xref: /freebsd/sys/netinet6/ip6_input.c (revision 595e514d0df2bac5b813d35f83e32875dbf16a83)
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 		if (m->m_next) {
470 			if (m->m_flags & M_LOOP) {
471 				IP6STAT_INC(ip6s_m2m[V_loif->if_index]);
472 			} else if (m->m_pkthdr.rcvif->if_index < IP6S_M2MMAX)
473 				IP6STAT_INC(
474 				    ip6s_m2m[m->m_pkthdr.rcvif->if_index]);
475 			else
476 				IP6STAT_INC(ip6s_m2m[0]);
477 		} else
478 			IP6STAT_INC(ip6s_m1);
479 	}
480 
481 	/* drop the packet if IPv6 operation is disabled on the IF */
482 	if ((ND_IFINFO(m->m_pkthdr.rcvif)->flags & ND6_IFF_IFDISABLED)) {
483 		m_freem(m);
484 		return;
485 	}
486 
487 	in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_receive);
488 	IP6STAT_INC(ip6s_total);
489 
490 #ifndef PULLDOWN_TEST
491 	/*
492 	 * L2 bridge code and some other code can return mbuf chain
493 	 * that does not conform to KAME requirement.  too bad.
494 	 * XXX: fails to join if interface MTU > MCLBYTES.  jumbogram?
495 	 */
496 	if (m && m->m_next != NULL && m->m_pkthdr.len < MCLBYTES) {
497 		struct mbuf *n;
498 
499 		if (m->m_pkthdr.len > MHLEN)
500 			n = m_getcl(M_NOWAIT, MT_DATA, M_PKTHDR);
501 		else
502 			n = m_gethdr(M_NOWAIT, MT_DATA);
503 		if (n == NULL) {
504 			m_freem(m);
505 			return;	/* ENOBUFS */
506 		}
507 
508 		m_move_pkthdr(n, m);
509 		m_copydata(m, 0, n->m_pkthdr.len, mtod(n, caddr_t));
510 		n->m_len = n->m_pkthdr.len;
511 		m_freem(m);
512 		m = n;
513 	}
514 	IP6_EXTHDR_CHECK(m, 0, sizeof(struct ip6_hdr), /* nothing */);
515 #endif
516 
517 	if (m->m_len < sizeof(struct ip6_hdr)) {
518 		struct ifnet *inifp;
519 		inifp = m->m_pkthdr.rcvif;
520 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
521 			IP6STAT_INC(ip6s_toosmall);
522 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
523 			return;
524 		}
525 	}
526 
527 	ip6 = mtod(m, struct ip6_hdr *);
528 
529 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
530 		IP6STAT_INC(ip6s_badvers);
531 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
532 		goto bad;
533 	}
534 
535 	IP6STAT_INC(ip6s_nxthist[ip6->ip6_nxt]);
536 
537 	/*
538 	 * Check against address spoofing/corruption.
539 	 */
540 	if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_src) ||
541 	    IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_dst)) {
542 		/*
543 		 * XXX: "badscope" is not very suitable for a multicast source.
544 		 */
545 		IP6STAT_INC(ip6s_badscope);
546 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
547 		goto bad;
548 	}
549 	if (IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) &&
550 	    !(m->m_flags & M_LOOP)) {
551 		/*
552 		 * In this case, the packet should come from the loopback
553 		 * interface.  However, we cannot just check the if_flags,
554 		 * because ip6_mloopback() passes the "actual" interface
555 		 * as the outgoing/incoming interface.
556 		 */
557 		IP6STAT_INC(ip6s_badscope);
558 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
559 		goto bad;
560 	}
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 			ia6->ia_ifa.if_ipackets++;
715 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
716 
717 			/*
718 			 * record address information into m_tag.
719 			 */
720 			(void)ip6_setdstifaddr(m, ia6);
721 
722 			bad = 0;
723 		} else {
724 			char ip6bufs[INET6_ADDRSTRLEN];
725 			char ip6bufd[INET6_ADDRSTRLEN];
726 			/* address is not ready, so discard the packet. */
727 			nd6log((LOG_INFO,
728 			    "ip6_input: packet to an unready address %s->%s\n",
729 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
730 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
731 		}
732 		IF_ADDR_RUNLOCK(ifp);
733 		LLE_RUNLOCK(lle);
734 		if (bad)
735 			goto bad;
736 		else {
737 			ours = 1;
738 			deliverifp = ifp;
739 			goto hbhcheck;
740 		}
741 	}
742 	if (lle != NULL)
743 		LLE_RUNLOCK(lle);
744 
745 	dst = &rin6.ro_dst;
746 	dst->sin6_len = sizeof(struct sockaddr_in6);
747 	dst->sin6_family = AF_INET6;
748 	dst->sin6_addr = ip6->ip6_dst;
749 	rin6.ro_rt = in6_rtalloc1((struct sockaddr *)dst, 0, 0, M_GETFIB(m));
750 	if (rin6.ro_rt)
751 		RT_UNLOCK(rin6.ro_rt);
752 
753 #define rt6_key(r) ((struct sockaddr_in6 *)((r)->rt_nodes->rn_key))
754 
755 	/*
756 	 * Accept the packet if the forwarding interface to the destination
757 	 * according to the routing table is the loopback interface,
758 	 * unless the associated route has a gateway.
759 	 * Note that this approach causes to accept a packet if there is a
760 	 * route to the loopback interface for the destination of the packet.
761 	 * But we think it's even useful in some situations, e.g. when using
762 	 * a special daemon which wants to intercept the packet.
763 	 *
764 	 * XXX: some OSes automatically make a cloned route for the destination
765 	 * of an outgoing packet.  If the outgoing interface of the packet
766 	 * is a loopback one, the kernel would consider the packet to be
767 	 * accepted, even if we have no such address assinged on the interface.
768 	 * We check the cloned flag of the route entry to reject such cases,
769 	 * assuming that route entries for our own addresses are not made by
770 	 * cloning (it should be true because in6_addloop explicitly installs
771 	 * the host route).  However, we might have to do an explicit check
772 	 * while it would be less efficient.  Or, should we rather install a
773 	 * reject route for such a case?
