a valid jumbo payload option is
1068  * included, the real payload length will be stored in plenp.
1069  *
1070  * rtalertp - XXX: should be stored more smart way
1071  */
1072 static int
1073 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
1074     struct mbuf **mp, int *offp)
1075 {
1076 	struct mbuf *m = *mp;
1077 	int off = *offp, hbhlen;
1078 	struct ip6_hbh *hbh;
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 		IP6STAT_INC(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 		IP6STAT_INC(ip6s_tooshort);
1100 		return -1;
1101 	}
1102 #endif
1103 	off += hbhlen;
1104 	hbhlen -= sizeof(struct ip6_hbh);
1105 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1106 				hbhlen, rtalertp, plenp) < 0)
1107 		return (-1);
1108 
1109 	*offp = off;
1110 	*mp = m;
1111 	return (0);
1112 }
1113 
1114 /*
1115  * Search header for all Hop-by-hop options and process each option.
1116  * This function is separate from ip6_hopopts_input() in order to
1117  * handle a case where the sending node itself process its hop-by-hop
1118  * options header. In such a case, the function is called from ip6_output().
1119  *
1120  * The function assumes that hbh header is located right after the IPv6 header
1121  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1122  * opthead + hbhlen is located in contiguous memory region.
1123  */
1124 int
1125 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1126     u_int32_t *rtalertp, u_int32_t *plenp)
1127 {
1128 	struct ip6_hdr *ip6;
1129 	int optlen = 0;
1130 	u_int8_t *opt = opthead;
1131 	u_int16_t rtalert_val;
1132 	u_int32_t jumboplen;
1133 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1134 
1135 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1136 		switch (*opt) {
1137 		case IP6OPT_PAD1:
1138 			optlen = 1;
1139 			break;
1140 		case IP6OPT_PADN:
1141 			if (hbhlen < IP6OPT_MINLEN) {
1142 				IP6STAT_INC(ip6s_toosmall);
1143 				goto bad;
1144 			}
1145 			optlen = *(opt + 1) + 2;
1146 			break;
1147 		case IP6OPT_ROUTER_ALERT:
1148 			/* XXX may need check for alignment */
1149 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1150 				IP6STAT_INC(ip6s_toosmall);
1151 				goto bad;
1152 			}
1153 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1154 				/* XXX stat */
1155 				icmp6_error(m, ICMP6_PARAM_PROB,
1156 				    ICMP6_PARAMPROB_HEADER,
1157 				    erroff + opt + 1 - opthead);
1158 				return (-1);
1159 			}
1160 			optlen = IP6OPT_RTALERT_LEN;
1161 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1162 			*rtalertp = ntohs(rtalert_val);
1163 			break;
1164 		case IP6OPT_JUMBO:
1165 			/* XXX may need check for alignment */
1166 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1167 				IP6STAT_INC(ip6s_toosmall);
1168 				goto bad;
1169 			}
1170 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1171 				/* XXX stat */
1172 				icmp6_error(m, ICMP6_PARAM_PROB,
1173 				    ICMP6_PARAMPROB_HEADER,
1174 				    erroff + opt + 1 - opthead);
1175 				return (-1);
1176 			}
1177 			optlen = IP6OPT_JUMBO_LEN;
1178 
1179 			/*
1180 			 * IPv6 packets that have non 0 payload length
1181 			 * must not contain a jumbo payload option.
1182 			 */
1183 			ip6 = mtod(m, struct ip6_hdr *);
1184 			if (ip6->ip6_plen) {
1185 				IP6STAT_INC(ip6s_badoptions);
1186 				icmp6_error(m, ICMP6_PARAM_PROB,
1187 				    ICMP6_PARAMPROB_HEADER,
1188 				    erroff + opt - opthead);
1189 				return (-1);
1190 			}
1191 
1192 			/*
1193 			 * We may see jumbolen in unaligned location, so
1194 			 * we'd need to perform bcopy().
1195 			 */
1196 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1197 			jumboplen = (u_int32_t)htonl(jumboplen);
1198 
1199 #if 1
1200 			/*
1201 			 * if there are multiple jumbo payload options,
1202 			 * *plenp will be non-zero and the packet will be
1203 			 * rejected.
