xref: /freebsd/sys/netinet6/ip6_input.c (revision d184218c18d067f8fd47203f54ab02a7b2ed9b11)
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 		IP6STAT_INC(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 		IP6STAT_INC(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 			IP6STAT_INC(ip6s_mext2m);
466 		else
467 			IP6STAT_INC(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 				IP6STAT_INC(ip6s_m2m[V_loif->if_index]);
473 			} else if (m->m_pkthdr.rcvif->if_index < M2MMAX)
474 				IP6STAT_INC(
475 				    ip6s_m2m[m->m_pkthdr.rcvif->if_index]);
476 			else
477 				IP6STAT_INC(ip6s_m2m[0]);
478 		} else
479 			IP6STAT_INC(ip6s_m1);
480 #undef M2MMAX
481 	}
482 
483 	/* drop the packet if IPv6 operation is disabled on the IF */
484 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
485 		m_freem(m);
486 		return;
487 	}
488 
489 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
490 	IP6STAT_INC(ip6s_total);
491 
492 #ifndef PULLDOWN_TEST
493 	/*
494 	 * L2 bridge code and some other code can return mbuf chain
495 	 * that does not conform to KAME requirement.  too bad.
496 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
497 	 */
498 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
499 		struct mbuf *n;
500 
501 		if (m->m_pkthdr.len > MHLEN)
502 			n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
503 		else
504 			n = m_gethdr(M_NOWAIT, MT_DATA);
505 		if (n == NULL) {
506 			m_freem(m);
507 			return;	/* ENOBUFS */
508 		}
509 
510 		m_move_pkthdr(n, m);
511 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
512 		n->m_len = n->m_pkthdr.len;
513 		m_freem(m);
514 		m = n;
515 	}
516 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
517 #endif
518 
519 	if (m->m_len < sizeof(struct ip6_hdr)) {
520 		struct ifnet *inifp;
521 		inifp = m->m_pkthdr.rcvif;
522 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
523 			IP6STAT_INC(ip6s_toosmall);
524 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
525 			return;
526 		}
527 	}
528 
529 	ip6 = mtod(m, struct ip6_hdr *);
530 
531 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
532 		IP6STAT_INC(ip6s_badvers);
533 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
534 		goto bad;
535 	}
536 
537 	IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]);
538 
539 	/*
540 	 * Check against address spoofing/corruption.
541 	 */
542 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
543 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
544 		/*
545 		 * XXX: "badscope" is not very suitable for a multicast source.
546 		 */
547 		IP6STAT_INC(ip6s_badscope);
548 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
549 		goto bad;
550 	}
551 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
552 	    !(m->m_flags & M_LOOP)) {
553 		/*
554 		 * In this case, the packet should come from the loopback
555 		 * interface.  However, we cannot just check the if_flags,
556 		 * because ip6_mloopback() passes the "actual" interface
557 		 * as the outgoing/incoming interface.
558 		 */
559 		IP6STAT_INC(ip6s_badscope);
560 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
561 		goto bad;
562 	}
563 
564 #ifdef ALTQ
565 	if (altq_input != NULL && (*altq_input)(m, AF_INET6) == 0) {
566 		/* packet is dropped by traffic conditioner */
567 		return;
568 	}
569 #endif
570 	/*
571 	 * The following check is not documented in specs.  A malicious
572 	 * party may be able to use IPv4 mapped addr to confuse tcp/udp stack
573 	 * and bypass security checks (act as if it was from 127.0.0.1 by using
574 	 * IPv6 src ::ffff:127.0.0.1).  Be cautious.
575 	 *
576 	 * This check chokes if we are in an SIIT cloud.  As none of BSDs
577 	 * support IPv4-less kernel compilation, we cannot support SIIT
578 	 * environment at all.  So, it makes more sense for us to reject any
579 	 * malicious packets for non-SIIT environment, than try to do a
580 	 * partial support for SIIT environment.
581 	 */
582 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
583 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
584 		IP6STAT_INC(ip6s_badscope);
585 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
586 		goto bad;
587 	}
588 #if 0
589 	/*
590 	 * Reject packets with IPv4 compatible addresses (auto tunnel).
591 	 *
592 	 * The code forbids auto tunnel relay case in RFC1933 (the check is
593 	 * stronger than RFC1933).  We may want to re-enable it if mech-xx
594 	 * is revised to forbid relaying case.
595 	 */
596 	if (IN6_IS_ADDR_V4COMPAT(&ip6->ip6_src) ||
597 	    IN6_IS_ADDR_V4COMPAT(&ip6->ip6_dst)) {
598 		IP6STAT_INC(ip6s_badscope);
599 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
600 		goto bad;
601 	}
602 #endif
603 #ifdef IPSEC
604 	/*
605 	 * Bypass packet filtering for packets previously handled by IPsec.