774 	 */
775 	if (rin6.ro_rt &&
776 	    (rin6.ro_rt->rt_flags &
777 	     (RTF_HOST|RTF_GATEWAY)) == RTF_HOST &&
778 #ifdef RTF_WASCLONED
779 	    !(rin6.ro_rt->rt_flags & RTF_WASCLONED) &&
780 #endif
781 #ifdef RTF_CLONED
782 	    !(rin6.ro_rt->rt_flags & RTF_CLONED) &&
783 #endif
784 #if 0
785 	    /*
786 	     * The check below is redundant since the comparison of
787 	     * the destination and the key of the rtentry has
788 	     * already done through looking up the routing table.
789 	     */
790 	    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
791 	    &rt6_key(rin6.ro_rt)->sin6_addr)
792 #endif
793 	    rin6.ro_rt->rt_ifp->if_type == IFT_LOOP) {
794 		int free_ia6 = 0;
795 		struct in6_ifaddr *ia6;
796 
797 		/*
798 		 * found the loopback route to the interface address
799 		 */
800 		if (rin6.ro_rt->rt_gateway->sa_family == AF_LINK) {
801 			struct sockaddr_in6 dest6;
802 
803 			bzero(&dest6, sizeof(dest6));
804 			dest6.sin6_family = AF_INET6;
805 			dest6.sin6_len = sizeof(dest6);
806 			dest6.sin6_addr = ip6->ip6_dst;
807 			ia6 = (struct in6_ifaddr *)
808 			    ifa_ifwithaddr((struct sockaddr *)&dest6);
809 			if (ia6 == NULL)
810 				goto bad;
811 			free_ia6 = 1;
812 		}
813 		else
814 			ia6 = (struct in6_ifaddr *)rin6.ro_rt->rt_ifa;
815 
816 		/*
817 		 * record address information into m_tag.
818 		 */
819 		(void)ip6_setdstifaddr(m, ia6);
820 
821 		/*
822 		 * packets to a tentative, duplicated, or somehow invalid
823 		 * address must not be accepted.
824 		 */
825 		if (!(ia6->ia6_flags & IN6_IFF_NOTREADY)) {
826 			/* this address is ready */
827 			ours = 1;
828 			deliverifp = ia6->ia_ifp;	/* correct? */
829 			/* Count the packet in the ip address stats */
830 			ia6->ia_ifa.if_ipackets++;
831 			ia6->ia_ifa.if_ibytes += m->m_pkthdr.len;
832 			if (ia6 != NULL && free_ia6 != 0)
833 				ifa_free(&ia6->ia_ifa);
834 			goto hbhcheck;
835 		} else {
836 			char ip6bufs[INET6_ADDRSTRLEN];
837 			char ip6bufd[INET6_ADDRSTRLEN];
838 			/* address is not ready, so discard the packet. */
839 			nd6log((LOG_INFO,
840 			    "ip6_input: packet to an unready address %s->%s\n",
841 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
842 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst)));
843 
844 			if (ia6 != NULL && free_ia6 != 0)
845 				ifa_free(&ia6->ia_ifa);
846 			goto bad;
847 		}
848 	}
849 
850 	/*
851 	 * FAITH (Firewall Aided Internet Translator)
852 	 */
853 	if (V_ip6_keepfaith) {
854 		if (rin6.ro_rt && rin6.ro_rt->rt_ifp &&
855 		    rin6.ro_rt->rt_ifp->if_type == IFT_FAITH) {
856 			/* XXX do we need more sanity checks? */
857 			ours = 1;
858 			deliverifp = rin6.ro_rt->rt_ifp; /* faith */
859 			goto hbhcheck;
860 		}
861 	}
862 
863 	/*
864 	 * Now there is no reason to process the packet if it's not our own
865 	 * and we're not a router.
866 	 */
867 	if (!V_ip6_forwarding) {
868 		IP6STAT_INC(ip6s_cantforward);
869 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
870 		goto bad;
871 	}
872 
873   hbhcheck:
874 	/*
875 	 * record address information into m_tag, if we don't have one yet.
876 	 * note that we are unable to record it, if the address is not listed
877 	 * as our interface address (e.g. multicast addresses, addresses
878 	 * within FAITH prefixes and such).
879 	 */
880 	if (deliverifp) {
881 		struct in6_ifaddr *ia6;
882 
883  		if ((ia6 = ip6_getdstifaddr(m)) != NULL) {
884 			ifa_free(&ia6->ia_ifa);
885 		} else {
886 			ia6 = in6_ifawithifp(deliverifp, &ip6->ip6_dst);
887 			if (ia6) {
888 				if (!ip6_setdstifaddr(m, ia6)) {
889 					/*
890 					 * XXX maybe we should drop the packet here,
891 					 * as we could not provide enough information
892 					 * to the upper layers.
893 					 */
894 				}
895 				ifa_free(&ia6->ia_ifa);
896 			}
897 		}
898 	}
899 
900 	/*
901 	 * Process Hop-by-Hop options header if it's contained.
902 	 * m may be modified in ip6_hopopts_input().
903 	 * If a JumboPayload option is included, plen will also be modified.
904 	 */
905 	plen = (u_int32_t)ntohs(ip6->ip6_plen);
906 	if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
907 		int error;
908 
909 		error = ip6_input_hbh(m, &plen, &rtalert, &off, &nxt, &ours);
910 		if (error != 0)
911 			goto out;
912 	} else
913 		nxt = ip6->ip6_nxt;
914 
915 	/*
916 	 * Check that the amount of data in the buffers
917 	 * is as at least much as the IPv6 header would have us expect.
918 	 * Trim mbufs if longer than we expect.
919 	 * Drop packet if shorter than we expect.
920 	 */
921 	if (m->m_pkthdr.len - sizeof(struct ip6_hdr) < plen) {
922 		IP6STAT_INC(ip6s_tooshort);
923 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
924 		goto bad;
925 	}
926 	if (m->m_pkthdr.len > sizeof(struct ip6_hdr) + plen) {
927 		if (m->m_len == m->m_pkthdr.len) {
928 			m->m_len = sizeof(struct ip6_hdr) + plen;
929 			m->m_pkthdr.len = sizeof(struct ip6_hdr) + plen;
930 		} else
931 			m_adj(m, sizeof(struct ip6_hdr) + plen - m->m_pkthdr.len);
932 	}
933 
934 	/*
935 	 * Forward if desirable.