1204 			 * the behavior may need some debate in ipngwg -
1205 			 * multiple options does not make sense, however,
1206 			 * there's no explicit mention in specification.
1207 			 */
1208 			if (*plenp != 0) {
1209 				IP6STAT_INC(ip6s_badoptions);
1210 				icmp6_error(m, ICMP6_PARAM_PROB,
1211 				    ICMP6_PARAMPROB_HEADER,
1212 				    erroff + opt + 2 - opthead);
1213 				return (-1);
1214 			}
1215 #endif
1216 
1217 			/*
1218 			 * jumbo payload length must be larger than 65535.
1219 			 */
1220 			if (jumboplen <= IPV6_MAXPACKET) {
1221 				IP6STAT_INC(ip6s_badoptions);
1222 				icmp6_error(m, ICMP6_PARAM_PROB,
1223 				    ICMP6_PARAMPROB_HEADER,
1224 				    erroff + opt + 2 - opthead);
1225 				return (-1);
1226 			}
1227 			*plenp = jumboplen;
1228 
1229 			break;
1230 		default:		/* unknown option */
1231 			if (hbhlen < IP6OPT_MINLEN) {
1232 				IP6STAT_INC(ip6s_toosmall);
1233 				goto bad;
1234 			}
1235 			optlen = ip6_unknown_opt(opt, m,
1236 			    erroff + opt - opthead);
1237 			if (optlen == -1)
1238 				return (-1);
1239 			optlen += 2;
1240 			break;
1241 		}
1242 	}
1243 
1244 	return (0);
1245 
1246   bad:
1247 	m_freem(m);
1248 	return (-1);
1249 }
1250 
1251 /*
1252  * Unknown option processing.
1253  * The third argument `off' is the offset from the IPv6 header to the option,
1254  * which is necessary if the IPv6 header the and option header and IPv6 header
1255  * is not contiguous in order to return an ICMPv6 error.
1256  */
1257 int
1258 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1259 {
1260 	struct ip6_hdr *ip6;
1261 
1262 	switch (IP6OPT_TYPE(*optp)) {
1263 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1264 		return ((int)*(optp + 1));
1265 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1266 		m_freem(m);
1267 		return (-1);
1268 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1269 		IP6STAT_INC(ip6s_badoptions);
1270 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1271 		return (-1);
1272 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1273 		IP6STAT_INC(ip6s_badoptions);
1274 		ip6 = mtod(m, struct ip6_hdr *);
1275 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1276 		    (m->m_flags & (M_BCAST|M_MCAST)))
1277 			m_freem(m);
1278 		else
1279 			icmp6_error(m, ICMP6_PARAM_PROB,
1280 				    ICMP6_PARAMPROB_OPTION, off);
1281 		return (-1);
1282 	}
1283 
1284 	m_freem(m);		/* XXX: NOTREACHED */
1285 	return (-1);
1286 }
1287 
1288 /*
1289  * Create the "control" list for this pcb.
1290  * These functions will not modify mbuf chain at all.
1291  *
1292  * With KAME mbuf chain restriction:
1293  * The routine will be called from upper layer handlers like tcp6_input().
1294  * Thus the routine assumes that the caller (tcp6_input) have already
1295  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1296  * very first mbuf on the mbuf chain.
1297  *
1298  * ip6_savecontrol_v4 will handle those options that are possible to be
1299  * set on a v4-mapped socket.
1300  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1301  * options and handle the v6-only ones itself.