606 	 */
607 	if (ip6_ipsec_filtertunnel(m))
608 		goto passin;
609 #endif /* IPSEC */
610 
611 	/*
612 	 * Run through list of hooks for input packets.
613 	 *
614 	 * NB: Beware of the destination address changing
615 	 *     (e.g. by NAT rewriting).  When this happens,
616 	 *     tell ip6_forward to do the right thing.
617 	 */
618 	odst = ip6->ip6_dst;
619 
620 	/* Jump over all PFIL processing if hooks are not active. */
621 	if (!PFIL_HOOKED(&V_inet6_pfil_hook))
622 		goto passin;
623 
624 	if (pfil_run_hooks(&V_inet6_pfil_hook, &m,
625 	    m->m_pkthdr.rcvif, PFIL_IN, NULL))
626 		return;
627 	if (m == NULL)			/* consumed by filter */
628 		return;
629 	ip6 = mtod(m, struct ip6_hdr *);
630 	srcrt = !IN6_ARE_ADDR_EQUAL(&odst, &ip6->ip6_dst);
631 
632 	if (m->m_flags & M_FASTFWD_OURS) {
633 		m->m_flags &= ~M_FASTFWD_OURS;
634 		ours = 1;
635 		deliverifp = m->m_pkthdr.rcvif;
636 		goto hbhcheck;
637 	}
638 	if ((m->m_flags & M_IP6_NEXTHOP) &&
639 	    m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) {
640 		/*
641 		 * Directly ship the packet on.  This allows forwarding
642 		 * packets originally destined to us to some other directly
643 		 * connected host.
644 		 */
645 		ip6_forward(m, 1);
646 		goto out;
647 	}
648 
649 passin:
650 	/*
651 	 * Disambiguate address scope zones (if there is ambiguity).
652 	 * We first make sure that the original source or destination address
653 	 * is not in our internal form for scoped addresses.  Such addresses
654 	 * are not necessarily invalid spec-wise, but we cannot accept them due
655 	 * to the usage conflict.
656 	 * in6_setscope() then also checks and rejects the cases where src or
657 	 * dst are the loopback address and the receiving interface
658 	 * is not loopback.
659 	 */
660 	if (in6_clearscope(&ip6->ip6_src) || in6_clearscope(&ip6->ip6_dst)) {
661 		IP6STAT_INC(ip6s_badscope); /* XXX */
662 		goto bad;
663 	}
664 	if (in6_setscope(&ip6->ip6_src, m->m_pkthdr.rcvif, NULL) ||
665 	    in6_setscope(&ip6->ip6_dst, m->m_pkthdr.rcvif, NULL)) {
666 		IP6STAT_INC(ip6s_badscope);
667 		goto bad;
668 	}
669 
670 	/*
671 	 * Multicast check. Assume packet is for us to avoid
672 	 * prematurely taking locks.
673 	 */
674 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
675 		ours = 1;
676 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_mcast);
677 		deliverifp = m->m_pkthdr.rcvif;
678 		goto hbhcheck;
679 	}
680 
681 	/*
682 	 *  Unicast check
683 	 */
684 
685 	bzero(&dst6, sizeof(dst6));
686 	dst6.sin6_family = AF_INET6;
687 	dst6.sin6_len = sizeof(struct sockaddr_in6);
688 	dst6.sin6_addr = ip6->ip6_dst;
689 	ifp = m->m_pkthdr.rcvif;
690 	IF_AFDATA_RLOCK(ifp);
691 	lle = lla_lookup(LLTABLE6(ifp), 0,
692 	     (struct sockaddr *)&dst6);
693 	IF_AFDATA_RUNLOCK(ifp);
694 	if ((lle != NULL) && (lle->la_flags & LLE_IFADDR)) {
695 		struct ifaddr *ifa;
696 		struct in6_ifaddr *ia6;
697 		int bad;
698 
699 		bad = 1;
700 #define	sa_equal(a1, a2)						\
701 	(bcmp((a1), (a2), ((a1))->sin6_len) == 0)
702 		IF_ADDR_RLOCK(ifp);
703 		TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
704 			if (ifa->ifa_addr->sa_family != dst6.sin6_family)
705 				continue;
706 			if (sa_equal(&dst6, ifa->ifa_addr))
707 				break;
708 		}
709 		KASSERT(ifa != NULL, ("%s: ifa not found for lle %p",
710 		    __func__, lle));
711 #undef sa_equal
712 
713 		ia6 = (struct in6_ifaddr *)ifa;
714 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
715 			/* Count the packet in the ip address stats */
716 			ia6->ia_ifa.if_ipackets++;
717 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
718 
719 			/*
720 			 * record address information into m_tag.