936 	 */
937 	if (V_ip6_mrouter &&
938 	    IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst)) {
939 		/*
940 		 * If we are acting as a multicast router, all
941 		 * incoming multicast packets are passed to the
942 		 * kernel-level multicast forwarding function.
943 		 * The packet is returned (relatively) intact; if
944 		 * ip6_mforward() returns a non-zero value, the packet
945 		 * must be discarded, else it may be accepted below.
946 		 *
947 		 * XXX TODO: Check hlim and multicast scope here to avoid
948 		 * unnecessarily calling into ip6_mforward().
949 		 */
950 		if (ip6_mforward &&
951 		    ip6_mforward(ip6, m->m_pkthdr.rcvif, m)) {
952 			IP6STAT_INC(ip6s_cantforward);
953 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_discard);
954 			goto bad;
955 		}
956 	} else if (!ours) {
957 		ip6_forward(m, srcrt);
958 		goto out;
959 	}
960 
961 	ip6 = mtod(m, struct ip6_hdr *);
962 
963 	/*
964 	 * Malicious party may be able to use IPv4 mapped addr to confuse
965 	 * tcp/udp stack and bypass security checks (act as if it was from
966 	 * 127.0.0.1 by using IPv6 src ::ffff:127.0.0.1).  Be cautious.
967 	 *
968 	 * For SIIT end node behavior, you may want to disable the check.
969 	 * However, you will  become vulnerable to attacks using IPv4 mapped
970 	 * source.
971 	 */
972 	if (IN6_IS_ADDR_V4MAPPED(&ip6->ip6_src) ||
973 	    IN6_IS_ADDR_V4MAPPED(&ip6->ip6_dst)) {
974 		IP6STAT_INC(ip6s_badscope);
975 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_addrerr);
976 		goto bad;
977 	}
978 
979 	/*
980 	 * Tell launch routine the next header
981 	 */
982 	IP6STAT_INC(ip6s_delivered);
983 	in6_ifstat_inc(deliverifp, ifs6_in_deliver);
984 	nest = 0;
985 
986 	while (nxt != IPPROTO_DONE) {
987 		if (V_ip6_hdrnestlimit && (++nest > V_ip6_hdrnestlimit)) {
988 			IP6STAT_INC(ip6s_toomanyhdr);
989 			goto bad;
990 		}
991 
992 		/*
993 		 * protection against faulty packet - there should be
994 		 * more sanity checks in header chain processing.
995 		 */
996 		if (m->m_pkthdr.len < off) {
997 			IP6STAT_INC(ip6s_tooshort);
998 			in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_truncated);
999 			goto bad;
1000 		}
1001 
1002 #ifdef IPSEC
1003 		/*
1004 		 * enforce IPsec policy checking if we are seeing last header.
1005 		 * note that we do not visit this with protocols with pcb layer
1006 		 * code - like udp/tcp/raw ip.
1007 		 */
1008 		if (ip6_ipsec_input(m, nxt))
1009 			goto bad;
1010 #endif /* IPSEC */
1011 
1012 		/*
1013 		 * Use mbuf flags to propagate Router Alert option to
1014 		 * ICMPv6 layer, as hop-by-hop options have been stripped.
1015 		 */
1016 		if (nxt == IPPROTO_ICMPV6 && rtalert != ~0)
1017 			m->m_flags |= M_RTALERT_MLD;
1018 
1019 		nxt = (*inet6sw[ip6_protox[nxt]].pr_input)(&m, &off, nxt);
1020 	}
1021 	goto out;
1022 bad:
1023 	m_freem(m);
1024 out:
1025 	if (rin6.ro_rt)
1026 		RTFREE(rin6.ro_rt);
1027 }
1028 
1029 /*
1030  * set/grab in6_ifaddr correspond to IPv6 destination address.
1031  * XXX backward compatibility wrapper
1032  *
1033  * XXXRW: We should bump the refcount on ia6 before sticking it in the m_tag,
1034  * and then bump it when the tag is copied, and release it when the tag is
1035  * freed.  Unfortunately, m_tags don't support deep copies (yet), so instead
1036  * we just bump the ia refcount when we receive it.  This should be fixed.
1037  */
1038 static struct ip6aux *
1039 ip6_setdstifaddr(struct mbuf *m, struct in6_ifaddr *ia6)
1040 {
1041 	struct ip6aux *ip6a;
1042 
1043 	ip6a = ip6_addaux(m);
1044 	if (ip6a)
1045 		ip6a->ip6a_dstia6 = ia6;
1046 	return ip6a;	/* NULL if failed to set */
1047 }
1048 
1049 struct in6_ifaddr *
1050 ip6_getdstifaddr(struct mbuf *m)
1051 {
1052 	struct ip6aux *ip6a;
1053 	struct in6_ifaddr *ia;
1054 
1055 	ip6a = ip6_findaux(m);
1056 	if (ip6a) {
1057 		ia = ip6a->ip6a_dstia6;
1058 		ifa_ref(&ia->ia_ifa);
1059 		return ia;
1060 	} else
1061 		return NULL;
1062 }
1063 
1064 /*
1065  * Hop-by-Hop options header processing. If a valid jumbo payload option is
1066  * included, the real payload length will be stored in plenp.