1302  */
1303 struct mbuf **
1304 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1305     int *v4only)
1306 {
1307 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1308 
1309 #ifdef SO_TIMESTAMP
1310 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1311 		struct timeval tv;
1312 
1313 		microtime(&tv);
1314 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1315 		    SCM_TIMESTAMP, SOL_SOCKET);
1316 		if (*mp)
1317 			mp = &(*mp)->m_next;
1318 	}
1319 #endif
1320 
1321 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1322 	/* RFC 2292 sec. 5 */
1323 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1324 		struct in6_pktinfo pi6;
1325 
1326 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1327 #ifdef INET
1328 			struct ip *ip;
1329 
1330 			ip = mtod(m, struct ip *);
1331 			pi6.ipi6_addr.s6_addr32[0] = 0;
1332 			pi6.ipi6_addr.s6_addr32[1] = 0;
1333 			pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1334 			pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1335 #else
1336 			/* We won't hit this code */
1337 			bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1338 #endif
1339 		} else {
1340 			bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1341 			in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1342 		}
1343 		pi6.ipi6_ifindex =
1344 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1345 
1346 		*mp = sbcreatecontrol((caddr_t) &pi6,
1347 		    sizeof(struct in6_pktinfo),
1348 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1349 		if (*mp)
1350 			mp = &(*mp)->m_next;
1351 	}
1352 
1353 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1354 		int hlim;
1355 
1356 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1357 #ifdef INET
1358 			struct ip *ip;
1359 
1360 			ip = mtod(m, struct ip *);
1361 			hlim = ip->ip_ttl;
1362 #else
1363 			/* We won't hit this code */
1364 			hlim = 0;
1365 #endif
1366 		} else {
1367 			hlim = ip6->ip6_hlim & 0xff;
1368 		}
1369 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1370 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1371 		    IPPROTO_IPV6);
1372 		if (*mp)
1373 			mp = &(*mp)->m_next;
1374 	}
1375 
1376 	if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1377 		int tclass;
1378 
1379 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1380 #ifdef INET
1381 			struct ip *ip;
1382 
1383 			ip = mtod(m, struct ip *);
1384 			tclass = ip->ip_tos;
1385 #else
1386 			/* We won't hit this code */
1387 			tclass = 0;
1388 #endif
1389 		} else {
1390 			u_int32_t flowinfo;
1391 
1392 			flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1393 			flowinfo >>= 20;
1394 			tclass = flowinfo & 0xff;
1395 		}
1396 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int),
1397 		    IPV6_TCLASS, IPPROTO_IPV6);
1398 		if (*mp)
1399 			mp = &(*mp)->m_next;
1400 	}
1401 
1402 	if (v4only != NULL) {
1403 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1404 			*v4only = 1;
1405 		} else {
1406 			*v4only = 0;
1407 		}
1408 	}
1409 
1410 	return (mp);
1411 }
1412 
1413 void
1414 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1415 {
1416 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1417 	int v4only = 0;
1418 
1419 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1420 	if (v4only)
1421 		return;
1422 
1423 	/*
1424 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1425 	 * privilege for the option (see ip6_ctloutput), but it might be too
1426 	 * strict, since there might be some hop-by-hop options which can be
1427 	 * returned to normal user.
1428 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1429 	 */
1430 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1431 		/*
1432 		 * Check if a hop-by-hop options header is contatined in the
1433 		 * received packet, and if so, store the options as ancillary
1434 		 * data. Note that a hop-by-hop options header must be
1435 		 * just after the IPv6 header, which is assured through the
1436 		 * IPv6 input processing.
1437 		 */
1438 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1439 			struct ip6_hbh *hbh;
1440 			int hbhlen = 0;
1441 #ifdef PULLDOWN_TEST
1442 			struct mbuf *ext;
1443 #endif
1444 
1445 #ifndef PULLDOWN_TEST
1446 			hbh = (struct ip6_hbh *)(ip6 + 1);
1447 			hbhlen = (hbh->ip6h_len + 1) << 3;
1448 #else
1449 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1450 			    ip6->ip6_nxt);
1451 			if (ext == NULL) {
1452 				IP6STAT_INC(ip6s_tooshort);
1453 				return;
1454 			}
1455 			hbh = mtod(ext, struct ip6_hbh *);
1456 			hbhlen = (hbh->ip6h_len + 1) << 3;
1457 			if (hbhlen != ext->m_len) {
1458 				m_freem(ext);
1459 				IP6STAT_INC(ip6s_tooshort);
1460 				return;
1461 			}
1462 #endif
1463 
1464 			/*
1465 			 * XXX: We copy the whole header even if a
1466 			 * jumbo payload option is included, the option which
1467 			 * is to be removed before returning according to
1468 			 * RFC2292.