721 			 */
722 			(void)ip6_setdstifaddr(m, ia6);
723 
724 			bad = 0;
725 		} else {
726 			char ip6bufs[INET6_ADDRSTRLEN];
727 			char ip6bufd[INET6_ADDRSTRLEN];
728 			/* address is not ready, so discard the packet. */
729 			nd6log((LOG_INFO,
730 			    "ip6_input: packet to an unready address %s->%s\n",
731 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
732 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
733 		}
734 		IF_ADDR_RUNLOCK(ifp);
735 		LLE_RUNLOCK(lle);
736 		if (bad)
737 			goto bad;
738 		else {
739 			ours = 1;
740 			deliverifp = ifp;
741 			goto hbhcheck;
742 		}
743 	}
744 	if (lle != NULL)
745 		LLE_RUNLOCK(lle);
746 
747 	dst = &rin6.ro_dst;
748 	dst->sin6_len = sizeof(struct sockaddr_in6);
749 	dst->sin6_family = AF_INET6;
750 	dst->sin6_addr = ip6->ip6_dst;
751 	rin6.ro_rt = in6_rtalloc1((struct sockaddr *)dst, 0, 0, M_GETFIB(m));
752 	if (rin6.ro_rt)
753 		RT_UNLOCK(rin6.ro_rt);
754 
755 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
756 
757 	/*
758 	 * Accept the packet if the forwarding interface to the destination
759 	 * according to the routing table is the loopback interface,
760 	 * unless the associated route has a gateway.
761 	 * Note that this approach causes to accept a packet if there is a
762 	 * route to the loopback interface for the destination of the packet.
763 	 * But we think it's even useful in some situations, e.g. when using
764 	 * a special daemon which wants to intercept the packet.
765 	 *
766 	 * XXX: some OSes automatically make a cloned route for the destination
767 	 * of an outgoing packet.  If the outgoing interface of the packet
768 	 * is a loopback one, the kernel would consider the packet to be
769 	 * accepted, even if we have no such address assinged on the interface.
770 	 * We check the cloned flag of the route entry to reject such cases,
771 	 * assuming that route entries for our own addresses are not made by
772 	 * cloning (it should be true because in6_addloop explicitly installs
773 	 * the host route).  However, we might have to do an explicit check
774 	 * while it would be less efficient.  Or, should we rather install a
775 	 * reject route for such a case?
776 	 */
777 	if (rin6.ro_rt &&
778 	    (rin6.ro_rt->rt_flags &
779 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
780 #ifdef RTF_WASCLONED
781 	    !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
782 #endif
783 #ifdef RTF_CLONED
784 	    !(rin6.ro_rt->rt_flags & RTF_CLONED) &&
785 #endif
786 #if 0
787 	    /*
788 	     * The check below is redundant since the comparison of
789 	     * the destination and the key of the rtentry has
790 	     * already done through looking up the routing table.
791 	     */
792 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
793 	    &rt6_key(rin6.ro_rt)->sin6_addr)
794 #endif
795 	    rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
796 		int free_ia6 = 0;
797 		struct in6_ifaddr *ia6;
798 
799 		/*
800 		 * found the loopback route to the interface address
801 		 */
802 		if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) {
803 			struct sockaddr_in6 dest6;
804 
805 			bzero(&dest6, sizeof(dest6));
806 			dest6.sin6_family = AF_INET6;
807 			dest6.sin6_len = sizeof(dest6);
808 			dest6.sin6_addr = ip6->ip6_dst;
809 			ia6 = (struct in6_ifaddr *)
810 			    ifa_ifwithaddr((struct sockaddr *)&dest6);
811 			if (ia6 == NULL)
812 				goto bad;
813 			free_ia6 = 1;
814 		}
815 		else
816 			ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
817 
818 		/*
819 		 * record address information into m_tag.
820 		 */
821 		(void)ip6_setdstifaddr(m, ia6);
822 
823 		/*
824 		 * packets to a tentative, duplicated, or somehow invalid
825 		 * address must not be accepted.
826 		 */
827 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
828 			/* this address is ready */
829 			ours = 1;
830 			deliverifp = ia6->ia_ifp;	/* correct? */
831 			/* Count the packet in the ip address stats */
832 			ia6->ia_ifa.if_ipackets++;
833 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
834 			if (ia6 != NULL && free_ia6 != 0)
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 (ia6 != NULL && free_ia6 != 0)
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 	u_int8_t *opt;
1080 
1081 	/* validation of the length of the header */
1082 #ifndef PULLDOWN_TEST
1083 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
1084 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1085 	hbhlen = (hbh->ip6h_len + 1) << 3;
1086 
1087 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
1088 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1089 #else
1090 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1091 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
1092 	if (hbh == NULL) {
1093 		IP6STAT_INC(ip6s_tooshort);
1094 		return -1;
1095 	}
1096 	hbhlen = (hbh->ip6h_len + 1) << 3;
1097 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
1098 		hbhlen);
1099 	if (hbh == NULL) {
1100 		IP6STAT_INC(ip6s_tooshort);
1101 		return -1;
1102 	}
1103 #endif
1104 	off += hbhlen;
1105 	hbhlen -= sizeof(struct ip6_hbh);
1106 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
1107 
1108 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1109 				hbhlen, rtalertp, plenp) < 0)
1110 		return (-1);
1111 
1112 	*offp = off;
1113 	*mp = m;
1114 	return (0);
1115 }
1116 
1117 /*
1118  * Search header for all Hop-by-hop options and process each option.