1067  *
1068  * rtalertp - XXX: should be stored more smart way
1069  */
1070 static int
1071 ip6_hopopts_input(u_int32_t *plenp, u_int32_t *rtalertp,
1072     struct mbuf **mp, int *offp)
1073 {
1074 	struct mbuf *m = *mp;
1075 	int off = *offp, hbhlen;
1076 	struct ip6_hbh *hbh;
1077 	u_int8_t *opt;
1078 
1079 	/* validation of the length of the header */
1080 #ifndef PULLDOWN_TEST
1081 	IP6_EXTHDR_CHECK(m, off, sizeof(*hbh), -1);
1082 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1083 	hbhlen = (hbh->ip6h_len + 1) << 3;
1084 
1085 	IP6_EXTHDR_CHECK(m, off, hbhlen, -1);
1086 	hbh = (struct ip6_hbh *)(mtod(m, caddr_t) + off);
1087 #else
1088 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m,
1089 		sizeof(struct ip6_hdr), sizeof(struct ip6_hbh));
1090 	if (hbh == NULL) {
1091 		IP6STAT_INC(ip6s_tooshort);
1092 		return -1;
1093 	}
1094 	hbhlen = (hbh->ip6h_len + 1) << 3;
1095 	IP6_EXTHDR_GET(hbh, struct ip6_hbh *, m, sizeof(struct ip6_hdr),
1096 		hbhlen);
1097 	if (hbh == NULL) {
1098 		IP6STAT_INC(ip6s_tooshort);
1099 		return -1;
1100 	}
1101 #endif
1102 	off += hbhlen;
1103 	hbhlen -= sizeof(struct ip6_hbh);
1104 	opt = (u_int8_t *)hbh + sizeof(struct ip6_hbh);
1105 
1106 	if (ip6_process_hopopts(m, (u_int8_t *)hbh + sizeof(struct ip6_hbh),
1107 				hbhlen, rtalertp, plenp) < 0)
1108 		return (-1);
1109 
1110 	*offp = off;
1111 	*mp = m;
1112 	return (0);
1113 }
1114 
1115 /*
1116  * Search header for all Hop-by-hop options and process each option.
1117  * This function is separate from ip6_hopopts_input() in order to
1118  * handle a case where the sending node itself process its hop-by-hop
1119  * options header. In such a case, the function is called from ip6_output().
1120  *
1121  * The function assumes that hbh header is located right after the IPv6 header
1122  * (RFC2460 p7), opthead is pointer into data content in m, and opthead to
1123  * opthead + hbhlen is located in contiguous memory region.
1124  */
1125 int
1126 ip6_process_hopopts(struct mbuf *m, u_int8_t *opthead, int hbhlen,
1127     u_int32_t *rtalertp, u_int32_t *plenp)
1128 {
1129 	struct ip6_hdr *ip6;
1130 	int optlen = 0;
1131 	u_int8_t *opt = opthead;
1132 	u_int16_t rtalert_val;
1133 	u_int32_t jumboplen;
1134 	const int erroff = sizeof(struct ip6_hdr) + sizeof(struct ip6_hbh);
1135 
1136 	for (; hbhlen > 0; hbhlen -= optlen, opt += optlen) {
1137 		switch (*opt) {
1138 		case IP6OPT_PAD1:
1139 			optlen = 1;
1140 			break;
1141 		case IP6OPT_PADN:
1142 			if (hbhlen < IP6OPT_MINLEN) {
1143 				IP6STAT_INC(ip6s_toosmall);
1144 				goto bad;
1145 			}
1146 			optlen = *(opt + 1) + 2;
1147 			break;
1148 		case IP6OPT_ROUTER_ALERT:
1149 			/* XXX may need check for alignment */
1150 			if (hbhlen < IP6OPT_RTALERT_LEN) {
1151 				IP6STAT_INC(ip6s_toosmall);
1152 				goto bad;
1153 			}
1154 			if (*(opt + 1) != IP6OPT_RTALERT_LEN - 2) {
1155 				/* XXX stat */
1156 				icmp6_error(m, ICMP6_PARAM_PROB,
1157 				    ICMP6_PARAMPROB_HEADER,
1158 				    erroff + opt + 1 - opthead);
1159 				return (-1);
1160 			}
1161 			optlen = IP6OPT_RTALERT_LEN;
1162 			bcopy((caddr_t)(opt + 2), (caddr_t)&rtalert_val, 2);
1163 			*rtalertp = ntohs(rtalert_val);
1164 			break;
1165 		case IP6OPT_JUMBO:
1166 			/* XXX may need check for alignment */
1167 			if (hbhlen < IP6OPT_JUMBO_LEN) {
1168 				IP6STAT_INC(ip6s_toosmall);
1169 				goto bad;
1170 			}
1171 			if (*(opt + 1) != IP6OPT_JUMBO_LEN - 2) {
1172 				/* XXX stat */
1173 				icmp6_error(m, ICMP6_PARAM_PROB,
1174 				    ICMP6_PARAMPROB_HEADER,
1175 				    erroff + opt + 1 - opthead);
1176 				return (-1);
1177 			}
1178 			optlen = IP6OPT_JUMBO_LEN;
1179 
1180 			/*
1181 			 * IPv6 packets that have non 0 payload length
1182 			 * must not contain a jumbo payload option.
1183 			 */
1184 			ip6 = mtod(m, struct ip6_hdr *);
1185 			if (ip6->ip6_plen) {
1186 				IP6STAT_INC(ip6s_badoptions);
1187 				icmp6_error(m, ICMP6_PARAM_PROB,
1188 				    ICMP6_PARAMPROB_HEADER,
1189 				    erroff + opt - opthead);
1190 				return (-1);
1191 			}
1192 
1193 			/*
1194 			 * We may see jumbolen in unaligned location, so
1195 			 * we'd need to perform bcopy().
1196 			 */
1197 			bcopy(opt + 2, &jumboplen, sizeof(jumboplen));
1198 			jumboplen = (u_int32_t)htonl(jumboplen);
1199 
1200 #if 1
1201 			/*
1202 			 * if there are multiple jumbo payload options,
1203 			 * *plenp will be non-zero and the packet will be
1204 			 * rejected.
1205 			 * the behavior may need some debate in ipngwg -
1206 			 * multiple options does not make sense, however,
1207 			 * there's no explicit mention in specification.
1208 			 */
1209 			if (*plenp != 0) {
1210 				IP6STAT_INC(ip6s_badoptions);
1211 				icmp6_error(m, ICMP6_PARAM_PROB,
1212 				    ICMP6_PARAMPROB_HEADER,
1213 				    erroff + opt + 2 - opthead);
1214 				return (-1);
1215 			}
1216 #endif
1217 
1218 			/*
1219 			 * jumbo payload length must be larger than 65535.