1469 			 * Note: this constraint is removed in RFC3542
1470 			 */
1471 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1472 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1473 			    IPPROTO_IPV6);
1474 			if (*mp)
1475 				mp = &(*mp)->m_next;
1476 #ifdef PULLDOWN_TEST
1477 			m_freem(ext);
1478 #endif
1479 		}
1480 	}
1481 
1482 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1483 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1484 
1485 		/*
1486 		 * Search for destination options headers or routing
1487 		 * header(s) through the header chain, and stores each
1488 		 * header as ancillary data.
1489 		 * Note that the order of the headers remains in
1490 		 * the chain of ancillary data.
1491 		 */
1492 		while (1) {	/* is explicit loop prevention necessary? */
1493 			struct ip6_ext *ip6e = NULL;
1494 			int elen;
1495 #ifdef PULLDOWN_TEST
1496 			struct mbuf *ext = NULL;
1497 #endif
1498 
1499 			/*
1500 			 * if it is not an extension header, don't try to
1501 			 * pull it from the chain.
1502 			 */
1503 			switch (nxt) {
1504 			case IPPROTO_DSTOPTS:
1505 			case IPPROTO_ROUTING:
1506 			case IPPROTO_HOPOPTS:
1507 			case IPPROTO_AH: /* is it possible? */
1508 				break;
1509 			default:
1510 				goto loopend;
1511 			}
1512 
1513 #ifndef PULLDOWN_TEST
1514 			if (off + sizeof(*ip6e) > m->m_len)
1515 				goto loopend;
1516 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1517 			if (nxt == IPPROTO_AH)
1518 				elen = (ip6e->ip6e_len + 2) << 2;
1519 			else
1520 				elen = (ip6e->ip6e_len + 1) << 3;
1521 			if (off + elen > m->m_len)
1522 				goto loopend;
1523 #else
1524 			ext = ip6_pullexthdr(m, off, nxt);
1525 			if (ext == NULL) {
1526 				IP6STAT_INC(ip6s_tooshort);
1527 				return;
1528 			}
1529 			ip6e = mtod(ext, struct ip6_ext *);
1530 			if (nxt == IPPROTO_AH)
1531 				elen = (ip6e->ip6e_len + 2) << 2;
1532 			else
1533 				elen = (ip6e->ip6e_len + 1) << 3;
1534 			if (elen != ext->m_len) {
1535 				m_freem(ext);
1536 				IP6STAT_INC(ip6s_tooshort);
1537 				return;
1538 			}
1539 #endif
1540 
1541 			switch (nxt) {
1542 			case IPPROTO_DSTOPTS:
1543 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1544 					break;
1545 
1546 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1547 				    IS2292(in6p,
1548 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1549 				    IPPROTO_IPV6);
1550 				if (*mp)
1551 					mp = &(*mp)->m_next;
1552 				break;
1553 			case IPPROTO_ROUTING:
1554 				if (!(in6p->inp_flags & IN6P_RTHDR))
1555 					break;
1556 
1557 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1558 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1559 				    IPPROTO_IPV6);
1560 				if (*mp)
1561 					mp = &(*mp)->m_next;
1562 				break;
1563 			case IPPROTO_HOPOPTS:
1564 			case IPPROTO_AH: /* is it possible? */
1565 				break;
1566 
1567 			default:
1568 				/*
1569 				 * other cases have been filtered in the above.
1570 				 * none will visit this case.  here we supply
1571 				 * the code just in case (nxt overwritten or
1572 				 * other cases).