1119  * This function is separate from ip6_hopopts_input() in order to
1120  * handle a case where the sending node itself process its hop-by-hop
1121  * options header. In such a case, the function is called from ip6_output().
1122  *
1123  * The function assumes that hbh header is located right after the IPv6 header
1124  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1125  * opthead + hbhlen is located in contiguous memory region.
1126  */
1127 int
1128 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1129     u_int32_t *rtalertp, u_int32_t *plenp)
1130 {
1131 	struct ip6_hdr *ip6;
1132 	int optlen = 0;
1133 	u_int8_t *opt = opthead;
1134 	u_int16_t rtalert_val;
1135 	u_int32_t jumboplen;
1136 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1137 
1138 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1139 		switch (*opt) {
1140 		case IP6OPT_PAD1:
1141 			optlen = 1;
1142 			break;
1143 		case IP6OPT_PADN:
1144 			if (hbhlen < IP6OPT_MINLEN) {
1145 				IP6STAT_INC(ip6s_toosmall);
1146 				goto bad;
1147 			}
1148 			optlen = *(opt + 1) + 2;
1149 			break;
1150 		case IP6OPT_ROUTER_ALERT:
1151 			/* XXX may need check for alignment */
1152 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1153 				IP6STAT_INC(ip6s_toosmall);
1154 				goto bad;
1155 			}
1156 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1157 				/* XXX stat */
1158 				icmp6_error(m, ICMP6_PARAM_PROB,
1159 				    ICMP6_PARAMPROB_HEADER,
1160 				    erroff + opt + 1 - opthead);
1161 				return (-1);
1162 			}
1163 			optlen = IP6OPT_RTALERT_LEN;
1164 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1165 			*rtalertp = ntohs(rtalert_val);
1166 			break;
1167 		case IP6OPT_JUMBO:
1168 			/* XXX may need check for alignment */
1169 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1170 				IP6STAT_INC(ip6s_toosmall);
1171 				goto bad;
1172 			}
1173 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1174 				/* XXX stat */
1175 				icmp6_error(m, ICMP6_PARAM_PROB,
1176 				    ICMP6_PARAMPROB_HEADER,
1177 				    erroff + opt + 1 - opthead);
1178 				return (-1);
1179 			}
1180 			optlen = IP6OPT_JUMBO_LEN;
1181 
1182 			/*
1183 			 * IPv6 packets that have non 0 payload length
1184 			 * must not contain a jumbo payload option.
1185 			 */
1186 			ip6 = mtod(m, struct ip6_hdr *);
1187 			if (ip6->ip6_plen) {
1188 				IP6STAT_INC(ip6s_badoptions);
1189 				icmp6_error(m, ICMP6_PARAM_PROB,
1190 				    ICMP6_PARAMPROB_HEADER,
1191 				    erroff + opt - opthead);
1192 				return (-1);
1193 			}
1194 
1195 			/*
1196 			 * We may see jumbolen in unaligned location, so
1197 			 * we'd need to perform bcopy().
1198 			 */
1199 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1200 			jumboplen = (u_int32_t)htonl(jumboplen);
1201 
1202 #if 1
1203 			/*
1204 			 * if there are multiple jumbo payload options,
1205 			 * *plenp will be non-zero and the packet will be
1206 			 * rejected.
1207 			 * the behavior may need some debate in ipngwg -
1208 			 * multiple options does not make sense, however,
1209 			 * there's no explicit mention in specification.
1210 			 */
1211 			if (*plenp != 0) {
1212 				IP6STAT_INC(ip6s_badoptions);
1213 				icmp6_error(m, ICMP6_PARAM_PROB,
1214 				    ICMP6_PARAMPROB_HEADER,
1215 				    erroff + opt + 2 - opthead);
1216 				return (-1);
1217 			}
1218 #endif
1219 
1220 			/*
1221 			 * jumbo payload length must be larger than 65535.