1220 			 */
1221 			if (jumboplen <= IPV6_MAXPACKET) {
1222 				IP6STAT_INC(ip6s_badoptions);
1223 				icmp6_error(m, ICMP6_PARAM_PROB,
1224 				    ICMP6_PARAMPROB_HEADER,
1225 				    erroff + opt + 2 - opthead);
1226 				return (-1);
1227 			}
1228 			*plenp = jumboplen;
1229 
1230 			break;
1231 		default:		/* unknown option */
1232 			if (hbhlen < IP6OPT_MINLEN) {
1233 				IP6STAT_INC(ip6s_toosmall);
1234 				goto bad;
1235 			}
1236 			optlen = ip6_unknown_opt(opt, m,
1237 			    erroff + opt - opthead);
1238 			if (optlen == -1)
1239 				return (-1);
1240 			optlen += 2;
1241 			break;
1242 		}
1243 	}
1244 
1245 	return (0);
1246 
1247   bad:
1248 	m_freem(m);
1249 	return (-1);
1250 }
1251 
1252 /*
1253  * Unknown option processing.
1254  * The third argument `off' is the offset from the IPv6 header to the option,
1255  * which is necessary if the IPv6 header the and option header and IPv6 header
1256  * is not contiguous in order to return an ICMPv6 error.
1257  */
1258 int
1259 ip6_unknown_opt(u_int8_t *optp, struct mbuf *m, int off)
1260 {
1261 	struct ip6_hdr *ip6;
1262 
1263 	switch (IP6OPT_TYPE(*optp)) {
1264 	case IP6OPT_TYPE_SKIP: /* ignore the option */
1265 		return ((int)*(optp + 1));
1266 	case IP6OPT_TYPE_DISCARD:	/* silently discard */
1267 		m_freem(m);
1268 		return (-1);
1269 	case IP6OPT_TYPE_FORCEICMP: /* send ICMP even if multicasted */
1270 		IP6STAT_INC(ip6s_badoptions);
1271 		icmp6_error(m, ICMP6_PARAM_PROB, ICMP6_PARAMPROB_OPTION, off);
1272 		return (-1);
1273 	case IP6OPT_TYPE_ICMP: /* send ICMP if not multicasted */
1274 		IP6STAT_INC(ip6s_badoptions);
1275 		ip6 = mtod(m, struct ip6_hdr *);
1276 		if (IN6_IS_ADDR_MULTICAST(&ip6->ip6_dst) ||
1277 		    (m->m_flags & (M_BCAST|M_MCAST)))
1278 			m_freem(m);
1279 		else
1280 			icmp6_error(m, ICMP6_PARAM_PROB,
1281 				    ICMP6_PARAMPROB_OPTION, off);
1282 		return (-1);
1283 	}
1284 
1285 	m_freem(m);		/* XXX: NOTREACHED */
1286 	return (-1);
1287 }
1288 
1289 /*
1290  * Create the "control" list for this pcb.
1291  * These functions will not modify mbuf chain at all.
1292  *
1293  * With KAME mbuf chain restriction:
1294  * The routine will be called from upper layer handlers like tcp6_input().
1295  * Thus the routine assumes that the caller (tcp6_input) have already
1296  * called IP6_EXTHDR_CHECK() and all the extension headers are located in the
1297  * very first mbuf on the mbuf chain.
1298  *
1299  * ip6_savecontrol_v4 will handle those options that are possible to be
1300  * set on a v4-mapped socket.
1301  * ip6_savecontrol will directly call ip6_savecontrol_v4 to handle those
1302  * options and handle the v6-only ones itself.
1303  */
1304 struct mbuf **
1305 ip6_savecontrol_v4(struct inpcb *inp, struct mbuf *m, struct mbuf **mp,
1306     int *v4only)
1307 {
1308 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1309 
1310 #ifdef SO_TIMESTAMP
1311 	if ((inp->inp_socket->so_options & SO_TIMESTAMP) != 0) {
1312 		struct timeval tv;
1313 
1314 		microtime(&tv);
1315 		*mp = sbcreatecontrol((caddr_t) &tv, sizeof(tv),
1316 		    SCM_TIMESTAMP, SOL_SOCKET);
1317 		if (*mp)
1318 			mp = &(*mp)->m_next;
1319 	}
1320 #endif
1321 
1322 #define IS2292(inp, x, y)	(((inp)->inp_flags & IN6P_RFC2292) ? (x) : (y))
1323 	/* RFC 2292 sec. 5 */
1324 	if ((inp->inp_flags & IN6P_PKTINFO) != 0) {
1325 		struct in6_pktinfo pi6;
1326 
1327 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1328 #ifdef INET
1329 			struct ip *ip;
1330 
1331 			ip = mtod(m, struct ip *);
1332 			pi6.ipi6_addr.s6_addr32[0] = 0;
1333 			pi6.ipi6_addr.s6_addr32[1] = 0;
1334 			pi6.ipi6_addr.s6_addr32[2] = IPV6_ADDR_INT32_SMP;
1335 			pi6.ipi6_addr.s6_addr32[3] = ip->ip_dst.s_addr;
1336 #else
1337 			/* We won't hit this code */
1338 			bzero(&pi6.ipi6_addr, sizeof(struct in6_addr));
1339 #endif
1340 		} else {
1341 			bcopy(&ip6->ip6_dst, &pi6.ipi6_addr, sizeof(struct in6_addr));
1342 			in6_clearscope(&pi6.ipi6_addr);	/* XXX */
1343 		}
1344 		pi6.ipi6_ifindex =
1345 		    (m && m->m_pkthdr.rcvif) ? m->m_pkthdr.rcvif->if_index : 0;
1346 
1347 		*mp = sbcreatecontrol((caddr_t) &pi6,
1348 		    sizeof(struct in6_pktinfo),
1349 		    IS2292(inp, IPV6_2292PKTINFO, IPV6_PKTINFO), IPPROTO_IPV6);
1350 		if (*mp)
1351 			mp = &(*mp)->m_next;
1352 	}
1353 
1354 	if ((inp->inp_flags & IN6P_HOPLIMIT) != 0) {