1573 				 */
1574 #ifdef PULLDOWN_TEST
1575 				m_freem(ext);
1576 #endif
1577 				goto loopend;
1578 
1579 			}
1580 
1581 			/* proceed with the next header. */
1582 			off += elen;
1583 			nxt = ip6e->ip6e_nxt;
1584 			ip6e = NULL;
1585 #ifdef PULLDOWN_TEST
1586 			m_freem(ext);
1587 			ext = NULL;
1588 #endif
1589 		}
1590 	  loopend:
1591 		;
1592 	}
1593 }
1594 #undef IS2292
1595 
1596 void
1597 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1598 {
1599 	struct socket *so;
1600 	struct mbuf *m_mtu;
1601 	struct ip6_mtuinfo mtuctl;
1602 
1603 	so =  in6p->inp_socket;
1604 
1605 	if (mtu == NULL)
1606 		return;
1607 
1608 #ifdef DIAGNOSTIC
1609 	if (so == NULL)		/* I believe this is impossible */
1610 		panic("ip6_notify_pmtu: socket is NULL");
1611 #endif
1612 
1613 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1614 	mtuctl.ip6m_mtu = *mtu;
1615 	mtuctl.ip6m_addr = *dst;
1616 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1617 		return;
1618 
1619 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1620 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1621 		return;
1622 
1623 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1624 	    == 0) {
1625 		m_freem(m_mtu);
1626 		/* XXX: should count statistics */
1627 	} else
1628 		sorwakeup(so);
1629 
1630 	return;
1631 }
1632 
1633 #ifdef PULLDOWN_TEST
1634 /*
1635  * pull single extension header from mbuf chain.  returns single mbuf that
1636  * contains the result, or NULL on error.
1637  */
1638 static struct mbuf *
1639 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1640 {
1641 	struct ip6_ext ip6e;
1642 	size_t elen;
1643 	struct mbuf *n;
1644 
1645 #ifdef DIAGNOSTIC
1646 	switch (nxt) {
1647 	case IPPROTO_DSTOPTS:
1648 	case IPPROTO_ROUTING:
1649 	case IPPROTO_HOPOPTS:
1650 	case IPPROTO_AH: /* is it possible? */
1651 		break;
1652 	default:
1653 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1654 	}
1655 #endif
1656 
1657 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1658 	if (nxt == IPPROTO_AH)
1659 		elen = (ip6e.ip6e_len + 2) << 2;
1660 	else
1661 		elen = (ip6e.ip6e_len + 1) << 3;
1662 
1663 	if (elen > MLEN)
1664 		n = m_getcl(M_NOWAIT, MT_DATA, 0);
1665 	else
1666 		n = m_get(M_NOWAIT, MT_DATA);
1667 	if (n == NULL)
1668 		return NULL;
1669 
1670 	m_copydata(m, off, elen, mtod(n, caddr_t));
1671 	n->m_len = elen;
1672 	return n;
1673 }
1674 #endif
1675 
1676 /*
1677  * Get pointer to the previous header followed by the header
1678  * currently processed.
1679  * XXX: This function supposes that
1680  *	M includes all headers,
1681  *	the next header field and the header length field of each header
1682  *	are valid, and
1683  *	the sum of each header length equals to OFF.
1684  * Because of these assumptions, this function must be called very
1685  * carefully. Moreover, it will not be used in the near future when
1686  * we develop `neater' mechanism to process extension headers.
1687  */
1688 char *
1689 ip6_get_prevhdr(struct mbuf *m, int off)
1690 {
1691 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1692 
1693 	if (off == sizeof(struct ip6_hdr))
1694 		return (&ip6->ip6_nxt);
1695 	else {
1696 		int len, nxt;
1697 		struct ip6_ext *ip6e = NULL;
1698 
1699 		nxt = ip6->ip6_nxt;
1700 		len = sizeof(struct ip6_hdr);
1701 		while (len < off) {
1702 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1703 
1704 			switch (nxt) {
1705 			case IPPROTO_FRAGMENT:
1706 				len += sizeof(struct ip6_frag);
1707 				break;
1708 			case IPPROTO_AH:
1709 				len += (ip6e->ip6e_len + 2) << 2;
1710 				break;
1711 			default:
1712 				len += (ip6e->ip6e_len + 1) << 3;
1713 				break;
1714 			}
1715 			nxt = ip6e->ip6e_nxt;
1716 		}
1717 		if (ip6e)
1718 			return (&ip6e->ip6e_nxt);
1719 		else
1720 			return NULL;
1721 	}
1722 }
1723 
1724 /*
1725  * get next header offset.  m will be retained.