1222 			 */
1223 			if (jumboplen <= IPV6_MAXPACKET) {
1224 				IP6STAT_INC(ip6s_badoptions);
1225 				icmp6_error(m, ICMP6_PARAM_PROB,
1226 				    ICMP6_PARAMPROB_HEADER,
1227 				    erroff + opt + 2 - opthead);
1228 				return (-1);
1229 			}
1230 			*plenp = jumboplen;
1231 
1232 			break;
1233 		default:		/* unknown option */
1234 			if (hbhlen < IP6OPT_MINLEN) {
1235 				IP6STAT_INC(ip6s_toosmall);
1236 				goto bad;
1237 			}
1238 			optlen = ip6_unknown_opt(opt, m,
1239 			    erroff + opt - opthead);
1240 			if (optlen == -1)
1241 				return (-1);
1242 			optlen += 2;
1243 			break;
1244 		}
1245 	}
1246 
1247 	return (0);
1248 
1249   bad:
1250 	m_freem(m);
1251 	return (-1);
1252 }
1253 
1254 /*
1255  * Unknown option processing.
1256  * The third argument `off' is the offset from the IPv6 header to the option,
1257  * which is necessary if the IPv6 header the and option header and IPv6 header
1258  * is not contiguous in order to return an ICMPv6 error.
1259  */
1260 int
1261 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1262 {
1263 	struct ip6_hdr *ip6;
1264 
1265 	switch (IP6OPT_TYPE(*optp)) {
1266 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1267 		return ((int)*(optp + 1));
1268 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1269 		m_freem(m);
1270 		return (-1);
1271 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1272 		IP6STAT_INC(ip6s_badoptions);
1273 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1274 		return (-1);
1275 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1276 		IP6STAT_INC(ip6s_badoptions);
1277 		ip6 = mtod(m, struct ip6_hdr *);
1278 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1279 		    (m->m_flags & (M_BCAST|M_MCAST)))
1280 			m_freem(m);
1281 		else
1282 			icmp6_error(m, ICMP6_PARAM_PROB,
1283 				    ICMP6_PARAMPROB_OPTION, off);
1284 		return (-1);
1285 	}
1286 
1287 	m_freem(m);		/* XXX: NOTREACHED */
1288 	return (-1);
1289 }
1290 
1291 /*
1292  * Create the "control" list for this pcb.
1293  * These functions will not modify mbuf chain at all.
1294  *
1295  * With KAME mbuf chain restriction:
1296  * The routine will be called from upper layer handlers like tcp6_input().
1297  * Thus the routine assumes that the caller (tcp6_input) have already
1298  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1299  * very first mbuf on the mbuf chain.
1300  *
1301  * ip6_savecontrol_v4 will handle those options that are possible to be
1302  * set on a v4-mapped socket.
1303  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1304  * options and handle the v6-only ones itself.
1305  */
1306 struct mbuf **
1307 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1308     int *v4only)
1309 {
1310 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1311 
1312 #ifdef SO_TIMESTAMP
1313 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1314 		struct timeval tv;
1315 
1316 		microtime(&tv);
1317 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1318 		    SCM_TIMESTAMP, SOL_SOCKET);
1319 		if (*mp)
1320 			mp = &(*mp)->m_next;
1321 	}
1322 #endif
1323 
1324 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1325 	/* RFC 2292 sec. 5 */
1326 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1327 		struct in6_pktinfo pi6;
1328 
1329 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1330 #ifdef INET
1331 			struct ip *ip;
1332 
1333 			ip = mtod(m, struct ip *);
1334 			pi6.ipi6_addr.s6_addr32[0] = 0;
1335 			pi6.ipi6_addr.s6_addr32[1] = 0;
1336 			pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1337 			pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1338 #else
1339 			/* We won't hit this code */
1340 			bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1341 #endif
1342 		} else {
1343 			bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1344 			in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1345 		}
1346 		pi6.ipi6_ifindex =
1347 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1348 
1349 		*mp = sbcreatecontrol((caddr_t) &pi6,
1350 		    sizeof(struct in6_pktinfo),
1351 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1352 		if (*mp)
1353 			mp = &(*mp)->m_next;
1354 	}
1355 
1356 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1357 		int hlim;
1358 
1359 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1360 #ifdef INET
1361 			struct ip *ip;
1362 
1363 			ip = mtod(m, struct ip *);
1364 			hlim = ip->ip_ttl;
1365 #else
1366 			/* We won't hit this code */
1367 			hlim = 0;
1368 #endif
1369 		} else {
1370 			hlim = ip6->ip6_hlim & 0xff;
1371 		}
1372 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1373 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1374 		    IPPROTO_IPV6);
1375 		if (*mp)
1376 			mp = &(*mp)->m_next;
1377 	}
1378 
1379 	if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1380 		int tclass;
1381 
1382 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1383 #ifdef INET
1384 			struct ip *ip;
1385 
1386 			ip = mtod(m, struct ip *);
1387 			tclass = ip->ip_tos;
1388 #else
1389 			/* We won't hit this code */
1390 			tclass = 0;
1391 #endif
1392 		} else {
1393 			u_int32_t flowinfo;
1394 
1395 			flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1396 			flowinfo >>= 20;
1397 			tclass = flowinfo & 0xff;
1398 		}
1399 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int),
1400 		    IPV6_TCLASS, IPPROTO_IPV6);
1401 		if (*mp)
1402 			mp = &(*mp)->m_next;
1403 	}
1404 
1405 	if (v4only != NULL) {
1406 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1407 			*v4only = 1;
1408 		} else {
1409 			*v4only = 0;
1410 		}
1411 	}
1412 
1413 	return (mp);
1414 }
1415 
1416 void
1417 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1418 {
1419 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1420 	int v4only = 0;
1421 
1422 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1423 	if (v4only)
1424 		return;
1425 
1426 	/*
1427 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1428 	 * privilege for the option (see ip6_ctloutput), but it might be too
1429 	 * strict, since there might be some hop-by-hop options which can be
1430 	 * returned to normal user.