1355 		int hlim;
1356 
1357 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1358 #ifdef INET
1359 			struct ip *ip;
1360 
1361 			ip = mtod(m, struct ip *);
1362 			hlim = ip->ip_ttl;
1363 #else
1364 			/* We won't hit this code */
1365 			hlim = 0;
1366 #endif
1367 		} else {
1368 			hlim = ip6->ip6_hlim & 0xff;
1369 		}
1370 		*mp = sbcreatecontrol((caddr_t) &hlim, sizeof(int),
1371 		    IS2292(inp, IPV6_2292HOPLIMIT, IPV6_HOPLIMIT),
1372 		    IPPROTO_IPV6);
1373 		if (*mp)
1374 			mp = &(*mp)->m_next;
1375 	}
1376 
1377 	if ((inp->inp_flags & IN6P_TCLASS) != 0) {
1378 		int tclass;
1379 
1380 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1381 #ifdef INET
1382 			struct ip *ip;
1383 
1384 			ip = mtod(m, struct ip *);
1385 			tclass = ip->ip_tos;
1386 #else
1387 			/* We won't hit this code */
1388 			tclass = 0;
1389 #endif
1390 		} else {
1391 			u_int32_t flowinfo;
1392 
1393 			flowinfo = (u_int32_t)ntohl(ip6->ip6_flow & IPV6_FLOWINFO_MASK);
1394 			flowinfo >>= 20;
1395 			tclass = flowinfo & 0xff;
1396 		}
1397 		*mp = sbcreatecontrol((caddr_t) &tclass, sizeof(int),
1398 		    IPV6_TCLASS, IPPROTO_IPV6);
1399 		if (*mp)
1400 			mp = &(*mp)->m_next;
1401 	}
1402 
1403 	if (v4only != NULL) {
1404 		if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1405 			*v4only = 1;
1406 		} else {
1407 			*v4only = 0;
1408 		}
1409 	}
1410 
1411 	return (mp);
1412 }
1413 
1414 void
1415 ip6_savecontrol(struct inpcb *in6p, struct mbuf *m, struct mbuf **mp)
1416 {
1417 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1418 	int v4only = 0;
1419 
1420 	mp = ip6_savecontrol_v4(in6p, m, mp, &v4only);
1421 	if (v4only)
1422 		return;
1423 
1424 	/*
1425 	 * IPV6_HOPOPTS socket option.  Recall that we required super-user
1426 	 * privilege for the option (see ip6_ctloutput), but it might be too
1427 	 * strict, since there might be some hop-by-hop options which can be
1428 	 * returned to normal user.
1429 	 * See also RFC 2292 section 6 (or RFC 3542 section 8).
1430 	 */
1431 	if ((in6p->inp_flags & IN6P_HOPOPTS) != 0) {
1432 		/*
1433 		 * Check if a hop-by-hop options header is contatined in the
1434 		 * received packet, and if so, store the options as ancillary
1435 		 * data. Note that a hop-by-hop options header must be
1436 		 * just after the IPv6 header, which is assured through the
1437 		 * IPv6 input processing.
1438 		 */
1439 		if (ip6->ip6_nxt == IPPROTO_HOPOPTS) {
1440 			struct ip6_hbh *hbh;
1441 			int hbhlen = 0;
1442 #ifdef PULLDOWN_TEST
1443 			struct mbuf *ext;
1444 #endif
1445 
1446 #ifndef PULLDOWN_TEST
1447 			hbh = (struct ip6_hbh *)(ip6 + 1);
1448 			hbhlen = (hbh->ip6h_len + 1) << 3;
1449 #else
1450 			ext = ip6_pullexthdr(m, sizeof(struct ip6_hdr),
1451 			    ip6->ip6_nxt);
1452 			if (ext == NULL) {
1453 				IP6STAT_INC(ip6s_tooshort);
1454 				return;
1455 			}
1456 			hbh = mtod(ext, struct ip6_hbh *);
1457 			hbhlen = (hbh->ip6h_len + 1) << 3;
1458 			if (hbhlen != ext->m_len) {
1459 				m_freem(ext);
1460 				IP6STAT_INC(ip6s_tooshort);
1461 				return;
1462 			}
1463 #endif
1464 
1465 			/*
1466 			 * XXX: We copy the whole header even if a
1467 			 * jumbo payload option is included, the option which
1468 			 * is to be removed before returning according to
1469 			 * RFC2292.
1470 			 * Note: this constraint is removed in RFC3542
1471 			 */
1472 			*mp = sbcreatecontrol((caddr_t)hbh, hbhlen,
1473 			    IS2292(in6p, IPV6_2292HOPOPTS, IPV6_HOPOPTS),
1474 			    IPPROTO_IPV6);
1475 			if (*mp)
1476 				mp = &(*mp)->m_next;
1477 #ifdef PULLDOWN_TEST
1478 			m_freem(ext);
1479 #endif
1480 		}
1481 	}
1482 
1483 	if ((in6p->inp_flags & (IN6P_RTHDR | IN6P_DSTOPTS)) != 0) {
1484 		int nxt = ip6->ip6_nxt, off = sizeof(struct ip6_hdr);
1485 
1486 		/*
1487 		 * Search for destination options headers or routing
1488 		 * header(s) through the header chain, and stores each
1489 		 * header as ancillary data.
1490 		 * Note that the order of the headers remains in
1491 		 * the chain of ancillary data.
1492 		 */
1493 		while (1) {	/* is explicit loop prevention necessary? */
1494 			struct ip6_ext *ip6e = NULL;
1495 			int elen;
1496 #ifdef PULLDOWN_TEST
1497 			struct mbuf *ext = NULL;
1498 #endif
1499 
1500 			/*
1501 			 * if it is not an extension header, don't try to
1502 			 * pull it from the chain.