1726  */
1727 int
1728 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1729 {
1730 	struct ip6_hdr ip6;
1731 	struct ip6_ext ip6e;
1732 	struct ip6_frag fh;
1733 
1734 	/* just in case */
1735 	if (m == NULL)
1736 		panic("ip6_nexthdr: m == NULL");
1737 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1738 		return -1;
1739 
1740 	switch (proto) {
1741 	case IPPROTO_IPV6:
1742 		if (m->m_pkthdr.len < off + sizeof(ip6))
1743 			return -1;
1744 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1745 		if (nxtp)
1746 			*nxtp = ip6.ip6_nxt;
1747 		off += sizeof(ip6);
1748 		return off;
1749 
1750 	case IPPROTO_FRAGMENT:
1751 		/*
1752 		 * terminate parsing if it is not the first fragment,
1753 		 * it does not make sense to parse through it.
1754 		 */
1755 		if (m->m_pkthdr.len < off + sizeof(fh))
1756 			return -1;
1757 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1758 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1759 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1760 			return -1;
1761 		if (nxtp)
1762 			*nxtp = fh.ip6f_nxt;
1763 		off += sizeof(struct ip6_frag);
1764 		return off;
1765 
1766 	case IPPROTO_AH:
1767 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1768 			return -1;
1769 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1770 		if (nxtp)
1771 			*nxtp = ip6e.ip6e_nxt;
1772 		off += (ip6e.ip6e_len + 2) << 2;
1773 		return off;
1774 
1775 	case IPPROTO_HOPOPTS:
1776 	case IPPROTO_ROUTING:
1777 	case IPPROTO_DSTOPTS:
1778 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1779 			return -1;
1780 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1781 		if (nxtp)
1782 			*nxtp = ip6e.ip6e_nxt;
1783 		off += (ip6e.ip6e_len + 1) << 3;
1784 		return off;
1785 
1786 	case IPPROTO_NONE:
1787 	case IPPROTO_ESP:
1788 	case IPPROTO_IPCOMP:
1789 		/* give up */
1790 		return -1;
1791 
1792 	default:
1793 		return -1;
1794 	}
1795 
1796 	return -1;
1797 }
1798 
1799 /*
1800  * get offset for the last header in the chain.  m will be kept untainted.
1801  */
1802 int
1803 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1804 {
1805 	int newoff;
1806 	int nxt;
1807 
1808 	if (!nxtp) {
1809 		nxt = -1;
1810 		nxtp = &nxt;
1811 	}
1812 	while (1) {
1813 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1814 		if (newoff < 0)
1815 			return off;
1816 		else if (newoff < off)
1817 			return -1;	/* invalid */
1818 		else if (newoff == off)
1819 			return newoff;
1820 
1821 		off = newoff;
1822 		proto = *nxtp;
1823 	}
1824 }
1825 
1826 static struct ip6aux *
1827 ip6_addaux(struct mbuf *m)
1828 {
1829 	struct m_tag *mtag;
1830 
1831 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1832 	if (!mtag) {
1833 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1834 		    M_NOWAIT);
1835 		if (mtag) {
1836 			m_tag_prepend(m, mtag);
1837 			bzero(mtag + 1, sizeof(struct ip6aux));
1838 		}
1839 	}
1840 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1841 }
1842 
1843 static struct ip6aux *
1844 ip6_findaux(struct mbuf *m)
1845 {
1846 	struct m_tag *mtag;
1847 
1848 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1849 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1850 }
1851 
1852 static void
1853 ip6_delaux(struct mbuf *m)
1854 {
1855 	struct m_tag *mtag;
1856 
1857 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1858 	if (mtag)
1859 		m_tag_delete(m, mtag);
1860 }
1861 
1862 /*
1863  * System control for IP6
1864  */
1865 
1866 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1867 	0,		0,		0,		0,
1868 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1869 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1870 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1871 	0,		0,		0,		0,
1872 	ENOPROTOOPT
1873 };
1874