1431 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1432 	 */
1433 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1434 		/*
1435 		 * Check if a hop-by-hop options header is contatined in the
1436 		 * received packet, and if so, store the options as ancillary
1437 		 * data. Note that a hop-by-hop options header must be
1438 		 * just after the IPv6 header, which is assured through the
1439 		 * IPv6 input processing.
1440 		 */
1441 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1442 			struct ip6_hbh *hbh;
1443 			int hbhlen = 0;
1444 #ifdef PULLDOWN_TEST
1445 			struct mbuf *ext;
1446 #endif
1447 
1448 #ifndef PULLDOWN_TEST
1449 			hbh = (struct ip6_hbh *)(ip6 + 1);
1450 			hbhlen = (hbh->ip6h_len + 1) << 3;
1451 #else
1452 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1453 			    ip6->ip6_nxt);
1454 			if (ext == NULL) {
1455 				IP6STAT_INC(ip6s_tooshort);
1456 				return;
1457 			}
1458 			hbh = mtod(ext, struct ip6_hbh *);
1459 			hbhlen = (hbh->ip6h_len + 1) << 3;
1460 			if (hbhlen != ext->m_len) {
1461 				m_freem(ext);
1462 				IP6STAT_INC(ip6s_tooshort);
1463 				return;
1464 			}
1465 #endif
1466 
1467 			/*
1468 			 * XXX: We copy the whole header even if a
1469 			 * jumbo payload option is included, the option which
1470 			 * is to be removed before returning according to
1471 			 * RFC2292.
1472 			 * Note: this constraint is removed in RFC3542
1473 			 */
1474 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1475 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1476 			    IPPROTO_IPV6);
1477 			if (*mp)
1478 				mp = &(*mp)->m_next;
1479 #ifdef PULLDOWN_TEST
1480 			m_freem(ext);
1481 #endif
1482 		}
1483 	}
1484 
1485 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1486 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1487 
1488 		/*
1489 		 * Search for destination options headers or routing
1490 		 * header(s) through the header chain, and stores each
1491 		 * header as ancillary data.
1492 		 * Note that the order of the headers remains in
1493 		 * the chain of ancillary data.
1494 		 */
1495 		while (1) {	/* is explicit loop prevention necessary? */
1496 			struct ip6_ext *ip6e = NULL;
1497 			int elen;
1498 #ifdef PULLDOWN_TEST
1499 			struct mbuf *ext = NULL;
1500 #endif
1501 
1502 			/*
1503 			 * if it is not an extension header, don't try to
1504 			 * pull it from the chain.
1505 			 */
1506 			switch (nxt) {
1507 			case IPPROTO_DSTOPTS:
1508 			case IPPROTO_ROUTING:
1509 			case IPPROTO_HOPOPTS:
1510 			case IPPROTO_AH: /* is it possible? */
1511 				break;
1512 			default:
1513 				goto loopend;
1514 			}
1515 
1516 #ifndef PULLDOWN_TEST
1517 			if (off + sizeof(*ip6e) > m->m_len)
1518 				goto loopend;
1519 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1520 			if (nxt == IPPROTO_AH)
1521 				elen = (ip6e->ip6e_len + 2) << 2;
1522 			else
1523 				elen = (ip6e->ip6e_len + 1) << 3;
1524 			if (off + elen > m->m_len)
1525 				goto loopend;
1526 #else
1527 			ext = ip6_pullexthdr(m, off, nxt);
1528 			if (ext == NULL) {
1529 				IP6STAT_INC(ip6s_tooshort);
1530 				return;
1531 			}
1532 			ip6e = mtod(ext, struct ip6_ext *);
1533 			if (nxt == IPPROTO_AH)
1534 				elen = (ip6e->ip6e_len + 2) << 2;
1535 			else
1536 				elen = (ip6e->ip6e_len + 1) << 3;
1537 			if (elen != ext->m_len) {
1538 				m_freem(ext);
1539 				IP6STAT_INC(ip6s_tooshort);
1540 				return;
1541 			}
1542 #endif
1543 
1544 			switch (nxt) {
1545 			case IPPROTO_DSTOPTS:
1546 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1547 					break;
1548 
1549 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1550 				    IS2292(in6p,
1551 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1552 				    IPPROTO_IPV6);
1553 				if (*mp)
1554 					mp = &(*mp)->m_next;
1555 				break;
1556 			case IPPROTO_ROUTING:
1557 				if (!(in6p->inp_flags & IN6P_RTHDR))
1558 					break;
1559 
1560 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1561 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1562 				    IPPROTO_IPV6);
1563 				if (*mp)
1564 					mp = &(*mp)->m_next;
1565 				break;
1566 			case IPPROTO_HOPOPTS:
1567 			case IPPROTO_AH: /* is it possible? */
1568 				break;
1569 
1570 			default:
1571 				/*
1572 				 * other cases have been filtered in the above.