1503 			 */
1504 			switch (nxt) {
1505 			case IPPROTO_DSTOPTS:
1506 			case IPPROTO_ROUTING:
1507 			case IPPROTO_HOPOPTS:
1508 			case IPPROTO_AH: /* is it possible? */
1509 				break;
1510 			default:
1511 				goto loopend;
1512 			}
1513 
1514 #ifndef PULLDOWN_TEST
1515 			if (off + sizeof(*ip6e) > m->m_len)
1516 				goto loopend;
1517 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + off);
1518 			if (nxt == IPPROTO_AH)
1519 				elen = (ip6e->ip6e_len + 2) << 2;
1520 			else
1521 				elen = (ip6e->ip6e_len + 1) << 3;
1522 			if (off + elen > m->m_len)
1523 				goto loopend;
1524 #else
1525 			ext = ip6_pullexthdr(m, off, nxt);
1526 			if (ext == NULL) {
1527 				IP6STAT_INC(ip6s_tooshort);
1528 				return;
1529 			}
1530 			ip6e = mtod(ext, struct ip6_ext *);
1531 			if (nxt == IPPROTO_AH)
1532 				elen = (ip6e->ip6e_len + 2) << 2;
1533 			else
1534 				elen = (ip6e->ip6e_len + 1) << 3;
1535 			if (elen != ext->m_len) {
1536 				m_freem(ext);
1537 				IP6STAT_INC(ip6s_tooshort);
1538 				return;
1539 			}
1540 #endif
1541 
1542 			switch (nxt) {
1543 			case IPPROTO_DSTOPTS:
1544 				if (!(in6p->inp_flags & IN6P_DSTOPTS))
1545 					break;
1546 
1547 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1548 				    IS2292(in6p,
1549 					IPV6_2292DSTOPTS, IPV6_DSTOPTS),
1550 				    IPPROTO_IPV6);
1551 				if (*mp)
1552 					mp = &(*mp)->m_next;
1553 				break;
1554 			case IPPROTO_ROUTING:
1555 				if (!(in6p->inp_flags & IN6P_RTHDR))
1556 					break;
1557 
1558 				*mp = sbcreatecontrol((caddr_t)ip6e, elen,
1559 				    IS2292(in6p, IPV6_2292RTHDR, IPV6_RTHDR),
1560 				    IPPROTO_IPV6);
1561 				if (*mp)
1562 					mp = &(*mp)->m_next;
1563 				break;
1564 			case IPPROTO_HOPOPTS:
1565 			case IPPROTO_AH: /* is it possible? */
1566 				break;
1567 
1568 			default:
1569 				/*
1570 				 * other cases have been filtered in the above.
1571 				 * none will visit this case.  here we supply
1572 				 * the code just in case (nxt overwritten or
1573 				 * other cases).
1574 				 */
1575 #ifdef PULLDOWN_TEST
1576 				m_freem(ext);
1577 #endif
1578 				goto loopend;
1579 
1580 			}
1581 
1582 			/* proceed with the next header. */
1583 			off += elen;
1584 			nxt = ip6e->ip6e_nxt;
1585 			ip6e = NULL;
1586 #ifdef PULLDOWN_TEST
1587 			m_freem(ext);
1588 			ext = NULL;
1589 #endif
1590 		}
1591 	  loopend:
1592 		;
1593 	}
1594 }
1595 #undef IS2292
1596 
1597 void
1598 ip6_notify_pmtu(struct inpcb *in6p, struct sockaddr_in6 *dst, u_int32_t *mtu)
1599 {
1600 	struct socket *so;
1601 	struct mbuf *m_mtu;
1602 	struct ip6_mtuinfo mtuctl;
1603 
1604 	so =  in6p->inp_socket;
1605 
1606 	if (mtu == NULL)
1607 		return;
1608 
1609 #ifdef DIAGNOSTIC
1610 	if (so == NULL)		/* I believe this is impossible */
1611 		panic("ip6_notify_pmtu: socket is NULL");
1612 #endif
1613 
1614 	bzero(&mtuctl, sizeof(mtuctl));	/* zero-clear for safety */
1615 	mtuctl.ip6m_mtu = *mtu;
1616 	mtuctl.ip6m_addr = *dst;
1617 	if (sa6_recoverscope(&mtuctl.ip6m_addr))
1618 		return;
1619 
1620 	if ((m_mtu = sbcreatecontrol((caddr_t)&mtuctl, sizeof(mtuctl),
1621 	    IPV6_PATHMTU, IPPROTO_IPV6)) == NULL)
1622 		return;
1623 
1624 	if (sbappendaddr(&so->so_rcv, (struct sockaddr *)dst, NULL, m_mtu)
1625 	    == 0) {
1626 		m_freem(m_mtu);
1627 		/* XXX: should count statistics */
1628 	} else
1629 		sorwakeup(so);
1630 
1631 	return;
1632 }
1633 
1634 #ifdef PULLDOWN_TEST
1635 /*
1636  * pull single extension header from mbuf chain.  returns single mbuf that
1637  * contains the result, or NULL on error.
1638  */
1639 static struct mbuf *
1640 ip6_pullexthdr(struct mbuf *m, size_t off, int nxt)
1641 {
1642 	struct ip6_ext ip6e;
1643 	size_t elen;
1644 	struct mbuf *n;
1645 
1646 #ifdef DIAGNOSTIC
1647 	switch (nxt) {
1648 	case IPPROTO_DSTOPTS:
1649 	case IPPROTO_ROUTING:
1650 	case IPPROTO_HOPOPTS:
1651 	case IPPROTO_AH: /* is it possible? */
1652 		break;
1653 	default:
1654 		printf("ip6_pullexthdr: invalid nxt=%d\n", nxt);
1655 	}
1656 #endif
1657 
1658 	m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1659 	if (nxt == IPPROTO_AH)
1660 		elen = (ip6e.ip6e_len + 2) << 2;
1661 	else
1662 		elen = (ip6e.ip6e_len + 1) << 3;
1663 
1664 	if (elen > MLEN)
1665 		n = m_getcl(M_NOWAIT, MT_DATA, 0);
1666 	else
1667 		n = m_get(M_NOWAIT, MT_DATA);
1668 	if (n == NULL)
1669 		return NULL;
1670 
1671 	m_copydata(m, off, elen, mtod(n, caddr_t));
1672 	n->m_len = elen;
1673 	return n;
1674 }
1675 #endif
1676 
1677 /*
1678  * Get pointer to the previous header followed by the header
1679  * currently processed.
1680  * XXX: This function supposes that
1681  *	M includes all headers,
1682  *	the next header field and the header length field of each header
1683  *	are valid, and
1684  *	the sum of each header length equals to OFF.
1685  * Because of these assumptions, this function must be called very
1686  * carefully. Moreover, it will not be used in the near future when
1687  * we develop `neater' mechanism to process extension headers.