1573 				 * none will visit this case.  here we supply
1574 				 * the code just in case (nxt overwritten or
1575 				 * other cases).
1576 				 */
1577 #ifdef PULLDOWN_TEST
1578 				m_freem(ext);
1579 #endif
1580 				goto loopend;
1581 
1582 			}
1583 
1584 			/* proceed with the next header. */
1585 			off += elen;
1586 			nxt = ip6e->ip6e_nxt;
1587 			ip6e = NULL;
1588 #ifdef PULLDOWN_TEST
1589 			m_freem(ext);
1590 			ext = NULL;
1591 #endif
1592 		}
1593 	  loopend:
1594 		;
1595 	}
1596 }
1597 #undef IS2292
1598 
1599 void
1600 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1601 {
1602 	struct socket *so;
1603 	struct mbuf *m_mtu;
1604 	struct ip6_mtuinfo mtuctl;
1605 
1606 	so =  in6p->inp_socket;
1607 
1608 	if (mtu == NULL)
1609 		return;
1610 
1611 #ifdef DIAGNOSTIC
1612 	if (so == NULL)		/* I believe this is impossible */
1613 		panic("ip6_notify_pmtu: socket is NULL");
1614 #endif
1615 
1616 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1617 	mtuctl.ip6m_mtu = *mtu;
1618 	mtuctl.ip6m_addr = *dst;
1619 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1620 		return;
1621 
1622 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1623 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1624 		return;
1625 
1626 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1627 	    == 0) {
1628 		m_freem(m_mtu);
1629 		/* XXX: should count statistics */
1630 	} else
1631 		sorwakeup(so);
1632 
1633 	return;
1634 }
1635 
1636 #ifdef PULLDOWN_TEST
1637 /*
1638  * pull single extension header from mbuf chain.  returns single mbuf that
1639  * contains the result, or NULL on error.
1640  */
1641 static struct mbuf *
1642 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1643 {
1644 	struct ip6_ext ip6e;
1645 	size_t elen;
1646 	struct mbuf *n;
1647 
1648 #ifdef DIAGNOSTIC
1649 	switch (nxt) {
1650 	case IPPROTO_DSTOPTS:
1651 	case IPPROTO_ROUTING:
1652 	case IPPROTO_HOPOPTS:
1653 	case IPPROTO_AH: /* is it possible? */
1654 		break;
1655 	default:
1656 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1657 	}
1658 #endif
1659 
1660 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1661 	if (nxt == IPPROTO_AH)
1662 		elen = (ip6e.ip6e_len + 2) << 2;
1663 	else
1664 		elen = (ip6e.ip6e_len + 1) << 3;
1665 
1666 	if (elen > MLEN)
1667 		n = m_getcl(M_NOWAIT, MT_DATA, 0);
1668 	else
1669 		n = m_get(M_NOWAIT, MT_DATA);
1670 	if (n == NULL)
1671 		return NULL;
1672 
1673 	m_copydata(m, off, elen, mtod(n, caddr_t));
1674 	n->m_len = elen;
1675 	return n;
1676 }
1677 #endif
1678 
1679 /*
1680  * Get pointer to the previous header followed by the header
1681  * currently processed.
1682  * XXX: This function supposes that
1683  *	M includes all headers,
1684  *	the next header field and the header length field of each header
1685  *	are valid, and
1686  *	the sum of each header length equals to OFF.
1687  * Because of these assumptions, this function must be called very
1688  * carefully. Moreover, it will not be used in the near future when
1689  * we develop `neater' mechanism to process extension headers.