1688  */
1689 char *
1690 ip6_get_prevhdr(struct mbuf *m, int off)
1691 {
1692 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1693 
1694 	if (off == sizeof(struct ip6_hdr))
1695 		return (&ip6->ip6_nxt);
1696 	else {
1697 		int len, nxt;
1698 		struct ip6_ext *ip6e = NULL;
1699 
1700 		nxt = ip6->ip6_nxt;
1701 		len = sizeof(struct ip6_hdr);
1702 		while (len < off) {
1703 			ip6e = (struct ip6_ext *)(mtod(m, caddr_t) + len);
1704 
1705 			switch (nxt) {
1706 			case IPPROTO_FRAGMENT:
1707 				len += sizeof(struct ip6_frag);
1708 				break;
1709 			case IPPROTO_AH:
1710 				len += (ip6e->ip6e_len + 2) << 2;
1711 				break;
1712 			default:
1713 				len += (ip6e->ip6e_len + 1) << 3;
1714 				break;
1715 			}
1716 			nxt = ip6e->ip6e_nxt;
1717 		}
1718 		if (ip6e)
1719 			return (&ip6e->ip6e_nxt);
1720 		else
1721 			return NULL;
1722 	}
1723 }
1724 
1725 /*
1726  * get next header offset.  m will be retained.
1727  */
1728 int
1729 ip6_nexthdr(struct mbuf *m, int off, int proto, int *nxtp)
1730 {
1731 	struct ip6_hdr ip6;
1732 	struct ip6_ext ip6e;
1733 	struct ip6_frag fh;
1734 
1735 	/* just in case */
1736 	if (m == NULL)
1737 		panic("ip6_nexthdr: m == NULL");
1738 	if ((m->m_flags & M_PKTHDR) == 0 || m->m_pkthdr.len < off)
1739 		return -1;
1740 
1741 	switch (proto) {
1742 	case IPPROTO_IPV6:
1743 		if (m->m_pkthdr.len < off + sizeof(ip6))
1744 			return -1;
1745 		m_copydata(m, off, sizeof(ip6), (caddr_t)&ip6);
1746 		if (nxtp)
1747 			*nxtp = ip6.ip6_nxt;
1748 		off += sizeof(ip6);
1749 		return off;
1750 
1751 	case IPPROTO_FRAGMENT:
1752 		/*
1753 		 * terminate parsing if it is not the first fragment,
1754 		 * it does not make sense to parse through it.
1755 		 */
1756 		if (m->m_pkthdr.len < off + sizeof(fh))
1757 			return -1;
1758 		m_copydata(m, off, sizeof(fh), (caddr_t)&fh);
1759 		/* IP6F_OFF_MASK = 0xfff8(BigEndian), 0xf8ff(LittleEndian) */
1760 		if (fh.ip6f_offlg & IP6F_OFF_MASK)
1761 			return -1;
1762 		if (nxtp)
1763 			*nxtp = fh.ip6f_nxt;
1764 		off += sizeof(struct ip6_frag);
1765 		return off;
1766 
1767 	case IPPROTO_AH:
1768 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1769 			return -1;
1770 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1771 		if (nxtp)
1772 			*nxtp = ip6e.ip6e_nxt;
1773 		off += (ip6e.ip6e_len + 2) << 2;
1774 		return off;
1775 
1776 	case IPPROTO_HOPOPTS:
1777 	case IPPROTO_ROUTING:
1778 	case IPPROTO_DSTOPTS:
1779 		if (m->m_pkthdr.len < off + sizeof(ip6e))
1780 			return -1;
1781 		m_copydata(m, off, sizeof(ip6e), (caddr_t)&ip6e);
1782 		if (nxtp)
1783 			*nxtp = ip6e.ip6e_nxt;
1784 		off += (ip6e.ip6e_len + 1) << 3;
1785 		return off;
1786 
1787 	case IPPROTO_NONE:
1788 	case IPPROTO_ESP:
1789 	case IPPROTO_IPCOMP:
1790 		/* give up */
1791 		return -1;
1792 
1793 	default:
1794 		return -1;
1795 	}
1796 
1797 	return -1;
1798 }
1799 
1800 /*
1801  * get offset for the last header in the chain.  m will be kept untainted.
1802  */
1803 int
1804 ip6_lasthdr(struct mbuf *m, int off, int proto, int *nxtp)
1805 {
1806 	int newoff;
1807 	int nxt;
1808 
1809 	if (!nxtp) {
1810 		nxt = -1;
1811 		nxtp = &nxt;
1812 	}
1813 	while (1) {
1814 		newoff = ip6_nexthdr(m, off, proto, nxtp);
1815 		if (newoff < 0)
1816 			return off;
1817 		else if (newoff < off)
1818 			return -1;	/* invalid */
1819 		else if (newoff == off)
1820 			return newoff;
1821 
1822 		off = newoff;
1823 		proto = *nxtp;
1824 	}
1825 }
1826 
1827 static struct ip6aux *
1828 ip6_addaux(struct mbuf *m)
1829 {
1830 	struct m_tag *mtag;
1831 
1832 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1833 	if (!mtag) {
1834 		mtag = m_tag_get(PACKET_TAG_IPV6_INPUT, sizeof(struct ip6aux),
1835 		    M_NOWAIT);
1836 		if (mtag) {
1837 			m_tag_prepend(m, mtag);
1838 			bzero(mtag + 1, sizeof(struct ip6aux));
1839 		}
1840 	}
1841 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1842 }
1843 
1844 static struct ip6aux *
1845 ip6_findaux(struct mbuf *m)
1846 {
1847 	struct m_tag *mtag;
1848 
1849 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1850 	return mtag ? (struct ip6aux *)(mtag + 1) : NULL;
1851 }
1852 
1853 static void
1854 ip6_delaux(struct mbuf *m)
1855 {
1856 	struct m_tag *mtag;
1857 
1858 	mtag = m_tag_find(m, PACKET_TAG_IPV6_INPUT, NULL);
1859 	if (mtag)
1860 		m_tag_delete(m, mtag);
1861 }
1862 
1863 /*
1864  * System control for IP6
1865  */
1866 
1867 u_char	inet6ctlerrmap[PRC_NCMDS] = {
1868 	0,		0,		0,		0,
1869 	0,		EMSGSIZE,	EHOSTDOWN,	EHOSTUNREACH,
1870 	EHOSTUNREACH,	EHOSTUNREACH,	ECONNREFUSED,	ECONNREFUSED,
1871 	EMSGSIZE,	EHOSTUNREACH,	0,		0,
1872 	0,		0,		0,		0,
1873 	ENOPROTOOPT
1874 };
1875