1690  */
1691 char *
1692 ip6_get_prevhdr(struct mbuf *m, int off)
1693 {
1694 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1695 
1696 	if (off == sizeof(struct ip6_hdr))
1697 		return (&ip6->ip6_nxt);
1698 	else {
1699 		int len, nxt;
1700 		struct ip6_ext *ip6e = NULL;
1701 
1702 		nxt = ip6->ip6_nxt;
1703 		len = sizeof(struct ip6_hdr);
1704 		while (len < off) {
1705 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1706 
1707 			switch (nxt) {
1708 			case IPPROTO_FRAGMENT:
1709 				len += sizeof(struct ip6_frag);
1710 				break;
1711 			case IPPROTO_AH:
1712 				len += (ip6e->ip6e_len + 2) << 2;
1713 				break;
1714 			default:
1715 				len += (ip6e->ip6e_len + 1) << 3;
1716 				break;
1717 			}
1718 			nxt = ip6e->ip6e_nxt;
1719 		}
1720 		if (ip6e)
1721 			return (&ip6e->ip6e_nxt);
1722 		else
1723 			return NULL;
1724 	}
1725 }
1726 
1727 /*
1728  * get next header offset.  m will be retained.
1729  */
1730 int
1731 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1732 {
1733 	struct ip6_hdr ip6;
1734 	struct ip6_ext ip6e;
1735 	struct ip6_frag fh;
1736 
1737 	/* just in case */
1738 	if (m == NULL)
1739 		panic("ip6_nexthdr: m == NULL");
1740 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1741 		return -1;
1742 
1743 	switch (proto) {
1744 	case IPPROTO_IPV6:
1745 		if (m->m_pkthdr.len < off + sizeof(ip6))
1746 			return -1;
1747 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1748 		if (nxtp)
1749 			*nxtp = ip6.ip6_nxt;
1750 		off += sizeof(ip6);
1751 		return off;
1752 
1753 	case IPPROTO_FRAGMENT:
1754 		/*
1755 		 * terminate parsing if it is not the first fragment,
1756 		 * it does not make sense to parse through it.
1757 		 */
1758 		if (m->m_pkthdr.len < off + sizeof(fh))
1759 			return -1;
1760 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1761 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1762 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1763 			return -1;
1764 		if (nxtp)
1765 			*nxtp = fh.ip6f_nxt;
1766 		off += sizeof(struct ip6_frag);
1767 		return off;
1768 
1769 	case IPPROTO_AH:
1770 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1771 			return -1;
1772 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1773 		if (nxtp)
1774 			*nxtp = ip6e.ip6e_nxt;
1775 		off += (ip6e.ip6e_len + 2) << 2;
1776 		return off;
1777 
1778 	case IPPROTO_HOPOPTS:
1779 	case IPPROTO_ROUTING:
1780 	case IPPROTO_DSTOPTS:
1781 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1782 			return -1;
1783 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1784 		if (nxtp)
1785 			*nxtp = ip6e.ip6e_nxt;
1786 		off += (ip6e.ip6e_len + 1) << 3;
1787 		return off;
1788 
1789 	case IPPROTO_NONE:
1790 	case IPPROTO_ESP:
1791 	case IPPROTO_IPCOMP:
1792 		/* give up */
1793 		return -1;
1794 
1795 	default:
1796 		return -1;
1797 	}
1798 
1799 	return -1;
1800 }
1801 
1802 /*
1803  * get offset for the last header in the chain.  m will be kept untainted.
1804  */
1805 int
1806 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1807 {
1808 	int newoff;
1809 	int nxt;
1810 
1811 	if (!nxtp) {
1812 		nxt = -1;
1813 		nxtp = &nxt;
1814 	}
1815 	while (1) {
1816 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1817 		if (newoff < 0)
1818 			return off;
1819 		else if (newoff < off)
1820 			return -1;	/* invalid */
1821 		else if (newoff == off)
1822 			return newoff;
1823 
1824 		off = newoff;
1825 		proto = *nxtp;
1826 	}
1827 }
1828 
1829 static struct ip6aux *
1830 ip6_addaux(struct mbuf *m)
1831 {
1832 	struct m_tag *mtag;
1833 
1834 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1835 	if (!mtag) {
1836 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1837 		    M_NOWAIT);
1838 		if (mtag) {
1839 			m_tag_prepend(m, mtag);
1840 			bzero(mtag + 1, sizeof(struct ip6aux));
1841 		}
1842 	}
1843 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1844 }
1845 
1846 static struct ip6aux *
1847 ip6_findaux(struct mbuf *m)
1848 {
1849 	struct m_tag *mtag;
1850 
1851 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1852 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1853 }
1854 
1855 static void
1856 ip6_delaux(struct mbuf *m)
1857 {
1858 	struct m_tag *mtag;
1859 
1860 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1861 	if (mtag)
1862 		m_tag_delete(m, mtag);
1863 }
1864 
1865 /*
1866  * System control for IP6
1867  */
1868 
1869 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1870 	0,		0,		0,		0,
1871 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1872 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1873 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1874 	0,		0,		0,		0,
1875 	ENOPROTOOPT
1876 };
1877